EViews.com

Hi,

I would like to test whether the parameters a2 and b2 are equal in the following separate linear OLS regressions:
Y1=a1+a2X1+a3X2+a4X3+...
Y1=b1+b2X4+b3X2+b4X3+...
The models have the same dependent and control variables; only the partitioning variables X1 and X4 differ and I have to test whether X4 is significantly stronger related to Y1 compared to X1.
However, X1 and X4 exhibit a very high correlation (r > 0.85), so I would like to estimate them in separate models to avoid multicollinearity, instead of using a dummy approach (I am also required to run two separate regressions so that my results are comparable with previous research).
I know that the equality of coefficients within a single equation can be tested using the Wald test in Eviews 7, but is there a way do this for separate equations?

Thank you in advance.

You might set up up a system and run seemingly unrelated regressions.

Thanks for your suggestion. SUR would be helpful, but I need standard errors that are robust to heteroskedasticity and autocorrelation, and ideally adjusted for clustering at firm level. From what I've read on this forum, Eviews 7 does not currently support this..is there a way to calculate a Wald test for the difference between two coefficients of two separately estimated equations?

There's nothing built-in. I'm sure it could be done with matrix operations, but it might not be easy.

Well, if an Eviews expert is saying something is not easy, I'm not even going to bother trying Thanks for your swift replies.