Master Thesis : European Productivity Growth: An Industrial Perspective on European, National and Regional Level

Faculty of Economics and Business

Master Thesis : European Productivity Growth:

An Industrial Perspective on European, National and

Regional Level

Polyxeni Vrettou

Master Programme: International Economics & Business

Supervisor 1

_:

Prof. dr. J.H.Garretsen

Supervisor 2

_:

TABLE OF CONTENTS

Abstract 3

1. INTRODUCTION 4

2. LITERATURE REVIEW 7

2.1 Neoclassical vs Endogenous Growth 7

2.2 Determinants of Growth and Regional Wealth 9

2.3 The regional Growth from an Industrial Pespective 12

3 THEORETICAL FRAMEWORK 17

4. METHODOLOGY 21

4.1.Sample and Indicators 24

5. DECOMPOSITION OF GROWTH ON EUROPEAN & NATIONAL LEVEL 26

5.1. Employment, Labour Productivity and GVA:Industrial Pespective 26 5.2. Productivity,Partitipation Rate, GVA per capita:Changes on European& National level 28 6. EMPIRICAL RESULTS

6.1 EUROPEAN & NATIONAL LEVEL 33

6.2. REGIONAL ANALYSIS 38

6.2.1. Industrial contribution of regional growth path 38

6.2.2. Regional Analysis: by Industry 42

i. Chemicals, Nuclear Fuel, Oil 42

ii. Mining / Quarrying 43

iii. Machinery/ Equipment/ Instruments 45

iv. Computer and Related Equipment 46

v. Wholesales and Retail Sales 46

7. CONCLUSIONS-DISCUSSION 48

Appendix 50

Abstract: The main objective of this research is to test the determinants of growth within an industry specific framework. The starting point is based on the acknowledgement that growth does not secure the convergence process among locations. Therefore, additionally to the convergence hypothesis, the contribution of specific elements on this catching up process is illustrated. Moreover, we are going to focus on human and physical capital, industrial concentration and specialization as potential engines of regional growth path; it is argued that these aspects of growth should be analyzed within an industry specific context, as the latter is considered as central to the way that growth determinants may contribute to the growth process . The first step in the analysis seeks to point out some major changes in the industrial mix on European (EU-15) and national level for the period 1979-2003. Then we shift our attention to the contribution of growth determinants (physical and human capital, industrial concentration and specialization) on productivity growth including both country and industry specific effects in our model. The final stage of the analysis is going to focus on the regional level, where the contribution of specific industries on regional wealth is tested. The overall results provide some evidence that all the mentioned growth determinants play a significant role on the determination of productivity level while their relative efficiency depends both on national and industrial specific context.

1.Introduction

One of the main reasons that led to the Single Market decision in European area was the need for improvement of European competitiveness as a response to the globalization trend. This core objective shifted the attention to the factors that can boost the European productivity as the most efficient way to a sustainable growth path. From its early beginning the European Community focused on the harmonization of economic growth among members as a necessary path to EU further development. The free market was seen as a market where the increased competition and the improvement of efficiency would result in the increase of welfare and that of living standards for the European Union, as well as for its individual members.

At the same time, the integration in a Single Market was decided among member states with substantial differences in the economic performance and economic structure. Somers (1994) points out that every country has its own industrial structure and business traditions, reflecting historical circumstances, comparative advantages, government decisions, geographical conditions and chance. Therefore the target of the increased competitiveness through an integration process appears to be a perplexing multitask target and it is referring to a whole bundle of different sub-actions; including the harmonization of national long-term policies, as major economic responsibilities at individual level remain to the national authorities. Somers (1994) classified a range of conditions that should be applied in the free-market. He underlines the importance of a large and domestic market, vigorous competition, favorable physical conditions, a high concentration of related industries, a good infrastructure, entrepreneurship, starting positions and government support.

by external economies because of abundance of high skilled factors, developed infrastructure, proximity to markets/customers/suppliers. These externalities can play a central role to the determination of growth path and these characteristics of the already developed regions are likely to maintain or even enhance their comparative advantage. Therefore, sectors that may take advantage of the establishment of their activities within an already developed regional context can be also favored by the economies of scale produced in the specific region. According to Pournarakis (1995), that can partly explain the industrial preferences for closeness to the centre observed in the late 80s and early 90s.

In the early 90s, the European Community (EU) was already considered to be as one of the key players in the international market. In 1992, EU’s GDP was higher than USA’s (5,422 billion ECU versus 4,523 billion ECU, respectively.). During the years followed EU’s performance has improved further. Trade and capital flows increased significant and foreign direct investment in EU now stands at one third of GDP (as compared to an initial one fifth). The intra-area developments have in turn produced major economies of scale, competition and they had noticeable effects on productivity level. (European Commission, 2008).

However, since 90s, the economic improvement has varied among members. Differences in terms of inflation and unit labour costs still keep alive the divergence scenarios in EU. Rigidities in price and wage market are still considered as strong barriers against the adjustment process across products, sectors and regions. The poor national performance -due to inefficiencies of domestic policies- was also considered in some cases as a major reason for the unbalanced development. At the same time the international strategy remains inappropriately defined (European Commission,2008) and the overall objectives and debates on the mentioned field remain more or less the same, as the hot subjects of divergence scenarios for the lagging regions and countries, and the backwardness of innovation and technological diffusion are still on the top of the agenda.

they were at the start of the integration process. What should also be considered as crucial is the fact that European Commission (2008) points out that there is some evidence that “structural reforms in countries sharing the single currency have higher multipliers than elsewhere; that is, those countries undertaking structural reforms can accrue more benefit while those falling behind may pay a higher price for their inaction”.

Therefore, taking into consideration the unreached convergence, the focus turns on the national performance and the overall change in the key economic indicators. Assuming that allocation of factors reflects the market response to the given incentives, the main question is how different sectors respond to a given context? In particular, the starting point of this analysis is the acknowledgement that the growth objective is not equal to the nominal or real convergence of wealth among countries. Partly, this is obvious from the European policy which chose to fund specific activities and lagging regions. Although EU stated that the increase in the competitiveness rate will come through the market integration and the specialization of activities, the total abandonment of industrial specific European policies will be difficult to achieve as regional distortions are still alive (see i.e. Pelkmans (1997), Chapter 14-15).

research aims, given the convergence/divergence among industrial productivity, to answer whether there is any evidence of the industrial contribution on regional wealth.

