How do regional labor markets adjust to immigration? A. dynamic analysis for post-war Germany

Transcription

1 How do regional labor markets adjust to immigration? A dynamic analysis for post-war Germany Sebastian Braun Henning Weber PRELIMINARY AND INCOMPLETE February 16, 215 Abstract We analyze how regional labor markets in West Germany adjusted to the mass inflow of German expellees after World War II. The inflow was large and very unevenly distributed across West German regions. Historical data, covering a period from just before to 25 years after the inflow, shows that both the relative unemployment rate of high- to low-inflow regions and regional migration jump up after the inflow and decline only gradually over the next decade. To interpret these empirical facts, we fit a dynamic search and matching model of two connected regional labor markets to the data. The model predicts that it takes seven quarters before half of the expellees are employed. Furthermore, for any ten expellees who arrive in the high-inflow region, a maximum of five native workers lose their job. Of those five workers, 1.5 workers leave the labor force, 2.5 become unemployed, and one worker migrates to the other region. The inflow decreased the expected lifetime income of the average native worker by 1.6%. JEL Code: J61, F22 Keywords: Immigration, labor market adjustments, dynamic model, post-war Germany This paper represents the authors personal opinions and does not necessarily reflect the views of the Deutsche Bundesbank, the Eurosystem, or its staff. The paper has benefited from comments and suggestions by Klaus Adam, Michael Binder, Marcus Böhme, George Borjas, Larry Christiano, Fabio Ghironi, Falko Juessen, Wolfgang Lechthaler, Guido Lorenzoni, Christian Merkl, Horst Raff, Claire Reicher, Michael Reiter, Dennis Snower, Ignat Stepnok, and participants of the 213 Norface Migration Network Conference, and seminars at the Kiel Institute for the World Economy and Georg-August-Universität Göttingen. Richard Franke provided excellent research assistance. Kiel Institute for the World Economy, Hindenburgufer 66, 2415 Kiel, Germany, Tel: , sebastian.braun@ifw-kiel.de. Deutsche Bundesbank, Wilhelm-Epstein-Strasse 14, 6431 Frankfurt am Main, Germany, Tel: , henning.weber@bundesbank.de.

2 1 Introduction This paper studies how regional labor markets adjust to a regionally concentrated inflow of migrants. It does so in the context of one of the largest population movements in modern history, the inflow of German expellees to West Germany after World War II. The eight million expellees that had arrived in West Germany by 195 were very unevenly distributed across the country, with expellee shares that ranged from 5% in the federal state of Rhineland-Palatinate to 33% in Schleswig-Holstein. We exploit this large, unexpected, and highly unequal inflow of expellees to take up two key questions: First, through which channels did regional labor markets adjust to the inflow of expellees? Second, how did the inflow affect natives labor income along the adjustment path? Our paper contributes to a large literature in labor economics that analyzes labor market adjustments to immigration. 1 Over the last two decades, interest in the issue has been fuelled by a sharp increase in the number of international migrants and rising public concern over the consequences of immigration. The existing literature focuses on changes in wages as the key mechanism through which labor markets adjust to immigration. Moreover, the literature typically analyzes the wage effects of immigration in a static framework. In contrast, remarkably little is known about the dynamic labor market effects of immigration and about the time that it takes the labor market to digest an immigration-induced labor supply shock (Borjas 214). Addressing this gap in the literature, our paper offers a dynamic perspective on labor market adjustments to immigration. It decomposes regional inflows into changes in employment, unemployment, non-participation and outflows to other regional labor markets, and sheds light on the relative importance of these different adjustment margins along the adjustment path. It also studies the effect of immigration on the lifetime income of natives, and contrasts the shortand long-run effects of immigration on income. Our specific historical setting, the influx of expellees to post-war Germany, significantly facilitates the identification of the effects of interest. There are various reasons for this. First, expellees could generally not choose their initial destination in West Germany based on local employment prospects. They either fled to safe regions close to their old homelands or, after the end of the war, were brought to post-war Germany in compulsory transports. Regional inflow rates were, therefore, largely independent of local economic conditions (Braun and Kvasnicka 1 Friedberg and Hunt (1995), Okkerse (28), Longhi, Nijkamp, and Poot (21), Kerr and Kerr (211) and Borjas (214) provide comprehensive reviews, meta-analyses, and critiques of the existing literature. 1

