Crime and neighborhood disorder: A spatial analysis in two cities

This paper examines the relationship between crime and neigborhood disorder, specifically physical disorder, in both Boston and Chicago. Publicly available geocoded records  of citizen requests to city 311 services, in addition to geocoded police incident reports, are mapped to both the census tract and census block group levels. Service requests are divided into four categories that can readily be characterized as contributing to physical disorder: abandoned buildings, abandoned vehicles, graffiti, and sanitation issues. Both crimes and service requests are aggregated monthly for both geographic levels, resulting in thirty-seven and forty-five month panels for Boston and Chicago, respectively.

The nature of the disorder data helps alleviate the potential response (or other) bias inherent in previous studies that have relied mainly on questionnaires, interviews, or field data collection by researchers. Perhaps more importantly, the panel and spatial nature of the data allows for the inclusion of neighborhood fixed effects in the analysis, eliminating systematic differences in what is considered “disorder” versus the everyday norm across neighborhoods, as well as testing for the presence of spatial spillovers in the dependent and independent variables.

Standard OLS and fixed effects models find large, positive associations between increases in disorder and increases in total and index crimes in both cities at the larger, census tract level. There is, however, a high degree of seasonality in disorder and crime; both tend to rise in the warmer months and fall in the colder months, a predictable and previously studied phenomenom. Utilizing time-period fixed effects with spatial panel models drastically reduces the predictive power of all measures of disorder on crime, completely eliminating any signficant relationship in Boston. In Chicago, a 10% increase in graffiti related service requests is associated with a 0.14% increase in total crime in the tract. There is also very little evidence of spillover effects of disorder on crime in neighboring tracts.

At the block group level, however, all four measures of disorder are positively related to crime in Boston at a similar magnitude. In Chicago, where there also is evidence of spatial spillover into surrounding areas, two of the measures are now associated with increases in crime. These results suggest that not only does a relationship between physical disorder and crime exist at a highly localized level, increases in physical disorder in one area may cause crime to spread into nearby communities.