We study inaccurate beliefs as a source of discrimination. Economists typically characterize discrimination as stemming from a taste-based (preference) or accurate statistical (belief-based) source. Although individuals may have inaccurate beliefs about how relevant characteristics (e.g., productivity, signals) are correlated with group identity, fewer than 7% of empirical discrimination papers in economics consider the possibility of such inaccurate statistical discrimination. Using theory and a labor market experiment, we show that failing to account for inaccurate beliefs leads to a misclassification of source. We outline three methods to identify source: varying observed signals, belief elicitation, and an intervention to target inaccurate beliefs.

Equal access to voting is a core feature of democratic government. Using data from hundreds of thousands of smartphone users, we quantify a racial disparity in voting wait times across a nationwide sample of polling places during the 2016 U.S. presidential election. Relative to entirely white neighborhoods, residents of entirely black neighborhoods waited 29% longer to vote and were 74% more likely to spend more than thirty minutes at their polling place. This disparity holds when comparing predominantly white and black polling places within the same states and counties and survives numerous robustness and placebo tests. We shed light on the mechanism for these results and discuss how geospatial data can be an effective tool to measure and monitor these disparities going forward.