Structural breaks in point processes: with an application to reporting delays for trades on the New York stock exchange

Abstract

In this paper some methods to determine the reporting delays for trades on the New York stock exchange are proposed and compared. The most successful method is based on a simple model of the quote revision process and a bootstrap procedure. In contrast to previous methods it accounts for autocorrelation and for variation originating both from the quote process itself and from estimation errors. This is obtained by the use of prediction intervals. The ability of the methods to determine when a trade has occurred is studied and compared with a previous method by Vergote (2005). This is done by means of a simulation study. An extensive empirical study shows the applicability of the method and that more reasonable results are obtained when accounting for autocorrelation and estimation uncertainty.