3 214). Second, expellees could not return home after they were forcefully uprooted, as their former homelands lay beyond the Iron Curtain during the Cold War. Their immigration to West Germany was, therefore, permanent. Third, expellees and natives were very close substitutes on the West German labor market. They both spoke German as their mother tongue and their pre-war level of education was almost identical (Bauer, Braun, and Kvasnicka 213). Taken together, these reasons will allow us to identify how regional labor markets adjust over time to a large, exogenous and permanent increase in the supply of homogenous labor. We start our analysis by presenting key empirical facts on the economic development of regions with high and low inflow rates of expellees. We do so by contrasting two stylized regions whose economic situations were similar in 1939, i.e., before the inflow. The high-inflow region H then received considerably more expellees than the low-inflow region L. By 195, the population share of expellees was 25% in region H but only 11% in region L. We compare the economic performance of the two regions before and after the inflow and document their subsequent relative performance until 197, a quarter century after the inflow. Our analysis thus covers a much longer time horizon than most papers do. We show that by 195, the unemployment rate of region H exceeded the rate in region L by a factor of two and a half. Regional unemployment rates then gradually converged during the 195s, and both regions recorded full employment at the beginning of the 196s. The decrease in unemployment went hand in hand with an increase in labor force participation. Data on GDP per capita show a pattern similar to that on unemployment: Region H first fell behind region L but then grew faster in the 195s. The convergence process in unemployment rates and relative GDP was accompanied by large regional migration flows. In 195 alone, net migration from region H to L amounted to 1.9% of H s population. Migration was, however, not sufficiently strong to fully restore the pre-inflow population ratio of the two regions. We also show that expellees were over-represented both among the unemployed and among regional migrants. To explain these novel facts, we develop a dynamic structural model of two connected regional labor markets and estimate it using the historical data. An important advantage of our structural model over a purely empirical reduced-form model is that the structural model allows us to identify the causal labor market effect of the expellee inflow even in the presence of regional migration. Regional migration diffuses the effect of the expellee inflow from region H to L, so that a simple comparison of labor market outcomes between regions H and L will not yield the causal effect of receiving high rather than low expellee inflows (even if the initial distribution of 2

4 expellees was random). 2 The backbone of our structural model is the search and matching model of unemployment that we extend in three directions to apply it to the particular historical episode. These extensions are migration between two regional labor markets subject to migration costs, endogenous labor force participation, and expellee inflow. We model expellee inflow as exogenous and permanent increase in the number of non-employed workers and calibrate its regional distribution to historical data. The remaining part of the model is standard. In each region, a large firm employs many workers and accumulates capital to produce output. The firm faces costs when it adjusts its workforce or capital stock. We use these adjustments costs, the migration costs, and the responsiveness of the participation margin to fit the model to the historical data. We then use the parametrized model to answer our two key research question. First, we analyze through which channels and how fast regional labor markets adjusted to the inflow of expellees. We find that upon impact, the inflow of expellees is mainly absorbed through increases in unemployment and decreases in participation. The response in job creation is not sufficiently fast to prevent unemployed workers from leaving region H in great numbers. We show that the employment response of natives in region H reaches its peak 1 quarters after the arrival of expellees. When measured at that time, we find that for any ten expellees who arrive in region H five native workers lose their job. Of those five workers, 1.5 leave the labor force, 2.5 enter the unemployment pool and one worker leaves region H for region L. While the effect on unemployment and participation disappears over time, the number of natives who leave H for L grows. Ten years after the shock, two natives have left region H for any ten expellees who entered. Second, we calculate the causal effect of the expellee inflow on native income along the adjustment path. We do so by comparing the income stream of natives in the actual treatment scenario to that in a counterfactual scenario, in which we set migrant flows to zero. We show that the wage effects in regions H and L are largest xxx and xxx quarters, respectively, after the shock. We then calculate the overall effect of the expellee inflow on the net present value (NPV) of workers expected lifetime labor income. We find that the inflow reduce lifetime labor income of the average native by xxx percent. Since in our model, workers also derive income from being unemployed and out of the labor force, the overall income loss is only around half the fall 2 Such simple comparison would violate the stable unit treatment value assumption (SUTVA), which in our case requires labor market outcomes in one region to be unaffected by expellee inflows to the other region. 3

5 in labor income. We also find strong heterogeneities in the magnitude of income losses: Losses are particularly strong for natives who are unemployed at the time of the inflow. In contrast, employed natives are initially shielded from the adverse effects that immigration has on the job finding rate. Finally, we use the model to perform a number of counterfactual exercises. We analyze how the regional distribution of the expellee inflow affect (...) Cohen-Goldner and Paserman (211) provide one of the rare attempts to study the dynamic effects of immigration on native wages and employment. The paper analyzes the impact of the large and unexpected inflow of Soviet Jews into Israel after the collapse of the Soviet Union. It relates the migrant share in a specific labor market segment to native wages and employment in the same segment, and allows the effects of immigrants to vary with their years of tenure in the Israeli labor market. The authors show that the initially negative wage effect of the inflow dies out after 5 to 7 years. For their empirical approach to be valid, labor market segments have to be isolated. 3 In contrast, our structural approach directly accounts for movements across regional labor market segments and allows us to quantify the relative importance of such movements as an adjustment mechanism. A number of papers have specifically studied the link between immigration and internal migration of natives. However, no clear picture has yet emerged from this literature. While some studies suggest that natives indeed respond to immigration by moving out to other areas (see, e.g., Filer (1992) or Borjas (26)), others find no such effect (see, e.g., Card and DiNardo (2) or Card (21)). Our paper demonstrates that the effect of immigration on native out-migration will crucially depend on the magnitude of the immigrant inflow, the degree to which the inflow is region-specific (rather than common to all regions of a country) and the time elapsed since the inflow. Our paper is also related to the seminal work of Blanchard and Katz (1992) who analyze adjustment of US states to regional labor demand shocks. The paper finds that adverse labor demand shocks drive up unemployment, inducing workers to emigrate to regions with lower unemployment rates. In the longer run, adjustment to adverse demand shocks occurs mainly through inter-regional migration. Employment will thus increase permanently in regions hit by 3 The authors use four different definitions of labor market segments. The first defines labor market segments based on two-digit occupations, the second based on one-digit occupations interacted with district of residence, the third based on one-digit occupations interacted with one-digit industries, and the fourth defines segments based on levels of schooling interacted with experience. 4

