1 polycentric urban systems in europe and the lisbon strategy: emerging territorial patterns...

1

Polycentric Urban Systems in Europe Polycentric Urban Systems in Europe and the Lisbon Strategy: Emerging and the Lisbon Strategy: Emerging

Territorial PatternsTerritorial Patterns

Francesco Pagliacci

PhD Program in Agrifood Economics and Statistics University of Bologna - Italy

Università di BolognaUniversità di Bologna 04 maggio 201204 maggio 2012Facoltà di Scienze StatisticheFacoltà di Scienze Statistiche

Politica Economica corso avanzatoPolitica Economica corso avanzato Prof. Cristina BrasiliProf. Cristina Brasili

1. Introduction (I)

Since 2000, the Lisbon Strategy (LS) has played a major role amongst the European Union (EU) policies. By 2010, most of its goals have not been achieved.

2

Several EU policies are targeted to the regional dimension (CSD, 1999; European Union, 2007). Today, EU’s regional policy holds second place as a share of EU total expenditures after the CAP.

In spite of these efforts, socio-economic differences across EU regions are wide.

These structural differences can affect the way regions achieve the LS targets.

A persistent drawback can be observed: LS applies at the EU level. The regional dimension is ignored.

1. Introduction (II)

3

1. The relationship between the achievement of the LS targets and the extent of regional polycentrism

• EU has considered the polycentric development as a main pre‐requisite for a more sustainable and balanced development assuring also greater competitiveness to the whole EU (CSD, 1999).

• Regional polycentrism can be considered as a specific output of political local governance. Thus, polycentric regions should be considered as economic and political actors as well as large metropolitan areas are.

2. The presence of spatial patterns in the performance of the EU regions according to the LS.

• An exploratory spatial analysis is performed to test the presence of the spatial dependence in the achievement of the Lisbon Strategy’s targets. In particular, the presence of a core-periphery pattern is suggested.

Two main hypotheses in explaining these differences in achieving the LS targets are tested:

Outline and Structure of the presentation

1. Introduction

2. Theoretical background The Lisbon Strategy The Polycentric development

4

3. Hp.1: Effects of regional polycentrism Methodology : rank-size

index & issues Some results

4. Hp.2: Spatial patterns in the achievement of the LS targets Global and Local Moran’s I Emerging territorial

patterns

3. Analysing Lisbon Strategy’s targets through PCA Methodology: multivariate

statistical analysis Main results: wide differences

in EU regions

5. Conclusions


1. Introduction


5




patterns



in EU regions

5. Conclusions

2. Theoretical background: the Lisbon Strategy (I)

LS main objective was to make the EU (European Council, 2000):

“the most competitive and dynamic knowledge-based economy in the world capable of sustainable economic growth with more and better

jobs and greater social cohesion”

6

LS mainly rests on three main pillars:

Economic pillar

Social pillar

Environ-mental pillar

2. Theoretical background: the Lisbon Strategy (II)

• Overall employment rate: 70%;• Employment rate for women: 60%;• Employment rate among older workers: 50%;• An annual economic growth around 3%;• More investments in research and innovation: 3% of total GDP in

R&D.

7

Some specific targets

LS has adopted the open method of coordination (OMC) between Member States, at different levels of decision-making.

OMC is an intergovernmental method of “soft coordination” by which Member States are evaluated by one another, with the Commission’s role being one of surveillance.

OMC was a source of peer pressure and a forum for sharing good practice.

The open method of coordination (OMC)

2. Theoretical background: the Lisbon Strategy (III)

By 2010, most of LS goals were not fully achieved. LS has been affected by several drawbacks:

1.LS was ‘a wrong strategy’ for the EU integration: convergence between different economies and risk for a ‘clash of capitalisms’ (Hopner and Schafer, 2007);

2.A ‘wrong agenda’: extremely liberal mark (Amable, 2009; Rodriguez, 2002) and shift towards a right-centred approach (Pochet, 2006);

3.Uneven participation in the LS: weakness and ambiguity of the OMC;

4.LS did not take into account: the differences amongst the 27 Member States. According to Sapir (2006), deep differences exist amongst the social models of Nordic, Anglo-Saxon, Continental and Mediterranean Countries; the existence of regional differences, within each EU Member States.

8

The end of the Strategy and its main biases

2. Theoretical background: Polycentrism (I)

Evolution of the concept of polycentrism

9

In the 1960s the concept of polycentrism was adopted as a theoretical tool in the analysis of the spatial organisation of US metropolitan regions (Ostrom et al., 1961).

