european company survey 2009 (ecs 2009) - uk data service · 2013-05-22 · tns network (see table...

69
European Company Survey 2009 (ECS 2009) Technical Report UK Data Archive Study Number 6568 - European Company Survey, 2009

Upload: others

Post on 23-Apr-2020

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

European Company Survey

2009 (ECS 2009)

Technical Report

UK Data Archive Study Number 6568 - European Company Survey, 2009

Page 2: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

2

European Company Survey (ECS 2009) Technical Report: Methodology, Questionnaire Development and Fieldwork prepared by

Arnold Riedmann (TNS Infratest Sozialforschung, Munich) With contributions from:

Dr. Martin Pfister (TNS Infratest Munich; chapter 6)

Kerstin Sleik (TNS Infratest Sozialforschung Munich)

Monika Fedor (TNS opinion, Brussels)

TNS Infratest Sozialforschung, Munich Landsberger Strasse 338 80687 Munich Germany on behalf of the

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

Munich, 20 October 2009 67.04.059813

Page 3: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

3

Contents Page

Preface 5

1. The design of the survey 6

1.1. The universe 6 1.2 The respondents 7 1.3 Data collection method 8 1.4 Topics 9 2. The questionnaires 11

2.1 Questionnaire development 11 2.2 Questionnaire translation process 12 3. Fieldwork preparations 16

3.1 Sampling 16 3.2 Programming of questionnaires 18 3.3 Briefing of interviewers 19 3.4 Material supporting fieldwork 20 4. Fieldwork 21

4.1 Fieldwork period 21 4.2 Number and structure of completed interviews 23 4.3 Non-response in the management interviews 26 4.4 Non-response in the employee representative interviews 30 4.5 Non-response analysis of the ER interviews: A multi-variate approach 34 4.6 Fieldwork monitoring 40 4.7 Specific fieldwork observations 40 5. Weighting 42

6. Level of confidence – Bootstrap variance test 44

6.1 Introduction 44 6.2 Bootstrap Theory 45 6.3 Results 46

Page 4: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

4

7. Data structure 49

7.1 Data processing and cleaning 49 7.2 Harmonization of sector codifications 49 7.3 SPSS data file 52 7.4 ASCII data file 55 7.5 Cross-tabulations 55 8. Recommendations for future surveys 56

8.1 Work organization 56 8.2 Timeframe 57 8.3 Identification and selection of respondents for the employee representative interviews 57 8.4 Reduction of the overload of questions asking for allowances at the end of the MM

interview 57 8.5 Questions asking for percentage shares 58

Annex A: List of tables in the report 59

Annex B: Country codes 60

Annex C: Data map for the ASCII file 61

Annex D: Syntax used for the NACE harmonization 67

Page 5: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

5

Preface

The European Company Survey (ECS 2009) is a large-scale representative survey among establishments in all EU-27 countries and three acceding and candidate countries (Croatia, Macedonia, Turkey). The survey was prepared and carried out on behalf of the European Foundation for the Improvement of Living and Working Conditions by TNS Infratest Sozialforschung, Munich (Germany) in co-operation with the international coordination centre of its sister company TNS opinion in Brussels (Belgium) and national fieldwork institutes of the TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated with an international group of renowned scientific experts in the subject matters under investigation as well in international survey research (see chapter 2.1). The ECS 2009 is the second EU-wide establishment survey launched by the European Foundation in recent years. The preceding survey – the European Establishment Survey on Working-time and Work-life Balance (ESWT 2004/05) had been conducted in 2004 in the EU-15 countries plus – in a second round in 2005 – in six of the ten countries that had newly joined the European Union on 1st May 2004 (Czech Republic, Cyprus, Latvia, Hungary, Poland and Slovenia). The ESWT 2004/05 had also been coordinated by TNS Infratest Sozialforschung. The main topics of the ECS 2009 are flexibility measures at the firm level and the involvement of employee representatives in decisions regarding their application (for more details on the topics cf. chapter 1.4). This “Technical Report” describes the development of the questionnaires, the translation process, the organization and outcome of the fieldwork and the basic principles applied for the sampling and the weighting of the data. It also contains some recommendations for future surveys of a similar kind.

Page 6: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

6

1. The design of the survey

1.1. The universe

The survey covers 30 European countries, namely the 27 EU Member States plus Croatia, Macedonia and Turkey. Interviews were carried out in establishments with 10 or more employees. Almost all sectors of activity were included in the sample, with the exception of the following sectors which are of limited quantitative importance considering the confinement of the universe to establishments with at least 10 employees: NACE Rev: 1.1 NACE Rev. 2 Agriculture, Forestry and Fishing A and B A Private Households P T Extraterritorial Organisations Q U All other sectors of activity, including Public Administration (NACE Rev. 1.1 L/ NACE Rev. 2 O) are part of the sample of the survey. All in all, the universe comprises about 3,2 million establishments with roughly 145 million employees. Table 1 below shows the size of the universe (establishments with 10 or more employees in the relevant sectors) for each of the countries involved. Figures are partially based on our own estimations, since exact statistical information about the universe is not available for several of the countries.1 Table 1.1.1: The size of the universe Country establishments employees (in '000) (in '000) Belgium 54 3.072 Bulgaria 37 1.641 Czech Republic 79 3.154 Denmark 34 2.115 Germany 564 27.842 Estonia 13 471 Ireland 20 1.359 Greece 57 1.637 Spain 282 10.015 France 327 18.629 Italy 300 11.221 Cyprus 5 178

1 For details cf. chapter 5 (“Weighting”).

Page 7: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

7

Latvia 17 697 Lithuania 24 1.080 Luxembourg 4 238 Hungary 69 2.439 Malta 2 111 Netherlands 96 4.968 Austria 48 2.262 Poland 176 8.213 Portugal 83 2.541 Romania 119 5.240 Slovenia 13 611 Slovakia 42 1.345 Finland 27 1.677 Sweden 74 3.297 United Kingdom 445 20.362 Subtotal EU-27 3.011 136.415 Croatia 18 892 Macedonia 6 260 Turkey 179 7.661 Subtotal 3 additional countries 203 8.813 Total all 30 countries 3.214 145.228

1.2 The respondents

In each establishment one management (MM) interview had to be conducted. The respondent for the MM interview was defined as the most senior person responsible for the personnel in the chosen establishment. In larger establishments this was normally the Human Resources Manager, in smaller units the Managing Director or – in the case of subsidiaries – the branch manager. In all organisations where a formal employee representation (ER) existed at the local level of the chosen establishment, additionally one interview with an employee representative was conducted wherever possible (ER interview). The proper respondent for this interview was the chairperson of the employee representative body in the establishment. In case of several bodies of employee representation coexisting within an establishment, the most important body was chosen for interviewing. This choice was made according to pre-defined general rules of selection. The programmed country specific filterings ensured that for each establishment the most appropriate ER respondent was selected, depending on the country situation and the specific constellation within the surveyed establishment.

Page 8: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

8

The country-specific definitions of the ER bodies eligible for interview were worked out in close cooperation between the network of EUROFOUND’s EIRO2 correspondents, TNS Infratest Sozialforschung and the European Foundation team. (For details on the chosen bodies please refer to the master questionnaires and its annexes.)

1.3 Data collection method

All interviews were carried out as Computer Assisted Telephone Interviews (CATI). All Interviews in the main data collection phase were carried out de-centrally by national fieldwork agencies located in the respective countries (see Table 1.3.1). Most of the institutes involved in the survey are part of the TNS group. In the few countries where no TNS fieldwork institute exists, long-standing cooperation partners of TNS (with cooperation experience from the Eurobarometer and other European studies) were responsible for fieldwork. Among the local fieldwork institutes carrying out the ECS 2009, 173 had already been responsible for the preceding survey ESWT 2004/05 and had thus collected experiences with this demanding survey design. Within each country, CATI interviewing was done centrally from one or more telephone studios in order to ensure an efficient sample management and a consequent quality control of the interviewing. Table 1.3.1 Local institutes in charge of field-work for the ECS 2009 Country Name Location and Website Austria Österreichisches

Gallup-Institut Vienna www.karmasin.at

Belgium TNS Dimarso Brussels www.tns-dimarso.com

Bulgaria TNS BBSS Sofia www.tns-bbss.org

Cyprus Cymar Market Research Ltd.

Nicosia www.cymar.com.cy

Czech Republic

TNS AISA Prague www.tns-aisa.cz

Denmark TNS Gallup DK Copenhagen www.tns-gallup.dk

Estonia TNS Emor Tallinn www.emor.ee

Finland TNS Gallup Oy Espoo www.tns-gallup.fi

France TNS SOFRES Montrouge Cedex www.tns-sofres.com

2 EIRO = European Industrial Relations Observatory 3 The ESWT 2004/05 had been carried out in 21 countries. In all countries with the exception of

Cyprus, Denmark, Luxembourg and Portugal the institutes that had been responsible for the ESWT 2004/05 were also responsible for the ECS 2009.

Page 9: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

9

Germany TNS Infratest Munich www.tns-infratest.com

Greece TNS ICAP Athens www.tns-global.com

Hungary TNS Hungary Budapest www.tns-global.hu

Ireland TNS mrbi Blackrock/ Co Dublin www.tnsmrbi.ie

Italy TNS Infratest Milano www.tns-italia.com

Latvia TNS Latvia Riga www.tns.lv

Lithuania TNS Gallup Lithuania

Vilnius www.tns-gallup.lt

Luxembourg TNS ILReS Luxembourg www.ilres.com

Malta Misco Valletta ettawww.miscomalta.com

The Netherlands

TNS NIPO Amsterdam www.tns-nipo.com

Poland TNS OBOP Warszawa www.tns-global.pl

Portugal TNS Euroteste Lisboa www.tns.pt

Romania TNS CSOP Bucharest www.csop.ro

Slovakia TNS Slovakia Bratislava www.tns-global.sk

Slovenia RM PLUS Maribor, www.rmplus.si

Spain TNS Demoscopia

Madrid www.tns-global.com

Sweden TNS Gallup Göteborg www.tns-gallup.se

United Kingdom

TNS UK Ltd London www.TNS-global.co.uk

Croatia Hendal d.o.o. Zagreb www.hendal.hr

Macedonia TNS Brima Skopje www.tnsglobal.com/global/europe/macedonia

Turkey TNS PIAR Istanbul www.tns-global.com.tr

1.4 Topics

Questionnaire development for the ECS 2009 took place in close cooperation between the European Foundation, its Advisory Committee, a group of researchers from various countries and TNS Infratest Sozialforschung. The thematic focus of the ECS 2009 is on different flexibility strategies used by firms in order to cope with challenges such as workload variations, problems in Human Resources Management or measures of restructuring or reorganisation. Among the flexibility measures

Page 10: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

10

dealt with in the survey there are different forms of working-time flexibility and contractual flexibility, variable forms of pay and new forms of work organisation such as work in autonomous teams. The management survey investigates the application of these flexibility measures, the shape they take in the establishment and – in some instances – the rationale behind their application. The employee representative questionnaire deals with the degree and form of influence the establishment level employee representatives exert on these strategies. In more general terms, it also deals with the role of the employee representation in general, with its resources and the degree of influence it exerts on important management decisions that are tackling the situation of the employees. More in particular, the survey deals with the following issues: a) Working time arrangements: − part-time work − extended operating hours (night work, week-end work, shift work) − flexible working time arrangements (e.g. flexi-time, working time accounts) − overtime These flexibility arrangements were already investigated in the ESWT 2004/05. Part of the questions in the ECS 2009 was formulated identically to the ESWT questions in order to allow for an analysis of trends over time. b) Other flexibility measures: − contractual flexibility (fixed-term contracts; temporary agency work; use of freelancers) − variable elements of pay (elements of pay depending on the performance of the individual,

the working group or the establishment, share ownership schemes) − new forms of work organisation (work in teams) − further training of employees c) Social dialogue practice: − existence of different types of formal employee representation at the establishment level

(trade union representation, works council, other types of representation) − resources of the employee representatives (available time, provision with information etc.) − Influence of the employee representation on establishment level decisions − Influence of the employee representation on the application and shape of the flexibility

arrangements dealt with in the management questionnaires − Forms of employee involvement (direct consultation of employees, queries of employees

directed to the formal employee representation)

Page 11: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

11

2. The questionnaires

2.1 Questionnaire development

The questionnaires were developed in close co-operation between TNS Infratest Sozialforschung and the following researchers: − the research team responsible for the ECS 2009 at the European Foundation (Greet

Vermeylen, David Foden, Branislav Miculic, Agnès Parent-Thirion, Kasia Jurczak, Radoslaw Owczarzak, Stavroula Demetriades, Felix Wolf),

− The Advisory Committee set up for this project, consisting of representatives from trade union federations, employer federations, governments, EU representatives and scientists

− experts from a number of different European countries and research institutes: − Dr. Guy van Gyes and Monique Ramioul (HIVA, Catholic University of Leuven,

Belgium) − Dr. Marcel Kerkhoffs, Prof. Dr. Peter Ester and Heejung Chung (OSA, Tilburg

University, Netherlands) − Prof. Dr. Dominique Anxo (Centre for European Labour Market Studies, Gothenburg,

Sweden) − Prof. Dr. Rafael Munoz de Bustillo Llorente (Universidad de Salamanca, Spain) − Peter Ellguth and Dr. Lutz Bellmann (Institut für Arbeitsmarkt- und Berufsforschung

der Bundesagentur für Arbeit, Nuremberg, Germany) − Prof. Dr. Colette Fagan (Manchester University/United Kingdom) − Dr. Spartak Keremidchiev (Institute of Economics at the Bulgarian Academy of

Sciences, Sofia, Bulgaria) − Dr. Jens Lowitzsch (Osteuropa Institut, Freie Universität Berlin, Germany)

At the contractor’s side, the main responsibility for the questionnaire concept in general and for the development of the MM questionnaire was with Harald Bielenski and Arnold Riedmann from TNS Infratest, while the main responsibility for the development of the ER questionnaire was shared between Guy van Gyes from the University of Leuven and Arnold Riedmann. All in all, the process of jointly developing the final master questionnaires lasted from February 2008 until the end of October 2008. During this period seven draft versions of both the MM-questionnaire and the ER-questionnaire were developed and circulated among the members of the research team. The team members commented via email on the various versions. Questionnaires were also personally discussed between TNS Infratest, the research team of EUROFOUND and the expert group in a meeting in Dublin on 07/08 April 2008. In another meeting on 27/28 May 2008, the Advisory Committee of the European Foundation for this project had the opportunity to personally discuss the questionnaire drafts with the

Page 12: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

12

EUROFOUND team and TNS Infratest and to propose revisions. The questionnaires were thus the result of a long and intense period of reflections, discussions and revisions.4 The 6th draft, the English master versions of the MM and ER master questionnaires of 22th July 2008, was pre-tested in a quantitative pilot carried out in about 320 establishments in 8 countries (40 establishments per country), representing different geographical regions and different economic structures within the European Union (Germany, Spain, France, Hungary, Italy, Poland, Sweden and the United Kingdom). For the pre-test, the master questionnaires were transmitted to the TNS Quest translation tool and translated by members of the pool of translators maintained by TNS Opinion. Each national version was then thoroughly checked by native speakers and revised where necessary (for details on the translations for the pilot phase see separate Pre-test Report of 15/10/2008, chapter 1.3). The pre-testing in the 8 countries revealed a need for substantial shortenings of the master questionnaires and for slight revisions of some of the questions to remain. On the basis of the recommendations in the Pre-test Report and the expert feedback received on these, TNS Infratest Sozialforschung and the research team of EUROFOUND discussed the revisions and shortenings of the questionnaire in a joint meeting held on 22/23 October 2008 in Dublin. At this occasion, the decision was made to substantially shorten the questionnaire parts dealing with working-time arrangements in order to allow for a sufficiently differentiated coverage of the newly introduced topics. Among the trend questions allowing a comparison of developments since 2004/05 (ESWT), therefore only a limited set could be maintained. By 10 November 2008, the decisions on the revisions were implemented in the international master questionnaires and the new master questionnaires were acknowledged by EUROFOUND. Due to the substantial shortenings of some parts of the survey it was necessary to rearrange some of the questions in order to maintain a straightforward and logic structure of the questionnaires. This rearrangement is the cause for the partly illogical sequence of question numbers.

