Everyone seems to like high recidivism rates. For the various competing interpretations of the American justice system and it effectiveness, recidivism rates serve as a policy Rorschach Test. They are routinely used by critics of mass incarceration as evidence of the system’s failures and just as frequently used by defenders as proof of the futility of rehabilitation. The argument has been going on for decades and since it seems to serve the purposes of both sides to point to high recidivism rates to support their views, there is little interest in the possibility that they might be wrong. Recent research, however, indicates that the national recidivism figures commonly cited may actually exaggerate the rate.
We routinely hear that about two-thirds of people released from prison recidivate. This number is based on studies that show that about that proportion of people are arrested and/or have their parole revoked eventually (usually somewhere up to 3 to 5 years later, depending on the follow up period for each study). Using data on arrests as a measure of recidivism is convenient for researchers because it is an official record that is cheaper and easier to get, it happens faster (so the follow-up period of the study can be shorter), and it’s a simple Yes-No outcome that is easier to code and analyze. Using arrest data ignores the fact that getting arrested or revoked does not mean you are guilty of a crime and that getting arrested is not that hard, particularly if you are back on the street and already have a criminal record.
Similarly, about half of people who are released from prison are said to eventually return to custody, either for a new offense or on a revocation. The one-half figure mostly comes from the various studies conducted by the Bureau of Justice Statistics (BJS), but a study released in 2011 by the Pew Charitable Trusts argued that it was more like 45 percent. A close examination of the differing methodologies between the BJS and Pew studies indicate the difference in the estimates is a most likely a result of the different study designs, participating states, and data sources.
The Pew study is noteworthy, however, in that it sought to detail, and then make sense of, state-by-state recidivism rates. The Pew study was a survey of state departments of corrections, asking them for estimates of their own readmission rates for two, three-year time periods. For a 1999 release cohort, 33 states provided estimates and for a 2004 cohort, 41 states responded. In keeping with Pew’s good-government-through-science credo, their report treats recidivism as a performance measure for a state department of corrections.
The report says that, “…many states are taking a hard look at their recidivism rate as a key indicator of the return they receive from their correctional investments.” The report notes that prisons serve many purposes, including retribution, incapacitation, deterrence, and “discouraging incarcerated offenders from committing new crimes once they are released” (a.k.a. specific deterrence). This last goal, “…is measured by the recidivism rate and has long been considered the leading statistical indicator of return on correctional investment.” In this context, “correctional investment” refers to punishing people and is meant, I suppose, to sound coolly rational, even if it actually sounds pretty cold-blooded.
Although the Pew report is titled, State of Recidivism: The Revolving Door of America’s Prisons, and uses recidivism rates as performance measures, the authors first warn readers not to read too much into the estimates. The report’s authors say that corrections agencies should be judged by rising or falling rates and that, “…a state where corrections agencies are strategically improving their release preparation and supervision strategies will see its recidivism rate drop.” But, they go on to write, “Policy makers should exercise caution, however, before merely accepting low or high recidivism numbers as evidence of successful or failing correctional programs.” They note that recidivism rates will be influenced by the risk levels of the people coming out of prison and can also be, “…influenced by larger social and economic forces”. To correct for this bias, the authors decided to present the data in alphabetical order, instead of ranking states by recidivism rate, which might be embarrassing for them. They also advise readers to focus on trends within states in order to, “…probe for reasons why one state’s recidivism rate might be higher than its neighbor’s rather than to make judgments about the performance of its corrections agencies based on this single indicator.” This statement appears on page 7 of a 48-page report, the balance of which is devoted to analyzing correctional agencies based on this single factor.
What if these estimates are all wrong? Or, more precisely, can they be seen in a very different way that leads to a very different picture? Enter researchers from Abt Associates and a study they published last year in Crime and Delinquency, based on an analysis of BJS’s National Corrections Reporting Program (NCRP) data. The lead author of the report, William Rhodes, recently wrote a blog post about their analysis titled, “American Prisons are not a Revolving Door: Most Released Offenders Never Return”. In his post, Rhodes says, “A common impression, reinforced by recent statistical reports, is that most offenders released from American prisons return repeatedly. The impression is wrong. Our analysis of offenders released between 2000 and 2012 shows that two of every three never return to prison. Many others reappear just once – typically for violating the technical conditions governing their community supervision instead of for new crimes.” Rhodes concludes, “…we contest the common perception that American prisons are a revolving door serving to cycle offenders through multiple terms until offenders get too feeble to victimize the public. The evidence is that most offenders serve their time and then avoid serious entanglement with the criminal justice system. Rather than asking the question of why so many offenders fail following release from prison, a more important question may be why so many offenders succeed and how corrections can promote success?”
According to Rhodes, the problem with past estimates of recidivism rates is that they are biased because people who repeatedly return to prison are overrepresented in the data. He writes, “If a statistician fails to weight his statistics to correct for this overrepresentation, offenders will appear to be highly recidivistic: One of every two will return to prison within five years. If the statistician correctly weights her statistics, offenders appear less recidivistic: Two of every three will never return.” So, according to this analysis, rather than being half full the glass is actually two-thirds empty.
In his post, Rhodes gives an example of doing a survey in a shopping mall to illustrate his point (the following is my adaptation of the analogy). Say a department store chain wants to use “repeat customers” as a performance measure to determine how well each of its stores does at getting customers to come back. It captures all of the sales during a set time period by retaining the credit card information for each sale during that period. Then it follows those customers to see if they make a future purchase at the same store. But what if there’s a big difference between stores in different places in the proportion of shoppers who frequent that store? A store in the suburbs might be the only department store around, while one in the city could be competing with several others within blocks. The suburban store is going to look better because it happens to be in the right place, not because it is any better at promoting customer loyalty.
If a prison system happens to be in a state where there are a lot of “repeat customers” because of its sentencing practices, those people are more likely to be in a release cohort because they are (again) getting out and are more likely to (again) be back soon. It may be true that half the people getting out in a year will be back, but that’s the individual recidivism rate. The system recidivism rate is really closer to one third, once you’ve adjusted for the bias created in the data by the overrepresentation of higher rate offenders. Paradoxically, both rates are true. But if you want, as Pew does, to use recidivism rates as an, “…indicator of return on correctional investment,” you ought to use the technique suggested by Rhodes and his colleagues.