Fixed-Term Work Contracts and Anti-Immigration Attitudes. A Novel Test of Ethnic Competition Theory

Whether labor market competition is shaping anti-immigration attitudes is a contentious issue. We conduct a novel test of ethnic competition theory by comparing the attitudes towards immigration of workers with fixed-term contracts to those with permanent jobs in Europe. Fixed-term contract workers are particularly at risk of competition as they have to compete for jobs in the foreseeable future. In the first step of our investigation, we analyze cross-sectional data (ESS, 2002–18) from 18 Western European countries. We find that—contrary to our expectation—fixed-term workers are less anti-immigration. The effect is substantively small. In the second step, we use a fixed-effects design with longitudinal data from the German Socio-Economic Panel (GSOEP, 1999–2015) to rule out time-constant unobserved heterogeneity. We find that transitioning from a fixed- to a permanent contract does not affect anti-immigration attitudes. Our combined results thus add to the evidence against labor market competition as an explanation of anti-immigrant attitudes.

Additional control variables and sample stratification Table A3 contains additional control variables, namely the number of working hours, marital status, household income, and area of residence, which are not strictly exogenous and might, for some assumed theoretical models, lead to biased estimates of the fixed-term contract-attitudes association. We therefore left them out of the models in the main text. However, in other assumed theoretical models, these variables may confound the fixed-term contract-attitudes association. In Table A3 we show that including these variables in the regression equations from Table 2 leads to substantively similar findings. Table A4 stratifies the sample in different ways to gauge the robustness of findings. The first two models are stratified by sex: for both men and women, fixed-term contracts are associated with somewhat more positive attitudes towards migrants, and there are no differences between men and women in the size of the association (χ 2 (1) = .31, p = .58).
The next three models in Table A4 compare participants in different partnership constellations: those who do not have a partner, single-earning members of couples, and members of dual-earner couples. Arguably, members of dual-earner couples are least likely to be affected by a potential job loss or job downgrading at the end of a fixed-term contract and consequently, contract status should have the smallest effect for this group. Conversely, single-earners in couples should be most financially at risk at the end of a fixed-term contract. Results show, that for all groups, workers on fixed-term contracts are less opposed to immigration than those on permanent contracts, though the relation is not significantly different for single-earners. The differences in the fixed-term contract-attitudes association between partnership status are not significant (χ 2 (2) = 3.4, p = .18).
The last column in Table A4 shows the fixed-term contract-attitudes association for the 'most likely' case, being low-educated, single-earners in working-class occupations. As job availability for this group is generally low, the possibility that migrants can take up their job comparatively high, and the financial consequences of non-renewal/not finding a new position comparatively large, one would expect a particularly pronounced relation 6 between contract type and anti-immigration attitudes. However as can be seen form the Table, the coefficient is negative, small and not significantly different from zero. Table A3: Anti-immigrant attitude regressed on contract type and control variables, OLS regression, unstandardized coefficients including controls for marital status, working hours, household income, and area of residence (standard errors in parentheses).

Additional comparison groups on the labor market
Analyses shown in the main text compare those on a fixed-term contract to those with a permanent contract. In this section, we compare these key groups with other groups on the labor market, namely those without a work contract, the unemployed-distinguishing by prior contract status-, and the self-employed. Workers without a contract are more opposed to immigration than those on a fixed-term or a permanent contract, although the difference with the latter is not significant. Workers without a contract are a heterogeneous group consisting of workers in the informal economy, helping family members, and civil servants with high job security. Consequently, the presumed exposure to labor market competition with immigrants varies within the no-contract group, making it difficult to interpret this result as evidence for or against ethnic competition theory. Figure A2, the unemployed are more opposed to immigration than those on fixed-term or permanent contracts. This finding is in line with the expectations of ethnic competition theory. Interestingly, the pattern across (prior) contract status-permanent, fixed or none-is similar to that of those currently employed. The self-employed are also more opposed to immigrants than those on fixed-term contracts. Similar to workers without a contract, this group is heterogeneous in its exposure to competition with immigrants, making it difficult to interpret this difference against the backdrop of ethnic competition theory. Predicted anti-immigration attitude Figure A2: Differences in anti-immigration attitudes between labor market groups Source: European Social Survey (2002( -18, ESS, 2002( , 2004( , 2006( , 2008( , 2010( , 2012( , 2014( , 2016( , 2018, own calculations. Notes: Vertical gray line indicates predicted antiimmigration attitude of those with a permanent contract. Error spikes denote 95 per cent confidence intervals. Based on a model accounting for all control variables of Model (2) of Table 2.

