From Reunification to Regional Integration: Productivity and the Labor Market in East Germany. Michael C. Burda Humboldt-Universität zu Berlin

Transcription

1 From Reunification to Regional Integration: Productivity and the Labor Market in East Germany Michael C. Burda Humboldt-Universität zu Berlin Jennifer Hunt University of Montreal July 2001 This modification version August 2001 Panel Version, 24 August 2001 Brookings Panel, September This research was supported by the German Science Foundation (DFG Project BU 921/1-1 "Transfers im Gefolge der deutschen Wiedervereinigung"), and by the Sonderforschungsbereich 373 of the DFG (Quantifikation und Simulation wirtschaftlicher Prozesse). We thank Bianca Brandenburg, Anja Heinze, Tom Krüger and especially Almut Balleer, Anja Schneider and Nils Schulze-Halberg for research assistance. Silke Anger performed some of the computations. The Institut für Wirtschaftsforschung Halle (Martin Rosenfeld, Rupert Kawka, Harald Kroll, Udo Ludwig) and DIW (Bernd Seidel, Dieter Vesper), the Statistisches Bundesamt, Bud Collier, Steffen Maretzke (Bundesamt für Bauwesen und Raumordnung), Erika Schulz, and Helmut Seitz provided valuable data. We thank Patricia Anderson, Antonio Ciccione and Harald Uhlig for conversations and discussions, and the editors for very useful comments and suggestions.

2 2 Introduction It is difficult to find a more dramatic episode of economic dislocation in 20 th century peacetime than that associated with the reunification of Germany. It is a sad irony of history that the plucky East Germans who toppled the dictatorship of the proletariat in the bloodless revolution of 1989 were rewarded in the aftermath with an economic bloodletting of such historic dimensions. From 1989 to 1992, overall GDP in East Germany declined by roughly 30%, value added in industry declined by more than 60%, employment fell by 35%. During the same period, unemployment rose from officially zero to 15% based on registered unemployment, or 33% if hidden unemployment (early retirement, short-time, make-work, training schemes for the unemployed, etc.) is included. Not since the Great Depression has such a precipitous collapse of economic activity been observed. Ten years after East Germany came in from the cold, the success of the transition cannot easily be summarized. In the year 2000, GDP per capita in the East including Berlin was 65.3% of the western level (excluding Berlin 60.6%), an impressive accomplishment by the yardstick of economists' more pessimistic forecasts. Thanks to generous transfers from the West, consumption per capita has converged even closer. Beblo, Collier and Knaus (2001) report that 81% of easterners' incomes rose in transition. However, convergence in productivity has slowed sharply, implying the need for continuing transfers from the West, and the labor market has yet to recover from the initial shock. The unemployment rate based on registered unemployed was 19.5% in 1998, or more than twice the rate in the West. Measures based on survey data, taking search and availability into account, show that the unemployment rate averaged 13% over the period. 1 In the period , labor force participation in East Germany for males fell from 86% to 79.8%, and from 77.2% to 72.2% for females (compared with overall constancy in the West). At the same time, the fraction of the eastern population 18 years and older in work declined from over 60% in 1990 to 52.5% in 1999 (compared with a steady 57-58% in the West over the same period). While the East German transition has attracted continuous attention from economists, the issues examined have changed since the early 1990s. The one to one exchange rate, the privatization and

3 3 restructuring of state enterprises, as well as the striking initial jump in real wages are no longer in the realm of policy, although they may have left their mark on the economy. Attention has shifted to sustainability growth in GDP per capita which is by definition the sum of growth in labor productivity and labor force participation. For this reason, the analysis of this paper will focus on two issues: the dramatic slowdown in productivity growth in East Germany, and the apparent dysfunctional nature of its labor market why unemployment (the underutilization of labor) is so high. We emphasize the spatial mobility of production factors and its implications for factor prices, as well as assessing hypotheses based on other economic forces. 2 We take the position that the ultimate measure of the economic success of German unification is no longer the introduction of a market economy, but it is the attainment of an efficient production pattern by made possible by the union of the two regions. German unification is a metaphor for the economic integration of any two neighboring regions of different levels of economic development. The mixed success of the transition shows the difficulty of development even under the most auspicious circumstances. East Germany was immediately able to import sound institutions, including the political, legal, monetary, banking and industrial relations systems. At a minimum, these have enabled it to avoid the anarchic equilibrium Russia finds itself in. Furthermore, East Germany has benefited from the largesse, labor market and expertise of a rich neighbor sharing a culture and language. Its experience serves as a crucible for understanding the ramifications of other larger-scale, less micro-managed regional integration projects. The milestone of German monetary, economic and social union now stands not only as a benchmark transformation (and as a foil, perhaps?) for the economic integration of Eastern Europe, but also the immediate consequences of European Monetary Union. Our original analysis has three core components. We construct measures of capital stocks in the new German states and proceed to estimate total factor productivity in both new and old states. We then study potential explanations of the slowdown in growth which began in We find that while East Germany has experienced an unprecedented increase in its capital stock per capita, this has been insufficient to force convergence. The problem lies in a slowdown in TFP growth since Next, we move to the labor market, using microeconometric evidence to assess the source of poor employment

