Testing order processing models of the bid-ask spread: a method of moments approach

Abstract

This paper tests all the implications of order processing models of the bid-ask spread jointly, in a GMM framework, using transactions data for the companies constituting the Dow Jones Index. The basic model, developed by Roll, does poorly because many of the observed serial covariances are positive, contrary to the prediction of the model. However, when the model is modified to allow for partial adjustment or overreaction of prices to new information, the resulting model fits the data much better. But, in either case, the average estimated spreads are lower than the average quoted spreads.