Abstract
Using Monte Carlo simulation, the size of the DF–GLS test is examined when there is a neglected level or trend break under the null hypothesis. Unlike the original DF test, the DF–GLS test has size which is not distorted/inflated.
Notes
1Note that this de-trending method goes back to Granger and Newbold (Citation1974) who propose regression in first differences.
2As it is easily observed the two de-trending methods are asymptotically equivalent.