Office of Research Transaction-Data Analysis of Marked Durations and Their Implications for Market Microstructure

We propose an Autoregressive Conditional Marked Duration (ACMD) model for the analysis of irregularly spaced transaction data. Based on the Autoregressive Conditional Duration (ACD) model, the ACMD model assigns marks to characterize events such as tick movements and trade directions (buy/sell). Applying the ACMD model to tick movements, we study the influence of trade frequency, direction and size on price dynamics, volatility and the permanent and transitory price impacts of trade. We also apply the ACMD model to analyze trade-direction data and estimate the probability of informed trading (PIN). We find that trade frequency has a critical role in price dynamics while the contribution of volume to price impacts, volatility, and the probability of informed trading is marginal. ∗Mitch Warachka gratefully acknowledges funding by the Wharton-SMU Research Center, Singapore Management University.