4 4 and unemployment performance in East Germany. Given our findings, migration the mobility of labor in the new German states assumes a central role in the economic integration process. It supplies an additional degree of freedom for labor market adjustment in the East, and may be needed if capital deepening slows and total factor productivity growth remain low. We sketch the interaction of capital formation and migration in the regional integration context in an empirical study of migration patterns in unified Germany. 3 German Unification, a Decade Later Brief History and Facts 4 The Berlin wall was irrevocably breached on November In March 1990, the first free elections since 1932 brought to power a conservative, market friendly government allied with then-chancellor Helmut Kohl. The election of a government favoring rapid unification reduced uncertainty about the future, and the economic, social and monetary union of July 1, 1990 ushered in the most economically important changes. The decision to choose a one to one exchange rate was a source of controversy. Political unification occurred on October 3, 1990, bringing the states of Berlin, Brandenburg, Mecklenburg-Vorpommern, Saxony, Saxony-Anhalt, and Thuringia into the federation. 5 Already by the spring of 1990 it had become evident that the East German economy was in shambles, belying even the most pessimistic estimates of the CIA and other US intelligence agencies. Economic and monetary union meant instant trade integration: not only could curious East Germans visit the West, but they could buy goods from enterprising western German traders. Domestic demand and foreign demand from former communist countries slumped. Industrial production fell by two-thirds within 18 months of the wall s opening up (Siebert 1992), while unprofitable production was propped up by a combination of short-time working and subsidization of loss-making enterprises. Unemployment rose quickly. Those who kept their jobs benefited from a rapid wage increase which was bargained by West German labor unions. Others gained large wage increases by moving to the west: more than one million people (6 percent of the population) did so in the period

5 5 The collapse of the labor market was cushioned by the introduction of the western social welfare system, and active labor market programs. 6 Firms meanwhile were taken over by the Treuhandanstalt, a trust which managed, held and disposed of state property. The privatization carried out was a fast one by transition standards (Roland 2000 argues that it was too fast). The Treuhandanstalt was wound down officially in 1995 and was replaced by a much slimmer and more unpronounceable version (Bundesvermögensamt für einigungsbedingte Sonderaufgaben, or BvS for short) with the primary task of controlling and enforcing thousands of contracts which had privatized the assets of central planning as well as privatizing the last dregs of East German industry and real estate. The evolution of the usual "headline indicators for East Germany is presented in Table 1. From eastern GDP growth was impressive, despite the loss of population, but it slowed, then fell below the western level in The rise of the eastern unemployment rate based on registered unemployed to almost 20% has already been mentioned. The western unemployment rate based on registered unemployed peaked at 11% in 1997, but this rate is known to overstate unemployment, as is shown by the column of Bureau of Labor Statistics estimates of western unemployment according to US concepts (this series has been discontinued). 7 The sharp rise in non-employment in the East, has been accompanied by sharply declining participation rates. <Table 1 here > Table 2 indicates how central European neighbors of East Germany have fared. East Germany is distinguished not by its output fall, but by its employment fall, which suggests the East German output fall is not primarily due to the overvaluation of the eastern currency at monetary union. Roland (2000) identifies price liberalization as the precipitating factor for output collapse in transition economies. While fascinating for monetary economists, the currency conversion merely represented a blip in the growth trend of M3, and subsequent corrective action by the Bundesbank proved that German unification would go down in history as a non-monetary event. 8 <Table 2 here >

6 6 How much convergence has actually occurred? The meteoric recovery of West Germany from the ashes of World War II inspired many commentators to expect the same from the new eastern states. An important difference is that the German capital stock had survived the war largely intact. Nevertheless, many saw the initial growth spurt evident in Table 1 as evidence of rapid convergence within a decade. 9 At the gloomier end of the spectrum, Barro s (1991, 1998) and Barro and Sala-i-Martin s (1992, 1993) invoke the behavior of the Solow growth model around the steady state as well as empirical observation of the US, Europe and Japan to argue that GDP growth closes only 2% per year of any gap in GDP per capita over long periods, conditioning on the usual variables. This implied that convergence would require two or more generations. CONVERGENCE IN CONSUMPTION. One of the most important benchmarks of success must be living standards, proxied in the data by aggregate consumption expenditures in the national income accounts. Unfortunately, the Federal Statistical Office stopped reporting disaggregated expenditure by region after 1995 in a step which appears politically motivated; available data is restricted to indirect estimates generated by research institutes and state statistical offices. The first column of Table 3 reports some estimates made by Blum and Scharfe (2000) and documents the great strides have been made, especially when the initial conditions are considered. On top of that, infrastructure in East Germany has been modernized to the tune of at least 438 billion DM in the period almost a third of the total accumulation physical capital - raising it to Western level in many consumption related categories. 10 Table 4 documents the same point at the microeconomic level, also showing the striking similarities in household behavior in the two countries after an initial adjustment period; this finding has been recently confirmed with detailed expenditure data by Grunert (2000). <Table 3 here> <Table 4 here> CONVERGENCE IN WAGES. A striking and highly visible consequence of the integration of East and West German labor markets was an unprecedented rise in nominal and real wages in the East. The second