6 a shock and will decrease permanently elsewhere. In contrast to Blanchard and Katz (1992), our paper assesses the relative importance of inter-regional migration and other mechanisms for adjusting to regional labor supply rather than labor demand shocks. Our specific historical setting, the mass displacement of ethnic Germans, has only lately gained attention among economists. Two recent empirical studies quantify the economic effects of the expellee inflow to West Germany. 4 Braun and Mahmoud (214) use data from the 195 population census to study the effect of the inflow on native employment. They find that in the short run, expellees had a substantial negative effect on native employment. Braun and Kvasnicka (214) analyze the effect of the expellee inflow on sectoral change and output growth between 1939 and 195. The authors find that the inflow fostered structural change away from low-productivity agriculture but reduced regional output per worker. In contrast to our paper, Braun and Mahmoud (214) and Braun and Kvasnicka (214) focus on the short-run effect of the expellee inflow and do not quantify the relative importance of different channels through which the West German economy adjusted over time to the expellee inflow. This paper proceeds as follows. Section 2 provides background on our historical setting, and Section 3 derives empirical facts on the economic development of regions with high and low inflow rates of expellees. Section 4 develops the structural model through which we interpret and analyze the empirical data. Sections 7 and 8 contain our main results: Section 7 analyze through which channels regional labor markets adjusted to the inflow of expellees, and Section 8 sheds light on the income effects of the inflow. Section 9 reports the results from our counterfactual exercises. Finally, Section 1 concludes. 2 Historical background The flight and expulsion of Germans from East and Central Europe took place between 1944 and 195, and occurred in three different phases. The first phase of the displacement began at the final stages of World War II. As the Red Army advanced westwards, hundreds of thousands of Germans from the Eastern provinces of the German Reich fled further inland. Most of these refugees planned to return home after the end of the war, and therefore fled to regions close to their old homelands. After Nazi Germany s unconditional surrender in May 1945, some refugees indeed managed to return home. However, 4 In addition, Bauer, Braun, and Kvasnicka (213) and Falck, Heblich, and Link (212) analyze the economic integration of expellees in post-war West Germany. 5

7 Figure 1: German territorial losses in World War I and II and the Sudetenland Memel Danzig Pomerania East Prussia Berlin West Prussia Posen East Germany East Upper Silesia Silesia West Germany Territories lost at Treaty of Versailles, 1919 Free City of Danzig, Territories lost to Poland / Soviet Union, 1945 Sudetenland, Czech Territory annexed by Germany in 1938 Polish authorities quickly prevented refugees from returning to their former homelands and began to expel the remaining German populations. The Czech authorities soon followed the Polish example. These so-called wild expulsions, which marked the second phase of the displacement, were not yet sanctioned by an international agreement. The third phase began after the Soviet Union, the United Kingdom and the United States concluded the Potsdam Agreement of August The Agreement legalized and sanctioned the expulsion of Germans from Eastern Europe and shifted Germany s border westwards to the Oder-Neisse line. The former eastern provinces of the German Reich were placed under Polish or Russian control (see Figure 1). Germans remaining east to the new border were brought to post-war Germany in compulsory and organized transfers. The German territory west to the Oder-Neisse line was divided into four occupation zones: a British, a French, an American and a Soviet zone. Overall, the mass exodus of Germans from East and Central Europe involved at least 12 million people. Most expellees re-settled in the territory of what was to become the Federal Republic of Germany (henceforth West Germany). The mass inflow of expellees dramatically increased the population of West Germany. Despite heavy war losses, the West German pop- 6

8 ulation grew from 39 million in 1939 to 48 million in 195. By September 195, there were 7.9 million expellees in West Germany, accounting for 16.5% of the West German population. The majority of expellees, around 4.4 million, came from Germany s former eastern territories. Another 1.9 million expellees had lived in Czechoslovakia before the war, most of them in the Sudentenland. 5 Most of the remaining expellees came from regions that Germany had lost after its defeat in World War I. The population share of expellees differed greatly across West German states, and ranged from 5.% in the state of Rhineland-Palatinate to 33.% in Schleswig-Holstein (see column (3) of Table 1). There are three main reasons for the very uneven regional distribution of expellees. First, German refugees who fled the approaching Red Army at the final stages of the war (first phase of the displacement) mainly sought shelter in regions close to their old homelands. Many Germans from the Sudetenland, for instance, fled to neighbouring Bavaria. The wild expulsions (second phase) only added to the regional imbalances, as Polish and Czech authorities often just drove Germans across the border into occupied Germany. Second, the French initially refused to admit any of the organized expellee transfers (third phase) to their zone of occupation (the French did not feel bound by the Potsdam Agreement, as they had not been invited to the Potsdam Conference). Therefore, expellees were initially transferred to the American and British occupation zones only. Third, the Allied bombing campaigns had destroyed large parts of the housing stock in West German cities. Expellees were thus mainly placed in rural areas (Connor 27). Importantly, the initial distribution of expellees in post-war Germany was not driven by local labor market conditions. 5 The Sudetenland, though mainly inhabited by German speakers, had become part of the independent Czechoslovak state after the collapse of the Austro-Hungarian Monarchy in It was annexed by Nazi Germany in September

