Abstract

In this study, we investigate how the attitude of natives—defined as the perceived trustworthiness of citizens from different countries—affects immigrants’ labor market outcomes in Germany. Evidence in the literature suggests that barriers to economic assimilation might be higher for some groups of immigrants, but the role of natives’ heterogeneous attitudes toward immigrants from different countries of origin has received little attention. Using individual-level panel data from the German Socio-Economic Panel covering the years 1984 to 2014, we apply survival analysis methods to model immigrants’ unemployment durations. We find that lower levels of trust expressed by natives toward the citizens of a given country, measured using Eurobarometer surveys, are associated with increased unemployment durations for immigrants from this country. We show that this result is not driven by origin-specific unobserved heterogeneity and that it is robust to different specifications and alternative explanations.

Introduction

Germany is currently confronted with the challenge of integrating sizable inflows of foreign-born populations, including economic migrants and refugees. According to the German Federal Statistical Office, 2,137,000 people immigrated to Germany in 2015, an increase by 672,000 arrivals (46 %) with respect to 2014 (Destatis, 2016). The public debate on the reception of immigrants and asylum seekers has sparked highly divergent reactions within the German population, ranging from warm welcome demonstrations to violent protests against this historical surge in foreign-born population. Such mixed feelings about immigration are not new, and the integration of the foreign-born population in the local labor market has been a source of concern for decades (Borjas 2014). In this context, an often-overlooked question is how the attitudes of natives affect the integration of immigrants.

Natives’ attitudes toward immigrants are conditioned by several factors, including trustworthiness. Specifically, a lack of trust toward the citizens of a given country could make natives reluctant to cooperate with immigrants originating from that country. Indeed, as a prerequisite for contracts in the absence of complete information, trust conditions the willingness to engage in economic transactions (Göran and Hägg 1994). Thus, in the presence of imperfect contract enforcement, the willingness of native employers to enter a contract may be determined by their perception of an immigrant worker’s trustworthiness. As a consequence, the number of job offers immigrants receive would depend on the level of trust that natives associate with the citizens of their origin country. In turn, standard job-search models predict that a reduced arrival rate of job offers entails increased unemployment durations.

Our study investigates empirically whether trust levels that Germans associate with the citizens of an immigrant’s country of origin influence his or her labor market outcomes. We carry out the empirical analysis building on an individual-level panel data set, the German Socio-Economic Panel (GSOEP). Specifically, we use monthly calendar information to construct labor market activity spells over the period 1984–2014. We then analyze immigrants’ unemployment spells using Cox and Weibull proportional hazard models and test whether the latter are associated with the level of trust that Germans express toward the citizens of an immigrant’s country of origin. We measure trust as the share of Germans declaring in Eurobarometer surveys that citizens of the country in question are trustworthy, exploiting variations both at the national and regional level. The analysis at the regional level allows us to control for origin-specific factors, such as selection into migration. Indeed, a major drawback of analyzing a self-selected stock sample is that it excludes potential migrants for whom a lack of trust is most costly. This in turn can confound the identification of the relationship between trust and immigrants’ labor market outcomes, inducing a downward bias in the estimated coefficients. Hence, our analysis highlights the relevance of the identification challenge posed by varying self-selection patterns across origin countries.

The results suggest that natives’ attitudes strongly influence the labor market outcomes of immigrants. Our findings indicate that if the level of trust that Germans associate with Turkish citizens increased by 150 %—reaching the level of trust associated with Austrian citizens—and with all other variables held constant, the average unemployment duration of Turkish migrants would be reduced by five months on average. Remarkably, our results also report that immigrants facing a lower level of trust are more likely both to experience longer unemployment spells and to slip into inactivity. Additional specifications reveal that these effects are robust to controlling for individual perceptions and attachment to Germany, reinforcing the interpretation that it is driven by the demand side rather than by individual characteristics.

In this article, we follow Gambetta’s (2000:217) definition of trust: “when we say we trust someone or that someone is trustworthy, we implicitly mean that the probability that he will perform an action that is beneficial […] is high enough for us to consider in engaging in some form of cooperation with him.” As Sapienza et al. (2013) noted, this definition considers trust to be a belief that can be measured as a probability. One could think that using the country of origin to infer trustworthiness would be itself a form of discrimination, which Lang and Lehmann (2012:2) defined as “profiling on the basis of race or ethnicity.” However, as Ashraf et al. (2006) showed in an experiment, a lack of trust can be driven by both tastes and expectations simultaneously. In other words, people may not trust others because they dislike them or because they think they are not trustworthy, implying that our results may be interpreted either through the spectrum of prejudices (taste discrimination; e.g., Becker 1957) or informational problems (statistical discrimination; e.g., Akerlof 1976; Arrow 1973; Phelps 1972).

This study is related to two strands of the literature that have developed largely in isolation from each other. First, it contributes to the literature investigating the labor market integration of immigrants by shedding light on natives’ trust as a significant channel driving the observed heterogeneity in labor market outcomes across different groups of immigrants. Indeed, it has been extensively documented that immigrants underperform natives in the labor market (Aldashev et al. 2009; Bisin et al. 2011; Borjas 2014; Casey and Dustmann 2010; Chiswick 1978). Although many factors have been shown to play a role—including lower abilities of immigrants, the difficulties that firms have to properly assess the qualifications obtained in a foreign country, lack of language skills, or weaker social networks—discrimination is often discussed as one of the most important. On the one hand, field experiments (such as correspondence studies), following the seminal contribution by Bertrand and Mullainathan (2004), have consistently shown that ethnic minorities face lower callback rates from employers.1 On the other hand, observational studies have also documented large ethnic penalties, consistent with labor market discrimination (see, e.g., Constant and Massey 2005; Couch and Fairlie 2010; Frijters et al. 2005; Fullin 2011).2 As far as Germany is concerned, evidence suggests that immigrants experience longer unemployment durations compared with native-born workers (Kogan 2004, 2011). The work closest to our analysis is Uhlendorff and Zimmermann (2014), who investigated the unemployment dynamics of immigrants in the German labor market and showed that immigrants do not differ from natives in terms of job quality but do experience longer unemployment spells.

Second, our study adds to the literature investigating the role of trust in economic transactions by documenting how natives’ trust contributes to shaping the labor market outcomes of immigrants. Indeed, several experimental studies have underlined the importance of shared group membership in determining the decision to trust others (Foddy et al. 2009; Platow et al. 2012; Tanis and Postmes 2005). In this way, Cettolin and Suetens (2018) showed that the country of origin is a relevant factor when it comes to evaluating the trustworthiness of others. Moreover, perceived trustworthiness of members of other groups defined by ethnicity or citizenship, which is rooted in group-specific cultural norms, beliefs, and values, has been shown to persist over time and to have strong influences on economic outcomes. For instance, Nunn and Wantchekon (2011) demonstrated that the consequences of the slave trade in terms of mistrust between groups of population in Africa are still observable. By conditioning the level of social capital and cooperation among agents, inherited trust facilitates economic transactions, which is mirrored in macroeconomic outcomes such as economic development (Fafchamps 2006; Knack and Keefer 1997) and economic growth (Algan and Cahuc 2010). Guiso et al. (2009) also showed that bilateral trust between European countries influences bilateral trade flows, portfolio investments, and direct investments.

