How to get R-squared in GMM regressions?

[Paco Rodriguez]

Hello everyone,
I am estimating Taylor rules of monetary policy using GMM and "gmm" command (Stata 14). I only estimate a single equation with a valid instrument set based on the J-Test, but the GMM output does not give me an R-squared or an adjusted R-squared.The point is that all the literature that uses GMM for Taylor estimates presents a very reasonable R-squared data, but I don't know how. The only alternative way is "ivregress command" but it does not allow to estimate equations.

Is there a way to get or calculate the R-squared in a GMM regression?

Thank you.

[Joro Kolev]

In general the R-squared is well defined only for OLS. For everything else the R-squared is a fudge.

You will probably not get a useful answer, because it is very unlikely that anybody here knows exactly what fudge the Taylor rule estimation literature does for the R-squared.

But to at least give yourself a chance to get an answer, ask your question properly by showing exactly what you typed at Stata, and exactly what Stata returned, and if possible by providing a sample of your data using -dataex-.

[Paco Rodriguez]

Thank you Joro. As an example of what I'm doing:

This is what I typed:

gmm (EONIA-((1-{b1})*({b2}+F12.Inflation*{b3}+F3.outputgap*{b4})+ EONIAt1*{b1})), igmm instruments(L(1/6).Inflation L9.Inflation L12.Inflation L(1/6).outputgap L9.outputgap L12.outputgap L(1/6).M3 L9.M3 L12.M3 L(1/6).yrbondyield L9.yrbondyield L12.yrbondyield) wmatrix(hac bartlett opt) variables(Inflation outputgap EONIAt1) vce(hac bartlett opt)

And what STATA returned:

note: iterative GMM parameter vector converged

GMM estimation

Number of parameters = 4
Number of moments = 33
Initial weight matrix: Unadjusted Number of obs = 96
GMM weight matrix: HAC Bartlett .
(lags chosen by Newey-West)

------------------------------------------------------------------------------
| HAC
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
/b1 | .9426539 .0055881 168.69 0.000 .9317015 .9536063
/b2 | .2459111 .1727983 1.42 0.155 -.0927673 .5845895
/b3 | 1.389057 .1030463 13.48 0.000 1.18709 1.591024
/b4 | .0209359 .0067138 3.12 0.002 .0077771 .0340947
------------------------------------------------------------------------------
HAC standard errors based on Bartlett kernel with . lags.
(Lags chosen by Newey-West method.)
Instruments for equation 1: L.Inflation L2.Inflation L3.Inflation L4.Inflation L5.Inflation
L6.Inflation L9.Inflation L12.Inflation L.outputgap L2.outputgap L3.outputgap
L4.outputgap L5.outputgap L6.outputgap L9.outputgap L12.outputgap L.M3 L2.M3 L3.M3 L4.M3
L5.M3 L6.M3 L9.M3 L12.M3 L.yrbondyield L2.yrbondyield L3.yrbondyield L4.yrbondyield
L5.yrbondyield L6.yrbondyield L9.yrbondyield L12.yrbondyield _cons

I completely understand your answer. Is there any other way to know how good the model fits in the sense of R2?