The remainder of this paper is organized as follows: Section 2 presents a brief overview of the existing literature review related to the main topic of this study. Firstly (in Sections 2.1 and 2.2), a short outline of some of the main perspectives regarding the regional growth path is presented, while the second part (Section 2.3) focuses mainly on the industrial approach of growth analysis. Section 3 presents the theoretical framework used in the analysis and it also contains the main hypotheses tested in the empirical part. Section 4 refers to the methodology and the data used in the analysis. Sections 5 and 6 contain the actual results produced in two main stages. During the first stage (Section 5) the growth path is decomposed into the productivity growth and the growth of participation rate. The decomposition of growth for the average industrial mix as well as for each industry is likely to provide the reader with some guidelines regarding the industrial and European/National performance. The most important industries are likely to be released for the national and European level. Secondly, the correlation among the per capita growth, the productivity growth and the growth in participation rate is presented in the final part so as to provide us with additional signs for the significance of our choice to test productivity as a proxy of per capita income. Section 6 is divided into two main sub sections. During the first the contribution of an additional investment of human and physical capital, as well as the influence of concentration and specialization of business activities on industrial and aggregated level (European and National) are tested; the second sub section mainly seek for additional evidence regarding the industrial contribution on regional wealth. In the final part (Section 7) a brief discussion referring to the main conclusions and limitations of this study is developed.

2. Literature Review

2.1 Neoclassical vs Endogenous Growth

rent ratio is lower and the opposite direction holds for the labour flows. Therefore, this theory suggests a convergence towards an “optimal” ratio of wage- capital ratio (w/r). A core point in the neoclassical analysis is that nations will converge to a steady state of income through the reallocation of labour and capital. Besides, it suggests returns to capital and labour to be diminishing when the amount of available capital and labour increases. This last prediction result in the assumption that the greater the initial gap among regions the faster will be the catching up process (see i.e. Dunford et al, 2000, Bennett T. McCallum, (1996)); in other words, the greater the initial distance among regions the faster the reallocation of factors will be and therefore the convergence process. Technological innovation and the rising frontiers in the model are made exogenous and moreover its effect on different units is thought to be homogeneous. Hence, the technological improvement does not change the behaviour of each independent region as the model will continue to reallocate factors since the marginal returns to the production factors meet the zero level and therefore the model have found its steady-optimal point.

The fact that technological change is made exogenous and affects individuals (locations or even industrial activities located in a specific region/nation) at the same exogenous growth rate is a strong assumption made by neoclassical model. This assumption is relaxed in endogenous growth models, where human capital, innovation and the connected technological improvement are likely to influence the regional growth rates. Therefore, an endogenous growth model does not nesseccarily predict convergence between the best performers with the rest of the world per se. The best performers by investing in human and physical capital and in innovatory activities can broaden their frontiers with respect to their production capacity. Contrary to the neoclassical analysis, in endogenous growth model, growth is not just the result of allocation of resources in a more efficient way in “a given regime”; Growth within this model lies mostly in the ability of an economic unit to shift its regime to a higher frontier by investing in the human, physical capital and technological upgrading (Bennett T. McCallum, (1996)).

On the other hand, increasing investments in new technologies and human capital can not secure a growth path again by itself. That would mean that each country which invests the same amount of money in innovation and human capital will experience the same growth rates. Obviously, this is not a reasonable assumption to be made. First of all, it is reasonable to assume that the quality and the type of investment matter and its return values vary among heterogeneous regional and business contexts. However, there are many other components which characterize the regional growth. While the interaction of the whole bundle or a sub group of determinants is likely to interact so as to establish different frontiers in each regional case. Therefore, the acknowledgement of the complex definition of growth determination made feasible the development of a vast number of nonoverlapped studies referring to the determinants of growth (see also Durlauf (2001)).

The section below (Section 2.2) aims to outline some main guidelines with respect to the different perspectives of analyzing the regional growth path. The final part (Section 2.3) presents some representative researches that focus on the industrial perspective of growth determination. In particular, we argue that most of the growth determinants (like accessibility, market proximity and so on) are likely to have an industry-specific effect. If the argument of the importance of productivity on national/regional wealth appears to be true then the industrial path of improving the regional productivity is considered as core element for a deeper insight on individuals’ wealth.

2.2. Determinants of growth and regional wealth

human capital and technological patterns (Lopez-Pueyo et al, 2008 ). On the other hand, the effectiveness of the above interaction also depends on determinants of growth already mentioned; the initial technological and human capital gap is considered as crucial component for the model ability to improve its performance. But in contrary to the neoclassical model, an extended endogenous growth model does not assume per se that initial gap will be pros for lagging behind regions which can now gain from the “exogenous” technological achievements first made in the advanced world.

Arbia et al (2003) developed a model that is tested for a sample of 119 NUTS 2 regions and they conclude that regions do not converge to a common value of per capita income. In their study, regions with no connectivity with other regions are excluded from the sample (British, Irish and Greek regions). Therefore, the effects of accessibility or the lack of it in a regional context are underestimated. However, many other studies (i.e. Aumayr et al 1996, Fischer et al 2004, Brauninger et al 2005, Navaretti et al 2004 pp.151-183) point out that accessibility and market proximity, as well as possible spillovers effects are important factors of regional growth and thus, of regional convergence .

Badinger et al (2005) underline the importance of human capital and factors’ participation in production, combined with the technological interaction between regions as a necessary condition to regional growth. Human capital is considered as the necessary constraint for the further positive effect if other conditions hold. Thus, investments in human capital -by upgrading the social capacity- make possible not only the development of innovation; but they also maximize the positive effects of the externalities that can be achieved through the technology transfer. Therefore, it is a determinant which can enforce or set barriers to the further regional growth path.

role which can shift activities in the opposite direction from the one that the agglomeration theory suggests.

Fischer and Stirböck (2004) and Aumayr (1996) found evidence of convergence within a given type of regions (clusters). This convergence process is accompanied with different speed of adjustment among different clusters. Therefore, multiple steady points are likely to coexist among the several types of regions, in contrast to the main assumption of neoclassical theory for a unique optimum steady point; Referring to the speed of the adjustment process, Fisher and Stirböck (2004) ends up with the estimation that “it will take, for example, 34.7 years in club A (173 EU 15 regions –NUTS 2) and 14.5 years in club B (CEE and Southern Europe) for half of the distance between the initial level of income and the steady-state level of the respective club to vanish. In addition, a higher convergence speed for regions in Central and Eastern Europe is evident, thus indicating a process of catching-up towards the richer Western regions” (Fisher and Stirböck, 2004 (pp 16)).

