Labour market effects of European integration in the Bavarian and Czech border regions

Moritz, Michael Book Labour market effects of European integration in the Bavarian and Czech border regions IAB-Bibliothek, No. 321 Provided in Cooperation with: Institute for Employment Research (IAB) Suggested Citation: Moritz, Michael (2009) : Labour market effects of European integration in the Bavarian and Czech border regions, IAB-Bibliothek, No. 321, ISBN 978-3-7639-4012-7, W. Bertelsmann Verlag (wbv), Bielefeld, https://doi.org/10.3278/300682w This Version is available at: https://hdl.handle.net/10419/280174 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-sa/4.0/ Labour Market Effects of European Integration in the Bavarian and Czech Border Regions Michael Moritz 321 gewidmet meiner Familie 321 Labour Market Effects of European Integration in the Bavarian and Czech Border Regions Michael Moritz Herausgeber der Reihe IAB-Bibliothek: Institut für Arbeitsmarkt- und Berufsforschung der Bundes agentur für Arbeit (IAB), Regensburger Straße 104, 90478 Nürnberg, Telefon (09 11) 179-0 R e d a k t i o n : Martina Dorsch, Institut für Arbeitsmarkt- und Berufsforschung der Bundesagentur für Arbeit, 90327 Nürnberg, Telefon (09 11) 179-32 06, E-Mail: [email protected] Gesamtherstellung: W. Bertelsmann Verlag, Bielefeld (www.wbv.de) R e c h t e : Kein Teil dieses Werkes darf ohne vorherige Genehmigung des IAB in irgendeiner Form (unter Verwendung elek tronischer Systeme oder als Ausdruck, Fotokopie oder Nutzung eines anderen Vervielfältigungsverfahrens) über den persönlichen Gebrauch hinaus verarbeitet oder verbreitet werden. © 2009 Institut für Arbeitsmarkt- und Berufsforschung, Nürnberg/ W. Bertelsmann Verlag GmbH & Co. KG, Bielefeld In der „IAB-Bibliothek“ werden umfangreiche Einzelarbeiten aus dem IAB oder im Auftrag des IAB oder der BA durchgeführte Untersuchungen veröffentlicht. Beiträge, die mit dem Namen des Verfassers gekenn zeichnet sind, geben nicht unbedingt die Meinung des IAB bzw. der Bundesagentur für Arbeit wieder. ISBN 978-3-7639-4013-4 (Print) ISBN 978-3-7639-4012-7 (E-Book) ISSN 1865-4096 Best.-Nr. 300682 www.iabshop.de www.iab.de Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http://dnb.ddb.de abrufbar. Dissertation der Wirtschaftswissenschaftlichen Fakultät der Universität Regensburg Berichterstatter: Prof. Dr. Dr. h. c. Joachim Möller Prof. Dr. Uwe Blien Tag der Disputation: 16. 02. 2009 Dieses E-Book ist auf dem Grünen Weg Open Access erschienen. Es ist lizenziert unter der CC-BY-SA-Lizenz. 3 IAB-Bibliothek 321 Contents Vorwort und Danksagung ......................................................................... 7 1 Introduction: Analysis of Labour Markets in Border Regions ... 9 1.1 Motivation ................................................................................................. 9 1.2 Research Questions ................................................................................. 11 1.3 The Approach and the Structure of the Thesis ................................ 13 Appendix to Chapter 1 ............................................................................................. 15 2 A Survey of the Literature ........................................................ 21 2.1 Theoretical Background: Integration Effects in Border Regions .... 21 2.1.1 Traditional Location Theory ................................................................... 21 2.1.2 Trade Theory .............................................................................................. 22 2.1.2.1 Traditional Trade Theory ......................................................................... 22 2.1.2.2 New Trade Theory .................................................................................... 23 2.1.2.3 Trade in Intermediate Goods: Model by Feenstra/Hanson ........... 24 Excursus: The Causes of Labour Demand Shifts – Skill-Biased Technological Change or International Trade?.................. .............. 24 2.1.3 New Economic Geography .................................................................... 33 2.1.3.1 The Basic Approach by Krugman ......................................................... 33 2.1.3.2 A Three-Region Model by Brülhart/Crozet/Koenig ......................... 36 2.1.4 Summary: Hypotheses for the Empirical Analysis .......................... 41 2.2 Empirical Studies on Integration Effects ........................................... 43 2.2.1 Studies on the U.S.-Mexican Integration Process .......................... 44 2.2.2 Studies on the European Integration Process .................................. 45 2.2.3 Summary..................................................................................................... 49 Appendix to Chapter 2 ............................................................................................. 50 3 Labour Market Effects in the Bavarian Border Region ........... 51 3.1 Introduction ............................................................................................... 51 3.2 Data and Basic Definitions .................................................................... 52 3.3 The Labour Market in Eastern Bavaria: Some Descriptive Evidence ...................................................................................................... 55 3.3.1 Selected Figures of VALA ....................................................................... 55 IAB-Bibliothek 321 4 Contents 3.3.2 Structural Change and Specialisation ................................................ 56 3.3.3 Relative Employment Share and Czech Employees in Eastern Bavaria .................................................................................... 62 3.3.4 Skill Structure of Employed People ..................................................... 65 3.3.5 Skill Structure of Unemployed People ................................................ 67 3.4 Econometric Analysis of Qualification Trends ................................. 69 3.4.1 Employed People ...................................................................................... 69 3.4.2 Unemployed People ................................................................................. 72 3.5 Econometric Analysis of Wage Differentials .................................... 73 3.5.1 Estimating Wage Differentials ............................................................. 73 3.5.2 Estimation Methods and Results ......................................................... 74 3.5.2.1 Tobit Regressions of Cross Sections (IABS Regional Scientific Use File) ...................................................... 74 3.5.2.2 Sensitivity Analyses: Weakly Anonymous IABS Version and BeH Extract ................................................................................................ 78 3.5.2.3 Propensity Score Matching ................................................................... 84 3.5.2.4 Pooling Cross Sections Over Time ....................................................... 88 3.5.2.5 Panel: Fixed Effects ................................................................................. 91 3.5.2.6 Additional Analysis: Splitting the Border Region ........................... 94 3.6 Conclusion.................................................................................................. 95 Appendix to Chapter 3 ............................................................................................. 98 4 Labour Market Effects in the Czech Border Region ................ 103 4.1 Introduction ............................................................................................... 103 4.2 Data and Basic Definitions .................................................................... 105 4.3 The Labour Market in the Czech Republic: Some Descriptive Evidence ...................................................................................................... 109 4.3.1 Relative Employment Share and Structural Change ...................... 109 4.3.2 Skill Structure of Employed People ..................................................... 114 4.3.3 Skill Structure of Unemployed People ................................................ 116 4.3.4 Wage Differentials between Border and Non-Border Region ..... 118 4.4 Econometric Analysis of Qualification Trends ................................. 122 4.4.1 Employed People ...................................................................................... 122 4.4.2 Unemployed People ................................................................................. 124 5 IAB-Bibliothek 321 Contents 4.5 Econometric Analysis of Wage Differentials .................................... 125 4.5.1 Standard OLS Regressions without Pooling Cross Sections ......... 125 4.5.2 Pooling Cross Sections Over Time ....................................................... 129 4.6 Conclusion.................................................................................................. 133 Appendix to Chapter 4 ............................................................................................. 135 5 Summary and Outlook ............................................................... 139 List of Figures ............................................................................................ 145 List of Tables.............................................................................................. 149 List of Abbreviations ................................................................................. 151 References .................................................................................................. 153 Kurzfassung ............................................................................................... 163 Summary .................................................................................................... 165 7 IAB-Bibliothek 321 Vorwort und Danksagung Das bayerisch-böhmische Grenzgebiet ist eine Region, die über Jahrhunderte hinweg vom enormen Austausch in kultureller, politischer und wirtschaftlicher Hinsicht geprägt wurde. Über vierzig Jahre verhinderten die Folgen des Zweiten Weltkrieges und die künstliche Barriere an der Demarkationslinie zweier politischer Systeme eine Fortsetzung des gemeinsamen Weges in der Geschichte. Als „Kind dieser Grenzregion“ konnte ich hautnah miterleben, welche Auswirkungen das Hemmnis der Grenze auf das gesellschaftliche und wirtschaftliche Leben in der nördlichen Oberpfalz hatte. Die damalige Tschechoslowakei war in meiner Jugendzeit ein unbekanntes Land. Nachdem ich nach dem Fall des Eisernen Vorhangs die Gelegenheit hatte, auch in Westböhmen beruflich tätig zu sein, war dies – neben dem allgemeinen Interesse an der Ökonomie der Arbeitsmärkte – Motivation genug für mich, diese Dissertationsschrift auszuarbeiten. Für die Ermöglichung, dieses Thema wissenschaftlich abzuhandeln, möchte ich mich zunächst bei meinen Betreuern, Prof. Dr. Dr. h. c. Joachim Möller und Prof. Dr. Uwe Blien, bedanken. Beide sind dem gewählten Thema von Anfang an mit hohem Interesse und Aufgeschlossenheit gegenübergestanden und haben mir stets das Gefühl vermittelt, mich mit einem wichtigen Forschungsgegenstand zu beschäftigen. Besonders betonen möchte ich weiterhin die Unterstützung durch das Institut für Arbeitsmarkt- und Berufsforschung der Bundesagentur für Arbeit (IAB) in Nürnberg. Die Bewerkstelligung der vorliegenden Dissertation wurde im Rahmen des IAB/WiSo- Graduiertenprogramms (GradAB) mit einem Stipendium gefördert. Großer Dank gebührt in erster Linie auch den Kolleginnen und Kollegen am IAB, insbesondere jenen im Forschungsbereich „Regionale Arbeitsmärkte“. Durch die enge Zusammenarbeit vor Ort konnte ich permanent auf kompetente und wertvolle Ratschläge zurückgreifen. Gleiches gilt für die Stipendiatinnen und Stipendiaten im GradAB. Außerordentlich möchte ich dabei Herrn Dr. Roman Lutz und Herrn Dr. Gerhard Krug hervor heben, mit denen ich mich kontinuierlich über fachliche Fragestellungen austauschen konnte. Eine wichtige Hilfe stellten auch die Mitarbeiterinnen und Mitarbeiter am Lehrstuhl von Prof. Dr. Claus Schnabel an der Wirtschafts- und Sozialwissenschaftlichen Fakultät in Nürnberg dar, die stets ein offenes Ohr für Problemstellungen beim Verfassen dieser Arbeit hatten. Herr Doc. Ing. Daniel Münich, Ph.D. (CERGE-EI) hat einen großen Anteil am Zustandekommen der Analysen über die regionale Entwicklung des tschechischen Arbeitsmarktes, indem er mir den Zugang zu den dazu notwendigen Datensätzen ermöglichte und mir bei deren Aufbereitung enorm half. Introduction: Analysis of Labour Markets in Border Regions IAB-Bibliothek 321 14 sition years. The main focus of this thesis clearly lies on the empirical analysis. However, in order to be able to derive hypotheses, the theoretical background will be dealt with in sufficient detail. This work is structured as follows: in Chapter 2 I present a survey of the literature on economic integration and border region studies. To begin with I address the main theoretical approaches, which can be used for generating hypotheses. After a short recap of traditional location, I focus on trade theory and New Economic Geography, which emanated from the former two strands of theories (Krugman 1991a, 1993). I concentrate explicitly on two models: first, the trade model by Feenstra/Hanson (1996a), which was developed against the background of Mexican trade liberalisation, and second, the NEG model by Brülhart et al. (2004), which deals with economic integration in the course of EU enlargement. Both models are largely used as a theoretical basis in borderland studies. Chapter 3 places emphasis on the analysis of the Bavarian border region. After showing some descriptive figures about the economic changes at the border before and after the fall of the Iron Curtain, I apply econometric models in order to analyse qualification trends and wage differentials in the eastern Bavarian border region. It is possible to add an extract from the employment register (BeH), which contains all observations (100 % instead of 2 %) from the border region, to the IABS. Thus, I am able to perform several sensitivity analyses applying advanced econometric methods (propensity score matching, DID, fixed effects) in order to estimate wage differentials. Chapter 4 investigates spatial effects of the transformation of the Czech Republic from a state-directed to a market economy and – similarly to the investigation on the German side – analyses the development of labour market indicators differentiating between districts close to Bavaria/Austria and non-border districts. In Chapter 5 I summarise the main findings of the thesis and integrate the results with the ongoing EU integration process and the further liberalisation of markets. It should be noted that the term “border region” is used in my thesis as a synonym for eastern Bavaria (the districts close to the Czech Republic) and/or the Czech districts close to Bavaria and/or Austria. Of course, both countries have districts bordering on other countries or federal states. Naturally, in the framework of my thesis these districts belong to the non-border (or interior) region, i.e. the rest of the country. After defining the outline of the regions in the respective chapter it is clear from the context which region is meant. 15 Chapter 1 Appendix to Chapter 1 Appendix to Chapter 1 Figure A 1.1: German exports to and imports from the Czech Republic (bn €, 1993–2007) Source: Sachverständigenrat zur Begutachtung der gesamtwirtschaftlichen Entwicklung. 1993 1995 1997 1999 2001 2003 2005 2007 Export Import 30 25 20 15 10 5 0 Figure A 1.2: FDI figures of German companies in the Czech Republic (1991–2006) Source: Deutsche Bundesbank. Notes: Until 1992 former Czechoslovakia; without dependent holding companies; due to adjustments to the reporting exemption limits in 1993 and 2002 a considerable number of enterprises were relieved of the obligation to report (Deutsche Bundesbank 1995, 2004). 1991 1994 1997 2000 2003 2006 20 18 16 14 12 10 8 6 4 2 0 (a) direct & indirect FDI (bn €) 1991 1994 1997 2000 2003 2006 1,400 1,200 1,000 800 600 400 200 0 (b) number of German companies 300 250 200 150 100 50 0 (c) employees ’000 1991 1994 1997 2000 2003 2006 60.0 50.0 40.0 30.0 20.0 10.0 0.0 (d) total revenue (bn €) 1991 1994 1997 2000 2003 2006 Introduction: Analysis of Labour Markets in Border Regions IAB-Bibliothek 321 16 Figure A 1.3: Bavarian exports to and imports from the Czech Republic (bn €, 1989–2003) Source: Bavarian Ministry of Economic Affairs, Infrastructure, Transport and Technology. Notes: Until 1992 former Czechoslovakia. 1989 1991 1993 1995 1997 1999 2001 2003 Export Import 6 5 4 3 2 1 0 Figure A 1.4: FDI of Bavarian companies in the Czech Republic (bn €, 1995–2006) Source: Deutsche Bundesbank. Notes: Direct and indirect FDI. 1995 1997 1999 2001 2003 2005 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0 17 Chapter 1 Appendix to Chapter 1 Figure A 1.5: Regional distribution of German holding companies operating in the Czech Republic Source: Author’s own calculations; German-Czech Chamber of Industry and Commerce. Notes: Regional classification on the basis of 4-digit postcode boundaries. Introduction: Analysis of Labour Markets in Border Regions IAB-Bibliothek 321 18 Figure A 1.6: Regional distribution of affiliated firms in the Czech Republic: (a) affiliated to German companies and (b) affiliated to Bavarian companies (a) affiliated to German companies (b) affiliated to Bavarian companies Source: Author’s own calculations; German-Czech Chamber of Industry and Commerce. 19 Chapter 1 Appendix to Chapter 1 Figure A 1.7: Market entry of German companies in the Czech Republic: (a) overall and (b) subdivided into economic sectors Source: Author’s own calculations; German-Czech Chamber of Industry and Commerce. Notes: Regarding economic sectors multiple responses were possible. 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 year 200 150 100 50 0 (a) overall number 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 year services manufacturing and services trading manufacturing and trading trading and services manufacturing, trading and services manufacturing 50 40 30 20 10 0 (b) subdivided into economic sectors number Introduction: Analysis of Labour Markets in Border Regions IAB-Bibliothek 321 20 Figure A 1.8: Distribution of affiliated firms between manufacturing, trading and service sector Source: Author’s own calculations; German-Czech Chamber of Industry and Commerce. Notes: Regarding economic sectors multiple responses were possible. 336 232 202 270 59 153 129 manufacturing, trading and services manufacturing and trading manufacturing and services manufacturing trading and services services trading 21 Chapter 2 2 A Survey of the Literature 2.1 Theoretical Background: Integration Effects in Border Regions There is no independent theory regarding the effects of the opening of borders on labour markets in frontier regions. However, empirically testable hypotheses with respect to border regions can be derived from traditional location theory as well as international trade (e.g. the model by Feenstra/Hanson 1996a) and New Economic Geography (NEG) models. In the model by Brülhart et al. (2004) extensions with explicit reference to border regions are elaborated. A review of theoretical approaches concerning integration effects in border regions is given by Niebuhr/ Stiller (2004) and the studies compiled in the course of the transnational project “Preparity” by Mayerhofer (2004) and Riedel/Untiedt (2001). In the next subsections I address the most relevant approaches and models, i.e. traditional location theory, both traditional and new trade theory and New Economic Geography. 2.1.1 Traditional Location Theory The economic situation of border regions and the effects of diminishing trade impediments are already stressed by Lösch (1962). Assuming transport costs directly proportionate to distance, a circular distribution area evolves around a central location in a homogeneous plain. Beyond this circle transport costs are prohibitive, so that there are no sales at all. The radius depends on the respective supply and demand functions of the specific products. As a consequence of a supposed frontier the circle is cut, e.g. as a result of impediments preventing any trade. Consequently, border regions do not attract enterprises in search of large market areas as long as the locational disadvantage is not compensated through lower costs (e.g. wages, rents etc.). After the opening of the border the distribution area expands: compared to non-border regions the improved market access should boost the attraction of the border region and lead to the establishment of enterprises, which are now able to supply the full market area. In other words the market potential – in simple terms the purchasing power inversely dependent on the distance (this term is also used in the New Economic Geography models) – of a firm increases (Figure 2.1).3 3 Furthermore, concerning traditional location theory, border region aspects are addressed by Giersch (1949/50), among others. IAB-Bibliothek 321 22 A Survey of the Literature Figure 2.1: Market potential in a border region 2.1.2 Trade Theory 7UDGLWLRQDO7UDGH7KHRU\ In the course of the economic integration of border regions, traditional trade theory plays a role when the potential structure of the intensified trade between formerly relatively closed countries can be predicted. In order to analyse effects of reduced trade costs regarding the European Union, old and new member states are each embraced as one country, so that the basic two-country, two-sector models can be applied. In Ricardo’s model (1817) every country will specialise in producing the commodity where it has a comparative advantage, accruing from different technologies. The different labour productivity in two sectors leads to different real wage costs and consequently to a competitive advantage in one sector and a competitive disadvantage in the other. In the Heckscher-Ohlin model (1919, 1933), as a result of different relative factor endowments a country has a comparative advantage in the sector which intensively uses the relatively abundant factor (labour or capital). Capital also comprises the endowment of a country with technology as well as human capital, i.e. highly qualified employees. With respect to the enlargement of the European Union, this implies that a comparative advantage for the Western European capital-abundant EU countries can still be assumed regarding the production of high quality products. The Eastern European countries possess a comparative advantage in the production of goods which uses intensively low-skilled labour. This pattern is valid for all accession countries, but is declining most notably for the CEEC-5 countries (Hungary, Poland, Czech Republic, Slovak Source: Lösch (1962). Home Foreign border accessible market potential A 23 Theoretical Background: Integration Effects in Border Regions Chapter 2 Republic, Slovenia) (Zarek 2006). The production factors are assumed to be totally mobile within the countries and between the sectors, but completely immobile between the countries, i.e. the factor endowments are fixed for the countries. After the introduction of trade, structural adjustments emerge through a reallocation of the production factors between the sectors. Concerning the development of wages, the specialisation on the activities with comparative advantages leads to effects according to the Stolper-Samuelson theorem. Assuming constant returns to scale and continuing production of both commodities, an increase in the relative price of a commodity will lead to a rise in the real return to the factor used relatively intensively in its production and to a fall in the real return to the other factor (Stolper/Samuelson 1941). With respect to the Western European countries, the implementation of distinctive trade relations with the accession countries suggests a relatively rising demand for more highly educated employees and consequently a deepening of the wage gap between high- and low-skilled workers. In contrast, wage differentials are supposed to be diminished in the Eastern European countries. According to the factor-price-equalisation theorem, free trade in commodities will equalise wage and interest rates in the participating countries. Trade works as a substitute for the mobility of production factors. Though the Heckscher-Ohlin-Samuelson (HOS) framework fails to explain important trade developments over the past decades, it is still used as a background for some significant studies on open border effects (see Chapter 2.2). One of the major problems in traditional trade theory is the assumption of factor immobility between countries, which is unrealistic at least with regard to capital. Moreover, in the traditional models, intra-industry trade, which accounts for a large majority of trade between industrial countries, cannot be explained. Thus, new models responding to this fact have been developed, summarised under the term “new trade theory”, which I comment on in the next subsection. In terms of regional science the absence of transport costs and the consideration of countries as “dimensionless points in space” without spatial diversification pose a striking drawback. 