Regional Competitiveness Indicators for Europe - Audit, Database Construction and Analysis

Transcription

1 Regional Studies Association International Conference Pisa, April, 2003 Regional Competitiveness Indicators for Europe - Audit, Database Construction and Analysis Ben Gardiner Cambridge Econometrics Abstract This paper reports on the construction of a comprehensive set of regional competitiveness indicators for Europe, including the European Union and candidate countries. Various sources have been investigated, included Regio and other Eurostat databases, the OECD Territorial Database, National databases, and data from private bodies, eg trade organisations. The paper covers the main tasks concerned with the construction of the database, and subsequent analysis that has taken place. Firstly, the contents of the various sources of regional data are reviewed and catalogued, with a check on availability, quantity and quality of data. Secondly, the data are classified by separating the various indicators into variables which are essentially measures of the outcome of competitiveness (output indicators, eg GDP per capita) and those which contribute towards this success (inputs). The input indicators are categorised into broad themes, covering basic infrastructure and accessibility, human capital, institutional features, and other factors, eg R&D, innovation demography, property, the information society. The construction of the database is the next major task to be reported, in particular the methods used to construct a system of regional accounts where both breadth (geographical coverage, time period) and depth (indicator coverage) extend upon existing sources. To conclude, some findings from analysis of the data are reported, looking at long-term trends within and across the countries covered. The results from this work have potentially profound implications for the way European regional data are analysed, as the resulting database opens the door for more in-depth analysis of regional competitiveness than has previously been possible. The paper has been prepared as part of projects for the European Commission (DG Regio and DG Enterprise) looking into factors affecting regional competitiveness, which in turn will feed into the 2003 Cohesion Report. The results presented here are with the permission of the Directorate General for Regional Policy. Contacts: Cambridge Econometrics Covent Garden, Cambridge CB1 2HS, United Kingdom. Tel: , Fax: bg@camecon.com Website:

2 Contents Page 1 Background Information Background to the study Remaining contents of the paper 3 2 Aspects of Regional Competitiveness Defining regional competitiveness Key influences from theory 4 3 Data Audit and Construction Data sources and availability European regional databank Classification of indicators 10 4 Data Analysis Decomposition analysis Associations and correlations 15 5 Conclusions Main findings Ongoing research Acknowledgements 20 6 References 21 2

3 1 Background Information 1.1 Background to the study The paper draws on the results of two projects which Cambridge Econometrics is coordinating for the DG Regio 1 and DG Enterprise 2 of the European Commission. The broad purpose of both studies is to identify and measure regional success, where success is defined as competitiveness. To measure success a set of indicators is required, which involves the construction of a European Regional database using sources from Eurostat and the various National Statistical Offices. Measuring success and attributing association and causation has required a range of techniques to be applied to the data, including decomposition, correlation and regression. 1.2 Remaining contents of the paper Section 2 looks at the issues involved with defining competitiveness, and what this means in a regional context. Some theoretical perspectives are used to supply possible indicators which may prove useful in understanding what drives competitiveness, and to provide a structure within which the empirical analysis can be placed. The audit of the regional data and its availability is the focus of Section 3, which concludes with a set of indicators to be used in the analysis stages in Section 4. The analysis includes: analysis of the components of competitiveness; decomposition of productivity trends by 15 sectors; correlation of productivity and competitiveness indicators over regions and time; proposals for Barro regression analysis. The final section summarises findings and reports the future research strategy. 1 A project entitled A Study on the Factors of Regional Competitiveness in collaboration with ECORYS-NEI (Rotterdam), Prof Ron Martin (Cambridge University), Wifo (Vienna), and CEET (Madrid), and due to be part of the 2003 Cohesion Report. 2 2 A project entitled Regional Aspects of Competitiveness in collaboration with ECORYS-NEI and the Centre for Economic and Managerial Research (Florence), which is to form a chapter in the 2003 Competitiveness Report. 3