Additionally, Aumayr (1996), explore the importance of labour share as a sign for the regional stage of development. In particular, under the acknowledgment that each region may enjoy a different phase of development, she assumes that the latter is connected with shifts of labour from the one sector to another. Thus, in the early industrialization we expect a shift of labour from agriculture to the high value activities and the deindustrialization stage would be connected with the labour shift from industry to services. Then she operates a comparative analysis based on the core distinction between urban and non-urban regions. 14 region types are classified by their economic and spatial indicators and she ends up with the conclusion that these two aspects matter for the determination of GDP growth. Finally, while she finds evidence that there is a different speed of adjustment (or even divergence) when the national boundaries are taken into consideration, the paper argues that the case of convergence appears when the analysis is developed within the boundaries of each specific regional type.

Pigliaru , 1999). Although several studies have shown for instance that the productivity gap between EU and US is one of the most important reasons of the per capita GDP between them, the restructure in sectoral mix is not covered as much as expected by convergence literature.

2.3 Growth path from an industrial perspective

Maudos et al (2008) distinct between the intra sectoral effect and the structural change effect and they argue that the mentioned gap between EU and US is a consequence of the gap in intra-sectoral effect. Intra-sectoral effect arises because of the differences in the level of productivity in a specific sector among regions (EU and US in the above paper).Structural change refers to the efficient or less efficient redistribution of inputs in highly productive activities/sectors or in activities/sectors with relative high productivity growth rates. According to the authors, EU modest performance in productivity growth originates in low productivity improvements within sectors. Therefore, a deeper insight in the sectoral productivity and its determinants appears to be important for the understanding of European and furthermore national and regional performance. The rest of this section gives a detail summary of the paper written by Ezcurra et al (2005), which provides the general framework for this research.

.

Ezcurra et al (2005), by running a shift share analysis on NUTS 2 data conclude that regional component –rather than industrial mix- has greater contribution to the inequality among regional productivity rates. They use data consisted of 197 NUTS 2 regions belonging to 15 Member States for the period 1977-1999. The authors first point out that different patterns and catching up speed adjustments appear among different regions. Some regions seem to experience higher growth rates than others.On the aggregate level, authors compute for Europe an annual average accumulative growth rate of 1.81% (Employment growth 0,41% and Average Growth Rate 2.22%).

each region. The first category includes regions which achieve a per capita growth higher than the European average (79 regions). The second category includes regions which achieve an average per capita growth lower than the average European (115 regions).

Additionally a further classification divides the former in three new group types. In the first group (type I-34 regions) initial production and employment growth was lagging behind and therefore these regions are likely to additionally benefit from the development already achieved in the advanced countries. The second group (type II) includes 22 areas which perform worse than the European average with respect to the output growth performance. However, due to the poor employment growth, the per capita production is higher than the Euro-average. Finally, the third “well performed” group includes 23 regions for which high rates of performance exist due to the negative employment growth (Type III regions).

Table 2.1. Regional Types within euro-15. (period 1977-1999) Performance TYPE No Regions % Total Populat (1999)] Description I 34 20,51

The initial production and employment growth lag behind. These regions are likely to be additionally benefited from the development achieved in the advanced countries

II 22 12,25

They perform worse than the European average with respect to the output growth. However, due to the poor employment growth, the per capita production is higher than the Euro-average

III 23 9,34 High performance rates are mostly due to the negative employment growth

a b o v e E U -a v e ra g e Total 79 42,1

IV 44 17,11 They perform worse both in terms of production and employment growth.

V 22 11,03

Their register level of production is below the Euro-average, but the employment growth rates is higher than the average European employment growth.

VI 49 27,3 They enjoyed a lagging starting level of production and employment

b e lo w E U -a v e ra g e Total 115 55,44 Eu-Average Brabant Wallon, Provence-Alpes-Coˆte d’Azur, Emilia Romagna 3 2,45 Source: Ezcurra R., Gil C., Pascual P., Rapún M., (2005), pp 681-682

Based on previous researches, the author builds their theoretical framework based on the argument that regional differences in per worker productivity are to be mostly blamed for regional inequalities. Therefore, the reasons that lead to these inequalities are considered as the crucial point of their regional analysis.

Productivity in each region is expressed by:

Where i=1,2,..n denote regions and j=1,2,..m sectors. Additionally, E denotes employment and X value added, hence

γ

s

presents the (industrial) employment share. Differences in productivity among regions can be due to regional specialization in more productive activities or due to differences in labour endowments which are likely to have the same effect on all the considered sectors.

The next step in the analysis is to decompose the productivity gap by comparing individual’s performance with the European average. What the authors seek to capture is the three determinants of regional performance, namely, industrial mix, region specific differential with homogeneous effect on all sectors and the interaction between these two factors. Using the formula illustrated above and by developing simple algebraic transformations, the authors compute the difference between the productivity of region i with the European average by using the formula (2.2):

2.2

(explained in Table 2.2 below.)

The above research study plays a central role to this paper. This work adapts Ezcurra’s et al (2005) view of analyzing the regional growth path. However, their final results based on the assumption that physical capital and technological improvements (through R&D expenditures), as well as human capital are considered as having equal effects to all sectors. Therefore, the contribution of these aspects on the regional productivity argued to be the most important determinants of regional growth. However, the present study, by relaxing this strong assumption, seeks to analyze the contribution of these elements under an industrial perspective. The starting point for this paper was the fact that the regressors used in the final analysis of Ezcurra’s study are likely to have different effects with respect to different sectoral contexts. Therefore, regional capacity in human and physical capital and so on can be just a preliminary sign of favourable regional conditions for activities experiencing a high sensitivity to these factors. Hence, given the convergence/divergence among the productivity level of different sectors (Ezcurra et al (2005)-pp 690), the regional productivity trends in the above work may underestimate the importance of the industrial contribution on regional wealth, just by arguing that the determinants of growth presented above have not any industry-specific effect.

