Goodbye Lenin, Hello Murat? 1 The Effect of Communism on Individual Attitudes Toward Immigration

Transcription

1 Goodbye Lenin, Hello Murat? 1 The Effect of Communism on Individual Attitudes Toward Immigration Matthew Carl May 12, 2017 Abstract This paper argues conceptually and demonstrates empirically that individual attitudes toward immigration are deeply affected by a country s politico-economic legacy. Drawing on data from the German Socio-Economic Panel and exploiting a quasi-natural experiment arising from the historic division of Germany into East and West, I show that former East Germans are, because of their exposure to communism, notably more likely to be very concerned about immigration than former West Germans that were never exposed to the communist regime. The effect of living in East Germany is driven by those former East Germans that were actually born during, and not prior to the imposition of, the communist rule. I further find that differences in attitudes persist over time following reunification and that the level of trust in strangers and contact with foreigners represent two salient channels through which communism affects individual preferences toward immigration. Keywords: immigration, individual attitudes, communism, Germany. JEL Classification Codes: F22, Z10, J18, P30. 1 Inspired by the popular film Good Bye Lenin (2003). Murat is a common Turkish name. I would like to thank first and foremost Professor Peter Grajzl for his mentorship and unconditional support of this project over the past year. I would also like to acknowledge Professors Joseph Guse, Christopher Handy, Art Goldsmith, and Neils-Hugo Blunch for their incredibly helpful comments and advice, all of which have inexorably woven their way into the pages of this paper. Finally, I am grateful to participants of the Carroll Round conference at Georgetown University and the Virginia Association for Economists Annual Meeting for helpful feedback. Without the assistance of these individuals, this paper would not have been possible; I am profoundly grateful to each of you. Undergraduate student of Economics, German, and Mathematics at Washington and Lee University, Lexington, VA carlm17@mail.wlu.edu.

2 1. Introduction A consequence of globalization and political upheaval, recent years have witnessed the highest volume of individual movement across national borders since World War II. Concomitantly, public discourse on the consequences of immigration has spurred a vigorous debate on the economic and social effects of immigration on the domestic landscape of receiving societies. Since policy choices in democracies reflect individuals preferences (see, e.g., Mayda 2006), a comprehensive understanding of individual attitudes toward immigration is key to understanding policy responses to immigration. Existing literature on the determinants of attitudes toward immigration points to a combination of economic and non-economic individual-level characteristics as determinants of variation in attitudes within and across countries (see, e.g., Ceobanu and Escandell 2010; Mayda 2006; and O Rourke and Sinnott 2005). First, empirical social scientific research suggests that traditional trade and labor market theories concerning the skill composition of natives in the receiving society relative to immigrants efficaciously predict attitudes. A second strand of literature suggests that important determinants of individual attitudes toward immigration are noneconomic, and that racist and xenophobic attitudes and cultural preferences are most salient in explaining in attitudes toward immigration. In this paper, I shed novel light on ongoing discourse about immigration by investigating the effect of politico-economic regimes, and, in particular communism, on individual attitudes toward immigration. Drawing on the socialization theory 2 (see, e.g., Campbell et al. 1960; Greenstein 1965; Langton and Jennings 1968; and Jennings and Markus 1984), I hypothesize that exposure to communism, via the regime s ubiquitous emphasis on external enemies and prolonged indoctrination of the population, exacerbates concerns about immigration. My argument therefore extends the argument by Alesina and Fuchs-Schündeln (2007), who demonstrate that communism critically shaped East Germans beliefs about the appropriate role of the State. 3 However, whether communism also affected individual values and beliefs concerning immigration has thus far not been explored in the literature. To test my hypothesis in the data, I exploit a quasi-natural experiment following the 2 Note, this theory was developed in political science to explain the effect of communism on individual political attitudes, which I adopt here for the conceptual framework concerning individual attitudes toward immigration. 3 That is, due to living under communism, empirical evidence demonstrates that individual preferences are more likely to fall in line with the ideology promulgated under communism rather than shift away from communist ideology. 1