4 2 Aspects of Regional Competitiveness 2.1 Defining regional competitiveness Competitiveness can be defined in many ways. The one used here was adopted by the European Commission (1999): [Competitiveness is defined as] the ability to produce goods and services which meet the test of international markets, while at the same time maintaining high and sustainable levels of income or, more generally, the ability of (regions) to generate, while being exposed to external competition, relatively high income and employment levels. and "In other words, for a region to be competitive, it is important to ensure both quality and quantity of jobs." A common representation of competitiveness which broadly fits this description is GDP per capita, which can be broken down into various component factors, each with an economic interpretation. The breakdown is not entirely independent either - some interrelation is likely between the indicators, eg highly productive regions using skilled labour may well display high rates of employment. GDP Population = Total GDP Hours Worked Total Hours Worked * Employment Employment * Working Age Population Working Age Population * Population Productivity Work-Leisure Employment Rate Dependency Rate There are a number of arguments about the use of GDP per employee or GDP per hour worked as a measure of productivity. The regional implications of choosing an hours worked measure are perhaps more profound than at the national level. Regions are more likely to be specialised in particular sectoral activities, eg agriculture, which means that adjusting for different hours worked profiles will more accurately represent the true labour effort involved in producing the output against which it is measured. A final issue to mention is that of sustainability. This involves the amount of resources, eg energy and emissions, used in enabling a region or country to become competitive. While important for measuring the long-term potential for high productivity growth, this is a dimension seen as outside of the main study area and so is not considered here. 2.2 Key influences from theory As regional competitiveness shares many features in common with its national counterpart, the theories involved are usually present in both areas of work. 4

5 Expounding these theories is not the purpose of this work, however. Instead, Figure 1 provides an overview. Figure 1: Theoretical Aspects of Regional Competitiveness From such an overview, a list of indicators can be established for many of the factors involved in the process of both national and regional competitiveness. In the table below they are split across three broad themes. Table 1: Factors of Regional and National Competitiveness Infrastructure & accessibility Basic Infrastructure - road - rail - air - property Technological Infrastructure - ICT - telecoms - internet Knowledge infrastructure - educational facilities Quality of Place - housing Human resources Demographic trends - migration of skilled workers - diversity High skilled workforce - knowledge-intensive skills Productive environment Entrepreneurial Culture - low barriers to entry - risk taking culture Sectoral Concentrations - balance / dependency - employment concentration - high value-added activities Internationalisation - exports/global sales - investment - business culture - nature of FDI Innovation - patents 5

6 Infrastructure & accessibility - natural surroundings - cultural amenities - safety Human resources Productive environment - R&D levels - research institutes and universities - linkages between companies and research - spillover effects (untraded interdependencies) Governance and institutional capacity Capital availability Specialisation Nature of competition Among the more interesting indicators are those relating to knowledge and innovation. When grouping these variables it is useful to have a structure into which these indicators can be placed. A study on benchmarking the regional knowledge economy of the South East (Huggins Associates, 2001) made some distinctions which can be used in this way. The study used three themes of knowledge creation, utilisation and capacity building, specified here in four components with an associated list of variable indicators. (a) Knowledge Capital - the region s capacity for creating new ideas. (b) Innovation Capacity the region s capacity for transforming such new ideas and creating commercial value through managerial, professional, and technical personnel within firms. (c) Knowledge Economy Outputs - the results in the form of knowledge-based goods and services that contain high value added outputs as measured by the number of patents and innovations. (d) Knowledge Economy Outcomes the wealth created as measured by indicators. Table 3 links some of the earlier-identified indicators into these new categories. Table 3: Indicators and the Knowledge-Innovation Process Knowledge Capital Innovation Capacity Knowledge Economy Outputs Working population, including age structure Education levels of working population Personnel employed in managerial, professional and technical occupations Personnel employed in R&D Personnel employed in Higher Education (HE) establishments Employment in knowledge-based sectors New business registration rates Number of HQs Number of businesses in knowledge-based sectors R&D expenditures in Government R&D expenditures in Business R&D expenditures in HE establishments Amount of venture capital Number of patents registered GDP Labour productivity Value of exports Knowledge Economy Outcomes GDP per capita. Weekly (monthly or annual) earnings Unemployment rates Household incomes 6

7 The importance of this classification is that one would not then choose one output measure to explain another, ie patents to explain productivity. Instead, one would choose indicators of knowledge capital and/or indicators of innovation capacity. 7

