Oh, Those Woeful Recidivism Numbers

Every decade or so, the U.S. Bureau of Justice Statistics releases a big national study of prisoner recidivism. The latest BJS research came out last week, and the numbers were no less depressing than they were in 2002. Here’s the report’s lead:

Overall, 67.8% of the 404,638 state prisoners released in 2005 in 30 states were arrested within 3 years of release, and 76.6% were arrested within 5 years of release.

Simply put, failure is the norm, not the exception, for individuals released from U.S. prisons.

High recidivism rates constitute the most difficult and important challenge for those of us who would like to see fewer long sentences and more generous opportunities for inmates to earn early release. If most prisoners are rearrested shortly after they get out, doesn’t that lead inexorably to the conclusion that we should err on the side of more, not less, time behind bars?

Certainly, the woeful recidivism numbers should take indiscriminate, mass releases off the table. On the other hand, I think it is possible to overstate the significance of national rearrest rates for sentencing and corrections policy. These numbers should be the start, not the end, of the conversation.

First, it is important to remember that not all post-prison failures are equally grave. For many readers, references to arrest after release will conjure up images of Willie Horton, the furloughed Massachusetts prisoner who committed a brutal assault and rape in 1987 and who was then featured in a notorious attack ad in the 1988 presidential campaign. However, the BJS data indicate that most rearrests are for nonviolent offenses. Homicides and rapes are particularly rare. Fewer than one percent of released prisoners were later arrested for homicide, and fewer than two percent for rape. Drug possession was the single most common charge for the rearrested ex-cons. Additionally, many rearrests were recorded as probation or parole violations. While some of these violations may have been serious crimes in their own right, many were likely of a merely technical nature.

None of this is to deny that some prisoners present serious public-safety risks. After all, nearly a quarter of the returning prisoners in the BJS study were arrested for assault within five years of release. This is a depressingly high percentage, but not so very high as to preclude serious discussion about the possibility of identifying low-risk inmates for early release.

Then, too, it is also important to remember that rearrest is the most expansive way to measure recidivism. Not everyone who is arrested is guilty, and not every guilty person who is arrested deserves serious punishment. The BJS study also provides alternative measures of recidivism that are a bit lower than the rearrest numbers. For instance, if conviction of a new offense is the measure, then the five-year recidivism rate drops from 77% to 55%. If a new prison sentence is the measure, then the recidivism rate drops even further to 28%. To be sure, no single measure of recidivism is perfect, but the rearrest numbers may be especially likely to create an exaggerated sense of the dangers posed by returning inmates.

Additionally, the BJS national numbers obscure intriguing variations from state to state. Although the BJS report does not provide state breakdowns, the Pew Center issued a large-scale recidivism study in 2011 that did just that. Pew found that three-year return-to-prison rates varied from 23% (Oregon) to 61% (Minnesota). To some extent, these variations are a product of how well states screen out low-risk offenders at the sentencing stage; a low recidivism rate for prisoners may indicate that a state is incarcerating a lot of people who should not be behind bars in the first place. However, some variation also likely reflects effective risk-reduction programming in prison and good supervision afterwards. For instance, this may help to account for the low recidivism rate in Oregon, which succeeded in reducing its returns to prison from 33% to 23% between the 1999 release cohort and the 2004 release cohort. The experience of Oregon and some other low-recidivism states suggests that high rates of failure may not be an unalterable fact of life for American prisoners, but can be brought down through the right corrections policies.

In sum, high rearrest rates overall should not detract from efforts to reduce sentence lengths and to move low-risk inmates out of prison more quickly.  Cost-savings from reduced prison utilization can be reinvested in evidence-based programming and community supervision.  In this way, there is hope for a “virtuous circle” of reduced incarceration and lower recidivism — a reversal of the vicious circle of the 1980s and 1990s, as exploding prison populations overwhelmed corrections and undermined the effectiveness of programming and supervision.

There is an ethical dimension to this, too.  At some point, long-term imprisonment that is based predominantly on the fear of recidivism becomes functionally indistinguishable from punishment for crimes that have not yet been committed, let alone charged and proven beyond a reasonable doubt.  Perhaps there are offenders for whom such incapacitation is justifiable, but this should never be more than a last resort.

In addition to these big-picture policy concerns, the BJS data also have implications for some more specific issues. First, the data reinforce conventional wisdom that the risks of failure are greatest in the first few months after release. More than 40% of prisoners were arrested in the first year after release. However, for those who avoided arrest in year one, the arrest rate was less than 30% in year two, with further declines in arrests each subsequent year. For those who avoided arrest over the first four years, the risk in year five was only 13%. Similar drop-offs are apparent with other recidivism measures. For instance, 50% were returned to prison within three years of release, but the cumulative percentage rose only to 55% over the next two years. The patterns suggest that long periods of post-prison supervision are probably unnecessary and a waste of scarce community corrections resources. It might make sense to adopt a presumptive cap of five years on post-prison supervision, with the potential for extension in unusual circumstances.

Second, the data cut against drawing sharp distinctions between violent and nonviolent offenders when it comes to alternatives to incarceration and early release from prison. Policymakers have frequently drawn such distinctions in recent years, for instance, in the eligibility criteria for drug treatment courts. However, those convicted of violent offenses do not have higher recidivism rates than those convicted of other offenses. Indeed, it is the property offenders who have the highest five-year rearrest rate (82%), as compared to 71% for the violent offenders. Homicide offenders have the lowest rearrest rate of all, at 51%. Nor, when they do recidivate, are violent offenders especially likely to commit fresh violent crimes. Only 33% of the violent offenders were rearrested for another violent crime. By comparison, almost as many property offenders (29%) were arrested for a violent crime. There seems little basis, then, for viewing those convicted of violent crimes as categorically riskier than those convicted of property or other offenses. In drawing sharp violent-nonviolent distinctions, policymakers almost certainly miss good opportunities for diverting low-risk and readily treatable offenders from prison and moving safe bets out of prison more quickly.