2.2 Questionnaire translation process

Questionnaires had to be translated into all official national languages of the countries covered by the study. All in all, 38 different country versions were produced (see table).

4 We would like to thank all colleagues who were involved in the development of the questionnaires

for their advice and their excellent co-operation.

Page 13: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

13

Table 2.2.1: Available national questionnaire versions

Country Language version(s)

AT German

BE French, Flemish

BG Bulgarian

CY Greek

CZ Czech

DE German

DK Danish

EE Estonian, Russian

EL Greek

ES Spanish

FI Finnish, Swedish

FR French

HR Croatian

HU Hungarian IE English (no translation required, only adaptations to national ER terminology and national

language peculiarities)

IT Italian

LT Lithuanian

LU Luxembourgish, French, German, English

LV Latvian, Russian

MK Macedonian

MT Maltese, English

NL Dutch

PL Polish

PT Portuguese

RO Romanian

SE Swedish

SI Slovenian

SK Slovakian

TR Turkish

UK English (no translation required, only adaptations to national ER terminology and national language peculiarities)

At the beginning of the translation process, the final master questionnaires were transmitted to the TNS Quest translation tool, hereby creating a separate master file for each country version. This was necessary because some of the questions related to the systems of formal employee representation have country specific variations, with differences in the shape of the items and in the filtering.

Page 14: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

14

24/11/2008 – 08/12/2008: The TNS Quest file was sent to the network of expert translators of TNS opinion. The translators of this network have years of experience in translating survey research documents for multi-country surveys. They are all native speakers and only work into their own language in their respective country. They were briefed on all the specific instructions for this questionnaire (use of specific expressions and terminology, etc.). 26/11/2008 – 15/12/2008: In a first step of quality control, all translated questionnaire versions were then sent to the corresponding local fieldwork institute in each country for proofreading. All the changes made in this step were flagged in the process. 05/12/2008 – 22/12/2008: In a second step, proofread translations were sent to the European Foundation to be submitted to their network of EIRO correspondents. The checks done by this group of experts focused on the correct implementation of the terminology for the national system of employee representation and on other specific terminology used in the survey. 16/12/2008 – 22/01/2009: The feedback from the EIRO correspondents was implemented by the coordination team of TNS Infratest and TNS opinion respectively sent to the national TNS institutes for implementation (for languages none of the team members was sufficiently familiar with). 19/12/2008 – 15/01/2009: The next step was the back translation of the local questionnaires into English by specific translators. These were different from those who did the initial translation. The back translators were new on the project and had not seen the source questionnaire in English. 14/01/2009 – 29/01/2009: Back-translations were then analyzed by native speakers in order to determine if differences between the original master in English and the back translated questionnaires were caused by errors in the translated local version. 15/01/2009 – 30/01/2009: TNS opinion screened the back translation analyses and then sent them to the national institutes so they could correct the questionnaire if an error was discovered. Changes made as result of the back-translations were documented.

Page 15: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

15

The following illustration shows a concluding overview of the whole translation process as described above. Table 2.2.2.: Translation process

Final English master questionnaire provided by TNS Infratest SozialforschungFinal English master questionnaire provided by TNS Infratest Sozialforschung

Transformation of the master questionnaires (word format) into the Excel-translation-tool (TNS Quest) by TNS Opinion and TNS Media, hereby producing a separate

English master version for each country

Transformation of the master questionnaires (word format) into the Excel-translation-tool (TNS Quest) by TNS Opinion and TNS Media, hereby producing a separate

English master version for each country

Translation of the master questionnaire (Excel-file) into 38 language versions by professional translators

Translation of the master questionnaire (Excel-file) into 38 language versions by professional translators

Checks of the translations by the local field institutesChecks of the translations by the local field institutes

Check and implementation of the field institutes' feedback by TNS Opinion and TNS Infratest Sozialforschung

Check and implementation of the field institutes' feedback by TNS Opinion and TNS Infratest Sozialforschung

Expert check of the translations by network of EIRO correspondentsExpert check of the translations by network of EIRO correspondents

Check and implementation of the expert feedback by TNS Infratest Sozialforschung, TNS Opinion or national fieldwork institutes

Check and implementation of the expert feedback by TNS Infratest Sozialforschung, TNS Opinion or national fieldwork institutes

Back translation of the thus revised language versions into English by professional translators different from original translators

Back translation of the thus revised language versions into English by professional translators different from original translators

Comparison of back translations and English master versions by independent language experts

Comparison of back translations and English master versions by independent language experts

Check of the back translation reports by TNS OpinionCheck of the back translation reports by TNS Opinion

Final language versions used for programmingFinal language versions used for programming

Implementation of the acknowledged changes by the local field institutesImplementation of the acknowledged changes by the local field institutes

Page 16: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

16

3. Fieldwork preparations

3.1 Sampling

Address registers Address registers to be selected for the ECS 2009 had to meet various requirements such as a consequent listing of both the headquarters and the subsidiaries in case of multi-site enterprises, completeness regarding the relevant sectors and availability of information on the sector and on the size-class for each listed address. In several countries address registers meeting all requirements of the study were not available. The best available registers often showed important shortcomings: − In 17 of the 30 countries address registers were available at company level only. In the

case of multi-site companies, these address registers contain only one entry (the headquarters), while the further local units (subsidiaries) are not listed.

− In some countries address registers did not include Public Administration and/or included the sectors “Education” and “Health and Social Work” only partially (mostly only the private organisations within these sectors).

These shortcomings were analysed before fieldwork start and were dealt with as follows: − In countries where no suitable establishment based register was available, the best

available company register was used and an additional screening procedure was applied in order to get a random selection of establishments from the company based sample (see questionnaire questions MM050 to MM053c). The screening procedure had to be applied in 17 of the 30 countries surveyed (see table 3.1.1).

− In countries where the best available register contains addresses of the sectors “Public

Administration”, “Education”, “Health and Social Services”, but under-represents one or more of these sectors, the under-representation was compensated for in the drawing of the gross sample: Additionally to the strictly randomized drawing of addresses within the sector “Services”, additional addresses were drawn from those sectors which were known to be underrepresented in the address source. By this way it was guaranteed that these sectors would be sufficiently represented in the final net samples.

− In countries where the best available register showed clear systematic shortcomings

regarding the coverage of the sectors with a high share of public organisations (“Public Administration”, “Education” and “Health and Social Services”), addresses for the under-represented sectors were drawn from additional address lists (such as lists provided by the respective Ministries or Yellow Pages telephone registers). For the sampling from these lists, the concerned countries received specific quota for NACE Rev. 1.1 L, M and/or

Page 17: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

17

N (respectively NACE Rev. 2 O, P and/or Q). These quota were based on the available statistical information about the quantitative relevance of these sectors.

− In some of the countries where addresses had to be drawn from two (or more) registers,

only the chosen main address register was company based while the additional register for the public sector(s) is establishment based. In these cases the screening procedure was applied only to the main address source and not to the addresses drawn from the additional lists.

Table 3.1.1: Address registers used for the survey

Country Main address sourceSampling unit

(C = Company, E = Establishment)

Screening applied (MM050-MM053c)

Additional address source

used (Public Sector(s))

Belgium Belfirst (Bureau Van Dijk) C yes xBulgaria Bulgarian National Statistical Institute C yes xCzech Republic Albertina C yes -Denmark KOB (Kobmandstandens oplysningsbureau) E no -Germany AMS (Arbeitsstätten Master Sample) E no -Estonia Estonian Business Register C yes -Ireland BILL MOSS Partnership C yes -Greece ICAP Business Databank C yes xSpain Schober E no xFrance SIRENE E no -Italy Dun & Bradstreet E no -Cyprus Business Register of Statistical Service C yes x

Latvia Business Register of Central Statistical Bureau E no -

Lithuania JSC Creditreform C yes -Luxembourg EDITUS E no -Hungary HBI LTD C yes -Malta Employment & Training Corporation 2007 C yes -

Netherlands Chamber of Commerce Establishment Register E no x

Austria Dun & Bradstreed B2B Austria 2006 E no xPoland PCM (Polskie Centrum Marketingowe) E no -Portugal Informa Dun & Bradstreet C yes xRomania www.ListaFirme.ro C yes -Slovenia IPIS Register of Slovene Companies C yes -Slovakia Albertina database C yes -Finland Profinder BtoB by Fonecta E no -Sweden SCB Företagsregistret E no -UK Dun & Bradstreet E no -

Croatia Institute for Business Intelligence C yes -Macedonia Central registry C yes xTurkey Chamber of Trade and Commerce C yes x

Page 18: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

18

Sampling matrix For sampling, a matrix consisting of five size-classes and two sectors of activity (“Producing Industries” and “Services”) was used. For each of the 10 cells of this matrix quota were given. The quota were adapted to the structure of the national economy and therefore differed somewhat between the countries. Table 3.1.2: Sampling matrix

Sector: Size class:

1. Producing Industries (NACE Rev. 1.1 C-F resp.

NACE Rev.2, B-F)

2. Service Sector (NACE Rev. 1.1 G to O resp.

NACE Rev.2, G-S) 10 – 19 employees 20 – 49 employees 50 – 249 employees 250 – 499 employees 500 + employees

The applied disproportional sampling design shows the following characteristics: − As regards the size-classes, the applied sample design reflects a mix of establishment-

and employee proportionality. This approach was chosen because a strict establishment proportional design would result in extremely few interviews in the largest size classes (most establishments are rather small) while a strictly employee proportional sample would be hard to put into practice in view of the limited absolute number of large establishments. The approach also has the big advantage that it keeps the weighting factors for the establishment and employee proportional weighting relatively homogenous.

− The “Producing Industries” were intentionally oversampled to a certain degree in order to

get a sufficiently high number of interviews from this important sector. − Within the 10 cells of the sampling matrix, addresses were always drawn at random.

3.2 Programming of questionnaires

For all fieldwork institutes using the software “NIPO Odin” or “Bellview”, the programming of the CATI script was done centrally by programmers of the international programming unit at TNS Dimarso, our Belgian sister company. Central scripting for NIPO Odin took place from 29/01/2009 to 04/02/2009. The scripting for NIPO Odin was based on the CATI script that had been produced for the quantitative pre-tests. Central scripting for Bellview was made between 09/02/2009 and

Page 19: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

19

24/02/2009. For Bellview, scripting had to be done from scratch since the pre-test had been programmed centrally in NIPO Odin only. The centrally programmed script was thoroughly tested by TNS Infratest and TNS Opinion and then distributed to the country institutes. The country institutes received a general script in their local languages. They were in charge of adapting the programme to their local needs (e.g. the locally used Sample Management System) and of testing and finalising the script. Countries using different CATI software (e.g. Quancept) had to do the programming on their own, based on the centrally provided master questionnaires. The local project managers were responsible for testing the programmed CATI scripts. Local scripting took place between 23/01/2009 and 10/03/2009, with some countries starting and finishing earlier than others. Table 3.2.1 Software versions used in the countries

Central scripting Odin Bellview

Local scripting

AT DK CY

BE ES EE

BG FI FR

CZ IE HR

DE IT MT

EL LT SE

HU LV SI

LU SK

MK UK

NL PL PT RO TR

3.3 Briefing of interviewers

The introduction of the interviewers to the specific challenges of this survey was made by the supervisors in the local telephone studios and by the local project managers in the fieldwork institutes. To this end, the following measures were taken: General written guidelines for briefing the interviewers on the specific challenges of the survey were provided to all institutes by the coordination team. Among other issues, these guidelines contained: − Information about the contents and aims of the survey − Information about the European Foundation (tripartite structure, tasks etc.)

Page 20: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

20

− Information about the supporting material available for this study (recommendation letter, link to the website of the European Foundation with information placed for respondents etc.)

− An emphasis on the importance of the survey and fieldwork quality − Strategies for getting high response rates − Explanations on the specific difficulties related to the ER interviews and the linkage

between both the MM and ER interviews − Hints on specific questions for which additional explanations were thought to be helpful. The CATI supervisors and fieldwork managers of all countries were additionally briefed on the survey via an interactive TV session broadcasted on 12th January 2009 from the offices of TNS Opinion. The TV conference was used by the coordination team of TNS Infratest Sozialforschung and TNS opinion for providing further information on the survey concept and on the specific challenges the interviewers might be faced with. The survey team of the European Foundation responsible for the ECS 2009 also actively participated in the TV conference, giving presentations on the nature and tasks of the European Foundation and on the importance of the survey in the context of their research activities. The presence of the EUROFOUND team in the TV session served as a measure to raise awareness and motivation of the responsible fieldwork managers and CATI supervisors for this demanding survey. During and at the end of the TV conference, questions raised by the fieldwork managers or CATI supervisors were directly answered by the project coordinators of TNS Infratest and TNS Opinion and by the European Foundation staff. The questions and answers were additionally summarized in written form and provided to the countries as additional input for the interviewer briefings.

3.4 Material supporting fieldwork

The European Foundation issued an official recommendation letter aimed at raising the acceptance of the survey among respondents by emphasizing its importance and by serving as verification of the project. The recommendation letter contained more information on the purpose of the survey and a reference to the website of the European Foundation where further information on the survey was available. It also mentioned the support of the survey by Businesseurope (The Confederation of European Businesses) and ETUC (The European Trade Union Confederation). The recommendation letter was translated by the local institutes. Additionally, further information on the survey (including links to the ESWT results) was presented on the website of the European Foundation. This information served both as a motivator for potential respondents in doubt about the usefulness of the survey and as a verification of the project.

Page 21: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

21

4. Fieldwork

4.1 Fieldwork period

The fieldwork period lasted from 27 January 2009 to 05 May 2009. In the majority of countries, fieldwork was mainly done between mid-February and end-April 2009. The start and end date of the fieldwork period vary somewhat between the countries: Sweden was the first country to start fieldwork (27 January 2009) while Estonia was the latest one (02 March 2009). The first country to finish interviewing was Latvia with fieldwork ending at the 06 March 2009, the last country to finish was Turkey (05 May 2009). Interviewing started in each country immediately after finalisation of CATI-programming. There were several reasons for the staggered start of fieldwork in the various countries:

• Some countries had to programme the questionnaires on their own because they do not use the CATI software in which the script was provided.

