Abstract

When assets do not trade as frequently as the market index, the standard ordinary least squares (OLS) beta exhibits thin trading bias. Several beta adjustment techniques exist to correct for this bias; however, no consensus exists as to which adjustment is best. This article compares the behavior of the most widely used beta adjustments proposed in the literature across emerging markets. Using a linear programming model, we form portfolios with equal risk characteristics, but different levels of censoring. Since beta is a measure of systematic risk, if most risk characteristics are kept constant across portfolios, the resulting betas should be approximately the same. Our results show that the best adjustments overall are the Scholes-Williams, trade-to-trade, and sample selectivity adjustments.