8 3 Data Audit and Construction 3.1 Data sources and availability Regio database The Regio domain is contained within Theme 1 of the NEWCRONOS database, and is the main location for regional data in Eurostat. The database reviewed here was obtained at the end of January The Regio domain contains a wide array of indicators, many which are not directly relevant to studying regional competitiveness. Listed below are the categories identified as most relevant to knowledge-based activities as identified in the review of literature. Economic accounts Relevant indicators include GDP, Gross Value Added (GVA), and employment. In particular, GVA and employment are available with a sectoral disaggregation, which can be used to help identify how sectoral productivity influences overall regional performance, as well as issues of specialisation. The disaggregation shown in Table 4 is close to the maximum available number of sectors in the Regio accounts, and also allows a blending of the data between the new (ESA95) and old (ESA79) accounting structure that is present in the database. Table 4: Sectoral Links Between NACE-CLIO AND ESA95 Sector Heading ESA95 Sections ESA79 Codes Agriculture, hunting, forestry and fishing A + B B01 Energy and Manufacturing C + D + E Mining and quarrying + Electricity, gas and water supply C + E B06 Manufacture of food products, beverages and tobacco DA B36 Manufacture of textiles and textile products + Manufacture of leather and leather products Manufacture of coke, refined petroleum products and nuclear fuel + Manufacture of chemicals, chemical products and man-made fibres + Manufacture of rubber and plastic products DB + DC B42 DF + DG + DH B13 + B17 + B50 (part) Manufacture of electrical and optical equipment DL B24 (part) Manufacture of transport equipment DM B28 Other Manufacturing (Manufacture of wood and wood products + Manufacture of pulp, paper and paper products; publishing and printing + Manufacture of other non-metallic mineral products + Manufacture of basic metals and fabricated metal products + Manufacture of machinery and equipment n.e.c. + Manufacturing n.e.c.) DD + DE + DN + DI + DJ + DK B15 + B24 (part) + B47 + B50 (part) 8

9 Sector Heading ESA95 Sections ESA79 Codes Construction F B53 Market Services G + H + I + J + K B 68 Wholesale and retail trade; repair of motor vehicles, motorcycles and personal and household goods G B58 (part) Hotels and restaurants H B58 (part) Transport, storage and communication I B60 Financial intermediation J B69 Real estate, renting and business activities K B74 Non-market services (Public administration and defence; compulsory social security + Education + Health and social work + Other community, social and personal service activities + Private households with employed persons + Extra-territorial organisations and bodies) L + M + N + O + P B86 Education The data come from two surveys, one in 1979 and the other in 1997, and include information on pupils and students by educational level, gender and age. Science and technology Indicators include R&D expenditure and personnel by institutional sector (eg government, business, etc), employment in high-tech sectors and those in the area of science and technology, and European patent applications (including a high-tech subset) Cambridge Econometrics European regional database Cambridge Econometrics maintains a European regional database which has been used to support the construction of the following indicators: GDP (m PPS and euro 1995m) GVA by 15 sectors (m PPS and euro 1995m) Employment by 15 sectors (000s) Hours worked (average hours per week) Employee compensation (m euro) Investment by five sectors (m PPS and euro 1995m) Population by various age groups (000s). These data are typically more complete in regional coverage than the original Eurostat data, and cover the period The long time series coverage is useful in the analysis of developments in competitiveness over time, which can then be linked to the various causes proposed in the review of literature. 9

10 3.2 European regional databank Table 5 provides a description of the information available to undertake subsequent data analysis and empirical work. Table 5: European Regional Database Coverage Indicator Units Years Disaggregation Gross Domestic Product m PPS and euro 1995m n/a (GDP) Gross Value Added (GVA) m PPS and euro 1995m sectors (ESA95) Employment 000s sectors (ESA95) Hours worked average number per total week Population 000s selected age cohorts within working age population Employment in high-tech areas 000s total and 3 sub-sectors (manufacturing, services and knowledge-intensive services) Human resources in science 000s total and high-tech and technology (HRST) Number of students 000s sporadic coverage total and those in tertiary education Investment m PPS and euro 1995m sectors (ESA95) Employee compensation Euros sectors (ESA95) Research & Development expenditure m PPS and euro 1995m and 1997 as most total and across three institutional sectors complete Research & Development personnel head count and 1995 as most total and across three institutional sectors complete Patent applications to EPO number of patents total and high-tech reasonable coverage across most Member States 3.3 Classification of indicators From the allocation and identification of indicators made in the literature review, the available indicators can be grouped as follows: Knowledge Economy Outputs/Outcomes. Alternative measures that could substitute for measures of competitive success are employee compensation, patent applications, or indeed the broader measure of GDP per capita. Innovation Capacity 10

