0
$\begingroup$

I am trying to run the following 3SLS regression in Stata, but I keep getting the error: "Equation is not identified: Does not meet first order conditions."

reg3 
(Y1 Y2       X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14) ///
(Y2 Y1 Y3    X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14) ///
(Y3 Y1    X1 X2 X3 X4 X5 X6 X7 X8 X9 X10 X11 X12 X13 X14)

Y1 through Y3 are the endogenous variables, and the rest are exogenous.

$\endgroup$
2
  • 3
    $\begingroup$ I think there's an econometric issue here about whether this system of equations is identified, and this is not just a software error interpretation question (which would be arguably off-topic). $\endgroup$
    – dimitriy
    Sep 20 '17 at 22:41
  • 3
    $\begingroup$ This error is a real error due to lack of identifiability as @DimitriyV.Masterov hypothesized, not a coding or other s/w error. For identifiability, the number of excluded exogenous variables must be $\geq$ the number of endogenous variables included in the RHS in each equation. So, for equation 2, there are two endogenous variables on the RHS but only one excluded exogenous variable. For equation 3, there is one endogenous variable on the RHS but no excluded exogenous variables. Equation 1 is OK. $\endgroup$
    – jbowman
    Sep 21 '17 at 21:04
1
$\begingroup$

Expanding on my comment:

This error is a real error due to lack of identifiability as @DimitriyV.Masterov hypothesized, not a coding or other s/w error. For identifiability, the number of excluded exogenous variables must be $\geq$ the number of endogenous variables included in the RHS in each equation. So, for equation 2, there are two endogenous variables on the RHS but only one excluded exogenous variable. For equation 3, there is one endogenous variable on the RHS but no excluded exogenous variables. Equation 1 is OK.

Note that you can't be indiscriminate about which exogenous variables to remove, for example, if you removed X1 from equation 3, you would have accidentally removed X1 from the set of exogenous variables, and now all of the equations would contain all of the exogenous variables. Since the full identifiability conditions are quite complex, it's good to have simple rules to protect you to some extent from that complexity, and one such simple rule of thumb is to have at least one exogenous variable per equation that is unique to that equation.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.