Measuring Systematic Risk with Neural Network Factor Model

TL;DR

This paper introduces a neural network-based nonparametric factor model to measure systematic risk, capable of automatically identifying relevant factors from asset returns without prior feature engineering.

Contribution

The paper presents a novel neural network factor model that automatically discovers systematic risk factors, offering an alternative to traditional parametric models without requiring feature engineering.

Findings

Model performs comparably to existing models on 20-year S&P 100 data.

It automatically identifies relevant features without prior knowledge.

It does not outperform parametric models due to inference limitations.

Abstract

In this paper, we measure systematic risk with a new nonparametric factor model, the neural network factor model. The suitable factors for systematic risk can be naturally found by inserting daily returns on a wide range of assets into the bottleneck network. The network-based model does not stick to a probabilistic structure unlike parametric factor models, and it does not need feature engineering because it selects notable features by itself. In addition, we compare performance between our model and the existing models using 20-year data of S&P 100 components. Although the new model can not outperform the best ones among the parametric factor models due to limitations of the variational inference, the estimation method used for this study, it is still noteworthy in that it achieves the performance as best the comparable models could without any prior knowledge.