11 The main available indicator to represent innovation capacity is R&D expenditure. Although this indicator can in principal be split into three sectors (Government, Business, and HE establishments) in practice the very sparse nature of the data mean that total R&D expenditure is the better choice to ensure reasonable coverage across space and time. Knowledge Capital To represent the stock of regional knowledge, a few more variables are available. These are, working population (by age structure if necessary), personnel employed in R&D, employment in high-tech sectors, total number of students and those involved in tertiary education. Additional influences Aside from the knowledge-based factors and their groupings, the literature suggests a number of other influences on competitiveness which need to be borne in mind when undertaking the empirical analysis. 1 Sectoral structure There are several references to the importance of assessing industrial structure in the analysis of regional performance. In the European Commission s Sixth Periodic Report (1999), for example: "An unfavourable sectoral structure together with a lack of innovative capacity seems to be among the most important factors underlying lagging competitiveness..." (p9) 2 Investment The size of a region s capital stock and the amount it invests in maintaining this capital stock are fundamental influences on the ability to produce more output per unit of labour, however this is measured. Again, there will be relations between investment and some of the other indicators described so far, such as market structure and R&D expenditure. 3 Spillover effects Spillover effects are seen as important factor when explaining regional productivity. The degree of spatial association can be measured using Moran s I Statistic (Moran, 1948) which is defined as follows: 11

12 I N N i j = N wij ( xi x)( x j x) 2 ( x x) i= 1 i where x i and x j are observations for region i and j of the variable being analysed, ie productivity. The key choice to be made in calculating the statistic is the construction of the weighting matrix, W, from which the individual elements, w ij, are chosen. One way to construct W is the binary contiguity matrix, whereby w ij = 1 if regions i and j share a border, and is zero otherwise. The method chosen for this paper is slightly more sophisticated and uses the reciprocal of the distance between the centre of each region, so that the greater the distance between regions i and j, the lower the value of w ij. 12

13 4 Data Analysis 4.1 Decomposition analysis GDP per capita Table 6 looks at the average levels of the components of GDP per capita for three bands of regional GDP per capita to assess variation across differing levels of competitiveness. The components showing most consistency with GDP per capita are productivity and the employment rate, while the dependency rate shows little variation across the GDP per capita bands. The hours-worked component also shows little variation, but what exists indicates a negative association with GDP per capita, ie the regions with lower GDP per capita have a higher than average level of working hours. TABLE 6: GDP COMPONENT ANALYSIS Number of regions GDP per capita Productivity Work vs Leisure Employment Rate < 75% EU average > 75% and below EU average > EU average Dependency Rate The link with these correlations is the sectoral structure of production, whereby regions which specialise in agriculture have a higher number of weekly hours worked, and due to the low productivity associated with such activities this leads to a low level of GDP per capita. This argument demonstrates another interdependence between the various components Productivity Productivity can be decomposed into sectoral components using the expression below. GVA total Employment total = 15 i= 1 GVA Employment i i Employemt * Employment i total The only complication is that, because hours worked is not available by sector by region, productivity must be defined as GVA per employee rather than GDP per hour worked. 13

14 Analysis of the results reveals that heterogeneity is a problem in assessing the contribution to productivity because the shares of sectors such as other market services, construction, and non-market services are relatively large. Chart 1 demonstrates this by looking at the sectoral contribution to productivity for the top-ten productivity regions in It is clearly seen that non-market services (L P) and other market services (K) dominate the contribution, although for Luxembourg (lu) and Brussels (be1) the relatively high proportion accounted for by financial intermediation is clearly evident. CHART 1: SECTORAL PRODUCTIVITY DECOMPOSITION, 2000 dk01 lu at32 at34 at13 be21 fr1 be1 be31 be GDP per hour w orked (PPS, 2000) A+B C+E DA DB+DC DF+DG+DH DL DM DD+DE+DN+ DI+DJ+DK F G H I J K L - P Another issue that can be investigated is that of traded versus non-traded sectors. Much of the argument supporting productivity convergence has its origins in trade and subsequent factor price equalisation and or/spillover effects. It has already been established that the services sectors, with their high share of regional employment account for the main proportion of the productivity level. One might expect that the traded sectors would display a more even level of productivity across the regions given that they are more open to competition, while less or un-traded sectors should display more regional variation. Chart 2 shows the difference between these groups, where the traded sectors are agriculture and manufacturing, while the remaining sectors are grouped as non-traded. The energy and mining sector is excluded as the productivity results from this activity are rather volatile. 14

