Ethnic Discrimination in Germany s Labour Market. A Field Experiment. By Leo Kaas and Christian Manger. Experimental & Behavorial Economics Discrimination in labor markets
Agenda 1. Introduction 2. Experiment Design 3. Results 3.1 Results 1: Callback based on ethnicity 3.2 Results 2: Callback based on ethnicity and application type 3.3 Results 3: Politeness of reaction 3.4 Results 4: Reaction time 4. Conclusion Seite 2
Introduction Experiment conducted by Leo Kaas and Christian Manger in 2007 and 2008 Motivation: Study ethnic discrimination in Germany s labour market in field environment Seite 3
Experiment Design Applications for internships in business and economics sent to 528 different companies Each company receives two letters of application The applications are equal in regard of qualification, skills, age, but with slight differences in details Seite 4
Experiment Design One application is sent with a Turkish-sounding, the other with a Germansounding name. Observation of ethnic discrimination Type B comes with two references from former employers, including positive remarks about personal traits like commitment and affaibilty. The combination of type and name is randomly chosen. Observation of statistical discrimination due to the absence of direct information Seite 5
Result 1: Callback based on ethnicity Callback No Callback Both Only One Equal Treatment Discrimination Non-observation Sample size Standard χ 2 -test of H0: Unequal treatment is equally likely for both ethnicities. Seite 6
Result 1: Callback based on ethnicity Seite 7
Result 1: Callback based on ethnicity ** Significant at 5% A German-sounding name raises the chance of a callback by 14% on average The callbacks by a small company for an application with a German-sounding are three times those of one with a Turkish-sounding name Discrimination in some other observed categories significant at 10% level, most not significant Seite 8
Result 2: Callback based on ethnicity and application type Almost identical callback rates for type B application Statistically significant (5%) difference for type A application we cautiously interpret this as evidence for statistical discrimination (p.11) Seite 9
Result 3: Politeness of reaction Callback to German AND no reaction to Turkish: 28 No reaction to German AND callback to Turkish: 12 Statistically significant difference (5%) strongest form of discrimination: the firm has a vacant post, it shows interest in the German candidate and does not even answer the Turkish one. (p.13) Seite 10
Results 4: Reaction time Reaction time (measured in workdays) for applications with Turkish name slightly longer, but not statistically significant. Seite 11
Conclusion Ethnic discrimination exists in Germany s Labour Markets Although according to these results it s lower than in experiments conduct in other countries. Why? Focus on specific high skill segment of the labour market Highly qualified applicants No language effects Calls for further research Seite 12
Conclusion Recommendations to fight discrimination: Optimization of recruitment process (anonymized applications) Providing references and other information on an early stage of application Seite 13
Thank you for your attention! Any Questions? Seite 14
Backup Vacancy traits Type: internships for students in economics and business => Automatization Location: Germany Duration: Six weeks to six months Source: large internet job sites (e.g. monster.de, jobscout24.de) Timeframe: First wave in December 2007/January 2008, second wave in December 2008. Heterogenousity of internships Different divisions within the firm (human resources, marketing, finance, controlling) Different company size (mostly firms with 500+ employees. Reasons: large firms and banks are most relevant employers for business and economics, are more likely to post their vanacies on the big internet sites) Way more vacancies in West Germany than East Germany since most have headquarters in West Germany Seite 15
Backup Specific application design Similarities: Personal: Age of 21 or 22, second-year student in business and economics, born, raised and educated in Germany Documents: CV, high-school certificate, certified grades of the first university year CV: Graduation -> no military service -> summer job -> university Two part-time jobs Reasonable computer skills Languages: German (native), English (fluent), further (basic knowledge); no Turkish Quality differences: Type B: certificates for part-time jobs with personal positive statements (affability, commitment, capacity for teamwork, conscientiousness) Seite 16
Backup Specific application design Slight differences: School and university grades Born, raised and educated in different regions of Germany: Baden-Wuerttemberg, North Rhine- Westphalia University orientation (A more business oriented) Photos (though both fit native German or one with a migration background) Sector and division specific paragraph (basic requirements, interest) Seite 17
