Measuring Hiring Discrimination

How economists use fake job applications to test whether employers treat candidates differently based on name, caste, religion or gender.

Discrimination in hiring is hard to prove: employers rarely say why they rejected someone. Economists developed a clever method to measure it: correspondence studies, also called audit studies.

How they work

Researchers send fictitious job applications that are identical in qualifications and experience, but differ in one characteristic, such as a name that signals caste, religion, race or gender. They then measure callback rates: how often each type of applicant is invited for an interview.

If equally qualified applicants get different callback rates, the difference points to discrimination.

Famous studies

In the United States, economists Marianne Bertrand and Sendhil Mullainathan sent about 5,000 resumes in 2001 and 2002 with names common among white or Black Americans. Applicants with white-sounding names received about 50 percent more callbacks.
In India, economists Sukhadeo Thorat and Paul Attewell sent applications to private sector jobs in 2005 and 2006 with names indicating high-caste Hindu, Dalit or Muslim backgrounds. Equally qualified Dalit and Muslim applicants were significantly less likely to be called for interviews.
Studies in Europe have found similar effects for immigrants and ethnic minorities.

Types of discrimination

Economists distinguish:

Taste-based discrimination: employers simply prefer certain groups, as analysed by Gary Becker in 1957.
Statistical discrimination: employers use group averages to guess individual traits when they lack information.

Both harm qualified individuals and waste talent.

Limits of the method

Studies measure only the first step, getting an interview.
Names may signal other things, such as social class.
Online platforms and AI hiring tools raise new questions, since algorithms trained on past data can repeat past biases.

Policy responses

Anti-discrimination laws.
Blind recruitment, removing names from initial screening.
Diversity monitoring and targets.

Two identical applications

Two applications for an accounting job list the same degree, marks and internships. Only the names differ. One applicant gets a call; the other does not. Repeated across hundreds of employers, a consistent gap reveals discrimination that would otherwise be invisible.

Thinking discrimination can't be measured

Correspondence studies provide strong, direct evidence by comparing identical applications that differ only in group identity.

Key takeaways

Correspondence studies send identical fake applications with different names.
A U.S. study found white-sounding names got about 50 percent more callbacks.
An Indian study found Dalit and Muslim applicants were less likely to be called.
Economists distinguish taste-based and statistical discrimination.