Diverse and Homogeneous Migrant Communities Upon Resettlement

WWII weakened both formal and informal institutions across Poland, but the situation in the formerly German territories was particularly dire. The German administration departed in 1945, and the emerging Polish institutions were “weak, facade-like structures unable to control the situation” (Grabowski Reference Grabowski2002, 148). The Citizens’ Police (Milicja Obywatelska, MO), tasked with maintaining law and order, had neither the manpower nor willingness to fulfill its duties. Initially, state authority was confined to cities and towns, and the newly established migrant communities were expected to fend for themselves.

In this environment, homogeneous migrant communities had an important advantage—shared norms and, in some cases, shared networks. Upon migration, such groups were able to quickly replicate the familiar patterns of associational behavior and reestablish order. One migrant describes his arrival to the village of Pyrzany in July 1945 as follows: “Houses in the very center of the village were occupied by the pastor, the organist, and others who deserved it. The poor settled in houses on the outskirts” (Halicka Reference Halicka, Halicka and Mykietów2011, 49). Although the village already had an elder (sołtys), a man from central Poland who had arrived earlier, the more numerous Galicja group elected their own representative. Migrants soon formed a volunteer fire brigade that doubled as a militia, founded a preschool, and reopened the German shops (Halicka Reference Halicka, Halicka and Mykietów2011). In this homogeneous community, shared norms and networks enabled the creation of private-order organizations for the provision of security and other public goods, such as the volunteer fire brigade, facilitating the resumption of private economic activity.

The localities settled by migrants from different regions faced greater challenges. Next to Pyrzany lies the village of Oksza, which was settled by diverse migrants from Central Poland and the USSR. A migrant described his first year in Oksza as follows: “A lot of destruction could have been avoided through better organization of resettlement, but there was no oversight of the process: Everyone was looting and pillaging, like a hungry wolf would do to his prey” (Solinska and Koniusz Reference Solinska and Koniusz1961). To this day, Oksza has no volunteer fire brigade or other communal organizations.

In the absence of shared norms and networks, heterogeneous communities were reluctant to invest in private-order organizations for the provision of public goods or norm enforcement. For many migrants, it was easier to opt out of collective activities altogether to pursue one’s self-interest, sometimes at the expense of others. Not surprisingly, crime rates were higher in heterogeneous areas for decades after the resettlement (see Appendix Section C.3).

Divergent Cooperation Mechanisms

Population transfers were followed by the gradual expansion of Communist rule. The state and its organizations soon played an outsized role in the formerly German territories, providing vital public goods and mediating intergroup conflicts. Stasieniuk (Reference Stasieniuk, Michalak, Sakson and Stasieniuk2011, 154) observes that in resettled communities, formal institutions “not only represented state authority, but also served as an intermediary between different societal groups. It is on the basis of these institutions that social contact was established.” Similarly, Sakson (Reference Sakson2011, 85–6) acknowledges that the new regime was “seen as an organizer of social life” in migrant communities; Kersten (Reference Kersten1991, 165) argues that displacement fostered “ties to new authorities.” Data on Communist Party membership support these conclusions (see Appendix C.1). Of course, not all state–society interactions were voluntary: Mass migration also weakened the resistance to communist authorities and their socioeconomic interventions, as evidenced, for example, by the greater number of collective farms and the higher infiltration of the Catholic Church by the so-called Red Priests (Nalepa and Pop-Eleches Reference Nalepa and Pop-Eleches2018).

I expect that state–society relations varied not only between the resettled and non-resettled areas, but also across migrant communities at different levels of cultural heterogeneity. In homogeneous communities, state institutions were more likely to compete with informal social and economic structures; as a result, their reach was limited. In heterogeneous communities, resistance to the new order—and the organizations that came with it—was weaker. Although the communist institutions encroached on private property and stifled political competition, the state filled an important need for the provision of public goods that heterogeneous communities struggled to provide on their own. Thus, I expect the demand for and the resulting reach of state institutions to increase with heterogeneity.