Backup - Applications Seite 18
Backup Name Design German: Tobias Hartmann, Dennis Langer. Both first and last name are among the 30 most common names for the birth years 1986-88 Turkish: Fatih Yildiz, Serkan Sezer. Both are typical for male descendants of Turkish immigration. Other requirements: No contradiction of common sense, not too stereotypical, don't exhibit other peculiarities (combination of Anatolian first name and Kurdish surname), not real names in studivz.de No check for social connotations Seite 19
Backup Application process design Creation of name-specific e-mail address and a mobile phone number No pick-up but polite voice mail answer Application via e-mail in merged PDF or via required form with same information as in PDF Two days time difference between first and second application Valid Callbacks: In the next four months, signaling interest (interview, direct job offer, leaving contact information). Date and time are noted. If further documents are requested, they get sent within 24 hours (not a valid callback yet) Seite 20
Backup Result 1: General Information Column (6): Net discrimination as (4) - (5). Thus, (1; no callback) is treated as non-observations (pro: firm was not considering hiring (already filled vacant spot); con: both applications reviewed and both not suitable). (3) is equal treatment. Basis for (7) Column (8): Includes (1) as equal treatment Seite 21
Backup Result 1: Tables Seite 22
Backup Result 1: Tables Seite 23
Backup Result 1: Tables Seite 24
Backup Result 1 Strong: All firms: discrimination (5%), 6 callbacks for every 15 (German) or 17 (Turkish) application; remarkably small in international comparison Small firms: Only 4 prefer the Turkish employee; possible explanation less-standardized recruitment process, more room for discrimination based on personal taste (in comparison to larger firms) Limited: East Germany: Twice as many prefer applications with German-sounding name over those with Turkishsounding ones. But too few vacancies(40) and callbacks (15) for truly significant result. Marketing: Weakly significant discrimination, not strong, probably based on large sample size (98) Human Resources: Same as in East Germany, too small sample size (49), thus not significant Counting rejection as observation: overall discrimination (10%, weakly significant) Seite 25
Backup Result 2: Table Seite 26
Backup Result 2 Almost identical callback rates for type B application (the one with the information on personal traits) Statistically significant (5%) difference for type A application Conclusion: Statistical discrimination (absence of information leads to the substitution of group average, which reveals prejudices on the ethnic background) Possibly due to other slight differences in the application, but they are minor and type B doesn't fare better than type A in general (applicants with German-sounding name got less callbacks under type B) Not due to region where the applicants lives in Germany Seite 27
Backup Result 3: Table Seite 28
Backup Result 3: Three categories (from most to least polite): Callback, rejection, no reaction German callback & Turkish no reaction: 28 Turkish callback & German rejection: 12 Difference significant (5%) -> "strongest form of discrimination Seite 29
Backup Result 4: Table Seite 30
Backup Result 5: Business cycle Two waves of applications: First in December of 2007/January of 2007 Financial crisis affected real economy during 2008 But no significant difference in callback rate (38.7% and 35.4%) or discrimination Seite 31
Backup - Conclusion Weak discrimination in comparison to experiments in other countries. Why? High-skill segment of labour market, can't afford to reject talents because of their ethnicity (in markets with more supply of applicants, personal taste can affect the choice more easily) Man of Turkish ethnicity was born and raised in Germany, no language effects (focus on second and third generation immigrants because they are the biggest minority in Germany) Good grades and interesting CVs (necessary to get a high enough number of callbacks in the first place). Mediocre grades would boost discrimination, but good grades from someone with Turkish-sounding name could produce an inconsistency with the employers' expectation about the applicant. Further research: Experiments in different sectors of the labour market Varying quality of applications (grades) Role of information provided in the application (differences in types of application) Seite 32
Backup - Conclusion Recommendations: Optimization of recruitment practices can help to provide equal opportunities for minority workers (e.g. anonymized applications stripped of names, age, adress, photos, etc.) Providing references and other information on an early stage of application (which is uncommon in many countries) Seite 33