Localized risk factors: Performance differentials between state-level and US factor models

We extend the literature on the debate on whether global or local factor models more accurately price assets by comparing US factor models with state-specific localized versions. We show performance differentials between localized and market-wide models even within a country. Using a comprehensive set of factor models and anomaly portfolios as test assets, we show that state-level risk factors tend to outperform their US-wide counterparts. Additionally, US-wide factor models do not span local factors in most cases but can explain correlations of portfolio returns across states. Finally, we show that state-level characteristics as well as the intra- and inter-state return comovement affect the performance gap between state-level and US factor models. Increases in return comovement across states reduce the performance gap between models, while increases in comovement within states raise the latter. The results have important implications for the estimation of the cost of capital as well as portfolio diversification. • We construct anomaly portfolios and factor models for individual states in the US. • State-level factor models fare better in pricing anomaly portfolios than US models. • Squared Sharpe ratio tests show that some US models outperform their local counterparts. • Performance differences are affected by state characteristics and return comovements. • Increased intra (inter)-state return comovement raises (reduces) the performance gap.