3 conclusion of World War II, when Germany was divided into two parts: East Germany was occupied by Soviet forces, which transformed the politico-economic institutional structure of the nation to communism; West Germany was occupied by Western Allied forces, and adopted a capitalist market system and democratic rule. Because the geographic border demarcating East and West Germany was drawn unrelated to individual preferences about immigration, it follows that Germans exposed to the politico-economic regime of West Germany constitute a meaningful control group for the East German treatment group exposed to communism. 4 Therefore, differences in individual attitudes toward immigration between East and West Germans following reunification can be attributed to East Germans prolonged exposure to the communist regime. Utilizing the German Socio-Economic Panel (G-SOEP), I first demonstrate that having lived under communism increases the probability of being very concerned about immigration by 3.4 percentage points on average, a result that obtains even after controlling for a vast array of economic and non-economic individual-level characteristics. Second, I examine the role that education under communism played in determining former East German attitudes toward immigration. In line with the socialization theory, I find that an additional year of schooling in East Germany, all else equal, translates into an increase in the probability of being very concerned about immigration; however, this result is statistically insignificant. Third, mirroring the analysis in Alesina and Fuchs-Schündeln (2007), I examine the convergence of attitudes toward immigration among former East and West Germans. The results suggest that, over time, former East and West German attitudes do not converge. Fourth, I conduct a series of exercises that further probe at the effect of communism on individual attitudes toward immigration. I find that having lived an additional year under communism, all else equal, translates into an increase in the prospects of being very concerned about immigration. I further find that the effect of communism is strongest for individuals born during the regime, and is statistically significantly less-pronounced for generations which were born prior to the institution of communism in East Germany (i.e., prior to 1945). Moreover, I find that the effect of communism on individual attitudes was homogenous across five of six East German states. 4 See Alesina and Fuchs-Schündeln (2007), Rainer and Siedler (2009), and Pop-Eleches and Tucker (2010), which discuss the literature of German history and its role in the exogenous character of the natural experiment of East and West Germany, utilizing this to address the causal effect of communism on individual attitudes and behavior. 2

4 Finally, I investigate the relevance of several potential channel variables that could explain the difference in concerns about immigration between former East and West Germans. The results indicate that differences in trust toward strangers and contact with foreigners are two quantitatively important pathways through which exposure to communism in East Germany shapes individual attitudes toward immigration. The subsequent sections are organized as follows. Section 2 reviews existing literature on the determinants of individual attitudes toward immigration, and section 3 delineates the conceptual framework. Section 4 introduces the G-SOEP dataset and variables utilized in the empirical analysis, the model for which is discussed in section 5. Finally, the results are presented in section 6, and implications of the findings are discussed in section The Determinants of Individual Attitudes Toward Immigration: Literature Review 2.1. Economic determinants of individual attitudes toward immigration Two stands of literature have emerged demonstrating that individual attitudes toward immigration are determined by both economic and non-economic factors (see, e.g., O Rourke and Sinnott 2005; Mayda 2006). First, several studies have shown that the relative composition of high-skilled and low-skilled native labor determines natives attitudes toward immigration (see, e.g., O Rourke and Sinnott 2005; Mayda 2006; Scheve and Slaughter 2001). Trade theory predicts that low-skilled workers will migrate from poor (i.e., low-skill abundant) to rich (i.e., high-skill abundant) countries; alternatively, high-skilled workers will migrate from rich to poor countries. Assuming high-skilled and low-skilled labor are complements, it follows that in rich countries, immigration will adversely affect low-skilled labor and benefit high-skilled labor. Thus, considering natives from a high-skill abundant country, theory predicts that low-skilled native labor should express negative views toward immigration. Indeed, empirical findings have demonstrated that high-skilled labor force participants in rich countries are less concerned about immigration compared to their low-skilled counterparts, where the effects are pronounced in countries with greater inequality (O Rourke and Sinnott 2005; Mayda 2006; Scheve and Slaughter 2001). Next, the unemployed could also hold negative attitudes toward immigration compared to the employed, given that such attitudes could manifest if native workers believe that immigrants are stealing their jobs. Moreover, the unemployed may feel that immigrants place a higher burden 3

5 on the welfare system, which disproportionately disadvantages them relative to the employed (Hillman 2002). As observed by O Rourke and Sinnott (2005), however, individuals currently holding jobs may also fear that increased migration will threaten job security, thereby leading to negative attitudes. Notably, Mayda (2006) finds that the economic determinants of attitudes toward immigration have explanatory power even after controlling for a host of non-economic factors (e.g., individuals perception of the impact of immigration on crime, culture, national identity, and racist attitudes toward political refugees and illegal immigrants) Non-economic determinants of individual attitudes toward immigration Empirical research shows that several non-economic factors, including age, education, race, political affiliation, and other socio-demographic characteristics, also determine individual attitudes toward immigration (see, e.g., O Rourke and Sinnott 2005; Hainmueller and Hiscox 2007; Dustmann and Preston 2007; Chandler and Tsai 2001). First, numerous studies demonstrate age to be a determinant of attitudes toward immigration (see, e.g., O Rourke and Sinnott 2005; Chandler and Tsai 2001). O Rourke and Sinnott (2005) observe that an aging Western society presents a challenge for funding public pension systems, while increased immigration results in a well-documented fiscal cost in terms of welfare for the receiving society (Razin and Sadka 2004; Nannestad 2004). Thus, pension and welfare systems differentially impact the young and the elderly. Age differences could also reflect discrepancies in culturally-determined values and beliefs. Particularly, older individuals may be more likely to adhere to traditional cultural values and norms. Given the above, one could intuit that, all else equal, older individuals would express more negative attitudes toward immigration compared to their younger counterparts. Empirical findings in O Rourke and Sinnott (2005) and Chandler and Tsai (2001) have demonstrated the legitimacy of this intuition. Second, educational attainment has been shown to affect attitudes (see, e.g., Hainmueller and Hiscox 2007; Chandler and Tsai 2001). Hainmueller and Hiscox (2007) find that individuals equipped with higher levels of education are more likely to favor immigration irrespective of the immigrants country of origin and whether they serve as complements or substitutes in labor. Moreover, Chandler and Tsai (2001) show that college education is a powerful agent engendering positive attitudes toward immigration. Together, these findings suggest that negative attitudes toward immigration are more likely to be associated with values or beliefs more prevalent in 4

