On the Limits of Social Control: Structural Deterrence and the Policing of “Suppressible” Crimes

Abstract

The present study examined the relationships between patterns of police arrests and subsequent variations in robbery, burglary, and aggravated assault in New York City police precincts from 1989 to 1998. Grounded in the structural deterrence theoretical perspective, and using a two‐stage fixed‐effects statistical framework, the study found that while controlling for indicators of social disorganization, increases in arrest vigor (i.e., arrests per officer for violent crimes in each precinct and raw arrest counts) predicted decreases in robbery and burglary, but that the relationships were non‐linear: as arrest vigor increased, robbery and burglary crime decreased; when arrest thresholds were reached, however, both robbery and burglary crime rates became positively associated with arrest aggressiveness. Conversely, variations in aggressive arrest patterns had no significant effect on aggravated assault, supporting the suppressible crimes arguments that primarily economically motivated crimes, and those that tend to occur in public settings, are most likely deterred by aggressive police practices.

Acknowledgments

The work for this article was completed while the author was an assistant professor in the Department of Justice, Law & Society at American University (and in part under NIJ grant #1996IJCX0053). The author gratefully acknowledges six years of support and friendship from a group of wonderful colleagues. Thank you. This article, and all subsequent work produced from these data, is dedicated to the memory of Dr. James J. Fyfe, who introduced me to the worlds of police scholarship and the New York City Police Department. Any professional success I might achieve is the direct result of his generous and careful mentoring; my professional failures and shortcomings are despite his best efforts. Rest in Peace, Jim.

Notes

1. For a variation on this measure, the interested reader might see Kane (2005), who used the arrest ratio to measure both over‐ and under‐policing in communities characterized by differential levels of structural disadvantage.

2. In an effort to control for possible serial correlation between robbery arrests per officer in theindependent variable and robbery crime rates in the dependent variable, the study tested the hypotheses using two versions of the arrest vigor variable: one that included robbery arrests in the composite, and one that did not. The models produced virtually identical results in all estimations, and the study reports on the results using the arrest vigor variable that included robbery.

3. The assignment of 1990 census measures (e.g., structural disadvantage and population mobility) to precincts from 1990–1998 assumes that precincts were invariant over time along these variables. This empirical assumption is likely inaccurate, but represents an inherent component and limitation of many longitudinal designs that incorporate decennially measured census indicators. Bivariate correlation analyses of the structural disadvantage and population mobility measures showed the following: between 1980 and 1990 structural disadvantage r = .85, population mobility r = .67; between 1990 and 2000 structural disadvantage r = .89, population mobility r = .65. That is, although precinct structure changed over time in New York City, levels of disadvantage and residential mobility correlated strongly with one another at 10 year intervals that both predated and postdated the temporal period under study. Nevertheless, it is likely most appropriate to consider this longitudinal test of macro‐deterrence while holding constant cross‐sectional differences in the antecedents of social disorganization.

4. The locations of precinct centroids were identified using Linear Referencing in ArcGIS by ESRI. Spatial Analyses were conducted with Spacestat and ESRI’s ArcMap; multivariate models were estimated in LIMDEP (version 7.0).

5. Including both a linear and non‐linear term in the same multivariate model creates special considerations for multicolinearity due to their functional dependence. In the present study, the quadratic form of the arrest aggressiveness variable (i.e., arrest aggressiveness‐squared) was correlated at .65 with the linear arrest aggressiveness variable, which falls under a generally acceptable threshold.

Additional information

Notes on contributors

Robert J. Kane

Robert J. Kane is an assistant professor in the College of Criminal Justice at Northeastern University. His primary areas of research are police authority and accountability, the social ecology of crime, the spatial distribution of crime and police coercion, and police‐citizen conflict in urban settings. Professor Kane has published articles on police legitimacy and procedural justice in structurally disadvantaged communities, police misconduct, police department deployment practices, community policing, the police response to domestic violence, the role and behavior of police in public housing, and patterns of substance abuse. He recently conducted a comparative study of police authority and accountability in England and the Netherlands. Professor Kane holds a doctorate degree in criminal justice from Temple University.