1HZ7UDGH7KHRU\ Since the 1960s new trade models have tried to explain intra-industrial trade, taking into account increasing returns to scale, countries’ different technological skills and imperfect competition. In the next subsections the most important models are described briefly. If the assumption of constant returns to scale is lifted, trade between countries can be explained without referring to comparative advantages. As scale economies generally do not fit in general equilibrium models, a more complex analysis must IAB-Bibliothek 321 30 A Survey of the Literature cost curves CC and C*C* intersect, it comes to a fragmentation of production with c(w,q,r,z*) = c(w*,q*,r*,z*) for input z*. Activities z’>z* require skilled labour relatively intensively. Since the relative wage of skilled employees is lower in the home country according to the above assumption, all activities z’>z* can be produced advantageously at home. On the other hand, all activities z’<z* can be produced at a cheaper rate in the foreign country. Feenstra/Hanson (1996a) proved that CC and C*C* can intersect once at most. As a consequence the foreign country specialises in activities [0, z*), whereas the home country specialises in activities (z*, 1]. Hence, the relative demand for skilled and unskilled labour can be calculated for each country. In the home country it holds that (2.8) In the foreign country it holds that (2.9) Since the final commodity as a composition of the intermediate inputs requires no additional costs, it is not significant where it is produced, as mentioned above. The Figure 2.4: Feenstra-Hanson model without capital mobility Source: Feenstra (2004). z C, C* C C* C C* z* 0 Foreign Home 1 31 Theoretical Background: Integration Effects in Border Regions Chapter 2 demand for skilled and unskilled labour depends on the factor prices. Figure 2.5 shows the gradient of D(z*) subject to the ratio of skilled and unskilled labour. Relative labour demand depends on factor prices which affect the prices for intermediate inputs influencing the equilibrium demand for x(z). In each country the gradients D(z*) and D*(z*) are decreasing in the relative wage of skilled/unskilled labour. If each country produces the range of activities which it has a comparative advantage in, equilibrium will be reached in the markets for both skilled and unskilled labour and capital (z* in Figure 2.4). The wage bill in the home country is represented by Z/T+. Since wages account for the share Θ of total costs, GDP in each country is Z/T+ Θ . Multiplied by the cost share of capital it follows that (2.10) Assuming a fixed capital endowment, the rental on capital r will in this manner be determined in the home as well as the foreign country. Assuming r<r* and capital mobility between the two countries, capital moves from the home to the foreign country, so that K decreases and K* increases. This results in a rising r and a shrinking r*, leading to an upward parallel shift of the CC curve and a downward parallel shift of the C*C* curve (Figure 2.6). Consequently, the equilibrium value of z increases from z* to z’ . The foreign country specialises in an expanded range of activities [0, z’), while the home country specialises in a contracted range of activities (z’, 1]. Regarding the demand for Figure 2.5: Relative labour demand without capital mobility Source: Feenstra (2004). H/L q/w D(z*) IAB-Bibliothek 321 32 A Survey of the Literature labour, this has the following implications: The production activities which are outsourced from the home to the foreign country (z*, z’) use skilled labour less intensively than the activities which are still produced at home. Hence, the relative demand for unskilled labour drops in the home country, while the relative demand for skilled labour rises. The relative labour demand curve shifts to the right from D(z*) to D(z’) (Figure 2.7). In the foreign country the relocated activities use relatively more skilled labour than was used hitherto, leading to a growing demand for skilled labour there too. In contrast to the implication of the Heckscher-Ohlin- Samuelson framework, relative wages of skilled labour are expected to rise in both countries. The conclusion from this theoretical model for border regions is that the presumed effects appear in the course of integration of economic markets in regions close to the frontier to a greater extent. This means that the rise in the demand for skilled labour is above average in border regions compared to the rest of the country, i.e. low-skilled workers residing close to the frontier should suffer extraordinarily losses, while, vice versa, more highly skilled workers there should particularly benefit from the opening of the border. The advantage of this model is the explicit proposition concerning the effects of integration on skill groups. The special effects on border regions, however, are derived indirectly assuming that distance should matter in cross-border trade relations. In order to focus also on models which by principle consider distance costs, I now turn to the strand of literature which is currently widely applied for regional analysis. Figure 2.6: Feenstra-Hanson model including capital mobility Source: Feenstra (2004). C, C* C C* C C* z 0 Foreign Home 1 z* z 33 Theoretical Background: Integration Effects in Border Regions Chapter 2 2.1.3 New Economic Geography In New Economic Geography (NEG) elements of location theory are combined with new trade theory. As one strand of new trade models, in which intra-industrial trade is explained through monopolistic competition, New Economic Geography also has its microeconomic basis in the Dixit-Stiglitz model, enriched by components of traditional location theory. Unlike conventional partial spatial models, in NEG models analysis is carried out in a general equilibrium model. Goods are not only heterogeneous with respect to their spatial availability, but also regarding their physical constitution and psychological perception and are produced under increasing returns to scale (Schöler 2005). 7KH%DVLF$SSURDFKE\.UXJPDQ Krugman’s seminal model (1991b) introduces interregional mobility of part of the workforce in his trade model (1980). Basically, he tackles the question of why the manufacturing sector is concentrated in some regions (core), whereas other regions lag behind (SHULSKHU\), i.e. both firms and people migrate to larger agglomerations. In Krugman’s basic NEG model the development of a core-periphery structure depends on the relationship between three parameters: transport costs, economies of scale (or the elasticity of substitution among differentiated goods) and the share of mobile workers in the population as a whole. Regarding my research topic I focus on the question of what happens to border regions according to NEG when transport costs (or trade costs) fall after the opening of a border. Does the basic model give any hints about this issue? Figure 2.7: Relative labour demand with capital mobility Source: Feenstra (2004). H/L q/w D(z*) D(z') IAB-Bibliothek 321 34 A Survey of the Literature The fundamental elements of the Krugman model are two regions and the two sectors agriculture and manufacturing. Agricultural goods are produced under perfect competition and constant returns to scale, while manufacturing goods are produced under monopolistic competition and with increasing returns to scale. Transport costs only accrue for manufacturing goods. The immobile peasants are symmetrically distributed between the two regions, whereas the workers in the manufacturing sector are interregionally mobile and locate in the region where the higher real wage is paid. The decisive parameters with respect to the development of a core-periphery structure are the share of manufacturing workers μ, the elasticity of substitution σ and the inverse index of transport costs τ . In the short run equilibrium, manufacturing workers are immobile, too, so that the distribution of the population between the two regions is given. If the manufacturing workers are equally distributed between the regions, wages are also equal in the regions. If the share of manufacturing workers in region 1 increases, two opposing effects influence the relative wage rate: on the one hand, according to the “home market effect”8, the wage rate tends to be higher in the larger market. On the other hand, there is less competition in the region with the smaller share of manufacturing workers, so that the reaction concerning the relative wage rate is ambiguous. In the long run equilibrium, it is important to distinguish between nominal and real wages. Due to transport costs, manufacturing workers pay a lower price for manufacturing goods in the larger market. Migration of manufacturing workers from region 2 to region 1 leads to a lower price index in region 1 and a higher price index in region 2, so that the divergence between the regions accelerates. The reaction of the real wage rate ω 1 ω 2 after a shift in the share of manufacturing workers in region 1 (denoted by f) is therefore decisive whether or not a core-periphery structure develops. If ω 1 ω 2 decreases with f, manufacturing regions will migrate out of region 1 until the share is equal in both regions. If ω 1 ω 2 increases with f, eventually all manufacturing workers will migrate into region 1 and only the immobile peasants will stay in region 2. With two factors working towards divergence (“home market effect” and the price index effect) and one factor working towards convergence (degree of competition in the smaller region), the determination of which effects dominate depends on the ratio of the values of μ, σ and τ . High transport costs, i.e. a relatively small value of τ , a low share of mobile workers (μ relatively small) and low (positive) returns to scale (thus a high elasticity of substitution σ ) favour convergence. Decreasing transport costs (apart 8 Assuming increasing returns to scale, the “home market effect” denotes the tendency of companies to produce in countries in which there is a large domestic demand (Krugman 1980). 35 Theoretical Background: Integration Effects in Border Regions Chapter 2 from zero, where distance does not matter at all), an increase in the number of mobile workers and increasing scale economies will lead to a concentration process in one region, which (by chance) has a head start. In Krugman’s basic model (1991b) this results in a ‘bang-bang’ solution, i.e. the complete agglomeration of the mobile workforce in one region if transport costs sufficiently decrease. In this context, it is important to mention the so-called “no-black-hole-condition” which rules out that agglomeration always prevails, even if transport costs were infinite. Therefore it is necessary to assume that , i.e. increasing returns to scale and/or the budget share of manufacturing workers are sufficiently low (Fujita et al. 1999). The catastrophic outcome of total agglomeration in the basic model has been criticised in the literature and subsequently several models were developed trying to explain partial agglomeration and an equal allocation of mobile workers across the two regions (Ottaviano/Puga 1998; Puga 2002; Ottaviano/Thisse 2004; Pflüger/Südekum 2008). The factors which pull economic activities together are characterised as centripetal forces (e.g. larger markets, spillover effects etc.). The factors which push economic activities apart are subsumed under centrifugal forces (e.g. higher stress of competition, labour costs, rents etc.). Venables (1996) referred to the generation of centripetal forces as forward linkages (workers as consumers prefer to be close to the producers) and backward linkages (producers prefer to concentrate where the market is larger). The key figure with respect to the attractiveness of a region for the potential settlement of firms is the relative impact of centripetal and centrifugal forces. The opening of a border or the abolition of trade impediments between two countries reduces the cross-border transaction costs and facilitates factor mobility, which could lead to a reallocation of economic activities not only EHWZHHQ the countries, but also ZLWKLQ the countries (Niebuhr/Stiller 2004). In the former case, migration leads to a spatial reallocation of people and firms at the international level. However, more important for research on border regions is the second case. While activities in a closed economy are inwardly orientated, after the opening of a border the domestic market loses its predominant role (Krugman/ Elizondo 1996). The centre becomes less important, and the periphery becomes more attractive due to the improved trade and distribution opportunities. The settlement of firms then leads to the immigration of workers, and a self-reinforcing circular process begins. Even if cost differentials between regions are not explicitly modelled, positive effects on border regions are implied in the literature (Hanson 1996). Border regions move from the periphery to the centre, and their “home market” and market potential increases. Thus, concerning the unequal development of regions the level of transport costs plays a central role. With regard to the question as to which effects a de- IAB-Bibliothek 321 36 A Survey of the Literature crease in transport (transaction) costs has on employment growth, the structure of economic activities, the qualification and the wage structure, results can hardly be derived from most NEG models. In the presence of multiple equilibria, the direction of effects depends on the parameter values of the variables. However, “New Economic Geography has come of age” (Neary 2001) and in recent years models have been developed which make it easier to constitute hypotheses for the economic deployment of border regions. There are several NEG models differentiating between regions and countries and thus dealing with increasing international integration (e.g. Krugman/Elizondo 1996; Monfort/Nicolini 2000; Behrens et al. 2006, 2007). In the next subsection I dwell on the model of Brülhart et al. (2004), which provides a basis for border region studies (Niebuhr 2006b, 2008) and from my point of view corresponds best to my research questions. $7KUHH5HJLRQ0RGHOE\%UÙOKDUW&UR]HW.RHQLJ One of the main problems of the basic NEG model with respect to my research questions is that it is not analytically solvable despite its restrictive assumptions. The equilibria can only be found by numerical computer simulations, with the consequence that hypotheses are hardly testable. A three-region model by Brülhart et al. (2004), which is derived from Krugman’s basic model and the analytically solvable core-periphery model of Pflüger (2004) based on it, differentiates between a foreign country (0) and two regions in the home country: an interior region (1) and a border region (2). As in the basic model there are two sectors: the perfectly competitive agricultural sector (A) uses only the immobile production factor labour (L). The monopolistically competitive manufacturing sector (X) produces differentiated goods and uses labour (L) as a variable input and human capital (K) as a fixed cost, which leads to increasing returns to scale. Manufactured goods are assumed to incur transport costs of the iceberg type, i.e. a fraction of a shipped commodity melts away and only the part 7 arrives at its destination, increasing the price of the unit received to pT. Transport costs arise in interregional domestic trade (with T12 = T21) and in each domestic region’s trade with the foreign country. The utility function for an individual has − in contrast to Krugman’s model − the quasi-linear form (2.11) where CX represents the consumption of the different varieties of the manufacturing commodity and CA the consumption of the agricultural commodity. 37 Theoretical Background: Integration Effects in Border Regions Chapter 2 The Index CX is defined by a CES function (2.12) where x represents the consumption of variety i of the manufacturing commodity and Q the number of varieties which are potentially available in a region, Q being proportionate to the region’s endowment with human capital. The elasticity of substitution between two varieties is denoted by σ . In zero-profit equilibrium the term σ  σ -1) represents the relation between the marginal and the average product of labour, thus the level of returns to scale. Production factors are assumed to be immobile between countries, so that L0 and K0 are exogenous. In the domestic regions the supply of the immobile factor labour is also fixed by L1 and L2. Human capital is mobile between the domestic regions and migrates according to the indirect utility differentials. The regional shares of human capital are denoted by K1 .  λ and K2 .  λ . The budget constraint of an individual is given by (2.13) where Y represents the income, pA is the price of the agricultural good and pi is the price of variety i of the manufactured good. Therefore, the demand of consumers in region s for variety i in region r is (2.14) Each of the three regions produces Qr varieties of the manufacturing commodity. Due to the iceberg transport costs the price for variety i produced in region r, which is sold in region s, is expressed by pirs = prTrs. Finally, the price index of the varieties of the manufacturing commodity which are sold in region s can be written in the form (2.15) and an individual’s demand is (2.16) IAB-Bibliothek 321 38 A Survey of the Literature The agricultural sector (A) produces goods under perfect competition using only the immobile factor labour (L). Transaction costs are not incurred for agricultural goods interregionally or internationally, so that pA1 = pA2 = pA0. It is assumed that in each region pA = wA. Since the agricultural commodity is used as a numéraire, wA = 1. According to these assumptions, a representative firm in region r will set the following profit-maximising price: (2.17) Market clearing for each variety leads to an equilibrium output of a firm producing in r: (2.18) and the profits of the firm are given by (2.19) where Rr represents the remuneration for human capital in region r. While in the short term human capital is also immobile between the domestic regions, in the long run it depends on the level of indirect utility whether domestic human capital owners will locate in region 1 or region 2.9 Finally, not nominal but real wages are essential for a higher utility of workers. The indirect utility function, derived from (2.11), has the form (2.20) and the consequential utility differential is (2.21) which depends on the distribution of human capital between the regions and the parameters of the model. Human capital owners migrate towards region 1 if the price index for manufacturing goods in region 2 and the compensation for human capital in region 1 respectively are sufficiently high. Brülhart et al. now distinguish between two scenarios: in the basic model the trade costs of the domestic regions with the foreign country are equal (T01 = T02). 9 For a detailed description of the short run equilibrium, see the appendix of Brülhart et al. (2004). 39 Theoretical Background: Integration Effects in Border Regions Chapter 2 Shrinking trade impediments cause two opposing effects: on the one hand, centripetal forces become weaker, because due to the new market in the foreign country the incentive to locate near domestic consumers is reduced. Besides, the income and also the supply from the foreign country become more important, lowering the domestic agglomeration force. On the other hand, centrifugal forces diminish, too. The higher competition from the foreign country relativises the possibility of evading domestic competition by locating in the periphery. It can be shown that with decreasing trade costs the effect on the centrifugal forces generally dominates, i.e. the probability of a core-periphery structure grows, although it is not determined which region will be the core. The same effect occurs if the foreign country gets bigger in economic terms. In the second scenario, which is more interesting and also more realistic with regard to spatial differences, the domestic regions are asymmetric. The trade costs between region 1 and the foreign country are higher (T01 > T02). Thus, region 1 is called the interior region, while region 2 is denoted as the border region. Hence, with respect to the growing market potential, the incentive for domestic firms to locate near the foreign consumers − thus in the border region − increases. Otherwise, as the domestic dispersion force is weakened by foreign supply, the interior region attracts domestic firms by reason of the distance to the foreign competitors. Though the model offers no clear-cut results, it can be concluded from simulations that for most parameter configurations the border region benefits from the abolition of trade impediments. Figure 2.8 shows the indirect utility differentials (V1 − V2 in equation 2.21) dependent on the share of manufacturing workers in the interior region ( λ ). An equilibrium of the allocation of manufacturing exists if either the indirect utilities are equalised or in the case of total agglomeration in one region and a lower potential indirect utility in the other region. Assuming L1 .1+K2 ) = L2 .1+K2 ) = 1, Figure 2.8a illustrates autarky for the parameter values σ = 6, α = 0.3, T12 = 1.46 and T01 = T02 = ∞ . In this configuration the only stable equilibrium is the symmetrically dispersed location of manufacturing. If the share of manufacturing workers is higher in one region, indirect utility is higher in the other, leading to migration until the share is balanced. Reduced trade costs with the foreign country lead to a completely different graph, above all regarding equilibria. In Figure 2.8b the effects of trade are depicted with T01 = 1.55 and T02 = 1.5. In addition to the preconditioned relation of labour and human capital in the domestic country in the previous figure, it is assumed that, with regard to the foreign workforce, L0 .1+K2 ) = K0 .1+K2 ) = 1. In this case only the complete agglomeration in one of the two regions constitutes a stable equilibrium. The implication of this outcome is that trade liberalisation increases the attractiveness of the border region, as the effect of an increased market potential is larger than the IAB-Bibliothek 321 46 A Survey of the Literature wicz (2005) finds a growing wage gap between high- and low-skilled workers in Poland. While Lorentowicz et al. (2005) – like Feenstra/Hanson working with nonproduction/production shares – confirm higher skill premiums in Poland caused by outsourcing as a result of fixed effects estimations for the period from 1994 to 2002, they surprisingly, and contradicting theory, also have estimates for Austria, pointing to lower skill premiums in the high-wage country due to international outsourcing. While this strong argument is not backed up by the literature, it is quite evident that skill upgrading also took place in the EU accession countries (Bruno et al. 2004). However, the question remains whether international outsourcing and trade in intermediates are the driving force behind this. According to Geishecker/Görg (2008), international outsourcing in German manufacturing increased between 45 and 60 % from 1991 until 2000. In one of the rare micro-level studies the effects of outsourcing on the wages of different skill groups are estimated by allowing for individual fixed effects. Geishecker/Görg (2008) find evidence for low-skilled workers being the losers in globalised production, since outsourcing reduced the real wages in this skill group by up to 1.8 %. On the other hand, high-skilled workers benefited from trade by increased wages of up to 3.3 %. These results are in line with the findings of Geishecker (2004), who states that with nearly stable relative wages in the 1990s, the decline in the relative demand for low-skilled labour can be explained to up to 24 % by international outsourcing. One of the most common measures used in studies on integration effects is the market potential, which goes back to Harris (1954) and for which several definitions exist in the meantime. Basically, according to this measure, the higher the sum of the purchasing power around a location (weighted by transport costs), the higher the demand for goods produced in a location, suggesting that economic activities in a region flourish as access to markets increases. In order to capture the effects of EU enlargement, Brülhart et al. (2004) use the market potential as their main explanatory variable. The market potential function in their version is defined as (2.22) where i and j represent regions out of a set of J regions, Y denotes the economic weight of a region (e.g. GDP or number of employees) and d denotes the distance between i and j. The estimated values of the market potential are incorporated as an exogenous variable in a regression model estimating