9 Table 1: Expellees inflows, pre-war differences and war destruction in West German states Expellee 1 inflows Pre-war differences War destruction expellee % expellees % expellees Population 1938 un- Share of national Share population population in 195 in 1946 change, employment labor force in income per destroyed (in 1) (in 1) population population (%) rate (%) 2 agriculture (%) capita (RM) 3 flats (%) 4 (1) (2) (3) (4) (5) (6) (7) (8) (9) Bavaria 9,184 1, Lower Saxony 6,797 1, Schleswig-Holstein 2, Region H 18,576 4, Baden-Wüerttemberg 6, Hesse 4, Rhineland-Palatinate 3, Region L 13,759 1, North Rhine Westphalia 13,196 1, Bremen Hamburg 1, Region L 29,119 3, Federal Republic 47,696 7, Data sources: Data on the population and the expellee share in 195, and on the population and the agricultural employment share in 1939 are from Statistisches Bundesamt (1954). Data on the expellee share in 1946 are taken from Statistisches Bundesamt (1952). Data on population in 1925 are from Hohls and Kaelble (1989). Data on national income and on 1938 unemployment rates are from Länderrat des Amerikanischen Besatzungsgebiets (1949), and data on the share of destroyed flats in 1946 are from Deutscher Städtetag (1949). Notes: 1 Expellees are defined as German nationals or ethnic Germans who on 1 September 1939 lived (i) in the former German territories east to the Oder-Neisse line, (ii) the Saarland or (iii) abroad, but only if their mother tongue was German. Figures for 1946 count all people who have lived abroad as expellees, as the census did not record their mother tongue. 2 The unemployment rate is expressed as a percentage of the dependent labour force. Pre-war unemployment data is not available for the West German states in their post-war borders. The unemployment rate of region H is approximated by the labor-force-weighted average of the unemployment rates in the employment agency districts of Bavaria, Lower Saxony, Nordmark, the unemployment rate of region L by the average of the unemployment rates in Hesse, Southwest Germany, the Rhineland and Westphalia. 3 Pre-war national income data is not available for the West German states in their post-war borders. National income of region H is approximated as the population-weighted average of national income in the Reich provinces of Bavaria, Hannover, and Schleswig-Holstein. National income of region L is approximated as the average income of Baden, Württemberg, Hesse, and Hesse-Nassau, the Rhine Province and Westphalia. 4 The share of destroyed flats is calculated as the share of completely destroyed flats in percent of the housing stock in mid-1943.

10 The occupying powers wanted all German residents, no matter whether they were expellees or not, to remain wherever they had arrived in Germany or had been located on the day of armistice. Therefore, relocation were banned altogether in the immediate post-war period. After the total ban was abolished in 1947, moving required permission from the military administration (permission was granted mostly for family reunification). The restrictions on regional mobility were only lifted in May 1949 (Müller and Simon 1959, Ziemer 1973). In addition to legal restrictions, the heavy war destructions and the resulting lack of housing space impeded regional mobility. Potential migrants also had little written information about local conditions elsewhere, as the Allies rationed paper and banned almost all German newspapers in the immediate aftermath of the war. For all these reasons, the regional distribution of expellees remained largely unchanged until 195. In fact, the state-level expellee shares in 1946 and 195 are highly correlated, with a correlation coefficient of.997 (see Table 1, columns (3) and (4)). Expellees were relatively close substitutes to native West Germans on the German labour market. They were all German native speakers, and had been educated in German schools. Expellees were also not a selected sub-group of their home regions, as virtually all Germans living east of the Oder-Neisse line were forced to migrate. In addition, expellees and natives had mostly lived in the same country for decades (the ceded eastern provinces, home to most expellees, had been part of the German Reich since its foundation in 1871). Table 2 compares the sex ratio, age structure, marital status and level of education of expellees and native West Germans in 195 and shows that the two groups were indeed very similar in their socio-demographic characteristics. As evident, there were more women than men among both expellees and non-expellees (a legacy of the two world wars). Expellees were slightly younger than natives and, therefore, also less likely to be married. Compared to other migration episodes, however, age differences were small. Expellees had also almost identical years of schooling than natives, a similar probability to have completed vocational training, and a slightly higher probability to have graduated from university. 3 Empirical facts on regional development In this section, we present empirical facts on how Germany s regions adjusted to the large inflow of expellees. To this end, we compare the demographic and economic development of two stylized regions, a high-immigration region H and a low-immigration region L, between 1939 and

