transformation of the employment structure in the eu and usa, 1995â€“2007

Transformation of the Employment Structure in the EU and USA, 1995–2007

Also by Enrique Fernández-Macías

THE SOCIAL ECONOMICS OF JOB QUALITY (co-authored with Rafael Muñoz de Bustillo, José-Ignacio Antón and Fernando Esteve)

Also by John Hurley

MORE AND BETTER JOBS? Patterns of Employment and Expansion in the EU, 1995–2006 (co-authored with Enrique Fernández-Macías)

Also by Donald Storrie

TEMPORARY AGENCY WORK IN THE EUROPEAN UNION CONTINGENT EMPLOYMENT IN EUROPE AND THE UNITED STATES (co-edited with Ola Bergstrom)

Transformation of the Employment Structure in the EU and USA, 1995–2007

Edited by

Enrique Fernández-Macías University of Salamanca, Spain

John Hurley European Foundation for the Improvement of Living and Working Conditions, Dublin, Ireland

and

Donald Storrie European Foundation for the Improvement of Living and Working Conditions, Dublin, Ireland

Selection and editorial matter © Enrique Fernández-Macías, John Hurley and Donald Storrie 2012 Individual chapters © their respective authors 2012 Softcover reprint of the hardcover 1st edition 2012 978-0-230-29779-1

All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission.

No portion of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting limited copying issued by the Copyright Licensing Agency, Saffron House, 6–10 Kirby Street, London EC1N 8TS.

Any person who does any unauthorized act in relation to this publication may be liable to criminal prosecution and civil claims for damages.

The authors have asserted their rights to be identified as the authors of this work in accordance with the Copyright, Designs and Patents Act 1988.

First published 2012 by PALGRAVE MACMILLAN

Palgrave Macmillan in the UK is an imprint of Macmillan Publishers Limited, registered in England, company number 785998, of Houndmills, Basingstoke, Hampshire RG21 6XS.

Palgrave Macmillan in the US is a division of St Martin’s Press LLC, 175 Fifth Avenue, New York, NY 10010.

Palgrave Macmillan is the global academic imprint of the above companies and has companies and representatives throughout the world.

Palgrave® and Macmillan® are registered trademarks in the United States, the United Kingdom, Europe and other countries

ISBN 978-1-349-33416-2 ISBN 978-0-230-36981-8 (eBook)DOI 10.1057/9780230369818 This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. Logging, pulping and manufacturing processes are expected to conform to the environmental regulations of the country of origin.

A catalogue record for this book is available from the British Library.

A catalog record for this book is available from the Library of Congress.

10 9 8 7 6 5 4 3 2 1 21 20 19 18 17 16 15 14 13 12

v

Contents

List of Illustrations vii

Preface and Acknowledgements xiii

Notes on Contributors xv

1 Introduction 1 Donald Storrie, John Hurley and Enrique Fernández-Macías

2 Methodology 16 Enrique Fernández-Macías, Terry Ward and Robert Stehrer

3 Patterns of Employment Expansion in Europe, 1995–2007 26 Enrique Fernández-Macías

4 Job Growth and Job Polarization in the United States and Europe, 1995–2007 52 Rachel E. Dwyer and Erik Olin Wright

5 Women’s Changing Job Structure in Europe: Patterns of Job Concentration, Low Pay and Welfare State Employment 75 Damian Grimshaw and Hugo Figueiredo

6 Immigration and Labour Market Segmentation in the European Union 111 Rafael Muñ oz de Bustillo and José-Ignacio Antón

7 Assessing Recent Employment Shifts in Europe Using a Multidimensional Job Quality Indicator 147 John Hurley, Enrique Fernández-Macias and Rafael Muñoz de Bustillo

8 Job Quality in Post-Socialist Accession Countries 180 Ágota Scharle

vi Contents

9 The Institutional Context of Skills–Wages Mismatches 201 Jean-Marie Jungblut and Philip O’Connell

10 The Changing Structure of Employment during Periods of Recession and Recovery in the EU 244 Donald Storrie, Terry Ward, Robert Stehrer and John Hurley

References 279

Index 293

vii

Illustrations

Tables

2.1 Number of jobs identified in each country 17 2.2 Correlation between wage and educational rankings

at the job level 21 2.3 Periods covered – breaks and adjustments 22 5.1 Change in average wage quintile measure for

women and men, 1995–2007 84 5.2 Percentage share of low-wage jobs among women

by type of wage-setting system 90 5.3 Changes in low-wage, middle-wage and high-wage

jobs in the EU23 92 5.4 The number of jobs that contributed to women’s

job growth and job destruction, 1997–2007 98 5.5 Wage and skill quintile measures for women’s top-ten

jobs by employment concentration, 2007 100 5.6 Common jobs with high female employment

concentration, 2007 102 5.7 Employment in welfare state jobs, by gender, 2007 104 5.8 Contribution of welfare state jobs to overall job

change, by gender 106 6.1 Methodological issues in the analysis of the

Jobs Project database 114 6.2 Stock of foreign-born population in a selection of

EU Member States, by country of birth, around 2006 120 6.3 Duncan Index of occupational segregation in Europe

for migrant status and gender, 2006 131 6.4 Correlation coefficient between change in native and

migrant employment growth by job cell 133 6.5 Evolution of total employment of immigrants and locals 137 7.1 Dimensions of job quality suggested by the

different traditions 157 7.2 Country rankings using the job amenities

ranking and the ETUI’s Job Quality Index 166 7.3 Bivariate ANOVAS with job amenities value as

dependent variable 167

viii List of Illustrations

7.4 Pairwise correlations of job rankings by skill, wage, job amenity and its four dimensions 169

8.1 Policy mix and expected labour-market outcome 189 8.2 Level and change of self-employment in the

accession countries 192 9.1 Contingency table of wage and skill quintiles for all

countries in the analysis 208 9.2 Skill–wage disparities: share of observations in cells

off the diagonal over time 214 9.3A Distributional characteristics of the contingency

tables for Germany and France 216 9.3B Distributional characteristics of the contingency

tables for the UK and Sweden 217 9.4A Descriptive statistics of characteristics of workers

in the skill–wage contingency tables 219 9.4B Descriptive statistics of characteristics of workers

in the skill–wage contingency tables 221 9.4C Descriptive statistics of characteristics of workers

in the skill–wage contingency tables 223 9.4D Descriptive statistics of characteristics of workers

in the skill–wage contingency tables 224 9.5 Model and fit statistics 228 9.5a Complete table with the raw data of the countries

used in the multivariate analysis 230 9.6 Estimates from the final fitted model number

(model 11) for prime age males in manufacturing 233 9.7 Estimates from the final fitted model number (model 11)

for prime age women in whole trade and services 236 A1 Comparison of association models 241 10.1 Employment structures and sectoral in

the EU15, 1980–2007 250 10.2 Changes in employment, EU27, by broad

sector and occupational groupings, 2008q2–2010q2 261 10.3 Patterns of employment change at national

level – comparison of pre- and post-crisis periods 268

List of Illustrations ix

Figures

1.1 The ranking of jobs and the allocation into quintiles 6 3.1 Patterns of employment expansion in Europe, 1995–2007 35 3.2 Contribution to job growth of the primary

sector and construction 41 3.3 Contribution to job growth of low- and

high-technology industries 43 3.4 Contribution to job growth of private services 45 3.5 Contribution to job growth of public services 46 3.6 A decomposition of the patterns of job growth

by employment status 47 4.1 Job growth in the United States by job-wage

quintile, 1995–2007 58 4.2 Job growth in the United States by gender, 1995–2007 61 4.3 Job growth in the United States by race

and nativity, 1995–2007 62 4.4 Job growth in the United States by nativity, 1995–2007 63 4.5 Contribution of manufacturing to job polarization,

1995–2007 64 4.6 Contribution of knowledge-intensive services to

job polarization, 1995–2007 65 4.7 Contribution of low-skill services to job polarization,

1995–2007 66 4.8 Contribution of construction to job polarization,

1995–2007 67 4.9 Job growth in the European Union, stacked by region,

1995–2007 69 4.10 Job growth in the United States by region, 1995–2007 69 4.11 Job growth in the United States in the construction

industry by region, 1995–2007 71 5.1 Job change by gender, full-time and

part-time, 1995–2007 79 5.2 Quintile distribution of jobs for women and men, 2007 82 5.3 Average wage and skill quintile measures for

women and men, 1995–2007 86 5.4 Country differences in the average wage–skill

quintile gap for men and women, 2007 87 5.5 Change in gender gaps in the average wage and

skill quintile measures, 1995–2007 89 5.6 Country patterns of low-wage job growth, by sex 93

x List of Illustrations

5.7 Comparing women’s low-wage job change with bottom quintile job change 95

5.8 Job concentration in the top-five and top-ten jobs, 2007 96 5.9 Women part-timers’ job concentration, top-ten jobs, 2007 97 5.10 The contribution of welfare state jobs to women’s

job change, 1995–2007 107 6.1 Immigrants as percentage of total population, by region,

1960–2005 115 6.2 Foreign population as percentage of total population

in EU countries, 2008 116 6.3 Time patterns of immigration: percentage of

foreign-born population in the EU27, 1960–2010 117 6.4 Distribution of male workers by educational level, 2006 123 6.5 Distribution of female workers by educational level, 2006 124 6.6 Employment rate among national and immigrants

in the EU, 2nd quarter, 2006 125 6.7 Unemployment rate among nationals and immigrants

in the EU, 2nd quarter, 2006 125 6.8 Distribution of total workers by job quintile in the EU,

percentage of each group, 2006 127 6.9 Distribution of male workers by job quintile in the EU,

percentage of each group, 2006 128 6.10 Distribution of female workers by job quintile in the EU,

percentage of each group, 2006 129 6.11 Job creation, destruction and immigration in the EU,

percentage of employment growth by job quintile 134 6.12 Job creation, destruction and immigration in the EU,

percentage of male employment growth by job quintile 135 6.13 Job creation, destruction and immigration in the EU,

percentage of female employment growth by job quintile 136 6.14 Over-qualification by migrant status and sex,

percentage of employed, 2006 140 6.15 Under-qualification by migrant status and sex,

percentage of employed, 2006 141 7.1 Job amenities index 160 7.2 Job amenities index by sector, occupation, gender

and legal status of employer broken down by four sub-dimensions 163

7.3 Job amenities index by country broken down by four sub-dimensions 164

List of Illustrations xi

7.4 Country employment distribution by job amenities quintiles 172

7.5 Net employment growth by quintile based on different job quality indicators 173

8.1 Level of employment in accession countries 184 8.2 Employment rate for the total working age population

and for the uneducated 184 8.3 Self-employment and industrial productivity

in the mid-1990s 186 8.4 Share of bottom quintile jobs in total employment

by level of employment 188 8.5 Level of the statutory minimum wage in

accession countries 191 8.6 Employment expansion in accession countries:

a mixed pattern 193 8.7 Absolute change in employment by sector and

wage quintile in Latvia and Lithuania 194 8.8 Employment expansion in the accession

countries: polarization 195 8.9 Absolute change in employment by sector and

wage quintile in Estonia and Hungary 196 8.10 Absolute change in employment by sector and wage

quintile in Slovakia, Slovenia and the Czech Republic 197 9.1 Design matrices showing which of the association

parameters of the models affect each cell of the table 210 9.2 Row and column scores estimated with the

log-multiplicative association model 5 231 10.1 Changes in the occupational structure of employment

in basic industries, EU15 256 10.2 Changes in the occupational structure of employment

in processing industries, EU15 257 10.3 Changes in the occupational structure of employment

in engineering industries, EU15 257 10.4 EU employment level shifts (000s) 2008q2–2010q2 –

actual, trend and predicted 263 10.5 Employment change by wage quintile, EU27,

2008q2–2010q2 267 10.6 Net change in employment (000s) by job–wage

quintile in manufacturing and construction, EU27, 2008q2–2010q2 270

xii List of Illustrations

10.7 Net change in employment in hi-tech and low-tech manufacturing by job–wage quintile, EU27, 2008q2–2010q2 271

10.8 Net change in employment in hi-tech and low-tech manufacturing by job–wage quintile, Germany, 2008q2–2010q2 273

10.9 Net change in employment in public and private services by job–wage quintile, EU27, 2008q2–2010q2 274

10.10 Net change in employment in knowledge-intensive/less-knowledge intensive services by job-wage quintile, EU27, 2008q2–2010q2 275

xiii

Preface and Acknowledgements

This book is the result of a long process, with the contribution of many people. It started with the JOBs project in the European Foundation for the Improvement of Living and Working Conditions (Eurofound) back in 2006. The project was initiated and led by Enrique Fernández-Macías with Donald Storrie and Branislav Mikulic in the original project team. Enrique, now on leave from Eurofound at the University of Salamanca, has been the main driving force throughout this long process. John Hurley joined the JOBs team in 2008 and is the other key person in progress towards publication of this book. Useful contributions by Jean-Marie Jungblut are also sincerely appreciated. Contributions from other Eurofound staff are gratefully acknowledged. Jan Vandamme has pro-vided his usual superb library services and thanks are also due to the then Head of the Working Conditions Unit, Agnès Parent-Thirion. We also gratefully appreciate the support of Jorma Karppinen, Director of Eurofound at the time, for both the project and the book initiative.

The JOBs project expert group met on three occasions, in Dublin, Paris and Salamanca. In addition to the authors of this book, Francis Green (Institute of Education, University of London), Mark Keese (Organisation for Economic Cooperation and Development) and Matias Scaglione (University of Wisconsin-Madison) provided much construc-tive comment and input during and between these meetings. We also appreciate the contribution of Maarten Goos (Catholic University of Leuven) at the Salamanca meeting.

We have received useful comments from two Eurofound Advisory Committees: Working Conditions and the European Monitoring Centre for Change (EMCC). Comments from two Eurofound evalua-tions of the JOBs project by Sangheon Lee (International Labour Office), Mark Keese, Christine Erhel (University of Paris 1) and Janine Leschke (European Trade Union Institute) are also gratefully acknowledged.

The task of collecting the original wage data was performed by Terry Ward (Applica) and Robert Stehrer (The Vienna Institute for Inter-national Economic Studies – wiiw). The project has benefited greatly from their intimate knowledge of EU data sets and good judgement calls in the difficult task of piecing together the various data sets to con-struct the wage ranking data in 23 European countries. We are of course highly indebted to our colleagues at Eurostat for making the relevant

xiv Preface and Acknowledgements

data sets available, for offering us guidance on data issues as well as valuable feedback on some of the more recent analysis.

Eurofound research on this topic does not end with this book. Other means of classifying jobs are being explored. Not least the possibility of applying multidimensional indicators of job quality, using Eurofound’s own European Working Conditions Survey. Moreover, Eurofound is now publishing regular annual updates of employment growth by wage quintiles. With each wave of the European Labour Force Survey, the European Jobs Monitor will provide very timely analyses of qualitative dimensions of employment growth in Europe.

The last few years have been turbulent indeed for European labour markets and so far the recovery is very unevenly spread across Europe. When one considers that the impact of the reorientation of the invest-ment strategies of companies in light of the recession has yet to become fully manifest in the employment data and the impending challenges to public sector employment, it is highly likely that, regardless of the pace of the recovery, we will continue to see rapid and significant shifts in the job structure in Europe. I am convinced that the European Jobs Monitor can provide policy makers with a robust and nuanced ring-side view as the structural transformation of European labour markets unfolds.

Donald Storrie Dublin, Ireland

xv

Contributors

José-Ignacio Antón is Lecturer of Economics at the University of Salamanca, Spain, with publications in the field of public economics, labour market, social policy and income distribution.

Rachel E. Dwyer is Assistant Professor of Sociology at Ohio State University. She studies rising inequality in American society in several social arenas, including housing segregation, neighbourhood inequality, youth indebtedness and the characteristics and determinants of employ-ment growth. Her recent publications include ‘Poverty, Prosperity, and Place: The Shape of Class Segregation in the US’, in the 2010 volume of Social Problems , ‘Expanding Homes and Increasing Inequalities’ in the 2007 volume of Social Problems and ‘The Patterns of Job Expansions in the USA: A Comparison of the 1960s and 1990s’, in the 2003 volume of Socio-Economic Review (with Erik Olin Wright).

Enrique Fernández-Macias is Lecturer in the Department of Sociology at the University of Salamanca, Spain, and Research Manager at Eurofound, Dublin, Ireland. His research interests lie in the fields of job quality, occupational structures and comparative labour market analysis.

Hugo Figueiredo is Lecturer in the Department of Social, Political and Territorial Sciences at the University of Aveiro and Research Associate at The Centre for Research in Higher Education Policies (CIPES), Portugal. Previously he worked at the European Work and Employment Research Centre (EWERC) at the Manchester Business School, University of Manchester. His research interests include higher education policies and the changing demand for graduate-level skills, gender and equal oppor-tunities and the comparative analysis of employment and innovation systems. His recent publications include the journal articles, ‘Towards a European Union Gender Equality Index’, Journal of European Social Policy , 19 (with J. Plantenga, C. Remery and M. Smith, 2009) and ‘How to Close the Gender Pay Gap in Europe’, Industrial Relations Journal , 36 (with J. Rubery and D. Grimshaw, 2005).

Damian Grimshaw is Professor of Employment Studies and Director of the European Work and Employment Research Centre (EWERC) at the Manchester Business School, University of Manchester. His

xvi Notes on Contributors

research interests include co-production and human resource man-agement, low-wage work in Europe, recruitment and retention in the UK care sector and gendered patterns of job growth in Europe. His recent publications include Fragmenting Work: Blurring Organizational Boundaries and Disordering Hierarchies (with M. Marchington, J. Rubery and H. Willmott, 2005), ‘Can Renewed Institutions Improve Low Wage Work?’, International Labour Review , 148 (2009) and ‘New HRM Practices in Knowledge-Intensive Service Firms: The Case of Outsourcing and Staff Transfer’, Human Relations , 62 (with M. Miozzo, 2009).

John Hurley is Research Manager in the Employment and Change Unit at Eurofound, Dublin, Ireland. His research interests lie in the area of comparative labour market analysis, restructuring and the changing world of work.

Jean-Marie Jungblut is a research officer in the Employment and Competitiveness unit of Eurofound, Dublin. He has a doctorate in Sociology from the University of Mannheim. He also worked for the Mannheim Centre for European Social Research (MZES) between 2002 and 2009. Earlier he worked for the Luxembourg Income Study and at the University in Trier where he graduated. His publications are in the areas of quality of life studies, vocational training and skill development.

Rafael Muñoz de Bustillo is Professor of Economics and Head of the Department of Applied Economics at the University of Salamanca, Spain. He has published extensively in the fields of labour market, social policy and income distribution.

Philip O’Connell is a programme coordinator of labour market research at the Economic and Social Research Institute, Dublin and the Head of the Social Research Division. He received his doctorate from Indiana University, Bloomington and taught at the University of North Carolina, Chapel Hill. He has served as a consultant on human resource development and labour market issues to the European Commission and the OECD. Much of his research focuses on education, training and the labour market.

Ágota Scharle is the co-founder and senior research partner of the Budapest Institute for Policy Analysis. She worked as the Head of the Economic Research division at the Hungarian Ministry of Finance. She has a D.Phil. in Economics from the University of Oxford. She has published extensively in the area of public policy and labour markets.

Notes on Contributors xvii

Robert Stehrer is Deputy Director of Research at the Vienna Institute for International Economic Studies (wiiw) and Lecturer of Economics at the University of Vienna. He participated in a number of international projects at the European level on economic and labour market issues. He has published articles and books in his main areas of research inter-est, including labour markets, productivity and trade structures and catch-up processes in the global economy.

Donald Storrie is Head of the Employment and Change Unit at Eurofound, Dublin, Ireland. He previously led the Centre for European Labour Market Studies at the University of Gothenburg, Sweden. He is a labour economist with particular interest in employment con-tracts, employment and health, and active labour market and European employment policy.

Terry Ward is a Research economist who, as Director of Alphametrics in the UK and Applica in Brussels, has undertaken a great many studies for the European institutions on employment, social policy and regional development. He was the main contributor to the annual Employment in Europe report throughout the 1990s and has edited and contributed to all five Economic and Social Cohesion Reports published since 1996. Before becoming an independent consultant, he was, for many years, senior research officer in the Department of Applied Economics, University of Cambridge, working on a range of economic policy issues.

Erik Olin Wright is Vilas Distinguished Professor of Sociology at the University of Wisconsin. His academic work has been centrally con-cerned with reconstructing the Marxist tradition of social theory and research in ways that attempt to make it more relevant to contempo-rary concerns and more cogent as a scientific framework of analysis. His empirical research has focused especially on the changing character of class relations in developed capitalist societies. Since 1992 he has directed The Real Utopias Project which explores a range of proposals for new institutional designs that embody emancipatory ideals and yet are attentive to issues of pragmatic feasibility. His important works include Class, Crisis and the State (1978), Classes (1985), The Debate on Classes (1990), Reconstructing Marxism: Essays on Explanation and the Theory of History (with Elliott Sober and Andrew Levine, 1992), Interrogating Inequality (1994), Class Counts: Comparative Studies in Class Analysis (1997), Deepening Democracy: Institutional Innovations in Empowered Participatory Governance (with Archon Fung, 2003), Envisioning Real Utopias (2010), American Society: How It Really Works (with Joel Rogers, 2010).

1

Despite recent shifts in global economic influence, much of the academic and policy debate on labour markets throughout the world still focuses on the relative merits of the institutions and the perfor-mance of labour markets in Europe and the United States of America. The European Union’s grand Lisbon Agenda with its commitment to ‘more and better jobs’ was often explicitly motivated by failure of Europe to generate jobs in the same numbers as the United States. 1 Certainly the US performance in the 1990s had been highly impres-sive. In that decade, it experienced its longest period of sustained economic growth in the 20th century. Unemployment fell to among the lowest levels in the OECD, and the employment rate peaked at an all-time historical high. There was no doubt that the ‘American Jobs Machine’ had been remarkably successful in creating more jobs, 20 million between 1991 and 2000. But were they better jobs? Some argued that most were low-paid, dead-end jobs in services, while the well-paid jobs in manufacturing were being destroyed by the combined forces of globalization and technical change. Others argued that, on the contrary, this unprecedented employment expansion was associ-ated with the creation of jobs with higher-than-average skill and pay levels, especially in professional and managerial occupations.

One reason for these different views was that their exponents used different types of evidence to support their views. The gloomy perspec-tive on the American Jobs Machine was often based on the evolution of wage inequalities, which clearly increased during this period. The rosy perspective was often sustained by broad analyses of the sectors and occupations that had expanded most over this period, which suggested that the majority of new jobs were of a relatively good quality (manage-rial and professional occupations, in sectors such as financial services

1 Introduction Donald Storrie, John Hurley and Enrique Fernández-Macías

2 Donald Storrie, John Hurley and Enrique Fernández-Macías

and information technology). But, surprisingly, most of this evidence was only indirectly related to the key issue in this debate: the actual quality of the jobs created. The evolution of wage inequality measured at the individual level is obviously related to the type of jobs created over the expansion, but it simply could not provide a clear answer to the underlying question of how different patterns of employment expan-sion by job quality levels could lead to similar outcomes in terms of wage inequalities. Moreover, the analysis of changes in the levels of employment by broadly defined sectors and occupations was too aggre-gated to provide a clear answer.

A report of the US Council of Economic Advisers (then chaired by Joseph E. Stiglitz), in 1996, provided a simple yet powerful method for evaluating more directly the quality of the jobs created during the expansion of 1994–1996. This method consisted in the sorting of all full-time workers into 45 detailed occupations within 22 industries, and then calculating the median weekly earnings of each of the result-ing economic sector and occupation cells. These cells defined a job. Jobs were then ranked by their median wage levels, and split into two groups, each holding half of total employment in 1994. The group that contained the sector and occupation cells with wages below the median were considered as holding ‘bad jobs’; the others were deemed as hold-ing ‘good jobs’. The report found that 68 per cent of the total net expan-sion of employment had taken place in the ‘good jobs’.

Erik Olin Wright and Rachel Dwyer revisited the same question with a similar jobs methodology as used in the Stiglitz report, but with a more sophisticated analysis (Wright and Dwyer 2000, 2003). They exploited more detailed sectors and occupations and extended the time span. Moreover, and perhaps most importantly, they used five job qual-ity quintiles. This revealed a much more nuanced picture of the evolu-tion of job quality than the dichotomy of just good and bad jobs. They found that between 1992 and 2000 job expansion was strongest at the top of the wage structure, followed by the bottom groups and with the middle either declining or stagnant. Thus while on average good jobs were in a majority, the distribution of jobs was sharply skewed to the high and low tails of the wage distribution. Another interesting result was the contrast to the previous periods of employment expansion in the 1960s and 1970s which was more unambiguously towards good jobs (‘upgrading’) with hardly any sign of simultaneous growth at the lower tail (‘polarization’).

Meanwhile in Europe towards the end of the 1990s, and especially after the turn of the millennium, employment growth had picked up

Introduction 3

appreciably. Between 2000 and 2007 employment growth in the European Union was marginally higher than in the United States, and for the more advanced economies (EU15) it was appreciably higher. In fact over the longer period between the aftermath of the recession of the early 1990s and the Great Recession (1995–2007), employment growth in the EU15 was 1.24 per cent per annum compared to 1.23 per cent in the United States. This simple statistic is important as it debunks the widely held misconception of sluggish European employ-ment growth relative to the United States. 2 Developments between 2007 and 2010 have further improved European performance vis-à-vis the United States. However, as had previously happened in the US and in light of the central slogan of the Lisbon Agenda of ‘more and better jobs’, the question then arose about the quality of the resurgent employment growth in Europe. To this end the European Foundation for the Improvement of Living and Working Conditions (Eurofound) applied the jobs methodology developed by Wright and Dwyer to the 23 EU member states for which comparable data could be obtained (Fernández-Macías and Hurley 2008).

The main empirical challenge of this Eurofound JOBs project was to obtain robust rankings of wages for each European country. Several European level data sources were used to construct these rankings: the European Structure of Earnings Survey (ESES), the European Survey on Income and Living Conditions (EU-SILC), the European Community Household Panel (ECHP) and the Structural Business Statistics (SBS) of the EU. This data, as well as the employment data of the European Labour Force Survey from 1995 to 2007, constitute the common empir-ical backbone for the European chapters in this book. However, the very significant changes of the Nomenclature statistique des activités économiques dans la Communauté européenne (Statistical classifica-tion of economic activities in the European Community) (NACE) sec-tor codes introduced by Eurostat in 2008 meant that the original wage rankings could not be used to study job growth after 2007. However, Eurostat were able to provide European Labour Force Survey (EU-LFS) wage data for 2008. This was used to generate a new 2008-based wage ranking and thus permitted a preliminary analysis of the evolution of employment between 2008 and 2010 (Hurley et al. 2011). This data is also used in the postscript final chapter ( Chapter 10 ) which gives some indication of how the Great Recession has impacted the job structure in Europe. However, before introducing the various chapters it is appro-priate to briefly outline the essence of the jobs methodology. It is developed in much more detail in Chapter 2 .


1.1 The jobs approach

The jobs approach focuses not on the grand aggregates of the ILO sta-tistical definitions of the labour force but, obviously, on jobs. A job is defined as an occupation in a sector. This is an intuitively attractive definition and corresponds to what people think of when describing their job – a secretary at a hospital, a salesperson in a car showroom or a chemist in the food processing industry. This definition is also conceptually very useful as occupation and sector relate to the two fun-damental dimensions of structural change. The sector describes what type of economic value is being created and the structure of occupa-tions gives some indication of how this value is being created. They have some correspondence to the division of labour within and outside the firm. A job so defined has also very practical empirical advantages as both occupation and sector codes have some degree of standardi-zation which is obviously convenient for comparative analysis. Also policy-makers throughout the world have recognized the value of fram-ing employment policy and analysis in terms of jobs. The influential OECD Jobs Study was perhaps the pioneer in this respect, but increas-ingly EU employment policy is phrased in the more concrete concept of a job. ‘More and better jobs’ was the headline phrase of the Lisbon Agenda and ‘New skills and jobs’ is central to its successor, the Europe 2020 strategy.

The jobs approach entails the definition of a job in an intuitive, con-ceptually coherent and empirically practical way and also requires some means of evaluating or classifying these jobs which is analytically use-ful. The original US research used the median wage. Wage is of course a useful classification of jobs in itself, not least in terms of the vital con-tribution of wages for the purchasing power of workers. Moreover, as productivity differentials are a fundamental driver of structural change, the association of wage and labour productivity is particularly relevant in the study of the long-run structural transformations of labour mar-kets that constitute a primary focus in this book.

The US research, however, interpreted wage as a proxy for quality of work (Levy and Murnane 1992; Ilg and Haugen 2000; Wright and Dwyer 2003). Apart from the important intrinsic monetary value of the wage as a dimension of quality of work, one very practical advantage is that wage income is more measurable, generally more widely avail-able and more comparable between different countries and data sets than most other dimensions of job quality. Furthermore, the wage is likely to correlate with other dimensions of work. Wages are certainly

Introduction 5

strongly related to other payments such as pensions and social-security benefits. Moreover, the status of a job in society is related to wage, as are other job characteristics such as autonomy, cognitive richness and job security. Wright and Dwyer (2003) conclude that wages are a ‘sufficient salient aspect of job quality’ to be used as a proxy even if the concept itself is multidimensional.

One might argue that in the more regulated European labour mar-kets wages may be less strongly correlated with a properly defined mul-tidimensional measure of quality of work than in the United States. However, recent European research on job or work quality has acknowl-edged the importance of the wage dimension. According to Leschke and Watt (2008), ‘wages are arguably the most important field in regard to job quality’. Work by Davoine et al. (2008b) also bears out the significance of wage in measures of overall job quality. However, it must be made clear that wage is only a proxy of quality. It may be the best one but any single indicator will be always imperfect. Chapter 7 addresses this issue, in a preliminary way, by analysing to what extent the results obtained using wage as an indicator of job quality are mir-rored by those obtained using other elements of job quality. Ongoing work at Eurofound is exploring other multidimensional indicators of job quality.

The essentials of the approach can be explained with the aid of Figure 1.1 . Jobs defined as an occupation in sector are first ranked according to median wage. This ranking is carried out once and remains fixed throughout. The jobs are then assigned to quintiles of employment, that is each accounting for 20 per cent of total employ-ment. The difference between the initial and final employment levels is then presented in the quintile charts showing how employment growth or change over the period has been distributed among the job–wage quintiles.

1.2 Outline of the book

The data compiled in the Eurofound JOBs project allows for a much more detailed analysis than performed in the hitherto, largely descrip-tive, Eurofound reports. Even if the scope of topics addressed in this book is very wide, the analysis is all based on a common or closely com-patible data sets. The methods used in the chapters are also essentially the same and the results can be presented in a relatively simple fashion. We expect that this provides both coherence and clarity throughout the book.


After a more detailed methodological outline of the jobs approach in Chapter 2 , Chapter 3 provides a detailed outline of the main empir-ical findings of the original Eurofound project (Fernández-Macías and Hurley 2008), updated to 2007, and expanded with a discussion of its theoretical foundations by Enrique Fernández-Macías. It argues that the classification of employment into jobs which underlies this approach can be understood as a stylized representation of the state of the divi-sion of labour in a specific moment of time. Hence, the results can be framed in the old Social Sciences debate of the implications of the changing division of labour for job quality, which goes back to Adam Smith and Karl Marx. After discussing its theoretical foundations, the author applies the approach to the long and virtually unbroken period of employment expansion in the EU between 1995 and 2007 in order to provide a qualitative assessment of the over 20 million new jobs created in net terms during this period. The chapter also introduces the distinc-tive graphic presentation of the empirical results – in the shape of job quality quintiles – that gives the jobs approach its capacity to convey so much information in a simple, intuitive format.

The main result of this chapter is that in the period studied there was a considerable variation of patterns of employment expansion across

Occupation

Corporate managers

Other professionals

Skilled agriculturel/fishery workers

Sales/services elementary occupations

Sales/services elementary occupations

Craft workers

Life science and health professsionals

Teaching professionals

Rank sector

1 Financial services

2 Legal/accounting

3 Education

4 Human health activities

.......

…..

1105Agriculture

1106Services to buildings

1107Education

1108Food manufacture

Highest 20%Mid-high 20%Middle 20%Mid-low 20%Lowest 20%

Figure 1.1 The ranking of jobs and the allocation into quintiles

Introduction 7

European countries, although most of them could be categorized in relation to two dominant patterns. The first (and most frequent) was a pattern of upgrading where net employment gains were more or less monotonically skewed to the top job quality quintiles; the second was a polarization pattern where growth was greater in both the top and the bottom quintiles compared to the middle quintiles. In fact, the top of the employment structure (the good jobs) expanded similarly everywhere, whereas most of the differences lay in the bottom (with some countries creating many bad jobs and others not creating bad jobs at all, or even destroying them in net terms) and to a lesser extent in the middle (with just a few countries creating middling jobs, and the majority remaining relatively stagnant in that segment of employment). This contrasts with some of the previous findings for the US and some European coun-tries, which tended to emphasize polarization. The chapter discusses the occupation and sector dynamics which are behind these patterns in the context of the Social Sciences debates on the implications for job quality of compositional changes in the division of labour.

As already indicated, the analytical approach adopted in this book owes much to the pioneering work of American sociologists Erik Olin Wright and Rachel Dwyer in their analysis of the Great American Jobs Machine of the 1990s (Wright and Dwyer 2003). Indeed one of the key motivations to repeating the analysis for the EU was to see the extent to which qualitative shifts in the EU employment structure in the period up to 2007 mimicked or diverged from the pattern of asymmet-rical polarization previously identified for the earlier but comparable period of employment expansion in the US. In Chapter 4 , Rachel E. Dwyer and Erik Olin Wright update their previous analysis and make an explicit comparison between the patterns in the US and in the EU as a whole. They find that the employment expansions in both the EU and US shared some common features which were, in order of impor-tance, higher growth in high-paid jobs, relatively lowest growth in middle-paid jobs and intermediate levels of growth in lower-paid jobs. However, some differentiation in the degree of asymmetrical polariza-tion was observed. The ‘declining middle’ or the ‘hollowing out’ of the labour market appears to have been somewhat sharper in the US than in the EU, probably related to more rapid deindustrialization and terti-arization in the US.

Dwyer and Wright draw attention to one of the obvious advantages of extending the jobs approach to the EU: it opens up possibilities for a more institutionalist framework of analysis. The availability of data from the US as well as 23 EU member states with in many cases quite


different sectoral employment compositions, distinctive forms of labour market and social welfare regulation as well as worker representation sys-tems may provide evidence of additional explanatory factors for differ-ing patterns of employment change complementing those which focus more or less exclusively on technological change. The authors point to the diversity of employment change patterns within the US where faster-growing regions such as the south and west have had less polar-ized growth compared to the older industrial regions in the Northeast and Midwest. This regional pattern in the US offers some similarities to that in the EU between core, original member states – Germany, France, the Netherlands – where polarization was stronger than in peripheral member states – Spain, Ireland – which enjoyed higher levels of growth during the 1990s and 2000s.

Over the period 1995–2007, employment growth in the EU was greater for women than for men, narrowing the gender employment gap. This is the latest episode of the generational labour-market revo-lution that has seen the share of female workers in national workforces approach and, in some cases, surpass that of men’s. As authors Damian Grimshaw and Hugo Figueiredo point out in Chapter 5 , the gender gap in terms of job quality also narrowed in all but one of the twenty three EU member states covered (Spain) as women took a higher share of well-paid new jobs than men. However, the patterns of employment growth for women display the same heterogeneity across countries as that for all workers. The authors also point out that women in many countries experienced job upgrading in terms of skills (based on highest level of education attained), but polarization in terms of wage (i.e. a net expan-sion of both the highest and lowest paid jobs, with stagnation in the middle). This supports the notion that many skilled jobs where wom-en’s employment has grown are in fact undervalued in terms of wage compared to jobs which are still dominantly male.

In addition to the narrowing gender gaps, the authors point to the level of concentration of female employment in a small number of jobs; developments in just ten jobs go a long way to explaining the shifts in women’s employment structure in most member states. Many of these jobs are ‘welfare state jobs’ – in education, health and social care – where female workers account for between 64 and 84 per cent of employment depending on country. One intriguing finding is that the significant growth in welfare state jobs has been a strong contributor to upgrading of women’s employment in most countries but has had a more polarizing influence in the liberal welfare regime countries such as the UK and Ireland.

Introduction 9

In Chapter 6 , Rafael Muñoz de Bustillo and José-Ignacio Antón use the jobs approach to analyse another dimension of labour-market seg-mentation: that separating native from immigrant or non-national workers. They highlight the fact that the current migrant flows in the developed world are low compared to those that occurred dur-ing the first great wave of globalization in the 40 years that preceded World War I. Nonetheless, in the period 1995–2007 Europe received a growing share of immigrants who now account for nearly one in ten of the population. Patterns of recent migration intra-EU have been very different from those that occurred in earlier decades. Flows into traditional host European countries such as Belgium and France have stabilized while former net emigration countries such as Spain and Ireland have faced sudden and intense periods of immigration since the late 1990s.

Taking advantage of the new perspective brought by the jobs approach, the authors confirm that immigrant workers tend to concen-trate in lower-paid jobs though with varying levels of intensity across countries; this pattern of segregation is much less marked in the UK and Belgium, for example, than in Germany, Spain or Italy. There is also evi-dence that this concentration is not solely explained by lower levels of human capital in the migrant population. The higher incidence of over-qualification amongst migrant compared to native workers shows that there is a significant untapped potential which may be underutilized for reasons of insufficient integration, poorer job-matching dynamics or discrimination.

The main proxy of job quality used in our approach has been the job wage, and a secondary proxy has been the average highest educational level attained by the job holders. But clearly job quality is not reducible to just wages or the educational background of individual workers. It has many other dimensions – job security, non-exposure to health risks, cognitive richness, autonomy, development opportunities and so on – which should also be taken into account. In Chapter 7 , John Hurley, Enrique Fernández-Macías and Rafael Muñoz de Bustillo develop a syn-thetic measure of job quality anchored in the Social Sciences literature and made operational using the European Working Conditions survey (2005) data set. The main result of this exercise is that the overall pat-terns of employment expansion by job quality in Europe do not change significantly using such multidimensional job quality index or using the wage- or skill-based measures. How we measure job quality does not appear to impact substantially on the observed patterns of employment expansion, as jobs in the same sectors – financial intermediation at the


top end and hotels/restaurants at the bottom – and in the same occupa-tions within those sectors – professionals, associate professionals and managers at the top end and elementary occupations at the bottom – tend to be similarly placed irrespective of whether we use a composite job quality measure, a wage-based measure or a skill-based measure.

One of the findings of previous analysis of European labour markets using the jobs approach (Fernández-Macías and Hurley 2008) was that the patterns of employment expansion were quite different in ‘older’ EU member states (those that joined before 1992) compared to those that were part of the major accession of the mid-2000s. Employment growth was in particular much less skewed towards higher-paid jobs in the new member states than the older member states. That the two groups of countries should exhibit different patterns is unsurprising. The experi-ence of transition post-1989 from a planned, command economy to a market economy was particularly brutal in many of the countries in the sphere of influence of the former Soviet Union and the period covered by the analysis (1995–2006) was close enough to still bear the marks of these traumas which included huge decreases in employment levels (of up to 30% in Hungary, Estonia and Latvia). The process of convergence with west-European economic and political models continued through-out the period, beyond accession in 2004, and is ongoing at the time of writing.

In Chapter 8 , Ágota Scharle sets labour-market developments in seven of the new member states against the backdrop of the transi-tion. Though somewhat poorer, these economies were roughly at the same stage of industrial development as the southern-European econ-omies had been in the mid-1990s. They also experienced similar bouts of sectoral reallocation of labour in the period prior to and after EU accession in 2004. Nonetheless, patterns of employment shift by job quality quintile varied from country to country possibly related to dif-ferent policy responses to the transitional shock. Evidence is suggest-ive of a trade-off between employment growth and job quality. Faster employment growth was associated with increases in lower-quality jobs (e.g. in Estonia) while qualitatively upgrading countries (e.g. the Czech Republic) recorded negative employment growth in the same period (the decade up to 2007).

Generally, higher-paid jobs are also jobs that require higher-level skills. As a result, the assignment of jobs to job quality quintiles across countries shows a high level of correlation for these two types of job ranking. Such ‘returns to education’ are usually interpreted in terms of increased levels of productivity of highly educated workers compared to

Introduction 11

workers with only compulsory or basic schooling. Discrepancies in the match between job-skill levels and job wages do exist, however. These are of interest in particular in identifying situations in which the labour market behaves in ways that depart from the predictions of classical eco-nomics; that is, where workers are paid more or less than their marginal productivity. In Chapter 9 , Jean-Marie Jungblut and Philip O’Connell outline various mechanisms that could lead to different wage levels for workers of similar skill levels.

Internal labour markets – characterized by long-term employment relationships with a strong focus on human capital development – may tend to underpay younger or entry-level workers on the basis that they will recoup and realize these ‘investments’ in the shape of seniority premia later in their career in the same company or group. Employees in such jobs accept a trade-off over time in the wage returns to edu-cation. The theory of segmented or dual labour markets on the other hand assumes a reserve pool of flexible labour where the returns to education may be low or non-existent. The fact of belonging to the less favoured, outsider or secondary labour market can consign even highly qualified workers to a succession of low-status and prospect-less jobs. The jobs approach can be used for comparing relative skill and wage levels for jobs using the quintile assignments. The authors take advan-tage of this and use multivariate techniques to investigate the extent to which job-skills mismatches are attributable to different types of labour market – internal, occupational or dual/segmented – and how such mis-matches vary by gender and across countries as well as according to other institutional variables.

The tenth and final chapter examines the impacts of recessions on the labour market going back to the early 1980s. It does so in the wake of the Great Recession of 2008–2009 in which financial crisis contrib-uted to destabilize the economies of the developed world more thor-oughly than at any time since the Great Depression of the 1930s. Instead of increasing by six million as trend growth would have pre-dicted, European labour markets shed five million jobs between the second quarter of 2008 and the second quarter of 2010. Terry Ward, Robert Stehrer and John Hurley compare this recent period of employ-ment decline with those that occurred in earlier recessions in particular as regards the relative employment impact by sector and occupation. They find, for example, that secular trends towards increasing shares of services employment have accelerated during periods of recession. This in turn is associated with a sharp divergence in employment growth between higher-skilled, white-collar occupations and blue-collar


occupations (both high and low skilled) where workers in the former category are relatively cushioned from the impacts of recession while those in the latter blue-collar category are disproportionately affected. Because of the location of these jobs in the (lower-) middle of the wage distribution, the impact of the Great Recession has been to sharpen the previously observed asymmetrical polarization of employment growth. Employment grew, for example, by nearly one million in top-quintile jobs in the EU during the recession.

This last chapter is something of a coda to the earlier chapters which concentrate on the preceding period of employment expansion from 1995 to 2007. One principal conclusion is that a wrenching recession such as the one that occurred in 2008–2009 while impacting very neg-atively on the quantity of jobs has broadly similar impacts in terms of the relative distribution of well-paid and medium-paid jobs. The distinctive effect of the recession was simultaneously to sharpen the relative growth of well-paid, well-skilled jobs and the relative decline of medium-paid jobs. Specific sectoral effects – the large-scale destruc-tion of construction jobs, resilience and growth of higher-grade pub-lic sector employment in health and education – have contributed to this accentuation of asymmetrical polarization during the recession. Ongoing austerity measures have nevertheless considerably diminished the durability of these higher-grade public sector jobs.

Perhaps the primary empirical value of the contributions to this book is the convincing evidence of the considerable variation in the patterns of employment growth by type of job throughout Europe and the United States. This is in contrast to a more homogeneous pic-ture painted by others for many European countries, for example in Goos et al. (2009). In Chapter 3 , the reasons behind this empirical dis-crepancy are discussed in some detail. It should be noted that other recent research has found similar heterogeneity in Europe (Oesch and Rodriguez Menes 2011) as is presented in this book. Nevertheless, many countries do show some degree of polarization and it occurs in the aggregate figures for both the United States and Europe. Moreover, it could be argued that polarization would have been even stronger had it not been for the extraordinary construction boom which held up the middle in many countries (see Chapter 3 ).

The most credible explanation of the polarization phenomenon is the theory of ‘task biased technological change’. Autor et al. (2003) argue that the relevant way of understanding how technology replaces vari-ous types of labour is not primarily by analysing the skill content of a job but the extent to which tasks can be made routine. Technology, and

Introduction 13

currently not least digital technology, can quite readily replace labour in routine tasks; that is, in tasks that can be codified into repetitive step-by-step procedures. Goos and Manning (2007) have shown for the UK that non-routine tasks tend to be concentrated at the two extremes of the skill and wage distribution. Thus the jobs containing tasks in the middle decline due to the introduction of labour productivity enhanc-ing technological change for these routine tasks. Thus while the previ-ously dominant skill-biased technological change hypothesis, reviewed in Katz and Autor (1999), predicted lowest growth in low-paid jobs, the recognition that many low-skilled jobs cannot be readily made routine leads to a polarized pattern of employment growth in developed econ-omies. However, it must be noted that the standard skill-biased tech-nological change hypothesis still has some explanatory staying power in that the results of this book, and elsewhere, show strong growth at the top in most countries, and certainly this was a very striking feature of the aggregate EU-level job growth. The middle jobs may not just be replaced by technology and disappear; they may also be offshored (i.e. existing jobs moved from high- to low-cost countries) or simply elim-inated by the competitive effects of international trade. Many of the lower-paid jobs remain as they entail non-routine tasks that are less tradable. In the trade context others focus more on the personal nature of these tasks and the fact that they require face-to-face interaction rather than routinization per se (Blinder 2006).

The differences between countries are more difficult to explain and the authors in this volume have been appropriately cautious in inter-preting the reasons for the heterogeneity. There is no lack of possible explanations. Obviously the different levels of economic development as reflected by the employment structure and wage levels would lead to different exposures to the technological and international compet-itive pressures outlined above. It is, however, interesting to observe that the countries differ mainly in the bottom wage quintile(s). This suggests that institutional features such as social-security systems, wage- determination mechanisms, minimum-wage regulation and public-sector employment are important in this respect. Indeed there is some suggestion that a deregulation of labour markets occurred in some of the more polarized countries. There is also some evidence that immigration flows impact on the variation in employment growth at the bottom end of the distribution.

Economic and social policy will not only determine growth at the bottom end of the wage distributions in different countries. Not least the experiences of the Great Recession show the negative impact of lax


financial regulation and practices (Ireland) and what happens to jobs with low employment protection (Spain). Moreover, countries that do promote skill formation, research and development, education and other social investments (Germany and Sweden) will succeed in grow-ing the top jobs even in a competitive global economy.

The evidence suggesting that institutions do affect the patterns of employment growth (and hence the observed diversity across coun-tries) has a very positive reading in terms of employment policies. For if institutions matter, there is nothing inevitable in the implications for job quality of the unrelenting march of technological change. Even if on its own, the current wave of technological change has a polarizing impact, institutional arrangements such as vocational training, collective bargaining coverage or employment protection can minimize such an impact or even reorient it in more socially desirable ways.

If polarization were to become a more widespread and deeper phe-nomenon it raises a number of real policy concerns. The middle clas-ses and the well-paid working class constitute the political basis for the welfare state of both the universal and the corporatist variety common in many European countries. 3 Furthermore, there are many negative welfare implications of inequalities, not least in terms of physical and psychological well-being (Wilkinson 2006). A disappear-ing middle may also undermine the possibility of stepwise social and occupational mobility. However, it is hard to imagine how the exist-ing middle-paid jobs in manufacturing and routine services can be protected in high-wage countries in a competitive global world. Are there other jobs that can fill the gap created by the shrinking middle? In the short-medium run, construction will recover but hardly back to the level of the years before the recent property and financial crash. In the longer term one can only speculate. For example, is it likely that the jobs that will emerge from the adaptation to global warming fill the middle?

Is it inevitable that personal and other service jobs are destined to remain at the bottom of the wage distribution? Could technological innovation boost productivity in many of the currently non-routine occupations? While it may be highly unlikely that people would agree to a personal care robot or a digital hairdresser, it is exceedingly dif-ficult to predict the innovations emerging from the dazzling pace of technological progress.

Introduction 15

Notes

1 . The Lisbon Strategy was an action and development plan for the economy of the European Union between 2000 and 2010. Its aim was to make the EU ‘the most competitive and dynamic knowledge-based economy in the world capable of sustainable economic growth with more and better jobs and greater social cohesion’, by 2010. It was set out by the European Council in Lisbon in March 2000.

2 . The employment growth data are from the EU KLEMS data set. 3 . The Social Democratic and Conservative welfare states in the terminology of

Esping-Andersen (1990).

16

The jobs approach is, above all, a methodology for studying the evolu-tion of the employment structures of advanced capitalist economies from a job-quality perspective. The main innovation and the methodo-logical foundation of this approach is the use of jobs (defined as specific occupations within specific sectors) as the unit of analysis, rather than individuals. This implies a structural view of the labour market, which (as will be argued in Chapter 3 ) can be linked to the classical Social Sciences concept of the division of labour. The other methodological foundation of this approach (which is not so innovative in itself) is the use of a relative (or positional) measure of job quality to characterize the jobs in each economy and to analyse the evolution of employment structures through such a lens.

One of the main advantages of this approach is that the methodo-logical ideas behind it are relatively simple and clear, which makes its results much easier to understand and communicate. But the opera-tionalization of these ideas for the comparative study of the evolu-tion of job quality across 23 different countries during a period of 13 years is understandably complex, as it involves dealing with data sets which were not designed for this purpose (and therefore require many adjustments which involve some difficult decisions). In this section, we will provide a detailed and step-by-step explanation of this process. 1

2.1 The jobs matrix

The methodological pillar of this study 2 is the construction of a jobs matrix in each one of the 23 countries analysed together with the col-lection of information on the number of workers in each job (defined

2 Methodology Enrique Fernández-Macías, Terry Ward and Robert Stehrer

Methodology 17

crossing occupation and sector at the two-digit level) for each year between 1995 and 2007.

The data source used for the construction of this matrix is the European Labour Force Survey (ELFS from now on), maintained by Eurostat. In fact, there is no European Labour Force Survey as such, but as many Labour Force Surveys as there are member countries of the EU (and some more, under specific agreements), each of them carried out by the respective national statistical office under the coordination of Eurostat on the basis of common criteria and definitions (see Eurostat, 2009). The ELFS data are representative of the population living in private households in each country: for our analysis, we will use a sub-sample of the ELFS covering the employed population (including the self-employed) in the 23 countries shown in Table 2.1 . The ELFS is a quarterly survey, but we will use the annual figures provided by Eurostat. All the papers in this book use the data grossed up to the total figures of the employed population in each country (based on the latest Eurostat estimations).

The standardized variables by sector of activity and occupation are crucial for our analysis: they provide the link with not just the wider concept of the division of labour, as will be argued in Chapter 3 , but also the basis for comparability across 23 different countries and for linking quantitative and qualitative information on each job in a single data matrix. The classifications used are the General Industrial Classification of Economic Activities in the European Communities (known as NACE), revision 1.1, and the International Standard Classification of Occupations (ISCO), version 88-COM.

Table 2.1 Number of jobs identified in each country

Germany 1,192 Belgium 686 Netherlands 891

Ireland 980 Cyprus 433 Slovenia 730

Finland 790 France 1,067 Denmark 610

Sweden 698 Italy 939 Czech Republic 986

Luxemburg 454 Austria 759 Slovakia 821

UK 1,079 Latvia 638 Greece 698

Hungary 977 Estonia 401 Lithuania 539

Portugal 781 Spain 988

18 Enrique Fernández-Macías, Terry Ward and Robert Stehrer

The NACE classification is based on three criteria: (1) the character of the goods and services produced; (2) the uses to which they are put; and (3) the inputs, processes and technology of production (UN 2002: 14). Although the concept is not explicitly mentioned in the available documentation, it is quite obvious that NACE is a classification of eco-nomic activities in terms of the horizontal division of labour. NACE has a nested structure, with different levels of aggregation. For the pur-pose here, a level was needed, which is detailed enough, for the cate-gories to be relatively homogeneous in terms of job content but not so detailed as to compromise the international comparability of the results. For these reasons, the two-digit level of NACE, which contains 60 divisions, was chosen.

ISCO, on the other hand, classifies workers according to the nature of the work they do. In fact, the main criterion for classification is the skills required for competent performance of each job, differentiating two dimensions: (1) the level of skills, distinguishing the range and com-plexity of tasks; and (2) the specialization of skills, referring to the field of knowledge associated, the tools, machines and materials used and the type of goods and services produced (Elias 1997: 6). Although according to its documentation ISCO is concerned with skills and not with author-ity or hierarchy, a simple glance at the one-digit level of this classification shows that authority is certainly there: the differentiation of a specific category for managers and administrators (category 1) can be justified only in terms of authority, as it is not linked with any formal skill level or specialization; and furthermore, the levels of ISCO are clearly associ-ated with the typical hierarchical structure of productive organizations (which should not be a surprise, because authority and skills are strongly associated in modern organizations). ISCO, therefore, can be used as a classification of workers along the vertical dimension of the division of labour. As with NACE, ISCO has a nested structure, with different levels of aggregation. The international comparability of ISCO is more problem-atic than that of NACE, because its classification criteria are more loosely defined: previous studies (Elias 1997: 11) recommend using the two-digit level for international comparisons, which is the level used here. In total, there are 28 occupations at the ISCO two-digit level.

Crossing 60 sectors and 28 occupations gives a total of 1680 potential jobs in the matrix. In practice, many of these jobs do not exist, or they account for such a small number of workers that the ELFS samples do not contain respondents for them. Depending on the size of each coun-try, the number of jobs identified can vary considerably (as shown in Table 2.1 : from 401 jobs in Estonia to 1192 in Germany).

Methodology 19

2.2 The job quality rankings

The second key element of our approach is the construction of an ordi-nal indicator of job quality, which enables each job in the matrix to be positioned against all the rest in terms of its quality, country by coun-try. As mentioned earlier, previous studies used a single indicator based on wage levels for this purpose: to this, we will add a second indicator based on the average educational level of workers, in order to add rich-ness and robustness to the analysis. 3

Both indicators are used in a particular manner in our analysis, not only because they are used for characterizing jobs rather than people, but also because they are static, relative and approximate measures of job quality:

● Static because we use a single ranking of jobs in terms of their wages or average educational levels for the whole period studied (1995–2007). Effectively, there is no base year for the rankings: as we will see, the wage ranking is based on different sources carried out in dif-ferent years (in 2000, 2002 and 2005), while the educational rank-ing is based on a weighted average of the whole period 1995–2007. This may mean that our characterization of job quality over time is affected by some minor bias since the relative position of jobs tends to change over time in terms of their quality, but this change is likely to be very small over a period of 13 years, which is the one covered here. 4 Moreover, this approach helps to focus our analysis on the nature of structural change, which is the real interest here. In other words, by fixing the job-quality characterization of jobs, all the change observed in the figures will result from structural changes in employment, as we will see later. ● Relative because the indicators of wages or education levels are used only for assigning a rank position to each job with respect to the rest. While this takes no account of the enormous disparities in absolute terms between jobs and countries, both in wages and education lev-els, it helps to focus attention on the change in the structure of jobs. The use of relative rather than absolute measures, and more gener-ally the use of a non-parametric approach, makes the analysis more robust and greatly facilitates international comparisons. ● Approximate because the indicators bear only a partial (in the case of wages) or indirect (in the case of education) relationship to the concept that they are intended to measure. In the following chapter, we will provide some theoretical justification for this approximation


to job quality. From a more applied perspective, we can say that currently there are no European-wide statistical sources that allow the construction of a composite indicator of job quality at the level of detail necessary here (for a partial approximation based on the Fourth European Working Conditions Survey, see Chapter 7 ). For the reasons given in the following chapter, we believe that our wage and educational indicators provide a more than reasonable approxima-tion to job quality, but, in any case, both indicators have enough intrinsic interest as to justify their use in this context.

As there is no single data source that allows us to construct a homo-geneous wage indicator at the EU level for the period under study, the wage ranking draws from four different sources: the European Structure of Earnings Survey (ESES), the European Survey on Income and Living Conditions (EU-SILC), the European Community Household Panel (ECHP) and the Structural Business Statistics of the EU (SBS). The basic structure of this ranking is provided by ESES, which is the best and most detailed European source available for this purpose (see Eurostat 2005) 5 : but, this survey does not cover the public sector nor does it pro-vide a breakdown of sectors at the two-digit level, 6 so it was necessary to complement this basic structure with data from the other sources. To add the relative wages of the public sectors, EU-SILC was used: the full matrix of wages by job was replicated on the basis of this source and the jobs only available in EU-SILC were added to the basic structure provided by ESES, after being corrected by a factor calculated by com-paring the jobs available in both sources. As with the ESES, EU-SILC does not permit a two-digit NACE breakdown, which is especially prob-lematic for the industrial sector (for services and the primary sector, the differences between one and two digit are minor). For a further break-down of manufacturing, ECHP was used (estimating the salary levels for each manufacturing sub-sector, aligning it to the previous estimates and applying the necessary correction factor). But ECHP covers only the EU15, so the SBS was used to breakdown manufacturing sector in the eight new Member States covered in this study. (In this case, the figures are less precise, because SBS does not include a breakdown by occupa-tion.) After this rather intricate process (for more details, see Stehrer and Ward 2008), estimates were obtained for the relative wage positions of each cell in the jobs matrix, for each of the 23 countries studied. While the resulting figures are undoubtedly an approximation to the real ones, they are probably as good as can be achieved with the exist-ing sources at the EU level and are, in our view, sufficiently reliable to

Methodology 21

serve the purpose here. The use of a static, relative and approximate approach, as explained earlier, should minimize any potential bias aris-ing from this complicated data-merging process.

On top of this, we have the second approximate measure of job quality, based on the average education level of workers in each job, which is much less problematic in its construction. This measure is simply based on the weighted average of the education level (using the three-level ISCED classification; for details of this classifica-tion, see UNESCO 2006) of those employed in each job and country, drawn from a pooled sample aggregating all ELFS from 1995 to 2007 (i.e. exactly the same data used for the quantitative jobs matrix, as explained earlier).

It is interesting to note that, despite their completely different under-lying logic and despite being based on different sources, there is a very high correlation between the two rankings (in most countries, such cor-relation is above 0.7: see Table 2.2 ). This suggests that both indicators are capturing the same underlying hierarchy of jobs. The correlation between the two job-quality rankings provides strong evidence of the robustness of our approach. 7

2.3 Data aggregation and analysis

In the two previous subsections, we have explained the construction of two separate jobs matrices, one quantitative (containing the number of workers in each combination of occupation and sector, and second-ary employment and demographic variables, for each country and year)

Table 2.2 Correlation between wage and educational rankings at the job level

Italy 0.87 Finland 0.69 France 0.73

Luxemburg 0.78 Latvia 0.57 Spain 0.71

Hungary 0.73 UK 0.83 Sweden 0.67

Belgium 0.69 Lithuania 0.76 Portugal 0.80

Cyprus 0.58 Slovakia 0.71 Austria 0.73

Germany 0.85 Ireland 0.68 Denmark 0.70

Netherlands 0.77 Greece 0.52 Estonia 0.63

Slovenia 0.72 Czech Republic 0.82

Tabl

e 2.

3 Pe

rio

ds

cove

red

– b

reak

s an

d a

dju

stm

ents

Co

un

try

Per

iod

cov

ered

B

reak

s id

enti

fied

A

dju

stm

ents

Au

stri

a 19

95–2

007

A

rec

lass

ific

atio

n o

f o

ccu

pat

ion

s in

ret

ail

trad

e in

20

04

(fro

m

sale

sper

son

s to

ass

oci

ate

pro

fess

ion

als)

led

to

a d

isp

lace

men

t of

10

0,0

00

wor

kers

fro

m q

uin

tile

2 t

o q

uin

tile

5

Eli

min

ate

stru

ctu

ral

chan

ge b

etw

een

20

03

and

20

04

from

th

e se

ries

B

elgi

um

19

95–2

007

N

one

Non

e C

yp

rus

1999

–20

07

Non

e N

one

Cze

ch

Rep

ubl

ic

1998

–20

07

Non

e N

one

Den

mar

k 19

95–2

007

N

one

Non

e E

ston

ia

1997

–20

07

Non

e N

one

Fin

lan

d

1997

–20

07

A r

ecla

ssif

icat

ion

of

occ

up

atio

ns

in t

he

hea

lth

sec

tor

in 2

002

(f

rom

pro

fess

ion

als

to a

sso

ciat

e p

rofe

ssio

nal

s an

d s

ervi

ce

wor

kers

) le

d t

o a

dis

pla

cem

ent

of 1

00,

00

0 w

orke

rs f

rom

q

uin

tile

5 t

o q

uin

tile

s 2

and

4

Eli

min

ate

stru

ctu

ral

chan

ge b

etw

een

20

01

and

20

02 f

rom

th

e se

ries

Fran

ce

1995

–20

07

Non

e N

one

Ger

man

y 19

95–2

007

N

one

Non

e G

reec

e 19

95–2

007

In

20

04,

a b

ig s

ud

den

dro

p i

n t

he

nu

mb

er o

f sk

ille

d a

gric

ult

ura

l w

orke

rs, b

ut

no

sim

ult

aneo

us

incr

ease

in

an

y ot

her

o

ccu

pat

ion

. Th

e d

rop

is

susp

icio

us

bu

t d

oes

not

aff

ect

mu

ch

the

tren

ds

Non

e

Hu

nga

ry

199 7

–20

07

Non

e N

one

Irel

and

19

95–2

007

In

199

8, t

her

e w

as a

big

dro

p i

n t

he

nu

mb

er o

f sk

ille

d

agri

cult

ura

l w

orke

rs, s

imu

ltan

eou

s w

ith

an

in

crea

se i

n

man

ager

s in

agr

icu

ltu

re (

look

s li

ke a

rec

lass

ific

atio

n)

Eli

min

ate

stru

ctu

ral

chan

ge b

etw

een

199

7 an

d 1

998

from

th

e se

ries

Ital

y 19

95–2

007

In

20

04,

th

ere

was

a d

rop

in

mor

e th

an 7

00,

00

0 sa

lesp

erso

ns

in

reta

il t

rad

e, w

ith

no

sim

ult

aneo

us

incr

ease

com

pen

sati

ng

it.

Th

ere

is a

lso

a d

ecre

ase

of n

ear

40

0,0

00

teac

hin

g p

rofe

ssio

nal

s an

d a

sim

ilar

in

crea

se i

n t

each

ing

asso

ciat

e p

rofe

ssio

nal

s

Eli

min

ate

stru

ctu

ral

chan

ge b

etw

een

20

03

and

20

04

from

th

e se

ries

Lat

via

1998

–20

07

Non

e N

one

Lit

hu

ania

19

98–2

007

N

one

Non

e Lu

xem

bou

rg

1995

–20

07

Non

e N

one

Net

her

lan

ds

1996

–20

07

Non

e N

one

Port

uga

l 19

95–2

007

In

199

8, t

her

e w

as a

big

dec

reas

e in

ele

men

tary

occ

up

atio

ns

in p

riva

te h

ouse

hol

ds,

sim

ult

aneo

us

wit

h a

n i

ncr

ease

in

h

ouse

keep

ing

serv

ice

wor

kers

(lo

oks

like

a r

ecla

ssif

icat

ion

)

Eli

min

ate

stru

ctu

ral

chan

ge b

etw

een

199

7 an

d 1

998

from

th

e se

ries

Sl

ovak

ia

1998

–20

07

Non

e N

one

Slov

enia

19

96–2

007

N

one

Non

e Sp

ain

19

95–2

007

N

one

Non

e Sw

eden

19

97–2

007

N

one

Non

e U

nit

ed

Kin

gdom

19

95–2

007

A

rec

lass

i fic

atio

n o

f o

ccu

pat

ion

s in

th

e h

ealt

h s

ecto

r in

20

01

(fro

m p

rofe

ssio

nal

s to

ass

oci

ate

pro

fess

ion

als

and

ser

vice

w

orke

rs)

led

to

a d

isp

lace

men

t of

50

0,0

00

wor

kers

fro

m

qu

inti

le 5

to

qu

inti

les

2 an

d 4

Eli

min

ate

stru

ctu

ral

chan

ge b

etw

een

20

00

and

20

01 f

rom

th

e se

ries


and the other qualitative (containing two ordinal indicators of job qual-ity for each combination of occupation and sector). Because these two matrices share the same structure (defined by the crossing of occupa-tion and sector), they can be combined at the job level, and this allows us to study the quantitative transformation of employment structures from a qualitative perspective. 8

The final analysis, nevertheless, is not carried out at the level of individual jobs (there are too many), but at a higher level of aggre-gation: by job-quality quintiles. These quintiles are constructed by aggregating in each of them 20 per cent of workers in each country for the year 2000, ranked by one of the quality measures (cf. Figure 1.1 ). That is, the lowest wage quintile for Spain contains the 20 per cent of Spanish workers whose jobs have the lowest median hourly wages and so on. Most of the analysis in the following chapters (though not all) looks at the evolution of employment levels in these quintiles; in some instances decomposed by third variables of interest such as gender and immigrant status. The use of a non-parametric and highly aggregated approach is not only adequate for synthesizing an enormous amount of data, but it also improves the comparability between countries, the consistency of time series and the validity of the aggregation of differ-ent data sources.

In most cases, the period covered is between 1995 and 2007, but in some cases, the effective period is shorter because of lack of data in the original sources (especially in the new Member States). Table 2.3 shows the periods covered in each country. In six countries, we had to adjust the series of employment in the ELFS for abrupt year-on-year changes in the number of workers in some jobs, associated with changes in survey or sampling methods. Such breaks affect the con-sistency of the change in the number of workers between two specific years, but they do not affect the rest of the series, and therefore can be adjusted by eliminating the problematic year from the total structural change for the whole period (substituting the change in that year by a constant factor equal to the global change in employment numbers). In Table 2.3 , the breaks in the series and the adjustments made for correcting them are explained (for more details, see Fernández-Macías 2010: chapter 3 , section 5.4).

Notes

1 . In fact, in this section we will only explain the process of constructing the European JOBS data set, covering 23 countries. With respect to the US

Methodology 25

data, which is only used in Chapter 4 , the interested reader should refer to this chapter and to the methodological appendices to the 2003 Wright and Dwyer paper (Wright and Dwyer 2003).

2 . The compilation of the basic data for this project, especially of the wage ranking which is explained in the following subsection, was mostly carried out by Terry Ward and Robert Stehrer (see Stehrer and Ward 2008).

3 . In Chapter 7 , we construct yet another ordinal indicator of job quality, based on a range of variables on the conditions of work and employment of workers in each job. Such multidimensional indicator of job quality is much richer in its content than the two indicators we use here, but also more problematic in terms of statistical validity because of the limitations of the sample from which it draws, so it can provide valuable complementary infor-mation but is not a substitute for the indicators used in this chapter.

4 . It is important to recall that we are speaking about the ranking of jobs by their wages and/or education, not about the absolute level of those two vari-ables. The absolute level of wages and education can change quite rapidly, so that even for a short period of 13 years as the one covered here it could be severely biased if a fixed structure was used: but the relative position of jobs (the wage/educational hierarchy between them) is unlikely to change signif-icantly over such a short period.

5 . We are grateful to Eurostat officials for providing data on median relative hourly earnings by sector and two-digit occupation which are not published and which have not been released before.

6 . The other important limitation of this survey is that it does not cover com-panies with less than ten employees: there is no way to circumvent this, but its implications should be minor considering that we are dealing here with rankings of average wages at the job level.

7 . In Chapter 7 , it is shown that these two indicators are also highly correlated with a third (tentative) indicator of job quality based on a multidimensional index drawn from the fourth EWCS.

8 . It should be noted that analysis was undertaken of the structure of employ-ment given by the ELFS and the sources of relative wages used to ensure that they were not significantly different – that a job was implicitly being defined in a similar way in terms of the sector of activity and occupation in the dif-ferent sources – which is key to our ability to merge these different sources in the way that was done and produce meaningful results.

26

3.1 Introduction

The debate on the structural evolution of the division of labour and its impact on job quality have been central in the Social Sciences for the last 200 years, and they remain so. The issue at stake is nothing less than the impact that technical and organizational change, the ulti-mate source of all economic progress, has on the condition of human beings as producers ( homo faber ). As we shall see in the next section, this debate was initiated in the late-18th century with a profound pessimism even amongst the most passionate defenders of the division of labour in industrial capitalism, such as Adam Smith; it later inclined towards a somewhat naïve optimism amongst the theorists of the post-industrial society; and recently, it resumed a more-pessimistic tone, although more nuanced this time, asserting that current technical and organizational change tends to generate polarization in employment structures.

In fact, it is a central issue not only in the scientific debate but also in the wider political or even social debate. In a context of global-ization in which governments and international institutions favour a (fallacious) rhetoric of competition between countries or regions, the evaluation of the structural change in national productive sys-tems becomes an evaluation of the success of the different countries in the Darwinian struggle for the highest value added activities (for an example of this kind of argument, see OECD 2007b: 19–23). In the European context, the long period of economic expansion initi-ated in the early 1990s (which coincides more or less with the period covered here) favoured a gradual shift in the emphasis of European employment policies, from an obsessive focus on quantity (see European Commission 1993) to a more balanced approach that, at

3 Patterns of Employment Expansion in Europe, 1995–2007 Enrique Fernández-Macías

Patterns of Employment Expansion in Europe, 1995–2007 27

least rhetorically, gave equal importance to the quantitative and qual-itative aspects (as evidenced in the motto ‘more and better jobs’ of the Lisbon strategy; see European Council 2000).

In this chapter, we will try to evaluate from a qualitative perspective the transformation of the employment structures of 23 European coun-tries between 1995 and 2007, linking this evaluation to the wider Social Sciences debate about the impact of structural change in the division of labour on job quality. To this end, as with the rest of contributions to this volume, we will use the ‘jobs approach’, a relatively new methodol-ogy which had never been applied to such a large sample of countries. This methodology was originally proposed by the economist Joseph E. Stiglitz when he was the Chief Economic Advisor to the US president in 1996 (US Council of Economic Advisors 1996), and was refined and consolidated by the sociologists Erik Olin Wright and Rachel Dwyer in two subsequent papers (Wright and Dwyer 2000, 2003). The main idea is simple, but very powerful analytically: it basically consists of shifting the unit of analysis of labour market from individuals to jobs, understanding jobs as specific occupations within specific sectors. The set of jobs defined in this way in an economy constitutes a jobs matrix, which can be understood as a stylized representation of the structure of the labour market, or in more general terms, as a structural snapshot of the state of the division of labour in a certain place and time. Using any data source which incorporates the variables of occupation and sector, we can add further information on each job, including ordinal meas-ures of their quality. Drawing from such measures, the jobs are then regrouped in categories according to their quality, and the evolution in the number of workers in each category is used to evaluate the nature of structural change over a specific time period.

