Recently, police departments, legislators, media, and the public at large in the U.S. have increasingly been concerned about racial disparities in officers' issuing traffic tickets. Ascertaining the extent to which an observed disparity reflects racial bias is the crucial issue. First, we use a theoretical model which borrows features from the recent literature regarding racial bias in vehicle searches. In our model, motorists, picking the speed to travel at, take into account the probability of getting ticketed and the speed that the officer will cite, while officers maximize a benefit function generically increasing in the speed of ticketed drivers; this benefit function, however, is general enough to allow officers to give certain drivers a break by citing them at a lower speed than they were traveling. Empirically, we exploit the existence of a massive accumulation of speeding tickets at 10 m.p.h. over the speed limit to elicit officers' discretionary behavior and leniency. Surprisingly,about 30% of all ticketed drivers were cited for driving exactly at this particular speed. Using our novel measure of officers' leniency, we find that especially white and male officers are heavily engaged in discretionary behavior. We also find officers' discretion is racially biased; minority officers are less lenient to minority drivers. This is interesting in comparison with Antonovics and Knight (forthcoming) who, using the same data set, found evidence on own-race preferences in vehicle searches.