• Some countries had difficulties in adapting the centrally provided script to their local sample management system, especially as far as the complex interface between MM and ER interviews is concerned.

• Estonia started late due to incompatibilities of their NIPO ODIN software version with the version in which the central script was provided. They therefore had to programme the questionnaires from the scratch.

• For some countries, the feedback from the expert questionnaire checks came in later than for others. For these countries, therefore the back-translation process and the delivery of their final script were subsequently also delayed.

The overall length of fieldwork period finally needed for accomplishing the ordered number of MM interviews and the maximum available number of ER interviews also varies: Some countries finished fieldwork after about 4 weeks already while others needed the full available period of up to 3 months (depending on the fieldwork start). Among the reasons for this are differences in the response behaviour between countries (response rates, importance respondents attribute to the survey etc.) or differences in the number of interviewers working for the study (see table 4.1.1 below). The fieldwork period was in each country dimensioned so that fieldwork for ER interviews continued for at least one week after the finalisation of the MM interviews. This measure ensured that there was sufficient time to try to get ER interviews for those establishments where the MM interview had been carried out only towards the very end of the fieldwork period.

Page 22: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

22

Table 4.1.1: Fieldwork period and number of interviewers working for the project, by country

Country Beginning End No. of

Interviewers Belgium 18.02.09 24.04.09 27 Bulgaria 11.02.09 21.04.09 18 Czech Republic 17.02.09 22.04.09 37 Denmark 18.02.09 15.04.09 22 Germany 05.02.09 03.04.09 52 Estonia 03.04.09 21.04.09 7 Ireland 26.02.09 24.04.09 7 Greece 20.02.09 22.04.09 31 Spain 18.02.09 27.04.09 38 France 18.02.09 15.04.09 85 Italy 19.02.09 28.04.09 63 Cyprus 09.02.09 13.04.09 25 Latvia 19.02.09 06.03.09 15 Lithuania 23.02.09 24.03.09 13 Luxembourg 18.02.09 25.03.09 12 Hungary 13.02.09 25.03.09 41 Malta 25.02.09 23.04.09 9 Netherlands 18.02.09 24.04.09 24 Austria 10.02.09 24.04.09 25 Poland 12.02.09 24.04.09 67 Portugal 12.02.09 22.04.09 50 Romania 17.02.09 01.04.09 12 Slovenia 02.03.09 19.04.09 12 Slovakia 16.02.09 11.03.09 50 Finland 16.02.09 20.04.09 11 Sweden 27.01.09 16.04.09 15 United Kingdom 09.02.09 24.04.09 39 Croatia 11.02.09 22.04.09 3 Makedonia 25.02.09 15.04.09 7 Turkey 20.02.09 05.05.09 25

After completion of approximately 30 interviews fieldwork had to be stopped and an interim data set had to be sent to the TNS opinion coordination centre for checking. The interim data sets for each country were checked with regard to technical correctness of the programmed CATI (completeness, filters etc.) and the structure of the data file (card-column-format, variable names, codes). Within about an hour, fieldwork institutes received an automated check report.

5 This figure is comparatively low because it refers only to the interviewers working permanently on

this survey. In peak times with numerous appointments, more interviewers were used.

Page 23: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

23

4.2 Number and structure of completed interviews

Number of MM and ER interviews All in all 27,160 MM interviews and 6,569 ER interviews were carried out within the fieldwork period. The foreseen number of MM interviews was reached in all countries, in a couple of countries even more MM interviews were conducted than originally foreseen. Therefore the total number of MM interviews across all 30 countries is notably higher than originally planned.6 Regarding the ER interviews, it was assumed that on the average of all countries it would be possible to conduct ER interviews in approximately 25% of the MM interviews. This aim was almost reached, with ER interviews being carried out in 24,2% of all surveyed establishments. Results for the single countries however vary largely: In the Nordic countries, the highest shares of ER interviews could be accomplished (Finland: 57%, Sweden: 54%, Denmark: 39%). The number of achieved ER interviews is well above the average also in Germany (37%), Luxembourg (33%) and Croatia (32%). On the other hand, in a couple of countries only in less than 10% of the establishments an ER interview could be completed (Portugal: 4%, Turkey: 4%, Malta: 7%, Greece: 7%, Portugal 4%). The national differences can be explained by two factors: On the one hand the incidence of employee representative bodies varies largely due to factors such as differences in the national history and culture of employee representation or differences in the legally defined size thresholds for the set-up of an employee representation. On the other hand there are considerable national differences in the willingness of managers and employee representatives to agree in an ER interview. It also has to be noted that the percentage share of completed ER interviews depends on the structure of the net MM sample – especially its distribution over the size classes: The larger an establishment is, the more likely it is that there is an employee representation.

6 It was planned to conduct MM interviews in 26.800 establishments across all 30 countries.

Page 24: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

24

Table 4.2.1: Number of completed MM and ER interviews per country

Country MM-Interviews ER-Interviews ER in % of MM Belgium 1,016 287 28% Bulgaria 502 128 25% Czech Republic 1,014 242 24% Denmark 1,023 394 39% Germany 1,500 558 37% Estonia 500 72 14% Ireland 503 93 18% Greece 1,005 76 7% Spain 1,509 375 25% France 1,500 441 29% Italy 1,502 320 21% Cyprus 505 60 12% Latvia 509 147 29% Lithuania 560 128 23% Luxembourg 501 164 33% Hungary 1,045 183 18% Malta 349 24 7% Netherlands 1,002 249 25% Austria 1,016 205 20% Poland 1,500 367 24% Portugal 1,012 39 4% Romania 500 137 27% Slovenia 536 153 29% Slovakia 520 100 19% Finland 1,000 565 57% Sweden 1,001 541 54% UK 1,510 166 11% Subtotal EU-27 24,640 ,214 25% Croatia 500 162 32% Makedonia 520 129 25% Turkey 1,500 64 4% Subtotal 3 new countries 2,520 355 14% Total all 30 countries 27,160 6,569 24%

Page 25: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

25

Sector and size structure of the net sample The table below shows the share of interviews realized in the various cells of the sampling matrix. Table 4.2.2: Structure of the net sample

Sector: Size class:

1. Producing Industries(NACE Rev. 1.1 C-F /

NACE Rev.2, B-F)

2. Service Sector (NACE Rev. 1.1 G to O / NACE Rev.2, G-

S)

Total size-class

10 – 19 employees 11% 16% 27% 20 – 49 employees 11% 16% 27% 50 – 249 employees 12% 17% 29% 250 – 499 employees 5% 5% 10%

500 + employees 3% 4% 7% Total sector: 42% 58% 100%

Due to the disproportionality of the sample design as regards both the size-classes and the ratio between the two sectors “Producing Industries” and “Services” (see chapter 3.1), a comparison of the unweighted net sample structures with the structures of the universe does not make sense. These “distortions” were intended and differences resulting from such a comparison would not allow for any conclusions on the quality of fieldwork.

Page 26: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

26

4.3 Non-response in the management interviews

For the description of the reasons for non-responses a difference has to be made between the countries where a special screening procedure was necessary in order to come from an originally company based address source to a sample at establishment level (see chapter 3.1 above) and the other countries where no such procedure was necessary. The screening countries had to undergo a 2-step selection process, implying a higher overall risk of non-response for all multi-site units within the sample. When comparing the structure of non-responses between countries it also has to be noted that differences might be due to given facts as well as due to different national habitudes in expressing a denial to participate.7 Another factor explaining some of the differences is that each national institute has developed own routines in coding non-responses. Even with a template provided for a harmonized non-response recording, therefore differences in the interpretation and application of the codes by the country institutes can not always be fully avoided. Table 4.3.1 shows the non-responses for the non-screening countries while table 4.3.2 shows non-responses for the screening countries. It can be seen that response rates vary considerably between the countries. Among the non-screening countries, Latvia has the highest response rate (54%), while the response rate in the Netherlands (11%) is the lowest within this group. For the screening countries, the analysis of non-responses is more complicated: For the interviews with multi-site organizations, in these a 2-step selection process had to be applied while the interviews with single-site establishments had the same 1-step selection process as the interviews from the non-screening countries (because in these cases the screening interview immediately led to the main interview, without the necessity of a further contact). To differentiate non-response reasons between multi-site and single-site organizations is not possible since this information can not be recorded separately. Overall, the response rates for the screening countries were in line with those from the screening countries or even a bit higher, with the highest rate reported from Greece (65%) and the lowest rates reported from Hungary and Croatia (17% each). The overall higher level reported from the screening countries is probably due to the fact that most of these countries are New Member States. There, the willingness of enterprises to participate in surveys is obviously generally still higher than in the old member states where a lot of company level research has been taking place over the years.

7 E.g. if the respondent says that he or she has “no time” for an interview during the fieldwork period

this can actually be due to time constraints, but may also be a sort of “polite” general refusal. Depending on the interviewer’s assessment this might therefore be coded as “no interview possible during fieldwork period” in some cases and in other cases as a “refusal due to time reasons” or even as a general refusal to participate in this type of surveys.

Page 27: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

27

Apart from differences in the habitudes of expressing refusals on part of the respondents and apart from possible differences in the interpretation of the non-response categories in the countries, there are a number of factors explaining differences in the response rates:

• Firm-level survey research has become more and more common in several of the countries involved in the survey. Therefore, more and more firms advise their managers generally not to participate in surveys (any more). This is e.g. among the most important reasons for the drop in response rates in Poland as compared to the ESWT 2005.

• There are national differences in the degree of acceptance of the European Union. In

countries where a positive attitude towards the European Union prevails, response rates in a survey conducted for an institution associated to the European Union can be supposed to be higher than in others.

• In some countries, there are widespread doubts about the confidentiality of the

answers given in such a survey. This was e.g. reported as a frequent reaction of respondents from Turkey.

• The current economic crisis has hit some countries more or earlier than others. It was

observed during fieldwork that several contacted persons (both on the MM and the ER part) refused their participation with explicit reference to this crisis. For management respondents saving measures applied as a reaction to this situation included a reduction of the participation in all “non-essential” activities. Employee representatives in turn sometimes expressed fears that their participation in the survey might be detrimental to their further employment prospects. It is however difficult to say in how far the crisis sometimes served as a pretext for the refusal.

Page 28: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

28

Table 4.3.1: Non-responses MM interviews non-screening countries

DK DE ES FR IT LV LU NL AT PL FI SE UKGross sample I:Total number of addresses used for the survey: 5.544 12.386 19.970 6.931 10.202 1.202 2.435 12.089 6.629 13.008 4.239 2.795 7.047

• Establishment does not exist: 579 27 2.208 0 0 20 28 1.087 69 376 0 80 128• Line dead; fax/modem, wrong telephone number: 155 438 909 393 867 130 65 283 33 378 77 221 86• Telephone not answered after a minimum of 10 futile contacts: 831 1.093 5.357 695 352 5 34 216 504 315 0 103 401• Line always busy or answering machine: 16 169 902 21 0 0 6 42 1.787 29 1.648 0 53• Private households (out of scope): 2 31 410 0 344 4 0 85 221 113 1 17 37• Less than 10 employees in MM102a/b (out of scope): 60 858 1.632 122 237 109 74 74 0 430 122 172 624• No answer in MM102a/b: 0 0 3 0 12 0 0 3 0 51 1 0 0• Quota for the cell already full: 0 125 3 0 12 0 0 821 0 3.202 0 0 0

Gross sample II:Total number of valid addresses: 3.901 9.645 8.547 5.700 8.378 934 2.228 9.478 4.015 8.114 2.390 2.202 5.718

• Refusal to participate in the interview: 796 5.926 3.564 2.229 5.106 146 513 6.119 582 4.100 863 865 313• No MM-interview possible within fieldwork period: 157 2.153 2.680 1.927 1.235 70 187 2.157 1.316 124 79 77 3.874• MM-interview incomplete: 28 16 201 37 547 22 232 200 6 0 4 43 13• No adequate respondent at the local establishment: 209 0 593 7 0 38 243 0 11 2.290 440 125 8• Other non-responses (please specify): 1.688 50 0 0 0 149 552 0 1.084 100 4 91 0

Complete MM interviews: 1.023 1.500 1.509 1.500 1.502 509 501 1.002 1.016 1.500 1.000 1.001 1.510

MM interviews as % of gross sample without quality neutral non-responses: 26% 16% 18% 26% 18% 54% 22% 11% 25% 18% 42% 45% 26%

Page 29: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

29

Table 4.3.2: Non-responses MM interviews screening countries

BE BG CZ EE IE EL CY LT HU MT PT RO SI SK HR MK TRGross sample I:Total No. of addresses used for the screening: 5.210 2.770 8.117 1.736 4.692 2.540 3.480 3.092 10.475 1.528 10.035 2.445 3.256 3.027 5.919 1.199 37.857

Quality neutral non-responses screening phase:• Establishment does not exist: 17 23 4.349 21 12 116 167 129 68 13 50 2 578 0 83 54 321• Line dead; fax/modem: 586 389 1 27 98 206 77 120 2.182 0 225 32 40 90 0 36 7.443• Telephone not answered after a minimum of 10 futile contacts : 335 157 3 312 0 501 54 111 0 175 1 1.076 506 16 65 47 213• Line always busy, answering machine 192 41 6 92 3.172 13 163 108 18 156 1.022 70 171 109 11 38 10.562• Private households (out of scope): 75 374 155 32 9 20 165 2 0 26 28 1 0 18 3 16 2.421• Less than 10 employees in MM050a (out of scope): 492 1 6 4 4 28 6 27 39 1 18 7 0 1 0 58 34• No establ. with 10+ employees acc. to MM051a (out of scope): 92 1 9 19 9 0 8 16 216 0 243 9 5 3 0 1 41

Total quality neutral non-responses screening: 1.789 986 4.529 507 3.304 884 640 513 2.523 371 1.587 1.197 1.300 237 162 250 21.035

Quality neutral non-responses main interviews:• Establishment does not exist; line dead; fax; modem: 49 0 0 0 0 3 0 0 0 0 0 0 133 0 0 29 0• Telephone not answered after a minimum of 10 futile contacts: 11 0 0 0 0 0 70 8 0 0 0 0 16 0 0 26 0• Line always busy or answering machine: 5 0 0 0 0 0 45 6 0 0 0 39 0 0 22 0• Private household (out of scope): 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0• Less than 10 employees in MM102a/b (out of scope): 82 8 75 177 0 97 196 163 1.072 51 731 54 10 60 2 16 1.593

• Valid, screened addresses not used for main interview (quota completed, not needed any more etc.): 99 383 233 0 0 0 306 0 691 303 3.224 0 88 68 2.746 17 10.738

Total quality neutral non-responses main interview: 246 391 308 177 0 100 617 177 1.763 354 3.955 54 286 128 2.748 117 12.331

Gross sample II:

Total No. of valid addresses used for main interviews 3.175 1.393 3.280 1.052 1.388 1.556 2.223 2.402 6.189 803 4.493 1.194 1.670 2.662 3.009 832 4.491

Refusals screening phase:• General refusal to answer screening interview: 1.008 441 2.004 166 662 470 771 433 5.097 124 203 229 214 2.058 135 95 2.812• No answer in MM050a, MM051a or MM052: 120 32 30 0 0 2 5 3 24 1 29 7 2 12 2 3 54• Refusal to give information on target person in MM053a/b/c: 36 3 5 0 0 1 2 1 10 1 12 0 0 4 0 0 11

