Testing for unit roots using the augmented Dickey-Fuller test: Some issues relating to the size, power and the lag structure of the test

Abstract The main purpose is to consider some of the operational aspects of the ADF test, and especially the question of how many autoregressive lags are needed to capture the underlying d.g.p. when the latter may be a mixed ARIMA process. Thus the results from Monte-Carlo experiments (using various stochastic processes) are reported to determine the lag structure and whether the ADF test consistently rejects the null and alternative when it is true.