In this chapter, we will follow in general terms the approach pro-posed by Erik Olin Wright and Rachel Dwyer in their analysis of the US employment expansion of 1990s, covering 23 European countries, over a period of 13 years (1995–2007). In Section 3.2, we will discuss the the-oretical foundations of this approach and the hypotheses that we will try to test in this chapter. In Section 3.3, we will evaluate the general patterns of structural change in European labour markets. In Section 3.4, we will compare our results to those of a recent similar study which reached rather different conclusions (Goos et al. 2009). In Section 3.5, we will analyse in some detail the sector dynamics which underlie the general patterns of structural change in employment, and we will relate them to the employment destandardization process which is taking place in some European countries.

28 Enrique Fernández-Macías

3.2 Theoretical foundations and hypotheses

3.2.1 The jobs matrix as a structural representation of the division of labour

Previous studies using the same approach did not include much discus-sion of its theoretical foundations: these studies were more oriented by policy than theory (with the possible exception of Goos and Manning 2007) and carried out with the specific objective of evaluating the quality of jobs created (and destroyed) in specific periods of economic expansion. The use of jobs , defined by crossing the variables of occupa-tion and sector, instead of individuals as unit of analysis was justified in these studies more as a heuristic convenience than as a substantially different way to analyse or conceptualize the labour market. But there must be some reason why using jobs rather than individuals as unit of analysis is heuristically convenient, and this reason must be related to the nature of the classifications of occupation and sector on which this approach is based.

The labour market is not an amorphous mass of individual agents who join and leave transactions depending on relative prices, but a relatively stable structure (though continuously changing) of relations between agents, formed by positions that transcend the individuals who occupy them. The principle which lies behind this structure is the division of labour: the subdivision of tasks and its allocation to specialized work-ers increase enormously the total productivity of labour, but it requires complex coordination mechanisms and stable labour positions, and hence a relatively stable structure of economic relations.

Adam Smith, who ‘discovered’ the principle of the division of labour for contemporary Social Sciences (Smith 1776: I.1.8), viewed all eco-nomic transactions from the perspective of the market, and hence spoke of a single principle of the division of labour (exemplified by his famous pin factory). It was Marx, in an explicit critique of Adam Smith, who argued that there were two radically different forms of the division of labour, with different subjects and coordination mechanisms. On the one hand, the market coordinates the division of labour between independent producers, or more in general, between independent pro-ducing units (firms): such division of labour corresponds to the dif-ferent branches of industry, and Marx called it the ‘social division of labour’ (Marx 1867: 14.4). But within companies, the mechanism that assigns and coordinates labour is not the market, but the direct author-ity of the capitalist (or the manager): hence, it involves a different form of division of labour, which Marx called ‘manufacturing division of


labour’, and which corresponds with the levels of skills and hierarchy within firms. A very similar distinction lies behind the institutionalist theory of the firm of Ronald Coase, which argues that firms exist at all (something which, strikingly, is at odds with the standard economic approach) because there are some types of transactions which are more efficient to coordinate by hierarchy than by markets (Coase 1996: 94). One implication is that there is a fundamental distinction between the mechanisms for coordinating labour within and between firms.

Our argument is that the variables of occupation and sector which define the jobs matrix provide the theoretical anchor for this approach, because these variables classify employment in terms of the two indi-cated dimensions of the division of labour. The jobs matrix, therefore, can be understood as a stylized representation of the labour market at a point in time, from the perspective of the division of labour along its horizontal and vertical dimensions. The ‘jobs’, defined as specific occupations within specific sectors, can be understood as the smallest unit in the division of labour from such a perspective. Our analysis in the following pages, then, is an evaluation of the transformations in the structure of the division of labour of several European countries over a specific period of time, from the perspective of job quality. 1

3.2.2 Job quality, relative wages and education levels

In most previous studies using the same approach, the jobs were sorted and classified in terms of their quality, using as a basis their median hourly wages. Again, the justification was more practical than theo-retical: without denying that job quality is about much more than just wages, the lack of more comprehensive statistics of job quality makes wages a reasonable proxy, or in any case a variable with sufficient inter-est in itself. In this chapter, we will follow this tradition and use median hourly wages as the basis for the sorting and classifying of jobs: but to add richness to our analysis, we will include a second ranking of jobs, based on the average educational level of workers.

The canonical theory of job quality in orthodox economics is the theory of compensating differentials, which was formulated by Adam Smith more than 200 years ago and which remains practically intact (which testifies to the lack of interest that this subject arouses in ortho-dox economics; see Smith 1776: I.10; Cahuc and Zylberberg 2004: 248). According to this theory, the utility that the worker derives from his or her employment depends, on the one hand, on the intrinsic unpleas-antness of work (‘the ease or hardship, the cleanliness or dirtiness, the honourableness or dishonourableness of the employment’ Smith


1776: I.10.5), and on the other, on the wage that the worker receives as compensation. Workers can have different preferences for money v. unpleasantness, which can lead to different levels of monetary com-pensation for the same level of total utility of employment. The wage in itself, therefore, would be a poor indicator of job quality from such perspective, because, ceteris paribus , it would move in the oppo-site direction to the rest of the attributes of employment (therefore, a high salary could be an indicator of a very unpleasant job). But as usu-ally happens, the devil is in the ceteris paribus : both the wage and the unpleasantness of work are costs for the firm, and therefore the max-imum possible combination of both elements which the worker can expect will be ultimately delimited by the productivity of his or her labour. Hence, from this perspective what really determines job quality (the utility derived from employment, defined by the specific combi-nation of wage and unpleasantness to which the worker can aspire) is productivity, which would be determined by the skills and aptitude of each worker. In this sense, the average educational level of workers in each job, to the extent that it serves as an approximation to the skills and productivity associated to each job, could be a better indicator of job quality than the median hourly wages. 2

The previous arguments illustrate why orthodox economic theory has never paid much attention to job quality: in a competitive labour mar-ket, wages would compensate perfectly for the unpleasantness of work (with all remaining differences resulting from different preferences) and the differences in the utility derived from each type of employ-ment (the combination of wages and unpleasantness) simply reflect their different productivity. In stark contrast, most of the sociological approaches to job quality, as well as those of non-orthodox economics, are based on the premise that in most cases the compensation of labour does not reflect the unpleasantness of work or its productivity, because there are mechanisms such as exploitation or discrimination which lead the good or bad attributes of jobs to accumulate rather than to compen-sate for each other. The traditional sociological approach, derived from the Marxist concepts of exploitation and alienation, focuses on the mechanisms for which labour compensation remains below the contri-bution of labour to production (the conditions of exploitation, such as deskilling and power relations at the workplace; the classical study in this tradition is Braverman 1974), and in the importance of work as an element of identity, realization and social integration of human beings (alienation; the classical study is Blauner 1964). On the other hand, the segmentation approach of institutionalist economics and industrial


sociology focuses on the mechanisms that break the unity of the labour market into different segments, governed by different rules and gener-ating differential access to wages and employment conditions which do not result from differences in skill or aptitude as the market paradigm assumes (see Doeringer and Piore 1971; Peck 1996: 56–79 for a review of recent segmentation debates). Although the sociological/institutional approaches tend to emphasize that job quality cannot be reduced to wages (job quality is assumed to be a complex and multidimensional phenomenon), they often argue that wages and working conditions are more likely to go together rather than compensate each other: hence, from these perspectives, wages could be used as a reasonable proxy for job quality. The average educational level of workers can be also a good approximation from this perspective, because of the empirical associa-tion between formal education and wages and working conditions in general (for a European review, see Asplund and Barth 2005).

3.2.3 The debate on the impact of structural change on the quality of employment

The Social Sciences debate on the impact of structural change in the division of labour on the quality of employment has been dominated by three main visions, associated with three different periods of con-temporary economic history: the first vision, profoundly pessimistic, is associated with the long period of development of the Industrial Revolution and to the predominance of the industrial sector; the sec-ond vision, of an almost evangelical optimism, is associated with the emergence of services as the predominant sector of employment and the development of the Welfare State after World War II; the third and most recent vision, which qualifies and synthesizes the previous two albeit with a generally pessimistic tone, is associated with the crisis of the Welfare State and the emergence of the new information technolo-gies in the last two decades of the 20th century.

The pessimistic vision dominated the debate on the implications for the labour condition of technical and organizational change since the beginning of the Industrial Revolution until, at least, World War II (in fact, it had a revival in the 1970s, when the swansong of this vision, Harry Braverman’s Labour and Monopoly Capital , was published). It is interesting to note that this pessimistic vision was shared by the most disparate range of social commentators, from staunch supporters of capitalism and industrialism (such as Adam Smith and F.W. Taylor) to its fiercest critics (such as Marx and Gramsci). This pessimism derived from the recognition of the ambivalent nature of the technical and


organizational change associated with the Industrial Revolution: on the one hand, it boosted productivity and in general, the capacity of human society to survive and prosper; on the other, such progress was produced at the expense of a continuous degradation of the conditions of work. The proud pre-industrial independent producer is transformed into a kind of mechanical brainless beast, who only contributes with brute energy to the productive process without even understanding it (Smith 1776: V.1.178; Marx 1867; Taylor 1911). The increasing returns of mass production and ‘scientific’ management principles tend to shift downwards the employment structures of industrializing economies. In the 1970s, Braverman would update these arguments to the incip-ient service economy: ‘there has been therefore an immense shift of labour out of the traditional manufacturing, mining and construction, and transportation industries into the very rapidly growing areas of real estate, insurance, finance, services and wholesale and retail trade. But these rapidly growing fields of industry are precisely the low-wage por-tions of the economy, while the higher-wage sector is the stagnant or declining portion’ (Braverman 1974: 323).

Such pessimism was slowly dispelled while, after World War II, most advanced Western economies initiated a series of institutional reforms that transformed them into more egalitarian and inclusive social sys-tems. In this context, a much-more optimistic vision about the impact of technical and organizational change starts to dominate the debate, a vision which is associated with the theories of post-industrialism. It is precisely the enormous leap in productive human capacity resulting from industrialization which had radically transformed the nature of labour, improving the conditions of work in two different ways. First, by reducing dramatically the amount of work necessary to produce the material goods required to support the population, which implies a grad-ual increase in the share of workers dedicated to the provision of imma-terial services, leisure and well-being (see Fourastié 1963; also Baumol 1967); second, by the substitution of the most arduous industrial tasks by machines, and the transformation of the unskilled, mass-produc-tion worker into high-qualified engineering workers, whose main task is the design and supervision of machines (see Bell 1976). The most degraded and alienating jobs of the industrial age tend to disappear or be upskilled, and the faster-expanding segments of employment are in services, especially those with a higher-informational content, good wages and working conditions. The labour markets of post-industrial societies tend, according to these theories, to experience a continuous process of structural upgrading. 3


But the pendulum swings again when, around the mid-1970s, most advanced countries entered a crisis which affected first the economy and then the identity of the state itself. The political consensus that had dominated the previous three decades and built the Welfare State started to deteriorate, slowly but steadily, until many of the previous social and labour policies were rolled back (in the 1980s in the UK and US, elsewhere in the 1990s). Social inequalities started growing again, and many argue that the conditions of work and employment deteriorated over this period as well (Doogan 2009: 194–206). Against this back-ground, a more-pessimistic view about the implications of technical and organizational change returns to the forefront of Social Sciences. It is, nevertheless, a more nuanced view, which in some ways can be understood as a synthesis between the pessimistic and optimistic views. According to this view, the structural evolution of the division of labour in advanced capitalist societies since the 1970s is marked by polariza-tion. The literature on this issue tends to focus on two explanatory fac-tors: technology and international trade. On the one hand, the nature of technical change since the informational revolution tends to substi-tute workers who traditionally occupied intermediate positions (routine production and administrative tasks) while boosting the demand for labour at the two extremes of the skill-job quality hierarchy ( knowledge-intensive tasks and manual non-routine tasks). These the-ories are in fact a derivation of the SBTC arguments (which predicted upgrading rather than polarization), which has been called Task-Based Technical Change (see Autor et al. 2003; Autor et al. 2006; Goos and Manning 2007). On the other hand, it is argued that the globalization of markets has also a destructive impact on the middle of the employ-ment structure of developed economies, especially on unskilled and semiskilled industrial jobs that can be equally performed in developing economies at a fraction of the cost (see Bluestone and Harrison 1982; Kuttner 1983; Harrison and Bluestone 1988).

Each of these three visions implies a specific image of change in the structure of employment in terms of job quality: the pessimistic view implies an image of degradation and downward bias in the transforma-tion of employment structures; the optimistic, an image of continuous upgrading; and the most recent nuanced view, an image of polarization in labour markets. In this chapter, using the jobs approach which was described earlier, we will construct an empirical snapshot of change in the structure of employment of 23 European countries and contrast it with the theoretical predictions based on each of these three visions of structural change in the Social Sciences literature.


3.3 General patterns of job expansion in Europe, 1995–2007

Figure 3.1 shows, for each of the 23 European countries for which we have data, the absolute change in the number of workers by quintile between 1995 and 2007. The dark grey bars represent the wage quin-tiles. For instance, the chart in the upper-left corner shows the Dutch figures: the quintile holding the lowest paid jobs (1st) expanded by roughly half a million in this period, almost the same growth as the highest (5th) quintile; the 2nd and 4th quintiles also expanded sim-ilarly, between 250,000 and 300,000 workers each; and the quin-tile with wages around the median (the 3rd) grew much less, around 50,000 workers in total. Overall, the pattern of job expansion was very polarized and notably symmetrical for the Netherlands. The light grey bars, on the other hand, show the change in the educational quin-tiles, whose main purpose is to support and contrast the results of the wage quintiles: in the case of the Netherlands, they show a very similar polarization pattern, though clearly more biased towards the top of the educational hierarchy. In most cases, the educational quintiles show a pattern which is very similar to the wage quintiles, but often more biased towards the top.

The first and maybe most important observation we can make look-ing at the national patterns shown in Figure 3.1 is that there is no sin-gle pattern that can characterize the change in the structure of the labour market of the 23 countries shown, not even approximately. There is instead a considerable plurality in these patterns of job expan-sion, which suggests that there is no single explanation valid for all of Europe, but a plurality of factors and diverging developments. This may seem obvious (after all, we are dealing with 23 countries which are rather different in many ways), but it can also be seen as contradict-ing the three hypotheses outlined earlier, each of which is presented, more or less explicitly, as a univocal explanation applicable in general to economies at a similar level of development. In fact, this plurality of patterns directly contradicts the results of a recent study based on a very similar approach, which argued that the employment structures of the old EU15 member states (except Portugal) suffered a similar process of polarization over the last decade and a half (see Goos et al. 2009). Such inconsistency of findings between two very similar studies, for the same period and the same countries, is quite striking, so we will return to this issue in the following section to try to identify the origins of the inconsistency.

05001,000 0 02040 –50050100 –10001002001,0002,000

Change in the number of workers per quintile (000s)

0 –2002000 –200200400001,0002,0003,000 –200200400

0100 –100 0100 –100 040 –2020 0200 –200

01020 0200400300100200 02,000100 0 –100200 1000

Lith

uani

aG

erm

any

Spa

inF

inla

ndN

ethe

rland

s

Bel

gium

Latv

iaP

ortu

gal

Luxe

mbu

rg

Fran

ce

Irel

and

Est

onia

Gre

ece

Sw

eden

Cyp

rus

UK

Aus

tria

Cze

ch R

epub

licD

enm

ark

Slo

vaki

a

Slo

veni

aIta

lyH

unga

ry

Edu

catio

nW

age

0100200300 0100200 1,000 0 –1,000 0501002,0001,000 0 –1,0002,000

Figu

re 3

.1

Patt

ern

s of

em

plo

ymen

t ex

pan

sion

in

Eu

rop

e, 1

995

–20

07


Our first important finding is that there was a plurality of patterns of job expansion in Europe between 1995 and 2007: once this is assumed, we can try and classify each national pattern according to its similar-ity with the images of change proposed by the hypotheses discussed earlier. In Figure 1.1 , the individual country charts have been orga-nized according to such a classification, by columns (drawing primar-ily on the wage quintiles). The first column (to the left) shows the five countries whose pattern of job expansion is more clearly polarized: the Netherlands, France, Cyprus, Slovakia and Hungary. The second column incorporates five countries whose pattern is simultaneously polarized and upwards biased: Germany, Belgium, Ireland, the UK and Slovenia. The third column shows the five countries characterized by a clear and unambiguous structural upgrading: Finland, Luxembourg, Sweden, Denmark and Italy. The fourth column shows four cases of very mild structural upgrading, in which the quintiles grew rather evenly or the expansion was bigger in the middle quintiles (in fact, in three of these four countries the upgrading is only the result of a strong destruction of employment in the first quintile): Spain, Portugal, Greece and the Czech Republic. Finally, in the last column we have put the Baltic States and Austria, which do not fit clearly into any category and which show a clear inconsistency in the two indicators of job quality, suggesting that their results may be less reliable (in the other countries, the two indicators are generally consistent).

Therefore, in terms of the three hypotheses discussed earlier, our results imply a complete rejection of the structural degradation hypoth-esis, a limited support for the hypothesis of polarization and a wider (though not complete) support for the structural upgrading hypoth-esis. Across most of Europe, the jobs with higher relative wages and higher average educational levels experienced the biggest expansion. Even most of the cases of polarization (with only the exception of the Netherlands and Slovakia) were clearly biased upwards. There are only two cases that could more or less fit the structural degradation image, and only for the wages indicator: Slovakia and Estonia.

An interesting question is to what extent the classification shown in Figure 3.1 coincides with the European institutional families that recurrently appear in the economic and sociological literature (Esping-Andersen 1990; Ferrera 1996; Fenger 2007). With different names (wel-fare models, institutional typologies or varieties or capitalism) these institutional families refer to groups of countries with similar and related institutional structures (because of geographic proximity, histor-ical and cultural affinity). In Europe, these are usually associated with


the big geographic regions of the North (Nordic countries), Isles (the UK and Ireland), Continental Europe (Germany, France, Belgium, the Netherlands, Luxembourg and Austria), South (Portugal, Spain, Italy, Greece and Cyprus) and East (the Baltic States, Czech Republic, Slovakia and Slovenia). To the extent that the institutional context affects job-creation patterns, some type of association between the classification shown in Figure 1.1 and these institutional families could be expected. The results suggest that such a relation exists, but it is weakened by some important exceptions. The polarization pattern seems to be asso-ciated with Continental Europe, including the Netherlands, France, Germany and Belgium. There are two Continental countries that do not fit this pattern, but they seem reasonable exceptions: Luxembourg is an exception in itself, a very small country with a very special economic structure; and there seems to be some problem with the Austrian data, highlighted by the (unexpected) discrepancy between the results of the wage and the educational rankings. (In fact, if we take the educational ranking, Austria becomes a case of polarization, similar to the rest of Continental Europe, which suggests that the problem may lie in the wage ranking.) The UK and Ireland, often put together within a ‘Liberal’ country cluster, also display a very similar pattern of hybrid polariza-tion/upgrading. The three Nordic countries for which we have data experienced a similar pattern of unambiguous upgrading. All Southern European countries followed a characteristic pattern of upgrading with a differential expansion of the middle tiers of the employment struc-ture. Finally, the six Eastern European countries for which we have data are dispersed across categories, without a clear, shared pattern. Overall, the association between the patterns of structural employ-ment change and European institutional clusters seems significant but not perfect: this suggests that the diversity of institutional frameworks in Europe is a determinant of the plurality of patterns discussed earlier, but that there are other confounding factors at work. In the following pages, we will hint at some of those factors, but the reader must be aware that our goal in this chapter is to identify the main patterns of structural change and explore some of its underlying trends: the task of specifying an explanatory model of structural change is left for further research.

3.4 Job polarization in Europe?

In a recent article based on a very similar methodological approach, for almost the same period (1993–2006) and a similar sample of countries


(EU15 plus Norway), Goos, Manning and Salomons (GMS from now on) conclude that all countries except Portugal experienced a polar-ization process in their employment structures. This is obviously in contradiction with the results that we have just discussed, which is striking considering the similarities mentioned between the two stud-ies. It is therefore worth considering in detail the reasons behind such inconsistency.

It is important to note that we are confident that the differences between the GMS results and ours do not stem from differences in the data used or from errors in the analysis of either party. We exchanged data and could broadly replicate our respective results with the approach of the other party. 4 The differences, therefore, are the result of a differ-ent analytic strategy, and to a lesser extent, of a different emphasis in the interpretation.

The first important difference concerns the ranking used for char-acterizing the quality of jobs. We use nationally specific wage rank-ings – in fact, we use two-nationally specific rankings: by wages and education – while GMS use a single ranking for all countries, based on wage data for the UK. The latter implies assuming that the wage struc-ture is very similar in all countries, so that using one particular rank-ing for all does not affect the national results: we believe this is clearly not the case and, therefore, that our approach is better. 5 Although it is true that there is an important degree of correlation between the wage rankings of the different countries, such correlation is far from perfect: and small differences in the rankings can generate very important dif-ferences in the patterns of job expansion (see, for example, Fernández-Macías 2010: 157–174 and 205–215).

The second important difference lies in the criteria used for the con-struction of the segments of job quality whose evolution is studied to evaluate the nature of the employment expansion. In this study, follow-ing Wright and Dwyer (2003), we have constructed five segments, each holding the same number of workers in the year 2000, with jobs ranked by their quality: such grouping of jobs in quintiles is, in our view, not only useful but also transparent, because it constitutes a kind of tabula rasa from which the change in the structure of employment in the 23 countries can be studied. GMS, on their paper, divide all jobs into three segments (bad, middling and good jobs), also according to their median wages, but the size of each of these three segments varies without any clear explanation as to why this is the case. 6 (At the EU level, the share of employment in each segment in the initial year studied is 22% for bad, 49% for middling and 29% for good jobs.) This may seem unimportant,


but it is not: in fact, the size of the segments can change considerably the growth rates of each job-quality tier. This is why it is so important to apply a consistent logic to the construction of the job-quality seg-ments, and to provide a clear justification of this logic. For instance, if we take the GMS figures for the EU and regroup the jobs so that the three job-quality segments are more even in the first year of the period (making them terciles of employment), the degree of polarization is con-siderably reduced, and the structural upgrading increased. 7

Finally, even if we accept their results as presented, GMS seem to emphasize too much the polarization pattern, ignoring the simulta-neous stronger structural upgrading. At the EU level, according to their own figures (GMS 2009: 59), the bad jobs expanded by 1.6 per cent, the middling jobs contracted by 7.8 per cent and the good jobs expanded by almost 8 per cent. There is no doubt that these figures involve some degree of polarization, but there is no doubt either that they involve a strong structural upgrading which is largely ignored in the paper.

Anyway, what is crucial is not so much the differences in the inter-pretation (which are always debatable), but the two earlier points made about differences in the analytic strategy. The use of a single ordinal measure of job quality drawn from a single country and the unexplained aggregation of jobs into uneven quality tiers are the main reasons for the differences between our results and the results of GMS. In both cases, the GMS approach tends to accentuate, somewhat artificially in our view, the polarization story.

3.5 Decomposing the patterns of job creation by sector and type of employment relation

As the jobs matrix is defined by a combination of the variables of sector and occupation at the two-digit level, it is possible to decom-pose the patterns of job expansion discussed in the previous pages for any reclassification of sector or occupation which is constructed at an aggregation level which is higher than the two-digit. In this section, we will take advantage of this possibility to decompose the patterns of job expansion at the national level for nine broad cate-gories of sector (constructed taking into account the arguments of the theoretical debate discussed earlier). This shall return us to the sphere of the division of labour explicitly (which we never actually abandoned, but it was largely concealed behind the quintile picture), because it will uncover some of the sector dynamics that underlie the patterns described in previous sections. It shall also help us to


evaluate in more detail the three hypotheses described in the second section of this chapter, because each of these hypotheses postulates a specific type of sector dynamic behind its image of change: the degradation hypothesis emphasizes the shifts in employment from mid-paid industrial jobs to low-paid industrial and service occupa-tions (Braverman 1974: 323); the upgrading hypothesis, a gradual disappearance of low-paid industrial jobs, and a constant expansion of good jobs in services, especially in high-technology and information-related activities (Bell 1976); and the polarization hypothesis, a reduction of mid-paid routine jobs in services and industry, an expansion of non-routine low-paid jobs in services (per-sonal services) and also of non-routine high-paid jobs in managerial and technical service occupations (Goos and Manning 2007).

Continuing a process that goes back to the Industrial Revolution (or earlier), the primary sector contributed negatively to employment in all countries, but especially in those where it still retains some importance (in the South and East of Europe, mainly). In terms of job quality, in most cases the primary sector contributed to the contraction of employment in the lowest quintiles (not only because this is where primary-sector jobs tend to concentrate, but also because these are the primary-sector jobs that are more likely to disappear).

The evolution of construction was rather diverse across Europe between 1995 and 2007, experiencing a strong expansion in some countries – most importantly in Spain and Ireland, but also in other Southern European and some Eastern European countries, a strong contraction in Germany, and hardly anything elsewhere. Employment in construction tends to concentrate in the middle-wage quintiles, and because as we will see later, the other sectors did not contribute much to the expansion of these segments of employment, this sector was (surprisingly) the one that determined most directly the dynam-ics of middling jobs. Construction is the one of the most cyclical sectors, which is another way of saying that its impact is likely to be the most ephemeral in structural terms: those countries that created more construction jobs before 2007 (as shown in Figure 3.2 ) are right now destroying those same jobs at even faster rates, so the impact of this sector on the structure of job quality shown in Figure 3.2 has reversed in the last three years. For instance, the recent dramatic col-lapse of construction in Spain is undoubtedly having a polarizing effect, because as Figure 3.2 shows, this sector contributed mostly to the expansion of middle-paid jobs in this country up to 2007 (for more details, see Chapter 10 ).


Gre

ece

Net

herla

nds

Fran

ceB

elgi

um

Cyp

rus

Irel

and

Slo

vaki

aU

KD

enm

ark

Cze

ch R

epub

lic

Sw

eden

Luxe

mbu

rgP

ortu

gal

Hun

gary

Slo

veni

aIta

ly

Ger

man

yF

inla

ndS

pain

Lith

uani

a

Latv

ia

Est

onia

Aus

tria

0200400600 0 02040 –50050100 –10001002001,0002,000

0 –2002000 –200200400001,0002,000 –200200

0100200–2002040–10001000100 –100

1,000 0 0100200 0100200–1,0002,000 1,000 500 0 –500 050100

01020 0200400–1,000100200 01,0002,000100 0 –100200

Prim

ary

Con

stru

ctio

nR

est

Figu

re 3

.2

Con

trib

uti

on t

o jo

b gr

owth

of

the

pri

mar

y se

ctor

an

d c

onst

ruct

ion


Figure 3.3 shows the industrial sector divided in two subsectors, accord-ing to the technological intensity of the productive process, following an OECD breakdown (see Hatzichronoglou 1997). Between 1995 and 2007, both types of industry remained basically stagnant across most of Europe. Low-technology industries (which includes subsectors such as manufacture of food, textiles, furniture) had a negative contribution to employment almost everywhere (destroying employment in net terms in the lowest quintiles, most notably in Germany, Austria, Denmark, Italy, Portugal and Belgium, as well as most of Eastern Europe). High-technology industries (which includes subsectors such as production of chemicals, cars, electrical equipment or computers) had a small contribution to the expansion of the highest layers of employment in some EU15 countries (especially in Germany, France, Ireland, Austria and Finland), and an important contribution to the expansion of the middle layers in the three Visegrad countries available in our sample (Czech Republic, Hungary and Slovakia). Although in absolute terms the contribution of both types of industry was rather small (in pos-itive or negative terms), it must be noted that this actually means a significant reduction of this sector in relative terms, because the period 1995–2007 was one of generalized employment expansion. In previ-ous historical periods, the industrial sector was the driver of the mid-dle (and mid-low) segments of employment: the inexorable decline of industry in terms of employment is one of the main reasons behind the stagnation of the middle layers of employment in most countries. Only in three of the New Member States of our sample (the Central-Eastern bloc of the Visegrad countries) was there a significant positive develop-ment in the industrial sector, and only in high-technology industries (these countries, as all New Member States, experienced a massive pro-cess of industrial restructuring that mostly affected employment in old and inefficient low-technology industries: Figure 3.3 shows that such restructuring destroyed large numbers of low-paid jobs): as expected, such development contributed to the expansion of the middle layers of employment in these countries.

Because of its large share of overall employment, we have split the ser-vice sector in five categories, two private (knowledge-intensive services and the rest, using the Eurostat classification; see Felix 2006) and three usually public (public administration, health and education). Private ser-vices , as Figure 3.4 shows, were responsible for most of the employment expansion between 1995 and 2007. The relevance of the distinction between knowledge-intensive services (which includes financial and business services, communications and non-land transport, research


Net

herla

nds

Fran

ceB

elgi

um

Slo

vaki

aU

KD

enm

ark

Cze

ch R

epub

lic

Cyp

rus

Irel

and

Gre

ece

Sw

eden

Luxe

mbu

rgP

ortu

gal

Hun

gary

Slo

veni

aIta

ly

Ger

man

yF

inla

ndS

pain

Lith

uani

a

Latv

ia

Est

onia

Aus

tria

0200400600 0 02040 –50050 –10001002001,0002,000

0 –2002000 –200200400001,0002,000 –200200

0 -100100–2002040–10001000100 –100

1,000 0 -1000100200 0100200–1,000 1,000 0 –1000 0 -5050100

010 0200400200100200 -100001,000100 0 –100200

Low

-tec

h in

dust

ryH

igh-

tech

indu

stry

Res

t

Figu

re 3

.3

Con

trib

uti

on t

o jo

b gr

owth

of

low

- an

d h

igh

-tec

hn

olog

y in

du

stri

es


and development among other activities) and the rest (which includes retail, hotels and restaurants and land transport among other activities) is illustrated in Figure 3.4 : knowledge-intensive services are strongly biased upwards, explaining in all countries more than half of the expan-sion of employment in the highest quintiles, whereas less knowledge-intensive services show the opposite image, strongly biased downwards and determining directly most of the expansion of the two lowest quin-tiles. It is interesting to note that private services hardly contributed at all to the middle segments of employment, despite the fact that they do account for a significant share of such segments (especially, in adminis-trative services). Private services tend to have a polarizing effect because it is those services on the extremes which are expanding faster. But this centripetal effect is uneven in different countries. Whereas knowledge-intensive services expanded very rapidly in all countries, which partly explain the near universal structural upgrading throughout this period, less knowledge-intensive services expanded significantly in just a few cases. Wherever the latter type of services grew, the bottom of the employment structure grew as well, and if there was no simultaneous expansion of the middle layers (generally driven by construction), this led to polarization.

Finally, Figure 3.5 shows the typically, though not necessarily, public-service sectors. Their contribution to employment was generally positive, but more moderate than that of private services, and more clearly biased towards the highest quintiles. The biggest and most generalized expan-sion was that of health, which contributed strongly to the expansion of the top-two quintiles in many countries; public administration and education had a more uneven contribution, both in terms of size and in terms of quality position (though in most cases biased upwards).

These sector dynamics facilitate a better understanding of what is general and what is specific in the patterns of job expansion discussed in Section 3.3. It seems clear that if most countries experienced a big-ger expansion of their higher job-quality segments of employment, it was the result of a generalized expansion of knowledge-intensive private services, and the health sector; that the sectors that had tradi-tionally propped up the middling segments of employment (industry and administrative services) remained stagnant, so that the evolution of these middling segments depended mainly of the fluctuations of the construction sector; and that wherever the lower quintiles grew, such growth was associated with personal services (or less knowledge- intensive services). But why did such expansion of personal services happen in some countries and not in others?

0 –2002000 –200200400001,0002,000 –200200

0 -100100–2002040–10001000100 –100

1,000 0 –1000100200 0100200–1,000 050100

010 –2000200400200100200 -100001,000100 0 –100200

Less

kno

wle

dge

inte

nsiv

e pr

ivat

e se

rvic

esK

now

ledg

e in

tens

ive

priv

ate

serv

ices

Res

t


Net

herla

nds

Fran

ce

Cyp

rus

Slo

vaki

a

Hon

gary

0200400600 0 02040 –50050 –10001002001,0002,000

Bel

gium

Irel

and

UK

Den

mar

kC

zech

Rep

ublic

Gre

ece

Sw

eden

Luxe

mbu

rgP

ortu

gal

Slo

veni

aIta

ly

Ger

man

yF

inla

ndS

pain

Lith

uani

a

Latv

ia

Est

onia

Aus

tria

1,000 500 0 –500

Figu

re 3

.4

Con

trib

uti

on t

o jo

b gr

owth

of

pri

vate

ser

vice

s

Change in the number of workers per quintile (000s)0200400600 0 02040 –50050 –10001002001,0002,000

0 –2002000 –2002004000 –20020040001,0002,000

50 0100–2002040–10001000100 –100

1,000 0 0100200 0100200–1,000 1,000 0500 –500 050100

0 –1010 0200400200100200 05001,000500100 0 –100200

Pub

lic a

dmin

istra

tion

Edu

catio

nH

ealth

Res

t

Lith

uani

a

Bel

gium

Irel

and

UK

Slo

veni

a

Ger

man

y

Latv

ia

Est

onia

Aus

tria

Spa

in

Por

tuga

l

Gre

ece

Cze

ch R

epub

licD

enm

ark

Sw

eden

Luxe

mbu

rg

Italy

Fin

land

Net

herla

nds

Fran

ce

Cyp

rus

Slo

vaki

a

Hun

gary

Figu

re 3

.5

Con

trib

uti

on t

o jo

b gr

owth

of

pu

blic

ser

vice

s


Lith

uani

a

Bel

gium

Irel

and

UK

Slo

veni

a

Ger

man

y

05001,000 0 02040 –50050 –10001002001,0002,000

0 –2002000 –200200400001,0002,000 –200200

Latv

ia

Est

onia

Aus

tria

Spa

in

Por

tuga

l

Gre

ece

Cze

ch R

epub

licD

enm

ark

Sw

eden

Luxe

mbu

rg

Italy

Fin

land

–10010 0200400200100200 02000100 0 –100200 5001000

Net

herla

nds

Fran

ce

Cyp

rus

Slo

vaki

a

Hun

gary

Non

-sta

ndar

d em

ploy

men

tS

tand

ard

empl

oym

ent

0100 –100 0100 –100 040 –2020 0100 –5050

0 –1000100200 0100200–2,0002,000 0500 –5050100 01,000

Figu

re 3

.6

A d

ecom

po

siti

on o

f th

e p

atte

rns

of j

ob

grow

th b

y em

plo

ymen

t st

atu

s


Figure 3.6 suggests a possible answer to this question, which goes in the same line as the inference made earlier about the possible rela-tionship between the different patterns of job expansion and different proposed European institutional ‘families’ or clusters. The decomposi-tion shown in Figure 3.6 is based on the type of employment relation, differentiating between workers with a standard employment relation (full-time and indefinite contract) and workers without (i.e. workers with part-time or non-indefinite contracts, or self-employed). Figure 3.6 can be linked to the argument of the destandardization of employment made, among others, by Ulrich Beck (1992: chapter 6 ), putting it in the context of our analysis of the transformations of the employment structure.

This approach seems, at first sight, quite fruitful: there is an obvi-ous relationship between the destandardization of employment and the expansion of low-paid jobs in recent years, a relationship which is stronger in Continental European countries – that is, precisely in the group of countries most clearly linked to the pattern of polarization. In the countries of the upper-left quadrant, there was between 1995 and 2007 a very strong process of destandardization of employment (all or most of the expansion of employment in net terms was non-standard, with net destruction of standard employment in the lowest quintiles), that affected all the employment structure but was strongly biased towards the lowest segments of employment. In other European coun-tries, either there was no destandardization (as in Finland, Denmark or the UK), or such destandardization was not so strongly biased down-wards, and therefore it could not generate polarization (as in Sweden or Italy). Nevertheless, there are important exceptions to such a general pattern (e.g. Cyprus, Ireland or Austria, for different reasons), so this idea must remain at the level of a line of orientation for further research but one which seems rather plausible: the de-standardization of employ-ment illustrated in Figure 3.6 would be capturing a series of changes in employment regulation that purposely facilitate the creation of low-paid, low-quality jobs. It should be no surprise if such changes would contribute to a polarization of employment.

3.6 Conclusions

In this chapter, we have studied the patterns of change in European employment structures from the perspective of job quality, using a rel-atively new methodology (the jobs approach ) and applying it to a big sample of countries. Our results show that, between 1995 and 2007,


there was a considerable diversity in the patterns of job expansion in Europe: some countries experienced a polarization, others a clear struc-tural upgrading, yet others a relatively flat expansion tilted towards the mid–high layers of employment. Such national diversity has important implications for the debate on the impact of technical and organiza-tional change on the nature of work, because it undermines the case for a single dominant or universal driver of changes in the employ-ment structure, suggesting that such impact is subject to a fundamental indeterminacy that is often absent in the literature. Although not in a totally conclusive manner, our results suggest that the existing insti-tutional diversity within Europe (an expression of conscious human agency) is one of the key determinants of such plurality of patterns of job expansion.

Assuming this fundamental diversity, we tried to evaluate empirically the predictive success of each of the three hypotheses of change that have dominated the Social Sciences debate on this issue. The most suc-cessful is, without any doubt, the structural upgrading hypothesis pos-tulated by the theories of post-industrialism and skill-biased technical change. In almost all countries, including some which experienced a simultaneous polarization, the highest job-quality segments grew more strongly than the rest of the employment structure. Although in this sense the structural upgrading was nearly universal, its actual form var-ied in crucial ways: there were some important cases of simultaneous polarization, while in other cases the upgrading took place together with an expansion of middling jobs.

Still, the more or less generalized trend towards a structural upgrad-ing in employment is an important finding, with interesting implica-tions. Similar research in the US shows that such structural upgrading was a constant in the US labour market over the last four decades at least (see Wright and Dwyer 2003), though with an increasing level of simul-taneous polarization over the years. It may be that the natural develop-ment of employment structures in capitalist economies is a continuous structural upgrading, as long as there is productivity growth (after all, both phenomena are related). If that is the case, it could be argued that this pattern should be taken as a given, and our focus should shift towards deviations from it (e.g. cases of polarization) or towards differ-ences in the degree or type of upgrading. (We have seen a wide variety in this chapter.) But for the time being, these cannot be more than more or less informed speculations: future similar studies should pro-vide more evidence on the nature of structural change in employment in the long run.


Structural upgrading in the employment structure, even in terms of job quality, should not be naively equated with social progress: they are totally different things, which can move in opposite directions. In fact, the type of structural upgrading analysed in this chapter has been often associated with increasing social inequalities and/or social exclu-sion (it is, in fact, one well-noted consequence of the skill-biased tech-nical change argument). If low- and mid-paid jobs are stagnant or even contracting, workers with low qualifications will have a hard time find-ing employment, and their wages and conditions of employment may deteriorate because of the competition of more people for fewer jobs. If simultaneously the higher layers of employment are expanding faster, wage inequality between high- and low-qualified workers is likely to grow. In fact, there is abundant evidence that this kind of process is one of the most important drivers of the current trend of increasing income inequalities in most developed economies (for a review, see Acemoglu 2002).

While there is no inexorable social progress in the type of structural upgrading discussed in this chapter, equally there is no inexorable social deterioration. The wider social implications of technical and organiza-tional change depend, in the last instance, on how society decides (in one way or another) to distribute the benefits of such change. There is nothing inevitable, good or bad, in this respect.

Notes

1 . Wright and Dwyer (2003: 205) provide an apparently different justification of this approach: ‘jobs are not just employment contracts to “work” at a given earnings level: they are contracts to perform sets of tasks to produce specific outputs’; ‘a job type ... , can be thought of as demarcating labour market opportunities for a particular kind of employment with a particular earnings potential’. In fact, what lies behind the task dimension of jobs and the opportunities associated with jobs is the same principle of the division of labour discussed above.

2 . These arguments are a bit forced in this context, because orthodox econom-ics always refers to individuals and not jobs: such an approach is probably incompatible with our structural approach to labour market analysis. But if it was not, it could be argued that beyond endogenous factors which are not of interest here, most of the differences in wage levels within a job would derive from the preferences of each worker for wage or unpleasantness: hence, as long as there is no bias in the preferences of workers across differ-ent jobs, the median wage level could be a good approximation to overall job quality at the job level. The differences in median salaries across jobs would reflect indirectly their differences in productivity, as would (even more directly) the average educational level of job-holders.


3 . Similar arguments are advanced by proponents of the theory of skill-biased technical change (SBTC), but with a rather different tone: these theories try to explain why after the 1970s there was an important expansion of inequalities in most of the industrialized world. SBTC theories argue that around the 1970s there was a technological revolution that increased the demand of skilled labour and depressed the demand of unskilled labour, thus increasing the wage differential between skilled and unskilled workers (for a review, see Acemoglu 2002; also, Violante 2008).

4 . This exchange took place in the meeting of the JOBs project in Salamanca in July of 2009 (see Fernández-Macías 2009; Goos 2009).

5 . In the GMS paper, the use of a single ranking is not defended as a best option, but just as a necessity because of lack of data (Goos et al. 2009: 59).

6 . It is also not clear in the paper whether the grouping of jobs is only based on occupations (which is apparently the case, but then, why is employment broken down by occupation and sector, as it says in p. 59?) or on the combi-nation of occupation and sector (as we do in this chapter).

7 . The original (uneven) segments produce the following growth rates: 1.23 per cent for bad jobs, -9 per cent for middling jobs and 7.75 per cent for good jobs. Regrouping the jobs so that they are approximately equal in the first year (using the data from GMS 2009: 59; the new shares would be 32% bad, 34% middling and 34% good jobs), we would get the following growth rates: -2.15 per cent for bad jobs, -5.43 per cent for middling jobs and +7.58 per cent for good jobs.

52

4.1 Introduction

This book presents important new evidence on the patterns of employ-ment growth in Europe, expanding previous research that has largely focused on the United States. The results show not only striking con-tinuities across countries, in a significant pattern of job polarization in many of the largest economies in Europe, but also important het-erogeneity with some nations experiencing very different patterns of employment growth. This research is crucial for developing institu-tionalist theories of employment change in modern economies that move beyond the focus on technological change in prior research. As the various chapters develop, employment growth is shaped by differ-ences between countries in their populations, employment policies, welfare states, and gender relations, not just the technical exigencies of production in modern economies. Studying employment patterns also highlights differences between countries in economic develop-ment and sectoral concentrations that are all too often overlooked in broad claims about post-industrial economies. No doubt, the division of labour in modern capitalism is defined by technological change, but the final shape of job structures is also directed by local and political circumstances that differ between places. In this chapter we compare job growth in the United States and Europe, focusing on variations in the degree of polarization.

4 Job Growth and Job Polarization in the United States and Europe, 1995–2007 Rachel E. Dwyer and Erik Olin Wright

Job Growth and Job Polarization 53

4.2 Theories of job growth and job polarization in the US

The character of employment growth critically reflects the prospects of any modern economy and has become a matter of some debate in turn of the 21st–century-modern economies (Ilg 1996; CEA 1996; Farber 1997; Autor et al. 2006). Economic restructuring in the US in the last decades of the 20th century significantly altered the employment structure, but scholars differ in assessing the impact on the quality of the jobs created. Some argue that the rise of a ‘new economy’ based on high-technology skills and high-end services has created an abun-dance of good jobs that are more fulfilling and autonomous than those provided in old-economy bureaucracies (Florida 2002). Others focus on the proliferation of low-wage service jobs that have few benefits, little security and restricted opportunities for advancement, and worry that workers without a college education have very little hope of attain-ing a decent standard of living and upward mobility for their children (Meisenheimer 1998; Kalleberg et al. 2000). Studies of employment growth support elements of both narratives of change – job growth was concentrated in both the highest and lowest wage positions (Reich 1992; Wright and Dwyer 2003; Autor et al. 2006). The new economy in the US thus generates good jobs and bad jobs and the nature of economic change looks very different depending on which end of the employ-ment structure is in view.

What is sometimes overlooked in the divergence between discussions of the growth of jobs at the top versus the bottom, however, is that neither view expects much growth in the middle . Indeed perhaps the most striking feature of economic restructuring has been the decline of middle-wage jobs, driven especially by de-industrialization and the col-lapse of manufacturing employment in the US (Harrison and Bluestone 1988; Ilg and Haugen 2000). This pattern of change has raised the spectre of a transition to an hourglass economy, where job growth is concentrated at the top and the bottom of the wage structure but not in the middle (Massey and Hirst 1998). The anaemic growth of middle-wage jobs is particularly worrying because these jobs were crucial to the expansion of the middle class in the post-war period (Hacker and Pierson 2010). While the best jobs are clearly at the top, it may be the ‘merely’ good jobs in the middle that were really key to making mod-ern America (Levy 1998). Middle-wage jobs may also serve as a bridge from the bottom to the top, and their decline may contribute to wors-ening prospects for social mobility in the United States (Rytina 2000).

54 Rachel E. Dwyer and Erik Olin Wright

Rising earnings inequality is the result, and indeed the majority of wage growth went to the most affluent employees in the 1980s and 1990s while middle- and lower-income workers suffered wage stagnation and even decline (Morris and Western 1999; Mishel et al. 2007).

Understanding job polarization is key to any effort to improve the life chances of middle- and lower-wage workers and build an economy that produces widely shared growth instead of gains for only a few. The trend has until recently received relatively little attention from scholars, however, and many important questions remain unanswered. Most importantly, the causes of job polarization are still unclear. While there is compelling evidence to support the influence of technolog-ical change, this explanation only explains some of the features of job polarization and is especially silent on the institutional condi-tions of employment in the United States and other places that have encouraged polarization (Card and DiNardo 2002; Autor et al. 2006; Lemieux 2008; Oesch and Menes (2011). Cross-national analysis is crucial for building an institutionalist theory of employment growth and this book makes a significant contribution to understanding how different national environments have shaped economic development under relatively similar technological developments. While there is a strong current of polarization in many European countries (Goos and Manning 2007; Goos et al. 2009), the research reported in this book uncovers important sources of heterogeneity in the quality of jobs cre-ated in turn of the 21st–century-capitalist economies. Identifying this national variation is crucial for developing policy that addresses the employment structure where people live.

In this chapter we compare job polarization in the United States to patterns of employment growth in Europe. We begin by discussing details of our method, which we have worked to make as comparable as possible to the jobs project database, given differences in the survey instruments. Then we present the results of our analysis, following four threads of comparison, drawing to varying degrees on earlier analyses in this book. First, we compare national patterns of job growth in the US and Europe, extending the discussion of the overall reach of job polari-zation across different national settings in Chapter 3 . Second, we exam-ine differential patterns by gender, race and country of birth, building on the analysis of gendered patterns of employment growth in Chapter 5 and the discussion of variation in immigration patterns in Chapter 6 . Third, we follow many earlier work on the subject in considering indus-trial patterns of growth, focusing on the key sectors of manufacturing, services and construction. Finally, we reflect on the proper scale of the


comparison between the US and Europe given the differences in size and integration of European national economies. We scale up to the EU and down to US regions to improve our understanding of the generality and heterogeneity of job polarization.

4.3 Methodology

We modify the methods we used in previous research (Wright and Dwyer 2003) to make our analysis more comparable to the European analysis. Some unavoidable differences remain, as is the case even within Europe, and we try to signal where those differences may affect our comparisons.

4.3.1 Data

The US data are from the Current Population Survey (CPS) annual outgoing rotation group (ORG) files for 1995 through 2007 collected by the US Bureau of Labor Statistics (BLS), the main data source for analyses of job polarization and wage inequality (Autor et al. 2003; Kim and Sakamoto 2008; Mouw and Kalleberg 2010). The CPS is con-ducted monthly, but respondents to the CPS rotate into the sample for four consecutive months, exit for four months and return for a final four months. The survey collects detailed wage data for households in their fourth and eighth months, the ‘outgoing rotation groups’. The analyses apply the BLS weight constructed for the earnings sample of the ORG.

We restrict the sample to jobs held by employees aged between 18 and 65. The CPS does not collect earnings data for the self-employed that is comparable to the data for employees, and thus it is difficult to create similar wage estimates for this segment of the labour force. We exclude self-employed workers in our employment estimates (unlike in the European studies) but self-employment is a relatively small percent-age of the US labour force, fluctuating between 7 and 8 per cent of the workforce over most of the period of this study (Hipple 2010). Like the Europe study, we include both full-time and part-time workers, counted as persons, not hours.

4.3.2 Measures

We define jobs as cells in an occupation by industry matrix, following Wright and Dwyer (2003). Jobs are more precisely captured by the cross-classification of occupation and industry than by occupation alone in part because occupations are defined at the national level with little


attention to variability across places in the details of work tasks and requirements (England et al. 1996; Tomaskovic-Devey and Skaggs 2002; Cohen and Huffman 2003). While occupations are clearly important categories of class position in developed economies, industrial sector still determines much about the skills, earnings and work conditions of jobs even within occupations. Inequality has increased within and between occupational categories since 1980, and some within-occupa-tion inequality may be due to between-industry differentiation (Kim and Sakamoto 2008; Mouw and Kalleberg 2010).

Occupations . The CPS follows the US Census coding scheme for occu-pations, which changes over time as the employment structure changes. There are two major coding regimes in the CPS over the 1995–2007 period – one, 1990s, based on the 1990 Census and another starting in 2000. 1 We use a Bureau of Labor Statistics crosswalk to make the 2000 occupational categories comparable to the 1990s categories so that all analyses are based on a similar coding scheme (Meyer and Osborne 2005). In order to limit the number of small cells, we collapse occu-pations into about 100 categories instead of the over 300 in the BLS scheme, but sensitivity analyses show that the pattern of results is sim-ilar when using three-digit occupations alone.

Sectors. The coding scheme for industry changes much less over time than the codes for occupation. We code industries into 23 categories. The 23-category classification is based on the standard 22-category scheme but splits ‘business and repair services’ into two separate sectors.

Job median wages. We measure job earnings using hourly wages, which is weekly earnings divided by usual hours worked per week for salaried workers. We convert earnings into constant 2007 dollars using the stan-dard Bureau of Labor Statistics CPI-U-X1 series. We combine all years within each analysis period and then calculate median earnings in con-stant dollars. We use median instead of mean to avoid concerns about top-coding in the CPS, and to limit the effect of outliers on wage calcu-lations. Combining years creates a very large sample for analysis so that there are significant numbers of cases in nearly all cells of the matrix, making the estimates of job median wage more valid and reliable. This procedure also means that if earnings change in a cell over the period of analysis, the rank-order position of the job in the hierarchy of job quality will be based on a weighted average of the earnings over the period (weighted by the number of people in the job in each year of the CPS sample). It turns out that although median earnings of job types do change over time, the rank-order of cells changes hardly at all, which is the key in this analysis as we explain next.


4.3.3 Strategy of analysis

As in the Europe studies, we analyse the amount of job growth across the wage structure. For much of the analysis, we group jobs into quin-tiles of median-hourly earnings and compare the share of growth across quintiles. This is a transparent method of identifying job polarization between the best and worst jobs. First, we rank-order jobs from the low-est- to the highest-median hourly earnings and then group them into five ordered-categories each containing about 20 per cent of the employ-ment in 1995. 2 The bottom quintile contains the roughly 20 per cent of employment that are in the jobs with the lowest-median earnings in 1995, the highest quintile contains the roughly 20 per cent of employ-ment in jobs with the highest-median-weekly earnings and so on. Then we calculate the net change in the number of jobs in each quintile from the beginning of the period to the end – this will show whether job growth was concentrated among high-paying or low-paying jobs, or whether it was more evenly spread. We do not drop small cells, since they contribute little to the overall pattern of change and any errors intro-duced by the small sample in estimates of wages or employment levels are likely randomly distributed (England et al. 1996). (In fact, sensitivity analyses that do drop very small jobs show almost identical patterns of results.) It is worth reiterating that the numbers in these graphs refer to net job expansion rather than job creation per se. That is, employment change involves both the creation of new jobs and the destruction of old jobs. If a particular cell in the occupation-by-sector job matrix increased by 10,000 over a period, this could mean the creation of 25,000 new jobs and the destruction of 15,000 previously existing jobs – we simply observe the net effect of these two processes. Note that this is also differ-ent from the number of job openings. Even a declining job category may have many openings as workers retire or move on to other jobs.

We analyse the patterns of job growth separately by sex and race as well as by occupational and sectoral groups as defined above. We con-struct consistent racial categories of non-Hispanic white, non-Hispanic black, non-Hispanic other race and Hispanic all races. The category ‘non-Hispanic other race’ is heterogeneous, but largely made up of Asian populations.

4.4 National patterns of job growth in the US and Europe

As previous research documented, employment growth in the United States is significantly polarized (Wright and Dwyer 2003; Autor et al.


2003; Autor et al. 2006). Figure 4.1 reports the pattern of US job growth over the period of the European study from 1995 to 2007, showing that jobs grew significantly at the top and the bottom of the employment structure, with very little growth in the middle. Job growth is weighted to the top, with the highest levels of employment growth in the top two quintiles, but there is substantial growth in the bottom quintile as well and a significant trough in the middle. This pattern is strikingly sim-ilar to our earlier findings for the 1990s, illustrating that polarization persisted into the 2000s economic expansion. There is somewhat less growth at the top for the entire 1995–2007 period than in the 1990s, as the high-technology boom that fuelled growth at the top cooled off (Wright and Dwyer 2003).

The pattern of employment growth in the United States is similar to the most polarized European countries in the Netherlands, France, Cyprus, Germany, the UK and Ireland. All of these cases show a pat-tern of polarization, with most showing some weighting to the top. Earlier chapters have linked these patterns with deregulating and de-standardizing labour markets in many of these countries, so it should be no surprise that the relatively low-regulation environment of US labour market institutions should produce a strikingly polarized pattern.

0

1,000

2,000

3,000

(000

s)

4,000

5,000

1 2 3 4 5

Figure 4.1 Job growth in the United States by job-wage quintile, 1995–2007


Another commonalty across countries that has been given perhaps less attention in the rest of the book is the anaemic job growth in the middle quintile. All the polarized countries including the US and the European cases with high or moderate polarization (in the first two columns of Figure 3.1 from Chapter 3 ) experienced less growth in the middle quintile than at the top or bottom, including countries where growth was more weighted to the top. Growth in the second or third quintiles was lower than in any of the other quintiles in all but two of the highly or partially polarized countries, and those two (Belgium and Hungary) still showed much lower growth in the middle than the top. Even countries with an upgrading pattern of employment growth with little expansion in the bottom quintile experienced quite low levels of job growth in the middle. In fact, only Spain, Portugal and Greece experienced growth in the mid-dle quintile that was at all on par with growth at the top and bottom.

Slow growth in the middle raises concerns about the structure of opportunities in an economy – whether or not there is substantial growth at the top or bottom of the labour market. One of the biggest worries about post-industrial economies is that they do not produce sufficient middle-wage jobs to support the maintenance of the large middle class that developed the United States and many European countries during the post-war period. The trough in the middle seen in both the United States and Europe vividly illustrates the source of these concerns. Opportunities for social mobility may also be dimin-ished in employment structures with growth only at the ends of the wage distribution. In the post-war period, middle-quintile jobs were the solid manufacturing and office jobs that enabled many people with high school educations to achieve decent pay and a middle-class stan-dard of living (Wright and Dwyer 2003). These jobs did not require especially high skill levels, but received decent compensation as a result of the institutions of the Fordist system like unions, internal labour markets, and relative job security (Bluestone and Harrison 1982). These institutions have been challenged in many countries by political shifts and pressures from globalization and technological change. The consis-tency across countries in the decline in the middle quintile illustrates how broad the institutional challenge has been.

Earlier chapters have suggested that the degree of job polarization in a country is shaped by demographic forces, and economic development reflected in the industrial composition of growth. Next we consider the US case as compared to Europe, beginning with an exploration of job growth by gender, race and nativity, and then moving on to the sectoral pattern of growth.


4.5 Job growth by gender, race and nativity

The degree of job polarization varies not only across countries, but also between different population groups within the same country. In fact, as earlier chapters have suggested, some of the variation between nations in patterns of employment growth appears to be linked to differences in demographic trends and population compositions. The US case pro-vides further support for this hypothesis, but additional research will be required in order to assess the causal order between employment growth and demographic trends. There are significant differences in patterns of employment growth by gender, race and nativity in the US.

Employment growth was polarized for both men and women in the United States in the 1990s and 2000s, though with some important dif-ferences. The overall pattern fits into Grimshaw and Figueiredo’s category of ‘gender parity – polarization’ similar to the Czech Republic, Slovakia, Estonia, France and the UK. The US is also similar to these countries in most job growth for men and women in full-time work, rather than women being concentrated in part-time work as in Germany, Austria and the Netherlands. The pattern for women is also quite similar to countries in the category ‘gender divide – female polarization, male upgrading’, especially in Ireland and Spain. Women saw strong growth at the top in many countries, consistent with other evidence of women’s gains in the labour market over this period (Neckerman and Torche 2007). What varied more was whether this was balanced by growth at the bottom. The US appears to be similar to other liberal countries like the UK and Ireland in seeing polarized growth for women ( Chapter 5 ). Polarized job growth among men in the US is even more distinctive as only one of the five classifications of gender patterns (the gender parity–polarization classification) shows evidence of male polarization.

Still, there are differences in the pattern of job growth between men and women in the US, just as in most European countries. As Figure 4.2 shows, the trough in the middle for women was entirely in the second and third quintile (with stronger growth in the 4th quintile), whereas men saw lower growth in the second, third and fourth quintiles. The trough is deeper for women, however; men experienced stronger growth in the second and third quintiles, while there was almost no job growth for women in those quintiles. This pattern of findings is significant because much of the worry over the decline of middle-wage jobs in the US has been expressed as a factor of de-industrialization, which has hit men’s employment much harder than women’s in the US (Bluestone and Harrison 1982; Reich 1992). This analysis shows that women too


had little growth in the middle, even less than men in the second and third quintiles from 1995 to 2007. In supplemental analyses we find that the strength of men’s growth in the middle in this period is largely due to the construction sector, a point we will return to below. Autor et al. (2006) suggest that women’s declining fortunes in the middle are linked to the mechanization and outsourcing of routine clerical work, a factor that may be particularly important in a country like the US with low unionization rates and weak labour market regulation.

The pattern of employment growth was even more differentiated by race and nativity status than by gender. Figure 4.3 shows that employ-ment growth among Hispanic immigrants was significantly concen-trated towards the bottom quintiles, while non-Hispanic immigrant growth was moderately polarized, with most growth at the top. Among US-born workers, employment growth for white workers was entirely in the top two deciles, with declining growth in the middle. This is in fact a form of polarization between the highest-wage jobs and all other jobs, though in a different pattern than the U- or J- shaped polar-ization that has received so much attention. US-born Hispanic growth was concentrated at the bottom and in the middle, with a much more upgraded pattern of growth than for Hispanic immigrants. This sup-ports evidence that Hispanics experience significant upward mobility across generations and after the immigrant generation, though whether

1,00

02,

000

3,00

04,

000

0

1 2 3 4 5

Women

(000

s)

01,

000

2,00

03,

000

4,00

0

1 2 3 4 5

Men

(000

s)

Figure 4.2 Job growth in the United States by gender, 1995–2007

01,0002,000

12

34

5

His

pani

c im

mig

rant

s

01,0002,000

12

34

5

Non

-His

pani

c im

mig

rant

s0200400600

12

34

5

US

-bor

n H

ispa

nic

-1,00001,0002,000

12

34

5

US

-bor

n w

hite

0500

12

34

5

US

-bor

n bl

ack

0200400

12

34

5

US

-bor

n ot

her

race

(000s)(000s)

(000s)(000s)

(000s)

(000s)

Figu

re 4

.3

Job

grow

th i

n t

he

Un

ited

Sta

tes

by r

ace

and

nat

ivit

y, 1

995

–20

07


this will persist for future cohorts is an open question (Bean and Leach 2004). Finally, US-born Black and Other Race workers show a moder-ately polarized pattern of employment, with higher growth at the tails than the middle.

Immigrant incorporation into the US labour market shows elements of what were two distinct patterns between countries in Europe. Clearly, the US is like most European countries in the large concentration of immigrant job growth at the bottom of the job–wage structure. The US also sees relatively strong growth at the top, however, like Austria, Belgium, and Luxembourg. The reason is the heterogeneous character of immigration into the United States (Bean and Leach 2004). Large numbers of poor and low-educated immigrants arrive and take the low-level service and labour jobs at the bottom of the wage structure, especially Hispanics. At the same time, the US has fairly robust immi-grant among highly educated middle-class populations, especially from Asian countries and find jobs at the top of the wage structure. These highly divergent patterns for different groups of immigrants mean that the foreign-born population overall is more evenly distributed across quintiles than the US-born population, as Figure 4.4 illustrates.

Strikingly, foreign-born workers account for almost all the growth that occurred in the middle quintiles over this period. Jobs that declined in the middle were disproportionately held by US-born workers. These

02,

000

4,00

06,

000

1 2 3 4 5

Foreign-bornUS-born

(000

s)

Figure 4.4 Job growth in the United States by nativity, 1995–2007


differential patterns of growth by race and nativity are related to impor-tant sectoral shifts in the US economy that have different implications for different population groups.

4.6 Sectoral patterns of growth

Industrial growth and decline in the United States look similar to Europe, but with some distinctive characteristics that suggest the strength of job polarization in the US is linked to the highly de-regulated institu-tional environment relative to the majority of European countries. We discuss manufacturing, knowledge-intensive services, low-skill services and construction by way of comparison with the sectoral analysis of European countries. 3

Figure 4.5 shows the pattern of growth in durable and non-durable manufacturing as a part of overall job growth. As in Europe, manufactur-ing did not contribute to job growth in the late 1990s and 2000s, in a striking departure from the strength of manufacturing earlier in the post-war period. The US is distinctive, however, in experiencing decline in manufacturing across the board – for both durable and non-durable manufacturing, and across every quintile. Most European countries saw

-2,000

0

2,000

4,000

6,000

1 2 3 4 5

Manufacturing, durable Manufacturing, non-durable Rest

(000

s)

Figure 4.5 Contribution of manufacturing to job polarization, 1995–2007


some employment growth in manufacturing – especially at the top – but manufacturing only brought net job decline in the US. 4 Manufacturing decline was concentrated in the middle quintiles, and Figure 4.5 illustrates that continuing de-industrialization in employment clearly held down growth in the middle without being replaced by growth in a rising sector.

The reason growth in rising sectors has not made up for declin-ing manufacturing is that the growing service sectors were polarized between the top and bottom of the jobs structure in the United States. Figure 4.6 shows the top-heavy growth in knowledge-intensive services and Figure 4.7 shows growth at the bottom in low-skill services, vividly illustrating the polarizing pressures in the US labour market. What’s more, these two large sectors account for a large percentage of over-all growth. Knowledge-intensive services accounted for 88 per cent of growth in the top quintile and a striking 97 per cent of growth in the fourth quintile. Low-skill service jobs make up 83 per cent of growth in the bottom quintile. European countries show a similar pattern to employment growth in knowledge-intensive and low-skill services, but with a variation in the level of growth, especially at the bottom. Knowledge-intensive services tended to grow at the top and grew fairly strongly everywhere. Low-skill services was typically weighted to the

0

1,000

2,000

3,000

4,000

5,000

1 2 3 4 5

Knowledge-intensive services Rest

(000

s)

Figure 4.6 Contribution of knowledge-intensive services to job polarization, 1995–2007


bottom and grew substantially in some places like the Netherlands, France and Germany, but grew much less strongly in other places like Finland, Sweden and Denmark. In fact, the degree of growth in low-skill services largely differentiates countries between highly polarized, moderately polarized and upgrading patterns of growth. Some of the countries that see more of an upgrading pattern also show a stronger pattern of growth in low-wage services in the middle of the quintile, like Spain and Portugal.

This discussion suggests that the fate of the less knowledge- intensive service sector will be important to understand variation in job polarization between countries. The role of small business and self- employment may be key here, as some places may see the retail and household service sector operated by more highly paid small proprie-tors than the low-wage unskilled labour that dominates in the US. Clearly, the institutions that regulate the low-wage labour market will be important for understanding variability in service job growth here as well (Osterman 2008).

Finally, the construction sector in the US is concentrated in the middle quintiles, just as in Europe, suggesting that unless some low-skill service jobs begin to upgrade, growth in the crucial middle of the jobs structure

0

1,000

2,000

3,000

4,000

5,000

1 2 3 4 5

Low-skill services Rest

(000

s)

Figure 4.7 Contribution of low-skill services to job polarization, 1995–2007


will be increasingly affected by the cycles of boom and bust characteris-tic of real estate. As Figure 4.8 illustrates, job growth in the construction sector in the US was highest in the second, third and fourth quintiles, with slight growth at the top and a small decline at the bottom. The construction sector made up about one-third of job growth in the mid-dle quintile and almost three-quarters of growth in the second quintile. Again, this same pattern of job growth in construction holds in Europe, with variability in levels of growth. Places that saw large growth in con-struction were less likely to show a polarized pattern of employment change. This suggests that as manufacturing declines, job growth in the middle quintile may be increasingly affected by the cycles of construc-tion, in the US and as in Europe. This has implications for the gender structure of the labour market as well since the vast majority of jobs in construction are held by men, contributing to the much larger trough in the middle for women compared to men discussed earlier.

In Europe, the strength of employment growth in the construction sector varied significantly across countries depending on the amount of growth in the housing sector, and this explained quite a bit of the vari-ation in the degree of job polarization as well. The housing boom in the United States also varied across the country, suggesting that the pattern of job polarization likely varied across regions. In the next section we compare the US and Europe at different geographic scales.

0

1,000

2,000

3,000

4,000

5,000

1 2 3 4 5

Construction Rest

(000

s)

Figure 4.8 Contribution of construction to job polarization, 1995–2007


4.7 The EU, the US, and regions: reflections on geographic scale

We have compared the US to individual European countries, but it is also useful to consider the European Union as a whole given its increas-ing economic integration, as well as the subdivisions of the United States into its economically diverse regions. The geographic scale of comparison affects the amount of variation that can be captured in the analysis. It also affects the political response to patterns of job growth as different levels of governance deal with different levels of geographic scale.

Job growth for all EU countries included together followed an asym-metrical pattern of polarization weighted to the top quintiles that looks very similar to the US as a whole. As Figure 4.9 shows, in the EU jobs grew most in the very top quintile, with strong growth in the fourth and bottom quintile, and weaker growth in the second and third quin-tile. (This figure is reprinted with permission from Fernández-Macías (2010).) Job growth in the EU is more weighted to the top than in the US, with less growth at the bottom, but similarly low growth in the mid-dle. The regional breakdown shows that Continental Europe strongly led polarization, as is evident in the country analysis. This analysis also highlights the relative contribution of large and small countries to the pattern of job polarization. Because countries in Continental Europe are so large, they strongly affect the EU pattern of growth, even though their relative rate of employment growth is slower than some of the smaller countries such as Spain or Ireland. It will also be important to think about whether and how more desirable patterns of job growth can be scaled up from smaller countries and regions.

The United States is a large and heterogeneous country and so the national pattern of employment growth may mask important regional differences, though this has not been studied in prior research. In the next analyses we separate the results into the four major regions of the country – the Northeast, Midwest, South and West – which have had different historic patterns of migration, demographic composition and industrial development. 5 Economic restructuring varied significantly across the country, with the Northeast and Midwest bearing the brunt of de-industrialization and the West reaping the benefits of the high-technology boom (Harrison and Bluestone 1988; Milkman and Dwyer 2002). In this analysis we use the national job quintile ranking since the US has an integrated labour market and an overarching federal system of employment regulation. Figure 4.10 shows that there is important


01,

000

2,00

0

1 2 3 4 5

Northeast

01,

000

2,00

0

1 2 3 4 5

South

(000

s)(0

00s)

01,

000

2,00

0

1 2 3 4 5

Midwest

01,

000

2,00

0

1 2 3 4 5

West

(000

s)(0

00s)

Figure 4.10 Job growth in the United States by region, 1995–2007

8,000

6,000

(000

s)

4,000

2,000

0

Cont East North South Ukie

Figure 4.9 Job growth in the European Union, stacked by region, 1995–2007

Note : This figure is reprinted from Fernández-Macías (2010).


regional variation within the United States, with two major patterns of employment growth.

The Northeast and Midwest experienced strongly polarized job growth since 1995. Job growth was weighted to the top in both regions, with the top two quintiles growing strongly, but comparably strong growth at the bottom as well. Most striking is the almost total lack of growth in the second and third quintiles, a very deep trough that is produced by significant declines in manufacturing employment in these old industrial regions. This pattern is closest to the polarized job growth in the Netherlands, France, and to some extent, Germany – all major economies at the end of the 20th century as well.

Job growth in the South and West was more evenly distributed in a pattern of muted polarization. The top and bottom quintiles show the strongest growth in these regions, consistent with the polarized pat-tern, but growth was much stronger in the middle compared to the Northeast and Midwest. This pattern of job growth is less common in Europe, but looks quite like the pattern in Ireland, and it is similar to Belgium in the middle and top.

It is interesting that job growth in the older post-war economic pow-ers in Europe was most polarized, like in the older industrial regions of the US. Countries in the middle of more robust economic development appear to see more growth in the middle like in the Sunbelt regions in the US. Indeed, the South and West experienced more than twice the amount of job growth than the Northeast and Midwest (the y axis is not standardized across the region graphs). Thus, the level of economic development during the post-industrial transition may be an impor-tant factor in explaining variability in the pattern of job growth.

The construction sector also plays an important role in regional dif-ferences in growth in the middle in the US as in Europe. Many of the countries where jobs grew most in the middle in Europe were places with strong construction booms, such as Ireland, Spain and Portugal. In the US too, differences in the level of construction growth were due to differences in the level of growth in the middle. Figure 4.11 shows that while jobs in the construction industry grew in the middle quintiles in all regions, growth occurred at much higher levels in the South and West than in the Northeast and Midwest (with a standardized y axis this time). This reinforces the findings in the sectoral analyses of both the US and European countries that construction appears to play a sig-nificant role in job growth in the middle.

This analysis illustrates some of the complexities of developing an institutionalist understanding of job polarization. Different patterns of


job growth across countries and between regions within countries are likely linked to different state structures and regulations, but they also emerge from industrial specializations that shape economic develop-ment. A more nuanced understanding of these issues will require devel-oping careful comparative studies that match countries and regions in something like a case-control design to help identify the source of variation in employment patterns (Byrne and Ragin 2009). The European jobs project database will be an invaluable resource in any such endeavour.

4.8 Conclusion

This chapter reinforces the theme in the rest of the book that patterns of job growth vary significantly across modern economies, demonstrating that institutional structures and differences in economic development shape employment growth, not just a universal technological change in capitalist economies. Even the commonalities across countries point to the importance of different political economic arrangements in shaping employment growth. Decline in the middle of the job–wage structure

0

100

200

300

400

500

1 2 3 4 5

Northeast

0

100

200

300

400

500

1 2 3 4 5

Midwest

0

100

200

300

400

500

1 2 3 4 5

South

0

100

200

300

400

500

1 2 3 4 5

West

(000

s)(0

00s)

(000

s)(0

00s)

Figure 4.11 Job growth in the United States in the construction industry by region, 1995–2007


is quite common, and further research is needed to understand what explains the more robust growth in the middle in some places. While growth at the top and bottom have received some of the most attention in previous research, it is the decline in the middle that most strongly marks the 1990s and 2000s as distinct from earlier decades in the post-war period (Wright and Dwyer 2003).

