Neighborhood Risks of Violence Against Women

Jeffrey Fagan, Columbia University

Research on the social epidemiology of violence against women has focused primarily on individual characteristics of victims or offenders, or event-level or situational characteristics. Although theory and research on interpersonal violence has long been concerned with social structural explanations and dynamic theories of neighborhood social organization, only recently has there been attention to the spatial dimensions of violence against women. Additionally, there is virtually no research that compares social or neighborhood risks of violence against women with violence against males. Nor has there been research that examines the links between the two patterns of victimization in a broader analysis of the spatial ecology of violence.

The research will examine violence in New York City neighborhoods from 1985-2000. Data from the Injury Surveillance System of the New York City Department of Health have been geocoded into new homogeneous neighborhood units. These recently developed spatial boundaries define 330 neighborhoods in New York City, areas that were identified based on survey methods where residents perceptually the boundaries of their residential and commercial neighborhoods. The data were aggregated to construct measures of violence against men, violence against women, intimate partner femicideslethal violence against women and non-lethal violence against women. The analyses will have several unique features. First, we will use data sets that permit estimates of violence over time. Second, we will estimate separate models for lethal and non-lethal injuries. Data on fatalities is available for the entire study period; data on non-fatal injuries are available beginning in 1990. Third, we will estimate separate models for gun and non-gun violence, an important feature of the epidemiology of gun violence in New York over the past two decades. Fourth, we will include measures of spatial autocorrelation to estimate the extent to which spatial dynamics affect patterns of violence. Fifth, because of the changing acharacter of New York City neighborhoods over the nearly two decades in this project, we will estimate neighborhood risks using time-varying covariates to determine how neighborhood change alters both the general and gender-specific risks of violence. All models will estimated using mixed-effects Poisson regressions with overdispersion components and controlling for spatial autocorrelation.

(Return to Program Resources)

Updated 05/20/2006