Explaining discrepant findings in cross-sectional and longitudinal analyses: An application to U.S. homicide rates

Cross-sectional studies often reach different conclusions regarding the association between key explanatory variables and outcomes than those of longitudinal approaches. This study considers possible explanations for discrepant findings using a decomposition approach with panel data on 404 U.S. counties for the period 1970–1999. The analysis establishes that there are important differences in the effects of independent variables on homicide rates across counties as opposed to within counties over time. Explanations offered for these discrepancies are that variables may have differing temporary (flow) and permanent (stock) influences on outcomes and possible omitted variable bias. The findings highlight the importance of distinguishing among possible stock and flow effects and are significant not only for the study of crime but also for other social phenomena.