@PPUNIT rejects, @LSUNIT (2 breaks) does not

[alexecon]

What are we to make of this situation? Thanks!

[TomDoan]

I'm not a fan of the Phillips-Perron test to start, but which options are you using with it? @LSUNIT, by design, allows for a trend, so it you're doing a non-trending model in @PPUNIT, the two might give conflicting results.

[alexecon]

I am using the DET=TREND option with the Phillips-Perron. Could I ask why you don't like this test? Many thanks.

[TomDoan]

The Phillips-Perron test does an inferior job (compared with an analogous ADF test) except for short-term serial correlation which is unlikely to be found in practice. You should probably use LSUNIT with 0 breaks (which is a Schmidt-Phillips test) as the base test. Note, however, that the LS unit test will have low power against the unit root if there really is no need for the breaks---you really should only do an LS test if you have accepted a UR without the breaks since the breaks tend to make it hard to reject a UR.