Title: Autoregressive conditional duration model with time-varying parameters for discrete trade durations
Authors: Petra Tomanova - University of Economics, Prague (Czech Republic) [presenting]
Abstract: Autoregressive conditional duration (ACD) models are useful for modelling the time between events, in particular the time between trading of stocks, which is known as trade durations. However, trade durations exhibit certain issues such as excessive zeros, discrete nature, overdispersion and intraday seasonality. We propose ACD model under the generalized autoregressive score framework allowing parameters to vary over time and capturing the dynamics of time-varying parameters by the autoregressive term and the scaled score of the conditional observation density. To deal with the discreteness of the data, we consider the trade durations as non-negative integer variables. This kind of variables is commonly analyzed using count data models based on specific underlying distribution, most notably the Poisson distribution. However, the feature that its expected value is equal to its variance is too strict in many applications as count data often exhibit overdispersion. To overcome this limitation we assume trade durations to follow the negative binomial distribution and we extend it to capture excessive zeros using the zero-inflated model. Simulation study and empirical analysis using high frequency data from NYSE TAQ database illustrate that the proposed model outperforms continuous ACD models.