TABLE 2.2. Decomposition of productivity gap

SYMBOL DERIVATION CLASSIFICATION DEFINITION

γi Productivity gap Productivity gap between

the region i and the Eu-average

εi Structural

Component

The impact of the

difference between regional industrial mix with the European average

ρi Regional or

Differential Component

Sector by sector

productivity gaps between region i with the European average αi Allocative Component Regional degree of specialization in sectors where productivity is higher than the European average

Source: Ezcurra R., Gil C., Pascual P., Rapún M., (2005), pp 683-684

Table 1 –Ezcurra et al, 2005 ( pp 68): Model & Parameter Estimates

Model Identification: Independent Variable: Regional Productivity Gap with the European average. Regressors: Structural Component (ε), Regional or Differential Component (ρ), Allocative Component (α).

Parameters: .α=constant term, b= Coefficient

Source: Ezcurra R., Gil C., Pascual P., Rapún M., (2005), pp 689

Model Identification: Independent Variable: Regional or Differential Component (ρ). Regressors: RnD (RD), Human Capital (H) and Physical Capital (I).

Neoclassical theory suggests that the production factors will be shifted from the activities with low marginal returns to activities with higher returns. Assuming that productivity reflects these returns on production factors, there should be an equalization trend for the individuals’ productivity performance, if everything else remains constant. On the other hand productivity growth in the neoclassical model will be directly linked with an increase in the capital-labour ratio which can be the result of an additional investment in physical capital or a change due to the exogenous technological improvement. An endogenous growth path does not necessary exclude the neoclassical point of view. However additional investments in physical capital or even the technological innovation taking place outside the industrial or regional boundaries are not a priori able to force individuals to grow faster. The quantity and quality of human capital are also likely to play a significant role for the determination of individuals’ ability to innovate or use properly the innovations produced elsewhere.

Taking into consideration that the liberalization in the European Union is expected to enhance the diffusion of ideas and new technologies, the ability of a lagging behind activity/region to catch up with the advanced productivity level depends on its ability to use this advanced technology. However we can not predict the way that human and physical capital can contribute to this growth path. In particular, it is not reasonable to assume that an increase in physical and human capital will have a homogeneous effect through countries and activities. Inter alia, the effectiveness of an additional investment is likely to be affected by national or regional effects as well as by industry-specific effects. National characteristics that reflect institutions, regulatory environment, geographical location, physical conditions and so on are likely to affect the way that the mentioned level of human and physical capital will influence the productivity growth. On the other hand, the regional industrial mix should also be an important constraint in the determination of the productivity level.

consideration the industrial perspective of the above statement, the argument, for instance, of perfect mobility of factors among various sectors seems even less reliable.

In the endogenous growth model, convergence and divergence not only on the national/regional but also on the sectoral level can not be argued ex ante. The increased importance given in R&D as a growth determinant is mainly driven by the importance of innovatories in shifting the model frontiers to an upper level. Within this study, we assume that high productive sectors should mostly “blamed” for the innovations taking place worldwide. The reason to assume that is that expenditures in R&D usually require a high budget (and therefore risk), high quality and/or quantity of embodied physical capital and a long run perspective for achieving the benefits of this investment. Therefore, first of all, low productive sectors are less likely to be able to afford such ambitious projects. Secondly, considering the investors as profit maximizers we expect that expenditures in innovation would have a potential beneficiary effect for those undertaking such long term project. These additional benefits should at least promise to cause the divergence in the productivity level among different sectors. Moreover, often the ability for this potential divergence (especially within an industry specific framework) is one reason that can lead to the decision for R&D expenditures. A sign for the significance of this statement is the increased importance given in the restructuring of the regulatory environment referring to the patents.

indicator that can boost the industrial productivity, with possible influence the broadening gap with the lagging behind industries.

H1: Human capital argued to play a significant role in the determination of productivity growth within an industry- and regional- specific framework

On the other hand the available physical capital is also likely to affect the productivity growth. Basic infrastructure and business facilities provided within the national/regional boundaries are likely to have a direct effect on the regional productivity growth, for example by affecting the speed of diffusion of ideas,minimizing the time needed for the industrial transactions and hence improving conditions that may increase the possibility for further technological achievements. At the same time, these favorable conditions may result in cutting down the production costs; thus, the expenditures in physical capital are also likely to affect the industrial competiveness. However, once more, we expect that the effect of an additional investment in physical capital has not a homogeneous effect on every industry. For example, investments in new technologies like ICT or Software are not likely to affect in the same way agriculture and Financial Services.

H2: Physical capital argued to be a significant determinant of productivity growth but its effect is likely to vary not only among locations but also among industries.

to perform better, by achieving higher growth rates in comparison with the non-agglomerated units (Brauninger- Niebuhr, 2005).

H3: We expect that concentration/specialization of activities has a positive effect on the industrial/regional productivity.

Therefore, using the productivity level as a well behaved proxy of per capita income and after testing the hypotheses one to three (H1-H3) mentioned above we would have some additional evidence that investments in human and physical capital and the concentration/specialization of activities affect the regional wealth. Additionally, as it is already mentioned we expect some of the industries (for example, the high tech industries) to be an important determinant of regional wealth. In detail, considering that some activities are likely to take advantage of their ability to innovate and use the new technological achievements, they are likely to achieve higher growth rates. Therefore, it is expected that the expenditures that contribute in the industrial productivity (especially those that may enhance a broadening productivity gap) would play an important role as fuelling or set barriers against the regional growth capacity.

H4: We predict that the concentration of high productive activities in one region, and the expenditures invested in that industry (In human and physical capital) systematically would influence systematically the level of regional per capita wealth.

4. Methodology

a benchmark case is based on some generalizations that make unfeasible a deeper insight on specific regions. (Aumayr, 1996). Some main general questions are likely to be answered from the informations revealed by the observed restructuring of business activities. Which are the main changes among industrial productivity, employment and output on national and European level? Which industries seem to achieve a divergent performance? Although, we are not going to develop a detailed analysis of industrial/national decomposition of growth, the informations provided in that part are likely to help the deeper understanding and facilitate the interpretation produced by the following empirical analysis. The observations from the decomposition of growth would be also our guideline, in the final part of the regional analysis (see below).

A core assumption through this research is that, given the decomposition of growth expressed by the productivity level and the participation rate (formula 2.1), productivity is a well behaved proxy of per capita income. Therefore, a general overview of the change in the magnitudes of productivity (output per employ), participation rate (ratio of workforce to total population) and Gross Value Added is important not only from an industrial perspective but also on European and National level. Thus, the average European and National performance regarding these indicators during the period considered will be our main focus on the next sub-part.