7 7 column of Table 3 gives details for the period for weekly gross wages. At the time of monetary union, earnings in the East were about a third of West German levels, given the 1:1 exchange parity of Ostmark for DM after already rising by several hundred percent in Ostmark terms up to June 1990; by 1996 wages had reached three-quarters of West levels. 11 Since then they have not increased in relative terms, and have actually fallen a bit. 12 This increase was not observed in any of the other transition economies, even in the Czech Republic, which had initial conditions quite similar to the German Democratic Republic (Burda 1991). CONVERGENCE IN PRODUCTIVITY. The last two columns of Table 3 document the development of productive economic activity, i.e. the extent to which the region "East Germany can generate value added at world market prices. The first indicator, labor productivity, has grown rapidly throughout the period from less than 45% in 1991 to 73% in In many sectors and establishments, East German productivity exceeds that in the West, a result of the newest investment and infrastructure available, as will be discussed below. On the other hand, a large part of the increase in productivity simply reflects firms discharging less productive labor and moving up along the marginal product of labor schedule (see Hunt 2002 for a description of this phenomenon for low wage female workers in ). The wide disparity in productivity performance in the first years after reunification (Burda and Funke 1995) suggest that firms were far from their efficient production frontiers and some efficiency gain was possible without layoffs. Still, employment reductions at firms as high as 80-90% show that a labor intensive production was not in the cards. These unemployed most entitled to receive a great deal of social benefits are the source of the transfer problem. As long as there are many unemployed in Eastern Germany, the transfer problem and the GDP per capita problem will continue to haunt the region. The last column of Table 4 shows that GDP per capita is behind the levels implied by wage, labor productivity and consumption levels. The cost of German unification By any measure, German unification has been an expensive proposition: over the period , total financial transfers to the region from the West in the sense of a current account deficit of the East

8 8 - have exceeded 1.5 trillion DM. The persistent failure of the East to produce enough to carry its weight means that transfers from the richer West have been necessary from the start. This was foreseen at the outset; the surprise has been the stubbornness of the transfers, raising the spectre of a "problem region such as the Appalachians in the US, parts of the eastern seaboard in Canada, or the Mezzogiorno in Italy. The current annual net transfer burden remains about 75 b per annum or about 5% of total German GDP. The lion s share (ca %) represents social entitlements and cannot be cut without fundamentally changing the nature of the German social contract; they result directly from the inherent generosity of the German welfare state and were triggered automatically by conditions following unification (Burda and Busch 2001). The risk of further transfers and dependency still exists. 13 It should be emphasized, however, that these transfers, both public and private, have not merely gone to support East German consumption. East Germany was a "thread bare economy in 1989, using obsolete technologies and machines which were past their productive prime. The evidence indicates that the GDR had failed after the mid-1970s to invest enough in the right sectors (see Busch 1983, Ludwig 2000) and the collapse of GDR economy was accelerated by a poor investment policies. Early backof-the-envelope estimates estimated a need of roughly $75 billion per annum to rebuild the capital stock (Burda 1990, Collins and Rodrick 1991, Siebert 1992). These estimates have been validated. Table 5 documents the striking extent to which physical capital has been accumulated in the East since Simply summing the columns reveals a remarkable fact: since 1990, East Germans have installed more than one half a trillion DM in equipment, and more than twice that amount in structures. Taken together, this is roughly 100,000 DM for every inhabitant of the country. It is thus incorrect to claim that the massive current account deficits of the region since 1990 have been used solely to finance consumption; in fact, the sum of the transfers current account deficits roughly equal the cumulative investment undertaken. Just as singular as the collapse of production in , the intense capital accumulation finds few parallels in modern economic history, if only because of the lack of financial backers. <Table 5 here >