Refusals main phase:• No answer in MM102a/b: 0 339 4 0 110 2 6 8 13 0 50 2 184 14 200 0 9• Refusal of MM-interview: 17 0 31 13 0 52 568 55 0 0 1.028 72 599 0 17 24 10• No MM-interview possible within fieldwork period: 247 46 0 101 0 20 261 18 0 65 2.159 377 102 0 179 131 0• MM-interview incomplete: 179 13 192 0 1 3 70 33 0 208 0 0 24 54 24 20 86• No adequate respondent at the chosen local establ.: 427 0 0 95 9 1 0 1 0 0 0 0 9 0 0 39 0• Other non-responses (please specify): 125 17 0 177 103 0 35 1.290 0 55 0 7 0 0 1.952 0 9

Complete MM interviews: 1.016 502 1.014 500 503 1.005 505 560 1.045 349 1.012 500 536 520 500 520 1.500

Response rate in %: 32% 36% 31% 48% 36% 65% 23% 23% 17% 43% 23% 42% 32% 20% 17% 63% 33%

Page 30: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

30

Fieldwork institutes were asked to assess whether the response rates reached for the ECS 2009 were higher, lower or about the same as those achieved for other recent CATI surveys among establishments or companies. In case of the rates being higher or lower than usual, they were also asked to name the (assumed) reason for this. The majority of countries rated the response rates obtained for the ECS 2009 as “in line with other surveys of this type and length”. The institutes in Austria, Belgium, the Czech Republic, Greece, Ireland, and Latvia however reported that response rates for the ECS 2009 were notably higher than those usually achieved. Some institutes attributed this to the recommendation letters, others to the topic of the study or to specific efforts in the fieldwork management. In two countries – the Netherlands and Turkey – response rates for the ECS 2009 were assessed to be lower than usual. For the Netherlands, this was attributed to the economic crisis and to the topic of the survey which was obviously perceived as little appealing. For Turkey, reasons were more fundamental: Many of the respondents were generally reluctant to give information on their personnel policies because of distrust regarding the way how their data would be used. Another factor is certainly an ambivalent attitude of many respondents in Turkey with regard to the European Union. Experience from fieldwork and further analyses of the response rates made in the preceding study ESWT shows that some differences in the responsiveness exist between the sectors of activity or between different size-classes. But there is no clear pattern discernable in this: The willingness to participate in the interview for example was clearly above average in the Public Administration in Italy, Germany and Sweden and interviewing in these sectors therefore had to be stopped after the interim sample checks8. In Spain, in turn, the willingness of the Public Administration to participate in the interview was lower than for other sectors. The quality of the net sample is not notably affected by eventual structural differences in the response rates. In all cells of the sampling matrix a sufficiently high number of interviews are available. Major differences in response rates between the cells of the sampling matrix were corrected in the course of the weighting process. Weighting in any case was necessary in order to redress the deliberately disproportionate structure of the net sample.

4.4 Non-response in the employee representative interviews

Identification of ER respondents The gross-sample for the ER interviews was defined by the outcomes of the MM questionnaire: Eligible as respondents for the ER interview were all those establishments where according to MM650ff. a relevant formal body of employee representation exists and where the management respondent did not clearly express that he/she does not want us to contact the employee representative for the purpose of an ER interview. 8 In Italy and Germany, the Public Administration sector was already notably over-represented at the

moment the interim sample checks were done (i.e. after 60% of fieldwork was completed). This over-representation in Italy and Germany was adjusted by the weighting, the relative weight of the Public Administration interviews within the weighting cell “Public Services” was reduced.

Page 31: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

31

It has to be noted that the management respondents were not explicitly asked for their permission on conducting an ER interview, but that they were asked in the open question MM800 to provide the contact details for the spokesperson of the employee representative body identified earlier in the interview. At this occasion, interviewers were advised to code the answers of the management respondent according to three categories:

(1) Respondent provides address of the ER (2) Respondent does not want to/cannot provide address of the ER, but does not

show any general opposition related to interviewing the ER (3) Respondent expresses a clear refusal for us to contact the ER

Success rate of ER interviews by availability of address details In all cases where code (1) was entered by the interviewer, efforts were made to get an interview with the named person by calling the indicated number. In total, in 8.568 establishments (65% of all establishments with an identified ER body) the management respondent provided the contact details of the ER. From these 8.568 ER addresses, 5.928 resulted (success rate of 69%). Table 4.4.1: Success rate for ER interviews by type of approach

Answer options in MM800/MM804/MM805 Answers (n) Answers (in %)ER interviews resulting from this base

Success rate in %

MM provides ER contact details 8.603 65% 5.963 69%

MM gives OK, but does not provide contact details 1.625 12% 552 34%

MM allows ER interview only in ER's leisure time (and 106 1% 54 51%MM refuses and maintains refusal in MM804 2.845 22% 0 0%

13.179 100% 6.569 50% Code (2) was entered in 1.625 cases (12% of all establishments with an identified ER body). These managers did not express any general objection against us interviewing the ER in the establishment, but could not or did not want to provide information on the contact details of the ER9. In these cases attempts to contact the identified ER were made via the switchboard of the firm (respectively the phone number used for the initial contact with the establishment). This approach was considerably less successful, with only 552 ER interviews resulting from it (success rate of 34%). Code (3) was entered by the interviewer only if the management respondent expressed a clear refusal to allow for an ER interview. Nevertheless, in these cases the MM respondent was asked in a subsequent question (MM804) whether it would be ok for us to try and make an interview appointment with the ER in his/her leisure time. This question had been introduced as a measure to further reduce the non-response for the ER interviews. The success of this measure of refusal conversion was however limited: Only 240 of the 2.951 refusals in MM800 could be converted by this way and among these 240 cases, 134 MM respondents refused to enable the ER contact in the subsequent question MM605. So the refusal conversion attempt resulted in not more than 106 additional allowances (on the

9 In several cases, managers did not want to provide the contact details for reasons of data protection.

Page 32: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

32

managers’ part) which in the end led to 54 additional ER interviews (i.e. a success rate of 51% of all posterior allowances, but only a 2% success rate from all MM800 refusals). Success of ER interviews by country On the average of all countries, ER interviews could be carried out in every second establishment (50%, see table 4.4.1 above) that has an eligible representation. Table 4.4.2 below shows however that the rate of ER interviews that finally resulted varies considerably between the countries. It is highest in Finland (ER interviews in 77% of all establishments with an ER), the Czech Republic (70%) and Latvia (70%) and it is lowest in Turkey (30%) and Cyprus (29%). Reasons for these differences in the final response rates reached for the ER are manifold. Differences in the culture of social dialogue within the countries are certainly one of the explaining factors, but also aspects such as differences in the availability of employee representatives (access of the ER to a telephone) might play a role.

Page 33: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

33

BE BG CZ DK DE EE IE EL ES FR IT CY LV LT LU

Number of establishments interviewed interviews) 1.016 502 1.014 1.023 1.500 500 503 1.005 1.509 1.500 1.502 505 509 560 501

• Establishments without ER (i.e. MM800 not asked) 400 283 668 268 612 375 287 871 561 438 559 297 300 344 203Establishments with employee presentation 616 219 346 755 888 125 216 134 948 1.062 943 208 209 216 298% Establishments with employee representation (unweighted!): 61% 44% 34% 74% 59% 25% 43% 13% 63% 71% 63% 41% 41% 39% 59%

• No OK for ER-interview by MM-respondent (if MM804 = 2 or 3 or MM805 = 3) 141 50 15 139 145 30 36 23 328 225 457 31 48 29 112

Establishment with OK of management for ER interview: 475 169 331 616 743 95 180 111 620 837 486 177 161 187 186

• Line dead; fax/modem: 17 0 1 1 5 0 0 0 4 69 0 3 0 1 0• Telephone not answered after a minimum of 10 futile contacts: 6 0 10 114 16 14 1 0 12 135 0 11 2 14 0• Line always busy or answering machine: 12 0 7 0 3 1 15 0 0 1 0 7 0 6 0• Refusal by ER: 40 16 37 25 120 8 29 22 155 18 166 93 7 20 1• No ER-interview possible within fieldwork period: 28 0 0 4 16 0 0 1 35 147 0 0 0 12 3• ER-interview incomplete: 3 16 1 0 5 0 0 2 15 16 0 3 0 3 0• Other non-responses: 82 9 33 78 20 0 42 10 24 10 0 0 5 3 18

ER interviews: 287 128 242 394 558 72 93 76 375 441 320 60 147 128 164ER interviews as % of interviewed establ: (unweighted!): 28% 25% 24% 39% 37% 14% 18% 8% 25% 29% 21% 12% 29% 23% 33%ER interviews as % of establ. with ER: 47% 58% 70% 52% 63% 58% 43% 57% 40% 42% 34% 29% 70% 59% 55%

HU MT NL AT PL PT RO SI SK FI SE UK HR MK TR

Number of establishments interviewed (MM interviews) 1.045 349 1.002 1.016 1.500 1.012 500 536 520 1.000 1.001 1.510 500 520 1.500

• Establishments without ER (i.e. MM800 not asked) 649 281 354 587 707 885 161 241 228 265 257 1.038 241 333 1288Establishments with employee representation 396 68 648 429 793 127 339 295 292 735 744 472 259 187 212% Establishments with employee representation (unweighted!): 38% 19% 65% 42% 53% 13% 68% 55% 56% 74% 74% 31% 52% 36% 14%

• No OK for ER-interview by MM-respondent (if MM804 = 2 or 3 or MM805 = 3) 107 24 102 92 138 25 108 80 112 67 43 72 19 15 32

Establishment with OK of management for ER interview: 289 44 546 337 655 102 231 215 180 668 701 400 240 172 180

• Line dead; fax/modem: 28 0 0 68 42 0 0 0 0 10 0 8 3 7 2• Telephone not answered after a minimum of 10 futile contacts: 5 4 5 21 34 0 0 2 0 0 10 3 5 7 9• Line always busy or answering machine: 1 0 4 6 3 0 0 0 5 41 0 13 5 2 23• Refusal by ER: 72 9 237 24 113 8 94 4 46 16 33 9 41 3 0• No ER-interview possible within fieldwork period: 0 7 10 8 11 0 0 2 0 35 22 201 20 7 0• ER-interview incomplete: 0 0 41 5 0 55 0 0 25 1 2 0 4 0 12• Other non-responses: 0 0 0 0 85 0 0 54 4 0 93 0 0 17 70

ER interviews: 183 24 249 205 367 39 137 153 100 565 541 166 162 129 64ER interviews as % of interviewed establ: (unweighted!): 18% 7% 25% 20% 24% 4% 27% 29% 19% 57% 54% 11% 32% 25% 4%ER interviews as % of establ. with ER: 46% 35% 38% 48% 46% 31% 40% 52% 34% 77% 73% 35% 63% 69% 30%

Table 4.4.2: Non responses ER Interviews

Page 34: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

34

4.5 Non-response analysis of the ER interviews: A multi-variate approach

In order to control for any non-response bias in the ER interviews, two multi-variate regression models were calculated which allow to analyse criteria with a potential influence on the response behaviour in isolation from others. In a first model, we analysed the refusals of the interviewed managers to allow for an ER interview in the establishment (all in all, 22% of the management respondents had refused this). A second model investigates the factors leading to structural differences between the gross sample available for the ER interviews (all establishments with an identified employee representation) and the net sample finally resulting from this. Potential differences between gross and net ER sample can be caused by a variety of non-response factors:

− The refusals of managers to allow for an ER interview (in these cases no attempt for an ER interview was made).

− Refusals on part of the contacted employee representatives to take part in the interview.

− Neutral non-responses for the ER interview, such as the unavailability of the employee representative during the fieldwork period, wrong telephone numbers etc.

− Aspects related to the organisation of fieldwork (Though fieldwork organisation and the training of interviewers were following common centrally provided standards, differences between countries might have occurred in some details, e.g. in the precise timing of the ER interviews).

The base for both calculated logit regression models were all establishments where an eligible employee representation was identified in the management interview (n = 13.179 cases). Refusals of MM respondents to allow for an ER interview This first model shows that the factor “country” has by far the strongest influence on the decision of the interviewed manager on whether or not to agree in an ER interview within the establishment. The probability that the management would refuse to allow this interview to take place proved to be by far the highest in Italy. Highly significant and clearly above the average probabilities for an MM refusal were also found for Slovakia, Malta, Slovenia, Luxembourg, Hungary, Romania, Latvia and Spain. The lowest probabilities for such refusals on part of the managers were found in the Czech Republic, Sweden, FYROM and Croatia. Table 4.5.1: Multi-variate analysis of management refusals to agree to an ER interview

Page 35: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

35

Management refusals to allow for an ER interview (dependent variable)(according to MM800/MM804/MM805)

Coefficient Exp (B)

ER improves performance (reference: agree)neither nor 0,16 * 1,17 *disagree 0,27 *** 1,31 ***no opinion 0,31 1,36ER delays decisions (reference: agree)neither nor 0,11 1,12disagree -0,10 0,91no opinion 0,06 1,07Direct link to employees better (reference: agree)neither nor 0,01 1,01disagree -0,12 * 0,89 *no opinion 0,01 1,00ER enhances staff commitment (reference agree)neither nor 0,11 1,12disagree 0,12 1,13no opinion 0,05 1,11Work climate (reference: good)strained -0,10 0,91no answer 0,61 * 1,84 *Economic situation (reference: very good)quite good -0,22 ** 0,80 **neither nor -0,26 ** 0,77 **quite bad -0,24 * 0,79 *very bad -0,31 0,73no answer 0,21 1,24Changes (ref.: no changes acc. to MM150_1 to 5) -0,23 *** 0,80 ***More than one ER (ref. one ER body) -0,36 *** 0,70 ***Wage agreement (ref. no)yes -0,19 0,82no answer 0,69 *** 2,00 ***Establishment type (reference: single independent)headquarters -0,07 0,94subsidiary site -0,01 0,99no answer 0,69 *** 2,00 ***Public sector company (reference: no)yes -0,25 ** 0,78 **no answer -0,20 0,82Size (reference: 10-19 empl.)20-49 -0,13 0,8850-249 -0,43 *** 0,65 ***250-499 -0,54 *** 0,58 ***500+ -0,64 *** 0,53 ***Country (reference: Germany)Belgium 0,32 * 1,37 *Denmark 0,01 1,01Greece -0,73 0,93Spain 0,55 *** 1,73 ***Finland -0,74 *** 0,48 ***France 0,19 1,21Ireland -0,01 0,99Italy 1,90 *** 6,39 ***Luxembourg 0,71 *** 2,04 ***Netherlands -0,22 0,80Austria 0,10 1,11Portugal 0,15 1,17Sweden -1,41 *** 0,24 ***United Kingdom -0,19 0,83Bulgaria 0,19 1,21Cyprus -0,64 ** 0,53 **Czech Republic -1,62 *** 0,20 ***Estonia 0,13 1,14Hungary 0,64 *** 1,89 ***Latvia 0,54 ** 1,72 **Lithuania -0,27 0,76Malta 0,82 ** 2,26 **Poland 0,02 0,98Romania 0,59 *** 1,80 ***Slovakia 0,99 *** 2,68 ***Slovenia 0,68 *** 1,98 ***Turkey -0,10 0,90Croatia -1,03 *** 0,36 ***FYROM -1,37 *** 0,25 ***Sector (reference: Manufacturing)Mining/ Quarrying -0,10 0,91Electricity/ Gas/ Steam/ Hot Water Supply -0,37 0,69Construction 0,27 ** 1,30 **Wholesale/ Retail Trade 0,02 1,02Hotels/ Restaurants 0,13 1,14Transport/ Storage/ Communication -0,06 0,94Financial Intermediation -0,19 0,83Real Estate/ Renting/ Business Activities 0,19 1,20Public Administration/ Defence/ Social Security -0,83 *** 0,44 ***Education -0,36 ** 0,70 **Health/ Social Work -0,52 *** 0,60 ***Community/ Social/ Personal Service Activities -0,14 0,87

Nagelkerke´s R² = 0,171Base: Establishments with an existing employee representation (at the level of the surveyed local unit) n = 13.179 casesSignificance: * p<0.05; ** p<0.01; *** p<0.001Source: ECS 2009, unweighted

Page 36: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

36

Apart from the factor “country”, the following further structural variables have a significant influence on the refusal behaviour of the managers:

• In establishments with a very good economic situation, managers were slightly more likely to refuse an ER interview than in establishments with a less favourable economic situation (a “very bad” economic situation did however not have a significant effect).

