Hi,
I'm using the -xtabond2- command to estimate a panel data growth model, but I have some problems with exactly how to use it.
I have read Roodman's paper on the estimator, but unfortunately I get problems with the Hansen test of overidentifying restrictions. The p-value seems to be 0 no matter how I specify the model. I have experimented both with the use of "collapse", the number of lags included and changes to which variables are treated as endogenuos, exogenous and predetermined.
So I was wondering if any of you might have an idea of what generally causes this problem. I have found several articles that use the approach on similar data without problems, but unfortunately they don't state their exact Stata codes. The code I use and the estimates I get look like this:
I would be very thankful for any advice or thoughts on my approach.
Best, Sebastian
I'm using the -xtabond2- command to estimate a panel data growth model, but I have some problems with exactly how to use it.
I have read Roodman's paper on the estimator, but unfortunately I get problems with the Hansen test of overidentifying restrictions. The p-value seems to be 0 no matter how I specify the model. I have experimented both with the use of "collapse", the number of lags included and changes to which variables are treated as endogenuos, exogenous and predetermined.
So I was wondering if any of you might have an idea of what generally causes this problem. I have found several articles that use the approach on similar data without problems, but unfortunately they don't state their exact Stata codes. The code I use and the estimates I get look like this:
:
. xtabond2 lev llev eb dp gr si ag indmed_lis i.year if listedf=="Listed", gmm(llev, lag(1 5) collapse) gmm(eb dp gr si, lag(2 5) collapse) iv(ag) iv(indmed_lis i.year) twostep robust
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
Warning: Two-step estimated covariance matrix of moments is singular.
Using a generalized inverse to calculate optimal weighting matrix for two-step estimation.
Difference-in-Sargan/Hansen statistics may be negative.
Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: id Number of obs = 28484
Time variable : year Number of groups = 3466
Number of instruments = 37 Obs per group: min = 5
Wald chi2(16) = 4139.68 avg = 8.22
Prob > chi2 = 0.000 max = 9
------------------------------------------------------------------------------
| Corrected
lev | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
llev | .7483798 .0236755 31.61 0.000 .7019766 .794783
eb | -.0469139 .061858 -0.76 0.448 -.1681534 .0743257
dp | .1397141 .1906327 0.73 0.464 -.2339191 .5133473
gr | .0506578 .0382378 1.32 0.185 -.024287 .1256026
si | -.0078096 .0080076 -0.98 0.329 -.0235041 .007885
ag | .0122202 .0060876 2.01 0.045 .0002888 .0241516
indmed_lis | .1778161 .0267415 6.65 0.000 .1254038 .2302285
|
year |
6 | 0 (empty)
7 | .0177745 .0078789 2.26 0.024 .0023322 .0332168
8 | .0491287 .013048 3.77 0.000 .0235551 .0747024
9 | .0081308 .0133151 0.61 0.541 -.0179663 .0342279
10 | -.0037079 .0095881 -0.39 0.699 -.0225002 .0150843
11 | .018332 .0120539 1.52 0.128 -.0052931 .0419572
12 | .0213359 .0126224 1.69 0.091 -.0034036 .0460753
13 | .0166446 .0126895 1.31 0.190 -.0082264 .0415155
14 | .0154747 .0108141 1.43 0.152 -.0057206 .03667
|
_cons | .0499256 .0763906 0.65 0.513 -.0997973 .1996484
------------------------------------------------------------------------------
Instruments for first differences equation
Standard
D.(indmed_lis 6b.year 7.year 8.year 9.year 10.year 11.year 12.year 13.year
14.year)
D.ag
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/5).(eb dp gr si) collapsed
L(1/5).llev collapsed
Instruments for levels equation
Standard
indmed_lis 6b.year 7.year 8.year 9.year 10.year 11.year 12.year 13.year
14.year
age_ln
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.(eb dp gr si) collapsed
D.llev collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -17.83 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = 0.20 Pr > z = 0.839
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(20) = 273.33 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(20) = 112.45 Prob > chi2 = 0.000
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(15) = 55.07 Prob > chi2 = 0.000
Difference (null H = exogenous): chi2(5) = 57.38 Prob > chi2 = 0.000
gmm(llev, collapse lag(1 5))
Hansen test excluding group: chi2(14) = 33.54 Prob > chi2 = 0.002
Difference (null H = exogenous): chi2(6) = 78.91 Prob > chi2 = 0.000
gmm(eb dp gr si, collapse lag(2 5))
Hansen test excluding group: chi2(0) = 19.72 Prob > chi2 = .
Difference (null H = exogenous): chi2(20) = 92.73 Prob > chi2 = 0.000
iv(ag)
Hansen test excluding group: chi2(19) = 112.44 Prob > chi2 = 0.000
Difference (null H = exogenous): chi2(1) = 0.01 Prob > chi2 = 0.933
iv(indmed_lis 6b.year 7.year 8.year 9.year 10.year 11.year 12.year 13.year 14.year)
Hansen test excluding group: chi2(11) = 25.00 Prob > chi2 = 0.009
Difference (null H = exogenous): chi2(9) = 87.45 Prob > chi2 = 0.000
Best, Sebastian