Mounting evidence indicates that students of color are more likely than their white peers to receive exclusionary discipline (out-of-school suspensions and expulsions), but alleviating these disparities requires an understanding of what drives them. In this study, we use seven years of student- and infraction-level data from every public school in Arkansas to examine disproportionalities within and across districts and schools. Prior analyses of the same data found that black students are more likely to receive exclusionary discipline than their white peers, even after controlling for the nature and number of disciplinary referrals, but most of the differences occur between schools, rather than within. We build on these findings in two ways: (a) by estimating the racial disparities in the likelihood of referral for particular infraction types and (b) by estimating whether the racial disproportionalities in the use of exclusionary discipline, after controlling for reported behavior, differ according to the types of infractions reported. We find that for common, subjective infractions, black students are at a higher risk of referral and at a higher risk of exclusionary discipline, conditional on referral. In addition, disparities appear to be driven by differences across school districts (rather than within). These findings have important implications for designing and targeting discipline reforms where they are needed most.