The shift in approaches to elderly care is illustrative. Jasiewicz (Reference Jasiewicz1972) writes that the traditional custom of families taking care of their elderly failed in diverse communities because social sanctions for noncompliance were ineffective. He finds that the elderly instead increasingly turned to the state, which guaranteed healthcare and a pension in exchange for getting titles to their land. The state was even more important for enforcing cooperation in agricultural work, a role resented in homogeneous villages but appreciated in diverse settings. In the heterogeneous village of Pracze in Lower Silesia, cultural differences restricted the mutual provision of neighborly help to the families and migrants from the same village. The scope of neighborly assistance widened, however, once the state instituted mandatory in-kind contributions to the social welfare fund. State institutions also began to commission and remunerate assistance to the sick, widowed, and elderly, historically provided through voluntary cooperation (Pawłowska Reference Pawłowska1968, 95). State enforcement was also used to maintain the commons and curb free-riding. In Pracze, the devastation of communal fields and orchards or the damage to communal equipment was increasingly addressed in court rather than through the village head or authority figures (Pawłowska Reference Pawłowska1968, 198). By contrast, the residents of homogeneous villages did not need state enforcement and continued to rely on informal enforcement strategies. In the village of Pyrzany, for example, the residents regretted not being “fully free in their activities […and having to] reckon with the new [Communist] political order and to negotiate with regional and county authorities” (Halicka Reference Halicka, Halicka and Mykietów2011, 53).

Institutions, State Capacity, and Economic Outcomes

Poland’s post-WWII history allows a test, however imperfect, of the relationship between heterogeneity, state capacity, and private economic activity in different institutional settings.

In 1950, Poland introduced central planning, nationalized industry, and attempted to collectivize agriculture. Although private farming remained dominant, large estates in the resettled regions were transformed into state farms. The focus on heavy industry undermined the service sector, and many small workshops and businesses were nationalized, taxed into bankruptcy, or forced underground. The state was most successful in quashing private economic activity in the formerly German territories, where its reach was deeper (see Appendix C.2).

Restrictions on private entrepreneurship were partially reversed in 1957, following worker protests in Poznań. Gradually, the “socialist” economic model was altered through the introduction of market elements. By 1980, 602,000 legal private businesses were officially registered, and many more operated informally (Kochanowski Reference Kochanowski2010, 185).

Even so, the reliance on state enforcement carried few advantages because the legal protections afforded to private property remained inadequate. Informal norms and networks, on the other hand, facilitated access to scarce goods and increased opportunities for economic exchange. Even the officially registered private businesses had to rely on informal connections to obtain tools and supplies (Kochanowski Reference Kochanowski2010, 185). Because shared culture facilitates informal coordination and enforcement, I expect homogeneous communities to have been more likely to sustain private economic activity under state socialism.

In 1988, the Polish government removed all remaining barriers to private entrepreneurship. “Shock therapy” economic reform was launched in 1989, expressly to reduce state influence over the economy. Between 1989 and 1994, the share of the private sector in total employment increased from 12% to 61%. Most private sector firms were small and established from scratch rather than by privatization.

Importantly, economic and political reform did not generate a void in formal enforcement. While the “party” element of the party-state disappeared, most other state organizations were subsequently strengthened. This is particularly evident with respect to revenue collection. Under state socialism, taxation was largely indirect, and the state extracted large revenues from industrial enterprises it owned or controlled. During the transition to a market economy, the state was forced to contend with outstanding debt and rising budget deficits, and also to offset declining revenues from state enterprises. In 1992, direct taxes on personal and corporate income were introduced. Tax collection depended largely on taxpayers’ compliance, secured in part through the continuing provision of public goods and services (Easter Reference Easter2002). Maintaining sufficient tax revenue, in turn, enabled the state to expand its administrative capacity and strengthen the institutions for the protection of private property and enforcement of contracts.

A key reform that strengthened state capacity at the local level was the March 1990 Law on Local Self-Government, which expanded the authority of over 2,400 municipalities (gminy). From then on, local bureaucrats were accountable exclusively to elected local councils and paid out of the municipal budgets. Municipalities were now authorized to collect and spend their own fiscal revenues from immovable property, agricultural income, and vehicles, within caps set at the national level.^{Footnote 3} Municipalities became responsible for financing the maintenance and improvement of local infrastructure, primary education, healthcare, and social assistance. Some negotiated numerous tax exemptions and special deals, yielding to local interests, while others set high tax rates and increased spending on key public goods, supporting private economic activity (Swianiewicz Reference Swianiewicz1996).

