Immigration and Citizenship: Labor Market Discriminations

To understand how citizenship can help reduce hiring discrimination among first- and second-generation immigrants, we will conduct a large-scale correspondence experiment in France varying the profiles’ immigration status and origin. We will send 6,400 fictitious resumes to occupation openings in eight professions, carefully selected based on variations in labor-market tightness and the level of French required for the occupation

2024-12-02

2025-12-31

Callback by nationality ; country of origin ; perceived country of origin ; gender and
cross-tabulations

Callback: Whether the application received a callback (receiving a positive personalized phone or e-mail response by the employer, including requests for additional information or an invitation to interview). (See Pre-Analysis Plan for more details).

- Intensity of callbacks (number of callbacks)
– Callback on the intensive margin
– Signals (gender, origin, civic engagement, experience, level of French, explicit
mention of type of residence permit, citizenship priming, place of education (as
a proxy for place of birth)
– Occupation characteristics (level of French proficiency, labor market tightness,
share of foreigners, share of women)
– Firm characteristics (share of foreigners, share of women, size of the firm, age
of the firm)
– Characteristics of HR (gender and origin – subject to DPO agreement)

(See Pre-Analysis Plan for more details).

In this project, we run a correspondence audit study to test whether and the extent to which candidates are discriminated at the hiring stage based on their origins and citizenship. To do so, we will send 6,400 fictitious resumes, varying citizenship, and origins, to occupation openings in eight professions, carefully selected based on variations in labor-market tightness and the level of French required for the occupation.

Not available

We will send four fictitious applications to each vacancy. Randomizing the CV templates, sending order and signals. Randomization will be performed by a computer program. (See Pre-Analysis Plan for more details).

Application within each occupation. Randomisation takes place to create the 4 applications with which we apply to
each job vacancy and the assignment of these four to each job vacancy for a given occupation. For each of the 8 occupations, a single application is characterized by a profile (which is a combination of nationality, lastname origin and place of work), a combination of core signals (gender and type of the third hobby), specific signals that vary according to each profile (firstname origin, french mastery, residency rights), a template (layout of the CV), a firstname and a lastname, the fact that nationality is mentioned at the top of the application, the fact that this mention is primed, and specific wording according to residency rights. CV templates are randomly selected at the vacancy level as well as the order in which the applications are sent. Nationality is explicitly written for all of the applications sent with 3 of the 4 profiles. However, only 1/4 of applications associated to the first profile mention the nationality. The priming of the nationality signal will be included whatever the profile (and only if the nationality is explicitly mentioned for profile 1) in the half of the CV templates that include a short text in the CV (only 2 of 4 templates among authorize short sentences). There are three hobbies for each of the four applications. The first two hobbies are “template dependent” while the third CV varies according to the type of the third hobby (core signal). Firstnames and lastnames are randomly selected from separate lists of first names and surnames associated with the different origins. Finally, specific wording for residency rights is selected, if necessary, from three different formulations.

No

6400 CVs sent to 200 applications in 8 occupations.

6400 CVs

1600 CVs for Profile 1
1600 CVs for Profile 2
1600 CVs for Profile 3
1600 CVs for Profile 4

Based on a correspondence study in France, the average call-back rate among individuals with French-sounding names for low-skilled jobs is 33%. For comparisons between any two profiles, there will be 3200 applications in total, with 1600 applications from profile A, and 1600 applications from profile B. With a standard level α = 0.05 and power = 0.8, we can detect an effect of 4.7 percentage points. This translate to a Minimum Detectable Effect (MDE) of 0.14%.