As shown in
No changes in the fixed-term contract-anti-immigration attitude association over time One might assume that there are changes in the fixed-term contract-antiimmigration attitude association over time: a fixed-term contract in the early 2000's might produce a very different effect than in the 2010's, since unemployment rates declined substantially over this period. To test this, we re-estimated Model (2) of Table 2 with an interaction between survey round and fixed-term contract. Results are shown in Figure A3, which reveals that there is no difference in the size of the gap in anti-immigrant attitude between fixed-term and permanent contracts.  Figure A3: Fixed-term gap in anti-immigration attitude is relatively constant over time Source: European Social Survey (2002-18, ESS, 2002, 2004, 2006, 2008, 2010, 2012, 2014, 2016, 2018, own calculations. Note: Error bands denote 95 per cent confidence intervals.

Gender and anti-immigration attitude in the ESS
While women are generally known to have a lower anti-immigration attitude than men, Table 2 shows an unexpected positive coefficient for the gender variable, suggesting that women have a stronger anti-immigration attitude than men. Generally, it can be misleading to interpret the coefficients of control variables in multivariable models, as they will be biased estimates of the causal effects (also known as the ' Table 2 fallacy, ' Westreich and Greenland, 2013).
To demonstrate that the direction of the gender association is not because of a coding error, Figure A4 shows the size and direction of the association between gender and anti-immigration attitudes for different model specifications. Model M1 shows the expected gender difference: when controlling only for country and survey round, women have a lower anti-immigration attitude than men. M2 shows that this is by virtue of women's generally higher education: When comparing men and women of the same educational level, we find no gender differences in anti-immigration attitude. This remains true when we add the remaining control variables from Table 2 to the equation (M3). Once we start comparing men and women in the same occupational group (M4), we see that women have a higher anti-immigrant attitude than men in the same occupational group. This holds true irrespective of the control variables (M5). Gender difference in anti-immigrant attitude Figure A4: Gender coefficient for anti-immigration attitude in five different model specifications Source: European Social Survey (2002-18, ESS, 2002, 2004, 2006, 2008, 2010, 2012, 2014, 2016, 2018, own calculations. Note: Error bars denote 95 per cent confidence intervals. Source: German Socio-Economic Panel (GSOEP, 1999-2015, Schupp et al., 2017, own calculations. Notes: FE: Fixed effects. * p <0.05, ** p <0.01, *** p <0.001 (two-tailed). All models include year fixed effects (not shown). Predictions based on Models (2) and (4) shown in Figure 4 in the main text. Source: German Socio-Economic Panel (GSOEP, 1999-2015, Schupp et al., 2017, own calculations. Notes: FE: Fixed effects. * p <0.05, ** p <0.01, *** p <0.001 (two-tailed). All models include year fixed effects (not shown). Predictions based on Models (2) and (4) shown in Figure 4 in the main text.

Additional comparison groups on the labor market
Analyses shown in the main text compare those on a fixed-term contract to those with a permanent contract. Here, we compare those key groups with other groups on the labor market, namely those without a work contract, the unemployed, and the self-employed. Those without a work contract are a heterogeneous group consisting of workers in the informal economy, helping family members, civil servants with very high job security, and workers who are protected by the relatively stringent labor legislation in Germany. Figure A5 shows that workers without a contract are slightly more concerned about immigration than those on permanent contracts, yet this difference is not significantly different from zero. Self-employed are very slightly less opposed than those on permanent contracts, but the difference is again not significant. As both workers without contracts and the self-employed are very heterogenous groups, it is unclear whether they on the whole face more or less competition than people on fixed-term and permanent contracts. This raises doubt about whether whether the coefficients vs. permanent contract for these groups should be interpreted as support for ethnic competition theory or not. Predicted concern about immigration Figure A5: Differences in concern about immigration between labor market groups Source: German Socio-Economic Panel (GSOEP, 1999-2015, Schupp et al., 2017, own calculations. Notes: Vertical gray line indicates the predicted concern about immigration for workers with permanent contracts. Error spikes denote 95 per cent confidence intervals. Based on a fixed effects model accounting for all control variables of Model (2) of Table 2.