This study used seemingly unrelated regression techniques to assess and compare family and other structural covariates of homicides by males and females across 1,600 counties.
The study found that both female and male homicide rates were significantly higher in counties with greater levels of family disruption (female-headed households), after controlling for other structural features of the community. A 1-percent increase in a community's level of family disruption multiplied homicides by females 11 percent and homicides by males 25 percent. Thus, a community that has a high level of family disruption (e.g., 20 percent of families are headed by females) can expect to have a 55 percent higher female homicide rate and a 125 percent higher male homicide rate than a community with average levels of family disruption (15 percent of families are female headed). The data for the independent and control variables were obtained mainly from U.S. Bureau of Census 2000 summary Tape Files 3. These files contain sample data weighted to represent the total population for small geographic units, including counties. The dependent variables were sex-specific adult homicide arrest rates, which were collected from the Uniform Crime Reporting Program for 2002. The key independent and control variables were family structure (male-female headed households and female-headed households). The sex-specific control variables involved an index of structural disadvantage, including standardized measures of female and male unemployment, the percentage of Black females and males, and family poverty. A sex-specific measure of age structure was included to control for county-to-county differences in the size of the population that was most prone to violence. Researchers used seemingly unrelated regression techniques that adjusted for autocorrelation among female and male models and the most current data from the U.S. Census Bureau's data and the Uniform Crime Reports. 4 tables, 12 notes, and 90 references