11 Table 2: Socio-demographic characteristics of expellees and non-expellees in West Germany, September 195 Expellees a Rest of the population b % females Age structure % aged % aged % aged % aged % aged 6 and above Marital status (aged 18 and above) % single % married % widowed or divorced Education (born ) c Years of schooling d % vocational training % university degree Data sources: All data except for educational attainment are from the census of 13 September 195, as published in Statistisches Bundesamt (1952). Figures on education are from own calculations based on a 1 percent sample of the census of 27 May 197 (FDZ der Statistischen Ämter des Bundes und der Länder 28). The table is reproduced from Braun and Kvasnicka (214). Notes: a Expellees are defined as German nationals or ethnic Germans who on 1 September 1939 lived (i) in the former German territories east to the Oder-Neisse line, (ii) the Saarland or (iii) abroad, but only if their mother tongue was German. b The education statistics distinguish between expellees and native West Germans (excluding foreigners). All other statistics distinguish between expellees and the rest of the population. c The education statistics are for those who were born between 1885 and 1927 (aged 23 to 65 in 195). The overwhelming majority of these persons should have completed their education by 195. d We only have data on the highest school degree. Years of schooling are inferred from the minimum years of schooling required to obtain a particular degree. The outcome variables that we consider are population, GDP per capita, and unemployment rates. We also describe migration between the two regions for the post-war period. 3.1 Regional classification and pre-war differences Table 1 shows how we classified the West German federal states (Bundesländer) into a high- and a low-immigration region. It also provides an overview of expellee inflows into these regions, of selected economic characteristics before the war and of the regional degree of war destruction. The H-region consists of the three states of Bavaria, Lower Saxony, and Schleswig-Holstein. These three states were called refugee states (Flüchtlingsländer) in contemporary publications. Although the refugee states accounted for only a third of West Germany s pre-war population, they hosted roughly 6% of all expellees in September 195. The population share of expellees 1

12 in the H-region was 25.% (see column (3) of Table 1). The L-region consists of the remaining six West German states 6, namely Baden-Wüerttemberg 7, Hesse, Rhineland-Palatinate, North Rhine-Westphalia and the city states of Bremen and Hamburg. The population share of expellees in the L-region was 11.1% in September 195, and thus less than half of the population share in the H-region. Before we compare the actual economic development of the H and the L regions after the expellee inflows, we briefly study whether the two regions already differed before the war, and document regional differences in the degree of war destructions. Any such (pre-existing) difference might provide an alternative explanation, other than differential expellee inflows, for observed differences in regional economic development after the war. Columns (5) to (8) of Table 1 report regional pre-war data on population growth, unemployment, agricultural employment, and national income per capita, and column (9) provides the share of houses destroyed in the war. Before the war, population growth rates were very similar in the two regions. The population of the H-region grew by 1.7% between 1925 and 1939, only.7 percentage points more than the population of the L-region. Likewise, unemployment rates were very similar before the war, reaching 1.6% in the H-region and 2.% in the L-region in 1938 (see column (5)). However, Table 1 also shows that the H-region was somewhat poorer than the L-Region (see column (7)). In fact, national income per capita was around 8% lower in H than in L in An important reason for the income differences between the two regions were differences in the sectoral structure of the economy. Region H was less industrialised and more rural in nature than region L. In 1939, 36.5% of the labour force in region H but just 21.8% in region L worked in the low productivity agricultural sector. Since the Allied bombing campaign targeted primarily German cities, the more rural region H also suffered less from war destructions than region L. Around 12% of all flats in region H were destroyed during the war, considerably less than the West German average of 2.3%. We deal with these pre-existing differences between the two regions in two ways. First, we 6 West-Berlin and the Saarland were not (yet) part of West Germany in Baden-Württemberg was formed on the territories of the formerly independent states of Baden, Württemberg- Baden, and Württemberg-Hohenzollern in Since neither unemployment nor income data is available for the West German states in their post-war borders, we had to approximate the values for the H- and L-region. We approximate 1936 national income in region H as the population weighted average of income in the former provinces of Baveria, Hannover, and Schleswig-Holstein. We then use the average national income in the rest of West Germany (excluding Hamburg and Bremen, Oldenburg, Braunschweig, Lippe, Schaumburg-Lippe) as a proxy for national income in region L. We discard the data for Bremen, Oldenburg, Braunschweig, Lippe, and Schaumburg-Lippe as data for these provinces are only reported as an aggregate figure. We must stress that the aggregation from the former provinces of the German Reich to our stylized regions is far from perfect. Therefore, the figures should be read with some caution. 11

