Thinking beyond averages: Quantile regression modelling and military expenditure

Abstract

The study of military spending has been an enduring concern within military sociology and political science. Methodologically, one of the biggest challenges lay in dealing with its heavy-tailed distribution influenced by the growing separation between China and the United States from the rest of the world. In the presence of outliers along the continuum of military expenditure, we should be paying more attention to portions of the distribution that don’t assume the values reported at the conditional mean. The article uses quantile regression modelling (QRM) to analyse the nuanced relationship between military expenditure and its predictors. It argues that classical linear regression produces average estimates that cannot predict values at different subsets of the data’s distribution, meanwhile QRM has relevant results in the search for noncentral values in the study of military expenditure often laying in the lower and the upper tails of the distribution.