In the 1990s, the concept assumed a normative relevance, especially at a broader scale of analysis. It played a key role in:

1. the “European Spatial Development Perspective” (CSD, 1999);

2. the “Territorial Agenda of the European Union: Towards a more Competitive and Sustainable Europe of Diverse Regions” (EU, 2007).

2. Theoretical background: Polycentrism (II)

10

A more polycentric urban development can counterbalance the central role still played by the so-called “Pentagon”.

Source: CSD (1999)

A polycentric development can improve the promotion of economic competitiveness, social cohesion and environmental sustainability (CSD, 1999).

2. Theoretical background: Polycentrism (III)

11

Some common features

In spite of the fuzziness of the concept, there is a general consensus about polycentric regions’ main features.

In polycentric urban regions, cities are (Kloosterman et al., 2001; Meijers, 2008; Cowell, 2010):

1. located in close proximity (generally within commuting distance);

2. well-connected and interrelated through co-operation flows;

3. historically different;

4. independent political entities;

5. lacking a leading city.

2. Theoretical background: Polycentrism (IV)

12Source: Cowell (2010)

Polycentric Urban Regions: some examples

Source: OTB in Romein (2004)

2. Theoretical background: Polycentrism (V)

13

Two approaches to the analysis of polycentrism

A polycentric region can be considered a way to manage larger urban regions that differs from larger single-core metropolitan regions. Both morphological and functional perspectives are relevant in analysing it (Nordregio et al., 2004; Meijers, 2008; Veneri et al., 2010):

1. Morphological approaches analyse the way cities differing in size and population are distributed across a given region (Lambooy, 1998; Parr, 2004; Meijers, 2008);

2. Functional approaches focus on the interactions among urban centres. Several kinds of flows can be used as a proxy for these interactions (Van der Laan, 1998; Hall et al., 2006; Limtanakool et al., 2007):

• the flows of commuters;

• the flows of goods;

• the flows of information/communication.

2. Theoretical background: Polycentrism (VI)

Some critical issues

1. Definitions about polycentrism are “vague” (Riguelle et al., 2007).

2.The concept is a typical multiscalar and multidimensional one: a region may be polycentric at a given spatial scale but monocentric at a different one.

3.The positive effects of polycentrism (according to EU documents) often lack a theoretical rationale and they have not been sufficiently investigated through empirical analysis (Meijers, 2008; Veneri et al., 2010).

4.The coherence of policies enhancing a polycentric development across EU with all the other EU policies (e.g., the Lisbon Strategy) is not straightforward.

14


1. Introduction


15




patterns



in EU regions

5. Conclusions

3. Measuring the extent of regional polycentrism: sample

The analysis focuses on four European Countries: France, Germany, Italy and Spain. The analysis is performed on a sample of 75 regions: NUTS 2 regions for France (Régions), Italy (Regioni) and Spain (Comunidades Autónomas); NUTS 1 regions for Germany (Länder).

16Source: personal elaboration

When computing the extent of polycentrism, the total sample is reduced from 75 to 72 regionsi. In Germany, 3 Länder are Stadtstaaten (city-states): due to this reason, they are considered as belonging to the Flächenländer (area states) containing them.

3. Analysing Lisbon Strategy’s targets through PCA: a list of variablesPrincipal components analysis (PCA) has been applied to a list of 25 variables, focusing on: demography, economy/labour market, innovation, environment.

17

Variable Source Refer. YearResident Population Eurostat 2009GDP per capita (EU-27 = 100) Eurostat 2008GVA agriculture (% on the total) Eurostat 2007GVA industrial sect. (% on the total) Eurostat 2007Employment in agriculture (% on the total) Eurostat 2007Employment in industrial sect. (% on the total) Eurostat 2007Total employment rate Eurostat 2008Employment rate (55-64 years) Eurostat 2008Female employment rate Eurostat 2008Unemployment rate Eurostat 2008Long-term unemployment rate Eurostat 2008Unemployment rate (15-24 years) 5 report cohesion 2008Population at risk of poverty after social transfers (% of total popul.) 5 report cohesion 2008Early school leavers aged 18-24 (in % on the total same age) 5 report cohesion 2007-2009Population aged 25-64 with low education (% on the total) 5 report cohesion 2008Population aged 30-34 with tertiary education (% on the total) 5 report cohesion 2008Expenditure on R&D (% of GDP) Eurostat 2008Patent application to EPO per million inhabitants 5 report cohesion 2006-2007Households with broadband connection (% of all households) 5 report cohesion 2009Land for artificial uses (% on total) Eurostat 2009Railroad accessibility (average value of Nuts 3) Espon 2001Road accessibility (average value of Nuts 3) Espon 2001Air accessibility (Nuts 3 with max accessib.) Espon 2001Passenger cars per 1000 inhabitants Eurostat 2008Yearly average concentration of PM10 (μg/m³) (average of Nuts-3) 5 report cohesion 2009