In fact there is wide agreement that improving the prospects of the middle class will require encouraging demand in middle-wage jobs. In the US policy circles it is quite common to focus on restoring manufacturing and encouraging physical infrastructure investments as a major source of growth of non-routine manual jobs in part because these are seen as not as likely to be moved offshore (Blinder 2007; Baily 2010; Pollack and Thiess 2010). There are many good reasons to pursue this strategy, and indeed the findings here supports the value of sup-porting construction jobs in particular, given the importance of this sector to growth in the middle in both the US and Europe. This growth was of course fuelled in many countries by a housing bubble that has since had devastating consequences for the global economy, but there are other sounder investments that would support these middle-wage jobs, such as infrastructure investments to repair ageing transportation system, or transitions to green technologies.

It is also important to pursue efforts to improve jobs in the expand-ing service occupations and sectors that have shown relatively robust growth but remain at the bottom of the earnings structure. These jobs too are less likely to be moved offshore because they require face-to-face interaction and so cannot be moved to places with lower labour costs (Blinder 2007; Moncarz et al. 2008). Many manufacturing jobs histori-cally began as bad jobs, with low pay and poor working conditions, and it was only the long struggle of organized labour to improve conditions that those became jobs that helped to build the largest expansion of the middle class in world history. Manufacturing jobs were also improved through the recognition of political leaders and industry heads that they depended on middle-class consumers. The housing crisis of the beginning of the 21st century holds many of the same lessons as the economic cataclysms at the beginning of the 20th century – that eco-nomic insecurity among the great mass of Americans threatens not only their life chances, but the holdings of elites as well. And specu-lative growth is far inferior to expansions based on the development of new technologies and goods and services that improve everyone’s standard of living, not just the bank accounts of financiers. Raising up service industry jobs will not be easy, but this may be crucial to slow


job polarization and remake the American middle class just as improv-ing manufacturing jobs did so in the past. Unionization and increasing the credentials required for service jobs, especially those related to edu-cation and health care, have been successful strategies in the past for improving wages of jobs such as teachers and nurses may be effective strategies (Milkman et al. 1998; Weeden 2002).

Policy efforts must also be sensitive to the differential impact of the polarized economy in different places. This analysis has shown that the polarized economy affects every region in the US, but not all in the same way. The sharp regional differences emphasize the variable impact of economic restructuring in different parts of the country. The sub-national level, in fact, is where much of the most significant economic changes occurred at the end of the 20th century and beginning of the 21st century (Harrison and Bluestone 1988; Jenkins et al. 2006; Lobao and Hooks 2007). The Northeast and Midwest are still rusting, even as there are positive signs of growth in other sectors. The West and South hold some promising developments, but would show pronounced job polarization without the growth of construction during the housing bubble. Future research and policy should also examine differences within regions. Milkman and Dwyer (2002) show, for example, that there were significant differences in job growth patterns between San Francisco and Los Angeles, linked to different industrial bases and Peck et al. (2009) find a distinct pattern in Chicago (see also Bernstein et al. 2002; Moller and Rubin 2008). Addressing the spatial dimensions of the polarized economy is especially important because industrial decline can become a downward spiral difficult to reverse as so many Rustbelt cities and towns in the US have discovered. Policy initiatives to address economic restructuring and rising inequality can learn from the dynamics that fill the top of the hourglass in some places and the bottom in other places. If the polarized economy is seen as a problem that can be solved rather than the inevitable result of economic restruc-turing, there may yet be a new day for the American middle class and new hope for upward mobility among the working poor.

These considerations are all the more important after the financial crisis of 2008 and the deep employment in the US and many European countries that followed. Crisis provides an opportunity to reconsider past policies and develop new responses to a changing economy. For too long, employment policy in the US has been an afterthought, little considered as an important focus for the federal government as the neo-liberal policy environment encouraged market solutions to social prob-lems. As we enter the third year of highly sustained unemployment, it


is well past the time to consider that the jobs picture will not improve without substantial government intervention. The studies here suggest that employment policy must consider not only the overall level of growth but also the quality and character of jobs created.

Notes

1 . In the United States, the 1990s expansion began in 1992 but job growth was anaemic for the first few years, with the majority of employment growth in the boom in 1995 and later (NBER 2010).

2 . Since jobs come in lumpy units, the quintiles do not amount to exactly 20 per cent of employment. None of the results is significantly affected by deviations from equal quintile categories.

3 . Knowledge-intensive services include FIRE (finance, insurance, and real estate), business services, hospital service, other medical service, educa-tional service, social service, other professional service and public adminis-tration. Low-skill services include retail trade, private household, personal and entertainment.

4 . The measures here are not strictly comparable to the sectoral analysis of European countries in Chapter 3 , which divided manufacturing into high-tech and low-tech categories. Most of the growth in manufacturing in Europe appears to have been in high-tech manufacturing and we might find more growth in the US in this category of manufacturing as well. Still, the differences noted here hold for the total pattern of job growth for the combination of high-tech and low-tech manufacturing in Europe compared to the combination of durable and non-durable manufacturing in the US.

5 . The Northeast includes Maine, New Hampshire, Vermont, Massachusetts, Rhode Island, Connecticut, New York, New Jersey and Pennsylvania. The Midwest includes Ohio, Indiana, Illinois, Michigan, Wisconsin, Minnesota, Iowa, Missouri, North Dakota, South Dakota, Nebraska and Kansas. The South includes Delaware, Maryland, the District of Columbia, Virginia, West Virginia, North Carolina, South Carolina, Georgia, Florida, Kentucky, Tennessee, Alabama, Mississippi, Arkansas, Louisiana, Oklahoma and Texas. The West includes Montana, Idaho, Wyoming, Colorado, New Mexico, Arizona, Utah, Nevada, Washington, Oregon, California, Alaska and Hawaii.

75

5.1 Introduction

The argument of this book is that the changing structure of jobs in a country, both across industries and between low-skill and high-skill occupations, is a critical factor shaping the nature of a country’s path of development and the distribution of opportunities for qual-ity employment. Currently received wisdom suggests that recent job change in many countries is characterized by polarization, with more ‘good jobs’ and more ‘bad jobs’, but fewer mid-ranking jobs in terms of skills required or wages paid (e.g. Goos and Manning 2007; Goos et al. 2009; Wright and Dwyer 2003). Mainstream economists argue that there is a universal logic at work, that advances in information and communication technologies increase the demand for high-skill jobs and growing economic wealth increases demand for non-routine service-sector jobs, while many routine and tradable jobs are offshored or outcompeted by imports from low-cost countries (Autor et al. 2006). In this chapter we take issue with this reasoning on two counts. First, it is not clear in the mainstream economics account why some workforce groups more than others are penalized in the changing distribution of job opportunities. Interrogation of the evidence for the US shows that the ostensibly neutral economic forces are in fact biased against certain socio-economic groups; while white male and female workers witnessed most job growth among the upper end of the job structure, Hispanic and black men and women saw net job expansion clustered

5 Women’s Changing Job Structure in Europe: Patterns of Job Concentration, Low Pay and Welfare State Employment Damian Grimshaw and Hugo Figueiredo

76 Damian Grimshaw and Hugo Figueiredo

among the worst-quality job types (Wright and Dwyer 2003). Second, the heterogeneous country patterns of job change reported in this book do not fit a universal explanatory model. There is evidence of a general upgrading, but this fits with a pattern of polarization in a minority of countries, hump-shaped change with large growth among middle-ranking jobs in others and a clear shift to middle- and high-quality jobs in others (see Chapter 3 ).

This chapter adds a further layer of detail to our knowledge about the causes and consequences of changing job structure by analysing the diverse patterns of changes in job structure among men and women in Europe. A key finding is that most countries experienced a gender divide in patterns of job change, but the pattern varies with combinations of male upgrading, female upgrading, male polarisation and female polar-isation with a strong role of part-time employment in some countries and not in others. Overall in all countries except Spain women’s job distribution improved relative to men’s. Furthermore, the changing job structure for women looks very different when jobs are ranked by skill as opposed to by wages. Women in many countries experienced upgrading of jobs on the skill measure but polarization of job structure on the wage measure. This supports the notion that many skilled jobs where women’s employment has grown are in fact undervalued in terms of the relative wage compared to men’s jobs. The chapter also identifies three factors that help in explaining some of the cross-country variation in patterns of women’s changing job structure. First, the strong concentration of wom-en’s employment into a small number of jobs in all countries means that tracking job change in the top-ten jobs (by size of female employment) sheds light on general patterns of job growth. The countries examined here share a common set of jobs with high female employment con-centration and, while each contributes strongly to overall job growth, comparison across countries reveals important differences in the rela-tive wage for similar jobs. Second, each country’s system for regulating low-wage work (involving a mix of minimum wage protection and col-lective bargaining coverage) is shown to influence the stock of low-wage jobs that employ men and women. There is no obvious sign, however, of a relationship between a country’s wage-setting institutions and the degree of change in the number of low-wage jobs. This finding accords with Freeman’s (2007) argument that labour market institutions shape distributional outcomes (i.e. the relative proportion of low-wage jobs) rather than economic performance outcomes (namely, job growth). Third, change in women’s employment in the welfare state sector – education, health and social-care jobs – is shown to be a major contributor to overall job change (following Esping-Anderson 1999 among others), especially

Women’s Changing Job Structure in Europe 77

in those countries with a well-developed welfare state. In most countries, the net impact of welfare sector job change is positive for women – with many jobs created at the upper end of the quality ranking. However, in countries with a liberal welfare state regime it is polarizing with strong job growth also recorded in the bottom quintiles.

5.2 Gender patterns underlying the changing job structure

Most countries in Europe experienced a marked gender divide in their changing job structure. However, as with the overall pattern of job change, there is no common pattern to gender differences. In some countries, polarization of job change among women was accompanied by a clear upgrading of job structure among men, while others display upgrading among women and polarization among men. Change in women’s part-time employment is also a dominant feature in several countries. Figure 5.1 sets out the different patterns, grouping countries into five categories.

Analysing change for 1995–2007 using job quintiles ranked by wages, a first pattern reflects polarization of women’s net job change with upgrading of men’s jobs. This conjunction of gendered patterns means women were not only over-represented among entrants into low-wage work during the period but also shared positive job opportunities with men at the upper end of the wage structure. This pattern is found in five countries – Ireland, Finland, Spain, Greece and Cyprus ( Figure 5.1a ). To illustrate, in Spain women accounted for three in four of net jobs cre-ated among the lowest two wage quintiles and one in two of the jobs in the upper two quintiles. Expressing a similar pattern in a different way, in Ireland while 41 per cent of all net job growth for women was found in the lowest two quintiles, for men this was only 21 per cent. A further notable feature of job change in this group of countries is the important role of part-time jobs for women in Ireland, Finland and Spain, with a stronger presence among the bottom two wage quintile jobs.

A second pattern of gender segmentation applies to four countries and involves relatively strong upgrading of jobs among women and a rather more mixed pattern among men ( Figure 5.1b ). Women’s net job growth occurs largely in the third, fourth and fifth quintiles. However, job growth for men is hump-shaped in Latvia and Lithuania and strongly polarized in Slovenia and Hungary. In the latter pair of countries, it is men not women who are over-represented among jobs created at the bottom of the wage structure; men account for three in four net jobs in the bottom two quintiles in Slovenia and almost nine in ten in Hungary.


A third type of gender divide applies to five countries where in each case changes in female part-time employment are the main driver of overall change in job structure ( Figure 5.1c ). A common pattern is the substitution of part-time jobs for full-time jobs among the lower quin-tiles – for men and women – and especially in Germany and Austria. Indeed, the pattern of net change in full-time jobs is relatively similar for men and women in these five countries. But it is clear that women’s part-time jobs dominate the picture. Expressed as a proportion of the total jobs created, the share accounted for by women’s part-time jobs ranges from 41 per cent in Luxembourg to 82 per cent in Austria. It is further notable that net growth in women’s part-time jobs appears to be distributed across all five job quintiles, ranked by wages, albeit with a tendency towards lower-job growth in the middle quintile in Germany and the Netherlands.

In the final two groups of countries, changes in job structure display no marked evidence of a gender divide. In four countries – Portugal, Denmark, Italy and Sweden – women share with men a general upgrad-ing of jobs characterized by strong job growth in the upper quintile categories and weak growth (or job losses) at the bottom ( Figure 5.1d ). In the other five countries men and women both experience a polar-ization of job change. France and Slovakia are the clearest exponents of this tendency. The other three countries combine loss of jobs at the bottom quintile with polarization between the second and fifth job quintiles ( Figure 5.1e ).

These divergent patterns of gendered job change raise the question as to what has happened to the distribution of women’s and men’s jobs (i.e. the stock of jobs) over the five quintiles, measured in 2007. In just over half the countries (13 of 23), two facts are notable: the share of women’s jobs in the bottom quintile is greater than the share of men’s jobs in the bottom quintile; and the share of women’s jobs in the top quintile is lower than the share of men’s jobs in the top quin-tile ( Figure 5.2 – the group from Denmark to Germany). This pattern of gender inequity is extended in four of these countries (Cyprus, the UK, the Czech Republic and Greece) where women’s jobs are less-favourably distributed than men’s in the first and second quintiles and the fourth and fifth quintiles.

In a next group of six countries, women still experience a higher share than men in bottom-quintile jobs, but in fact enjoy a higher share of jobs in the top quintile than do men. For example, in Hungary, 20 per cent of women’s jobs and 14 per cent of men’s jobs are in the bottom quin-tile, yet 27 per cent of women’s jobs are in the top quintile compared to


a. Gender divide – female polarization, male upgrading

020406080

100120

(000

s)

(000

s)(0

00s)

(000

s)(0

00s)

140

1 2 3 4 5 1 2 3 4 5Women Men Women Men

Women MenWomen Men

Women Men

FT PT

Cyprus

–5

0

5

10

15

20

25

1 2 3 4 5 1 2 3 4 5

Greece

–200–150–100–50

050

100150200250

1 2 3 4 5 1 2 3 4 5

Spain

0200400600800

1,0001,2001,400

1 2 3 4 5 1 2 3 4 5

Finland

–60–40–20

020406080

100

1 2 3 4 5 1 2 3 4 5

Ireland

b. Gender divide – female upgrading, male mixed patterns

Hungary

–50

0

50

100

150

200

250

1 2 3 4 5 1 2 3 4

Latvia

–40–20

020406080

(000

s)(0

00s)

(000

s)(0

00s)

100

1 2 3 4 5 1 2 3 4 5

Lithuania

–80–60–40–20

0204060

1 2 3 4 5 1 2 3 4 5

Slovenia

–30–20–10

01020304050

1 2 3 4 5 1 2 3 4

Women Men Women Men

Women Men Women Men


c. Gender divide – female part-time dominates

Germany

–1000

–500

0

500

1000

1500

1 2 3 4 5 1 2 3 4 5

Women Men Women Men

Women Men Women Men

Women Men

Austria

–60

–40

–20

0

20

40

60

1 2 3 4 5 1 2 3 4 5

Belgium

–100

–50

0

50

100

150

1 2 3 4 5 1 2 3 4 5

Luxembourg

–10

–5

0

5

10

15

1 2 3 4 5 1 2 3 4 5

Netherlands

–100–50

050

100150200250300350

1 2 3 4 5 1 2 3 4 5

(000

s)(0

00s)

(000

s)(0

00s)

(000

s)

d. Gender parity – upgrading

Portugal

–200

–150

–100

–50

0

50

100

(000

s)

150

1 2 3 4 5 1 2 3 4 5

FT PT

Denmark

–60–40–20

020406080

100

1 2 3 4 5 1 2 3 4 5

Italy

–300–200–100

0100200300400500600

1 2 3 4 5 1 2 3 4 5

Women

Sweden

–150–100–50

050

100150200

1 2 3 4 5 1 2 3 4 5

Men Women Men

Women MenWomen Men

(000

s)

(000

s)(0

00s)


just 20 per cent of men’s. Sweden is the only country where women in fact are less likely than men to be employed in bottom quintile jobs but also less likely than men to be employed in top quintile jobs. Indeed, Sweden’s job distribution is especially notable for the very high share of women working in middle quintile jobs – 34 per cent; no other coun-try comes close to such a high share – Hungary has 23 per cent and Lithuania 22 per cent. Finally, in three countries women’s distribution is more favourable than men’s at both ends of the quintile range of jobs. In Slovenia, Belgium and Lithuania a smaller share of women than men work in bottom-quintile jobs and a higher share of women than men work in top quintile jobs.

Overall, therefore, is it possible to state whether the gendered distri-bution of jobs has moved towards greater or lesser equity? One method of addressing this question is to estimate the change in the average wage quintile measure – comparing the weights of male and female workers, respectively, across the five quintiles at the start and end of the period

e. Gender parity – polarization

Czech Republic

–200

–150

–100

–50

0

(000

s)

50

100

1 2 3 4 5 1 2 3 4 5

Estonia

–15–10–505

101520253035

1 2 3 4 5 1 2 3 4 5

France

–400–200

0200400600800

1,0001,200

1 2 3 4 5 1 2 3 4 5

Slovakia

–20

0

20

40

60

80

1 2 3 4 5 1 2 3 4 5

UK

–400

–200

0

200

400

600

800

1 2 3 4 5 1 2 3 4 5

Women Men Women Men

Women Men Women Men

Women Men

(000

s)

(000

s)(0

00s)

(000

s)

Figure 5.1 Job change by gender, full-time and part-time, 1995–2007

Ger

man

y

0%10%

20%

30%

40%

Au

stri

a

0%10%

20%

30%

40%

Cyp

rus

0%10%

20%

30%

40%

Cze

ch R

epu

blic

0%10%

20%

30%

40%

Den

mar

k

0%10%

20%

30%

40%

12

34

5

12

34

5

12

34

5

12

34

5

12

34

5

12

34

5

Wom

enM

en

Est

on

ia

0%10%

20%

30%

40%

Fra

nce

0%10%

20%

30%

40%

Gre

ece

0%10%

20%

30%

40%

Lat

via

0%10%

20%

30%

40%

Net

her

lan

ds

0%10%

20%

30%

40%

Fin

lan

d

0%10%

20%

30%

40%

UK

0%10%

20%

30%

40%

12

34

5

12

34

5

12

34

5

12

34

5

12

34

5

12

34

5

Lu

xem

bo

urg

0%10%

20%

30%

40%

12

34

5

Po

rtu

gal

Irel

and

0%10%

20%

30%

40%

Bel

giu

m

0%10%

20%

30%

40%

Hu

ng

ary

0%10%

20%

30%

40%

Ital

y

0%10%

20%

30%

40%

Lit

hu

ania

0%10%

20%

30%

40%

Slo

vaki

a

0%10%

20%

30%

40%

Slo

ven

ia

0%10%

20%

30%

40%

Sp

ain

0%10%

20%

30%

40%

Sw

eden

0%10%

20%

30%

40%

12

34

51

23

45

12

34

5

12

34

50%10%

20%

30%

40%

12

34

51

23

45

12

34

5

12

34

51

23

45

12

34

5

Figu

re 5

.2

Qu

inti

le d

istr

ibu

tion

of

job

s fo

r w

omen

an

d m

en, 2

007


investigated. The results shown in Table 5.1 reveal that in all countries except Spain women’s average wage quintile measure improved rela-tive to men’s, as much as five percentage points and more in Slovenia, Denmark, Luxembourg and the Czech Republic. Did countries with a relatively wide gender gap at the start of the period experience an above-average improvement in women’s distribution across job quin-tiles – in other words, is there evidence of convergence? The results suggest not. Finland, Cyprus, Estonia, Portugal and Ireland made only marginal improvements in women’s relative job distribution compared to men’s despite a large gap between men and women at the start of the period. Moreover, countries where women’s job distribution was actually more favourable than men’s, such as Slovenia and Italy, experi-enced above-average improvement.

Table 5.1 Change in average wage quintile measure for women and men, 1995–2007

Women Men % gender gap Percentage point change in gender gap 1995 2007 1995 2007 1995 2007

Country average 2.80 2.98 3.06 3.15 91.8 94.5 2.7 Slovenia 2.98 3.25 2.90 2.95 102.7 110.1 +7.4 Denmark 2.76 3.02 3.05 3.14 90.5 96.3 +5.8 Luxembourg 2.72 3.17 2.88 3.17 94.5 100.0 +5.6 Czech Rep. 2.72 2.95 3.15 3.20 86.5 92.0 +5.5 Netherlands 2.81 2.89 3.18 3.12 88.2 92.7 +4.5 Sweden 2.78 3.06 2.97 3.12 93.4 97.8 +4.4 UK 2.64 2.78 3.13 3.13 84.3 88.7 +4.4 Slovakia 2.67 2.73 3.13 3.06 85.2 89.1 +3.9 Greece 2.58 2.82 3.14 3.30 82.3 85.5 +3.2 France 2.74 2.90 3.12 3.20 87.7 90.5 +2.8 Italy 2.91 3.07 2.85 2.94 101.8 104.5 +2.7 Portugal 2.76 3.06 3.05 3.30 90.4 92.7 +2.3 Finland 2.67 2.79 3.10 3.18 86.3 87.7 +1.5 Lithuania 3.05 3.28 3.04 3.22 100.2 101.6 +1.4 Estonia 2.81 2.84 3.09 3.07 91.0 92.3 +1.4 Germany 2.88 2.98 3.08 3.14 93.7 95.0 +1.3 Austria 2.85 2.93 3.04 3.10 93.6 94.7 +1.1 Ireland 2.79 2.95 3.07 3.21 90.9 92.0 +1.1 Hungary 3.07 3.18 2.94 3.02 104.2 105.1 +0.9 Belgium 2.99 3.11 2.98 3.08 100.4 100.9 +0.5 Latvia 2.91 3.11 3.12 3.33 93.1 93.5 +0.3 Cyprus 2.59 2.70 3.29 3.42 78.9 79.0 +0.1 Spain 2.76 2.87 2.99 3.14 92.1 91.4 -0.6

Note : start and end dates vary for several countries.


Nevertheless, the almost universal improvement in women’s job dis-tribution compared to men’s is notable, and complements the more general finding that both men and women experienced a rise in their respective average wage quintile measure (with the exceptions of men in the Netherlands, Slovakia, Estonia and the UK). The issue of how such trends contrast with gendered changes in job structure ranked by skill quintiles is explored in the next section.

5.3 Gender differences in wage and skill quintile measures

In light of the observation that women are more likely than men to experience a mismatch between relative pay and relative skill level in their job (Grant et al. 2005; Grimshaw and Rubery 2007), in this sec-tion we compare the distribution of women’s and men’s jobs by wage quintile and by skill quintile (see also, Stehrer et al. 2009). In line with expectations, we find a strong inverse relationship of wage and skill indicators between men and women; while for women the skill quin-tile measure generally exceeds the wage quintile measure, for men the reverse is true. For all 23 countries combined, the country average wage and skill measures are 2.98 per cent and 3.29 per cent, respectively, for women and 3.15 per cent and 2.94 per cent, respectively, for men. This is a striking finding.

Figure 5.3 plots the individual country results for men and women at the start and end of the period (1995 and 2007 for most countries). For each graph, the results for women tend to lie to the upper left of the diagonal line and for men to the lower right. In 2007, all countries except Slovenia show an average wage quintile measure for women that is consistently less than the corresponding average skill quintile meas-ure. The gap is especially large in Cyprus, Greece, Ireland, Spain and Estonia where the average skill measure for women exceeds the wage measure by more than 0.5 points. Yet the opposite is true for men in all 23 countries, for both years shown in Figure 5.3 .

It is notable that certain countries display a relative congruity between the average wage and skill quintile measures. This includes a selection of CEE countries – Hungary, Slovenia and the Czech Republic, as well as Germany, Austria and Luxembourg. In the case of Germany and Austria, this finding may reflect the particularity of their wage-setting and skill formation systems which tend to follow what Marsden (1999) calls a ‘qualification-based rule’, such that wages align with qualifications far more so than in other countries largely as a result of the widespread


dual system of vocational training and strong reputational standard of qualifications as a signal of worker skill (Culpepper and Finegold 1999). Our data suggest that the qualification-based rule favours gender parity to the extent that it encourages a better fit between women’s relative skill position and their wage. 1

Figure 5.4 presents the same country patterns of gender differenti-ation across countries in a different manner by plotting the female average wage–skill quintile gap against the male wage–skill quintile gap using 2007 data. Again, the gap for men is positive in all countries (i.e. men’s average wage quintile measure exceeds their average skill quintile measure) and the opposite holds for women except in Slovenia. There is a general tendency for the positive wage–skill gap among men to be offset by the negative wage–skill gap among women. Examples of female and male wage–skill gaps include −0.24 and +0.23 in the UK,

Women, 1995

2.5

2.6

2.7

2.8

2.9

3.0

3.1

3.2

3.3

3.4

3.5

3.6

2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6

Wage quintile measure

Ski

ll q

uin

tile

mea

sure

Women, 2007

2.5

2.6

2.7

2.8

2.9

3.0

3.1

3.2

3.3

3.4

3.5

3.6

2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6


Ski

ll q

uin

tile

mea

sureSE

UK

SISK

FINL

LU

LTLVBE

ITES

HU

GR FR

EE

DKCZDE

CY

AT

IE

PT

3.39

UK

FI

SE

SISK

LTLV

HULU

GR

FR

EE

NL DK

CZ

ESCY BE

AT

DE

PTIE

Men, 1995

2.5

2.6

2.7

2.8

2.9

3.0

3.1

3.2

3.3

3.4

3.5

3.6

2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6


Ski

ll q

uin

tile

mea

sure

Men, 2007

UKFI

SE

ESSI

SK

NL

LULT

LVIT

HU GR

FR

EE

DKCZ

CY

BE

AT

DE

IE

PT UK

FISE

ESSI

SK

NL

LU

LT

LVIT

HU

GRFR

EE

DK

CZ

CY

BE

AT

DE

IE

PT

2.5

2.6

2.7

2.8

2.9

3.0

3.1

3.2

3.3

3.4

3.5

3.6

2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 3.5 3.6


Ski

ll q

uin

tile

mea

sure

Figure 5.3 Average wage and skill quintile measures for women and men, 1995–2007


−0.37 and +0.37 in Finland, −0.25 and +0.26 in Lithuania and −0.53 and +0.48 in Estonia. The unweighted country average measures are −0.32 for women and +0.21 for men. Overall, therefore, the system of gender bias operates in both directions with a tendency for women to be employed in jobs where the quintile measure of average skill exceeds the quintile measure of average wage, combined with a tendency for men to work in jobs where the average quintile wage exceeds the aver-age quintile skill.

Not all countries fit this general pattern, however. Several countries display a different kind of gender bias where men experience a close match between their average wage and skill quintile measures, but women a wide gap. These countries are located towards the bottom left hand corner of Figure 5.4 – Spain, Greece, Portugal and Slovakia. In Spain, for example, the wage–skill gaps for men is less than 0.2 points (3.14 and 2.96, respectively) but for women is more than 0.5 (2.87 and 3.41, respectively). Also, re-stating the above observation regarding

AVG

UK

FI

SI

SK

FR

NL

LU

LT

LVIT

HU

GR

SE

ES EE

DKCZ

CY

BE

IE

PT

Men’s wage–skill gap

Wo

men

’s w

age–

skill

gap

–0.8

–0.7

–0.6

–0.5

–0.4

–0.3

–0.2

–0.1

0.1

0.2

0.3

0.0–0.3 –0.2 –0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8AT

DE

Figure 5.4 Country differences in the average wage–skill quintile gap for men and women, 2007


Germany and Austria, these countries (along with Luxembourg and Slovenia) can be found close to the point of origin reflecting the close match of wage and skill quintile measures for both men and women.

In common with the trend towards a closing of the gender gap in average wage quintile measures ( Table 5.1 ), the data also point to an improvement in women’s average skill quintile measure compared to men’s. However, there is no straightforward association between the degree of change in gender–wage gap and gender–skill gap across countries – once again suggestive of an absence of convergence trends. Figure 5.5 collects country data in four groups, distinguishing between relative size of gender–wage gap (using the wage quintile measure) and relative size of improvement in women’s average wage quintile com-pared to men’s. The figure also displays the change in the gender–skill gap for each year. There are two significant findings concerning the relationship between skill and wage quintile gender gap measures. First, there is no simple association between the relative size of gender gap in wage quintile measure and the size of gender gap of skill quintile meas-ure. The difference between women’s and men’s average skill quintile measures is in fact exactly the same, some 0.12 points on average, both for countries with a relatively wide gender–wage gap and for countries with a relatively narrow gender–wage gap. Second, countries that expe-rienced a relatively small improvement in women’s relative wage quin-tile measure over the period are more likely to exhibit a higher positive difference between women’s and men’s average skill quintile measure.

5.4 Low-wage jobs

As outlined in Chapter 2 the data set compiled for this project does not incorporate country differences in the scale of differentiation of job quality. A quintile ranking of jobs is generated for all countries but this says nothing about the fact that most bottom quintile jobs in Sweden, for example, are paid higher relative to the median wage in Sweden than bottom quintile jobs in the UK relative to its median wage. This is an important limitation if we are interested in exploring the causes of change in women’s representation among low-wage jobs since the liter-ature typically defines low-wage employment as those jobs paid below some fixed fraction of the median wage, rather than simply the bottom 20 per cent (or some other fixed proportion) of all jobs (e.g. Grimshaw 2011; Gautié and Schmitt 2010; Lucifora et al. 2005). Therefore, in this section, we utilize the more conventional method for ranking jobs according to the level of the wage relative to the median. For each


country, we categorize jobs as ‘low wage’ where pay is 75 per cent or less of the median wage, ‘middle wage’ where pay is between 75 per cent and 125 per cent of the median and ‘high wage’ if pay is 125 per cent or more of the median.

a. Wide gender–wage gap with strong improvement in women’s wage

–0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15 0.2 0.25 0.3

–0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15 0.2 0.25 0.3

–0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15 0.2 0.25 0.3

–0.25 –0.2 –0.15 –0.1 –0.05 0 0.05 0.1 0.15 0.2 0.25 0.3

GRUKSKCZFRNLDK

b. Wide gender–wage gap with weak improvement in women’s wage

CY

FI

PT

IE

EE

c. Narrow or positive gender–wage gap with strong improvement in women’s wage

SE

LU

IT

SI

d. Narrow or positive wage gap with weak or negative improvement in women’s wage

ES

LV

AT

DE

LT

BE

HU

2007 skill gap1995 skill gap2007 wage gap1995 wage gap

Figure 5.5 Change in gender gaps in the average wage and skill quintile measures, 1995–2007


A serious note of caution is required before we present the results since unlike other earnings data sources the jobs data are highly aggre-gated (the average wage per job, defined by the matrix of industries and occupations, disaggregated by sex) and therefore substantially underes-timate wage variation compared to individual earnings data. Moreover, there is reason to suspect the validity of results for three countries (Greece, Spain and Hungary) because the incidence of low-wage jobs is inexplicably low compared to evidence from other data sources; 2 these countries have therefore been excluded from the comparison of coun-try patterns below.

In terms of the stock of women’s employment in low-wage jobs in 2007, we find a considerably wide range across countries: at one end, Sweden and Finland register 9 per cent and 10 per cent, respectively, whereas Austria and Cyprus register more than 30 per cent. Other stud-ies suggest these differences in distributional outcomes are likely to be shaped by country wage-setting institutions, especially the inclusive-ness of collective bargaining coverage and the presence and relative level of a statutory minimum wage (Bosch et al. 2009; Rubery et al. 2005). Table 5.2 groups countries according to the character of their wage-setting institutions. The unweighted country averages reveal that the proportion of all women who are employed in low-wage jobs is

Table 5.2 Percentage share of low-wage jobs among women by type of wage-setting system

1. High relative minimum wage

(40%+ of average)

2. Strong collective bargaining coverage

(80%+)

3. Weak CB coverage and no or low minimum wage

*BE 14 AT 34 CZ 22

*FR 13 BE 14 CY 43 IE 26 DK 14 DE 18 LU 29 FI 10 EE 33 *SI 25 FR 13 LT 20 IT 15 LV 30 NL 23 PT 13 SE 9 SK 37 SI 25 UK 21 Average (%) 21 17 26

Source: 2007 OECD minimum wage database; 2000 ICTWSS data selected for collective bargaining coverage.

Note : *Belgium, France and Slovenia appear twice since they have both strong collective bargaining coverage and a high relative minimum wage.


lowest in countries with strong collective bargaining coverage (defined as at least 80% coverage), followed by countries with a high minimum wage (defined as 40% or more of the country’s average wage). Women’s risk of low pay is in fact highest – at 26 per cent – among countries with neither strong collective bargaining coverage nor a high-value statutory minimum wage. These wage system characteristics also add explana-tory power to comparisons of the gender inequity of low-wage employ-ment – that is, women’s risk of low pay compared to men within each country. Among group 1 countries, women’s relative risk (defined as the share of all women employed in a low-wage job divided by the share of all men employed in a low-wage job) averages 1.61, among group 2 countries it is 1.49, and among group 3 countries it is 1.72. While there are important country differences within each group identified in Table 5.2 , the general stylized pattern nevertheless lends support to the argument that institutional arrangements for regulating low-wage work can make a difference in reducing women’s vulnerability to low pay (see also EC 2008a: 93–94).

Given the relatively large stock of low-wage jobs among women and the widely held view that job change has generally been polarizing, with growth at the bottom and top of the job structure, it is surprising that low-wage jobs in Europe in fact make up only a small fraction of total net job growth. Among all employees, low-wage jobs accounted for just 3.6 per cent of net job growth across the 23 countries combined (approximately 890,000 jobs out of 24.5 million) over the 1995–2007 period. The vast bulk of job growth occurred among middle-wage jobs (40% of total net job growth) and high-wage jobs (56%) ( Table 5.3 ). We believe part of the explanation for this low share reflects problems with the estimation of low-wage jobs using the jobs data set, especially since Spain, the driver of job growth in Europe during the period, registers a level far below that from other earnings data sets.

Analysing the contribution of male and female employees to job growth by wage category we find that while women contributed a greater volume of low-wage jobs, this was approximately proportional to the overall greater contribution of women’s employment to job growth in Europe. In other words, women were not over-represented among low-wage job growth. We find that women’s presence in low-wage jobs increased by 540,000 over the period compared to a rise of 350,000 for men, but this accounted for a similar share of overall job growth for women and men – 3.6 per cent and 3.7 per cent, respectively. Figures for part-time job change fit a similar pattern; women entered 1.28 million additional low-wage, part-time jobs – 17 per cent of the total net growth


of women’s part-time jobs – and men entered 660,000 additional part-time, low-wage jobs – 24 per cent of all male part-time jobs.

As might be expected, the 20 countries (excluding Spain, Greece and Hungary for reasons described above) do not share a common pattern in trends in women’s low-wage employment. A key result is that the total number of low-wage jobs increased in nine countries, registered very little change (less than 2% of total net change in employment) in five countries and decreased in six countries. The rising number of low-wage jobs was no more likely to be experienced by women than by men: nine countries record a rise in women’s low-wage employment compared to ten countries with a rise in men’s low-wage employment. Moreover, in almost all countries (the three exceptions are Austria, Cyprus and Estonia), women’s increased employment in high-wage jobs (true in all 20 countries) significantly exceeds the change in low-wage jobs.

Country patterns of change do not cluster around varieties of wage-setting institutions in the same way as our analysis of the stock of low-wage jobs at a given point in time. This is because much of the change in low-wage employment is caused by large sectoral shifts, such as the decline in agriculture, textiles and wearing apparel industries – especially notable in Portugal, Lithuania and the Czech Republic. Thus we find the two countries that largely contribute to women’s low-wage job growth (in terms of numbers of jobs) are Germany with 500,000 additional jobs, where there is no statutory minimum wage and collective bargaining coverage is patchy, as well as France with 440,000 jobs where the statutory minimum wage is relatively high and collective bargaining coverage the highest in Europe. Moreover, the

Table 5.3 Changes in low-wage, middle-wage and high-wage jobs in the EU23

Total job growth Part-time job growth

Total Male Female Total Male Female

Low-wage jobs 890,000 350,000 540,000 1,940,00 660,000 1,280,000 3.6% 3.7% 3.6% 18.8% 23.9% 16.9% Middle-wage jobs

9,840,000 3,260,000 6,580,000 5,390,000 1,290,000 4,100,000

40.2% 34.0% 44.3% 52.2% 46.7% 54.2% High-wage jobs 13,730,000 5,980,000 7,750,000 3,000,000 820,000 2,190,000 56.1% 62.4% 52.1% 29.0% 29.7% 28.9% Total jobs 24,450,000 9,580,000 14,870,000 10,330,000 2,760,000 7,570,000 100% 100% 100% 100% 100% 100%

Note : Low-wage jobs include those paid at 75 per cent or less of the median wage for each country, middle-wage jobs include those paid between 75 per cent and 125 per cent of the median and high-wage jobs include those paid at 125 per cent or more of the median.


countries accounting for the largest declines in low-wage jobs among women include the UK (210,000 net job loss) and the Czech Republic and Italy (150,000 each). The lesson for comparative institutional analysis is that it must incorporate a twin focus on industry structure (and changing developmental trends) and labour market institutions in order to understand changing patterns of job quality (e.g. Galbraith and Berner 2001).

A further key finding is that most countries witnessed a divided pat-tern of low-wage job change by sex. While in six countries the country pattern is shared among both men and women, in fourteen countries men and women experienced divided patterns of low-wage job change ( Figure 5.6 ). Among those countries where the number of women’s

a. Shared gender pattern

–1.4–1.2

–1–0.8–0.6–0.4–0.2

00.20.40.60.8

PT LT IT BE FI NL

Inde

xed

to to

tal n

et jo

b gr

owth

Male Female

b. Divided gender patterns

–1.4–1.2–1

–0.8–0.6–0.4–0.2

00.20.40.60.8

CZ SE LV UK SI DK SK LU IE FR DE CY AT EE

Inde

xed

to to

tal n

et jo

b gr

owth

Figure 5.6 Country patterns of low-wage job growth, by sex

Note : For each country, total net job growth (decline) is assigned an index of +1 (−1). A ‘shared gender pattern’ is one where male job change is not less than half female job change. Spain, Greece and Hungary are excluded due to unreliability of low-wage job data.


low-wage jobs has risen significantly, only Slovakia records a higher rise for men. In the rest, women’s low-wage job growth outnumbers men’s. This is especially striking in Estonia and Luxembourg (where the number of low-wage jobs for men in fact fell), as well as in Austria and Cyprus. For example, in Austria, the net changes for women over the period were 94,000 low-wage jobs compared to 11,000 middle-wage jobs and 68,000 high-wage jobs; also, the vast bulk of women’s low-wage job growth, some 90,000 jobs, were part time. In countries where the num-ber of women’s low-wage jobs fell during the period, this was shared among men in five of the eight countries; while in Slovenia and the UK men experienced a significant rise in low-wage jobs and in Sweden there was no significant change in the number of men’s low-wage jobs.

The question remains as to what extent trends in the number of low-wage jobs follow a similar country pattern as that described above where job change is decomposed into wage quintiles. Here we wish to understand the significance of changes in the number of low-wage jobs among women compared to the direction and magnitude of change of the 20 per cent of jobs located in the bottom wage quintile. We do not anticipate a one-to-one matching given the large inter-country differ-ences in wage structure and resulting variation in positioning of the bottom wage quintile relative to the median. In this case we include Spain, Greece and Hungary for the purposes of comparison. Figure 5.7a ranks countries by the size of change in low-wage jobs for women. The 23 countries divide into three groups: nine with a ‘shrinking bottom’ (decline in number of low-wage jobs); five registering no significant change; and nine with a ‘growing bottom’. Figure 5.7b plots the coun-tries in the same order so as to compare with data on change in the bottom quintile jobs for women. Only four countries appear to be sig-nificantly out of line. Two of these countries fit into our suspect group – Spain and Greece – for which we have doubts about low-wage data validity: Greece shows no change in women’s low-wage jobs but regis-ters a significant overall decline for bottom quintile jobs; and for Spain we see a small rise of low-wage jobs for women but a massive increase in bottom quintile jobs for women of some 900,000. In both cases, the issue is that only a small proportion of women are picked up in the low-wage data compared to their much larger representation among bottom quintile jobs. The other two misfits are Italy, which records a signifi-cant fall in the number of low-wage jobs (some 150,000 jobs) but only a minor fall in bottom quintile jobs for women, and Austria, where the data on women’s low-wage work picks up a far greater increase than that registered by data for the bottom quintile (approximately 90,000


compared to close to 20,000, respectively). In this case, it is likely that the rise in low-wage work overshoots change in bottom quintile jobs since more than one in three women are employed in low-wage jobs compared to 22 per cent in the bottom quintile in Austria.

5.5 Job concentration effects

Unlike men’s employment, women’s employment is strongly concen-trated into a relatively narrow range of jobs – as defined in this project by the matrix of ISCO two-digit and NACE two-digit categories. This means that the changing employment trends in a handful of jobs can make a substantial difference to the overall shape of changing job struc-ture for women (see also, Grimshaw and Rubery 1997). Our focus on the category of job in this project is especially useful here since unlike the category of occupation it illuminates the dual tendency for employ-ment to be delineated by horizontal sex segregation (men and women

a. Low wage job change

-300

-200

-100

0

100

200

300

400

500

600

(000

s)(0

00s)

700

UK CZ IT PT SE LT HU LV SI GR DK BE LU FI EE SK CY ES IE AT NL FR DE

b. Q1 job change

-300

-200

-100

0

100

200

300

400

500

600

700

UK CZ IT PT SE LT HU LV SI GR DK BE LU FI EE SK CY ES IE AT NL FR DE

Shrinking bottom No change

Growing bottom

Figure 5.7 Comparing women’s low-wage job change with bottom quintile job change


concentrated in different industries and different types of occupations) and vertical segregation (men more likely to work in higher-grade occupations).

In all 23 countries, alternative measures of job concentration confirm women’s concentration into a small number of jobs. For example, the data for 2007 show that half of all female employment in each country is accounted for by between just 10 and 28 jobs across the various coun-tries. In each country men’s job concentration is considerably less than women’s (34 jobs, on average, account for half of male employment). Using an alternative measure of job concentration, we also find that the share of all employment concentrated in the five largest jobs is consis-tently greater among women than among men. The highest concentra-tion is found in Greece (38% of all female employment), followed by Sweden (35%) and Denmark (35%). Figure 5.8 presents this data for men and women for the top-five and top-ten largest jobs. Across all coun-tries, between 34 per cent and 52 per cent of women are concentrated in the top ten jobs. Again, women’s high job concentration is largely a result of sex segregation and not simply a reflection of the employment

a. Female job concentration in top-five and top-ten jobs

0

10

20

30

40

50

60

AT BE CY CZ DE DK EE ES FI FR GRHU IE IT LT LU LV NL PT SE SI SK UK

Per

cent

age

shar

e of

all

fem

ale

empl

oym

ent

b. Male job concentration in top-five and top-ten jobs

0

10

20

30

40

50

60

AT BE CY CZ DE DK EE ES FI FR GR HU IE IT LT LU LV NL PT SE SI SK UK

Per

cent

age

shar

e of

all

mal

e em

ploy

men

t

Top-five jobsTop-ten jobs

Top-five jobsTop-ten jobs

Figure 5.8 Job concentration in the top-five and top-ten jobs, 2007


size of jobs; men’s job concentration in the top-five and top-ten jobs is significantly lower than women’s in all countries.

In most countries, many of the largest jobs employ women in part-time work. Limiting the sample to the 14 countries where part- timers account for more than one in ten employees 3 (all except Greece, Cyprus and the seven CEE countries), we find that the top-ten jobs for women part-time workers also follow a pattern of strong job concentra-tion. Among these 14 countries the top-ten part-time jobs for women account for between 7 per cent and 36 per cent of all women’s employ-ment, with the Netherlands at an extreme end of this range ( Figure 5.9 ); indeed, in the Netherlands women’s part-time work in the largest five jobs accounts for 27 per cent of all women’s employment – a mix of health and education professionals and sales and personal services workers. Germany is also notable since it has shifted considerably in the last decade to be a major employer of part-time workers, more so than the UK by 2007. 4 Moreover, in all 14 countries part-time jobs are more concentrated than women’s full-time jobs; between 41 per cent and 67 per cent of women’s part-time work is concentrated in just ten jobs in these countries (compared to 37–47% considering all women’s job concentration, Figure 5.8 ).

Women’s high job concentration means that a very small number of jobs make a significant contribution to patterns of overall job growth and job decline. Across all 23 countries, an average of just four jobs

0

10

20

30

40

50

60

70

80

IT PT FI ES FR AT IE LU BE DE DK UK SE NL

Per

cent

age

of e

mpl

oym

ent

Share of FPT employmentShare of all F employment

Figure 5.9 Women part-timers’ job concentration, top-ten jobs, 2007


contributed more than 25 per cent of women’s gross job growth and a little over 13 jobs accounted for at least 50 per cent of women’s job growth. The same holds true for the pattern of job decline; on average, in each country close to four jobs accounted for 25 per cent of the gross decline in female employment and approximately 13 jobs for half-gross job decline. Table 5.4 shows that all countries share this characteristic. The pattern is unsurprisingly stronger in some of the smaller countries (Cyprus, Estonia and Luxembourg) where the mix of jobs is small, as well as countries where the shift out of agriculture has been strong during the period with the loss of skilled jobs for women in agriculture accounting

Table 5.4 The number of jobs that contributed to women’s job growth and job destruction, 1997–2007

Share of gross job growth CY EE LU AT GR LT SE SK

25 (%) 2 4 4 3 4 5 2 4 50 (%) 7 11 9 10 12 15 8 12 75 (%) 20 22 23 29 31 33 33 33

SI IE NL DK LV PT FR HU

25 (%) 3 3 5 6 4 3 3 5 50 (%) 11 11 12 13 15 12 13 15 75 (%) 34 36 38 39 40 40 41 44

DE BE ES UK FI CZ IT

25 (%) 5 5 4 4 5 5 6 50 (%) 14 15 13 17 18 20 21 75 (%) 47 50 50 50 51 55 66

Share of gross job growth GR ES PT LU CY IT LV IE

25 (%) 1 1 1 3 2 2 3 4 50 (%) 1 2 2 7 5 5 9 10 75 (%) 2 7 9 15 16 19 26 27

BE LT SE EE FR NL DK FI

25 (%) 4 1 2 2 2 4 5 5 50 (%) 12 8 9 13 9 15 18 18 75 (%) 38 38 41 43 43 49 51 51

HU DE SI AT SK UK CZ

25 (%) 6 5 4 8 6 8 6 50 (%) 18 18 18 25 24 24 25 75 (%) 51 53 55 63 67 67 73


for more than 25 per cent of women’s decline in employment during the period in Greece, Spain and Portugal. The high concentration of women’s part-time work in a handful of jobs sheds further light on pat-terns of job growth. In four countries – Belgium, Germany, Luxembourg and the Netherlands – the top-ten jobs for women’s part-time work con-tributed more than 15 per cent of total employment growth (men and women) – as much as 25 per cent in the Netherlands.

Sex segregation by job concentration raises questions regarding not only equity of employment opportunity for men and women, but also equity of reward. On the one hand, women’s concentration into few sectors or types of jobs may in certain circumstances provide shelter from competition against male unemployed labour, or access to jobs such as in the public services that offer relatively secure employment and attractive opportunities for skill development or access to family friendly HRM practices. On the other hand, women’s job concentration may reflect a restricted pool of job opportunities or be associated with a pay penalty relative to jobs where men tend to work.

Other research points to the wage penalty associated with wom-en’s employment concentration (OECD 1998). Using the wage quintile measures, and distinguishing between the average wage quintile for women’s top-ten jobs and the average wage quintile measure for all remaining jobs (in each case weighted by women’s job concentration), we find that women’s job concentration is associated with a wage pen-alty in 16 of the 23 countries. In other words, the average wage quintile for women’s top-ten jobs is lower than the average wage quintile for all remaining jobs ( Table 5.5 ). The wage penalty is highest in Greece and Finland where the average wage quintile measure for women’s top-ten jobs is 2.26 and 2.23, respectively, compared to wage quintile measures for women’s remaining jobs of 3.42 and 3.20.

In most countries where women’s job concentration carries a wage penalty, the relative skill of the top-ten jobs is also less than the aver-age for remaining jobs. More surprising is the finding that in seven countries we find evidence of a wage penalty accompanied by a skill premium among women’s top-ten jobs.

Only seven countries display a wage premium in women’s top-ten jobs. Here, job growth in the top-ten jobs can make a very positive contribution to the overall structure of women’s employment. Sweden stands out with an average wage quintile measure some 0.74 points higher than that estimated for all remaining jobs. Other countries with a relatively large premium include Slovenia, Denmark and Germany. In five of these seven countries, the wage premium is associated with


a high average skill quintile measure – again, this is highest in Sweden where the skill quintile for the top-ten jobs is 3.87 compared to a meas-ure of 3.02 for remaining jobs.

How do these findings relate to our preliminary evidence on the nature of changing job structure? Those countries with a wage penalty for women’s job concentration are far more likely to have experienced job polarization characterized by a significant rise in women’s jobs in

Table 5.5 Wage and skill quintile measures for women’s top-ten jobs by employ-ment concentration, 2007

Female employment concentration in top

10 jobs (%)

Women’s top 10 jobs

Women’s remaining jobs

Wage quintile

Skill quintile

Wage quintile

Skill quintile

a) Wage penalty and negative skill gap

GR 52 2.26 3.34 3.42 3.71 FI 42 2.23 2.97 3.20 3.30 PT 42 2.56 3.17 3.42 3.77 CY 45 2.33 3.39 3.00 3.43 UK 41 2.47 2.92 2.99 3.09 LT 39 3.03 3.27 3.44 3.69 ES 45 2.71 3.24 3.00 3.55 FR 38 2.86 3.05 2.92 3.30 LU 39 3.15 3.24 3.18 3.25

b) Wage penalty and positive skill gap

LV 35 2.78 3.51 3.29 3.50 IE 45 2.69 3.67 3.17 3.53 SK 40 2.47 3.27 2.90 3.12 EE 35 2.57 3.71 2.98 3.19 HU 37 3.05 3.47 3.25 3.13 IT 40 2.99 3.52 3.12 3.31 NL 43 2.86 3.26 2.91 3.14

c) Wage premium and negative skill gap

SI 39 3.65 3.10 3.00 3.27 AT 45 2.99 2.89 2.89 3.04

d) Wage premium and positive skill gap

SE 47 3.45 3.87 2.71 3.02 DK 43 3.33 3.31 2.79 3.09 DE 37 3.21 3.20 2.85 2.89 CZ 34 3.14 3.35 2.85 2.97 BE 40 3.26 3.77 3.00 3.15


the first and/or second quintile (as in Figure 5.1 ). Nine of the sixteen countries with a wage penalty experienced polarization of women’s job structure compared to just one – the Czech Republic – out of seven countries where women’s employment concentration in the top-ten jobs earns a wage premium on average.

Three phenomena underlie these country differences: the variety of jobs among the top ten; the weighting of women workers among the top-ten jobs; and the wage quintile ranking associated with similar jobs. Here we address this third factor – country differences in wage quintile ranking for similar jobs. Table 5.6 ranks jobs according to how often they appear among the top-five and top-ten jobs for women in each country. Five jobs stand out as especially common, each appearing among the top-ten jobs in at least 19 countries. Two jobs – teachers in education and retail salespersons – are in fact among the top-five largest jobs for women in all 23 countries.

Table 5.6 also identifies the number of countries that define the job according to each wage quintile measure, 1 to 5. Despite similarity in patterns of sex segregation, there is strong evidence of inter-country variety in the relative position of jobs in the overall jobs wage ranking. For six jobs, differentiation spans neighbouring wage quintiles – the job of retail salespersons ranks among the bottom 20 per cent lowest paid in 14 countries but at the second wage quintile in 8 countries 5 (and, somewhat surprisingly, is ranked in the middle quintile in Lithuania), and personal services jobs in hotels and restaurants are also differenti-ated between the bottom wage quintile in 13 countries and the second quintile in 10 countries. Given the very high concentration of women in these jobs, such differences in relative pay exert considerable influ-ence on the overall pattern of women’s changing job distribution as presented in Figure 1a–d .

A further five jobs display the most varied pattern of wage ranking across countries. Health associate professionals are ranked across wage quintiles three to five, public administration office clerks from quin-tiles two to five, and skilled agricultural and fishery workers, personal services jobs in education and other business office clerks each include examples of country rankings along the quintile range of one to four. In part, the data are picking up problems of cross-country comparison of job categories. This is unavoidable given the country specific struc-turing of jobs; countries do define jobs differently – what appears as a similar job from one country to the next will often in practice comprise differences in required skills, social status, varied bundles of tasks and relative pay. In addition, the wide range of relative wages of these jobs


also reflects sector- and occupation-related wage contours, reflecting the unique character and dynamics of country systems of wage determina-tion. We know from many studies of inter-industry wage differentials, that industry wage contours persist after controlling for productivity differences (e.g. Gannon et al. 2007; Rycx 2003). One recent European study analysing matched employer–employee earnings data concludes that it is country differences in the wage returns to firm characteristics, not in individuals’ productivity-related characteristics, which are most influential in shaping wage structures (Simon 2008). Thus, similar jobs are likely to be paid at different points of the wage structure in different countries. Our results therefore emphasize the need for further research to explore how institutions shape the wage position of large job catego-ries in order to understand the range of factors that influence the shape of the changing job distribution (see also, Grimshaw et al. 2001).

Table 5.6 Common jobs with high female employment concentration, 2007

Job code and name

No. of countries with job in

No. of countries with a wage quintile for the

job measuring

Top 5 Top 10 1 2 3 4 5

8023 Teachers in education 23 23 – 1 – 3 19 5252 Retail salespersons 23 23 14 8 1 – – 5551 Hotels & restaurants –

personal services workers 15 20 13 10 – – –

8551 Health & social work – personal services workers

11 20 – 13 9 1 –

8532 Health associate professionals

17 19 – – 4 14 5

7541 Public administration office clerks

3 10 – 3 11 8 1

8522 Health & social work life science/health professionals

2 10 – – – 1 22

0161 Skilled agriculture and fishery workers

4 8 12 6 2 3 –

7534 Public administration other associate professionals

4 8 – – 2 12 9

9591 Household activities, elementary sales and services workers

4 6 12 11 – – –

8051 Education personal/protective services workers

2 6 4 13 5 1 –

7441 Other business office clerks 1 8 2 10 8 3 –

Note : Job code specified as NACE classification (two-digit) ISCO classification (two-digit).


Finally, for two jobs in Table 5.6 , there is relatively strong inter-country commonality in the wage ranking. Both are professional jobs predomi-nantly found in the public sector. Teachers in education tend to be paid at the top wage quintile, as do life-science and health-professional jobs in the health and social work sector. There are only a couple of country deviants from this pattern. Teaching professionals in Greece are ranked at the second wage quintile, which may in part reflect a problem of job classification, or a problem of stagnating salaries for teachers since the mid-1980s (EIRO 2006, cited in Fernández-Macias 2010: 174). The three countries which rank teachers at the fourth quintile are Estonia, Finland and Latvia. And only Belgium pays less than the top quintile wage to jobs for life-science and health professionals.

We turn now to the role of a particular subset of jobs, those associ-ated with the welfare state, in shaping patterns of women’s job change. Again, while a common driver of employment growth in all countries, we find that welfare state jobs can be a force for upgrading and for polarization, depending on the specific country system.

5.6 Welfare state effects

A final distinguishing feature of women’s employment patterns across different countries is associated with welfare state jobs, including those in education and health and social care. Because women are over- represented in welfare state jobs, the sector plays a key role in shaping patterns of women’s job change, as well as contributing to gender dif-ferences in job change given the distinctive economic pressures on the welfare state sector compared to other sectors of employment.

Across the 23 member states, women’s share of welfare state jobs – defined here as NACE 80 (education) and NACE 85 (health and social care) – ranges from 64 per cent to 84 per cent ( Table 5.7 ). Of course, the significance of women’s welfare state jobs as a contributor to job change also depends on the relative size of the education and health/social care sector in the particular country. On this measure, variation across countries is far greater and broadly clusters around the type of welfare regime (Esping-Anderson 1990; Trifiletti 1999). Countries with socio-democratic welfare regimes (Sweden, Finland and Denmark) have the highest share of women’s employment in welfare state jobs, 37–45 per cent. Countries with a conservative regime have a medium share, ranging from 24–39 per cent, as do countries with liberal regime types (34% in the UK and 31% in Ireland). The five countries categorized as Mediterranean welfare states have the lowest share of welfare state jobs,


ranging from 19 per cent to 24 per cent, along with the seven Central and Eastern European countries where the share ranges from 21 per cent to 28 per cent ( Table 5.7 ). In all five country clusters, the share of male employees in welfare state jobs is very low, with only Denmark, Sweden and the Netherlands registering a share higher than 10 per cent.

Analysis of change in welfare state employment for the 1995–2007 reveals three key findings. First, welfare state jobs are a major con-tributor to women’s overall job change, accounting for approximately

Table 5.7 Employment in welfare state jobs, by gender, 2007

Female share of welfare state

jobs (%)

Welfare state jobs as a share of all jobs

Women (%) Men (%)

Socio-democratic Denmark 75.3 41.5 12.1 Sweden 79.5 45.1 10.4 Finland 82.1 36.9 7.5

Conservative Germany 73.2 27.8 8.6 Austria 74.2 24.3 7.0 Belgium 73.6 34.5 9.9 France 75.7 31.3 9.2 Netherlands 75.6 39.4 10.8

Liberal UK 77.4 33.9 8.5 Ireland 78.7 31.1 6.4

Mediterranean Portugal 78.8 23.5 5.3 Cyprus 71.5 18.5 6.0 Greece 63.5 20.5 7.6 Italy 68.0 23.9 7.0 Spain 71.0 21.1 6.0

CEE countries Hungary 77.7 25.3 6.1 Czech Republic 77.8 23.2 5.0 Estonia 83.5 24.1 4.1 Latvia 83.8 20.9 3.9 Lithuania 83.4 27.7 5.4 Slovakia 80.5 24.8 4.7 Slovenia 78.9 23.6 5.3

Note : Welfare state jobs include education (NACE 80) and health and social care (NACE 85).


27 per cent of all positive job growth, expressed as a country average (compared to 7% for men). Second, the size of the welfare state in a country is a relatively good predictor of the size of contribution such jobs make to women’s overall job growth. In countries with a medium or high share of welfare state jobs (conservative, liberal and socio- democratic types), such jobs account approximately for between one-third and two-fifths of women’s overall positive job growth. In countries where the welfare state sector is less developed (Mediterranean and CEE countries), education and health and social-care jobs contribute far less to women’s overall positive job gains, at around one in five job gains or less. Four countries are exceptions to this general pattern: In Austria, welfare state jobs for women accounted for just 25 per cent of total job gains, significantly less than the average of 36 per cent among other countries with conservative welfare regimes. Also, despite their relatively low share of welfare state jobs among women (around 24% in each case), Portugal, Estonia and Slovenia nevertheless registered a relatively strong contribution of welfare state jobs to women’s positive job gains, between 24 per cent and 28 per cent ( Table 5.8 ).

The third key result associated with welfare state jobs concerns the distribution of female job growth across the five quintiles, ranked by relative wage. While in most countries welfare state jobs are an impor-tant contributor to the upgrading of women’s distribution of jobs over the period, in a small but significant group of countries welfare state jobs have clearly contributed to women’s job polarization. Furthermore, the job polarization effect seems to be related to the type of welfare state system. Figure 5.10 presents the graphic results for countries, again following the grouping according to type of welfare state.

Three groups of countries display a clear upward bias of welfare state job growth for women. The six conservative type countries record a pattern of upgrading whereby more than two in three of women’s job gains (from 65% to 79%) in education and health and social care are located in the upper two wage quintiles. By contrast, female job gains in the bottom two quintiles account for far less – between 14 per cent and 28 per cent – of welfare sector job creation. Mediterranean countries, with the exception of Greece which fits a job polarization pattern, also display an upgrading bias, albeit distributed over quin-tiles three to five. And the three socio-democratic countries show a pattern of female job growth closer to an inverted U-shape, but with an upward bias; across Denmark, Sweden and Finland, the bulk of wom-en’s job growth occurred in quintiles three and four (65%, 79% and 76%, respectively)

Tabl

e 5.

8 C

ontr

ibu

tion

of

wel

fare

sta

te j

obs

to o

vera

ll j

ob c

han

ge, b

y ge

nd

er

Soci

oD

emo

crat

ic

Den

mar

k

Swed

en

Fin

lan

d

M

F M

F

M

F

% o

f jo

b cr

eati

on

13.3

0 37

.60

12.4

0 4

4.8

0 6.

00

29.2

0

%

of

job

des

tru

ctio

n

6.7

0 9.

20

7.4

0 41

.60

4.10

10

.00

Co

nse

rvat

ive

Ger

man

y A

ust

ria

Bel

giu

m

Fran

ce

Net

her

lan

ds

M

F M

F

M

F M

F

M

F

% o

f jo

b cr

eati

on

11.1

0 3

0.0

0 7.

00

24.

80

9.4

0 33

.70

8.5

0 36

.40

6.2

0 42

.10

% o

f jo

b d

estr

uct

ion

1.

20

1.4

0 5.

80

8.8

0 4.

50

4.0

0 10

.20

23.

00

2.2

0 3.

60

Lib

eral

UK

Ir

elan

d

M

F M

F

% o

f jo

b cr

eati

on

9.70

41

.60

4.8

0 37

.20

% o

f jo

b d

estr

uct

ion

1.

80

8.9

0 2

.40

2.8

0

Med

iter

ran

ean

Po

rtu

gal

Cy

pru

s G

reec

e It

aly

Spai

n

M

F M

F

M

F M

F

M

F

% o

f jo

b cr

eati

on

7.2

0 27

.70

7.10

2

0.10

10

.90

22.0

0 3.

00

18.0

0 5.

90

20.

90

% o

f jo

b d

estr

uct

ion

1.

30

1.6

0 2

.10

3.4

0 1.

20

2.8

0 6.

50

3.10

1.

60

0.70

CE

E C

ou

ntr

ies

Hu

nga

ry

Cze

ch R

epu

bli

c E

sto

nia

L

atv

ia

Lit

hu

ania

Sl

ov

akia

Sl

ove

nia

M

F M

F

M

F M

F

M

F M

F

M

F

% o

f jo

b cr

eati

on

3.4

0 17

.80

7.3

0 19

.90

1.3

0 2

4.10

1.

60

9.50

1.

70

20.

90

2.9

0 14

.10

6.8

0 27

.70

% o

f jo

b d

estr

uct

ion

4.

70

13.9

0 3.

40

7.6

0 5.

20

19.5

0 8.

90

11.1

0 6.

40

11.6

0 4.

00

14.8

0 2

.40

8.70

Not

es :

“jo

b cr

eati

on”

is d

efin

ed a

s th

e su

m o

f al

l in

div

idu

al jo

bs

(isc

o2 3

nac

e2)

wh

ich

had

po

siti

ve e

mp

loy

men

t gr

owth

du

rin

g th

e re

spec

tive

per

iod

; “j

ob

des

tru

ctio

n”

is d

efin

ed a

s th

e su

m o

f all

ind

ivid

ual

job

s (i

sco2

3 n

ace2

) w

hic

h h

ad n

egat

ive

emp

loy

men

t gr

owth

du

rin

g th

e re

spec

tive

per

iod

; th

e ta

ble

show

s th

eref

ore

the

shar

e of

th

at j

ob

crea

tion

/des

tru

ctio

n a

cco

un

ted

by

wel

fare

sta

te s

ecto

rs, d

efin

ed a

s N

AC

E 8

0 an

d 8

5.


The liberal welfare regime grouping of the UK and Ireland displays a clear pattern of female job polarization among welfare state job change and this provides an important explanation for the overall pattern of women’s job polarization reported at the beginning of this chapter ( Figure 5.1 ). In both countries a large share of jobs created in the wel-fare sector (some 40% in both countries) is concentrated in the second quintile. The occupation of ‘personal and protective services workers’ in both the education and health and social work sectors, in particular, stands out as the most important job behind this pattern of female job growth. Unlike other countries, therefore, the welfare state sector has a much weakened upgrading impact on overall changing job structure, despite job gains for women in the top quintile.

Conservative

Mediterranean

Socio-democratic

1 2 3 4 50

20406080

100120140

Belgium

1 2 3 4 5–20–10

0102030405060

Austria

1 2 3 4 502468

101214

Luxembourg

1 2 3 4 50

50100150200250300350

Netherlands

1 2 3 4 5–200–150–100–50

050

100150

Portugal

1 2 3 4 50

5

10

15

20

25Cyprus

1 2 3 4 5–100–50

050

100150200250

Greece

1 2 3 4 5–100

0

100

200

300

400

500Italy

1 2 3 4 50

100200300400500600700800900

1,000Spain

1 2 3 4 5–60–40–20

020406080

100Denmark

1 2 3 4 5–100

–50

0

50

100

150Sweden

1 2 3 4 5–20–10

01020304050607080

Finland

1 2 3 4 5

0

400600800

1,0001,200

Germany

200

–200

(000

s)

1 2 3 4 5–200

0

200

400

600

800

1,000France

(000

s)(0

00s)

(000

s)(0

00s)

(000

s)(0

00s)

(000

s)(0

00s)

All jobs

Welfare state jobs

Welfare state jobs includingpublic administration

(000

s)(0

00s)

(000

s)(0

00s)

(000

s)

Figure 5.10 Continued


Finally, CEE countries reveal a mixed pattern. In four countries (Estonia, Hungary, Lithuania and Slovenia) welfare state job change for women follows an upgrading pattern, in two (the Czech Republic and Slovakia) it is polarizing, and in Latvia, female job creation in educa-tion and health and social care is only apparent in the second wage quintile.

5.7 Conclusion

This chapter provides further clear evidence of the heterogeneity of patterns of job change in Europe. In most countries, women’s job structure has changed in a manner quite different to that of men’s, although there is no universal pattern of either male job upgrading at the expense of female downgrading, for example, or female job upgrad-ing at the expense of male polarization. The evidence points to a mix of country patterns. Nevertheless, across all countries in the jobs data

Liberal

CEE countries1 2 3 4 5

0

20

40

60

80

100

120Ireland

1 2 3 4 5–15

–10

–5

0

5

10

15 Estonia

1 2 3 4 5–20

020406080

100120140

Hungary

1 2 3 4 5–15–10–505

1015202530

Latvia

1 2 3 4 5–80

–60

–40

–20

0

20

40Lithuania

1 2 3 4 5–20

–10

0

10

20

30

40Slovakia

1 2 3 4 5–20–10

01020304050

Slovenia

1 2 3 4 5–300–200–100

0100200300400500600700

UK

1 2 3 4 5–200

–150

–100

–50

0

50

100Czech Republic

(000

s)(0

00s)

(000

s)(0

00s)

(000

s)

All jobs

Welfare state jobs

Welfare state jobs includingpublic administration

(000

s)(0

00s)

(000

s)(0

00s)

Figure 5.10 The contribution of welfare state jobs to women’s job change, 1995–2007


set, with the sole exception of Spain, the gender gap in wage quality narrowed over the 1995–2007 period – evidence of overall net gains for women compared to men. Many jobs have changed in form, however, and women in several countries experienced growth in part-time jobs at the expense of full-time jobs; part-time jobs contributed strongly to growth among the bottom quintile jobs in those countries experienc-ing polarization of women’s jobs, and also fuelled overall job growth in several countries, especially Germany and Austria where part-time jobs substituted for full-time jobs.

Our analysis of patterns of job change among women during 1995–2007 complements and extends other studies of women’s employment and gender inequality in European employment systems (e.g. Rubery et al. 1999; Plantenga and Hansen 2008). First, women are still far more likely than men to be employed in low-wage, or bottom quintile, jobs; in 19 of the 23 countries, women were more likely than men to be in bottom quintile jobs in 2007. While the type of wage-setting institution helps to explain why some countries retain a larger stock of low-wage jobs than others, it does not match with evidence of change in the share of low-wage jobs largely due to massive sectoral shifts in employment composition; we take this as an indication of the need to marry institu-tional analysis with sectoral change (drawing on, for example, Kuznets 1955; Galbraith and Berner 2001). Second, the highly concentrated nature of women’s employment – much more so than among men – means that a handful of jobs make a massive contribution to patterns of job change (OECD 1998). Further qualitative investigation of these jobs would be valuable. Our limited analysis reveals some interesting patterns, such as the wage penalty experienced by women employed in the top-ten jobs in 16 of 23 countries despite these same jobs register-ing a skill premium in half the countries. This confirms the very strong gender bias of a job’s position in the skill ranking and the wage rank-ing – further evidence of undervaluation of women’s work (see also, Grimshaw and Rubery 2007). Third, the jobs data set complements evi-dence that points to the important role of welfare state jobs, such as education, health and social work, for women’s employment (Esping-Anderson 1999). Countries with strongly developed welfare states wit-nessed welfare state jobs contributing to anything from one-third to two-fifths of gross job creation among women. Moreover, welfare state jobs have been an important contributor to the general upgrading of women’s job structure, albeit with a polarizing influence in the liberal welfare regime countries of the UK and Ireland.


Notes

1 . It is notable that in other cross-national studies on gender bias in labour markets, Germany and Austria (as archetypal ‘coordinated market econo-mies’) are shown to display relatively strong patterns of gender inequal-ity (such as sex segregation or female share of managers) because of their specific form of vocational training and emphasis on firm/industry spe-cific training, supported by a relatively strong commitment to employment protection (Estévez-Abe 2005). Our analysis instead points to important features of gender parity in both Germany and Austria (compared to the UK, for example). Germany and Austria rank favourably in terms of the matching of women’s skill and wage quintile measures (i.e. a relatively good balance between skill level and wage level in jobs, compared to a large dis-crepancy in the case of the UK) and in terms of women’s higher average wage quintile measure (and narrower gap with men).

2 . According to the jobs dataset, the share of female employees in low wage jobs (as defined above) in 2007 was 0% in Greece, 6% in Spain and 8% in Hungary. Using a slightly different definition of low-wage work – 70% of the median of all employees rather than 75% - SILC data for 2008 produce very different shares: 24% for Greece, 27% for Spain and 27% for Hungary. The remaining countries produce far more comparable results.

3 . Using data from the annex of the EC (2008) Employment in Europe report. 4 . Eurostat data show a rising part-time employment share from 17.6 per cent

to 26.0 per cent during 1997–2007 in Germany compared to a trend of 24.6 per cent to 25.5 per cent in the UK. Among women, the part-time employ-ment share in 2007 was 46 per cent in Germany and 42 per cent in the UK (EC 2008b).

5 . Countries where retail sales jobs are ranked 2 include Germany, Austria, Sweden and Slovenia, along with the Mediterranean countries Spain, Greece, Italy and Portugal.

111

6.1 Introduction

Although not historically unprecedented, Europe is going through a time of intense change in terms of both the intensity of immigration and the public perception of this demographic phenomenon. This chap-ter aims to analyse the labour market participation of immigrants in the European Union, both from a national comparative perspective and from a European point of view. With that aim, the chapter is organized as follows. Section 6.2 deals with several methodological issues linked to the study of immigration, particularly regarding the specificities of the Jobs Project database. In order to provide a background to the current migration flows, the Section 6.3 discusses how the current immigration levels compare to previous waves both in the Old Continent and in other regions of the world. The following section presents a demographic and economic characterization of immigrant population in the European Union. The backbone of the chapter, Section 6.5, addresses the different patterns of participation of foreigners in national labour markets, includ-ing the allocation of migrants along the national job structures and the incidence of over-qualification among this group. Finally, Section 6.6 concludes and summarizes the main findings of the chapter.