• Refusals were less likely in establishments where any structural change had taken place in the past 3 years (such as a takeover, a relocation or the acquisition of another organisation) as compared to establishments where no such change had taken place.

• Managers of establishments in which more than one type of employee representation exists (e.g. both a trade union and a works council representation) were less likely to refuse than managers where there is only one sole type of representation.

• Managers from public sector organisations were less likely to refuse than managers of private establishments.

• The likeliness of a refusal was found to be lowest in establishments of the sectors “Public administration”, “Education”, “Health and social services” (which are all sectors with a high proportion of public units). The “Construction” sector, on the other hand, turned out to have a significantly higher probability of refusals on part of the management.

• Managers of middle-sized and large establishments (from 50 employees onwards) were significantly less likely to deny the allowance for an ER interview than those of small establishments with up to 49 employees. Between the middle-sized and the large organisations, however, there is little difference in the response behaviour.

Surprisingly, the general work climate in the establishment had no significant influence on the refusal behaviour of the managers. Also, the type of the organisation (single-site, headquarters of a multi-site or subsidiary) did not have any significant influence. A further potential factor of influence on the refusal behaviour is the quality of the social dialogue at the level of the establishment. In order to be able to measure this type of influence, managers had been asked to describe their view on shopfloor employee representation against the background of experiences with the representation in their own establishment. To this end, they were asked to express their approval or disapproval to the following four statements capturing different aspects of the quality of social dialogue on a Likert scale (question MM702_1 to MM702_4):

(1) “The employee representation helps us in a constructive manner to find ways to improve workplace performance.”

(2) “The involvement of the employee representation often leads to considerable delays in important management decisions.”

(3) “We would prefer to consult directly with our employees.” (4) “Consulting the employee representation in important changes leads to more

commitment of the staff in the implementation of changes.” One would normally expect managers of establishments with a poor social dialogue to be much less willing to allow for an ER interview than managers of establishments with a good social dialogue. The probability of management refusals was indeed found to be significantly

Page 37: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

37

higher among managers disagreeing to the first statement, though with an odds ratio of 1.31 the extent of this influence is not huge. It means that managers disagreeing to the view that the ER helps in a constructive manner to improve the workplace performance are about 30% more likely to refuse the contact with the ER than those who agree with this statement. A certain correlation was also found for statement (3) indicating a preference for a direct contact with the employees, but this influence is not very significant and its extent is quite weak. Composition of the net sample of ER interviews More important than the analysis of the refusals on part of the managers is the second model analysing the factors leading to differences between the composition of the gross and the net ER sample since in this perspective different types of non-response behaviour are taken into account which in the end might reinforce or neutralize each other. Table 4.5.2: Multi-variate analysis of factors influencing the composition of the net ER sample

Page 38: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

38

Existence of an employee representation (ER) interview (dependent variable)

Coefficient Exp (B)

ER improves performance (reference: agrees)neither nor -0,08 0,99disagrees -0,10 0,90no opinion 0,20 1,22ER delays decisions (reference: agrees)neither nor 0,06 0,95disagrees 0,11 * 1,11 *no opinion 0,06 1,05Direct link to employees better (reference: agrees)neihter nor 0,01 1,10disagree 0,18 *** 1,20 ***no opinion 0,21 1,24ER enhances staff commitment (reference: agrees)neither nor -0,21 *** 0,82 ***disagree -0,10 0,91no opinion -0,39 ** 0,68 **Work climate (reference: good)strained 0,15 ** 1,16 **no answer -0,55 0,58Economic situation (reference: very good)quite good 0,10 1,10neither nor 0,09 1,10quite bad 0,13 1,14very bad 0,26 1,29no answer -0,30 * 0,74 *Changes (reference: no changes acc. to mm150_1-5) 0,07 1,07Number of ER bodies (reference: one ER body)Several ER bodies 0,30 *** 1,35 ***Wage agreement (reference: no collective wage agreement)yes -0,23 0,80no answer -0,58 *** 0,56 ***Establishment type (reference: single independent company)headquarters of a multi-site company 0,18 ** 1,20 **subsidiary site 0,33 1,03no answer -0,23 0,79Public sector company (reference: no)yes 0,26 *** 1,29 ***no answer 0,22 1,24Size (reference: 10-19 empl.)20-49 0,18 ** 1,20 **50-249 0,38 *** 1,46 ***250-499 0,44 *** 1,55 ***500+ 0,39 *** 1,47 ***Country (reference: Germany)Belgium -0,55 *** 0,58 ***Denmark -0,30 ** 0,75 **Greece -0,13 0,88Spain -0,59 *** 0,55 ***Finland 0,73 *** 2,07 ***France -0,83 *** 0,44 ***Ireland -0,78 *** 0,46 ***Italy -1,47 *** 0,23 ***Luxembourg 0,06 1,06Netherlands -0,84 *** 0,43 ***Austria -0,39 ** 0,68 **Portugal -1,33 *** 0,26 ***Sweden 0,69 *** 2,00 ***United Kingdom -1,10 *** 0,33 ***Bulgaria 0,60 1,06Cyprus -1,03 *** 0,36 ***Czech Republic 0,47 ** 1,60 **Estonia 0,07 1,08Hungary -0,67 *** 0,51 ***Latvia 0,33 1,40Lithuania -0,07 0,93Malta -0,90 ** 0,41 **Poland -0,62 *** 0,54 ***Romania -0,63 *** 0,53 ***Slovakia -1,02 *** 0,36 ***Slovenia -0,43 ** 0,65 **Turkey -1,34 *** 0,26 ***Croatia 0,06 1,07FYROM 0,82 *** 2,02 ***Sector (reference: Manufacturing)Mining/ Quarrying 0,07 1,07Electricity/ Gas/ Steam/ Hot Water Supply 0,37 * 1,45 *Construction -0,35 *** 0,70 ***Wholesale/ Retail Trade 0,30 1,03Hotels/ Restaurants -0,16 0,85Transport/ Storage/ Communication -0,16 0,86Financial Intermediation 0,11 1,11Real Estate/ Renting/ Business Activities -0,14 0,87Public Administration/ Defence/ Social Security 0,65 *** 1,91 ***Education 0,16 1,18Health/ Social Work 0,19 * 1,20 *Community/ Social/ Personal Service Activities 0,11 1,12

Nagelkerke´s R² = 0,149Base: Establishments with an existing employee representation (at the level of the surveyed local unit) n = 13.179 casesSignificance: * p<0.05; ** p<0.01; *** p<0.001Source: ECS 2009, unweighted

Page 39: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

39

In this model, again the factor “country” is the most important one. FYROM, Finland and Sweden stand out with the highest probability of having the existing employee representatives interviewed, with a beta coefficient of around 2, meaning that for establishments with an existing employee representation in these countries the chance of finally getting an ER interview is about twice as high as in the reference country (which in the model is Germany, a country which has neither a particularly good nor a particularly bad success rate for the ER interviews). The lowest probability for an ER interview, in turn, was found for establishments in Italy, Portugal, Turkey, the United Kingdom and Cyprus, with beta coefficients of less than 0,4. Apart from the country factor, behind which there are differences e.g. in the national culture of social dialogue or the general openness towards company surveys, the following factors were found to have an influence on the existence of an ER interview:

• Establishments with more than one type of ER body were more likely to provide an interview than those with only one type of body10.

• Headquarters of multi-site companies were more likely to provide an ER interview than single independent companies or subsidiary sites.

• Public sector organisations were more likely to provide an ER interview than private sector establishments.

• Middle-sized and large establishments were more likely to provide an ER interview than the smallest ones; this is probably mostly because representatives in larger establishments have more time available for their representative duties and can better be reached by phone.

• Organisations of the “Public administration” were more likely to provide an ER interview than establishments of other sectors. In the “Construction” sector, the probability to get an ER interview was lowest – probably to a large degree owed to difficulties in contacting these representatives, many of whom are working on construction sites where they are difficult to reach by phone.

• Establishments where the current work climate is strained were slightly more likely to give an ER interview than those with a good work climate.

Among the variables describing the quality of the social dialogue, two were found to have some influence on the existence of an ER interview: Establishments where the manager disagreed to the statement “The involvement of the employee representation often leads to considerable delays in important management decisions.” were slightly more likely to provide an ER interview than those agreeing to this statement, with a beta coefficient of 1,11. Also, establishments where the manager disagreed to the statement “We would prefer to consult directly with our employees.” were found to be more likely to provide an ER interview. The disagreement is in both cases an indication for a rather positive relationship between management and employee representation. For the most meaningful of these indicators in

10 This is not due to technical reasons: There were no additional attempts made to interview the

second type of representation in establishments where the ER type of first choice refused or was not available. The existence of several types of representation can however be supposed to be related to the size of the establishment – larger establishments will tend to have more often several types of representation than smaller ones. The possibility of having more than one type of representation also is given only in part of the countries.

Page 40: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

40

terms of the quality of social dialogue – statement (1) describing the constructive contribution of the ER to the workplace performance – there was however no significant influence discernable. Summarizing, it can be said that the factor “country” and some structural variables such as “size” or the ownership of the organisation (public vs. private) have an influence on the composition of the net sample of ER interviews. The influence of the quality of the social dialogue at the establishment level, in turn, is comparably low. The employee representative interviews are not systematically biased towards establishments with a particularly good social dialogue and they can thus be considered as representative in this regard.

4.6 Fieldwork monitoring

In order to follow the progress of the fieldwork, an online reporting tool was developed especially for this survey. Using this tool, the number of interviews and the structure of the sample (sampling matrix) were monitored on a weekly basis. The tool also allowed for a regular monitoring of the number of establishments eligible for an additional interview with the employee representation. This information was also provided to the European Foundation team on a weekly updated basis. The supervision of the telephone interviewers in the CATI studios was the task of the local supervisors. As a routine measure, the interviewers in the CATI studios of our network of local institutes are regularly supervised as regards the contact behaviour and the quality of interviewing in general.

4.7 Specific fieldwork observations

Fieldwork as such ran smoothly, with little noteworthy occurrences or problems. Specific issues worth mentioning are: − In some countries, the quota set for the largest size-classes were quite ambitious and

much higher than in most other business surveys. In some countries, quota for larger size-classes had to be lowered in the course of fieldwork because the number of addresses available for this survey was exhausted.

− The regular monitoring of the results from the ER interviewing had revealed very low rates

of employee representation for Portugal (rates considerably below the results measured in the ESWT 2004). Interviewing in Portugal was therefore interrupted in order to search for the causes of this phenomenon. It turned out that an inaccuracy in the formulation of an item in the question identifying the existing bodies of employee representation (MM650_01) might be a cause for this. Therefore, this question was corrected and all

Page 41: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

41

establishments that had been called up to that moment were recalled in order to ask them question MM650_01 in the new, amended version11.

− The final data-check revealed a small filtering error that had occurred in Belgium. This

error concerned questions MM654 and MM657 which had not been asked in some constellation of multi-site establishments. The mistake occurred in the adaptation of the screening procedure at the beginning of the interview (MM050ff.) to the local interviewing standards of the Belgian institute. The error was corrected by way of back-calls made in the concerned establishments in the period from 12 May 2009 to 09 June 2009.

− In Turkey, fieldwork faced several specific challenges related to the extremely difficult

sampling situation in the country:

• In the interim sample checks it turned out that the address list from the Chambers of Commerce which was originally foreseen for the survey proved to be strongly biased on manufacturing industries and some other NACE sectors. Therefore additional address lists had to be used, mostly also stemming from the Chambers of Commerce.

• Addresses in the additional address sources did mostly not contain information on the

sector of the establishment. Therefore a specific screening questionnaire was designed in which the sector of activity is asked for on the NACE Rev. 1.1 one-digit level12. This additional sector screener was asked before the “normal” ECS interview started. By this way it was ensured that reliable sector information is available for all interviews from Turkey.

• Due to the need to use several address registers of which none was really fully

representative of the economy, the Turkish fieldwork institute was provided with minimum quota of interviews to be realized for each NACE 1-digit sector of activity. By this – very costly and time consuming - way it could be ensured that the weaknesses of the address lists are not reflected in the structures of the net sample.

• Some of the available lists showed a strong regional bias on establishment from

Istanbul. This bias was corrected by limiting the number of addresses from Istanbul in the gross sample drawn for the survey and by closely monitoring the regional structures of the interviews during the fieldwork period.

The coordination team of TNS Infratest closely followed up the process in Turkey during the whole fieldwork period in order to guarantee a good quality sample in spite of the adverse sampling conditions.

11 The original version of MM650_01 was asking for „Uma commisao sindical” while the new, amended

version asks for “Um delegado sindical ou uma commisao sindical”. 12 This additionally required sector screening for Turkey is the main reason for the large number of

completed, but unused screeners due to “quota completed” in Turkey (see non-response table 4.3.2). If the sector is not known a priori from the address source, large efforts have to be made to come to an evenly distributed sample.

Page 42: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

42

5. Weighting

Weighting of the data is necessary for descriptive analyses because of two reasons. Firstly it was decided to build disproportional samples (see chapter 3.1) and the disproportional structure of the samples has to be redressed for analysis. Secondly experience shows that the willingness to participate in a survey depends – among others – on size and sector of the establishment. This can induce disproportional non-responses, which are to be corrected in the weighting procedure. For the weighting, the 10-cell matrix used for the sampling was broadened to a 15-cell matrix by further differentiating within the “Services” section between sectors of activity mainly made up by private organisations (“Private Services”) and sectors made up mainly by public entities (“Public Services”). It should be noted that the sector shortly called “Public Services” here is not exclusively made up by public organisations. Especially in NACE Rev. 1.1 M & N /Rev 2 O & P (Education and Health and Social Work) also many private organisations are contained (e.g. private schools and private hospitals). Table 5.1: Weighting matrix13:

Sector: 1. Producing Industries 2. Private Services 3. Public Services

NACE Rev.1.1 C-FNACE Rev. 2 B-F

NACE Rev.1.1 G-K & ONACE Rev. 2 G-N & R-S

NACE Rev.1.1 L-NNACE Rev. 2 O-Q

10 - 19 employees

20 – 49 employees

50 – 249 employees

250 - 499 employees

500 + employees For weighting of the data two different weighting factors were calculated. The establishment proportional weighting factor (“est_wei”) redresses the net sample so that the distribution of establishments over the 10 cells of the stratification matrix used for sampling (see chapter 3) is reproduced. Analyses based on establishment proportional weighted data allow to make statements of the following type: “X % of the establishments surveyed have part-timers among their staff”. Since the majority of the establishments are small and medium sized units analyses of this type are strongly influenced by the situation of these smaller units.

