NLSYSTEM Mask option

[onem]

Dear Tom,

I have nine moment conditions and two instruments. I want to have the Kroneker product of conditions x instruments for the first six conditions only, and the last three conditions will not use the instruments. I am inclined to do the following:

Code: Select all

instruments constant su1
dec rect mask(2,9)
compute %do(j,1,9,mask(1,j)=(j<=6))
compute %do(j,1,9,mask(2,j)=(j<=6))
nlsystem(mask=mask,instruments,parmset=baseparms,trace) / M11 M12 M13 M21 M22 M23 M31 M32 M33

However, it seems like when I do this, the last three conditions are ignored in the optimization. Is there a way to do this?

Thank you,

Onem

[TomDoan]

If you're not using (any) instruments, how are you estimating them?

[onem]

Hi Tom,

Thank you for your response. I am trying to estimate the model of Evans and Lyons (2008), "How is macro news transmitted to exchange rates" , Journal of Financial Economics, 88 26-50. In the model, there are two instruments, three moment conditions, and - to my understanding- only two of the moment conditions are Kroneker products of the instruments (there are five parameters to estimate and they mention that with this specification the model is exactly identified). The moment conditions are specified on page 44 of the attached paper (equations 30, 31 and 32).

I would greatly appreciate your feedback on this.

Thank you,

Onem

[TomDoan]

(30) uses x(t) as the "instrument". (Least squares is equivalent to instrumental variables with instruments equal to the explanatory variables).

[onem]

Dear Tom,

Thank you very much! Everything makes sense to me now.

Best regards,

Onem