Crime and the City

(The following is an excerpt from my conference paper, Crime and the City: An Economic Model and Evidence, presented at International Conference on “Improving the quality of human life: multidisciplinary approach on strategic relevance for urban issues”
September 6-7,2007, JW Marriot, Surabaya, Indonesia).

ABSTRACT: Using the Metropolitan Police Force’s data on London crime, this paper aims to assess whether the observed spatial concentration of crime is due to underlying people-place characteristics or to spatial spillovers. To explain spatial variations of urban crime, this paper employs some techniques of spatial econometrics and put forward two perspectives, namely: economic model and social interactions. This paper finds evidence of spatial spillovers of crime between neighbouring areas and argues that the channels for such spillovers are peer effects between single and less educated people. The peer effects are found to be larger for economically motivated crime.

The first murderer (Cain) built the first city
(Genesis 10:6, as in Glaeser and Sacerdote (1999: S243))

1 INTRODUCTION
It has long been observed that big cities have higher crime rates than in small cities or rural areas. Glaeser and Sacerdote (1999) find the positive association between city size (measured as natural log of city population) and crime rates per capita as well as murders per capita in the U.S., while Gibbons (2004) presents a cross-countries comparison, showing that crime is positively correlated with the countries’ degree of urbanisation. On the whole, there is a general pattern that crime tends to be geographically concentrated in big cities, metropolitan areas, or highly urbanised areas.

In addition, it is also identified that there is a spatial concentration of crime within cities. In other words, there is unevenness in crime distribution within cities or there are indeed ‘hot spots’ of crime within cities. As cited in Zenou (2003), Grogger and Willis (2000) point out that central cities are more prone to most crimes than suburbs. It is documented that, in U.S. between 1985 and 1992, central cities’ average crime victimisation rates were 0.409 per household, while they were much lower in suburbs, i.e. 0.306 per household (Bearse, 1996 in Zenou, 2003). Gibbons (2004) shows that in Greater London, there are 44.3 burglaries per 1000 households in Hackney, while the rates in Sutton are 12 burglaries per 1000 households. In New York City (NYC), the rates of serious crimes per capita vary, from 0.022 serious crimes per capita in the 123rd precinct of NCY to 0.21 serious crimes per capita in the 1st precinct of NCY (Glaeser et al., 1996).

A simple explanation suggests that these stylised facts are due to the variances of economic and social conditions across areas within cities. The study by Hakim (1980) finds that property crimes have a direct relation with an area’s relative wealth, while Costello and Wiles (2001: 31) review some researches that generally find that “offenders offend in areas dominated by the same ethnic group as the offender”. In this regard, in a single city, we can expect that personal assault crimes will be much higher in certain areas than the others.

However, more recent study by Glaeser et al. (1996) finds that these demographic and social variables are only able to account for little (no more than 30%) of the variation in crime across areas in NYC. This finding is in line with the ‘neighbourhood effects’ or ‘peer interactions’ story in explaining the spatial patterns of crime within cities. This theory argues that it is not only economic incentives that determine individual behaviour, but it may also depend on the behaviour of peers or neighbours (Freeman et al., 1996). In other words, provided that committing crime may give the same expected returns, individuals may be prone to crime, if their peers commit crime than they do otherwise (Freeman, 1999). Therefore, social interaction models should be viewed more complementary than substitute to economic models.

With regard to such phenomena, this paper is aiming at investigating the spatial pattern of crime across London wards, based on the Metropolitan Police Force’s data and the 2001 U.K. Census Data. In order to be able to provide explanation on what might constitute the crime disparities in London, this paper develops models that incorporate both wards’ individual characteristics and spatial spillover effects. The computed value of Moran’s I suggests that the spatial autocorrelation in crime among London wards does exist, hence, following Anselin (1988), spatial lag variables should be taken into account.

To summarise the results of this paper briefly: spatial spillovers do account for sizeable differences in crime clusters across London wards and improve explanatory power of the model that is otherwise only explained by underlying population characteristics alone. This paper starts by reviewing related literature on the determinants of crime in Section 2, followed by empirical framework in Section 3. Section 4 presents and discusses the descriptive and estimation results, while Section 5 concludes and provides some policy implications.