With mounting support for dramatic criminal justice reform, the question is no longer whether we should decarcerate American prisons but how. This question is far more complicated than it might seem. We could cut the prison population in half, for example, by drastically shortening sentences. Or we could reduce prison admissions. Or we could do both. And we could do either or both for countless combinations of criminal offenses. Moreover, even when they reach the same numeric target, these strategies are not equivalent. They would have vastly different consequences for both prisoners and the public, and widely varying timeframes to take effect. To pick among them, we need richer metrics and more precise empirical estimates to evaluate their consequences.

This Article begins by proposing metrics to evaluate the relative merits of competing decarceration strategies. The public debate has focused almost exclusively on how we might decarcerate while minimizing any increases in crime and has, therefore, underappreciated the costs of prison itself. We should consider at least three more metrics: the social harm of incarceration, racial disparity, and timing. Next, the Article develops an empirical methodology to identify the range of strategies that would reduce the national prison population by 25, 50, and 75%. Finally, it identifies the best performing strategies against each metric.

The results have several broader takeaways. First, the optimal approach to decarceration depends heavily on which metrics we value most. The results thus quantify a stark set of policy choices behind a seemingly simple objective. Second, the results confirm that, to dramatically shrink prisons, it is critical to decarcerate a substantial number of people convicted of violent offenses—a fact that may surprise the majority of Americans who believe people convicted of drug offenses occupy half of prison beds. Finally, the results show that race-neutral decarceration strategies are likely to exacerbate rather than mitigate racial disparities. Armed with the conceptual tools and methodologies developed in this Article, we can make more informed decisions about how to best scale down prisons, given our priorities and constraints.