The results of this article can also be linked to theoretical contributions in the literature. Indeed, the rationale underlying the effect of trust on employers’ willingness to enter a work contract can be conveyed by a simple sequential game that Bohnet et al. (2000) proposed. In this model, the first mover (the employer) decides whether to enter a contract without knowing whether the second mover (the worker) will perform.3 Upon receiving an offer, the worker might either comply with the contract or breach, the latter of which triggers a litigation process with two possible outcomes: the worker can be held liable and bear the legal costs, or the contract might not be enforced, providing a benefit to the worker and a net loss to the employer.4 Workers are heterogeneous in their psychological cost of breaching. Monetary payoffs are such that workers with high psychological costs always perform, whereas workers with low psychological costs breach if the expected gains, determined by the probability of contract enforcement, exceed the yields of complying with the contract. The risky payoff structure implied by imperfect contract enforcement, combined with uncertainty around worker’s preferences with respect to honesty, entails that the employer will enter a contract only if a signal is received that the potential worker is trustworthy. Then, if the country of origin is used as a signal for trustworthiness, immigrants from countries that German employers perceive as less trustworthy will receive fewer job offers and end up with lower exit rates from unemployment, as predicted by a standard job search model.5 Because standard job-search models ignore the reaction of firms, one might expect the labor market to adjust through wages and differences in the exit rate from unemployment between groups of workers to cancel at equilibrium. In a search-matching model, Rosén (1997) showed that groups of workers with a lower probability of being hired can end up with lower wages and a higher unemployment rate in a stable equilibrium. Lang and Lehmann (2012) showed that a simple extension of this model can also explain longer unemployment spells and a higher turnover for discriminated groups.

Unravelling the determinants of immigrants’ unemployment is a highly relevant issue given that immigrants are overrepresented in the unemployed population in Germany: the unemployment rate reached 14.6 % in 2017 for foreigners (with a large heterogeneity among origin countries) versus 4.7 % for the native population (German Federal Employment Agency 2018). Trust levels could plausibly affect different labor market outcomes, such as wages, inactivity, or job quality. We therefore mainly focus on unemployment spells because they impose high costs on society in the form of poorer health of the unemployed, skill depreciation, forgone tax revenue, slower assimilation, and so on.6 In addition, from a theoretical point of view, trust conditions the willingness to engage in cooperation, such as a work contract, suggesting that its effects are mainly concentrated in the job searches of immigrants. A simple correlation between trust and a set of immigrants’ labor market outcomes at the origin level shows that the latter is more correlated with the unemployment rate than, for instance, annual labor income or number of hours worked (see Table A7 in the online appendix). Additional estimates in this article also show that beyond unemployment and labor force participation, trust has no significant impact on the probability of leaving employment. Leopold et al. (2017) also showed that unemployed immigrants in Germany have experienced a greater decline in their subjective well-being compared with natives.

Data

The GSOEP is the most extensive tracking survey of private households and persons in Germany.7 We combine annual longitudinal demographic information on immigrants with monthly calendar data in order to precisely reconstruct individual labor market activity from January 1984 to December 2014. With the exception of Berlin, our sample contains only regions from western Germany because the bulk of surveyed immigrants reside in these regions. Information on labor market activity is then matched with Germans’ levels of trust toward the citizens of immigrants’ origin countries in the sample, taken from the Eurobarometer (for 1976–1997) and the European Election surveys (for 2004). Immigrants are defined on the basis of country of birth, which is a relevant criterion for the analysis because employers’ attitudes are more likely to be influenced by individuals’ observable characteristics—such as name or physical traits—shaped by country of origin rather than by citizenship.

Labor Market Activity

In each wave of the GSOEP, respondents are asked to provide information on their monthly activities of the previous year. Figure A1 in the online appendix provides an example taken from the questionnaire. We build on this information to generate spells of activity for the year preceding the survey. In particular, we obtain individual-specific monthly activity spells by recoding the 15 activity categories into three labor market statuses: employed, unemployed, and out of the labor force.8 Individuals are classified as employed for the months when they declared that they were full-time employed, part-time employed, or marginally employed. Following the GSOEP classification, a second job, a mini-job (earning up to 400 euros per month), and short work hours are also considered as employment. The definition of unemployment is fuzzy because of the difficulty of identifying discouraged workers—that is, workers who are not officially registered as unemployed but who are still available for work. Therefore, following the International Labour Organization definition of unemployment, our sample comprises individuals who are officially registered as unemployed as well as individuals who are not officially registered as unemployed but who declare that they are actively looking for work and/or are available for work in the two weeks following the interview. Both pieces of information are taken from the annual biographical questionnaires. We allocate these annual answers to all 12 months of the survey year. Taking this particular group into account in our analysis is crucial because these individuals could have left the active population precisely because of a lack of trust. To assess the sensitivity of our results to the definition of unemployment, we construct alternative samples with different definitions of unemployment. The results are reported in the online appendix, Table A9. Individuals who do not correspond to either activity status are considered to be out of the labor force. Finally, our sample is restricted to the working-age population—that is, individuals aged 16–64. We also drop individuals once they declare that they have retired.

Trust Data: Eurobarometer and European Election Surveys

The trust data are taken from different surveys sponsored by the European Commission and designed to measure public opinions on various topics. They were conducted on a representative sample of the total population aged 16 and older (approximately 1,000 individuals per country and per year). Specifically, we use waves of Eurobarometer surveys between 1976 and 1997, which collected self-reported trust information from Germans with respect to citizens from 15 countries. We complement this information with the European Election Survey, which collected similar information in 2004. In the Eurobarometer surveys, respondents are asked the following question: “I would like to ask you a question about how much trust you have in people from various countries. For each, please tell me whether you have a lot of trust, some trust, not very much trust, or no trust at all.” To construct a measure of bilateral trust from Germans toward other nations, we use the share of positive answers among the total answers—that is, the share of Germans who answered “very trustworthy” or “fairly trustworthy.” In the 1995 and 1997 waves of the Eurobarometer surveys and in the 2004 European Election Survey, the question was slightly different. The wording of the question was, “Do you trust citizens from country X?” Only two answers were possible: “I trust them,” or “I do not trust them.” For these years, the share of those who answered “I trust them” is used as an indicator of positive opinions.