per capita GDP or the share of the population employed in the manufacturing sector. The results of the simulations suggest that EU enlargement only caused small effects on per capita income 47 Chapter 2 Empirical Studies on Integration Effects in Objective 1 regions12, but a significant rise in manufacturing employment, above all in the regions neighbouring the new member states. In a cross-section analysis of European regions by Niebuhr (2006a) covering the period between 1985 and 2000, the relationship of regional wages rising with an increasing market potential is confirmed, though – compared with other studies (e.g. Roos 2001; Brakman et al. 2000, 2004; Hanson 2005) – the geographic scope of demand linkages is significantly larger than in the previous studies and market access seems to become less important over time. Interestingly, the estimates in the study, which includes only Western European countries, suggest that there are no substantial border impediments affecting the regional wage structure, indicating the high integration of the European Union. In two further studies Niebuhr investigates the impacts of an increased market access on per capita income, explicitly referring to border regions, against the background of the Brülhart et al. model. Concerning only EU15 regions (Niebuhr 2006b) the simulation analysis suggests that integration benefits for internal EU border regions are significantly higher than for non-border and external EU border regions. Investigating integration effects at the EU27 level, Niebuhr (2008) concludes that the Eastern European countries above all will benefit from the reduction of trade impediments. The impact of enlargement seems large even on the market potential of CEEC external border regions. Though the estimated effects on per capita income remain small, at least the CEEC regions along the former Iron Curtain exhibit a significant increase through enlargement. The results of these studies are corroborated by Huber et al. (2006). By explicitly modelling border effects, they find that regions in the accession countries which are close to the EU15 countries are especially gaining from enlargement, with the Czech Republic being most affected. However, EU15 regions close to the accession countries are also predicted to benefit from the abolition of trade impediments. In a study that focuses on border regions, Mayerhofer (2004) finds that the Austrian regions close to the EU accession countries have developed favourably since the opening of the border. However, he concludes that this positive process does not stem from integration with Eastern European countries, but from the decentralisation and de-concentration process which proceeded in Austria in the 1990s. Barjak/Heimpold (2000b) explore the economic development of the regions situated along the German-Polish border, finding that below-average rates of exports, sales and investments in the manufacturing sector do not provide evidence for positive locational effects of integration on the labour market in eastern German border districts. In contrast, three out of four Polish border regions register a 12 Objective 1 regions are EU areas lagging behind in their development where the gross domestic product (GDP) is below 75 % of the Community average. IAB-Bibliothek 321 48 A Survey of the Literature rise in the share of nationwide per capita investments. Regarding the unfavourable development in the New Laender the authors do not blame the opening of the border but refer to specific locational disadvantages, such as inappropriate traffic infrastructure and the lack of qualified personnel. Riedel/Untiedt (2001) estimate the trade potential of the German federal states (%XQGHVOÁQGHU) using a gravitation model for the years 1991 to 1997. They conclude that exports from Bavaria to the Czech Republic are disproportionately high, expressing the benefits of the shorter distance. In the eastern German states, though the rate of exports coming from former COMECON connections is above-average, this effect is only marginal, and, added to that, decreases over the years. However, for eastern German border regions export activities increase from 1993 on, indicating a positive integration process. Concerning the imports, the federal German states that have a common border with Poland and/or the Czech Republic exhibit higher growth rates, also signifying the higher importance of foreign trade to these regions. Some studies focus on the development at the former inner-German border. Blien et al. (2003) place special emphasis on the eastern German districts close to western Germany, among other regions. Using a shift-share approach for the period from 1993 to 1999, they find that − compared to other eastern German regions − these districts exhibit a relatively positive employment trend. However, this favourable development can only partly be attributed to the geographical position, but mainly to the firm size structure and special influences like above-average government aid. Using a difference-in-differences approach, Büttner/Rincke (2007) compare western German regions situated close to the former German-German border with other western German regions after re-unification. They come up with the result that cross-border mobility led to an increase in employment, a drop in the relative wages and higher local unemployment rates in the western German border regions. According to Redding/Sturm (2005), the western German cities close to the inner-German border suffered substantially in the aftermath of the German division due to the loss of market access. Controlling for the industrial structure and war-related destruction, population growth was significantly lower in the border cities compared to other West German cities. After the German reunification the border cities experienced a relative recovery. Fuchs-Schündeln/Izem (2007) inquire into the causes of the low labour productivity in eastern Germany. Differentiating between job and worker characteristics, they analyse unemployment rates in the eastern German districts. Since the unemployment rates increase with the distance to the former inner-German border, they conclude that East Germans living close to the border can easily commute and find jobs in western German districts. Thus, there are indications that the eastern German workforce does not have a lack of human capital, but job characteristics in eastern Germany are less favourable than 49 Chapter 2 Empirical Studies on Integration Effects in the western part. Besides this, they confirm the results of Büttner/Rincke (2007), finding that western German districts close to the East suffer significantly higher unemployment rates after reunification than the rest of western Germany. Generally, as far as the implications of international trade and outsourcing are concerned, the focus of most studies is on the effects on the developed countries. Both theoretical and empirical analyses with respect to the consequences on the labour market in transition or developing countries are rather scarce. Egger/Egger (2002) investigate the impact of trade in both final and intermediate manufacturing goods on gross wages in seven CEEC countries and find surprising results. According to their analysis, trade in final goods does not have a significant effect on wages. Interestingly, the impact of intermediate goods exports is clearly negative, whereas the effect of intermediate goods imports is significantly positive. The rather unexpected signs of the wage effects can be explained by two opposing effects. On the one hand, outsourcing to CEEC low-wage countries potentially raises the demand for low-skilled workers and thus the remuneration for low-skilled labour probably increases. On the other hand, as an effect of the shift in labour demand, the composition of the work force concerning skill groups may be pushed towards low-skilled labour, which in turn can lead to an overall shrinking wage bill. 2.2.3 Summary Recapitulating the empirical results from the different integration areas it can be said that there are no clear-cut results. Unfortunately, in many cases the analyses are related to whole countries, while there is no reference to regional information, often due to a lack of available data. However, in both the U.S.-Mexican case and regarding the European integration process, the basic message of the bulk of the studies is that border regions are either not exposed to special changes or border regions benefit from integration to a higher than average degree. Many studies refer to the higher market potential after trade liberalisation which should lead to a favourable position of border regions. Only few studies show negative effects for specific skill groups in border regions. Interestingly, the studies which contradict the Feenstra-Hanson model refer to Stolper-Samuelson effects, i.e. in the so-called low-wage country it is not high-skilled workers who gain from integration, but low-skilled workers due to their comparative advantage. Though it is beyond all doubt that skill premiums went up in both Western and Eastern European countries, the main cause of this is controversial. Since it is above all the effects of free trade and international outsourcing on the development of skill structures and wage differentials that are disputed, I will focus on these topics in Chapters 3 and 4. IAB-Bibliothek 321 50 A Survey of the Literature Appendix to Chapter 2 Table A 2.1: Implications of the Feenstra-Hanson trade model and the Brülhart et al. NEG model Feenstra/Hanson Brülhart/Crozet/Koenig Eastern Bavaria activities which are offshored to the Czech Republic are least skillintensive relative increase in sectors where import competition from the Czech Republic is supposed to be relatively low increasing structural change and specialisation increasing structural change and specialisation lower relative labour demand for unskilled and low-skilled workers change in labour demand for unskilled and low-skilled workers depends on strength of opposing effects (import competition vs. higher market potential) higher relative labour demand for more highly skilled workers higher relative labour demand for more highly skilled workers Czech border region activities, which are offshored from high-wage countries, are relatively skill-intensive in the Czech Republic relative increase in sectors where import competition from Germany and Austria is supposed to be relatively low increasing structural change and specialisation increasing structural change and specialisation lower relative labour demand for unskilled and low-skilled workers higher relative labour demand for unskilled and low-skilled workers higher relative labour demand for more highly skilled workers change in labour demand for more highly skilled workers depends on strength of opposing effects (import competition vs. higher market potential) 51 Chapter 3 3 Labour Market Effects in the Bavarian Border Region 3.1 Introduction The fall of the Iron Curtain fundamentally changed the economic relationships between Western and Eastern European countries. The regions situated on the border with the new EU member countries are particularly affected by reduced restrictions in trade and an increased division of labour, which lead to intensified attention and multitudinous syndicates within the EU countries. Remarkable examples are the working community of chambers of economy in the EU regions bordering the Central and Eastern European candidate countries (ARGE28), the SPIRIT initiative of the German labour unions and the association of European border regions (AEBR). Due to the unexpected events of 1989 the labour market along the border between the Federal State of Bavaria (western Germany) and the Czech Republic, which has one of the world’s largest wage differentials, can be regarded as a natural experiment. Despite existing restrictions on labour mobility, which for Czech employees will probably be limited until 2011, effects of international trade have been obvious since the opening of the border. Just as the introduction of the North American Free Trade Association (NAFTA) led to cross-border linkages along the U.S.-Mexico border (Hanson 2001), a gradual integration process is leading to transnational outsourcing of economic activities in the Bavarian-Czech borderland. As a result of trade and outsourcing I expect structural shifts in the labour demand, especially in the border region, and therefore changes in the skill and wage structures which are more distinctive than those at national level. There is controversial debate as to whether or not economic integration positively influences the development of border regions. Basically I distinguish between integration of product markets and free movement of capital and above all labour. Regarding fully integrated economic areas Büttner/Rincke (2007) show significant results concerning negative labour market effects in western German regions near the former German-German border, which suffer from a fall in the relative wage position and higher unemployment rates than other western German regions. Blien et al. (2003) investigate eastern German regions along the former inner-German border, concluding that though the situation on the labour market in these regions is better than in the remaining eastern German regions, this positive development is only partly caused by the favourable geographical location. Redding/Sturm (2005) provide results about western German border cities before and after reunification. The decline in population growth due to the reduced market potential during the period when Germany was divided was followed by a recovery after reunification. Contrary to the abovementioned studies, Barjak/Heimpold (2000b) and Stiller (2004) analysed IAB-Bibliothek 321 52 Labour Market Effects in the Bavarian Border Region German-Polish border regions, where full liberalisation of labour mobility has not yet happened, and do not find specific effects on the labour market there. Riedel/ Untiedt (2001) conducted investigations on several economic factors with respect to German regions bordering on new EU member states. Studies by Niebuhr (2006b, 2008) provide findings about higher integration benefits for border regions caused by declining trade impediments. Surprisingly few studies use individual-level data to estimate spatial effects in regions situated along the former Iron Curtain. In this chapter I investigate the development on the Bavarian side of the border. After providing some descriptive figures I analyse the shifts in skill group shares of employed and unemployed people comparing the eastern Bavarian border region to the other western German districts. Furthermore I apply several econometric methods in order to estimate spatial wage effects which I expect to be more distinctive in the region close to the Czech Republic. Even without transnational labour mobility I established hypotheses backed by trade theory and New Economic Geography predicting a higher demand for more highly skilled labour in the eastern Bavarian borderland. Thus, the central object of investigation in this chapter is which skill group benefits from the opening of the border: low-skilled, skilled or high-skilled workers? The paper is organised as follows: section 3.2 provides a description of the regional classification and the dataset I use for my investigations. Section 3.3 contains descriptive evidence of the labour market in the Bavarian border region before and after the fall of the Iron Curtain, compared with the development at national level in parts. In section 3.4 I apply a balanced panel model in order to estimate qualification trends. In section 3.5 I introduce econometric models to estimate wage differentials and present the results. Section 3.6 concludes. 3.2 Data and Basic Definitions I use micro data from the IAB Employment Sample (IABS) for the years 1980 to 2001, which are provided by the Institute for Employment Research (IAB) and contain information about a two percent random sample of all employees covered by the German social security system (for a description of the dataset see Hamann et al. 2004; Hamann 2005). Two versions of the IABS are available: The scientific use file (I use the regional file) and the weakly anonymous version. Both versions, which differ, for instance, in some content characteristics (e.g. classification of industries, occupation, employment size, average remuneration), are in line with the requirements used in my analysis. In order to avoid there being too few and insufficient observations for robust estimates for the border region, I include an extract from the employment register (%HVFKÁIWLJWHQKLVWRULN %H+) in some estimation versions 53 Chapter 3 Data and Basic Definitions which covers all observations that pertain to social insurance contributions in the eastern Bavarian border region (i.e. 100 % instead of 2 %). For all estimations I eliminate from the data apprentices, marginal part-time and part-time workers, homeworkers and all observations where information about education and/or professional status is missing. I concentrate on full-time employees, aged 16 to 65, who will have been employed for at least one year on June 30 (the reference date). I distinguish between the following three skill groups of workers (Table 3.1):13 Table 3.1: Classification of German skill groups Skill group Qualification low-skilled people with no occupational qualification regardless of the educational level reached, i.e. with or without a certificate of upper secondary education ($ELWXU) skilled people with an occupational qualification whether or not they have a certificate of upper secondary education ($ELWXU) high-skilled people with upper secondary education and a degree from a university or polytechnic According to the regulation regarding foreign commuters in German border regions ($QZHUEHVWRSSDXVQDKPHYHURUGQXQJ $6$9 1997, regulation on the granting of employment permits to foreigners), the Bavarian borderland consists of the eastern parts of the regions of 2EHUIUDQNHQ (Upper Franconia), Oberpfalz (Upper Palatinate) and 1LHGHUED\HUQ (Lower Bavaria) including the university towns of Bayreuth and Passau and the towns with polytechnics ()DFKKRFKVFKXOHQ) Hof, Weiden, Amberg and Deggendorf (16 districts and seven autonomous municipal authorities, Figure 3.1). In contrast to the border region in ASAV, § 6 para. 1, my analysis covers the city and district of Regensburg. Taking into account the differences between urban and rural areas I use the classification scheme of the Federal Office for Building and Regional Planning (%XQGHVDPWIÙU%DXZHVHQXQG5DXPRUGQXQJ%%5), which differentiates between regions with large agglomerations (BBR 1-4), regions with features of conurbation (BBR 5-7) and regions of rural character (BBR 8-9) (Table 3.2). As a control group I only use the observations from the remaining western German districts (without Berlin), since the inclusion of data from the eastern German states (which are available from 1992 onwards) would lead to biased results. This reduces the basic dataset to around 100,000 observations per year from 265 regions, some of which consist of aggregated districts. 13 The education variable in the IABS has shortcomings in the form of missing values and inconsistencies. Nevertheless, I use the original IABS data in my calculations and do not apply imputation procedures, which are proposed e.g. by Fitzenberger et al. (2006). IAB-Bibliothek 321 54 Labour Market Effects in the Bavarian Border Region Table 3.2: Regional classification scheme based on BBR classification Structural region type District type Description of district type (BBR) Regions with large agglomerations (basic type 1) BBR 1 Core cities BBR 2 Highly urbanised districts in regions with large agglomerations BBR 3 Urbanised districts in regions with large agglomerations BBR 4 Rural districts in regions with large agglomerations Regions with features of conurbation (basic type 2) BBR 5 Central cities in regions with intermediate agglomerations BBR 6 Urbanised districts in regions with intermediate agglomerations BBR 7 Rural districts in regions with intermediate agglomerations Regions of rural character (basic type 3) BBR 8 Urbanised districts in rural regions BBR 9 Rural districts in rural regions Source: Federal Office for Building and Regional Planning (%XQGHVDPWIÙU%DXZHVHQXQG5DXPRUGQXQJ%%5). Figure 3.1: Eastern Bavarian border region Eastern Bavarian Border Region District Types (BBR) 55 Chapter 3 The Labour Market in Eastern Bavaria: Some Descriptive Evidence 3.3 The Labour Market in Eastern Bavaria: Some Descriptive Evidence 3.3.1 Selected Figures of VALA A comparative study of labour markets in German federal states (9HUJOHLFKHQGH $QDO\VHYRQ/ÁQGHUDUEHLWVPÁUNWHQ9$/$) carried out by the regional research unit of the IAB investigates the differences between regional employment growth in Germany. In a separate study for eastern and western Germany the employment growth from 1993 until 2001 is broken down into several economic determinants using a two-step shift-share regression model. Besides the impact of the structure of economic activity, the firm size, the skill level and the wage level on the development of regional employment growth, a regional fixed effect is also identified, which accounts for idiosyncratic characteristics of the respective district that cannot be explained by the other variables. For instance, a particularly favourable combination of industrial sectors possibly leads to spillover effects with profits for the overall economy. Specific skills which are not represented by the formal qualification structure, the closure or settlement of large companies, the geographical location and not least the opening of a border can also play an important role for the economic perfor mance of a district. In short, it turns out that these locational effects are a very essential source of regional disparity. In Table 3.3 I present some outstanding results for some Bavarian districts, which are described in detail in Böhme/Eigenhüller (2005). Table 3.3: Locational fixed effects for Bavarian districts (as percentage points) Munich (city) –0.72 Schwandorf 2.01 Munich (district) 2.45 Cham 1.71 Nuremberg (city) –1.82 Passau (district) –0.67 Freising 5.28 Hof (district) 0.99 Regensburg (city) 2.14 Tirschenreuth –1.22 Regensburg (district) 2.66 Wunsiedel –1.86 Source: Böhme/Eigenhüller (2005). The overall locational effect for the Federal State of Bavaria is 0.53 percentage points. As in the rest of western Germany, the core cities in Bavaria, Munich and Nuremberg (both BBR1), also exhibit negative locational effects, an indication of IAB-Bibliothek 321 62 Labour Market Effects in the Bavarian Border Region 3.3.3 Relative Employment Share and Czech Employees in Eastern Bavaria Calculating the share of employees in eastern Bavaria shows the virtually continuous growth of the relative size of the economy in this region (Figure 3.6). While around 2.9 % of all workers in western Germany were employed in the eastern Bavarian border region in 1980, this proportion increased to 3.2 % by 2001. Only in the mid 1990s can a temporary decline be observed. Altogether this was not a substantial rise, but the figures show that the opening of the border did not stop the trend. Bearing in mind that German workers could possibly be substituted by Czechs, I only calculate the share for German workers, but, as the graph shows, the development proceeds analogously, i.e. the relative size of the economy in the border region also grew in this respect. According to the abovementioned $QZHUEHVWRSSDXVQDKPHYHURUGQXQJ (ASAV), Czech commuters, i.e. people who cross the border every day or work no more than two days per week in Bavaria, have a facilitated access to the Bavarian labour market compared to other German regions and to conventional migrants. The ASAV came into force in 1997. In the early 1990s it was even easier for Czechs to work in Bavaria. However, due to rising unemployment rates in Germany in the mid 1990s, residence and work permits were then only granted or renewed if no German worker could be found for the job. As a consequence the estimated number of Czech commuters working in Germany declined from 16,000 in 1993 to 5,000 in 2000 (Andrle/Dupal 1997; Prager Zeitung 2001). The employment figures for Czech workers in Bavaria developed correspondingly – apart from the trend in the hotel and catering industry with an obvious lack of qualified German personnel. Fig- Figure 3.6: Employment share of workers employed in eastern Bavaria (as %, 1980–2001) 3.4 % 3.3 % 3.2 % 3.1 % 3.0 % 2.9 % 1980 1983 1986 1989 1992 1995 1998 2001 share of all workers employed in eastern Bavaria share of German workers employed in eastern Bavaria Source: Author’s own calculations using the regional scientific use file of the IABS. 63 Chapter 3 The Labour Market in Eastern Bavaria: Some Descriptive Evidence ure 3.7 depicts the absolute number of Czech commuters in Bavaria, which shows that naturally most of them were employed in the Bavarian border