9 9 This remarkable achievement is not without blemishes. One salient feature of the table is the lopsided pattern of investment in structures in the East when compared with the West. This is mostly residential housing, but also infrastructure and of course buildings in which productive activity takes place. The overly intensive subsidization of construction has been sharply criticized by Sinn (1995, 2000) and Begg and Portes (2001) among others. Convergence and Integration: A Framework for Analysis In what follows, we adopt a two-pronged approach to understanding the stalled per capita productivity in the East. Since overall productivity represents the product of labor productivity and the employment ratio, one avenue is to investigate why labor productivity remains lower in the east, given the current functioning of the labor market. Another, equally important avenue is to ask why the East German labor market fails to put people to work in the same way as its western counterpart can. We thus seek inefficiencies in the labor market, i.e. the causes of high unemployment. At the same time, understanding stalled per capita productivity in East Germany must be put into context of economic integration of the two regions. Economic integration can be defined as the achievement by two or more geographic regions of the efficient production pattern made possible by their union, using world market prices for output and inputs as the appropriate metric. 14 In the case of regions in Germany or Europe, large gaps in GDP levels per capita (adjusting for purchasing power) or in factor prices are usually taken to be evidence of incomplete integration. Integration may be achieved by five mechanisms: 1) internal accumulation of production factors such as physical or human capital; 2) labor mobility from capital-poor to capital-rich regions (migration); 3) capital mobility from capital-rich to capital-poor regions (foreign direct investment); 4) Heckscher-Ohlin trade between the two regions, which in the absence of complete specialization implies the equalization of factor prices; 15 and 5) adoption of leading technologies by the backward region.

10 10 What we have in mind is a model of integration which stresses the first three mechanism and is described in more detail elsewhere (Burda 2001). Figure 1 depicts the outset of the integration process between a capital-rich West (W) and a capital-poor East (E), the union of which comprises a small open economy. For the moment, assume that both regions have access to the same constant returns production function F(K,L) for producing output with capital (K) and labor (L). The economy can borrow and lend in unlimited amounts for projects with positive net present value at the given world interest rate r; for simplicity labor mobility from the rest of the world will be set to zero. As a result, the steady state of the economy in Figure 1 will lie along the capital-labor ratio equating capital-labor ratios in both regions to that given by the slope of the factor price frontier of the economy at the world interest rate. Integration is represented by various adjustment paths to the diagonal. Path A represents an adjustment in which capital rapidly locates in the East, while labor moves from East to West rather slowly; the resulting "size of the East is not affected by the integration process. Since the West is assumed to operate at its steady state capital-labor output at the outset, this is equivalent to bringing capital from abroad. In Path B in contrast, labor moves rapidly from the East to the West, while capital sluggishly moves to the east and instead locates in the West. Path C represents a path in which investment is so small that the net evolution of the Eastern capital stock is negative, leading to a "national park in the East. <Figure 1 here> The assumption of constant returns suggests that there is no unique resting point of the economy; production is possible at any point along the diagonal, with no loss of per capita productivity, since both regions are producing at the same capital-labor ratio. So what criteria should guide society s choice? Are there perhaps other criteria which are also relevant for selecting the allocation of production factors? Not surprisingly, relative costs of adjustment will play a crucial role in determining the steady state. In the rest of this paper we seek to examine how total factor productivity, capital accumulation, and labor mobility have characterized the integration of Germany. Assessing Labor Productivity

11 11 Convergence in productivity one of the key open questions in macroeconomics remains the central policy question for eastern Germany s convergence. In their seminal paper Akerlof et al. (1991) showed convincingly that productivity in East Germany was much lower than West German level when the borders were opened. Even after a decade of economic integration and remarkable strides, productivity remains lower in the East, as Table 4 has documented. Yet the sectoral dimension reveals that the story is not clear cut. Table 6 examines sectoral growth in labor productivity over the past decade, which has also been examined in detail by Klodt (2000). Not only is the development uneven across sectors, it is also not monotonic in time. Most striking is the contrast between the monotonic rise in the aggregate, matched by most service sectors, with agriculture, which quickly reached parity with the West; with manufacturing, which has risen slowly since 1995 after sharp initial increases; with construction, which rose sharply but declined since 1996 by percentage points. This mirrors with a drop in relative wages and a decline in employment in the sector over the period. The decade of the 1990s was not one of stable catch-up. <Table 6 here> In this section we undertake to explain labor productivity in Germany from both a microeconometric as well as a macroeconomic perspective. In addition to an analysis of total factor productivity in the German states, we will study the German Socio-Economic Panel data (GSOEP), described in detail in SOEP Group (2001). The initial western sample was drawn in 1984, while the initial eastern sample was drawn in June 1990, just before monetary union. We use the sample years , primarily as cross-sections. The wage used is monthly earnings divided by 4.33 times weekly hours. Unless otherwise specified, wages are deflated with consumer price indices for east and west, and made comparable in purchasing power using results reported in Krause (1994). The samples of workers include those aged who are not self-employed or doing an apprenticeship, and who are not employed in agriculture. We exclude older workers since their employment rates are low due to early retirement. Quantity of Inputs

