Agan (2011) used a retrospective, quasi-experimental design to determine the impact of sex-offender registration on the rate of sex crimes, rate of sex-offender recidivism, and on the location of sex crimes. Three separate data sets and study designs were used; however, for this review, the focus was only on one data set, which included a comparison of registered and non-registered sex offenders.
Using data collected by the Bureau of Justice Statistics (BJS) on a cohort of prisoners from 15 states (Arizona, California, Delaware, Florida, Illinois, Maryland, Michigan, Minnesota, New Jersey, New York, North Carolina, Ohio, Oregon, Texas, and Virginia) released in 1994, Agan investigated the impact of sex-offender registries on recidivism. This cohort included 10,510 individuals convicted of a sex offense. After juvenile offenders, those arrested in other states, and those returning to Arizona (where conflicting reports made it impossible to determine the date that the registry went into effect) were removed from this cohort, a final group of 9,623 sex offenders from 14 states was included in the analysis. The final treatment group consisted of sex offenders who (as determined by the study author) would have been required to register (N = 5,195), depending on the crimes that the offenders committed, the states in which they were released, and the dates that sex-offender registries went into effect in those states. The comparison group consisted of released sex offenders who would not have been subject to this requirement (N = 4,428). Because no individual-level data on registration was available for the offenders, the study author had no way of knowing whether offenders who were required to register had actually done so.
The treatment group was 25.7 percent African American, 23.3 percent Hispanic, and 51 percent other racial/ethnic groups. The majority (99.2 percent) were male, and the average age at release was 37.1 years with an average sentence length of 5.6 years. The comparison group was 37.4 percent African American, 6.7 percent Hispanic, and 55.9 percent other. The majority (98.6 percent) of the comparison group were male. The average age at release was 36.9 years with an average sentence length of 8 years. To account for the differences across the treatment and comparison groups, the study author used a Cox proportional hazard mode.