A semiparametric approach to nonlinear ACD model

Abstract

This paper introduces a new semiparametric approach to a nonlinear ACD model, namely the Semiparametric ACD model. This new model is relatively flexible in the sense that the data are allowed to speak for themselves, without assumptions being imposed arbitrarily on its key component. Moreover, it enables a more thorough examination of the intertemporal importance of the conditional duration on the ACD process. Experimental analysis suggests that the new model possesses sound asymptotic characters, while its performance is also robust across data generating processes. Furthermore, the empirical analysis illustrates clear advantages of our model over its parametric counterparts