The increased fiscal autonomy and responsibilities of local governments widened subnational differences in the degree and quality of governance. The ability to pay bureaucrats affected the privatization of communal property, the distribution of social benefits, and the process of licensing for new businesses and regulating private economic activity more broadly. Localities with the highest levels of fiscal and bureaucratic capacity also promoted entrepreneurship by funding business incubator programs, guaranteeing start-up loans, and offering training programs for the unemployed (Misiag Reference Misiag2000). For example, in the heterogeneous town of Bartoszyce, the municipal government created the Department for the Promotion and Social Affairs, which centralized information about all registered businesses and offered training to the unemployed (Gorzelak and Jalowiecki Reference Gorzelak and Jalowiecki1996). In Dzierzgoń, settled by migrants from Kresy and Central Poland, the government established the Society for the Development of Dzierzgoń to offer legal and economic advice to those interested in starting new firms, as well as the Mutual Guarantee Fund to help small and medium enterprises overcome capital barriers (Gorzelak and Jałowiecki Reference Gorzelak and Jałowiecki1999).

I expect the transformation of national institutions and the decentralization reform to have different implications in homogeneous and heterogeneous communities. Heterogeneous communities, where informal institutions were less effective, would benefit from the improvement in the quality of formal institutions to a greater extent than homogeneous communities. Economic actors in heterogeneous communities could now take advantage of the increased protection of property rights and contract enforcement to engage in market transactions and register new firms. Greater demand for government-provided public goods in heterogeneous communities would also increase their investment in fiscal capacity and thus their ability to finance public goods necessary to support private economic activity. Both channels would contribute to higher private entrepreneurship and incomes in more heterogeneous communities under a market economy.

Historical Data on Population Transfers

The data on the origin of population in 1,217 historic municipalities (gminy) of the former German territories come from a census commissioned by the Ministry for the Recovered Territories in December 1948, after the population transfers were largely completed. The data were never published and are contained in the Polish Archiwum Akt Nowych.

The census recorded the size of four distinct population groups: repatriates from the USSR, settlers from Central Poland, reemigrants from Western Europe, and the autochthonous population (see additional information about these groups in Appendix A.1). Polish sociologists generally agree that these are the main categories of residents in the former German territories after WWII (Chmielewska Reference Chmielewska1965). However, this categorization does not encompass all cultural cleavages and thus understates the diversity of the population in this region, which biases against the hypotheses outlined above. The autochthonous population included Protestant Mazurians and Catholic Warmiaks in the north and Catholic Silesians in the southwest, all of whom spoke different dialects and sometimes not a word of Polish. Reemigrants from Western Europe were a diverse group originating in France, Germany, Yugoslavia, and other states. The repatriates from rural Galicia (Ukrainian SSR) included both Catholic and Greek Orthodox populations and had few cultural similarities with repatriates from the more urbanized Lithuanian SSR. The settlers from Central Poland arrived from all three former imperial partitions and were also an internally heterogeneous group.

However, in a given municipality usually only one of these subgroups would be present because the population movement occurred from east to west and followed three major railway routes (see Figure A.2 in the Appendix). As a result, at the micro level migrants in each of the three categories typically came from the same prewar province.

Measuring Diversity

I decompose diversity into (1) the heterogeneity of the migrant population and (2) the share of migrants, following Alesina, Harnoss, and Rapoport (Reference Alesina, Harnoss and Rapoport2016). The first component, Div_mig, measures the diversity of migrant groups in each municipality. If s _j is the share of migrants from region j from the total population of migrants, with j = 1,…, J, then migrant heterogeneity can be expressed as ${\rm{Di}}{{\rm{v}}_{mig}} = \sum\limits_{j = 1}^J {\left[ {{s_j} \times \left( {1 - {s_j}} \right)} \right]}$. This expression does not depend on the share of the indigenous population.

The second component, Div_resettled, measures the share of migrants in the total population. In the Polish context, it serves as a control variable, capturing the effects of displacement and separating migrants from the natives. The latter were concentrated in specific regions, decided against emigration to Germany in the 1940s for economic and political reasons, and could obtain German passports and work abroad in the 1990s, before Poland joined the EU. Div_mig was a product of arbitrary assignment of migrants during population transfers, whereas Div_resettled can be explained by the historic patterns of nation-building in Germany and Poland.^{Footnote 4}

Figure 3 shows that migrant diversity varied considerably across the formerly German municipalities while the indigenous population was heavily concentrated in the south and northeast. In 82% of the local communities, the share of migrants exceeded 80%.