One could be concerned that attitudes toward the citizens of a given country differ from attitudes toward immigrants from this country. Unfortunately, natives’ attitudes toward immigrants at the origin level are not available. Furthermore, asking about natives’ trust toward the citizens of a given country is more suitable for our empirical analysis because it mitigates the concern that natives’ attitudes are shaped by the labor market performance of a given diaspora. Thus, for each survey wave, we first calculate the mean value of the share of Germans who declare that they trust citizens of the origin country in question. Second, we take the average of the calculated mean values over available waves for each region-origin dyad in order to obtain time-invariant, region-origin–specific mean values. Hence, we obtain a unique value of the trust variable for each region-origin dyad. We also calculate this variable at the country-of-origin level. To avoid recording the opinions of immigrants, we consider responses from only German citizens.9,10 Given the wording of the question, there may be some ambiguity in the interpretation of this measure of trust. Guiso et al. (2009) provided a clarification: the correlation with other questions in separate surveys suggests that the level of trust captured in Eurobarometer surveys reflects the subjective probability that a random person from a given country is trustworthy rather than the respondent’s ability to identify trustworthy people in another country.

Finally, one could be concerned that the Trust variable computed at the regional level reflects statistical noise because of the small number of annual respondents when the Eurobarometer surveys are split between 11 German regions. This concern is mitigated by the fact that the mean value is computed over several waves of Eurobarometer surveys. On average, the mean value of the Trust variable for each region is computed over a sample of 422 individuals.

The factors that shape the perception of the trustworthiness toward citizens from a given country are all rather stable over time. This is illustrated in Figure A4 in the online appendix. The upper panel of the figure shows that the evolution of levels of trust over time is driven by common shocks that do not affect the ranking of countries much. This appears even more clearly in the lower panel of the figure, in which we partial out year fixed effects that capture shocks, such as economic or political conditions in Germany affecting Germans’ general levels of trust toward others.

Table 1 offers further evidence showing the stability of trust over time. We cannot reject the hypothesis that, on average, the share of Germans with positive opinions toward citizens of a given origin in a specific region is not different in two consecutive survey waves. The stability of trust toward citizens from other countries over time is not surprising given that such perceptions are deeply rooted. It confirms that differences in levels of trust are indeed determined by many factors, including historical events, such as wars, cultural differences, differences in political systems, and the quality of law and its enforcement (Guiso et al. 2009).

Descriptive Statistics

Our main sample of analysis is restricted to unemployment spells that do not exceed 48 months.11 This leaves us with a sample of 208,292 individual-month observations between January 1984 and December 2014. Our benchmark sample contains 2,501 individuals from 15 countries, living in 11 regions (Länder) in Germany. On average, 40.3 % of Germans declare that they trust citizens from the countries included in our sample of analysis (see Table 2). This mean value hides a lot of variability among countries of origin: only 24 % of Germans consider Romanians to be trustworthy, whereas as much as 80 % of Germans perceive Austrians as trustworthy.

Figure A5 in the online appendix shows that trust toward citizens of a given country also varies greatly among regions. The variability across regions appears to be in line with the country level average: the minimum and maximum values of Trust at the regional level are generally within a 20 percentage point range of the country level mean. Hence, the relative level of trust toward citizens of a given country is reflected at the regional level with varying intensity.

Median unemployment spells last for eight months in our main sample of analysis, and this figure reaches nine months when considering only those unemployment spells that end with a return to employment. However, immigrants from different countries of origin experience very different lengths of unemployment, and longer average spells of unemployment tend to coincide with lower perceived trustworthiness (see Table 2). For instance, the median unemployment spell for individuals from Turkey lasts 10 months, and the corresponding figure stands at 5 months for individuals from the United States. Considering employment as the only possible failure, and everything else that defines the end of an unemployment spell as right-censored, Fig. 1 shows that immigrants from trusted origin countries experience shorter unemployment spells. This relationship is not systematic because individuals from different countries of origin are very different in terms of observable characteristics. Thus, we seek a more sophisticated analysis that can account for such confounding factors.

Empirical Analysis

We model immigrants’ unemployment duration in Germany using survival analysis methods with unemployment duration reported in months. Additional issues related to our empirical equation—and particularly the implications of migrants’ self-selection, due to the stock-sample nature of our data set—are discussed in the Empirical Strategy section.

Duration Model

Immigrants in our benchmark specification enter the analysis as unemployed, and a failure is defined as the transition from unemployment to employment. Unemployment spells that are not stopped by hiring are treated as right-censored. This corresponds to inactivity, return migration, or panel attrition, for instance.12

We rely on proportional hazard models, which assume that the hazard faced by an individual i, in response to individual’s characteristics, is multiplicatively proportional to a baseline hazard h0(m), faced by all individuals. Defining xim as a vector of covariates with subscripts i for individual and m for month, our main specification can be written as follows:
himTrusto,xim=homexpβ0+β1Trusto+βxxim,
1
where Trusto represents, for an individual i, Germans’ trust level toward his or her origin country o. A positive coefficient of interest, β1, means that a higher level of trust toward the origin country of a given individual is associated with a higher instantaneous probability of leaving unemployment (i.e., a reduced expected length of unemployment).13 Our benchmark specification takes into account year, quarter, and regional fixed effects. The full vector of individual characteristics xim includes age, sex, number of children, an interaction between number of children and sex, education, marital status, number of years since migration, and whether the unemployed individual received social assistance.14,15
Our empirical analysis uses two specifications to estimate these hazards. Each specification makes different assumptions about the shape of hazard over time. First, the semiparametric Cox proportional hazard model makes no assumption on time dependency. Thus, hazard can be an increasing, a decreasing, or a constant function of time. Second, we use the parametric Weibull model, which allows us to rewrite the hazard function as follows:16
himTrusto,xim=homexpβ0+β1Trusto+βxxim=pmp1expβ0+β1Trusto+βxxim,
2
where p is a parameter, estimated from the data, that models the time dependency of the hazard. If p > 1 (p < 1), then the hazard is an increasing (decreasing) function of time. Also, our model allows for multiple unemployment spells. To avoid time-dependency for the unemployment spells of the same individual, we correct the covariance matrix by clustering the errors at the individual level (Lin and Wei 1989).17

Empirical Strategy

Equation (2) does not account for origin-specific factors that might influence the exit rate of immigrants from unemployment. Indeed, an important concern arises from the fact that our analysis builds on a stock sample of migrants who chose to migrate to Germany despite the potential distrust they would face. Specifically, trust might influence the composition of the self-selected group of observed immigrants given that different labor market opportunities may lead immigrants from highly trusted origin countries to be drawn from a different part of the population than their counterparts from less-trusted origin countries.

To the extent that trust influences the distribution of wage offers faced by potential migrants, the standard Roy model (Roy 1951), applied to the analysis of the migration decision by Borjas (1991), predicts that when other determinants of individual earnings are held constant, immigrants from a less-trusted origin country will have lower average reservation wages than immigrants from highly trusted origin countries. This in turn leads to higher acceptance rates of job offers and shorter unemployment spells, implying a downward bias in the estimated coefficient for the variable Trusto. The self-selection of migrants regarding trust levels at destination entails that controlling for origin-specific effects is crucial.

