On autoregressive order selection criteria

This study investigates the performance of various commonly applied order selection criteria in selecting order of Autoregressive (AR) processes. The most important finding of this study is that Akaike’s information criterion, Schwarz information criterion, Hannan-Quinn criterion, final prediction error and Bayesian information criterion perform considerably well in estimating the true autoregressive order, even in small samples. Besides, there is no significant gain in differentiating these criteria unless one has a considerably large sample size. This study contributes to the empirical literature by providing helpful guidelines regarding the use of order selection criteria in determining the autoregressive order.