When Bureaucratic Discretion Leads to Discrimination: Evidence from France
[Thesis]
Emeriau, Mathilde
Hainmueller, Jens
Stanford University
2019
122 p.
Ph.D.
Stanford University
2019
While economists have been studying discrimination on the labor market for the past fifty years, political scientists are only getting started in measuring the scope of discrimination taking place in courtrooms and administrations in democratic regimes around the world. Recent studies have documented variation in adjudication between decision makers, but we still know little about the mechanisms through which bureaucratic discretion, or the leeway that bureaucrats have in making decisions, results in discrimination, or unequal treatment of otherwise equal applicants. Yet understanding mechanisms of discrimination is a crucial first step to design interventions to reduce discrimination. One of the reasons we know so little about the conditions under which bureaucrats discriminate is the difficulty in getting micro-level administrative data. In this thesis, I exploit a new source of data which allows me to open the black box of bureaucratic decision-making. With the support of the Stanford Immigration Policy Lab, I was able to negotiate access to the administrative database and the archives of the French asylum office. Over the last five years, I digitized 4,000 asylum applications and read most of their administrative archives. Existing studies generally conclude that discrimination in administration in courtrooms and administration results from taste-based discrimination. In other words, judges and bureaucrats discriminate when they pursue personal objectives, like preferences, or political interests, instead of doing their jobs. In this thesis, I show that bureaucrats discriminate even when they are doing their job the best they can. In the first two papers, I examine two alternative mechanisms of discrimination. The first one I call stereotype-based discrimination. Bureaucrats discriminate because they have biased priors. It is similar to statistical discrimination but with one exception: priors are inaccurate, they are not based on actual data. I argue that this is happening at the French asylum office. I show using micro-level data, that French bureaucrats discriminate against Muslims in the attribution of refugee status but that they stop after a year. I argue that over time bureaucrats learn to update their biased priors. The second mechanism through which discretion leads to discrimination is through a priming effect. Using high-frequency administrative data, the second paper shows that in 2015 and 2016 bureaucrats at the French asylum office were more generous after a migrant shipwreck and less generous after a terrorist attacks but only for Syrians. I argue that these events affect bureaucrats by priming them to think of asylum seekers either as victims or as a threat. Moving away from the French asylum office, the third paper coauthored with David Laitin, directly speaks to the broader implications of bureaucratic discrimination for immigrant integration. This third paper establishes a connection between the fact that Muslims integrate more slowly than Christians to their experience of discrimination in France, and their mistrust in institutions like the Police.