Predictive distributions and the market return: The role of market illiquidity

Abstract

This paper evaluates the role of volatility-free stock market illiquidity proxies in forecasting monthly stock market returns. We adopt a probabilistic approach to multivariate time-series modelling using Bayesian nonparametric vector autoregressions. These models flexibly capture complex joint dynamics among financial variables through data-driven regime switching. Out-of-sample forecasts maintain accuracy as the horizon increases. Adding illiquidity generates statistical improvements in out-of-sample predictive accuracy. We highlight the operational importance of market illiquidity after selecting the most appropriate forecasting model that delivers profitable strategies that outperform a range of multivariate models; as well as the historical mean.