A Review and Analysis of the Impact of Homicide Measurement on Cross-National Research

Abstract

The number of cross-national homicide studies is increasing rapidly. Many scholars, however, do not consider the details of how individual nations and the four main centralized homicide data sources—raw estimates from the World Health Organization (WHO) Mortality Database, adjusted estimates from the WHO Global Health Observatory, United Nations Office on Drugs and Crime, and World Bank World Development Indicators—generate national homicide rates and the impact this may have on results and the scientific record. We tested whether homicide trends, levels, and structural covariates are dependent on data source. We used 1990–2018 data in 5-year groupings and pooled them over time and nation. We utilized exploratory data analysis techniques to look for differences in homicide rates and trends. Then we employed seemingly unrelated regression (SUR) to determine whether associations with homicide of typical structural covariates were dependent on homicide data source. Finally, we examined Wald Tests to determine whether differences in the sizes of the SUR coefficients from each data source were significantly different from zero. We found differences in homicide trends and rates by data source and that associations with homicide rates of structural covariates varied in significance, magnitude, and even direction depending on homicide data source. Cross-national homicideresearch has a promising future for understanding short- and long-term global and regional trends and population-level covariates and constructing theoretical explanations for geographical and temporal variation. However, researchers must better understand how national homicide data are generated by nations and these four data sources. All four systems possess limitations, but homicide data from the WHO Mortality Database present the most attractive option.