The Impact of Economic Integration on
Structural Change and Productivity Growth:
The Case of the Central and Eastern European Countries
Master Thesis
M.A. in International Economics
M.Sc. in International Economics and Business
15th June, 2015
Author: Benedikt Christian Walter E-Mail: benedikt.c.walter@gmail.com Student ID: S2801426
Supervisor:
Associate Professor Dr. Bart Los
Faculty of Economics and Business at the University of Groningen
Co-Assessor:
Assistant Professor Dr. Astrid Krenz
Abstract
Has structural change contributed to productivity growth in the Central and Eastern European Countries (CEECs) that joined the EU? Has European economic integration fostered labor shifts from low to high productive sectors? Using a shift share analysis this paper examines, in a first step, the contribution of labor reallocation between sectors to aggregate productivity growth (structural change term). In a second step a panel data regression is used to identify variables, related to economic integration, which have contributed to the structural change term. Other studies on the determinants of structural change have often solely concentrated on factors within countries that drive the shift of labor between sectors, neglecting the role of external factors, like labor migration, FDI and trade. In contrast to that, this paper finds that structural change has in some countries contributed to productivity growth and is significantly related to variables and indicators of economic integration.
iv
1 Introduction ... 1
2 Measuring Economic Integration ... 2
3 Economic Integration of the New EU Member States ... 3
4 Structural Change and Productivity Growth ... 6
4.1 Structural Change and Productivity Growth: Theoretical Relations ... 7
4.2 Structural Change and Productivity Growth: Empirical Findings ... 8
5 Drivers of Structural Change: The Integration Variables ... 10
5.1 Foreign Direct Investment ... 10
5.2 Trade Integration ... 11 5.3 Labor Migration ... 12 6 Data ... 13 7 Decomposition Analysis ... 15 8 Decomposition Results ... 17 9 Regression Analysis... 23 9.1 Independent Variables ... 24 9.2 Control Variables ... 27 9.3 Model Specification ... 29
9.4 Descriptive Statistics and Correlations ... 30
9.5 Methodological Issues ... 33
10 Regression Results and Discussion ... 37
11 Limitations and Robustness ... 43
12 Final Conclusions ... 44
13 References... 46
List of Figures
Figure 1: Intra-EU exports as share of GDP ... 4
Figure 2: Real FDI inflow as share of GDP ... 5
Figure 3: Stock of high skilled emigrants in the EU-15. ... 6
Figure 4: Correlation between change in labor productivity and change in employment shares ... 21
Figure 5: Shift term by country ... 32
Figure 6: Outliner detection. ... 33
Figure 7: The marginal effect of FDI on the structural change term ... 39
Figure 8: FDI variable ... 53
Figure 9: Emigration variable ... 54
Figure 10: Trade variable ... 54
Figure 11: Undervaluation variable ... 55
Figure 12: Labor flexibility variable ... 55
List of Tables
Table 1: Overview of ISIC rev. 3 Industry Classification ... 14Table 2: Comparison of conventional versus modified decomposition analysis ... 18
Table 3: Decomposition of aggregate labor productivity growth ... 19
Table 4: Decomposition of the productivity growth rates for five 4-year periods ... 20
Table 5: Decomposition results for the CEECs before and after 2004 ... 22
Table 6: Overview and non-statistic descriptions of the variables ... 30
Table 7: Bivariate correlation coefficients ... 31
Table 8: Regressions for the full sample ... 38
Table 9: Regressions for the subsample before 2004 ... 41
Table 10: Regressions results for the subsample after 2004 ... 42
Table 11: Regressions for the full sample ... 52
Table 12: Descriptive Statistics ... 53
1
Introduction
More than ten years after the 2004 EU-enlargement it is time to take stock: What was the effect of economic integration on structural change and productivity growth? In January 2015 Leszek Balcerowicz, the father of the famous “Balcerowicz Plan” that transformed Poland into a market economy, gave an interview in which he identified productivity growth as one of the most important aims for future economic reform in Poland. Asked for the biggest challenge Balcerowicz’s answer was, “Reallocating labor form low to high productive sectors” (Benz 2015). Balcerowicz’s statement exactly describes what this paper is about: structural change and productivity growth.
Two research questions will be addressed: Has structural change contributed to productivity growth in the Central and Eastern European Countries (CEECs) that joined the EU? 1 And if it did so, has European economic integration fostered productivity enhancing structural change? Measuring GDP in per capita terms and in purchasing power standards Warsaw, Prague and Bratislava have today a GDP higher than Vienna (Darvas 2014). This powerful illustration of the economic success of the CEECs gives an intuition for the deep structural changes the CEECs have gone through. Overall productivity growth has been much higher in the CEECs than in the rest of the EU. From 1995 to 2014 labor productivity grew on average by 3.3% per year while in the EU-15 it grew by only 0.8%.2 Studies found little or no contribution of structural change to productivity growth before the EU accession in 2004 and 2007 (Havlik 2008). Less evidence is available for the time after 2004. This paper fills this gap by analyzing developments in the economic structure of the CEECs and examining the contribution of structural change to productivity growth for the time 1995-2010.
Many studies acknowledge the contribution of structural change to productivity growth but miss the opportunity to look into the factors driving structural change, and even if they do, structural change is understood mainly as a process determined by factors, which solely depend on local structures and circumstances (see e.g. McMillan et al. 2014). External factors of structural change, which are directly related to a country´s integration in the regional or world economy are hardly examined in the literature. Given the huge steps in economic integration of the CEECs,
1 The analysis includes the Central and Eastern European Countries that joined the EU in 2004 (the Czech
Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, and Slovenia) and 2007 (Bulgaria and Romania). Two other countries joined the EU in 2004 (Cyprus and Malta). However, the focus is on the CEECs as they share a common cultural and historic background, excluding Cyprus and Malta (and Croatia, which joined in 2013) from the analysis. Furthermore, EU-15 is used as a term to characterize the “old” member states, which constituted the EU before the 2004/2007 enlargement.
2 Data taken from The Conference Board Total Economic Database, real GDP in PPP per person employed,
accomplished in the last two decades, it seems to be reasonable that external factors had played a crucial role in the process of economic reconstructing. E.g. exports from the CEECs to the EU-15 as share of GDP have on average increase from 20% in 1995 to more than 50% in 20EU-15 (IMF 2015). In this paper a systematic analysis will identify external drivers of structural change related to European economic integration.
Using a comprehensive sectoral data set from the World Input-Output Database (WIOD) of the University of Groningen a decomposition analysis will identify whether structural change had enhancing or reducing effects on productivity growth. The analysis decomposes aggregate labor productivity growth into an increase in labor productivity at the sectoral level (within term) and labor productivity growth due to a shift of labor across sectors (structural change term). In a second step a regression analysis is used to identify the variables that influenced the structural change term. European economic integration is characterized by the free movement of labor, capital, services and goods. Three variables Exports to the EU-15, FDI Inflow and Labor Emigration - closely related to European economic integration - are included as independent variables in the analysis. The efficient reallocation of labor is a driver of development. The results of the analysis can therefore contribute to the discussion on economic convergence within the EU. Convergence of living standards is one of the main objectives of European integration defined by the treaties of the EU. The question if European economic integration can give an impetus to productivity enhancing structural change is, therefore, of interest for researcher and policy maker (see e.g. Albu (2012) or Bongardt et al. (2013) for an overview on the convergence discussion).
