Hi! Is it valid to use the Bai-Perron test to detect a structural break in the constant in a regression with non-stationary variables, and then use the break dates as step dummies in an Engle-Granger co-integration test?
Yes, you use normally that test to detect break points in the constant. In eviews is the first option. After, you could use the break point with a dummy with zero in previous value and one in the break point and after. You could use Johansen-Mosconi and Nielsen test.
Zivot-Andrews Unit Root Test is also there to detect structural break automatically. In Eviews8 this can be done. then you can do Bayer and Hanck (2013) cointegration test with structural break....
Bai-Perron is an algorithm for efficiently finding the least squares break points in a linear regression model of the form yt=Xtβ+ut y t = X t β + u t .