The fallacy of global comparisons based on per capita measures

Abstract

Social and environmental scientists often present comparisons across countries and policy makers use such comparisons to take evidence-based action. For a meaningful comparison among countries, one often needs to normalize the measure for differences in the population size. The first choice is usually to calculate per capita ratios. Such ratios, however, normalize the measure for differences in population size directly only under proportional increase of the measure with population size. Violation of this assumption frequently leads to misleading conclusions. To illustrate this issue, we compare per capita ratios with an approach based on regression, which eliminates many of the problems with ratios and allows for a straightforward data interpretation. The per capita measures in three global datasets (GDP, COVID-related mortality, and CO2 production) systematically overestimate values in countries with small populations, while countries with large populations tend to have misleadingly low per capita ratios due to the large denominators. Despite their biases, comparisons based on per capita ratios are still ubiquitous and are used for influential recommendations by various global institutions. Their continued use can cause significant damage when employed as evidence for policy actions and should be replaced by a more scientifically substantiated method, such as a regression-based approach.