Page 1 of 1
Testing for increased impact?
Posted: Mon Sep 13, 2010 11:29 pm
by jjnull
Hi all,
I'm working on my econometrics project and would like to ask if it is possible to test for a "stronger impact" of an explanatory variable on the dependent variable during different time periods.
For example, before the financial crisis, firm size has a positive impact on stock price. Is it possible to test if during the financial crisis, firm size has a stronger, positive impact on stock prices than before the crisis? Btw I'm using panel data which stretches from 2004 to 2010 Q2. Regression method is OLS.
I'm new to econometrics, would appreciate if anyone can help me with this. Thanks! :)
Re: Testing for increased impact?
Posted: Tue Sep 14, 2010 6:56 am
by startz
Define a dummy variable, DF, that equals one during the financial crisis and zero otherwise. Then something like
ls y c x DF DF*x
and test is the last coefficient is significantly greater than zero.
Re: Testing for increased impact?
Posted: Tue Sep 14, 2010 10:01 am
by jjnull
Hey Startz,
thanks!
Hmm if I do that then I can't do a time period fixed effect LSDV anymore right?? Is there anyway to do the dummy thing plus a time period fixed effects regression?
Define a dummy variable, DF, that equals one during the financial crisis and zero otherwise. Then something like
ls y c x DF DF*x
and test is the last coefficient is significantly greater than zero.
Re: Testing for increased impact?
Posted: Tue Sep 14, 2010 10:15 am
by startz
Hey Startz,
thanks!
Hmm if I do that then I can't do a time period fixed effect LSDV anymore right?? Is there anyway to do the dummy thing plus a time period fixed effects regression?
Define a dummy variable, DF, that equals one during the financial crisis and zero otherwise. Then something like
ls y c x DF DF*x
and test is the last coefficient is significantly greater than zero.
Actually, you can take out the DF term, leave in DF*x, and keep in the fixed effects. (I think)
Re: Testing for increased impact?
Posted: Wed Sep 15, 2010 9:17 pm
by jjnull
Hey Startz,
thanks!
Hmm if I do that then I can't do a time period fixed effect LSDV anymore right?? Is there anyway to do the dummy thing plus a time period fixed effects regression?
Define a dummy variable, DF, that equals one during the financial crisis and zero otherwise. Then something like
ls y c x DF DF*x
and test is the last coefficient is significantly greater than zero.
Actually, you can take out the DF term, leave in DF*x, and keep in the fixed effects. (I think)
Hmmm makes sense, since in a fixed period panel, all periods have their own intercept.
But sadly the results doesn't tally with the regressions that i ran seperately.
Anywyay, thanks startz! :)