To account for selection at the national level, we estimate a second equation that considers natives’ attitudes at the regional level. In particular, we compute the region-level variable Trustor for 15 origin countries o in 11 German regions r. Our estimated equation becomes
himTrustor,xim=pmp1expβ0+β1Trustor+βxxim.
3

This specification exploits the variation between origin-region pairs. It allows us to control for unobserved origin-specific factors, such as quality of education, productivity, and self-selection patterns by including origin fixed effects interacted with year fixed effects. Following earlier discussion, we therefore expect an increase in our coefficient of interest, β1.

A legitimate concern would be that similar self-selection effects might occur at the regional level, inducing a downward bias in the coefficient of Trustor. Indeed, from a theoretical point of view, self-selection at the regional level could yield a bias similar to self-selection at the origin level. This implies that because we cannot control for self-selection into German regions, our empirical analysis can determine only a conservative estimate of the effect of trust on unemployment spells. However, our data set allows us to compare migrants from the same origin country across regions. To the extent that selection on unobservable characteristics is correlated with selection on observable characteristics, looking at the distribution of immigrants’ observable characteristics may be informative with respect to the selection on unobservable characteristics. Large differences can give an indication on the magnitude of the bias in our empirical analysis due to self-selection at the regional level. For each origin group of immigrants, Fig. A6 in the online appendix presents observable characteristics when comparing regions with relatively high levels of trust to regions with relatively low levels of trust toward citizens of a given origin country. Importantly, no striking differences emerge, suggesting that immigrants in our sample do not systematically self-select into regions that express higher levels of trust toward citizens of their origin country. Although this is not perfectly informative about the distribution of unobserved characteristics, it strongly reduces the concerns regarding self-selection at the regional level and the extent of the bias induced by the latter.

Finally, the way we define right-censoring in our benchmark specification induces a dependent censoring mechanism that may also bias our results. Indeed, our benchmark specification considers only employment as a possible failure, and everything else that stops an unemployment spell, such as inactivity, return migration, or panel attrition, is considered right-censored. Because right-censoring in these cases is not randomly distributed across immigrants, we complement our analysis with additional estimates considering alternative definitions of entrance and/or failure. Specifically, we look at transitions from unemployment to inactivity and from employment to other statuses. This set of estimates gives us an idea of the direction of the bias induced by right-censoring due to inactivity. However, for other forms of right-censoring (such as return migration) that we are not able to properly observe, it can theoretically be expected that immigrants who decide to leave Germany are those facing the lowest levels of trust. Indeed, individuals suffering the most from low levels of trust may have greater incentives to leave Germany for another destination or to return to their origin country. This would imply that the remaining pool of immigrants in Germany is composed of those who have been able to mitigate the adverse effects of a lack of trust. Self-selection at the regional level may induce a downward bias in our estimated coefficient for Trust.18

Results

Benchmark Specification

Columns 1 and 2 in Table 3 report the effect of Germans’ trust levels toward the different countries of origin of migrants at the national level. These estimates include a full set of individual controls and several sets of fixed effects. Focusing on our variable of interest, the Cox estimate suggests that a higher level of natives’ trust toward a given origin country is weakly associated with a higher instantaneous exit probability from unemployment for immigrants from this country. On the contrary, the Weibull estimate in column 2 does not show any significant association between Trust and immigrants’ unemployment durations.19 Columns 3 and 4 show estimates of the effect of Trust at the regional level. The additional variability obtained from the differences observed between the 15 origin countries across the 11 regions has the advantage of improving the precision of the estimated parameters.20 These new estimates show a positive and significant effect of Trustor on the probability of leaving unemployment. The reported effect is statistically significant at the 5 % level.21

As discussed earlier, we are concerned that the estimated coefficient of our variable Trustor in these first estimates may reflect country-specific factors, such as different incentives to migrate, that determine the selection of migrants in the population of the country of origin; or the fact that different immigrants have faced different levels of education quality in their origin countries. Column 5 of Table 3 includes origin fixed effects interacted with year fixed effects, which absorb the effect of all time-varying and time-constant origin-specific characteristics affecting immigrants’ unemployment duration. The coefficient for trust becomes significant at the 1 % level, and its magnitude dramatically increases.22 This is in line with the theoretical intuition given by the Roy model: because lower trust levels reduce the expected gains from migration, immigrants from low-trusted origins self-select into the lower parts of the origin country reservation wage distribution. Not accounting for origin-specific factors therefore introduces a downward bias in the estimates, significantly reducing the estimated impact of natives’ attitudes on immigrants’ unemployment duration. In terms of magnitude, if, ceteris paribus, the level of trust that Germans associate with Turkish citizens increased by 150 % (reaching the level of trust associated with Austrian citizens), then the average unemployment duration of Turkish immigrants would be reduced by five months on average.

Figure 2 shows the predicted survival functions of these two groups. As expected, the survival function of Austrian immigrants is always above the survival function of Turkish immigrants regardless of unemployment duration. After seven months of unemployment, our model predicts that more than 85 % of unemployed Austrians should have found a job compared with less than 65 % of Turkish immigrants. Thus, the effect of natives’ attitudes is also economically significant.

Estimates in columns 6 and 7 of Table 3 address whether the adverse effects of lower levels of trust are mitigated by higher educational level or migration duration.23 Our results suggest that years since migration do not reduce the negative effects of trust on immigrants’ unemployment duration. Conversely, compared with individuals holding a high or low level of education, individuals with an intermediate level of education seem less affected by lower levels of trust as suggested by the interaction term between Intermediate education and Trustor. This result can be interpreted in light of the model proposed by Lang and Manove (2011), which assumes that education is used as a productivity signal by employers. The signal is less precise for discriminated groups, and the noise decreases as the level of education increases. Hence, firms put more weight on the education signal to assess worker productivity in jobs requiring low and medium education. This generates incentives for members of discriminated groups to overinvest in the education signal, driving a wedge between returns to ability and returns to education. The consequence is that individuals from low-trust countries with intermediate levels of education actually have lower ability and lower wages than individuals from high-trust countries with intermediate levels of education.24 To the extent that lower wages also reduce the reservation wage, we would expect individuals with medium-level education from more trustful origins to be less likely to exit unemployment compared with individuals with medium-level education from less-trustful origins.

Finally, in column 8 of Table 3, we introduce regional fixed effects interacted with year fixed effects that account for all the time-varying, regional unobserved characteristics that affect immigrants’ unemployment duration and do not vary across origin countries. Therefore, we ensure that our estimations are not biased by yearly heterogeneous dynamics in regional labor markets. Indeed, regions with more favorable labor markets may attract particular groups of immigrants. Still, the coefficient of Trustor remains highly positive and significant.

