The sentencing guidelines movement focused on reducing extralegal disparity at the sentencing stage. There is a concern that an unintended consequence of presumptive guidelines is a shift in power to prosecutors, who gain influence on the sentence through their charging behaviors. This shift cannot be seen empirically if sentencing models use charges at conviction to control for the type of crime. Recent studies on the federal system suggest that using charges at arrest rather than conviction to control for type of crime dramatically changes the story about discretion and the source of racial disparities. In this chapter, we use data from two counties in New York to test whether this same pattern can be found in non-guidelines jurisdictions. We found large racial disparities in the probability of incarceration in the standard conviction model. Although we found substantial charge bargaining, disparities became lower when charges at arrest or arraignment were controlled for. At least in our data, the use of conviction information does not hide evidence of racial disparity created earlier in the process.