# RE:st: Part II: An error in the xtabond-description

 From "David M. Drukker, StataCorp" To statalist@hsphsun2.harvard.edu Subject RE:st: Part II: An error in the xtabond-description Date Mon, 23 Aug 2004 19:02:40 -0500

Giovanni Bruno <giovanni.bruno@uni-bocconi.it> pointed out that there is an
issue with

> [the formula for] S2 (last formula at page 32 in the xt manual), which
> is based on A2 that necessarily uses 1-step residuals. This is not
> the 2-step Sargan in Arellano and Bond (1991) paper, formula (10) p.
> 282 that employs 2-step residuals.

Giovani is correct that equation (10) in Arellano and Bond (1991) uses the
two-step residuals to define the Sargan statistic for the two-step
estimator.  However, in this equation Arellano and Bond (1991) deviated from
the common definition of the two-step Sargan statistic as the value of the
GMM objective function at its two-step optimum.  The definition implemented
in -xtabond- is the more common definition of the Sargan-Hansen statistic.

To clarify define,

A_2 = (\sum_i Z_i'e1_i'e1_i Z_i)^{-1}

S_2 = (\sum_i e2_i'Z_i)A_2(\sum_i Z_i' e2)

where 	e1_i are one-step residuals
e2_i are two-step residuals
Z_i is the matrix of instruments for each panel i.

The coefficients beta that minimize S_2 with residuals e2_i are the GMM
estimates.  This is the Hansen (1982) formulation in which the weight matrix
A_2 is a function of the residuals from the previous step, but not the
current step.

As Mark Schaffer <M.E.Schaffer@hw.ac.uk> pointed out, the Sargan statistic
with A_2 as function of e2_i still converges to a chi-squared distribution,
so there is nothing wrong with equation in Arellano and Bond (1991).

However, as Mark also noted, that definition differs from the standard
textbook version in which A_2 is a function of e1_i which is discussed in
Hayashi(2000) and Davidson and MacKinnon(2004, 1993).  Thus, the A_2 as a
function of e1_i is well documented.

--David
ddrukker@stata.com

References

Arellano, M. and Bond, S. (1991) "Some tests of specification for panel data:
Monte Carlo evidence and an application to Employment Equations", The Review
of Economic Studies, 58(2) 277-297.

Davidson, R and MacKinnon J. (2004) Econometric Theory and Methods.
New York:Oxford University Press.

Davidson, R and MacKinnon J. (1993) Estimation and Inference in Econometrics.
New York:Oxford University Press.

Hansen, L.P. (1982) "Large Sample properties of GMM estimators" Econometrica
50(4) 1029-1054.

Hayashi, F. (2000) Econometrics. New York:Princeton University Press.
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/