Texas Recidivism Rates: Is "why are they so low?" the wrong question?

On Wednesday I asked why are Texas recidivism rates so low? State data showed Texas' three-year re-incarceration rate is half that of California's, and more than 40% below the national average. I received some very interesting comments to that post, as well as several emails, and believe I can shed a little more light on that subject - or at least cast light from different angles.

I'd at first suggested that Texas' numbers might simply be underreported, but a commenter referred me to this 2005 study (pdf) by the UC-Irvine Institute on Evidence Based Corrections that purported to make apples-to-apples comparisons of recidivism between California, Texas and other large states. On page four of that report, graphs show Texas with half California's new-conviction rate, and about a 40% lower re-arrest rate.

That seems to confirm Texas statistics relative to other states are likely accurate. If so, we should be looking for reasons why the rates are lower, not to debunk the premise. Indeed, I may have posed the wrong question, failing to grasp all the assumptions that underly how we view recidivism rates. Possibly the question shouldn't be why are Texas' rates so low, but are low recidivism rates a good thing?

That sounds counterintuitive - under TDCJ's definition, recidivism means formerly incarcerated people either violate parole or commit new crimes. We don't want that, right? So how is it possible that higher recidivism rates might actually be a sign of a healthier system? Here's an insightful hypothesis from an anonymous Grits commenter:

My guess would be that Texas is putting a lot of low level people in prison who aren't really crooks and don't need to be there. Since many weren't actually bad guys to begin with, when they get out fewer re-offend. Just a guess.

Good guess! Michael Connelly, an academic who blogs at Corrections Sentencing, agreed that this could explain the numbers, and noted that Oklahoma's re-incarceration rates are similar to Texas'. Said Michael:

One way to check would be to look at the LSI-R scores of the recidivists or to take a cohort of releasees and do a survival analysis of them by low, medium, and high LSI-R scores. In OK, the numbers are telling, roughly 3/4ths of the "low" are still out 3 years later while only about 7% of the "high" are. It would make a good study.

Okay, now we're getting somewhere - an hypothesis and a suggested methodology for testing it. I love Grits commenters!! :-) LSI-R scores are a risk categorization tool developed by community supervision experts to help weigh public safety risks that must be managed by probation and parole departments.

Michael's right that would be a fascinating study. If Texas' breakdown was similar to Oklahoma's that would likely explain the disparity, but it wouldn't mean we're doing something right - it would mean Texas overincarcerates low-risk offenders who probably weren't going to re-offend in up to 3/4 of the cases.

Data presented in response to this post suggest at least two factors contribute to Texas' lower recidivism rates compared with California and the national average: overincarceration of low-risk offenders, and more Texas' prisoners serving their full sentence than being released on parole. About half of California's re-incarcerations stem from parole revocations, but as the UC-Irvine study pointed out,

If someone isn’t on parole supervision, they can’t be violated for technical parole violations. The fact that nearly all California prisoners report to parole agents partially explains our high technical violation rates.

In Texas, though, according to this Urban Institute study an emailer pointed me to, most state jail felons serve their full term and are released without any supervision. (The max penalty for a state jail felony is 2 years incarceration, and those offenders cannot be paroled.) Said the Urban Institute:

most (84 percent) state prisoners are subject to postrelease supervision, compared with less than 3 percent of confinees [state jail offenders].

So a large number of prisoners - mostly low-level drug and property offenders - go unsupervised upon release in Texas, while California supervises nearly everyone they release. Since Texas state jail felons aren't being supervised, they can't be revoked, but I'm not sure that means the public is any safer. It begs the question, would it improve public safety if they were incarcerated less and supervised more intensively during a re-entry period?

I don't know the answer, but it's an interesting question. The conservative Texas Public Policy Foundation has called for giving state jail felons parole in part for this reason.

What else might explain the difference? I think California's famous Prop 36, or the "Substance Abuse and Crime Prevention Act" of 2000, might be a big factor. That law mandated community supervision and drug treatment instead of incarceration for low-level drug offenders. Not only has it saved Califonia money, it dramatically reduced the number of low-level drug offenders entering prison in the first place.

To complete the loop, if we assume many Prop 36 offenders would reside in the low-level LSI-R risk categories, anyway, they're likely among the class of offenders who, when incarcerated in Oklahoma, did not reoffend 3/4 of the time.

So are low recidivism rates necessarily a good thing if they mean we're incarcerating low-level offenders who would be unlikely to re-offend whether they went to prison or not? Is that best for the taxpayers and public safety, especially if truly violent offenders must be released to house them? By that analysis, a system that supervises more low-level offenders in the community may increase recidivism rates, but only because we'd be reserving prison for the real bad guys, not simply people struggling with addiction.

I don't know if this analysis is correct: I'm merely exploring the question, here, with the help of Grits commenters, and obviously there are a number of variables to consider. But before now I've only ever considered higher recidivism rates a bad thing. This discussion makes me think the matter is really a lot more complex than that simple construction would lead us to believe.

Thanks, Grits commenters, for the great information. And let me know what you think of this new open-sourced hypothesis.