The first section of the paper provides an overview of the economic integration process of the CEECs and describes developments in terms of FDI flows to and emigration and exports from the CEECs. Theoretical arguments and recent studies on the relationship of economic integration, structural change and productivity growth are presented in the subsequent part. This sets the scene for the empirical analysis, which is presented in section 7 and 8. The core of the paper, empirical results and their discussion, is presented in section 9. The last section outlines the limitations of the analysis and draws the conclusions.
2
Measuring Economic Integration
Rayp and Standaert 2015). Measures of economic integration based on the law of one price often make the assumption of perfect competition and homogenous goods, which causes problems in the practical use of the concept. The second approach, focusing on quantities, uses indicators of openness as measures for integration, e.g. often simply trade (export plus imports) relative to GDP. This measures clearly neglect the complexity of economic integration with its multidimensional flows (investment flows, labor flows, export flows). Other measures acknowledge the multidimensional characteristics by constructing indices taking into account the economic, political and social interactions between regions or countries (see e.g. Martens and Zywietz 2006, Dreher 2006, König and Ohr 2013). Further indirect measurements of economic integration exist, e.g. examining the degree of openness in terms of barriers to trade, either due to legal circumstances (tariffs, quotas, etc.) or due to natural barriers like geographic distance. However, this de jure measures are not true measures of economic integration rather an indicator for obstacles to integration (Arribas et al. 2007).
This paper uses measures based on three central features of European economic integration: Trade integration, foreign direct investment (FDI) and labor market integration. European economic integration is a complex process and not limited to the mentioned aspects. To draw a complete picture of European economic integration is beyond the scope of this paper and it will be argued that this basic variables offer a useful tool to shed light on the role of economic integration in structural change.
3
Economic Integration of the New EU Member States
The transition process of the CEECs from planned to market based economies started with a deep recession. GDP and employment slumped to levels, which wiped out the economic progress of decades as output declined by about 15% per year from 1989 until 1992 (Havlik 2014). But looking back the development process was a success story. From 1995 to 2012 real GDP grew on average by 3.4% while the EU-15´s real GDP growth rate amounted to only 1.7% on average (Dobrinsky and Havlik 2014). The process was unequal within the CEECs with Slovenia, Poland, the Czech Republic and the Baltic States as the avant-garde of the catching-up process. Romania and Bulgaria still lag behind in terms of GDP (Eurostat 2015a).
manufactured goods down to zero by 2002 (Crespo and Fontoura 2007). Restrictions on agricultural goods were maintained and the accession countries did not yet become part of the customs union, which applies a common external tariff on non-EU products (Baas and Brücker 2011). Figure 1 shows a remarkable increase of CEECs exports to the EU-15 relative to GDP. In 1995 the CEECs had on average an export-GDP-ratio of 20%. This figure increased to more than 50% in 2013, what is a clear indicator of the increased importance of European trade for the CEECs. However, there is a huge variation between the countries, as indicated by the vertical bars, which show the maximum and minimum values among the CEECs.
Foreign Direct Investment: Already since the early 1990s FDI inflow to the transition countries had followed an upwards trend. A key role in the explanation of FDI inflow at the first stages of transition had the privatization of former state owned enterprises, which got acquired by foreign firms (Havlik 2013). This process was completed in most of the CEECs around 1995. However, the rise of FDI inflow in the early 2000s was still related with the privatization process in Bulgaria and Romania. In recent years FDI was driven by enterprises from the EU-15 setting up new investments, so called greenfield investments (Kornecki and Raghavan 2011). The increase of FDI inflow in the early 2000s was stimulated by the precursors of the EU accession like the clarification of accession dates and the accompanying positive expectation of companies (Kalotay 2010). The de jure accession of eight countries to the EU in 2004 was the beginning of a period of increased FDI inflow which was abruptly interrupted by the global economic and financial crisis in
2008/2009 (see Figure 2). Looking on the sectoral composition of the FDI, it becomes apparent that FDI to the tradable sector played an important role in almost all CEECs especially in Slovakia, Romania, Hungary and Poland. Estonia and Latvia were remarkable exceptions (Kinoshita 2011).
Labor market integration: The free movement of labor has been regarded as the most controversial step in the economic integration of the CEECs and sparked heated discussions in the mid-2000s. Concerns about job security in the EU-15 led to a system with a gradual opening of the labor market for CEECs migrants. Ireland, Sweden and the UK abolished their labor market restrictions immediately after the accession. Eight member states opened their labor markets between 2006 and 2009, while Germany and Austria lifted the restrictions not until 2011 (Anderson 2015). In 2004 the number of registered person from the CEECs in the EU-15 was about one million. Until 2007 this figure grew up to 1.8 million. In the UK the number of people from the CEECs grew almost fivefold up to 0.7 million in 2009. Post-enlargement immigration has increased the EU-15 population by approximately 1%, and had a remarkable impact on the population of the CEECs with an emigration ranging from 1.8% up to 4.8% of the respective country´s population (Zaiceva and Zimmermann 2012). A study for the European Commission conducted by Holland et al. (2011) estimates the emigration from the CEECs to the EU-15 to 1.8% of the CEECs population.
The CEECs made substantial progress in increasing the qualification levels of their population during the last decades. The number of university graduates increased sharply, while dropout rates declined (Baas and Brücker 2011). In 2006 25% of Polish emigrants had a tertiary education. In the whole population this number was 15%. Education among usually young emigrants differs from education levels of the whole population with another age distribution. Brücker and Baas (2009) therefore adjust this number to a comparable age distribution and confirm that emigrants had a higher average education than the home country population. The authors conclude that emigration, at least from Poland, was mainly high skilled. Figure 3 shows the share of high skilled emigrants relative to the labor force of the CEECs. A clear increase in high skilled emigration relative to the home country labor force can be observed from 2004 onwards. The emigration stock drastically declined when the economic and financial crisis hit the EU-15 labor markets and triggered high numbers of migrants to return to their home countries.
4
Structural Change and Productivity Growth
The last section has given an indication of how the economic integration process was accompanied by an inflow of FDI, increased trade and the movement of labor between the EU-15 and the CEECs. In a next step the effects of the economic integration variables on structural change are introduced, based on theoretical considerations.
