A Simple, Statistically Robust Test of Discrimination
Imagine auditing a bank for discrimination. If one group of borrowers repays their loans more than other groups, that suggests the bank held them to a more stringent standard, only lending to the most creditworthy applicants from that group. Such “outcome tests” are one of the most common empirical approaches for detecting double standards in lending, hiring, criminal justice, and other domains. Despite their simplicity, outcome tests suffer from well-known statistical issues and often miss discrimination or even mistake discrimination for favorable treatment. “Benchmark tests”-where, for example, researchers directly compare lending rates across different groups-are the most common alternative, but suffer from their own deep limitations. Here we introduce a robust outcome test, which considers both the rate at which decisions are made (benchmark tests) and how often those decisions succeed (outcome tests). Despite their individual fallibility, carefully combining benchmark and outcome tests retains the simplicity of the original approaches while yielding surprisingly strong statistical guarantees under a common non-parametric assumption that we show-both empirically and theoretically-likely holds approximately in real-world data. We illustrate our robust outcome test with 2.8 million police stops from across California, finding evidence of widespread racial discrimination, a pattern that the standard outcome test would miss.
Bio: Johann D. Gaebler is an assistant professor in the Technology Group in NYU Stern’s Department of Technology, Operations, and Statistics. He develops statistical and computational methods to improve organizational decisions and public policy, and to make AI systems safer and more trustworthy. His research spans responsible AI and alignment, algorithmic decision-making, and computational social science, building conceptual and practical tools to address issues in areas like education, hiring, criminal justice, elections, and news media. Johann studied statistics (Ph.D. and M.S.) and mathematics (A.B.) at Harvard University and was a Knight-Hennessy Scholar at Stanford University.