6.2 Methodological issues

Several remarks must be made in order to guide the reader through the chapter regarding the definition of immigrant and the time period con-sidered in the analysis. Some of the issues are controversial and in most

6 Immigration and Labour Market Segmentation in the European Union Rafael Muñoz de Bustillo and José-Ignacio Antón

112 Rafael Muñoz de Bustillo and José-Ignacio Antón

cases our choice is determined by the possibilities and characteristics of the database.

First, there is no standardized definition of who is an immigrant. In fact, in most states, this concept simply does not exist in legal terms. However, beyond anthropological considerations, there are basically two major criteria to define who is to be considered an immigrant in applied Social Sciences research: nationality and country of birth. Both methodological choices involve advantages and disadvantages. On the one hand, the former criterion is almost always preferred to the latter because naturalization laws vary a great deal depending on national-ities and across host countries. 1 Therefore, two foreign-born workers with the same time of residence in a European country might receive a different treatment in the analysis depending on their country of birth. 2 On the other hand, using nationality often allows us to distin-guish returned emigrants and expatriates, as well as to assess the effects of naturalization in relation to specific economic outcomes. 3 There are some practical limitations that must also be taken into account; for instance, the choice of the criterion based on country of birth often contributes to enlarging the available samples, an advantage that also applies here.

The approach followed in the chapter is essentially eclectic and empirically oriented: while certain international organizations, statis-tical institutes, surveys and national or international public authorities adopt one definition, other institutions and databases – exclusively – use the other one; where possible, we have tried to favour the criterion based on the place of birth, but, when it is not possible to carry out such a strategy because of the mentioned data limitations, we have instead used the citizenship criterion. We believe that this choice is reason-able for practical considerations and better than the alternative which would be to largely restrict our analyses. A specific case can exemplify the issue. While the United Nations Population Division allows the use of both criteria, OECD databases usually favour the country-of-birth criterion and Eurostat, the approach based on citizenship. In add-ition, in the Jobs Project database, based on the European Labour Force Survey, there is not a uniform criterion either. The particular variables used for each country are presented in Table 6.1. While the data avail-able for most countries allow the researcher to use the country-of-birth criterion, this is not possible in others (Germany and Ireland) where citizenship has been employed instead. In spite of these caveats, for-tunately, the correlation between citizenship and country of birth is remarkably high, making the results obtained in the analysis robust to

Immigration and Labour Market Segmentation 113

both methodological choices. Using the Jobs Project database the corre-lation between the number of foreign people and foreign-born people in a certain job and country is roughly 95 per cent.

Another different – and controversial – issue is whether all non- native-born people should be treated in the same way in the analysis. The consideration of who is an ‘immigrant’ according to the popular perception in OECD countries is often linked to the arrival from less-developed regions. Nevertheless, the level of disaggregation available in the Jobs Project database does not allow us to consider in depth this issue in the analysis, as it is only possible to distinguish between EU and non-EU-born individuals. Brücker et al. (2002), when exploiting the European Community Household Panel data, analyse people born in the European Union (which at that time comprised 15 members) jointly with natives. This perspective is appealing, but, since our data cover years before and after the EU enlargements of 2004–2007, in order to work with a time-consistent concept of the foreign-born population we have considered all those workers born abroad as immigrants. The alternative – to completely exclude EU citizens from the category of migrants – does not strike us as reasonable or sensible as it would lead to us to ignoring foreign-born groups as important as the Romanians in Spain, Poles in Ireland and the UK and so on. To consider Finns in Sweden or EU civil servants in Brussels as (highly qualified) migrants may stretch our standard preconceptions of what a migrant worker is but such an approach does nonetheless offer a more coherent and com-prehensive approach than excluding such workers from the migrant category.

The third methodological problem has to do with the temporal dimension: not all annual waves contain information on country of birth or nationality, so the analysis has to necessarily restrict itself to the available information. As explained in detail by Fernández-Macías and Hurley (2008), the database suffers from some structural breaks (e.g. methodological changes in occupational or sectoral taxonomies). When using data by job quintiles from a dynamic perspective, we dealt with such problem using the procedure suggested by these authors: to compute both the change until the year just before the structural break and the variation of employment from that year onwards. Finally, in two Eastern European countries, the Czech and the Slovak Republics, immigrants account for less than 2 per cent of total working popula-tion. Therefore, we have decided not to include these two countries in the analyses, given the small sample sizes and the – inherent – complexity of presenting, organizing and reading data from more than


twenty countries. In order to focus on the more relevant cases, we have excluded these two states from analyses.

The definition of migrant and time period considered for each country are presented in Table 6.1.

6.3 Migration in Europe

Migration is the great absentee of the current process of globaliza-tion. During the previous globalization wave, from the mid-1850s to the beginning of World War I, financial movements, trade and people underwent similar trends in terms of increasing transnational mobility. During that period, in a context of virtually free movement of peo-ple, massive migration led to dramatic demographic changes (Hatton

Table 6.1 Methodological issues in the analysis of the Jobs Project database

Country Structural breaks Migration variables

Austria 2003–2004 Country of birth (1995–2006) Belgium Country of birth (1995–2006) Cyprus Country of birth (1999–2006) Czech Republic Country of birth (2002–2006),

<2% of foreign-born workers Germany Nationality (1995–2006) Denmark Country of birth (1995–2006) Estonia Country of birth (1998–2006) Spain Country of birth (1995–2006) Finland 2001–2002 Country of birth (1997–2006),

<5% of foreign-born workers France 2003–2004 Country of birth (1996–2006) Greece Country of birth (1995–2006) Hungary Country of birth (2001–2006),

<2% of foreign-born workers Ireland 1997–1998 Nationality (1998–2003 and 2006) Italy 2003–2004 Country of birth (2006) Lithuania Country of birth (1998–2006) Luxembourg Country of birth (1995–2007) Latvia Country of birth (2004–2006) Netherlands Country of birth (1999–2006) Sweden Country of birth (1997–2006) Portugal 1997–1998 & 2000–2001 Country of birth (1999–2006) Slovenia Country of birth (2002–2006) Slovak Republic Country of birth (2003–2006),

<2% of foreign-born workers United Kingdom 2000–2001 Country of birth (1995–2006)

Source : Authors’ analysis from Jobs Project database.


and Williamson 1998). In contrast, the world immigration rate (immi-grants/total population) has remained basically stable since 1990, at a relatively low 3 per cent. 4 Nevertheless, this stability hides an important change in terms of the direction of the immigration flows. As depicted in Figure 6.1 , since 1985 there has been an increase in the importance of Europe as a receiving region. In fact, in 2010, included in the analy-sis, Europe as a continent hosted more immigrants than North America (69.8 million compared to 50 million); traditionally the immigrants’ promised land (United Nations 2009). In this context, during the last decade, while many of the classic European host countries such as Belgium experienced a period of immigration stability, other states – for instance, Spain and Ireland – experienced sudden immigration flows of very high intensity, comparable in scale to that experienced by the US at the turn of the last century.

6.3.1 Immigration in Europe: major facts

The diversity of the Old Continent also applies to their migration patterns, since the percentage of foreign-born population varies a lot

0

2

4

6

8

10

12

14

16

18

20

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

%

World EuropeNorth América AsiaLA and the Caribbean AfricaOceania

Figure 6.1 Immigrants as percentage of total population, by region, 1960–2005

Note : Immigration figures (generally) refer to the foreign-born population.

Source : Authors analysis from United Nations (2009).


across EU Member States ( Figure 6.2 ). In this respect, Luxembourg, as a country where almost half of its population is immigrant, stands clearly out from the rest. The next three countries at the top of rank-ing – two Baltic countries plus Cyprus – have far lower percentages of immigrant stock than does Luxembourg. Their position reflects very specific national circumstance. While in the Baltic countries these figures are mainly associated to workers arriving from the Soviet Union in the second half of the 20th century (Schmid 2004), in the Cypriot case this outcome is the result of a more dynamic economy and a more open immigration policy at the beginning of the 1990s (Trimikliniotis and Demetriou 2005). A second group of countries, composed of Ireland and Spain, was until relatively recently better known as sending than receiving countries. After them, there is a continuum of traditional host countries, such as Belgium, Germany, France and Sweden, while the bottom of the spectrum is predomi-nantly composed of new Member States (other than the Baltic Member States, for reasons indicated above). 5

The second major fact is associated with the existence of very differ-ent time patterns of immigration among the EU countries. According to Figure 6.3 , which reproduces the stock of immigrants as the percentage

0

5

10

15

20

25

30

35

40

45

RO PL

BG SK LT HU FI

CZ SI

MT

NL

PT

DK

SE IT FR

EU

27E

U25 UK

EU

15 GR

DE

BE AT ES

IE CY

EE LV LU

%

Immigrants extra EU27 Immigrants from the EU27

Figure 6.2 Foreign population as percentage of total population in EU countries, 2008 Source : Authors’ analysis from Eurostat database.


510

1520

05

1015

20

2010

05

1015

20

05

1015

20

1960 1970 1980 1990 20001960 1970 1980 1990 2000 20101960 1970 1980 1990 2000 2010

05

1015

20

010

2030

40

05

1015

20

05

1015

20

1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 20101960 1970 1980 1990 2000 2010

05

1015

20

05

1015

20

05

1015

20

1960 1970 1980 1990 2000 20101960 1970 1980 1990 2000 20101960 1970 1980 1990 2000 2010

05

1015

20

1960 1970 1980 1990 2000 20101960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

05

1015

20

05

1015

200

1020

3040

Hungary

France

Malta

Latvia

05

1015

20

05

1015

20

1960 1970 1980 1990 2000 20101960 1970 1980 1990 2000 20101960 1970 1980 1990 2000 2010

Estonia

Germany

Lithuania

Ireland

010

2030

40

1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 20101960 1970 1980 1990 2000 2010

Finland

Greece

Italy

Luxembourg

BulgariaAustria Belgium

DenmarkCyprus Czech Rep.

Figure 6.3 Continued


of total population from 1960 (or 1990, depending on the countries) to 2005, one can distinguish four different groups of countries:

● Group A , comprising all eastern European countries with the excep-tion of the Czech Republic, Slovakia and Bulgaria, with decreasing immigration rates. ● Group B , classic host countries, including Belgium, France, the Netherlands and Germany, where immigration rates rise very early, reaching a plateau in the 1980s or 1990s, depending on the case, and remaining relatively stable since then (or even decreasing as in the case of Belgium). ● Group C , with the latecomers, that is, Spain, Greece, Ireland, Italy, Cyprus, Finland and probably Austria, showing a late but steep rise in their immigration rates. ● Group D , involving countries as different as Sweden and Portugal, showing, for very different reasons, a continuous, almost linear increase of immigrant rates.

The existence of such a different time pattern is important, as the type and rate of labour market integration of immigrants might be very

PolandNetherlands

SloveniaSlovakiaRomania

Portugal

UKSpain Sweden

1960 1970 1980 1990 2000 2010

05

1015

200

510

1520

05

1015

20

1960 1970 1980 1990 2000 2010

1960 1970 1980 1990 2000 2010

1960 1970 1980 1990 2000 2010

05

1015

200

510

1520

05

1015

20

1960 1970 1980 1990 2000 2010

1960 1970 1980 1990 2000 2010

1960 1970 1980 1990 2000 2010

05

1015

200

510

1520

05

1015

20

1960 1970 1980 1990 2000 2010

1960 1970 1980 1990 2000 2010

Figure 6.3 Time patterns of immigration: percentage of foreign-born population in the EU27, 1960–2010

Source : Authors’ analysis from United Nations (2009).


different for those immigrants newly arrived (as in Spain) compared to immigrants with many years of residency in their host country.

A third major fact relates to the different origins of the immigrant population in Europe. Depending on the timing of the immigration waves, the geographical location and the cultural and historical links of the countries, the EU Member States differ widely in terms of the coun-try of origin of their immigrants. As shown in Figure 6.2 , while some countries – such as Luxembourg, Cyprus and Ireland – host mostly non-national EU citizens, in others the majority of the foreign population comes from outside of the EU. This is the case, for example, in Latvia, Germany, Austria and Spain. Detailed data of the foreign-born popula-tion by country of origin ( Table 6.2 ) show the relevance of geographical proximity and of cultural, historical and linguistic relations as well as the role of immigrant networks in the development of such specific patterns of migration. 6 For instance, Austria receives its extra-EU immi-grants mostly from the former Yugoslavia and Turkey, France from the countries of the Maghreb (almost 1/3 of immigrants are from Algeria, Morocco and Tunisia), Greece from Albania (36% of immigrants), Poland from Ukraine (40%), and Spain from South America. Even the United Kingdom, with a broader diversity of immigrants in comparison with the rest of the EU countries, shows an important concentration of immigrants from India, Pakistan and Bangladesh.

6.4 Characteristics of immigrants: age, gender, education and employment

The analysis of immigrants’ performance in the European labour mar-kets must be necessarily preceded by a brief review of the main socio-economic characteristics of this population group, particularly, when compared to locals. The following pages, which present a descriptive analysis of the foreign population by age, gender, employment status and educational level, aim to accomplish such an objective.

Age structure. Immigrants are to a large extent young people. This is unsurprising if one considers migration as an investment decision where this demographic segment can expect higher benefits (more years ahead to recoup the investment and better physical conditions, among other rea-sons) and faces lower costs (e.g. lower attachment to their home countries). According to Eurostat, half of immigrants are aged between 15 and 39, compared to only one-third of nationals. This difference is even sharper in countries such as Spain, where immigration is a very recent and intense phenomenon and a large portion of immigrants are newcomers and still young (62% of the total immigrant population belong to this age group).

Tabl

e 6.

2 St

ock

of

fore

ign

-bor

n p

opu

lati

on i

n a

sel

ecti

on o

f EU

Mem

ber

Sta

tes,

by

cou

ntr

y of

bir

th, a

rou

nd

20

06

Au

stri

a B

elg

ium

D

enm

ark

Fi

nla

nd

Fr

ance

G

reec

e H

un

gary

Ir

elan

d

Ex-

Yu

gosl

avia

14

.7

Fran

ce

12.1

Tu

rkey

9.

8 E

x-U

SSR

22

.3

Alg

eria

13

.7

Alb

ania

36

.0

Rom

ania

49

.5 U

nit

ed

Kin

gdo

m

44.

2

Turk

ey

12.7

M

oro

cco

11.8

G

erm

any

8.0

Swed

en

15.9

M

oro

cco

12.6

G

erm

any

9.0

Ex-

Cze

cho

slov

akia

8.8

Pola

nd

10

.4

Ger

man

y 12

.7

Ital

y 9.

4 Ir

aq

3.8

Est

onia

7.

7 Po

rtu

gal

11.4

Tu

rkey

6.

8 E

x-Y

ugo

slav

ia

8.3

Un

ited

St

ates

4.

1

Bo

snia

H

erze

gov

ina

12.6

N

eth

erla

nd

s 8.

8 B

osn

ia

Her

zego

vin

a 6.

3 So

mal

ia

2.8

It

aly

6.8

Ru

ssia

n

Fed

erat

ion

6.

5 E

x- U

SSR

7.

9 L

ith

uan

ia

4.1

Ex-

Cze

cho

slov

akia

5.

8 Tu

rkey

6.

5 Po

lan

d

3.6

Ex-

Yu

gosl

avia

2

.8

Spai

n

5.6

Geo

rgia

6.

4 G

erm

any

7.1

Nig

eria

2

.7

Pola

nd

5.

1 G

erm

any

6.3

No

rway

4.

5 G

erm

any

2.6

Tu

rkey

4.

5 B

ulg

aria

3.

5 A

ust

ria

1.8

Lat

via

2

.3

Cro

atia

4.

2 D

. R. o

f th

e C

on

go

5.3

Swed

en

4.3

Ch

ina

2.4

Tu

nis

ia

4.5

Egy

pt

2.9

U

krai

ne

1.4

Ger

man

y 1.

9

Rom

ania

4.

2 Sp

ain

2

.7

Leb

anon

4.

0 Ir

aq

2.4

G

erm

any

2.6

R

oman

ia

2.4

C

hin

a 1.

4 C

hin

a 1.

8 H

un

gary

2

.6

Pola

nd

2

.6

Iran

3.

8 T

hai

lan

d

2.2

U

nit

ed

Kin

gdo

m

2.5

K

azak

hst

an

2.2

U

nit

ed S

tate

s 1.

2 Ph

ilip

pin

es

1.6

Ital

y 1.

8 Se

rbia

&

Mo

nte

neg

ro

2.4

Fo

rmer

Y

ugo

slav

ia

4.3

Un

ited

K

ingd

om

2

.0

Bel

giu

m

2.1

U

nit

ed

Stat

es

2.1

Po

lan

d

1.0

Ind

ia

1.5

Slov

enia

1.

4 R

uss

ian

Fe

der

atio

n

2.3

U

nit

ed

Kin

gdo

m

3.7

Turk

ey

2.0

Po

lan

d

1.8

Cy

pru

s 2

.0

Oth

ers

11.6

Fra

nce

1.

5

Oth

ers

22.2

U

nit

ed

Kin

gdo

m

1.8

Paki

stan

3.

4 U

nit

ed

Stat

es

1.8

Vie

tnam

1.

5 A

ust

rali

a 1.

8

R

oman

ia

1.4

Port

uga

l 1.

8 So

mal

ia

3.7

Iran

1.

8 Se

neg

al

1.4

Ukr

ain

e 1.

5

So

uth

A

fric

a 1.

3

Alg

eria

1.

5 A

fgh

anis

tan

0.

8 V

iet

Nam

1.

8 C

hin

a 1.

3 Po

lan

d

1.4

Au

stra

lia

1.1

Rom

ania

1.

2 V

ietn

am

2.8

In

dia

1.

3 M

ali

1.1

Un

ited

K

ingd

om

1.

2

Sp

ain

1.

0

Oth

ers

23.

6 O

ther

s 33

.3

Oth

ers

28.

2 O

ther

s 26

.8

Oth

ers

14.5

O

ther

s 19

.1

Lu

xem

bo

urg

P

ola

nd

P

ort

uga

l Sl

ov

ak R

epu

bli

c Sp

ain

Sw

eden

U

nit

ed K

ingd

om

Port

uga

l 2

8.8

Ukr

ain

e 4

0.2

An

gola

2

6.7

Cze

ch

Rep

ubl

ic

51.9

M

oro

cco

11.8

Fi

nla

nd

15

.4

Ind

ia

9.9

Fran

ce

13.0

B

elar

us

13.6

Fr

ance

14

.6

Hu

nga

ry

10.8

R

oman

ia

9.7

Iraq

7.

0 Ir

elan

d

7.2

Bel

giu

m

10.2

G

erm

any

12.7

M

oza

mbi

qu

e 11

.7

Ukr

ain

e 6.

4 E

cuad

or

8.3

Ex-

Yu

gosl

avia

6.

3 Pa

kist

an

4.8

Ger

man

y 8.

9 L

ith

uan

ia

10.3

B

razi

l 7.

7 Po

lan

d

3.5

Un

ited

K

ingd

om

6.

1 Ir

an

4.7

Ger

man

y 4.

7

Ital

y 8.

5 R

uss

ian

Fe

d.

7.1

Cap

e V

erd

e 6.

9 R

uss

ian

Fe

der

atio

n

2.8

C

olo

mbi

a 5.

6 B

osn

ia

Her

zego

vin

a 4.

7 Po

l an

d

4.0

Serb

ia &

M

on

ten

egro

4.

5 Fr

ance

4.

4 G

erm

any

3.7

Ger

man

y 2

.3

Arg

enti

na

5.2

Pola

nd

4.

4 B

angl

ades

h

3.8

Net

her

lan

ds

2.3

U

nit

ed

Stat

es

1.1

Ven

ezu

ela

3.4

Mac

edo

nia

2

.2

Ger

man

y 4.

2 N

orw

ay

3.8

Sou

th A

fric

a 3.

4

Un

ited

K

ingd

om

2

.2

Cze

ch

Rep

ubl

ic

0.8

Gu

inea

-B

issa

u

3.3

Rom

ania

2

.1

Fran

ce

4.0

Den

mar

k 3.

8 U

nit

ed S

tate

s 2

.9

Cap

e V

erd

e 1.

7 A

ust

ria

0.5

Spai

n

2.1

A

ust

ria

1.9

Bol

ivia

3.

8 G

erm

any

3.7

Ken

ya

2.4

Spai

n

1.5

Kaz

akh

stan

0.

5 Sw

itze

rlan

d

2.0

U

nit

ed

Stat

es

1.7

Peru

2

.6

Turk

ey

3.2

Jam

aica

2

.3

Bo

snia

H

erze

gov

ina

1.2

Serb

ia &

M

on

ten

egro

0.

5 Sa

o To

me&

P

rin

. 1.

9 Fr

ance

1.

6 V

enez

uel

a 2

.5

Ch

ile

2.4

N

iger

ia

2.0

Den

mar

k 1.

0 R

oman

ia

0.4

Sou

th A

fric

a 1.

7 V

iet

Nam

1.

1 B

ulg

aria

2

.3

Leb

anon

1.

9 A

ust

rali

a 2

.0

U

nit

ed S

tate

s 0.

8 It

aly

0.4

Un

ited

K

ingd

om

1.

5 B

ulg

aria

0.

8 B

razi

l 2

.2

Th

aila

nd

1.

7 Fr

ance

1.

9

Ch

ina

0.7

Bo

snia

H

erze

gov

ina

0.4

Can

ada

1.1

Bel

giu

m

0.4

Port

uga

l 2

.1

Un

ited

K

ingd

om

1.

5 Sr

i L

anka

1.

8

Pola

nd

0.

7 U

nit

ed

Kin

gdo

m

0.4

Un

ited

Sta

tes

1.1

Serb

ia &

M

on

ten

egro

0.

4 C

hin

a 2

.1

Syri

a 1.

5 Ph

ilip

pin

es

1.7

Ot h

ers

14.3

O

ther

s 6.

8 O

ther

s 10

.4

Oth

ers

10.1

O

ther

s 27

.5

Oth

ers

33.9

Oth

ers

45.1

Not

e : N

atio

nal

dat

a so

urc

es a

re n

ot h

arm

oniz

ed i

n t

he

way

th

ey d

eal

wit

h c

ou

ntr

ies

wit

h r

ecen

tly

chan

ged

bo

rder

s su

ch a

s th

e fo

rmer

Yu

gosl

avia

, C

zech

osl

ovak

ia a

nd

th

e fo

rmer

USS

R.

Sour

ce :

Au

thor

s’ a

nal

ysis

fro

m O

EC

D (

20

08:

22

5–2

36).


Gender. Few doubts can be cast on the relevance of taking into account gender when analysing the functioning of labour markets. While ear-lier waves of immigration were to a large extent male dominated, that is no longer the case. In the EU27 women make up half of the immi-grant population, and this is generally true across all EU Member States. Nevertheless, this does not apply to specific groups of immigrants. For example, in France, 65 per cent of immigrants from Mali are men, in contrast, two-third of Polish immigrants are women. In Spain, women are over-represented in the Latin-American group, making up 57 per cent of Ecuadorian immigrants, for example, but under-represented in the Moroccan group (36%).

Educational level. One of the main concerns of host countries is the skill level of foreign-born workers, as evidenced by the restrictions and quotas imposed on lower-educated immigrants by countries such as the United States or the United Kingdom, among others. In this respect, the general pattern observed across EU countries ( Figures 6.4 and 6.5) points to the lower schooling levels among foreign-born workers than among locals, irrespective of gender. However, there are some exceptions worth mentioning. First, in some countries –particularly, Latvia, Estonia, Ireland, Portugal and the United Kingdom – immigrants exhibit similar schooling attainment levels to nationals. This circumstance is proba-bly related to the fact that many of these migrants do not correspond with the popular stereotype of foreign workers. For example, there are a lot of British and Irish people working in the Ireland and the United Kingdom, respectively, while in Portugal return migration from France constitutes one of the major foreign-born population groups. In the case of Estonia and Latvia, as mentioned before, the Russian-speaking pop-ulation accounts for a very significant share of the non-native labour force, and this population tends to have higher qualifications than the ‘standard’ foreigner. In other countries, such as Spain or Belgium, even though the proportion of individuals with a college degree is higher among employed nationals, the percentage of immigrant workers with low levels of schooling is lower than among locals. The opposite pattern is observed in Luxembourg and Belgium.

Employment and unemployment. Although often fuelled by political reasons, the main drive of immigration is to improve the economic outcomes of immigrants and their families. This, together with the characteristic mentioned above of a majority of immigrants belong-ing to the prime age working group, helps to explain why in most EU countries foreign workers have a high labour force participation rate and, thus, a high employment rate, even if immigrants also suffer from higher unemployment rates ( Figures 6.6 and 6.7). However, there are

0 20 40 60 80 100%

UK

SI

SE

PT

NL

LV

LU

LT

IT

IE

GR

FR

FI

ES

EE

DK

DE

CY

BE

AT

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

Low Medium High

Figure 6.4 Distribution of male workers by educational level, 2006 Note : Low = ISCED-1 and ISCED-2; Medium = ISCED-3 and ISCED-4; High = ISCED-5 and ISCED-6

Source : Authors’ analysis from Jobs Project Database.

0 20 40 60 80 100

%

UK

SI

SE

PT

NL

LV

LU

LT

IT

IE

GR

FR

FI

ES

EE

DK

DE

CY

BE

AT

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

MigrantsNatives

Low Medium High

Figure 6.5 Distribution of female workers by educational level, 2006

Note : Low = ISCED-1 and ISCED-2; Medium = ISCED-3 and ISCED-4; High = ISCED-5 and ISCED-6

Source : Authors’ analysis from Jobs Project Database.


DK NL

SE

PT

SI

UK

AT

EE

CY

FI

CZ

LV

DE

IE

EU15

EU27

RO

ESSK

FR

GR

BG

BE

HU

LU

IT

PL

MT

30

35

40

45

50

55

60

65

70

555045403530

Employment rate (nationals)

Em

ploy

men

t rat

e (im

mig

rant

s)

En < Ei

En > Ei

Figure 6.6 Employment rate among national and immigrants in the EU, 2nd quarter, 2006 Note : The employment rate is defined as employment by total population (of each popu-lation group)

Source : Authors’ analysis from Labour Force Survey data.

0

5

10

15

20

25

LU NL

DK

CY AT IE UK

EE IT CZ

PT

EU

15 HU

BE

SE

EU

27

EU

25 ES

FR FI

GR

DE

%

Immigrants Nationals

Figure 6.7 Unemployment rate among nationals and immigrants in the EU, 2nd quarter, 2006

Source: Authors’ analysis from Labour Force Survey data.


remarkable exceptions to this general picture: in many of the classic immigration countries, such as Belgium, Germany, Denmark or the Netherlands, immigrants have lower employment rates than locals and unemployment is more acute among nationals in Greece and the Czech Republic. It is also worth mentioning the lack of a consistent relation-ship (of any sign) between the employment rates of both groups of pop-ulation. Denmark and the Netherlands, for example, have employments rates well above the EU average, but immigrants’ employment rates are below the EU average, while the opposite is true for Greece or Italy.

6.5 Immigrants performance in European labour markets

6.5.1 Occupational segregation, job creation and job destruction

This section tries to answer the question of what types of jobs for-eign-born workers fill and how this pattern compares to the locals’. Figures 6.8 , 6.9 and 6.10 show the distribution of immigrant and local workers – total workers, male and female, respectively – by quintiles of jobs, taking 2006 as the reference year for constructing the rank. Several interesting facts should be highlighted. First, the general rule is the concentration of migrants in the lowest quintiles of the distri-butions, which suggests that foreign-born workers enjoy jobs of lower quality than locals. However, there are non-negligible differences across countries, pointing to the existence of several different types of labour market participation of foreign-born workers. While in some countries such as Austria, Cyprus, Germany, Spain, Cyprus, France and Italy the concentration of immigrants at the bottom is particularly intense, there are other countries where the immigrant population does not seem to face such a severe disadvantage compared to natives. Countries such as Finland (where a high share of the foreign population is Swedish), the United Kingdom (with an important presence of Irish workers) and Belgium (with Brussels being the headquarters of most of European institutions employing thousands of high-skilled foreigners) exemplify this point. A look at the same data from a gender perspective reveals that women, and particularly female immigrants, are, in general terms, under-represented in the top two quintiles of jobs.

The quantitative analysis presented above can be complemented with some qualitative information about what specific jobs are more com-mon among immigrants and how this picture compares with locals. For reasons of space, we limit the discussion to a few remarks derived from


detailed analyses carried out separately for men and women, since male and female labour markets are clearly segregated, with women special-ized in different activities than men. 7

First, regarding male workers, it is remarkable that in all countries, low- or medium-skilled jobs in the construction sector are always

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

010203040

1 2 3 4 5

Cyprus

Lithuania

Slovenia

Estonia

Greece

Belgium

Germany

Austria

Ireland

Luxembourg

Portugal

United Kingdom

Spain

Sweden

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

1 2 3 4 5

Netherlands

Latvia

Italy

Finaland France

Denmark

Natives Migrants

Figure 6.8 Distribution of total workers by job quintile in the EU, percentage of each group, 2006



among the five most common jobs for both natives and migrants, although in most cases more important among the latter group. In add-ition, while jobs in education or health care seem to play an important role for native workers, this is not always the case for migrants. Only in

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

0

10

20

30

Natives Migrants

Cyprus

Lithuania

Netherlands

Slovenia

Estonia

Greece

Belgium

DenmarkGermany

Austria

Ireland Italy

LatviaLuxembourg

Portugal

United Kingdom

Spain Finaland France

Sweden

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

Figure 6.9 Distribution of male workers by job quintile in the EU, percentage of each group, 2006



some particular cases, such as the Nordic countries, do foreign work-ers have a significant participation in these kinds of jobs. In contrast, especially in France and other Mediterranean countries, employment in low-qualified jobs in the services sector (particularly, in hotels and

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

0

20

40

60

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5

Cyprus

Lithuania

Slovenia

Estonia

Greece

Belgium

Germany

Austria

Ireland

Luxembourg

Portugal

United Kingdom

Spain

SwedenNetherlands

Latvia

Italy

Finaland France

Denmark

Natives Migrants

Figure 6.10 Distribution of female workers by job quintile in the EU, percentage of each group, 2006



restaurants) is especially significant and this sometimes includes those managing small firms.

In relation to women, the pattern is different. Among nationals, the services associated with the welfare state – health care, education and other social services – and the public administration – especially, office clerks – have a major relevance in terms of female employment. Although these types of jobs have also a non-negligible role in the female immigrant job structure, foreign-born women are over-represented in low-skill jobs in the retail sector, hotels and restaurants industry and, particularly, among domestic servants.

A closer look at job structure tells us that immigrant employment is more concentrated in the five top-employing jobs (understood in the jobs project sense as occupation x sector cells) than locals: while often less than 30 per cent of native-born workers are concentrated in the five top-employing jobs among the local population, immigrants have a higher concentration in their corresponding sectors. In this respect, Portugal, Cyprus and Spain represent extreme cases, with roughly half of the foreign-born labour force working in only five types of jobs. This might be related not only to different socio-demographic characteris-tics of locals and migrants or occupational segregation resulting from discriminatory employer hiring practises, but also to the role played by occupational networks of migrants in the process of labour market entry by immigrants. It is the case, for example, that the foreign-born population tends to locate, other things being equal, in the same jobs as previously established immigrants from the same country, whose advice and help are very important in the process of finding a job. 8 This pattern might have non-trivial consequences: for example, the higher the concentration of immigrants in a specific sector, the higher will be their risk of facing specific employment problems in case of a downturn in the specific sector (e.g. construction).

From a more rigorous and quantitative point of view, it is possible to formally assess how different is the job allocation between immigrants and locals. There is a considerable and increasing volume of literature, starting as early as the 1950s, focused on analysing how workers are allocated across jobs or occupations; that is, occupational segregation, jointly with their causes and potential implications (discrimination, lower wages, limited opportunities of advancement, working conditions etc.). This academic interest has fostered the development of better and better measurement tools for depicting and analysing this issue, which has received particular attention from the gender perspective. As it is beyond the scope of this work to construct new or more sophisticated


indicators, we will rely on the Duncan Dissimilarity Index (DDI) (Duncan and Duncan 1955), an index widely used by social scientists, that yields information about how different is the distribution of two population groups (in our case, locals and immigrants) across jobs. The DDI can be formally expressed as follows:

12

i iN I

i

DDI p p=

where p i j ( j = N , I ) denotes the proportion of individuals of group j placed in the job i . The statistic is bounded by 0 (no different occupa-tional pattern) and 1 (complete segregation across jobs). Therefore, the higher the DDI, the larger is the occupational segregation; that is, the more different is the distribution of migrants and natives across jobs.

The main results of the application of the DDI are shown in Table 6.3 , which reproduces the DDI by country of origin (foreign-born v. native-born workers) by country of origin and gender and, in the last column,

Table 6.3 Duncan Index of occupational segregation in Europe for migrant status and gender, 2006

Segregation by migrant status Segregation by gender Total Men Women

AT 0.372 0.399 0.395 0.539 BE 0.238 0.264 0.265 0.516 CY 0.477 0.462 0.570 0.595 DE 0.377 0.413 0.405 0.534 DK 0.328 0.412 0.337 0.530 EE 0.388 0.458 0.417 0.661 ES 0.423 0.412 0.503 0.564 FI 0.415 0.492 0.460 0.594 FR 0.317 0.337 0.360 0.533 GR 0.587 0.588 0.614 0.465 IE 0.353 0.385 0.361 0.580 IT 0.398 0.397 0.435 0.487 LT 0.472 0.534 0.459 0.623 LU 0.515 0.543 0.541 0.534 LV 0.367 0.448 0.400 0.639 NL 0.254 0.304 0.262 0.511 PT 0.326 0.379 0.346 0.523 SE 0.257 0.338 0.235 0.526 SI 0.419 0.466 0.472 0.523 UK 0.249 0.310 0.235 0.526



solely by gender (i.e. men v. women) in order to see whether segregation is higher by gender (men v. women) or by country of origin (foreign-born v. native-born). No clear pattern emerges from this picture, but several issues can be highlighted. First, Belgium, the United Kingdom, Netherlands, Sweden and France are the countries where migrants work in jobs that are more similar to native ones. In contrast, in Greece, Luxembourg, Cyprus, Latvia and Spain the occupational segregation between migrants and locals is the highest. Second, in general terms, differences by gender are not large (the correlation between DDI for male and DDI for female across countries is nearly 95%), but there are remark-able differences in several countries, such as Cyprus (where women are much more segregated than men) and Sweden (where the female job structure is more homogenous than the male). Finally, when compar-ing these figures with occupational segregation by gender in order to calibrate the magnitude of the differential patterns of employment for migrants and locals, the only case where segregation by migrant status exceeds dissimilarity by gender is Greece.

According to standard labour market analysis, immigration produces an increase in the supply of labour, and, ceteris paribus , a reduction in wages in comparison with a zero immigration situation. The impor-tance of this impact will depend on the immigration rate and the sub-stitutability between immigrant and local workers. In the hypothetical case of a single labour market (i.e. not segmented) and identical charac-teristics of immigrants and locals, this impact would be highest. In con-trast, if immigrant and locals have completely different characteristics and are employed in completely different segments in the labour mar-ket, the effect might be null. Finally, it can be argued that if immigrants and locals complement each other, the result can be an increase in the productivity of local workers and, in a purely competitive framework, an increase in their wages. 9

Is it then possible to evaluate from the previous analysis to what extent immigration affects the situation in the labour market of local workers? Unfortunately the answer is no, as it is possible, and even probable, that, in the presence of a high increase of immigrant labour supply in specific niches of the labour market, local workers might react changing jobs, climbing up the labour ladder if employment is grow-ing and moving to different niches of the labour market less affected by immigrant labour supply. This means that after comparing the job structure of immigrants and locals we can say to what extent are both groups of workers competing in the same market niches (as defined in the jobs approach) in a given moment of time, but we cannot evaluate


whether they have not competed in the past. A possible, although very indirect, way to address this problem is to calculate the correla-tion between changes in immigrant and local employment by jobs, to see to what extent, in a given period of time, local and immigrant employment follow different patterns of job creation. In this respect, as we can see in Table 6.4 , the correlation between the growth of native and immigrant employment at the job level reveals (once again) no clear pattern: although employment growth among both populations groups is positively correlated (with the exception of Greece, with a negative but almost null value), the range of results goes from the prac-tically zero correlation in Lithuania, Portugal and Sweden to values around 65 per cent in Ireland and Austria. In general, migrant job crea-tion is not related with native job destruction; on the contrary, it seems that the demand side is a much more important force when trying to explain the relationship between employment changes among both population groups.

Given the impressive immigration flows experienced by some EU countries such as Spain or Austria, it seems of particular interest to look at the dynamics of job creation and job destruction, establishing where employment is being created and destroyed for immigrants and locals. Figures 6.11 –6.13 reproduce the employment growth as a percentage of the number of workers in each quintile in 1995, by native and migrant status. As concluded from the analysis of employment structure, the dynamics of job creation and destruction for immigrants and locals (both for men and women) are to a large extent country specific.

Nevertheless, new migrant jobs tend to be created mainly in the two lowest quintiles. Austria, Cyprus and Spain are the main exponents of this trend. Luxembourg represents a noteworthy exception, as there is an impressive growth of jobs performed by foreign-born individuals in the top quintile.

Connecting with the analysis of the employment dynamics of local and immigrant workers discussed in the previous section, using Figure 6.11 , it is possible to identify different patterns of job creation and

Table 6.4 Correlation coefficient between change in native and migrant employment growth by job cell

GR LT PT SE CY FR IE LU DK EE FI DE ES NL BE UK AT

–0.040 0.044 0.067 0.073 0.163 0.227 0.264 0.264 0.336 0.348 0.380 0.441 0.475 0.507 0.509 0.615 0.650

Note : Changes are referred to time periods considered in Table 6.1 .



020

4060

–10

0

10

20

0

10

20

30

–40

–20

0

20

–20–10

01020

–10

0

10

20

–20

0

20

40

010203040

0

20

40

60

0

10

20

30

–100

102030

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

–10

0

10

20

–10

0

10

20

020406080

–40–20

02040

–50

0

50

100

–20

0

20

40

Austria (1995–2006) Belgium (1995–2006) Cyprus (1999–2006)

Germany (1995–2006) Denmark (1995–2006) Estonia (1998–2006)

Spain (1995–2006) Finland (1997–2006) France (1996–2006)

Greece (1995–2006) Ireland (1998–2006) Lithuania (1998–2006)

Luxembourg (1995–2006) Netherlands (1999–2006) Portugal (1999–2006)

Sweden (1997–2006) United Kingdom (1995–2006)

Natives Migrants

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5

Figure 6.11 Job creation, destruction and immigration in the EU, percentage employment of growth by job quintile



0

20

40

–20

0

20

40

–10

0

10

20

–40

–20

0

20

–30–20–10

010

–10

0

10

20

–10

0

10

20

–10

0

10

20

0

20

40

60

–40

–200

2040

–20

0

20

40

010203040

–10

0

10

20

–20

0

20

40

0

20

40

60

–100

102030

–10

0

10

20

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

Belgium (1995–2006) Cyprus (1999–2006)






Natives Migrants

Austria (1995–2006)

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5

Figure 6.12 Job creation, destruction and immigration in the EU, percentage male employment of growth by job quintile



0

20

40

60

–100

102030

010203040

–20

0

20

–20

–100

1020

0

10

20

30

0

10

20

30

0

50

100

150

–200

204060

0

50

100

150

–50

0

50

100

–20

0

20

40

–50

0

50

100

010203040

020406080

0

10

20

30

–20

0

20

40

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

1 2 3 4 5 1 2 3 4 5

Austria (1995–2006) Belgium (1995–2006) Cyprus (1999–2006)






Natives Migrants

Figure 6.13 Job creation, destruction and immigration in the EU, percentage female employment of growth by job quintile



destruction in terms of quintiles of job quality and distinguishing between local and foreign workers. In order to facilitate the interpre-tation of the figure, Table 6.5 summarizes such tendencies focusing on whether the evolution of employment of immigrants and locals by job quintiles follows opposing (columns 2 and 3) or similar paths (columns 4 and 5). As we can see, in the first quintile there is a gen-eral pattern of substitution of local labour by immigrant labour (10 countries), although in five countries (Spain, France, Ireland, the Netherlands and Cyprus) both immigrant and local employment grows. (In Spain and Cyprus the intensity of growth is higher among immigrants, while in France and Ireland it is the opposite.) The nota-ble exception is Germany, where employment in this quintile grows among locals and decreases, albeit very little, among immigrants. In quintile two, the case is more diverse; in half a dozen countries there is substitution with growth in immigrant employment and reduction in local employment, while in a majority of cases there is growth in both local and immigrant employment. The upper quintiles show in most cases the same pattern of increase in both groups with few excep-tions, notably Estonia, with reduction in immigrant employment in the three upper quintiles.

Table 6.5 Evolution of total employment of immigrants and locals

Opposite direction Same direction

Reduction in local employment and growth in immigrant employment

Growth in local employment and reduction in immigrant employment

Growth in local and immigrant employment

Reduction in local and immigrant employment

Q1 AT*, BE, DK, FI, FR, GR, LT**, LU, PT, SW, UK

DE ES, FR, IE, NL, CY EE

Q2 AT, BE, DK, PT, SE FR, EE, DE* CY, ES, FI, GR, UK, LT, IRL, NL

LU

Q3 NL, UK FR, EE Rest of countries DE Q4 – EE Rest of countries Q5 – EE Rest of countries LT

* Local employment stagnant, ** Immigrant employment stagnant

Note : Changes are referred to time periods considered in Table 6.1 .

Source : Authors’ analysis from Figure 6.11 .


Finally, we can speculate about the profile of employment creation in the different countries in the absence of immigration. By compar-ing the profile of the black part of the bars with the general profile of the full bars, a very simplistic answer could be given assuming that in absence of immigration the jobs held by immigrants would have not been created at all. Under such assumption, as in many cases the growth in employment in the lowest quintiles relies heavily on immigrant work (e.g. Spain, Cyprus, Austria or Belgium), most coun-tries would show a lower degree of polarization and a higher degree of upgrading. Obviously, the assumption that the absence of immigration would result in the elimination of the jobs held by immigrants is a gross simplification. Most probably, some such jobs would have been created anyway, while other jobs currently held by locals would have not been created at all – their existence being dependent on migration in one way or another. Furthermore, especially in countries such as Ireland, Greece, Cyprus or Spain, with large immigration inflows in the period considered, the lack of immigrants would have shown in a shortage of labour and the change of production methods in favour of more capital intensive technology (e.g. residential care for the elders instead of personnel hired to help at home). It is also possible that the entrance of immigrants, filling the jobs at the bottom of the rank, in a context of general economic growth, acts as a push factor, pushing up the labour ladder the existing local workers, contributing thus to the increase in employment in middle quintiles. Lastly, especially in countries with a less-developed welfare state, the lack of immigrants could increase the restriction faced by women in their work–life bal-ance (as migrants, especially women, often do domestic work or care for children, elderly or disabled individuals), reducing the labour mar-ket supply of locals.

6.5.2 Job matches and immigration

From a European perspective, over-qualification of workers is definitely one of the hot topics in both labour and education economics and prob-ably the most worrying side of the mismatch between labour supply and demand. Therefore, it is not surprising that, during the last years, it has become one of the biggest sources of concern among both academ-ics and policy-makers. 10

The measurement of over- and under-qualification is itself a much-debated issue as proved by the variety of perspectives adopted to analyse these phenomena (Hartog 2000). There are basically three possible approaches: worker self-assessment (WA), job analysis (JA),


and realized matches (RM). The WA perspective cannot be adopted since no information about workers’ opinions on skills requirement is available in the Job Projects database. 11 The JA approach consists in comparing systematic evaluations carried out by professional job analysis about the required level of skills for a certain job with the educational level of the individual holding the job. Since we do not have a detailed catalogue of the education requirements for each job in each country, the OECD (2007a) approach to measuring immi-grants’ over-education has been followed. It consists in re-codifying both occupational (using ISCO classification) and educational levels (using ISCED taxonomy) into three categories of skills and educa-tional attainments (low, intermediate and high), respectively. The cross-comparison of both categories logically defines the over- or under-qualification status of an individual. According to the RM, the required level of education for a job is derived from the mode of the distribution of educational attainment within the job. Therefore, this perspective is necessarily relative and, unless everybody holding a certain job has the same education, implies the existence of over- and under-educated workers in the economy whatever the average skills level. This should not be a major problem as long as the main interest is comparing how immigrants and natives fare in the labour market.

The characteristics of the Jobs Project database allow only implement-ing the JA and RM approaches. For reasons of brevity, only the main results of the approach based on the realized matches in the labour market are presented here. The results obtained using the JA approach offer different absolute figures for some countries, but the situation of immigrants in relation to locals, the main focus of the analysis, is very similar to the picture offered by the RM methodology.

In order to estimate the degree of over- or under-qualification of work-ers, jobs (as defined in the Jobs Project) are taken as the basic units of analysis, determining the representative educational category that cor-responds to each job. Those individuals whose schooling level is higher than the mode are labelled as over-qualified workers. Similarly, those employed individuals with an educational level below the mode are considered as under-educated workers.

The incidence of over-qualification on the basis of the RM criterion is depicted in Figure 6.14 . Unsurprisingly, the incidence of over- qualification is higher among immigrants than among nationals, with the exception of Germany. This is completely consistent with the distribution by job quintile and may be related to several phenomena:


AT BE CY DE DK EE ES FI FR GR IE IT LT LU LV NL PT SE SI UK



Total

Men

Women

Natives Migrants

010

2030

4050

010

2030

4050

1020

3040

500

Figure 6.14 Over-qualification by migrant status and sex, percentage of employed, 2006



1020

3040

010

2030

400

1020

3040




Total

Men

Women

Natives Migrants

Figure 6.15 Under-qualification by migrant status and sex, percentage of employed, 2006



limited transferability of skills acquired in other countries, discrimi-natory practices by employers – whatever the source of such behaviour (taste, statistical discrimination, market power etc.) –, lack of language proficiency or possibly lower reservation wages and job-search time. The second fact worth mentioning is the existence of substantial differ-ences by gender: in particular, immigrant female workers show higher levels of over-qualification than men, although there are exceptions, notably Greece. This difference by gender is specific for immigrants: over- qualification is higher among native men than among women in 13 countries, and in those countries where it is not so, the difference (with the exception of Cyprus and France) is not large. A third and final remark has to do with cross-country differences: Mediterranean and Baltic EU members and Ireland exhibit significantly high rates of over-qualification, pointing to the existence of a non-negligible mismatch between human capital resources and national production structures.

A glance at under-qualification reveals a roughly inverted image of the over-qualification rates ( Figure 6.15 ). In general, differences between foreign-born and local workers are much smaller than in the case of over-qualification. While in some countries – like Estonia, Ireland, Spain, Italy or Portugal – the incidence of under-qualification among foreign-born population is lower than among their local counterparts according to both criteria, in others (Belgium, Germany, Finland and France, among others) the opposite picture is observed. In this case dis-crepancies by gender are not particularly acute, with men displaying, in general, higher rates of under-qualification than women.

6.6 Summary and conclusions

Immigration is a very sensitive and multifaceted issue. Important dimensions include the impact on the sending country, the human experience, the economic and social impact on the receiving country including its impact on natives’ labour market outcomes, their effect on the financial sustainability of the welfare state, and so on. From this panoply of issues, this chapter has aimed at studying the labour mar-ket participation of immigrants, their similarities and differences with local workers, using the database and job ranking developed by the Jobs Project. Several conclusions arise from this analysis.

First, although migration to a large extent is the great absentee of the process of globalization, the last two decades have witnessed a change in the geography of migration, with an important increase in


the European immigration rate which has risen to almost 9 per cent. The stock of foreigners and the timing of immigration flows vary widely across EU Member States. While several of the once-typical immigra-tion countries (Belgium and France, for example) have been compar-atively untouched by the latest immigration wave, others considered until recently as countries of emigration, such as Spain, Ireland and Greece, have been subject to a very intense and sudden inflow of for-eign workers. Countries also differ in terms of the country of origin of immigrants. In some countries, such as Luxembourg or Belgium, most immigrants come from other EU countries; in others, such as Greece or Spain, most immigrants are from outside of the Union. With few exceptions, in most EU countries a majority of immigrants come from a relatively small number of sending countries.

Second, regarding the socio-economic and demographic characteris-tics of immigrants, they tend to be younger than nationals, especially in new immigration countries, and to be gender balanced on aggregate (but not always in every particular group of immigrants), and have, on average and with some notable national exceptions, lower human capi-tal levels and higher employment and unemployment rates.

Third, the most interesting contributions of the chapter have to do with the analysis of the labour market participation of foreign workers:

(a) As a rule, immigrants tend to concentrate in the lowest quintiles of the job distribution. As a consequence foreign-born workers tend to have jobs of lower quality than locals. Nevertheless, the analysis suggests the existence of non-negligible differences across countries, pointing to the existence of several different patterns of labour market participation of foreign-born workers. In some Member States (Austria, Cyprus, Germany, Spain, Cyprus, France, Greece and Italy) the differences are quite stark while in other coun-tries the immigrant population does not seem to face a severe dis-advantage compared to the native population (Finland, the United Kingdom and Belgium, for example). A look at the same data from a gender perspective reveals that women in general – and particularly female immigrants – are under-represented in the top two quintiles of jobs. In addition, immigrants show a much higher concentration in a relatively small number of jobs (defined by occupation and sec-tor). In this respect, Portugal, Cyprus and Spain represent extreme cases, with roughly half of foreign-born labour force working in only five types of job.


(b) Consistent with the diversity of the characteristics of immigra-tion in Europe, the Duncan Dissimilarity Index does not show a homogeneous pattern of segregation of immigrants and natives in the labour market of different EU Member States. Belgium, the United Kingdom, Netherlands, Sweden and France are the states where migrants and natives work in jobs that are more similar. At the other end of the scale, in Greece, Luxembourg, Cyprus, Latvia and Spain, the occupational segregation of migrants and locals is the highest. In general terms, differences by gender are not large, but there are remarkable differences in several countries, such as Cyprus (where women are much more segregated than men) and Sweden (where the female job structure is more homogenous than the male one). Although immigration is in some countries an important element of labour market segregation, in comparative terms, with the exception of Greece, other factors, principally gen-der, are more important in explaining segregation in the labour market.

(c) The dynamic of job creation and destruction for immigrants and locals is to a large extent country specific. Nevertheless, new migrant jobs tend to be mainly in the two lowest quintiles. Austria, Cyprus and Spain are the main exponents of this trend. Luxembourg, with an impressive growth of jobs performed by foreign-born individ-uals at the top quintile, represents a notable exception. In contrast, immigrants’ job creation is spread more or less evenly across the five quintiles in Ireland.

(d) Immigrant workers have worse job matches than nationals. In particular, the incidence of over-qualification is higher among migrants than among nationals, for both men and women. This might reflect a less–than-perfect transferability of immigrant’s skills and education, the lack of sufficient time to make a suitable match in the case of newly arrived immigrants or, alternatively, the existence of discrimination in the labour market. 12

Summing up, the analysis performed in this chapter is consistent with the existence of segmentation in terms of the type (quality) of jobs held by national and foreign-born workers, with immigrants over-represented at the bottom of the job-quality distribution. However, this general conclusion should make allowances for differences in seg-regation intensity among the different EU Member States. This high concentration of immigrants in a relatively reduced number of activ-ities and occupations has important implication in times of crisis, as


immigrants, as a group, may tend to have lower resilience to adverse economic circumstances.

Notes

1 . See, among many others, Borjas and Trejo (1991) and Brücker et al. (2002). 2 . Spanish law clearly illustrates this problem: while the standard procedure

takes ten years of continued residence in the country, Latin American migrants can get the Spanish nationality in two years.

3 . This sort of exercise is carried out, for example, by Borjas (2003a) for health insurance in the US.

4 . In the words of Hatton and Williamson (1998: 3) ‘contemporary numbers are relatively small compared with the mass migration of a century ago ... mass migration in the 40 years prior to World War I raised the New World labor force by a third and lowered the Old World labor force by an eighth’.

5 . Note that Figure 6.2 uses citizenship as the criterion to define immi-grant status, which results in lower immigration rates than the alterna-tive measure based on the country of birth, as it excludes from this group all those immigrants acquiring the nationality of the host country. In some countries, this discrepancy can be very relevant: for example, in the Netherlands, from 2001 to 2007, the number of foreigners that acquired Dutch nationality amounted to 5.6 per cent of the total population in 2008.

6 . The analysis of geographical patterns of migration and the factors behind them have received much attention from immigration scholars. For a detailed analysis, see Pedersen et al. (2008).

7 . For details see section 4 of the technical annex at http://web.usal.es/~janton/annex.pdf.

8 . This phenomenon is illustrated by Mullan (1989) and Patel and Vella (2007) for the United States.

9 . The impact of immigration on the labour markets of the host countries is one of the major fields of debates among academic economists. On the one hand, the works of some authors like Borjas (1994 and 2003b) support the existence of a negative effect of immigration on labour market outcomes of (at least) some segments of workers of the host countries. On the other, there is also a remarkable body of literature that defends the opposite view. This position is mainly based on evidence from ‘natural experiments’, like the massive arrival of Cubans to Florida after the Mariel boat lift (Card 1990), the increase in labour supply in France resulting from the return of the so called pied noirs to France with the independence of Algeria (Hunt 1992) or the migration of Russian Jews to Israel with the fall of the Berlin wall (Friedberg 2001). See Longhi et al. (2005 and 2006) for a meta-analysis of empirical findings and Friedberg and Hunt (1995), Borjas (1999) and Bodvarsson et al. (2009) for detailed literature reviews.

10 . See Hartog (2000) and Groot and Maassen van den Brink (2000) for an extensive review of this issue.

11 . Other possible databases, such as the European Social Survey , contain very few observations of foreign-born workers.


12 . In this last respect, according to a recent Eurobarometer 64 per cent of Europeans (EU25) considered the existence of discrimination based on ethnic origin (in many countries related to immigration) widespread. In Sweden, the Netherlands or France the percentage was equal or higher than 80 per cent (Special Eurobarometer 263 / Wave 65.4, Discrimination in the European Union, 2007)

147

7.1 Introduction

The main aim of this chapter is to enrich our analysis of changes of the employment structure in the EU during 1998–2007 by develop-ing a new broad-based job quality measure to view those changes. Specifically, we formulate a multidimensional measure of job amenities 1 based on non-pecuniary job dimensions which derives from the sociological literature on job quality. We then operational-ize this measure by matching survey items from the Fourth European Working Conditions survey (2005) data set to the job amenity dimen-sions identified in order to generate a job quality value at individual respondent level. This allows us to extend the jobs project analysis in new directions using a richer indicator of job quality measured at the individual as well as the job level.

In the original European jobs project analysis (Fernández-Macías and Hurley 2008; Stehrer and Ward 2008 ), two proxy measures of job qual-ity were used – median hourly job wage and average skill level of job holders – but priority was given in practice to the wage-based meas-ure. In part, this decision was based on a concern to be faithful to the methods used by previous research in the USA (Council of Economic Advisors 1996; Wright and Dwyer 2003), partly at least with a view to facilitating comparison between EU and US labour market develop-ments over recent periods. But it is reasonable to point out that the choice of a standardized wage measure as proxy of job quality is more

7 Assessing Recent Employment Shifts in Europe Using a Multidimensional Job Quality Indicator John Hurley , Enrique Fernández-Macías and Rafael Muñoz de Bustillo

148 John Hurley, Enrique Fernández Macias and Rafael Muñoz de Bustillo

readily accepted in the US than in Europe. Indeed as Vidovic (2007) argues, studies conducted in the US focused mainly on wages as a job quality measure (e.g. Levy and Murnane (1992), Council of Economic Advisors (1996), Ilg (1996), Farber (1997), Ilg and Haugen (2000) and particularly Wright and Dwyer (2003)). The principal reasons for this preference are twofold. First, wage income is more widely measured than many other individual job quality indicators. Other dimensions of job quality such as cognitive richness of work or work autonomy are less easy to model or measure and are addressed less often and less sys-tematically in surveys. Second, on the assumption that the wage of a job is likely to correlate with other unobserved, or less easy to observe, dimensions of work, earnings are a ‘sufficient salient aspect of job qual-ity’ (Wright and Dwyer 2003) to be used as a proxy even if the concept itself is multidimensional.

In the EU, policy initiatives in the area of job quality (e.g. the European Commission’s Laeken indicators) and related academic and policy-oriented research (e.g. European Commission (2008b); Davoine et al. (2008b); Smith et al. (2008); Dahl et al. (2009); Leschke and Watt (2008); Muñoz de Bustillo et al. (2011a)) tend to emphasize the com-plex and multidimensional nature of job quality and eschew more reductionist approaches based exclusively on wages or education level. Despite including 31 job quality indicators across ten job quality dimen-sions, the Laeken job quality indicators did not include a single wage or earnings based measure. 2 European approaches draw attention to many different facets of job quality identified in both the economic and sociological literature – income, autonomy, flexibility, cognitive richness, job satisfaction and so on. They also tend to combine vari-ables measured at both individual and aggregate level. Unfortunately, the bulk of the Laeken indicators are labour market indicators rather than job quality indicators as such and incorporate variables such as the long-term unemployment rate and the gender–pay gap, which weakens considerably this proposal and makes it almost useless for the purposes of measuring and comparing job quality across Europe (see Muñoz de Bustillo et al. 2011a; also European Commission 2008b).

In this chapter we develop a job amenities index based exclusively on indicators of job attributes that according to the specialized literature have an impact on worker well-being. In contrast to some of the qual-ity of work indicators already mentioned, the index will exclude con-textual information or aggregate labour market indicators. It will also focus on the attributes of jobs rather than information specific to the workers who work in those jobs. To construct the index, we use the data

Assessing Recent Employment Shifts 149

from the European Working Conditions Survey (EWCS), which is broad in scope, covers a broad range of quality of work issues and does so in a way that is comparable across countries at EU level. This survey is conducted every five years by Eurofound and is based on self-reported worker assessments on a hundred work-related question items (Parent-Thirion et al. 2007). We develop a multidimensional measure of job quality using nearly half of the Fourth European Working Conditions survey (henceforth, 4EWCS) items and use this to generate an EU-based ranking of jobs in terms of non-pecuniary job amenities.

It is important to stress from the outset that the main goal of this chapter is not to develop a multidimensional job quality index per se, but to use such an indicator to complement the analysis carried out in other chapters of this book, where job quality, for the above- mentioned reasons, is proxied by wage and the average educational level of job hold-ers. In fact, the job amenities index that we will use here is a simplified version of a recent proposal of a multidimensional job quality index for policy purposes in the EU: the reader interested in more details of this proposal should refer to Muñoz de Bustillo et al. (2011b). Using this job amenities index for analysing the structure and change of employment in Europe from the perspective taken in this book will help us to follow the evolution of job creation and destruction from a different but com-plementary perspective and will allow us to evaluate whether wage is an appropriate proxy variable of job quality.

7.2 Theoretical background: approaches to measuring job quality

Jobs have many quality attributes ranging from remuneration to the potential risk of accidents and the social and physical conditions of the places where the job is actually performed. This multidimension-ality involves a considerable difficulty of measurement, which helps to explain why we have consolidated indicators of prices, production or employment, for example, while lacking an equivalent generally accepted indicator of job quality. Besides this multidimensionality, job quality is a normative concept for which there is no agreed definition: people may and do have different ideas about what the most desirable attributes of a job are. The same job, with the same attributes, could be evaluated differently by different people. On top of these technical dif-ficulties, there are some political obstacles for the creation of an inter-national index of job quality: the most important of them is the almost obsessive focus of European employment policies (both at the national


and EU level) on job creation. These difficulties explain the lack of an agreed indicator of job quality. But the inherent difficulty of measuring job quality does not mean it cannot be measured.

In this section we will critically present two different approaches that can be taken to design an indicator of job quality. These approaches will be used in the following section to develop a job amenities index that can be operationalized using the 4EWCS (2005) data set.

The first approach to the measurement of job quality is based on the recognition of the complexity of the task: if (and this is an important if) we are not interested in measuring job quality per se but because we want to know the impact that different characteristics of jobs have on the well-being of the workers, then instead of focusing on the input (the characteristics of jobs) we can centre our efforts on measuring the output, that is the well-being of the worker at his or her job. Lacking a method to directly measure the well-being, we have to turn to an indi-rect measure of well-being: and the most frequently used measure in this context is job satisfaction.

This strategy of using satisfaction as an indicator of job quality has several advantages. First, it allows the researcher to turn a multidimen-sional concept into a single index which is much easier to manage and interpret. Second, this approach takes into account differences in taste in relation to what a good job is: instead of using a rigid framework of good and bad characteristics for every worker, in this approach it is the worker, himself or herself, who applies his/her own criteria about what is positive and negative of a job. For example, if someone likes working at night, or at weekends, then he or she will consider it as a positive characteristic of work, and it will contribute to his or her well-being and satisfaction with the job. Third, it avoids the need to measure and weight the different characteristics as each worker will do that in an intuitive way. Fourth, when answering a question on job satisfaction the worker will consider many attributes of his or her job, more char-acteristics that can ever be included in any multidimensional model of job quality. Last, but not least, the information can be gathered easily and at a low cost.

Probably for all these reasons, there is a growing literature on job satisfaction, its determinants and differences among groups of workers and countries: in practice, job satisfaction is often used as an indicator of job quality, even for the purposes of international comparisons. But even a quick comparison of the broad levels of job satisfaction across countries shows that this approach has some important problems. For instance, from the information on the different levels of job satisfaction


around the world produced by the International Social Survey Program (33 countries of different cultural background and level of develop-ment), three facts are worth mentioning. 3 First is the existence of a high level of job satisfaction everywhere, an average of 7.08 (in a scale of 1–10) for the 32 countries of the sample. 4 Second, the very small range of variation of the average level of job satisfaction in the sample of countries: a range going from 6.2 in South Korea to 8.1 in Mexico. Third is the ordering of countries: France is the second worst country in terms of job satisfaction, and Mexico the best, well above countries such as Denmark or Norway, for example. These results are at odds with most evidence on the levels of job quality across countries, and raises important questions concerning the use of this indicator as a proxy of job quality for international comparisons, notwithstanding its merit for other types of research.

Does job satisfaction correlate with other usual indicators of job qual-ity? There is an abundant literature, especially in Social Psychology, which attempts to relate job satisfaction with different qualitative aspects of the job, such as autonomy (Spector 1997: 31), stress (ibid.: 42), usefulness of work (Mangione and Quinn 1975) and so on. These stud-ies have usually found an important degree of correlation between the characteristics of the job and job satisfaction. Nevertheless, this type of study usually has a serious starting problem which casts doubt on the reliability of its results. Many of the qualitative indicators used to measure job characteristics are based on the worker’s own evaluation, and it is very likely that the evaluation made by the worker of his/her job may be affected by the degree of job satisfaction. The degree of cor-relation found between these indicators and job satisfaction does not necessarily have to involve a causal relation, but may be merely due to the fact that they are to some extent different ways of measuring the same thing. In order to avoid this problem it is convenient to use indi-cators of job quality not ‘contaminated’ by the worker’s subjectivity. When indicators that are less problematic in this sense are used, such as wage, sector, size of firm or job stability (Clark and Oswald 1996; Brown and McIntosh 1998; García Mainar 1999), the correlations are usually very low and the results not very conclusive. Indeed different studies often find contradictory relationships between certain indicators and job satisfaction. Spector (1997: 42–46), in an extensive review of the literature on the determinants of job satisfaction, found a pronounced inconsistency among the results of different studies on wages, work-load and organization of working time. A similar conclusion has been presented by Muñoz de Bustillo and Fernández-Macías (2005) from a


detailed analysis of the differences of job satisfaction among Spanish workers: ‘the variability of job satisfaction is surprisingly low and ... the little variability there is bears practically no relation to any relevant social or economic variable’ (p. 670).

There are different mechanisms that might explain both the rel-atively high level of job satisfaction and its limited variability com-pared with the high differences in job quality existing in all societies. The first possibility is that the existence of compensating differentials (Rosen 1986) by which jobs would have different combinations of good and bad characteristics producing a similar level of overall satisfaction. Although, as we will see further on, this could explain the existence of similar levels of satisfaction for similar workers in jobs with differ-ent characteristics, it cannot explain the similar level of satisfaction of workers with different endowments (human capital, skills etc.). The sec-ond, and probably more relevant explanation, is related to the existence of a process of adaptation of the expectation of the worker to his or her objective conditions. For years, social psychologists have been study-ing psychological mechanisms (the best-known theory in this respect is that of cognitive dissonance) that make people tend to adapt their expectations and even their perceptions of the environment to their actual conditions. 5 This behaviour could be interpreted as a strategy of emotional survival as it is very difficult to maintain a conception of the world (or in this case, of work) that is too discordant with reality. As a consequence, if one cannot change things, one adapts to them. The third mechanism is related to the existence of processes by which objec-tive conditions adapt to expectations: if a person has a job that does not fit his/her expectations and is not able to change such expectations and adapt them to the facts of the jobs then he/she will probably end up leaving the job. Finally, it could well be, especially in times of economic crisis and high unemployment, that for many workers in many places the fact of having a job is in itself a source of satisfaction, indepen-dently of how ‘satisfactory’ the job is in terms of quality.

Therefore, a high level of job satisfaction could simply be the result of a process of adaptation of the expectations of workers to the charac-teristics of their jobs and vice versa and as such cannot be interpreted in terms of the existence of good conditions of work and employment. Obviously, this interpretation does not mean that there is no informa-tion in the job satisfaction data, or that such information is useless. It suggests that the information of job satisfaction cannot be taken as a reliable indicator of the quality of the job performed (Muñoz de Bustillo and Fernández-Macías 2005).


An alternative method is to select different dimensions of job qual-ity using different theoretical perspectives and research findings in the Social Sciences about the impact of different jobs’ attributes on workers’ well-being. In this sense, sociology has more to say than economics, at least in its mainstream tradition. Mainstream economics has never paid much attention to the issue of job quality. The closest thing to a theory of job quality in this tradition is the theory of compensating differen-tials. The formulation of this theory comes from Adam Smith himself, in The Wealth of Nations (book 1, chapter 10 ). In fact, after 200 years of economic literature, Adam Smith’s formulation remains the canonical model of job quality in this tradition, as well as the canonical theory for explaining differences in wages for workers with similar skills and capacities (see Rosen 1986 or Cahuc and Zylberberg 2004: 248).

The theory of compensating differentials argues that the utility derived from a job (which we can equate to our concept of job qual-ity) depends on the combination of two separate but substitutive elem-ents: on the one hand, the disamenities (the characteristics of the job which negatively affect the well-being of the worker) associated with such job, and on the other, the monetary compensation (the wage) that the worker receives for doing such job. If the labour market is open and free, and agents have perfect information, these two elements (wages and disamenities) will tend to compensate each other, so that the over-all level of utility derived from each job will be roughly the same for workers with the same skills and capacity.

The idea of compensating differentials is probably behind the lack of attention paid to the issue of job quality in mainstream econom-ics. This is because a corollary of this theory is that for a given level of skills and abilities, and in the absence of external interventions, with full information and perfect competition, workers will have the com-bination of amenities and wage that they prefer. (Their utility will be maximized.) There is, therefore, no problem with job quality within a capitalist economy unless there are distortions resulting from external interventions or lack of competition: to increase job quality, therefore, we just have to ensure that markets operate freely and that agents have enough information. 6

A crucial element of this theory for our purposes is that the compen-sating mechanism only applies to workers with similar levels of skills and abilities . This is because in mainstream economics wage levels (again, assuming a competitive labour market) are determined by the marginal productivity of labour – and the same goes for amenities. Because ame-nities and wages are substitutive goods for employers, the bundle of


wages and amenities that each worker can get is determined by his or her productivity (therefore, the difference between the combination of wage and amenities in two different jobs will be proportional to the productivity difference between both types of workers). So the theory of compensating differentials does not say that all workers will get the same level of utility from their jobs (i.e. that all jobs will be of the same quality), but that all differences will result from the different produc-tivities of workers with different skills and abilities (which again, is the optimal outcome: a wage above marginal productivity of labour would be unsustainable, a wage below would be exploitative). All the differ-ences that exist beyond those based on productivity are only appar-ent, different combinations of wages and amenities chosen by workers themselves. Job quality is, again, simply not problematic from this perspective. 7

From the perspective of this study, the theory of compensating differ-entials is important because, if present in the daily working of the labour market, ceteris paribus, a higher wage could not always be interpreted in terms of a better job, as it could be accompanied by a high level of dis-amenities, leaving a lower than expected level of total utility.

In the sociology of work literature, job quality has always been a cen-tral issue. Because, on the one hand, this Social Sciences tradition does not assume (as mainstream economics does) that salaries will automat-ically compensate for any unpleasant working condition. On the other, because with or without compensation, researchers within this tradi-tion have tried to understand the implications of existing conditions of work on the well-being of workers. The fact that harsh working condi-tions are willingly accepted by workers in exchange for a high salary does not eliminate the negative implications of such conditions for the worker, which have to be studied and if necessary regulated. In a way, this second point derives from the first one, because once we reject the assumption of fair pay, once we accept that workers may be forced by whatever reasons to accept otherwise unacceptable conditions of work, those conditions by themselves become a central social issue.

The sociological concern with job quality, as with many other key sociological themes, goes back to the work of Karl Marx. In particular, two core concepts of Marxist theory have important implications for the sociological analysis of job quality: first, the concept of exploitation , which links an initial inequality in endowments between capitalists and workers to an intrinsically unfair compensation for work; second, the Marxist concept of alienation , a philosophical concept that involves an anthropological interpretation of the nature of labour. For Marx,


productive work is the essence of the human condition: it is through work that humans achieve realization and fulfilment. The capitalist labour relation, by separating workers from the means and fruits of their work, alienates them from their very nature and impedes their realiza-tion as human beings (Marx 1959). Exploitation implies that workers never receive a full compensation for their work, and alienation that the nature of work under capitalism is inherently dehumanizing. The first concept implies a focus on the conditions that maintain wages low: deskilling and power relations in the labour market and at the workplace. The second concept implies a focus on the intrinsic qualities of work and the social conditions in which they are carried out (auton-omy, skills, the social environment).

Drawing on these Marxist concepts (and updating them), Braverman developed his theory of the degradation of work in the 20th century (Braverman 1998, originally 1974), emphasizing the importance of skills and autonomy as the key elements of job quality: his ideas inspired a large body of research on the implications for job quality of the orga-nization of the labour process, an objectivist tradition of studying the intrinsic quality of work. On the other hand, Blauner (1964) proposed an empirical operationalization of Marx’s concept of alienation into four dimensions: powerlessness, meaninglessness, social alienation and self-estrangement. This analysis, although open to criticism (Edgell 2006: 36), initiated a subjectivist tradition of studying intrinsic job quality which has been applied to different countries and work environments (see, for example, Shepard 1977; Hull et al. 1982; Vallas 1988).