Assuming that the production function is a function of labour(L), human (H) and physical (K) capital, the productivity level can be expressed as a function of the physical and human capital per worker.

(

Y/L)

A*B*f

Where Y=Output , Y/L=Output per worker/labour productivity, K/L & H/L=Physical and human capital per worker and A&B sign the national and industrial effects that are likely to interact in the final determination of the level of productivity .

Assuming a Cobb Douglas general functional form and after taking logarithms we end up with our main function represented above

Log (Productivity)= β1* log (Human Capital) + β2*log (Physical Capital)

+δ1*C1+δ2*C2+ v (4.2)

v.= the error term & “C1”- “C2”=national and industrial fixed effects., β,δ= Coefficients

Finally, we wxpect that one of the determinants that is included in “v” in the above function is the specialization/concentration of activities which is likely to play a significant and positive role on the determination of productivity level. Adding the past performance (one year lagged productivity level) for the reasons already explained the final general functional form is :

Log (Productivity)=k1* log (Past Performance) + k2* log (Human Capital) + k3*log

(Physical Capital) + k4* log (Concentration/Specialization) +δ1*C1+δ2*C2+ u

u.= the error term & “C1”- “C2”=national and industrial fixed effects., k,δ= Coefficients (4.3)

Equivalent, if we want to express (4.3) in terms of productivity gap the the former can be rewritten as :

Log (Productivity Gap)=k1* log (Gap in Past Performance) + k2* log (Gap Human

Capital) + k3*log (Gap Physical Capital) + k4* log (Gap Concentration/Specialization)

+

ε

ε

Where gap denotes the relevant ratio between the industrial productivity in region i . i.e. Gap in Human Capita ratios l between Industry 1 and 2 = (Human Capital in Industry 1)/(Human capital In Industry 2), where the industry 1,2 can be refer to the same industry in two different regions ie. (region i.-EU average).

The second stage of the study will be concentrated on regional (Nuts 2) level. Similar to the previous analysis what we are going to test here is the effect of investments made in specific industry and region, and that of domestic public expenditures that invested in the specific industry but not exclusively in a specific region. Therefore the second indicators will be a sign of the capital expenditures made by country i for enhancing the industrial development. Instead of repeating the previous analysis on a regional level, what we attempt to capture here is the regional response to specific industrial characteristics. The first part of analysis should provide us with some evidence whether there is significant effect of the specific indicators on national and European level with respect to specific industries. At this step, we focus on the classification of rich and poor national performers, regarding that the ratio of regional to European wealth can be decomposed into two ratios, this of national to European wealth and that of regional to national wealth. The first stage of analysis refers to the first ratio while the second part may contain some additional informations about the second one.

Unfortunately, the inconsistency of historical time series led to the decision for individuals regressions with respect to the presence or not of a specific bundle of industrial characteristics (concentration, social and regional capital expenditures). The decomposition of growth will be used as a guideline for the label of the key industries that will be tested in the final part. The general functional form (Formula 4.3-4.4) illustrated above (used in the first part) will also applied in that part. What additionally we are going to test here, apart from the industrial contribution on regional wealth is whether there is systematical difference in the response of the best and lagging behind national performers to the given industrial growth determinants.

4.1 Sample and Indicators

providing clear picture for the sub-sectoral ones. Moreover, taking into consideration the importance that is given to the high tech industries as well as the expenditures on R&D and education , an additional classification of these industries was decided. Thus, using as a starting point the 10-sector classification (Agriculture, Mining, Manufacturing, Public Utilities, Construction, Wholesales and Retail Trade, Transport and Communication, Finance-Insurance and Real Estate, Community Social and Personal Services, Government Services) and combining this narrow labeling with the 60-industry classification (both are provided by Groningen Growth and Development Centre) the final grouping includes 18-sector categories (Appendix Table 4.1).

The indicator(s) of physical capital will be extracted from the Total Economy Database (available at Groningen Growth and Development Centre (GGDC)) which provide detailed information about fixed capital stock and formation, by type of investment, for the period 1980-2003. Additionaly, the participation rate (number of employees divided with total population) and the population growth is based on Total Population Database: Historical Statistics for the World Economy (Copyright Angus Maddison). As proxy for the human capital will be used the ratio of enrolments (per worker) in the tertiary education (OECD Database 1985-2003).

The first indicator will be the concentration of activities, taking into consideration both the specific industrial context and the total concentration of activities within the specific region. We are going to use as a proxy of concentration index the share of industrial units that are concentrated in the specific region and the same calculation applies to the regional concentration of total activities. The investments per worker made in specific industry and region and the public expenditures in industry (as a percentage of GDP) will be our indicators for the social and regional capital creation. Our independent variable will be based on the GDP per capita measurement for income (which is available at the Nuts 2 Level by Eurostat) for the period 1995-2003. However our sample is likely to vary among each part, this limitation is driven by the availability of historical series for industries and regions.

5. Decomposition of Growth On European And National Level

5.2. Employment, Labour Productivity and GVA:Industrial Pespective

As mentioned above, the objective of growth in Europe, as this is illustrated by the average weighted sectoral performance, is expected to be enhanced first of all due to the reallocation of factors from low productive to sectors with high productivity level. Therefore, there should be a convergence among the productivity of specific factors within different sectors, as far as the reallocation of factors is possible within a given bundle of industries, assuming everything else remains constant. On the other hand, in case that reallocation of these units requires an adjustment in labour characteristics (for instance by promoting the employees’ education through industry-specific seminars etc.), the reallocation of this procedure can be extremely time demanding, even in the absence of market or other rigidities.

Obviously, these flows highlight the shifting labour shares to tertiary sector. It should also be mentioned that on the European level the total persons engaged increased from 147.983 thousands to 171.167 employees in 2003 (15,7%) when the European total population increased only by 8,7%(Based on United Nation’s Database).

