Nonstationary autoregressive conditional duration models

Recently, there has been a growing interest in studying the autoregressive conditional duration (ACD) models, originally introduced by (Engle, R. F., and J. R. Russell. 1998. “Autoregressive Conditional Duration: A New Model for Irregularly Spaced Transaction Data. Econometrica 66: 1127–1162). ACD models are useful for modeling the time between the events, especially, in financial context, the time between trading of stocks. In this paper, we propose a specific type of nonstationary ACD model, viz., time varying ACD model (tvACD), by allowing the parameters of the usual ACD model to vary as functions of time. Some probabilistic and inferential aspects of such models have been investigated. We also develop a local polynomial procedure for the estimation of the parameter functions of the proposed tvACD model. Asymptotic properties of the estimators have been investigated, including the asymptotic normality. The asymptotic distribution being dependent on the parameters of the original distribution, a weighted bootstrap estimator is suggested and its validity is established. Simulation study and empirical analysis using high frequency data (HFD) from National Stock Exchange (NSE, INDIA) illustrate the application of the proposed tvACD model.