FIGURE 3.

Migrant Diversity and the Autochthonous Population at the Municipal Level in 1948

Unit of Analysis

I test my theoretical predictions using historic and contemporary data at the level of municipalities. This is the smallest unit for which both historical and contemporary data are available. Municipalities are small enough to ensure that migrants from one region will interact with migrants from a different region at high values of Div_mig. Furthermore, contemporary municipalities are self-contained social units with legislative and governing bodies and are thus appropriate for studying the effects of diversity on social and economic outcomes. One challenge with adopting this unit, however, is that municipal boundaries have changed considerably since 1939. To account for these changes, I aggregated historical communal (1939) and municipal (1948) data to the level of 630 contemporary municipalities, which are larger in size than historical units (see details in Appendix B.3). If contemporary municipality borders split historical municipalities, I weighted the historical data by the proportion of the overlapping area.^{Footnote 5}

Empirical Strategy

As noted above, officials tasked with resettlement were unfamiliar with the new territory and assigned migrants to locations in an arbitrary manner. I draw on Polish and German data to examine potential determinants of the resulting local heterogeneity. The intuition is that assignment was as-if random, conditional on covariates discussed below, and thus the estimated coefficient on Migrant Diversity can be interpreted causally.

I estimate the following model: ${y_{i}} = \alpha + \beta \,\times {\rm{Di}}{{\rm{v}}_{mi{g_{i}}}} + \gamma \,\times\, {\rm{Di}}{{\rm{v}}_{resettle{d_{i}}}} + \theta \,\times\, {X_{i}} + {D_j} + {\varepsilon _{i}}$, where y _i is the outcome in municipality i; Div_mig and Div_resettled are the two components of diversity discussed above, X _i is a set of municipality-level covariates, and D _j is a vector of district-level fixed effects, and ε _i are errors. When spatial dependence is present, Moran eigenvectors m _i are included as covariates to filter out residual autocorrelation (Thayn and Simanis Reference Thayn and Simanis2013).^{Footnote 6}

I supplement this specification with regressions that adjust for weights obtained from nonparametric covariate balancing generalized propensity score (npCBGPS) analysis. This approach optimizes covariate balance between the treatment and the control units, reduces model-dependence, and accommodates continuous treatments (Fong, Hazlett, and Imai Reference Fong, Hazlett and Imai2018). I also explore sensitivity of the results to the unobserved confounding.

To establish whether reliance on informal enforcement strategies and resulting differences in state capacity are indeed the mechanisms behind the effects of migrant diversity in 1948 and contemporary economic outcomes, I use sequential g-estimation analysis (Acharya, Blackwell, and Sen Reference Acharya, Blackwell and Sen2016).

Dependent Variables

The theory has a number of observable implications that are summarized in Table 1 and discussed in more detail below.

TABLE 1.

Theoretical Concepts, Measurement, and Hypothesized Effects of Migrant Diversity

Covariates

Estimating the effects of Migrant Diversity on economic activity is challenging because it is possible that heterogeneous migrants settled in economically more developed areas. There were some instances of skilled migrants from far-flung regions assigned to more industrial localities to facilitate production. Thus, I use the proportion of population employed in industry (Share in Industry) collected from the 1939 German census as a proxy for industrial potential.^{Footnote 11} The variable explains 12% of the variance in Migrant Diversity (see Appendix B.2).

Localities near the railway were easier to reach for migrants from any point of origin. Proximity to the railway also facilitated state control and economic development, biasing in favor of the main hypotheses. Thus, I account for distance to the nearest railway station, measured in 1948, which explains 9% of the variance in Migrant Diversity.

Distance to post-1945 international borders is another possible confounder. For security reasons, military families were encouraged to settle in the border regions, while migrants of certain origin (such as Lemkos) were prohibited from settling near the borders. Distance to the border also had divergent economic implications before and after 1989. Thus, I calculated the distance to the international border, in kilometers (km), from the centroid of each municipality, which is negatively correlated with Migrant Diversity and explains 11% of the variance.