Pointed decrease in employment rates within the manufacturing (Industry No 17),

seems to be less important than the sharp increase within service sector (No18). The number of employees in manufacturing decreased by 6.775 thousands employees (-27,7% decrease) while the expansion in labour employment in services count for 16.409 thousand of employees (134,7% increase in comparison with the number of employees in 1979). Thus, Industry No 17-Rest Of Manufacturing experienced a fall in its labour share (persons engaged) from 17% of total persons engaged in 1979 to 10% in 2003. A second sector with considerable large decrease in its relative labour shares is Number 1 (Agriculture, Forestry Fishing) where there was a decline in the relevant employment share from 10% in 1979 to 4% in 2003, a percentage which is at the same time equivalent with 52,7% decrease in the number of persons engaged. The number of employees per sector, the relative share of each sector to the total person engaged and the change of the relative magnitudes between 1979 and 2003 are given in table 5.1 below. Additionally, what might worth to be outlined is that eight out of 18 sectors cut the number of employees engaged during the period 1979-2003.

Furthermore, Computer and related activities remained relative stable. Machinery and related equipment and Instrument, Telecommunications/Communications, and Electricity, gas and water supply are the three factors with the sharpest increase in productivity levels. (See also Appendix Fig 5.2.1 and 5.2.2). Only for Agriculture productivity remained under the EU average during the period 1979-2003. Transport also remained under the EU average but in general its value converges to EU average productivity during the considered period.

The share of sectoral to total employment multiplied by the labour productivity will provide us with a measure of sectoral contribution to the EU-15 per capita performance (Appendix Part II. European Performance:Figure and Table No 4.3). Alternatively, the number of employees in industry multiplied by the industrial productivity will provide us with the industrial contribution to the aggregate total performance. Telecommunications/Communications, Computer and related activities, Machinery/Related Equipment/Instruments seem to be the sectors which expand their shares to the aggregated performance as this is measured by the industrial output. Table 5.4 (Appendix), summarizes the overall sectoral performance for the period 1979-2003. First of all, during the period 1979-2003, all the sectors -with the exemption of No2 Industry (Mining and Quarrying) - increase their Output (Volume Indices 1995). Finance Insurance/Real Estate and the Rest of Services are the two sectors that hold a positive change in their contribution to the aggregate performance in comparison with their initial shares. For all the other sectors, their percentage held a negative change during the above period. Among the latter Agriculture-Forestry-Fishing, Mining-Quarrying, Construction, Public Administration-Defence and Education demonstrate a negative change in their shares to the value added by 30%, 53%, 34%, 23% and 25% respectively.

specialization trends are easily pointed out. For example, Mining/Quarrying is one industry that although is considered as a declining one on European level, the increase of the sectoral importance for some countries (i.e. Denmark) outlines an observable specialization pattern. (Detailed Notes and Tables for the national performance, presented by country, are illustrated in the Appendix:Part II: National Performance).

5.2. Productivity,Partitipation Rate and GVA per capita:Changes on European and National level

Most researches mention that a lagging productivity growth is a core determinant of the outlined gap between the European Union with the rest of the key international players. Literature review argues that the growth in per capita income is driven mainly by the growth of productivity. Productivity is seen as the most important determinant and the best proxy if we seek to explain the change in per capita income. However, a boom or a sharp reduction in participation rate is what can make the trend between the two variables to differ. Elbers and Gunning (2002) argue that in general the path of GDP per capita growth is close to GDP per worker, except for countries for which the initial participating rate is extremely low.

By plotting the time series of productivity and per capita income, we can argue that there is a clear sign of correlation between the trends of the two curves over the period considered (1979-2003). On the other hand participation rate shows some ups and downs and its time series’ values do not produce the steady increasing trend that productivity and GDP per capita shows. A sharp boom or a sharp declining rate is obviously able to influence the growth in per capita income, especially when productivity follows a steady and smooth path of development. The graph 5.2.1 illustrates the trends for productivity and GDP growth. The two indicators seem to have the same growth path during the period 1979-2003. On the other hand, the shocks in the participation rate in Europe (second right figure in the same table) should have played a significant role during the period considered.

the change in per capita income. However, after 1985 and till the early 90s the growth in productivity remained relative stable while the participation rate grew fast and hence intuitively its effect on GDP per capita growth is likely to be more important than the growth in productivity rates during this period. The third period (till 2003) signs an increasing tendency for both the participation and productivity rate and intuitively, for GVA per capita rate. However the growth pattern in GDP per capita seems to be more closely related to the productivity growth. Especially, after 2000 when the participation rate seems to experience the third downturn during the period considered; the growth in productivity seems to lead the change in GVA per capita. As a final general conclusion, we can sum up that although the productivity seems to be a better proxy for the overall GVA performance, the relative importance of productivity and participation rate seems to vary over time

Table 5.2.1

Productivity and Per Capita Income (1979-2003)

(Volume Indices 1995=100)

`

*Productivity=Value Added Per Worker , *GVApc=(Productivity)*(Participation Rate)

*Source 1:60-industry Database (GGDC), * Source 2:Total Population: Historical Statistics for the World Economy: 1-2006 AD (Copyright Angus Maddison)

Participation Rate (1979-2003)

*Participation Rate=Employment/Population *Source 1:60-industry Database (GGDC), *Source 2:Total Population: Historical

Statistics for the World Economy: 1-2006 AD

(Copyright Angus Maddison)

To examine the relative overall importance of productivity and participation rate on the determination of per capita GVA(GVA.pc), we will test the correlation between these two variables. We can argue that overall productivity growth is strongly correlated with the GVApc with an overall correlation 0.7291 vs 0.7060 correlation between participation and GDP.pc growth. Nevertheless, it is important also to point out that the

correlation between participation and GVA.pc for the first period (till 90s). The correlation of productivity and GVA .pc expanded after 90s and the opposite happened between participation and GVA growth for which the correlation slightly decreased after 90s. However the overall correlation of participation growth with GVApc growth is weaker than that of productivity. A last point that is highlighted by the correlation test is that the correlation between productivity and participation growth turned from a negative to a positive relation after 90s but the correlation remains relative weak.

