Threshold Stochastic Conditional Duration Model for Transaction Data

This paper proposes a threshold stochastic conditional duration (SCD) model for financial data at the transaction level. In addition to assuming that the innovations of the duration process follow a threshold distribution with positive support, we also assume that the latent first-order autoregressive process of the log conditional durations switches between two regimes. The regimes are determined by the levels of the observed durations and the threshold SCD model is specified to be self-excited. Markov Chain Monte Carlo methods within a Bayesian framework are then developed for parameter estimation. For model comparison, we employ a deviance information criteria, which does not depend on the number of model parameters directly. Duration forecasting is constructed by using an auxiliary particle filter based on the fitted models. Simulation studies demonstrate that our proposed model and estimation approach work well in terms of parameter estimation and duration forecasting. Lastly the proposed models and estimation approach are applied to two benchmark data sets that have been studied in the literature, namely IBM and Boeing transaction data