Monday, August 13, 2007

Predicting crime

Some information in predicting crime. There are a several interesting areas:

1) Predicting changes in overall crime rates (say within a nation or community). The two major factors useful for doing so are either demographic or economic, as indicated in this article from the Canadian Justice Department. The key demographic factor is the number of male between 15-25; crime rates also correlate with economic conditions (although there seems to be sme ambiguity as to whether the link is positive or negative, which makes me wonder what is really known...some argue more wealth leads to more crime ---there's more to steal -- while others say more wealth makes for less crime -- fewer people need to steal).

2) Predicting slightly more specific changes by looking for meaningful patterns. For example, there are a number of well documented seasonal cycles, as described at Crimepsychblog. For example, I quote:

Peaks in the summer months and troughs in the winter months: […Two sexual] offences follow very similar seasonal patterns to each other with a large peak in July, when indecent assault on a female is 21 per cent above trend and rape of a female is 14 per cent above trend. Theft of a pedal cycle has a very clear seasonal pattern; peaks start in May and continue to reach 29 per cent above trend in September… Arson, unlike other criminal damage offences, shows rises in summer months…
Peaks in the winter and troughs in the summer months: Just four crime types display seasonal peaks in the winter and falls in the summer. These are
all property crimes. Domestic burglary peaks to 11 per cent above trend in January. … It is important to note that domestic burglary has an opposite seasonal pattern to nondomestic burglary.


3) Predicting more specific changes by treating crime in an epidemiological kind of way, realising that it has correlations in space and time. See, for example, the work of Kate Bowers, Shane Johnson and Ken Pease. Or work showing that one burglary in an area increases the chance of another crime within a few hundred yards for several months. For example:


Crime Patterns of Risk: A Cross National Assessment
of Residential Burglary Victimization
Shane D. Johnson Æ Wim Bernasco Æ Kate J. Bowers Æ Henk Elffers Æ
Jerry Ratcliffe Æ George Rengert Æ Michael Townsley
 Springer Science+Business Media, LLC 2007

Abstract Using epidemiological techniques for testing disease contagion, it has recently been found that in the wake of a residential burglary, the risk to nearby homes is temporarily elevated. This paper demonstrates the ubiquity of this phenomenon by analyzing space–time patterns of burglary in 10 areas, located in five different countries. While the precise patterns vary, for all areas, houses within 200 m of a burgled home were at an elevated risk of burglary for a period of at least two weeks. For three of the five countries, differences in these patterns may partly be explained by simple differences in target density. The findings inform theories of crime concentration and offender targeting strategies, and have implications for crime forecasting and crime reduction more generally.


There is, of course, also the notable work of Glaeser, Sacerdote and Scheinkman, showing that crime has significant spatial fluctuations, more than can be accounted for by economic fluctuations:

The high degree of variance of crime rates across space (and across time) is one of the oldest puzzles in the social sciences (see Quetelet (1835)). Our empirical work strongly suggests that this variance is not the result of observed or unobserved geographic attributes. This paper presents a model where social interactions create enough covariance across individuals to explain the high cross- city variance of crime rates. This model provides a natural index of social interactions which can compare the degree of social interaction across crimes, across geographic 1units and across time. Our index gives similar results for different data samples and suggests that the amount of social interactions are highest in petty crimes (such as larceny and auto theft), moderate in more serious crimes (assault, burglary and robbery) and almost negligible in murder and rape. The index of social interactions is also applied to non-criminal choices and we find that there is substantial interaction in schooling choice.


In another paper, Glaeser and Sacerdote note that crime is much more pronounced in larger than in smaller cities, although they don't go a long way to explaining why. I wonder: could this simply reflect another scaling law like those seen in many other variables such as energy consumed, the average speed of walking, etc? [I can't find the link to this at the moment, but it's the work, I think, of Geoffrey West, Luis Bettencourt and others].

A few more random observations on crime patterns:

* Another link from Crimepsychblog
to work showing that the sites where crimes are committed are not equally. Rather, they tend to be power law distributed in terms of the likelihood of crime:

This one points out the maths of crime: it is almost always a small fraction of any kind of place (convenience store, bar, etc) that accounts for most of the crimes. Which is maybe not so surprising as most everything works this way. But this is undoubtedly useful for policy, as in treating the spread of diseases.


The paper to see is Understanding Risky Facilities published by the US Department of Justice.

* Another interesting idea is that of Hot Products -- particular products that, in a physics sense, act as catalysts for crime. Think mobile phones or other expensive electronics stuff. A report from the UK Home Office gives more detail:

The mobile phone
A good example of a hot product is the mobile phone. Expansion of the mobile phone market has been rapid. The targeting of mobile phones is already a factor in increasing rates of street robbery. As mobile phone handsets incorporate internet technology mobile phone crime is likely to continue and increase. The ‘no-contract’ mobile phones are particularly attractive to criminals because: they allow greater anonymity for callers; there are loopholes in the ‘pay as you go’ mobile phone schemes that enable knowledgeable users to switch to other networks and avoid payment; the vouchers used to pay for calls can also be targeted for theft; vouchers are easy to reproduce; and, criminals have reprogrammed ‘prepaid’ mobile phones to obtain free calls (reported in Association of British Insurers 2000: 14). http://www.insurance.org.uk/ResearchInfo/ 
Examples of future hot products include (Association of British Insurers 2000) 
The launch of Digital Television and the switch off of analogue transmissions may spark a crime epidemic in the rush to replace obsolete televisions. However, the offer of set-top boxes free of charge is likely to reduce their attractiveness to thieves.
Portable Digital Virtual Disc (DVD) players are now available weighing 900 grams and costing up to £1000.

No comments: