A useful way to display recidivism rates is to graph the cumulative percentage of recidivists over the time since release from prison, creating a recidivism rate curve, as in Chart 1. These rate curves are the complement of estimated survival functions.
Chart 1 displays the reoffense and reimprisonment rates for state prison inmates released from July 1995 through June 2001 (release cohort). The lower axis indicates the number of months (up to 60) since release from prison (follow-up period). The side axis shows the percentage of inmates who recidivated. The graphed curves mark the percentage of total inmates released who reoffended and who returned to prison for a new crime over the available follow-up period.
The recidivism rates are the values on the graphs. To illustrate, recidivism rates at selected follow-up periods are noted on the chart: 6, 12, 18, 24, 36, 48, and 60 months after release. For example, at three years (36 months) after release the reoffense rate is 39.9% and the reimprisonment rate is 25.7%. In other words, roughly 40 of 100 inmates released from prison were convicted of a new crime within three years and roughly 26 of 100 were reimprisoned within 3 years. Rates in this report can be interpreted as probabilities of recidivating: there is a 39.9% probability that a released inmate will commit a new offense within three years and a 25.7% probability that a released inmate will be reimprisoned within three years for a new offense.
This chart shows that reoffense rates increase most quickly immediately after release and grow more slowly as the follow-up period increases. For example, in the first six months after release the reoffense rate grows to 12.1%, but between 36 and 48 months after release the rate grows only 4.9 points from 39.9% to 44.8%.
The pattern of reimprisonment rates is somewhat different, with the largest growth (5.8 points) occurring between 12 and 18 months after release. This difference reflects the length of time required for the recidivists to be arrested, convicted, sentenced, and delivered to the Department.
Note that reimprisonment rates are lower than reoffense (reconviction) rates because not all reconvictions result in a new prison sentence—only a subset of inmates who commit new crimes are sentenced to prison for them.
Recidivism rate curves are also useful for displaying the influences certain factors may have on recidivism rates. For example, Chart 2 below displays the reoffense rates for male and female inmates. Clearly, female inmates appear to reoffend at lower rates than males, suggesting that gender influences recidivism. For information about what influences recidivism rates, see Factors Affecting Rates.
Whether this apparent difference between males and females is real can be determined by comparing these rate curves using tests of statistical significance. These tests demonstrate whether a factor, like gender, has a meaningful influence on recidivism rates and estimate the size of that influence.
This research method for determining whether significant differences exist between the rates of two subsets of the release cohort provides a basis for analyzing whether programs or functions of the Department reduce recidivism. For example, one study using Department data separated inmates into two groups—releases from privately operated prisons and releasees from Department operated prisons—to determine whether reoffense rates differed meaningfully between these two groups. The same statistical technique used to calculate recidivism rates reported here was used to test for meaningful differences while controlling for other factors know to influence recidivism rates (see Farabee and Knight, 2002).
An important aspect of recidivism rate curves generated from this data is that there is a relationship between the when inmates were released and when the data are collected. This relationship is displayed in Chart 3 below using reoffense rates of male inmates, as an example. This data appears to indicate that reoffense rates for male inmates have declined steadily from fiscal year 1995-96 releases to fiscal year 2000-01 releases; but that cannot be concluded from this data. Instead, the Department's research shows that this relationship results from how close the releases are to the time when the data were collected. The reliability of these recidivism rate estimates depends on the size of the subset of releases that recidivates. For fiscal year 1995-96, the rates are based on 5,985 reoffenses (53.9%) of 11,106 releases that year, whereas fiscal year 2000-01 rates are based on only 1,366 reoffenses (8.4%) of 16,352 releases. As the number of recidivist cases increases relative to the release population, the estimated rates become more stable and comparable. Caution must be used when analyzing recidivism rate differences based on either small numbers of cases or large numbers of releases with small percentages of recidivists.
For information about the statistical technique used to estimate recidivism rates (survival functions) and to evaluate significant differences between rates, see Statistical Analysis. For more information about analyzing recidivism rate changes over time, see the Technical Appendix.