Source: elaboration on Eurostat (2011), © ESPON Database (2006), European Commission (2010b)

3. Analysing Lisbon Strategy’s targets through PCA: methodology

18

PCA belongs to multivariate statistics (Hotelling, 1933; Pearson, 1901): it transforms a group of p indicators into a much smaller group of variables (k), still explaining a high level of variance. A correlation matrix was used. The k principal components (where k < p) come from the following linear combinations, expressed as a matrix:

Y= X A (1) where, Y is the n-by-k matrix, containing the scores of the n statistical units in the k components;

A is the vector matrix p-by-k of the normalized coefficients;

X is the n-by-p matrix of the standardized data.

The selection of the “right” number (k) of principal components (where k < p)

19

3. Analysing Lisbon Strategy’s targets through PCA: model specification

Source: personal elaboration on Eurostat (2011), © ESPON Database (2006), European Commission (2010b) [Software: R 2.13.0]

According to these results, 6 PCs are selected. They account for 81.9% of total variance;

Test KMO = 0.7633 can be considered good.

20

3. Analysing Lisbon Strategy’s targets through PCA: interpretation of results (I)Factor loadings for the 6 PCs (after VARIMAX rotation)

PC 1 PC 2 PC 3 PC 4 PC 5 PC 6Resident Population 0.716 GDP per capita 0.481 -0.568 0.247 GVA of agriculture -0.636 0.225 GVA of manufacture 0.213 0.236 0.909 Employment in agriculture -0.643 0.246 -0.287 0.258 Employment in manufacture -0.325 0.904 Total employment rate -0.528 0.765 Total employment rate (55-64 y) 0.381 0.659 Female employment rate -0.430 0.759 -0.346 0.240Unemployment rate 0.896 0.266 -0.225 Long-term unemployment rate 0.892 -0.283 Unemployment rate (15-24 y) -0.252 0.560 -0.520 0.397 -0.232 Population with low education -0.386 -0.527 0.700 Population with tertiary education -0.227 0.920Early school leavers 0.963 R&D expenditures 0.561 -0.271 0.426Patents per million inhabitants 0.687 0.247 -0.297 0.209 Household with broadband connection 0.496 0.501 0.405Population at risk-of-poverty (after social transfers) 0.794 -0.300 0.284 Concentration of PM10 0.666 0.205 0.203 Land for artificial uses (% on total) 0.599 Passenger cars per 1000 inhabitants -0.284 -0.211 -0.582Railroad accessibility 0.606 0.248 -0.516 Road accessibility 0.562 0.266 -0.519 Air accessibility 0.905

Source: elaboration on Eurostat (2011), © ESPON Database (2006), European Commission (2010b)

21

3. Analysing Lisbon Strategy’s targets through PCA: interpretation of results (II)

A short definition for each PC

PC1 (21.9% of total variance): regional urbanization and accessibility

PC2 (15.3%): weak economic performance and social exclusion

PC3 (12.2%): well-performing labour market

PC4 (11.6%): low-skilled workers

PC5 (8.2%): role of manufacturing activities

PC6 (7.3%): human capital and innovation

Regional performance according to the LS target is assessed by assigning a standardized score on each extracted PC to each region.

Regional performance according to the LS: Standardized scores for the 6 PCs

22Source: elaboration on Eurostat (2011), © ESPON Database (2006), European Commission (2010b)

3. Analysing Lisbon Strategy’s targets through PCA: main results (I)

When considering different pillars of the LS, different patterns emerge at the EU scale.

A unique relationship among the three different pillars of the LS does not exist. A well performing labour market and high investments in human capital are not always positively linked to the general economic performance.

Examples:

1. Eastern German Länder

2. Rural French regions

23

3. Analysing Lisbon Strategy’s targets through PCA: main results (II)

Therefore, other features could explain these differences in the regional performance according to the LS…


1. Introduction


24




patterns



in EU regions

5. Conclusions

4. LS and regional polycentrism: methodology (I)

Morphological extent of regional polycentrism. The rank-size index can be a crude but useful tool to measure regional polycentrism (Haggett, 1965; Nordregio et al., 2004; Meijers, 2008).