region. Table 3.5: Czech commuters in the Bavarian employment office districts in 1995 employment office district total employment office district total Ansbach 0 Deggendorf 1,510 Aschaffenburg 0 Donauwoerth 1 Bamberg 4 Freising 0 Bayreuth 127 Ingolstadt 0 Coburg 3 Kempten 0 Hof 1,078 Landshut 6 Nuremberg 6 Memmingen 0 Regensburg 8 Munich 4 Schwandorf 1,806 Passau 2,063 Schweinfurt 0 Pfarrkirchen 4 Weiden 829 Rosenheim 0 Weissenburg 0 Traunstein 0 Wuerzburg 1 Weilheim 0 Augsburg 1 Bavaria 7,451 Source: IAB data base. Notes: Reference date: June 30, place of residence/nationality: Czech (Republic), place of work: Bavaria. Figure 3.7: Czech commuters in Bavaria covered by the German social security system (persons ‘000, 1992–2002) 10,000 9,000 8,000 7,000 6,000 5,000 4,000 3,000 1990 1992 1994 1996 1998 2000 2002 2004 Bavaria thereof border region Source: IAB data base. Notes: Reference date: June 30, place of residence/nationality: Czech (Republic), place of work: Bavaria. IAB-Bibliothek 321 64 Labour Market Effects in the Bavarian Border Region Table 3.5 shows, exemplified for the year 1995, that the Czech commuters were almost exclusively employed in five of the 27 Bavarian employment office districts overall, many of them in the building industry, the hotel and catering industry and the wholesale and retail industry. Concerning the qualifications of the commuters, many of them do not have an occupational qualification. This can be discerned by taking a look not only at the commuters, but at all Czech employees in Germany covered by the social security system Figure 3.8: Share of Czech employees in the eastern Bavarian workforce (as %, 1980–2001) 7 % 6 % 5 % 4 % 3 % 2 % 1 % 0 % 5.0 % 4.5 % 4.0 % 3.5 % 3.0 % 2.5 % 2.0 % 1.5 % 1.0 % 0.5 % 0.0 % 5.0 % 4.5 % 4.0 % 3.5 % 3.0 % 2.5 % 2.0 % 1.5 % 1.0 % 0.5 % 0.0 % 5.0 % 4.5 % 4.0 % 3.5 % 3.0 % 2.5 % 2.0 % 1.5 % 1.0 % 0.5 % 0.0 % 5.0 % 4.5 % 4.0 % 3.5 % 3.0 % 2.5 % 2.0 % 1.5 % 1.0 % 0.5 % 0.0 % 5.0 % 4.5 % 4.0 % 3.5 % 3.0 % 2.5 % 2.0 % 1.5 % 1.0 % 0.5 % 0.0 % (a) low-skilled, male (b) low-skilled, female (d) skilled, female(c) skilled, male (f) high-skilled, female(e) high-skilled, male 1980 1983 1986 1989 1992 1995 1998 2001 share of Czech workers 1980 1983 1986 1989 1992 1995 1998 2001 share of Czech workers 1980 1983 1986 1989 1992 1995 1998 2001 share of Czech workers 1980 1983 1986 1989 1992 1995 1998 2001 share of Czech workers 1980 1983 1986 1989 1992 1995 1998 2001 share of Czech workers 1980 1983 1986 1989 1992 1995 1998 2001 share of Czech workers Data source: Author’s own calculations with employment register %HVFKÁIWLJWHQ+LVWRULN%H+ 65 Chapter 3 The Labour Market in Eastern Bavaria: Some Descriptive Evidence (Figure 3.8). Using the employment register (%HVFKÁIWLJWHQ+LVWRULNBeH) I calculate the share of Czech workers among all workers in the eastern Bavarian border region by including those reported as Czechoslovak (due to the common state until 1992), Czech or Slovak (since a lot of Slovak people still live in the Czech Republic). Plausibly, the share was close to zero before the fall of the Iron Curtain. Only a small number of Czechoslovaks who are identified as highly skilled worked in the Bavarian border region. Not surprisingly, the picture changes at the beginning of the 1990s. More than 5 % of the eastern Bavarian male, low-skilled workforce was reported Czech (Czechoslovak, Slovak) in 1992 and 1993.14 Corresponding to the development of commuters, the figures decline in the mid 1990s. Regarding female, low-skilled and male, skilled workers the share did not increase to such a large extent, but still reaches 2 %. While only a small upswing in the share of female, skilled, Czech workers is observable, the proportion does not increase for high-skilled workers at all. 3.3.4 Skill Structure of Employed People15 In order to evaluate the shifts in the skill structure in eastern Bavaria, my analysis is motivated by the need to contrast the development in the border region with the average level in western Germany. The absolute deviation (as %) of skill group shares in the border region from the national averages is shown in Figure 3.9. It has to be emphasised that the shares of the single district types in the border and the non-border region are different. 14 To simplify matters from now on I use only the term “Czech”. 15 Parts of this chapter have been published in Moritz/Gröger (2007). Figure 3.9: Deviation of skill group shares for workers in the eastern Bavarian border region from the averages in western Germany (as %, 1980–2001) 8 % 6 % 4 % 2 % 0 % –2 % –4 % –6 % 1980 1983 1986 1989 1992 1995 1998 2001 low-skilled skilled high-skilled Source: Author’s own calculations using the regional scientific use file of the IABS. IAB-Bibliothek 321 66 Labour Market Effects in the Bavarian Border Region The share of low-skilled workers, which stood at 6.58 % above the national average in 1980, was virtually equal to the national level in 2001 (0.39 % above the national average). In contrast to this development I observe a sharp rise in the share of skilled employees in eastern Bavaria, which, starting from a disproportionately low value (–4.41 % below the average in 1980), reached the national level at the beginning of the 1990s and stood at 4.31 % above the average in 2001. The shortfall of high-skilled employees in the border region compared with the national level increased from –2.16 % (1980) to –4.70 % (2001). Though the growth rate of the high-skilled share is higher in the border region, the absolute difference between eastern Bavaria and the rest of western Germany increases due to the low base level. This result indicates the presence of β-convergence, i.e. the share in the region lagging behind grows faster than the share in the region being ahead in the beginning. But there is no tendency towards σ-convergence concerning highskilled employees, i.e. the standard deviation of the shares in the borderland and in the rest of the country is not decreasing over time.16 Taking the rather rural structure of eastern Bavaria into consideration and comparing the regional values only with district types 5-9 at national level results in a similar diagram (see Figure 3.10). From 1980 to 2001 the share of low-skilled workers in eastern Bavaria approximates the level of district types which are typical for this region. The share of skilled workers, which was distinctly below the average in 1980, exceeded the national level moderately in 2001. In contrast to the values in Figure 3.9, the deficit of high-skilled workers in eastern Bavaria compared 16 The concepts of β- and σ-convergence were introduced by Sala-i-Martin (1990). Figure 3.10: Deviation of skill group shares for workers in the eastern Bavarian border region from the averages in the district types 5-9 in western Germany (as %, 1980–2001) 1980 1983 1986 1989 1992 1995 1998 2001 low-skilled skilled high-skilled Source: Author’s own calculations using the regional scientific use file of the IABS. 6 % 4 % 2 % 0 % –2 % –4 % –6 % 67 Chapter 3 The Labour Market in Eastern Bavaria: Some Descriptive Evidence with the average level of the district types 5-9 increased only marginally. It seems that the growing negative deviation of the share of high-skilled workers in eastern Bavaria from the national level can be explained by the structure of centrality and population density. Border region effects resulting from the fall of the Iron Curtain are obviously not observable at the descriptive level. For an overview of absolute differences see Figure 3.11. 3.3.5 Skill Structure of Unemployed People The IAB Employment Sample (IABS) not only covers information about employment relationships, but also contains the periods in which people who were once employed were unemployed and received benefits, respectively. Though there are far fewer unemployment spells in the dataset, I also compare the skill group shares for unemployed people in eastern Bavaria with the shares in the district types 5-9 at the national level (Figure 3.12: for an overview of absolute differences see Figure 3.13). Hereby, in spite of there being no obvious structural break in the development of skill group shares of employed people, I explore whether there are inappropriate shifts in 17 The low value for high-skilled employees for 1996 can be traced back on the lack of registration of doctors in this year. Figure 3.11: Shares of skill groups for workers in eastern Bavaria and western Germany (as %, 1980–2001) 17 40 % 35 % 30 % 25 % 20 % 15 % 10 % (a) low-skilled 1975 1980 1985 1990 1995 2000 2005 Western Germany Eastern Bavaria Source: Author’s own calculations using the regional scientific use file of the IABS. (b) skilled 80 % 75 % 70 % 65 % 60 % 1975 1980 1985 1990 1995 2000 2005 Western Germany Eastern Bavaria (c) high-skilled 8 % 6 % 4 % 2 % 0 % 1975 1980 1985 1990 1995 2000 2005 Western Germany Eastern Bavaria IAB-Bibliothek 321 68 Labour Market Effects in the Bavarian Border Region the skill group shares of unemployed people. Naturally, due to the dependence on a former employment relationship, these shares are strongly correlated with the employment shares in former years. However, it is possible to investigate, for instance, whether the share of low-skilled unemployed in eastern Bavaria rose dramatically after the opening of the border or at least did not decrease further. In line with my hypothesis, the relative labour demand for low-skilled workers should above all fall in the border region. However, there is no evidence of above-average job losses of lowskilled workers in the Bavarian borderlands. Figure 3.12 corresponds to the results for employed people exhibiting the relatively falling share of low-skilled, and rising share of skilled workers in eastern Bavaria. Taking into consideration the fact that the share of low-skilled people generally shrinks over the years, it comes as no surprise that the share of unemployed low-skilled people also decreases from 1980 to 2001. Interestingly, while the nationwide share fluctuates around 35 % in the early and mid 1990s, the share in the border region diminishes continuously (Figure 3.13a). Figure 3.13b shows that the share of skilled unemployed people in the border region exceeds the nationwide reference proportion, while Figure 3.13c displays the constantly lower share of high-skilled unemployed people in the border region. Summarising the descriptive figures, there are no hints for particular border effects after the fall of the Iron Curtain. The labour market trends in eastern Bavaria proceed in the 1990s without a structural break. I do not find any indications that low-skilled employees in the border region lose out relatively speaking from increased trade relations with the Czech Republic. In the next subsection I analyse the qualification trends by means of an econometric model. First I estimate the influencing factors for the employees, and afterwards, although there are few observations in the dataset, I turn towards the unemployed people. Figure 3.12: Deviation of skill group shares for unemployed people in the eastern Bavarian border region from the averages in the district types 5-9 in western Germany (as %, 1980–2001) 1980 1983 1986 1989 1992 1995 1998 2001 low-skilled skilled high-skilled Source: Author’s own calculations using the regional scientific use file of the IABS. 15 % 10 % 5 % 0 % –5 % –10 % –15 % 69 Chapter 3 Econometric Analysis of Qualification Trends 3.4 Econometric Analysis of Qualification Trends 3.4.1 Employed People18 In order to investigate the changes in the skill structure of employees I use the following econometric approach: for each of the 265 regions I calculate the shares of the skill groups for each year so that I obtain a “balanced panel” from 1980 to 2001. Then I split the dataset according to the skill groups and estimate three equations using the standard OLS method, i.e. I have 5,830 observations in each of the estimations (265 regions times 21 years). 6+$5(rt = α + β %$<(51r + γ 1 DRTYP2r + γ 2 DRTYP3r + δ BORREG r + λ 1 75(1't + λ 2 75(1'B BORREG rt + λ 3 75(1'B BAYERN rt (3.3) + λ 4 75(1'B'57<3rt + λ 5 75(1'B'57<3rt + τ 1 75(1'23(1%t + τ 2 75(1'23(1%B BORREG rt + ε rt 18 Parts of this chapter have been published in Moritz/Gröger (2007). Figure 3.13: Shares of skill groups for unemployed people in the district types 5-9 in eastern Bavaria and western Germany (as %, 1980–2001) 60 % 55 % 50 % 45 % 40 % 35 % 30 % 25 % 20 % (a) low-skilled 1980 1983 1986 1989 1992 1995 1998 2001 Western Germany Eastern Bavaria Source: Author’s own calculations using the regional scientific use file of the IABS. 75 % 70 % 65 % 60 % 55 % 50 % 45 % 40 % (b) skilled 1980 1983 1986 1989 1992 1995 1998 2001 Western Germany Eastern Bavaria 4.5 % 4.0 % 3.5 % 3.0 % 2.5 % 2.0 % 1.5 % 1.0 % 0.5 % 0.0 % (c) high-skilled 1980 1983 1986 1989 1992 1995 1998 2001 Western Germany Eastern Bavaria IAB-Bibliothek 321 70 Labour Market Effects in the Bavarian Border Region 6+$5(rt stands as a placeholder for the share of low-skilled (/2:6.rt ), skilled (6.,//('rt) and high-skilled (+,*+6.rt) workers in region r in year t (as %). My estimation approach includes (0.1) dummy variables for the Federal State of Bavaria (%$<(51), since the proportion of school leavers with an upper secondary education in Bavaria is significantly smaller than the proportions in other federal states. Furthermore, I include dummy variables for the district types (DRTYP2: district types 5-6; DRTYP3: district types 3-4 & 7-9) and the eastern Bavarian border region (%255(*). Using 75(1' and 75(1'23(1% I control for temporal trends. 75(1'= 1,…22 if 1 ≤ t ≤ 22, 75(1'23(1%= 0 if t ≤ 10 (until 1989) and 75(1'23(1%= 1,…12 if 11 ≤ t ≤ 22 (since 1990). In addition, interaction variables of 75(1' and the dummy variables are included. The interaction term of 75(1'23(1% and %255(* represents the most important variable in the estimation. The coefficient of this variable provides information about the changes in the skill differentials in eastern Bavaria after the opening of the border. According to my hypothesis I expect negative values for low-skilled employees and positive values for skilled and high-skilled workers, i.e. low-skilled workers in eastern Bavaria should lose compared to low-skilled workers in the rest of the country after the fall of the Iron Curtain, more highly skilled workers in the border region should benefit by the opening of the border. Table 3.6 contains the results of the estimates. The adjusted R2 ranges between 0.32 for skilled workers and 0.44 for low-skilled ones. The signs of the coefficients for the dummy variables correspond with my theoretical expectations: positive values in the estimation with low-skilled workers, negative signs in the cases of more highly skilled workers – with the exception of DRTYP3 in the regression for skilled workers and %255(* in the regression for high-skilled workers. The coefficient for 75(1' being significantly negative for low-skilled workers and significantly positive for skilled and high-skilled workers shows a general trend towards higher qualifications. The results for the interaction terms of 75(1' and the dummy variables indicate the trend towards decreasing shares of low-skilled workers in the border region, in general in Bavaria and in more peripheral areas. In accordance with these estimates, the coefficients for the interaction variables exhibit a positive effect in the estimation for skilled employees. In the case of high-skilled workers I observe differentiated results. 71 Chapter 3 Econometric Analysis of Qualification Trends Table 3.6: Estimation results for the share of low-skilled, skilled and high-skilled workers low-skilled skilled high-skilled Variable Coef. Std.Err. Coef. Std.Err. Coef. Std.Err. BAYERN 0.0251*** 0.0035 –0.0234*** 0.0032 –0.0017 0.0022 DRTYP2 0.0110*** 0.0037 –0.0037 0.0034 –0.0073*** 0.0016 DRTYP3 0.0055 0.0034 0.0124*** 0.0032 –0.0180*** 0.0017 BORREG 0.0401*** 0.0099 –0.0415*** 0.0090 0.0014 0.0023 TREND 0.0081*** 0.0003 0.0057*** 0.0003 0.0025*** 0.0002 TREND_BORREG –0.0016 0.0013 0.0019 0.0012 –0.0003 0.0003 TREND_BAYERN –0.0007*** 0.0002 0.0001 0.0002 0.0006*** 0.0002 TREND_DRTYP2 –0.0006** 0.0003 0.0013*** 0.0002 –0.0007*** 0.0002 TREND_DRTYP3 –0.0005** 0.0002 0.0021*** 0.0002 –0.0015*** 0.0002 TRENDOPENB 0.0040*** 0.0004 –0.0047*** 0.0004 0.0007*** 0.0002 TRENDOPENB_BORREG –0.0008 0.0019 0.0006 0.0016 0.0002 0.0005 Constant 0.2941*** 0.0026 0.6685*** 0.0023 0.0374*** 0.0012 Test statistics Number of observations 5,830 5,830 5,830 R-squared 0.4425 0.3229 0.3681 Root MSE 0.0471 0.0450 0.0278 Source: Author’s own calculations using the regional scientific use file of the IABS. Notes: Regression with heteroskedasticity-robust standard errors; */**/*** significant at the 10/5/1 percent level. What is more striking and even stands in contrast to the theoretical predictions is the significantly positive sign of the coefficient for 75(1'23(1% for low-skilled workers and the significantly negative value for skilled employees. This suggests a trend towards low-skilled labour in the years after the opening of the border. The coefficient of the variable in the focus of my analysis, 75(1'23(1%B%255(*, lies close to 0 and is not statistically significant in all cases. Therefore, I am not able to identify an effect of the opening of the border on the skill structure in eastern Bavaria. Though I observe an upward movement in the skill structure in general, I find no evidence of the theoretically predicted increase in skilled labour in the region under review, which is in line with the descriptive results.19 19 The estimation was also carried out using Seemingly Unrelated Regression (SUR) methodology, yielding no fundamental changes in the results. IAB-Bibliothek 321 78 Labour Market Effects in the Bavarian Border Region Figure 3.14 shows that the wage differential of low-skilled workers in the border region oscillated around –3% until the mid 1990s and then deepened to about –5 % in the year 2000. This means that on average low-skilled workers in eastern Bavaria initially earned approximately 3 % less than their counterparts in the rest of western Germany. From 1994 onwards the wage gap tended to grow to 5 %. Though I observe that the wages in the border region were significantly below the wages at national level in the whole period under review, the trend line does not indicate a significant widening of the wage differential after the opening of the border. For this purpose, the upper confidence bound in the 1990s had to drift below the 1980s’ values of the lower confidence bound. The results for skilled and high-skilled employees are shown in figures 3.15 and 3.16. During the period under observation the wage differential of skilled workers narrowed from nearly –5 % in the early 1980s to about –3 % in the late 1990s, i.e. a general catching-up process in favour of skilled employees took place in eastern Bavaria in the 1980s and 1990s, despite setbacks in some years (Figure 3.15). The coefficient for %255(* declined noticeably in 1992 and 1993, but in general no consistent open-border effect is identifiable. Regarding high-skilled employees in eastern Bavaria I observe similar results but on a larger scale (Figure 3.16). Due to there being far fewer people with higher education in the dataset, the results vary to a greater extent than in estimations with low-skilled or skilled employees. While the wage differential of people employed in the border region was permanently negative during the 1980s, positive values have appeared sporadically since 1991. As in the estimation for skilled employees I do not observe an effect caused by the opening of the border. 6HQVLWLYLW\$QDO\VHV:HDNO\$QRQ\PRXV,$%69HUVLRQDQG%H+([WUDFW In order to check the robustness of the results in the previous subsection I apply some sensitivity analyses. To do this I use the weakly anonymous original IABS version. In contrast to the scientific use file this version contains information about the firm size and the country of origin (not only divided into natives and foreigners). Besides, the branches of economic activity are not aggregated into 16 industries, but are available on the basis of a 3-digit code (WZ 73). I transform this very detailed classification into a scheme of 28 industries. Since the eastern Bavarian border region is rather small and the IABS provides only few observations for some years and skill groups, I use all observations from the employment register (Be- VFKÁIWLJWHQ+LVWRULN%H+) for the border region and combine this extract with the weakly anonymous IABS version. In order to avoid a bias in the results I weight the observations adequately according to their representativeness and estimate wage 79 Chapter 3 Econometric Analysis of Wage Differentials equations. Contrary to the regressions in the previous subchapter I do not now estimate a Tobit model, but use wages that, in the case of censoring, are imputed on the basis of Tobit estimates of the distribution parameters (Gartner 2005), so that I can then apply the standard OLS method. The results from the larger dataset do not differ in principle from the outcomes using only the weakly anonymous IABS, but are naturally less fluctuating. Thus, I only present the results obtained using the BeH extract. Including the firm size in the equation as an exogenous variable does not yield a substantial change in the results. Since the effect of this variable is far from being significant, I remove it from the regression. In an alternative specification I split the skill groups according to sex and include a dummy variable for nationality ()25(,*1), which equals one if the employee is not German. Moreover, in order to control for the Czech commuters who are allowed to work in the Bavarian borderlands, I interact the nationality dummy with the border region dummy (%255(*B )25(,*1). The coefficient of this variable measures the wage effect of foreigners in the border region relative to foreigners in the rest of the country. As in the previous regressions the estimation results of the control variables are sensible. The coefficient of the additional variable controlling for foreign workers exhibits negative values for all male skill groups and for low-skilled female workers during the observation period. This means that foreigners belonging to these groups earn less than the relevant domestic German workers, notably evident for male skilled employees with a rather constant wage differential for foreigners of –16 % to –11 %. Positive wage differentials for non-Germans are identified for female skilled workers in the 1990s and in some years for female high-skilled workers. Regarding the interaction term between foreign workers and the border region I return to this variable below. Due to the highly divergent number of observations in the skill and sex groups, the scaling of the results for %255(* in Figure 3.17 varies. In the case of domestic low-skilled workers, an interesting difference between males (Figure 3.17a) and females (Figure 3.17b) is observable. In the 1980s male low-skilled employees in the eastern Bavarian border region earn about 5 % less than their colleagues in the non-borderlands. This wage differential considerably narrows at the beginning of the 1990s to around –3 %. Until the end of the decade the wage gap widens again, approaching the original level of –5 %. In contrast to this surprising development there is no similar trend for low-skilled female workers in the border region. Starting from a far smaller wage gap of approximately –2 %, the difference increases to about –4 % in the 1980s. This level stabilises during the 1990s, with only one negative outlier in 1999. IAB-Bibliothek 321 80 Labour Market Effects in the Bavarian Border Region Regarding skilled employees, the general catching-up process in eastern Bavaria, which the results of the regression with the scientific use file already imply, is clearly confirmed in my further estimates, but interestingly, again only for male employees (Figure 3.17c). With earnings of nearly 5.5 % less in 1980, the differential becomes continuously smaller until the mid 1990s, where it settles at a level of about 3 %. Then, similarly to the trend for low-skilled male workers, the graph Figure 3.17: Gender-specific wage effect for German employees in eastern Bavaria (as %, 1980–2001) Source: Author’s own calculations using the weakly anonymous version of the IABS and BeH extract. Notes: In the case of censoring, wages are calculated in the framework of an imputation procedure using the Tobit estimation method; regression with heteroskedasticity-robust standard errors. 