Additional Threats to Identification

A legitimate concern is that omitted variables could simultaneously influence natives’ attitudes toward immigrants and the opportunities for the foreign-born to find a job. Indeed, immigration rates can be correlated with natives’ attitudes toward a particular origin and promote (through networks) or deter (through competition) access to employment for immigrants from this country. Therefore, for estimates shown in the first three columns of Table 4, we include the logarithm of the annual stock of immigrants of each origin in each region. We also include the logarithm of the stock of unemployed residents at the regional level in order to better control for regional market dynamics that can affect the access of immigrants to the local labor market. These two variables are taken from the German Federal Statistical Office (2016). In both cases, regional fixed effects interacted with time fixed effects absorb the variation in the size of the native population over time at the regional level. Our analysis is constrained by the availability of unemployment data, which cover only the 1991–2014 period. The effect of natives’ attitudes on immigrants’ unemployment duration is robust to the introduction of these two control variables. The coefficient of Trustor remains positive and highly significant at the 1 % level. As expected, the coefficient of the logarithm of the stock of unemployed residents is negative and significant, confirming that lower hiring opportunities decrease the likelihood for all individuals of exiting unemployment. However, we find no significant effect of immigration rates on immigrants’ unemployment duration.

Another important source of concern is that variables capturing natives’ attitudes are also strongly correlated with linguistic distance. Thus, our results could reflect the fact that migrants for whom it is more costly to learn German also struggle more to find a job. To tackle this potential issue, we introduce categorical variables capturing self-reported fluency in speaking and writing German. The results, reported in columns 4 and 5 in Table 4, show that the command of German at the individual level does not drive our main results. As expected, immigrants with a lower ability to speak or write German experience longer unemployment spells.

Another source of concern is that our results may be driven by the fact that migrants from different origins specialize in specific sectors with varying labor market dynamics. We investigate this possibility by controlling for industry fixed effects; these estimates are shown in columns 1 and 2 in Table 5. Last industry and Next industry are, respectively, the last and the next available (nonmissing) sectors reported by immigrants for a given unemployment spell.25 The results suggest that our main findings are not due to the clustering of immigrants in specific sectors. The changes in coefficient are mainly due to the fact that the sample size is reduced because of missing values in the additional regressors. In the same way, we control for the last nonmissing—or zero—wage reported by the individual. Controlling for previous productivity levels, we still find a positive and significant effect (p < .05) of natives’ attitudes on the probability of exiting unemployment, as shown in column 3.

Finally, measuring natives’ attitudes at the regional level may exacerbate the concern that the ability of a given diaspora to perform in local labor markets could shape these attitudes. This concern is greatly mitigated by the fact that the questions in Eurobarometer surveys ask about trust toward citizens of a given country, not toward migrants in Germany. Still, in order to address the possibility of reverse causality, we rely on an instrumental variable (IV) approach: namely, the two-stage residual inclusion method (2SRI), which is widely used to address endogeneity issues in nonlinear models.26 We use a proxy for the cultural distance between each German region r and each origin country o as an instrument for Trustor. Indeed, Guiso et al. (2009) underlined that cultural distance is a strong determinant of trust given that individuals tend to have more trust in people who share their beliefs and values. The exclusion restriction of our IV strategy implies that conditional on the other covariates included in the regression, cultural distance has no impact on an individual’s probability of exiting unemployment other than through the discrimination channel. We obtain bilateral cultural distances using the World Value Surveys (2015), which explore values and human beliefs through individual questionnaires conducted in almost 100 countries across the world by asking individuals to express their views on a number of practices. We select individual views on homosexuality, abortion, divorce, and suicide as instruments. Hence, our identification strategy relies on the assumption that tolerance toward homosexuality, abortion, divorce, and suicide is very unlikely to influence the individual hazard rate of finding a job other than through cultural distance with natives. For each of these four dimensions, we define Sir and Sio as the share of individuals either living in German region r or in origin country o declaring that the ith dimension is justifiable. The variability of our instrument comes from differences in beliefs and values of origin countries as well as variation among individuals living in different German regions in their attitudes toward homosexuality, abortion, divorce, or suicide, for instance. Indeed, German regions are very heterogeneous with respect to cultural values and norms, such as family values (Bertram 2013; Bertram et al. 2013; Nauck and Bertram 1995; Silbereisen and von Eye 1999, for references). We exploit these discrepancies between German regions in order to obtain a bilateral measure of cultural distance between German regions and immigrants’ origin countries:
Disor=iSirSio2,
4
where i ∈ {Homosexuality; Abortion; Divorce; Suicide} is the vector of views on practices. We report the results of these estimates in Table 5. We replicate our main result in column 4, excluding Austrian, Greek, and Portuguese immigrants because their origin countries do not have data for this particular question. The coefficient of Trustor is still significant and not statistically different from the one reported in Table 3, column 5. In column 5 of Table 5, we introduce the residuals of the first-stage estimates regressing Trustor on Disor; the instrument is significant with the expected sign. An increase in the cultural distance between a given German region and a given origin country decreases the share of Germans expressing positive views toward immigrants from this country. Moreover, the coefficient of Trustor remains positive and highly significant. However, the first-stage residual capturing the determinants of Trustor, which are not captured by our instrument, is not significant. This strengthens the argument that the variability in German attitudes captured in our regressions is not shaped by the local labor market performance of immigrants. In other words, unobserved origin-region characteristics correlated with Trustor, such as the average labor market performance of migrants from a given origin country, are not driving individuals’ hazard rates of exiting unemployment.27

Alternative Failures and Entrances

So far, our study has considered only employment as a possible failure for unemployment. We now investigate whether trust may also affect additional labor market dimensions—notably, immigrants’ probability of slipping into inactivity. Indeed, immigrants facing lower levels of trust and therefore longer unemployment spells may be discouraged in their job search. Such a mechanism could induce them to stop searching for a job earlier than immigrants exposed to higher levels of trust, implying a downward bias in our benchmark estimates. We reestimate our benchmark model using inactivity as a new failure. Thus, any other form of failure, such as employment, is considered right-censored. The results are reported in Table 6, where column 1 replicates our benchmark specification. Column 2 shows that trust has a significant and negative impact on the probability of leaving unemployment for inactivity, which is in line with the interpretation that our baseline sample is a positively self-selected sample of immigrants who decided to keep searching for a job despite the level of trust they faced.

Another legitimate concern is that after finding employment, immigrants from less-trusted origin countries experience a higher probability of losing their job and returning to either unemployment or inactivity. We test this assumption in columns 3 and 4 of Table 6, where individuals enter the analysis as employed and where failures are (respectively) unemployment and inactivity. In both cases, we do not observe significant differences in immigrants’ probability of losing their job, regardless of the level of trust they face in the German labor market. These results shed additional light on the underlying mechanism, suggesting that natives’ attitudes are mostly relevant in the job search process where they affect the arrival rate of job offers. After immigrants are employed, individual performance seems to outweigh group characteristics, such as country of origin.

