Forecasting stock return distributions around the globe with quantile neural networks

TL;DR

This paper introduces a novel machine learning approach using quantile neural networks with spline interpolation to accurately forecast the entire distribution of stock returns, capturing complex features like fat tails and asymmetries across global markets.

Contribution

It develops a two-stage quantile neural network method that constructs smooth, flexible return distributions without restrictive assumptions, improving forecast accuracy of distributional statistics.

Findings

Improved out-of-sample performance for mean and variance forecasts

Robustness demonstrated across US and international markets

Effective modeling of non-Gaussian features like fat tails and asymmetries

Abstract

We propose a novel machine learning approach for forecasting the distribution of stock returns using a rich set of firm-level and market predictors. Our method combines a two-stage quantile neural network with spline interpolation to construct smooth, flexible cumulative distribution functions without relying on restrictive parametric assumptions. This allows accurate modelling of non-Gaussian features such as fat tails and asymmetries. Furthermore, we show how to derive other statistics from the forecasted return distribution such as mean, variance, skewness, and kurtosis. The derived mean and variance forecasts offer significantly improved out-of-sample performance compared to standard models. We demonstrate the robustness of the method in US and international markets.