Homicide mortality in the United States, 1935-1994: age, period, and cohort effects.

The authors analyzed homicide mortality data for the United States from 1935 to 1994, to delineate temporal trends and birth cohort patterns. This study included 850,822 homicide-attributed deaths documented by the National Center for Health Statistics, and incorporated graphical presentation, median polish, and Poisson regression modeling in an age-period-cohort analysis. Death rates from homicide in the United States doubled in the past four decades, with most of the increase having occurred during the 1960s and early 1970s. Poisson regression models confirmed that the rise of homicide mortality in both men and women was largely attributable to a significant period effect between 1960 and 1974. No discernible cohort patterns were found among women. However, homicide rates for recent male birth cohorts appeared to peak at younger ages and at higher levels. A significant increase in homicide mortality risk beginning with males born around 1965 was found by examining the residuals of median polish, and the second-order changes in the regression coefficients from the age-period-cohort model. The hike of homicide mortality during 1985 and 1994 was explained by this cohort effect. Increased prevalence of substance abuse and availability of firearms are two likely factors underlying this disturbing cohort pattern.