In rural areas, the resettlement process was influenced by the presence of large German estates, transformed into state-owned farms during the early 1950s.^{Footnote 12} State farms attracted primarily landless migrants from Central Poland and, in addition to employment, provided important social services during the communist period. After 1989, the closure of state farms increased unemployment rates and generated a host of social ills. However, the location of large German estates, proxied by Share Farms over 100 ha, is only weakly correlated with Migrant Diversity.

I measure access to state public goods provision as distance (in km) between the center of each municipality and the county seat, which contained state police and fire brigade, hospitals, and courts (Distance to County Seat).^{Footnote 13} Other covariates are population size, as informal enforcement weakens with the size of a community, and the share of the urban population in 1948, when the population transfers were completed. I also collected data on the gender and age of migrants, which do not vary with Migrant Diversity.

Sources and descriptive statistics for all historic and contemporary variables used in the analyses are presented in the Online Appendix (Sections B.1 and B.2). As shown in the Appendix, most correlations between Migrant Diversity and socioeconomic covariates are near zero, which speaks to the arbitrariness of migrant assignment. Nevertheless, I include all covariates that predict some variation in Migrant Diversity or bias in favor of my hypotheses. Regression models also include fixed effects for pre-WWII German districts (Regierungsbezirke), which varied in infrastructure and industrial potential.

Enforcement Strategies and Fiscal Capacity

My theory predicts a negative relationship between heterogeneity and the prevalence of volunteer fire brigades, which provide a local public good through bottom-up collective action, and a positive relationship between heterogeneity and the prevalence of municipal guards, a contemporary organization for the provision of security through taxation. I test these predictions in Table 2.

TABLE 2.

Migrant Diversity, Organizations for Public Goods Provision, and Fiscal Capacity

+ p < 0.1; * p < 0.05; ** p < 0.01.

Model 1 in Table 2 regresses Volunteer Fire Brigades (per 10,000 people) on Migrant Diversity and covariates. The coefficient on Migrant Diversity is negative and significant: Diverse migrant communities have fewer OSPs than homogeneous migrant communities. A change in heterogeneity from zero (homogeneous) to 0.66 (heterogeneous, comparable to a community with three equally sized culturally distinct groups) predicts a decrease in the number volunteer fire brigades by 2.8 per 10,000 people, equivalent to approximately two thirds of a standard deviation, when comparing communities within German districts. The coefficient on the Share of Migrants is negative but does not reach statistical significance.

Model 2 in Table 2 regresses the presence of a Municipal Guard on Migrant Diversity and covariates. The coefficient on Migrant Diversity is positive and statistically significant: Predicted probability of establishing a municipal guard in most heterogeneous migrant communities (Migrant Diversity = 0.66) is approximately 10% greater than in homogeneous communities (Migrant Diversity = 0), as shown in Figure 4. The estimate is conditional on the proximity to the state police and other covariates and captures variation only within formerly German districts. The coefficient on the Share of Migrants is also positive and statistically significant. The predicted probability of creating a municipal guard increases from 3% to 8% as the share of migrants rises from 50% to 100%, which is consistent with the greater need for state enforcement following the disruption of social ties.

FIGURE 4.

Predicted Probability of Establishing a Municipal Guard at Different Levels of Migrant Diversity (Left) and Density of Volunteer Fire Brigades (Right). Based on Models 2 and 3 from Table 2

Model 3 examines whether public goods provision through informal enforcement predicts lower demand for public goods provided through taxation by regressing Municipal Guard presence on the density of volunteer fire brigades. The coefficient on Volunteer Fire Brigades is negative and statistically significant. The model suggests the predicted probability of establishing a municipal guard is 3.4% lower, on average, in communities with the average prevalence of volunteer fire brigades (4.17 per 10,000 people) than in communities without any (see Figure 4).

Did greater demand for state-provided public goods affect tax policy and revenues in historically homogeneous and heterogeneous communities? Models 4–7 in Table 2 explore the divergence of fiscal policies in 1993–95, the period when municipalities began determining their own revenues and expenditures. First, I regress property tax rates and revenues on Migrant Diversity and covariates. In Model 4, with Property Tax Rate as a dependent variable, the coefficient on Migrant Diversity is positive but does not reach statistical significance; the coefficient on Share Migrants is negative and significant at the 10% level.^{Footnote 14} Thus, contrary to expectations, tax rates are similar in communities at different levels of Migrant Diversity. However, in Model 6, with ln(Property Tax Revenue) as a dependent variable, the coefficient on Migrant Diversity is positive and significant at the 5% level. This model suggests that predicted revenues from property tax in the most heterogeneous communities exceed those in the most homogeneous communities by 85 Złoty per capita, on average, an effect that is equivalent to a quarter of the standard deviation in the dependent variable. By contrast, the coefficient on Share Migrants does not reach statistical significance in these models and changes signs, which suggests that the observed differences in fiscal capacity are due to heterogeneity rather than to the uprooting of population.