A different, but equally important strand of sociological analysis of job quality is segmentation theory. According to this theory the labour market is split into different segments which function according to dif-ferent rules and organizational principles (the competitive model being one of them, but not the only one). The original formulation of this theory by Doeringer and Piore, two institutional economists, argued that contemporary labour markets are dual in nature, composed by a primary and a secondary sector (Doeringer and Piore 1971). The pri-mary sector is associated with the economic activities which make use of the most advanced technologies and highly skilled labour: work in this sector is coordinated by internal labour markets, in which employ-ers and employees develop long-term relations (workers are protected from competition from outsiders), and wages and working conditions are above the standard. The secondary sector is associated with more backward economic activities which require relatively low- (or gen-eral-)skilled workers: this sector functions much more as the standard


model of a competitive labour market (with intensive competition between workers and salaries and conditions determined by supply and demand), employment relations are much more short term and unsta-ble, and working conditions and remunerations are considerably worse than in the primary sector. Overall, the segmentation approach implies that there is no automatic compensation between wages and working conditions (rather, they tend to correlate when different segments are compared) and that the determination of wages does not depend only on skills (although skills are a crucial segmenting driver in this theory); therefore, that job quality can vary substantially in non-optimal ways. The segmentation approach emphasizes the importance of the contrac-tual relation and the conditions of employment for job quality.

A third strand of sociological analysis (with key contributions from occupational medicine and epidemiology) which is also relevant for the study of job quality is the health and safety literature. This tradition is strongly empirical in character (there is no unified theoretical approach behind it), and it focuses on the analysis of the direct effect that the conditions of work have on workers’ health. Up to the 1970s, most research on this tradition focused on physical risks and hazards typi-cally associated with industrial work; after the 1970s, the focus shifted towards psychosocial risks and their outcomes, reflecting the increas-ing importance of service employment (see Wilkinson 2001).

A fourth strand of sociological research relevant for job quality, which is relatively new but has become quite important in recent years, is that of work–life balance. Although this issue has been in one way or another present in sociological research on work and employment from the very beginning, it is only in the last 25 years that this subject has become a really central issue. Behind this increasing centrality of work–life balance is the increasing incorporation after the 1970s of married women and women with children into the workforce: 8 this has raised to the surface a conflict between work and family demands which before had been precariously resolved by the gender division of labour inside and outside home – men worked for pay, women look after reproduc-tive duties at home. This approach is especially important from a gen-der perspective, because the persistence of unequal gender roles implies that women are much more likely to suffer from conflicting work and family demands. The aspects of work which have a larger impact on the work–life balance are those related to the duration and scheduling of working time (working time marks the boundary between work and life), working time flexibility and secondarily the intensity of the work effort (for a review, see Guest 2002).


The final strand of sociological analysis of job quality emphasizes the importance of worker representation and participation as a support for improvements in working and employment conditions. In this case, the impact on job quality is largely indirect and mediated. The assumption, supported in much research (see Broughton 2007 for overview of recent research in different EU Member States), is that strong collective inter-est representation tends, ceteribus paribus , to be associated with higher wages, greater job security as well as access to training and broader work–life balance possibilities and, overall, to enhance the previously mentioned dimensions of job quality. It is also associated with greater compression of the wage distribution and therefore with lower levels of wage inequality at the macro level.

This very brief overview of the main strands of sociological research on job quality has two main implications for this chapter. First, from a sociological perspective job quality is a multidimensional phenomenon and therefore requires a multidimensional approach for its study and measurement. Although there is no encompassing theory of job quality that attempts to link all its components and their interactions in a single model, we can derive a tentative model based on the elements consid-ered important by the different sociological traditions discussed in this section. The main elements of such a model are shown in Table 7.1 . We will use this list of attributes determining job quality, excluding wages for which we have a specific ranking (used in all the other contributions to this book), to construct an alternative multidimensional job ‘ameni-ties’ ranking that will be compared to the two main rankings (based on wages and education level) of the European Jobs Project data set.

Table 7.1 Dimensions of job quality suggested by the different traditions

The orthodox economic approach: compensating differentials

The traditional sociological approach: alienation and intrinsic quality of work

The institutional approach: segmentation and employment quality

Occupational medicine and health and safety literature: risks and impact of work on health

Work–life balance studies

Industrial democracy and participation

Labour compensation (1): wages

Objective strand : (2) skills (3) autonomy Subjective strand: (4) powerfulness (5) meaningfulness (6) social support (7) self-fulfilment

(8) Contractual status and stability of employment (9) Opportunities for skills development and career progression

Working conditions: (10) Physical risks (11) Psychosocial risks;

Working time: (12) Duration (13) Scheduling (14) Flexibility (15) Regularity (16) Clear boundaries Intensity: (17) Pace of work and workload

Voice: (18) Union membership (19) Collective bargaining coverage


The second implication is that despite job quality being much more than about just wages, the sociological tradition does not assume any automatic compensation mechanism linking wages and disamenities. In fact, many sociological theories assume exactly the opposite: mecha-nisms such as exploitation, discrimination or segmentation would tend to make intrinsic and extrinsic rewards correlate positively rather than negatively. Wages and amenities are often assumed to correlate, and therefore in contemporary labour markets there would be neatly dif-ferentiated good and bad jobs which accumulate negative or positive characteristics. Hence the paradox that although sociological theories have often argued against a purely instrumental perspective of work, there is nothing in these theories against using wages as a proxy of job quality. What these sociological theories would argue is that it is wrong to consider that remuneration is the one and only defining variable of job quality, but being a variable that correlates highly with many other elements of job quality, they would not argue against using it as an approximation to an overall job quality measure in the absence of better alternatives.

7.3 Data and construction of the job amenities index

The only existing European survey which covers the range of issues necessary to construct a multidimensional job amenities indicator as outlined in Table 7.1 is the EWCS: although this survey is wide and deep in its content, it is very shallow in its sampling, not allowing for the detailed level of disaggregation that would be adequate for our purposes. The core requirement of our approach is that the infor-mation can be broken down by sector and occupation, ideally at the two-digit level. We can do this with the EWCS, but only to a limited extent. Given a country sample size of 1,000 respondents in the major-ity of countries, we have insufficient observations to produce country-specific rankings of jobs by their amenities (the wage and educational rankings which are used in the other chapters of this book are country specific). Instead, in order to have cell sizes at two-digit level capable of generating reliable estimates, we are forced to aggregate all EU coun-tries together in one large sample and generate EU-wide rankings of the job amenities index. 9 Even combining the data for all 27 Member States, a further limitation is that for many jobs, the number of cases remains too small for the estimation to be reliable. In practice, we have settled on a minimum count of 15 respondents 10 as the basis of ranking an individual job.


So, adding a multidimensional measure of job amenities or job qual-ity based on the EWCS obliges us to carry out our analysis at the EU level rather than the country level. This is of course problematic to the extent that, despite the freedom of movement of labour and capital in the EU, there is not (yet) anything that can be called an EU labour mar-ket. The patterns of employment growth are naturally limited by the frontiers of the national labour markets: within each country (broadly speaking), a single labour regulation applies, there is (in most cases) a cultural and linguistic unity that allows real mobility of labour, and there are shared and relatively circumscribed economic conditions (notwithstanding the ongoing processes of globalization and European integration).

Still, nothing prevents us from taking the EU labour market as a kind of heuristic tool, to construct a what if type of analysis. After all, even if there is still not an EU labour market strictly speaking, there is cer-tainly the will to construct it: the EU project is largely about building a huge market of capital, goods and labour. And there is already a set of shared broad principles of labour regulation, increasingly harmonized. This makes the holistic analysis of EU employment based on common European rankings, at least, a potentially interesting exercise.

As a further empirical justification we can cite a high level of correla-tion between the national jobs rankings by wages and especially by skill level (Fernández-Macías, E. 2010 11 ). In practice, this means that a single cross-European job ranking is likely to represent reasonably accurately the relative positions of jobs in individual labour markets.

One benefit of carrying out the analysis at EU level 12 is also that the aggregate data (e.g. on changes in employment level by quintile) will tend to smooth out distinctive national patterns of employment growth which are very country specific in the time period we are examining. In this way, they may provide a more robust indication of properly struc-tural labour market changes as opposed to cyclical effects unique to individual countries. 13 A further advantage is that with an EU-level job ranking, we can benchmark (of course with many qualifications) indi-vidual Member States against a common ranking.

Comparing the jobs ranking we will construct in this chapter with the rankings used in the rest of this book, we can say that the wage and educational rankings are wider in their coverage of jobs, and they are country specific; but the EWCS-derived amenities index is wider in its coverage of job quality, provides data at individual as well as aggre-gate level and can be further decomposed by (sub)dimensions and (sub)indicators.


7.3.1 The construction of the job amenities index

The job amenities index was constructed based on the principles dis-cussed in the theoretical section of this chapter, but omitting two dimensions – those of pay and industrial democracy/participation for reasons outlined below. The idea of (dis)amenities derives from Adam Smith’s theory of compensating differentials. The specific set of job- related amenities is drawn from the various strands of sociological anal-ysis of job quality described in Table 7.1 . The structure of this model and the way we operationalized it into a job amenities index based on the EWCS is summarized in Figure 7.1 .

We have made a deliberate choice to leave out pay/remuneration from our operational model of job amenities. This is because our specific purpose in this chapter is to elaborate an alternative job quality measure to those already created for the jobs project based on wage and skill-level data. The inclusion of a pay dimension in our multidimensional index would blur the boundaries between the different rankings by

Job amenities index

(43 items – Cronbach’sAlpha = 0.80

1. Intrinsic quality of work (25%) (14 items) a) Objective (12.5%)

I. Skills (7.5%; q23c q23d q23e q23f) II. Autonomy (7.5%; q24a q24b q24c q17a q23b)

b) Subjective (12.5%) I. Powerfulness (3.125% = autonomy)

II. Meaningfulness (3.125%; q25k) III. Social support (3.125%; q25a q37f) IV. Self-fulfillment (3.125%; q25h q25i)

2. Employment quality (25%) (4 items)a) Contractual stability (12.5%; q3b q37a) b) Development opportunities (12.5%; q28a_1 q37c)

3. Health and safety (25%) (17 items) a) Physical risks (12.5%; q10a-q10e q11a-q11j) b) Psychosocial risks (12.5%; q29a-q29d)

4. Work–life balance (25%) (8 items) a) Working time (12.5%)

I. Duration (4.13%; q8a) II. Scheduling (4.13%; q14a q14b q14c)

III. Flexibility (4.13%; q17a) b) Intensity (12.5%; q20b_aq20b_bq25f)

Figure 7.1 Job amenities index


including shared components and would complicate correlations and other comparisons based on them.

We also omit the dimension of industrial democracy from our operational model as the survey items in the EWCS address issues of worker representation indirectly rather than directly. There is no direct question, for example, on whether the respondent is a union mem-ber or has employment conditions covered by collectively bargained agreements. As such the EWCS does not allow us to generate a reli-able measure for such a dimension. In any case, it can be argued that participation/industrial democracy – important as it has been for the development and maintenance of work and employment rights – is a determinant rather than a component of job quality and that its impact should be implicit in the other four dimensions of our model. So the actual construction of the amenities index draws only from the four main sociological traditions studying the impact of the conditions of work and employment on the well-being of workers.

In order to allow aggregation, each variable was rescaled to a 0–1 scale based on ‘0’ being the most negative outcome for job amenity and ‘1’ the most positive or desirable. To give a simple example, a dichotomous variable such as Q28a_1, ‘Over the past 12 months, have you undergone any of the following types of training to improve your skills or not? – A Training paid for or provided by your employer ... ’, was coded as ‘1’ if the respondent said yes and ‘0’ if the respondent said no. This is the most appropriate method for the normalizing of such diverse variables (the EWCS includes Likert-scale, ordinal, categorical, continuous as well as dichotomous category variables) as the 43 items used for the con-struction of the overall job amenities index. 14

The general principle behind the weighting of items was one of equal weighting within each level and component . This means that each of the four main dimensions of the job quality model – intrinsic job quality, employment quality, health and safety and work–life balance – account for 25 per cent of the overall index. Of course, the assignment of items to different levels within the overall structure means that in fact the weights assigned to each variable are by no means equal: they depend on the position of each variable within the structure of the model, whether it is the main basis for an indicator or one of several variables to be aggregated into a single sub-indicator. Therefore, the weighting of each variable is directly determined by its place in the nested structure of the amenities ranking. 15

The aggregation process follows step by step and from bottom to top the structure of Figure 7.1 : from variables to sub-indicators, from


sub-indicators to indicators and so on, until the final aggregation from dimensions to the overall composite index. This way of constructing the index has the advantage of making the final result easily decomposable: each dimension can be easily broken down into its sub-dimensions in a very transparent manner.

Once the job amenities index was computed at the individual level of the EWCS (i.e. for each respondent), we proceeded to calculate sum-mary measures for the various indicators and their sub-dimensions at the basic level of observation of the jobs project, the job (ISCO-2 digit x NACE-2 digit cells). This in turn allowed us to merge the EWCS sum-mary data with the original jobs project data set thereby linking the job amenities rankings/indices to the evolution of employment by jobs (the jobs matrix, drawn from the ELFS 1998–2007) as well as the original wage and educational rankings derived from other Eurostat sources. We used the full EWCS-27 data set 16 including employees as well as self-employed- 17 and without differentiating by country.

7.4 Analysis

7.4.1 Descriptive outputs

Having constructed the job amenities index, we would like to test how job amenities vary by occupation, sector and country according to our model using the EWCS data and how this compares with findings from external sources. This is a useful preliminary test of plausibility. Figures 7.2 and 7.3 show the results for the job amenities index and its four sub-dimensions for all one-digit sectors and one-digit occupations as well as gender and legal status of employer (private or public sector). The reason for re-aggregating back to one-digit sector and occupational detail is purely pragmatic; it would be impossible to summarize our data if we use our basic level of observation at two digits.

The sectors and occupations have been ranked according to their value in the overall index (shown in the first column on the left): the absolute value of the index ranges from 0.74 (financial intermediation) to 0.56 (hotels and restaurants) for sector, and from 0.71 (professionals) to 0.55 (elementary occupations) for occupation. The actual range of variation of the absolute scores of the index across sectors and occu-pations is relatively small (around 15% of the total possible variation, which would go from 0 to 1).

The absolute values of the index are largely secondary to us (and can always be contested, being based on ultimately arbitrary decisions with respect to the coding of the original variables). The important thing


here is the relative position of each sector and occupation in terms of the overall ranking. And the rankings of sectors and occupations accord-ing to their multidimensional job quality seem reasonable according to our previous knowledge of the distribution of working conditions (see Parent-Thirion et al. 2007). For instance, the ranking of ISCO codes matches quite closely the implicit occupational hierarchy of this classi-fication, with a few minor variations: the top of the job amenities index is occupied by professionals rather than managers. The extremes of the index are also plausible at sectoral level with financial services at the top and hotels/restaurants and agriculture/fishing at the bottom.

Public sector workers report higher levels of job amenity than private-sector workers notably in relation to employment quality (contractual stability and training opportunities) and work–life balance. The former finding relates primarily to greater levels of job security amongst public employees compared with their private counterparts. The public sector is also often considered a forerunner in relation to work–life balance policies. 18 Female workers report somewhat higher levels of overall job amenity than males. Greater work–life balance, lower levels of work intensity and lower work-related health risks more than compensate

0.2

0.4

0.6

0.8

1.0

Job

am

enit

ies

ind

ex

overall amenities index dim1 work q dim2 emp q dim3 h&s dim4 wlbH

OR

EC

AA

gric

ultu

reC

onst

ruct

ion

Man

ufac

ture

Tran

spor

t/com

ms

Ret

ail

Hea

lthO

ther

Ser

vice

s

Edu

catio

nF

inan

ceP

lant

ope

rato

rsE

lem

enta

ry o

ccs

Cra

ft w

orke

rsS

kille

d ag

ricS

ervi

ce w

orke

rsC

lerk

sTe

chni

cian

sS

enio

r m

anag

ers

Pro

fess

iona

ls

Mal

eFe

mal

e

Priv

ate

Pub

lic

Util

ities

Rea

l est

ate

Pub

lic a

dmin

istr

atio

n

Figures 7.2 Job amenities index by sector, occupation, gender and legal status of employer broken down by four sub-dimensions

Source : 4EWCS, 2005 (authors’ calculations).

0.4

0.5

0.6

0.7

0.8

0.9

EL

RO

CZ

ES

HU

BG

CY

SI

LT

PT

PL

SK

DE

EE

MT

LV

FI

AT

IT

FR

IE

UK

BE

LU

SE

NL

DK

Job amenities index

amen

ities

dim

1 w

ork

qdi

m3

h&s

dim

4 w

lbdi

m2

emp

q

Figu

re 7

.3

Job

amen

itie

s in

dex

by

cou

ntr

y br

oken

dow

n b

y fo

ur

sub

-dim

ensi

ons

Sour

ce :

4EW

CS,

20

05 (

auth

ors’

cal

cula

tion

s).


for lower intrinsic quality of work and employment quality though the overall gender gap is marginal in this case and would be reversed if a pay component had been included in our ranking.

At country level, the results are consistent with previous research on cross-national comparisons of quality of work. Job quality tends to be higher in the older Member States with higher GDP per head and lower in the newer Member States with lower GDP per head. For example, Denmark, which has the highest job amenities index value, has an enviable propensity to come at or near the top of many cross-national comparisons of social amenity including of work or employ-ment quality.

The contribution of the four dimensions to the overall amenities index value varies between the countries in suggestive ways. For instance, the countries placed first and third on the overall index – Denmark and Sweden – differ from that of the second place, the Netherlands, primar-ily in having lower scores for work–life balance and higher scores for employment quality. The high Dutch scores for work–life balance reflect in part the prevalence of part-time work, uniquely high in this Member State, but part-time work is also associated with less secure employ-ment status as well as lower access to career-development possibilities. This suggests some element of a trade-off between the dimensions of employment quality and work–life balance and this is also supported by comparatively low levels of correlation between these two dimensions as we will see later. Of the larger Member States, the high position of the UK is perhaps more surprising but other attempts to measure job qual-ity cross-nationally have generated similar results (Leschke and Watt 2008) for the UK.

Table 7.2 sets out the country rankings based on European Trade Union Institute’s (ETUI’s) overall job quality index (see Leschke and Watt 2008) and the job amenities index based on the 4EWCS and also includes a basic sensitivity analysis of the country amenities rank-ing by applying different weighting formulae for the sub-dimensions. We set out the range of rankings for each country under the differ-ent weighting methods in column three. They are highly consistent in most cases.

7.4.2 The ‘job’ as a predictor of the level of job amenities

One advantage of our job amenities index is that it gives us an individ-ual-level measure that can also be aggregated at job level. This allows us to do something that the original wage and skills measures in the jobs project (all based on pre-aggregated data) did not: that is, to assess the


extent to which job amenities vary within and between jobs. Ideally, in justification of the entire jobs-based approach, we would expect a significant measure of the overall individual variance in job amenities to be accounted for by the job. That said, we would also expect that a large measure of the variation of job amenities is likely to relate to per-sonal, individual circumstances and preferences which our model will not capture, as well as differences related to career effects within the

Table 7.2 Country rankings using the job amenities ranking and the ETUI’s Job Quality Index

Country JQI ranking

(ETUI)

Job amenities ranking (based

on 4EWCS)

Amenities ranking range using different weighting methods

DK 1 1 1–2 NL 2 2 1–5 SE 4 3 2–4 LU 6 4 2–4 BE 7 5 4–6 UK 3 6 6–9 IE 8 7 5–9 FR 11 8 8–10 IT 15 9 6–11 AT 9 10 7–11 FI 5 11 7–13 LV 19 12 10–13 MT 10 13 9–14 EE 16 14 13–15 DE 14 15 12–15 SK 23 16 15–21 PL 27 17 16–18 PT 18 18 17–19 LT 20 19 17–22 SI 12 20 17–23 CY 13 21 18–24 BG 24 22 19–25 HU 22 23 22–25 ES 21 24 21–25 CZ 17 25 19–25 RO 26 26 26 GR 25 27 27

Note : Spearman’s rho correlation index = 0.84 for main rankings in column one and two; Column 3 ranking ranges based on five alternative weighting schemes: equal weighting for all 43 individual items (x1) and then double weighting of each main dimension in turn so that the dimensions contrib-ute in the ratio 40:20:20:20 to the final ranking (x4).

Sources : EWCS, ETUI (Leschke and Watt 2008).


same job (even within the same job, conditions often change with ten-ure, normally for the better).

So, to test our hypothesis – that the job (i.e. the combination of a specific occupation in a specific sector) is a significant determinant of job quality or job amenities – we carried out an Analysis of Variances (ANOVAs) of our job amenities index comparing different group-ing variables such as gender, age or country with the job. In the first instance, this was done bivariately. The question we are trying to ask, beyond confirmation of statistically significant effects, is how much of the overall variation in the job amenities index can be explained by differences in the average values across jobs, and how much remains within jobs?

In Table 7.3 , we can see the results of an ANOVA contrasting the total variance of the job amenities index (at the individual level) with the variance that exists between the averages by job (jobs holding less than 15 workers have been eliminated from the sample, as explained earlier). The job (defined as ISCO occupation in NACE sector at two-digit level) accounts for just over a quarter of the variation in our index (adjusted r-sq = 0.26). For comparison, the adjusted r-sq for other potentially important background variables are much lower though, indepen-dently, occupation (ISCO) has an adjusted r-sq of 0.19 and NACE sector has an adjusted r-sq of 0.12.

What does this tell us? Clearly, the job as a construct accounts for significantly more variation in job amenities than country, gender, establishment size or seniority. The majority of this variation (though not all) is accounted for by the occupational component rather than by the sectoral component. This is consistent with Figure 7.2 where the

Table 7.3 Bivariate ANOVAS with job amenities value as dependent variable

Dependent variable Independent variable

Independent variable categories Adjusted r-sq

Job amenities index (43 items)

ISCO * NACE (‘job’) 565 0.26 Country 27 0.04 Gender 2 0.01 ISCO 2d 28 0.19 NACE 2d 59 0.13 Education (isced) 7 0.10 Establishment size 5 0.01 Seniority 5 0.03

Note : All independent variables significant at the 0.01 level.

Source : 4EWCS (n = 23,259, author’s calculations).


occupational gradient of job amenities was steeper than the sectoral gradient. This in turn is consistent with the fact that the ISCO occu-pational classification has an implicit hierarchical dimension (from white-collar managers to blue-collar operatives) whereas the NACE sector classification is horizontal and non-hierarchical. Indeed, the fact that sector accounts for as high a degree of the variation of job ameni-ties as it does (more, for example, than the average education level of the job holders) may be considered surprising. It does, however, pro-vide strong support to the jobs approach whose basic assumption is that the horizontal (i.e. sectoral) and vertical (i.e. occupational) division of labour provides a relevant framework for evaluating job quality.

Although the adjusted r-sq value of 0.26 is not exceptionally high, in our view it is more than high enough to justify the use of the jobs approach in this case. Taking into account the amount of statistical noise that exists in any cross-national survey such as the EWCS, the problems derived from having such a small sample, and also the high degree of subjectivity involved in some of the variables used for con-structing the job amenities index, an adjusted r-sq of 0.26 is far from disappointing. On the contrary, it empirically proves that the position of the worker in the jobs matrix has important implications for the overall level of amenities or quality of his or her job.

In our final merged data set, we have three main proxies of job quality at the job level: education- and wage-based rankings from the origi-nal jobs project data set and a job amenities ranking derived from the EWCS. In addition, we have indicators for the main sub-dimensions of the job amenities index. All of the rankings have been computed at the aggregate EU level. In this section, we look at the correlations of the different job rankings.

7.4.3 The relationship between job skill level, wage and amenities

Behind the construction of the three main job quality proxies there is the assumption that each measures the same thing (job quality) albeit from different angles. In this perspective, job quality is a common, unobserved factor; important aspects of which each of our three proxy measures seek to capture. Measures of correlation are an obvious way to test this assumption and we would expect, in general, high levels of positive correlation. On the other hand, given that each of the three main measures is qualitatively distinct and that they capture differ-ent elements of job quality, we would not necessarily expect a very high level of correlation (r > 0.9). Additionally, the fact that the three


rankings are constructed using totally different data sources and with different underlying rationales makes the correlation test quite power-ful as there is no reason to suspect any spurious correlations.

The three main rankings in the correlation matrices are highlighted. Each has a pairwise correlation of over 0.70. The strong positive asso-ciation suggests that the three main indicators are indeed approximat-ing some underlying common property. The fact that the correlations are no higher than 0.7–0.85 is consistent with the basic assumption of our approach which is that each of the three main indicators (skill/ education, wage and job amenities) are measuring the same general property, job quality, but each from a different perspective.

The first observation that we can make from Table 7.4 are that the correlations are positive for all job rankings as well as for the job ame-nity sub-dimensions. There is no evidence of compensation at the job level between, for example, wages and job amenities. In fact, jobs that are high ranked on wages or skills tend also to be high ranked on job amenities.

The correlation between the wage and job amenities indicator can be considered a job-based test of the theory of compensating differentials. Clearly, at job level, such compensation does not exist as there is a high level of positive correlation between the two rankings. There is no evi-dence of a mechanism between jobs whereby higher wages compensate for lower overall job amenity though of course our approach ignores within-job variations.

Table 7.4 Pairwise correlations of job rankings by skill, wage, job amenity and its four dimensions

Job amenities sub-dimensions

Skill Wage Amenities Work q emp q h & s wlb

Skill * Wage 0.87 * Amenities 0.82 0.71 * Work q 0.74 0.70 0.79 * emp q 0.84 0.81 0.82 0.67 * h & s 0.57 0.42 0.85 0.45 0.53 * wlb 0.36 0.23 0.65 0.37 0.29 0.57 *

Notes : Skill and wage are the original jobs project rankings based on education (ELFS) and mean hourly wage (SES, ECHP-SILC, SBS). Amenities is the ranking using the job amenities index based on the 4EWCS. Correlations are weighted by employment totals by job, EU23 2000.

Source : European Jobs project data set, 4EWCS.


A second observation is that education level has a higher correlation with our job amenities indicator than wage. Why should wage have a lower correlation than skill with a more broad-based measure of job quality? One possible explanation could be that higher-educated work-ers put more emphasis on having nicer than better-paid jobs. Another is that the educational level of jobs may be in itself a proxy for an important dimension of job quality, the intrinsic work dimension – which measures the extent to which work is cognitively challenging, meaningful and healthy (or at least largely risk free).

At the level of the job amenities dimensions, there are also uni-formly positive associations though those for the health and safety and work-life balance dimensions tend to be weaker than for employ-ment and work quality. Correlation between the work–life balance and both the wage ranking and the employment quality ranking is, for example, relatively weak (though still positive). This is in part at least related to the diversity of jobs with long working hours and relatively high levels of working time flexibility or autonomy. Senior managers and administrators as well as agricultural labourers and some categories of hotel/ restaurant workers tend to be similar in these respects but are less similar in most other work dimensions. In general, many jobs with high amenity in relation to intrinsic work dimensions (autonomous, flexible, cognitively rich/stimulating) as well as employment quality (stable and well paid) may also have very long working hours. And the opposite is also true especially in more segmented labour markets: shorter-hours work tends to be of poorer quality not only in relation to employment conditions but also in relation to intrinsic work quality.

In general, the weak positive associations in the correlations matrix for some of the components of the index may buttress the case for some partial forms of compensation mechanism where positive job quality characteristics such as higher wage are traded-off against negative char-acteristics such as atypical or irregular working schedules (shift, night or weekend work premia, for example) or higher physical risk exposures (hazard pay). But even if this supposition is correct, the impact is lim-ited and serves only to weaken the positive association between wages and the specific amenities sub-dimension at job level but not to change its sign from positive to negative (as the theory of compensating dif-ferentials would imply).

It is important to note that these generally high correlations are based on summaries of the various EWCS-based indicators at job level (i.e. having eliminated through averaging the sizeable within-job variation


already noted in the ANOVA). They tend as such to over-represent the level of correlation between the main job quality proxies and dimen-sions. Nonetheless, the high level of correlation of the rankings – each covering different aspects of job quality using different data sources – reassures us that our approach is generating a valid qualitative discrim-ination between jobs.

7.4.4 Employment distribution by country using EU job amenity quintiles

The three EU-level job rankings allow us to compare not just employ-ment change from a qualitative perspective in a given period but also to do a cross-national comparison of employment structures. To do so, we use our job quintile assignments based on the amenities rankings and simply decompose employment in a given year (2005) by country and quintile. This amounts to comparing each country to an implied, even EU distribution of employment across quintiles. In Figure 7.4 , we see the resulting employment distribution for each Member State.

The same countries (generally speaking older, non-Mediterranean Member States) that demonstrate higher levels of amenity at the individ-ual level ( Figure 7.4 ) also tend, perhaps unsurprisingly, to have employ-ment skewed towards the top quintiles. High-ranking countries on both measures include two Scandinavian countries, the Benelux countries and the UK. Each of these countries also features among the top-ranked countries in the ETUI’s job quality index (Leschke and Watt 2008). The corollary also holds true in particular with respect to the new Member States as well as the Southern Member States where employment is skewed towards lower-amenity jobs.

The countries thus tend to cluster in ways that are consistent with previous efforts to make a typology of employment regimes (Muffels et al. 2002) or job quality regimes (Davoine et al. 2008a) in Europe, most of which are adaptations of the most commonly known typology of social welfare regimes (Esping-Andersen 1990). The Benelux countries rank high in terms of employment structure, perhaps unexpectedly ahead of the Scandinavian Member States but along with liberal UK/Ireland cluster and the other corporatist countries – DE, AT, FR, IT 19 – they constitute a high job quality Northern European axis reflecting the findings of Davoine et al. (2008a).

Differences in country rankings between the mean job amenities rank-ing ( Figure 7.3 ) and the employment distribution rankings ( Figure 7.4 ) are suggestive of the extent to which job amenity differences at country level result from labour market composition on the one hand or specific


national differences in the level of job amenity for a given job, on the other. This is an important question given EU policy commitments to improving job quality. The fact that Greece, for example, has the lowest rank for job amenities of all EU27 despite having a middling position in terms of employment distribution by quintiles is indicative of disen-chantment with conditions of work and employment in this country whose explanation is unlikely to be structural. There may be cultural biases in a self-reported survey with a number of subjective questions – the propensity of Greece, for example, to finish comparatively low in cross-national surveys of work or life well-being is as marked as that of

0% 20% 40% 60% 80% 100%

ROBGESPTSKLTPLCYCZEELVSI

GRHUMTITIEATFIFRDEDKUKBENLSELU

Q1

Q2

Q3

Q4

Q5

Figure 7.4 Country employment distribution by job amenities quintiles

Notes : Countries sorted by employment weighted by job amenities quintile. Totals have been grossed up to 100 per cent as no job quintile assignments were made in the case of 10 per cent of EU27 employment for reasons of small sample and cell sizes.

Source : ELFS (2005 annual data), 4EWCS.


Denmark to finish high (Anderson et al. 2009). Nonetheless, the rela-tively low position of countries such as Greece and Germany in average job amenities argues the need for greater policy attention to job quality in these countries.

7.4.5 Net employment growth, 1998–2007, in the EU by quintile using three job quality measures

Finally we assess net employment growth in 1998–2007 using our job amenities indicator to assign jobs to quintiles following the original jobs project approach. What we find is that the pattern of employment growth is quite similar regardless of which measure we use to rank jobs – wage, skills or amenities (see Figure 7.5 ). There is a common pattern of upgrading with the top quintile experiencing most growth in the EU(23) and the second quintile experiencing the second highest growth.

02,

000

4,00

06,

000

Em

ploy

men

t cha

nge

by q

uint

ile

8,00

010

,000

AmenitiesSkill

Wage

Figure 7.5 Net employment growth by quintile based on different job quality indicators

Note : The job amenities bars are based on net employment shifts in those 565 jobs for which we were able to generate a ranking from the EWCS. These 565 jobs cover approximately 90% of total employment.

Source : Jobs Project data set, EWCS.


The wage ranking produces a pattern of asymmetrical job polariza-tion with some significant growth in the bottom quintile outweighed by larger growth in the top two quintiles. 20 This pattern is not dissimi-lar to that recorded for the US labour market in 1992–2000, also using a wage-based job ranking (Wright and Dwyer 2003; see also Chapter 4 in this volume). To a certain limited extent each of the three rankings demonstrates the hollowing out of the employment structure identi-fied in the earlier analysis of US labour markets as well as more recent analysis of European labour markets (Goos et al. 2009). This is char-acterized by comparatively small growth in medium-ranking jobs (in low-medium quintile two jobs for wage and amenities and in medium quintile three jobs for skill levels) with larger growth at the extremes. But while there is some limited polarization of employment growth, it is very asymmetrical with greater and more consistent growth at the top than at the bottom of the job distribution, according to each of our three job rankings.

Using the education ranking, we see a more pronounced picture of employment upgrading with only very limited growth in the bot-tom quintile. Using skill/education level as our measure of job quality therefore generates the most positive picture of European employment growth. The job amenities ranking produces hollowing out in the sec-ond (medium-low) job quintile but again limited growth in the bottom quintile and most growth in the top quintile (though comparatively less than the skill- or wage-based rankings).

There is clearly less variation between the different rankings at the top of the employment structure. Each ranking conveys a very similar message of upgrading skewed to the top quintile. What differences exist relate primarily to the bottom three quintiles; this reflects findings at the country level where variation between countries is greatest in the lower quintiles while employment expansion tends to be concentrated in the top quintiles in the majority of countries (see Chapter 3 in this volume). The main conclusion is that using either the skill or the job amenities rankings does not seriously alter the conclusions drawn from the original jobs project analysis using the wage ranking. Each ranking generates a picture of overall upgrading of the European employment structure with to varying extents some hollowing out in the middle. The outcomes are consistent with the skill-biased technological change (SBTC) hypothesis which predicts greater relative demand in higher-skilled jobs and lower relative demand in lower-skilled jobs. At the same time, the existence of some hollowing out of the employment is par-tially consistent with refinements of the SBTC hypothesis which hold


that routinized jobs most prone to capital replacement by automation (and/or geographical displacement via offshoring) tend to be middle ranking (see, for example, Goos and Manning 2007) leading to greater relative employment loss in the middle than the bottom of the wage distribution.

7.5 Conclusions

Job quality is inherently an abstract, multidimensional and complex concept and as such presents great challenges in terms of both def-inition and measurement. One common approach to date has been to bypass this complexity by using wage, skill-level or job-satisfaction measures as proxies of job quality. While such approaches are pragmatic and take advantage of the comparatively ready availability of skill, job satisfaction and wage data, they necessarily tend to deliver one-dimensional indicators of job quality. Are low-paid jobs or low-skilled jobs necessarily bad jobs? Not in some, possibly many, cases but they would be characterized as such if we restrict ourselves only to wage- or skill-based measures. In this chapter, we have developed a more broad-ranging measure of job quality using a representative European data source. We have done so both as a means of testing the validity and applicability of other single indicator proxies of job quality and also as a worthwhile exercise in its own right – linking different data sources with complementary strengths at the job level – and one with suggest-ive descriptive outcomes.

Our first finding is that even though there is substantial variation in the multidimensional job amenities indicator from individual to individual, even where individuals perform the same job, nonetheless the job itself is the most influential determinant of the level of job amenities among a list of plausible independent variables including age, tenure, country, educational attainment and gender. It accounts for over a quarter of the variation in job amenity at the individual level. This is important as it offers empirical justification of the jobs approach which is based on the assumption that jobs (again, under-stood as a given occupation in a given sector) represent reasonably well-demarcated structural units of observation of the labour market categorizing of workers with broadly similar workplace functions and responsibilities.

A second finding of the analysis is that at the job level the correlation between wage, skill/education level and job amenities is positive and high. Jobs that rank high or low on any one measure tend also to rank


correspondingly high or low on the others. While (superficially at least) this outcome goes counter to the theory of compensating differentials – which predicts wages should compensate for disamenities – it does serve to justify previous approaches to job quality predicated on using job wage or skill level as proxies. Assuming that our broad based job ame-nities measure offers a valid approximation of job quality, then so too will job wage and skill levels.

Our approach does, however, imply disregarding within-job trade-offs and, as observed, there is a large individual and within-job var-iation of job amenities. It is also fair to add that the conditions for a rigorous testing of compensating differentials are unlikely to be met by cross-sectional survey estimates at aggregated level using different survey sources.

That said, it does appear that more contemporary, sociological approaches to job quality – notably those describing labour market seg-mentation and the importance of internal labour markets – are more consistent with the findings of recent European survey data on the dis-tribution of job quality than the canonical economic approach which dates back in its original form to before the French Revolution. Good jobs tend to be well paid, highly skilled (or at least to employ predomi-nantly well-educated individuals) and to have higher levels of job ame-nity measured using other non-pecuniary elements of job quality. And the opposite applies for bad jobs.

This has important policy implications. If job amenities and disamen-ities cluster together respectively in good and bad jobs, this may be both manifestation and cause of greater inequality in the labour market and is likely to have persistent effects. Already most accounts of structural labour market change identify growing polarizing forces in the supply and demand for labour as well as increasing levels of wage inequality (markedly so in the case of the top half of the wage distribution). The absence of compensating mechanisms at the job level may mean that both more- and less-measurable dimensions of job quality are tending increasingly to partition between good jobs and the rest. This widening qualitative gap is mirrored in growing inequalities of wage- and career-development potential. That this should happen at a time when the stagnating middle of the labour market limits possibilities for upwards transitions is a cause for concern. A healthy labour market requires horizons of individual improvement and mobility and these may be becoming more limited.

The fact that our three approaches to ranking jobs in terms of job quality correlate well means that they also contribute to telling a


consistent story about structural change in the EU labour market in during 1998–2007. On the positive side, the ranks of good and very good job holders have tended to grow faster than those of poor- and especially of medium-quality jobs. The analysis in this chapter offers a generally positive representation of structural shift from lower- to higher-value employment in line with known trends of tertiariza-tion, technological/IT advances and increasing knowledge-intensity of work.

Notes

1 . Job amenities in this chapter refer to the positive elements of the work experience, or alternatively the absence of negative elements (disamen-ities), cited in the economics and sociological literature as being conducive to worker well-being. We use the original formulation from The Wealth of Nations (Smith, 1904) though the customary term ‘utility’ in present-day economics has a similar meaning.

2 . Davoine et al. (2008b) suggest the comitology behind this curious omis-sion. ‘ ... As a result of the position adopted by the United Kingdom and the Scandinavian countries, the definition does not include wage level as a component of work quality, whereas other countries (e.g. France) were in favour of taking this indicator into consideration.’

3 . These results are drawn from authors calculations based on 2005 ISSP data. For more details, see Muñoz de Bustillo et al. (2011b).

4 . The level of satisfaction in the 1997 wave of the ISSP for the countries in both samples was also very high, in fact higher than in 2005 (7.43 v. 7.11).

5 . See Festinger (1957). 6 . In fact, from this perspective, any government attempting to regulate work-

ing conditions (e.g. regulating health and safety) will either reduce wages below the optimum level or generate unemployment, thus decreasing over-all utility.

7 . From an empirical perspective, and despite its continued theoretical importance, the support for the argument of compensating differentials is extremely thin (Purse 2004), except for some very specific sectors and occupations.

8 . Unmarried women were part of the workforce in advanced capitalist econo-mies well before the 1970s.

9 . Based on data from all EU27 countries. 10 . With a small refinement to raise cell-counts by combining similar jobs

where counts are < 15, the ranking ends up taking into account over 90 per cent of the EWCS sample and includes jobs representing over 90 per cent of the working population in the EU.

11 . Only the Baltic countries as a cluster show lower levels of correlation in the job/skill rankings of jobs (see Fernández-Macías, 2007).

12 . In parallel, and to be consistent, we have also constructed EU-level rankings based on the wage and skill rankings for each country derived from various


Eurostat sources (SES, ECHP/SILC, SBS) and used in the other chapters of this book. These were created by weighting each ranking by the number of people in that particular job in that particular country. This takes into account not only the different sizes of national labour markets, but also the different distributions of employment by occupation and sector.

13 . One example is the very divergent trends in the level of construction sector employment in two large Member States, Germany and Spain, in the period in question. The post-unification construction sector boom in Germany had ended by 1998 and the sector was a net loser of jobs in the following decade. During the same period, a construction boom in Spain almost dou-bled employment in this sector which rose by 1.3m. The collapse of this boom took place precisely at the end of the period covered by this study and therefore is not covered.

14 . Full details of normalization of variables for the construction of the job amenities index are available upon request to the authors.

15 . The inclusion of the Q24a–d variables under two separate headings is designed in part to reflect the fact that the individual questions are the most appropriate proxies for the two indicated sub-dimensions, ‘autonomy’ and ‘powerfulness’, and also to give an adequate weight to what are in any case unambiguously important variables.

16 . To generate the EWCS-derived job amenities measure at the job level, the individual data have been weighted using the post-stratification factor included in the EWCS data set (w4). We avoided using the EU27 weighting variable as this would have assigned undue weight to the larger Member States and reduced the quality of the estimations especially for jobs with comparatively few observations. The estimations are not strictly represen-tative of the European workforce as a whole but instead represent an aver-age of country-representative estimations of job quality. This is in any case coherent with the fact that we are in principle ignoring country effects, assuming that what determines job quality is primarily the position in the division of labour along the horizontal and vertical axes (occupation and sector).

17 . In the case of the job amenities index, this implies a small logistical prob-lem: the dimension of employment quality is half missing for the self-em-ployed, for obvious reasons (they have no labour contract). Our solution was simply to ignore that indicator for the self-employed, aggregating the rest of the indicators into the dimensions and overall index. In fact, whenever some item was missing for any individual respondent, the item was disre-garded in the aggregation process.

18 . Though in practice this perception may be based on the experience of certain Member States such as Belgium, Finland, Germany, Ireland, Italy, Latvia and Sweden, see Riedmann et al. (2006: 11).

19 . Italy is acknowledged to be a difficult country to assign to a specific regime type. Muffels et al. (2002) assign it to the southern grouping while Leschke and Watt (2008) generate two separate typologies assigning Italy to the cor-poratist group in one and to the southern group in the other comment-ing that ‘ ... a Salamonic verdict, that also takes into account the significant north-south dichotomy within Italy, would probably conclude that the


country exhibits a mixture of “southern” and “continental” features, and that this is also reflected in our job quality indicator.’ Our two indicators in this chapter locate Italy within the corporatist rather than the southern group.

20 . At EU aggregate level. This pattern is far from pervasive at Member State level as we have seen in Chapter 3 .

180

The eight post-socialist accession (A8) countries went through a sys-temic change in the early 1990s, with profound effects on their labour markets. 1 The evolution of wages and employment in the A8 are likely to differ from the EU15 for at least three reasons: the compressed wage structure of the pre-transition era, the transitional shock and the policy response to the shock. Questioning the possible interpretation of the transition to a market economy as simply a case of exceptionally rapid technological and organizational change, we argue that other factors played an important role, and that this has implications for the evalu-ation of the jobs distribution as well. In this chapter we focus mostly on labour market policies, showing that these may explain some of the difference in jobs distribution between the A8 and the EU15, and also within the A8.

Wage dispersion was bound to rise during the systemic transition for several reasons. The loss of former trade partners and price liberal-ization accelerated industrial restructuring, but there were some other forces at play. The egalitarian wage grid of central planning was dis-mantled and unions were weak in most accession countries, which led to relatively flexible wage-setting practices. This allowed the previously distorted wage distribution to adjust relatively fast to productivity levels and changes in demand. The few studies documenting wage develop-ments in the early years of transition all find large shifts in relative wages between 1989 and 1994 (Rutkowski 1996; Kertesi and Köllő 2000; Newell 2001). The relative importance of the factors behind is difficult to trace due to the complexity of the transitional process and the lack of reliable micro-data. However, several studies found that returns to schooling increased fast and accounted for a large part of rising wage dispersion in most transition economies. 2

8 Job Quality in Post-Socialist Accession Countries Ágota Scharle

Job Quality in Post-Socialist Accession Countries 181

This chapter focuses on changes in the jobs distribution induced by sectoral shifts in employment. During the transition such shifts may be caused by industrial restructuring and also by recovery from the tran-sitional shock. By definition these shifts do not reflect wage dispersion within occupations induced by rising returns to schooling or improved productivity. It is important to note, therefore, that the focus is not on explaining wage dispersion during the transition, but on understand-ing why the wage distribution of accession countries may have evolved differently from old EU Member States.

Post-socialist countries in Europe all opened their markets to the out-side world and dismantled (or lost) their Soviet trade relations during the systemic transition in the early 1990s. By 1995, post-socialist Eastern Europe, though poorer, was roughly at the same stage of industrial development as Southern Europe. The share of agriculture in employ-ment ranged between 7 per cent in the Czech Republic and 23 per cent in Poland compared to 7 per cent in Italy and 20 per cent in Greece. Services were somewhat lagging behind, ranging between 45 per cent in Poland and 59 per cent in Hungary compared to their 56–60 per cent share in Southern Europe. Employment shifted from agriculture and industry to services in all post-socialist countries and followed roughly the same path as in Greece during the past 15–20 years.

Further, the impact of the information revolution was either relatively modest (in Latvia and Lithuania), or was accompanied by increased access to markets in advanced European economies. Foreign direct investment attracted by relatively cheap skilled labour also helped to preserve or even create new jobs in routine industrial production (Radošević et al. 2003).

Based on the above, one would expect the eight accession countries to exhibit some signs of polarization but generally follow a pattern of employment expansion similar to Greece, Portugal or Spain, where a modest upgrading of job quality was observed between 1995 and 2007 (see Chapter 3 ). However, we find that to apply only to the Czech Republic, while the other A8 countries either exhibit considerable polar-ization (the Visegrad group) or a mixed pattern (the Baltic States).

This chapter reviews some of the factors that may help in explaining why the jobs distribution would vary within the A8 and why it should differ from Southern Europe in the past 15 years. In particular, we seek an explanation for two questions: (1) the large losses at the bottom combined with gains in the middle of the jobs distribution in Latvia and Lithuania; (2) the polarization of the jobs distribution in the other A8, except for the Czech Republic. The first two sections give a brief

182 Ágota Scharle

account of the nature and size of the transitional shock and its evo-lution in time. The next three sections describe the policies that may have affected the jobs distribution: the initial choice of unemploy-ment benefits and job subsidies in response to the transitional shock, minimum wage levels, income taxation and measures to encourage self-employment. The sixth section describes how differences in the above factors may have contributed to the varied evolution of the jobs structure in accession countries. The last section offers some tentative conclusions.

8.1 The transitional recession

Political changes in Central and East European countries were followed by dramatic changes in their economies over the 1990s. Output fell by 15–25 per cent and there were large shifts in the ownership structure, in the sectoral composition of GDP and in firm-size distribution. Much of the discussion of the transition process has centred on the causes of the dramatic decline in output and on the depth of economic and insti-tutional adjustment. The early explanations focused on the loss of trade relations (demand) or the disruption in price structures and hardening budget constraints (supply), or the combination of the two. More recent analyses have examined the role of government policies in anticipation of the recession, showing how the timing of privatization and the mix of social and employment policies may lead to very different labour market outcomes.

In the neo-classical interpretation of Bofinger (1994), firms switched from maximizing output (and over-employing labour) rather than prof-its, which led to a gradual shedding of labour and an immediate reduction of output. 3 In a Keynesian approach, Bhaduri and Laski (1994) claim that the main cause was the squeeze on demand through tight monetary and fiscal policy and a restraint on wages. Blanchard et al. (1994) conclude that macro-stabilization, the collapse of CMEA 4 trade and mismeasure-ment of GDP were the main causes of output decline, suggesting that demand contraction rather than supply disruptions was the ultimate cause. Later studies tend to put more emphasis on the impact of insti-tutional disruption. Jackman (1995) points to the end of planning as a mechanism for resource allocation while the institutional arrangements (exchange, distribution, finance) for a market economy are not yet fully in place. Blanchard et al. (1997) suggest disruptions in supply caused by shifts in relative costs and relative demand played an increasingly


important role after the initial demand shock. 5 In a similar vein, Kornai (1993) enlists five causes: the shift from a sellers’ to a buyers’ market, the transformation of the real structure of the economy (induced partly by the removal of price subsidies), the disturbances in the co-ordination mechanisms, the macro consequences of the tightening of financial dis-cipline, and the backwardness of the financial system. Gomulka (1998) even claims that the only effect of macroeconomic policy was to hasten or delay the fall in output and the start of recovery.

The fall in output was smaller, and recovery started earlier in the Visegrad countries and in Slovenia. In the gloomiest three years between 1990 and 1993, the cumulative fall in real GDP amounted to 18 per cent in Hungary and in Slovenia, that is slightly more than in the Czech Republic (15%) and in Poland (16%). Other countries in the region suf-fered larger declines of 22–25 per cent. 6

In the Baltic States the Russian crisis in 1998 caused a second shock wave that disrupted trade relations and with it, the recovery of their economies. Lithuania and Latvia were hit harder than Estonia as their initial share of export to Russia was relatively higher. 7 The nature of the disruption was rather similar to the initial demand shock of 1989–1990 following the breakup of the soviet block and the dismantling of ‘Comecon’ trade agreements. Employment declined in the sectors that were most affected by the Russian crisis: especially in fishing, agricul-ture, manufacturing and construction. The economy adjusted relatively fast, partly with the help of increased FDI inflows: trade was redirected towards the West and productivity levels increased. As a result, how-ever, demand for low-productivity blue-collar workers sharply declined (Varblane et al. 2003).

The transition entailed a drastic fall of up to 30 per cent points in the level of employment in all accession countries, though the decline was much larger in the three countries that chose fast privatization (Estonia, Hungary and Latvia). Socialist economies had achieved close to full employment across all levels of education and, as Figure 8.1 shows, this has not been regained since. There was a corresponding rise in wage and income inequalities, with considerable variation across countries, as documented by, for example, Flemming and Micklewright (2000) or Forster et al. (2005).

The employment rates declined deeper among uneducated workers and have been persistently and significantly below the EU15 average in all the post-socialist accession countries ( Figure 8.2 ). The socialist labour market was also characterized by a relatively compressed wage

184 Ágota Scharle

distribution, though with considerable variation within the social-ist bloc (Atkinson and Micklewright 1992). Compressed wages, full employment, an extensive system of price controls and subsidies, and in kind provisions, ensured a low level of income inequalities.

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 200770

75

80

85

90

95

100

PolandLithuaniaCzech RSlovakiaSloveniaLatviaEstoniaHungary

Figure 8.1 Level of employment in accession countries, 1989 = 100 Sources : for 1989–1995: ILO (2011); for 1996–2007: Eurostat on-line database (LFS employment, population aged 15–64.

45 50 55 60 65 70 75 80 8530

35

40

45

50

55

60

65

70

75

80

EU15

Be

Bg Cz

Dk

Dk

Ee

Ie

Gr

Sp Fr

It

Cy

Lv

Lt

Lu

Hu

Mt

Nl

At

Pl

P

Ro Si

Sk

Fi

S

UK

No

Md

Tr

Total employment in 2009, %

Em

ploy

men

t of b

elow

upp

er s

econ

dary

, %

Figure 8.2 Employment rate for the total working age population and for the uneducated, 2009 Source : Eurostat on-line database. Age 25–64.


Ironically, the present misery of uneducated workers has been aggra-vated by their good fortunes in the past. In socialist economies the demand for low-skilled workers was much higher than in most devel-oped market economies. Köllő (2006) examines the distribution of jobs by level of reading and writing skills required and shows that in the mid-1990s the share of undemanding jobs was still 2–3 times higher in post-socialist economies compared to Western economies. This had two important consequences: first, the socialist education system had not been challenged to change and focus more on skills. Second, many workers spent much of their working careers in jobs that made little use of their competencies, which eroded even the poor skills they had had when leaving school. As a result, the typical post-socialist economy entered into the economic transition with a relatively large proportion of low-skilled workers (larger than what the educational composition of the workforce would suggest) and a traditional educational system that continued to produce workers poorly equipped for working with new technologies.

The composition of the new jobs created in the newly emerging market economies was however close to that of Western economies in terms of skills requirements, or even exceeded it, where foreign invest-ment entailed green-field investments and the introduction of new technologies (Köllő 2006). Moreover, while market institutions were established relatively fast, many features of the socialist economy and society have been slow to change. Most importantly, the lack of genu-ine, independent entrepreneurs and the disrupted tradition of family businesses could not be regenerated overnight. In market economies it is typically the Small and Medium Enterprises (SME) sector that absorbs most unskilled workers, but in Eastern Europe it could not fill this role as it was too small and too slow to develop even after the transition. Maloney (2004) finds that in the mid-1990s, the Central and Eastern Europe (CEE) self-employment rate was less than half of what it should have been, given the level of labour productivity in the formal sector ( Figure 8.3 ). 8

In this context the loss of low-paid jobs – though it may be a posi-tive development in a mature market economy – is not unquestionably favourable in a post-socialist economy as it implies a continuation of the dramatic decline in employment opportunities for unskilled workers during the transitional recession. This also implies that an expansion of employment in low-wage jobs may be driven by the much needed adjustment of labour markets rather than shifts in demand induced by technological change or globalization.

186 Ágota Scharle

8.2 Duration of the transitional period

There have been various attempts to measure the degree of restructuring and to identify the end of the transitional period, when market struc-tures have converged to mature economies. Jackman (1995) proposes industrial composition as the most important dimension (and meas-ure) of restructuring. On that measure, the transition was continued well into the late 1990s. Comparing data for 1989 and 1994, Jackman and Pauna (1997) find that the sectoral allocation of labour in Central and Eastern European countries was not sufficiently close to that in the European Union: only 35–60 per cent of the required reallocation across sectors had taken place.

Svejnar (2001) argues that the transition process may also be pro-longed by the weakness of governments. A mature market economy requires not only the dismantling of planning and intrusive subsi-dies – which most post-socialist economies quickly accomplished – but also the establishment of a legal framework and efficient public administration that provides a level playing field for the market economy. The latter requires an ability to collect taxes and minimize

.461S

hare

of l

abou

r fo

rce

in s

elf-

empl

oym

ent

.0576

6.63332 10.5322

Log of industrial V.A./worker

Peru

Bol

EISCol Ven

ArgGre

Kor

Por

Bra

Spa

New

Ire

Ita

Ice

SwitJap

USAust Den

Can Fra GerNeth

NorLux

FinSwe

BelUK

Aus

Hun

CzePot

UruParGuaHon

PanCR

ChiMex

Turk

Figure 8.3 Self-employment and industrial productivity in the mid-1990s Note : The Czech Republic, Hungary and Poland are the outliers in the bottom left quarter.

Source : Maloney (2004).


corruption and rent-seeking behaviour, which weak governments may lack.

Assuming that the evolution of employment is a suitable indicator of restructuring, the turning point in the recession ranged between 1997 and 2003, but typically happened in 2000. As the time span of the prin-cipal data set used in this volume covers the years between 1996/1998 and 2007, the window we are looking through will capture only the last few years of the transition period. For Hungary, we only get the last year of the recession, for Estonia, Lithuania and Slovenia we get the last three years. To illustrate the importance of this ‘window’ effect, Figure 8.4 plots total employment against the share of the bottom quin-tile of jobs (using the quintile thresholds in the 2000). To show the depth of the recession, employment is expressed in proportion to its initial pre-transition level in 1989. It is important to note that for most of the countries and most of the recession years, the share of low-wage jobs was decreasing and only Estonia and Slovakia show some reversal of this trend during the recovery.

8.3 Policy response to the transition

Balla et al. (2008) argue that ill-designed employment policies and more specifically the lack of incentives to facilitate the re-employment of low-productivity workers was a further important factor contribut-ing to the persistence of low employment and large wage inequalities. Extending the model of Aghion and Blanchard (1994), they assume two segments of the emerging private sector that differ in workers’ produc-tivity. Governments may alleviate the social impact of the initial shock by slowing down the privatization process, providing benefits to the unemployed and/or by subsidizing the employment of low- productivity workers. Simulation results show that lower benefits induce higher aggregate employment and inequalities throughout the redeployment process, while higher subsidies are conducive to lower inequalities and higher aggregate employment. 9 The marginal effect of the employment subsidy is largest on employment and income when job destruction is fast and benefits are high.

Policy response to the transitional shock varied considerably across the region. Table 8.1 summarizes the two main dimensions of policy choice based on Balla et al. (2008), which sorts the accession countries into four groups. In this taxonomy, only Estonia and Latvia appear to have pursued an optimal strategy of rapid privatization and high share of wage subsidies (instead of unemployment benefits), which would

188 Ágota Scharle

19981999

2000

2001

2002

2003 2004

2005

2006

2007

1997

19981999

2000

2001 2002

200320042005

2006200720

212223

Sha

re o

f Q1

jobs

, %

70 75 80

Employment, 1989 =100

Estonia

1998

19992000

20012002 2003

20042005

2006

20071012

1416

1820

Sha

re o

f Q1

jobs

, %

82 84 86 88 90 92


Lithuania14

1618

20

Sha

re o

f Q1

jobs

, %

70 75 80 85


Latvia

19971998

199920002001

20022003

20042005

2006200716

1718

1920

21

Sha

re o

f Q1

jobs

, %

87 88 89 90 91 92


Czech R.

1996

19971998 1999

200020012002

200320042005 2006

20071618

2022

24

Sha

re o

f Q1

jobs

, %

70 72 74 76 78


Hungary

19981999

2000

2001

2002 20032004

2005 20062007

19.520.020.521.021.522.0

Sha

re o

f Q1

jobs

, %

80 82 84 86 88 90


Slovakia

19961997

19981999

20002001 2002

20032004

2005

20062007

1819

2021

22

Sha

re o

f Q1

jobs

, %

78 80 82 84 86


Slovenia

Figure 8.4 Share of bottom quintile jobs in total employment by level of employment, 1997–2007

Source : Calculations of Enrique Fernández-Macías and Ágota Scharle based on the EJM database.


minimize the cost of the transition in terms of national income and wage inequalities. 10

The model thus predicts the largest increase in job dispersion in Hungary, and considerable increase in Poland and Slovenia, which would come mostly from a rise at the top of the wage distribution, that is a growth of employment in good jobs. 11 In all other A8 countries wage inequalities are expected to increase less, but wages at the bot-tom end are more likely to drop (since benefit levels are relatively low).Wage cuts at the bottom end may shift some jobs to the bottom quintile and may also boost labour demand: both factors imply an increase of employment in bad jobs.

8.4 Wage costs: minimum wages and taxation

Minimum wage policies, and to some extent the flat tax reforms (in the Baltics and in Slovakia), have also impacted on employment and the jobs distribution.

All the eight accession countries have a statutory minimum wage and seven of the A8 (with Estonia as the only exception) executed a steep rise in it at some point between 1989 and 2010. The increases have been typically justified by ensuring decent wages, boosting labour supply and reducing poverty, or by the need to approximate local wages to Western levels on the eve of EU accession. A less-often cited but perhaps more important motivation has been to increase government revenues. 12

Table 8.1 Policy mix and expected labour-market outcome

Policy mix Country Inequality in wages

Low-speed dismantling of public sector

Low benefits High ALMP

Czech R., Slovakia Lithuania

Small

High benefits Low ALMP

PolandSlovenia

Medium

High-speed dismantling of public sector

Low benefits High ALMP

Estonia Latvia

Small

High benefits Low ALMP

Hungary Large

Notes : Voucher type privatization where shares were distributed to the population had little effect on firms’ budget constraints and is therefore not considered as privatization (even when it was used early on). The grouping by level of active labour market policy spending (ALMP) is based on OECD.stat Public expenditure and participant stocks on LMP. The expected change in wage inequality is based on the model simulations of Balla et al. (2006).

190 Ágota Scharle

Minimum wage levels in the A8 have converged around 35–45 per cent of the average wage during the past decade, which is comfortably within the range observed in the EU15 (cf lowest at 36.5 in Spain and highest at 46.9 in France in 2007). However, minimum wage legis-lation has been more volatile in the A8 than in old Member States. Between 1997 and 2007, the Czech Republic, Hungary and Slovakia experienced unusually fast and steep increases in the minimum wage (see Figure 8.5 ; in other A8 similar increases happened earlier or later in time.)

If large enough, a minimum wage rise would typically eliminate some jobs in the bottom quintile. We expect to find such an effect in the Czech Republic, Hungary and Slovakia and, to some extent, in Lithuania, where the minimum wage was almost doubled in 1997. To a lesser extent, relatively high minimum wage levels may have acceler-ated the destruction of low-wage jobs in Slovenia as well.

Flat tax reforms were implemented in Estonia, Lithuania (1994), Latvia (1997) and Slovakia (2004). Such reforms are likely to affect the wage distribution if the flat rate is low and replaces a highly progressive personal income tax system. This applies especially to Slovakia, where high incomes had been taxed at 38 per cent before the 19 per cent flat rate was introduced in 2004. As a result of the reform, above average wages are likely to increase slower, as employers gradually appropriate some of the tax gain of employees; that is, the tax cut will first increase net wages but this gain is gradually reduced as employers adjust gross wages. The rise in the net wage may increase labour supply and the fall in the (relative) gross wage may increase labour demand, but both effects may induce some expansion in employment at the top end of the wage distribution.

8.5 Self-employment

As already mentioned above, the accession countries have a below opti-mum self-employment rate, which have hindered the recovery of low-skilled employment. Policies to improve the business environment, especially if specifically tailored to small businesses, may have low-ered the inherited constraints (lack of capital and experience) to SME growth. The self-employment rate (in proportion to total employment) tended to be relatively high and increasing in the Czech Republic, and markedly increased from a relatively low level in Slovakia. Importantly, in these countries the self-employed typically work in industry or ser-vices rather than in agriculture. That implies that Czech and Slovak

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

20253035404550

Slo

veni

aS

lova

kia

Hun

gary

Lith

uani

aC

zech

Rep

ublic

Latv

iaP

olan

dE

ston

ia

Figu

re 8

.5

Lev

el o

f th

e st

atu

tory

min

imu

m w

age

in a

cces

sion

cou

ntr

ies,

per

cen

tage

of

aver

age

wag

e

Sour

ces :

Eu

rost

at o

n-l

ine

dat

abas

e (e

arn

_mw

_avg

r1)

for

20

01–2

007

, Fu

nk

and

Les

ch (

20

05)

for

1998

–20

01,

for

1997

: E

riks

son

an

d P

ytl

ikov

a 2

00

4, W

orl

d B

ank

20

02, S

kled

ar 2

00

4.

192 Ágota Scharle

entrepreneurs have had some potential to create jobs for low-skilled workers while this was less likely in the case of the petty farmers of Latvia, Lithuania and Slovenia, who constitute a large share of the self-employed in these countries ( Table 8.2 ).

8.6 Patterns of jobs distribution in the accession countries

Let us now turn to the explanation why the jobs distribution would evolve so differently within the accession countries and as compared to the EU15 during the past 15 years.

Latvia and Lithuania experienced huge losses at the bottom combined with gains in the middle of the jobs distribution. In the mid-1990s, both countries had a large primary sector (around 20% of employment) and a relatively small services sector (around 50–55% of employment), very similar to the sectoral composition of the Greek economy. By 2007, agricultural employment was halved in all three countries, and the share of services grew by around 14–17 per cent. Changes in the jobs distribution were, however, markedly different: while Greece (as other Southern European economies) showed a mild upgrading of jobs, Latvia and Lithuania gained jobs mostly in the middle of the jobs distribution ( Figure 8.6 ).

The loss of jobs in the bottom quintile (the bottom two in the case of Latvia) seems relatively easy to account for: the transition shock elimi-nated many bad jobs and since the shock was elongated by the Russian crisis, the data spanning from 1998 to 2007 capture a relatively large

Table 8.2 Level and change of self-employment in the accession countries

Change during recovery** Level in 2000 Level in 2007

Czech Republic 1.08 14.51 15.62

Estonia 1.11 7.95 8.80 Lithuania 0.72 16.67 12.02* Latvia 0.86 10.77 9.26* Hungary 0.73 14.71 12.05 Slovenia 0.88 11.20 11.22* Slovakia 1.65 7.79 12.86

Notes : *At least one-fourth are small holders in agriculture; **Recovery is understood to begin when total employment begins to rise.

Source : Calculations of Enrique Fernández-Macías and Ágota Scharle based on the EJM database.


part of the recession. However, a more important question would be to understand why there was no job creation at the bottom end until 2007, despite the rise in low-skilled unemployment during the transi-tional recession and the relatively favourable benefits policy. A possible answer (waiting to be tested) may lie in high minimum wages and the lack of a non-agricultural SME sector.

Most of the job growth in the middle of the distribution came from construction and less knowledge intensive private services. There was also some growth in knowledge intensive services, mostly in the top two quintiles, but it was too small to impact on the overall pattern of the jobs distribution (Figure 8.7).

Lithuania (1998–2007)

Latvia (1998–2007)

Education

Wage

100

0–1

00

Cha

nge

in th

e nu

mbe

r of

wor

kers

per

qui

ntile

(00

0s)

100

0–1

00

Figure 8.6 Employment expansion in accession countries: a mixed pattern

Source : Calculations of Enrique Fernandez Macias based on the EJM database.

194 Ágota Scharle

The other A8 countries (except for the Czech Republic) experienced a polarization of the jobs distribution (Figure 8.8). Around 1997–1998, the sectoral distribution of employment in Estonia, Hungary, Slovakia and the Czech Republic was similar to Spain’s: agriculture employed around 8–9 per cent, and services around 52–59 per cent of the workforce (the corresponding figures are 9 and 61% for Spain in 1995, the base year in the EJM data). In ten years, the share of agriculture dropped to 4–5 per cent, exactly as in Spain, and that of services grew to 52–63 per cent (cf

100Lithuania (1998–2007)

Latvia (1998–2007)

KIS

LKIS

HTI

LTI

Construction

Primary

1 2 3 4 5

1 2 3 4 5

50

0

–50

–100

–150

150

100

50

0

–50

–100

Figure 8.7 Absolute change in employment by sector and wage quintile in Latvia and Lithuania

Notes : HTI and LTI stand for high and low technology industry (defined according to the technological intensity of the productive process, following an OECD proposal – see Hatzichronoglou 1997), KIS and LKIS stand for knowledge intensive and less knowledge intensive private services.

Source : Calculations of Enrique Fernández-Macías based on the EJM database.


65% in Spain). In contrast to Spain, however, the gradual shift in the structural composition of employment was preceded by a large drop in the employment rate. Between 1989 and 1998 total employment fell by around 30 per cent in Estonia and Hungary and by 20 per cent in Slovakia and Slovenia. Most of the decline was in low-paid jobs of uned-ucated workers. In these countries a large pool of unemployed (actively looking for work, or discouraged by the prolonged recession) were avail-able to be re-employed and most of them were uneducated workers who could only expect to be hired at low wages. In this context, the rise in low-wage employment (especially if combined with growth in total employment) would be a signal of successful labour market readjust-ment rather than an unfavourable side effect of sectoral reallocation of labour .

–200

020

0Czech Republic (1998–2007)

050

100

Slovenia (1996–2007)–5

00

100

50

Slovakia (1998–2007)

–20

2040

0

Estonia (1997–2007)

–100

010

020

0

Hungary (1997–2007)

Cha

nge

in th

e nu

mbe

r of

wor

kers

per

qui

ntile

(00

0s)

EducationWage

Figure 8.8 Employment expansion in the accession countries: polarization


196 Ágota Scharle

Estonia pursued an optimal policy mix during the transition that fos-tered a fast recovery with suppressed growth in wage inequalities and long-term unemployment in the low-skilled population. It kept both unemployment benefits and the minimum wage at a relatively low level and encouraged job creation with active labour market policies. Compared with the other Baltic States it also had a relatively high and increasing level of non-agricultural self-employment, most probably helped by a well-functioning public administration and early efforts to reduce administrative burdens (Masso and Eamets 2004). These factors together with an investment boom in construction are likely to explain

–150

–100

–50

0

50

100

150

200Hungary (1996–2007)

1 2 3 4 5

50

–30

–20

–10

0

10

20

30

40

1 2 3 4 5

Primary

Estonia (1998–2007)

Construction

LTI

HTI

LKIS

KIS

Figure 8.9 Absolute change in employment by sector and wage quintile in Estonia and Hungary

Notes : HTI and LTI stand for high and low technology industry, KIS and LKIS stand for knowledge intensive and less knowledge intensive private services.



why employment tended to grow mostly at the bottom end of the jobs distribution (Figure 8.9).

Hungary fared the worst in terms of labour market readjustment following the transitional shock. The combination of fast privatiza-tion, high unemployment benefits and low spending on active labour market policies led to a fast and large drop in employment and a rise in wage inequality. Low-wage job creation was further discouraged by the rise in minimum wages, increasing macroeconomic instability and feeble attempts to reduce the administrative burden on SMEs. The rapid and early increase in wages at the top of the jobs distribution may explain why there was little job growth in the fourth quintile: apparently knowledge intensive services with more potential to expand already paid relatively high wages in 2000 13 so that new jobs were created either in these high wage knowledge intensive occupa-tions or the low-wage construction industry and other services. The

–150

–100

–50

0

50

100Slovakia (1998–2007)

1 2 3 4 5

–50

–25

0

25

50Slovenia (1998–2007)

1 2 3 4 5

–300–250–200–150–100–50

050

100150200

Czech Republic (1998–2007)

1 2 3 4 5

Primary

Construction

LTI

HTI

LKIS

KIS

Figure 8.10 Absolute change in employment by sector and wage quintile in Slovakia, Slovenia and the Czech Republic

Notes : HTI and LTI stand for high and low technology industry, KIS and LKIS stand for knowledge intensive and less knowledge intensive private services.


198 Ágota Scharle

considerable job loss in low-technology industries reinforced these developments: the shedding of such jobs affected mostly the bottom and the fourth quintile.

Finally, the Czech Republic, Slovakia and Slovenia are somewhere between best-performing Estonia and worst-performing Hungary (Figure 8.10). Slovakia was fairly successful at regaining its pre-transition level of employment. Slow privatization mitigated job destruction dur-ing the transitional recession, while labour market policies (together with the flat tax) helped to contain the rise in wage inequality. Slovakia also had the largest increase in self-employment in the region between 2000 and 2007. Self-employed jobs grew mostly in the second quintile, but possibly generated some more employment by hiring as well. This would explain why knowledge intensive services (beside construction) account for most of the rise in low-wage employment. However, there was no increase in jobs for workers with primary education, which may at least partly be explained by the relatively high minimum wage. The Slovenian case is similar in that the loss of jobs was moderate during the transition, but markedly different in that agriculture played an impor-tant role in the regeneration of low-wage employment. While Estonians created jobs for uneducated workers in new SMEs in the service sector, Slovenians employed them on small (and inefficient) farms.

The Czech Republic seems to be closest to the Southern European pattern of modest but clear upgrading in the jobs distribution. A look at the sectoral breakdown of employment gains and losses confirms this: job shedding in the bottom quintile in low-technology industries and less knowledge intensive services was made up for by job creation in better-paid jobs in high-technology industries and knowledge intensive services. However, this also implies that the Czech economy did not create new jobs for uneducated workers in the past ten years, and hence could not return to its pre-transition level of employment.