Table 5.2.2

Growth in GDP per Capita, Productivity and Participation Rate

*Productivity=Value Added Per Worker ,*Participation Rate=Employment/Population *GDPpc=(Productivity)*(Participation Rate),

*60-industry Database (GGDC),

*Total Population: Historical Statistics for the World Economy: 1-2006 AD (Copyright Angus Maddison)

The first step in the above analysis is to examine the sectoral path of growth in Europe as this is likely to provide us with a representative picture for the European level; at the same time, we can derive some additional information about the inter-sectoral rather than intra-industrial adjustment flows. On the other hand, even under the general assumption that improvements in productivity are directly connected with the level of capital stock and capital formation, nothing can verify that this trend is homogeneous on European, national and regional level. However, the above statement is linked with the acknowledgement that the European average in principle is likely to represent the general trends of large sectors and large countries. A good performance achieved by a smaller sector can be largely underestimated in terms of productivity growth on the European level (see also Jan van der Linden, 1998 Chapter 5). The same argument can be assumed for the case of small countries in terms of their share in the European

GVA. However, small sectors may play an important role on national level and therefore the deeper insight on the sectoral performance may be valuable for a broader understanding of trends explored on national level.

Repeating the analysis made above on national level returns quit different results. In the graph below we plot the growth in per capita Gross Value Added, the growth in participation and productivity level, illustrated by country (Appendix:Part 2:Figure 5.2.3). A first comment based on the figure is that some countries seem to be more volatile than others. For example Austria and Belgium (no1 and no 2), seems to follow a steady growth path; the ups and downs in per capita GVA growth seem to be more closely related to productivity than to the participation growth. However, these countries present a relative constant growth rate with respect to all the three indicators. Other countries, i.e. the United Kingdom seem to have steeper ups and downs in terms of productivity growth but in general the growth in per capita GVA seems to follow the smooth growth path illustrated by its participation growth line. Generalizations obviously are not allowed in that part of the analysis. A general comment that can be derived is that countries that present great volatility in their participation rates, they also experience the same growth path in their productivity rates while the opposite does not appear to be significant. The opposite is true for countries like Portugal (No 11) for which the GVA growth seems to be more closely related to the volatility of productivity growth rates.

importance of the productivity growth on national wealth, the participation rate seems to be more significant on national level than in the average performance illustrated by the European average (75,9% correlation on the national level versus 70,6% on the European level).

6. Empirical Results

6.1. European and National Level

First of all referring to the physical capital, we consider two types of physical capital: Fixed Capital Stock and Fixed Capital Formation (Source: M. P. Timmer, G. Ypma, B. van Ark, Research Memorandum GD-67, Groningen Growth and Development Centre, 2003). The reason to do so is that the capital stock is an indicator of how sectors react to a given existing level of physical capital while the fixed capital formation is a variable which gives a sign for further investments and creation of physical capital. A strong assumptions at this stage is that total stock and fix capital formation is equivalent available to all sectors and therefore we use as measurement the ratio of total capital stock and capital formation divided by the number of employees.

development in a specific area. On the other hand, high tech investments and innovatory projects are usually seen as projects with high budget and long term perspective. Therefore they are likely to be undertaken by the most efficient sectors.

The use of different variables in our model may serve the pinpointing of specific growth engines for each sector. Unfortunately, the correlation between the different categories of capital stock and capital formation holds values above 0.7794. The colinearity between variables prevents us from using different variables in our model. The next step is to check the correlation between the two chosen indicators: The correlation between capital formation and capital stock is as expected extremely high (0.9990) and therefore we can not include both variables. After running regressions for each of the two indicators (which end up almost with the same results) per worker capital stock would indicate the level of physical capital in the empirical part above.

Similar to the case of different types of physical capital, investments in human capital possibly influence in a different way each industrial category with respect to time. However, there is a strong limitation in the available historical data. Although many different indicators are available, there is a lack of historical data series. The chosen indicator of human capital refers to the number of enrollments in tertiary education (OECD Stat Database) as a percentage of total employees. Although there are different type of indicators for each level of education, the missing values makes inconsistent the estimation for the period 1985-2003. Therefore, the period considered is finally limited in the period 1990-2003, while limitations still exist especially for the case of Luxembourg where we estimate a constant (in values) increase for the period 1991-1996.

of human capital should illustrated as relative weak for them. On the other hand, capital stock is likely to have an increasing effect at the early stage of business restructuring when countries lack even of basic public and social facilities but after that stage, when the country reaches an minimum level of capital then its contribution can be relative weak or not significant. The values of the enrolments (per worker) in tertiary education and that of capital stock (per worker) are the final chosen indicators for the human and physical capital.

What is important in that stage of analysis is whether or not an increase in human and physical capital is likely to enhance or not the industrial catching up process and in which way. Furthermore, in contrast to Ezcurra’s assumption we seek for evidence that there are significant not only national but also industrial constraints affecting the influence of human and physical capital on productivity growth. Therefore, we can not argue that the overall increase of human and physical capital have homogeneous influence within different sectors.

What finally we are going to test is the significance of concentration / specialization on national and industrial growth path. In that part we are going to use as indicator of specialization the gap between the industrial share in country’s i gross value added relative to the share that the specific industry has on the European level. While the national concentration will be symbolize the national share on the European Gross Value Added.

investments in human and physical capital, and the specialization of activities, after controlling for the level of past performance.

After rejecting the hypothesis of homoscedasticity (chi2(1) = 514.66,Prob>chi2 = 0.00), we would build a model where dummies for countries and industries will be introduced. After dropping one out of the 15 countries and one out of 18 industries, we estimate a robust regression and the parameter estimates are presented in the table below. Besides, the test for the null hypothesis that the dummies for industries are not differ systematically is rejected (F(14,3534) =12.53,Prob>F=0.00) and the same appears to be true for the national dummies (F(13, 3532) = 9.13, Prob>F=0.00). Therefore, at 5% of significant level we may argue some evidence that both industrial and national constraints play an important role in the determination of productivity level. Besides, the industrial dummies appear to have greater influence on productivity level than the national ones.

Table 6.1.1 Parameters Estimates

Independent Variable: Productivity Gap between Industry i. in country j. and the average European Productivity level for industry i.

Regressors Coefficient t-Statistic Pvalue>|t|

Gap 1: Human Capital :Number of Enrollments

Per Worker 0 .123 3.19 0.001

Gap 2:Physical Capital: Stock Per Worker _{-.0124 -5.65 0.000} Gap 3: Concentration: Industrial Share _{0.1280 3.74 0.000}

Constant _{-2.662 -25.48 0.000}

Past productivity level _{0 .554 19.89 0.000}

Results Provided by Stata Software, All the regressors are expressed in Logarithmic Values, 5% Level of significance

• The Results are based on a robust regression which includes dummies for countries and industries. *Hausman Test used suggested the significance of a fixed- against the random-effects model -chi2(4) =1012.71, Prob>chi2 = 0.0000 *Source: Groningen Growth and Development Centre, 60-Industry Database, October 2005, http://www.ggdc.net/

on the productivity gap. Besides the larger is the industrial share in country j the larger would be the industrial gap between country j and European average. Finally the past productivity level is positively related to the industrial productivity gap between country and the European average industrial productivity level; therefore the most productive industrial units are likely to hold and broaden the gap with the average industrial performance.