12 12 PHYSICAL CAPITAL From the perspective of a neoclassical, constant returns technology, a gap in labor productivity can result from too little physical capital per unit labor in the economy. This was certainly the situation in 1990 in East Germany and elsewhere in Eastern Europe, when the revaluation downward of national capital stocks led to a radical revision upward of the time to necessary for convergence (see for example Collins and Rodrik 1991). Any study of the role of private capital in the evolution of productivity should focus on investment expenditures on equipment (as opposed to structures). Equipment capital is known to be the key bottleneck for development, and high rates of investment in equipment are robustly associated with rapid economic growth (DeLong-Summers 1991, 1992). The robust role of equipment investment in economic development is supported by the fact that individual Western states evidence little or no timeseries or cross sectional variation of equipment capital to GDP ratios over the past two decades, while the same ratio for structures varies widely across the cross section dimension. The capital-output ratio in West states computed using capital stocks estimated by the Federal Statistical Office at 1991 prices is roughly , and has been so for two decades (Bavaria in 1998: about.88). 16 Low capital-output and capital-labor ratios for equipment imply that rapid catch-up in capital intensity is technically feasible, especially if there is a lot of outside help. To put things in perspective, Table 5 showed that cumulative investment in equipment in the period totaled DM billion in 1995 prices; real GDP in the new states including Berlin was DM billion. Assuming an initial equipment stock of zero in the East and a depreciation rate of 7.5%, one can cumulate the standard accumulation equation to arrive at a lower bound of the aggregate eastern K/Y ratio of 75-90% (for details, see the Appendix). Reality is more favorable, since much Eastern equipment could be employed in market activities in Many eastern enterprises already worked with Western capital goods, and investment was already occurring in 1990 and has since The assumption that the Eastern states had reached the lower bound of the Western equipment ratio by 1999 appears plausible; Klodt (2000), Ragnitz (2001:182) as well as Ragnitz et al (2001:71ff), argue as we do that physical capital per se is no longer a bottleneck to economic growth in Eastern Germany.

13 13 HUMAN CAPITAL. Eastern workers could also be less productive because they possess less human capital. However, the means of the GSOEP worker samples reported in Table 7 show that both in 1990 and 1999, Eastern workers are in fact more educated than western workers, confirming a number of other observers. 17 Although fewer easterners have degrees from tertiary institutions ("university"), more have vocational degrees that typically imply study beyond the usual dual-system apprenticeship, and the proportion with none of these degrees ("general schooling") is much lower than in the West. The unweighted means reflect the oversampling of western foreigners, but the latter observation is true even if only German nationals in the West are considered. (Other variables are little affected by the oversampling.) Although it is not possible to measure actual experience for the eastern sample, for women particularly average experience was much higher in the East in 1990, due to much higher employment rates under communism. This is reflected in the table in the average tenure with the employer in Burda and Schmidt (1998) show more formally using an Oaxaca decomposition that average characteristics tended to favor the East in <Table 7 here > Total Factor Productivity Eastern workers could also be less productive conditional on input quantities: that is, total factor productivity could be lower in the East. To assess this possibility, we apply the standard Solow decomposition of output growth in the West and the East since 1991, using our own estimates of the total capital stock for each Bundesland (details in the appendix). 19 Summary statistics are presented in Table 8, which performs the standard Solow procedure described in the Appendix to Eastern and Western data. In addition, the point raised recently by Ragnitz (1999a) and others that terms of trade (the relative prices of regional output) is addressed in the first column for both periods considered. We show the relative change in the "terms of trade of each state or region s output compared with German GDP deflator. Finally we display estimated total factor productivity (TFP) growth in both Eastern and West German states.

14 14 The most striking finding is the dramatic slowdown in TFP growth in the latter half of the 1990s, beginning in This development is systematic and hits all the states in the same qualitative fashion, although Saxony-Anhalt a state with capital-intensive chemical, machinery, and energy sectors exhibits low TFP growth for almost the entire period. 20 The Solow decomposition also reveals that while overall factor input growth for the East grew by only 2.4 log points per annum over the period, employment s per annum contribution fell by 1.8 points, while capital s rose by 4.2 %! Not only was the East German isoquant shifting, but a massive move was taking place along that isoquant in the Northeast (i.e. West German) direction. It is also noteworthy that West Germany and almost all of the individual West German states is hardly affected by the post unification decade. The second half of the period sees a return to normal TFP growth in the West. <Table 8 here> The Solow decomposition for the Eastern German economy since 1992 points to deficient total factor productivity growth as the main culprit for slowing convergence. This result is robust to various measures of the capital stock and to different employment measures. We proceed now to examine possible reasons for lower TFP. Quality of Inputs A first approach to explaining lower TFP in the East is to question whether the quality of the inputs being controlled for might not be lower in the East. Although, as documented, the amount of capital is lower in the East, its quality is almost surely higher than in the West as it is so much newer. Several aspects of the quality of labor must be considered. Experience gained under communism may be less valuable than that in a capitalist economy. Using the GSOEP, Krueger and Pischke (1995) document that the return to (potential) experience in the East fell from , and we have confirmed this using our sample and definitions (although the problem arises that only the wage and not the covariates are known for 1989). Curiously, however, Krueger and Pischke document using a 1988 dataset that the return was lower in 1988 than It is thus not quite clear whether the low return in