8.7 Conclusions

This chapter has reviewed patterns of change in the jobs distribution of seven of the eight post-socialist countries that joined the EU in 2004. Though somewhat poorer, these economies were roughly at the same stage of industrial development as the Southern European economies in the mid-1990s, and followed a similar path of sectoral reallocation of labour during the past 15 years. However, the observed patterns of change in the jobs distribution shows marked differences both within the accession countries and in comparison to Southern Europe, despite


the fact that all countries experienced considerable reallocation of jobs across sectors and the consequences for the jobs distribution were the same as in more developed EU Member States.

Most of the differences seem to be attributable to the transitional shock and the variations in the policy choices of post-socialist govern-ments: their initial response to the transitional recession and later, the setting of minimum wages and attempts to create a business environ-ment favourable to SME growth. It has been argued that the optimal choice of policies would lead to considerable job creation at the bot-tom end of the distribution resulting in a polarized pattern of change over the past 15 years. Thus, while the shift away from low-wage jobs in agriculture and low-technology industries towards well-paid jobs in services may seem a favourable development in Southern Europe, it is not so fortunate if it is a result of insufficient job creation in SMEs, as is likely to have been the case in post-socialist economies.

The Czech Republic appeared to be closest to the benchmark Spanish case, where the jobs distribution showed a modest but clear upgrad-ing in the past 15 years. However, during this period total employment grew by almost 19 per cent in Spain while it dropped by over 1 per cent in the Czech Republic. By contrast, the seemingly unfavourable growth in bad jobs in Estonia contributed to an increase in total employment of almost 5 per cent.

Notes

1 . Helpful comments by Enrique Fernandez Macias on earlier versions of this chapter as well as generous help with producing some of the Figures are gratefully acknowledged.

2 . See especially Newell and Reilly (1999) and Kertesi and Köllő (2000). 3 . Pre-transition levels of productivity were indeed very low. Kornai (1992)

reports that East German levels of productivity in mining and energy were around 40–46 per cent of West German levels. The energy intensity of out-put in 1979 was twice or three times higher in European planned economies than in Western Europe (Kornai 1992: 293–294).

4 . The Comecon Council for Mutual Economic Assistance, abbreviated as CMEA or Comecon, was founded in 1949 to co-ordinate economic develop-ment among Soviet-oriented economies including among others Hungary, the former Soviet Union, the former Czechoslovakia, Poland, and Romania. It was disbanded in 1991.

5 . Blanchard (1997) develops a two-phase transition model to explain how and why restructuring happens in the second phase following the initial shock. This turns out to be largely dependent on the amount of unemploy-ment in the initial phase.

6 . Data from EBRD Transition update (2000: table 1 , page 4).

200 Ágota Scharle

7 . From 1996 to 1998, exports to Russia dropped from 16 to 13 per cent in Estonia, compared to 23 to 12 per cent in Latvia and 24 to 17 in Lithuania (table 5 in Dezséri 2001).

8 . Reasons for the slow growth in SMEs are under-researched but most likely include overregulation, lack of capital (including social capital) and a rela-tively extended welfare system. The EU15 are more often used as a bench-mark than Latin American countries and, hence, few observers have noticed the insufficient speed of SME growth in the CEE.

9 . The underlying process is the following: the transition entails a shift from centrally planned (or at least strictly regulated) wage setting to market wages that correspond to marginal products. Low-productivity workers tend to set their reservation wages above the unemployment benefit, which is likely to be too high compared to their marginal product. It takes time for workers to notice this (through their failure to get a job) and lower their expectations. Inequalities continue to persist until wages adjust sufficiently. The adjust-ment process is longer if the cost of hiring low-wage workers is increased by taxes, and shorter if it is reduced by subsidies (Balla et al. 2008).

10 . Income inequalities nevertheless increased relatively faster than in the other accession countries, partly because the Baltic States were the poor-est among the A8 and could not (or chose not to) spend as much on social transfers (especially on pensions). There is also a lack of reliable data to com-pare pre- and post-transition levels of inequality. For example, Milanovic (1999) finds that income inequality increased faster in Latvia than in the Visegrad group, but admits that data for Latvia suffer from a strong upward bias (mostly due to biased sampling).

11 . Recall that in the model the rise of inequalities follows from relatively high benefits and part of the problem is that low skilled wages do not fall enough.

12 . The underlying expectation is that firms underreport wages and a mini-mum wage rise would reduce the undeclared part. The conclusion however that this would boost net government revenues is not sufficiently supported by theory and empirical evidence. For a summary of the Hungarian debate, see Benedek et al. (2006). 13 The ranking of jobs is based on the wage dis-tribution in 2000 in the EJM database. Note that in Latvia, Lithuania, and Slovakia, there was considerable job creation in lower-paid jobs within knowledge intensive services.

201

Adam Smith in The Wealth of Nations argued that wage differences are determined not only by competitive factors (compensating differentials) but also by differences in individual abilities and institutions, which are non-competitive. He saw the latter factors as arising from the ‘laws of Europe’, regulating wages, restricting labour mobility and the crea-tion of barriers to entry in internal labour markets (see also Katz and Autor 1999). This is a widely described phenomenon and accepted fact by scholars who criticize the orthodox neo-classical market approach to labour markets and wage determination (e.g. Kerr 1977; Doeringer and Piore 1971). Other institutions impact on the way education/skills are traded for wages which relate to industrial relations systems (Dunlop 1958). In summary, there are two mechanisms which affect wage deter-mination: those mechanisms that reflect economic rationality and other that are more linked to institutions, politics and custom. In this chapter we review selective literature that deals with skill–wage match, then we look at the skill–wage match in European countries and some of the mechanisms that might influence their association or dissoci-ation. Finally, we test a set of hypotheses using log-linear association models developed by researchers in the field of social mobility research, to explore the association between skills and wages at country level, bringing our findings together with other contextual variables relating to the institutional framework in which labour markets operate.

9.1 Economic mechanisms determining wages

In the orthodox economics approach, labour markets have a double function: they offer a mechanism to recruit suitably skilled workers and a mechanism of wage determination in accordance with skill demand

9 The Institutional Context of Skills–Wages Mismatches Jean-Marie Jungblut and Philip O’Connell

202 Jean-Marie Jungblut and Philip O’Connell

and supply. In a fully competitive labour market with perfect informa-tion about workers’ skills and wages, individual workers would be sorted and matched to jobs according to their education, experience and skill levels while each job would have a wage level reflecting the average workers’ productivity. Thus the matching of workers to jobs would be aligned with demands and supply of skills and the remuneration should be in accordance with the expected value added by each job. This situ-ation would typically lead to market equilibrium. However information asymmetries and labour market institutions prevent this equilibrium. Many mechanisms inherent to labour markets sort workers into jobs and attribute wages to jobs in a way that does not necessarily reflect marginal productivity. We would therefore expect that countries would typically differ in terms of assigning or sorting workers to jobs and in the way jobs are remunerated.

Some of the more economic mechanisms shall be mentioned briefly. Lazear (1979) points to mechanisms of a universal nature, such as in the agency hypothesis which gives rise to steeply sloped earnings profiles to prevent workers from shirking. In this approach, younger workers are paid less than their marginal productivity while there is a seniority premium for workers with long tenure. This results in internal labour markets with job hierarchies and differential remunerations according to seniority and/or tenure. According to this theory, the employment relationship is seen as a reward structure for employees to ‘invest’ in the company at the start of their career and realize these investments over an extended period of time. This connection is supported by the literature on internal labour markets and long-term wage relationships. Farber (1999) believes that firms appear willing to pay to encourage long-term employment relationships, and they may do so because it is efficient to invest in their workforce. Recent analysis on average tenure of workers shows that the attachment between workers and employers has hardly changed over time, contrary to popular belief (Mayer et al. 2010; Huff Stevens 2008).

Another related form of labour market differentiation which is reflected in matching disparities and differential remuneration is described by the theory of dual labour markets (Barbieri 2009; Dickens and Lang 1985). This approach holds that there is a distinct low-wage sector in which there are no returns to schooling, where workers do not receive on-job training and there are noneconomic barriers which prevent incumbents of secondary labour position from obtaining better jobs (in the primary sector). This second tier of the labour market is used as a resource pool of labour in order to generate flexibility. If companies

The Institutional Context of Skills–Wages Mismatches 203

need intermittent workers to meet temporary higher demands of labour, they will temporarily recruit workers from this pool, often with-out the intent to turn the employment relationship into a long-term wage relationship. These noneconomic barriers between sectors are of a contractual nature, often achieved through the help of labour market intermediaries, which provide a temporary low-cost solution to labour shortage. The level and nature of the barriers are influenced by other labour market institutions such as employment protection and rights of workers by law as well as collective agreements covering the workforce.

Lindbeck and Snower (1988, 2002) examine the behaviour of eco-nomic agents in markets where some participants have more privileged positions than others. Incumbent workers enjoy favourable employ-ment opportunities while the outsiders are excluded from the labour market. The reason for this disparity is that firms incur labour turn-over costs when they replace insiders with outsiders. Examples of labour turnover costs are the costs of hiring, firing and providing firm-specific training, and the potential loss of accumulated knowledge and skills. Insiders may resist competition with outsiders by refusing to cooperate with or harassing outsiders who try to underbid the wages of incum-bent workers. Thereby insiders have also the power to increase their wages and make new hires less attractive to employers, often via wage bargaining involving their union representation. This fact is also linked to the efficiency wage hypothesis which argues that wages are deter-mined by more than simply supply and demand. Specifically, it points to the incentive for managers to pay their employees more than the market-clearing wage in order to increase their productivity or effi-ciency (Akerlof and Yellen 1986; Krueger and Summers 1988).

Other mechanisms of matching workers to jobs and wages to jobs seem to be dictated by factors such as product markets, the availability of scarce specific skills, technical circumstances and so on. Institutional characteristics of national labour markets, such as institutions of col-lective bargaining, may result in specifically differentiated associations of wages by skill levels. Moreover, not all jobs are freely available to everyone having the right level of skills, but may, instead, be arranged in promotional chains in a framework of internal labour markets. Therefore jobs tend to be ranked in a hierarchy where each job is either an entry position or accessible by incumbents of a limited series of other jobs, that is by incumbents that are likely to have the necessary expe-rience and set of skills to effectively perform the job and/or simply the requested level of seniority to be promoted. This means that some jobs are more or less freely accessible while others are very limited in access.


This might be a universal feature of internal labour markets, but, again, the extent might differ from one country to another. We will see below how this has been elaborated in industrial relations research.

9.2 Institutional mechanisms of wage determination

Another type of mechanism that impacts the skill–wage disparity, in a limited way, is wage setting at the national, sector or company level. Wage-setting mechanisms are typically responsible for wage differen-tials that are at least not directly related to the level of productivity but to the power inherent to certain professions to cause disruption of economic life when using industrial action. Furthermore, some profes-sionals may have a skill monopoly, some form of protection from com-petition or possess insider knowledge that they can leverage to increase rewards. Such professions are often reserved for persons with higher levels of education but sometimes they are accessible only if incum-bents have spent a certain amount of time in a company: such positions might be considered as bonuses for extra achievement during a career. These circumstances eventually lead to promotion ladders in internal labour markets (Doeringer and Piore 1971) and worker segregation.

Following Dunlop (1958) an industrial relations system can be seen to consist of three actors: management organizations, workers’ repre-sentatives (including the formal and informal ways they are organized) and government agencies that play a role in the system (also including specialized professional intermediaries such as tripartite organizations etc.). These actors and their hierarchies or organizations are located within an environment defined in terms of technology, labour and product markets, and the distribution of power in the larger society as it impacts the actors and the workplace. In this environment actors inter-act with each other, negotiate, and use economic and political power or influence in the process of determining the rules that constitute the output of the industrial relations system. In a later publication, Dunlop et al. (1964) conclude that ‘economic development tends to narrow differentials’ (p. 248) and that ‘above all, the new technology must be served’ (p. 72) and that this was particularly true for job and wage clas-sifications and job evaluation ‘in a sense that they are most universal’ (Dunlop 1958).

Maurice et al. (1982, 1984) oppose this emphasis on the universal classifications of jobs and wages in explaining the difference between Germany and France while other scholars have pointed out the diffe-rence between the industrial relations systems of France and the UK


as well (e.g. Eyraud et al. 1990). These authors highlight the role of skill production regimes and the importance of specific skills for the protection of insiders to the disadvantage of outsiders and their deci-sive impact on the way wages are associated with jobs. This relates to the strategy of efficiency wages discussed above. Contrary to Dunlop (1958), Maurice (1984) questions whether rules and decisions are shaped by technical and economic contexts rather than by ‘power contexts in the larger society or ideologies’ (p. 351) and argued that the ‘processes creating job structures (the classification of jobs and wages, training, the relationship between promotion and job, worker mobility between employment categories, and the promotion)’ is not dictated by neces-sity or reason. Later Eyraud et al. (1990) argue that ‘training practices, labour market structure, industrial relations systems and patterns of labour management are mutually interdependent and reinforcing’ (p. 502). Indeed, the way skills – other than general skills – are provided or acquired plays a significant role in this context. If workers have to ‘shop’ for skills by frequently changing employers at the beginning of their career accumulating work experiences through internships and invest to acquire skills on the market themselves, they will be seeking a premium for these investments at a later stage of their professional life. If, however, skills needed to work are more likely to be obtained on-the-job and linked to a succession of jobs in the internal labour market, the incentive structure of pay is quite different.

In recent research, scholars dealing with the Varieties of Capitalism (VoC) approach (Hall and Soskice 2000; Kelly and Hamann 2007) have also shed light on how countries diverge in their wage-setting practices in relation to the processes and actors involved. The VoC approach has grouped capitalist systems – especially those of the developed OECD economies – into two categories: (1) Liberal Market Economies (LMEs), and (2) Coordinated Market Economies (CMEs). Although defined in various ways, neo-corporatism was generally associated with the capac-ity of a state to negotiate durable bargains with employers and the trade union movement regarding wages, working conditions and social or economic policy. Accordingly, a nation’s capacity for neo-corporatism was generally said to depend on the centralization or concentration of the trade union movement, following an Olsonian logic of collect-ive action which specifies that more encompassing unions can better internalize the economic effects of their wage settlements (Olson 1965; Calmfors and Driffill 1988). The first cooperation problem to solve is the sphere of industrial relations, where the problem that companies face is how to coordinate bargaining over wages and working conditions with


their workers or their representatives. Outcomes of the coordination are wage and productivity levels that determine the success of the firm and help in the rating of unemployment or inflation in the economy. In the sphere of education and training, firms face the problem of securing a suitably skilled workforce, while workers face the problem of deciding how much to invest in their competencies. The outcome of this co-ordination problem turns out to be not only the fortunes of individual workers and companies but the skill levels and competitiveness of the overall economy. This also links with the arguments of Maurice et al. (1982) who see one of the major distinctions in skill production between Germany and France in the way they provide training to their workers.

Sociologists often make a distinction between occupational labour markets and firm’s internal labour markets. Occupations represent groups of individuals who perform similar activities and therefore have a similar level of skills and training (Kalleberg and Sorensen 1979). Marsden (1990) proposed to look at the German labour market as a predominantly occupational labour market where workers’ skills are largely marketable and workers lose little by moving between compa-nies, while companies lose little when they have to hire new workers. Marsden contrasts occupational labour markets with internal labour markets, in which workers remain for long periods of time acquiring company-specific skills. If a skilled employee is made redundant, the certificate enables him to obtain alternative skilled employment via the external occupational labour market (see also Soskice 1994).

Eyraud et al. (1990) contrasted France and Britain as two examples where either internal labour markets or external occupational markets were predominant. Occupational labour markets such as the German one provide a fairly general level of occupational skills through appren-ticeships, while France hires different categories of generally skilled staff and provides them with company-specific skills. In this respect, per-haps surprisingly, Germany is more comparable to the UK while France is more comparable to Sweden (see, for example, Lazear and Oyer 2004, for Sweden). This debate can also be found in Streeck’s (2009) criticism about the variety of capitalism approach, where he argues for common-alities of capitalism and regards Germany as much closer to the UK than many other commentators appear to believe.

The focus of our investigation is thus on the institutional mecha-nisms that impact on the dissociation of skills and wages, contrary to what is believed to be a natural close association – that is, each job requires a certain level of skills and is remunerated in accordance with average worker productivity. There is sound empirical evidence in our


data to suggest that a substantial share of the jobs have wages that are higher (or lower) than one would expect, given the average required skill level of its incumbents. We contend that jobs where skills and wages are dissociated vary in a non-random fashion across countries and that in countries like the liberal market economies the incentives for insiders have to be much higher than it would be possible for employers to pay in countries which have more regulated/coordinated labour market. We expect that the skill–wage dissociations in occupational labour markets such as Britain and Germany are therefore higher than in the more reg-ulated labour markets of France and Sweden. Especially, wage bargain-ing and the bargaining coverage in the latter countries sets limits to what job holders can claim to earn.

In the former two countries which have more flexible labour markets, skills are traded for wages: there need to be on the one hand higher incentive structures for workers to stay with their employers or, on the other, an attractive pay level to recruit workers with the required level of skills. This phenomenon eventually leads to the unexpected outcome that in countries where wages are centrally agreed, they are more in line with productivity or lower, while in countries with more ad hoc wage setting, wages may be less in line with what would be expected in terms of economic equilibrium.

To give a further rational explanation for this, we may also refer to theories about endogenous determinants of wage levels. For example, on the one hand, Akerlof and Yellen (1990) argue that, in a firm where the workers’ characteristics are not possible to be fully observed and where the monitoring of their actions is not possible, employers have to find well-designed incentives to maximize the workers’ effort. According to the ‘fair wage-effort’ hypothesis proposed by these authors, workers often compare their wages either with workers within the same firm or with workers in other firms. And therefore firms have a strong incentive to compress their wage distribution, which may improve labour rela-tions and enhance average workers’ productivity. This is, we argue, the case in France and Sweden.

On the other hand, the ‘tournament’ model which seems to con-tradict the former theoretical explanation, developed by Lazear and Rosen (1981), points to the benefits of a more dispersed wage struc-ture, derived from a performance-based pay system. A simple version of the tournament model suggests that managers should introduce a large spread in the rewards of workers – in the form of a promotion or a bonus – in order to stimulate their effort and should award the largest prize to the most productive workers. This, we argue, explains the case


of more flexible labour markets and thus helps to explain the nexus between skill and wage and the observable dissociations.

Finally, there seems to be at least two different endogenous wage-set-ting mechanisms which are in line with exogenous factors: Companies operating in flexible occupational labour markets opt for a tournament model while companies operating with internal labour markets with strong wage-setting mechanisms opt for the ‘fair wage’ alternative.

We can now formulate a series of hypothesis to be tested in the ana-lytical part of the chapter in line with the theories presented so far.

Hypothesis 1: Workers are sorted into jobs in relation to their qualification and experience and wages are assigned to jobs according to the presupposed level of productivity of the job.

The first hypothesis (H1) posits that, ceteris paribus , each level of skill that is common for the incumbents of a job corresponds to a typical wage level. Table 9.1 represents a contingency table where cells are jobs defined by skill and wage quintiles weighted by the number of workers. In a per-fectly competitive economy with perfect information on qualifications and wages, all individuals should be on or at least around the diagonal. This is obviously not the case. However, there is some validity to H1 because there is some evidence of a symmetric clustering around the diag-onal; cell frequencies are lowest for the cells most distant to the diagonal.

Hypothesis 2: However, significant disparities between skill profiles of jobs and wages do prevail in all economies. Matching is more or less imperfect due to certain mechanisms inherent to the matching process under institutional constraints and information asymmetries. In short, there are national differ-ences when it comes to skill and wage differentials of jobs.

The second hypothesis (H2) refers to the argument made by indus-trial relations scholars that beyond economic and technical necessities,

Table 9.1 Contingency table of wage and skill quintiles for all countries in the analysis

Wage quintiles

Skill quintiles 1 2 3 4 5 Total

1 650 714 584 270 28 2246

2 179 500 644 155 10 1488 3 218 170 388 287 60 1123 4 11 54 141 413 824 1443 5 4 7 37 83 1,529 1660

Total 1062 1445 1794 1208 2451 7960


there are certain cultural contexts which impact on the way workers are sorted into jobs and how jobs are rewarded. This is our central argu-ment in this chapter. It states that, ceteris paribus , apart from the perfect match of skill levels and wage, there is a substantial number of jobs for which wage does not meet the expected level of skills typical for each of these jobs. This is either because the jobs require an initial level of skills which is higher than would be typical for the wage level we observe, or because the wage level is higher than we would expect, given the typi-cal level of skill required. The difference, we presume can be explained by national institutional contexts, such as the nature of the labour mar-ket, wage bargaining institutions, skill production arrangements and the specificity of skills needed for each job.

Among these jobs characterized by skill–wage disparities, there are yet other factors which are not down to national particularities, but of a more universal type. These are more ‘economically motivated’ wage differentials as the theories in the first section describe such as incentive wages, agency theory and insider–outsider considerations and internal labour markets. Accordingly, we can find two distinct exceptions to our general hypothesis number one:

Hypothesis 3: Certain jobs can be considered as ‘entry-level jobs’ (ELJ) which are characterized by a systematic skill–wage disparity; that is incumbents have a higher level of skills and presumably productivity than their typical remuneration or inconsistent with their actual (lower) remuneration.

This pattern is consistent with the presence of internal labour markets (Doeringer and Piore 1971; Siebert and Addison 1991) and with wages earned by ‘entrants’ in insider–outsider theory (Lindbeck and Snower 1988, 2002). These positions are usually associated with apprentice-ships and with spells of on-the-job training in less-secure and low-wage positions, trading training for wages. Some of these jobs would also be located in the so-called secondary labour market in contrast to the primary labour market (Dickens and Lang 1985).

Hypothesis 4: Some jobs are privileged in the sense that they are reserved to workers promoted in the internal labour market. Such jobs pay salary levels above marginal productivity. (Seniority premium jobs: SPJ)

To proceed to these estimates we need a series of design matrices reflecting the entry-level jobs and jobs with seniority premiums. As we will see in the tables with the descriptive statistics, there are some combinations of skill requirements of jobs and wage levels which seem to be occupied by either mostly young job entrants while others are occupied by more experienced workers with higher levels of seniority. Accordingly the following two matrices represent topological effects


to be estimated in order to cover hypothesis 3 (entry-level jobs: ELJ) and 4 (seniority premium jobs: SPR).

The two matrices ( Figure 9.1 ) represent the five-by-five skill–wage con-tingency and the ones in the table represent the cells for which effects are separately estimated. There is one parameter for the entry-level jobs in the first column (wQ1) in the table and another parameter estimated as seniority premia represented by the pattern above the diagonal. The choice of the cells is in line with the descriptive statistics provided in the tables to follow. When we average the descriptive statistics pre-sented in Tables 9.2A –D across all countries, the proportion of young workers, senior workers, tenure, contracts and training arrangements yield the patterns given in the above two design matrices. Not every job below the diagonal is an entry-level job and not all jobs above the diagonal are jobs with seniority premiums. We have also produced esti-mations with different cells being ELJ and SPJ, but the most conclusive results were obtained for the cells included in the matrices above.

First, there are the so-called entry-level jobs in Q(2,1) and Q(3,1) where intermediate levels of skills are associated with the lowest level of remuneration. As discussed above, such jobs are mostly occupied by young workers subject to training arrangements, apprenticeships or just on probation. These jobs are the entry jobs leading to better remuner-ated jobs. We do not have longitudinal data at our disposition and thus no information of careers and progression of workers across jobs, there-fore we can only assume that some jobs are mainly accessible through promotions, while others are not. So they can also be considered as the beginning of a job queue, where incumbents wait until they are pro-moted if successful in a tournament (see also Thurow 1975) or after the required years of seniority with the same employer in the case of an internal labour market. Second, jobs with seniority premia are situated

Entry-level jobs

1 1 1 1 1

2 1 1 1 1

2 1 1 1 1

1 1 1 1 1

1 1 1 1 1

Diagonal

2 1 1 1 1

1 3 1 1 1

1 1 4 1 1

1 1 1 5 1

1 1 1 1 6

Seniority premia

1 1 2 1 1

1 1 1 2 1

1 1 1 1 2

1 1 1 1 2

1 1 1 1 1

Figure 9.1 Design matrices showing which of the association parameters of the models affect each cell of the table. Notes : H1: Five distinct parameters for the diagonals; H3: Entry-level jobs and H4: Seniority premia.


above the diagonal, but not adjacent to the diagonal combining Q(1,3) to Q(3,5) and Q(4,5); that is, jobs that have a higher wage than expected given the level of skill which is usually required in such jobs in terms of initial education. It can be argued that in addition to their initial level of education, most of the incumbents in such positions have learned on-the-job the skills that qualify them to earn a higher wage.

The four hypotheses will be assessed in a manner similar to a regres-sion-analysis approach. We will first control for the distribution on the diagonal to take the perfectly matched jobs out of the equation (H1), then we will introduce the topological models (H3 and H4) and finally estimate the remaining variance in the table using a multiplicative layer effect. This last effect can thus be considered the dissociation between skills and wages in the table that can only be explained by institutional contexts or national circumstances.

We will describe the data used to test these hypotheses and briefly provide a description of other variables associated with the skill–wage data matrices by country. We select, in line with the methodology used in the multivariate analysis of four countries which are supposed to represent the variety of labour markets in the European Union. The UK is used as a typical country with a flexible labour market and a weak system of worker representation; Germany represents the typical conti-nental country with an apprenticeship system, strong worker represen-tation and an occupational labour market. France represents the type of a country with strong internal labour markets and Sweden is the prototype of a Nordic country having a very high level of protection for permanent jobs but a very low protection for fixed-term and temporary agency workers. Sweden has also a well-developed system of industrial relations and strong internal labour markets.

9.2.1 Data and methods

In this chapter we use the wage quintiles along with the skill quintiles from the European Jobs Project data set to investigate the relationship between wage and skill levels, taking into account the series of hypoth-eses described above. We first investigate the overall association (or lack thereof) of skills and wages and then use log-linear modelling to examine what model is the best prediction of observed disparities. Subsequently, we look closely at association patterns in our model estimates and other labour market indicators such as union density, bargaining coverage and employment-protection legislation to summarize our findings and examine further evidence of the effects of national idiosyncrasies and types of labour markets: Internal labour markets, occupational labour markets and mixed forms. The universe of our study are male prime age


workers (aged 25–54) working as employees in full-time jobs in compa-nies in mining, quarrying and manufacturing with at least ten workers at the establishment level. This is our chosen population because we want to have as homogenous a population as possible. There are obvi-ously many caveats when choosing such a restricted population and the results from this analysis cannot be generalized to whole countries as such but should hold good for the traditional sectors of the economies in the presented countries. Our study covers about 7.8 million prime age male workers working in the sectors selected in the four countries which represent slightly above 8.1 per cent of the total workforce. The number of jobs covered is 26 per cent of all jobs in the original jobs file. It is thus clear that we can only draw conclusions about the population selected for the present analysis.

To measure skill and wage levels, the European Jobs Project uses a combination of sector of economic activity and the occupation to represent distinct jobs. Each job has at least two attributes: the average level of skills of workers in this job and a typical wage. This informa-tion is ranked from high to low for all jobs in a country and these are then ranked into five percentiles such that 20 per cent of the work-force is assigned to the lowest quintile when their typical wage (or skill level) is lowest. The second highest, 20 per cent, is assigned to the second quintile and so on. The skill quintiles are thus based on the average educational attainment for employees in a job while the wage levels represent the typical remuneration – in theory related to the worker’s marginal productivity – in each job. Tabulation of the two dimensions yields a contingency table with five rows for the skill dimension and five columns for the wage dimension. Each cell gives the number of workers in jobs having one out of five levels of skill combined with one out of five levels of income. We have in total four such tables for each of the countries we present in this analysis.

For the purpose of this chapter, we use log-linear models, in order to be able to test hypotheses against each other. One downside of such models is that they cannot handle very large data sets, so we re-weight each country in the analysis from the EU Labour Force Survey (using the population weights derived by Eurostat) to generate reduced sample sizes for each country. We have therefore samples ranging from 323 observa-tions in Sweden to 4,047 observations in Germany and a total sample of 7,950 observations across the four countries under observation.

A contingency table for both skills and wages in quintiles shows what wage level is associated with skill levels of jobs in each country weighted by the number of employees for each combination.


Table 9.1 shows that 650 workers in jobs with the lowest average level of skills are in the lowest wage quintile, 714 workers with the same skill level are working in jobs of the second wage level and so on. For ease of presentation we address the cells as Q(s,w) where s and w stand for the index of the row and column respectively. 1 Q(3,4) is the cell that combines the third skill quintile with the fourth wage quintile and, in Table 9.1 , corresponds to 287 job holders. It can also be seen that the distribution of workers across the cells is roughly symmetric; that is, most of the workers can be found on the diagonal and in the cells adjacent to the diagonal. This provides a first support for Hypothesis number 1, which holds that most workers are in jobs that are remuner-ated in line with the skill demand of the job. Nevertheless, there is a significant proportion of workers with skill–wage disparities. For exam-ple, there are 28 workers in the highest wage quintile while their job is in the lowest skill quintile. Such disparities are what we would propose to explain in this chapter.

Table 9.2 shows the proportion of observations (jobs weighted by the number of workers) off the diagonal from 1997 to 2007 for all coun-tries for which we have data in the European Jobs Project. This measure summarizes the frequency of workers for which we observe a skill–wage disparity. The figures for 2007 range from a low 55 per cent in Germany, Italy and the Netherlands to as high as 67 per cent in Finland and Greece. Several explanations can be given for these differences. We can observe in general that at the top of the table are countries from continental Europe and Sweden, which have either dual apprenticeship systems (DE and NL) or countries with vocational school provision (IT, FR, SE) – contents and curricula of the educational achievements are usually well known by employers. Also these latter three countries are known for having a tradition of internal labour markets. At the other end of the continuum we find countries with more liberalized labour markets (UK, DK and FI) or countries that do not have a tradition of either vocational schools or apprenticeships (PT and GR) where mar-ginal productivity and wages are presumably more dissociated as a consequence of substantial dual labour markets.

We might thus expect, on the one hand that liberal market econo-mies with less initial education and more on-the-job training experi-ence a higher level of disparities between skills and wages because they need a functioning incentive structure. On the other hand, countries with protected insider labour markets which lead to labour market dual-ism in the so-called coordinated economies with a strong tradition in initial training, also experience a high level of disparities but these are


due to skill–wage mismatches. For example the presence of an over-qualified workforce in the secondary labour market and long-tenured insiders with a substantial seniority premium. The career of a worker depends on the amount of training and the increase of formal levels of qualifications during work life, be it as an insider in a firm internal labour market or on an occupational labour market. Not surprisingly the UK, Denmark and Finland are such countries, where the participa-tion in on-the-job training, career breaks to return to college is most frequent. Hence the career path is less influenced by initial education and more by the professional orientation in work life itself (see Maurice et al. 1982; Eyraud et al. 1990 for theoretical explanations and national differences in this regard). It is hardly surprising that Germany and the Netherlands with its dual apprenticeship systems, where occupa-tions are well defined and standardized across the country and sectors of industry, have low levels of skill–wage disparities. Therefore, we can assume that countries where skill acquisition is delayed and happens in a work context are subject to a higher skill–wage disparity.

Looking at column (5), there is some evidence in Table 9.2 that skill–wage disparities have converged in Europe from 1997 to 2007. For Germany, for example, the skill–wage disparity (off-diagonal cells) in 1997 was 0.58 but decreased to 0.55 in 2007. At the same time it has increased in the Netherlands from 0.50 to 0.55. We observe that in

Table 9.2 Skill–wage disparities: share of observations in cells off the diagonal over time

Country

Off-diagonals by country Difference

1997–2007 (4)–(2) 1997 2002 2007

(1) (2) (3) (4) (5)

NL 0.50 0.54 0.55 0.05

IT 0.60 0.57 0.55 –0.05 DE 0.58 0.58 0.55 –0.03 FR 0.60 0.60 0.55 –0.05 SE 0.59 0.53 0.57 –0.02 ES 0.59 0.57 0.58 –0.01 BE 0.56 0.61 0.59 0.03 AT 0.62 0.64 0.59 –0.03 UK 0.59 0.59 0.60 0.01 PT 0.68 0.66 0.62 –0.06 DK 0.70 0.66 0.66 –0.05 FI 0.67 0.67 0.67 0.00 GR 0.72 0.74 0.67 –0.05


most of the countries under scrutiny the share of workers which have skill–wage disparities has decreased between 1 per cent and 6 per cent, in other words, the overall disparity has declined. This finding could be explained by the increased competitiveness of the European econ-omies leaving less room for remunerations that are not in line with (initial) skill requirements and in the general environment of liberali-zations on the labour market in many countries which may have led to a restructuring of the associations between skills and wages. This is also evidence for Dunlops assumption that economic development tends to ‘narrow differentials’, and thus reducing the impact of cultural idiosyncrasies.

This is not a refutation of findings on increased wage inequality or skill-biased technological change, as our methodology is ‘blind’ to such processes. It is important to note that we are not measuring skill–wage mismatch on an individual level, so we leave references to the literature on skill mismatch aside. The figures reported here are jobs which have a typical skill requirement and a typical remuneration that is associated with them. The bulk of wage increases have occurred above median-wage level and particularly at the top earnings while in most countries the real median wage has not changed much over time (see Card and DiNardo 2002; Katz and Autor 1999, for detailed explanations), there-fore wage increases have happened mostly in the top quintile and con-cern just a small minority of the workforce studied here.

The log-linear layer association models used in this chapter unfor-tunately limits the number of levels to row or column dimension whichever is lowest minus one. For the remainder of the chapter, we will focus on four countries in order to explain in greater depth skill–wage disparity and the association with other factors. As both row and column dimensions are equal to five, we have to limit the number of layers to four (see Clogg 1994, chapter 5 for details on this limitation). Tables 9.3A and 9.3B deal with the distribution in the original tables. Panel i in Tables 9.3A and 9.3B show the standardized Pearson residuals, that is observed frequencies minus expected frequencies in the event of no association divided by the standard deviation. High positive num-bers show where the observed values lie above the expected values and negative values show the opposite. Apart from the diagonals, which lie systematically above the theoretical values in general, all countries show a high affinity between skills and wages in Q(5,5) and Q(4,5) and, on the other extreme, in Q(1,1) and Q(1,2).

On the one hand we can observe that certain cells above the diag-onal and below the diagonal are also way above the expected cell counts in the event of no association. In Germany, in particular, the


cells adjacent to the diagonal show a higher frequency than expected ( especially Q(1,2), Q(2,3) and Q(3,4)). On the other hand, we observe excess frequencies below the diagonal too, for example jobs that are remunerated systematically below the skill level. This is particularly the case in Sweden while the phenomenon is non-existent in the UK and France.

Table 9.3A Distributional characteristics of the contingency tables for Germany and France

Wage Quintiles

1 2 3 4 5 1 2 3 4 5

I. Germany

i. Standardized Pearson Residuals ii. Outflow table

Skill quintiles

1 0.83 0.53 0.00 0.04 –1.41 0.32 0.34 0.26 0.07 0.00

2 –0.34 0.81 1.22 –0.31 –1.38 0.09 0.40 0.51 0.00 0.00 3 0.50 0.07 –0.09 0.27 –0.74 0.34 0.27 0.22 0.17 0.00 4 –0.30 –0.35 –0.14 0.14 0.64 0.01 0.07 0.19 0.13 0.59 5 –0.69 –1.06 –1.00 –0.14 2.88 0.00 0.00 0.04 0.03 0.92

iii. Inflow table

Skill quintiles

1 0.54 0.38 0.26 0.30 0.00

2 0.14 0.43 0.50 0.01 0.00 χ 2 = 3900 3 0.31 0.15 0.12 0.36 0.00 Pr. = 0.000 4 0.01 0.03 0.08 0.22 0.23 d.f. = 16 5 0.00 0.00 0.03 0.12 0.77

II. France


Skill quintiles

1 1.23 0.62 0.39 –0.25 –1.98 0.31 0.18 0.23 0.23 0.05

2 –0.06 0.16 0.18 0.58 –0.85 0.09 0.15 0.24 0.51 0.01 3 –0.18 –0.21 0.83 0.31 –0.75 0.06 0.02 0.41 0.35 0.15 4 –0.54 –0.26 –0.73 0.41 1.12 0.00 0.02 0.02 0.35 0.60 5 –0.44 –0.30 –0.67 –1.04 2.45 0.00 0.00 0.00 0.00 0.99

iii. Inflow table

Skill quintiles

1 0.81 0.67 0.40 0.25 0.04 2 0.09 0.21 0.17 0.22 0.00 χ 2 = 1500 3 0.09 0.04 0.40 0.22 0.07 Pr. = 0.000 4 0.00 0.07 0.03 0.30 0.38 d.f. = 16 5 0.00 0.01 0.00 0.00 0.51


Panel ii in each of the sub tables of Tables 9.3A and 9.3B report outflows, that is the share of skill levels which are associated with distinct wage lev-els. For example, in Germany, out of those who work in jobs with average skill level (sQ3), 34 per cent are in the first wage quintile (wQ1), 27 per cent in second wage quintile (wQ2) and so on. This shows that a substantial

Table 9.3B Distributional characteristics of the contingency tables for the UK and Sweden

Wage quintiles

1 2 3 4 5 1 2 3 4 5

III. UK


Skill quintiles

1 0.46 1.63 0.54 –0.70 –1.94 0.12 0.44 0.31 0.13 0.00

2 –0.03 0.08 0.34 0.09 –0.47 0.02 0.21 0.45 0.31 0.01 3 –0.07 –0.28 0.83 0.27 –0.74 0.01 0.04 0.55 0.36 0.03 4 –0.21 –0.85 –0.92 0.72 1.26 0.00 0.00 0.05 0.38 0.56 5 –0.15 –0.57 –0.79 –0.38 1.90 0.00 0.00 0.00 0.14 0.85

iii. Inflow table

Skill quintiles

1 0.88 0.85 0.44 0.16 0.00

2 0.04 0.10 0.16 0.10 0.00 χ 2 = 1600 3 0.04 0.04 0.33 0.19 0.01 Pr. = 0.000 4 0.03 0.01 0.07 0.44 0.49 d.f. = 16 5 0.01 0.00 0.00 0.11 0.49

IV. Sweden


Skill quintiles

1 1.30 –0.05 –0.08 –0.33 –0.84 0.69 0.15 0.14 0.01 0.01

2 1.25 0.25 –0.62 –0.18 –0.70 0.71 0.21 0.03 0.03 0.02 3 –0.32 0.04 0.31 0.03 –0.06 0.15 0.20 0.45 0.10 0.10 4 –1.66 –0.09 0.55 0.52 0.68 0.03 0.14 0.30 0.20 0.34 5 –0.57 –0.16 –0.16 –0.04 0.93 0.04 0.04 0.04 0.04 0.85

iii. Inflow table

Skill quintiles

1 0.50 0.31 0.30 0.05 0.02

2 0.46 0.39 0.06 0.14 0.04 χ 2 = 253 3 0.02 0.08 0.18 0.09 0.04 Pr. = 0.000 4 0.01 0.20 0.44 0.68 0.48 d.f. = 16 5 0.01 0.02 0.02 0.05 0.42


proportion of people work in jobs which pay less than the skill require-ments would suggest. On the other hand, we have a majority of people (34% and 51%) paid more than the skill requirements in Germany located in cells Q(1,2) and Q(2,3). These are the jobs which we believe having a seniority premium. A similar pattern is observable for France and the UK.

The inflow tables represented in panels iii of Tables 9.3A and 9.3B show what type of skills are most common for each of the wage quintiles. Again, in Germany, we can see that apart from the 54 per cent where skill is in line with wages (Q1,1), 31 per cent of the workers in wQ1 work in jobs with average skill level (sQ3), an indication for entry-level jobs. What is striking is that for wage quintiles three and above, the most fre-quent cells are the ones just above the diagonal for Germany while in France this seems to happen along the first skill quintile. This suggests that in France with a tradition of internal labour markets, many work-ers working in jobs requiring low levels of initial education but succeed to move on the jobs with higher wage levels after spending a substan-tial amount of time in the internal labour market, acquiring company-specific skills through on-the-job training.

The UK seems to be a mix of the German and French case, as the most frequent cells are adjacent to the diagonal as well as along the first skill quintile. Sweden is the big exception here and shows that there is an almost symmetrical pattern around the diagonal, which leads us to believe that there is a strong element of dualism in the Swedish labour market. In contrast to the other countries, each wage quintile has a bimodal distribution when compared to the level of skills jobs required: above and below the diagonal. The jobs with positive skill–wage dispar-ity are good jobs while the jobs below the diagonal are obviously poorly paid for the level of initial skills required.

Table 9.4 in its sub tables provide descriptive statistics of jobs in the skill–wage contingency tables disaggregated by age (a and b), tenure (c and d) and permanent (e) or temporary contracts. The latter are also shown with the share of workers on probation or fixed-term training contracts(f).

We note first that figures in the cells change systematically from left to right and top to bottom. These patterns suggest that the tables reflect a movement of workers through a succession of jobs as they move through their careers; they also move from one income level to the next along the level of general skill acquired before entering the labour market and sometimes, if they acquire new skills with certifi-cation, from top to bottom of the table. This is of course only a guess,


as we do not have longitudinal data to prove this (for more details on this, see Sorensen and Tuma 1981 and DiPrete et al. 1997).

There is another difference mirrored in the table concerning skill level Q5 where wage decreases with increasing age, contrary to the lower skill quintiles where wage increases with age. The skill quintile Q5 represents jobs that for the most part require tertiary-level educa-tion, such as engineering, MBAs and so on. A significant number of young professionals with sQ4 and sQ5 seem to enter the labour market at high wage levels which gives evidence for a high demand for skills in line with the skill biased technological change hypothesis.

Table 9.4A Descriptive statistics of characteristics of workers in the skill–wage contingency tables

Germany

Wage quintiles

1 2 3 4 5 1 2 3 4 5

a. Ratio of workers aged 15–24 in each category

b. Mean age of workers in each category

Skill quintiles

1 0.09 0.11 0.08 0.08 – 40.61 41.09 41.18 41.00 –

2 0.12 0.19 0.21 0.00 0.52 41.49 38.89 37.73 47.00 28.64 3 0.20 0.13 0.15 0.15 0.00 37.33 40.01 40.08 39.68 53.22 4 0.16 0.12 0.08 0.10 0.04 41.85 42.47 40.87 42.30 42.48 5 – 0.00 0.03 0.00 0.02 – 50.79 44.48 45.29 42.23

c. Ratio of workers with more than ten years tenure

d. Rate of workers with less than one year tenure

Skill quintiles

1 0.45 0.52 0.61 0.65 – 0.16 0.14 0.09 0.00 –

2 0.47 0.43 0.51 1.00 0.48 0.10 0.16 0.11 0.00 0.00 3 0.47 0.55 0.54 0.57 1.00 0.13 0.09 0.08 0.10 0.00 4 0.50 0.54 0.53 0.58 0.59 0.00 0.13 0.07 0.08 0.05 5 – 0.70 0.46 0.60 0.50 – 0.00 0.08 0.09 0.10

e. Rate of workers with a permanent contract

f. Rate of workers on temporary probation or training contracts

Skill quintiles

1 0.89 0.88 0.95 0.92 0.64 0.50 0.68 0.84

2 0.88 0.84 0.86 1.00 0.48 0.63 0.85 0.88 1.00 3 0.86 0.85 0.88 0.89 1.00 0.94 0.86 0.87 0.82 4 1.00 0.88 0.91 0.89 0.94 1.00 0.67 1.00 0.80 5 1.00 1.00 0.97 0.95 1.00 0.68


When considering that wages increase with experience (or age), a movement from the left to the right of the skill/wage contingency table is the norm. These patterns are largely consistent with the human capital literature (Ben-Porath 1967; Mincer 1974) stating that wages increase with schooling and experience. For example, someone with just compulsory education would typically enter with the combination Q(1,1) or Q(1,2); he will then be promoted along the cells from left to right, as the declining share of 14–24 year olds show in Table 9.4.A.a. The same can be discovered when looking at Table 9.4.A.b where the average age of workers increases from left to right with the notable exception of Q(2,5) in France and Germany.

Further, with increasing work experience, the rate of workers with ten or more years of tenure in Table 9.4.A .c increases also from left to right to reach a table maximum whereby 100 per cent of German work-ers (in Q(2,4) and Q(3,5)) have more than ten years of tenure. Again in Table 9.4.A.iv, the upper right corner has very few workers who have started to work for their employer less than a year ago, compared to the figures around 15 per cent in the upper left corner. Most of the other cells in the table see an almost linear decrease from Q1 to Q5 of the wage continuum, except for the highest skill quintile (Q5) where the association seems to be U shaped when we reflect tenure and wage. This suggests substantial job mobility if workers have high levels of general skills that are marketable.

Indeed Topel and Ward (1992) argue that the main method of increasing salary is for workers to change jobs frequently, especially at the beginnings of their careers. In Table 9.4A.e, with the rate of workers with permanent contracts, we see that the table minima with workers on temporary contracts are in Q(3,1), Q(3,2) and Q(2,2) while the table maxima are in Q(2,4) and Q(3,5), that is the workers with average initial education such as craftsmen or master craftsmen with higher levels of earnings. This is, however, also the case with incum-bents of jobs with sQ5, where most workers seem to have permanent contracts.

In Table 9.4.A.f, we see the distribution of workers on probation and training contracts. These are again at a high level in Q(3,1) and the low-est in the upper right corner of the table with the exception of Q(5,2), where not a single worker seems to be on probation or training.

Table 9.4B shows the same statistics for France, indicating, a very similar pattern to that found in Germany. The first two wage quintiles of sQ3 seem to reflect the entry-level jobs with 18 per cent of 15–24 year olds in Q(3,1) followed by 10 per cent in Q(3,2). Consequently,


Table 9.4B.b shows that the lowest average age is found in Q(3,1) too. We also find the same anomaly as for Germany in Q(2,5); jobs that are occupied by relatively young workers averaging 32 years of age with jobs where skills required are below average but at the highest level of earnings.

There are very few jobs filled by workers below the age of 25, con-trary to Germany where over half of the incumbents were younger than 25. Table 9.4B .c reveals the share of workers with more than ten years of tenure in France. We notice that the lowest level is found again in Q(3,1) with 31 per cent while the jobs with highest rates of seniority are found just above the diagonal and in the upper right part of the table in Q(1,5), Q(3,5) and Q(4,5) but not in Q(2,5), very similar to the

Table 9.4B Descriptive statistics of characteristics of workers in the skill–wage contingency tables

France

Wage quintiles

1 2 3 4 5 1 2 3 4 5



Skill quintiles

1 0.15 0.11 0.06 0.17 0.06 37.4 38.7 38.3 36.7 39.8 2 0.06 0.00 0.09 0.07 0.00 38.1 38.7 41.4 40.0 32.0 3 0.18 0.10 0.09 0.09 0.10 31.8 43.0 40.1 39.9 38.8 4 0.00 0.08 0.07 0.04 41.3 32.9 37.4 41.9 5 0.02 42.0

c. Ratio of workers with more than 10 years tenure


Skill quintiles

1 0.42 0.44 0.49 0.39 0.60 0.21 0.26 0.18 0.25 0.09 2 0.45 0.52 0.60 0.46 0.00 0.11 0.15 0.12 0.19 0.00 3 0.31 0.73 0.53 0.65 0.75 0.14 0.00 0.13 0.15 0.14 4 0.43 0.28 0.44 0.61 0.28 0.00 0.11 0.07 5 0.54 0.09



Skill quintiles

1 0.81 0.82 0.83 0.74 0.89 0.00 0.00 0.05 0.03 0.00 2 0.84 0.92 0.94 0.82 1.00 0.39 0.00 0.32 0.02 3 0.92 1.00 0.85 0.88 0.83 0.00 0.17 0.14 0.31 4 0.85 1.00 0.94 0.96 0.00 0.20 0.16 5 0.98 0.65


pattern found for Germany. As for Germany as well, the highest share of workers with less than one year of tenure is found in the upper left corner of the table but the pattern is different as the most new recruits are found in jobs with sQ1 along the first three wage levels in addition to cell Q(4,2). This might indicate that job mobility in France is not so in line with job’s skill level as it seems to be the case in Germany but also across skill levels as workers start in low-level jobs and seem to progress to higher-skill-level jobs, which is in line with the distinction traditionally made between the two countries as being occupational and internal labour markets which are predominant in Germany and France, respectively.

In other words, while the pre supposed movement of workers in France goes predominantly from top to bottom of the table, in Germany the presupposed movement is more from left to right. This is also confirmed when looking at sub-tables b and c for both countries. The overall level of contract permanency in France is found in the centre of the table and adjacent to the diagonal, a stark contrast to Germany, where permanent jobs are more likely to be found, as we have seen, on the outer cells of the table. The numbers of job incumbents which are on probation and/or training are much lower in France than in Germany and no systematic pattern emerges. In Germany, training/probation periods are much higher and predominant below the diago-nal, that is, for jobs that are paid lesser than their required skill level. This finding provides more support to conventional understandings of the functioning of the labour markets in the two countries.

The next table (9.4C) reflects the worker characteristics for the UK. We can see again that entry-level jobs ( Table 9.4C .d) are located on the left side of the table, below the diagonal in Q(2,1) and Q(3,1) and Q(4,2); jobs for which skill requirements are higher than pay. The pattern is, however, in stark contrast to Germany and France, as the jobs where incumbents are younger are much more confined to this region whereas there are many exceptions in Germany (particularly in Q(2,5)), and in France, along sQ2. Incumbents in the entry-level jobs in the UK seem to be younger than in Germany and France. Jobs with highest levels of tenure, as shown in Table 9.4C .c are mainly located in the highest wage quintile, which is in stark contrast with France and especially Germany, where most of the jobs with highest level of seniority are located just above the diagonal. Only in France do we find in addition to the jobs above the diagonal, many jobs with high levels of seniority on the right margin of the table in Q(1,5) and Q(3,5).

When we consider jobs with less than one year of tenure, we can local-ize these again in Q(2,1) and Q(3,1), where most workers seem to be rather


young, providing further support for the thesis that these are entry-level positions. Permanent positions are very frequent in the UK; this should be interpreted with caution, as employment protection in the UK is very weak. It is however noteworthy, that, most of the cells below the diag-onal are permanent jobs, as are the jobs in wQ5. The lowest shares are to be found on the diagonal and the top left corner. In Table 9.4C .f, the proportion of workers on fixed-term probation or training contracts is highest on the diagonal and in the lower right corner above the diag-onal. This finding is in line with human capital theory on on-the-job training which maintains that employers in competitive markets only provide training for specific skills and share the cost with employees. This happens when there is a situation of mutual investment and an interest in maintaining the work relationship after the training.

Table 9.4C Descriptive statistics of characteristics of workers in the skill–wage contingency tables

UK

Wage quintiles

1 2 3 4 5 1 2 3 4 5



Skill quintiles

1 0.14 0.15 0.10 0.05 39.7 39.7 41.1 42.1 2 0.28 0.13 0.09 0.09 0.00 36.5 38.1 40.5 40.8 47.0 3 0.21 0.17 0.12 0.15 0.11 35.0 41.8 41.7 40.0 41.6 4 0.00 0.36 0.04 0.10 0.02 53.4 38.2 42.2 40.3 43.5 5 0.09 0.04 40.2 41.7



Skill quintiles

1 0.32 0.30 0.39 0.44 0.25 0.19 0.15 0.11 2 0.14 0.29 0.33 0.47 0.49 0.73 0.16 0.14 0.07 0.00 3 0.38 0.40 0.39 0.38 0.32 0.41 0.18 0.12 0.12 0.18 4 1.00 0.35 0.39 0.35 0.45 0.00 0.14 0.06 0.13 0.09 5 0.37 0.48 0.17 0.10



Skill quintiles

1 0.90 0.94 0.95 0.95 0.18 0.22 0.30 0.00 2 1.00 0.94 0.96 0.98 1.00 0.24 0.00 0.00 3 1.00 1.00 0.95 0.97 1.00 0.34 0.39 4 1.00 1.00 1.00 0.97 1.00 0.34 0.58 5 0.97 0.99 0.00 0.00


Table 9.4D shows the same summary statistics as above for Sweden. At a first glance, we find the same pattern as we have seen so far for entry-level jobs not only in Q(2,1) and Q(3,1), but also in Q(4,1) and Q(3,4). In Table 9.4D .c, we observe that again the jobs with highest average senior-ity are above the diagonal in Q(1,3), Q(2,4) and Q(3,5), close to the pat-tern we found in Germany but different to the other two countries. The pattern in Table 9.4D .d with the share of workers in each job having less than a year tenure shows again, that the entry-level jobs are located in the wQ1 for sQ1, 3 and 4, and in Q(1,4) and Q(3,4). Again, this is similar to what we observed for Germany and the UK but different from France, where such jobs are mostly located along sQ1. When we look at the pat-tern for permanent contracts in Table 9.4D .e, we see a familiar pattern

Table 9.4D Descriptive statistics of characteristics of workers in the skill–wage contingency tables

Sweden

Wage quintiles

1 2 3 4 5 1 2 3 4 5



Skill quintiles

1 0.14 0.11 0.08 0.33 0.00 39.2 40.7 41.7 38.9 47.0 2 0.14 0.08 0.09 0.09 0.11 39.3 40.5 41.0 43.5 42.6 3 0.15 0.10 0.02 0.13 0.00 38.8 38.9 42.7 41.5 46.5 4 0.06 0.01 0.03 0.03 0.01 45.6 43.6 43.0 41.4 44.4 5 0.00 0.00 0.01 59.5 47.0 40.0



Skill quintiles

1 0.39 0.55 0.64 0.52 0.54 0.19 0.13 0.08 0.35 0.00 2 0.48 0.50 0.48 0.69 0.61 0.14 0.13 0.12 0.10 0.04 3 0.44 0.38 0.52 0.42 0.64 0.16 0.14 0.13 0.24 0.15 4 0.59 0.62 0.59 0.49 0.53 0.16 0.10 0.09 0.09 0.10 5 1.00 0.30 0.42 0.00 0.00 0.13



Skill quintiles

1 0.89 0.90 0.93 0.72 1.00 0.44 0.28 0.15 0.41 2 0.92 0.91 0.96 0.92 0.97 0.37 0.42 0.32 0.00 0.51 3 0.91 0.88 0.95 0.90 0.94 0.58 0.11 0.43 0.26 1.00 4 0.93 0.97 0.98 0.97 0.96 1.00 0.49 0.31 0.28 0.08 5 1.00 1.00 0.95 0.27


emerging (e.g. Germany and UK but again not similar to France) with most permanent contracts in jobs with either high wage or high levels of skill requirements. Fixed-term contracts with probation periods or training contracts are confined to the entry-level jobs and two catego-ries of jobs in wQ5 in sQ2 and 3.

The findings presented so far favour the interpretation that there are different types of jobs that are held at different stages of workers’ careers. We can identify jobs that are held in the beginning of work relation-ship and other jobs that reflect more seniority, experience and tenure of incumbents working in them. Jobs in the middle of wQ1 and sQ2 and 3 are likely to be mainly ‘entry-level jobs’ where younger workers are hired and trained, while jobs in the upper right corner are occupied by permanent, older and experienced workers with seniority premiums.

Other patterns in the table are so far unexplained and will find an explanation in country-specific arrangements and institutions. These jobs are either the consequence of substantial bargaining power of these workers vis-à-vis their employers or of underlying incentive mechanisms in internal labour markets or simply training arrange-ments. The differences between Germany and France are striking in this regard. This pattern will become even more evident when we will discuss the skill disparities estimates later in the chapter. Both France and Sweden have a high level of bargaining coverage and voca-tional schooling while UK and Germany are usually characterized by occupational labour markets with apprenticeship systems and low bargaining coverage. This will be explained more systematically in the following.

9.2.2 Multivariate analysis

In order to test the hypotheses we have presented in the first section of this chapter, we use log-linear models to fit the log of the frequen-cies given in the original tables to find an estimate for the general wage–skill disparity in each table. This technique is mostly known from cross-national mobility research and the methodology has been developed and used, for example, by Clogg and Shihadeh (1994) and Yu Xie (1992), Ganzeboom, Luijkx and Treiman (1989) among others. The models explain the association found in a series of contingency tables in a so-called Goodman association model II or RC(M) model which was first developed by Leo Goodman (1979a, 1979b, 1986, 1991). In the log-multiplicative framework while it is possible to control for other effects, one comparison parameter ( b m ) is obtained for each sub-table representing the association of that particular table and to test


whether four parameters yield a better fit than one single association parameter across all the sub-tables taking into account a loss of degrees of freedom.

In their analysis of mobility patterns in 16 countries, Grusky and Hauser (1984) attributed cross-national variations in mobility to four explanatory variables – industrialization, educational enrolment, in come inequality and social democracy – and found that these exog-enous variables explained systematic differences between countries. Similarly, Ganzeboom et al. (1989) analysed 149 intergenerational mobility tables from 35 countries and found significant differences in the mobility regimes of the countries by separating the variation within countries from the variation across countries and rejected the so-called ‘common social fluidity’ hypothesis (sometimes also called the Featherman, Jones and Hauser or FJH hypothesis, see, for example, Erikson and Goldthorpe 1985, 1987 and 1992).

For technical details of the methodology used, and for more details on the association models used in this chapter, refer to the Technical Annex to this chapter and recent publications in the field of social mobility such as Richard Breen’s (2004) volume on Social Mobility in Europe.

The fit statistics of the consecutive models estimated are represented in Table 9.5 . All the models are intermediate between the independence model, assuming no association between skills and wages and the full-saturated model with the following formula:

R C L RL CL RC RCL

i jk i j k ik jk ij ijkF = tt t t t t t t

The first parameter t of the log-multiplicative model is the geometri-cal mean or main effect of the complete table which is 5.82. The next parameters, ,R C

i jt t and Lkt control for the simple adjustment to rows, col-

umns and layers while the additional parameters, ,RL CLik jkt t and RC

ijt allow estimating the interaction effects which are combinations of columns, rows and layers. The last parameter RCL

ijkt is usually ignored as it saturates the model; only the three two-way interactions are thus included for reasons of parsimony. This model yields a variation (G 2 ) of 672.15 (not shown in the table) and leaves around 8.5 per cent of the observations misclassified and is not significantly better than the saturated model, that is it fails to explain as much of the variation while being more parsi-monious. If we leave out the interaction effect of skill and wages RC

ijt , we get the so-called independence model given by the following formula:

R C L RL CLijk i j k ik jkF =tt t t t t


This model renders only the marginal distributions; that is control-ling for the skill and wage distributions and their interaction with the layer variable which is country, setting the association parameter for skills and wages 1RC

ij =t . This is the model M0 in Table 9.5 and yields a very large variation between the fitted values and the observed values (G 2 = 8134) and leaves over 41 per cent of the observations misclassi-fied. This model serves as a benchmark for other models which aim to reduce the unexplained variance by introducing specific parameters successively. Thus, each of the successive models tests a hypothesis as presented earlier in the chapter.

The first two models (1 and 2) block out the diagonals and thus adds a parameter for each combination of skill and wage when the index for row i is equal to the index for the column (i = j) first ignoring the layer effect (homogeneous model with five parameters) and then taking into account that each combination of skill and wage when i = j is different for each country (heterogeneous model with 5 * 4 = 20 parameters).

The homogeneous model estimates five distinct parameters in addition to the independence model (d.f. = 59 v. 64), while the hetero-geneous model estimates 4 * 5 = 20 parameters and leaves 44 degrees of freedom. The diagonals are estimated exactly as in the original table. In mobility research, this is called the inheritance parameter; that is when origin equals destination. This is of as much interest to us here as in mobility research as it controls for immobility or in our case it repre-sents the confirmation of hypothesis number 1, that skill requirements are in line with remuneration. We can call this the skill–wage parity explanation. Hence these two models substantially decrease the varia-tion in the estimated cell frequencies.

Model 2 reduces the unexplained variation of the independence model (0) by 57 per cent and leaves only 19.5 per cent of the observations misclassified. Model 2 is less parsimonious than model 1 but explains significantly more than model 1 and is thus better. However, both mod-els fail to explain as much as the saturated model and hypothesis 1 is far from being enough to explain the variation in the four national cross tabulations. We can thus conclude that the way skills and wages are associated on the diagonal (skill–wage parity) are significantly different in each country. However, our main focus is on skill–wage disparities, and therefore the following models try to estimate the off-diagonal associations using a combination of log-linear association models R + C and log-multiplicative association models (RC) and in addition some topological parameters which try to explain systematic divergence in the original tables as explained by our hypothesis 2–4.

Tabl

e 9.

5 M

od

el a

nd

fit

sta

tist

ics

Mo

del

Ty

pe

Des

crip

tio

n

Dia

gs

Top

olo

gic

al

G 2

df

Δ G

2 D

.I.

BIC

Sig

nif

ican

ce o

f th

e in

crem

enta

l x

2 te

st

0 –

Nu

ll a

sso

ciat

ion

bet

wee

n

S an

d W

–

81

34

64

– 0.

410

7559

.00

–

1 –

Dia

gon

al

hom

og.

38

85

59

0.52

2 0.

222

3354

.90

0.0

00

2 –

Dia

gon

al

her

erog

.

3492

4

4 0.

571

0.19

5 30

96.5

0 0.

00

0 3

UA

U

nif

orm

Ass

. (h

omog

.)

her

erog

.

1090

43

0.

866

0.0

86

703.

78

0.0

00

4 U

A

Un

ifor

m A

ss. (

het

er.)

h

erer

og.

10

25

40

0.87

4 0.

08

0 66

5.87

0.

00

0 5

RC

(M)

RC

(h

omog

.) h

erer

og.

52

7 37

0.

935

0.05

9 19

4.33

0.

00

0 6

RC

(M)

RC

(h

eter

og.)

h

erer

og.

35

4 2

5 0.

956

0.0

42

129.

51

0.0

00

7 R

C(M

) h

omog

enou

s R

C, e

qu

al s

core

s h

erer

og.

65

2 4

0 0.

920

0.07

1 29

3.4

0 –

8 R

+ C

(M)

Sem

i-h

eter

og. R

+ C

(X

ie, 1

992)

h

erer

og.

42

7 34

0.

947

0.0

47

121.

79

0.0

00

9 R

C(M

) Se

mi-

het

erog

. RC

(G

oo

dm

an)

her

erog

.

137

18

0.98

3 0.

018

–24.

46

0.0

00

10

R +

C(M

) +

T1

Sem

i-h

eter

og. R

+ C

(X

ie,

1992

) +

ELJ

h

erer

og.

ELJ

30

7 30

0.

962

0.38

1 37

.78

0.0

00

11

R +

C(M

) +

T1

+ T

2 Se

mi-

het

erog

. R +

C (

Xie

, 19

92)

+ E

LJ +

IL

M

her

erog

. E

LJ +

SPR

2

03

26

0.97

5 0.

031

–30.

48

0.0

00


The first model to estimate the association between skills and wages is the so-called homogeneous R + C uniform association model. This model uses just one more degree of freedom compared to model 2 while it reduces the unexplained variance of the independence model by over 86 per cent. The uniform association model computes just one single parameter to explain the association between skill levels and wages across all the tables, setting all the association between 2-by-2 adjacent cells of each table to just this parameter which is 0.71, a moderately strong association (no association would be 1) between skills and wages.

Model 4 allows country differences in the association between skills and wages and yields 0.63, for Germany 0.64 for France, 0.60 for Sweden and 1.10 for the UK. Thus, the association is similar for Germany and France, highest in Sweden and lowest in the UK after controlling the diagonals. Indeed, off-diagonal cells not adjacent to the diagonal are more frequent in France and Germany than in the UK while Sweden seems to have the highest spread of observation throughout the table, although the number of observed cases in Sweden is very low. This can easily be verified by the pattern in the table with the raw data in Table 9.5a . The variance explained, compared to the independence model is now 87.4 per cent and the share of misclassified observations are now reduced to less than 10 per cent. This is ample evidence that there clearly is still a significant association between skills and wages even when controlling (blocking out) the diagonals. The last model only adds little compared to the homogeneous UA model. However, when computing incremental x2 tests, 2 we find that all the more-complex models are significant when compared to the more-parsimonious ones using fewer estimators.

The best fit model is, however, not found so far. The Bayesian Information Criterion 3 (BIC) is 665.87 for the current model, which tells that the model(s) used so far do not explain the distribution better than the saturated model. If the BIC is negative with surplus degrees of freedom we have a parsimonious model explaining at least as much as the saturated model does.

Models 5 to 8 are different types of log-linear and log-multiplicative layer effect models where the interaction between rows (skills) and col-umns (wages) enters multiplicatively into the estimation. The rows and columns are attributed separate scores in the estimation process. The scores replace the ordinal dimensions of the rows and columns assign-ing a metric to these, similar to correspondence analysis. The scores for the row(s) and column(s) are estimated together with the models. Figure 9.2 shows the scores that are assigned to skills and wages by the


homogenous RC model (5); that is the same metric is applied to all coun-tries. The scores for skills show that there is no discrimination between the two lowest (highest) levels of skills, reducing the range of skill levels actually to three. This can be explained by the fact that the jobs on both extremes of the skill continuum are not so different from the next highest (lowest) skill category. It also makes sense in the light of high-est level of education provided by national educational systems which usually discriminate between three levels: lower secondary, upper sec-ondary and tertiary levels. In other words, jobs that are ranked in sQ1 (sQ5) are not significantly different from those ranked in sQ2 (sQ4). When considering the scores given to wages, we see that wQ1 is scored

Table 9.5a Complete table with the raw data of the countries used in the multivariate analysis

Wage quintiles

1 2 3 4 5

Skill quintiles France

1 170 97 127 123 25 2 19 31 52 108 3 3 20 6 126 110 48 4 1 10 8 148 251 5 1 1 1 1 341

Sweden

1 74 16 15 1 1 2 68 20 3 3 2 3 3 4 9 2 2 4 2 10 22 15 25 5 1 1 1 1 22

United Kingdom

1 66 237 168 69 1 2 3 29 61 42 2 3 3 10 127 82 8 4 2 2 26 193 288 5 1 1 1 49 291

Germany

1 340 364 274 77 1 2 89 420 528 2 3 3 192 150 126 93 2 4 6 32 85 57 260 5 1 4 34 32 875


higher than wQ2, indicating that the return on investment in educa-tion and training is higher for wQ1 than for wQ2. The scores attributed to the wage quintiles are linked to the distributional shape of the wage distribution which is usually right skewed while the scores of the skill quintiles reflect the returns on investment to education.

The RC model finds the ‘best’ score parameters to maximize the agree-ment between the data and the hypothesis or assumption of linear-by-linear interaction (Clogg 1994: 46). In model 7, the row and column scores are constrained to be equal and homogenous across countries. This model scores much worse than the previous two, and model 6 is significantly better, giving evidence that the scaling of row and columns have to be different. The models 5 to 8 successively reduce the variation to the empirical distribution with model 8 reducing the unexplained variation in the independence model by over 96.8 per cent (G 2 = 257). However, the BIC (113.30) still does not indicate a satisfactory level of explanation of the original contingency table.

–1.50

–1.00

–0.50

0.00

0.50

1.00

1.50

2.00

0 1 2 3 4 5 6

Adjusted scoresattributed toskills and wages

Skill / wage quintiles

Skills Wages

2.50

Figure 9.2 Row and column scores estimated with the log-multiplicative association model 5

Source : Jobs Project data set (authors’ calculations).


Continuing with the idea of finding a single odds-ratio for four adjacent cells per sub-table by country leads us to model 8 which is the ‘log-multiplicative layer effect model’ introduced by Xie (1992) with the origin–destination association varying multiplicatively with tables. Xie assumes that ( )expRC RCL

ij ijk ij k+ =t t c f where c’s describe the origin– destination association and w’s indicate table-specific deviations in association by country. The log-multiplicative layer effect model specifies the RC two-way and RCL three-way interactions as the log- multiplicative product of two things: the overall origin–destination two-way association and a deviation parameter for the k th table.

The model is a generalization of Goodman’s (1979a) row and column effects association model II, which is the base for model 9. The model is also a special case of a general class of models introduced by Goodman for multidimensional contingency tables and on the relationship between correspondence analysis and log-linear analysis (Goodman 1986: 262–266). See the Technical Annex for details about the two vari-ants of association models.

We also adopt the simple heterogeneous variant as researchers gener-ally recognize that the use of association models to assess differences in the strength of association between row and column categories is not effective if they allow scores of row and/or column to vary freely with mobility tables. This is because two sets of row/column scores are not comparable in any simple way (Yamaguchi 1987: 483). Therefore the association parameter w will vary by sub-table, but the scores for skills and wages in each of the sub-tables will be the same. assuming that skills and wages are ‘valued’ similarly in the different countries (i.e. returns to education and job rewards/productivity levels). Comparing models 8 and 9, we can see that apart from being more parsimonious, the Xie model also yields a slightly better fit.

Models 10 and 11 are based on the simple heterogeneous log-multi-plicative layer effect model (8) while adding two topological parameters to it. Model 8 yields satisfactory results with G 2 equal to 427.19 using 34 degrees of freedom. It reduces the variation of the null-association model by the same amount as model 9 and leaves 4.7 per cent of the observations misclassified while the Bayesian information criterion is equal to 121.79. Hence models 8 and 9 yield very close results, but still not convincing when compared to the saturated model. The phi param-eter estimated by the latter two models (see phi parameters in Table 9.6 ) give substance to hypothesis number 2, saying that there are substan-tial differences across countries in their association of skills and wages or the skill–wage disparity. However, we have still significant deviations


in some regions of the table to be covered by additional topological models. These topological design effects are aligned with our hypothe-ses 3 and 4. In addition to the blocking out of the diagonal parameters, the log-multiplicative association parameters from model 8, we intro-duce three series of parameters reflecting the existence of entry-level jobs and jobs with seniority premium.

Models 10 and 11 successively test these additional hypotheses. Model 10 based on model 8 (semi-heterogeneous log-multiplicative layer effect model), taking into account entry-level jobs, that is one parameter per country using four degrees of freedom (compared to model 8) brings the BIC to 37.78 and renders a model that is still more parsimonious than model 9. This model explains 96.2 per cent of the variation not explained by the independence model and leaves 3.8 per cent of the observations misclassified. Finally, model 11, including parameters for jobs with seniority premium brings G 2 down to 203, a reduction of the non-explained variance of the independence model by 97.5 per cent. This model yields a better explanation of the observed data with less parsimony than model 9. Model 11 also yields a BIC of −30.48, making this the most successful candidate to explain the frequencies observed in the tables and better than the saturated model.