6 individuals with less educational attainment. A third potential non-economic factor contributing to attitudes toward immigration is race. While Chandler and Tsai (2001) find evidence that race does not play a statistically significant role in determining attitudes toward immigration, Dustmann and Preston (2007) use richer data to present empirical evidence that British respondents view immigration from countries with ethnically different populations more negatively. This paper contributes to the existing literature on the determinants of attitudes toward immigration by exploiting the panel data structure of the G-SOEP to study the enduring effect of communism on individual attitudes toward immigration. Controlling for the economic and noneconomic determinants of attitudes toward immigration, I demonstrate that greater concerns about immigration to Germany held by former East Germans are a result of the historical legacy imparted by the communist regime of East Germany. 3. The Effect of Communism on Individual Attitudes Toward Immigration: Conceptual Framework There is a wealth of psychological, political-, and social-scientific theory suggesting that exposure to communism could lead to greater concerns about immigration. The underlying mechanism through which communism engenders a change in individual preferences is explained by the socialization theory (see, e.g., Campbell et al. 1960; Greenstein 1965; Langton and Jennings 1968; Jennings and Markus 1984), an institutional theory originally introduced in political science. 5 The socialization theory posits that individual attitudes are critically shaped by communism through a socialization process, in which citizens develop preferences for conformity at a young age that remain fixed over the course of their lives. Thus, by conceptually understanding precisely how individual attitudes developed under communist rule, I examine whether, and if so, to what extent, these attitudes persisted into the post-communist era (Pop-Eleches and Tucker 2010). I hypothesize that exposure to communism will, all else equal, make individuals more concerned about immigration because immigration represents one salient threat to conformity, an aspect underscored by communist rule in East Germany. The East German regime accomplished 5 Here, I draw from the analysis in Pop-Eleches and Tucker (2010). While this theory is applied in a traditional context to explain individual political attitudes, I argue that there exists a certain level of endogeneity between individual political attitudes and individual attitudes toward immigration. In this way, I view the theoretical basis of this theory as translational to the current conceptual framework. 5

7 conformity with communist ideology through indoctrination 6 in educational, media, familial, religious, and local governmental spheres, and enforced conformity via a unique and pervasive secret intelligence agency: the East German Ministry of State Security (Ministerium für Staatsicherheit) secret service, also known as the Stasi 7 (Almond 1983; Aviram and Milgram 1977; Lottich 1963). The result was an unprecedented degree of State control over the daily lives and thoughts of East German citizens, where surveillance and indoctrination efforts were exacerbated by the simultaneous repression and cooptation of most civil society organizations by the communist regime (Pop-Eleches and Tucker 2010). Given that studies in social psychology have demonstrated empirically that behavioral norms develop primarily within the first 10 years of life, changing only marginally thereafter (see, e.g., Harbaugh and Krause 2000; Fehr et al. 2008; Sutter et al. 2009), it is crucial to understand to what extent socialization occurred within educational institutions. Clearly reflected in the youth law (Jugendgesetz) of 1987, the objective of education in East Germany was the following: The most important task in the formation of the developed socialist society is to bring up all young people to be citizens who are true to the ideas of socialism, who think and act as patriots [and] who strengthen socialism and staunchly defend it against all enemies. 8 Thus, the communist regime s emphasis on ideological conformity in educational institutions, coupled with the Stasi s indefatigable enforcement of conformity, resulted in East German socialization tantamount to conformity with communist ideology. 9 Individuals valuing conformity are sensitive to perceived threats to conformity (Feldman 2003). One threat to conformity consistently emphasized under communist rule was the threat represented by external enemies : the threat of people residing outside the regime to ideological and social conformity within the regime. This concept is echoed in the latter statement of the Jugendgesetz, in which citizens of the communist regime are called to staunchly defend [communist ideology] against all enemies [thereof]. Therefore, immigration, as a pluralistic 6 Note, I use indoctrination and socialization interchangeably to reflect the process by which individual preferences are shaped in harmony with communist ideology. 7 Failure to conform to communist ideology subjected one to serious penalties as severe as imprisonment or psychological torture by the Stasi. 8 Jugendgesetz (9th Edition, Berlin, 1987), p. 9. The Jugendgesetz was an act on the participation of the youth of the German Democratic Republic (GDR) in the furthering of the developed socialist society and its indefatigable encouragement of the GDR (translation is my own). 9 Importantly, the longevity of the communist regime of East Germany, which lasted 45 years, afforded the regime a means by which it could efficaciously change the attitudes and behavior of East German citizens (Pop-Eleches and Tucker, 2010). Specifically, regime duration is a critical consideration not only in the transformation of institutions, but also in the degree to which individual exposure to these various institutions ultimately shaped attitudes. 6

