Thin-Trading and Beta Estimation: Results From a Simulated Environment

Abstract

This study contributes to the literature on common stock Beta estimation in an environment of varying levels of infrequent trading, or market frictions. In this study a stock market environment with a range of Beta relationships and varying levels of trading infrequency is artificially created using Normal and Bernoulli-distributed Monte Carlo simulations. Four different modified Beta estimation techniques are compared to the OLS-Beta. The rate of deterioration of the various models is investigated with respect to systematic bias (bias of the mean) as well as a measurement bias (bias of the estimate's standard error). The Trade-to-Trade model performed best overall followed by the simple OLS-adjustment with respect to systematic bias; however, both models showed increasing levels of measurement bias from an early stage. Due to the data-intensive nature of the Trade-to-Trade model it does not serve as a one-size-fits-all solution. The strongest alternative is shown to be the adjusted OLS estimate. The choice of an appropriate thin-trading filter is a function of the choice of the Beta estimator which is in-turn determined by the researchers’ tolerance for either systematic or measurement bias. The contribution of this study is that it provides guidance to the use of alternative methods to estimate risk coefficients.