Driven by a combination of concern for historically high suspension rates and substantial disproportionalities in suspension use, a recent wave of education reforms encourages schools to reduce their use of suspensions for student behavior management. Both academic and political discourse has focused on the extensive use of suspension for relatively minor behavioral infractions, with an implicit assumption or explicit articulation that suspension could still be used for severe infractions. This paper tests that assumption, providing evidence that reductions in suspensions for severe infractions may produce positive impacts without harming school safety. Using data from high schools in the Chicago Public Schools (CPS), 2007-2014, we examine how declines in school reliance on suspensions for severe infractions are associated with changes in academic performance, attendance, and student reports of school climate for all students in the school. Recognizing the substantial methodological difficulty in obtaining impact estimates, we exploit a series of official and unofficial policy-induced changes to suspension practice, using school and student fixed effects models with extensive controls to reduce potential sources of bias in the estimates. We find the reduction in out-of-school suspension for severe infractions was associated with small but statistically significant increases in student test scores, consequential attendance improvements (beyond the impact of fewer days suspended), and heterogeneity in changes to students' perceptions of school safety. Test score impacts are concentrated in racially diverse schools and those with low baseline suspension use. Attendance impacts are driven by schools predominantly serving African American students (which also had the highest baseline suspension rates); these schools also had large, significant improvements in perceptions of school climate. [This paper was published in "Peabody Journal of Education" v93 n2 2018.]