This paper examines the impact of jury racial composition on trial outcomes using a unique dataset of all felony trials in Sarasota County, Florida between 2004 and 2009. We utilize a research design that exploits day-to-day variation in the composition of the jury pool to isolate quasi-random variation in the composition of the seated jury. We find strong evidence that all-white juries acquit whites more often and are less favorable to black versus white defendants when compared to juries with at least one black member. Using the Anwar-Fang rank order test, we find strong statistical evidence of discrimination on the basis of defendant race. These results are consistent with racial prejudice on the part of white jurors, black jurors, or both. Using a simple model of jury selection and decision-making, we replicate the entire set of empirical regularities observed in the data, including the fact that blacks in the jury pool are just as likely as whites to be seated. Simulations of the model suggest that jurors of each race are heterogeneous in the standards of evidence that they require to convict and that both black and white defendants would prefer to face jurors of the same race.