Time Varying Three Way Decompositions of Posted Spreads

Abstract

In this study, Huang and Stoll's (1997) three way decomposition model for posted spreads is estimated using a data set with 10 stocks traded at Stockholmsborsen (SB), randomly selected from in the OMX stock index. The Huang and Stoll (1997) model encompasses statistical and indicator models by for example Ross (1984), Stoll (1989), George et al (1991), Glosten and Harris (1988), decomposing posted spreads into order processing, adverse selection and inventory cost components. In this study fixed probabilities of reversed trades ranges from 12.38 percent to 21.25 percent. This means that trade continuations are highly probable. Even so, this study supports the Huang and Stoll (1997) model. The signs of estimated adverse selection cost component coefficients are positive for all stocks ranging from 0.31 percent to 10.75 percent. No bunching procedure is used, which probably would have yielded higher adverse selection coefficients. Also, this study finds time of day, or diurnal, patterns in the probability of reversed trades. However, this pattern does not affect the coefficients for the adverse selection or inventory cost components. But diurnal probabilities of reversed trades do have an important implication for the size of expected changes in the mid quote. Diurnal patterns in the probability of reversed trades support earlier notions of larger spreads and higher transaction volumes at the opening and supports earlier findings of informed trading occurring during the first 15 minutes after the opening.