Models 5 and 7 examine intermediate relationships between voluntary provision of public goods, on the one hand, and investment in fiscal capacity, on the other hand. Regression analysis suggests that both property tax rates and revenues were lower in localities with greater voluntary provision of public goods, proxied by the density of volunteer fire brigades. Thus, culturally homogeneous communities that are effective at informal cooperation appear to have lower demand for state-provided public goods and are less likely to invest in fiscal capacity.

I replicate the results from the main linear models in Table 2 using OLS with weights from nonparametric CBGPS^{Footnote 15} and demonstrate their robustness to unobserved confounding (see Appendices D.2 and D.3).

Higher prevalence of volunteer fire brigades in historically more homogeneous communities mirrors the conclusions in the literature on the importance of homogeneity for voluntary provision of public goods (e.g., Habyarimana et al. Reference Habyarimana, Humphreys, Posner and Weinstein2009; Miguel and Gugerty Reference Miguel and Gugerty2005). At the same time, the positive effects of Migrant Diversity on the creation of the Municipal Guards and on Property Tax Revenue raise questions about the broader economic implications of lower levels of voluntary cooperation in heterogeneous settings. In particular, the results suggest that homogeneous communities that are effective at collective action may have lower demand for state-provided public goods and thus lower incentives to invest in state capacity, an important determinant of long-run economic growth.

Economic Outcomes before 1989

My theory predicts that greater reach of the state in heterogeneous communities carried few economic advantages and may have even hampered private entrepreneurship during state socialism. By contrast, informal norms and networks, historically stronger in homogeneous communities, created additional opportunities for economic exchange and provided access to scarce resources. Regressions in Table 3 explore the observable implications of the theory using economic outcomes measured in the early 1980s, when private entrepreneurship was legal, but market-supporting formal institutions remained inadequate.

TABLE 3.

Migrant Diversity and Economic Outcomes (per 1,000 People) at the Municipality Level in 1980–82. OLS Regression

+ p < 0.1; * p < 0.05; ** p < 0.01.

The coefficient on Migrant Diversity is not statistically significant in the regression models that focus on employment in the Socialized Economy or Private Handicrafts when all covariates are included (Models 1 and 2). Migrant Diversity also does not predict the prevalence of TVs per 1,000 people after conditioning on covariates (Model 5). The coefficient on Migrant Diversity is negative and statistically significant at the 5% level in Models 3 and 4, which regress Shops and Phones on diversity and historic covariates, however. These two models imply that homogeneous (Migrant Diversity = 0) communities have an additional shop and 4.5 more phones per 1,000 people than most heterogeneous communities (Migrant Diversity = 0.66), which is slightly below half a standard deviation for shops and one-fifth of a standard deviation for phones. The coefficient on Share Migrants changes signs depending on the model. Models 1 and 2 suggest lower levels of private economic activity and higher levels of state employment in migrant communities, relative to the indigenous communities in the formerly German territories. Models 4 and 5 imply greater wealth in communities with higher shares of migrant population.

I examine the robustness of the coefficients on the main explanatory variable, Migrant Diversity, to using the npCBGPS methodology in Appendix D.2. The magnitude and statistical significance of the coefficient on Shops increases, while the coefficient on Phones is attenuated and loses statistical significance. The relationship between Migrant Diversity and Shops and Phones is relatively robust to the unobserved confounding, as discussed in Appendix D.3.

In sum, regression analysis shows that homogeneous communities had slightly more shops and phones per capita than diverse communities in the 1980s, on average. At the same time, homogeneity was not associated with the prevalence of TV sets or with employment in the private and socialized sectors. The interpretation of the economic consequences of Migrant Diversity clearly depends on measurement choices. Unfortunately, data do not allow for studying the extent of the shadow economy across communities at different levels of diversity. It is safe to conclude from the analysis above, however, that heterogeneity produced no economic advantages during the communist period and that, if anything, homogeneous communities were slightly better off economically than heterogeneous communities.

