Tuesday, February 1st @ 12:30-2:00 PM (1011 Evans Hall)
Interpretable proximate factors for large dimensions
Markus Pelger, Stanford
This papers deals with the approximation of latent statistical factors with sparse and easy-to-interpret proximate factors. Latent factors in a large-dimensional factor model can be estimated by principal component analysis, but are usually hard to interpret. By shrinking the factor weights, we obtain proximate factors that are easier to interpret. We show that proximate factors consisting of 5-10% of the cross-sectional observations with the largest exposure are usually sufficient to almost perfectly replicate the population factors, even if these do not have a sparse structure. We derive an asymptotic lower bound for the correlation and generalized correlations of proximate factors with the population factors providing guidance on how to construct the proximate factors. Simulations and empirical applications to financial single- and double-sorted portfolios illustrate that proximate factors provide an excellent approximation to latent factors while being interpretable.