Defining and measuring homicide rates for birth cohorts: Methodological and theoretical challenges and solutions

Abstract

Social scientists have long been interested in understanding how age, period, and cohort effects shape long-term homicide trends. Yet fundamental measurement challenges remain pervasive in estimating age-specific homicide rates for birth cohorts. We begin by defining multiple alternative approaches for measuring homicide rates for birth cohorts, highlighting their advantages and disadvantages. We empirically demonstrate substantial differences in the estimated cohort-specific homicide rates across alternative measurement approaches. These discrepancies are particularly large when using 5-year age categories and 5-year birth cohorts, with a nearly twofold difference in estimated homicide rates for 15- to 19-year-olds in some cohorts. Substantively, these discrepancies have consequences even for basic questions about homicide trends, including which birth cohorts had the highest homicide rates. We provide a framework for understanding why these discrepancies arise and propose the reconstructed cohort measurement approach as a practical solution to produce valid estimates given available data. Simulation analyses confirm this approach substantially outperforms approaches used in the prior literature. Lastly, we demonstrate that these measurement challenges can lead to fundamentally different results in regression models predicting cohort effects on homicide rates. Our results underscore how accurately measuring cohort-specific homicide rates is an essential, yet often overlooked, part of understanding homicide trends.