Robustness to the Labor Supply Channel

We further investigate whether longer unemployment spells are actually attributable to attitudes of immigrants themselves. In Table 7, we provide a complementary analysis exploring the relationship between immigrants’ perceived discrimination, the self-perception of German identity, and immigrants’ unemployment duration. Specifically, we use a variable that captures the answer to the following question: “How often have you experienced disadvantages in the last two years because of your origin?” The corresponding variable takes the value 0 if the answer is never, 1 if it is seldom, and 2 if it is often.28 We also look at how the feeling of belonging to the German nationality influences labor market integration. Indeed, respondents in the GSOEP are asked to reply to the following question: “How much do you feel like a German?” The German identity variable takes the value of 0 if the response is completely, 1 if the response is mostly, 2 if the response is in some respects, 3 if the response is barely, and 4 if the response is not at all. Perceived discrimination and unemployment duration certainly strongly affect each other, and these results should be interpreted with great caution. Still, the correlations presented here are interesting in that they complement the main results by exploring alternative explanations.

Columns 1–3 present the effect of perceived discrimination on immigrants’ unemployment duration. After controlling for potentially confounding factors through the inclusion of origin-year fixed effects in column 2, higher levels of perceived discrimination are negatively associated with the hazard rate (i.e., the expected length of unemployment) only for immigrants who declare having often been discriminated against because of their origin. However, in column 3, including perceived discrimination with Trustor does not statistically affect the coefficient of the latter. The results are less straightforward when it comes to German identity, as shown in columns 4–6. Although in the first specification, higher levels of German identity seem to be associated with shorter unemployment spells among immigrants, this effect does not remain significant in more sophisticated specifications. After we control for origin-specific unobserved heterogeneity, feeling more or less German is no longer associated with immigrants’ unemployment durations. This result is in line with the intuition that although immigrants may have a strong feeling of belonging to Germany, negative attitudes from natives may still hinder their entrance in the local labor market. As for perceived discrimination, German identity does not cancel the effect of Trust on immigrants’ unemployment durations, as shown in column 6.

In addition, we use two proxies for immigrants’ affiliation with Germany: intention to stay and holding German citizenship. Intention to stay is a categorical variable with four response options indicating the desire to stay in Germany for each respondent. Column 7 in Table 7 shows that our results are robust to the inclusion of this variable. In addition, it shows that beyond the effect of trust, immigrants who are more likely to remain in Germany also experience shorter unemployment spells. Finally, our results remain unchanged after the information on citizenship is included in the benchmark specification, as shown in column 8. The coefficient for Trustor remains positive and highly significant. Also, we do not find any significant difference in terms of probability of remaining unemployed for individuals with or without German citizenship.

Overall, these results suggest that the negative impact observed on immigrants’ unemployment durations is mainly related to natives’ attitudes (demand side) and cannot be attributed only to immigrants’ attitudes (supply side). This is confirmed by the last regression, shown in column 9, which includes all the previous controls simultaneously.

Conclusions

In the debate on the integration of immigrants in the labor market of a destination country, the role of natives’ attitudes has often been overlooked. In particular, varying attitudes across German regions toward immigrants from different countries of origin might contribute to explaining the heterogeneity observed in terms of immigrants’ labor market performance.

In this article, we investigate how natives’ attitudes relate to immigrants’ labor market outcomes. Our empirical analysis finds that positive German attitudes are associated with shorter spells of unemployment and inactivity for migrants. This result is particularly important given the large number of migrants from developing countries expected in Germany and other high-income countries in the coming years. It underlines the fact that the assimilation of foreigners at destination is the responsibility not only of newcomers but also of the native population. This effect is robust to different specifications and alternative definitions of unemployment.

Reducing negative attitudes in migrants’ host countries toward the foreign-born should be at the heart of integration policies because it affects returns to education and the incentives for immigrants to invest in human capital at destination. This aspect has been stressed as crucial in the assimilation process (Borjas 2014). Raising awareness on these issues has led policy-makers to introduce anti-discrimination policies with the goal of overcoming the negative effects of discrimination on immigrants’ labor market outcomes. As a matter of fact, Germany ranked 22nd of 38 in the 2014 Migration Integration Policy index, which measures efforts of to integrate immigrants in OECD countries.29 Also, if natives’ attitudes reflect cultural, historical, and political differences, then the main focus of policies focused on integration should be on public beliefs and resentment toward immigrants from different countries.

Acknowledgments

We thank Simone Bertoli, Jean-Louis Combes, Vianney Dequiedt, Pascale Phélinas, Anne Viallefont, Pedro Vicente, Ekrame Boubtane, and Ababacar Gueye; all participants at the 28th SOLE conference in Raleigh, North Carolina, United States and at the workshop Migration and the Labour Markets in Edinburgh, UK; as well as four anonymous referees for their helpful comments and suggestions. We also thank Herbert Brücker for providing us data on immigration and unemployment rates in Germany. The data set used in this article was made available to us by the German Institute for Economic Research (DIW), Berlin. All remaining errors are our own responsibility.

Notes

1

See Riach and Rich (2002), Zschirnt and Ruedin (2016), and Bertrand and Duflo (2017) for extensive reviews of the literature on field experiments on discrimination.

2

Ethnic penalties refer to a strong relationship between ethnicity and gaps in labor market outcomes (such as wages, job quality, or employment dynamics) that cannot be explained by demographic and other human capital variables.

3

In this context, performing refers to a choice rather than to productive ability. One can think of a level of effort required by the contract that can be only imperfectly observed by the employer. The worker then chooses to exert the level of effort that maximizes his or her utility, which can be below the effort required by the contract.

4

The contract might not be enforced for several reasons. For example, if effort cannot be perfectly observed, proving that the worker did not exert the level of effort stated in the contract might be impossible or too costly for the employer.

5

Reducing the arrival rate of job offers has two opposite effects. On the one hand, the unemployment duration of discriminated workers decreases because they become less picky and reduce their reservation wage. On the other hand, the lower expected number of opportunities to leave unemployment increases immigrants’ unemployment duration. Under relatively weak conditions, the latter effect dominates (van den Berg 1994).

6

Still, we also present suggestive correlations between trust and individuals’ annual labor earnings in the online appendix, Table A6.

7

We use the version v.32, accessible at 10.5684/soep.v32 (Goebel et al. 2018).

8

Two difficulties arise from GSOEP calendar data. First, individuals may report several activities for the same month, which implies overlapping spells. Second, the number of activities has marginally changed over time to include more categories. We provide a detailed explanation of how we address these two problems in the online appendix.

9

Before 1995, Eurobarometer surveys were exclusively administered to German citizens. From 1995 onward, they were also administered to EU citizens. For these waves, we drop all observations from respondents who are not German citizens.

10

One might expect that the opinion of German citizens with a migration background could induce a bias in our measure of trust. The potential influence of this on our results is limited because only 12 % of Germans had a migration background in 2017, and this share was even lower in the decades that constitute our period of analysis. Moreover, because some employers are also likely to have a migration background, a measure including German citizens with migration background is probably closer to the level of trust that immigrants actually face in the labor market.

11

We suspect unemployment spells above this threshold to be potentially artificial (due to early retirement for example). Nonetheless, such instances correspond to less than 1 % of the total observations, and all results are robust to the inclusion of unemployment spells above this threshold.

