The effects of prison sentences on recividism

As Scott Greenfield properly points out, I am big on empiricism at sentencing, particularly the use of actuarial data, including age, race, and gender. See Scott H. Greenfield, Prisons, Off The Hook (November 16, 2014). In a forthcoming article in the Federal Sentencing Reporter (which I have shared with Scott but asked him to withhold until publication), I will develop my thoughts further.

Nonetheless, using his powerful analytical skills, Scott now poses an important question: For those like me who push empiricism, where is the empirical analysis of the failure of prisons? He answers his own question. In a response to a comment, Scott cites to Paul Gendreau and Claire Goggin, The Effects of Prison Sentences on Recidivism, Corrections Research, Department of the Solicitor General Canada (1999).

Here is the executive summary of that article:

The use of prisons to control crime has increased in frequency in the last decade. Most recently, mandatory minimum sentencing policies have gained widespread popularity throughout the United States, severely limiting judicial discretion in sentencing. The principle rationale for mandatory minimums is the belief that length of time in prison acts as a deterrent to future recidivism.

Three schools of thought dominate the area. The first is that prisons definitely suppress criminal behaviour. Given the unpleasantness of prison life and the negative social stigma associated with incarceration, these should serve as deterrents to later criminal behaviour. The second, the “schools of crime” viewpoint, proposes just the opposite, that is, that prisons increase criminality. By this account, the barren, inhumane, and psychologically destructive nature of prisonisation makes offenders more likely to recidivate upon release. The third school of thought, which we label the “minimalist/interaction” position, contends that the effect of prison on offenders is, for the most part, minimal. This view states that prisons are essentially “psychological deep freezes”, in that offenders enter prison with a set of antisocial attitudes and behaviours which are little changed during incarceration. This perspective also suggests that lower risk offenders may be more adversely affected by greater lengths of incarceration through exposure to an environment typically dominated by their higher risk, more hard core peers.

Fifty studies dating from 1958 involving 336,052 offenders produced 325 correlations between recidivism and (a) length of time in prison and recidivism or (b) serving a prison sentence vs. receiving a community-based sanction. The data was analysed using quantitative methods (i.e., meta-analysis) to determine whether prison reduced criminal behaviour or recidivism.

The results were as follows: under both of the above conditions, prison produced slight increases in recidivism. Secondly, there was some tendency for lower risk offenders to be more negatively affected by the prison experience.

The essential conclusions reached from this study were:

1. Prisons should not be used with the expectation of reducing criminal behaviour.

2. On the basis of the present results, excessive use of incarceration has enormous cost implications.

3. In order to determine who is being adversely affected by prison, it is incumbent upon prison officials to implement repeated, comprehensive assessments of offenders’ attitudes, values, and behaviours while incarcerated.

4. The primary justification of prison should be to incapacitate offenders (particularly, those of a chronic, higher risk nature) for reasonable periods and to exact retribution.

For now, my response to this article and Scott’s larger point is muted. Suffice it to state that for the utilitarian judge like me, and keying upon 18 U.S.C. § 3553(a)(2)(C) commanding that the “court, in determining the particular sentence to be imposed, shall consider . . . the need for the sentence imposed . . . to protect the public from further crimes of the defendant . . . .,” I am in general agreement with the authors of the study* that the primary goal of sentencing ought to be incapacitation, with a dollop of retribution thrown in for good measure.  As a matter of fact, I believe that an empirical approach to sentencing may allow us to refine our decisions about offenders who require longer incapacitation versus offenders who require shorter prison sentences (or perhaps none).

In summary, thanks to Scott for his thoughtful views on this subject.


*I note that the article relies on data that is fairly old–86% of the studies examined in the meta-data analysis were from the 1970s.

10 responses

  1. There are more than 50 criminal justice systems in the US and there are significant differences between them and none of them resemble the Canadian CJS. I have studied Iowa prisons and jails and I don’t think that conclusions based on those system can be generalized. In particular they cannot be applied to the federal prison system.

    I agree with 1,2 and 4 but I have no idea what 3 involves or if would work in the US.

  2. It would be interesting to see a study of how the liberal approach to incarceration is working out these days. As it was stated, the data is 40 + years old. Virtually useless in making any sort of analysis on the subject. I don’t think that we need to leave certain sentencing guidelines up to judges, as we’ve seen in the past, in liberal bastions such as Vermont the results of liberal judges who have more pity for the criminal offenders than they do for the victims of the crimes that many of these animals commit. And once the liberal judge decides that he wants to play around with the sentencing guidelines, and allows rapists and child molesters to go free, there is little recourse that the public has to ensure the safety of the citizenry.

  3. Another good source of empirical data is the Bureau of Justice Statistics (go to A study released in 2010 indicates that “Within 5 years of release 84.1% of inmates who were age 24 or younger at release were arrested, compared to 78.6 % of inmates ages 25 to 39 and 69.2 % of those age 40 or older.” See “Recidivism of Prisoners Released in 30 States in 2005; Patterns from 2005 to 2010,” by Matthew R. Durose, Alexia D. Cooper, Ph.D., and Howard N. Snyder, Ph.D.