4.1 Structural Change and Productivity Growth: Theoretical Relations
Productivity is a measure for the efficiency with which inputs are converted into outputs. In general there are two ways in which productivity can be measured: Labor productivity and Total Factor Productivity (TFP) (Gordon et al. 2015). TFP is defined as the difference between the growth in inputs and the growth in output. In other words: TFP causes the growth in output, which is left when for all growth in inputs is accounted. For example TFP is measured as output growth not caused by a growth in the production factors labor and capital. Rising labor productivity has two sources, it can stem from growth in TFP (e.g. through better production technologies or management techniques) and growth of capital relative to labor (capital intensity). Measuring productivity as TFP thus allows to distinguish between the sources of labor productivity growth: TFP growth or capital accumulation. However, the research question of this paper asks if structural change was directed to high productive sectors. The question why some sectors are more productive than others e.g. if labor productivity growth stems from increases in the capital stock or increase in TFP is different from that. In this analysis productivity is therefore defined as gross value added per worker employed. Nevertheless, it needs to be acknowledged that the chosen productivity measure can alter the results. Brown and Earle (2008) apply a decomposition analysis to measure the contribution of labor shifts to aggregate productivity growth in transition countries, among them some EU-accession countries. The authors use two different productivity definitions - TFP and labor productivity - and find e.g. higher contributions of labor reallocations to TFP growth than to labor productivity growth for Lithuania during 1995-2005. The reverse is true for Romania, where the contribution of labor shifts to labor productivity growth is higher than the contribution of labor shifts to TFP growth.
excess labor in the low productive sector than in the high productive sector, reallocation reduces the output in the low productive sector less than it increases it in the high productive sector. It should be noted that labor reallocation not necessarily means that worker formerly employed in a certain sector start working in another, rather is the exit and entry of worker in the long-run a typical way through which structural change unfolds (Abowd et al. 1999).
The hypothesis of productivity increasing shifts in labor also implies the possibility of productivity decreasing structural change. This “structural change burden” (Baumol 1967) has a negative impact on productivity if labor moves from high productive parts of the economy to parts with low productivity growth. Theoretically this reverse effect arises due to the restricted inherent ability of industries to increase their productivity. E.g. in the service sector where the scope for labor productivity increases is normally assumed to be small due to a high labor intensity. In expanding service industries productivity increases through technological progress are limited what, on the other hand, might not be the case in more capital intensive industries. Increasing shares of labor employed in such “productivity constrained” industries cause the economy to observe productivity decreasing structural change. The shift from industry to services is a distinctive feature of many developed countries since the 1970s (Jorgenson and Timmer 2011). Different arguments have been raised to explain this phenomenon. Demand-driven factors focus on shifting consumer preferences towards services, which become more demanded when living standards rise and basic needs are satisfied. For example, differences in income elasticities over time lead to changes in the structure of demand and allow some sectors to grow and forces others to shrink. Other explanations focus on the change in the demand for intermediate products. Services like marketing, R&D and management have become the main source of competitive advantage of firms (Peneder et al. 2001). However, there are counterarguments against Baumol´s proposition of a “structural change burden”. The service industry is very heterogeneous e.g. driven by advances in ICT some branches of the service sector saw high productivity growth. Empirical support for this argumentation comes from Timmer and de Vries (2009). The authors find that productivity increases in the services sector have been more important for growth than productivity increases in the manufacturing sector for a set of countries from Asia and Latin America.
4.2 Structural Change and Productivity Growth: Empirical Findings
confirmed. For example, Timmer and de Vries (2009) apply a shift-share analysis for the time between 1950 and 2005 to gauge the long-term effect of structural change on productivity growth. Measuring productivity as GDP per worker they find productivity increases mainly within sectors and small contributions of productivity growth caused by reallocations of employment towards more productive sectors. In a study of 38 countries McMillan et al. (2014) find positive contributions of structural change on productivity growth for the emerging economies of Asia with labor moving from low-productive to high-productive sectors. They confirm the finding of the relative lower importance of structural change on productivity growth for developed nations. These findings give support for the theoretical prediction that for countries on lower stages of development and therefore more labor in low productive sectors the reallocation of labor is more important than for more developed countries.
The empirical literature on structural change and productivity growth for the CEECs is less comprehensive. Havlik (2008) provides evidence for the CEECs and concludes that shifts of labor among individual sectors had no marked impact on aggregate productivity growth. Productivity growth was mainly rooted in productivity increase within sectors. A shift share analysis reveals that, except Poland, more than 90% of aggregate productivity growth in the period 1995-2004 can be attributed to productivity growth within individual sectors. This finding is interesting given the considerable reconstruction of the CEECs in the transition time and other findings on emerging and transition countries which support the hypothesis of productivity enhancing structural change. This result also stands in contrast to the case of Russia where positive effects of labor shifts to productivity are found and which had similar starting conditions as the CEECs (Havlik 2008). However, productivity growth in the transition period of Russia was well below that of the CEECs and it can be assumed that huge parts of the structural adaption in the CEECs, undergone during the transition, have already been at an advanced stage in 1995.
5
Drivers of Structural Change: The Integration Variables
Section 3 has shown that economic integration of the CEECs was accompanied by increased trade, capital and labor flows. Given this reasoning for the use of the variables FDI, trade and emigration as a proxy for economic integration the question emerges how these variables influence structural change.
5.1 Foreign Direct Investment
The role of external factors in productivity enhancing structural change has not gained much attention in the literature despite the fact that productivity enhancing effects of trade liberalization and economic integration are well documented. Those results go hand in hand with the comprehensive literature, which has thoroughly investigated the effect of FDI on the host country´s economic performance. FDI has an effect directly on labor productivity as well as indirectly by influencing TFP growth.
Labor productivity growth is defined as increased output per worker. By augmenting a neoclassical framework, in which output growth is dependent on investment, FDI can be regarded as an output increasing factor. However, the neo-classical theory offers only a limited view on the role of FDI in productivity growth. In those models FDI puts the economy on a higher growth path but does not lead to long-run growth effects due to diminishing returns to capital. Endogenous growth models offer a more differentiated approach, which has examined further channels through which FDI stimulates productivity. Despite increased investment FDI has an impact on the host country´s productivity through technology transfer and spillover effects (Nair-Reichert and Weinhold 2001). External capital provided by FDI is connected with the transfer of technology and management skills. The transfer of knowledge, through training (training of local workers by foreign firms) and organizational development has a positive influence on FDI receiving industries. Improvements in technology and management techniques can stimulate TFP growth and external effects, like knowledge spillovers, can spread productivity enhancing knowledge throughout the industry (Chowdhury and Mavrotas 2006). Generally the literature confirms the assumption that for all channels host country characteristics play a decisive role. Minimum thresholds e.g. the development level of the host country concerning human capital, which is part of the absorptive capacity, are a crucial prerequisite for the productivity enhancing effects of FDI.
of Central and Eastern Europe and the CIS [Commonwealth of Independent States 3] have undergone important structural changes, linked with the entry of FDI” (Kalotay 2010, p.72). He compares different groups of East European countries and finds that FDI had the deepest impact on structural change in the countries, which joined the EU. His explanation is mainly build on the different timing of intensified FDI inflow starting earlier in the CEECs that joined the EU and later in other transition countries of Eastern Europe. The study further highlights the effect of European economic integration for the nexus of FDI and structural change. East European Countries that have joined the EU attracted sectoral FDI inflows more similar to the industrial structure of the economic union.
