A q-Weibull Autoregressive Conditional Duration Model with an Application to NYSE and HSE Data

41 Pages Posted: 2 Nov 2011  

Tommi A. Vuorenmaa

Triangle Intelligence

Date Written: August 1, 2009

Abstract

This paper generalizes the ACD models of Engle and Russell (1998) using the so-called q-Weibull distribution as the conditional distribution. The new specification allows the hazard function to be non-monotonic. We document that the q-Weibull distribution recently suggested in physics as a generalization of the Weibull distribution is closely related to the much older Burr Type XII distribution in statistics. The nested, more heavy-tailed, q-exponential distribution is also being introduced. Data from the New York and the Helsinki Stock Exchange with different market microstructures and data types show that the q-Weibull specification outperforms the standard specifications used in econometrics and performs as well as the Burr specification of Grammig and Maurer (2000). The more parsimonious q-exponential specification typically provides a reasonable fit, improving the fit over the most commonly applied Weibull specification. We also find that the price threshold used affects the shape of the hazard function and thus the relative success of the models and must be taken into account when modeling price durations.

Keywords: ACD model, hazard function, q-Weibull distribution, transactions data

JEL Classification: C16, C22, C41

Suggested Citation

Vuorenmaa, Tommi A., A q-Weibull Autoregressive Conditional Duration Model with an Application to NYSE and HSE Data (August 1, 2009). Available at SSRN: https://ssrn.com/abstract=1952550 or http://dx.doi.org/10.2139/ssrn.1952550

Tommi A. Vuorenmaa (Contact Author)

Triangle Intelligence ( email )

Helsinki
Finland
+358-40-7757766 (Phone)

HOME PAGE: http://tommiavuorenmaa.net/

Paper statistics

Downloads
118
Rank
193,947
Abstract Views
584