25

Methodological aspects

1. Within each region, cities are ranked according to their population.

2. The logarithms of both rank and population are taken.

3. An example: the Emilia-Romagna region (Italy)

Rank# City Populat. Log (Rank#)

Log (Pop.)

1 Bologna 371,217 0.00 12.802 Modena 175,502 0.69 12.103 Parma 163,457 1.10 12.00

4 Reggio Emilia 141,877 1.39 11.90

5 Ravenna 134,631 1.61 11.806 Ferrara 130,992 1.79 11.807 Rimini 128,656 1.95 11.808 Forlì 108,335 2.08 11.609 Piacenza 95,594 2.20 11.50

10 Cesena 90,948 2.30 11.40

4. LS and regional polycentrism : methodology (II)

Rank-size equation of cities is estimated (OLS method):

Ln (pop) = + β Ln (rank) (2)

The equation is expressed in the Lotka form (Parr, 1985), a special application of the Zipf’s Law (Zipf, 1935; 1949).

When cities are arrayed by their size on double-log graph paper, the ‘log-normal’ distribution takes the form of a straight line, whose slope is close to -1.

The law holds for big countries (e.g., India, China, the USA) as well as for the EU, but explanations about it pose some difficulties. The framework is similar to that proposed by the Gibrat’s Law for firms’ size distribution (Gabaix, 1999).

26

4. LS and regional polycentrism : methodology (III)

Estimations for the coefficient β in (2) provide a proxy for the level of polycentrism within a given region. The slope of the OLS regression line is:

o greater than -1 (regression line is flatter) in polycentric regions.

o smaller than -1 (regression line is steeper) in monocentric regions.

27

Rank-size distribution: a polycentric region and in a monocentric one

Source: personal elaboration on Istat (2001) and Insee (1999)

-1 0 1 2 3 4

91

01

11

21

31

4

log(rank)

log

(po

pu

latio

n)

y= 13.840 - 0.548x

Koln

Nordrhein-Westfalen (DE)

-1 0 1 2 3 4

91

01

11

21

31

4

log(rank)

log

(po

pu

latio

n)

y=12.161 - 1.266x

Zaragoza

Aragón (ES)

4. LS and regional polycentrism: some issues (I)

i. Which definition of city should be used to provide international comparison? The concept of Functional Urban Region (FUR) should be more appropriate in this identification problem. Due to the lack of comparable data, administrative units are used (Italian comuni; Spanish municipios; German gemeinden; French communes/ communautés d’agglomeration)

ii. National Census are the main sources for data about population

iii. Estimations are affected by the number of cities included in the OLS analysis (Meijers, 2008)Different methods:• A fixed number of towns per region?• A fixed size threshold of inhabitants?• A size above which the sample accounts for some given proportion of

regional population?Following Meijers (2008), a fixed number of towns per region is chosen (regions are largely heterogeneous): the 5, 8, 10, 12 and 15 largest cities within each region are used in the OLS models.

28

29

4. LS and regional polycentrism: the sample of cities

Source: personal elaboration on Insee (1999), Istat (2001), Ine (2001), Destatis (2008)

4. LS and regional polycentrism: main results (I)

30Source: personal elaboration on Insee (1999), Istat (2001), Ine (2001), Destatis (2008)

The extent of mono-/ polycentrism, estimated for samples of the 5, 8, 10, 12 and 15 largest cities per region

4. LS and regional polycentrism: some correlations (I)

Can the sharp differences emerging from PCA be explained through a different extent of the polycentrism at the regional level?

31

Polyc. Index 5 cities

Polyc. Index 8 cities

Polyc. Index 10

cities

Polyc. Index 12

Cities

Polyc. Index 15

citiesPC1: urbanization and accessibility

-0.215 -0.16 -0.101 -0.049 0.020(0.064) (0.1695) (0.3893) (0.6745) (0.8664)

PC2: weak economic performance/ social exclusion

0.101 0.086 0.091 0.116 0.142(0.3897) (0.4613) (0.4361) (0.3232) (0.2256)

PC3: performance of the labour market

-0.128 -0.109 -0.118 -0.112 -0.090(0.2729) (0.3515) (0.3114) (0.3379) (0.4447)

PC4: presence of low-skilled workers

0.007 0.091 0.159 0.216 0.265(0.9526) (0.4393) (0.1726) (0.06219) (0.02151)

PC5: extent of manufacture0.409 0.428 0.430 0.436 0.430

(0.00027) (0.00013) (0.00012) (0.00009) (0.00012)PC6: human capital and innovation

-0.187 -0.288 -0.351 -0.417 -0.484(0.1072) (0.01219) (0.00199) (0.0002) (0.00001)

