Predictive Validity of Lott & Varney's (2022) OLS Model of European Homicide Rates

Abstract

Purpose: To assess the predictive validity of Lott and Varney's (2022) ordinary least squares (OLS) multiple regression (MRA) causal model of the effect of immigrant population numbers, as a proportion of the total population, on the homicide victim rate for 31 European countries.

Methods: The out-of-sample errors (the 315 errors from forecasting each observation using a model estimated from the other n-1 (314) observations) from the Lott and Varney specified MRA model forecasts were compared with the out-of-sample forecast errors from five alternative models. Three of the five were estimated using one of two alternative estimation methods, and one of those alternatives differed from the Lott and Varney model only in the estimation method.

Findings: Lott & Varney’s ordinary-least squares regression estimated model provided out-of-sample forecasts that were more accurate than the forecasts from a naïve model that predicted that the murder victimisation rate in a country would be the same as median figure for that country, but were less accurate than when the Lott and Varney model was estimated using the conceptually simpler and more conservative least absolute deviation method.

Limitations: The findings in this research note are for only one OLS MRA model, and the forecast errors calculated using alternative error measures found that none of the models was best against all accuracy criteria.

Implications: The result is consistent with my ongoing research that suggests that the predictive validity of causal models can be improved by using simpler more conservative alternatives to OLS regression estimation.

Keywords: causal models, conservatism, model estimation, predictive valdity, simplicity, regression analysis

JEL Classification: C1, C2, C5, Z18