Ethnic Discrimination on the Labor Market in Comparative Perspective
Theoretical background and objectives
Survey data are one way to study labour market disadvantages of immigrants. But they have the disadvantage that not all differences with natives can be explained away with the available variables. Hence, there is no way to determine with certainty whether the residual gaps are due to discrimination or to other unobserved variables. Audit and correspondence studies have become popular responses to this problem and have demonstrated for a wide range of ethnic groups and countries that discrimination occurs. So far studies have almost exclusively used a paired application design, in which two applications, one native and one from a selected minority group, are sent, which apart from cosmetic details differ only in the ethnicity of the applicant. Widespread as it may be, this design has the major disadvantage that it is diagnostic rather than analytic. It can demonstrate beyond reasonable doubt that discrimination occurs – at least for a selected ethnic group – but not whether taste or statistical mechanisms are behind discrimination, nor which characteristics of applicants – their race, religion, cultural or linguistic distance, or specific ethnicity – provoke discrimination. In this project that was started in late 2014 we want to move beyond these limitations by using an unpaired multiple-group, multiple-treatment design in which we vary racial phenotype, religion, as well as ethnicity. Native ethnics are compared to second generation applicants from 34 immigrant ethnic groups. For her dissertation, Ruta Yemane will implement a similar design in the USA in cooperation with Harvard University. The German study allows a direct measurement of racial discrimination because in Germany photographs are allowed or required in the application process. In the USA race will be indirectly signaled by names and ethnic language. The multiple-group design allows regression analyses testing for taste or statistical discrimination, for instance by relating callback rates to cultural distance to the countries of origin (using World Values Survey data) or to group educational and labour market status averages (e.g., using the German Mikrozensus).
In order to investigate the drivers of discrimination against second generation immigrant job applicants, we sent thousands of applications from fictitious persons to real job openings in eight professions all over Germany. Next to job applicants’ ethnicity (German or migration background in one out of 34 origin countries), phenotype (Asian, Black, White), and religious affiliation (none, Buddhist or Hindu, Christian, or Muslim), we varied several other characteristics of the applications, such as applicants’ gender, final grades, whether or not a reference letter was included, as well as information about applicants’ current contract. Our results confirm that employers discriminate against immigrant job applicants. The magnitude of discrimination, however, varies strongly between origin groups. Whereas employers do not discriminate against Western and Southern European and East Asian immigrants, other origin groups experience significant disadvantages. In addition, we observe substantial disadvantages for Black and Muslim job applicants. With respect to classic theories about the drivers of discrimination on the labor market, that is, taste-based and statistical discrimination, we find that the cultural distance between origin countries and Germany explains discrimination against different groups much better than productivity-related group characteristics, such as average levels of education. Consequently, our empirical findings are more supportive of taste-based discrimination than they are of statistical discrimination theories.