I am hoping that someone can give me insight into the relationship between reg3 and ivregress with regards to overid.
I have a three equation system:
Y1= X1 (X1 = exogenous variables).
Y2 = Y1 + X2 (X2 = exogenous variables). Excludes more than one exog variable for identification. Therefore equation is overidentified
Y3 = Y1 + Y2 + X3 (X3 = exogenous variables) - excludes more than two exog variables for identification. i.e. is overidentified.
When I run ivregress (2sls) on Y3, using instruments to generate fitted values for Y1 and Y2, the Sargan test passes.
When I run ivregress on Y2, using instruments to generate fitted values for Y1, the Sargan test passes.
But when I run reg3 on all three equations simultaneously, Sargan fails.
Any help would be appreciated. Thanks!
I have a three equation system:
Y1= X1 (X1 = exogenous variables).
Y2 = Y1 + X2 (X2 = exogenous variables). Excludes more than one exog variable for identification. Therefore equation is overidentified
Y3 = Y1 + Y2 + X3 (X3 = exogenous variables) - excludes more than two exog variables for identification. i.e. is overidentified.
When I run ivregress (2sls) on Y3, using instruments to generate fitted values for Y1 and Y2, the Sargan test passes.
When I run ivregress on Y2, using instruments to generate fitted values for Y1, the Sargan test passes.
But when I run reg3 on all three equations simultaneously, Sargan fails.
Any help would be appreciated. Thanks!