13 directly account for them in our model. In particular, we allow for regional differences in the degree of war destructions. Second, we use an alternative classification of federal states that levels out pre-existing differences between the high- and the low-inflow region. We then show that both the empirical facts and our theoretical results are robust to the use of this alternative classification. How does the alternative classification look like? Table 1 shows that the differences between regions H and L in the average sectoral structure and the degree of war destruction come mainly from the state of North Rhine-Westphalia, the most populous state of West Germany, and from the two city states of Hamburg and Bremen. North Rhine-Westphalia comprises the Ruhr region, Germany s pre-war industrial center. Therefore, only 14.1% of North Rhine-Westphalia s labour force was in agriculture before the war (compared to a national average of 27.%). North Rhine-Westphalia is not only highly industrialized but also highly urbanized, and thus suffered over-proportionally from war destructions. The same is true for the city states of Bremen and Hamburg, which comprise only urban areas, and had almost no agriculture in The alternative classification, therefore, excludes North Rhine-Westphalia and the two city states from the low-immigration region. The resulting geographical entity, to which we refer to as the L region, consists only of Baden-Württemberg, Hesse, and Rhineland-Palatinate. Table 1 illustrates that regions H and L are very similar not only in terms of pre-war population growth and unemployment but also in terms of agricultural employment and war destructions. The agricultural employment share in region L was only slightly lower than in region H in 1939 (32.1% vs. 36.5%). Likewise, the share of destroyed flats, our measure of wartime destructions, was virtually identical in the two regions (12.1% and 12.8%). While the L- and the H-region were thus similar in their pre-war economic structure and their degree of war destruction, the L -region experienced a much smaller inflow of expellees (12.6% vs. 25.%). The main problem with using the alternative classification is that it excludes almost a third of the West German population. Moreover, inner-german migration can no longer flow only between the high- and the low-inflow regions (but also between the two regions and the rest of the country). 9 9 When using the alternative classification in our two-region model, we deal with this problem by directly feeding into the model the observed migration flows between L and the rest of the country and between H and the rest of the country (see Section xxx). Flows from and to the rest of the country are thus exogenously given and not endogenously explained by the model. Alternatively, we could have extended our model to incorporate a third region. However, extending our model would have considerably complicated our analysis. Moreover, the third region, consisting of North Rhine-Westphalia, Hamburg and Bremen, considerably differs from the other two regions in its pre-inflow characteristics (therefore bringing us back to our initial problem). 12

14 3.2 Empirical facts We compare the relative demographic and economic development of the H- and L-regions in 1939, i.e., immediately before the war, and in 195, i.e., shortly after the expulsion was complete. The pre- to post-war comparison of outcome variables in the H- and L-regions is followed by a description of the subsequent development of the two regions until 197. Of particular interest to us is the question of whether the relative size and economic performance of the two regions returned over time to their pre-immigration value and, if so, how quickly they did. Appendix A.1 contains the same analysis for regions H and L. Importantly, the empirical facts that we obtain by comparing H and L are very similar (in a qualitative sense) to those that we obtain by comparing H and L. Population and Internal Migration. Figure 2 shows the population size of region H relative to that of region L between 1939 and 197. It distinguishes between the actual development and a hypothetical one. The latter shows how the relative population size of the two regions would have evolved if the only reason for changes in the relative population size was inter-regional migration between H and L. 1 Thereby, it abstracts from other potential influences, such as differences in fertility rates or migration from abroad, on the relative population size of the two regions. As evident, the inflow of expellees changed the relative size of the two regions dramatically. The ratio of region H s to region L s population jumped up from.56 in 1939 to.65 at the end of It then gradually came down again without reaching its pre-inflow value (the population ratio stood at.537 in 197). The hypothetical line shows that inter-regional migration from H to L was by far the most important factor in moving the relative population size of the two region back towards its pre-war level. Figure 3 depicts the net emigration rate from region H to L, i.e., net emigration from H to L in a given year in percent of H s population in the previous year. The graph also provides 1 The hypothetical series is calculated by adding to the actual population figures on 31 December 1949 (cumulated) net migration between the two regions. 11 The inflow of expellees was, by far, the most important driver of this dramatic increase in relative population (see Braun and Mahmoud (214) for a comprehensive overview of regional population changes in West Germany between 1939 and 195). However, it was not the only one. Even without the inflow of expellees, region H s population would have grown by more than 7, people between 1939 and 195 (or by 5.4 percent). Region L s population, in contrast, would have decreased by.9 percent. There are two main reasons for this difference. First, the states of Lower Saxony and Schleswig-Holstein, both part of region H, border the Soviet Occupation Zone, and, therefore, also received an over-proportional share of migrants from what was to become the German Democratic Republic. Second, the number of civilian casualties was lower in rural than in urban areas, and many city dwellers, especially from Bremen and Hamburg, were evacuated to more rural areas during the war. 13

