The Effectiveness of Information Criteria in Determining Unit Root and Trend Status

This paper compares the performance of using an information criterion, such as the Akaike information criterion or the Schwarz (Bayesian) information criterion, rather than hypothesis testing in consideration of the presence of a unit root for a variable and, if unknown, the presence of a trend in that variable. The investigation is performed through Monte Carlo simulations. Properties considered are frequency of choosing the unit root status correctly, predictive performance, and frequency of choosing an appropriate subsequent action on the examined variable (first differencing, detrending, or doing nothing). Relative performance is considered in a minimax regret framework. The results indicate that use of an information criterion for determining unit root status and (if necessary) trend status of a variable can be competitive to alternative hypothesis testing strategies.