Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: ivregress with gmm vs 2sls
From
Tunga Kantarcı <[email protected]>
To
[email protected]
Subject
st: ivregress with gmm vs 2sls
Date
Thu, 28 Jul 2011 12:13:56 +0200
Hello,
I am estimating an IV model using the gmm and 2sls estimation methods.
They produce substantially different results for some reason I am
trying to figure out. Please note the following.
1. My data is panel but, for some reason, I am not exploiting the
panel nature of the data (by using a FE or RE model) and I pool the
cross sections.
2. When I exclude the variable AGET from the regression, then the
results from gmm and 2sls are very similar. Hence, the problem seems
to be about specification. When AGET is included, the coefficient to
the variable HWP0134 is estimated to be zero in the 2sls regression.
Hence, I am trying to figure out why I receive a zero coefficient.
3. I tried the ivreg2 command with the 2sls estimation method to see
if it also gives me a zero coefficient estimate. It did not. And the
results using the ivreg2 command with the 2sls methid are rather
similar to the results from the ivregress command using the gmm
method. Hence, it seems that there is something going on with the
ivregress with 2sls method.
I would appreciate any clues...
Estimation results:
ivregress gmm SRH (HWP0134 HWP3570=ELA2 ELA3 ELA4 ELAS2 ELAS3 ELAS4)
AGE AGES AGET NCL, vce(cluster HHIDPN)
Instrumental variables (GMM) regression Number of obs = 46648
Wald chi2(6) = 337.29
Prob > chi2 = 0.0000
R-squared = .
GMM weight matrix: Cluster (HHIDPN) Root MSE = 1.4876
(Std. Err. adjusted for 12164 clusters in HHIDPN)
------------------------------------------------------------------------------
| Robust
SRH | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
HWP0134 | 2.987662 1.233466 2.42 0.015 .5701128 5.405211
HWP3570 | .8978863 .3963741 2.27 0.023 .1210073 1.674765
AGE | -1.407141 .6269281 -2.24 0.025 -2.635898 -.1783846
AGES | .0230615 .0101469 2.27 0.023 .0031739 .0429491
AGET | -.0001195 .0000531 -2.25 0.024 -.0002236 -.0000155
NCL | .0355435 .0064052 5.55 0.000 .0229895 .0480975
_cons | 28.60861 12.23743 2.34 0.019 4.623689 52.59352
------------------------------------------------------------------------------
Instrumented: HWP0134 HWP3570
Instruments: AGE AGES AGET NCL ELA2 ELA3 ELA4 ELAS2 ELAS3 ELAS4
ivregress 2sls SRH (HWP0134 HWP3570=ELA2 ELA3 ELA4 ELAS2 ELAS3 ELAS4)
AGE AGES AGET NCL, vce(cluster HHIDPN)
Instrumental variables (2SLS) regression Number of obs = 46648
Wald chi2(6) =23412.02
Prob > chi2 = 0.0000
R-squared = 0.0338
Root MSE = 1.0329
(Std. Err. adjusted for 12164 clusters in HHIDPN)
------------------------------------------------------------------------------
| Robust
SRH | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
HWP0134 | 0 .0157014 0.00 1.000 -.0307742 .0307742
HWP3570 | -.0393177 .0930963 -0.42 0.673 -.221783 .1431476
AGE | .0668445 .0136595 4.89 0.000 .0400725 .0936166
AGES | -.0005923 .0006007 -0.99 0.324 -.0017697 .000585
AGET | 2.46e-06 2.65e-06 0.93 0.354 -2.74e-06 7.65e-06
NCL | .0387995 .0042354 9.16 0.000 .0304982 .0471008
_cons | (omitted)
------------------------------------------------------------------------------
Instrumented: HWP0134 HWP3570
Instruments: AGE AGES AGET NCL ELA2 ELA3 ELA4 ELAS2 ELAS3 ELAS4
*
* 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/