13 For Belgium a slightly different matrix was used for the weighting since there statistics were

available only in size bands 50-199/200-499 instead of 50-249/250-499.

Page 43: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

43

The employee proportional weighting factor (“emp_wei”) redresses the net sample so that the distribution of employees over the 10 cells of the stratification matrix used for sampling is reproduced. Analyses based on employee proportional weighted data allow statements of the following type: “Y % of all employees covered by this survey14 work in establishments which have part-timers among their staff”.15 Since many employees work in large establishments, analyses of this type are much more influenced by the situation in larger units. The weights are calculated as a projection on the universe, i.e. they take into account not only the structure but also the absolute size of the universe in each of the countries. The actual ratio between the countries is thus taken into account. The same weighting factors can be used for both, national and cross-national analyses. In countries with screening procedure the number of establishments (with 10 or more employees) per company was taken into consideration when calculating the weights for the interviewed establishments in order to compensate the lower statistical chances of the establishments of multi-site companies to be selected for interview. The basis for the weighting was information on the universe that had been collected by the national fieldwork institutes. The source of the information is in almost all countries the statistical information maintained by the national statistical office. For several countries however, reliable statistical data on the universe of establishments was not available from the statistical offices. This holds especially for the NMS where statistics were mostly available only for the unit of the company, but not for establishments. In these cases, best estimates on the distribution of establishments had to be made by TNS Infratest Sozialforschung. These estimates were mainly based on data from the Labour Force Survey. Likewise, for the universe of employees working in establishments of the relevant sizes and sectors reliable statistics do not exist in part of the countries. In these countries, too, estimates were made on base of the Labour Force Survey.

14 The universe was restricted to establishments with 10 or more employees. 15 Please note that this does not mean that Y% of all employees are part-timers.

Page 44: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

44

6. Level of confidence – Bootstrap variance test16

6.1 Introduction

The different random samples of this large European survey are characterized by the strong disproportional design and by the different sample sizes. These designs have to be corrected by weighting according to the respective population structure of each country. Due to the variance of the weighting factors, this process brings additional variance into the sample. The level of confidence within this large and complex survey is of main interest. To evaluate the influence of different sample sizes as well as of the strong disproportional sample with regard to the levels of confidence, we have chosen a resampling design, namely the bootstrap estimate of standard error (Efron und Tibshirani, 1993)17. The bootstrap estimate requires no theoretical calculations based on sampling theory since it is calculated directly on base of the obtained survey results. A bootstrap evaluation of all samples and all answers within this survey would be too large. Therefore we selected a set of 10 questions of different types from the survey as typical variables and 6 countries with different sample size as typical for the survey too. Selected Samples for Bootstrap Analysis: 500 cases: Latvia and Ireland 1.000 cases: Sweden and Greece 1.500 cases: Poland and Germany Selected Questions for Bootstrap Analysis: MM201: Existence of shift-workers MM250: Proportion of employees working part-time MM300: Possibility to adapt working hours MM302: Proportion of employees entitled to flexible hours system MM351: Mode of compensation for overtime hours MM403: Further contracts after end of fixed-term employment MM459_1 to MM459_6: Reasons for the application of profit sharing schemes MM558: Teamwork MM561: Check of need for further training MM500: economic situation of the establishment

16 Chapter written by Dr. Martin Pfister, TNS Infratest Munich, statistical department. 17 Efron, B. and Tibshirani, R., An introduction to the bootstrap, Chapman & Hall, 1993

Page 45: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

45

6.2 Bootstrap Theory

Given are the sample x1, x2, ..., xn with the unknown distribution F and the estimate Θ = t (F). F describes the empirical distribution of the estimate in our sample.

The bootstrap sample is defined as a random sample of the size n drawn from F . The

bootstrap data **

2*1 ,...,, nxxx are a random sample of the size n, drawn with replacement of the

population of n objects (x1, x2, ..., xn). The standard error according to the bootstrap algorithm has to be calculated in the following way:

1. Selection of B independent bootstrap samples **

2*1 ,...,, nxxx B, each consisting of n values

drawn with replacement from x1, x2, ..., xn . The estimation of confidence level requires a value of B of 1.000 (Efron und Tibshirani, 1993).

2. Analysis and calculation of estimates of the bootstrap samples including complete

weighting scheme:

)()(ˆ ** bxtb =Θ Bb ,...,2,1=

3. Estimation of the standard error )ˆ(ΘFse using the variance within the B replications:

[ ]2/1

1

2** )(ˆ)(ˆ1ˆ⎭⎬⎫

⎩⎨⎧

⋅Θ−Θ= ∑=

B

bB b

Bes

, with BbB

b/)(ˆ)(ˆ

1** ∑ =

Θ=⋅Θ

4. The calculation of the confidence level for Θ according to the bootstrap algorithm is based

on the assumption that Θ is approximately normally distributed. According to that, the confidence interval within the usually 90% probability of evidence for Θ is calculated by

es645.1ˆ ⋅±Θ .

Page 46: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

46

6.3 Results

A first analysis has to look at the results of the single questions. Question MM201 was about the existence of shift-workers within the establishment. The interesting answer to this question was “yes”. The following Figure shows the result of the 1000 bootstrap simulations for the six selected country samples. The histograms give the projected result of answer “yes” of question MM201: the projected number of establishments where shift workers exist. The figures show almost perfect normal distributions, therefore the calculation of confidence levels is allowed, using simple equations for variance and standard error.

The following table summarizes the calculation of confidence level for the question MM201. The table gives the value of universe for each country (Universe). Furthermore the measured and projected value of establishments with existence of shift-workers (Mean Value “yes”) as well as the standard error of this value determined according to the distributions shown in the figure above. Out of these values the confidence level on the 90% probability level of evidence can be calculated. The last two columns transfer these values back into proportions of the universe.

150.000,000

160.000,000

170.000,000

180.000,000

190.000,000

200.000,000

MM201est

0

20

40

60

80

100

Häu

figke

it

Mean = 174.469,62797Std. Dev. = 8.849,728711N = 1.000

Establishments

Germany

17.500,000 20.000,000 22.500,000 25.000,000

MM201est

0

20

40

60

80

100

120

Häu

figke

it

Mean = 19.693,57171Std. Dev. = 1.346,256301N = 1.000

Establishments

Sweden

15.000,000 17.500,000 20.000,000

MM201est

0

20

40

60

80

100

Häu

figke

it

Mean = 17.041,90373Std. Dev. = 1.456,206607N = 1.000

Establishments

Greece

5.000,000

5.500,000

6.000,000

6.500,000

7.000,000

7.500,000

MM201est

0

20

40

60

80

100

Häu

figke

it

Mean = 6.207,29149Std. Dev. = 514,791042N = 1.000

Establishments

Ireland

6.000,000

6.500,000

7.000,000

7.500,000

8.000,000

8.500,000

9.000,000

MM201est

0

20

40

60

80

100

Häu

figke

it

Mean = 7.172,60172Std. Dev. = 437,017583N = 1.000

Establishments

Latvia

70.000,000 75.000,000 80.000,000 85.000,000

MM201est

0

20

40

60

80

100

120

Häu

figke

it

Mean = 76.424,57461Std. Dev. = 2.869,732833N = 1.000

Establishments

Poland

Page 47: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

47

Table 6.3.1: Summary of results of bootstrap analysis

Country Universe

(establish-

ments)

Mean Value

“yes”

MM201

Standard

Error

Confidence

Level 90%

probability

Mean Value

in %

(of universe)

Confidence

Level in %

(of universe)

Ireland 19.691 6207 515 +/- 847 31,5% +/- 4,3%

Latvia 17.400 7173 437 +/- 719 41,2% +/- 4,1%

Greece 56.697 17042 1456 +/- 2395 30,1% +/- 4,2%

Sweden 73.879 19694 1346 +/- 2215 26,7% +/- 3,0%

Germany 564.021 174470 8850 +/- 14558 30,9% +/- 2,6%

Poland 175.828 76425 2870 +/- 4721 43,5% +/- 2,7%

Obviously the confidence level is larger with smaller sample size. But sample size is not the only factor of influence. The proportion itself also influences the confidence level and one can suppose that the quality of sampling has an influence, too. Even the question itself can be answered in different quality in different countries. Because of these reasons we selected the 10 typical questions out of the survey for the bootstrap analysis. Doing so one gets an impression of the variability of the confidence levels depending on various factors like sample size, country, sample quality and question. The next step of the bootstrap analysis was therefore to look at all distributions of the selected 10 questions within the 6 countries considering both weighting schemes: establishment weight as well as employee weight. Almost all results show normal distributions, only some answers with very small measured percentage values cannot be considered for further analysis. The following figure shows all results transferred into proportions of the universe and confidence levels of these proportions. More than 400 single points of projected answers resulted and were put together in this figure. At first, the confidence intervals show a clear dependence of sample size, a phenomenon which is also known from theory. The samples with size of 1500 have the smallest confidence intervals. The intervals depend on the measured percentage value itself: values of 50% have the largest intervals.

Page 48: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

48

Range of Level of Confidence [+/-]

0,0%

1,0%

2,0%

3,0%

4,0%

5,0%

6,0%

7,0%

8,0%

9,0%

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

Percentage Value in the Sample

Con

fiden

ce in

terv

al [+

/-]

Sample Size 1500(Germany, Poland)

Sample Size 1000(Greece, Sweden)

Sample Size 500 (Latvia,Ireland)

Samples with the size of 1.500 interviews are characterized by confidence intervals that can reach +/- 1,5% for values of 7% to 10% and at maximum +/- 3% for values of 35% to 65%. Samples with the size of 1.000 interviews can reach intervals of +/- 2,5% for values of ca. 10% and at maximum +/- 4,5% for values of 35% to 65%. Samples with the smallest size of 500 interviews can reach intervals of +/- 4,5% for values of ca. 10% and at maximum +/- 6% for values of 35% to 65%. The confidence interval itself is characterized by some variability: the same percentage value can have a different confidence interval within the same sample, only depending on the question itself. The sample size influences mainly this variability: the confidence interval vs. percentage value is a much narrower band for the sample size of 1.500 than for the size of 500. For small sample sizes differences of 3% within the intervals for the same percentage value can be measured whereas in the large samples only 1% differences are observed.

In summary, it could be demonstrated that the levels of confidence in this survey determined by bootstrap algorithm show relationships like a dependence on sample size, a phenomenon known from theory. Furthermore, a certain variability of the confidence interval should be taken into consideration which is also depending on the sample size.

Page 49: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

49

7. Data structure

7.1 Data processing and cleaning

An automated checking program was built for this survey. This program checked the filters of the questionnaire, the coherency, the codification and the correct storage of the data. The local institutes sent their data files in ASCII format via a secured TNS Info web portal that was created specifically for this survey. These files were checked and automated reports were sent to the local institutes. They cleaned the data file if necessary and sent a corrected version of the data file. The process was repeated as many times as was necessary.

7.2 Harmonization of sector codifications

Background Since 1st January 2008 a new NACE version (NACE Rev. 2) has come into effect which significantly differs from the former version NACE Rev. 1.1 even on the 1-digit level. The national statistical offices of the EU member states are obliged to use the new NACE version since the beginning of 2008 for a series of official economic statistics. Business address registers and statistics on the distribution of establishments by sectors of activity have however often not yet been adapted to the new NACE classification. The NACE codification in the address sources is of great importance for this survey because it had been decided not to include any question on the sector of activity into the questionnaires. Instead, the sector code from the address source had to be added to each interview. The NACE codes were also used for the weighting of the data-set. Sector codes in the original country data files Only for one third of the countries it was possible to deliver the data with the new NACE Rev. 2. Two thirds of the countries delivered NACE Rev. 1.1 because the address source of first choice in these countries has not yet been adapted to the new codification. This leads to the following distribution: − 10 countries with the new NACE Rev. 2 in the address registers − 20 countries with the old NACE Rev. 1.1 in the address registers. The local fieldwork institutes were advised to deliver the finest possible NACE distribution for their data in order to facilitate efforts to transform Rev. 2 into Rev 1 coded data or vice versa. The following table shows the information on the NACE type and the number of digits in which the codifications of the sector was provided by the local fieldwork institutes:

Page 50: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

50

Table 7.2.1: NACE-type and number of digits by country Sector information delivered with the data-set

country NACE type digits AT 1 6 BE 1 1 BG 1 4 CY 1 4 CZ 2 5 DE 2 4 DK 1 6 EE 2 5 EL 1 4 ES 2 4 FI 1 6 FR 2 4 HR 1 2 HU 1 4 IE 2 3 IT 1 4 LIT 1 6 LU 1 5 LV 2 4 MK 1 2 MT 1 2 NL 1 4 PL 1 4 PT 2 5 RO 1 6 SE 2 5 SI 2 2 SK 1 5 TR 1 1 UK 1 4

As the table shows, all countries except for Belgium and Turkey provided the interviews with a finer sector differentiation than the NACE 1-digit level contained in the data-set (variables “nace1_1di” respectively “nace2_2di”). These finer differentiations can however not be delivered by us within the data-set because they would allow for a de-anonymization especially of the larger establishments in smaller countries and would thus be an offence against existing data-protection legislation18.

18 For further harmonization efforts that might be planned for the future (e.g. a conversion from the

NACE Rev. 1.1 codes to the now valid NACE Rev. 2 codes in the context of the preparation of a new wave of the survey) all existing digits are saved at TNS Infratest Sozialforschung.

Page 51: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

51

Harmonization of NACE Rev.1.1 and NACE Rev. 2 codes in the data-set For analyses to be done with the data, it is important that establishments with the same type of activity are marked with the same sector code. Otherwise, for example comparing interviews from NACE L in countries using the old NACE Rev. 1classification with interviews from countries in sector L according to the new classification would lead to a comparison of apples with pears since NACE L is “Public Administration” in the old classification while the letter stands for “Real estate activities” in the new NACE Rev. 2 version. On the finer levels, there are also a series of further changes The problem of having national data-sets with different NACE versions was tackled by creating a new variable with a harmonized code. To this end, it was decided to recode the data delivered with the NACE Rev. 2 code into NACE Rev. 1.1. The recoding was done from NACE Rev. 2 to NACE Rev. 1.1 (and not the other way round) for several reasons: − On the one and two-digit levels, the NACE Rev. 2 codification is more differentiated than

the NACE Rev. 1.1 used to be. While e.g. there are 21 NACE one-digit level codes (letter codes from A to U) in the NACE Rev. 2 version, the number of NACE one-digit codes in NACE Rev. 1.1 is only 17. This means that some of the former NACE Rev. 1.1 one-digit codes were further differentiated in NACE Rev. 2. To redress a finer differentiated codification back into a less differentiated one is obviously more feasible than the other way round.

