This website uses cookies.

By clicking the "Accept" button or continuing to browse our site, you agree to first-party and session-only cookies being stored on your device to enhance site navigation and analyze site performance and traffic. For more information on our use of cookies, please see our Privacy Policy.

  1. Home
  2. ASSA Annual Meeting
  3. 2023
  4. Program
  5. Factor Models, Machine Learning, and Asset Pricing
« Back to Results

Factor Models, Machine Learning, and Asset Pricing

Paper Session

Saturday, Jan. 7, 2023 2:30 PM - 4:30 PM (CST)

Sheraton New Orleans, Borgne
Hosted By: American Finance Association
  • Chair: Dacheng Xiu, University of Chicago

Structural Deep Learning in Conditional Asset Pricing

Jianqing Fan
,
Princeton University
Tracy Ke
,
Harvard University
Yuan Liao
,
Rutgers University
Andreas Neuhierl
,
Washington University-St. Louis

Abstract

We develop new structural nonparametric methods for estimating conditional asset pricing models using deep neural networks, by employing time-varying conditional information on alphas and betas carried by firm specific characteristics. Contrary to many applications of neural networks in economics, we can open the \black box" of machine learning predictions by incorporating financial economics theory into the learning, and provide an economic interpretation of the successful predictions obtained from neural networks, by decomposing the neural predictors as risk-related and mispricing components. Our estimation method starts with period-by-period cross-sectional deep learning, followed by local PCAs to capture time-varying features such as latent factors of the model. We formally establish the asymptotic theory of the structural deep-learning estimators, which apply to both in-sample t and out-of-sample
predictions. We also illustrate the \double-descent-risk" phenomena associated with over-parametrized predictions, which justies the use of overfitting machine learning methods.

Textual Analysis of Short-Seller Research Reports, Stock Prices, and Real Investment

Jules van Binsbergen
,
University of Pennsylvania
Xiao Han
,
City University London
Alejandro Lopez-Lira
,
University of Florida

Abstract

We construct a comprehensive database of short-sell research reports and investigate to what extent these reports impact real economic activity. We document that firms mentioned in the reports significantly reduce their real investment and stock issuances. On average, each report is associated with a reduction of corporate investment equal to \$118 million and stock issuances equivalent to \$179 million. We attribute this effect to a substantial increase in the cost of capital since we find that, on average, target firms earn abnormal returns of -4\% on the publication day, and the subsequent price revisions equal -20\% and may take up to 6 months to materialize. Furthermore, the cost of capital implied from the dividend-discount model increases substantially for target firms. Finally, by applying textual analysis, we show that a large percentage of the text relates to the revelation of accounting fraud or mismanagement. We compare the results against firms that commit accounting frauds but have not been identified in short-seller research reports. Consistent with our hypothesis, we find that only the firms whose accounting frauds are mentioned by short-seller research experience a downward change in real activities.

High Dimensional Factor Models with an Application to Mutual Fund Characteristics

Martin Lettau
,
University of California-Berkeley

Abstract

This paper considers extensions of 2-dimensional factor models to higher-dimension data that can be represented as tensors. I describe decompositions of tensors that generalize the standard matrix singular value decomposition and principal component analysis to higher dimensions. I estimate the model using a 3-dimensional data set consisting of 25 characteristics of 1,342 mutual funds observed over 34 quarters. The tensor factor models reduce the data dimensionality by 97% while capturing 93% of the variation of the data. I relate higher-dimensional tensor models to standard 2-dimensional models and show that the components of the model have clear economic interpretations.

Semiparametric Conditional Factor Models: Estimation and Inference

Qihui Chen
,
Chinese University of Hong Kong-Shenzhen
Nikolai Roussanov
,
University of Pennsylvania
Xiaoliang Wang
,
University of Pennsylvania

Abstract

This paper introduces a simple and tractable sieve estimation of semiparametric conditional factor models with latent factors. We establish large-N-asymptotic
properties of the estimators and the tests without requiring large T. We also develop a simple bootstrap procedure for conducting inference about the conditional
pricing errors as well as the shapes of the factor loadings functions. These results
enable us to estimate conditional factor structure of a large set of individual assets
by utilizing arbitrary nonlinear functions of a number of characteristics without the
need to pre-specify the factors, while allowing us to disentangle the characteristics’
role in capturing factor betas from alphas (i.e., undiversifiable risk from mispricing). We apply these methods to the cross-section of individual U.S. stock returns
and find strong evidence of large nonzero pricing errors that combine to produce
arbitrage portfolios with Sharpe ratios above 3.

Discussant(s)
Seth Pruitt
,
Arizona State University
Asaf Manela
,
Washington University-St. Louis
Alberto G. Rossi
,
Georgetown University
Robert Korajczyk
,
Northwestern University
JEL Classifications
  • G1 - Asset Markets and Pricing