Emerald Group Publishing Limited
Copyright © 2008, Emerald Group Publishing Limited
Policing, drugs and the homicide decline in New York City in the 1990s
Steven F. Messner, Sandro Galea, Kenneth J. Tardiff, Melissa Tracy, Angela Bucciarelli, Piper Markham, Victoria Tinka Frye and David VlahovCriminologyVol. 45 No. 22007pp. 385-414
Messner et al. (2007) examined New York City homicide data at the precinct level to test Wilson and Kelling’s broken windows theory. This study attempted to improve on previous research in three ways. First, Messner et al. (2007) disaggregated homicide measures to indicate whether or not a gun was used. Second, they created a new indicator of cocaine use. Third, they added felony arrests into the model to examine whether or not areas with increased misdemeanor arrests have more arrests generally. Messner et al. (2007) used data from the Office of the Chief Medical Examiner (OCME), the NYPD, and the US Census Bureau from 1990 to 1999.
Messner et al. (2007) only examined homicides as their measure of crime. However, unlike previous studies, the authors separated homicides that were gun-related versus those that were non-gun-related. They also departed from previous measures of cocaine use. Previous studies measured cocaine use by hospital discharge records; Messner et al. (2007) measured cocaine use by the number of deaths related cocaine overdose. The authors measured both the misdemeanor arrest rate and the felony arrest rate. The study used the following control variables: percent male, percent under age 35, and percent unemployed. They use a composite socio-economic status (SES) composed of: percent female-headed households, percent of the population on public assistance, percent under 200 percent poverty, percent persons with less than high-school education. Messner et al. (2007) also controlled for the availability of firearms (measured as the suicide rate as a proxy) and the change in police manpower allocations.
The bivariate analysis indicated that areas with higher misdemeanor arrests had a lower homicide rate than those with fewer misdemeanor arrests. Model 2 adds change in cocaine, percent male, percent under age 35-years-old and percent unemployed. Messner et al. (2007) found that changes in misdemeanor arrest rate, changes in cocaine use and percent unemployed were significant predictors of the homicide rate. Model 3 added the remaining control variables: change in firearm availability, percent black, SES and change in manpower. This model revealed that only change in misdemeanor arrest rates and change in cocaine use significantly predicted changes in homicide rates.
Messner et al. (2007) tested whether the type of homicide is relevant to the effect of misdemeanor arrests on homicide rates. They found that significant predictors in model 3 also had effects on gun-related homicides but none of the predictors in model 3 had significant effects on non-gun-related homicides. The authors were surprised that firearm availability had no effect on gun-related homicides. Messner et al. (2007) offered two possibilities 1) that the proxy measure of suicide-related deaths is not a good indicator, or 2) gun availability may be indirectly related to homicide rates. Finally, Messner et al. (2007) examined whether or not changes in felony arrest rates mitigate the effects of misdemeanor arrest rates. They found that changes in felony arrest rates had no significant impact on homicide rates.
The authors concluded that broken windows theory policies did have a limited impact on the New York City homicide decline of the 1990s. Messner et al. (2007) found some support for the idea that crack-cocaine markets were involved with homicide rates. However, they do make the concession that their measure, cocaine overdose deaths, cannot distinguish between crack and powdered cocaine related deaths.
Daniel J. Lytle MS University of Cincinnati, Cincinnati, USA