Between 2003 and 2014, the majority of people stopped under the New York Police Department's policy of stop-and-frisk were non-white. This led to charges of racial bias. This paper examines whether biases persist after a stop occurs. Data on 587,479 stops from 2010 are analyzed to examine differences by race for six outcomes: use of force, being frisked, being searched, being issued a summons, being arrested, and yielding a productive stop. Multilevel logistic models are then estimated to examine the effect of precinct-level residential racial composition and crime rates on the odds of the six outcomes. Results show that blacks and Latinos are more likely to be frisked and to have force used against them; however, this risk depends on precinct-level characteristics. A supplementary analysis of stops from 2014 shows that decreased reliance on stop-and-frisk reduces the odds of force being used against blacks and Latinos to non-significance.