In the light of the research question these results seem to confirm that a link between the inflow of FDI and structural change exists. Especially in the case of the CEECs this relationship is most probably of relevance for structural change as the transformation and integration process of the CEECs was accompanied by substantial FDI inflows. A confirmation of the positive relation of FDI and productivity enhancing structural change, also in this study, would give support to the proposition that European economic integration has fostered labor movements from low to high productive sectors. To sum up the theoretical reasoning: The main mechanism between FDI and structural change can be described as a productivity stimulating effect of FDI (through various channels), which in turn can encourage the process of structural change as labor demand in high productive industries increases and wages rise. This reconstructing of the economy supports further productivity growth through the reallocation of labor. The following hypothesis can be formulated: Hypothesis 1: Productivity enhancing structural change is positively related to FDI inflow in the CEECs.
5.2 Trade Integration
Trade integration can lead to productivity increasing inter-sectoral labor movements (Wacziarg and Wallack 2004). In general terms trade integration is accompanied by a decline in trade costs, which will reveal the comparative advantage of countries and lead to specialization effects. Emerging pattern of specialization in turn induce labor movements to the industries in which a country has a comparative advantage. In classical trade theory the movement of production factors to sectors with a comparative advantage is driven by relative differences in factor endowments (e.g. in the Heckscher–Ohlin model) or differences in technology (e.g. in the Ricardo model). These by specialization triggered labor movements offer a channel through which trade integration is connected with productivity enhancing structural change. Productivity differences between sectors
3 The Commonwealth of Independent States includes the countries of the former Soviet Union except of the Baltic
gives some industries a comparative advantage, as a consequence those sectors will increase their exports when trade barriers fall. Increased exports cause a higher demand for labor. Labor will thus move to the productive sectors of the economy. Productivity differences also translate into wage differences, which attract labor. Trade integration is also connected to the demand-driven factors of structural change. Falling trade costs alter the relative prices of goods and affect the sectoral spending of consumer. Foreign demand for home country exports induces increased labor demand in the export industry.
Also the new trade theory predicts effects of intensified trade on the sectoral composition of an economy (see e.g. Dixit and Norman 1980; Krugman 1991). E.g. economies of scale can lead to an agglomeration of industries, which increases their productivity and causes the shift of labor between sectors. Recent theoretical modeling emphasizes that even within industries productivity differences prevail in which labor shifts can lead to overall more productivity (see for example Melitz 2003; Bernard et al. 2003). The intra-sectoral movement of labor implies that it is not observable at a given level of sectoral disaggregation. The choice of sectoral differentiation therefore effects the ability to observe inter-sectoral labor shifts and firm level data would be necessary to analyze such effects. However, the predictions of the classical and new trade theory concerning inter-sectoral labor movements might well be captured by a sectoral dataset.
Hypothesis 2: Productivity enhancing structural change is positively related to intra-EU exports of the CEECs.
5.3 Labor Migration
emigration. Emigration of labor can be regarded as a change in the domestic labor supply. With the emigration of skilled labor the supply of worker, which are able to work in expanding high productive industries, is declining and thus dampening the potential for domestic labor shifts from low to high productive sectors. Immigration is speeding-up the structural change process, while emigration is slowing it down, conditional on the human capital of the migrants. Further theoretical argumentation for the importance of skill levels in the reallocation process of labor comes from Poirson (2001). The author develops a model in which labor reallocations from the agricultural sector to modern sectors influences labor productivity growth and controls for the initial level of education. The findings suggest, for a sample of African countries, that the lower productivity growth compared to developed countries can be explained by lower education levels, which hinder the reallocation of labor. Masson (2001) develops a model of migration from rural to urban areas and emphasizes the importance of skill acquisition for the reallocation process. In his model migration from rural areas to cities is dependent on the ability to acquire the skills necessary to be employed in the urban (modern) industries. Credit constrains and urban unemployment make it difficult for migrants to acquire the necessary skill level and thus hinder labor reallocation in his model. The reasoning on the importance of human capital for productivity enhancing structural change bases in many aspects on the discussion on brain drain, which has emphasized negative, but occasionally also positive effects, of high skilled labor migration from the CEECs (see e.g. Tung 2006). The above considerations lead to Hypothesis 3:
Hypothesis 3: Productivity enhancing structural change is negatively related to high skilled labor emigration from the CEECs.
6
Data
The definition of labor productivity in this paper is value added per person employed. Data on employment and sectoral gross value added (GVA) is available from the Socio Economic Accounts of the World Input-Output Database (WIOD) of the University of Groningen (Timmer et al. 2015). The Socio Economic Accounts of the WIOD database provide a comprehensive collection of data from various sources.4 WIOD contains data for the time 1995-2011 for all CEECs. Using the data on GVA and employment sectoral labor productivity has been calculated by dividing sectoral GVA by the corresponding employment. The result is a panel data set on labor productivity for 35 sectors based on the United Nations ISIC rev. 3 classification (see Table 1). Overall the dataset includes 10 countries over 15 years. As GVA is expressed at current prices it has been deflated by a price index for GVA (1995=100).5 GVA is expressed in local currency, thus value
4 The underlying sources on GVA and employment for the CEECs are derived from Eurostat and the OECD´s
Structural Analysis Database (STAN) (Timmer 2012).
5 The price index in WIOD is not available for 2010 and has been calculated using the 2009-2010 growth rate of the
added at 1995 prices has been converted in US dollar making use of 1995 exchanges rates provided together with the data on the website of the WIOD database. The main advantage of the WIOD database is the high sectoral differentiation covering a sufficiently long time period before and after the EU-accession. The industry-level differentiation of 35 sectors is very detailed compared to other frequently used databases on sectoral productivity. The 10-sector-database of the Groningen Growth and Development Center (GGDC) distinguishes by 10 sectors. The EU KLEMS data base contains data on up to 72 different sectors but for the CEECs only until 2005.
The sample countries’ labor productivity in 2010 ranges from $50.10 per employed person for Bulgaria to $297.60 per employed person for Slovenia, where labor productivity is almost six times as high (at 1995 prices). Table 13 in the appendix shows detailed descriptions of the labor productivity in the CEECs. The table also shows that there are considerable productivity gaps between the economy-wide labor productivity and the sector with the highest labor productivity. This is true for both the countries with relatively lower productivity and those with high labor productivity.