8 phenomenon, represents one particularly prominent threat to conformity, insofar as it epitomizes social, ideological, ethnic, and cultural diversity that directly opposes communist ideology. Given that former East Germans preferences adapted to value conformity due to socialization, this paper thus investigates the hypothesis that former East Germans are more likely to be ceteris paribus very concerned about immigration compared to former West Germans. 4. Data and Variables 4.1. Dataset: The German Socio-Economic Panel I utilize panel data from the German Socio-Economic Panel (G-SOEP), a wide-ranging, nationally representative longitudinal panel dataset of private households which consists of former East and West Germans, foreigners, and recent immigrants to Germany. The G-SOEP began in 1984 with a random sample of individuals residing in private households in West Germany. Each survey year, the G-SOEP samples an average of 11,000 households for a total of 30,000 individuals. Former East Germans were first added to the G-SOEP in June of 1990 following the fall of the Berlin Wall and just prior to the reunification of Germany. The motivation underlying the addition of former East Germans at this point was multifold. In sampling the first group of former East Germans prior to the currency, economic, and social union of East and West Germany on October 3, 1990, the study sought to observe the imminent social and economic changes and their impacts in East Germany (SOEP Group 2011). The G-SOEP is one of the primary studies utilized for empirical research by social scientists in Germany. Created in the 1980s with the intention of facilitating academic research in the fields of micro-analytical sociological and economic research as well as micro-econometrics, the G-SOEP offers nationally representative data on both objective living conditions as well as subjective perceptions (SOEP Group 2011). As a result, the G-SOEP is employed actively to explore research questions related to the labor market, income development, educational attainment, social outcomes, and life satisfaction among many other individual- and householdlevel questions by social scientists around the world. The G-SOEP dataset has been utilized in four previous studies to investigate attitudes toward immigration (see Avdeenko and Siedler 2016; Poutvaara and Steinhardt 2015; Schüller 2013; Calahorrano 2011). To the best of my knowledge, however, my analysis is the first to employ the G-SOEP or any data to explore the effect of politico-economic regimes on individual attitudes toward immigration. 7

9 Variables utilized in the regression analysis are defined in Table 1, and Table 2 presents descriptive statistics for these variables Outcome variable The outcome variable I use is based on the following survey question, asked in every survey year since 1999: What is your attitude toward the following areas are you concerned 10 about them? Respondents were provided a list of various political, economic, and social phenomena and could respond in one of three ways: as (1) very concerned, (2) somewhat concerned, or (3) not concerned at all. Among others, respondents were asked to indicate their level of concern about immigration to Germany. Hence, I define the outcome variable, Y, as a dummy equal to 1 if the respondent answered that she was very concerned about immigration and 0 otherwise. 11 As a robustness check (see section 6.2), to demonstrate that the results are not sensitive to the definition of the outcome variable, I also group respondents that were either very concerned or somewhat concerned about immigration to Germany (assigned 1) and view those observations relative to respondents who were not concerned at all (assigned 0). The resulting sample size contains 220,396 person-year observations from 1999 until Here, 61,711, or approximately 28-percent of person-year observations, are coded as very concerned about immigration to Germany, with the remaining 158,685 person-year observations indicating responses of somewhat concerned or not concerned at all about immigration Focal explanatory variable My focal explanatory variable is based on the following survey question, asked in every survey year since 1990: Where did you live in 1989? In response to this question, survey participants indicated whether they lived in East Germany (including East Berlin), West Germany (including West Berlin), or abroad. Hence, respondents answers to this question offers a time-invariant characterization of whether they lived in East or West Germany prior to reunification. Because I am interested in the effect of communism on attitudes toward immigration, I 10 The word concern used in the survey question, is a translation from the original German word used in the survey, Sorge. It is important to note that Sorge in German has an associated negative connotation, and thus, the word concern as it translates into English is not that of positive concern, but rather indicates a negative attitude toward immigration. 11 Importantly, the extent of knowledge about individual attitudes toward immigration is restricted to this interpretation. Thus, I am unable to establish whether concern about immigration translates to behavior toward immigrants, or whether opinions and beliefs in this capacity translate to other actions or behavior more generally. Ceobanu and Escandell (2010) provide a thorough review of the literature related to individual attitudes toward immigration and emphasize the importance of this semantic distinction and the associated interpretation. 8

