Non-standard errors in asset pricing: Mind your sorts

Abstract

Non-standard errors capture variation due to differences in research design choices. We document large variation in design choices in the context of asset pricing factor models and find that the average ratio of the non-standard error to the standard error across factors exceeds one. Using NAN breakpoints instead of NYSE breakpoints improves the average Sharpe ratios the most, from 0.46 to 0.63. Other important design choices relate to excluding microcaps, industry-adjusting, and the rebalancing frequency, which highlights the need for researchers to clearly describe and motivate these choices.