I am using Eviews 12 and programming a model that spans from 1971 to 2018 (yearly). My problem is the following:
While the vast majority of the explanatory variables have available data from 1971-2018, some dependent variables depend on explanatory variables only available from 1998 onward. So, the non-problematic equations with explanatory variables spanning 71-18 would be like this:
Code: Select all
Y1= B0+B1X1+B2X2+B3X3 +...
And the few (but important) problematic equations look like this:
Code: Select all
Y1=(B0*+B1*X1*+B2*X2*+B3*X3*+...)
with X3 spanning from 1998-2018.
We previously tried to solve the issue by the following method, suggested in this forum:
First, we estimated two different equations, one from 71-97 and another 98-18. Then we write this in the model window:
Code: Select all
Y1= (@year<1998) *(B0+B1X1+B2X2+B3X3+...) + (@year>=1998)* (B0*+B1*X1*+B2*X2*+B3*X3*+...)
It worked as long as the dependent variable for both equations had the same form, say: Y1 for both equations are in logs, or in first difference, or levels... Otherwise, it became problematic, i.e. the model crashes. So, as a simple solution, valid for running the model for out-of-sample forecasting purposes, we kept the 98-18 equation and excluded in scenarios that variable from 71-97. However, this is not good enough for counterfactual analysis.
Therefore, I was looking into 'if then else' statements in Eviews, but I am not being able to make it work. So, this is what I am writing:
Code: Select all
if @year>=1998 then
Y1=B0*+B1*X1* + B2*X2* + B3*X3* +...
else
Y1=B0+B1X1 + B2X2 + B3X3 +...
endif
It does not work either when I write the equations as :eq1 and :eq2 instead of the "substituted coefficients" form.
I hope this all makes sense. Could I get help to solve this issue? Many thanks in advance.
Kind regards