Page 1 of 1
Structural breaks with panel data
Posted: Mon Jul 29, 2019 6:54 am
by Estefaniame
Dear all
I would like to know if there is any procedure to test for structural breaks in case of having panel data. Thank you very much.
Re: Structural breaks with panel data
Posted: Mon Jul 29, 2019 10:13 am
by TomDoan
Structural breaks in what? Breaks are in models, not in data.
Re: Structural breaks with panel data
Posted: Mon Jul 29, 2019 1:38 pm
by Estefaniame
Yes, in the model. The behaviour of the involved variables do not seem to be the same before and after some moments in time.
Thank you beforehand.
Re: Structural breaks with panel data
Posted: Mon Jul 29, 2019 3:27 pm
by TomDoan
In what model? There are very different techniques depending upon what the original model is, where the breaks are located (are they assumed to be at the same time period in each individual, or can they vary), etc.
Re: Structural breaks with panel data
Posted: Tue Jul 30, 2019 4:33 am
by Estefaniame
In a fixed-effects model. A very simple one with panel data: one variable depending on other two. We suspect there are changes in certains moments in time because they make sense and we would like to REFLECT this fact in the model. We do not date or assume them before, we want the model to tell us. Is there any possibility to do it?
Thank you.
Re: Structural breaks with panel data
Posted: Tue Jul 30, 2019 8:56 am
by TomDoan
The usual way of handling that with panel data is to include time effects as well, which will allow for common time shifts of any form (trends, breaks, etc.). Note that those are for level shifts. If you are talking about something else, you need to be much more specific about what your model is when you allow for a break.