10 drop all respondents from the sample that indicated they had lived abroad in 1989 (21,867 personyear observations), and consider only those who resided either in East or West Germany. Consequently, the focal explanatory variable in this study, East, is an indicator variable equal to 1 if respondents resided in East Germany and 0 if respondents resided in West Germany. In the regression analysis, I restrict the sample to only German citizens, excluding non-german nationals. 12 In doing so, I obtain 220,396 person-year observations over which I regress, in which 68,322, or approximately 31-percent of observations are coded as East German, with the remaining 152,074 consisting of individuals who lived in West Germany. One of the key identifying assumptions made in this paper is that West Germans constitute a suitable control group for the East German treatment group. Several studies offer thorough and comprehensive discussions on the validity of post-war Germany as a natural experiment, and conclude, in effect, that East and West Germany were indistinguishable until the exogenously imposed separation in 1945 (see, e.g., Alesina and Fuchs-Schündeln 2007; Rainer and Siedler 2009; Pop-Eleches and Tucker 2010). There are two potential concerns which could jeopardize the integrity of the exogenous division of Germany with respect to the outcome of interest, attitudes toward immigration. First, the initial division of Germany could have been imposed non-randomly vis-à-vis attitudes toward immigration, whereby more individuals with existing concerns about immigration were endowed to East rather than West Germany. Second, non-random movement between East and West Germany following the division, particularly consisting of individuals less concerned about immigration in East Germany migrating to West Germany, could have led to self-selection, thus threatening the validity of the communism treatment effect. I will address each concern in turn, arguing that there exists no substantial evidence for either claim. 13 First, an empirical analysis conducted by Alesina and Fuchs-Schündeln (2007) in pre- World War II Germany reveals no systematic differences between the two regions of East and West Germany, suggesting that individuals existing concerns about immigration attributable to economic anxieties were as good as randomly assigned between the two countries. Next, most cultural and political heterogeneity (e.g., language, political parties, Protestants and Catholics, etc.) in Germany up to 1945 was organized along a North to South rather than East to West divide, 12 The findings are not sensitive to the choice to exclude non-german nationals (result not reported). 13 Note, the following discussion closely adopts the argument presented in Alesina and Fuchs-Schündeln (2007). 9

11 historically fashioned by Roman versus Germanic occupation. Therefore, the East-West division can be viewed as initially exogenous. Second, because three million individuals migrated from East to West Germany between 1949 and 1961 when the Berlin Wall was established an event which stymied subsequent flows to just 600,000 people in total between the years of 1962 and 1989 it is critical to address the concern that self-selection occurred following the exogenous division (Storbeck 1963; Schumann et al. 1996). 14 Sociological literature suggests two primary mechanisms which motivated East to West migration: (1) familial reasons and (2) economic prosperity in the West (see, e.g., Heidemeyer 1993; Storbeck 1963). Thus, while it is possible that preferences endogenous to attitudes toward immigration played a role in migratory decisions, it is unclear why these motivations would introduce external validity concerns severe enough to dismantle the character of the random assignment in attitudes toward immigration between East and West Germany (Alesina and Fuchs-Schündeln 2007). In conclusion, the variable East appropriately captures the exogenous variation existing between former East and West Germans due to political regime type. Finally, in Section 6.2, I also check the robustness of the variable East by generating a series of federal state of residence (Bundesländer) variables. Here, the six Bundesländer variables (Brandenburg, Mecklenburg-Vorpommern, Sachsen, Sachsen-Anhalt, Thüringen, and East Berlin) characterize in which East German Bundesland an East German lived in June of Like the variable East, this again offers a time-invariant characterization of which Bundesland a former East German lived in prior to reunification, and allows for an examination of the homogeneity of the effect of communism across the various regions it affected Control variables In my regression analysis, I include several controls for individual socio-economic and demographic characteristics influencing individual attitudes toward immigration. I control for education, income, household size and composition, age, gender, marital status, employment status, occupation type, and survey year fixed-effects. First, as discussed in section 2, there are several economic and non-economic determinants of individual attitudes toward immigration. O Rourke and Sinnot (2005), Scheve and Slaughter 14 By contrast, just 300,000 individuals migrated from West to East in the 1950s, and West to East migration was negligible after 1961 (Alesina and Fuchs-Schündeln 2007). 10

12 (2001), and Mayda (2006) empirically verify classical trade and labor market theories suggesting that occupation type, employment type, and income affect individual attitudes. To this end, I include two dummy variables indicating individuals who are high-skilled and low-skilled workers, as well dummies for employment status (employed, unemployed and not working, unemployed and receiving pension, and self-employed). Finally, I also include the logarithm of annual household post-government income to control for any potential variation in individual attitudes toward immigration which could be driven by income. Second, a wealth of empirical evidence demonstrates that non-economic factors affect individual attitudes toward immigration. Following Hainmueller and Hiscox (2007) and Chandler and Tsai (2001), who demonstrate that individuals with lower educational attainment exhibit morenegative views toward immigration than those with higher educational attainment, I include as a control variable the total years of education a respondent has attained. Because education may operate qualitatively differently in East and West Germany with respect to attitudes toward immigration, I examine this relationship more carefully in a subsequent exercise (see section 6.2). Furthermore, age affects attitudes toward immigration (O Rourke and Sinnott 2005; Chandler and Tsai 2001). Thus, I include a degree-two polynomial in age to account for any non-linearity between age and individual attitudes toward immigration. Additionally, gender, marital status, and household composition and size are plausible determinants of individual attitudes toward immigration. To this end, I include in the regression analysis dummy variables that control for these individual- and household-level characteristics. Finally, to control for year-specific variation in attitudes, I also include as a control in each regression a series of 15 survey year dummy variables for the years 1999 through 2013, the period over which I exploit the G-SOEP. By controlling for each of these economic and non-economic determinants of attitudes toward immigration, the regression analysis indicates (under the assumption that the selection on observables holds) the causal effect of communism on individual attitudes toward immigration. 5. Empirical Model I estimate several variations of the following linear probability model: Y = β East it i + X it δ + t + ε it, (1) where, in the preferred specification, Y is a dummy variable indicating respondents i who are very concerned about immigration to Germany in year t, East is the focal explanatory variable 11