15 reflected a decline or not. Nevertheless, the experience gap in favor of the East may be lower than it appears. On the other hand, case study evidence suggests that at least in some contexts (and possibly with positively selected workers), the cooperative (brigade) system of working under communism prepared eastern workers well for new production techniques. Eastern workers adapted more readily to team-based production systems introduced in manufacturing in the 1990s than western workers did (Turner 1998). The schooling system worked somewhat differently in the West and the communist East, with, for example, less tracking in secondary classroom schooling, and more coordination between the classroom and firm components of the apprenticeship training in the East. Apprenticeships generally lasted two years in the East compared to three in the West, and took place in less well-equipped firms. On the other hand, easterners tended to complete more skilled apprenticeships than westerners. Tertiary education in practice was completed in less time in the East, but this reflects to some degree the inefficiencies of the western university system, where students repeat many courses (see Scheuer 1990 and Krueger and Pischke 1995). Productivity Gaps by Skill Level We can exploit the GSOEP to gain more insights into possible causes of the productivity gap, in particular by measuring how it varies with worker skill. We assume that the wage in both East and West represents the marginal product of labor, which will be the measure of productivity for the analysis. We also assume that western wages reflect the true market skill of a worker. We assign a skill level to each eastern worker by predicting what he or she would have earned in the West with his or her characteristics. We then plot the actual wages the individuals earn in the East against their predicted wage (skill). If skill were already rewarded as in the West, the actual wages would align along the 45 degree line. In reality we expect them to fall below the 45 degree line, indicating a productivity gap, and we are interested in the size of the productivity gap at different skill levels. Since we are conditioning on the worker characteristics, the composition of the workforces has been removed as a source of productivity difference. We cannot, however, condition on the worker s capital, so productivity gaps

16 16 here cannot be thought of as TFP gaps. To calculate the skill measure we run median log wage regressions for the West for men and women separately, pooling The "basic" covariates are year dummies, age, age squared, a foreign dummy, education dummies, and the interaction of the education dummies with age. The "extended" covariates also include tenure with the firm, firm size dummies, industry dummies, and dummies for whether a person reported they were working part-time or sporadically. The coefficients are used to predict wages (skill) for easterners. We set the value of the year dummy for 1988 to one for prediction. Skill is calculated for easterners for 1990 and The four panels of Figure 2 shows the results using the extended controls, for men and women, and for 1990 and People working in the West are excluded from the eastern sample. Rather than plotting the wage versus skill for each individual, we have plotted the median regression line through the points (solid line), and the 95% confidence intervals (dotted lines). The 45 degree line is also plotted. As expected, the regression line is below the 45 degree line in 1990 for men and women, and is also flatter. This indicates that easterners were less productive than westerners, and particularly so at high skill levels: the return to skill was lower in the East. By 1999 for both men and women the line has shifted upward and become steeper. The productivity gap has thus clearly diminished, and particularly so for the skilled: the return to skill has risen. <Figure 2 here> Each graph indicates the slope of the productivity (regression) line (a slope of one would be equivalent to the 45 degree line), and the log wage gap between the productivity line and the 45 degree line at the median skill. Men and women both begin with a median gap of 0.8 log points (a wage ratio of 2.2), which shrinks to 0.20 for men (a gap of 22%) and to only 0.08 (8%) for women. The male slope rises from 0.64 to 0.98, with the latter number being insignificantly different from one. The female slope rises from 0.79 to 1.26, the latter being significantly greater than one. In unreported graphs we have attempted to incorporate bonuses into the wage measure (thus reducing the sample size). For 1999 this

17 17 increases the median gaps by 0.02, and raises the female productivity slope to 1.31 and the male to To address the question of how productive Eastern workers are in the East, we restricted the analysis in Figure 2 to those working there. It might be argued that following up on Easterners in the West would reveal more about how productive easterners could potentially be. The difficulty is that these workers are not a random sample of easterners, but are likely to be positively selected, and the sample is changing over time. In addition, sample sizes are small. Nevertheless, in Figure 3 we repeat the analysis for the sample of easterners working in the West, due to emigration or commuting. For this analysis we simply use nominal wages for all workers. The initial year is 1991, since by construction of the sample there are no emigrants in 1990, and the commuting question was first asked in In 1991 men have a very low return to skill (slope of 0.5) and a 0.25 log point median gap. Large standard errors make it difficult to say anything about the women in By 1999, however, both slopes and gaps are insignificant for both men and women. 23 <Figure 3 here> Implications of Productivity Gaps by Skill Level Overall, it seems appropriate to characterize the 1999 return to skill as fairly similar in the East and West, with the productivity gap thus constant across skill. 24 This constancy rules out certain explanations for the remaining productivity gap. It is unlikely to be due to deficiencies in capital, for example, which we would expect to lead to a higher productivity gap for the more skilled. In this section we have deliberately not set the East a moving target, and are rather comparing with the pre-unification West in Klodt (2000) estimates that by 1999, the East had reached the western capital intensity of 1988 (see Klodt 2000). The implication is that if the East does close the contemporaneous capital intensity gap, the return to skill may be higher in the East. It is also unlikely that the remaining gap could be explained by a mismatch between eastern skills and imported western technology, since this is likely to affect different skill levels differently (see Acemoglu and Zilibotti 2001). We therefore concentrate below on explanations which could plausibly affect the productivity of all skills equally, such as

