An Analysis Of The Correlation Between Population And Crime Rates

Social scientists have tried to explain geographic concentrations of crime for more than a century. Two macro-sociological explanations are often investigated.

First, social disorganization theory suggests that some ecological factors, such as social deprivation and residential mobility, lead to a lack of collective efficacy, which in turn leads to crime. The theory is “one-sided” — the presence of some community factors decreases collective efficacy, and that increases community crime. Second, routine activity theory states that crime is more frequent where likely offenders meet suitable targets in the absence of capable guardians.

In terms of population, routine activity theory offers conflicting propositions: an increased human presence in a given area is expected to be associated with both an increase and a decrease in criminal activity.

Expectations from urban studies are consistent: Jacobs’ optimistic model expects that mixed land use draws large numbers of visitors who provide effective informal control of public space – they act as guardians. Taylor’s territorial model rather predicts that mixed land use increases the probability that offenders and targets will converge – and thus, increase crime. Furthermore, empirical studies testing social disorganization theory have, so far, provided strong support; the support for routine activity theory is somewhat mixed, with some studies finding that population is positively linked to crime and others, that it is negatively linked to crime.

The current study aims to investigate a potential explanation: crime might be positively linked to population in some areas while being negatively linked to population in others. The study is also a challenge to conventional statistical modeling that assumes that relationships are the same for all units of analysis, here that the effects of key factors on neighborhood crime are the same for all areas. On the contrary, this study aims to “think locally,” that is, to calculate a local coefficient for all study areas.

To do so, the Toronto Police Service provided police-recorded statistics for 2011 at census-tract (CT) level. Counts of crimes in a CT were linked to Census indicators (residential population size, a scale of social disorganization and residential mobility) and to estimated numbers of visits taken from a large transportation survey (visits for school, shopping, work, and entertainment). A continuous index of mixed land use was also computed to account for everything from the completely segregated areas (e.g. residential neighborhoods) to the notably diverse activity areas. Firstly, conventional global analyses confirmed previous findings: for example, the level of social disorganization was positively associated with the number of assaults, auto thefts, robberies, and sexual assaults in a CT. Also, analyses provided strong support for Jacobs’ model, because mixed land use was negatively associated with all CT measures of crime.

But the analysis also revealed unexpected results: the level of social disorganization was negatively associated with the number of burglaries and thefts. So instead of arguing that social disorganization is sometimes wrong, another type of analysis was conducted: geographically weighted regression models. This type of analysis produces local estimates, one for each of the 547 census tracts of Toronto. Simple calculations allowed us to determine if observed relationships were statistically significant.

It turned out that for four of the six crime types under study, social disorganization was (as expected) positively linked to crime, that for burglary, the level of social disorganization was usually not related, and that it was negatively related to the number of thefts in a CT. For mixed land use, only with robbery were significant negative relationships found for a majority of tracts. For other crime types, a positive relationship was found for up to a third of tracts. In other words, the analysis found support for both Jacobs’ and Taylor’s models, despite the fact that they predicted opposite relationships.

This paper investigated the empirical relevance of opposite propositions but did not attempt the spatial distribution of significant and non-significant relationships. Future research could, for example, attempt to answer questions such as, “What is the desired amount of mixed land use in terms of crime and security?”

These findings are described in the article entitled Routine activity, population(s) and crime: Spatial heterogeneity and conflicting propositions about the neighborhood crime-population link, recently published in the journal Applied Geography. This work was conducted by Rémi Boivin from the Université de Montréal and the International Centre for Comparative Criminology.