Changing placement policy may help to improve developmental education student outcomes in community colleges, but there is little understanding of the impacts of these reforms. We take advantage of heterogeneous placement policy in a large urban community college district in California to compare the effects of math remediation under different policy contexts. District colleges either switched from using math diagnostics to using computer-adaptive tests, or raised placement cutoffs. We use quasi-experimental methods to identify the impact of remediation under each policy and the change in impact following placement policy experimentation. We find that switching to a computer-adaptive test exacerbated the penalty of remediation for marginal students and resulted in more placement errors. Modestly raising placement cutoffs had no significant effects.