Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: How to beat the Sargan test? (gmm for dummies)

 From db10 <[email protected]> To [email protected] Subject st: How to beat the Sargan test? (gmm for dummies) Date Wed, 21 Mar 2012 10:52:27 -0700 (PDT)

```Hi.

In my masters thesis I need to do some estimations for the effect different
determinants have on capital structure. The problem is that my knowledge for
GMM estimation is scarce.

I have managed to construct a model, with xtabond and xtdpdsys, which gives
me significant results for capital structure that is supported by earlier
empirical findings. The problem is that when I preform Sargan test of
overidentifying restrictions the H0 for overidentifying restrictions are
valid is confirmed.

I don’t understand how to implement endogenous variables and instrumental
variables. How to decide what variables are endogenous, exogenous and
instrumental is also a mystery to me. Hope somebody can help me with some
celerity.

Depandant variable: nclta= none current libilities to total assest
Explanatory variables: ebitta=profitability, lnoper=size, tfixta=collateral,
growth= growth

My model is:

. xtdpdsys nclta ebitta lnoper tfixta growth, lags(1) twostep artests(2)

System dynamic panel-data estimation         Number of obs         =
1902
Group variable: company                      Number of groups      =
480
Time variable: year
Obs per group:    min =
3
avg =
3.9625
max =
4

Number of instruments =     14               Wald chi2(5)          =
5263.26
Prob > chi2           =
0.0000
Two-step results
------------------------------------------------------------------------------
nclta |      Coef.   Std. Err.      z    P>|z|     [95% Conf.
Interval]
-------------+----------------------------------------------------------------
nclta |
L1. |   .6037445   .0644683     9.36   0.000      .477389
.7301001
|
ebitta |  -.1904182   .0056967   -33.43   0.000    -.2015836
-.1792528
lnoper |   .0542803   .0154855     3.51   0.000     .0239293
.0846312
tfixta |   .2824556   .0701347     4.03   0.000      .144994
.4199171
growth |   .0004994   .0003116     1.60   0.109    -.0001113
.0011102
_cons |   -.476992   .1435811    -3.32   0.001    -.7584057
-.1955783
------------------------------------------------------------------------------
Warning: gmm two-step standard errors are biased; robust standard
errors are recommended.
Instruments for differenced equation
GMM-type: L(2/.).nclta
Standard: D.ebitta D.lnoper D.tfixta D.growth
Instruments for level equation
GMM-type: LD.nclta
Standard: _cons

.
. *AUTOCORRELASJON TEST
. estat abond

Arellano-Bond test for zero autocorrelation in first-differenced errors
+-----------------------+
|Order |  z     Prob > z|
|------+----------------|
|   1  |  -5.82  0.0000 |
|   2  |-.49855  0.6181 |
+-----------------------+
H0: no autocorrelation

.
. *SARGAN TEST
. estat sargan
Sargan test of overidentifying restrictions
H0: overidentifying restrictions are valid

chi2(8)      =  26.56045
Prob > chi2  =    0.0008

--
View this message in context: http://statalist.1588530.n2.nabble.com/How-to-beat-the-Sargan-test-gmm-for-dummies-tp7393001p7393001.html
Sent from the Statalist mailing list archive at Nabble.com.

*
*   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/
```