12

Interruptions of unemployment spells (for training or maternity leave for instance) are considered as right-censored spells. Table A9 in the online appendix shows that our results are robust when recoding these short leaves as unemployment.

13

Our tables report coefficients rather than exponentiated coefficients (hazard ratios). The difference is that coefficients must be compared with 0 instead of 1.

14

Age and years since migration are defined at the beginning of the unemployment spell in order to avoid the collinear variation of these two variables with duration. Other variables, such as marital status, assistance, and number of children, are updated yearly. All covariate definitions and descriptive statistics are reported in Tables A1 and A3, respectively, in the online appendix.

15

Our main conclusions are robust to stratification by gender as well as to separating samples for men and women. These results are reported in Table A14 in the online appendix.

16

The Weibull model is preferable to the gamma, log-logistic, log-normal, and exponential models because of its lower AIC and BIC criteria.

17

Successive failures are assumed to be unordered and of the same type. In the baseline sample, 57 % experienced only one unemployment spell. Our results are robust to clustering at the origin-region level, as reported in Table A8 in the online appendix. Table A12 in the online appendix shows that our results are robust to controlling for past unemployment history.

18

Consistent with this mechanism, Table A11 in the online appendix shows that removing right-censored spells from the analysis leads to a dramatic increase in our estimated coefficient. Our results remain robust to the exclusion of left-censored spells.

19

The estimated shape parameter ln(ρ) in Weibull regressions is significantly positive, which means that the probability for immigrants of finding a job increases with time in unemployment.

20

We actually use 112 of the 165 possible origin-region pairs because migrants from some countries are not observed in all regions.

21

Unlike findings by Uhlendorff and Zimmermann (2014), our results are not exclusively driven by Turkish immigrants, the largest group of immigrants in Germany. The results with Trust and Trustor are robust to estimates excluding Turkish immigrants (available upon request).

22

Karlson et al. (2012) showed that unlike in linear models, the changes in coefficient of the variable of interest cannot be straightforwardly attributed to the inclusion of confounding variables. We take advantage of the fact that with their accelerated failure-time (AFT) metric, Weibull models can be seen as regular linear regression models, albeit with extreme values of residual terms. AFT models no longer model hazards (as parametric proportional hazard Weibull models do) but instead model the logarithm of the duration. Thus, using coefficients from Weibull in the AFT metric allows us to compare coefficients between models with different covariates. The results are perfectly equivalent in the two metrics given that βAFT = −βPH / ρ. The results with the AFT metric are available upon request.

23

Our results are also robust to the inclusion of years of education abroad. These results are reported in Table A13 in the online appendix. All our results with the education variable are also robust to the inclusion of more detailed categories, with separate groups for degrees specific to the German education system, such as vocational degrees.

24

The model also predicts that individuals with low and high levels of education have similar levels of ability regardless of the level of discrimination they face.

25

These two variables are computed irrespective of whether the immigrants had a long period of unemployment or inactivity. By definition, immigrants without prior work experience or first recorded spells are excluded from this analysis. The number of individuals in the sample is therefore reduced by roughly 30 % compared with our benchmark specification.

26

The 2SRI estimator corrects for the inconsistency of the estimated parameters obtained with the two-stage least squares (2SLS) method applied to nonlinear models. Although the 2SLS and the 2SRI share the same first-stage equation, the latter does not replace the endogenous variable by its predicted value but instead includes the first-stage residuals as additional regressors (Terza et al. 2008).

27

When included in our main specification as the explanatory variable, our instrument is not significant, supporting the validity of the exclusion restriction.

28

Figure A7 in the online appendix shows a strong and negative correlation between the variable Trust and the perceived discrimination variable. The correlation coefficient between the two variables is –.17 and is statistically significant at the 1 % level.

29

The index can be found online at http://www.mipex.eu/anti-discrimination.

