Autoregressive Conditional Duration: A New Model for Irregularly Spaced Transaction Data

Abstract

This paper proposes a new statistical model for the analysis of data that do not arrive in equal time intervals, such as financial transactions data, telephone calls, or sales data on commodities that are tracked electronically. In contrast to fixed interval analysis, the model treats the time between events as a stochastic time varying process. We propose a new model for point processes with intertemporal correlation. Because the model focuses on the time interval between events it is called the Autoregressive Conditional Duration (ACD) model. Strong evidence is provided for transaction clustering for the financial transactions dataanalyzed, even after time-of-day effects are removed. Although the model is most naturally applied to the arrival of transactions, we suggest a thinning algorithm to model characteristics associated with the arrival times, allowing the investigator to model processes that are observed in irregular time intervals, not just the arrival times of the data. Models for transaction events, the flow of volume, and the rate of change for prices are estimated.

JEL Classification: C2, C22, G1