1980 1983 1986 1989 1992 1995 1998 2001 BORREG upper 95 % CI limit lower 95 % CI limit –1 % –2 % –3 % –4 % –5 % –6 % –7 % (a) low-skilled, male 1980 1983 1986 1989 1992 1995 1998 2001 BORREG upper 95 % CI limit lower 95 % CI limit 1 % 0 % –1 % –2 % –3 % –4 % –5 % –6 % –7 % –8 % (b) low-skilled, female –2.0 % –2.5 % –3.0 % –3.5 % –4.0 % –4.5 % –5.0 % –5.5 % –6.0 % (c) skilled, male 1980 1983 1986 1989 1992 1995 1998 2001 BORREG upper 95 % CI limit lower 95 % CI limit 0.0 % –0.5 % –1.0 % –1.5 % –2.0 % –2.5 % –3.0 % –3.5 % –4.0 % –4.5 % (d) skilled, female 1980 1983 1986 1989 1992 1995 1998 2001 BORREG upper 95 % CI limit lower 95 % CI limit 1980 1983 1986 1989 1992 1995 1998 2001 BORREG upper 95 % CI limit lower 95 % CI limit 4 % 3 % 2 % 1 % 0 % –1 % –2 % –3 % –4 % (e) high-skilled, male 1980 1983 1986 1989 1992 1995 1998 2001 BORREG upper 95 % CI limit lower 95 % CI limit 15 % 10 % 5 % 0 % –5 % –10 % –15 % –20 % (f) high-skilled, female 81 Chapter 3 Econometric Analysis of Wage Differentials turns to a deeper wage gap of around 4 % in 1999 with a slight recovery in the following two years. For female skilled workers the picture is completely different again (Figure 3.17d). Relative wage losses for employees in the borderlands in the 1980s (from –2 % to –3 %) are succeeded by a reduction in losses in the 1990s (from –3 % to –2 %). As far as high-skilled workers are concerned, the distinctively smaller number of observations still poses a problem (apparent in the figures through the huge confidential bounds). Male high-skilled workers in eastern Bavaria (Figure 3.17e) earn about 1 % below average until the mid 1980s. In the following ten years the wage differential is positive and then oscillates around the 0 % line. Female highskilled workers in the border region obviously catch up in the 1980s from belowaverage to above-average values. In the 1990s the wage level stabilises around the reference value for western Germany (Figure 3.17f). The question arising is which forces are behind the relative wage gains for male low-skilled workers. One possible explanation could lie in the strengthened position of low-skilled employees in supermarkets and hypermarkets. These stores did not exist in the early years after the fall of Communism in the Czech Republic and, like in Poland (Ullmann 2006), were largely established as recently as the second half of the 1990s. Masses of Czech consumers flocked to the nearby eastern Bavarian stores in these years in order to satisfy their demand for western-type consumer goods. Although the relative employment share in the retail industry grew to a smaller extent by an above-average rate in the border region in the early 1990s (see Figure 3.2), this did not have a resounding effect on low-skilled workers’ wages in eastern Bavaria. Running the regression for low-skilled male workers without those employed in the retail and wholesale industries does not change the result essentially. Another reason for the catching-up of low-skilled workers, which is already ruled out here by the specification of the regression model, could be the more productive Czech commuters who potentially substituted less productive German workers at the beginning of the 1990s. Since I control for foreign workers, this cannot cause the positive change in the wage differential. But, as it turns out, it pays off to take a closer look at Czech employees in the Bavarian borderlands. First of all, I explore the outcomes for the interaction term %255(*B)25(,*1 (Figure 3.18). As mentioned above, this variable controls for all foreign workers in eastern Bavaria, measuring their relative wage position compared to all foreign workers in the rest of western Germany. An alternative explanation is the additional wage differential for foreign employees compared to German employees in the border region. IAB-Bibliothek 321 82 Labour Market Effects in the Bavarian Border Region Disregarding high-skilled employees, there are evidently significant changes for the other skill groups after the opening of the border. Slightly positive additional wage differentials for foreign workers in eastern Bavaria in the 1980s turn into negative values of down to –10 % in the early years after the fall of the Iron Curtain. After these drastic changes, which are mostly apparent for low-skilled and skilled male Figure 3.1 8: Additional wage effect for foreign employees in eastern Bavaria (as %, 1980–2001) 10 % 5 % 0 % –5 % –10 % –15 % 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 6 % 4 % 2 % 0 % –2 % –4 % –6 % –8 % –10 % –12 % Source: Author’s own calculations using the weakly anonymous version of the IABS and BeH extract. Notes: In the case of censoring, wages are calculated in the framework of an imputation procedure using the Tobit estimation method; regression with heteroskedasticity-robust standard errors. (a) low-skilled, male 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 6 % 4 % 2 % 0 % –2 % –4 % –6 % –8 % –10 % –12 % (b) low-skilled, female 10 % 5 % 0 % –5 % –10 % –15 % (c) skilled, male (d) skilled, female 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 20 % 15 % 10 % 5 % 0 % –5 % –10 % –15 % –20 % (e) high-skilled, male 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 60 % 50 % 40 % 30 % 20 % 10 % 0 % –10 % –20 % –30 % –40 % –50 % (f) high-skilled, female 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 83 Chapter 3 Econometric Analysis of Wage Differentials workers (Figures 3.18a and 3.18c), the wage differentials stabilise in the mid 1990s, or the trend even reverses, as in the case of low-skilled male workers. The first suggestion with regard to the reason for these relative wage losses of foreigners in eastern Bavaria is that Czech commuters, for whom it was relatively easy to become employed in the Bavarian borderlands at the beginning of the 1990s, caused the drop in relative wages. Understandably, for Czech workers it was highly appealing to work in Bavaria and earn the multiple higher wage, though it was less than the average wage of other foreigners. Moreover, Czech commuters had a particular incentive for cross-border jobs, since they could keep their homes in the Czech Republic and take advantage of the lower cost of living there. Recalling the shares of Czech employees in the different skill groups in eastern Bavaria (Figure 3.8), the essential point so far is that the share of Czech workers in the border region is considerably high for precisely those skill groups where I noticed a fall in the relative wages for foreigners after 1990. In order to check definitively whether Czech workers caused the relative wage drop, I estimate the wage differentials excluding Czechs. The revealing results are presented in Figure 3.19. The graphs for the skill groups, in which the share of Czech workers did not exceed 1 % in the 1990s (female skilled and both sexes of high-skilled workers), do not change substantially. For the skill groups which exhibit a perceptible increase in the share of Czech employees, however, the graph changes fundamentally if Czechs are excluded from the estimation. For male low-skilled and skilled workers (Figures 3.19a and 3.19c), as well as for low-skilled female workers (Figure 3.19b) – to a smaller degree –, the drop in the relative wage for foreigners estimated above now disappears. Summarising the sensitivity analyses, the estimates with the larger datasets shed light on differences between male and female workers. For male low-skilled and skilled workers, who represent more than 60 % of the eastern Bavarian workforce, a catching-up process can be verified until the mid 1990s. From 1995 onwards the trend reverses and workers in the border region lose out in relation to employees in the rest of western Germany. Neither the Czech consumers shopping in the supermarkets of the Bavarian borderlands nor the substitution of less productive German workers with more productive Czech commuters can account for the relative wage gains in the early 1990s. Neither can the cause of the growing wage gap in the late 1990s be explained by the available data. Interestingly, Czech workers employed in eastern Bavaria bear the responsibility for the drop in the wage differential for foreign workers in the border region. However, the estimation methods applied up to now control neither for selection biases nor unobserved heterogeneity, which leaves enough scope for research to be done in the next subsections. IAB-Bibliothek 321 84 Labour Market Effects in the Bavarian Border Region Figure 3.19: Additional wage effect for non-Czech foreign employees in eastern Bavaria (as %, 1980–2001) 10 % 5 % 0 % –5 % –10 % –15 % 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 6 % 4 % 2 % 0 % –2 % –4 % –6 % –8 % –10 % –12 % (a) low-skilled, male 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 8 % 6 % 4 % 2 % 0 % –2 % –4 % –6 % –8 % –10 % –12 % (b) low-skilled, female 10 % 5 % 0 % –5 % –10 % –15 % (c) skilled, male (d) skilled, female 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 20 % 15 % 10 % 5 % 0 % –5 % –10 % –15 % –20 % (e) high-skilled, male 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 60 % 50 % 40 % 30 % 20 % 10 % 0 % –10 % –20 % –30 % –40 % –50 % (f) high-skilled, female 1980 1983 1986 1989 1992 1995 1998 2001 BORREG_FOREIGN upper 95 % CI limit lower 95 % CI limit 3URSHQVLW\6FRUH0DWFKLQJ The disadvantage of the applied estimation method lies in the assumption of a certain functional specification in the regression equation. An evaluation of economic relationships based on a dummy variable has the drawback that only differences in levels are captured. Otherwise, additional interaction terms would have to be incor- Source: Author’s own calculations using the weakly anonymous version of the IABS and BeH extract. Notes: In the case of censoring, wages are calculated in the framework of an imputation procedure using the Tobit estimation method; regression with heteroskedasticity-robust standard errors. 85 Chapter 3 Econometric Analysis of Wage Differentials porated into the estimation equation, making the interpretation of the results quite difficult. Besides this, large differences in the range of variable values between observations in the border region and the interior region could pose a problem. This would be the case if employees in eastern Bavaria were a specific subgroup of all employees in western Germany, e.g. caused by a lower infrastructural endowment in the border region. If employees in eastern Bavaria differed from workers in the non-border region and if people with certain characteristics tended to live in the border or non-border region, then their relative wage would have risen or fallen, independently of whether they worked in the (non-)border region. These obstacles can be mitigated by using a matching approach (Rubin 1974), which is particularly applied for the evaluation of labour market policies (e.g. Heckman et al. 1999). In order to evaluate the wage effect for workers in the border region I calculate the DYHUDJHWUHDWPHQWHIIHFWRQWKHWUHDWHG (att), which – relating to border region workers – is defined as the difference between the expected outcome for having a job in the borderland and having a job in the nonborderland. Δ att = ( Δ | BORREG = 1) (OQ:$*(1 | BORREG = 1) t(OQ:$*(0 | BORREG = 1) (3.6) The fundamental evaluation consists in the hypothetical outcome of (OQ:$*(0 | BORREG = 1) , i.e. the potential wage for border region workers in the non-borderland which is not observable. Naturally, it is only possible to observe one status, i.e. a person has a job either in the border region or in the non-border region. In my case, I can only observe the wage of an individual in one region and not the counterfactual outcome in the other region. The wage of non-border region workers in the non-borderland (OQ:$*(0 | BORREG = 0) is observable. However, people in the two areas of observation could differ systematically, i.e. as it turns out, workers in the non-borderland have greater potential experience on average, which in turn raises the probability of a higher wage. Therefore, (OQ:$*(0 | BORREG = 1) has to be estimated. For this purpose the group of border region employees is contrasted with a group of workers not working in the border region, which is preferably similar regarding relevant variables and is therefore called “statistical twins”. Basically, all variables which have an effect on the state of working in the border region and on the wage are relevant. The potential wage for working in the non-borderland has to be equal for border region and non-border region employees OQ:$*(0 ⊥ BORREG or (OQ:$*(0 | BORREG = 1) = (OQ:$*(0 | BORREG = 0) respectively, i.e. the &RQGLWLRQDO,QGHSHQGHQFH$VVXPSWLRQ (&,$) is fulfilled and the non-border region employees serve as an adequate IAB-Bibliothek 321 86 Labour Market Effects in the Bavarian Border Region control group (Rubin 1977). In other words, it is assumed that all influencing factors are included in the analysis and thus wage differentials can be traced back exclusively to the region. A second necessary requirement consists in the CRPPRQ6XSSRUW$VVXPSWLRQ, i.e. employees with the same X values have a positive probability of working either in the border region or the non-border region. Since it is not possible to condition on all relevant observed covariates X, Rosenbaum/Rubin (1983) suggest the use of a balancing score b(X), i.e. the conditional distribution of X given b(X) is independent of belonging to the treatment group (%255(* ) or the control group (%255(* ), that is X ⊥ BORREG | b(X) (3.7) As a further condition I assume that the wage of a person working in the border region is not affected by other people’s state of working in the borderland or the non-borderland (6WDEOH8QLW7UHDWPHQW9DOXH$VVXPSWLRQ6879$). Admittedly, applying a matching approach is problematic, insofar as no real assignment to treatment (“working in the border region”) can be identified. I assume that the opening of the border had a particular effect on workers employed in an area of up to 70 km from the Czech border, considering that it is easier in this area to outsource production activities or purchase goods and services in the Czech Republic. I am not able to manipulate the treatment “working in the border region”, since only workers who are either employed in eastern Bavaria (and mostly already were so before the fall of the Iron Curtain) or hold a job in the non-border region are included. The condition that there is no unobserved influencing factor concerning the variables involved in constituting the control group is questionable. The qualitative education level, motivation or career planning could possibly determine whether or not a person is living in the borderland. Consequently, I cannot assign the treatment clearly, because “working in the border region” is dependent on several unobservable individual characteristics. Hence, it can be argued whether causal conclusions are possible and whether the quantification of a causal effect seems appropriate. Nevertheless, I am able to examine changes in wage differentials before and after the opening of the border and I regard the application of the matching approach as a complement to the abovementioned conventional regression method. I apply “propensity score matching” as a matching method, i.e. I match people from the treatment group (“border-region workers”) with people from the comparison group (“non-border-region workers”). The propensity score is one of several possible implementations of the above-mentioned balancing score, in this case the 87 Chapter 3 Econometric Analysis of Wage Differentials probability of “working in the border region” given observed characteristics X. This probability is calculated by probit estimation with “working in the border region” as an endogenous variable. Considering the results of the previous subsection I assume that the “foreigners in the border region” group is fundamentally influenced by Czech commuters, for which I am not able to find equivalent counterparts in the non-border region. Hence, I decided to restrict the matching analysis to German employees. The control group is identified using the nearest neighbour algorithm with replacement on the basis of the variables experience, district type and sector of economic activity (primary, secondary or tertiary sector).21 Since border region employees are highly overrepresented in the dataset due to the observations from the BeH, every worker from the border region is matched with 20 workers from the non-border region who exhibit the most similar propensity score. As the matching is successful, the means of the exogenous variables do not differ significantly between the employees in eastern Bavaria and the rest of western Germany. This is important with respect to the district types, for instance, since district type 5 is in the border region only represented by the city of Regensburg. Using imputed wages again I estimate the DYHUDJHWUHDWPHQWHIIHFWRQWKHWUHDWHG (att) for every year between 1980 and 2001. As mentioned above, this relevant parameter is defined as the difference of the average wage of the border region workers and the average wage of the matched twins. Figure 3.20 shows the results of the DYHUDJHWUHDWPHQWHIIHFW (att) for the different skill and sex groups. The outcome for male workers in the border region corresponds to my previous findings: a narrowing of the wage gap in the first half of the 1990s for low-skilled and skilled employees in eastern Bavaria, whereas in the late 1990s the wage differentials rise again (Figures 3.20a and 3.20c). According to this matching analysis, high-skilled workers in the border region earn below-average wages before the opening of the border and enjoy a positive wage differential in the 1990s – a result that is in line with my theoretical expectations (Figure 3.20e). Regarding the analysis for female workers, the outcome deviates somewhat from the regression estimates: low-skilled employees now exhibit rising relative wages in the borderland (Figure 3.20b) in the 1990s, contradicting the continuous downward trend in the regression analysis. The outcome for the other two skill groups is however consistent with the preceding results, with growing relative wages for skilled women in eastern Bavaria (Figure 3.20d) and not very informative values for high-skilled female employees due to fewer observations (Figure 3.20f). 21 Using the classification of 28 industries instead of three sectors leads to poor matching results. IAB-Bibliothek 321 94 Labour Market Effects in the Bavarian Border Region dividuals, the relative wages in the eastern Bavarian border region are decreasing, though not significantly everywhere. It is interesting that the catching-up process for male low-skilled and skilled workers is also confirmed using this estimation method. However, the trend from the mid 1990s onwards seems generally unfavourable for eastern Bavaria. $GGLWLRQDO$QDO\VLV6SOLWWLQJWKH%RUGHU5HJLRQ In an additional variant of the estimation I subdivide the eastern Bavarian border region into two subgroups: the southern and the northern part (see appendix, Figure A 3.1). The southern part embraces the border region of Lower Bavaria (1LHGHUED\HUQ) and the southern districts of the Upper Palatinate (Oberpfalz), i.e. the districts and cities along the river Danube, the Bavarian forest and the districts of Cham and Schwandorf, which belong to the catchment area of Regensburg. The districts of Upper Franconia (2EHUIUDQNHQ) and the northern districts of the Upper Palatinate form the northern part of the border region. This subdivision is motivated by indications that the development in the two parts of the border region differs. Unemployment rates, for instance, are traditionally higher in the northern border region. In part, the discrepancies between the north and the south are obvious from the VALA results (see chapter 3.3.1 and Böhme/Eigenhüller 2005). Locational fixed effects are mostly negative in the northern districts, while the economic situation looks more favourable in the southern part of the region. As in the previous subchapter I use the fixed effects estimation approach controlling for unobserved heterogeneity. The results are shown in the appendix in Tables A 3.1 and A 3.2 and Figures A 3.2 and A 3.3. There are noticeable differences between the two parts of the border region. First of all, as shown in the tables, the basic wage differentials in the year 1980 in northeastern Bavaria are positive for female low-skilled and skilled workers, but not significant. The high significance of the wage gap of male skilled and high-skilled employees disappears if only the northern part is considered. Regarding the changes in the 1980s and 1990s, which are depicted in the figures, to some extent a South-North divide becomes apparent. While the break in the catching-up process is observable for male lowskilled and skilled workers in both parts, the size of the decline is much larger in the north. There, compared to around 1–2 percentage points in the south, the employees in the skill groups mentioned relatively lose about 4 percentage points until 2001. The same descent from 1995 onwards is evident for female low-skilled workers in the northern border region. Interestingly, female skilled employees continuously lose in the south, while the wage differential is kept relatively constant in northeastern Bavaria suggesting even an overall positive wage compared to the rest of western Germany. Concerning high-skilled workers, the widening of 95 Chapter 3 Conclusion the wage gap seems to be more distinctive in the north, with obviously incisive relative losses for females. All in all, these results are not surprising. As mentioned above, the northeastern border region faces structural adjustment problems to a far higher degree than the prosperous cities along the Danube. However, the results indicate that the integration effects are partly superposed by the fundamental development of the districts. While the situation in the southern part of eastern Bavaria seems quite acceptable, the position in the northeastern borderland gives reason for concern. 