    It has always seemed to me that restraint (removal from the community) is the primary purpose of criminal punishment. I believe in rehabilitation of the offender, but rehabilitation takes considerable time with violent offenders. I think that as we age we often get more reasonable and accepting of life. I don’t think someone at say 21 who has engaged in excessively ant-social behavior will be reformed within a short period of time. However, they may be salvageable by the time they are 50. I deplore seeing a young person in prison, but sometimes that is just plain necessary to protect the community.

    On another note, I’m curious about the non-violent offenders. For example, I don’t like fraud and political corruption cases. That said, I thought 15 years for former Gov. Blagojavich of Illinois was excessive. 5 to 10 years (along with a very stiff financial penalty) seemed to me more reasonable given the rule of parsimony. I know that deterrence is important in these kinds of cases, but I think 5 to 10 is pretty stiff. Being from Connecticut, our former Gov. Rowland is about to be sentenced (for a second time) for campaign finance violations. He is looking at 8 to 10 years. He previously served approximately 1 year for a conviction involving acceptance of gifts. This recent conviction involves his taking $ 35,000 from a congressional candidate for his “consultant” services and having them hidden as consulting fees for the candidate’s husband’s nursing home business. While I am very unimpressed by his parlous conduct, and his failure to get the message the first time, 8 to 10 years seems pretty stiff to me.

    What say everyone else?


  4. Judge, Incapacitation for how long, until you are certain the offender will not offend again, or his statistical class will not offend again, or there is a certain probability of not reoffending, not clear what happens to the idea people are punished for what they did rather than who they are, do not want to go all C>S> Lewis, H>L>A> Hart on you but isn’t this just a new variant of old idea that social sciences should have the say, with actuarial sentences now taking the place of indeterminate sentences and wise rehabilitators. Also Results are produced by prison system as much as personal factors, so you are holding prisons constant and then really punishing convict for prison system. I remain dubitante, though I am please to see how my almost 50 years out of date knowledge of the pris bis is not out of date.

  5. Judge, I tried to restrain myself, but gave into weakness of the will, are you proposing LSAT for cons.

  6. jsneff,

    With regard to item 3, there is an 80 question empirically validated self-assessment that is used when the offender transistions to federal supervised release. Essentially, the assessment gets at criminal thinking and the different styles of criminal thinking that may impact community supervision. In my FSR article (that is awaiting publication), I discuss that assessment in some detail. All the best.


  7. While this data may be fun to put into columns and so forth, it suffers, in my opinion from a fundamental scientific flaw. The assignment of prisoners to sentences is not random — it is intentionally nonrandom. Because judges attempt to sentence more severe convicts (who can be presumed to have higher recidivism) to longer sentences we should expect those sentenced to longer sentences too have higher recividism. No matter how we try, attempts to massage the data to remove this bias are doomed to failure because the judge’s specific charges is to consider all the difficult to measure intangibles that will never make it int a experimenter’s dataset.

    You could try to account for this in your experimental design — perhaps you could compare two jurisdictions with different guidelines for the same offense and hope that differences in severity of crime will wash out with large numbers. The problem there, however, is that there may be other differences between the jurisdictions that will mess up your results.

    There is no doubt that the experimenters are aware of this bias — this is really a science 101 kind of bias that would be taught in a fundamental research curriculum. The problem is that this is exactly the kind of subtly that gets lost when you do meta-analysis and combine a bunch of studies.

    The way you like to do this study is to sentence would be to sentence offenders to a 50% chance of either a 10 year or 5 year incarceration, and then flip a coin to decide which sentence they served. This would be unethical, and probably illegal. Much of the fun of social science research is trying to ask your question in a way that you can ethically perform the experiment.

    So its an interesting suggestion, but I would need to see MUCH more discussion of such a fundamental bias before I am going to bite.

  8. JDM,

    For now, let me say that I believe actuarial data that is statistically validated as predictors of risk to reoffend can help judges in the selection of sentences. In particular, such data can help judges predict the lack of such a risk. When my FSR article is published, I will say more about it in this blog. Thank you for your engagement.

    All the best.


  9. Gee, Richard, those are pretty inflammatory comments re. Vermont! Do you have any particular statistics or rational basis to accuse those liberal VT judges of allowing “rapists and child molesters to go free…”? Or is it just a feeling?

    Here, I’ll save you the effort:
    “According to a 2006 U.S. Department of Justice report, Vermont, one of the least populous states in the country, was among the states with the fastest growing prison populations in the nation. Vermont’s largest city is home to only 40,000 people, and unlike many other states, its resident population is not growing. Still, between 1996 and 2006, the state’s incarceration rate increased 80 percent, nearly doubling its prison population from 1,058 to 2,123.”

%d bloggers like this: