Comparing coefficients of two models

[cucu]

Hi, all

Not particularly savvy when it comes to econometrics, so I would appreciate your help with a problem related to comparing coefficients across models. I have a model like Y = c(0) + c(1)*X1 + c(2)*X2... I want to see the different behaviour of the model in two different periods.

I guess I have the option to use a time dummy for period 2 and run a regression like

Y = c(0) + c(1)*X1 + c(2)*X2 + c(3)*dummy + c(4)*X1*dummy + c(5)*X2*dummy

where, I assume, c(3) shows the effect of the second period on Y and c(4) and c(5) show the additional contribution of X1 and X2 during period 2 compared to period 1, so whether they become more or less important (please confirm this to me).

However, I was thinking of running the same model (Y = c(0) + c(1)*X1 + c(2)*X2) in each period, so I would get different coefficients for X1 and X2. However, how could I test the difference between coefficients (for example how do I know that the difference between C(1) in the first period and C(1) in the second period is statistically significant). Can I do this or am I forced to use the period dummy method?

Thank you!

[startz]

You can do it either way and will get the same answer. Look up "Chow test."

[cucu]

Thank you for your response.

What I can understand from the users guide is that the Chow test shows whether there's a structural change occurring between the two periods, which is good. The way I interpret this is that it tells me a change in how the IV (taken together) explain the DV. I guess it doesn't tell me which of the IV suffer from this structural break and to what extent. Or does it?

I'm asking because I find out that I can't do the test on panel data. Guess I'm forced, in the end, to go the dummy route. I would've preferred not too for some (subjective) reason. Any suggestions?

Thanks again

[startz]

The dummy variable test and the Chow test give numerically identical results. So if it's easier to use dummies, go right ahead.

If you like, you can use dummies for some variables but not others. That's a way of modeling the idea that some coefficients change while others don't.