Lost in the multiverse: Methodological uncertainty in studying global equity returns

Abstract

We examine methodological uncertainty in studies of the cross-section of country equity returns, analyzing 15 predictors across up to 13,824 research implementations. Varying nine key design choices, we find that many classic signals—such as momentum, beta, or idiosyncratic risk—are surprisingly fragile. Setups emphasizing small, segmented countries enhance performance, while those focused on liquid, investable markets tend to weaken it. Applying bootstrap and out-of-sample tests, only a few factors, such as market size, dividend yield, short-term momentum, and sovereign risk, consistently emerge as robust. Our evidence calls for cautious interpretation of country-level return patterns and for robustness checks across alternative research designs.