Placing Constraints on Constant Terms

Davis2010

New Member
I am running several iterations of a system of related equations. The first is four separate OLS equations estimating:

reg E_share lnp_cap lnp_lab lnp_energy lnp_mat
reg K_share lnp_cap lnp_lab lnp_energy lnp_mat
reg L_share lnp_cap lnp_lab lnp_energy lnp_mat
reg M_share lnp_cap lnp_lab lnp_energy lnp_mat

The second is a SUR regression of three of the four equations (the fourth is implicit):

sureg (eq1: E_share lnp_cap lnp_lab lnp_energy lnp_mat) (eq2: K_share lnp_cap lnp_lab lnp_energy lnp_mat) (eq3: L_share lnp_cap lnp_lab lnp_energy lnp_mat)

Finally, I would like to run a SUR regression as above, but with a constraint that the constants terms for all four equations sum to 1. I'm not sure what the syntax for this constraint is, as the variables I am imposing the constraint on are parameters and not actual variables in my data, and was wondering if someone has an idea of how to proceed. Thanks!

Etienne

New Member
Why don't you use the "constaint" command ? for instance :

constraint 1 : [eq1]_cons + [eq2]_cons + ... = 1
sureg ..., constaints(1)

This probably works...

Davis2010

New Member
Thanks for your help Etienne, that is very informative. One further question:
since the SUR only has 3 of our 4 equations (the fourth is redundant), how do I impose a constraint that the constants for all 4 equations sum to zero?

Etienne

New Member
Hi there,

Since your different equations contain exactly the same set of regressors on the right-hand-side, what you are doing is in fact equivalent to four separate OLS. However, I understand that you want to constrain the constants terms for all four equations to sum to 1. So my suggestion is :

constraint 1 : [eq1]_cons + ... + [eq4]_cons = 1
reg3 (eq1: y = ...) ... (eq4: y = ...) , constraints(1)

Hope this helps !