Economic Outcomes after 1989

I further argue that after 1989, greater reach of the state in historically heterogeneous communities contributed to higher entrepreneurship and higher incomes. Figure 5 presents the relationship between migrant heterogeneity and gross income per capita as well as gross domestic product per capita, measured in 1995, at the province level. The N = 14 is small, but there is a clear positive relationship between these economic indicators and migrant diversity.

FIGURE 5.

Gross Income and GDP in Provinces Affected by Population Transfers

I explore the relationship between heterogeneity and economic outcomes at the municipal level using data on entrepreneurship rates and personal income tax per capita in Table 4. As noted earlier, tax rates on personal incomes do not vary across municipalities, so the latter measure closely tracks municipal differences in individual earnings. All regressions include a full set of covariates, district fixed effects, and control for spatial autocorrelation where it is present.^{Footnote 16} Model 1 indicates that Migrant Diversity does not predict Personal Income Tax in 1993, the earliest year for which data exist and a year after the tax was introduced. Income differences across resettled communities gradually widened over time, however. The coefficient on Migrant Diversity reaches statistical significance in 1995 and increases in magnitude in subsequent years (Models 2–4). The results for 1995 imply that homogeneous communities collected 13 Zł. less in personal income tax, on average, than most heterogeneous communities, equivalent to more than half a standard deviation in income tax during this period. This is a conservative estimate within regions that accounts for pretreatment covariates and spatial autocorrelation. In 2000, the predicted differences in personal income tax collected from homogeneous and most heterogeneous communities reach 21 Zł, or 40% of a standard deviation in income tax for that year (Model 4). Thus, the differences between homogeneous and heterogeneous communities widen as overall income inequality rises. By contrast, the coefficient on Share Migrants is negative in three out of four models and reaches statistical significance only in Models 2 and 3, for 1995 and 1998 outcomes. This may be because the indigenous population could legally work in Germany during this period, enabling higher earnings.

TABLE 4.

Diversity and Personal Income Tax per Capita (1–4) and Private Entrepreneurship (5–7). OLS Regression

+ p < 0.1; * p < 0.05; ** p < 0.01.

Models 5–7 show that Migrant Diversity also predicts higher rates of private entrepreneurship. Substantively, Model 5 implies that within a formerly German administrative district and accounting for all pretreatment covariates, an increase in historical levels of heterogeneity from zero (homogeneous) to 0.66 (most heterogeneous) predicts an increase in private enterprises by five per 1,000 people, equivalent to nearly a quarter of a standard deviation in the outcome. These differences remain statistically significant and substantively meaningful, but decrease slightly over time. By 2000, the predicted differences between most homogeneous and most heterogeneous communities increase to eight enterprises per 1,000 people, on average. Table A.10 in the Appendix shows that the coefficient on Migrant Diversity nearly doubles in size when excluding cities, which have both high levels of cultural diversity and private entrepreneurship. One possible interpretation is greater geographic mobility of urban population. Another is that cultural heterogeneity matters more in rural areas, where interactions with outgroups and the state are otherwise less frequent. There is no statistically significant relationship between Share Migrants and entrepreneurship rates.

Table A.11 in the Appendix explores the relationships between voluntary provision of public goods, taxation, and economic outcomes. All relationships are in the expected direction and statistically significant: The prevalence of volunteer fire brigades and tax rates and revenues in the early 1990s are strong predictors of incomes and entrepreneurship rates in subsequent years.^{Footnote 17}

In Appendix D.2, I show that the magnitude of the coefficient on Migrant Diversity increases when the data are weighted to achieve full covariate balance at different levels of treatment. Sensitivity analysis in Appendix D.3 also demonstrates that findings are robust to unobserved confounding within plausible sensitivity parameters.

Persistent Cultural Differences

Migrants often bring attitudes toward formal and informal institutions as well as human capital and entrepreneurial initiative with them. In the Polish case, migrants from different regions differed in their institutional heritage. Relatedly, migration experiences varied: Some were forced to relocate while others selected into migration. These factors may have influenced migrants’ attitudes toward the state or their economic behavior. For example, Becker et al. (Reference Becker, Grosfeld, Grosjean, Voigtländer and Zhuravskaya2018), who also examine the effects of post-war population transfers in Poland, show that although migrants from different regions did not differ in their education levels before WWII, experiencing forced displacement incentivized the descendants of migrants from the USSR to increase their investment in human capital, a mobile asset.