13 indicating exposure to communism, X is a vector of individual-level controls as introduced in section 2 and discussed in section 4.4., and is survey year fixed-effects Here, the test of the effect of being East German is a test that the treatment of communism increased individual concern about immigration relative to the West German control group. Thus, the hypothesis tests are two-tailed tests, and measure whether β, the coefficient on the variable East, is statistically significantly different than zero. Demonstrating positive statistical significance for the coefficient β is tantamount to establishing that former East Germans are both statistically and economically significantly more likely to be very concerned about immigration to Germany relative to former West Germans, ceteris paribus. Given the identification assumption that the variable East is indeed exogenous, the OLS estimate of β captures the causal effect of the politico-economic institution of communism on individual attitudes toward immigration. In sections 6.2 and 6.3, I estimate several alternative specifications of Equation (1). First, I examine the robustness of the specification of the outcome variable by pooling together respondents that were either very concerned or somewhat concerned about immigration and viewing those observations relative to respondents who were not concerned at all. Second, I include numerous other variables in the vector of covariates, X, to examine the (1) effect of education in East Germany, (2) convergence of attitudes over time, (3) effect of living under communism longer, (4) homogeneity of the effect of communism over various regions, and (5) potential channel variables explaining why communism increases individual concerns about immigration. I rely on the linear probability model in favor of an econometric model like the ordered probit for the following reasons. First, the use of the linear probability model renders the interaction effects between the focal explanatory variable and a subset of controls, an exercise I explore in a subsequent section, more readily interpretable. Additionally, the baseline estimates obtained using the ordered probit model are qualitatively identical to those obtained via the use of the linear probability model Results 6.1. Main results Table 3 reports the main results and builds the baseline specification introduced in the previous 15 The results from the ordered probit regressions are available upon request. 12

14 section, in which I include as explanatory variables the variable East in addition to several individual-level controls. As discussed above, the outcome variable is a dummy variable indicating whether the respondent was very concerned about immigration to Germany. The analysis which follows restricts the sample to only German citizens, thus excluding non-german citizens (11,610 person-year observations). The main findings are qualitatively identical when including and controlling for non-german citizens (results not reported). The first explanatory variable, the East dummy variable indicating the communism treatment effect, is most critical and the focus of the present analysis. Considering column (1) of Table 3, which reports the bivariate linear regression of Very concerned about immigration on East, the coefficient on the variable East is highly statistically significant with a magnitude of This result should be interpreted as follows: Having lived under communism increases the probability that one is very concerned about immigration by 4.0 percentage points, controlling only for survey year fixed-effects. Columns (2) through (8) examine the robustness of this correlation to the inclusion of a vast array of covariates. Each specification controls for survey year fixed-effects. First, in columns (2) through (5), I introduce several non-economic characteristics, including a degree-two polynomial in age (column (2)), gender (column (3)), marital status (column (4)), and the number of adults and children residing in the household (column (5)). Notably, the variable East remains robust to the inclusion of this set of covariates, and the magnitude of the East coefficient remains effectively unchanged. Further, the coefficient on East is statistically significant at the one-percent level in each of columns (2) through (5). The statistically significant coefficient on the focal explanatory variable, East, in column (5) indicates that, after controlling for respondents age, gender, marital status, and household composition, having lived under communism increases the prospects of being concerned about immigration by 4.0 percentage points. Second, I examine the robustness of the East dummy variable to an array of economic characteristics, including occupation type and labor status (column (6)), total years of educational attainment (column (7)), and the logarithm of post-government household income (column (8)). Column (6) controls for a series of dummy variables indicating skilled labor, unskilled labor, unemployed and not working, unemployed and receiving pension, and self-employed. The interpretation of each reported occupation type should be interpreted relative to the heterogeneous 13