ISIC rev. 3 Industry Classification
Name/Description Code Name/Description Code
Agriculture, hunting, forestry and fishing AtB Construction F
Mining and quarrying C Sale, maintenance and repair of motor vehicles
and motorcycles; retail sale of fuel 50 Food , beverages and tobacco 15t16 Wholesale trade and commission trade, except
of motor vehicles and motorcycles 51
Textiles 17t18 Retail trade, except of motor vehicles and
motorcycles; repair of household goods 52
Leather and footwear 19 Hotels and restaurants H
Wood and products of wood and cork 20 Other inland transport 60
Pulp, paper, printing and publishing 21t22 Other water transport 61
Coke, refined petroleum and nuclear fuel 23 Other air transport 62
Chemicals and chemical products 24 Other supporting and auxiliary transport
activities; activities of travel agencies 63
Rubber and plastics 25 Post and telecommunications 64
Other non-metallic mineral products 26 Financial intermediation J
Basic metals and fabricated metal 27t28 Real estate activities 70
Machinery and equipment, n.e.c 29 Renting of machinery and equipment and other
business activities 71t74
Electrical and optical equipment 30t33 Public administration and defense; compulsory
social security L
Transport equipment 34t35 Education M
Manufacturing n.e.c; recycling 36t37 Health and social work N
7
Decomposition Analysis
Following the methodology applied by Timmer and Szirmai (2000), Peneder (2003), Havlik (2008) and Timmer and de Vries (2009) a shift-share analysis is used to decompose aggregate productivity growth into the contributions of labor shifts and of productivity growth within sectors. Aggregate labor productivity at time 𝑡 is defined as LPt=VAt/Lt where, VAt denotes gross value added
aggregated over all sectors and Lt denotes total employment. In a first step the productivity levels
derived from the WIOD data have been used to calculate aggregate productivity growth:
gLP= LP1-LP0 LP0
where the subscript t=1 indicates the final year and t=0 the base year. Aggregate labor productivity LPt=VAt/Lt can equivalently be written as:
LPt= ∑ni=1VAi,t Lt = VA1,t L1, t * L1,t Lt + VA2,t L2,t * L2,t Lt +…+ VAn,t Ln,t * Ln,t Lt = ∑ LPi,t* Si,t n i=1
where ∑ VAni=1 i,t is the sum of sectoral value added over n sectors and Si,t is the employment share of
industry i in total employment at time t. Calculating the difference of the labor productivity in t=1 and t=0 yields:
LP1-LP0= ∑ (LPi,1-LPi,0) Si,0 n
i=1 + ∑ (Si,1-Si,0)LPi,1
n i=1
Dividing both sides of the term by the initial labor productivity LP0 gives the decomposition of the
growth rate of aggregate labor productivity between the base year and the final year into two effects:
gLP= ∑ (LPi,1-LPi,0)
n
i=1 Si,0
LP0 +
∑ni=1(Si,1-Si,0)LPi,1
LP0
Within productivity growth (the first term of Equation 4) captures the sum of productivity growth that occurs within an industry between the base year and the final year weighted by the initial
employment share of that sector. For a given share in employment aggregate productivity increases when sectors become more productive. This term contributes positively to the aggregate productivity growth if labor productivity in sector 𝑖 grows between the base year and the final year, assumed that labor shares stay constant. The shift effect (the second term of Equation 4) is calculated as the difference between the employment share in industry 𝑖 in the base year and the final year weighted by the final year labor productivity. The intuition is that assuming a given level of labor productivity aggregate productivity increases when more labor is engaged in high productive sectors. A positive shift effect indicates that labor has shifted to higher productive sectors and had thus a positive contribution to aggregate labor productivity growth. However, it remains an open question why changes in labor productivity should not be weighted by the final year labor shares or changes in labor shares with initial labor productivity. Foster et al. (1998) point out that this method will lead to an over- or underestimation of the results, depending on the weights. Alternatively the weights could be calculated as arithmetic mean S̅ and LP̅̅̅̅ over the period between t=0 and t=1. This would make the decomposition invariant to a certain base year (Timmer and de Vries 2009).
A modified decomposition method has been suggested by Timmer and Szirmai (2000). The authors add an interaction term to the conventional decomposition method, which captures the effect of labor movements towards sectors with higher productivity growth. Mathematically the method can be described as:
gLP= ∑ (LPi,1-LPi,0)Si,0
n i=1
LP0 +
∑ni=1(Si,1-Si,0)LPi,0
LP0 +
∑ni=1(Si,1-Si,0)(LPi,1-LPi,0)
LP0
In Equation 5 aggregate productivity growth is decomposed into within industry growth (first term) and two shift terms: a static shift term (second term) and a dynamic shift term (third term). The static shift term indicates the movement of labor to sectors with a higher labor productivity level. The dynamic shift term, on the other hand, measures labor shifts to more “dynamic” sectors or put differently to sectors with higher labor productivity growth. In this method, described in detail in Peneder (2003), the interaction term or dynamic shift effect contributes positively to aggregate productivity growth when industries increase both labor productivity and their share of total employment. On the other hand, the dynamic shift effect contributes negatively to aggregate productivity growth if industries with increasing productivity cannot keep up their shares in employment (Havlik 2005). The next section, which will present the decomposition results, will also test them for robustness regarding the decomposition method in Equation 4 and 5.
An implicit assumption of the presented methods is that labor productivity is not dependent on changes in employment. This assumption only holds if all labor inputs have the same productivity i.e. marginal and average labor productivity in a sector equal. If marginal labor productivity is lower than average labor productivity a decline in labor rises average labor productivity. In such a case the decomposition method attributes this rise in labor productivity to the within term, despite the fact that it actually stems from labor movements between sectors (Timmer and de Vries 2009). This fact arises because labor productivity in the calculation of the within effect does not depend on changes in labor shares. Taken this considerations into account it can be assumed that the shift effect underestimates the effect of labor reallocation to productivity growth. Some more sophisticated decomposition methods exist, which address this problem by estimating the ratio of marginal to average productivity (see e.g. Timmer and de Vries 2009).
A further remark has to be made in the context of the presented decomposition method as pointed out by Timmer and Szirmai (2000). The shift-share analysis takes a supply-side perspective (allocation of production factors). As already indicated in section 4.1 changes in the demand structure, e.g. through changing income elasticities, influence the expansion and decline of industries. Such demand-driven factors of structural change are taken as exogenous and cannot be analyzed using the presented decomposition methods. The methods should be regarded as “accounting methods”, which indicate what the effects of changing employment shares are irrespective of the underlying causes.
8
Decomposition Results
COMPARISON OF DECOMPOSITION METHODS
Period Within Effect Static Shift Effect Dynamic Shift Effect Productivity growth
1995 – 1998 1.80% 0.54% - 2.32% 1.82% 0.56% -0.03% 2.32% 1998 – 2001 2.80% 0.02% - 2.83% 2.90% 0.12% -0.20% 2.83% 2001 – 2004 3.18% 1.13% - 4.21% 3.34% 1.29% -0.34% 4.21% 2004 – 2007 2.67% 0.79% - 3.41% 2.64% 0.76% -0.34% 3.41% 2007 – 2010 -0.54% 0.46% - -0.07% -0.46% 0.54% -0.16% -0.07%
Table 2: Comparison of conventional versus modified decomposition analysis depicted in Equation 4 and 5. The CEECs sectoral GVA has been aggregated converting GVA in 1995 US dollar. Five four year subperiods and yearly averages. Source: Own calculations based on the Socio Economic Accounts of the World Input-Output Database (Timmer et al. 2015)
Table 2 reveals that productivity growth has been large in the subperiods 2001-2004 and 2004-2007; the time around the EU-accession. Productivity grew on average by around 4.2% during the time between the years 2001-2004 and 3.4% from 2004-2007. Productivity actually stagnated between 2007-2010, what can be explained by the effect of the global financial and economic crisis starting in 2008. Looking at the within effect it can been seen that productivity growth has, to a large extend, been driven by the growth of productivity within sectors. The contribution of the within effect to aggregate productivity growth is bigger than the contribution of the dynamic and static shift effect in all periods, except 2007-2010. In the time of 2007-2010 within productivity growth was negative while the static shift term remained positive. A finding, which is worth to emphasize, is the positive sign of the static shift term and the negative sign of the dynamic shift term in all periods. This result indicates that worker moved to sectors with initially high labor productivity, but that those sectors, which saw fast growing productivity, could not maintain their labor shares.
structural change than countries on higher stages of development (see section 4.1 and 8.2). Dynamic shift effects were in all cases negative indicating that labor moved to industries, which saw low increases in productivity. It is notable that in all countries the finding of a dominating within growth effect is confirmed. The contribution of the dynamic and static shift effect taken together (combined shift effect) to productivity growth show a diverse pattern. High negative net-effects are found for Estonia, Lithuania and Latvia. However, those countries where also the countries with the highest within productivity growth. The results suggest that for the Baltic States productivity gains rooted in within sector growth rather than in the shift of labor. On the other hand positive contributions of the combined shift effect are found for the Czech Republic, Hungary, Poland and Slovenia. Positive static shift effects can be observed for all countries. This finding suggests that labor moved to initially high productive industries.