Table 9.6 shows the resulting estimates of the last discussed models (8, 10 and 11) based on Xie’s log-multiplicative layer effect models. The table also covers some institutional variables that are highly likely to impact the skill–wage association: Employment-Protection Legislation (from OECD 2008), Union Density and Bargaining Coverage. The phi parameter (w) for models 8, 10 and 11 shows the level of wage–skill disparity for each of the four countries under scrutiny. Model 8 does not take into account any topological factors which might explain some of the disparity in the consecutive models. Phi is the remain-ing unexplained disparity once we control for entry-level jobs and

Table 9.6 Estimates from the final fitted model number (model 11) for prime age males in manufacturing

Country

Model 8 Model 10 Model 11

w w w ELJ SPR UD BACOV EPL

DE 0.70 1.29 1.16 3.26 –1.50 20.10 63.00 2.63 FR 0.47 0.88 0.67 1.60 0.21 8.00 95.00 3.00 SE 0.34 0.40 0.31 3.98 0.26 75.00 92.00 2.06 UK 0.81 1.53 1.31 0.66 –0.26 28.70 34.80 1.09


seniority premium. Sweden now appears to be the country with the lowest level of disparity, once the other factors are taken into account, followed by France. The two countries have a political climate that is supportive of equity and egalitarianism as it is embedded in the repub-lican ideals in France and the social-democratic welfare state arrange-ments in Sweden. Also a high level of employment protection fosters internal labour markets, as in France. Internal labour markets, as has been demonstrated by Maurice et al. (1982) and Eyraud et al. (1990) use in-company training of specific skills while occupational labour markets rely more on general skills and the external labour market to recruit their workforce. Consequently, the United Kingdom as a liberal market economy relies more on individualism and meritocracy as the rules to fix wages and hence has a higher skill–wage disparity. This is also true for Germany, but to a lesser degree, as Germany also has a non-negligible share of internal labour markets.

To compute the relative magnitude of the layer effect between two coun-tries the following formula must be used: (w 2 − w 1 )/w 1 (see Xie 1992: 387). Thus, compared to Germany, Sweden’s disparity is only half in model 8, because (0.34−0.70)/0.70 = −0.51 while France’s disparity in model 8 is one-third less pronounced compared to Germany. The disparity in Britain is, however, 15 per cent higher than in Germany. Once we control for entry-level jobs (model 10), Sweden’s skill–wage disparity becomes almost 70 per cent lower than Germany’s while France’s does not change and the gap between the UK and Germany increases to 18 per cent.

Finally, when controlling for seniority wage premiums, the gap between France and Germany increases to 42 per cent less disparity in France than in Germany while the gap between Germany and Sweden grows to 73 per cent less disparity in Sweden. The gap between Germany and the UK, however, narrows to 13 per cent.

We can thus conclude that there is more similarity between Germany and the UK than with the other two countries. The position of France regarding skill–wage disparity lies half way between the low skill–wage disparity in Sweden and the high skill–wage disparity in Germany and the UK. When confronting these results with contex-tual factors, we can see that countries with high bargaining coverage and internal labour markets (France and Sweden) have lower dispar-ities while the contrary is true for countries which are traditionally occupational labour markets and low bargaining coverage, but also less employment protection.

If wage agreements are reached and applicable to most workers in a country (which is especially the case in manufacturing), there is less room for individual wage negotiations. The more so as internal labour


markets prevent workers from changing jobs, and wage claims are mod-erate. When the skills are obtained in apprenticeships or on the mar-ket by workers themselves, something that is common in Britain and Germany, a premium has to be earned when workers take up jobs with an employer and thus substantial wage gains can be obtained when a worker switches employers in the context of flexible labour markets relying on outside recruitment.

Table 9.6 also provides the estimated parameters for entry-level jobs (ELJ); that is jobs that pay relatively low salaries for average-skilled workers, and seniority premium (SPR), that is senior positions occupied by skilled workers who have been promoted frequently in the inter-nal labour market. 4 On the one hand, there are positive associations between union density, bargaining coverage, employment protection on one side, and the strength of the coefficient for ELJ. One could explain this fact by the presence of strong internal labour markets fostered by traditional industrial relations systems with employment protection and a need for entry-level positions to the internal labour markets.

On the other, the same holds true for indicators of industrial rela-tions (union density and bargaining coverage) and seniority premium. If union density and/or bargaining coverage are high, the coefficient for SPR tends to be high too. Internal labour markets are facilitated by a high-level or employment-protection legislation and industrial rela-tions are important to find non-market arrangements and agreements (e.g. wage bargaining) as a functional equivalent to markets for wage setting.

Workers themselves also need a certain level of protection to be will-ing to invest in specific skills and a career progression as an incen-tive. This institutional arrangement typically found in CME, however, appears to lower skill–wage disparity in general. The contrary is true with countries that rely more on markets for wage setting and skill pro-duction as LMEs do. In order to function properly, occupational labour markets need more flexibility in their wage-setting procedures in order to attract workers with the right skills, motivate them and increase their attachment to their employer therefore wage incentives (one might call them efficiency wages, although these are also found in the former institutional arrangement) are important.

9.2.3 Further evidence from a sample of female workers in retail and service

As the above analysis has only been applied using a very restricted share of the labour force (prime aged full-time employed males in large


establishments in manufacturing), we have also taken a different sam-ple of the population and replicated the above analysis. We chose a pop-ulation of 25–54–year-old women in full-time employment, working in larger establishments in wholesale and retail, hotels and catering as well as in personal services which are sectors of industrial activity that are dominated by female employment. In general, there are more women on the diagonal than off the diagonal and also much more clustered around the diagonal compared to men which could mean that women are more often paid their marginal rate of productivity, but also that there are less fringe benefits and so on for women than for men and less returns to experience (seniority premia). Thus women are less likely to work in jobs characterized by skill–wage disparity (off-diagonals) when compared to men (see Table 9.3 ).

We could also interpret this as a lower level of return to education which is a common finding in human capital literature. However, the same pattern of converging countries ( Table 9.3 ) as for men is discovered for women. However the level of convergence is much less pronounced for women when compared to men. The distribution of younger women in the skill–wage table is very similar to the men’s distribution in the UK and Sweden but very different in the other two countries. The age pattern for women is much more symmetric to the diagonal in Q(3,1) to Q(5,3) across the four countries than this is the case for men. Women with most tenure are to be found in the lower right corner of the table while the top right corner is mostly empty. The entry-level jobs for women – in terms of the share of workers with less than one year tenure – are in the top left corner of the table. The distribution of women in terms of permanency of contracts is very similar to men and the same holds true for probation/training contracts.

The coefficients of the multivariate analysis for the data set using only women in wholesale trade and services are presented in Table 9.7 . Applying the association model developed by Xie (1992) to the data set of

Table 9.7 Estimates from the final fitted model number (model 11) for prime age women in whole trade and services

Country

Model 8 Model 10 Model 11

w w w ELJ SPR UD BACOV EPL

DE 0.19 0.48 0.64 –7.54 –1.16 20.10 63.00 2.63 FR –0.05 –0.16 0.36 1.83 6.12 8.00 95.00 3.00 SE 0.02 0.04 0.08 –2.27 –1.03 75.00 92.00 2.06 UK 0.27 0.60 0.55 –4.97 –5.01 28.70 34.80 1.09


women, using model 11 described above, we find association parameters that are very close to those for men. When compared to Germany the skill–wage disparity is 44 per cent lower in France (men: –42%), 88 per cent lower in Sweden (men: -73%) and 14 per cent lower in Britain (men 13% higher). Hence, with the exception of Britain, the estimates are very close to what we have estimated for the male sample and the ranking of the countries when we compare skill–wage disparity is only alternated for the top two countries: Germany becoming the country with most skill–wage disparities.

In general, the sectors we cover when looking at the female workforce are traditionally much less organized or covered by bargain agreements than are the male sectors in the main part of this chapter. This possibly explains why the layer-effect coefficients are lower for female workers than those observed for the male workers and why there is a higher share of women on the diagonal. We can thus confirm the results we presented for the male sample with findings from the female sample, which gives even more evidence that there are systematic differences across the countries when we look at the way workers are sorted to jobs and the way jobs are remunerated, which go beyond the simple rationale of economic processes.

9.3 Conclusion and discussion

If the association between skills and wages is less than what we would expect from perfectly competitive markets, this may reflect the influ-ence of systemic contextual differences between labour markets in dif-ferent countries for the selection of jobs in traditional sectors which we analyse. In this chapter we have tried to disentangle which factors may drive skill–wage disparities. We found a substantial level of association between skill and remuneration levels for jobs, around 30–40 per cent for men and 50–60 per cent for women (i.e. on the diagonals). The remu-neration in the sectors we analyse seems to be much more in line with skill levels of jobs for women than is the case for men. Furthermore, these disparities seem to have reduced since the mid-1990s for both men and women. Nevertheless, there subsists a substantial part of the workforce employed in jobs that have lower remuneration than the required skill level would predict. The opposite is also true. When we take account of certain discrepancies that might be expected, such as entry-level jobs which may be underpaid, and seniority positions which, by definition have a seniority premium, we still find skill–wage disparities in all the countries analysed. The differences between


the countries in terms of skill–wage disparities suggest an interesting conclusion: countries with occupational labour markets require more discretion to pay higher salaries while countries which rely more on internal recruitments in internal labour markets have a lower dispar-ity of skills–wages, but are clearly distinctive when it comes to entry-level jobs and seniority premium. When we replicated the analysis with another sample of the EU-LFS – women in retail, catering and ser-vices – we find similar patterns in skill–wage disparities, with a broadly similar ranking of countries, with the exception that Germany and the UK switch positions. The comparison of our estimates with contex-tual information on industrial relations and labour market institutions underlines the presence of institutional arrangements that support our hypothesis even though the number of countries in our multivariate analysis does not provide enough support to be conclusive. With the caveat that it would be imprudent to generalize our findings to the entire working populations of the countries we analyse, nonetheless the Jobs Project data set has allowed us to investigate a new field for research yielding interesting results combining economic and socio-logical perspectives on the way jobs combine levels of skills and remu-neration varying systematically across countries.

Technical annex

In categorical data analysis, the term association has a specific mean-ing. 5 Association models are a class of statistical models that fit observed frequencies in a contingency table with the objective of measuring the strength of association between two or more ordered categorical varia-bles. In a two-way table, the strength of association is measured between the two categorical variables that make up the table. For a three-way or higher-way table, the strength of association being measured can be between any pair of ordered categorical variables that comprise the cross-classified table.

Some association models assume an a priori ordering of the categories while other models reveal the ordering of the categories through esti-mation. Association models are a special case of log- linear models.

In 1979 Leo Goodman published a paper in the Journal of American Statistical Association which can be regarded as the starting point of this research field. We first present the case of a two-way table before extending to three-way and multi-way tables.

In a two-way contingency table with R rows and C columns, the cell of the i th row and the j th column (i = 1, ... I, and j = 1, ... J) gives the observed frequency f ij while F ij shall be the expected frequency estimated under


some model. In general, a log-linear model for whatever two-way table can be written as:

( )log R C RCij i j ijF =s s s s� � �

where s is the ‘main effect’, Ris the ‘row effect’, C

js the ‘column effect’, and RC

ij+s the ‘interaction effect’, on the logarithm of the expected fre-quency. All parameters in the equation are subject to ANOVA-type nor-malization constraints (see Powers and Xie 2000: 108–110). It is common to leave R

is and Cjs unconstrained and estimated non- parametrically.

Note that the terms Ris and C

js are fitted to saturate the marginal dis-tributions of the row and column variables. The focus of the analysis is on the parameter RC

ijs At the one extreme, RCijs is set to nil, resulting

in an independence model, or a model which does assume no associa-tion between rows and columns. At another extreme, RC

ijs is estimated and may be ‘saturated’, taking (I − 1) (J − 1) degrees of freedom, yielding exact predictions so that F ij = f ij for all i and j. However, typically, the researcher is interested in fitting models between these two extremes by altering specifications for RC

ijs which are more parsimonious because less degree of freedom is used.

Local odds ratios in a two-way table are function of the interaction parameter RC

ijs If we denote u ij the local log-odds-ratio for a 2 3 2 sub-table formed by four adjacent cells obtained from two adjacent row cat-egories and two adjacent column categories:

u ij = log( F ( i + 1)( j + 1) F ij )/( F ( i + 1) j F i ( j + 1) )

A simple model to estimate if the two scales are directly observed or imputed from external sources is the so-called linear-by-linear association:

( )log R Cij i j i iF x y= + + +s s s b

where b is the parameter measuring the association between the two dimensions x and y representing row and column, easily estimated by maximum likelihood estimation for log-linear models. If we do not have extra information about the two scales x and y, we can estimate the scales internally. If the categories of the variables are correctly ordered, we may make a simplifying assumption that the ordering positions form the scales, that is, x i = i, y j = j. This is the integer-scoring simplifi-cation which results in the uniform association model:

( )log R Cij i j ijF = + + +s s s b


While the uniform association model is based on integer-scoring for both dimensions (rows and columns), we may also use integer scoring for only the row or only the column variable. When integer-scoring is used only for the column variable, the resulting model is called the ‘row-effect model’. Conversely, when integer-scoring is used only for the row variable, the resulting model is called the ‘column-effect model’. Taking the row-effect model as an example, we can derive the following model from the linear-by-linear association model:

( )= + + +s s s flog R Cij i j iF j

The model is called the ‘row-effect model’ because latent scores of the row variable ( w i = b xi ) are estimated after we apply integer-scoring for the column variable. That is, w i is the ‘row effect’ on the association between the row variable and the column variable.

We are now going to take a step further with Goodman’s RC Model which treats both the row and column scores as unknown. Two of Goodman’s (1979) association models are designed to estimate such latent scores. Goodman’s Association Model I simplifies the above equa-tion ( )= + + +s s s slog R C RC

ij i j ijF to:

( )log R Cij i j i jF j i= + + + +s s s f w

Where w i and w j are respectively the unknown row and column scores as in the row-effect and column-effect models. However, it is necessary to add three normalization constraints in order to uniquely identify the (I + J) unknown parameters of w i and w j . Goodman’s Association Model I requires that the row and column variables are correctly ordered a priori , since integer-scoring is used for both. If we have no knowledge that the categories are correctly ordered this association model (Model I) is not applicable. Therefore Goodman also formulated another association model which is known as the Association Model II which has received the most attention. It is of the form:

( )log R Cij i j i jF = + + +s s s bfw

where b is the association parameter, and w i and w j are unknown scores to be estimated. The parameters w i and w j are subject to four normaliza-tion constraints, since each requires the normalization of both location


and scale. The interaction component ( RCijs ) of Goodman’s Association

Model II is in the form of multiplication of unknown parameters, and is also known as the ‘log-multiplicative model’ or simply the RC model. The RC model has remarkable advantages because it allows us to esti-mate unknown parameters even when the categories of the row and the column variables may not be correctly ordered. All that needs to be assumed is the existence of the ordinal scales. The model can reveal the orders through estimation.

Table A1 summarizes the above association models. The second column displays the model specification for the interaction parameters ( RC

ijs ). The number of degrees of freedom for each RCijs specification is given

in the third column (d.f.). If there are no other model parameters to be estimated, the degrees of freedom for a model is equal to (I−1) (J−1)-d.f. The formula for calculating the local log-odds-ratio (u ij ) is shown in the last column.

Goodman’s Association Model II (RC model) can be extended to have multiple latent dimensions and RC

ijs may be specified as the following

RCij m im jm=s b f w

where the summation applies to all possible m dimensions, while the parameters are subject to normalization constraints. This class of mod-els is called RC(M) models (see Goodman 1986).

In this chapter we are interested in understanding how the two-way association between R and C (skills and wages) depends on levels of L (countries). We therefore denote F ijk the expected frequency in the i th row, the j th column, and the k th layer. The saturated log-linear model is written as:

( )log R C L RC RL CL RCLijk i j k ij ik jk ijkF = + + + + + + +s s s s s s s s

Table A1 Comparison of association models

Model RCijs d.f. u ij

Uniform Association b ij 1 b

Row-Effect jf i (I−1) f i + 1 − f i Column-Effect iw j (J−1) w i + 1 − w i Association Model I jf i + iw j I + J-3 (f i + 1 − f i ) + (w i + 1 − w i ) Association Model II (RC) bf i w j I + J-3 ( f i + 1 − f i )( w i + 1 ) − w i


We focus in our analysis on the variation of the RC associations across layers and therefore the baseline (for the null hypothesis) is the follow-ing conditional independence model:

( )log R C L RL CLijk i j k ik jkF = + + + + +s s s s s s

which reflects our model 0 in Table 9.5 . That is to say, we thus need to specify RC

ijs and RCLijks in order to understand the association between

skills and wages within countries. There are two broad possibilities to extend the association models for a two-way table to a three-way table.

First, we can specify an association model for the typical association pattern between R and C and then estimate parameters that are spe-cific to layers or test whether they are invariant across layers (see Clogg 1982). The general case of the approach is to specify RC

ijs and RCLijks in

terms of the RC model and adapt the following model:

( )log R C L RL CLijk i j k ik jk k ik jkF = + + + + + +s s s s s s b f w

That is, the b , f and w parameters may be layer-specific or layer- invariant. We have also specified a number of other model speci-fications above like the uniform association model, column and row-effects models, adapting the formulas for a layer effect.

The second approach called the ‘log-multiplicative layer-effect model’ or ‘unidiff model’ allows a flexibility in the specification for the typical association pattern between R and C and then to constrain its cross-layer variation to be log-multiplicative (Xie 1992). Therefore we give a flexible specification for RC

ijs but constrain RCLijks so that the equation

becomes:

( )log R C L RL CL

ijk i j k ik jk k ijF = + + + + + +s s s s s s f c

The second approach allows the RC association to become saturated with (I-1)(J-1) dummy variables, without following a particular model. Powers and Xie (2000: 140–145) provide a more detailed discussion of the variations and the practical implications of the second approach.

Estimation

When the RC interaction takes the form of the product of unknown parameters – the log-multiplicative or log-bilinear specification is tricky.


In our case, a reiterative estimation procedure is required. The basic idea is to alternately treat one set of unknown parameters as known while estimating the other and to continue the iteration process until both are stabilized. For the estimation of the models in this chapter we use the software Lem, a program provided by Jeroen Vermunt, that can estimate different forms of the log-multiplicative model while retain-ing flexibility. The program is freely available from its author website at http://spitswww.uvt.nl.

Notes

1 . To identify the cells more easily, in the following we use Q for quintiles and numbers in parenthesis reflecting the row–column cell identification. Each cell is identified by Q(r,c) numbering such as Q(1,1) will be the upper left cell or the intersection of the first row and the first column of a table and Q(5,5) identifies the bottom right of each contingency table, or the intersec-tion of the fifth row and fifth column. When referring to rows or columns alone we designate them by Q1 with the prefix s or w when speaking of skill or wage levels. Wage level wQ4 means the fourth quintile or fourth column of the wage continuum while sQ4 will identify the fourth row or the fourth skill quintile.

2 . Incremental Chi square tests are computed using the explained variance divided by the degrees of freedom used: for example, between R + C UA homogeneous and heterogeneous we explain 1090–1025 = 64 of variance using 43–40 = 3 degrees of freedom. The Chi square test yields a signif-icance level of. 0000. If the more-parsimonious model would explain as much as the more-complex model (Chi square test significance > 0.05), the former model would have to be accepted.

3 . The Bayesian Information Criterion is a criterion for model selection among a class of parametric models with different numbers of parameters. When estimating model parameters using maximum likelihood estimation, it is possible to increase the likelihood by adding parameters, which may result in over-fitting. The BIC resolves this problem by introducing a penalty term for the number of parameters in the model. The formula for the BIC is: BIC = G 2 – df log(N).

4 . It has to be considered that the results and association presented here have to be taken with caution as we are only presenting results for four countries. The patterns arising from this rather qualitative assessment, however, yield some interesting results that would need further investigation.

5 . Most of this summary reflects the work of Yu Xie as described, for example, in Powers and Xie 2000.

244

10.1 Introduction

The period of employment expansion during what used to be called ‘the great moderation’ – from the early 1990s to 2007 – has been the subject of much of the analysis in this book. The focus in this chap-ter will be on the effect on the structure of employment of the ‘great recession’ initiated by the global financial crisis which started in the EU towards the end of 2007 and gathered momentum during 2008. The aim is to compare the effect of this recession both against that of other recent recessions in Europe and against the likely outcome had growth continued at its trend rate.

In common with much of the developed world, the European Union underwent a severe recession in 2008–2009. This came after a preced-ing long period of growth that stretched back to the mid-1990s with only a minor downward adjustment following the dotcom bubble and crash of 2000–2001. Between 2008 and 2010, employment declined by an average of 1.1 per cent a year in the EU (2.6% in the US). There were five million fewer people in paid employment in the EU27 in the sec-ond quarter of 2010 than in the second quarter of 2008 as a result of the economic crisis; the most severe in over a generation. This was a direct consequence of a financial crisis whose causes were various but linked: deregulated global capital markets, credit and savings gluts as well as global trade imbalances, excessive financial engineering, irresponsible

10 The Changing Structure of Employment during Periods of Recession and Recovery in the EU Donald Storrie, Terry Ward , Robert Stehrer and John Hurley

The Changing Structure of Employment 245

banking practices and inadequate bank supervision inter alia (Krugman 2008; Stiglitz 2010).

Recessions have negative effects on labour markets reducing demand for labour and increasing unemployment. Recessions borne of finan-cial crises tend to be especially severe. Reinhart and Rogoff (2009), for example, estimate that unemployment typically increases by 7 per cent after such recessions and the downtrend lasts on average for four years. Policy intervention – in the form of stimulus packages to sus-tain demand and labour market measures such as short-time working, prolonged holidays and the like – had a significant effect in tempering the impact of the crisis on jobs. Large-scale reductions in output were accompanied in many continental Member States by some decline in employment but generally less than might have been expected on past experience (Cazes and Verick 2011: 2).

While recent downturns – those in the early 1980s and the early 1990s, which affected all EU economies, and the downturn in the early part of the 2000s, which was much less general in its incidence – differ from ‘the great recession’ in many respects, most especially in terms of the initial causes, many of the differential effects on sectors and jobs can be expected to be similar, precisely because the nature of the output which was most affected was very similar.

This chapter examines the experience in previous downturns in the EU economies and the different consequences for employment they had, first, for sectors of activity because of the varying nature of the goods and services produced – for investment goods sectors , for exam-ple, which produce output whose purchase can essentially be postponed as opposed to basic consumer goods or services – and, secondly, for the different types of job within sectors. For reasons of data availability, much of this analysis is at a broad sectoral level.

It also examines the kinds of jobs which were most affected by the recent downturn and the extent, if any, to which these differed from those of similar experiences in the past. It also considers how far the change in the job structure in Europe during the Great Recession diverged from that over the preceding growth period and how far, on the contrary, it represented a continuation (or even an acceleration) of long-term trends.

The next section outlines the data sources used for analysing the employment effects of recent recessions. Section 10.2 examines the employment effects at sectoral level of the economic downturns prior to the 2008–2009 ‘great recession’ going back to the early 1980s. Section 10.3 considers patterns of occupational change within manufacturing

246 Donald Storrie, Terry Ward, Robert Stehrer and John Hurley

during periods of job loss, before drawing some lessons from previ-ous experiences in Section 10.4. The consequences of the 2008–2009 recession for the structure of employment in the EU are then described in Section 10.5, 1 while Section 10.6 describes the aggregate changes in employment in terms of relative wages – that is, by wage quintile. Section 10.7 presents the breakdowns of employment change by wage quintile in the different EU Member States and outlines a typology of patterns of employment shift. Section 10.8 decomposes the change in employment to examine shifts at the broad sectoral level – in services, construction, manufacturing and agriculture – and how they were dis-tributed across the relative wage spectrum. Section 10.9 presents some conclusions.

10.2 Data sources used

An essential requirement for undertaking such an exercise is a detailed set of data which enables the different kinds of activity to be suffi-ciently distinguished and which goes back ideally to the early 1980s – any earlier the nature of the sectors and their interrelationship with each other may be too different to provide a guide to current poten-tial developments. Fortunately, such a data set is available as a result of the work undertaken comparatively recently to construct a database for analysing productivity developments in different sectors of activity across Europe. This disaggregated sectoral database, compiled under the EU-KLEMS project, 2 contains annual data for 31 NACE rev1.1 sectors (the NACE one-digit service sectors plus 14 sectors within manufactur-ing together with agriculture and mining) on gross value-added at con-stant prices, employment and annual hours worked, as well as other variables, for each year from 1980 to 2005/2006 for all EU15 coun-tries. The data in question are from the national accounts in each of the countries concerned and are compatible with the Eurostat national accounts data, which are also disaggregated by the 31 sectors, though far more complete (the Eurostat data go back to 1980 only for a very few countries).

These data are combined with data on labour inputs and productivity from the Short-term Business Statistics database in relation to the more recent period (2008–2010). More detailed data on employment from the EU Labour Force Survey (LFS) are used to describe employment shifts at broad sector and occupational level during the Great Recession. The LFS data are also used to characterize these employment shifts in terms of job–wage quintiles using the jobs approach.


An impediment to comparing previous recessions with the most recent one is the revision of the NACE sectoral classification which became operational in 2008. Although NACE rev 2 marks a radical departure from the previous system of classification, for many of the sectors most affected by the crisis – such as construction and various manufacturing sub-branches – there is a reasonable level of comparability between the two classification systems.

The analysis begins by examining the changes in value-added, pro-ductivity and employment by sector during past periods of economic downturn in EU Member States, or more precisely in EU15 countries since for the countries which entered the EU in 2004 and 2007, the experience of economic downturn is more limited and different in character. For these countries, therefore, the main experience is that which occurred over the years immediately following the fall of the former communist regime around the end of the 1980s–beginning of the 1990s. This experience, however, was during the very initial stages of the transition to market economies and was a result largely of a collapse in trade with the Soviet Union. Accordingly, it is of limited rel-evance in assessing the recent recession.

In the EU15 countries, the focus is mainly on the economic down-turns which occurred in the early 1980s and early 1990s and to a more limited extent on that in the early part of the present decade since this was less widespread and, in most cases, less pronounced than in the earlier periods.

The variables examined are

changes in value-added at constant prices in the individual sectors; ●

changes in employment, or more accurately, in the volume of labour ●

input, as measured by total hours worked; changes in labour productivity, defined as changes in value-added ●

per hour worked, or volume of labour input; changes in average annual hours worked by those employed, in order ●

to move from the change in labour input to the number of people in work; changes in the number of people employed as an outcome of changes ●

in the preceding four variables.

In each case, the concern is to examine the changes in each sector of activity relative to the overall change in order to identify the differen-tial effect of the downturn.

This analysis is supplemented by an examination of the changes in the structure of employment by (ISCO two-digit) occupation within


sectors, though in this case the focus is on the period since the mid-1990s, because the data available from the LFS make it difficult to go back before this (largely because of changes in the ISCO system of classi-fication in the mid-1990s).

10.3 Sectoral developments during previous downturns

The concern here is to examine the changes in the variables listed above – in GDP, productivity, average hours worked and employment – in EU15 Member States over the three periods of downturn which had occurred before the present recession since 1980. For each country, the period of downturn examined relates to the years, or the year, in which the growth of total value-added in the economy in question slowed down by most or became negative. These years are in most cases the same or broadly the same, though they can differ slightly because of differences in the timing of the downturn across Member States. For example, the downturn at the beginning of the 1990s began at least a year earlier in Sweden and Finland, as well as the UK, than in most other EU countries.

The changes over these three periods of recession are compared in Table 10.1 with changes over a non-recessionary period in order to identify the differential effect of the recession on sectors and to take account of any long-term trends in their growth – or decline. These lat-ter changes (labelled ‘trend’ in the tables) which are based on changes over the second half of the 1990s and up to the downturn in 2001 (or in some countries longer because they were not affected by a down-turn) are intended to indicate the longer-term trend situation and to provide a benchmark against which the effect of the recession on the different sectors can be measured. For example, in the case of textiles and clothing (NACE rev 1.1 DB), value-added in most countries has declined since 1980 even in normal years, so the fact that value-added declined in each of the recession periods in itself is to be expected and the interesting question is whether the fall in these periods was larger than normal or not and, if so, how much larger. On the other hand, in the case of business services (NACE K), there was continuous growth in value-added throughout the period and in this case the relevant question concerns the extent to which the rate of growth during reces-sions was reduced, if at all.

The main question of interest, of course, is the effect of the develop-ments in value-added on employment and, in particular, on the number


employed in the different sectors. As outlined above, this depends not only on the extent of any fall in value-added, or slowdown in growth, but also on the extent of the change in productivity – which during periods of recession, might also differ from the normal, or trend, rate of growth (as measured by the volume of labour input – i.e. the total hours of work) and changes in the average hours worked by those employed. In other words, any reduction in labour inputs during periods of reces-sion might be compensated in some degree by a reduction in hours worked by the people employed, so diminishing the extent to which the number in work is cut back.

The first two columns of the table show the share of total employ-ment accounted for by each of the sectors in order to indicate the rela-tive importance for overall employment of any reduction. Four of the 31 sectors have been amalgamated because of their typically small size to form larger sectors – specifically, agriculture (A) and fisheries (B) and the two mining and quarrying sectors (CA and CB). In addition, two sectors, employment in private households (e.g. cleaning and garden-ing) and extra-territorial organizations have been omitted because of the small numbers employed in most countries and a lack of data.

Table 10.1 for the EU15 indicates, first, that the composition of employment over the period 1980–2005 changed significantly in a number of respects – in particular, that the importance of agricul-ture for jobs declined markedly (from accounting for 9% of employ-ment to accounting for 4%), that there was a decline in the share of employment in all manufacturing sectors, most notably in textiles and clothing, and that by far the biggest increase in employment occurred in business services (from 5% of the total to 13%). (The shaded rows denote sectors in which the share of employment in 2005 was especially small – under 1%.)

Secondly, it shows that the effect of the downturn in the early 1980s was particularly severe, total value-added falling by just over 2 per cent in 1981. The sectors affected by this were mostly in manufacturing, with value-added in glass and non-metallic mineral products falling by over 7 per cent and in metal manufacture by just under 7 per cent. At the same time, value-added in construction declined by just over 4 per cent and in the distributive trades by just over 1 per cent (as against trend growth of around 3%). On the other hand, value-added in both busi-ness services and public administration expanded by over 3 per cent a year, in the latter, by well over the trend increase (this is the case even if account is taken of the fact that the upward trend in the 1980s was more than in the 1990s – at around 1–1.5% a year).


Table 10.1 Employment structures and sectoral in the EU15, 1980–2007

% total employed

Change in value-added Change in labour input

(total hours worked)

Recession years Trend Recession years Trend

Sector 1980 2005 1980s 1990s 2000s 1990s 1980s 1990s 2000s 1990s

AB 9.2 3.8 –0.4 0.5 –1.7 1.1 –3.3 –4.7 –3.2 –1.5 C 0.8 0.2 –2.9 2.3 0.1 0.2 –4.6 –10.6 –2.6 –3.4 DA 2.9 2.1 0.2 –0.5 0.5 0.6 –2.9 –1.5 –0.4 0.0 DB 3.0 1.0 –3.9 –3.1 –4.4 –0.7 –6.8 –6.7 –5.7 –2.5 DC 0.6 0.2 1.8 –5.5 –6.6 –2.2 –5.1 –5.2 –4.9 –2.6 DD 0.8 0.5 –6.0 –2.7 1.3 2.9 –4.6 –2.5 –3.3 –0.4 DE 1.9 1.3 –3.1 0.1 0.2 2.1 –2.4 –1.9 –2.1 –0.3 DF 0.2 0.1 –8.4 –20.5 –6.7 –3.1 –1.9 –4.1 0.2 –2.5 DG 1.5 0.9 0.0 1.5 3.6 4.3 –3.8 –4.6 –1.4 –0.7 DH 0.9 0.8 –1.8 0.7 2.2 4.4 –4.9 –2.8 –2.3 1.1 DI 1.3 0.7 –7.3 –3.1 –1.2 2.1 –5.3 –4.2 –2.9 –0.2 DJ 3.8 2.4 –6.7 –2.4 –0.7 2.6 –5.5 –4.5 –1.5 0.3 DK 2.7 1.8 –4.0 –5.0 –1.0 1.5 –4.2 –5.4 –2.0 0.0 DL 2.7 1.7 1.1 –1.1 12.1 10.2 –4.3 –5.4 –5.7 0.4 DM 2.4 1.5 1.6 –5.2 3.1 5.2 –5.9 –6.2 –2.1 0.8 DN 1.3 0.9 –3.1 –2.5 –3.7 2.3 –4.1 –2.3 –2.9 0.3 E 0.9 0.6 0.5 1.5 0.6 0.9 –0.5 –1.7 –2.1 –2.5 F 8.3 7.3 –4.3 –1.9 0.0 0.9 –3.9 –2.3 –0.1 1.5 G 14.3 15.1 –1.2 1.4 1.5 3.0 –0.1 –0.7 –0.2 0.8 H 3.1 4.9 1.3 –0.9 –0.9 3.0 1.2 0.3 2.1 2.1 I 6.0 5.7 1.9 1.0 2.1 4.6 –.8 –0.8 –0.3 0.9 J 2.7 2.9 1.3 1.4 3.3 2.9 1.1 –0.4 –0.6 0.5 K 5.1 13.0 3.3 2.3 1.0 3.0 0.7 2.0 1.6 4.6 L 7.1 6.7 3.8 1.1 0.8 0.5 0.4 0.4 –0.5 –0.5 M 5.6 6.7 2.3 1.2 1.4 0.8 0.1 0.9 2.6 1.1 N 6.4 9.8 2.4 1.6 2.2 1.8 0.8 1.6 2.1 1.2 O 3.1 4.9 1.0 1.8 0.4 1.7 0.9 1.1 1.3 1.9 Total –2.2 0.5 1.4 2.6 –1.8 –1.2 0.0 1.0

AB Agriculture DF Petroleum refining

C Mining DG Chemicals and pharmaceuticals DA Food, drink, tobacco DH Rubber and plastics DB Clothing and textiles DI Glass and non-metallic mineral prods DC Leather and footwear DJ Metal manufacture DD Wood and wood products DK Machinery and equipment DE Paper, pulp, printing DL Electrical and electronic equipment

Source : EU KLEMs database.


Change in labour productivity

Change in average hours worked

Change in number employed

Recession years Trend Recession years Trend Recession years Trend

1980s 1990s 2000s 1990s 1980s 1990s 2000s 1990s 1980s 1990s 2000s 1990s

3.0 5.4 1.5 2.7 –0.3 0.3 –0.6 –0.2 –3.0 –5.0 –2.6 –1.3 1.8 14.5 2.7 3.7 –2.3 0.2 –1.6 0.2 –2.4 –10.8 –0.9 –3.6 3.1 1.0 1.0 0.6 –0.7 –0.1 –0.8 –0.6 –2.2 –1.4 0.4 0.6 3.0 3.9 1.3 1.9 –0.5 –0.2 –0.8 –0.1 –6.3 –6.5 –4.9 –2.4 7.2 –0.3 –1.7 0.4 0.0 –0.2 –1.6 –0.2 –5.1 –5.0 –3.4 –2.4

–1.4 –0.3 4.8 3.3 –1.0 –0.2 –1.7 0.0 –3.6 –2.2 –1.6 –0.3 –0.8 2.1 2.3 2.4 –0.3 –0.2 –0.5 0.3 –2.0 –1.7 –1.6 –0.6 –6.6 –17.1 –6.9 –0.6 –0.5 –0.8 1.0 –0.7 –1.4 –3.3 –0.8 –1.8

3.9 6.4 5.0 5.0 –0.4 –0.5 –0.5 –0.2 –3.4 –4.1 –0.8 –0.5 3.3 3.7 4.6 3.3 –1.0 –0.5 –0.6 –0.2 –3.9 –2.3 –1.7 1.3

–2.2 1.2 1.8 2.3 –1.2 –0.2 –0.8 –0.3 –4.1 –4.0 –2.1 0.0 –1.2 2.2 0.8 2.3 –1.1 –0.6 –0.5 0.0 –4.4 –3.9 –1.1 0.4 0.2 0.5 1.0 1.5 –1.5 –0.3 –0.4 –0.3 –2.8 –5.1 –1.6 0.3 5.7 4.6 18.9 9.8 –1.1 –0.1 –1.1 –0.2 –3.3 –5.4 –4.6 0.6 8.0 1.1 5.3 4.4 –1.3 –1.1 –0.7 –0.7 –4.7 –5.1 –1.4 1.4 1.0 –0.2 –0.8 2.0 –0.5 –0.1 –0.4 –0.3 –3.7 –2.2 –2.5 0.6 1.0 3.3 2.8 3.5 –1.4 0.2 –0.6 –0.6 1.0 –1.9 –1.5 –1.9

–0.5 0.4 0.1 –0.6 –1.2 –0.5 –0.4 0.1 –2.7 –1.8 0.3 1.4 –1.0 2.1 1.7 2.1 –0.2 –0.6 –0.7 –0.5 0.0 –0.1 0.6 1.3 0.1 –1.2 –3.0 0.9 –0.3 –0.8 0.0 –0.5 1.5 1.1 2.1 2.6 2.7 1.8 2.4 3.7 –0.3 –0.2 –0.4 –0.6 –0.6 –0.6 0.1 1.5 0.2 1.8 4.0 2.4 –0.9 –0.5 –0.5 –0.2 1.9 0.1 –0.2 0.7 2.5 0.3 –0.6 –1.5 –0.9 –0.4 –0.1 –0.5 1.7 2.5 1.8 5.1 3.4 0.8 1.3 0.9 –1.4 0.1 –0.5 –0.4 1.8 0.3 0.0 0.0 2.2 0.3 –1.2 –0.3 –1.6 0.2 0.2 –0.1 1.7 0.7 2.4 1.2 1.5 –0.1 0.0 0.6 –1.8 –0.5 –0.3 –0.5 2.6 2.2 2.5 1.7 0.2 0.7 –0.8 –0.3 –0.7 –0.4 –0.4 –0.4 1.5 1.5 1.6 24

–0.4 1.7 1.4 1.6 –0.9 –0.4 –0.5 –0.4 –0.9 –0.8 0.6 1.4

DM Motor vehicles and transport equip. J Financial services

DN Furniture and other manufactures K Business services E Electricity, gas, water L Public administration F Construction M Education G Retail and wholesale distribution N Health and social services H Hotels and restaurants O Personal and community services I Transport


The decline in value-added was accompanied by a slightly smaller fall in the volume of labour input, so that labour productivity also fell a lit-tle in 1981, thus cushioning the effect on employment. The reduction in productivity seems to have been concentrated in the sectors show-ing the biggest falls in value-added – in glass and non-metallic mineral products, metal manufacture and construction, especially, as well as in the distributive trades – though the reduction was not enough to com-pensate for the fall in value-added and the volume of labour inputs still declined significantly.

In other manufacturing sectors, productivity tended to increase, especially in the electrical engineering industry and motor vehicles, so that the volume of labour input declined sharply in all manufacturing industries (by 4–6% in most cases). Productivity also increased in most service sectors, though by not enough, except in transport, to offset the positive effect of the growth in value-added on labour input.

The depressing effect of the reduction in value-added on the number employed was further cushioned by a decline in average hours worked during the year by those employed, of almost 1 per cent overall, and by more than 1 per cent in most parts of manufacturing showing the biggest falls in labour input (the exception being textiles and clothing). Average hours worked also declined markedly in the public sector, in public administration, education and healthcare.

The number employed, therefore, fell by only half the decline in the volume of labour input during the year because of the reduction in average working time, though, nevertheless, still by almost 1 per cent and by between 3 and 5 per cent in most manufacturing industries and by over 6 per cent in textiles and clothing. This fall was partly offset by a rise in the number employed in service sectors, with the number fall-ing only in transport; in the distributive trades the number employed remained unchanged, in contrast to a trend increase of over 1 per cent a year.

In the subsequent recession in the early 1990s, the growth of val-ue-added in the EU15 as a whole slowed down in 1990–1992 and fell slightly in 1993. In these three years, therefore, growth averaged only 0.5 per cent a year. The differential effect of this on different sectors was similar to that in the previous recession, though in this case motor vehi-cles did not escape a decline in value-added. The decline in value-added in construction was smaller in line with the more-modest scale of the downturn, as was the reduction below the trend rate of growth in the distributive trades (in this case, by not enough to cause a fall). During this period, however, hotels and restaurants were hit much harder than


in the earlier period, value-added falling by around 1 per cent. The other service sectors continued to expand if in most cases by less than the trend rate of growth.

Despite the generally shallower downturn than ten years before, the reduction of labour input was only slightly less, at just over 1 per cent a year, as a result of productivity continuing to increase (which might reflect the longer duration of the downturn). The decline in labour input was especially marked in manufacturing (nearly as great as in the preceding period), though it was also significant in both transport (as previously) and the distributive trades (which unlike in the previous period showed an increase in productivity), while there was a decline as well in financial services. In hotels and restaurants, however, produc-tivity fell and labour input rose slightly.

Although average hours worked declined, the extent was similar to that in trend periods of growth, and only around half as much as in the earlier period. This was equally true in most sectors, especially in manufacturing and, in consequence, the reduction in the num-ber employed was much the same overall in the economy as a whole (just under 1% a year) and larger in manufacturing, where the number employed in the engineering industries fell in each case by over 5 per cent a year. It was still the case, however, that in services the only sector to show a significant decline in the number in work was transport.

In the much shallower downturn in the early part of the present decade, when a number of EU15 Member States experienced hardly any slowdown at all, the effect on the different sectors was much the same as before only generally smaller with motor vehicles manufacturing escaping any significant effect. The exceptions were textiles and cloth-ing, furniture and other manufactures and hotels and restaurants, in all of which the fall in value-added was as large as or larger than in the early 1990s.

Overall, there was a slight slowdown in productivity growth which was enough to prevent labour input from being reduced. Labour input, however, declined in all manufacturing industries, especially in elec-trical and electronic equipment where the high growth of value-added was accompanied by an even larger increase in productivity. Labour input also declined in construction, marginally, the distributive trades, transport and, more significantly, in financial services, as well as in public administration.

This decline in labour input was offset in part by a reduction in aver-age hours worked, which was on much the same scale as in the preced-ing years but which was accordingly responsible for all the increase in


the number employed which occurred over the period (0.6% a year). The reduction in average hours was especially marked in manufactur-ing, in electrical and electronics equipment, in particular, and served to moderate the fall in the number employed, which was still close to 5 per cent a year in this industry as well as in textiles and clothing. The only service sector to show a decline in the number employed was financial services, though there was only a marginal increase in transport and no increase at all in public administration.

In comparison to trend growth, the recessions of the 1980s to early 2000s tended to have a stronger differential impact on manufactur-ing than service-sector employment. The more severe recessions of the early 1980s and 1990s led to annual average declines of employment of between 3 and 4 percentage points greater than trend in several such sectors. For example, the employment decline in clothing and textiles manufacture – already contracting throughout the 1990s at a rate of around 2.5 per cent a year – accelerated to a rate of decline of around 6.5 per cent in recession periods. Manufacturing sectors that enjoyed trend growth of jobs in the 1990s – such as motor vehicle manufacture recorded even bigger declines in employment compared to trend. The net effect of the 1980s/1990s recessions on employment in this sector was to turn trend employment growth of 11.5 per cent a year into a decline of around 5 per cent a year. Employment in other capital pro-ducing manufacturing sectors (electrical equipment, machinery and equipment) also tended to be affected disproportionately compared to basic manufacturing activities such as food, drink and tobacco produc-tion and paper, pulp production where net job losses (i.e. compared to trend) were less marked. In large part, this reflects the structure of demand for the output of basic manufacturing industries as compared with capital producing ones, which makes, for example, the food and drink industry less vulnerable to a reduction in demand than invest-ment goods industries such as cars.

The impact of recessions on employment in the construction sector tended to be somewhat smaller than in manufacturing as a whole (job losses of 2–3% a year as compared with trend growth of around 1.5% a year in the 1990s).

Employment in service sectors, on the other hand, was generally much less affected. The majority of service activities continued to show employment growth even during recession years. The two exceptions – retailing and transport – had only marginal employment losses (less than 1% a year) during periods of recession and even in these sectors the reduction in employment compared to trend growth was only around


2 percentage points a year as against 6 percentage points a year for the manufacturing industries most affected.

Business services was the sector that experienced the fastest employment growth between 1980 and 2005 (5% a year), though its job growth was reduced significantly during downturns (by around 2 percentage points a year). Employment in (mostly) state-funded service sectors, on the other hand, tended to compensate to some extent for job losses elsewhere. Employment growth in education, public administration and, most especially, in healthcare was greater during periods of recession than in periods of economic growth. Employment in healthcare increased by around 2.5 per cent a year during recession years as compared with trend growth of 1.7 per cent a year. This in some degree reflects government efforts to counter the downturn in economic activity through increased public expendi-ture and was also apparent during the great recession of 2008–2009 as indicated below.

10.4 Changes in the structure of occupations during previous downturns

As indicated above, no comparable set of data on occupations exists at the EU level for the years before the mid-1990s, so it is not possi-ble to examine developments in the structure of occupations within sectors during periods of downturn before then. The only data avail-able come from EU Labour Force Survey from the mid-1990s onwards. What emerges from an examination of these data is that there has been a fairly uniform shift in most sectors of activity from lower lev-els of occupations to higher-level ones, or, in other words, from those not requiring high levels of education, though perhaps extended voca-tional training, to those for which educational attainment is, in most cases, essential. The question here, however, concerns the tendency, if any, for the structure of occupations to change over the economic cycle – whether, for example, the relative number of people employed in higher-level occupations tends to increase or decline as economic activity falls.

This question is particularly relevant in respect of manufacturing, which, as indicated above, is more susceptible to being affected by eco-nomic downturns than services. To examine this question without investigating in detail changes in the occupational structure in each of the industries distinguished above, manufacturing can be divided into three groups of industries according both to their technical


characteristics and, related to this, to the structure of occupations within them. The three groups in question are

basic manufacturing industries, which are those such as food and ●

drink, textiles and clothing, metal manufacture, wood and furniture, in which skilled and semi-skilled manual workers tend to account for 60–70 per cent of employment and managers and professionals for around 20 per cent; processing industries, which are chemicals and pharmaceuticals and ●

pulp and paper, together with electrical and electronic equipment in which the occupational structure is similar, with skilled and semi-skilled workers representing 30–40 per cent of the workforce and managers and professionals, 40–50 per cent; engineering industries, which in this case are machinery and ●

equipment and motor vehicles and other transport equipment, in which managers and professional represent around 35 per cent of employment and skilled and semi-skilled manual workers around 50 per cent.

In the case of the basic industry group, there is some sign in the EU15 as a whole of the share of managers and professionals (engineers, account-ants, marketing managers and so on) increasing as total employment declined from 2001 onwards, matched by a reduction in the share of both skilled (such as toolmakers or mechanics) and semi-skilled workers (plant and machine operators and assemblers) ( Figure 10.1 ). Much the same is the case in the processing industries, where the share of skilled

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Percentage of total employed

Managers/professionals

Plant and machine operators

Clerks/salesmen

Elementary

Crafts/tradeworkers

Total employed, index

Figure 10.1 Changes in the occupational structure of employment in basic industries, EU15


and semi-skilled manual workers in employment declined closely in line with each other as the share of managers and professionals rose ( Figure 10.2 ).

In the engineering industries, the decline in employment after 2001 was less than in other parts of manufacturing, but a similar increase in the share of managers and professionals is evident. In this case, the counterpart decline is largely concentrated among skilled manual work-ers ( Figure 10.3 ). In all three cases, however, it is difficult to disentangle the effect of the downturn from long-term trends.

It is worth noting that there is much less sign of any effect of a decline in employment (or indeed of the long-term trend for higher-level

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

40.0

45.0

1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007



Clerks/salesmen

Elementary

Crafts/tradeworkers



Figure 10.3 Changes in the occupational structure of employment in engineer-ing industries, EU15

0.05.0

10.015.020.025.0

30.035.040.0

45.050.0

1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007



Clerks/salesmen

Elementary

Crafts/tradeworkers



Figure 10.2 Changes in the occupational structure of employment in process-ing industries, EU15


occupations to increase) on the share of unskilled manual workers (labourers, cleaners and so on), which in all three groups changed rela-tively little over the period.

It seems, therefore, that insofar as a reduction in employment – and, by inference, a downturn in economic activity – affects the occupa-tional structure of the workforce in manufacturing, it is the more skilled manual workers who tend to lose out rather than the least skilled.

10.5 Some lessons from previous recessions

Experience during the downturns which have occurred in the EU15 countries over the relatively recent past suggests that some sectors are affected much more than others. In particular, on past experience, value-added in investment goods industries and construction declines more than in the rest of the economy. How far, however, this decline is translated into a reduction in the number employed depends on what happens to labour productivity – whether the upward trend reflecting advances in technology and know-how is maintained or is moderated as output falls – and average working time. Both, in turn, depend on the reaction of employers to a fall in sales and the extent to which they attempt, and have the financial capacity, to keep their workforce intact and to avoid shedding jobs – or, in the cases of those countries with established short-time working schemes, such as Germany, Belgium and others (see Mandl et al. 2010), that governments can continue to subsidize companies who maintain people in employment on reduced hours of work rather than making them redundant.

Past experience also suggests that the behaviour of labour productiv-ity and average hours worked and, accordingly, the effect of the down-turn on employment, are likely to be influenced both by the scale of the downturn and its duration, or perhaps more relevantly its expected duration. In both Finland and Sweden, therefore, where the scale of the fall in output in the early 1990s was greater than in other EU15 Member States and where the downturn was more persistent, labour productiv-ity increased over the period at much the same rate as in more normal periods while the average working time of those in employment either increased or fell only slightly.

In a number of other countries, on the other hand, such as, in partic-ular, Italy in the early part of the 2000s and Belgium, the Netherlands, Austria and Portugal in the early 1990s, growth of labour productiv-ity declined markedly during the recession years, while in France in the early part of the present decade, as well as in the early 1980s and


in the UK, in the early 1990s, average hours worked were reduced. In each case, the effect was either to moderate the scale of job losses or to maintain the growth in employment, even if at a slower rate than before.

10.6 Changes in the employment structure in the EU during the 2008–2009 recession

Developments in the 2008–2009 recession are in line with the relative changes in the different sectors indicated above. This is particularly the case as regards production, or value-added, which shows especially large reductions in the engineering industries and other producer goods industries, such as basic metals.

In the EU15 as a whole, therefore, according to the short-term business statistics compiled by Eurostat, 3 production in the basic metals industry (iron and steel) fell by 35 per cent between the second quarter of 2008 and the second quarter of 2009, in motor vehicles by 33 per cent and in machinery and equipment by almost 30 per cent, while in electrical equipment, it declined by 26 per cent. In most of the service sectors, on the other hand, the reduction – in this case in turnover – was less than 10 per cent, the main exception being retailing and wholesaling, where the decline was just over 11 per cent.

These falls in production, however, were not reflected in a decline in employment on anywhere the same scale in most cases. Only in wood and wood products and textiles was the decline in employment over 10 per cent and while it was relatively large in motor vehicles (just over 7%), electrical engineering (7%) and basic metals (just over 6%), it was still considerably smaller than the fall in production. By implication, employers in these industries absorbed much of the decline in output by effectively assuming a large reduction in productivity, which in practice was accompanied by the extensive use of short-time working in a number of countries. By contrast, in construction, the decline in employment (just over 8%) was on much the same scale as the reduc-tion in production, to a large extent reflecting the concentration of this reduction in Spain, where the fall in output was quickly followed by job losses.

In what follows, we trace the employment impacts of the Great Recession using the jobs approach, covering the period 2008q2–2010q2. Technically, recession in the EU27 began in 2008q2 and ended in 2009q3 after five quarters of consecutive negative growth. In this chap-ter we opt for the two-year timeframe, 2008q2–2010q2, for analysis as


it covers the period from before the collapse of Lehman Brothers bank (September 2008), seen by many as the trigger for the global crisis, to just after the stabilization of EU27 unemployment rates at around 9.5 per cent. There are other justifications for opting for this timeframe. Employment declines in recession typically lag output declines by two- to three-quarters which justifies extending the period to beyond that of quarter-on-quarter output declines. In addition, since the focus here is on changes in employment levels – which are not seasonally adjusted – it makes sense to select the same quarter for comparison to minimize seasonal effects.

The geographical coverage is extended to the EU27, that is, includ-ing the EU12 countries which have entered the EU since 2004, which were excluded from the foregoing analysis of the employment effects of previous economic downturns partly because of a lack of data but also because of the very different circumstances which prevailed in these countries at the time.

This section therefore presents a broad picture of employment shifts in the European Union between the second quarters of 2008 and 2010 when the number employed declined by just over five million. A ques-tion of particular interest is whether the change in the job structure in the EU during the Great Recession was different not only from the experience of previous downturns but also from that of the growth years leading up to it or whether, on the other hand, it represented an acceleration of previously observed structural trends. The answer seems broadly to be the latter as the analysis below shows.

The impact of the recession varied considerably between EU Member States. Employment began to decline as early as 2007q3 in some countries (Romania, Hungary, Spain, Ireland and Lithuania) and as late as 2008q4 in others (Cyprus, Germany, Poland, Netherlands and the Czech Republic). There were as many as twelve consecutive quarters of employment decline in Lithuania and as few as two or three in the case of Austria, Luxembourg and Cyprus. In broad terms, those countries where employment dipped earliest were also those with the largest peak-to-trough reductions over the extended period.

Some countries emerged comparatively unaffected – Belgium, for example, where employment has barely declined, and many of the larger Member States including Germany, where short-time working measures have been particularly important, and Poland, where there was no fall in GDP – while others such as Spain, Ireland and the Baltic States have expe-rienced large-scale increases in unemployment. Peak-to-trough declines


in employment were over 15 per cent in each of the Baltic countries and over 10 per cent in Ireland, Bulgaria and Spain.

The impact has also varied in terms of the effect on different cate-gories of workers. Men under 25, those with low-education levels or on temporary contracts have been most affected. Those in higher-skilled occupations, especially experienced, older workers have largely escaped being affected by the crisis so far, though an exception has been a decline in employment among corporate and general managers (ISCO 12 and 13). In sum, the recession has been markedly skill-biased, sparing in large part – at least in the period covered by this analysis – higher-skilled, non-manual workers.

Employment in the predominantly state-funded sector – public administration, education and healthcare – actually increased for non-manual workers while reductions in employment have been relatively modest in private services with the exception of retailing ( Table 10.2 ). In every sector except agriculture – which accounts for little over 5 per cent of total employment – employment has fallen by more or risen by less for lower-skilled than for higher-skilled workers and for manual than non-manual ones. The number of manual workers – both low and

Table 10.2 Changes in employment (%), EU27, by broad sector and occupa-tional groupings, 2008q2–2010q2

Sector/occupation

Non-manual Manual

Total High skill Low skill High skill Low skill

Primary sector –7.7 –2.6 –0.9 2.9 –0.7 Construction –3.6 –10.5 –11.2 –16.4 –10.7 Manufacturing –6.9 –7.8 –10.1 –14.0 –10.2 Retail –2.7 –3.4 –5.8 –6.3 –3.7 Other private services 1.6 –0.8 –2.0 –0.9 0.1 Public services and utilities

4.3 3.0 –3.7 –0.6 3.1

Total 0.9 –1.0 –7.3 –6.0 –2.3

Notes : Occupational breakdown: non-manual high skilled = ISCO main groups 1–3 (legislators, managers, (assoc) professionals). Non-manual low skilled = ISCO 4–5 (clerks, service workers). Manual high skilled = ISCO 7 (craft workers). Manual low-skilled = ISCO 6, 8–9 agricultural workers, plant operators, elementary occupations. Sector breakdown: Primary sector = NACE A-B. Manufacturing = NACE C. Construction = NACE F. Retail = NACE G. Other private services = NACE H-N, R-U. Public services and utilities = NACE E-F, O-Q.

Source : ELFS (authors’ calculations).


high skilled – especially in construction and manufacturing – fell by most during the recession.

The contraction in manual employment during 2008–2010 shows some similarities but also some differences as compared with those recorded during the preceding period of employment growth. First, con-struction accounted for a large share of the decline in manual jobs during the Great Recession after enjoying significant, in some cases very high, growth beforehand. The bursting of property bubbles and construction booms were responsible for much of the rise in unemployment in the Baltic States, Ireland and Spain and around a third of the overall net reduction in employment. Secondly, manual job losses in manufactur-ing fell more heavily on low-skilled workers than on high-skilled ones which represent a departure from the experience of earlier recessions. Labour hoarding by employers may have been greater during the Great Recession with employers tending to maintain the jobs of those likely to be the most difficult to recruit once the recovery came. The differen-tial impact of short-time working schemes may also have contributed as these tend to benefit a higher proportion of those in higher-skilled occupations such as craft and trades workers than those in elementary occupations (Mandl et al. 2010: 41). Thirdly, however, high-skilled man-ual workers were the single most affected broad occupational group in terms of employment decline (-7.3%), since job losses occurred across all sectors, in public and private services as well as in manufacturing.

The recession therefore appears to have accentuated the long-run shift of employment away from primary sector (agriculture, fishing and extractive industries) and manufacturing activities towards service activities; these accounting for 69 per cent of EU27 employment in 2008 and 71 per cent in 2010. How far, however, this represents a permanent shift rather than the short-term effect of the downturn in activity being concentrated on the former two sectors remains open to question.

In what follows, the overall EU trends are examined before turning to the varying patterns of change in the individual Member States. Employment change is then decomposed into its components in terms of major sectoral aggregations to see how these have been affected by the recession.

10.6.1 Aggregate EU employment developments by wage quintile during the great recession

In Figure 10.4 , the distribution of the actual decline in employment in the EU across wage quintiles 4 between quarter two of 2008 and quar-ter two of 2010 is compared with two hypothetical trend scenarios.


The black bar shows what actually happened, that is a very significant employment decline concentrated in medium- and lower-medium pay-ing jobs. The hollow white bar shows trend growth without the reces-sion and the grey bars show how actual job losses would have been distributed across quintiles assuming the same patterns of relative change observed in the pre-recession period.

Despite the recession, jobs in the top quintile actually increased by nearly 1 per cent per year over this period. These findings are compara-ble to the results for 1995–2007 presented in Chapter 3 , which showed employment growth firmly skewed towards higher-paid jobs. Higher-paid and higher-skilled jobs were much more resilient to the effects of the recession than lower-paid jobs – and those types of jobs were also the main beneficiaries of employment growth during the long preced-ing period of EU job expansion. Those in low-paid jobs in the bottom quintile, that experienced only modest losses, were also less affected.

Employment change during the crisis in the EU as a whole can be characterized as polarized with some upgrading. This compares to the overall picture from the earlier period, which showed more distinct upgrading – employment growth in the top two quintiles accounted for over two-thirds of the overall increase – and some polarization, in that

0–2

,000

2,00

0

EU27 08q2–10q2 EU27 predictedEU27 trend

Figure 10.4 EU employment level shifts (,000s) 2008q2–2010q2 – actual, trend and predicted

Source : ELFS (authors’ calculations), European Jobs Project dataset.


growth was lowest in the middle of the wage distribution and some-what higher in the lowest-paid jobs. The recession, therefore, adversely affected employment both by destroying (low)-medium-paid jobs and by stemming the net creation of higher-paid jobs which had been the hallmark of the long pre-crisis expansion.

Figure 10.4 also makes clear that the recession has contributed to ‘hollowing out’ the labour market by affecting disproportionately jobs in the middle of the wage distribution. This is a common finding of previous analyses in both the US and the EU using a similar method-ology (Wright and Dwyer 2003; Goos and Manning 2007 for the UK; Fernández-Macías 2010 for the EU as a whole), though these related more to periods of employment expansion than of employment decline. In those expansion periods, employment trends are positive in jobs at the top and bottom of the wage distribution and much less positive or negative in the middle, giving rise to an overall polarization of jobs as the ‘middle disappears’. It is interesting that these trends, previously observed during growth periods, should also be evident during a severe downturn.

The trend growth bars (hollow) show how net employment shifts would have been distributed across the quintiles had each quintile continued to experience the same annual rate of growth as during the period 1998–2007. 5 The predicted growth bars (grey) show how the net employment decline between 2008q2 and 2010q2 – approximately five million jobs – would have been distributed across quintiles assuming the same average relative annual shifts between quintiles as experienced during 1998–2007.

Assuming that employment growth had continued on trend, over six million net new jobs would have been created between 2008 and 2010. Viewed from this perspective, the net decline actually experienced rep-resents a shortfall of over 11 million jobs in relation to what would have been expected had not the recession occurred. Even on the most opti-mistic forecasts, it is likely to take some time to make good this deficit.

The second hypothetical scenario (characterized by the grey bars) shows that the 2008–2009 recession has resulted in a sharper polari-zation of employment than would have been predicted had the same rate of shifts of employment between quintiles occurred as during the previous decade of employment expansion. As noted above, the broad shape of both the predicted and actual employment shifts is polarized, though skewed towards higher-paid jobs. There is greater job loss in the medium- and low-medium paid jobs, somewhat smaller job loss in low-paid and medium-high paid jobs and the best relative performance


in the highest-paid jobs. However, the employment loss in the low-(medium) paid jobs was larger than predicted while the relative per-formance of both top and bottom quintiles was more positive than predicted.

The great recession, therefore, both witnessed accelerated negative employment shifts and led to an accentuation of the polarization of these shifts. The former reflects the scale of the economic downturn, the latter its concentration on construction and manufacturing indus-tries, especially those in which the output can essentially be postponed being made up to a large extent of investment goods. Some of this job loss may prove transitory in the sense that the sectors in which the reduction in output was concentrated are likely on past experience to be also those that gain most as the recovery takes place. Some of the investment at least, which has been postponed, should occur as demand picks up. Conversely, the resilience and growth of public sec-tor employment – especially in higher-grade occupations in the health-care and education sectors – is also likely to have been to some extent recession-specific as governments across Europe attempted to counter the effects of the recession. This too may be reversed during the recov-ery phase. On past experience of economic downturns over the past 30 years, it is to be expected that growth in public sector employment would gradually slow down. This time, however, the severity of the recession coupled with the countercyclical measures taken have left public finances in most EU countries in a bad state and virtually uni-versal efforts are being made at the time of writing to improve them by cutting public expenditure and, in many cases, reducing public sector employment, especially in public administration. These could have a pronounced effect on the pattern of employment and the structure of jobs over the medium and longer term.

10.7 Developments at Member State level

Figure 10.5 presents net employment changes in 2008q2–2010q2 by wage quintile for all Member States as well as the EU27 overall. In vir-tually all countries, quintiles showing employment losses outnumber those showing gains, with those countries where the recession hit less hard – such as Germany, Poland and the Benelux countries – tending to have countervailing growth, especially in the highest-paid jobs.

Countries are ordered in Figure 10.5 from top to bottom in terms of the relative size of employment decline during the recession. Those countries that suffered the biggest declines are at the top; those with the


smallest declines at the bottom. In addition, neighbouring countries are grouped together so far as possible, with the larger Member States on the right-hand side. The figure illustrates the large variation between countries in the size of the impact of the recession on employment. The six countries that recorded peak-to-trough reductions in employment of 10 per cent or more each have a concentration of job loss in low-skilled but medium-paid jobs.

This pattern is attributable in large part to the plight of construction jobs in countries where booms in construction collapsed from 2007 onwards. The construction-led volatility of employment in Ireland and Spain in particular is remarkable. Both countries experienced similarly frenetic employment growth in the decade preceding the crisis. The number employed in Ireland almost doubled between 1992 and 2007 while Spain accounted for over one in three net additional jobs in the EU15 between 1995 and 2006 (Fernández-Macías and Hurley 2008: 14). Much of the growth was in an overheated construction sector. The col-lapse has been even more dramatic than what preceded it. In Spain alone, a million construction jobs were lost between the beginning of 2008 and 2010. Construction-sector employment fell by 40–60 per cent in the countries along the top row of Figure 10.5 in just two years. The bursting of national property bubbles has had wider repercussions, but it was the single most important factor behind declining employment in the countries where the recession struck hardest.

In the analysis covering the EU23 during the period 1995–2007, three main patterns of employment growth were identified at national level – polarized growth, upgrading and growth in the middle – with two fur-ther hybrid or mixed categories. Labelling different national growth patterns in this way makes sense over a 12-year period as the change can be considered in large part as structural. The short time frame of the current analysis – two years – and the fact that it is self-evidently an exceptional period of job destruction may make a repetition of the same exercise somewhat artificial. Nonetheless, with this and other caveats, 6 it is interesting to see the extent to which the patterns identified earlier still apply. The definition of employment polarization used here is adjusted to reflect changed circumstances: it refers to any country where job destruc-tion has been especially concentrated in medium-paid jobs. Table 10.3 summarizes the pre- and post-crisis patterns of employment growth.

The first point to make is that the list of countries with polarizing employment change has lengthened. Two of the original countries – Cyprus and France – have been joined by eight EU12 Member States in this group. The list is affected in particular by the decline in employment

–15

–10–50

LV

–15

–10–50

EE

–15

–10–50

IE

–6–4–20

BG

–15

–10–50

ES

–10–50

LT

–505

DK

–4–202

FI

–2–101

–50

–6–202

SK

––4

HU

–10–505

SI

5

RO

–2012

SE

–1

–402 –2

–100510 –5

LUC

ZG

R

–6–4–20

PT

–3–101

IT

–2

–202

NL

–4–202

AT4

–2024

PL

–50510

UK

FR

DE

–2

BE

–4024

CY

–2

MT

EU

–1005 –5 –6–202 –4

–2012 –1–2012 –1

–204

–3–101 –2

Figu

re 1

0.5

An

nu

al e

mp

loym

ent

chan

ge (

%)

by w

age

qu

inti

le E

U27

, 20

08q

2–2

010

q2

Not

e : Q

uin

tile

s b

ased

on

nat

ion

al jo

b–w

age

ran

kin

gs f

or B

E, D

K, E

E, F

R, G

R, H

U, I

T, L

T, L

U, L

V, P

L, P

T, U

K

and

co

mm

on E

U j

ob

–wag

e ra

nki

ng

for

all

oth

er M

emb

er S

tate

s (c

alcu

late

d a

s th

e w

eigh

ted

mea

n o

f th

e st

and

ard

ized

ran

kin

gs o

f th

e ab

ove

13 c

ou

ntr

ies)

.

Sour

ce :

EL

FS (

auth

ors’

cal

cula

tion

s).


in construction which tends to be concentrated in medium-paid jobs. This is clearly the reason why Ireland, Spain, the UK and two of the Baltic countries appear in this group in 2008–2010 but not previously. Indeed, it is likely that the fast-growing employment in construction in 1995–2007 (Fernández-Macías and Hurley 2008: 25) served to disguise polarization in the overall employment structure for the earlier period in some countries. By contrast, its sharp decline in the recession has accentuated any underlying polarization, as already noted.

The category of upgrading and hybrid polarization/upgrading coun-tries is of similar size in the two periods. In these countries, employ-ment loss was concentrated in lower-paid jobs while better-paid jobs experienced growth. In the recession, this was decisively the case in ‘pure’ upgrading countries such as Luxembourg, Sweden and Germany, while hybrid polarization/upgrading countries, such as Austria and Belgium, show significant employment loss in medium-paid jobs, lit-tle change in the lowest-paid jobs and growth of employment in the highest-paid ones.

A new term, ‘downgrading’, is necessary to cover the change from growth of employment to decline between the two periods. No coun-tries experienced ‘growth in the middle’ during the crisis period but a number of countries experienced an overall downgrading in the struc-ture of employment structure according to our methodology. By this we mean that job destruction was greater in higher-paid jobs than in lower-paid ones. There was no equivalent of this pattern in the period 1995–2006 as only one of 23 countries recorded marginally greater job growth in the bottom quintile than in the top quintile (the Netherlands, the most obviously polarizing country). For the recession period, the

Table 10.3 Patterns of employment change at national level – comparison of pre- and post-crisis periods

Pattern of employment change 1995–2007 2008–2010

Polarization CY, FR, HU, NL, SK

BG, CY, ES, FR, GR, IE, LV, PT, SI, UK

Hybrid polarization/upgrading AT, BE, DE, SI, UK AT, BE, FI, NL, PL Upgrading DK, FI, IE, LU, PT DE, LU, SE, SK Hybrid upgrading/growing middle CZ, ES, IT, SE RO Growth in middle EE, GR, LT, LV – – – Downgrading – DK, CZ, HU, IT, LT Not classified – EE, MT


group covers – perhaps surprisingly – Denmark, the Czech Republic and Hungary as well as Italy and Lithuania.

There is no obvious common explanation for this development in these five countries. If the actual jobs which are primarily responsi-ble for these quintile shifts are examined, a varied picture emerges. In Italy, the largest growing job by some margin has been in low-paid household services (+150,000). The decline in the highest-paid quintile jobs in Italy is attributable in large part to job losses in public adminis-tration and education. Italy is also the only Member State in which all of the ‘higher’ non-manual occupational grades (aggregated to ISCO one digit – officials, /legislators, managers, professionals and associate professionals) experienced job reductions. Cumulatively these high-er-level occupations accounted for just over half-a-million job losses, that is, equivalent to the net employment loss in Italy over the two-year period. This is quite a different pattern compared to most other Member States where employment losses were concentrated lower down the occupational and wage distribution.

In terms of job growth at the lower end of the wage distribution, four of the five fastest growing jobs in the Denmark were low-paid jobs in the bottom quintile, including in retail sales and restaurants. Meanwhile, the highest growth in Hungary was among lower-level public adminis-tration workers (in ‘elementary occupations’).

The picture is therefore disparate and it would be unwise to speculate about structural trends given the short time frame of the current analy-sis. Nonetheless, certain countries exhibit a clear downgrading of their employment structure during the recession. Overall, patterns of shifts in employment are even more varied across countries than those in the previous expansion period, presented in Chapter 3 .

10.7.1 Patterns of employment change in manufacturing, construction and services

This final section is concerned with the broad sectors which have contributed most to net overall employment losses – construction and manufacturing – and employment gains – services – during the recession.

Construction / manufacturing

The two broad sectors that suffered the brunt of the impact of the recession on employment are construction and manufacturing. Over 10 per cent of pre-crisis jobs were destroyed in each between 2008q2 and 2010q2. The concentration of net employment decline in middle- and


lower-ranked jobs relates primarily, if not exclusively, to employment reduction in these two sectors (as does the diminishing gender-employ-ment gap). Construction-sector job losses were heavily concentrated in medium and medium-low paid jobs while in manufacturing sector, job losses were spread across the wage distribution but, again, most heavily concentrated in the same two quintiles ( Figure 10.6 ).

The four individual jobs in which employment declined the most were all in construction. Between them, construction-sector net job losses amounted to more than 1.9 million over the two-year period – over 35 per cent of the total net decline.

Even non-manual workers, engineers, other professionals and corpo-rate managers in the construction sector were amongst those experienc-ing the largest reductions in employment. Employment fell most sharply in those Member States where construction had grown as a result of property booms to account for 11–13 per cent of total employ-ment (compared to 8% in the EU as a whole) at the beginning of 2008 ( Figure 10.7 ).

The biggest falls in employment in absolute terms, however, were in manufacturing between 2008q2 and 2010q2 (3.8 million jobs). The bulk

EU

–1,0

00–3

,000

1,00

0–2

,000

0

Manufacture ConstructionAll

Figure 10.6 Net change in employment (,000s) by job–wage quintile in manufacturing and construction, EU27, 2008q2–2010q2

Source : EU-LFS (authors’ calculations).


of the losses were, in low-tech, traditional manufacturing (clothing and textiles, furniture and wood products, etc.) where global competition and offshoring continue to reduce the share of production carried out in the EU. These jobs were predominately in the three lowest-wage quin-tiles. There was also significant job loss in the high-tech manufacturing sectors, in jobs located in the middle quintile.