15 CHART 2: TRADED AND UNTRADED SECTOR PRODUCTIVITY Traded (1995 euro) y = x Regions ordered by total productivity (2000) Untraded (1995 euro) y = x Regions ordered by total productivity (2000) The results show firstly that productivity in the non-traded sector is less volatile across regions than in the traded sectors and secondly that there is more of an association between productivity in the traded sectors and the overall productivity ranking across regions, ie there is a larger gap in productivity in the traded sectors than in the untraded sectors when looking at the overall regional productivity ranking. 4.2 Associations and correlations Input indicators The preceding sections identified a number of key indicators which could be used as explanatory factors of competitiveness, and for which data were available at the NUTS- 2 level. Tables 7 and 8 show the average level of these indicators for the two components of GDP per capita showing most variation, ie productivity and the employment rate. 15

16 < 75 % EU TABLE 7: INPUT INDICATORS AND PRODUCTIVITY Patents per Capita R&D Intensity (%) Tertiary students per capita (%) Hi-tech employment (%) Investment- Output Ratio (%) average > 75 % EU average and < EU average > EU average < 75 % EU TABLE 8: INPUT INDICATORS AND THE EMPLOYMENT RATE Patents per Capita R&D Intensity (%) Tertiary students per capita (%) Hi-tech employment (%) Investment- Output Ratio (%) average > 75 % EU average and < EU average > EU average Population Density Population Density The results for productivity show higher values for the indicators for the three groupings of GDP per capita. For the employment rate the picture is less clear, with only half of the indicators (patents, R&D intensity, and hi-tech employment specialisation) showing a consistent pattern. It is of course possible that some of the input indicators are correlated over time, which could give cause for concern when attempting to attribute explanatory power in the econometric estimations. Table 9 presents the correlation matrix for the various input indicators, using selected time periods to obtain a reasonable coverage across the regions. TABLE 9: INPUT INDICATOR CORRELATIONS R&D Intensity Tertiary Students High-Tech Employment Investment Ratio R&D Intensity Tertiary Students High-Tech Employment Investment Ratio Population Density Population Density 16

17 With the exception of investment, for which poor data quality may explain some of the results, most indicators show a positive correlation, albeit not particularly strong. As well as looking at a single cross section, it would also be useful to assess correlation of growth rates, as many of the variables are likely to be trended however, the small sample sizes currently available do not permit this analysis to take place Productivity correlations The indicators chosen as inputs to the competitive framework are expected to largely operate through productivity. Chart 3 shows the correlation of productivity its level at the start of the period. The expected negative correlation does exist, while the degree of variation implies there are a multitude of other factors (mentioned in the literature section) which can help to improve the explanation. CHART 3: PRODUCTIVITY LEVEL AND GROWTH 6 Productivity growth ( , %pa) Correlation Coefficient = Productivity Level in 1987 (GDP per hour worked, 000s, PPS) Chart 4 looks at interdependence among regions through the concept of spillovers, a topic which is central to new economic geography as a mechanism through which productivity gains are dispersed. The weighting mechanism is the inverse of the physical distance between regions, ie the further regions are apart, the smaller the weight attached to them. The results show a reasonably high degree of correlation, indicating that the potential for spillover effects should be assessed in further empirical work. 17

18 CHART 4: PRODUCTIVITY SPILLOVERS 1.2 Productivity spillover effect Correlation Coefficient = fi Productivity Level (PPS), 2000 at13 se01 lu fr1 dk01 18

