A generalized least squares estimation method for the autoregressive conditional duration model

Abstract

A generalized least squares estimation method with inequality constraints for the autoregressive conditional duration model is proposed in this paper. The estimation procedure includes three stages. The final generalized least-squares estimator is consistent and \(\sqrt{T}\)—asymptotically normal distributed. Our estimator has the advantage over the often used quasi-maximum likelihood estimator in which it easily implemented and does not require the choice of initial values for the iterative optimization procedure. A large number of simulation studies confirm our theoretical results and suggest that the proposed estimator is more robust compared to quasi-maximum likelihood estimator. An application to IBM volume duration shows that the performance of the proposed estimation is better than quasi-maximum likelihood estimation in forecasting.