3.6 Conclusion The aim of this chapter was to analyse the relative effects of the fall of the Iron Curtain on the eastern Bavarian borderland with respect to shifts in the structure of employment, skills and wages. To do this, I compared the situation on the labour market before the opening of the border in the 1980s to the years afterwards until 2001. Both theoretical strands on which I based the analyses predict integration effects which should be larger in the border region. However, concerning labour demand, the Feenstra-Hanson trade model and the Brülhart et al. NEG model come to opposing conclusions in some aspects. While both theories expect high-skilled employees to profit especially in eastern Bavaria due to comparative advantages and a higher market potential respectively, the situation of low-skilled workers is seen differently. On the one hand, according to Feenstra/Hanson, international trade and outsourcing hits the low-skilled workforce in the border region to a higher degree, leading to relative wage losses. On the other hand, according to the NEG model, the higher attractiveness of the region after the abolition of trade impediments could also benefit low-skilled employees. The results of my analyses are the following: with respect to the structural change and specialisation in the border region, the opening of the German-Czech border did not change trends that had already been under way in the years before. The structural change in eastern Bavaria proceeds analogously to the development in the rest of the country. Caused by the continuous decline of the consumer goods industry, the economic structure in eastern Bavaria conforms to the western German pattern. Regarding the relative employment share, the development in eastern Bavaria is positive overall: the share of the workforce in Germany which is employed in the border region increased continuously before and after the introduction of free trade with the Czech Republic and outsourcing to it. This outcome holds for only German workers observed, not only where all workers are considered. IAB-Bibliothek 321 96 Labour Market Effects in the Bavarian Border Region Concerning the skill structure of employed and unemployed people, I do not find clear evidence of disproportionate shifts in the descriptive figures or in the estimates of an econometric model. The general trend towards more skilled labour also led to substantial shifts in the employment structure in the Bavarian border region. In the period under review, from 1980 to 2001, the share of low-skilled workers, which was substantially larger in eastern Bavaria in 1980, adjusted to the average value of the districts comparable with the border region observed. In accordance with this development, the share of skilled workers in the border region increased up to the average of the comparable districts. The share of high-skilled employees remained just below the aggregate level. The regression results correspond to these descriptive figures in the sense that there is no evidence of a special skill-upgrading process after the opening of the border. Having a look at the unemployment spells in the IABS does not change this picture. A large part of my study deals with wage differentials between eastern Bavaria and the rest of western Germany. I applied several estimation methods in order to identify the effects of the open border. While using the IABS scientific use file and estimating a Tobit model yields no significant results of relative wage losses for low-skilled workers and gains for more highly skilled workers, advanced methods and a larger dataset shed more light on the changes. In the former case there are no significant indications that workers in eastern Bavaria would profit or lose from integration with the nearby Czech market. Despite a slight catching-up trend, i.e. a decrease in the wage gap, which started as early as in the 1980s, I cannot identify any positive effects for skilled and high-skilled workers as a result of integration. The wage differential of low-skilled employees seems to have grown in the late 1990s, but the results are not significant. This prima facie impression suggests that the opening of the border did not have a profound impact on the districts situated immediately on the border. Adding additional observations for the border region from the employment register makes it possible to take a closer look. Using imputed wages and separating male and female employees, I apply several estimation methods (standard OLS, propensity score matching, DID and fixed effects). The outcome is very robust, at least in some cases: male low-skilled and skilled workers in the borderland caught up until the mid 1990s and then lost ground compared to employees in the rest of western Germany. The findings for all female skill groups and also male high-skilled employees are not so clear-cut. The results of the fixed effects estimation are most striking: for all sex and skill groups the wage gap widens in the observation period. The question remains whether or not the opening of the border caused this development, and if so, why different skill and sex groups are affected in an unequal manner. After all, there is no change for the worse immediately after the opening of the border: indeed, the indications are 97 Chapter 3 Conclusion actually quite positive in the early 1990s. Interestingly, precisely in this period it was relatively easy for Czech commuters to obtain a work permit in Bavaria. The catching-up process above all of male low-skilled employees after the fall of the Iron Curtain is not caused by the substitution of German workers with more productive Czech workers (nor by the surge of supermarkets in the borderland). In fact, Czech employees in eastern Bavaria earned wages substantially below the average in eastern Bavaria. However, from 1995 or so onwards, the development of wages in the border region gives cause for concern. In short, there are no profound border effects of economic integration observable apart from the wage differentials. With respect to the theoretical models, the relative losses for low-skilled workers, which are detected by using the larger dataset, are in line with the Feenstra-Hanson model. International trade and outsourcing apparently took their toll in the border region. So far, there are no obvious positive effects due to a higher market potential predicted on the basis of Brülhart et al. The results prompt the ongoing research on this topic. Since the observation period ends in 2001, the effects of the accession of the Czech Republic to the EU are not investigated in this framework. Moreover, free movement of labour is still restricted between Germany and the Czech Republic, i.e. the abolition of tariffs and the full liberalisation of labour markets in 2011 at the latest might cause deeper effects in eastern Bavaria. IAB-Bibliothek 321 98 Labour Market Effects in the Bavarian Border Region Appendix to Chapter 3 Figure A 3.1: Southern and northern part of the eastern Bavarian border region (a) Southern part of the eastern Bavarian border region (b) Northern part of the eastern Bavarian border region Eastern Bavarian Border Region District Types (BBR) Eastern Bavarian Border Region District Types (BBR) 99 Chapter 3 Appendix to Chapter 3 Table A 3.1: Basic wage differential for German employees in fixed effects estimations in the southeastern Bavarian border region BORREG (low-skilled) BORREG (skilled) BORREG (high-skilled) sex Coef. τStd. Err. Coef. τStd. Err. Coef. τStd. Err. male –0.0905*** 0.0173 –0.0565*** 0.0060 –0.1084*** 0.0392 N = 939,867 n = 157,411 R2 = 0.7521 N = 3,247,504 n = 388,752 R2 = 0.7530 N = 227,656 n = 38,652 R2 = 0.5605 female –0.0518 0.0608 –0.0248 0.0199 –0.0654 0.0663 N = 694,266 n = 118,528 R2 = 0.6135 N = 1,234,669 n = 214,035 R2 = 0.5638 N = 50,221 n = 12,562 R2 = 0.5227 Source: Author’s own calculations using the weakly anonymous version of the IABS and BeH extract. Notes: In the case of censoring, wages are calculated in the framework of an imputation procedure using the Tobit estimation method; regression with heteroskedasticity-robust standard errors; */**/*** significant at the 10/5/1 percent level; N: number of observations, n: number of individuals; R2 within. Table A 3.2: Basic wage differential for German employees in fixed effects estimations in the northeastern Bavarian border region BORREG (low-skilled) BORREG (skilled) BORREG (high-skilled) sex Coef. τStd. Err. Coef. τStd. Err. Coef. τStd. Err. male –0.0697*** 0.0191 –0.0156 0.0115 –0.0566 0.0417 N = 810,398 n = 129,645 R2 = 0.7624 N = 3,036,192 n = 350,090 R2 = 0.7627 N = 190,724 n = 31,997 R2 = 0.5824 female 0.0321 0.0470 0.0139 0.0229 –0.1082 0.1297 N = 731,382 n = 114,928 R2 = 0.6540 N = 1,154,173 n = 190,594 R2 = 0.5953 N = 34,995 n = 9,001 R2 = 0.5157 Source: Author’s own calculations using the weakly anonymous version of the IABS and BeH extract. Notes: In the case of censoring, wages are calculated in the framework of an imputation procedure using the Tobit estimation method; regression with heteroskedasticity-robust standard errors; */**/*** significant at the 10/5/1 percent level; N: number of observations, n: number of individuals; R2 within. IAB-Bibliothek 321 100 Labour Market Effects in the Bavarian Border Region Figure A 3.2: Fixed effects estimations: wage effect in southeastern Bavarian border region (as %, 1981–2001) 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 4 % 3 % 2 % 1 % 0 % –1 % –2 % –3 % –4 % Source: Author’s own calculations using the weakly anonymous version of the IABS and BeH extract. (a) low-skilled, male, South 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 4 % 3 % 2 % 1 % 0 % –1 % –2 % –3 % –4 % –5 % (b) low-skilled, female, South (c) skilled, male, South (d) skilled, female, South 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 4 % 3 % 2 % 1 % 0 % –1 % –2 % –3 % –4 % (e) high-skilled, male, South 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 10 % 5 % 0 % –5 % –10 % –15 % (f) high-skilled, female, South 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 3 % 2 % 1 % 0 % –1 % –2 % –3 % 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 3 % 2 % 1 % 0 % –1 % –2 % –3 % –4 % –5 % 101 Chapter 3 Appendix to Chapter 3 Figure A 3.3: Fixed effects estimations: wage effect in northeastern Bavarian border region (as %, 1981–2001) 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 2 % 1 % 0 % –1 % –2 % –3 % –4 % –5 % –6 % Source: Author’s own calculations using the weakly anonymous version of the IABS and BeH extract. (a) low-skilled, male, North 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 2 % 1 % 0 % –1 % –2 % –3 % –4 % –5 % –6 % –7 % (b) low-skilled, female, North (c) skilled, male, North (d) skilled, female, North 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 4 % 3 % 2 % 1 % 0 % –1 % –2 % –3 % –4 % (e) high-skilled, male, North 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 10 % 5 % 0 % –5 % –10 % –15 % –20 % (f) high-skilled, female, North 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 3 % 2 % 1 % 0 % –1 % –2 % –3 % 1980 1983 1986 1989 1992 1995 1998 2001 border effect upper 95 % CI limit lower 95 % CI limit 5 % 4 % 3 % 2 % 1 % 0 % –1 % –2 % –3 % 103 Chapter 4 4 Labour Market Effects in the Czech Border Region 4.1 Introduction Of course, the fall of the Iron Curtain had an effect not only on Western European labour markets, but also on the transition countries. The employees in the Central and Eastern European Countries (CEEC) had to undergo deep changes during the first years on the way from plan to market. Not only did the formerly dependable delivery areas of the COMECON break away, but many state-owned enterprises were also not ready for competition. In the years after the opening of the border, trade impediments vanished and, just as crucial, the transition countries opened their markets for foreign capital. Apart from other motives like developing new markets, foreign direct investment (FDI) also occurs in order to relocate production activities to low-wage countries, also referred to as international outsourcing. Concerning the evaluation in the literature Egger/Egger (2002: 83) critically note that “… the theoretical analysis and empirical assessment … of international outsourcing is rather new and at least concerning its implications for developing countries it seems to be still in its infancy.”22 Obviously, investigating integration effects in former Eastern Bloc countries is quite different from analysing Western European countries. Until the fall of the Communist regimes a real labour “market” did not exist, i.e. unemployment was basically hidden and education-related wage differentials were extremely low (Münich et al. 2005). Moreover, in contrast to the research on the German labour market, I am not able to approach spatial differences on the Czech labour market by stressing the “natural experiment” situation before and after the introduction of free trade and capital mobility due to the lack of suitable data. Datasets containing appropriate regional information only provide data from the beginning of the 1990s onwards. However, it is exceedingly interesting to discover whether the Czech border region close to the Western European high-wage countries benefits from its geographical position during the increasing integration of markets. It is important to notice that even without transnational free labour mobility (which will probably be restricted for Czech workers until 2011), trade and international outsourcing of production activities can lead to shifts in the labour demand and wage structure regarding different skill groups. According to my hypothesis, these integration effects should be stronger in border regions. Using two data sources, I investigate whether free trade with Western European countries led to special effects on the labour market in the districts neighbouring Bavaria and Austria. 22 See also Pusterla/Resmini (2007: 839): “The Central and Eastern Europe region has been only marginally considered in the empirical literature on firm location choice.” IAB-Bibliothek 321 110 Labour Market Effects in the Czech Border Region Figure 4.2: Share of full-time employees working in the Czech border region: (a) including Prague and Mladá Boleslav, (b) without Prague and Mladá Boleslav (as %) indicates the relative importance of the border districts as an economic location. In 1992, 26.5 % of all fully employed people worked in the border region (Figure 4.2a). This proportion increased to 28.6 % in 1996 and then slightly declined to 28.3 % in 2002. This means that in the early transition years the districts near Bavaria and Austria gained in attractiveness as locations for employers and employees. From 1996 to 2002 the non-border districts including Prague recaptured three tenths of a percentage point of relative employment. Since the outstanding importance of Prague and Mladá Boleslav23 possibly distorts the outcome, I also calculate the border region share without these districts (Figure 4.2b). In this case the proportion of employees working in the border region is naturally far higher. However, the conclusion does not change. Starting from an employment share in the border region of 32.1 % in 1992, the proportion rose to 34.3 % in 1996 and fell again to 34.1 % in 2002, signifying the stabilisation of the regional employment share. Similar to the investigation with the German data, I also inspect the Czech Microcensus with respect to the structural change and specialisation in the border and non-border region. Of course, since I only have two points in time containing information about industries I do not observe a structural change indicator over time using this variable. However, I can alternatively analyse the changes using the differences in the distribution of occupations. First, I take a look at the relative shares of occupational and industrial sectors. Due to the predictions of the models of Feenstra/Hanson and Brülhart et al., free trade and international outsourcing should lead to spatial effects regarding the distribution of economic activities within a country. The border region particularly should attract economic activities 23 The automotive manufacturer Škoda Auto a.s. has its main production location in Mladá Boleslav, employing around 20,000 staff members. Source: Author’s own calculations from Czech Microcensus 1992, 1996, 2002. 30 % 29 % 28 % 27 % 26 % 25 % (a) with Prague and Mladá Boleslav 1990 1992 1994 1996 1998 2000 2002 2004 35 % 34 % 33 % 32 % 31 % 30 % (b) without Prague and Mladá Boleslav 1990 1992 1994 1996 1998 2000 2002 2004 111 Chapter 4 The Labour Market in the Czech Republic: Some Descriptive Evidence in industries which have comparative advantages over the foreign country (Barjak/ Heimpold 2000b), i.e. Germany and Austria. The two theoretical strands point in the same direction: while the Feenstra-Hanson model refers to offshore activities conducted outside the high-wage country, the NEG model suggests a relative increase in sectors where import competition from Germany and Austria is supposed to be relatively low. In any case, the predicted effects should be reflected in the descriptive figures and in indicators displaying structural change and specialisation. Table 4.3: Employment shares of occupations in the Czech non-border and border region (as %) ISCO-88 major groups non-border region border region 1992 1996 2002 1992 1996 2002 1 Legislators, senior officials and managers 3.45 2.95 3.57 2.74 2.36 3.87 2 Professionals 7.13 5.87 7.77 7.04 5.22 9.76 3 Technicians and associate professionals 20.44 19.74 24.89 20.17 19.20 24.84 4 Clerks 10.47 12.78 13.74 10.51 14.32 6.94 5Service workers, shop & market sales workers 10.20 10.44 13.35 9.95 10.21 12.98 6 Skilled agricultural and fishery workers 1.29 1.52 1.07 2.26 2.15 1.61 7 Craft and related workers 25.95 26.08 19.48 26.12 27.25 21.94 8 Plant and machine operators, assemblers 12.14 11.83 9.86 11.18 11.53 13.06 9 Elementary occupations 8.94 8.79 6.26 10.03 7.75 5.00 Total 100 100 100 100 100 100 Source: Author’s own calculations from Czech Microcensus 1992, 1996, 2002. Concerning occupations, Table 4.3 shows the employment shares of the nine ISCO major groups in the three years of observation divided into the border region and the rest of the country. Not surprisingly, as the Czech proficiency with respect to engineering and manufacturing is well-known, technicians and craft workers (major groups 3 and 7) constitute the bulk of the workforce, followed by clerks, service workers and plant and machine operators (major groups 4, 5 and 8). As the ISCO corresponds to the International Standard Classification of Education (ISCED), it pays off to analyse the shifts in this context, too. Elementary occupations (major group 9) are defined as the lowest skill level. Major groups 4–8 are considered to be at the second level, major group 3 forms the third level and major group 2 the highest level. There is no skill reference for major group 1, since this group embraces significant skill differences. Obviously there are no outstanding differences between the districts near Bavaria and Austria and the rest of the Czech Republic. From 1992 until 2002 major IAB-Bibliothek 321 112 Labour Market Effects in the Czech Border Region groups 1–5 exhibit increasing employment shares in both the non-border and border region (with one exception), while the shares fell for major groups 6–9 in both objects of investigation (with one exception). This indicates a general professional skill upgrading, which, interestingly, happened not from 1992 until 1996 but from 1996 until 2002. The employment shares remained relatively stable in the early transition years, but after the recession years, the occupations which correspond to higher skill levels recorded higher values. Possibly, employment relationships were relatively stable in the upswing years, but the years from 1997 onwards brought a lot of restructuring. I will get back to this point below. Another striking figure is the severely decreasing share of clerks in the border region from 1996 until 2002. This is apparently closely related to the advancement of Prague as a financial centre. If Prague is excluded from the dataset, the share of clerks in the non-border region falls from 1996 until 2002. Besides, the share of plant and machine operators and assemblers (major group 8) rose in the border region in contrast to the rest of the country. The increase in this occupation group is potentially connected with some cross-border relations in industries which are also important at least in the Bavarian borderlands.24 In Table 4.4 the shares of 14 industries subject to the NACE classification are recorded for border and non-border districts. Though this variable is not available in 1992, it is nevertheless interesting to investigate the shifts between 1996 and 2002, since this period embraces the years of recession and, as the figures for occupations have shown, a lot of changes happened during this period of time. First of all − as in the case of occupations − the relative figures for the border and non-border region are very similar. As is also common in transition countries, most industries in the primary and secondary sector became less important in relation, while the shares of the service industries in the tertiary sector increased. The sign of the change is identical in the non-border and border region in 11 of the 14 industries, which indicates that the structural change proceeded in the same direction. Only in the industries E (electricity, gas and water supply), G (wholesale and retail trade etc.) and N (health and social work) did the share in the non-border region rise, but declined in the border region. The only really outstanding change is the relative shrinking of the largest industry, which comprises all sorts of manufacturing. This industry decreased by about 8 percentage points in the non-border districts, but only marginally in the border districts. Possibly − as mentioned above − the dominant position of manufacturing in the border region is maintained due to trade relations of large manufacturing locations such as Pilsen, for instance, which is closely affiliated with the Bavarian industry. 24 Regarding the shares of clerks and plant and machine operators & assemblers the differences between the borderland and the non-border region in 2002 are statistically significant. 113 Chapter 4 The Labour Market in the Czech Republic: Some Descriptive Evidence Table 4.4: Employment shares of industries in the Czech non-border and border region (as %) NACE industries non-border region border region 1996 2002 1996 2002 AB Agriculture, hunting and forestry & fishing 4.51 3.65 6.37 5.16 C Mining and quarrying 3.53 1.18 1.31 0.65 D Manufacturing 35.16 27.20 32.94 32.18 E Electricity, gas and water supply 2.32 2.55 2.63 2.34 F Construction 8.42 7.31 9.75 7.26 G Wholesale and retail trade; repair of motor vehicles, motorcycles and personal and household goods 10.21 11.62 9.32 8.71 H Hotels and restaurants 2.12 3.46 2.56 2.66 I Transport, storage and communication 7.37 8.16 7.37 8.23 J Financial intermediation 2.34 3.21 1.68 1.69 K Real estate, renting and business activities 2.85 3.52 2.86 4.35 L Public administration and defence; compulsory social security 6.55 8.96 6.92 8.79 M Education 5.72 7.14 6.32 7.74 N Health and social work 5.28 6.73 6.70 6.61 O Other community, social and personal service activities 3.63 5.30 3.26 3.63 Total 100 100 100 100 Source: Author’s own calculations from Czech Microcensus 1996, 2002. Table 4.5 comprises the values for an indicator of structural change (ISC) and the Krugman specialisation index (KSI). The indicator of structural change measures the absolute deviations of the employment shares of occupations or industries in year t+1 (ai,t+1) from the figures in year t (ai,t ). Adding up all absolute deviations and dividing by 2, the ISC equals 0 if the shares in t+1 are identical to the shares in t and equals 1 if the structure in t+1 deviates maximally from the structure in t (formula 4.1). ,6&t, t+1 = ∑ | ai, t+1 – ai, t | (4.1) The values for the occupational structure are higher in the border region for both time periods, which can potentially be traced back to the smaller number of observations in this area. The ISC for the industrial structure, which can only be calculated once, however, has a higher value for the non-border region, probably caused by the high decrease in manufacturing. N i = 1 IAB-Bibliothek 321 114 Labour Market Effects in the Czech Border Region The KSI is defined as the sum of the absolute deviations of the employment shares in the border region (ai,t, border) from the employment shares in the rest of the country (aLWQRQERUGHU) for all occupations or industries in year t. Divided by 2, the index equals 0 if the employment shares in the two areas are identical and equals 1 if the structure in the border region deviates maximally from the structure in the nonborder region (formula 4.2). .6,t = ∑ | ai, t, border – aLWQRQERUGHU| (4.2) Regarding both occupations and industries, the KSI exhibits increasing values, i.e. the specialisation of the border region grew over the years. Including the data from Tables 4.3 and 4.4, this development can be explained by a higher persistence of manufacturing occupations (e.g. major group 8 in Table 4.3) and industries (Table 4.4) in the border region, while the change towards the tertiary sector is stronger in the non-border region. Moreover, the results of both indices (ISC and KSI) corroborate the impression that in the uneasy years from 1996 onwards the economy underwent more profound change than in the four years before. Table 4.5: Indicator of structural change and Krugman specialisation index for occupations and industries in the Czech Republic 1992/1996 1996/2002 Indicator of structural change (occupations) non-border 0.029 0.056 border 0.115 0.160 Indicator of structural change (industries) non-border 0.123 border 0.061 1992 1996 2002 KSI (occupations) 0.023 0.033 0.085 KSI (industries) 0.063 0.080 Source: Author’s own calculations from Czech Microcensus 1992, 1996, 2002. 4.3.2 Skill Structure of Employed People Regarding the distribution of skills, I investigate whether there is a different development in the skill structure between border and non-border districts. While the predictions of the two models were consistent in the previous subsection, they are not so with respect to the skill structure of the labour demand: according to Feenstra/Hanson the activities which are shifted to the foreign low-wage country should lead to a skill upgrading process, since these production steps are relatively skill-intensive there. If distance matters, border regions will be particularly affected N i = 1 115 Chapter 4 The Labour Market in the Czech Republic: Some Descriptive Evidence and the demand for more highly skilled labour is assumed to increase at an aboveaverage rate in the districts near Bavaria and Austria. In contrast, on the basis of the NEG model unskilled and low-skilled labour especially should have comparative advantages in the borderland, as import competition from beyond the frontier is relatively low for activities requiring less human capital in relation. The descriptive figures are contained in Figure 4.3. The share of unskilled employees generally decreases from about 12 % in 1992 to about 6 % in 2002, in both the border and non-border districts. Only in 1996 are unskilled workers slightly overrepresented in the districts close to Bavaria and Austria. Regarding low-skilled workers, the share remains fairly stable from 1992 to 1996, oscillating around 45 % in both regions under review. However, in 2002 it then declines to 42.7 % in the border districts and to 40.1 % in the non-border districts. While the fraction of medium-skilled employees shifts in parallel from about 30 % in 1992 to 40 % in 2002, the proportion of high-skilled workers initially falls between 1992 and 1996, but then rises to 13.2 % in the non-border region and 11.4 % in the border region. Altogether, unskilled and low-skilled workers are slightly overrepresented in the Figure 4.3: Shares of skill groups of full-time workers comparing the Czech border region to the rest of the country (as %) Source: Author’s own calculations from Czech Microcensus 1992, 1996, 2002. 