This occurred in both the more- and less-developed EU economies and was largely in the heavy capital-intensive sectors such as fabricated metals, motor vehicles and machinery and equipment. These sectors account for much of the considerable high-tech net job loss in the mid-dle quintiles (among machine operators and other metal and machin-ery workers) in Sweden and the Czech Republic in particular and to a lesser extent in France, Slovenia and Slovakia.

It is striking that only very limited high-tech job loss in the middle of the wage distribution occurred in Germany. This is almost certainly due to a large extent to the reduction in average hours in Germany through short-time working schemes, which preserved jobs more effectively than in other countries. Indeed, Germany continued to be the major source of high-tech manufacturing employment growth. Four of the

EU

–1,0

00–3

,000

1,00

0–2

,000

0

High-tech Manufacture Low-tech ManufactureAll

Figure 10.7 Net change in employment (,000s) in hi-tech and low-tech manufacturing by job–wage quintile, EU27, 2008q2–2010q2



top eight growing jobs in the highest-wage quintile in Germany were among physical, mathematical and engineering science professionals.

Services

Given that over 70 per cent of employment is in the service sector, it is to be expected that services would have a sizeable effect on our aggre-gate representation of employment changes during the recession. This was certainly the case in the decade preceding the crisis when services accounted for virtually all of the growth at the top and the bottom ends of the wage distribution (Fernandez-Macias and Hurley, 2008: 29). However, the slowdown in service employment growth during the recession means that the ongoing polarization tendencies are, to a larger extent than before, accounted for by the collapse of the middle wage ranking manufacturing and construction jobs.

Similarly, the rapid decline in manufacturing and construction means that even if the rate of employment growth in services declined appreciably, it nonetheless increased in absolute terms during the recession and the service share of overall employment continued to increase.

The large, predominantly state-funded, education and healthcare sec-tors have been important sources of employment growth both between and, most especially, during recessions going back to the 1980s. This pattern was repeated during the 2008–2009 recession which saw the health and education sectors contribute significantly to job growth, not least in the highest-wage quintile. Even during the last quarter of 2008 and the first of 2009, when the downturn was greatest, employ-ment in education and healthcare continued to increase and over the period, 2008q2–2010q2, recording net job growth of around 3 per cent and 5 per cent, respectively. As Figure 10.8 illustrates, 7 private-service employment declined markedly in middle-ranking jobs while it grew across the board in the predominantly state-funded sector, with a strong bias towards higher-paid jobs ( Figure 10.9 ).

Another way of decomposing the change in service employment is in terms of the ‘knowledge intensity’ of the sectors, following the Eurostat distinction between knowledge-intensive and less-knowledge intensive services. 8 Aside from health and education professionals, a major con-tribution to the increase in knowledge-intensive service (KIS) jobs in the top quintile was made by science professionals in computer services, consultancy and other business services. Indeed KISs have remained rel-atively unaffected during the crisis, enjoying employment growth across the wage distribution. They alone account for all of the net growth in


high- and medium-high paid jobs in the EU. Growth was also evident in lower-paid KIS jobs, primarily in residential care but also in smaller, expanding activities such as gambling and betting, information ser-vices and head offices and corporate headquarters ( Figure 10.10 ).

By contrast, less-knowledge intensive services (LKIS) suffered employ-ment losses across the board. The retail sector contributed most to these losses especially in the lower quintiles while losses were also notable in postal services (in quintile 2, reflecting the concluding phase of dereg-ulation in the EU), in warehousing (in quintile 3) and transport and personal services (in quintile 4). There was some countervailing LKIS employment growth in food and beverage services but not enough to offset the losses elsewhere.

10.8 Conclusions

This chapter has described the occupational and sectoral shifts of employment that occurred in recent recessions going back to the 1980s and compared them to those that occurred during the Great Recession of 2008–2009. As in previous recessions, the sectors that contributed most

DE

0–4

0040

0–2

0020

0

High-tech Manufacture Low-tech ManufactureAll

Figure 10.8 Net change in employment (,000s) in hi-tech and low-tech manufacturing by job–wage quintile, Germany, 2008q2–2010q2



to overall employment declines were construction and manufactur-ing, the latter a sector where recent recessions have tended to accen-tuate long-run structural decline. The service sector, on the contrary, expanded both in absolute and relative terms during the recession and accounts for just over seven of every ten jobs in the EU. The most sig-nificant contribution to these gains came from predominantly state-funded sectors such as healthcare and education – sectors in which employment growth going back to the early 1980s has tended to be higher during than between recessions.

At occupational level, the impact of the recession on employment was also differentiated with manual workers suffering greater job losses than non-manual ones. Skilled manual workers were – as in previous recessions – somewhat more exposed to job loss than their unskilled or low-skilled counterparts. High-skilled and high-paid non-manual occu-pations (mainly professionals and associate professionals) were the prin-cipal source of employment gains during the recession.

EU–1

,000

01,

000

Public Private

Figure 10.9 Net change in employment (,000s) in public and private services by job–wage quintile, EU27, 2008q2–2010q2

Note : Private services comprise retail and other private services as in Table 10.2 (i.e. NACE G-N, R-U) while public services comprise utilities, health, education and public adminis-tration (NACE E-F, O-Q).



Some interesting empirical conclusions are to be drawn by comparing the results of the overall EU-wide shifts in the two years (2008–2010) with the previous Eurofound analysis of the patterns of job expansion between 1998 and 2007. Up until the recession, the EU saw strong over-all employment growth but with appreciably higher growth in the top (fifth) wage quintile followed by the next highest (fourth) quintile. However, there was also appreciable growth in the bottom quintile, with low employment growth in the middle quintiles two and three. This pattern was characterized as upgrading with some polarization. Thus compared to the strong US job growth a decade earlier, the EU growth was somewhat less polarized, with the highest-paid jobs show-ing clearly more growth than the ones at the bottom.

It is interesting that the period between 2008 and 2010, one of employ-ment contraction, exhibits a similar pattern – at the aggregate EU level if not at individual country level – to that of the preceding decade in terms of the distribution of employment shifts across the wage struc-ture. The change in the structure of employment during the recession represents an acceleration or accentuation of previous trends. The top

EU

0–2

,000

2,00

0

KIS LKISAll

Figure 10.10 Net change in employment (,000s) in knowledge-intensive/less- knowledge intensive services by job-wage quintile, EU27, 2008q2–2010q2



quintile was the only sector to show any net growth of employment, while the reduction in employment was largest in the middle second and third sector, that is similar to the pattern of change in decade lead-ing up to the recession.

During the preceding decade of job growth, the polarization tendency was accounted for largely by the growth of service jobs at the two tails of the wage distribution. However, it is the massive job loss in manufactur-ing and construction and the ensuing collapse of the middle-ranking jobs that underlie the accentuation of the polarization tendency in the recession. While some of the fall in employment in the middle quin-tiles in manufacturing was in high-tech sectors (e.g. in Sweden, France, Slovakia and Slovenia) most was in low-tech manufacturing. The contin-ued growth in high-tech manufacturing in Germany, mostly in the top wage quintile, is also striking. Construction is the other sector account-ing for job losses in the middle of the wage distribution. Just as the construction boom held up employment in the middle quintiles in the preceding decade the collapse of construction in many counties accen-tuated the decline in the middle quintiles in the recession. Though the loss of jobs in both construction and manufacturing is cyclical, it is less likely that this loss will be reversed, at least to a substantial extent, in the latter than the former.

The persistence of job growth in the top quintile, while not occurring in all Member States, is striking. Even in Spain and Ireland, countries experiencing among the biggest overall reduction in employment in the EU, the number of jobs in the top quintile increased or remained virtually unchanged.

As in the preceding decade, job growth in the top quintile was mainly due to an increase in Knowledge-Intensive Services. These include both public services (mainly education) and private services (business ser-vices). During the recession, in common with previous downturns going back to the 1980s, the relative importance of public services for employ-ment growth increased and they contributed markedly to growth in the top quintile.

Growth in the highest-paid jobs was overwhelmingly in knowledge-intensive services, while the decline in the number of the lowest-paid jobs was comparatively modest.

Looking ahead, the most likely prognosis for employment growth across sectors and across the wage spectrum during the recovery phase is that some of sector-specific impact of the recession will be reversed during the recovery, though it is unlikely on past experience to be fully


reversed. Higher-level occupations in health and education accounted for much of the resilience of top-quintile employment growth during the recession while low-medium and medium-paid jobs in construction and manufacturing suffered disproportionate reductions, notwithstand-ing public-support measures such as widespread short-time working schemes. Together, these developments contributed to an accentuation of the patterns of employment polarization already observed in the period up to 2007. If in the recovery, employment in the predominantly state-funded sectors of education, and healthcare and, above all, public administration, especially in higher-level occupations, grows relatively less strongly, or even declines as now seems possible, while construction and manufacturing regain some of the jobs lost, the outcome could be some moderation in the pattern of polarized employment growth. Over the longer term, however, it is to be expected that the shifts in the struc-ture of employment which were evident in the period leading up to the recession will once again re-emerge in EU countries.

Notes

1 . This chapter was inspired by an original contribution by Terry Ward and Robert Stehrer, ‘The structure of employment and job quality in periods of recession and recovery in the EU’ which forms most of the first five sections. The remaining sections summarize findings from a more recent paper ‘Shifts in the employment structure during the Great Recession’ by John Hurley and Donald Storrie, which updates the analysis to mid-2010.

2 . For details, visit http://www.euklems.net 3 . These are published in the industry, trade and services section at http://epp.

eurostat.ec.europa.eu/portal/page/portal/statistics/search_database. 4 . The job–wage quintiles in this case are based on allocating jobs (again,

defined as occupations in sectors) to quintiles based primarily on 2008 EU Labour Force Survey wage data. A common EU ranking was generated from the 13 Member States for which adequate wage data was available. This served to rank jobs which were then assigned to quintiles in such a way that as close to possible as 20 per cent of EU employment was assigned to each quintile.

5 . With the qualification that we are applying the 1998–2007 per annum growth rate for each quintile from EU23 (all Member States except PL, MT, RO, and BG) to the EU27.

6 . Four Member States were not covered in the earlier period due to data non-availability – PL, MT, RO, BG. Also, the quintile assignments were based on national wage data for all 23 Member States for the 1995–2006 analysis. This is our preferred method. In this report, in order to cover employment devel-opments in all 27 Member States, we use a national ranking for 13 Member States and a common European job–wage ranking (based on an average of


the standardized rankings of the 13 Member States for which we do have data) for those 14 countries where adequate wage data was unavailable.

7 . This figure uses the following breakdown: private = retail + other private services (NACE G-N, R-U); public = health, education, public admin and utilities = NACE E-F, O-Q.

8 . See http://epp.eurostat.ec.europa.eu/cache/ITY_SDDS/Annexes/htec_esms_an3.pdf for full listing.

279

References

Acemoglu, D. (2002) ‘Technical Change, Inequality, and the Labor Market’, Journal of Economic Literature , Xl(March): 7–72.

Akerlof, G.A. and J.L. Yellen (1986) Efficiency Wage Models of the Labor Market . Cambridge: Cambridge University Press.

Akerlof, G.A. and J.L. Yellen (1990) ‘The Fair Wage-Effort Hypothesis and Unemployment’, The Quarterly Journal of Economics , 105(2): 255–283.

Anderson, R., M. Branislav, V. Greet, L.-Y. Maija and Z. Valentina (2009) ‘ Second European Quality of Life Survey ’ (overview report), Dublin: Eurofound.

Asplund, R. and E. Barth (eds) (2005) Education and Wage Inequality in Europe: A Literature Review. Taloustieto Oy.

Atkinson, A. and J. Micklewright (1992) Economic Transformation in Eastern Europe and the Distribution of Income . Cambridge: Cambridge University Press.

Autor, D.H., F. Levy and R.J. Murnane (2003) ‘The Skill Content of Recent Technological Change: An Empirical Exploration’, Quarterly Journal of Economics , 118(4): 1279–1333.

Autor, D.H., L.F. Katz and M.S. Kearney (2006) ‘The Polarization of the U.S. Labor Market’, The American Economic Review: Papers and Proceedings , 96: 189–194.

Baily, M.N. (2010) ‘The Next Economy and the Growth Challenge for the United States’. Paper Prepared for the Scholars Strategy Network Conference, Harvard University, Cambridge MA, 30 September 2010. Washington: Brookings Institution.

Balla, K., J. Köllő and A. Simonovits (2008) ‘Transition with Heterogeneous Labour’, Structural Change and Economic Dynamics , 19(3): 203–220.

Barbieri, P. (2009) ‘Flexible Employment and Inequality in Europe’, European Sociological Review , 25(6): 621–628.

Baumol, W.J. (1967) ‘Macroeconomics of Unbalanced growth: The Anatomy of Urban Crisis’, The American Economic Review , 57(3): 415–426.

Bean, F. and M. Leach (2004) ‘Immigrant Job Quality and Mobility in the United States’, Work and Occupations , 31: 499–518.

Beck, U. (1992) Risk Society: Towards a New Modernity , London: Sage. Bell, D. (1976) The Coming of Post-Industrial Society: A Venture in Social Forecasting .

New York: Harper. Benedek, D., M. Rigó, Á. Scharle and P.A. Szabó (2006) ‘Increases in the minimum

wage in Hungary, 2001–2006’. Working Papers in Public Finance, No. 16. Ben-Porath, Y. (1967) ‘The Production of Human Capital and the Life Cycle of

Earnings’, Journal of Political Economy , 75: 352–365. Bernstein, J., H. Boushey, E. McNichol and R. Zahradnik (2002) Pulling Apart:

A State-by-State Analysis of Income Trends. Center on Budget and Policy Priorities and Economic Washington, D.C.: Policy Institute.

Bhaduri, A. and K. Laski (1994) ‘The Relevance of Michal Kalecki Today’, Economic and Political Weekly, 7, Feb. 12–18, p. 356.

Blanchard, O.J. (1997) The Economics of Post Communist Transition . Oxford: Oxford University Press.

280 References

Blanchard, O.J., K.A. Froot and J.D. Sachs (1994) The Transition in Eastern Europe . Chicago: The University of Chicago Press.

Blanchard, O.J., S. Commander and F. Coricelli (1995) Unemployment and Restructuring in Eastern Europe and Russia . Washington: Economic Development Institute of The World Bank.

Blauner, R. (1964) Alienation and Freedom . Chicago: The University of Chicago Press.

Blinder, A. (2006) ‘Offshoring: The Next Industrial Revolution?’ Foreign Affairs , March/April: 113–128.

Blinder, A. (2007) ‘How Many U.S. Jobs Might be Offshorable?’ Working Paper No. 142. Princeton University Center for Economic Policy Studies.

Bluestone, B. and B. Harrison (1982) Deindustrialization of America . New York: Basic Books.

Bodvarsson, Örn B. and H. Van den Berg (2009) The Economics of Immigration: Theory and Policy. Heidelberg: Springer-Verlag.

Bofinger, P. (1994). ‘Macroeconomic transformation in Eastern Europe: the role of monetary policy reconsidered’, in Herr et al. (eds.) Macroeconomic problems of transformation . Aldershot, Hants: Edward Elgar.

Borjas, G.J. (1994) ‘The Economics of Immigration’, Journal of Economic Literature , 32(4): 1667–1717.

Borjas, G.J. (1999) ‘The economic analysis of immigration’, in A. Orley and D.E. Card (eds.) Handbook of labor economics , vol. 3. Amsterdam: Elsevier, 1697–1760.

Borjas, G.J. (2003a) ‘Welfare Reform, Labor Supply, and Health Insurance in the Immigrant Population’, Journal of Health Economics , 22(6): 933–958.

Borjas, G.J. (2003b) ‘The Labor Demand Curve Is Downward Sloping: Reexamining the Impact of Immigration on the Labor Market’, Quarterly Journal of Economics , 118(4): 1335–1374.

Borjas, G.J. (2004) ‘Increasing the Supply of Labor through Migration. Measuring the Impact on Native-born Workers’, Backgrounder , May. Center for Immigrant Studies.

Borjas, G.J. and S.J. Trejo (1991) ‘Immigrant Participation and the Welfare System’, Industrial and Labor Relations Review , 44(2): 165–182.

Bosch, G., K. Mayhew and J. Gautié (2009), ‘Industrial relations, legal regula-tions, and wage setting’, in J. Gautié and J. Schmitt (eds), Low-wage work in the wealthy world . New York: Russell Sage Foundation.

Braverman, H. (1998, originally 1974) Labor and Monopoly Capital . New York: The Monthly Review Press.

Breen, R. (2004) Social Mobility in Europe . Oxford: Oxford University Press. Broughton, A. (2007) Working Conditions and Social Dialogue . Dublin: Eurofound. Brown, D. and M. McIntosh (1998) If You’re Happy and You Know It ... Job

Satisfaction in the Low Wage Service Sector . Centre for Economic Performance paper, London School of Economics, London.

Brücker, H., G. Epstein, B. McCormick, G. Saint-Paul, A. Venturini and K. Zimmermann (2002) ‘Welfare state provision’, in T. Boeri, G.H. Hanson and B. McCormick (eds) Immigration policy and the welfare system . London: Oxford University Press, 66–90.

Byrne, D. and C. Ragin (eds) (2009) The Sage Handbook of Case-Based Methods . Thousand Oaks, CA: Sage Publications.

References 281

Calmfors, L. and J. Driffill (1988) ‘Bargaining Structure, Corporatism and Macroeconomic Performance’, Economic Policy , 3(6): 12–61.

Cahuc, P. and A. Zylberberg (2004) Labor Economics . Cambridge, MA: MIT Press. Card, D.E. (1990) ‘The Impact of the Mariel Boatlift on the Miami Labor Market’,

Industrial and Labor Relations Review , 43(2): 245–257. Card, D.E. and J.E. DiNardo (2002) ‘Skill-Biased Technological Change and Rising

Wage Inequality: Some Problems and Puzzles’, Journal of Labor Economics , 20(4): 733–783.

Cazes, S. and S. Verick (2011) ‘What has happened to Okun’s Law in the US and Europe? Insights from the Great Recession and longer time trend’. Paper given at ILO research conference, 16–17 February 2011, Geneva, Key Lessons from the Crisis and Way Forward.

Clark A.E. and A.J. Oswald (1996) ‘Satisfaction and Comparison Income’, Journal of Public Economics , 61: 359–381.

Clogg, C.C. (1982) ‘Some Models for the Analysis of Association in Multiway Cross-Classifications Having Ordered Categories’, Journal of the American Statistical Association , 77: 803–815.

Clogg, C.C. and E.S. Shihadeh (1994) Statistical Models for Ordinal Variables . Thousand Oaks, CA: Sage.

Coase, R. (1996) ‘The nature of the firm’, in L. Putterman and R. Krozsner (eds) The economic nature of the firm: a reader . Cambridge: Cambridge University Press.

Cohen, P.N. and M.L. Huffman (2003) ‘Individuals, Jobs, and Labor Markets: The Devaluation of Women’s Work’, American Sociological Review , 68: 443–463.

Council of Economic Advisors (CEA) (1996) Job Creation and Employment Opportunities: The US Labour Market 1993–1996. Washington DC: Office of the Chief Economist.

Culpepper, P.D. and D. Finegold (1999) The German Skills Machine . New York: Berghahn Books.

Dahl, S-A., T. Nesheim and K. Olsen (2009) ‘Quality of Work: Concepts and Measurement’. RECWOWE working paper.

Davoine, L., C. Erhel and M. Guergoat-Lariviere (2008a) ‘A Taxonomy of European Labour Markets using Quality Indicators’, Rapport de Recherche No. 45, Centre d’études de l’Emploi.

Davoine, L., C. Erhel and M. Guergoat-Lariviere (2008b) ‘Monitoring Quality in Work: European Employment Strategy Indicators and Beyond’, International Labour Review , June 2008.

Dezséri, K. (2001) ‘Abandoned Brotherhood: Declining Economic Relations between the CEECs and Russia in the 1990s’, Eastern European Economics, 39(3) (May– June, 2001): 5–44

Dickens, W.T. and K. Lang (1985) ‘A Test of Dual Labor Market Theory’, The American Economic Review , 75(4): 792–805.

DiPrete, T.A., P.M.D. Graaf, R. Luijkx, M. Tahlin and H.-P. Blossfeld (1997) ‘Collectivist versus Individualist Mobility Regimes? Structural Change and Job Mobility in Four Countries’, The American Journal of Sociology , 103: 318–358.

Doeringer, P.B. and M.J. Piore (1971) Internal Labor Markets and Manpower Analysis . Lexington, MA: MIT Press.

Doogan, K. (2009) New Capitalism? The Transformation of Work . Cambridge: Polity. Duncan, O.B. and B. Duncan (1955) ‘A Methodological Analysis of Segregation

Indexes’, American Sociological Review , 20(2): 210–217.

282 References

Dunlop, J.T. (1958) Industrial Relations Systems . Boston, MA: Harvard Business School Press.

Dunlop, J.T., C. Kerr, F.H. Harbison and C.A. Myers (1964) Industrialism and Industrial Man: The Problems of Labor and Management in Economic Growth , 2nd edition. New York: Oxford University Press.

EC (2008) Employment in Europe , DG for Employment, Social Affairs and Equal Opportunities, Brussels: EC.

Edgell, S. (2006) The Sociology of Work: Continuity and Change in Paid and Unpaid Work . CA: Sage Publications: Pine Forge Press.

Elias, P. (1997) ‘Occupational Classification (Isco-88): Concepts, Methods, Reliability, Validity and Cross-National Comparability’, OECD Labour Market and Social Policy Occasional Papers, No. 20, OECD Publishing.

England, P., L.L. Reid and B.S. Kilbourne (1996) ‘The Effect of the Sex Composition of Jobs on Starting Wages in an Organization: Findings from the NLSY’, Demography , 33: 511–521.

Erikson, R. and J.H. Goldthorpe (1985) ‘Are American Rates of Social Mobility Exceptionally High? New Evidence on an Old Question’, European Sociological Review , 1: 1–22.

Erikson, R. and J.H. Goldthorpe (1987) ‘Commonality and Variation in Social Fluidity in Industrial Nations (I): A Model for Evaluating the “FJH hypoth-esis”’, European Sociological Review , 3: 54–77.

Erikson, R. and J.H. Goldthorpe (1992) The Constant Flux: A Study of Class Mobility in Industrial Societies . Oxford: Clarendon Press.

Eriksson, T. and M. Pytlikova (2004) ‘Firm-level Consequences of Large Minimum-wage Increases in the Czech and Slovak Republics’, Labour , 18(1): 75–103.

Esping-Andersen, G. (1990) The Three Worlds of Welfare Capitalism . Cambridge: Polity Press & Princeton: Princeton UP.

Esping-Anderson, G. (1999) Social Foundations of Post-Industrial Economies . Oxford: OUP.

Estévez-Abe, M. (2005) ‘Gender Bias in Skills and Social Policies: The Varieties of Capitalism Perspective on Sex Segregation’, Social Policy , 12(2): 180–215.

European Commission (1993) White Paper on growth, competitiveness, and employment: The challenges and ways forward into the 21st century, COM(93) 700 final.

European Commission (2008a) Industrial Relations in Europe. Brussels: European Commission (DGV).

European Commission (2008b) Employment in Europe 2008 . Luxembourg: Office for Official Publications of the European Communities.

European Council (2000) ‘Lisbon Summit Presidency Conclusions’, European Parliament Website.

Eurostat (2006) Quality Report of the European Labour Force Survey 2005, Eurostat. Downloaded from: http://epp.eurostat.ec.europa.eu/cache/ity_off-pub/ks-ra-07-017/en/ks-ra-07-017-en.pdf [accessed May, 2011].

Eurostat (2009) The European Union Labour Force Survey: Main characteristics Of the national surveys. Downloaded from: http://epp.eurostat.ec.europa.eu/cache/ity_offpub/ks-ra-09-002/en/ks-ra-09-002-en.pdf [accessed April, 2011].

References 283

Eyraud, F., D. Marsden and J-J. Silvestre (1990) ‘Occupational and Internal Labour Markets in Britain and France’, International Labour Review , 129(4): 501–517.

Farber, H.S. (1997) ‘Job Creation in the United States: Good Jobs or Bad?’, Working Papers 285, Industrial Relations section, Princeton University.

Farber, H.S. (1999) ‘Mobility and stability: The dynamics of job change in labor markets’, in O. Ashenfelter and C. David (eds) Handbook of Labor Economics , volume 3, part 2. Amsterdam: Elsevier, 2439–2483.

Felix, B. (2006) ‘High Tech Industries and Knowledge Based Services’, Statistics in Focus 13/2006, Eurostat.

Fenger, H.J.M. (2007) ‘Welfare Regimes in Central and Eastern Europe: Incorporating Post-Communist Countries in a Welfare Regime Typology’, Contemporary Issues and Ideas in Social Sciences , 3(2).

Fernández-Macías, E. (2009) ‘Job Polarization in the UK And Europe? A Test of the Original Jobs Wage Ranking for the UK’, Paper Presented in the Jobs Project Meeting, Salamanca, 2–3 July 2009.

Fernández-Macías, E. (2010) ‘Changes in the Structure of Employment and Job Quality in Europe’, PhD Thesis, University of Salamanca.

Fernández-Macías, E. and J. Hurley (2008) More and better jobs: Patterns of Employment Expansion in Europe , Dublin: Eurofound. Available at http://www.eurofound.europa.eu/pubdocs/2008/50/en/1/EF0850EN.pdf

Ferrera, M. (1996) ‘The “Southern Model” of Welfare in Social Europe’, Journal of European Social Policy , 6(1): 17–37.

Festinger, L. (1957) A Theory of Cognitive Dissonance . Stanford, CA: Stanford University Press.

Flemming, J.S. and J. Micklewright (2000), ‘Income distribution, economic sys-tems and transition’, in A.B. Atkinson and F. Bourgignon (eds), Handbook of income distribution , volume 1, Amsterdam: Elsevier.

Florida, R. (2002) The Rise of the Creative Class . New York: Basic Books. Forster, M., D. Jesuit and T. Smeeding (2005) ‘Regional poverty and income

inequality in Central and Eastern Europe: evidence from the Luxembourg income study’, in R. Kanbur and A.J. Venables (eds) Spatial inequality and devel-opment . Oxford: Oxford University Press.

Fourastié, J. (1963) Le Grand Espoir Du XXe Siecle . Gallimard. Freeman, R.B. (2007) ‘Labor Market Institutions around the World’, NBER

Working Paper no. 13242. Friedberg, R.M. (2001) ‘The Impact of Mass Migration on the Israeli Labor

Market’, Quarterly Journal of Economics , 116(4): 1373–1408. Friedberg, R.M. and J. Hunt (1995) ‘The Impact of Immigrants on Host Country

Wages, Employment and Growth’, Journal of Economic Perspectives , 9(2): 23–44.

Funk, L. and H. Lesch (2005) Minimum wages in Europe, Eurofound, 2005 August, http://www.eurofound.europa.eu/eiro/2005/07/study/tn0507101s.htm.

Galbraith, J.K. and M. Berner (2001) Inequality and Industrial Change: A Global View . Cambridge: Cambridge University Press.

Gannon, B., R. Plasman, F. Rycx and I. Tojerow (2007) ‘Inter-Industry Wage Differentials and the Gender Wage Gap: Evidence from European’, Economic and Social Review .

284 References

Ganzeboom, H.B.G., L. Ruud and D.J. Treiman (1989) ‘Intergenerational Class Mobility in Comparative Perspective’, Research in Social Stratification and Mobility , 8: 3–84.

García Mainar, M.I. (1999) ‘La satisfacción con el trabajo en España’, Documentación laboral , 59: 113–129.

Gautié, J. and J. Schmitt (eds) (2010) Low Wage Work in the Wealthy World . New York: Russell Sage Foundation.

Gomulka, S. (1998) ‘Output: Causes of the Decline and the Recovery’. CASE-CEU Working Paper 8.

Goodman, L.A. (1979a) ‘Multiplicative Models for the Analysis of Occupational Mobility Tables and other Kinds of Cross-Classification Tables’, American Journal of Sociology , 84(1): 804–819.

Goodman, L.A. (1979b) ‘Simple Models for the Analysis of Association in Crossclassifications Having Ordered Categories’, Journal of the American Statistical Association , 74: 537–552.

Goodman, L.A. (1986) ‘Some Useful Extensions of the Usual Correspondence Analysis Approach and the Usual Log-Linear Approach in the Analysis of Contingency Tables’, International Statistical Review , 54: 243–309.

Goodman, L.A. (1991) ‘Measures, Models, and Graphical Displays in the Analysis of Cross-Classified Data’, Journal of the American Statistical Association , 86: 1085–1111.

Goos, M. (2009) ‘Differences in Data, Analysis and Interpretation between Jobs Project and GMS’. Paper presented at Jobs project meeting, Salamanca, 2–3 July 2009.

Goos, M. and A. Manning (2007) ‘Lousy and Lovely Jobs: The Rising Polarization of Work in Britain’, The Review of Economics and Statistics, 89(1): 118–133.

Goos, M., A. Manning and A. Salomons (2009) ‘Job Polarization in Europe’, American Economic Review: Papers and Proceedings , 99: 58–63.

Grant, L., S. Yeandle and L. Buckner (2005) ‘Working Below Potential: Women and Part-time Work’. EOC Working Paper No. 40. Manchester: Equal Opportunities Commission.

Grimshaw, D. (2011) ‘What Do We Know about Low-wage Work and Low-wage Workers?’, ILO Conditions of Work and Employment Series, No. 28, Geneva: ILO.

Grimshaw, D. and J. Rubery (1997) ‘The Concentration of Women’s Employment and Relative Occupational Pay: A Statistical Framework for Comparative Analysis’, OECD Labour Market and Social Policy Occasional Papers, No. 26.

Grimshaw, D. and J. Rubery (2007) ‘Undervaluing Women’s Work’. EOC Working Paper Series, No. 53. Manchester: Equal Opportunities Commission.

Grimshaw, D., G. Whitehouse and D. Zetlin (2001) ‘Changing Pay Systems, Occupational Concentration and the Gender Pay Gap: Evidence from Australia and the UK’, Industrial Relations Journal , 32(3): 209–229.

Groot, W. and H. Maassen van den Brink (2000) ‘Overeducation in the Labor Market: A Meta Analysis’, Economics of Education Review , 19(2): 149–158.

Grusky, D.B. and R.M. Hauser (1984) ‘Comparative Social Mobility Revisited: Models of Convergence and Divergence in 16 Countries’, American Sociological Review , A49: 19–38.

Guest, D. (2002) ‘Perspectives on the Study of Work-Life Balance’, Social Science Information , 41(2): 255–279.

References 285

Hacker, J.S. and P. Pierson (2010) Winner-Take-All Politics: How Washington Made the Rich Richer – and Turned its Back on the Middle Class. New York: Simon and Schuster.

Hall, P.A. and D. Soskice (2001) Varieties of Capitalism. The Institutional Foundations of Comparative Advantage . Oxford: Oxford University Press.

Harrison, B. and B. Bluestone (1988) The Great U-Turn: Corporate Restructuring and the Polarizing of America. New York: Basic Books.

Hartog, J. (2000) ‘Overeducation and Earnings: Where Are We, Where Should We Go?’, Economics of Education Review , 19(2): 131–147.

Hatton, T.J. and J.G. Williamson (1998) The Age of Mass Migration: An Economic Analysis. Oxford: Oxford University Press.

Hatzichronoglou, T. (1997) ‘Revision of the High-Technology Sector and Product Classification’. OECD Science, Technology and Industry Working Papers, 1997/2, OECD Publishing.

Hipple, S.F. (2010) ‘Self-Employment in the United States’, Monthly Labor Review, 113(9): 17–32.

Hull, F., S. Friedmas and T. Rogers (1982) ‘The Effects of Technology on Alienation from Work’, Work and Occupations , 9(1): 31–57.

Hunt, J. (1992) ‘The Impact of the 1962 Repatriates from Algeria on the French Labor Market’, Industrial and Labor Relations Review , 45(3): 556–57l.

Hurley, J., D. Storrie and J. Jungblut (2011) Shifts in the Structure of Jobs in Europe during the Great Recession . Dublin: Eurofound.

Ilg, R.E. (1996) ‘The Nature of Employment Growth, 1989–95’, Monthly Labor Review , 119: 29–36.

Ilg, R.E. and S.E. Haugen (2000) ‘Earnings and Employment Trends in the 1990s’, Monthly Labor Review , 123(March): 21–33.

ILO (2011) ILO LABORSTA Labour Statistics Database Table 2A Employment, general level http://laborsta.ilo.org/ [accessed 4 March 2011].

Jackman, R. (1995) ‘Economic Policies, Employment and Labour Markets in Transition in Central and Eastern Europe’. CEP Discussion Paper 265.

Jackman, R. and C. Pauna (1997) ‘Labour Market Policy and the Reallocation of Labour Across Sectors’. CEP Discussion Paper 338.

Jenkins, J.C., K.T. Leicht and H. Wendt (2006) ‘Class Forces, Political Institutions and State Intervention: Subnational Economic Development Policy in the U.S., 1971–1990’, American Journal of Sociolo gy, 111: 1122–80.

Kalleberg, A.L. and A.B. Sørensen (1979) ‘The Sociology of Labor Markets’, Annual Review of Sociology , 5: 351–379.

Kalleberg, A.L., B.F. Reskin and K. Hudson (2000) ‘Bad Jobs in America: Standard and Nonstandard Employment Relations and Job Quality in the United States’, American Sociological Review , 65: 256–278.

Katz, L.F. and D.H. Autor (1999) ‘Changes in the wage structure and earnings inequality’, in O. Ashenfelter and D. Card (eds) Handbook of labour economics , volume 3. Amsterdam: Elsevier, 1463–1555.

Kelly, J. and K. Hamann (2007) ‘Does varieties of capitalism explain national patterns of labor relations?’ in E. Heery, N. Bacon, P. Blyton and J. Fiorito (eds.) The SAGE handbook of industrial relations . London: Sage.

Kerr, C. (1977) Labor Markets and Wage Determination: The Balkanization of Labor Markets and Other Essays . Berkeley and Los Angeles, CA: University of California Press.

286 References

Kertesi, G. and J. Köllő (2000) ‘Wage Inequality in East-Central Europe’. Budapest Working Papers on the Labour Market BWP 2000/7.

Kim, C., H. Kim and A. Sakamoto (2008) ‘The Rise of Intra-Occupational Wage Inequality in the United States, 1983–2002’, American Sociological Review , 73: 129–157.

Köllő, J. (2006) ‘Workplace literacy requirements and unskilled employment in East-Central and Western Europe: Evidence from the International Adult Literacy Survey (IALS)’, Budapest Working Papers on The Labour Market BWP. 2006/7, Institute of Economics, Hungarian Academy of Sciences Department of Human Resources–Corvinus University of Budapest, Budapest, http://www.econ.core.hu/doc/bwp/ bwp/bwp0607.pdf.

Kornai, J. (1992) The Socialist System: The Political Economy of Communism . Oxford: Clarendon Press.

Kornai, J. (1993) ‘Transformational Recession: A General Phenomenon Examined Through the Example of Hungary’s Development’. Collegium Budapest Working Paper No. 1.

Krueger, A.B. and L.H. Summers (1988) ‘Efficiency Wages and the Inter-Industry Wage Structure’, Econometrica , 56(2): 259–293.

Krugman, P. (2008) The Return of Depression Economics and the Crisis of 2008 . London: Penguin.

Kuttner, R. (1983) ‘The Declining Middle’, The Atlantic Monthly , July: 60–72. Kuznets, S. (1955) ‘Economic Growth and Income Inequality’, American Economic

Review , 45(1): 1–28. Lazear, E.P. (1979) ‘Why Is There Mandatory Retirement’, Journal of Political

Economy , 87(6): 1261–1284. Lazear, E.P. and P. Oyer (2004) ‘Internal and External Labor Markets: A Personnel

Economics Approach’, Labour Economics , 11: 527–554. Lazear, E.P. and S. Rosen (1981) ‘Rank-Order Tournaments as Optimum Labor

Contracts’, Journal of Political Economy , 89(5): 841–864. Lemieux, T. (2008) ‘The Changing Nature of Wage Inequality’, Journal of

Population Economics , 21: 21–48. Leschke, J. and A. Watt (2008) ‘Putting a number on job quality: construct-

ing a European job quality index’. European Trade Union Institute working paper.

Levy, F. (1998) The New Dollars and Dreams: American Incomes and Economic Change . New York: Russell Sage Foundation.

Levy, F. and R. Murnane (1992) ‘US Earnings Levels and Earnings Inequality: A Review of Recent Trends and Proposed Explanations’, Journal of Economic Literature , September.

Lindbeck, A. and D. Snower (1988) The Insider-Outsider Theory of Employment and Unemployment . Cambridge, Mass: MIT Press.

Lindbeck, A. and D. Snower (2002) The Insider-Outsider Theory: A Survey . IZA Discussion Paper No. 534. Bonn: IZA.

Lobao, L.M. and G. Hooks (2007) ‘Advancing the Sociology of Spatial Inequality: Spaces, Places, and the Subnational Scale’, in L. Lobao, G. Hooks and A. Tickamyer (eds) The Sociology of Spatial Inequality . Albany: The State University of New York Press, pp. 29–61.

Longhi, S., P. Nijkamp and J. Poot (2005) ‘A Meta-Analytic Assessment of the Effect of Immigration on Wages’, Journal of Economic Surveys , 19(3): 451–477.

References 287

Longhi, S., P. Nijkamp and J. Poot (2006) ‘The Fallacy of “Job Robbing”: A Meta-analysis of Estimates of the Effect of Immigration’, Journal of Migration and Refugees Issue , 1(4): 131–152.

Lucifora, C., A. McKnight and W. Salverda (2005) ‘Low-Wage Employment in Europe: A Review of the Evidence’, Socio-Economic Review , 3(2): 259–292.

Maloney, W.F. (2004) ‘Informality Revisited’, World Development , 32(7): 1159–1178.

Mandl, I., D. Storrie, J. Hurley, M. Mascherini, A. Broughton, R. Owczarzak, S. Riso and L. Salvatore (2010) ERM Annual Report 2010, Extending Flexicurity – The Potential of Short-Time Working Schemes. Dublin: Eurofound.

Manglione T.W. and R.P. Quinn (1975) ‘Job Satisfaction, Counter-Productive Behavior and Drug Use at Work’, Journal of Applied Psychology , 60: 114–116.

Marsden, D. (1990) ‘Institutions and labour mobility: Occupational and inter-nal labour markets in Britain, France, Italy and West Germany’, in R. Brunetta and C. Dell’ Aringa (eds) Labour Relations and Economic Performance . London: Palgrave Macmillan.

Marsden, D. (1999) A Theory of Employment systems: Micro-Foundations of Societal Diversity . Oxford: OUP.

Marx, K. (1959, first published in 1932) Economic & Philosophical Manuscripts of 1844 . Progress Publishers of the Soviet Union. Consulted through the inter-net: http://www.marxists.org/archive/marx/works/1844/manuscripts/preface.htm.

Massey, D.S. and D.S. Hirst (1998) ‘From Escalator to Hourglass: Changes in the U.S. Occupational Wage Structure 1949–1989’, Social Science Research , 27: 51–71.

Masso, J. and R. Eamets (2004) ‘How flexible is the Estonian labour market? Micro evidence from firm demographics data’ – Berlin; Tallinn: Berliner Wissenschaftsverlag GmbH; Mattimar OÜ, 2004, pp. 237–245.

Maurice, M., F. Sellier and J-J. Silvestre (1982) Politique d’eéducation et organisation industrielle en France et en Allemagne . Paris: PUF.

Maurice, M., F. Sellier and J-J. Silvestre (1984) ‘Rules, Contexts and Actors Observations Based on a Comparison between France and Germany’, British Journal of Industrial Relations , 22(3): 346–363.

Mayer, K.U., D. Grunow and N. Nitsche (2010) ‘Mythos Flexibilisierung? Wie instabil sind Berufsbiographien wirklich und als wie instablil werden sie wahrgenommen?’, Kölner Zeitschrift für Soziologie und Sozialpsychologie , 62(3): 369–402.

Meisenheimer II, J.R. (1998) ‘The services industry in the “good” versus “bad” job debate’, Monthly Labor Review , 22–47.

Meyer, P.B. and A.M. Osborne (2005) ‘Proposed Category System for 1960–2000 Census Occupations’. BLS Working Paper no. 383. Washington D.C.: U.S. Department of Labor, Bureau of Labor Statistics.

Milanovic, B. (1999) ‘Explaining the Increase in Inequality during Transition’, Economics of Transition , 7(2): 299–341.

Milkman, R., E. Reese and B. Roth (1998) ‘The Macrosociology of Paid Domestic Labor’, Work and Occupations , 25: 483–510.

Milkman, R. and R.E. Dwyer (2002) ‘Growing Apart: The “New Economy” and Job Polarization in California, 1992–2000’, in The State of California Labor . Berkeley, CA: University of California Press, pp. 3–35.

288 References

Mincer, J. (1974) Education, Experience, and Earnings . New York: Columbia University Press.

Mishel, L., J. Bernstein and S. Allegretto (2007) The State of Working America 2006/2007 . Ithaca, NY: Cornell University Press.

Moller, S. and B.A. Rubin (2008) ‘The Contours of Stratification in Service-Oriented Economies’, Social Science Research , 37: 1039–1060.

Moncarz, R.J., M.G. Wolf and B. Wright (2008) ‘Service-Providing Occupations, Offshoring, and the Labor Market’, Monthly Labor Review , December 8: 71–86.

Morris, M. and B. Western (1999) ‘Inequality in Earnings at the Close of the Twentieth Century’, Annual Review of Sociology , 25: 623–57.

Mouw, T. and A.L. Kalleberg (2010) ‘Occupations and Wage Inequality in the United States’, American Sociological Review , 75: 402–431.

Muffels, R., T. Wilthagen and N. van den Heuvel (2002) ‘Labour market transi-tions and employment regimes: Evidence on the flexibility-security nexus in transitional labour markets’. WZB Discussion Paper FS I 02-204.

Mullan, B.P. (1989) ‘The Impact of Social Networks on the Occupational Status of Migrants’, International Migration , 27(1): 69–86.

Muñoz de Bustillo, R. and E. Fernández Macías (2005) ‘Job Satisfaction as an Indicator of the Quality of Work’ Journal of Socio-economics , 34(5): 656–673.

Muñoz de Bustillo, R., E. Fernández-Macías, J.I. Antón and F. Esteve (2011a) ‘E Pluribus Unum? A Critical Survey of Job Quality Indicators’, Socio Economic Review , 9(2): 447–475.

Muñoz de Bustillo, R., E. Fernández-Macías, J.I. Antón and F. Esteve (2011b) Measuring More Than Money: The Social Economics of Job Quality . London: Edward Elgar.

National Bureau of Economic Research (NBER) (2010) ‘U.S. Business Cycle Expansions and Contractions’. NBER Business Cycle Dating Committee . http://www.nber.org/cycles/cyclesmain.html [accessed July 31, 2010].

Neckerman, K.M. and F. Torche (2007) ‘Inequality: Causes and Consequences’, Annual Review of Sociology , 33: 335–357.

Newell, A. (2001) ‘The Distribution of Wages in Transition Countries’. IZA Discussion Paper No 267, March 2001.

Newell, A. and B. Reilly (1999) ‘Rates of Return to Educational Qualifications in the Transitional Economies’, Education Economics , 7(1).

OECD (1994) The OECD Jobs Study: Facts, Analysis, Strategy. Paris: OECD. OECD (1998) The Future of Female-Dominated Occupations . Paris: OECD. OECD (2007a) International Migration Outlook 2007 , Paris: OECD. OECD (2007b) Moving up the Value Chain: Staying Competitive in the Global

Economy , Paris: OECD. OECD (2008) International Migration Outlook 2008 , Paris: OECD. Oesch, D. and R.M. Jorge (2011) ‘Upgrading or Polarization? Occupational

Change in Britain, Germany, Spain and Switzerland, 1990–2008’, Socio-economic Review , 9(3): 503–531.

Olson, M. (1965) The Logic of Collective Action : Public Goods and the Theory of Groups . Harvard: Harvard University Press.

Osterman, P. (2008) ‘Improving the Quality of Low Wage Work: The Current American Experience’, International Labor Review , 147: 115–134.

Parent-Thirion, A., E. Fernández-Macías, J. Hurley and G. Vermeylen (2007) Fourth European Working Conditions Survey . Dublin: Eurofound.

References 289

Patel, K. and F. Vella (2007) ‘Immigrant Networks and Their Implications for Occupational Choice and Wages’, IZA Discussion Paper, 3217.

Peck, J. (1996) Work-Place: The Social Regulation of Labour Markets . New York: Guilford.

Peck, J., M. Doussard and N. Theodore (2009) ‘After Deindustrialization: Uneven Growth and Economic Inequality in “Postindustrial Chicago”’, Economic Geography , 85: 183–207.

Pedersen, P.J., M. Pytlikova and N. Smith (2008) ‘Selection or Network Effects? Migration Flows into 27 OECD Countries, 1990–2000’, European Economic Review , 52(7): 1160–1186.

Plantenga, J. and J. Hansen (2008) ‘Assessing Equal Opportunities in the European Union’, International Labour Review , 138(4): 351–379.

Pollack, E. and R. Thiess (2010) ‘Impact of Alternate Public Transit and Rail Investment Scenarios on the Labor Market’, Issue Brief #285. Washington D.C.: Economic Policy Institute.

Powers, D.A. and Y. Xie (2000) Statistical Methods for Categorical Data Analysis . New York: Academic Press.

Purse, K. (2004) ‘ Work-Related Fatality Risks and Neoclassical Compensating Wage Differentials’ Cambridge Journal of Economics, 28(4): 597–617.

Radošević, S. and U. Varblane and T. Mickiewicz (2003) ‘The Patterns of Foreign Direct Investment and Employment in Central Europe during Economic Recovery’, Transnational Corporations , 12(1): 53–90.

Reich, R.B. (1992) The Work of Nations: Preparing Ourselves for 21st Century Capitalism. New York: Vintage Books.

Reinhart, C. and K. Rogoff (2009) ‘The Aftermath of Financial Crises’, American Economic Review, May, 99(2): 466–472.

Riedmann, A., H. Bielenski, T. Szczurowska and A. Wagner (2006) Working Time and Work Life Balance in European Companies . Dublin: Eurofound.

Rosen, S. (1986) ‘The theory of equalizing differences’, in O. Ashenfelter and R. Layard (eds) Handbook of Labor Economics , volume 1. Amsterdam: Elsevier Science Publishers BV, pp. 641–692.

Rubery, J. (2003) ‘Pay Equity, Minimum Wage and Equality at Work’. Working Paper No. 19, Geneva: ILO.

Rubery, J., C. Fagan and M. Smith (1999) Women’s Employment in Europe: Trends and Prospects . London: Routledge.

Rubery, J., D. Grimshaw and H. Figueiredo (2005) ‘How to Close the Gender Pay Gap in Europe: Towards the Mainstreaming of Pay Policy’, Industrial Relations Journal , 36(3): 184–213.

Rutkowski, J.J. (1996) ‘Changes in the Wage Structure During Economic Transition in Central and Eastern Europe’. World Bank Technical Paper, Washington D.C.: The World Bank.

Rycx, F. (2003) ‘Industry Wage Differentials and the Bargaining Regime in a Corporatist Country’, International Journal of Manpower , 24(4): 347–366.

Rytina, S. (2000) ‘Is Occupational Mobility Declining in the United States?’ Social Forces, 78: 1227–1276.

Schmid, C. (2004) ‘Beached Russian “Immigrants”: The Politics of Citizenship and Language in Latvia’. Paper presented at the Annual Meeting of the American Sociological Association, San Francisco, United States.

Shepard, J. (1977) ‘Technology, Alienation, and Job Satisfaction’, Annual Review of Sociology , 3: 1–21.

290 References

Siebert, W.S. and J.T. Addison (1991) ‘Internal Labour Markets: Causes and Consequences’, Oxford Review of Economic Policy , 7: 76–92.

Simon, H. (2008) ‘International Differences in Wage Inequality: A New Glance with European Matched Employer-Employee Data’. MPRA Working Papers, University Library of Munich.

Skledar, S. (2004) Minimum wage under debate: Slovenia http://www.eurofound.europa.eu/eiro/2004/05/feature/si0405102f.htm [accessed 8 August 2011].

Smith, A. (1904, originally 1776) An Inquiry into the Nature and Causes of the Wealth of Nations . London: Methuen. Consulted on the internet: http://www.econlib.org/library/smith/smwn.html.

Smith, M., B. Burchell, C. Fagan and C. O’Brien (2008) ‘Job Quality in Europe’, Industrial Relations Journal , 39(6): 586–603.

Sorensen, A. and N.B. Tuma (1981) ‘Labor Market Structures and Job Mobility’, Research in Social Stratification and Mobility , 1: 67–94.

Soskice, D. (1994) ‘Reconciling Markets and Institutions: The German Apprenticeship System’, in L.M. Lynch (ed.) Training in the Private Sector . Chicago: University of Chicago Press.

Spector, P.E. (1997) Job Satisfaction. Application, Assessment, Causes and Consequences . London: Sage.

Stehrer, R. and T. Ward (2008) Recent Changes in the Job Structure of the EU: Technical Report . Dublin: Eurofound.

Stehrer, R., T. Ward and E. Fernández-Macías (2009) Changes in the Structure of Employment in the EU and their Implications for Job Quality . Vienna Institute for International Economic Studies Research Reports N. 354.

Stiglitz, J. (2010) Freefall, Why the Economic Crash Happened . London: Penguin. Streek, W. (2009) Re-Forming Capitalism: Institutional Change in the German

Political Economy . Oxford and New York: Oxford University Press. Svejnar, J. (2001) ‘Transition Economies: Performance and Challenges’. William

Davidson Working Paper Number 415, December 2001. Taylor, F.W. (1911) The Principles of Scientific Management . Harper & Row. Consulted

on the internet: http://www.marxists.org/reference/subject/ economics/taylor/principles/index.htm [accessed May, 2011].

Thurow, L. (1975) Generating inequality . New York: Basic Books. Tomaskovic-Devey, D. and S. Skaggs (2002) ‘Sex Segregation, Labor Process

Organization, and Gender Earnings Inequality’, American Journal of Sociology , July, 108(1): 102–128.

Topel, R.H., and M.P. Ward (1992) ‘Job Mobility and the Careers of Young Men’, Quarterly Journal of Economics , 107(2): 439−479.

Trifiletti, R. (1999) ‘Southern European Welfare Regimes and the Worsening Position of Women’, Journal of European Social Policy , 9(1): 49–64.

Trimikliniotis, N. and C. Demetriou (2005) ‘Active Civil participation of Immigrants in Cyprus’. Country report prepared for the European research project POLITIS, Oldeburg.

UNESCO (2006) International Standard Classification of Education . Paris: UNESCO. United Nations (2002) ‘International Standard Industrial Classification of

Economic Activities Revision 3.1’. United Nations Statistical Papers Series M., N. 4, Rev. 3.1.

United Nations (2009) International Migrant Stock: The 2008 Revision Population Database . New York: United Nations. Available at http://esa.un.org/migration.

References 291

US Council of Economic Advisors (1996) Job Creation and Employment Opportunities: The United States Labour Market, 1993–1996 . Office of The Chief Economist of the U.S.A.

Vallas, S.P. (1988) ‘New Technology, Job Content and Worker Alienation’, Work and Occupations , 15(2): 148–178.

Varblane, U., R. Eamets and K. Sõstra (2003) ’External Macroeconomic Shocks and the Estonian economy: How did the Russian Financial Crisis affect Estonian Unemployment and Foreign Trade?’, Baltic Journal of Economics , 3(2): 5–24.

Vidovic, H. (2007) Recent Changes in the jobs structure of the EU: Literature Review . Dublin: Eurofound.

Violante, G.L. (2008) ‘Skill-biased technical change’, in S.N. Darlauf and L.E. Blume (eds) The new Palgrave dictionary of economics . London: Palgrave Macmillan.

Visser, J. (2010) ICTWSS: Database on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts in 34 countries between 1960 and 2007 . Amsterdam: AIAS, http://www.uva-aias.net/208.

Weeden, K. (2002) ‘Why Do Some Occupations Pay More than Others? Social Closure and Earnings Inequality in the United States’, American Journal of Sociology , 108: 55–101.

Wilkinson, C. (2001) Fundamentals of Health at Work: The Social Dimensions . London: Taylor and Francis.

Wilkinson, R.G. (2006) The Impact of Inequality: How to Make Sick Societies Healthier . London: Routledge.

World Bank (2002) Lithuania: Country Economic Memorandum, Report No. 25005-LT http://www-wds.worldbank.org/servlet/WDSContentServer/WDSP/IB/2002/11/22/000094946_02111304010422/Rendered/INDEX/multi0page.txt.

Wright, E.O. and R.E. Dwyer (2000) ‘The American Jobs Machine: The Trajectory of Good and Bad Jobs in the 1960s and 1990s’, Boston Review , December Issue.

Wright, E.O. and R.E. Dwyer (2003) ‘The Patterns of Job Expansions in the USA: A Comparison of the 1960s and 1990s’, Socio-Economic Review , 1(3): 289–325.

Xie, Y. (1992) ‘the Log-Multiplicative Layer Effect Model for Comparing Mobility Tables’, American Sociological Review , June, 57: 380–395.

Yamaguchi, K. (1987) ‘Models for Comparing Mobility Tables’, American Sociological Review , 52: 482–494.

293

agency hypothesis, 202age structure, of immigrant

populations, 119agriculture, 6, 92, 98–9, 101, 163,

181, 184, 190, 194–5, 198–9, 246, 249

alienation, 30, 154–5American Jobs Machine, 1–2, 7apprenticeships, 206, 209, 210, 211,

213, 214, 225, 235association models, 238–42Austria, 37, 42, 78, 85, 88–9, 94, 109,

110n1, 119, 268automation, 175autonomy, 151, 155, 170

Baltic countries, 36, 116, 183, 196, 260–1

see also specific countriesbarriers to entry, 201Belgium, 42, 59, 81, 99, 103,

260, 268blue-collar workers, 11–12Braverman, Harry, 31, 155business services, 248, 249, 255

Central Europe, 42, 104see also specific countries

clerical work, 61, 101Coase, Ronald, 29cognitive dissonance, 152collective bargaining, 90, 161,

205–6, 235college education, 53, 122

see also educational levelcompensating differentials, 29–30,

153–4, 176, 201competition, 26construction sector, 14, 40, 41, 61,

66–7, 70–2, 127–8, 249, 254, 266, 269–72

contract employees, 48

coordinated market economies (CMEs), 110n1, 205, 207, 235

coordination problems, 205–6country of birth, 112, 119Current Population Survey (CPS), 55Cyprus, 58, 77, 84, 89, 116, 119,

137, 266Czech Republic, 42, 84, 85, 92, 93,

101, 113, 181, 183, 190, 194–5, 198, 199

degradation hypothesis, 32–3, 36, 40, 155

de-industrialization, 7, 53, 60demographics

immigration and, 111–46job growth by, 60–4, 75–6

Denmark, 42, 78, 84, 104, 165, 269deskilling, 155destandardization, of employment,

48, 49disamenities, 153, 158, 176division of labour, 26

in modern capitalism, 52structural changes in, 27, 31–3structural representation of, 28–9

downgrading, 108, 268–9see also degradation hypothesis

dual labour markets, 11, 202–3, 213Duncan Dissimilarity Index (DDI),

131–2Dwyer, Rachel, 2, 7–8, 27

Eastern Europe, 40, 42, 104, 181, 185, 186

see also post-socialist accession countries; specific countries

economic growth, 1, 138, 255economic mechanisms, as wage

determinants, 201–4economic policy, 13–14, 183,

187–9, 205

Index

294 Index

economic stimulus, 245educational level, 9–11, 29–31, 53,

149, 204, 218, 219, 255of immigrant populations, 122,

123–4job amenities index and, 170wages and, 21

education sector, 73, 101, 103, 105, 107, 109, 128–9, 255, 273

efficiency wage hypothesis, 203, 205

employmentby immigrants, 122, 125, 126in post-socialist accession

countries, 183–5employment conditions, 29–31, 32employment expansion/growth, 1–3,

6–7, 185, 244see also job creation/growthfactors affecting, 52impact of recessions on, 11–12,

244–78institutionalist theory of, 14, 54patterns of European, 1995–2007,

8, 10, 26–51, 52–74in post-socialist accession

countries, 192–8by sector and employment

relation, 39–48structural upgrading, 36in US, 8, 53–5using job quality measures, 173–5

employment policy, 4, 26–7, 72–4, 187–9

employment relations, 156, 202job creation by, 39–48

employment status, 47, 48employment structures

changes in EU, 147–79, 244–78destandardization of, 48, 49gender and, 75–110during recession and recovery,

244–78in US, 53–5

engineering industries, 32, 252–3, 256–8

entry-level jobs, 209–11, 224, 225, 235, 236, 237

Estonia, 36, 84, 94, 105, 137, 183, 184, 187, 189, 194, 196

EU-KLEMS project, 246EU-LFS project, 238Eurofound JOBS project, 3, 5, 54, 71,

111–14, 139, 142, 157, 173, 211, 212, 213, 238

European Community Household Panel (ECHP), 3, 20

European Labour Force Survey (ELFS), 3, 17

European Structure of Earnings Survey (ESES), 3, 20

European Survey on Income and Living Conditions (EU-SILC), 3, 20

European Trade Union Institute (ETUI), 165

European Union (EU)see also specific countriesemployment expansion in, 6–7,

26–51, 52–74employment shifts in, 147–79gender and job structure in, 75–110immigration to, 111–46impact of recession in, 244–78labour markets, 1–3, 159Lisbon agenda, 1, 3regional patterns of growth, 68–71

European Working Conditions Survey (EWCS), 149, 158–9

exploitation, 30, 154–5

fair wage-effort hypothesis, 207, 208female employment, see womenfinancial crises, 244–5

see also recessionsfinancial regulation, 14Finland, 42, 77, 84, 89firm, institutional theory of the, 29fisheries, 6, 101, 249flexibility, 170foreign-born workers, 61–4, 111–46foreign direct investment, 181France, 42, 58, 66, 78, 119, 151, 206,

207, 211, 216, 218, 220–2, 224, 234, 266

full-time workers, 48

Index 295

genderemployment gap, 8of immigrant populations, 122job concentration effects, 95–103job growth by, 60–4occupational segregation by,

95–103, 131–2patterns of job structures by,

75–110wage differences and, 81, 84–8,

99–103, 109work-life balance and, 156

General Industrial Classification of Economic Activities (NACE), 17–18, 244–5

geographic scales, 68–71Germany, 40, 42, 58, 66, 78, 85, 88,

92, 99, 109, 110n1, 206–7, 211, 212–13, 214, 215–16, 217, 218, 220–2, 224, 234–5, 260, 268, 271–2

globalization, 1, 9, 26, 33, 59, 142Gramsci, Antonio, 31Great Recession, 73

causes of, 244–5employment effects of, 244–78impact on jobs of, 3impact on labour markets, 11–12

Greece, 59, 77, 87, 94, 99, 103, 142, 181, 192

health and safety literature, 156health care services, 44, 73, 101, 103,

105, 107, 109, 128–9, 255, 273high-skill jobs, 75, 261–2, 274high-tech jobs, 271–2high-technology industries, 42, 43high-wage jobs, 40, 91, 92, 94Hispanics, 61, 63hospitality sector, 25203housing boom, 66–7Hungary, 42, 59, 77, 81, 85, 181, 183,

187, 189, 190, 194, 195, 197, 198

immigrants/immigration, 9, 61–4characteristics of, 119–26country of origin, 119definition of, 112–13

impact on wages of, 132job creation and destruction and,

132–8, 144job matches and, 138–42, 144labour market segmentation and,

111–46occupational segregation and,

126–38as percentage of total population,

115, 116income inequalities, 50, 54industrial democracy, 161industrial relations, 204–6, 235Industrial Revolution, 31, 32, 40industrial sector, 31, 40, 42–3industrial sociology, 30–1inequalities, 14, 33, 54

see also wage inequalitiesinformation technologies, 31, 33, 181infrastructure investment, 72institutional context, 36–7, 54,

64, 93employment growth and, 14as wage determinant mechanism,

204–38institutionalist economics, 30–1institutional theory of the firm, 29internal labour markets, 11, 59,

201–4, 206, 209–10, 211, 213, 218, 238

International Social Survey Program, 151

International Standard Classification of Occupations (ISCO), 17–18

international trade, 33Ireland, 40, 42, 58, 70, 77, 84, 107,

116, 119, 260, 266Italy, 42, 78, 93, 94, 181, 269

job amenities, 147–79skill level, wages and, 168–71

job amenities indexanalysis, 162–75data and construction of, 158–62employment distribution by

country using, 171–3at job level, 165–8net employment growth by, 173–5

296 Index

job classifications, 4, 17–18, 55–6, 204–5

job concentrationgender and, 95–103, 109immigrants and, 126–38

job creation/growth, 1–3see also employment expansion/

growthby employment relation, 39–48in Europe, 57–9by gender, race, and nativity,

60–4, 75–6at geographic scales, 68–71immigrants and, 126–38, 144low-wage, 92–4in middle quintile, 59patterns of, 1995–2007, 26–51,

52–74in post-socialist accession

countries, 185, 192–8by sector, 39–48sectoral patterns of, 64–7in US, 53–5, 57–74

job decline, 98job destruction, immigrants and,

126–38, 144job polarization, 7, 8, 12–14, 40,

44, 48, 49, 52, 174, 264–5, 268–9, 276

causes of, 54in construction sector, 66–7in Europe, 37–9, 268–9gender and, 60–4, 75–110in manufacturing sector, 64–5race, nativity, and, 60–4in service sector, 65–6in US, 53–74

job quality, 1–3, 29–31approaches to measuring, 9–10,

149–58compensating differentials and,

153–4dimensions of, 9, 147–8, 157gender gap in, 8impact of structural change on,

31–3job satisfaction and, 150–2multidimensional measure of,

147–79

net employment growth by, 173–5

in new economy, 53in post-socialist accession

countries, 180–200quintiles, 6, 10–11, 24ranking of, 6rankings, 19–21wages and, 4–5, 29–31, 147–8,

175–6jobs approach, 4–5, 7–8, 27, 48

methodology, 16–25theoretical foundations of,

28–33job satisfaction, 150–2job security, 59, 165jobs matrix, 16–18, 27–9, 57Jobs Study (OECD), 4, 113, 139, 212

knowledge-intensive services, 42, 44, 65, 74n3, 197–8, 273–5

labour, division of, see division of labour

labour conditions, 29–32, 154, 156, 158, 205–6

Labour Force Survey (LFS), 245labour markets

differentiation, 202–3dual, 11, 202–3, 213Europe, 2–3, 159flexible, 208functions of, 201–2immigrants in, 111–46impact of recessions on, 11–12,

244–78internal, 11, 59, 201–4, 206,

209–10, 211, 213, 218, 238jobs approach to, 28–33polarization. see job polarizationpolicy debates on, 1–2segmentation of, 8–9, 11, 31,

155–6, 176US, 1–2, 7–8, 49, 174as wage determinant

mechanism, 201–4labour mobility, 159, 201, 220, 222labour productivity, 4, 153–4, 203,

207, 251–3, 258

Index 297

labour unions, 59, 61, 73, 161, 235Laeken indicators, 148Latvia, 77, 108, 181, 183, 187,

190, 192–4liberal market economies (LMEs),

205, 207, 213, 235Lisbon Agenda, 1, 3, 4, 15n1, 27Lithuania, 77, 81, 92, 181, 183,

187–90, 192–4low-skill jobs, 65–6, 129–30,

185, 261low-technology industries, 42,

198, 271low-wage jobs, 40, 48, 50, 53, 66, 76,

265, 269gender and, 88–95, 109in post-socialist accession

countries, 187, 197Luxembourg, 37, 78, 84–5, 88, 94, 99,

116, 119, 268

manual job losses, 261–2, 274manufacturing jobs, 1, 53, 64–5, 72,

249, 252–8, 262, 269–72market economies, 10, 180,

185–6, 247see also coordinated market

economies (CMEs); liberal market economies (LMEs)

market equilibrium, 202Marx, Karl, 28, 30, 31, 154–5mass production, 32meaninglessness, 155median wages, 4, 29, 56, 88–9men

job concentration effects for, 96–7, 127–8

job growth for, 60–1job structures for, 77–83in low-wage jobs, 92, 94wages for, 85–8wage-skill gap, 85–8

Mexico, 151middle class, 14, 53, 72–3middle-wage jobs, 40, 50, 53,

59, 60, 72, 91, 92, 94, 174–5, 264, 265

migrant flows, 9, 13, 114–19, 142–3see also immigrants/immigration

migrant workers, see immigrants/immigration

minimum wage, 90, 92, 189–90mining, 249

NACE, 17–18, 244–5nationality, 112nativity, job growth by, 61–4neo-corporatism, 205Netherlands, 58, 66, 78, 97, 99,

104, 165new economy, 53New Member States, 42, 116

occupational mobility, 14occupational segregation

gender and, 131–2for immigrants, 126–38for women, 95–103

occupations, 56changes in structure of, 255–8

offshoring, 13, 175on-the-job training, 209, 211,

213–14, 218, 223organizational change, 26, 33over-qualified workers, 138–42

part-time employment, 77, 78, 80, 92, 97, 109

personal services, 14, 101, 107Poland, 181, 260polarization, see job polarizationPortugal, 42, 59, 66, 70, 78, 84, 87,

92, 99, 105post-industrialism, 32, 49, 59post-socialist accession countries

duration of transitional period in, 186–7

industrial development in, 181job distribution patterns in, 192–8job quality in, 180–200policy responses, 187–9self-employment in, 190–2, 196tax reforms in, 189–90transitional recession in, 182–6wage costs in, 189–90wage dispersion in, 180–1

powerlessness, 155power relations, 155

298 Index

primary sector, 40, 41, 155, 156, 262

private services, 42, 44, 45, 249processing industries, 256–7productivity, 32, 132, 153–4, 187,

203, 207, 251–3, 258public service sector, 44, 46, 76–7, 99,

103–8, 163, 261, 265

race, job growth by, 60–4, 75–6recessions

see also Great Recessionemployment effects of, 244–78impact on labour markets, 11–12lessons learned from previous,

258–9policy responses to, 245sectoral changes during, 248–55transitional, 182–6

retail sector, 101, 235–7Russia, 183

scientific management, 32secondary sector, 155–6segmentation theory, 155–6, 176segmented labour markets, 8–9, 11,

31, 155–6, 176self-employment, 48, 66, 185, 190–2,

196, 198self-estrangement, 155seniority, 209–11, 222, 224, 235services sector, 32, 262, 272–3

growth of, 65–6impact of recessions on, 254–5job creation in, 42, 44–8jobs in, 1, 11, 31, 40, 53, 72–3, 75knowledge-intensive services, 42,

44, 65, 74n3, 197–8, 273–5women in, 235–7

shirking, 202short-term work schemes, 277skill-biased technological change, 13,

49, 51n3, 174–5, 215skill levels

amenities and, 168–71gender differences in, 85–8immigration and, 138–42wages and, 168–71

skill production regimes, 205, 206skills-wages mismatches, 11,

201–43Slovakia, 36, 42, 78, 87, 94, 113, 187,

189–92, 194–5, 198Slovenia, 77, 84, 85, 88, 105, 183, 187,

189–90, 192, 195, 198small and medium enterprises

(SMEs), 66, 185, 197, 198Smith, Adam, 26, 28–31, 153, 201social exclusion, 50social inequalities, 33, 50social mobility, 14, 53, 59social policy, 13–14social psychology, 151, 152Social Sciences, 16, 26, 28, 153Social Sciences debate, 6, 7, 27, 31,

33, 49social services, 105, 107, 109sociology of work, 154Southern Europe, 40, 181, see also

specific countriesSpain, 40, 59, 66, 70, 77, 84, 87, 94,

99, 116, 119, 137, 190, 194–5, 199, 260, 266

standard employment relations, 48

Stiglitz, Joseph E., 2, 27stress, 151Structural Business Statistics

(SBS), 3, 20structural change, job quality

and, 31–3structural degradation hypothesis,

32–3, 36, 40, 155structural upgrading, 2, 7–8, 10,

32–3, 36–40, 44, 49, 50, 263, 266, 268, 275

Sweden, 78, 81, 89, 100, 104, 165, 206, 207, 211, 212, 13, 216, 218, 224–5, 234, 268

task biased technological change, 12–13, 33

task complexity, 150taxation, 189–90Taylor, F.W., 31teachers, 101, 103

Index 299

technological change, 8, 12–14, 26, 32–3, 49, 51n3, 52, 54, 59, 174–5, 215, 258

tenure, 202, 218, 220, 221, 222, 224

tertiarization, 7tournament model, 207–8transitional economies

duration of transition period in, 186–7

employment growth in, 10job quality in, 180–200policy responses, 187–9recession in, 182–6

turnover costs, 203

under-qualified workers, 138–42unemployment, 73–4

among immigrants, 122, 125, 126in post-socialist accession

countries, 196–7during recessions, 245US, 1

United Kingdom (UK), 58, 93, 94, 107, 206, 207, 211, 213–14, 216, 218, 222–4, 234

United Statesconstruction sector in, 66–7,

70–2economic growth, 1immigrant workers in, 61–4job growth in, 52–74job polarization in, 53–9labour market, 1–2, 7–8, 49, 174manufacturing sector in, 64–5regional patterns of growth,

68–71, 73service sector in, 65–6unemployment, 1

upgrading, 2, 7–8, 10, 32–3, 36–40, 155, 263, 266, 268, 275

U.S. Bureau of Labor Statistics (BLS), 55

utility, 153, 154

value-added, 246–50, 252–3varieties of capitalism (VoC)

approach, 205–6

wage inequalities, 1, 2, 50, 157, 176wages

determinants of, 201–38educational rankings and, 21endogenous determinants of,

207, 208gender differences in, 81, 84, 85–8,

99–103, 109impact of immigration on, 132institutional determinants of,

204–38job amenities and, 168–71job concentration effects for, 99job quality and, 4–5, 29–31, 147–8,

175–6median, 4, 29, 56, 88–9minimum, 90, 92, 189–90mismatch between skills and,

201–43in post-socialist accession

countries, 180–1productivity and, 4, 153–4qualifications and, 85–8relative, 29–31skill level and, 168–71working conditions and, 156, 158

welfare state, 8, 14, 31, 33, 76–7, 103–9, 138

well-being, 153, 156, 161white-collar workers, 11–12women

alienation of, 154–5employment growth for, 8job concentration effects for,

95–103, 109, 129, 130job growth for, 60–1job structures for, 75–110in low-wage jobs, 88–95, 109in part-time employment, 77, 78,

80, 92, 97, 109in retail sector, 235–7in service sector, 235–7skills-wages mismatches and, 85–8,

235–7wages for, 85–8in welfare state jobs, 103–9

worker participation, 157, 161worker representation, 157, 161

300 Index

workersblue-collar, 11–12exploitation of, 154–5immigrant. see immigrants/

immigrationqualifications of, 138–42well-being of, 153, 156, 161white-collar, 11–12

work experience, 209, 213, 220work hours, 170, 249–52, 258working class, 14working conditions, 29–32, 154, 156,

158, 205–6work-life balance, 138,

156, 163Wright, Erik Olin, 2, 7–8, 27

transformation of the employment structure in the eu and usa, 1995â€“2007

Documents