Estimating Latent Asset-Pricing Factors

Abstract

and the conference participants at the NBER-NSF Time-Series Conference, SoFiE, Western Mathematical Finance Conference and INFORMS for helpful comments. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications.

Get rich or die trying… finding revenue model fit using machine learning and multiple cases

“…Multi‐case theory‐building recognizes theoretical patterns by systematically analyzing a small number of cases. Since this method can reveal empirically grounded theoretical patterns, it can improve ML with theoretical guidance that both narrows the search space (and so upgrades model performance; Lettau & Pelger, ) and enhances interpretability with meaningful theoretical constructs and mechanisms. Machine learning is also a pattern recognition technique.…”

“…Second, ML complements multi‐case theory building (Table 1). Case studies (a) reveal constructs that narrow the search space (making ML more meaningful and efficient; Lettau & Pelger, ) and (b) provide theoretical mechanisms (complementing MLs atheoretical pattern recognition). In complement, ML adds (a) potential for large‐scale corroboration and elaboration of multi‐case insights and (b) precise identification of size effects, nonlinearities, equifinal paths, and configurations (all challenging for case studies).…”

Factors That Fit the Time Series and Cross-Section of Stock Returns

“…Thus "good" asset pricing factors need to have small cross-sectional pricing errors and explain the variation in the data. 4 In this large-dimensional context the limit is more complicated and studied in Lettau and Pelger (2018a). 5 The Arbitrage-Pricing Theory developed by Chamberlain and Rothschild (1983) assumes that only strong…”

“…In this paper we consider a simplified factor model with the purpose to make the weak and strong factor models comparable. The assumptions of the strong factor models can be relaxed considerably without affecting the results, as shown in Lettau and Pelger (2018a).…”

“…Third, RP-PCA identifies factors with high Sharpe-ratios that fit the cross-section of expected returns and not only co-movement in the time series dimension: This is important as low Sharpe-ratio factors that can only explain time-series comovements, for example industry factors, do not seem to be important for the cross-section of expected returns and hence do not capture the risk-return trade-off. 1 This paper focuses on the empirical estimation using RP-PCA while the econometric asymptotic theory for the estimator is developed in Lettau and Pelger (2018a). That paper shows that RP-PCA dominates conventional estimation based on PCA along a number of dimensions, especially when factors are "weak".…”