Experimental Design Details
In the standard correspondence study, matched pairs of qualitatively identical job applications are sent to employers that have advertised a job opening. The only difference between the fictitious applications is the name of the applicant, which usually signals ethnicity or gender. The degree of discrimination in hiring is then quantified by calculating the difference in the callback rate (i.e., the fraction of invitations) to a job interview between the groups.
In order to test the interaction between gender discrimination (measured by the difference in call-backs between applications according to the gender of the candidate signaled by the first name) and all other dimensions considered relevant, we opted for the following experimental protocol.
The occupations selected for applications differ on the following 3 dimensions:
1. the level of qualification of the occupation (3 categories: low skilled / medium skilled (without management function) / high skilled (with management function);
2. the degree of feminisation of the occupation (3 categories: predominantly male / balanced / predominantly female);
3. the degree of labour market tightness of the occupation (2 categories: low labor market tightness / high labor market tightness).
Based on these categories, the study covers a total of 12 occupations: 4 low-skilled occupations, one of which has a high labor market tightness; 4 medium-skilled occupations (without managerial functions), one of which has a high labor market tightness; 4 high-skilled occupations (with managerial functions). The 10 occupations with low labor market tightness were chosen according to their degree of feminisation, by retaining in each qualification level a predominantly male occupation, a balanced one and a third predominantly female one. As for the occupations with high labor market tightness, they were chosen preferably from predominantly male and/or balanced occupations.
Applications sent in response to job offers of these 12 occupations are distinguished by age, gender(male/female) and origin(french/maghreb). Age is determined by work experience in three groups: 4 to 6 years of work experience (23 - 30 years); 14 to 16 years of work experience (33 - 40 years), 29 to 31 years of work experience (48 - 55 years).
Based on these variables we apply to each job offer with 4 distinguished identities which combine gender and origin in a factorial way (1: female/french, 2: male/french, 3: female/maghreb, 4: male/maghreb). To preserve the statistical power of the tests, we respond to each job offer with a fixed age range (all candidates are young in response to the first offer, older in response to the second offer, etc.) including only two age groups per occupation (young and middle for low skilled and middle and older for medium/high-skilled).
Given the large scale of the study (number of occupations, number of observations required, see below, and territorial coverage) we have chosen to not send out more than 4 applications per job offer due to the associated increase in costs and the higher risk of detection.
In order to take into account the possible differentiated effect of the content of CVs according to the gender of the candidates (and possibly their origin), we add signals to all 4 CVs sent in response to an offer (all CV's in one quartet will have the same signal). Given the well documented effect of maternity on gender discrimination we hence include three blocks of signals. The first of which crosses period of inactivity and presence of children; the second block crosses marital status and presence of children; the third block will contain CVs without additional signals. The final signals included are thereby: a: no signal, b: in a couple and 2 children, c: period of inactivity, d: in a couple, 2 children and period of inactivity, e: in a couple, f: single, g: in a couple and 2 children, h: single and 2 children).
A last dimension we include is the variation of gender discrimination with the social status of responding candidates, by varying french and maghrebin names based on their social status (advantaged/disadvantaged).
We further created 4 CV profiles (not including signals or identities) with respective experiences, education etc.
We also provide an ordered list of paired CV's to meet offers with detection risk. For these job offers, we draw from the ordered list of couples rather than the list of quartets. The ordered list of couples contains only identities with french origin and only profiles A and B. The first couple is (A1, B2), the second (A2, B1) then again (A1, B2), etc. Responses to offers with couples do not include signals.