Learning to be Unbiased: Evidence from the French Asylum Office

What determines whether some asylum seekers are granted refugee status while others are rejected? I draw upon archival records from a representative sample of 4,000 asylum applications filed in France between 1976 and 2016 to provide new evidence on the determinants of asylum decisions. Comparing accepted applicants with those who were rejected, I find that Muslim applicants are 30 percent less likely to be granted asylum than otherwise similar Christian applicants. In addition, linking archival records to detailed administrative data, I show that bureaucrats at the French asylum office initially discriminate against Muslims but stop after about a year on the job. Assessing potential mechanisms of discrimination, I do not find support for the claim that discrimination is driven by bureaucrats' preferences or ideology. Instead, I argue that discrimination occurs because bureaucrats underestimate the probability that Muslims are persecuted. This novel finding has direct implications for strategies to curtail discrimination in courtrooms and administrations.