15 .66 Figure 2: Population in H over population in L, Actual Hypothetical Data source: Statistisches Bundesamt Note: Population is measured at the end of each year. The hypothetical series is calculated by adding to the actual population figure of the H- and L-region on 31 December 1949 (cumulated) net migration between the two regions. separate emigration rates for the expellee and non-expellee population, which are, however, only available until The figure shows that the net emigration rate was almost 2% in 195. Therefore, 2% of region H s population migrated to region L in 195. The emigration rate then slowly leveled off but stayed positive throughout the 195s. In the 196s, net emigration from region H to L was essentially zero (or even slightly negative). Figure 3 also shows that expellees were much more prone to leave region H for region L than the rest of the population. In fact, the emigration rate of expellees stood at a stunning 4.4% in 195 and was thus four times as high as the emigration rate of the non-expellee population (1.1%). It fell in tandem with the overall emigration rate, but stayed ahead of it until GDP per capita. Figure 4 shows GDP per capita of region H relative to region L between 195 and 197. The data shows that in 195, GDP per capita of region H reached just 75.4% of region L s level. The gap between the two regions then narrowed considerably in the 195s and early 196s, and relative GDP stood at 84.2% in The mid- and late 196s saw no further improvement in region H s relative GDP per capita. If anything, the gap to region L widened again. In 197, GDP per capita of region H reached 83.% of region L s level. How did GDP per capita change between 1939 and 195? Unfortunately, comparable GDP data from before the war does not exist for West German regions. To still come up with an 14

16 5.% Figure 3: Net emigration rate from region H to L, % 3.% 2.% 1.%.% 1.% Overall Expellees Non expellees Data sources: Data for 195 is from the Institut für Raumforschung, data for all other years from the Statistische Bundesamt. Notes: The emigration rate expresses net emigration from region H to region L in a given year in percent of regions H s population in the previous year. The emigration rate of expellees (non-expellees) expresses net emigration of expellees from region H to region L in percent of regions H s expellee (non-expellee) population in the previous year. estimate of the pre- to post-war change in relative GDP, we use two proxies. First, we resort to national income data from 1936, as also reported in Table 1. Our calculation suggests that in 1936, national income in region H was 92% the value of region L. Judged by this measure, the high-immigration region suffered a significant fall in relative GDP following the inflow of expellees (Figure 4 shows the 1936 relative income data along with GDP data for 195-7). One problem with the income data is that the provinces of the German Reich, for which the data is available, do not correspond to the later West German states. We therefore had to approximate the values for the H- and L-region (see footnote xxx for the details). Second, we use firm sales as a proxy for production as suggested by Vonyó (212). The data comes from published sales tax statistics (Statistisches Bundesamt 1955a, Statistisches Bundesamt 1955b). Total sales are defined as domestic deliveries and other services of a business for money and own consumption of the business. The sales data has two advantages over the national income data: First, we have comparable data on firm sales for both the pre- and the post-war period (although, unfortunately, only until 1955). Second, sales data is available at a county level and can thus be precisely aggregated to the federal state level (and thus also to the level of our stylized regions). At the downside, sales are not a direct measure of the production value, and certain exemptions 15

17 for businesses with low revenues apply. However, firm sales per capita correlates strongly with national income. 12 The pre- to post-war change in relative sales per capita between the two regions gives at least an indication of the change in GDP per capita. As sales statistics are not available for 1939, we use data for 1935, along with population figures for The data suggest that relative sales per capita in region H fell from 69.5% of region L s value in 1935 to 59.1% in 195, and then increased to 63.6% in Figure 4: GDP per capita in region H relative to L, Data sources: Data for 1936 is from Länderrat des Amerikanischen Besatzungsgebiets (1949), data for from the Statistische Bundesamt. Notes: The data point for 1936 is the (approximated) national income of region H relative to region L. We calculate the 1936 national income in region H as the population weighted average of income in the former German Reich provinces of Bavaria, Hannover, and Schleswig-Holstein. We then use the average national income in the rest of West Germany (excluding Hamburg and Bremen, Oldenburg, Braunschweig, Lippe, Schaumburg-Lippe) as a proxy for national income in region L. See the text for further explanation. All other data points give GDP per capita of region H relative to region L. All in all, the evidence suggest that relative GDP per capita decreased significantly in the high-immigration region between 1939 and 195, before bouncing back in the 195s (and early 196s). Since we do not have comparable data on GDP per capita from before the war, we cannot conclusively decide whether and how quickly relative GDP per capita returned to its pre-inflow level. 12 The correlation coefficient between sales per capita in 1935 and national income per capita in 1936 is.92 for the 19 regions of the German Reich, for which both type of data are available. And for 195, the correlation coefficient between sales per capita and GDP per capita of the nine West German states is As evident, the relative level of sales per capita is lower than the relative level of GDP per capita. This is mainly due to the fact that the share in sales per capita of the two city states, which are both part of region L, is much larger than their share in GDP per capita. 16

