Wednesday, March 03, 2004

Stopping First Crimes Best Way to Halt Criminality

Now here's a shocking discovery: deterrence works! All joking aside, this paper provides some evidence for a notion some might be tempted to dismiss as unfounded in reality.

The best way to combat casual crime is not to search for persistent offenders but to deter people from committing their first crime.

So say researchers at the London-based company Volterra Consulting who have studied the statistics of criminal acts. "The single most important thing is to persuade people not to commit a crime in the first place," says Paul Ormerod of the Volterra team.

They came to this conclusion after finding that people who do and don't commit crimes appear to be governed by slightly different statistical rules. The two types live in different mathematical worlds, the researchers say. A switch in statistical behaviour occurs just once, when a young person crosses the divide. Once having committed a single crime, a youth is statistically likely to go on to commit any number of further crimes, they report in a paper on the physics website arXiv1.

It might seem obvious that the way to cut crime is to stop people from doing it. But their recommendations contrast with some common approaches to crime prevention.


The team considered two very different studies. The Cambridge Study in Delinquent Development recorded the number of serious convictions in a group of about 400 males in North London over a 20-year period, beginning in 1961 when the boys were just 8 or 9 years old. In the Pittsburgh Youth Study, about 1500 boys in Pittsburgh schools were interviewed at regular intervals between 1986 and 2001. The participants were asked to report the number of acts of criminality or delinquency they had committed during each period.

The researchers expected that the number of crimes committed per person would fit a statistical distribution shaped like a bell if the criminal acts were being committed by random people in the selection: only a tiny fraction of boys would commit no crimes or lots of crimes, and most boys would fall into the average slot of committing a medium number of criminal acts.

Instead they found that that crime rates fell into a mathematical pattern called a power law, in which large deviations from average behaviour are more common. In both studies, most of the boys committed no crimes at all. In the Pittsburgh study, quite a few boys reported over 1,000 criminal acts during the study period, while the average number was just 90.

Physicists often find power-law statistics in systems with many interacting parts. This suggests that the young boys in the study are not responding randomly and independently to criminal opportunities that come their way. Instead they are probably influencing one another, presumably through strong peer pressure. (emphases added)

One fairly straightforward conclusion that might be drawn from this it isn't a good idea to concentrate a lot of poor young people together in inner-city public schools or in high-density housing "projects." In addition, whatever the direct educational benefits school vouchers may or may not have, they are probably still a good idea, in that they can help inner-city parents remove their children from negative influences by sending them to schools outside the areas in which they reside. Even a full voucher to attend the likes of Groton or Andover Academy is probably much cheaper in the long run than allowing a young person to fall into a life of crime; housing, clothing and feeding an incarcerated person can be surprisingly expensive.

1: W. Cook, P. Ormerod, E. Cooper, "Scaling Behavior in the Number of Criminal Acts Committed by Individuals", 2004.