Bail reform rollbacks are driven by anecdotes, not recidivism data

Several states and cities that enacted bail reform between 2017 and 2020 have since rolled back those reforms. New York's 2019 bail reform eliminated cash bail for most misdemeanors and nonviolent felonies. Within months, a series of high-profile crimes committed by people released pretrial generated intense media coverage and political pressure. By 2020, the legislature had already amended the law to give judges more discretion to set bail, and further rollbacks followed in 2022. The problem is that these rollbacks are driven by individual anecdotes rather than aggregate data. The actual data from New York showed that rearrest rates for people released under the new law were comparable to pre-reform rates. A study by the New York City Criminal Justice Agency found no statistically significant increase in pretrial rearrest. But a single dramatic case, amplified by media and political opponents, generates more political pressure than a thousand data points showing the reform is working. This pattern persists because criminal justice policy is uniquely susceptible to availability bias. Voters remember the one headline about a released defendant committing a violent crime. They do not see the thousands of people who were released, showed up for court, and were never rearrested. Politicians face asymmetric risk: no one gets credit for the crimes that did not happen because someone was released pretrial, but every politician gets blamed for the one crime that did happen. The structural issue is that there is no institutional mechanism to force data-driven policymaking in criminal justice. Unlike drug approval (FDA) or aviation safety (FAA), there is no agency that must certify bail policy changes based on evidence before they are enacted or repealed. Policy is made by legislatures responding to political incentives, and the incentive structure systematically favors appearing tough on crime over following the data.