References

Akerlof, G. (
1976
).
The economics of caste and of the rat race and other woeful tales
.
Quarterly Journal of Economics
,
94
,
599
617
.
Aldashev, A., Gernandt, J., & Thomsen, S. L. (
2009
).
Language usage, participation, employment and earnings: Evidence for foreigners in West Germany with multiple sources of selection
.
Labour Economics
,
16
,
330
341
.
Algan, Y., & Cahuc, P. (
2010
).
Inherited trust and growth
.
American Economic Review
,
100
,
2060
2092
.
Arrow, K. (
1973
).
The theory of discrimination
. In Ashenfelter, O., & Rees, A. (Eds.),
Discrimination in labor markets
(pp.
3
33
).
Princeton, NJ
:
Princeton University Press
.
Ashraf, N., Bohnet, I., & Piankov, N. (
2006
).
Decomposing trust and trustworthiness
.
Experimental Economics
,
9
,
193
208
.
Becker, G. S. (
1957
).
The economics of discrimination
.
Chicago, IL
:
University of Chicago Press
.
Bertram, H. (
2013
).
Das Individuum und seine familie: Lebensformen, familienbeziehungen und lebensereignisse im erwachsenenalter
(Deutsches Jugendinstitut Familien-Survey Band 4) [The individual and his family: Life forms, family relationships and life events in adulthood (German Youth Institute Family Survey Vol. 4)].
Berlin, Germany
:
Springer-Verlag
.
Bertram, H., Bayer, H., & Bauereiß, R. (
2013
).
Familien-atlas: Lebenslagen und regionen in Deutschland: Karten und zahlen
[Family atlas: Life situations and regions in Germany: Maps and numbers].
Berlin, Germany
:
Springer-Verlag
.
Bertrand, M., & Duflo, E. (
2017
).
Field experiments on discrimination
. In A. V. Banerjee & E. Duflo (Eds.),
Handbook of economic field experiments
(Vol.
1
, pp.
309
393
).
Amsterdam, the Netherlands
:
North-Holland
.
Bertrand, M., & Mullainathan, S. (
2004
).
Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination
.
American Economic Review
,
94
,
991
1013
.
Bisin, A., Patacchini, E., Verdier, T., & Zenou, Y. (
2011
).
Ethnic identity and labour market outcomes of immigrants in Europe
.
Economic Policy
,
26
,
57
92
.
Bohnet, I., Frey, B. S., & Huck, S. (
2000
).
More order with less law: On contract enforcement, trust and crowding
.
American Political Science Review
,
95
,
131
144
.
Borjas, G. J. (
1991
).
Immigration and self-selection
. In J. M. Abowd & R. B. Freeman (Eds.),
Immigration, trade, and the labor market
(pp.
29
76
).
Chicago, IL
:
University of Chicago Press
.
Borjas, G. J. (
2014
).
Immigration economics
.
Cambridge, MA
:
Harvard University Press
.
Casey, T., & Dustmann, C. (
2010
).
Immigrants’ identity, economic outcomes and the transmission of identity across generations
.
Economic Journal
,
120
,
F31
F51
.
Cettolin, E., & Suetens, S. (
2018
).
Return on trust is lower for immigrants
.
Economic Journal
, Advance online publication. https://doi.org/10.1111/ecoj.12629.
Chiswick, B. R. (
1978
).
The effect of Americanization on the earnings of foreign-born men
.
Journal of Political Economy
,
86
,
897
921
.
Constant, A., & Massey, D. S. (
2005
).
Labor market segmentation and the earnings of German guestworkers
.
Population Research and Policy Review
,
24
,
489
512
.
Couch, K. A., & Fairlie, R. (
2010
).
Last hired, first fired? Black-white unemployment and the business cycle
.
Demography
,
47
,
227
247
.
Destatis
. (
2016
,
July
14
).
Immigration and net immigration peaked in 2015
[Press release]. Retrieved from https://www.destatis.de/EN/PressServices/Press/pr/2016/07/PE16_246_12421
Fafchamps, M. (
2006
).
Development and social capital
.
Journal of Development Studies
,
42
,
1180
1198
.
Foddy, M., Platow, M. J., & Yamagishi, T. (
2009
).
Group-based trust in strangers: The role of stereotypes and expectations
.
Psychological Science
,
20
,
419
422
.
Frijters, P., Shields, M. A., & Wheatley Price, S. (
2005
).
Job search methods and their success: A comparison of immigrants and natives in the U.K
.
Economic Journal
,
115
,
F359
F376
.
Fullin, G. (
2011
).
Unemployment trap or high job turnover? Ethnic penalties and labor market transitions in Italy
.
International Journal of Comparative Sociology
,
52
,
284
305
.
Gambetta, D. (
2000
).
Can we trust trust
. In Gambetta, D. (Ed.),
Trust: Making and breaking cooperative relations
(pp.
213
237
).
Oxford, UK
:
Department of Sociology, University of Oxford
.
German Federal Statistical Office
. (
2016
).
Regional statistical database
.
Wiesbaden, Germany
:
Statistisches Bundesamt
.
Goebel, J., Grabka, M. M., Liebig, S., Kroh, M., Richter, D., Schröder, C., & Schupp, J. (
2019
).
The German Socio-Economic Panel (SOEP)
.
Jahrbücher für Nationalökonomie und Statistik/Journal of Economics and Statistics
. Advance online publication. https://doi.org/10.1515/jbnst-2018-0022
Göran, P., & Hägg, T. (
1994
).
The economics of trust, trust-sensitive contracts, and regulation
.
International Review of Law and Economics
,
14
,
437
451
.
Guiso, L., Sapienza, P., & Zingales, L. (
2009
).
Cultural biases in economic exchange?
.
Quarterly Journal of Economics
,
124
,
1095
1131
.
Karlson, K. B., Holm, A., & Breen, R. (
2012
).
Comparing regression coefficients between same-sample nested models using logit and probit: A new method
.
Sociological Methodology
,
42
,
286
313
.
Knack, S., & Keefer, P. (
1997
).
Does social capital have an economic payoff? A cross-country investigation
.
Quarterly Journal of Economics
,
112
,
1251
1288
.
Kogan, I. (
2004
).
Last hired, first fired? The unemployment dynamics of male immigrants in Germany
.
European Sociological Review
,
20
,
445
461
.
Kogan, I. (
2011
).
New immigrants—Old disadvantage patterns? Labour market integration of recent immigrants into Germany
.
International Migration
,
49
(
1
),
91
117
.
Lang, K., & Lehmann, J. Y. (
2012
).
Racial discrimination in the labor market: Theory and empirics
.
Journal of Economic Literature
,
50
,
959
1006
.
Lang, K., & Manove, M. (
2011
).
Education and labor market discrimination
.
American Economic Review
,
101
,
1467
1496
.
Leopold, L., Leopold, T., & Lechner, C. M. (
2017
).
Do immigrants suffer more from job loss? Unemployment and subjective well-being in Germany
.
Demography
,
54
,
231
257
.
Lin, D. Y., & Wei, L. J. (
1989
).
The robust inference for Cox proportional hazards model
.
Journal of the American Statistical Association
,
84
,
1047
1078
.
Nauck, B., & Bertram, H. (Eds.). (
1995
).
Kinder in Deutschland: Lebensverhältnisse von kindern im regionalvergleich (Deutsches Jugendinstitut Familien-Survey Band 5)
[Children in Germany: Living conditions of children in a regional comparison (German Youth Institute Family Survey Vol. 5)].
Opladen, Germany
:
Leske + Budrich
.
Nunn, N., & Wantchekon, L. (
2011
).
The slave trade and the origins of mistrust in Africa
.
American Economic Review
,
101
,
3221
3252
.
Phelps, E. S. (
1972
).
The statistical theory of racism and sexism
.
American Economic Review
,
62
,
659
661
.
Platow, M. J., Foddy, M., Yamagishi, T., Lim, L., & Chow, A. (
2012
).
Two experimental tests of trust in in-group strangers: The moderating role of common knowledge of group membership
.
European Journal of Social Psychology
,
42
,
30
35
.
Riach, P. A., & Rich, J. (
2002
).
Field experiments of discrimination in the market place
.
Economic Journal
,
112
,
F480
F518
.
Rosén, A. (
1997
).
An equilibrium search-matching model of discrimination
.
European Economic Review
,
41
,
1589
1613
.
Roy, A. D. (
1951
).
Some thoughts on the distribution of earnings
.
Oxford Economic Papers
,
3
,
135
146
.
Sapienza, P., Toldra-Simats, A., & Zingales, L. (
2013
).
Understanding trust
.
Economic Journal
,
123
,
1331
1332
.
Silbereisen, R. K., & von Eye, A. (Eds.). (
1999
).
Growing up in times of social change
(Vol.
7
).
Berlin, Germany
:
Walter de Gruyter
.
Tanis, M., & Postmes, T. (
2005
).
A social identity approach to trust: Interpersonal perception, group membership and trusting behavior
.
European Journal of Social Psychology
,
35
,
413
424
.
Terza, J. V., Bazu, A., & Rathouz, P. (
2008
).
A two-stage residual inclusion estimation: Addressing endogeneity in health econometric modeling
.
Journal of Health Economics
,
27
,
531
543
.
Uhlendorff, A., & Zimmermann, K. F. (
2014
).
Unemployment dynamics among migrants and natives
.
Economica
,
81
,
348
367
.
van den Berg, G. J. (
1994
).
The effects of changes of the job offer arrival rate on the duration of unemployment
.
Journal of Labor Economics
,
12
,
478
498
.
World Value Surveys
. (
2015
).
World Value Survey 1981–2014 longitudinal aggregate v.20150418, 2015
.
Madrid, Spain
:
JDSystems Data Acrchive [aggregate file producer]
. Retrieved from www.worldvaluessurvey.org
Zschirnt, E., & Ruedin, D. (
2016
).
Ethnic discrimination in hiring decisions: A meta-analysis of correspondence tests 1990–2015
.
Journal of Ethnic and Migration Studies
,
42
,
1115
1134
.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary data