Deep Learning in Characteristics-Sorted Factor Models

TL;DR

This paper introduces a deep learning-based augmented factor model that generates latent factors for asset pricing, improving the modeling of cross-sectional returns using nonlinear characteristic relationships.

Contribution

It proposes a structural deep learning framework that captures nonlinearities in characteristic-based asset pricing, enhancing predictive accuracy over traditional linear models.

Findings

Robust asset pricing performance demonstrated on high-dimensional data

Identification of key raw characteristic sources improves investment strategies

Model reduces pricing errors compared to conventional methods

Abstract

This paper presents an augmented deep factor model that generates latent factors for cross-sectional asset pricing. The conventional security sorting on firm characteristics for constructing long-short factor portfolio weights is nonlinear modeling, while factors are treated as inputs in linear models. We provide a structural deep learning framework to generalize the complete mechanism for fitting cross-sectional returns by firm characteristics through generating risk factors -- hidden layers. Our model has an economic-guided objective function that minimizes aggregated realized pricing errors. Empirical results on high-dimensional characteristics demonstrate robust asset pricing performance and strong investment improvements by identifying important raw characteristic sources.