Using a model of student dropout with only two possible outcomes--"still in school" or "dropout"--hides the complex reasons that students leave high school. We offer a model with three outcomes: in school, pushed out or pulled out. Using data from the Educational Longitudinal Survey, we find that for black students, differences in SES explain higher likelihoods of being either pushed or pulled out as compared to white students, but Latino students remain more likely to be pulled out even after we control for SES. We also find that SES moderates the relationship between race/gender and being pushed out, and that higher levels of SES may be detrimental to students of color in the context of high poverty schools.