1.000 Value for Hansen Test (GMM). Is this normal?

[Matthew Garrett]

Hello everyone,

I recently ran a GMM regression on my data and when attempting to analyze the AR(1), AR(2), Sargan and Hansen test results I realize that my Hansen test result came back as 1.000. Is this normal? It says its robust, but weakened by many instruments.

Thank you.

This is the code I am running for the GMM: xi: xtabond2 NCSKEW L.LaggedNCSKEW LaggedPC LaggedSize LaggedMB LaggedLev LaggedROA LaggedSTURN LaggedRet LaggedSIGMA COVID, gmm(L.LaggedNCSKEW LaggedPC LaggedSize LaggedMB LaggedLev LaggedROA LaggedSTURN LaggedRet LaggedSIGMA) iv(COVID) small robust

When altering the model (amount of variables), I get different values for AR(1), AR(2), the Sargan test, but the Hansen Test still gives me a value of 1.000. Is this normal?

All help is appreciated, thank you very much!

[Sebastian Kripfganz]

This is likely a result of too many instruments. You need to collapse and/or curtail the instruments. See the following presentation for further information:

• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.

[Matthew Garrett]

This is likely a result of too many instruments. You need to collapse and/or curtail the instruments. See the following presentation for further information:

• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.

Thank you for your help! I read through it and collapsing it did solve my issue.

However, certain coefficients and the degree of significance of certain variables changed after adding the collapse command (some became less significant, some became insignificant).

Should I be worried about this? Or do I now have better estimates and better values from my regression? I understand that it does say "Too many instruments relative to the cross-sectional sample size can cause biased coefficient and standard error estimates and weakened specification tests (Roodman, 2009a)", so I should have better values now due to collapsing the instruments.

I apologize for the beginner question - I am a Stata beginner and am learning as I go. Thanks once again.

[Sebastian Kripfganz]

That results change when you modify the specification is not surprising. As you noticed, they can go either way. Your initial results are not a reliable basis for comparison because of the too-many-instruments problem.

[Matthew Garrett]

That solves it. Thanks a lot!