Estimation of the Bid–Ask Spread and Its Components: A New Approach

TLDR

The study introduces a new approach to estimate the components of the bid‑ask spread unbiasedly and efficiently. The approach employs unbiased and efficient estimators to separate the spread into its constituent components. The study finds that time‑varying expected returns cause most of the downward bias in spread estimates, that adverse‑selection accounts for only 8–13 % of the quoted spread for small trades, and that order‑processing costs dominate the spread.

Abstract

We show that time variation in expected returns and/or partial price adjustments lead to a downward bias in previous estimators of both the spread and its components. We introduce a new approach that provides unbiased and efficient estimators of the components of the spread. We find that between 77 and 97 percent of the downward bias in previous spread estimates is caused by time variation in expected returns. More importantly, the adverse-selection component, though significant, accounts for a much smaller proportion (8 to 13 percent) of the quoted spread, at least for small trades, than the proportion (over 40 percent) previously reported in the literature. Order processing costs are the predominant component of quoted spreads.