18 18 infrastructure and business skills. It is important to bear in mind that the wage structure and the decline in employment in the East are unlikely to be independent of one another. For example, Hunt (2002) documents the large decline in the gender wage gap in East Germany, and shows that several percentage points of the relative wage gain for women are due to employment declines among low-paid women. The employment decline will affect the median gap in the graph if the declines are disproportionately due to people whose actual wage is above or below their predicted wage. If this selection effect varies by skill, the slope of the line could be affected too. Other factors explaining the TFP growth slowdown LABOR HOARDING AND THE CYCLE. The data in Table 1 show that the Eastern State have experienced a sharp slowdown in growth starting in In the following year unemployment began rising. One explanation might then be a classic case of labor hoarding, as firms attempt to hold on to their best workers and avoid severance costs in the event the downturn is brief. To explore the possibility that the cycle may influence the behavior of the Solow residual, we regressed the benchmark estimates summarized in Table 8 on a complete set of fixed effects (time and state), plus an Eastern set of time effects, plus log first difference of unemployment and its lag. We also interacted the unemployment covariates with an East dummy. The results are presented in the first two columns of Table 9, and show that indeed there is an effect of unemployment and it is qualitatively different for East and West over the period examined. Unlike the West, where increases in unemployment are associated with cyclical declines in productivity, TFP in the East responded positively to changes in productivity. This must represent the effect of restructuring and moving out to the efficient production frontier. < Table 9 here> INFRASTRUCTURE. Another candidate systemic explanation for the East German TFP growth

19 19 slowdown after 1995 is the changing evolution of economic infrastructure in the East. By infrastructure we mean the stock of public or semi-public goods which also contribute to production outcomes. Examples are highways roads and bridges, telecommuncation networks, airports and harbor terminals, as well as universities, hospitals, police and fire departments and public utilities---which tend in Germany to be publicly funded and/or highly regulated. Although controversial, there is some evidence that sustained infrastructure investment may promote economic growth. 25 As with private productive capital, it is widely agreed that infrastructure in eastern Germany in 1990 was severely deficient compared with the West. Despite immense outlays almost a third of all investment since reunification - in the productive dimension it continues to be inferior in many respects (DIW (2000b, 2001)). 26 Moreover, the evidence indicates a decrease in spending by states and local authorities which set in during the mid-1990s (DIW 2000a,b). As it is thought to affect the productivity of all factors, infrastructural backwardness is a candidate deficiency that could affect skill levels equally and thus account for the striking results on productivity above. In the third and following columns, we add the first difference in log per capita infrastructure stocks as estimated by the DIW (2001a,b). We find a robust positive sign on the total and, when disaggregated, of equipment infrastructure capital stocks (as opposed to streets, buildings, sewage, and tunnels). MANAGERIAL TALENT, ENTREPRENEURSHIP AND MARKETING SKILLS. Certain business skills either did not exist under communism, or did not transfer well to the post-communist period, and these deficiencies could explain part of the TFP gap. Writers of case-studies have noted that eastern managerial and organizational skills were lower in the early years of transition. Managers had to adapt to working in much smaller firms, working more cooperatively with workers, and working with workers who resented them if they had also been successful under communism. The move towards capitalism also demanded a knowledge of marketing, law and economics, (non-russian!) foreign language skills, and a culture of entrepreneurial risk-taking. Many analysts believe these considerations still hold true. Ragnitz (1999a, 2001) argues that a continuing lack of marketing skills leads to eastern products being sold at a discount, which directly

