Unemployment change and homicide: an exploration of the national violent death reporting system

Two studies were undertaken as part of this project. We used homicide data collected by CDC's National Violent Death Reporting System (NVDRS), a state-level active surveillance system that provides data on all violent deaths in 16 US states. Data were obtained for 2003-2009. We used the NVDRS, unemployment data from the Bureau of Labor Statistics (BLS), and Current Population Survey (CPS) to estimate associations between unemployment level and homicide through three research aims. First, we used all NVDRS homicide cases to examine the association between monthly change in unemployment and homicide rates. Information on homicides and population estimates were analyzed by Poisson regression to estimate rate ratios as a function of change in unemployment level over month and quarter in which a homicide event occurred (Aim 1). After adjustment for age, gender, race, median household income, and population density, county-level homicide rates increased an average of 2% (Rate ratio = 1.02; 95% CI: 1.00 - 1.05) per percentage point increase in unemployment level over the prior month. Unadjusted rate ratios for unemployment decreases of 2.5 percentage points or greater were dramatically more protective against homicide as compared to any other unemployment decreases (Rate ratio: 0.19; 95% CI: 0.15 - 0.25). We used a case-crossover design to examine the change in risk of experiencing a workplace homicide as unemployment levels changed over a 1-month period (Aim 2). We examined unemployment change data for the month homicide event occurred (case period) and the two months before and after the case period (control periods). Conditional logistic regression models estimated the unemployment change-workplace homicide association across strata of community and victim-level characteristics. Third, we assessed heterogeneity in the association by characteristics of the victim and workplace violence type (Aim 3). A 1-percentage point increase in unemployment over one month was associated with a small increase in the odds of a workplace experiencing a homicide (OR = 1.03; 95% CI = 0.94 - 1.12). County-level population density modified the odds ratio, and homicide risk was heterogeneous among victim race and workplace violence type; however, no measure of the unemployment-workplace homicide association resulted in a statistically significant effect measure.Doctor of Philosoph