15 omitted category of routine non-manual labor, routine service-sales, and farm labor. Of importance to this study, the variable East remains positive and statistically significant to the inclusion of occupation type and labor status, where the magnitude of the coefficient on East is now In column (7), I add as a control respondents total educational attainment in years. Consequently, the coefficient on the variable East increases in magnitude to and remains statistically significant at the one-percent level. Finally, by controlling for income, I obtain the preferred baseline regression. The preferred specification, which includes controls for age, gender, marital status, household size and composition, occupation type and labor status, total educational attainment, and income, is portrayed in column (8). After controlling for respondents post-government household income, which captures the combined income after taxes and government transfers of all individuals in the household, the coefficient on East remains positive and statistically significant at the one-percent level, and thus robust to the inclusion of economic determinants of individual attitudes toward immigration. The magnitude of the coefficient on the focal explanatory variable, East, is 0.034, indicating that having lived under communism results in a 3.4 percentage point increase in the probability of being very concerned about immigration even after controlling for the full range of non-economic and economic determinants of individual attitudes toward immigration. These results indicate that regardless of specification, there is no change in the statistical and economic significance of the focal explanatory variable, East, as it explains individual concern about immigration. This is evidence that the treatment of communism in East Germany constitutes a causal determinant of individual attitudes toward immigration Further results Tables 4 and 5 report further results based on five exercises which expand on the baseline regression from column (8) of Table 3. To economize on space, the covariates introduced in Table 3 are no longer reported, but are included in each of the remaining specifications. First, following my discussion in section 5, I explore whether the main qualitative result discussed above is sensitive to the definition of the outcome variable. Second, because education under communism could have functioned qualitatively differently than in West Germany, I examine the effect of education in East Germany on attitudes toward immigration. Third, I explore if attitudes between former East and West Germans convergence over time. Fourth, I investigate the role that differential exposure times under communism play in determining individual attitudes toward 14

16 immigration. Finally, in Table 5, I replace the variable East with a series of dummies indicating respondents federal state of residence (Bundesland) in 1990 to verify that the attitudes over former East and West German regions were homogenous as a result of different historical legacies. Column (1) of Table 4 examines the robustness of the primary result discussed in section 6.1 to the definition of the outcome variable. To this end, I replace the outcome variable by grouping respondents that were either very concerned or somewhat concerned about immigration to Germany (assigned 1) and viewing those observations relative to respondents who were not concerned at all (assigned 0). Relaxing the necessary condition of being concerned about immigration, 73 percent of person-year observations are now coded as very or somewhat concerned with the remaining 27 percent coded as not concerned at all. The result in column (1) of Table 4 indicates that the significance of the East dummy is robust to the definition of the outcome variable, where having been exposed to communism increases the prospects of being very or somewhat concerned about immigration by 4.8 percentage points, ceteris paribus. Second, it is plausible that educational institutions in East and West Germany operated differently in determining attitudes toward immigration, whereby higher education in East Germany could have actually increased concerns about immigration due to prolonged exposure to indoctrination. 16 I examine the effect of education in East Germany on individual attitudes toward immigration in column (2). To this end, I restrict the sample to respondents born prior to (and including) the year 1971, ensuring that most respondents analyzed completed their education prior to reunification. The OLS estimate of the interaction between the East dummy and the continuous variable measuring respondents educational attainment in years captures the percentage point change in the probability of being very concerned about immigration due an additional year of education under communism. The coefficient on the East Years of education interaction term is positive, indicating that an additional year of education in East Germany increases the probability of being concerned about immigration, all else equal. However, the obtained result is statistically insignificant (pvalue 0.186). Third, since former East Germans are, on average, more concerned about immigration to 16 Unfortunately, I am unable to investigate the differential effect of education within different fields (e.g., the social and natural sciences, humanities, etc.) on individual attitudes toward immigration. There is reason to believe that education in East Germany could have operated differently in altering preferences related to immigration depending on the field of study. 15

17 Germany than former West Germans, a natural question arises: Has the difference in attitudes between these groups converged since reunification (Alesina and Fuchs-Schündeln 2007)? Column (3) from Table 4 indicates that the magnitude of difference between former East and West German concerns about immigration is not diminishing over time, a finding consistent with the socialization theory discussed in section 3. Here, I generate three survey year cohorts to investigate the magnitude of East German concern from 2002 to 2005, 2006 to 2009, and 2010 to 2013, where the omitted category is the survey year cohort 1999 to First, note that the coefficient on East, which indicates the average effect of communism on attitudes toward immigration from 1999 to 2001 is positive and statistically significant with a magnitude of The interaction terms are central to the analysis of convergence among former East versus West German attitudes. The interaction of the variable East with the survey years 2002 to 2005, 2006 to 2009, and 2010 to 2013 has a coefficient of , , and 0.006, respectively. The interpretation of these coefficients is as follows: While having lived in East Germany resulted in a 3.5 percentage point increase in the probability of being very concerned about immigration to Germany on average in the survey years 1999 to 2001, the effect weakens to 3.4, 2.6, and increases to 4.1 percentage points on average in the survey year cohorts 2002 to 2005, 2006 to 2009, and 2010 to 2013, respectively. The coefficients are statistically insignificant. Fourth, I investigate the impact of increased exposure to communism on individual attitudes toward immigration. Here, I posit that respondents with more exposure to communism, and thus greater levels of socialization within the regime, will exhibit more concerns about immigration, ceteris paribus. The variable Years lived in East indicates the total number of years a respondent spent in the communist regime, and thus takes on a minimum value of 0 (for West Germans) and a maximum value of 45 (or how old the East German regime was at the time of reunification). To allow for non-linearity, I include a quadratic in years. Reported in column (4), the result indicates that living an additional year under communism increases (concavely) the probability of being very concerned about immigration by 0.6 percentage point. The result reported in column (5) of Table 4 probes more carefully at the effect of time spent under communism on attitudes toward immigration. Here, I construct five bins which characterize respondents year of birth as follows: (1) born after 1975 (the omitted category), (2) born from 1961 to 1975, (3) born from 1946 to 1960, (4) born from 1931 to 1945, and (5) born before First, note that the East coefficient, representing the percentage point difference in 16

