XTABOND with windmeijer correction for the two-step estimator

[mvanbenthem]

Hey all,

I am estimating my dynamic panel data models with among others, the two-step difference GMM framework (i.e. the Arellano-Bond estimator), using the command xtabond.
I added vce(robust) to apply the windmeijer correction .

When i now want to run the Sargan test using estat sargan i get the following message:

cannot calculate Sargan test with vce(robust)


Any idea what i am doing wrong, and how i can work around this to still retrieve the sargan test?


Thanks advance,
Maarten

[Sebastian Kripfganz]

The Stata manual for "xtabond postestimation" says: "The distribution of the Sargan test is known only when the errors are independently and identically distributed. For this reason, estat sargan does not produce a test statistic when vce(robust) was specified in the call to xtabond."

You may consider the user-written command xtabond2 by David Roodman. For the two-step estimator the Hansen test should be used instead of the Sargan test. See also the discussion in the accompanying Stata journal article:
Roodman (2009). How to do xtabond2: An introduction to difference and system GMM in Stata. The Stata Journal, 9 (1), 86-136.