Table 6.1.2 Parameters Estimates

Independent Variable: Productivity Gap between County i. and the average European Productivity level

Regressors Coefficient t-Statistic Pvalue>|t|

Gap 1: Human Capital :Number of Enrollments

Per Worker 0.043 3.14 0.002

Gap 2:Physical Capital: Stock Per Worker _{-0.007 -2.88 0.004} Gap 3: Concentration: Industrial Share 0.190 3.18 0.002

Constant _{-0.607 -2.11 0.036}

Past productivity level 0.130 2.12 0.035

Results Provided by Stata Software, All the regressors are expressed in Logarithmic Values, 5% Level of significance

* The Results are based on a robust regression which includes dummies for

countries. Hausman Test suggested the significance of a fixed- against the random-effects model (Chi=76.40, Prob>chi2 = 0.0000)

*Groningen Growth and Development Centre, 60-Industry Database, October 2005, http://www.ggdc.net/

6.2 Regional Analysis

6.2.1 Industrial contribution on the regional growth path

The detailed analysis of each sector on a regional base is beyond the scope of this part of the analysis. Our main interest is the correlation between the high value added activities and their contribution on the regional growth path and on the divergence/convergence scenarios among regions. The Regional Database (provided by Eurostat) although includes a variety of different indicators with respect to the NACE industrial classification, it suffers from missing values and inconsistency in the observed time period for each country and industry. Hence, we run independent regressions for specific industries so as to test their relative importance on the regional growth pattern.

development during the period considered ( changes in productivity, employment and gross value added level). Agriculture/Forestry/Fishing & Mining/Quarrying are the first pair of industries with a similar trend within the period 1979-2003. At the end of the period, they ended up with a reduction in the number of total employees and a much higher change (in the opposite direction) of the productivity level. However, the final result is a large reduction of the sectoral value added. A second pair of industries with the similar pattern to the one described above is the pair of Chemicals/nuclear fuel/Oil & Electricity/ Gas/Water Supply. These four sectors present similarities taking into consideration two elements of industrial contribution to gross value added (productivity and employment) and the resulted change in Industrial Value Added. Additionally, all the four industries are industries that may have some additional similarities as i.e. the dependence with the natural area of production and therefore different sensitivity in the transport costs and the market proximity/accessibility.

A third pair of industries with similar behaviour is the pair of Communication/Telecommunication & Machinery/related Equipment/Instruments, which are two core industries with respect to the productivity improvement. Both Industries expanded their productivity performance while they experienced at the same time a reduction (in a much lower rate) in their employment level, which final led to the increase of their relative importance on the European Average Performance. Another interesting industry is the industry signed here as Computers/related Equipment/Instruments which is the only industry which rose their importance in the GVA performance through the expansion of employment rate but without suffering at the same time from a fall in its productivity score.Services (Wholesales/Retail Sales, Finance/Insurance/Real estate) also present a similar growth path, labeled by a relative steady share in GVA, linked with a rise in employment and productivity level.

industries on regional wealth. Expenditures in defence and public Administration may be subject to a different bundle of determinants which can lie on the political, geographical, historical and cultural context rather than presenting the industrial response to the given economical constraints.

Therefore, we will test the contribution of Chemicals/Nuclear Fuel/Oil, Mining and Quarrying, Computer/related activities and Machinery/related Equipment/instruments, Computer/related activities on a regional level. Our objective is to pinpoint the positive or negative effect of an additional investment in industry (total investments per employ) as well as the concentration of units and therefore the concentration of industrial activities in the region. We measure the concentration as the share of industrial units established in the specific region to the total number of units (established on national level). However, we should underline that a possible limitation is that we don’t control for the size of local units. As much as the size matters in the industrial productivity, we will use the gap of industrial productivity relative to European average productivity for the specific industry so as to tackle or at least limit the downfall due to the above problem. Another control variable will be the concentration of total industrial units in the region, where there is no problems due to multicollinearity with the determinants under consideration. At this point, it is important also to mention that total concentration is strongly correlated (above 99%) with the concentration of population and therefore we can not include both the parameters as a control variable in our model. Finally another strong limitation, mainly driven by the missing values in our database is that we are going to pick up specific sub categories of our industries instead of the average concentration and investments made in the specific region and industry. However, there is also an advantage in that specification as it could also be misleaded to summarize as homogeneous, units of different activities. Considering –in principle- that the sub categories of one industry are characterized with the same average productivity score level, finally the above limitation may be bias less our model than if we include all the heterogeneous units as an homogeneous cluster.

Therefore our main function is represented as

(Growth In Per Capita income)=β1*log(Lagged Per Capita Income) + β2*log

(Concentration of industrial Units) + β3*log (Investments Per Employ)+β4*log(Public

Additionally, we are going to control for the concentration of total units and the sectoral productivity gap as well as the national productivity gap with the european average where the correlation test provide evidence of low correlation rate among regressors and growth rate.

6.2.2. Regional Analysis: by Industry

i. Chemicals, Nuclear Fuel, Oil:

The first industry on which we turn our focus on a regional-NUTS 2 level is industry No 15 (Chemicals, Nuclear Fuel, Oil). The first question that we seek to answer is whether Chemicals/Nuclear Fuel/Oil has a significant effect on the determination of the regional income and in which way. Regions where industrial units are concentrated are more or less likely to perform best? Secondly, investments made on a national or regional level enhance or not the regional growth path? In other words, the concentration of units and the investments in the specific industry have a positive influence on the regional performance or not? The last question that we seek to answer in that part is whether or not there is systematical difference in the key growth determinants among high and low score performers.

The first test so as to choose our control variables is the correlation test. The test shows strong correlation among national gap between total productivity and industrial productivity as well as among them with the regional growth rate. The correlation between industrial concentration and total business concentration exceeds the 90% denoting the establishments of concentrated units where there is concentration of activities in general.