Nonparametric identification of factors for the cross-section of Latin American stock returns

Abstract

Factor Investing, a widely recognized investment approach, remains relatively underexplored in Latin American stock markets. This study investigates the determinants of stock returns in Latin America employing the Group Adaptive Elastic Net within a nonparametric framework from 2000 to 2020. Initially assessing 34 widely recognized financial factors, our analysis identifies that a set of six factors captures most of the variance in cross-sectional returns in the proposed model: Volatility, Assets-to-Market ratio, Cash Flow to Price, Earnings to Price, Intermediate Momentum, and Turnover. Furthermore, an active Factor Investing strategy derived from this framework demonstrates substantial outperformance relative to a benchmark index in out-of-sample testing. Overall, we find evidence of short-term predictability of returns in Latin American stocks based on nonlinear and dynamic factor effects.