Non-standard Errors in Portfolio Sorts

Description

We systematically study the variation in returns induced by varying 14 methodological
decisions in portfolio sorts. These non-standard errors range between 0.14
and 0.39 percent per month and are larger than standard errors. However, for most
sorting variables, mean return differentials and alphas are pervasively positive, statistically
significant, and increase monotonically. Decisions such as excluding firms
with negative earnings or the information time lag have an impact comparable to
size-related ones. Non-standard errors are countercyclical, raising concerns about
non-classical measurement error in predictive regressions. Using our publicly available
code to report distributions of estimated premia provides an easy remedy.

Period	19 Aug 2023
Event title	50th Annual Meeting of the European Finance Association (EFA)
Event type	Conference
Location	AmsterdamShow on map