Selection bias and recidivism

Selection bias is one of many problems people have when forming a world view. There is a nice explanation of selection bias in the Wall Street Journal Europe.

But doesn’t the high recidivism rate show that prison is not an effective deterrent after all? It does not. Testing the deterrence effect of prison by observing the proportion of ex-prisoners who commit crimes is a bad case of the statistical error of “sample bias.” Prisoners are, by hypothesis [I think he meant to say ‘by definition’, Sinc], people for whom the threat of prison is an insufficient deterrent to crime. That prison does not deter ex-prisoners tells us nothing about how much it deters the rest of the population, nor therefore by how much it reduces crime.

Once you think of criminal punishments as deterrents, 100% recidivism is unsurprising, because the first conviction is the most expensive for a criminal. This is when he incurs the one-off, irrecoverable costs of becoming a known criminal, such as diminished career and social prospects. If the chance of incurring these costs (in addition to the penal costs) did not deter him from committing a crime, then the inevitably lesser costs of subsequent convictions are unlikely to deter him. This is true whatever the legal penalty for crime—be it torture, prison or “community service”—and however effectively it deters first crimes. Recidivism is a red-herring.

The point is that looking at criminals tells us nothing about law-abiding individuals. So far so good. Yet I find the example unconvincing.

Prison is meant to be a deterrent against all crime, including recidivism. The author of the article, Jamie Whyte, is unconvinced by rehabilitation.

The new British government plans to replace punishment with attempted rehabilitation. Since this will reduce the cost of committing crimes, it will increase the number of people who become criminals. In other words, the policy will encourage criminality for the sake of then curing it. This would be absurd even if a cure for criminality existed, even if attempts at rehabilitation were generally successful. Given that rehabilitation remains no more than a fantasy of penal reformers, however, the policy is not merely absurd but positively wicked.

But rehabilitation is only one mechanism to inhibit recidivism.* As he points out, at least a quarter of convicted criminals do not reoffend. The challenge that we face as a society is how to structure the criminal justice system so that the three-quarters of convicted criminals who are recidivists are further incentivised to live within the law while not overly punishing the quarter who do mend their ways.

On that that note – Andrew Leigh spent some time behind bars. I have long believed that not enough politicians and public servants end up in gaol; but alas, he was only visiting. He makes the point that prison populations are rising.

Building prisons is a growth industry because the number of inmates continues to grow. As James Eyers pointed out in this newspaper recently, the growth in Australia’s prison population has been driven not by a rise in crime, but by law changes such as tougher bail conditions and mandatory non-parole periods.

He seems to be suggesting that this is ‘a bad thing’ – yet it would be entirely consistent Whyte’s argument about recidivism. Longer gaol terms inhibit crime by known criminals.

* One mechanism would be to decriminalise drugs. I am not a fan of substance abuse and addictive substances, but the ‘war on drugs’ has clearly failed, while the ‘war on tobacco’ seems to be succeeding. The difference being that the government and health authorities are succeeding in inhibiting tobacco usage while it remains a legal product, but they have failed to inhibit drug usage while it is an illegal product. Now we can debate whether the government should be waging ‘war’ on any private behaviour – but by its own objectives it must be obvious that the war on drugs has failed.

This entry was posted in Uncategorized. Bookmark the permalink.

7 Responses to Selection bias and recidivism

  1. conrad says:

    It’s seems pretty obvious that it’s getting the balance right which is important. The obvious problem with just sticking more people in jail longer is that it costs a huge amount (to everyone involved, including the people going in), and the threat of going to prison is only one factor driving crime — if you look at cross time and cross country comparisons, other reasons like unemployment are very important.

    I imagine the crime-perception problem would be fixed if people were punished longer for just some types of crime, like violent ones (and unfortunately probably for drug ones also).

  2. rodney says:

    In NSW select criminals have an affair with a prison Psychologist and become academics in Sociology (or possibly Peace Studies). Alternately the can join the Labor Party by marriage and become head od a major Govt Department at 32.

    More generally the fashion of closing mental hospitals has meant that increasingly the difference between crims and the mentally ill, always blurred ,has been lost.
    A rational analysis of behaviour of the mentally ill is not going to work.

  3. daddy dave says:

    That’s a really interesting point. Recidivism is not a good measure of overall crime deterrence.

    “Prison is meant to be a deterrent against all crime, including recidivism. “

    But if you’re reducing the total crime rate, maybe recidivism is an acceptable cost?
    In other words, it’s better to have 3 robberies committed by the same guy than 30 robberies all by different people.

    Also, you got low rates of recidivism, that might not be anything to cheer about, because they can be caused by factors other than reform or a desire ‘not to go back there.’
    Crime is committed mostly by young men. As a man gets older, he’s less likely to reoffend.
    Crime is sometimes committed as a one-off event (such as crimes of passion, or rare opportunity). These are unlikely to be repeated.
    Convicts are sometimes innocent. These people, too, are unlikely to reoffend.

  4. Sinclair Davidson says:

    Also, you got low rates of recidivism, that might not be anything to cheer about, because they can be caused by factors other than reform or a desire ‘not to go back there.’

    I understand the point, but so what? Do we care why criminals don’t reoffend?

  5. daddy dave says:

    Do we care why criminals don’t reoffend?

    Yes, if you’re using recidivism rates to guide policy.

  6. conrad says:

    It would be good to have data that you can’t get also — there’s a fair bit of data now (mainly from the US) showing that quite a reasonable proportion of people convicted for some types of crime (generally the type we like least) have quite identifiable frontal lobe problems (which basically means that, compared to normals, they won’t be able to inhibit inappropriate behaviour, they will act without thinking, and they will not understand consequences well). If you want the best outcomes, then these people clearly need to be treated differently to those that don’t have such problems since changing possible penalties you can dish out and then giving a few people penalties on the harsh end of the scale is not going to have any great effect on the behavior of these guys.

  7. Adrian says:

    Recall reading an article in the city journal, i think, which argued that one the main reasons US crime has stayed low during the recent economic downturn is because of high rates of incarceration from the 1990s onwards.

Comments are closed.