19 5 Conclusions 5.1 Main findings This paper reports an ongoing investigation into regional competitiveness, defined as GDP per capita: its measurement and its the causes, subject to data availability. Economic theory suggests several factors that distinguish the concepts of regional and national competitiveness. Foremost amongst these are the issues of localisation and specialisation that come from the field of New Economic Geography. A number of studies (eg Audretsch and Feldman, 1998) link spillover effects, in particular knowledge spillovers, with productivity gains providing a bridge to the ideas of new growth theory where such effects are regarded as important sources of externalities. Recent research on knowledge and innovation (Huggins, 2001), has proposed a process into which various indicators fit whereby categorisations such as knowledge capital, innovation capacity, and knowledge economy outputs and outcomes help to allocate indicators, while recognising that many of the measures will be related/correlated to one another. A review of data availability suggested a limited number of indicators available, both across regions and across time, with which to undertake empirical analysis. However, sufficient indicators were available to measure productivity by 15 sectors across the NUTS-2 regions over , and to identify a selection of knowledge-based proxies such as R&D expenditure, and students enrolled in higher education. Analysis of the components of GDP per capita revealed a strong association between average levels of productivity and the employment rate with different groupings of GDP per capita, but little relationship with hours worked or the dependency rate. Decomposition of the regional productivity data showed that the majority of successful regions were supported by large contributions from market services, in particular other market services and financial intermediation heterogeneity of employment share caused interpretation problems, however. Splitting sectors into traded and untraded revealed that productivity levels are more volatile across regions in their traded sectors, and that productivity displayed a flatter regional profile across untraded sectors. Correlation analysis of productivity and the available indicators again suggested strong links with productivity, but less pattern for the employment rate. Most input indicators 19

20 show a positive bi-variate correlation with each other, suggesting that they are often explaining the same phenomena, while a negative relationship between productivity growth and its starting level, providing support for (unconditional) regional convergence. Investigation for spillover effects revealed that they were also likely to be present, which will affect the structure of the forthcoming econometric modelling. 5.2 Ongoing research The progress of the work has been hampered by the difficulty in obtaining the regional data to supplement the Eurostat Regio database, but the aim is to complete this as quickly as possible for as many indicators as possible. A thorough audit has taken place across the Member States and Candidate Countries, and the purchasing of data is now occurring. The establishment of such a database will help to improve understanding of the causes of regional competitiveness in Europe, which has for a long time been limited by incomplete time series and missing variables. The work will shortly progress from the data analysis stage to work on econometric models using Barro regressions and growth accounting methods to establish a more robust framework, followed by case studies of successful regions to assess the importance of more qualitative factors, eg regional governance, and the ability to transfer the factors driving success to those areas which are currently less competitive. 5.3 Acknowledgements Thanks are due to the collaborators on both DG Regio and DG Enterprise projects, and the peer reviewer on the DG Enterprise report, Dr Gerhard Untiedt of GEFRA, Germany. Also Matthew Brooke of the European Commission (DG Regio) provided additional constructive comments and gave permission to present what is essentially work in progress. Responsibility for the contents of this paper lies with the author and the views therein do not necessarily reflect those of the European Commission. 20

21 6 References Audretsch, D. B. and Feldman, M. P., R&D Spillovers and the Geography of Innovation and Production. Industrial policy and competitive advantage, in Audretsch, D B., Copy-editing: Elgar Reference Collection. International Library of Critical Writings in Economics, vol. 84. Cheltenham, U.K. and Northampton, Mass., Elgar; distributed by American International Distribution Corporation Williston Vt, European Commission, European Competitiveness Report, European Commission, Sixth Periodic Report on the Social and Economic Situation of Regions in the EU, Huggins Associates, Global Index of Regional Knowledge Economies: Benchmarking South East England, Final Report, prepared for The South East England Development Agency (SEEDA), November Krugman, P., Development, Geography and Economic Theory, MIT, Cambridge (Mass.), Krugman, P., Increasing Returns and Economic Geography, Journal of Political Economy, 1991, Vol. 99, pp Porter, M., The Competitive Advantage of Nations, Free Press, New York, Seyit, Kose and Moomaw, R, Knowledge spillovers and regional growth in Europe, paper presented at the ERSA congress, Dortmund Simmie, J., Sennett, J., Wood, P. and Hart, D., Innovation in Europe: A Tale of Networks, Knowledge and Trade in Five Cities, Regional Studies, 2002, Vol. 36, pp Tondl, G., The Changing Pattern of Regional Convergence in Europe, Jahrbuch für Regionalwissenschaft, 1999, Vol. 19, No. 1, pp Tondl, G., Convergence after Divergence? Regional Growth in Europe, Springer, Vienna-New York, Tondl, G., What Determined the Uneven Growth of Europe s Southern Regions? An Empirical Study with Panel Data, Working Paper des Forschungsinstituts für Europafragen, April 1999, No. 30, 49 pp. 21