18 the probability of being very concerned about immigration among former East Germans born after 1975 relative to former West Germans, is positive and statistically significant with a magnitude of The coefficients critical to revealing the differences in attitudes due to age are the interaction terms. Here, the results imply that there is no statistical difference in attitudes toward immigration among former East German respondents who were born in the regime (i.e., the cohorts 1961 to 1975 and 1946 to 1960) relative to those born after Interestingly, however, former East Germans born prior to the imposition of communist rule do not appear to have attitudes which differ substantially from former West Germans. Namely, the coefficients for the East born and East born before 1931 interactions are negative and statistically significant at the one-percent level with magnitudes of and , respectively. This result suggests that, relative to former East Germans born after 1975, respondents born in East Germany from 1931 to 1945 and before 1931 are 0.6 and 2.7 percentage points less likely to be very concerned about immigration, respectively. While an F-test of joint significance does not allow for the rejection of the null that East Germans born from 1931 to 1945 are statistically indistinguishable from former East Germans born after 1975 (p-value = 0.509), a second F-test reveals that East Germans who were born prior to 1931 exhibit concerns about immigration that are statistically significantly smaller than former East Germans born after 1975 (p-value = 0.093). Finally, to ensure that the measured effect truly captures the effect of different historical legacies imparted by different politico-economic regimes, I investigate the homogeneity of attitudes across former East and West German regions. In column (1) of Table 6, I introduce in place of the focal explanatory variable, East, six new dummy variables which indicate the federal state (Bundesland) of residence in which former East German respondents lived when they were first surveyed in June of 1990, where the omitted category consists of the former West German states. To account for movers following reunification, I code these dummy variables only considering respondents residence at the time they were sampled in the 1990 wave of the G- SOEP, prior to the formal currency, economic, and social union of East and West Germany in October of Thus, the resulting sample is restricted to respondents from the 1990 wave of the G-SOEP that remained in the survey throughout at least a portion of the period of analysis (

19 to 2013). 17 Like the East dummy, the Bundesland dummy variables represent time-invariant characterizations of respondents state of residence prior to reunification. As expected, five of the six Bundesland dummy variables have positive coefficients, four of which are statistically significant, demonstrating that the effect of communism on increased concerns about immigration was effectively uniform across the East German Bundesländer. The coefficients on the Bundesland dummy variables, which should be interpreted as the average percentage point difference in the probability of being very concerned about immigration among respondents within a particular East German Bundesland relative to West Germany, are as follows: Brandenburg (0.032), Sachsen (0.047), Sachsen-Anhalt (0.033), Thüringen (0.015), East Berlin (0.154), and Mecklenburg-Vorpommern (-0.041). An F-test of joint equality leads to a rejection of the null that the effect was equivalent across the six former East German regions (p-value < 0.01). Column (2) of Table 5 introduces a series of dummy variables for former West German states in addition to former East German states, where the omitted category is now the West German state, Saarland. Again, the effect of communism seems to have been relatively uniform across East German states, where having lived in a former East German state in 1990 increases the probability of being concerned about immigration relative to concerns held in Saarland. Further, the results in column (2) suggest that attitudes across West German regions are mostly homogenous except for the regions Nordrhein-Westfalen and West Berlin where the probabilities of being concerned about immigration are statistically significantly greater than in Saarland and in Rheinland-Pfalz and Baden-Württemberg where the probabilities of being concerned about immigration are statistically significantly less than in Saarland Potential channel variables Table 6 contains results from the inclusion of an array of potential channel variables into the baseline specification reported in column (8) of Table 3. As before, I do not report the coefficients on the covariates from Table 3 although they are still included in each specification. This exercise offers a more precise explanation for the observed higher concerns about 17 The definition of the Bundesland dummy variables is like the definition of the focal explanatory variable (a necessary condition for living in a former East German state is that the East dummy be equal to 1), but does not include respondents who entered the survey after Unlike the variable used to generate the East dummy, the survey question inquiring about respondents federal state of residence is asked annually (and not retrospectively about 1989, specifically), and thus differs in many instances across time within the panel as respondents move within Germany. 18