− As mentioned, the number of countries where codes were available for NACE Rev. 2 is

much lower than the number of countries for which NACE Rev. 1.1 is the available codification. Since the conversion leads to somewhat less accurate results than the original codification, it seemed advisable to do the conversion for the type of codification provided by the lower number of countries and to leave the majority of countries with their original code.

The harmonisation from NACE Rev. 2 to NACE Rev. 1.1 was carried out on the 2-digits level. A change on the one-digit level would have caused a problem concerning the comparability of the sectors of activity in the several countries. For example “Sewage and refuse disposal, sanitation and similar activities” was in section O “Other community, social and personal service activities” in NACE Rev. 1.1 while in NACE Rev. 2 this code is in section E “Water supply; sewerage, waste management and remediation activities”. Likewise, in NACE Rev. 2 section J “Information and communication” is a newly formed section that did not exist in NACE Rev. 1.1 before. It is composed of diverse two-digits stemming from different one-digits. In most cases the harmonisation was obvious since NACE Rev. 2 has a finer differentiation than NACE Rev. 1.1. Most two-digits remained in the same one-digit level group (though these groups mostly changed the letters with which they are labelled), even when on the 2-digits level they changed the group to which they are attributed. Particular difficulties have arisen where the two-digits of NACE Rev. 2 are formed from more than one 2-digits of NACE Rev. 1.1. In these cases, decisions were taken on base of

Page 52: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

52

quantitative aspects: The two-digits of NACE Rev. 2 were classified into the two-digits of NACE Rev. 1.1 from which largest number of establishments had “migrated”19. The syntax used for the conversion from NACE Rev. 2 codes to NACE Rev. 1 codes can be found in Appendix D. Sector codes available in the delivered data set The variable “rev1_1di” contains data with original NACE Rev. 1.1 codes from the address source (for all countries that delivered NACE Rev. 1.1 data) as well as data which were converted into these codes by way of the described harmonization. The variable “rev_type” provides information on whether the codes for a particular country are original codes (for these rev_type = 1 or 3) or converted codes (rev_type = 2).

7.3 SPSS data file: General structure and hints on specific variables

Data for all 30 countries where the survey was carried out were delivered to the European Foundation in SPSS format. The integrated SPSS save file is named “ECS2009_20091001_EF.sav”. The file contains labels for all variables and values. The SPSS file contains the management (MM) and the employee representative (ER) interviews for all establishments of the included countries. ER data are already linked to the MM data of the corresponding establishment. Variables from the MM questionnaire are named “MM…”, variables from the ER questionnaire are named “ER…”. Variable names are identical with the question numbers in the questionnaires. Weighting factors in the data-set The weighting factors are stored in the following variables: − “est_wei” establishment proportional weight − “emp_wei” employee proportional weight For the ER interviews, there is an additional weighting factor “er_wei” available. This factor corrects country differences in the response rates obtained for the ER interviews. It is calculated as the ratio between the total of establishments with an employee representation in country x (unweighted) and the number of ER interviews realized in the country. The factor is highest for those countries where response rate is particularly low.

19 In these cases, an alternative would have been to randomly distribute the establishments of the

concerned NACE Rev. 2 sector among the different NACE Rev. 1.1 codes this sector is formed. But by this way, more not fully accurate attributions would have resulted.

Page 53: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

53

An example: In the German sample, a relevant employee representation was identified in 888 establishments and the number of finally completed ER interviews is 558. The “er_wei” factor for Germany is therefore 888/558 = 1,59. In the Italian sample, a relevant employee representation was identified in 943 establishments, but the number of finally completed ER interviews is only 320. The “er_wei” factor for Italy is therefore higher, with 943/320 = 2,95. The factor does not replace the employee or establishment proportional factor, but is an additional factor with which each ER interview can be multiplied if this type of additional adjustment is needed for an analysis. It has repercussions on the ER results in cross-national analyses because all interviews from countries with a low ER response rate are multiplied with a higher factor. For national analysis, the application of the factor does not make sense since it is equal for all interviews from a country. Please note: The cross-tabulations produced for this survey are NOT weighted with this additional ER weight! If using this ER weight in any analyses, this should be clearly marked because it can lead to substantially different results! Further variables added to the data-set For reasons of data protection (risk of de-anonymization), the figures on the number of employees working in the surveyed establishments (MM102a/b) are delivered only in a summarized form as size-classes (variables “size_5” and “size_10”), but not as numerical values. The variables MM102a/b are therefore missing in the data-set. In addition to the variables directly resulting from the interview, the SPSS data-file contains some additional variables which were generated newly on base of the interview data:

Page 54: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

54

Variable name Description

Id Unique identification number for each interview country Country variable, in the Eurobarometer order (EU-15, NMS, candidate

countries) country2 Country variable, in the official EU order, sorted alphabetically by country

name in national language eu27 Differentiates between EU27 countries (eu27 = 1) and the 3 candidate

countries surveyed (eu27 = 2) size_5 Size of the establishment in 5 size-classes (taken from MM102a/b) size_10 Size of the establishment in 10 size-classes (taken from MM102a/b) rev_type NACE version in which data were provided (NACE Rev.1.1/ Rev. 2) rev1_1di NACE Rev. 1.1 code at the one-digit level; for countries delivering NACE

Rev. 2 data, this variable was created by a conversion of the Rev. 2 codes into Rev. 1.1 codes

sec3 3 main sectors of activity (Industry, Private Services, Public Services), created from variable rev1_1di

sec11 Sector differentiation used for the overview report; in this sector variable, the quantitatively very small NACE 1-digit sectors C (Mining and Quarrying” and E (“Electricity, gas and water supply”) are collapsed with the large sector NACE D (“Manufacturing”) into the sector “Manufacturing and Energy”, consisting of NACE C, D and E; it is recommended to use this variable (not rev1_1di) for analyses by sector

ER_RESP This variable is based on the combination of answers to the question related to the identification of ER bodies at the establishment (MM650_1 to MM650_5 to MM654). It is defined country specifically and shows which type of body and spokesperson was finally chosen for the ER interview. For an interpretation of this variable see Annex 2.5 of the MM questionnaire.

ER_exist Variable indicating whether an eligible ER representation exists at the establishment (according to information from the MM interview)

ER_exist2 This variable is based on ER_exist, but additionally includes multi-site organisations with no representation at the local unit, but with a representation at the level of the headquarters which also represents the employees of the chosen unit (MM654 = 1)

ERint Variable indicating whether an ER interview exists for the establishment pt_exist Indicates the existence of part-timers as a simple yes/no question,

created on base of MM250/MM250x MM250XN etc. For each numerical question, an XN-variable was created in the data-set;

it summarizes the answers to the open numerical question asked in the first place (e.g. MM250) and the answers to the x-questions asked to all respondents who could not answer the numerical question (e.g. MM250x).

Page 55: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

55

7.4 ASCII data file

The data-file is also available in ASCII-format. The ASCII-file is named “ECS2009_20091001_EF.dat”. A description of the card-column-format (data map) is documented in the Annex of this report (Annex C).

7.5 Cross-tabulations

For this survey, cross-tabulations are available for all questions of content, showing results by country. The cross-tabulations were produced for both the establishment proportional weighting and for the employee proportional weighting. The ER interviews in the weighted tables received the same weights as the MM interviews they correspond to. All tables show both the unweighted base and the weighted base. Please note that especially in the very small countries the weighted base can be “0” for some of the filtered questions. A “0” in the weighted base does not mean that the question was not relevant for any establishment in the country, but it indicates that the question was relevant for less than 500 establishments (due to the divisor 1‘000 and a rounding effect).

Page 56: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

56

8. Recommendations for future surveys

On the basis of the experiences made with the ECS 2009 (and with the ESWT 2004/05) we would like to give the following recommendations for comparable future surveys:

8.1 Work organisation

The design of the survey methodology, questionnaire design, field work (data collection) and analysis are separate working steps in the research project. But these steps are closely linked to each other. Questionnaire design has to reflect the information needs on one hand and the possibilities and limitations of data collection on the other hand. Questions have to be worded in a way that they can easily be understood and answered by the respondent in the specific interview situation. For establishment surveys this means – among others – that questions about statistical data (like the size and structure of the workforce) should not go too much into detail. Otherwise there is the risk that respondents – especially in medium-sized and large establishments – cannot answer these questions without making more or less extensive investigations in their files beforehand. If they are willing to do this, this makes the interview considerably longer; if they are not willing to do it, the result will be an increased number of item non-responses or – in the worst case – of interrupted interviews. Therefore in the design phase of a survey it is important to bring together those who want to get answers on specific political or scientific questions, those who have in-depth knowledge of the substance matter and those who are responsible for field work. In the present case and in the preceding survey ESWT 2004/05 this was organised in a way that could serve as a model: The survey design and the questionnaires were developed in close cooperation by representatives of EUROFOUND, a group of experts from different countries and the institute which was responsible for field work. In a nutshell this meant that those who developed the questionnaires were at the same time responsible for the fact that the questionnaires actually meet the information needs and that they “work” in the fieldwork phase. Since questionnaire development usually is made against the background of certain hypotheses – even if these are not always made explicit – it is also advisable to have those who developed the questionnaires involved in the analysis phase. This ensures that on the one hand the information contained in the questionnaire (not only in single questions, but also by a certain combination of questions) is really used for the analysis. On the other hand, it makes sure that the data are not “over-interpreted”.

Page 57: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

57

8.2 Timeframe

The timeframe foreseen for the preparatory work to be done for the survey, namely for the questionnaire development, proved to be too short. In the original planning of EUROFOUND laid down in the calls for tender, the questionnaire development period had been dimensioned under the assumption that a large part of the ESWT questions would be repeated as trend questions in the ECS 2009. In the course of the questionnaire discussions and as reaction to the input of the newly formed Advisory Committee, it then turned out that there were many ideas for new topics so that finally large parts of both the MM and the ER questionnaire had to be developed newly. The timeframe foreseen for the questionnaire development did not allow sufficient time for this process. Since it was not possible to postpone the delivery date for the final dataset accordingly, this implied a high time pressure in the further survey preparation, especially in the multi-step translation phase. For a future survey of this kind, therefore the timeframe for questionnaire development should be dimensioned more generously from the beginning, taking into account the complexity of the discussion processes between EUROFOUND team, the Advisory Committee and the external researchers involved in the conceptualisation.

8.3 Identification and selection of respondents for the employee representative interviews

Due to the large variety of systems of employee representation existing in European countries, the mapping of ER bodies existing at the establishment level and the choice of the most appropriate body for interviewing resulted in a very complex survey instrument. The relevant questions (MM650 to MM653) had to be designed country specifically, filtering rules had to be worked out separately for each of the countries etc. This led to a highly complex CATI programme in these parts which was a real challenge for countries that had little experience in more complex survey designs as well as for the central coordinating team. For an eventual next round of the survey, it should be discussed whether there are ways on how to simplify this part of the questionnaire as well as the interface between MM and ER questionnaire without loosing important information.

8.4 Reduction of the overload of questions asking for allowances at the end of the MM interview

At the end of the MM questionnaire (MM800ff.) there is a series of questions asking the respondents for allowances:

Page 58: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

58

MM800: “allowance” to contact the employee representative for an interview MM804 and MM805: further question on the allowance for contacting the ER in his/her

leisure time for those who originally refused MM807: question on the allowance to be re-contacted for further questions MM808: question on the allowance for merging the survey data with other data. While MM800 is to our assessment and within the given survey design essential for the access to the employee representative, the questions trying to convert MM refusals in MM800 (enquiring contact details for the ER interview) into affirmations by asking for allowance to do the interview in the representative’s leisure time (MM804 and MM805) was of very limited success: all in all only 54 additional ER interviews resulted from this effort (cf. chapter 4.4). For the 2.800 respondents that maintained their refusal, this question meant an annoying prolongation of the interview. In view of the repercussions such annoying questions at the end of the questionnaire may have for response rates in further establishment level interviews, we propose to delete this “refusal conversion” question in an eventual next round of the survey. Likewise, the question asking for the linkage of the survey data with other data (MM808) should be considered for deletion if there is no precise and feasible idea yet at the very beginning of the survey project on what to do with this allowance.

8.5 Questions asking for percentage shares

Experiences with the new concept applied in this survey for the questions asking for estimates on certain distributions (e.g. share of part-timers, share of female employees etc.) in two steps – first as an open numerical question and then additionally as a prompted question with percentage categories – proved to be ambiguous. On the one hand, most respondents (usually more than 90%) could indeed spontaneously give a percentage estimate in numerical terms. Offering percentage categories subsequently to those who could not give a more precise figure then led to a couple of further valid answers so that in the end there were only few cases with “DK/NA”. On the other hand, there are hints that some respondents might have made mistakes in the spontaneous rough calculation of percentages on the telephone so that the accomplished higher degree of exactness in the results (as compared to the ESWT approach of asking in percentage categories only) has in some cases probably led only to an enhanced spurious accuracy. For an eventual next round, the approach to the numerical questions should therefore be discussed.

