st: How to compare results from -reg-, -xtreg-, and -xtabond2-?

[Misha Spisok <misha.spisok@gmail.com>]

Hello, Statalist,

In Roodman's article, "How to Do xtabond2" (SJ, Vol. 9, No. 1, pp.
86-136, 2009), he mentions that the coefficient on the lagged
dependent variable from an ostensibly superior estimator ought to lie
between those from the naive OLS and LSDV (citing, in turn, Bond,
2002), and, as intuition would suggest, that it ought to be below
1.00.

This makes sense, but I'm wondering about the case of additional
regressors.  Are there similar guidelines for other "independent"
variables?

For example, in the above article, one of the other regressors is 'w.'
 Here are some relevant commands (mostly from the article).

webuse abdata
/* naive OLS */
. regress n nL1 w wL1 k kL1 kL2 ys ysL1 ysL2 yr*
. di _b[nL1]
1.0446429
. di _b[w]
-.52367267
/* LSDV */
. xi: regress n nL1 w wL1 k kL1 kL2 ys ysL1 ysL2 yr* i.id
. di _b[nL1]
.73294765
. di _b[w]
-.5597445
/* or, similarly, */
/* fixed effects regression */
. xtreg n nL1 w wL1 k kL1 kL2 ys ysL1 ysL2 yr*, fe
. di _b[nL1]
.6671535
. di _b[w]
-.56561777
. xtabond2 n L.n L2.n w L.w L(0/2).(k ys) yr*, gmm(L.(n w k))
iv(L(0/2).ys yr*) nolevel robust small
. di _b[L.n]
.81798666
. di _b[w]
-.68166853

Is there any reason to expect that the coefficient on 'w' from
-xtabond2- ought to lie between the results from naive OLS (-reg-) and
LSDV (either -xi: reg- or -xtreg, fe-)?  Are there any guidelines for
such a case (i.e., for variables besides the lagged dependent
variable)?

Best,

Misha
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/