18 Unemployment and labor force participation. Figure 5 shows the unemployment rate in regions H and L in 1938, i.e., just before the war, and between 195 and 1963 (since timeconsistent regional employment data is not available thereafter, our data series ends in 1963). Right before the war, both regions recorded almost full employment. Unemployment increased dramatically in the immediate post-war period. Yet, the situation was much more severe in the high-immigration region, where the unemployment rate was at 16.7% in 195, than in the low-immigration region, where it was just 6.4%. Expellees were much more likely to be unemployed than natives: In 195, every third unemployed person in West Germany was an expellee (compared to a population share of expellees of 16.5%). 14 Unemployment then gradually came down during the 195s. Both regions were basically back at full employment at the beginning of the 196s. Nevertheless, unemployment remained slightly higher in region H than in region L until Expellees benefited over-proportionally from the fall in unemployment. Their share among the unemployed fell from 34.3% in 195 to 22.% in 1958 (the unemployment statistic does no longer distinguish between expellees and non-expellees thereafter). At the end of the 195s, therefore, the share of the expellees among the unemployed was only slightly above their population share (which was 18.3% in 1958). One problem of the official unemployment data is that it only includes persons officially registered as unemployed. People searching for a job without being registered with the employment agency (e.g. because they would not receive benefits anyway) are not counted as unemployed. In addition, persons who would like to work but have nevertheless withdrawn from the labor market (e.g. because they consider their chances of finding a job to be very small) are also not counted as unemployed. The group of persons who are not officially registered as unemployed but would like to work is likely to be larger when labor market prospects are dire. Therefore, hidden unemployment might have been quite high at the beginning of the 195s. As the labour market recovered, and the probability of finding a job grew, formerly discouraged workers might have more and more chosen to re-join the labor force. In fact, the labor force participation rate increased strongly from 66.9% in 195 to 7.6% in 1957 (see green line in Figure 5). 15 Summary of stylized facts. To sum up the above, our comparison of the demographic and economic development of the H- and L-regions has uncovered the following five stylized facts: 14 We have data on the number of unemployed expellees but not on the number of employed expellees. Therefore, we cannot calculate the unemployment rate of expellees and non-expellees separately. 15 There are no direct estimates of hidden unemployment in West Germany before the mid-196s. This is because detailed annual labor force statistics by sex and age became only available at the end of the 195s. 17

19 Figure 5: Unemployment rate in region H and L and labor force participation rate in West Germany, % 71% 16% 14% 12% 1% 8% 6% 4% 2% % Unemployment rate, region H Unemployment rate, region L Labor force participation rate, West Germany 7% 69% 68% 67% 66% 65% Data sources: The unemployment data come from Länderrat des Amerikanischen Besatzungsgebiets (1949) (for 1938) and from various issues of the Amtliche Nachrichten of the Bundesanstalt für Arbeitsvermittlung und Arbeitslosenversicherung (for ). Data on economically active persons, used to calculate the labor force participation rate, is taken from Sensch (24), Table B3.1. Data on the total population aged come from the Statistische Bundesamt. Notes: The unemployment rate is expressed as a percentage of the dependent labour force. The unemployment rate of region H in 1938 is approximated by the (labor force weighted) average of the 1938 unemployment rates of Bavaria, Lower Saxony, Nordmark, the unemployment rate of region L by the average of the unemployment rates of Hesse, Southwest Germany, Rhineland and Westphalia. The labor force participation rate is the ratio of all economically active persons to the population aged The relative population size of region H increased from.56 in 1939 to.65 at the end of It then gradually decreased in the 195s and stood at.536 in 196. The 196s saw no further change in relative population. 2. Migration from region H to L was the decisive factor in moving relative population back towards its pre-war level in the 195s. The net migration rate of expellees was much higher than that of natives. 3. Relative GDP per capita of region H decreased between 1939 and 195. Region H then experienced considerably faster economic growth than region L until the early 196s. It remains unclear whether relative GDP per capita returned to its pre-war level. 4. Unemployment rates jumped up between 1938 and 195 both in region H and L. However, the increase was much more pronounced in H. Expellees were much more likely to be unemployed than natives. Unemployment then gradually decreased during the 195s, and 18

20 both regions again recorded full employment at the beginning of the 196s. 5. The labour force participation rate in West Germany increased strongly between 195 and 1957, and remained constant thereafter. In what follows, we develop a dynamic partial equilibrium model to analyze whether these empirical facts can be understood as a consequence of the expellee inflows. 4 A dynamic model of regional labor markets The backbone of our model is the textbook Diamond-Mortensen-Pissarides (DMP) search and matching model of unemployment. 16 We extend this model along three directions that are motivated by the particular historical episode that we study. Our first extension is regional migration, and we model it by considering two regional labor markets one in region H and another one in region L that interact via migration. Workers that search for a job select themselves endogenously into one region. These migration decisions are forward looking and subject to migration costs. Our second extension endogenizes labor force participation because it moves sharply in historical data. We model participation as a decision of non-employed workers to participate or not in the labor market in the current period, which yields a parsimonious formulation of endogenous participation. Our third extension incorporates expellee inflow into the model. We model this inflow as exogenous increase in the number of non-employed workers, i.e. as labor supply shock, and calibrate the regional distribution of this shock to historical data. Therefore, this shock is large in region H and small in region L. The remaining bits and pieces of the model are standard. In either region, a large firm employs many workers and accumulates a capital stock to produce output, which is homogenous across regions and thus serves as numeraire. Adjusting employment or the capital stock is subject to adjustment costs. Furthermore, we consider a competitive equilibrium in which firms and workers bargain over the wage, since we are also interested in wage dynamics.our description of the model focuses on region H with the understanding that region L is formulated symmetrically to region H. In general, variables that pertain to region L are superscripted by a star and/or subscripted by L. 16 E.g. see Sections 1 and 3.1 in Pissarides 2 XX REF. 19