Ethan O'Leary

I propose and test a novel mechanism through which discrimination persists: the misperception of statistical discrimination as taste-based discrimination. I first derive a theoretical model of self-confirming discrimination which arises in the job-application setting to demonstrate this mechanism. Correctly attributing the source of discrimination leads to multiple equilibria such that arriving at inequality represents a coordination error. However, misattributing discrimination to taste-based origins leads only to a discriminatory equilibrium and reduces the perceived return to deviating from discrimination-inducing application strategies. In a pre-registered laboratory experiment, I find that subjects who experience discrimination from an ambiguous origin overestimate the probability that this inequality is caused by taste-based discriminators. This misattribution directly leads to a higher probability that application strategies sustain statistical discrimination: a 1 percentage point increase in believed probability that discrimination is taste-based leads to a 0.35 (0.24) percentage point decrease (increase) in the probability that a subject in the discriminated (favoured) group applies to a job, perpetuating statistical discrimination. Moreover, misattribution leads to stronger perceptions of market unfairness and lower demand for affirmative action among the favoured group. This demonstrates that the misattribution of the source of discrimination can both directly and indirectly lead to sustained discrimination and inequality.

Trust—the belief in another individual or organization's intent to act in good faith—is a cornerstone of social prosperity. However, cultural norms significantly influence trust levels within and between groups, often leading to discriminatory behaviours that undermine cooperation. This paper investigates the impact of culture on trust using a novel experimental approach that isolates trust from confounding factors such as risk preferences and distributional concerns. The setting of the experiment is England — a culturally and economically divided nation. Focusing on self-identified Southerners and Northerners, I find evidence of a cultural trust gap. Specifically, while Southerners exhibit equal trust towards both groups, Northerners trust Southerners significantly less than their own group. By decomposing trust into its conditional and unconditional components, I show that this disparity is driven by unconditional trust: Northerners perceive Southerners as more selfish than their own culture.

We study whether workers anticipate gender discrimination by evaluators and how such anticipation distorts competitive choices in job applications. We conduct a large-scale, pre-registered online experiment in which approximately 9,000 workers choose between a non-competitive piece rate and a competitive payment scheme whose payout depends on being selected by an incentivised human evaluator. By randomly varying whether the evaluator can observe the worker's gender across a male-stereotyped and a female-stereotyped task, we isolate the causal effect of anticipated discrimination on labour supply. In the male-stereotyped task, gender revelation breaks the link between women's objective performance beliefs and their beliefs about being selected: conditional on the same win belief, women report significantly lower selection expectations when their gender is known. This distortion reduces female entry rates, with the strongest effects among above-average performers for whom competing would have been the correct decision. We complement the worker experiment with a manager pilot experiment that measures actual evaluator behaviour. Comparing worker expectations to evaluator decisions reveals a double mismatch: in the male-stereotyped task, managers show no discrimination against women, yet workers anticipate it; in the female-stereotyped task, female managers favour female workers, yet workers fail to anticipate this advantage. Our findings suggest that the welfare costs of gender gaps in job applications can arise from distorted beliefs rather than from discriminatory evaluator behaviour, pointing to information-based interventions as a promising policy lever.