Intermediate Lags and the Power of Unit Root Tests

Walter Enders, University of Alabama
Yu Liu, University of Texas at El Paso

Abstract

We propose a simple modification to the general-to-specific lag-length selection method typically employed in a standard augmented Dickey-Fuller (ADF) test. Instead of using the entire set of lags selected by the general-to-specific method, we suggest using only the lags that are statistically significant. Our Monte Carlo experiments show that the modified test has superior power to the standard ADF test with only minor size distortions. The power of Dickey-Fuller generalized least squares (DF-GLS) test can also be improved by discarding insignificant intermediate lags. The appropriate use of the modified test is illustrated by using the inflation rates of OECD countries.