15 % 12 % 9 % 6 % 3 % 0 % 1990 1992 1994 1996 1998 2000 2002 2004 non-border border unskilled 50 % 47 % 44 % 41 % 38 % 35 % 1990 1992 1994 1996 1998 2000 2002 2004 non-border border low-skilled 43 % 40 % 37 % 34 % 31 % 28 % 1990 1992 1994 1996 1998 2000 2002 2004 non-border border medium-skilled 15 % 12 % 9 % 6 % 3 % 0 % 1990 1992 1994 1996 1998 2000 2002 2004 non-border border high-skilled IAB-Bibliothek 321 116 Labour Market Effects in the Czech Border Region border districts at the end of the observation period, but there is no fundamental difference identifiable in the development of skill group shares. Disregarding the decreasing share of high-skilled workers from 1992 until 1996, the figures show evidence of a skill upgrading process in the Czech Republic which is in line with the relative changes in the ISCO major groups (see chapter 4.3.1). The share of unskilled and low-skilled workers declines over time, while the share of more highly skilled employees rises. 4.3.3 Skill Structure of Unemployed People Using quarterly unemployment data provided by the district labour exchanges, I take a similar look at the shares of unemployed people in order to investigate whether the distribution of skill groups in the two areas of observation exhibits fundamental differences compared to the figures for employed workers in chapter 4.3.2. The absolute numbers show the tremendous growth in unemployment across all skill groups in the late 1990s recession years (Figure 4.4). The number of unskilled unemployed increased from less than 60,000 at the beginning of the 1990s to more than 160,000 ten years later. The number of low-skilled unemployed, which was also about 60,000 in 1992, even rose to nearly a quarter of a million in the first years of the new century. The groups of medium-skilled and high-skilled unemployed quadrupled from approximately 30,000 to nearly 120,000 and from under 5,000 to almost 20,000 respectively. Interestingly, the development of the shares of the different skill groups in unemployment prima facie seems quite surprising (Figure 4.5). The share of unskilled people out of the total unemployment share declines – after a rise in the early 1990s – from nearly 40 % to 30 %. Equivalently, the fractions of the other three skill groups increased in the recession years. While the proportion of medium-skilled unemployed fell after 1998, the share of low-skilled and high-skilled unemployed grew moderately. Taking into account the economic transformation process in the Czech Republic, the figures are quite plausible. Before 1997, unemployment was rather an exception. The recession affected a much larger spectrum of the labour force across education groups and the growing denominator (growing faster than the number of unskilled unemployed) led to a lower share of unskilled workers. The pool of unskilled workers is limited and given that many members of this skill group were already unemployed before 1997, the proportion of unskilled unemployed could not grow so fast. In other words, unemployment became an issue of the “masses”, as was common in other EU countries. After the recession years, the proportion of unskilled people increases again compared to the total number of unemployed. Comparing the border region to the non-border districts, the fraction of unskilled and low-skilled unemployed in the 117 Chapter 4 The Labour Market in the Czech Republic: Some Descriptive Evidence border districts remains slightly below the level for the rest of the country, while the opposite applies for the medium- and high-skilled unemployed. Bearing in mind the figures for employees (see chapter 4.3.2), the results could indicate a slightly higher labour demand for more highly skilled workers in the non-border region. This is quite clear intuitively, since Prague belongs to the non-border region and possibly absorbs qualified personnel from other parts of the country. In the econometric part of this chapter I will control for this and other factors. Figure 4.4: Number of unemployed in the different skill groups 140 120 100 80 60 40 20 0 180 160 140 120 100 80 60 40 20 0 Source: Author’s own calculations from quarterly unemployment data of Czech district labour exchanges (1/1992–2/2006). 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 unskilled 250 200 150 100 50 0 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 low-skilled 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 medium-skilled 20 18 16 14 12 10 8 6 4 2 0 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 high-skilled (a) Number of unemployed (persons ‘000) (b) Number of unemployed (Index: year 1992 = 100) 1992 1994 1996 1998 2000 2002 2004 2006 unskilled low-skilled medium-skilled high-skilled 400 350 300 250 200 150 100 50 0 IAB-Bibliothek 321 118 Labour Market Effects in the Czech Border Region 4.3.4 Wage Differentials between Border and Non-Border Region Differences in the labour demand are assumed to be also reflected in the development of wages. As mentioned in the previous subsections, relative labour demand and thus relative wages should rise for more highly skilled employees in the border region compared to non-border districts if the Feenstra-Hanson trade effects play a dominant role. In contrast, according to the model of Brülhart et al., unskilled and low-skilled workers are particularly supposed to benefit in the borderlands due to the higher market potential and relatively low import competition. Regarding wage differentials between the Czech border and non-border region at the descriptive level, I use the gross wages available from the Microcensus in 1996 and 2002 and contrast the figures for the two areas. Table 4.6 shows that annual nominal gross wages substantially increased in the observation period for all three calculated deciles and region types with growth rates from about 32 to 62 percent. In general, border region workers earn less than Figure 4.5: Relative shares of skill groups of unemployed people comparing the Czech border region to the rest of the country (as %) Source: Author’s own calculations from quarterly unemployment data of Czech district labour exchanges (1/1992–2/2006). 42 % 39 % 36 % 33 % 30 % 27 % 1990 1995 2000 2005 2010 non-border border unskilled 48 % 45 % 42 % 39 % 36 % 33 % 1990 1995 2000 2005 2010 non-border border low-skilled 30 % 27 % 24 % 21 % 18 % 15 % 1990 1995 2000 2005 2010 non-border border medium-skilled 5 % 4 % 3 % 2 % 1 % 0 % 1990 1995 2000 2005 2010 non-border border high-skilled 119 Chapter 4 The Labour Market in the Czech Republic: Some Descriptive Evidence their peers in the non-border region. In 1996, the wage gap between non-border and border districts monotonically widens for all skill groups with the decile considered. The relative wage gap in the groups of medium- and high-skilled workers is larger (from about 3 to 11 percent), while the only decile in which border region employees are ahead is the second decile for low-skilled workers. In 2002 the wage differential widens for three skill groups in all deciles. However, concerning unskilled employees the picture is completely different. All deciles of this skill group are higher in the border region with a maximum difference of 15.1 % for D5. The differences between the years of observation are shown in Figure 4.6. Regarding wage differentials between different skill groups, I calculate the skill premium for low-, medium- and high-skilled workers compared to unskilled employees (Table 4.7). In most cases the wage differentials are higher at the top of the distribution. With the exception of low-skilled workers in 1996 the skill premium is higher in the non-border region. Bearing in mind the previous results it is not surprising that the wage gap concerning unskilled employees decreases considerably in the border region in 2002. Figure 4.7 shows the size and development of the skill premium for low- and medium-skilled employees relative to unskilled workers in the border region (horizontal axis) and the non-border region (vertical axis). Using the same scale for both regions the 45° line represents points where the skill premia are equal in the border and non-border region. In 1996 all values are relatively close to the 45° line, so that it can be concluded that the border region differs only marginally from the non-border region. The arrows in the figure indicate the changes between 1996 and 2002. In 2002 all points are above the 45° line, representing a higher skill premium in the nonborder region. All arrows point north-west, i.e. the skill premium shrinks in the districts close to Western Germany and Austria, while it increases in the districts relatively distant from the border.25 Summarising the results for the descriptive wage differentials, I conclude that the border districts suffered relative wage losses in three out of four skill groups. Interestingly, in the group of unskilled workers the development differs substantially. However, the informative value of these figures is restricted, since the non-border region, for example, contains Prague and Mladá Boleslav, featuring special developments which I have to control for in the econometric analysis. 25 For reasons of clarity I do not map the wage premia of high-skilled workers, which are in line with the results for low- and medium-skilled workers and do not change the general findings. IAB-Bibliothek 321 126 Labour Market Effects in the Czech Border Region marginally (see section 4.3.2). In order to evaluate the changes in the wage differentials I estimate the following Mincerian wage equation (Mincer 1974) separately for the years 1992, 1996 and 2002: OQ:$*(i = α + β ')(0i + γ 1 (;3(5i + γ 2 (;3(5i 2 + γ 3 (;3(5B)i + γ 4 (;3(52B)i + ∑ δ j 0$567$7ji + ∑ ϕ m 2&&83mi (4.5) + η 325'(16i + ϕ 35$+$i + τ BORREG i + ε i :$*(i denotes the individual i ’ s annual gross wage in the regular occupation in the relevant year. Unfortunately, data concerning gross wage are not available for 1992. On the other hand, the variable for the net wages includes a lot of missing values in 2002 (almost two-thirds of the 4,880 observations with regard to full-time employment). Since net and gross wages are almost perfectly correlated in 1996 and 2002 (correlation coefficient > 0.99), I decided to use the net wage in 1992 as a proxy for the gross wage.27 In addition to the conventional variables of the Mincerian wage equation (')(0, (;3(5, (;3(52, interaction terms), I use dummies for marital status (0$567$7) and occupational status (2&&83). In this estimation I do not control for industries since this information was not available for 1992. As in the estimations of qualification trends, I control for the population density of the districts (323'(16) and the special labour market situation in Prague and Mladá Boleslav (35$+$). For a detailed definition of the variables see Tables 4.10 and 4.11. The results of the coefficients for the control variables correspond to the theoretical expectations (Table 4.12). Female workers ceteris paribus earn about 20 % less than male employees in the lower skilled group and 25 % less in the more highly skilled group. These values hardly change over time. One additional year of potential experience yields a significant wage increase, but the significant negative coefficient for (;3(52 signifies that the benefit of experience declines over time. For female workers these effects are less distinctive. 27 One possible explanation for the high value of the correlation coefficient is the fact that in socio-scientific surveys “people tend to respond by estimating net rather than gross earnings, even if they are asked for the latter” 9HèFHUQÊN1HYHUWKHOHVV,ZLOOGRVRPHVHQVLWLYLW\DQDO\VHVVHHEHORZLQRUGHUWRFKHFNZKHWKHUWKH results are robust. J = 3 j = 1 0  m = 1 127 Chapter 4 Econometric Analysis of Wage Differentials Table 4.10: Variables of the wage equation (Czech Republic) ln WAGE logarithm of individual wage DFEM dummy for sex (female = 1) EXPER potential job experience EXPER2 potential job experience 2/100 EXPER_F potential job experience, female EXPER2_F potential job experience 2/100, female MARSTAT* 3 dummies for the marital status (married, divorced, widowed) OCCUP* 8 dummies for occupations BRANCH* 13 dummies for industries (only in 1996 and 2002) POPDENS population density in thousand inhabitants/km² PRAHA dummy for Prague and Mladá Boleslav BORREG dummy for border region Constant constant Table 4.11: Values of EXPER (Czech Republic) Qualification Potential experience Skill group primary education not complete EXPER = AGE – 6 – 6 unskilled primary education EXPER = AGE – 6 – 9 unskilled occupational qualification with lower secondary education EXPER = AGE – 6 – 11 low-skilled occupational qualification with (lower) secondary education (without PDWXULWD) EXPER = AGE – 6 – 12 low-skilled occupational qualification with upper secondary education (with PDWXULWD) EXPER = AGE – 6 – 13 medium-skilled higher technical education EXPER = AGE – 6 – 15 medium-skilled university degree EXPER = AGE – 6 – 19 high-skilled PhD degree EXPER = AGE – 6 – 21 high-skilled Notes: The workers’ potential on-the-job experience (EXPER) is measured in years as age minus average duration of education minus six. I impose six years as the average duration of education for unskilled workers without primary education, nine years for unskilled workers with primary education, 11, 12 and 13 years for workers with secondary education depending on the respective level of secondary occupation and 15, 19 and 21 years for workers with higher technical education or university graduates. IAB-Bibliothek 321 128 Labour Market Effects in the Czech Border Region Table 4.12: Estimation results for the wage effect of lower and more highly skilled workers without pooling cross sections variable lower skilled (unskilled & low-skilled) more highly skilled (medium- & high-skilled) 1992 1996 2002 1992 1996 2002 DFEM –0.2277*** (0.0242) –0.2172*** (0.0211) –0.2125*** (0.0516) –0.2840*** (0.0300) –0.2713*** (0.0256) –0.2587*** (0.0481) EXPER 0.0313*** (0.0019) 0.0177*** (0.0015) 0.0155*** (0.0036) 0.0296*** (0.0026) 0.0200*** (0.0023) 0.0190*** (0.0047) EXPER2–0.0741*** (0.0042) –0.0402*** (0.0032) –0.0334*** (0.0081) –0.0733*** (0.0064) –0.0480*** (0.0055) –0.0471*** (0.0114) EXPER_F –0.0266*** (0.0025) –0.0143*** (0.0021) –0.0146*** (0.0054) –0.0175*** (0.0036) –0.0062** (0.0029) –0.0070 (0.0058) EXPER2_F 0.0651*** (0.0060) 0.0334*** (0.0049) 0.0354*** (0.0126) 0.0555*** (0.0095) 0.0189** (0.0074) 0.0175 (0.0148) MARSTAT1 (married) 0.1754*** (0.0140) 0.1136*** (0.0106) 0.0298 (0.0222) 0.1612*** (0.0173) 0.0806*** (0.0141) –0.0160 (0.0258) MARSTAT2 (divorced) 0.1900*** (0.0189) 0.0977*** (0.0148) 0.0148 (0.0285) 0.1630*** (0.0245) 0.0504*** (0.0194) –0.0186 (0.0311) MARSTAT3 (widowed) 0.1877*** (0.0287) 0.1166*** (0.0260) –0.0203 (0.0492) 0.1141*** (0.0372) 0.1475*** (0.0313) –0.0104 (0.0760) OCCUP* yes yes yes yes yes yes PRAHA 0.0520*** (0.0175) 0.0708*** (0.0151) 0.1083*** (0.0361) 0.0733*** (0.0186) 0.0850*** (0.0157) 0.1910*** (0.0349) POPDENS 0.0379*** (0.0081) 0.0576*** (0.0071) 0.0408** (0.0174) 0.0352*** (0.0080) 0.0503*** (0.0069) 0.0238 (0.0155) BORREG –0.0072 (0.0082) –0.0087 (0.0065) 0.0009 (0.0157) –0.0160 (0.0104) –0.0152* (0.0085) –0.0360* (0.0186) Constant 10.6998*** (0.0769) 11.7313*** (0.0428) 12.1240*** (0.1103) 1.0280*** (0.0310) 12.1292*** (0.0279) 12.5134*** (0.0494) Test statistics N 7,479 10,967 2,190 5,485 8,555 2,689 R20.4010 0.3390 0.3301 0.3726 0.3598 0.3138 Dependent variable: ln Wage Source: Author’s own calculations from Czech Microcensus 1992, 1996, 2002. Notes: Regression with heteroskedasticity-robust standard errors; */**/*** significant at the 10/5/1 percent level. There are wage premia for married, divorced and widowed employees in 1992 and 1996, which, interestingly, disappear for both skill groups in 2002. Maybe the first generation of young single employees, educated after the fall of Communism 129 Chapter 4 Econometric Analysis of Wage Differentials compensated for the wage premia of non-singles with their higher productivity.28 Significant outcomes for almost all occupation dummies indicate the differences between the various professions. The wage differential for workers in Prague and Mladá Boleslav increases over time, from 5.2 % to 10.8 % in the lower skilled group and from 7.3 % to 19.1 % in the more highly skilled group. With the exception of one case, the population density has a significant positive effect on the wage. The variable which I am most interested in is the border region dummy. In 1992 and 1996 the coefficient indicates negative, but – in three out of four cases – insignificant wage differentials for border region workers (Table 4.12 and Figure 4.8). In 2002 the wage gap seems to disappear for unskilled and low-skilled workers and to widen for medium- and high-skilled employees in the districts near Bavaria and Austria. However, since there are far fewer observations in 2002, the confidence interval is very large for this year, so that it is not possible to derive deeper conclusions from this estimation. Therefore, in a next step I apply a difference-in- differences approach in order to obtain more exact results. 4.5.2 Pooling Cross Sections Over Time In contrast to the former estimations I now pool all observations for each of the original four skill groups over time, i.e. I have an independently pooled cross-section for unskilled, low-skilled, medium-skilled and high-skilled employees. Estimating only one equation in each case leads to a larger sample size which, in turn brings more precise estimators and test statistics with more power. Keeping the control 28 Using net wages as an endogenous variable does not change this result (despite the large amount of missing values in 2002). Figure 4.8: Wage effect for (a) lower and (b) more highly skilled workers in the Czech border region (as %) Source: Author’s own calculations from Czech Microcensus 1992, 1996, 2002. (a) lower skilled 1990 1992 1994 1996 1998 2000 2002 2004 BORREG upper 95 % CI limit lower 95 % CI limit (b) more highly skilled 1990 1992 1994 1996 1998 2000 2002 2004 BORREG upper 95 % CI limit lower 95 % CI limit 1 % 0 % –1 % –2 % –3 % –4 % –5 % –6 % –7 % –8 % 4 % 3 % 2 % 1 % 0 % –1 % –2 % –3 % –4 % IAB-Bibliothek 321 130 Labour Market Effects in the Czech Border Region variables of the previous regressions, I include year dummies for the years 1996 and 2002 (<($5, <($5) with the reference year 1992. Furthermore, I include interaction terms of the year dummies with the border region dummy. The variables %255(*<($5 and %255(*<($5 measure the change of the wage differential in the border region from 1992 to 1996 and 2002 respectively. As these interaction terms control for the difference (over time) in the difference (wage gap in the border region), the coefficients ω1 and ω2 represent the difference-in- differences estimators (Wooldridge 2002). The equation now has the following form: OQ:$*(i = α + β ')(0i + γ 1 (;3(5i + γ 2 (;3(5i 2 + γ 3 (;3(5B)i + γ 4 (;3(52B)i + ∑ δ j 0$567$7ji + ∑ ϕ m 2&&83mi + η 325'(16i (4.6) + ϕ 35$+$i + τ BORREG i + υ 1 YEAR1996 t + υ 2 YEAR2002 t + ω 1 %255(* YEAR1996i t ) + ω 2 %255(* YEAR2002i t ) + ε i The results are shown in Table 4.13. In this case the coefficient values of the control variables also correspond to the theoretical expectations. The gender wage gap is most distinctive for unskilled workers, i.e. female unskilled employees earned ceteris paribus 36.9 % less than their male counterparts. Compared to unskilled females the differential is only half as large for low-skilled female workers, but then increases with the skill level. The coefficient values for the variables concerning experience indicate that depending on the skill group, one additional year of potential experience yields a wage increase, which mitigates over time and is smaller for female workers. With the exception of low-skilled workers, the wage bonus in Prague and Mladá Boleslav oscillates around 10 %. The population density, which controls for agglomeration effects, has a positive effect on wages, but which is only significant in the case of low- and medium-skilled workers. The coefficient for %255(* shows a negative, but insignificant wage differential for employees from all skill groups in the districts near Bavaria and Austria in 1992. This wage gap did not change considerably until 1996, as shown by the outcome for %255(*<($5. However, the values for %255(*<($5 indicate that things changed between 1996 and 2002. By adding the basic wage effect for the border region and the effect until 2002, which is captured by the interaction term, it turns out that unskilled workers in the border districts in 2002 earned about 12 % more than employees in districts remote from Bavaria and Austria. In all other skill groups the wage differential for border region employees deteriorated over time. Although − apart from the unskilled group − only the value for medium-skilled workers is significant at the 5 percent level, it is striking that the wage differential deepens with the skill level. While the total wage effect in the low-skilled group amounts 1.9 % in 2002, i.e. workers in the border region earned 1.9 % less, the effect for medium- and high- J = 3 j = 1 0  m = 1 131 Chapter 4 Econometric Analysis of Wage Differentials skilled workers adds up to 5.1 % and 6.1 % respectively. This means that regarding skill levels, a clear structure with respect to wage differentials emerged by 2002: the higher the skill level, the more disadvantageous was a job in the border region. Table 4.13: Estimation results for the wage effect in difference-in-differences estimations variable unskilled low-skilled medium-skilled high-skilled coef. t-Stat. coef. t-Stat. coef. t-Stat. coef. t-Stat. DFEM –0.3688*** –6.58 –0.1975*** –10.37 –0.2446*** –9.55 –0.2768*** –5.97 EXPER 0.0109*** 2.65 0.0257*** 18.70 0.0238*** 10.19 0.0210*** 5.67 EXPER2–0.0255*** –3.10 –0.0592*** –19.32 –0.0543*** –10.21 –0.0525*** –5.39 EXPER_F –0.0004 –0.08 –0.0236*** –11.42 –0.0125*** –4.42 –0.0071 –1.24 EXPER2_F 0.0047 0.47 0.0588*** 11.58 0.0356*** 5.08 0.0330** 2.14 mar. status yes yes yes yes occ. status yes yes yes yes PRAHA 0.0993*** 3.84 0.0650*** 4.45 0.1116*** 7.34 0.1013*** 3.28 POPDENS 0.0114 0.91 0.0531*** 7.99 0.0342*** 5.15 0.0164 1.29 YEAR1996 0.8460*** 68.23 0.8615*** 136.79 0.9197*** 114.51 1.0083*** 59.88 YEAR2002 1.1759*** 49.43 1.2279*** 122.35 1.3457*** 113.86 1.4054*** 61.15 BORREG –0.0190 –1.17 –0.0007 –0.08 –0.0075 –0.65 –0.0165 –0.80 BORREG~96 0.0009 0.04 –0.0053 –0.45 –0.0077 –0.54 –0.0053 –0.19 BORREG~02 0.1382*** 3.07 –0.0187 –0.99 –0.0435** –1.97 –0.0445 –1.04 Constant 11.0608*** 71.53 10.7842*** 245.63 11.0350*** 375.89 11.1996*** 282.95 Test statistics N 4,000 16,636 12,855 3,874 R² 0.7697 0.7705 0.7606 0.7326 Dependent variable: ln Wage Source: Author’s own calculations from Czech Microcensus 1992, 1996, 2002. Notes: Regression with heteroskedasticity-robust standard errors; */**/*** significant at the 10/5/1 percent level. Some sensitivity analyses do not change this finding. Restricting the difference-in- differences approach to only two years of observation (1992 & 1996, 1992 & 2002, 1996 & 2002) yields very similar results. In the most interesting version, excluding the observations of 1992, it is possible to include dummy variables for the 14 industries in the estimation. It could be important to control explicitly for industries such as manufacturing, for instance, which is notably represented above-average in the border region in 2002. However, the wage differentials for the different skill groups do not deviate substantially from the outcomes above (Table 4.14, see Ta- IAB-Bibliothek 321 132 Labour Market Effects in the Czech Border Region ble A 4.2 in the appendix for control variables): a remarkable relative wage gain of 13.2 % for unskilled workers in the borderlands, while all other skill groups exhibit relative wage losses from 1996 until 2002, downgrading with the skill level. According to this version, the relative wage of medium-skilled workers in the border region decreased by more than five percentage points (significant at the 1 percent level) and the wage of high-skilled workers by more than six percentage points (significant at the 10 percent level). Table 4.14: Wage effect in the border region controlling for industries variable unskilled low-skilled medium-skilled high-skilled coef. t-Stat. coef. t-Stat. coef. t-Stat. coef. t-Stat. BORREG(1996) –0.0029 –0.20 0.0106 1.45 0.0036 0.40 0.0112 0.50 BORREG~2002 0.1317*** 2.98 –0.0209 –1.21 –0.0512*** –2.65 –0.0646* –1.67 Test statistics N 2,402 10,755 8,826 2,418 R² 0.4570 0.4930 0.5022 0.4642 Dependent variable: ln Wage Control variables: see Table A 4.2 in the appendix Source: Author’s own calculations from Czech Microcensus 1996, 2002. Notes: Regression with heteroskedasticity-robust standard errors; */**/*** significant at the 10/5/1 percent level. In an alternative specification I do not split the dataset according to skill groups but again according to years. In contrast to the former estimations I now run regressions including all skill groups in one year. Thus, I generate dummy variables for low-skilled, medium-skilled and high-skilled employees (/B6.,//, 0B6.,//, +B6.,//) with unskilled workers as the reference group. Furthermore, all skill group dummies interact with the border region dummy (/6.,//%255(*, 06.,//%255(*, +6.,//%255(*). Consequently, I now analyse not only the deviations of the wage differential in the borderlands, but also the development of the wage differentials between the different skill groups. The results for the variables with respect to the skill level and the region are summarised in Table 4.15 (see Table A 4.3 in the appendix for control variables). The values for the coefficient of %255(* show that in 2002 unskilled workers in the border region earned significantly more (11.2 %) than unskilled workers in the non-border region. The wage differentials between the skill groups increased above all in the early transition years from 1992 until 1996 133 Chapter 4 Conclusion and remained almost stable afterwards.29 Regarding the interaction terms between the skill and border region dummies, the outcome only yields significant results in 2002. Based on the wage differential for the reference group (the unskilled workers), all other skill groups are in an inferior position in the border region, which is consistent with my previous results. Table 4.15: Regression results for qualificational wage differentials variable 1992 1996 2002 coef. t-Stat. coef. t-Stat. coef. t-Stat. BORREG –0.0181 –1.11 –0.0170 –1.18 0.1121*** 2.61 L_SKILL 0.0418*** 3.69 0.0612*** 6.32 0.0904*** 3.66 M_SKILL 0.1635*** 11.62 0.2169*** 18.82 0.2299*** 8.37 H_SKILL 0.3364*** 16.32 0.4360*** 24.79 0.4381*** 12.09 LSKILL*BORREG 0.0164 0.88 0.0126 0.79 –0.1218*** –2.66 MSKILL*BORREG –0.0028 –0.14 –0.0030 –0.18 –0.1481*** –3.17 HSKILL*BORREG –0.0155 –0.63 –0.0053 –0.21 –0.1491** –2.58 Test statistics N 12,964 19,522 4,879 R² 0.4454 0.4486 0.4343 Dependent variable: ln Wage Control variables: see Table A 4.3 in the appendix Source: Author’s own calculations from Czech Microcensus 1992, 1996, 2002. Notes: Regression with heteroskedasticity-robust standard errors; */**/*** significant at the 10/5/1 percent level. 