Table 3: Decomposition of aggregate labor productivity growth into parts due to labor productivity growth within sectors and productivity growth due to shifts of labor using Equation 5. Percentages may not add up to 100 due to rounding. Source: Own calculations based on the Socio Economic Accounts of the World Input-Output Database (Timmer et al. 2015)
DECOMPOSITION RESULTS 1995-2010
(1) (2) (3) (4) (5) (6) (7)
Percentage of labor productivity growth explained by
Country Annual average productivity growth rate
Combined Shift
Effect Within Effect Static Shift Effect Shift Effect Dynamic Total effect
For the two countries with the highest and lowest productivity growth between 1995- 2010 Table 4 presents the decomposition results of five subperiods. The figures show a mixed picture as the static shift effect had in some cases a negative and in some cases a positive effect on productivity growth. In the time after the EU-accession (2004-2007 for Estonia and Lithuania, 2007-2010 for Bulgaria and Romania) the static shift effect was positive. Again the dynamic shift effect is negative in almost all countries and all time periods. This is an important intermediate finding, which indicates that sectors with high labor productivity growth have seen decreasing labor shares. Or put differently: labor moved away from sectors with productivity increases.
SELECTED NATIONAL DECOMPOSITION
Country Time Period Within effect Static Shift Effect Dynamic Shift Effect
Estonia 1995 - 1998 6.13% 1.53% -0.56% 1998 - 2001 7.11% -0.26% -1.32% 2001 - 2004 5.05% 0.39% -0.42% 2004 - 2007 4.81% 0.74% -0.81% 2007 - 2010 -1.48% 0.54% -1.06% Lithuania 1995 - 1998 5.65% -0.08% -0.94% 1998 - 2001 5.04% 0.07% -0.47% 2001 - 2004 3.87% 1.35% -0.73% 2004 - 2007 3.70% 1.38% -0.47% 2007 - 2010 1.06% 0.14% -0.70% Romania 1995 - 1998 -0.70% 0.65% -0.36% 1998 - 2001 1.17% -1.51% -1.96% 2001 - 2004 4.99% 10.03% -9.10% 2004 - 2007 3.96% 1.83% -1.01% 2007 - 2010 -0.76% 1.60% -0.30% Bulgaria 1995 - 1998 0.33% -0.58% 0.09% 1998 - 2001 3.17% 2.40% -0.94% 2001 - 2004 1.75% -0.12% -0.39% 2004 - 2007 0.85% 0.41% -0.09% 2007 - 2010 2.18% 1.17% -0.32%
Table 4: Decomposition of the productivity growth rates for five 4-year periods: The first two rows show the two countries with the highest productivity growth over the period 1995–2010, the last two rows the two countries with the lowest productivity growth; yearly averages. Source: Own calculations based on the Socio Economic Accounts of the World-Input Output Database (Timmer et al. 2015)
and sectors with limited labor productivity growth (often services). Agriculture has distinct characteristics compared to other parts of the economy concerning the production function and the technical rate of change (Timmer and de Vries 2009). Including agriculture thus might interfere with examining the “structural change burden hypothesis”.
A negative and significant correlation can be observed. 6 Figure 4 underlines the findings on the dynamic shift term: sectors, which could increase labor productivity saw decreasing employment shares (mainly industry and manufacturing).7 For example, the electronic (29), textiles (17t18), food (C) and chemical industry. On the other hand there are some sectors where employment increased, but labor productivity remained almost constant (mainly service sectors). For example, real estate activities (70), renting of machinery and equipment and other business activities (71t74), wholesale trade and commission trade (51), construction (F) and public administration (L). This finding suggests the existence of a “structural change burden” in the context of the CEECs.
6 Including agriculture no significant relationship between productivity changes and employment changes has been
detected. This can be explained by the fact that agriculture was a sector with low productivity in 1995 and 2010 and saw by far the biggest loss in employment (see Figure 14 in the appendix and the next footnote).
7 A negative dynamic shift term indicates from a theoretical standpoint the “structural burden hypothesis”, as labor
moves to sectors with limited productivity increases. On the other hand a positive static shift effect illustrates a “structural bonus” hypothesis, which states that labor moves to high productive sectors (Havlik 2005). Figure 14 in the appendix illustrates the “structural bonus hypothesis”.
Figure 4: Correlation between change in labor productivity and change in employment shares, Source: Own calculations based on the Socio Economic Accounts of the World Input-Output Database (Timmer et al. 2015)
Change in employment share 1995-2010
As there are no studies for the time after 2004 the question arises if pattern of structural change and productivity growth have changed in the time after 2004. Table 5 shows the annual contribution of the combined shift effect to aggregate labor productivity before and after 2004 for the countries, which joined the EU in 2004. No clear pattern emerges by looking at the combined shift effect in absolute and relative terms. In five countries the absolute effect was smaller after 2004 than before. The reverse is true for the combined shift effect relative to total labor productivity growth. Five out of eight countries had a relative contribution of the shift effect that was bigger in the time after 2004. For Latvia and Lithuania a positive effect before 2004 turned into a negative. In the Czech Republic it turned slightly positive after 2004.
COMPARISON OF COMBINED SHIFT EFFECT BEFORE AND AFTER 2004
1998 – 2004 2004 – 2010
Country Absolute productivity growth Relative to total Absolute productivity growth Relative to total
Czech Republic -0.10% -3.19% 0.02% 0.76% Estonia -1.25% -16.13% -0.78% -50.50% Latvia 0.68% 11.66% -0.37% -14.64% Lithuania 0.22% 3.48% -0.38% -13.33% Hungary 0.79% 16.87% 0.26% 82.37% Poland 0.97% 17.55% 0.39% 23.77% Slovenia 0.69% 20.25% 0.27% 23.24% Slovakia -0.69% -18.79% 0.00% 0.00%
Table 5: Decomposition results for the CEECs before and after 2004, yearly averages. Source: Own calculations based on the Socio Economic Accounts of the World Input-Output Database (Timmer et al. 2015)
It has been found that sectors with initially high productivity could attract labor (positive static shift effect). Over the whole period 1995-2010 all countries had a positive static shift effect . However, sectors with high labor productivity growth could not maintain their labor shares and labor actually moved from “high productivity growth” industries to industries with stagnating productivity (negative dynamic shift effect), which is an indication of a structural change burden in the CEECs. This finding is again confirmed by the literature. Studies of Havlik (2005) (for the CEECs) and Peneder (2003) (for the whole EU) find evidence on a negative dynamic shift effect. Labor shifted mainly from agriculture and manufacturing towards business services, what is a process observed for many countries. Is there a link to the EU-accession? For the CEECs it can be assumed that the rising importance of services has to some extend been fueled by the outsourcing of business services from EU-15 firms to the CEECs (Stare and Rubalcaba 2009). Further evidence on the role of the EU-accession and the business sector in the CEECs comes from Krenz and Gehringer (2004), which analyze the determinants of firm localization in Poland. The authors find that the agglomeration of services industries can explain the choice of firms to operate in a region, using panel data on local and international firms. For a set of new firms EU-accession has positively influenced the localization decision and increased the attractiveness of Polish regions. The authors underline in their explanation the role of the single market, which has for new firms shaped the decision to locate in a certain region.