Page 59: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

59

Annex A: List of tables in the report

Table 1.1.1 The size of the universe Table 1.3.1 Local institutes in charge of field-work for the ECS 2009 Table 2.2.1 Available national questionnaire versions Table 2.2.2 Translation process Table 3.1.1 Address registers used for the survey Table 3.1.2 Sampling matrix Table 3.2.1 Software versions used in the countries Table 4.1.1 Fieldwork period and number of interviewers working for the project, by

country Table 4.2.1 Number of completed MM and ER interviews per country Table 4.2.2 Structure of the net sample Table 4.3.1 Non-response MM interviews non-screening countries Table 4.3.2 Non-response MM interviews screening countries Table 4.4.1 Success rate for ER interviews by type of approach Table 4.4.2 Non-response ER interviews Table 4.5.1 Multi-variate analysis of management refusals to allow for an ER interview Table 4.5.2 Multi-variate analysis of factors influencing the composition of the net ER

sample Table 5.1 Weighting matrix Table 6.3.1 Summary of results of bootstrap analysis Table 7.2.1 NACE-type and number of digits by country

Page 60: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

60

Annex B: Country codes

Country Code Abbreviation Austria 12 AT Belgium 01 BE Bulgaria 31 BG Croatia 46 HR Cyprus 32 CY Czech Republic 33 CZ Denmark 02 DK Estonia 34 EE Finland 06 FIN France 07 FR Germany 03 DE Greece 04 EL Hungary 35 HU Ireland 08 IE Italy 09 IT Latvia 36 LV Lithuania 37 LT Luxembourg 10 LU Macedonia 63 MK Malta 38 MT Netherlands 11 NL Poland 39 PL Portugal 13 PT Romania 40 RO Slovakia 41 SK Slovenia 42 SI Spain 05 ES Sweden 14 SE Turkey 43 TR United Kingdom 15 UK

Page 61: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

61

Annex C: Data map for the ASCII file

Variable Rec Start End Format

id 1 1 10 F10.0

Country 1 11 12 F2.0

MM100 1 13 20 F8.0 MM101 1 21 28 F8.0

MM103 1 29 29 F1.0

MM104 1 30 30 F1.0 MM106 1 31 31 F1.0

MM150_1 1 32 32 F1.0

MM150_2 1 33 33 F1.0 MM150_3 1 34 34 F1.0

MM150_4 1 35 35 F1.0

MM150_5 1 36 36 F1.0 MM157_1 1 37 37 F1.0

MM157_2 1 38 38 F1.0

MM157_3 1 39 39 F1.0 MM157_4 1 40 40 F1.0

MM157_5 1 41 41 F1.0

MM157_6 1 42 42 F1.0 MM157_7 1 43 43 F1.0

MM158_1 1 44 44 F1.0

MM158_2 1 45 45 F1.0 MM158_3 1 46 46 F1.0

MM159 1 47 47 F1.0

MM200_1 1 48 48 F1.0 MM200_2 1 49 49 F1.0

MM200_3 1 50 50 F1.0

MM201 1 51 51 F1.0 MM250 1 52 59 F8.2

MM250X 1 60 67 F8.0

MM250XN 1 68 75 F8.2 MM252_1 1 76 76 F1.0

MM252_2 1 77 77 F1.0

MM252_3 1 78 78 F1.0 MM252_4 1 79 79 F1.0

MM254 1 80 80 F1.0

MM256 1 81 81 F1.0 MM257 1 82 82 F1.0

MM300 1 83 83 F1.0

MM302 1 84 86 F3.0 MM302X 1 87 94 F8.0

MM302XN 1 95 95 F1.0

MM303 1 96 96 F1.0

Page 62: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

62

MM304 1 97 97 F1.0 MM305 1 98 98 F1.0

MM350 1 99 101 F3.0

MM350X 1 102 109 F8.0 MM350XN 1 110 110 F1.0

MM351 1 111 111 F1.0

MM400_1 1 112 112 F1.0 MM400_2 1 113 113 F1.0

MM400_3 1 114 114 F1.0

MM401 1 115 119 F5.0 MM401X 1 120 127 F8.0

MM401XN 1 128 128 F1.0

MM402 1 129 131 F3.0 MM402X 1 132 139 F8.0

MM402XN 1 140 140 F1.0

MM403 1 141 141 F1.0 MM450 1 142 144 F3.0

MM450X 1 145 152 F8.0

MM450XN 1 153 153 F1.0 MM451 1 154 154 F1.0

MM452 1 155 155 F1.0

MM454 1 156 156 F1.0 MM455 1 157 159 F3.0

MM455X 1 160 167 F8.0

MM455XN 1 168 168 F1.0 MM456 1 169 171 F3.0

MM456X 1 172 179 F8.0

MM456XN 1 180 180 F1.0 MM460 1 181 181 F1.0

MM461 1 182 182 F1.0

MM458 1 183 183 F1.0 MM459_1 1 184 184 F1.0

MM459_2 1 185 185 F1.0

MM459_3 1 186 186 F1.0 MM459_4 1 187 187 F1.0

MM459_5 1 188 188 F1.0

MM459_6 1 189 189 F1.0 MM463 1 190 190 F1.0

MM464 1 191 191 F1.0

MM550 1 192 194 F3.0 MM550X 1 195 202 F8.0

MM550XN 1 203 203 F1.0

MM553 1 204 206 F3.0 MM553X 1 207 214 F8.0

MM553XN 1 215 215 F1.0

MM558 1 216 216 F1.0 MM559 1 217 217 F1.0

Page 63: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

63

MM561 1 218 218 F1.0 MM562_1 1 219 219 F1.0

MM562_2 1 220 220 F1.0

MM562_3 1 221 221 F1.0 MM562_4 1 222 222 F1.0

MM563new 1 223 223 F1.0

MM564_1 1 224 224 F1.0 MM564_2 1 225 225 F1.0

MM564_3 1 226 226 F1.0

MM650_1 1 227 227 F1.0 MM650_2 1 228 228 F1.0

MM650_3 1 229 229 F1.0

MM650_4 1 230 230 F1.0 MM650_5 1 231 231 F1.0

MM650_7 1 232 232 F1.0

MM651a 1 233 233 F1.0 MM651b 1 234 234 F1.0

MM651c 1 235 235 F1.0

MM652 1 236 236 F1.0 MM653a 1 237 237 F1.0

MM653b 1 238 238 F1.0

MM654 1 239 239 F1.0 MM657 1 240 240 F1.0

MM602_1 1 241 241 F1.0

MM602_2 1 242 242 F1.0 MM602_3 1 243 243 F1.0

MM602_4 1 244 244 F1.0

MM700_1 1 245 245 F1.0 MM700_2 1 246 246 F1.0

MM700_3 1 247 247 F1.0

MM700_4 1 248 248 F1.0 MM702_1 1 249 249 F1.0

MM702_2 1 250 250 F1.0

MM702_3 1 251 251 F1.0 MM702_4 1 252 252 F1.0

MM701 1 253 253 F1.0

MM500 1 254 254 F1.0 MM501 1 255 255 F1.0

MM502 1 256 256 F1.0

MM800 1 257 257 F1.0 MM804 1 258 258 F1.0

MM805 1 259 259 F1.0

MM807 1 260 260 F1.0 MM808 1 261 261 F1.0

ER104 1 262 264 F3.0 ER104X 1 265 272 F8.0

Page 64: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

64

ER104XN 1 273 273 F1.0 ER106 1 274 275 F2.0

ER107 1 276 278 F3.0

ER107X 1 279 286 F8.0 ER107XN 1 287 287 F1.0

ER300 1 288 289 F2.0

ER301 1 290 290 F1.0 ER304 1 291 291 F1.0

ER151_1 1 292 292 F1.0

ER151_2 1 293 293 F1.0 ER151_3 1 294 294 F1.0

ER151_4 1 295 295 F1.0

ER200_1 1 296 296 F1.0 ER200_2 1 297 297 F1.0

ER200_3 1 298 298 F1.0

ER202 1 299 299 F1.0 ER203 1 300 300 F1.0

ER204 1 301 301 F1.0

ER207_1 1 302 302 F1.0 ER207_2 1 303 303 F1.0

ER207_3 1 304 304 F1.0

ER207_4 1 305 305 F1.0 ER207_5 1 306 306 F1.0

ER207_6 1 307 307 F1.0

ER207_7 1 308 308 F1.0 ER207_8 1 309 309 F1.0

ER207_9 1 310 310 F1.0

ER260 1 311 311 F1.0 ER261a_1 1 312 312 F1.0

ER261a_2 1 313 313 F1.0

ER261a_3 1 314 314 F1.0 ER261a_4 1 315 315 F1.0

ER261b_1 1 316 316 F1.0

ER261b_2 1 317 317 F1.0 ER261b_3 1 318 318 F1.0

ER261b_4 1 319 319 F1.0

ER262a_1 1 320 320 F1.0 ER262a_2 1 321 321 F1.0

ER262a_3 1 322 322 F1.0

ER262a_4 1 323 323 F1.0 ER262a_5 1 324 324 F1.0

ER262a_6 1 325 325 F1.0

ER262b_1 1 326 326 F1.0 ER262b_2 1 327 327 F1.0

ER262b_3 1 328 328 F1.0

ER262b_4 1 329 329 F1.0 ER262b_5 1 330 330 F1.0

Page 65: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

65

ER262b_6 1 331 331 F1.0 ER263a 1 332 332 F1.0

ER263b 1 333 333 F1.0

ER350 1 334 334 F1.0 ER351 1 335 335 F1.0

ER352 1 336 336 F1.0

ER353a 1 337 337 F1.0 ER353b 1 338 338 F1.0

ER400_1 1 339 339 F1.0

ER400_2 1 340 340 F1.0 ER400_3 1 341 341 F1.0

ER400_4 1 342 342 F1.0

ER400_5 1 343 343 F1.0 ER400_6 1 344 344 F1.0

ER400_7 1 345 345 F1.0

ER400_8 1 346 346 F1.0 ER401 1 347 347 F1.0

ER402_1 1 348 348 F1.0

ER402_2 1 349 349 F1.0 ER402_3 1 350 350 F1.0

ER402_4 1 351 351 F1.0

ER402_5 1 352 352 F1.0 ER402_6 1 353 353 F1.0

ER402_7 1 354 354 F1.0

ER402_8 1 355 355 F1.0 ER404_1 1 356 356 F1.0

ER404_2 1 357 357 F1.0

ER404_3 1 358 358 F1.0 ER404_4 1 359 359 F1.0

ER404_5 1 360 360 F1.0

ER404_6 1 361 361 F1.0 ER404_7 1 362 362 F1.0

ER404_8 1 363 363 F1.0

ER404_9 1 364 364 F1.0 ER404_10 1 365 365 F1.0

ER406_1 1 366 366 F1.0

ER406_2 1 367 367 F1.0 ER406_3 1 368 368 F1.0

ER406_4 1 369 369 F1.0

ER406_5 1 370 370 F1.0 ER406_6 1 371 371 F1.0

ER406_7 1 372 372 F1.0

ER406_8 1 373 373 F1.0 ER406_9 1 374 374 F1.0

ER406_10 1 375 375 F1.0

ER501 1 376 377 F2.0 ER502 1 378 378 F1.0

Page 66: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

66

ER503 1 379 381 F3.0 ER503X 1 382 389 F8.0

ER503XN 1 390 390 F1.0

ER504 1 391 391 F1.0 ER_resp 1 392 393 F2.0

ER_wei 1 394 401 F8.2

est_wei 1 402 420 F19.5 emp_wei 1 421 439 F19.5

rev_type 1 440 440 F1.0

sec3 1 441 441 F1.0 size_5 1 442 442 F1.0

size_10 1 443 444 F2.0

ER_exist 1 445 452 F8.0 er_exist2 1 453 460 F8.0

ERint 1 461 468 F8.0

rev1_1di 1 469 469 A1 sec11 1 470 473 A4

country2 1 474 481 F8.0

eu27 1 482 489 F8.0 pt_exist 1 490 497 F8.0

Page 67: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

67

Annex D: Syntax used for the NACE harmonization

*** Recoding of NACE Rev. 2 codes into NACE Rev. 1.1 codes on the two-digits level. * B Mining and quarrying. fre nace_2di. comp rev1_2di = 0. recode rev1_2di (0 = sysmis). exe. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '05' rev1_2di = 10. if nace_2di = '06' rev1_2di = 11. if nace_2di = '07' rev1_2di = 13. if nace_2di = '08' rev1_2di = 14. if nace_2di = '09' rev1_2di = 14. end if. * fre rev1_2di nace_2di. * C Manufacturing. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '10' rev1_2di = 15. if nace_2di = '11' rev1_2di = 15. if nace_2di = '12' rev1_2di = 16. if nace_2di = '13' rev1_2di = 17. if nace_2di = '14' rev1_2di = 18. if nace_2di = '15' rev1_2di = 19. if nace_2di = '16' rev1_2di = 20. if nace_2di = '17' rev1_2di = 21. if nace_2di = '18' rev1_2di = 22. if nace_2di = '19' rev1_2di = 23. if nace_2di = '20' rev1_2di = 24. if nace_2di = '21' rev1_2di = 24. if nace_2di = '22' rev1_2di = 25. if nace_2di = '23' rev1_2di = 26. if nace_2di = '24' rev1_2di = 27. if nace_2di = '25' rev1_2di = 28. if nace_2di = '26' rev1_2di = 33. if nace_2di = '27' rev1_2di = 31. if nace_2di = '28' rev1_2di = 29. if nace_2di = '29' rev1_2di = 34. if nace_2di = '30' rev1_2di = 35. if nace_2di = '31' rev1_2di = 36. if nace_2di = '32' rev1_2di = 36. if nace_2di = '33' rev1_2di = 29. end if. * fre rev1_2di nace_2di. * D Electricity, gas, steam and air conditioning supply. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '35' rev1_2di = 40. end if.

Page 68: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

68

* E Water supply, sewerage, waste management and remediation activities. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '36' rev1_2di = 41. if nace_2di = '37' rev1_2di = 90. if nace_2di = '38' rev1_2di = 90. if nace_2di = '39' rev1_2di = 90. end if. * F Construction. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '41' rev1_2di = 45. if nace_2di = '42' rev1_2di = 45. if nace_2di = '43' rev1_2di = 45. end if. * G Wholesale and retail trade, repair of motor vehicles and motorcycles. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '45' rev1_2di = 50. if nace_2di = '46' rev1_2di = 51. if nace_2di = '47' rev1_2di = 52. end if. * fre rev1_2di nace_2di. * H Transportation and storage. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '49' rev1_2di = 60. if nace_2di = '50' rev1_2di = 61. if nace_2di = '51' rev1_2di = 62. if nace_2di = '52' rev1_2di = 63. if nace_2di = '53' rev1_2di = 64. end if. * I Accomodation and food service activities. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '55' rev1_2di = 55. if nace_2di = '56' rev1_2di = 55. end if. * J Information and communication. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '58' rev1_2di = 22. if nace_2di = '59' rev1_2di = 92. if nace_2di = '60' rev1_2di = 92. if nace_2di = '61' rev1_2di = 64. if nace_2di = '62' rev1_2di = 72. if nace_2di = '63' rev1_2di = 72. end if. * K Financial and insurance activities. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '64' rev1_2di = 65. if nace_2di = '65' rev1_2di = 66. if nace_2di = '66' rev1_2di = 67. end if. * fre rev1_2di nace_2di.

Page 69: European Company Survey 2009 (ECS 2009) - UK Data Service · 2013-05-22 · TNS network (see table 1.3.1). For questionnaire development, TNS Infratest Sozialforschung co-operated

69

* K Real estate activities. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '68' rev1_2di = 70. end if. * M Professional, scientific and technical activities. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '69' rev1_2di = 74. if nace_2di = '70' rev1_2di = 74. if nace_2di = '71' rev1_2di = 74. if nace_2di = '73' rev1_2di = 74. if nace_2di = '74' rev1_2di = 74. if nace_2di = '72' rev1_2di = 73. if nace_2di = '75' rev1_2di = 85. end if. * N Administrative and support service activities. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '77' rev1_2di = 71. if nace_2di = '78' rev1_2di = 74. if nace_2di = '80' rev1_2di = 74. if nace_2di = '81' rev1_2di = 74. if nace_2di = '82' rev1_2di = 74. if nace_2di = '79' rev1_2di = 63. end if. * O Public administration and defence, compulsory social security. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '84' rev1_2di = 75. end if. * P Education. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '85' rev1_2di = 80. end if. * fre rev1_2di nace_2di. * Q Human health and social work activities. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '86' rev1_2di = 85. if nace_2di = '87' rev1_2di = 85. if nace_2di = '88' rev1_2di = 85. end if. * R Arts, entertainment and recreation. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '90' rev1_2di = 92. if nace_2di = '91' rev1_2di = 92. if nace_2di = '92' rev1_2di = 92. if nace_2di = '93' rev1_2di = 93. end if. * S Other service activities. do if rev_neu = 1 and v141 ne 43 and v141 ne 1. if nace_2di = '94' rev1_2di = 91. if nace_2di = '95' rev1_2di = 52. if nace_2di = '96' rev1_2di = 93. end if.