Correlations amongst extracted PCs and polycentricity indexes

Source: elaboration on Insee (1999), Istat (2001), Ine (2001), Destatis (2008) and on Eurostat (2011), © ESPON Database (2006), European Commission (2010b)

4. LS and regional polycentrism: some correlations (II)

A positive correlation is found between “Polycentricity Index” and PC5; a negative one between “Policentricity Index” and PC6

32

Source: elaboration on Insee (1999), Istat (2001), Ine (2001), Destatis (2008) and on Eurostat (2011), © ESPON Database (2006), European Commission (2010b)


1. Introduction


33




patterns



in EU regions

5. Conclusions

5. Spatial patterns in achieving the LS targets: methodology (I)

34

According to the distribution of the scores obtained by regions in the 6 PCs, some spatial patterns seem to emerge. Thus, an exploratory spatial data analysis is performed.

Both Global and Local Moran’s I statistics are computed.

Weight matrix

A row standardized spatial weights matrix W, defined as:

The generic element can take different values:w*ij= 0 if i=j w*ij = 0 if j N(i) w*ij = 1 if j N(i) where: N(i) is the list of neighbours of the region i, according to a first order queen contiguity matrix.

5. Spatial patterns in achieving the LS targets: methodology (II)

35

The chosen first order contiguity matrix: map and main features

First Order Contiguity

Source: personal elaboration – Software GeoDa and Software R (package: spdep)

Number of nonzero links: 312 Share of nonzero weights: 5.55Average number of links: 4.16

5. Spatial patterns in achieving the LS targets: Global Moran’s I statistics

36

Each PC shows high values for the Global Moran’s I statistics: the test is well above the null hypothesis of no spatial correlation.

Moran's I p-valuePC1: urbanization and accessibility 0.3500 6.02E-06PC2: weak economic performance and social exclusion 0.6333 3.65E-15PC3: performance of labour market 0.7355 <2.2E-16PC4: low-skilled workers 0.7053 <2.2E-16PC5: extent of manufacture 0.1880 0.0078PC6: human capital and innovation 0.5738 8.26E-13

Global Moran’s I statistics and p-value for the 6 extracted PCs

Source: personal elaboration

5. Spatial patterns in achieving the LS targets: Local Moran’s I statistics

37

In order to detect geographic patterns, local Moran’s I tests are performed.

Local Moran’s I cluster maps for the 6 extracted PCs

Source: personal elaboration

5. Spatial patterns in achieving the LS targets: some results

38

The spatial analysis suggests the existence of some territorial patterns:

1.EU-level: a core-periphery pattern. More ‘central’ regions perform better than peripheral ones. According to the economic/labour market performance, poor performing regions are spatially clustered in peripheral regions. The differences pointed out by Sapir (2006) between the Continental social model and Mediterranean one also hold at the regional level.

2.National level: strong differences are observed within each Country (especially across Italy and Germany, where the divide between central regions and lagging behind ones has stronger historical roots).


1. Introduction


39




patterns



in EU regions

5. Conclusions

6. Conclusions (I)

40

In spite of the strong criticism against the Lisbon Strategy, it has played a key role among EU policies, since 2000. A major bias deals with the absence of any regional approach in the strategy.

Polycentrism plays a key role within EU planning policies. Polycentrism should foster inclusion, economic competitiveness and environmental sustainability across EU regions. It should counterbalance the key role which is still played by more central regions.

Unfortunately the analysis does not support this hypothesis. As polycentrism is deeply related to manufacturing activities, more polycentric regions perform worse than monocentric ones according to the investments in R&D and innovation.

Spatial and geographical patterns seem to play a more important role in describing regional performances according to the LS.

6. Conclusions (II)

41

The lack of any regional approach within the LS remain the most important bias. It has been unrealistic to consider the whole EU as a homogeneous area, able to tackle the same challenges in a similar way.

This lack has hindered the fully achievement of the LS’s targets by 2010. Therefore, there is now a stronger need for a general re-framing of the policy agenda of the EU: regions should be treated separately and Europe 2020 Strategy should take into account these region-specific features and issues.

Europe should become a stronger global economy thanks to its heterogeneity and not in spite of it.

Thanks for your attention!

Dott. Francesco Pagliacci

Università di BolognaDipartimento di Scienze Statistiche

e-mail: [email protected]

42

1 polycentric urban systems in europe and the lisbon strategy: emerging territorial patterns...

Documents

lisbon strategy ls

eu csd

eu regionsconclusionsoutline

eu regionsconclusions2

eu level

ls targetsglobal

eu european council

polycentric regions