Why no results for "estat abond" when running an xtabond mondel with "vce(gmm)"?

[Brenden Beck]

Hi all,

I'm estimating a model with the xtabond command, and when I run the post-estimation command "estat abond" I get the error message:

artests not computed for one-step system estimator with vce(gmm)

If I run the same model with the "vce(robust)" or the "two-step" gmm options, it generates an abond test result no problem. I've read through the help files, but can't find a reason that Stata shouldn't be generating a result. Can anyone shed any light on this? Syntax below. Thanks a bunch! Syntax below. I'm using Stata 15.

xtabond var1 var2 var3 var4, endog(var5, lagstr(1,4)) artest(2) maxlag(4) maxldep(4) vce(gmm)
estat abond

[William Lisowski]

Among the help files, did you look at the writeup in

help xtabond postestimation

for estat abond? It confirms exactly what you described, with the explanation

Remarks for estat abond

    After the one-step system estimator, the test can be computed only when
    vce(robust) has been specified.  (The system estimator is used to estimate the
    constant in xtabond.)

Perhaps the parenthetical comment explains the reason for this restriction to someone who is familiar with the Arellano-Bond test for serial correlation, but I am not that someone.

In the PDF documentation for xtabond postestimation linked to from the top of the help output we are told to see the Remarks and examples in the PDF documentation for xtabond for more remarks about estat abond that are made in the context of the examples analyzed therein.

[Brenden Beck]

Thanks so much for your reply, William. Yes, my question is about why the test can only be computed with robust estimators and not GMM estimators. I've seen it done in the lit, so I was assuming it was a Stata quirk, but maybe I need to look harder at the lit to figure out why. Thanks again.

[Sebastian Kripfganz]

The official Stata commands for dynamic panel models lack some functionality. It is possible to compute the Arellano-Bond test after one-step first-difference GMM with non-robust standard errors, for example using my xtdpdgmm command:

• XTDPDGMM: new Stata command for efficient GMM estimation of linear (dynamic) panel models
• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.

[Brenden Beck]

Perfect. Thanks so much, Sebastian.