Police numbers

First post for a while: I thought it was about time to start the blog up again.

The number of police officers is becoming more and more of a hot issue (see for example here, here and here). In fact I guess it always is. The prime reason for believing that the number of serving police officers should be maintained, or even increased, is of course always that crime will increase if they are reduced.

It seems such a simple equation: more police officers equals less crime, and less police officers equals more crime. So simple, in fact, that most criminologists assume it must be (deceptively) wrong. In general I think I agree. The police are far from the being the only institution that might affect the causes of crime, which lie in a complex mix of personal, social, institutional and environmental factors. Certainly self-report surveys find only weak associations between offending and the risk of sanction - in other words, when people think about offending, they don't really think about how likely they are to be caught by the police if they do break the law. Even when it comes to more definite preventative measures this seems to be much more to do with 'target-hardening' by individuals and organisations than it does with police patrols or other activities.

All in all it seems rather unlikely that police 'crime-fighting' activity can have much impact on the general level of crime, for all that it might be effective in relation to specific areas and/or specific crimes. This was certainly the consensus in academic work up until about 2000. Studies regularly found either no association between numbers of police and levels of crime, or indeed a positive association (i.e. more police were associated with higher levels of crime, even in studies which included relevant statistical controls).

But I've been doing a piece of work in the last few days that has challenged my assumptions. Looking at a small body of papers published in the last 15 years or so, it seems the number of police officers may indeed have a negative association with recorded crime rates. Based on a number of observational studies, quasi-experiments and natural experiments, there's a surprising consensus around an elasticity of about -0.3, which would imply that a 10 per cent increase in police officers leads to a 3 per cent reduction in crime.

Now, there are a number of problems with the research I've been looking at. Most obviously, it usually looks at associations between numbers of police and recorded crime, which is obviously only a small fraction of actual crime. Another is that much of it has been produced by economists and has as such an impoverished theory of crime causation, essentially rational choice theory, that may not provide a robust enough causal mechanism for the observed correlations. In other words there are significant doubts that the research can explain the associations it uncovers. Yet if nothing else these studies raise some very interesting questions.

I'll write more when I'm finished. It's always interesting to have your preconceptions put under the spotlight, I think ......