Standard deterrence theory argues that the costs of committing a crime are determined by the probability and severity of punishment. Using data from the National Incident-Based Reporting System (NIBRS), I document that the share of reported rape cases ending in arrest fell from 23 to 12 percent between 2000 and 2024, with a similar decline for fondling. The decline is not explained by changes in the (observable) case characteristics of reported incidents, changes in NIBRS-participating agencies, or broader shifts in policing: arrest probabilities for aggravated assault rose over the same period. These findings suggest that the deterrent power of the criminal justice system has eroded over time, specifically for sex crimes. I additionally characterize which reported incidents are most likely to result in arrest, finding that the strongest predictors are the presence of a co-occurring non-sex offense and the availability of basic identifying information about the offender. These estimates provide empirical benchmarks for a planned incentivized survey measuring perceptions of arrest probabilities.

Callisto is an online and anonymous sexual assault reporting platform, which allows survivors to find others victimized by the same perpetrator. While Callisto has been introduced on college campuses to lower barriers to reporting sexual assault, there is no systematic evidence on its effectiveness or student engagement. In this study, I conduct a survey of Callisto users to gather qualitative insights on their experience with the platform and their perceptions of barriers to reporting. I then conduct an online experiment of 1,923 women to test whether updating beliefs about social norms on victim-blaming increases engagement with Callisto. In the experiment, subjects complete an incentivized social norm task in which they guess the percent of prior male participants who agreed with a victim-blaming statement. In a belief correction treatment, half of subjects receive information regarding the actual percent of prior male subjects who agreed.  I find that among platform users, Callisto provides social and emotional benefits at least as good as other reporting options. In addition, 93% of respondents agreed that it was likely that a sexual assault survivor would be afraid of not being believed or being accused of lying if she reported—and over 60% of respondents ranked this as a top barrier to reporting. Experimental evidence reveals that while there are large misperceptions of the victim-blaming norm, the belief correction treatment has no impact on a respondent’s likelihood of engaging with Callisto. While social norms appear to be a major barrier to reporting, correcting misperceptions alone does not shift behavior.