EViews.com

Hi,

I want to make a regime switching regression with Eviews 8. I wish to post a restriction such that coefficents of switching regressor and non-switching one are the same.
For example, x1 is non-switching regressor while c is switching one, y is dependent. I wish to regress
y=b*(x1-c)
b is the coefficient I want.
Could anyone help me? Thanks so much!

best regards
James

I think I'm confused. You say that you want the coefficient on the non-switching variable to be the same as the coefficient on the switching variable. I don't understand can this is possible if the switching variable has different coefficients in diferent regimes and the non-switching doesn't.

Dear Glenn,

Sorry for the confusion caused by me. I didn't express my problem clearly. I want to estimate
y=b1*(x1-c1)+b2*(x2-c2)
x1 and x2 are non-switching regressors, c1 and c2 are switching constants to be estimated. b1 and b2 are coefficients to be estimated.
Could you please help advise me how to conduct this estimation for b1, b2, c1 and c2?
Appreciate your help!

best regards
James

I'm still not sure I understand. If I expand your equation out as

y = b1*x1 + b2*x2 - b1*c1 - b2*c2

we have the first two parts are the non-switching regressors. and the second are products of switching and non-switching coefficients. First off, I don't see how the both of the products are identified since they enter the mean equation identically. Second, if we remove the redundancy and if I'm not missing something, can't you just estimated in the unrestricted form and then solve for the restricted value? As in

y = b1*x1 + b2*x2 - b1*c1 ---> y = b1*x1 + b2*x2 - c3

where c3 is switching and, then estimate the regime coefficients by dividing by b1. Note that EViews won't allow you to specify the first form since we only support specification by list.

Thanks Glenn, you are really helpful!