Alejandro Lopez-Lira and Nikolai Roussanov
First version: May 2020; This version: February 2023
I skimmed this paper (71 pages), but my last course in Statistics was roughly 40 years ago. I'm not sure if the results are meaningful, or that the methods are practical for the average investor not managing fund of some sort.Abstract
We document challenges to the notion of a trade-off between systematic risk and expected returns when analyzing stock characteristics’ empirical ability to predict excess returns. First, we measure individual stocks’ dynamic exposures to all common latent factors using efficient high-dimensional methods. These factors explain virtually all of the common time-series variation in stock returns. However, exposure to these latent factors appears to earn negligible risk premia. Next, we construct out-of-sample forecasts of stock returns based on a wide range of characteristics using machine learning methods and linear models. A zero-cost beta-neutral portfolio that exploits this predictability but hedges all undiversifiable risk delivers a Sharpe ratio above one with no correlation with any systematic factor, thus rejecting the central prediction of the arbitrage pricing theory.
Can anyone with better statistical chops tell me if this is actionable for mere mortals?