4.6 Conclusion In this chapter I analysed the development of several labour market indicators in the Czech Republic after the fall of the Iron Curtain, comparing the districts close to Bavaria and/or Austria with the rest of the country. Hypotheses can be derived from two theoretical strands: the Feenstra-Hanson trade model dealing with the skill intensity of outsourced production activities and the Brülhart et al. NEG model referring to the market potential and import competition. In the early transition years (from 1992 until 1996) the relative employment share of the border region increased and then stabilised by 2002. Contrary to my hypotheses I do not find clear evidence of disproportionate shifts in the economic 29 7KHVHUHVXOWVFRUUHVSRQGWRWKHILQGLQJVRI9HèFHUQÊNq,QWKHtSHULRGWKHHIIHFWRIHGXFDWLRQ stagnated …” IAB-Bibliothek 321 134 Labour Market Effects in the Czech Border Region structure in the Czech districts bordering on Bavaria and Austria compared to the non-border districts. With respect to both industries and occupations the shifts proceeded more or less in a similar way with some exceptions, e.g. clerks and the manufacturing sector. Calculating an indicator of structural change and a specialisation index yields higher values in the period from 1996 until 2002. This is not surprising, not only because of the longer time span, but also due to the troubling recession years. In the period under review a skill-upgrading process took place all over the country. Distinguishing between four skill groups, the skill structure of employed and unemployed people changed analogously in both areas of observation, i.e. the trend towards more skilled labour led to noticeable shifts in the Czech border region as well as in the remaining districts. The descriptive statistics are confirmed by the results of econometric estimations in each case (employed and unemployed). Regarding wage differentials between workers employed in the border region and workers in the rest of the country, I first took a look at the descriptive figures and then ran several regressions, obtaining robust results: in 1992 border region employees generally earned slightly less than their peers in the non-border districts (about 1–2 %). While there was not so much variation until 1996, the picture changed between 1996 and 2002. The workers with the lowest skill degree exhibit a positive wage differential of around 12 % in the border region compared to their counterparts in the non-border region. For all other skill groups in the border region the spatial wage gap is negative and, in absolute value, increases with the skill level. These results clearly contradict the predictions of the Feenstra-Hanson model, according to which the skill upgrading should be especially noticeable in the border region. However, the findings meet with the expectations of the NEG model by Brülhart et al., according to which above all industries and employees are in a favourable position in the border districts where import competition from Germany and Austria is low, i.e. less human-capital-intensive activities benefit. Of course, these results only indicate the effects of economic integration in an ongoing process, which is far from being completed. The effects of the Czech Republic’s accession to the EU still have to be analysed, not to mention the impact of free movement of labour, which will bring new opportunities to the Czech workforce in 2011 at the latest. Since the Czech Republic is surrounded by old and new EU member states, the country is predestined for further research on integration effects. 135 Chapter 4 Appendix to Chapter 4 Table A 4.1: Distance from district capital to next Bavarian or Austrian international border crossing (in minutes by car) District min border crossing District min border crossing Praha 120 :DLGKDXV Liberec 194 :DLGKDXV %HQHxRY 114 *UDPHWWHQ Semily 195 :DLGKDXV Beroun 90 :DLGKDXV Hradec Králové 191 *UDPHWWHQ Kladno 119 :DLGKDXV -LèFÊQ 184 :DLGKDXV .ROÊQ 147 *UDPHWWHQ Náchod 227 *UDPHWWHQ Kutná Hora 131 *UDPHWWHQ 5\FKQRYQDG.QèH{QRX 189 'UDVHQKRIHQ 0èHOQÊN 156 :DLGKDXV Trutnov 232 *UDPHWWHQ Mladá Boleslav 155 :DLGKDXV Chrudim 148 *UDPHWWHQ Nymburk 153 :DLGKDXV Pardubice 167 *UDPHWWHQ Praha-východ 120 :DLGKDXV Svitavy 116 'UDVHQKRIHQ Praha-západ 120 :DLGKDXV ·VWÊQDG2UOLFÊ 155 'UDVHQKRIHQ 3èUÊEUDP 105 3KLOOLSVUHXW +DYOÊèFNX °v Brod 94 *UDPHWWHQ 5DNRYQÊN 119 :DLGKDXV Jihlava 86 .OHLQKDXJVGRUI è &HVNÆ%XGèHMRYLFH 41 :XOORZLW] 3HOKèULPRY 61 *UDPHWWHQ è Ceský Krumlov 33 :XOORZLW] 7èUHEÊèF 68 .OHLQKDXJVGRUI -LQGèULFKX °v Hradec 24 *UDPHWWHQ nG’ár nad Sázavou 103 'UDVHQKRIHQ 3ÊVHN 88 :XOORZLW] Blansko 84 'UDVHQKRIHQ Prachatice 42 3KLOOLSVUHXW %UQRPèHVWR 49 'UDVHQKRIHQ Strakonice 56 3KLOOLSVUHXW Brno-venkov 49 'UDVHQKRIHQ Tábor 76 *UDPHWWHQ %èUHFODY 27 'UDVHQKRIHQ 'RPD{OLFH 19 )XUWKL: +RGRQÊQ 50 'UDVHQKRIHQ Klatovy 47 )XUWKL: 9\xNRY 65 'UDVHQKRIHQ 3O]HèQPèHVWR 52 :DLGKDXV Znojmo 15 .OHLQKDXJVGRUI 3O]HèQMLK 52 :DLGKDXV -HVHQÊN 205 'UDVHQKRIHQ 3O]HèQVHYHU 52 :DLGKDXV Olomouc 100 'UDVHQKRIHQ Rokycany 66 :DLGKDXV 3URVWèHMRY 81 'UDVHQKRIHQ Tachov 26 :DLGKDXV 3èUHURY 107 'UDVHQKRIHQ Cheb 13 6FKLUQGLQJ Šumperk 149 'UDVHQKRIHQ Karlovy Vary 54 6FKLUQGLQJ .URPèHèUÊ{ 87 'UDVHQKRIHQ Sokolov 35 6FKLUQGLQJ 8KHUVNÆ+UDGLxWèH110 'UDVHQKRIHQ 'èHèFÊQ 199 6FKLUQGLQJ 9VHWÊQ 160 'UDVHQKRIHQ Chomutov 104 6FKLUQGLQJ =OÊQ 126 'UDVHQKRIHQ /LWRPèHèULFH 164 :DLGKDXV Bruntál 162 'UDVHQKRIHQ Louny 134 6FKLUQGLQJ )UÚGHN0ÊVWHN 165 'UDVHQKRIHQ Most 126 6FKLUQGLQJ Karviná 203 'UDVHQKRIHQ Teplice 152 6FKLUQGLQJ 1RYÚ-LèFÊQ 141 'UDVHQKRIHQ ·VWÊQDG/DEHP 168 6FKLUQGLQJ Opava 169 'UDVHQKRIHQ è &HVN¾/ÊSD 202 :DLGKDXV 2VWUDYDPèHVWR 168 'UDVHQKRIHQ Jablonec nad Nisou 185 :DLGKDXV Source: Author’s own calculations from Internet Route Planner ViaMichelin. Notes: District: 77 Czech NUTS 4 level districts; min: distance in minutes by car; border crossing: next Bavarian or Austrian international border crossing. 142 IAB-Bibliothek 321 Summary and Outlook employees in the Czech border region are the relative winners in the course of increasing trade relations with Germany. What implications do the results provide for regional policy? First, it can be said that the fears raised by the population and politicians in eastern Bavaria about the region falling behind were exaggerated. The main conclusion of the results is that the Bavarian border region did quite well after the opening of the border. The increasing shares of relative employment and skilled employees indicate that human capital – a vital growth factor – plays a growing role in eastern Bavaria. However, the negative wage differential becoming larger for low-skilled and skilled workers in the second half of the observation period is cause for concern with regard to the future attractiveness of the borderland. Assuming that the prospects for activities using low-skilled labour relatively intensively are rather poor, regional policy could underpin the lasting creation of highly skilled jobs in the region. Examples of possible starting points are the strengthening of the regional universities, the durable promotion of centres for founders and technology and the provision of risk capital. Since both the Bavarian and the western and southern Bohemian border region hold a strong position in the manufacturing and engineering industries, cross-bor- der co-operations in these fields pose promising opportunities. The completion of the A6 motorway has improved the traffic and transport opportunities considerably. The extension of international railway connections would certainly make an additional contribution to the economic development of eastern Bavaria. Regarding the Czech Republic it seems obvious that the capital city of Prague absorbs a sizeable fraction of the available human capital. Apart from this agglomeration pull the results for wage differentials suggest that the Czech border region is running the risk of becoming less attractive for qualified workers. In order to counteract the potential outflow of skilled workers it might be important to support the attraction of locations like Brno, Pilsen and Budweis, which could then also bring forward less densely populated areas. Of course, due to the current state of data availability this thesis only marginally addresses the consequences of the EU enlargement. Since the accession of the Czech Republic to the EU in May 2004 is one large part of an ongoing integration process, its effects should however act in the same direction as those described above. In general, the interactions between the German and the Czech labour markets leave scope for further research. Now surrounded only by EU member states, the Czech Republic holds a favourable position. Germany (and Austria) is obliged to lift the restrictions on the free movement of labour from the Eastern European countries by 2011 at the latest. This, certainly, will affect the economies of the countries which have a common border with CEE countries. However, in contrast to the widely negative assessment of the popular press, it 143 Chapter 5 Summary and Outlook is not so obvious which countries will profit and which will suffer from labour mobility. The shortage of skilled labour (“brain drain”), for instance, is becoming an increasing problem in the Czech Republic which could worsen by the easier access on the western European labour markets. On the other hand, it is hardly predictable to what extent the incentive of higher incomes in a neighbouring country will influence the Czechs, who have so far been rather immobile. In any case, there is no lack of research issues. 145 IAB-Bibliothek 321 List of Figures Figure 1.1: Unemployment rate of low-skilled persons in EU25 countries ..................................................................................................... 12 Figure A 1.1: German exports to and imports from the Czech Republic .......... 15 Figure A 1.2: FDI figures of German companies in the Czech Republic ............ 15 Figure A 1.3: Bavarian exports to and imports from the Czech Republic ........ 16 Figure A 1.4: FDI of Bavarian companies in the Czech Republic ........................ 16 Figure A 1.5: Regional distribution of German holding companies operating in the Czech Republic ......................................................... 17 Figure A 1.6: Regional distribution of affiliated firms in the Czech Republic: (a) affiliated to German companies and (b) affiliated to Bavarian companies ........................................................................... 18 Figure A 1.7: Market entry of German companies in the Czech Republic: (a) overall and (b) subdivided into economic sectors .................... 19 Figure A 1.8: Distribution of affiliated firms between manufacturing, trading and service sector ..................................................................... 20 Figure 2.1: Market potential in a border region ................................................... 22 Figure 2.2: Production-possibility frontier between inputs 1 and 2 .............. 27 Figure 2.3: Isocost lines of input activities ............................................................ 28 Figure 2.4: Feenstra-Hanson model without capital mobility ......................... 30 Figure 2.5: Relative labour demand without capital mobility ......................... 31 Figure 2.6: Feenstra-Hanson model including capital mobility....................... 32 Figure 2.7: Relative labour demand with capital mobility ................................ 33 Figure 2.8: Utility differential: (a) in the case of autarky and (b) in the case of trade with the foreign country ...................................... 40 Figure 3.1: Eastern Bavarian border region ........................................................... 54 Figure 3.2: Employment shares of 16 industries in eastern Bavaria and western Germany ............................................................................. 58 Figure 3.3: Structural change indicator in eastern Bavaria and western Germany ...................................................................................................... 61 Figure 3.4: Krugman specialisation index for eastern Bavaria ........................ 61 Figure 3.5: Correlation of the differences between the growth of the relative employment shares in eastern Bavaria and western Germany (from 1980 until 2001) and the differences in the relative employment shares in 1980 .................................................. 61 Figure 3.6: Employment share of workers employed in eastern Bavaria ...... 62 Figure 3.7: Czech commuters in Bavaria covered by the German social security system ............................................................................. 63 IAB-Bibliothek 321 146 List of figures Figure 3.8: Share of Czech employees in the eastern Bavarian workforce .................................................................................................... 64 Figure 3.9: Deviation of skill group shares for workers in the eastern Bavarian border region from the averages in western Germany ..................................................................................... 65 Figure 3.10: Deviation of skill group shares for workers in the eastern Bavarian border region from the averages in the district types 5-9 in western Germany ...................................... 66 Figure 3.11: Shares of skill groups for workers in eastern Bavaria and western Germany ............................................................................ 67 Figure 3.12: Deviation of skill group shares for unemployed people in the eastern Bavarian border region from the averages in the district types 5-9 in western Germany ................................. 68 Figure 3.13: Shares of skill groups for unemployed people in the district types 5-9 in eastern Bavaria and western Germany ...................................................................................................... 69 Figure 3.14: Wage effect for low-skilled workers in the border region .......... 77 Figure 3.15: Wage effect for skilled workers in the border region ................... 77 Figure 3.16: Wage effect for high-skilled workers in the border region ......... 77 Figure 3.17: Gender-specific wage effect for German employees in eastern Bavaria .................................................................................... 80 Figure 3.18: Additional wage effect for foreign employees in eastern Bavaria ......................................................................................... 82 Figure 3.19: Additional wage effect for non-Czech foreign employees in eastern Bavaria .................................................................................... 84 Figure 3.20: Estimation of average treatment effect on the treated (att) for German employees in eastern Bavaria ........................................ 88 Figure 3.21: Difference-in-differences wage effect in eastern Bavaria .......... 91 Figure 3.22: Fixed effects estimations: wage effect in eastern Bavaria ......... 93 Figure A 3.1: Southern and northern part of the eastern Bavarian border region ............................................................................................. 98 Figure A 3.2: Fixed effects estimations: wage effect in southeastern Bavarian border region ........................................................................... 100 Figure A 3.3: Fixed effects estimations: wage effect in northeastern Bavarian border region ........................................................................... 101 Figure 4.1: Czech NUTS 3 and NUTS 4 regions ..................................................... 108 Figure 4.2: Share of full-time employees working in the Czech border region: (a) including Prague and Mladá Boleslav, (b) without Prague and Mladá Boleslav .......................................................................................... 110 147 IAB-Bibliothek 321 List of Figures Figure 4.3: Shares of skill groups of full-time workers comparing the Czech border region to the rest of the country....................... 115 Figure 4.4: Number of unemployed in the different skill groups .................... 117 Figure 4.5: Relative shares of skill groups of unemployed people comparing the Czech border region to the rest of the country .................................................................................................... 118 Figure 4.6: Wage differential between Czech border and non-border region .................................................................................... 121 Figure 4.7: Skill premium in the border and the non-border region by skill group and decile ........................................................................ 122 Figure 4.8: Wage effect for (a) lower and (b) more highly skilled workers in the Czech border region .................................................... 129 149 IAB-Bibliothek 321 List of Tables Table A 2.1: Implications of the Feenstra-Hanson trade model and the Brülhart et al. NEG model .............................................................. 50 Table 3.1: Classification of German skill groups................................................. 53 Table 3.2: Regional classification scheme based on BBR classification ...... 54 Table 3.3: Locational fixed effects for Bavarian districts ................................ 55 Table 3.4: Change in the employment shares in eastern Bavaria and western Germany. ..................................................................................... 60 Table 3.5: Czech commuters in the Bavarian employment office districts in 1995 ........................................................................................................ 63 Table 3.6: Estimation results for the share of low-skilled, skilled and high-skilled workers ................................................................................ 71 Table 3.7: Estimation results for the share of low-skilled, skilled and high-skilled unemployed people .......................................................... 73 Table 3.8: Variables of the wage equation (Germany) ...................................... 74 Table 3.9: Values of EXPER (Germany) ................................................................... 75 Table 3.10: Results for the coefficient τ of BORREG in the wage equation for three skill groups ............................................................................... 76 Table 3.11: Basic wage differential for German employees in 1980 in difference-in-differences estimations .......................................... 90 Table 3.12: Basic wage differential for German employees in fixed effects estimations ................................................................................................. 92 Table A 3.1: Basic wage differential for German employees in fixed effects estimations in the southeastern Bavarian border region ............ 99 Table A 3.2: Basic wage differential for German employees in fixed effects estimations in the northeastern Bavarian border region ............. 99 Table 4.1: Sample size of the Czech Microcensus in 1992, 1996 and 2002 ............................................................................................................. 107 Table 4.2: Classification of Czech skill groups .................................................... 107 Table 4.3: Employment shares of occupations in the Czech non-border and border region ..................................................................................... 111 Table 4.4: Employment shares of industries in the Czech non-border and border region ..................................................................................... 113 Table 4.5: Indicator of structural change and Krugman specialisation index for occupations and industries in the Czech Republic ...... 114 Table 4.6: Gross wages in the Czech border and the non-border region ... 120 Table 4.7: Skill bonus by regional type in the Czech Republic ....................... 121 IAB-Bibliothek 321 150 List of Tables Table 4.8: Estimation results for the share of skill groups of Czech employees ................................................................................................... 123 Table 4.9: Estimation results for the share of skill groups of Czech unemployed people .................................................................................. 125 Table 4.10: Variables of the wage equation (Czech Republic) .......................... 127 Table 4.11: Values of EXPER (Czech Republic) ..................................................... 127 Table 4.12: Estimation results for the wage effect of lower and more highly skilled workers without pooling cross sections ................. 128 Table 4.13: Estimation results for the wage effect in differencein-differences estimations ................................................................... 131 Table 4.14: Wage effect in the border region controlling for industries ....... 132 Table 4.15: Regression results for qualificational wage differentials ............ 133 Table A 4.1: Distance from district capital to next Bavarian or Austrian international border crossing ............................................................... 135 Table A 4.2: DID estimation controlling for industries – full set of variables ...................................................................................................... 136 Table A 4.3: Qualificational wage differentials – full set of variables............. 137 151 IAB-Bibliothek 321 List of Abbreviations DV  DNFLRY¾VSROHèFQRVWMRLQWVWRFNFRPSDQ\ BeH Beschäftigten-Historik (employment register) CEPR Centre for Economic Policy Research CEEC Central and Eastern European Countries CERGE-EI Center for Economic Research and Graduate Education – Economics Institute CES Constant Elasticity of Substitution CESifo Center for Economic Studies – Institut für Wirtschaftsforschung COMECON Council of Mutual Economic Assistance è CSÚ è &HVNÚ6WDWLVWLFNÚ·èUDG&]HFK6WDWLVWLFDO2IILFH DID difference-in-differences ed(s). edition/editor(s) EU European Union FDI Foreign Direct Investment GDP Gross Domestic Product Hrsg. Herausgeber (editors) HWWA Hamburgisches Welt-Wirtschafts-Archiv (Hamburg Institute of International Economics) IAB Institut für Arbeitsmarkt- und Berufsforschung (Institute for Employment Research) IABS IAB-Beschäftigtenstichprobe (IAB employment sample) IES Institute of Economic Studies, Prague ILO International Labour Organization ISC Indicator of Structural Change ISCED International Standard Classification of Education ISCO International Standard Classification of Occupations ISSP International Social Survey Programme IW Institut der deutschen Wirtschaft IZA Institut zur Zukunft der Arbeit (Institute for the Study of Labor) Jg. Jahrgang (year) .èF  .RUXQèFHVNÚFK&]HFKFURZQV KSI Krugman Specialisation Index LFS Labour Force Survey NACE Nomenclature of Economic Activities NBER National Bureau of Economic Research NEG New Economic Geography