20 20 influences the productivity statistics. The interesting question, however, is why such deficiencies have not been remedied in the eleven years since monetary union. A high return to these scarce skills should have induced investment by easterners and perhaps mobility by westerners. In the early transition, western firms sent managers on temporary assignments to the east, and we show evidence below that easterners were sent west for training. Possibly building up easterners stock of business knowledge is simply a very slow process, and western managers could be an inefficient solution if they impose management structures that are inappropriate given the skill mix of the workers. 27 CREDIT CONSTRAINTS. Another hypothesis for low productivity in the East is that eastern firms suffer from credit constraints. This could also have a skill-neutral effect. Neuberger (2000), for example, finds that there is a "finance-gap" between East and West German firms. In line with the new literature on financial multipliers and credit squeezes, firms tend to get credit as an increasing function of their capitalization, meaning that the new firms in the East are at a natural disadvantage. Turner (1998) gives example of the "Niles" machine tools plant in Berlin which had difficulties obtaining credit before it was privatized in May SECTORAL SHIFTS AND FIRM SIZE. Many analysts point to the industrial structure of East Germany and the predominance of small firms as reasons for its low productivity. It is well-known that observationally similar workers are paid more in larger firms (at least, in large firms that are profitable in market economies), and in certain capital-intensive industries. The GSOEP means in Table 7 show that in 1990 eastern workers were in larger firms than western workers, but by 1999 the reverse was true. Our own calculations using data from Institut für Wirtschaftsforschung-Halle confirm this trend: the fraction of workers in industrial establishments with 20 or fewer employees rose in the East from 1.4% in 1991 to 6.6% in 1999, while it actually fell from 4.5% to 4.3% over the same period in the West. Table 10 shows the collapse of manufacturing's share in employment to only 16%, and the rise of construction's share to 17.4% in Employment in services has risen, but still lags behind the West (some of the rise represents a relocation of services that used to be provided by manufacturing firms). The data show, however, that the construction boom has peaked. Thus the relatively constant

21 21 employment level since 1993 conceals significant sectoral shifts. It is interesting to note, in Table 11, that the shift in value added by sector have sometimes exceeded the shifts in employment. These shifts are the most promising sign we can detect that more aggregate indicators for East Germany may yet resume convergence. 28 <Tables 10,11 "Employment Share and Value Added Share here> It is not clear to what extent small firms and a poor industrial mix are causes or symptoms of slow convergence. Firms in the East may be small because they are unsuccessful, possibly specializing in non-tradables or small markets for goods with high transport costs, as suggested by Von Hagen and Strauch (2000). Alternatively, their small size may reflect policy errors that have led to low investment from abroad, such as high wages. Our own findings, reported in the last columns of Table 9, indicate only a weak negative influence of firm size on TFP growth. As for the industrial composition, if it is not a policy choice, noting that the industrial mix is poor may not advance analysis much. Sinn (1995) and Klodt (2000), however, stress that to some degree the industrial mix has been a policy choice through the medium of subsidies to capital. Our analysis thus far has controlled for industry and firm size, directly in the wage-based analysis, and in the Solow residuals indirectly via capital. Industry and firm size are thus not part of the residual productivity gaps we have been trying to explain. Furthermore, controlling for industry and firm size in the wage-based analysis only reduces the 1999 median gaps by 0.01 (this is not shown). Ragnitz (2001), by contrast, finds using a different methodology that firm size alone accounts for 17 percentage points of the raw productivity gap. Despite our finding that industry and firm size are not so important in a static accounting sense, we believe that insights for long run growth can be gained by considering the industry mix issue. Hunt (2001a) observes that after the first year of transition, incentives to change jobs and industries appear to have been low, and job-changing rates fell. Using the GSOEP, she calculates that over the

22 22 period only 18% of real wage growth went to job changers within the East, and 7% was due to movement to and from jobs in the West. This is surprisingly low compared to the 22% of real wage growth in the West that went to job-changers. In the first year of transition, voluntary movers within the East gained 15% more than similar workers who stayed with the same employer, and the whole gain was due to changing industry. Subsequently wages continued to rise, but movers had no significant gain over stayers. These results suggest that wage-growth patterns after 1991 were not conducive to sectoral shifts, as stayers were too well rewarded. In the period it may be appropriate to think of the labor unions choosing wages (this is discussed further below). In some industries wage rises for the period were laid out in advance in The unions, bargaining at the industry level, did take the prospects of the industries into account when setting wages, but it is nevertheless possible that they did not pick the right relative wages across industries. High wages in "bad" industries could have delayed restructuring in a fashion similar to the model of Aghion and Blanchard (1994). However, after 1993 the power of the unions was greatly reduced. In order to inculpate the unions, it is necessary to argue that nominal wage rigidities and low inflation coupled with a high overall wage level made it hard for relative wages to adjust. MIGRATION AND COMPOSITION EFFECTS. While our analysis shows that East Germans are as productive as westerners conditioning on human capital attributes, it may be the case that migration has removed these individuals from the sample. In this respect we can identify an important link between productivity and migration: if the most productive individuals leave or have the greatest propensity to leave, the Solow residual as calculated in Table 8 will mismeasure true input (by considering workers unweighted by human capital input). An obvious modification is to redo the Solow analysis for more disaggregated labor types. Using data available from the Mikrozensus since 1991 on employment by age, we construct an extended version of the Solow decomposition which weights different employment age groups by their share in labor income estimated using the GSOEP (for details, see the Appendix). The results, which presented in Table 12, show that TFP rises after this correction, which is consistent with the hypothesis that input of more productive individuals (in the year old range) is declining in