slope dummy?

[Scriptie_Caspar]

Dear all,

With Eviews 5, I am trying to perform a regression on a model which analyses the effect of (among others) the R&D intensity (R&D/Sales) on firmlevel employment. I'd like to know whether or not the industry (sector) affects this. So I created a couple of dummy variables like 'sector=1', 'sector=2', etc. which has the value of 1 if the firm is in resp. sector one or two and the value of zero otherwise. These dummies are together in one group (called 'g').
Does anyone know how I can check in Eviews if the sector to which a firm belongs also affects the effect of R&D on Employment?
As the title suggests, I suppose you can do this with a slope dummy, but I don't know how to implement this into the regression...

I would really appreciate it!

Kind regards,
Caspar

[startz]

Something like

ls E D1 D1*RD D2 D2*RD

then check if c(2)=c(4)

[Scriptie_Caspar]

Thanks for your reply!

So, now when I use the slope dummies I have another problem/question.

I entered the estimation: 'empl sales rd sector0 rd*sector0 c'

now I do get some plausible results, however somebody told me there is a chance for multicollinearity now.
Do you perhaps know how I can test for that in EV (and solve it?)?
I have very little experience with EViews, so could you please explain it slowly :).

Thanks a lot!

Regards,

[Scriptie_Caspar]

oh, and the reason why I ask this is because I don't understand why the p-value is different if you estimate 'empl sales rd*sector0' so just without the rd and sector0 separately...

[startz]

Thanks for your reply!

So, now when I use the slope dummies I have another problem/question.

I entered the estimation: 'empl sales rd sector0 rd*sector0 c'

now I do get some plausible results, however somebody told me there is a chance for multicollinearity now.
Do you perhaps know how I can test for that in EV (and solve it?)?
I have very little experience with EViews, so could you please explain it slowly :).

Thanks a lot!

Regards,

EViews 7 does have some measures of multicollinearity. Having said that, all that's going on is that when you require the coefficients to be the same for each sector you get stronger results than if you don't. Effectively, there is more data for each coefficient. If you find that the coefficients are different you should probably stick with that model. If not, returning to the "all the same" model is okay.