Recidivism rates must be calculated over a specified follow-up period. This is the time—measured in months for this report—between release from prison and when the data on new offenses or prison admissions is collected (March 2002). This report provides data on recidivism rates up to 60 months following release.
As more time passes since a cohort of inmates was released from prison, the number of inmates who recidivate grows, so the percentage of released who recidivate increases. For example, the recidivism rate measured at 36 months after release is higher than the rate measured at 12 months after release. Recidivism rates can be compared between follow-up periods for the same release cohort. However, when comparing rates for different release cohorts, one should use rates based on the same follow-up time after release.
More than one-third (36.1%) of released inmates spend some time on supervision following release from prison. Some inmates under supervision (technical violators) spend part of their follow-up period back in prison. During this time they are not at risk of committing new offenses in public. This report accounts for such time "not on the street" by reducing both the follow-up period and the time to reoffense and reimprisonment by the amount of intervening time spent back in prison for a technical violation. This prevents the "not at risk" time from artificially lowering recidivism rates.
Many recidivism studies use a fixed follow-up period. For example, all releases might be tracked for 3 years following release, and only recidivism events that occurred within 3 years of release will be counted. This data is usually analyzed based on whether or not each inmate recidivated—generating a certain percentage that recidivated. If the timing of the recidivism events (such as an offense date) is measured, they may also analyze the time to failure within the 3-year period, which is more useful information.
A fixed follow-up study has two limitations, especially if the time to recidivism is not examined. First, the entire follow-up time must pass before the collected data can be properly analyzed. This may delay results beyond a useful time to inform decision-making, such as when an evaluation might help make budget choices about a particular program. Second, no information is available about recidivism over a longer-term. To the extent that a program does or does not reduce recidivism more than three years after release, an evaluation using the fixed period will fail to show this important result.
In contrast and to overcome these limitations, this study uses all follow-up time available rather than a fixed follow-up period and reports rates from six months to five years after release. However, not all released inmates used to calculate these rates have been out of prison equally long. To correct for differences among inmates in the time since release, this study relies on a statistical procedure (survival analysis) for estimating rates that accounts for the available follow-up period for each inmate released. The method includes all inmates in the release cohort but only counts them for recidivism rates based on their actual individual follow-up periods.
The main benefit of this approach is that the maximum possible number of cases can be used to estimate rates for each follow-up period. Nevertheless, rate estimates for cases released within two years of data collection are known to be unstable (biased low). They should be considered preliminary and subject to revision upward as additional time passes, allowing data on more recidivists to enter the Department's information system. To compensate for this small bias, all follow-up period rates are estimated using all releases (over six years) in the cohort. Also comparisons of shorter term rates will still allow meaningful evaluation of recidivism differences between groups unless evidence shows that this known, predictable bias affects the comparison groups differently. The advantage of this analytical approach is that recidivism evaluations can be done sooner, using shorter follow-up periods, and with confidence in any statistically significant results.
For more information on measuring follow-up time and the statistical procedure used to estimate recidivism rates, see the Technical Appendix. For recidivism rates over various follow-up periods, see Recidivism Rate Curves.