Most risk assessment articles in my field show small results. The studies examine how well a particular "instrument," that is, a structured test or evaluation, predicted recidivism in violent and/or sexual offenders. Here is an example from this spring, "What does it mean when age is related to recidivism among sex offenders?" I chose it because it is in this month's pile and it's pretty good. There isn't anything earth-shaking here, but it adds to our sum of knowledge a bit. The takeaway: the age of an offender at first offense or index offense is a better predictor of recidivism than his age at release. There is this thought that offenses decrease as offenders age, and this is true, However, not that much, and age at first offense tells us more - the younger, the worse.
All this according to a the predictions of a single test. There are dozens of tests. Why not just one, really good test? Well, because none of them fit that description. Some tests are better predictors among adolescents, or among the mentally ill, or the developmentally delayed. Some predict future sexual offense poorly but future violent offense well. Some are sensitive to changes, and so can tell us if a treatment is working. Others are actuarial and the scores don't change much: the number of years you were in school, the amount of abuse you received as a child, how long you lived with a bio parent - these numbers are going to be the same 5 years later.
I notice in online discussions that people bring their opinions to the data quite a bit - even the professionals in related disciplines will strongly declare that the presence of a strong father is strongly protective against offending, e.g "I have never treated an offender who was raised in an intact, nonabusive family!" Really? I can think of lots, including my own father, and this isn't even my specialty.
One of the difficulties in measuring is that these are events that don't occur often, even among the worst offenders. If you went to a psychologist and said "I'd like some help in reducing a behavior I only do every few years or so," she would shake her head and say it would be hard to know if treatment were working. Whether the treatment was behaviorist, insight-oriented, or medical, there just isn't enough data to gauge its effect. Yet we would think that someone raping or murdering every few years was quite a lot. Therefore, we try to look at large groups of offenders to see if there are clues. But offenders are not all alike, and we don't know what groups to divide them into to measure our predictive ability. Does this test work on all races? All ages? In different countries?
My personal turnaround in looking for that kind of one-stop shopping for recidivism answers came in the Q & A of a conference in the 1980's. A woman who was trying to get a non-profit started to do something-or-other about violent pornography - ban it, restrict it, keep it away from inmates, I don't remember - asked the presenter at the end of the day "Do you think exposure to pornography (and here she went into a tangent about tits negative portrayal of women, its availability, etc) contributes to recidivism?" She clearly wanted only one answer. The presenter looked down and shook his head. "Well, it might. Some studies have shown a weak effect. I'll say possibly yes. But I'll tell you this: I'd give any of the guys in my program a whole stack of magazines before I let him have even a single can of beer."
Of course. Duh. What was I thinking?
Among those who have been convicted, brain injury, developmental delay, and substance abuse are the biggest predictors of future offense. Well ahead of the others. However, those populations, taken as a whole, don't show much greater incidence of offense. It's just that if they are offenders, it's really hard to fix. Arson and abusing animals are warning signs. Age of first offense. Parental abuse or neglect are in the mix, but it is hard to sort out whether this is because you received the parent's abuse or their genes.
Anyway, it's not simple.