Page 1 of 1

bai perron structural break tests

Posted: Thu Aug 24, 2017 8:33 am
by justin_m
Suppose I am regressing one variable on 5 other variables. I expect breaks to occur in the slopes of the 5 variables but at different dates - and I am interested in the dates at which the breaks occur.

The Bai-Perron sequental L+1vsL test can do this by including all 5 regressors. However, I can no longer identify the breaks specific to each variable. A possibility is to, for each variable, test for equality of coefficients before and after each break. If we reject the null of equality, I find evidence for a break.

I could run the regression on 1 variable at a time, but I have sound motives to want to control for all the other variables.

Is there a better way to deal with this?

Thanks