9
Regression Analysis
9.1 Independent Variables
Foreign Direct Investment: The independent variable FDI is defined as change in FDI inflow as share of GDP over a four year period. Data on FDI inflow is taken from the United Nations Conference on Trade and Development (UNCTAD 2015). UNCTAD provides a database of FDI inflows for the CEECs for the time 1995 to 2010. In order to obtain real FDI inflows the UNCTAD Statistics Consumer Price Index has been used to deflate the FDI flows and to obtain FDI data in constant prices. It would have been preferable to distinguish between FDI sourcing from the EU-15 and the rest of the world to identify the contribution of FDI inflow directly related to European economic integration. As bilateral FDI data is only available for the time after 2001 such a variable could not be used. However, there are also arguments to believe in a connection between FDI inflow from non-EU countries and European economic integration. An increased single market is more attractive for investors as they can now sell their products without tariffs in the whole EU. European economic integration therefore provides incentives for FDI inflow from outside of the EU. Galgau and Sekkat (2004) note that increased economic integration e.g. by abolishing trade barriers is likely to increase FDI inflow because multinational corporations (MNEs) can act more efficiently across borders. Some empirical studies confirm this argument and find that the single market has increased the attractiveness of the EU for foreign investors (Barrell and Pain 1999). The coefficient of 𝐹𝐷𝐼 is expected to be positive as the theory predicts that increased foreign investment is associated with a positive contribution of structural change to productivity growth (see the reasoning for Hypothesis 1).
Intra-EU trade: How can the integration of CEECs economies into the common market for goods and service be quantified? Many studies using growth regression simply use dummy variables for being part of the single market or not (Badinger 2005). Another frequently used measure for market integration is intra-EU trade as percentage of total trade or exports as percentage of total exports. Alternatively market integration can be measured by the country’s exports as share of GDP. This analysis uses exports to the EU as share of GDP as a proxy for EU market integration, as this is common measurement in the literature (Arribas et al. 2007). Data on intra-EU exports in US dollar is taken from the IMF (2015) and converted in constant prices. The variable on intra-EU exports is labeled 𝑇𝑅𝐴 and expected to have a positive coefficient (see Hypothesis 2).
origin and educational level for the years 1980-2010 (5 year intervals). For all EU-15 countries figures on migration stocks sourcing from the CEECs are available. Restricting the migrant age to above 25 years comes with an advantage as it excludes e.g. student mobility within the EU. Emigration stocks in the IAB data are based on foreign born population, a further advantage, as this definition better captures the concept of labor migration compared to the alternative principle of citizenship. The county of birth is a time invariant characteristic, which is not the case for citizenship. Emigrants, which acquire a foreign citizenship might therefore not be counted anymore as labor migrants, but their emigration still had an influence on the labor market conditions of their home countries (Docquier et al. 2007). Figures on the CEECs labor force are taken from the Annual Macro Economic Database (AMECO) of the European Commission (DG ECFIN 2015). Comparing migrant stocks from the CEECs to the high skilled labor force in the source country would have been a preferable measure as it provides an indicator for the pressure caused by high skilled emigration on the labor market for high skilled worker. It would have been an alternative to set the number of migrants relative to the size of the high skilled population of the CEECs as this is also a measure for the available potential of high skilled workers. As this data is not available for many CEECs for the time before the accession, the total labor force seems to be a feasible proxy for the size of high skilled labor reserves.
A big disadvantage of the IAB data is that it not provides annual data. Data is only given in five year intervals. To the knowledge of the author no database exists, which provides consistent migration data for the whole time period, in particular, which provides information on the skill level of migrants.8 The scarcity of migration data makes it necessary to interpolate the data gaps between the five year intervals. This is a procedure frequently used in studies on labor emigration from the CEECs (see e.g. Layard 1994, Baas and Brücker 2009, 2011). Using the growth rate of the stocks of emigrants from the CEECs in the EU-15 the missing values for the stock of high skilled emigrants between the years 𝑡 − 1 and 𝑡 can be interpolated. The missing values have been derived using the following formula, described in depth in De Vries et al. (2013):
EMtH=EMt-1H*EXP [ln (EMt e EMt-1e )- (ln ( EMb2e EMb1e )/(b2-b1)) + (ln ( EMb2H EMb1H )/(b2-b1))]
8 Eurostat provides data on migration stocks but not for the whole time period of interest.
The high skilled migration stock is denoted by EMtH and the overall migration stock
by EMe. EMb1H and EMb2H are the benchmark migration stocks of high skilled labor given for
the years 1995, 2000, 2005 and 2010 by the IAB data. b2-b1 denotes the time between the
benchmark years. This interpolation method bases on the assumption that the high skilled emigration stock is growing with an average logarithmic growth rate. This growth rate is calculated by adjusting the annual average growth rate of the IAB data (term 3 in the brackets) by deviations from a growth trend derived from external data (term 1 and 2 in the brackets). The external growth trend is calculated by the difference of the external growth rate and the average external growth rate between two benchmark years, which is a procedure that counteracts annual fluctuations. External data is taken from Kahanec (2012). Kahanec provides data on emigrants from the CEECs in the EU-15 for the years 1997 to 2009, based on own calculations, data from Eurostat and Holland et al. (2011). For the years which did not lie between two benchmark years, data was derived using extrapolation and back-casting methods, assuming the growth rate of the external data. Mathematically the extrapolation method can be denoted as:
EMtH=EMt-1H ( EMe t
EMet-1
)
The missing emigration data, which has been calculated using the back-casting method, was derived by the following equation:
EMtH=EMt+1H (EMe t+1
EMet )
Interpolation methods always imply some sort of uncertainty about the correct value of the data points and usually lead to interpolation errors. But for the problem at hand the advantages though the gain in data range outweighs the loss in precision and given the lack of data the applied approach seems to be promising.9 It might be assumed that the resulting bias underestimates high skilled emigration. Section 3 has given the intuition that migration
9 As Kahanec (2012) notes the lack of comprehensive data for European labor mobility mainly arises because of an
inadequate infrastructure to collect data and a non-harmonized methodology of national statistical authorities.
