The Curious Case of Percent Young: Replications and an Atypical Meta-Analysis

• In Event: Cross-National Studies and Theories of Homicide and its Characteristics (Sponsored by Division of International Criminology)

Abstract

Objective: In the cross-national homicide literature, the percent of the population that is young is inconsistently associated with homicide victimization. The inconsistency is likely due to the partialling fallacy. I test if there is evidence of collinearity and a latent construct. Then using an atypical meta-analysis, I explore if the heterogeneity of effects sizes in percent young can be accounted for by including redundant variables.

Methods: I replicated previously published studies generating measures of collinearity and test for a latent construct. Using articles published before May of 2021, I obtained 95 partial correlation coefficient effect sizes for percent young. Moderator variables are the presence of economic, education, income inequality, and/or poverty variables within published models. I estimated RVE meta-regression models.

Results: Collinearity existed in previously published studies. Across the replications, the variables of interest reduced to a singular latent construct. The effect sizes for percent young were significantly smaller in the published models that included economic and education variables.

Conclusions: There is a systematic downward bias of the effect sizes of percent of the population that is young because of the inclusion of economic and education variables. I conclude by situating these findings within the larger literature.

Author

• Meghan Rogers, University of Iowa