Several prominent applications of the Veil of Darkness (VOD) test, where solar variation is used to identify racial profiling in traffic stops, have found reverse discrimination in cities widely purported to disproportionately target minorities. We develop a theoretical model of traffic enforcement and demonstrate that the VOD test for racial profiling cannot distinguish between discrimination and reverse discrimination. In our model, this problem arises because motorists rationally alter their driving behavior when faced with discriminatory policing. For groups that face discrimination, our model implies that motorists who previously did not speed choose to speed in darkness, when demography cannot be observed, thus creating the possibility that the share of stopped minority motorists increases in darkness. We develop a follow-up test for identifying the direction of differential treatment by examining the speed distribution of motorists across daylight and darkness. Using data on traffic stops in Massachusetts made by State and Local Police, we reject the VOD test for equal treatment and demonstrate that driving speeds of stopped African-Americans are higher in darkness consistent with discrimination.