Could the presence of a specific cultural group, rather than the diversity of the migrant population, explain the differences in the local provision of public goods or economic outcomes at the municipal level? The prevalence of each group varied by region, with more migrants from Central Poland settling in the north and more migrants from the USSR settling in the south. Thus, including fixed effects for German districts helps address the concern that the main findings are driven by cultural differences rather than the extent of cultural heterogeneity. As an additional test, I regress the main social and economic outcomes on the shares of migrants from each region. Analysis in Appendix F.1 shows that while cultural differences do matter for some social and economic outcomes, they cannot provide a consistent explanation for all of the findings in the article. Furthermore, the main results hold when controlling for the origins of the dominant group in each municipality.

Human Capital and Skills

Could differences in human capital or perhaps the complementarity of skills resulting from divergent migration experiences explain the economic success of diverse resettled communities following the transition to a market economy?

While detailed occupational data are lacking, the 1978, 1988, and 2002 national censuses contain data on education at the municipal level. If the differences in education among diverse and homogeneous migrant communities existed in 1948, they should still be visible in the 1978 census and may either fade away or increase in subsequent periods. Regressions in Appendix F.2 show no statistically significant differences in education across different levels of heterogeneity in 1978 or 1988, but the coefficient on Migrant Diversity is a positive and significant predictor of education levels in 2002. This suggests that higher state capacity and/or greater economic prosperity in historically heterogeneous communities increased human capital, but not the reverse.

Research on positive economic effects of heterogeneity also emphasizes the variety of ideas and abilities that drive productivity and innovation (e.g., Alesina, Harnoss, and Rapoport Reference Alesina, Harnoss and Rapoport2016; Bove and Elia Reference Bove and Elia2017; Ortega and Peri Reference Ortega and Peri2014). These studies measure the diversity of skills as heterogeneity of migrants’ origins, a measure equivalent to Migrant Diversity. They also assume that generational turnover weakens returns to diversity. For example, Alesina, Harnoss, and Rapoport (Reference Alesina, Harnoss and Rapoport2016, 102) state that unlike “people of different ethnic or genetic origins who were born, raised and educated in the same country,” “people born in different countries are likely to have been educated in different school systems, learned different skills, and developed different cognitive abilities.” Accordingly, they find that birthplace diversity in 1990—but not in 1960—affects output in 2000. Alesina, Harnoss, and Rapoport (Reference Alesina, Harnoss and Rapoport2016, 120) interpret this as evidence that diversity increases productivity “primarily through first-generation effects.” By contrast, this article finds that diversity of migrants in 1948 predicts economic outcomes in the 1990s, when the labor force in western and northern Poland consisted largely of second- and third-generation migrants, born in Poland and subjected to homogenizing communist schooling. While the immigrants’ children inherit their skills and abilities, generational change should produce convergence in productivity across homogeneous and heterogeneous communities over time. Nevertheless, the residual heterogeneity of skills in historically heterogeneous localities may still explain their superior economic performance under a market economy, which enabled gains from specialization and trade, a possibility also suggested by the results from sequential g-estimation for the personal income tax.

State Policies

While it is generally agreed that the communist government ruled in a centralized manner and neglected local needs, it is possible that greater resources flowed to more heterogeneous migrant communities with the aim of facilitating integration and reducing social tensions. Regressions in Appendix F.4 show that heterogeneity does not predict higher prevalence of public schools and libraries, employment in the state sector, or the size of municipal budgets and compensatory subsidies from the central government during the communist period.

Sorting

It is possible that migrants from the initially more heterogeneous communities gradually sorted into more homogeneous communities, which were better at voluntary public goods provision. This dynamic could explain the post-1989 reversal of fortunes. Analysis in Appendix F.5 rejects this explanation. Regression results indicate that more people are moving into heterogeneous communities than into homogeneous communities. Thus, rather than becoming more homogeneous over time, such communities might be becoming even more diverse, a pattern also consistent with the greater openness to outsiders due to weaker informal networks.