(7)
from the CEECs was to a large extend high skilled, thus interpolating high skilled emigration by overall emigration most probably underestimates the true values of high skilled emigration. The 4-year change of the variable EMtH is introduced as an explanatory variable and it´s
coefficient is expected to have a negative sign (see Hypothesis 3). 9.2 Control Variables
Initial structural change gap: Following McMillan et al. (2014) the initial structural change gap is calculated as the share of employment in the least productive sector on total employment at the beginning of the period. The initial structural change gap is used to account for potential catching-up effects. Countries with a high amount of labor in low-productive industries have more room to reallocate labor to high low-productive sectors than countries, where the scope for further reallocation of labor is limited. The initial structural change gap thus proxies for the excess of labor, which can potentially be reallocated. The underlying idea follows the classical dual-economy models (see section 4), where the marginal productivity of labor in the “traditional” sector is lower than in the “modern” sector and reallocation thus leads to productivity gains. There are further possibilities to control for catching-up effects. Labor productivity gaps between sectors of an economy can directly be taken into account. Excess labor is attracted by higher wages in the “modern” sector, these wage differentials, in turn are determined by productivity differences. Thus countries with higher productivity gaps potentially observe a higher labor reallocation to high productive sectors. Another way to indicate the potential of a country to observe productivity enhancing structural change is the calculation the ratio of marginal to average labor productivity with a low ratio indicating a high potential for productivity gains. The idea behind this is that the reduction of workers will increase labor productivity as long as marginal productivity is below average productivity (Timmer and de Vries 2009).
Interaction with the real effective exchange rate index: There has been extensive research on the effect of exchange rate fluctuations on FDI (see e.g. Kiyota and Urata 2004, Russ 2007). Two basic channels through which the exchange rate and FDI are connected can be identified: First, the depreciation of the FDI host countries exchange rate reduces production cost and stimulates the inflow of FDI. Second, the depreciation of the host countries currency changes the relative wealth position, favoring the source country compared to the host country. The increased wealth position can lead to raising acquisitions of host country assets and investment due to relatively lower costs of capital (Blonigen 1997). Given the fact that FDI is expected to promote structural change conditional on the exchange rate an interaction term is introduced. The term 𝐹𝐷𝐼𝑖𝑡 ∗ 𝑈𝑉𝑖𝑡 captures the effect of undervaluation
This procedure is common in studies on the role of FDI on growth. Those studies frequently use interaction terms in order to account for prerequisites on which the effect of FDI depends like e.g. human capital or financial markets development (see e.g. Alfaro et al. 2004; Zhang 2002; Levine 2005). Undervaluation is measured by a real effective exchange rate (REER) index and the variable 𝑈𝑉 is defined as the 4-year change in the REER of a CEEC against a trade-weighted index of the other EU member currencies. The REER not only measures exchange rate fluctuations, it is also an indicator for price competitiveness relative to a countries major trading partners. It equals the nominal effective exchange rate deflated by the consumer prices. A depreciation (increase in competitiveness and falling REER index) is expected to amplify the positive effect of FDI on the contribution of structural change to productivity growth, while an appreciation (loss of competitiveness and rising REER index) is expected to hamper the effect. Data is taken from Eurostat (2015c).
Labor market flexibility: Institutional obstacles like tight labor market regulation can hamper the smooth movement of labor between sectors (Bernal-Verdugo et al. 2012). On the other hand, flexible labor markets allow for the shift of labor to high productive sectors with less frictions and less costs (Sala-i-Martin et al. 2014). The analysis takes this into account by making use of a labor rigidity index. The index is derived from the Fraser Institute’s Economic Freedom of the World (EFW) database (Fraser Institute 2015). The index ranges from 0 to 10 (higher values indicate more flexible labor markets) and is composed of indicators on labor market regulation (de jure) and labor market indicators (de facto) from six areas: 1) minimum wage legislation, 2) hiring and firing legislation, 3) collective wage bargaining, 4) costs of hiring, 5) costs of work dismissal and 6) duration of conscription (with a higher labor market flexibility for countries without or low conscription). Data is based on the World Bank Doing Business Report and the World Economic Forum’s Global Competitiveness Report.
The problem with the Fraser Institute’s labor rigidity index is that yearly data for the CEECs is only available from 2000 onwards. The literature also discusses alternative indices intended to measure the flexibility of labor markets.10 A very comprehensive dataset is compiled by Campos and Nugent (2012). The authors provide an employment protection law index (LAMRIG) for more than 140 countries in 5-year averages since 1960. LAMRIG is a de jure measurement of employment rigidity taking national labor market legislation as an indicator for labor market flexibility. However, the LAMRIG database has the same
10 For a comprehensive discussion on the use of labor market indices and their advantages and downsides see
shortcoming as it does not provide annual data. Another index intended to measure labor market flexibility is a subindex of the Global Competitiveness Index developed by the World Economic Forum. This index also includes de facto and de jure measures but the methodology has changed over the years making comparisons over time difficult (Aleksynska and Cazes 2014). Again, the time coverage is limited to the time after 2004. The third frequently used index for labor market flexibility is compiled by the IMD World Competitiveness Center. This index includes a time series for the time 1995-2013 with a set of indicators indented to measure labor market flexibility. Unfortunately this database is not publically available. Given the limited data range of all mentioned indices the choice has been made to use the Fraser Institute´s database, as it offers yearly data at least for the time after 2000.
9.3 Model Specification
The decomposition analysis in section 8 has decomposed aggregate productivity growth in a within, static and dynamic effect. Combining the static and the dynamic shift effect gives an indicator, which allows to assess the overall effect of labor movements on productivity growth. The static and dynamic structural change term are combined (∆𝐶𝑖𝑡) and used as a dependent
variable in the following panel data regression:
∆Cit=β0+β1∆FDIit+β2∆EMitH+β3∆TRAit+β4∆UVit+β5∆FDIit*∆UVit+β7Xit+eit
The dependent variable ∆𝐶 stands for the structural change term (combined shift effect) as calculated in the previous section for 4-year time periods. The independent variables are expressed as change over the 4-year period from t to t+1. ∆𝐹𝐷𝐼 and ∆𝐸𝑀𝐻denote the
independent variables FDI inflow and high skilled labor emigration. ∆𝑇𝑅𝐴 denotes the trade indicator, operationalized as share of intra-EU exports to GDP. 𝑿 is a vector that contains the above introduced control variables. ∆𝐹𝐷𝐼 ∗ ∆𝑈𝑉 represents the interaction term of ∆𝐹𝐷𝐼 with the 4-year change in the real effective exchange rate index. The subscripts 𝑖 and 𝑡 denote the country and the base year, respectively. 𝑒 is the error term, which consists of a component 𝑢𝑖 and 𝜖𝑖𝑡 where 𝑒𝑖𝑡 = 𝑢𝑖+ 𝜖𝑖𝑡. 𝑢𝑖 includes unobservable individual effects, which
are time invariant and not captured by the independent variables. 𝜖𝑖𝑡 is an error term varying
over time and the countries. A detailed explanation on the econometric issues arising is given in section 8.5. Table 6 gives an overview of the variables.
