Work

When Homicide Statistics Contradict Themselves: A Reproducible Simulation of Within–Between Reversals in Comparative Criminology

Year: 2026
Type: article
Abstract: International homicide comparisons routinely place countries with different income levels on the same graph. The association across those countries is sometimes discussed as though it described what happens when income changes inside a country. The two comparisons, however, draw on different information. We examine their distinction through a reproducible simulation, alongside published regional homicide estimates from the United Nations Office on Drugs and Crime. The experiment contains 120 synthetic units observed over 14 periods. A negative between-unit coefficient and a positive within-unit coefficient are imposed in the generating process, then estimated under sampling noise and alternative analytic choices. The reference estimates are −0.606 and +0.516, with a within-minus-between contrast of +1.122 and a unit-bootstrap 95% interval of +1.041 to +1.203. Further simulations examine sampling precision, incomplete coverage and specification changes. The UNODC estimates establish that homicide rates vary across regions. They do not identify an income–homicide relationship. Opposing coefficients in the generated panel demonstrate an inferential possibility, not a reversal observed among actual countries. The resulting contribution is a replicable test of whether the statistical comparison reported in comparative homicide research matches the proposition used to interpret it.
Language: English
Open Access status: hybrid
OpenAlex ID: W7227516835
FWCI:
Cites: 49
Cited by:
Related to: 10