A Hidden Markov Process Approach to Information-Based Trading

Abstract

This paper proposes a novel approach to information-based trading, incorporating both asymmetric information and symmetric order-flow shocks. It focuses on the dynamics of securities trading and postulates that trading activities are determined by the state of nature. A two-dimensional Markov chain is used to model the hidden information states of the market and the state set is allowed to vary across time and companies. Distinguished from the prevailing approaches to information-based trading, the Hidden Markov Model (HMM) approach updates the prior belief of information states using newly observed order flows and identifies trading motives in a data-driven manner. Each trading day is associated with dynamic measures of probability of information based trading (PIN) and probability of symmetric order-flow shock (PSOS). In addition, it allows multiple occurrences of information events rather than limit the frequency of information arrival to once a day. The HMM approach does not rely on a particular market structure and it can be applied to various markets. To evaluate the HMM approach, we conduct extensive Monte Carlo simulation experiments. It shows superior performance in dynamic daily PIN and PSOS estimates as well as cumulative estimates over any time interval in all simulation scenarios, evidenced by their negligible errors compared with true values and higher accuracy than the estimates obtained from alternative approaches. Using a sample of 30 NYSE stocks, we examine the properties of the dynamic PIN estimate as a proxy for information asymmetry and the dynamic PSOS estimate as a measure of illiquidity. We also show how they can be used to explain and predict realized volatility of returns. Dynamic PSOS is a significant contributor to the realized volatility for all the stocks in the sample and its effect is more profound than dynamic PIN, reflecting dominant effect of investors' dispersed beliefs on price fluctuations.