Dynamic asymmetries in bid-ask responses to innovations in the trading process

This paper proposes a flexible structural model of quote formation to jointly study the dynamics between buyer-initiated and seller-initiated trades and the posterior bid and ask quote revisions. The empirical reduced form counterpart is a vector error correction (VEC) model for bid and ask quotes, with the spread as the cointegrating vector, that extends the bivariate vector autoregresive (VAR) model introduced by Hasbrouck (1991a). The empirical results for several NYSE common stocks reveal that there are informational gains by not averaging the quote revision process through the quote midpoint. We find evidence of asymmetric behavior in the responses of both ask and bid prices to the innovations in the trading process. Under similar market conditions, the bid dynamics after an unexpected seller-initiated trade show significant deviations relative to the ask dynamics after a similar unexpected buyer-initiated trade