Hello!
Dear all,
I couldn't pass Hansen and Sargan tests. The P-values of both of H-S is lesser than 5%, which means that I have over identifying issues in instrument variables.
I'm using xtabond2 by Stata 14.0 in MacBook Air.
My current model is like this:
Dependent variable: 141 different home countries inflows to China from 2003 to 2016
Independent variables: All China's factors, for example, GDP, wage cost and education levels from 2003 and 2016.
For this model, the issue is that on the right side of the equation, there are no variations except for the time variance.
Most of the variables are statistically significant but such results are very biased according to both Hansen and Sargan (both P is consistently lesser than 5% even in a few robust tests). In the last few days, I tried all the possible solutions to solve this biased issue, for example, put some key variables of investing countries in the model, using different proxies, using xbond2 vec robust. But Hansen and Sargon are still < 5%.
I was wondering that:
1. Is it possible to use my current model to get unbiased results, meaning get P value of H-S greater than 5%?
2. However, if there is nothing to do with Stata codes, my next potential solution is to change the econometric model into the gravity model, which will include a few countries in the sample. The gravity model is often used in cross-country analyses of trade and investment flows. A dummy variable will be used to specifically look at China's determinants.
In this case, will it be sufficient to use a dummy variable to identify the true determinants of FDI in China in a dynamic cross-country panel model?
Is it appropriate to use cross-country analysis to address my real intention, which is to explore the determinants of inward FDI in China?
Dear all,
I couldn't pass Hansen and Sargan tests. The P-values of both of H-S is lesser than 5%, which means that I have over identifying issues in instrument variables.
I'm using xtabond2 by Stata 14.0 in MacBook Air.
My current model is like this:
Dependent variable: 141 different home countries inflows to China from 2003 to 2016
Independent variables: All China's factors, for example, GDP, wage cost and education levels from 2003 and 2016.
For this model, the issue is that on the right side of the equation, there are no variations except for the time variance.
Most of the variables are statistically significant but such results are very biased according to both Hansen and Sargan (both P is consistently lesser than 5% even in a few robust tests). In the last few days, I tried all the possible solutions to solve this biased issue, for example, put some key variables of investing countries in the model, using different proxies, using xbond2 vec robust. But Hansen and Sargon are still < 5%.
I was wondering that:
1. Is it possible to use my current model to get unbiased results, meaning get P value of H-S greater than 5%?
2. However, if there is nothing to do with Stata codes, my next potential solution is to change the econometric model into the gravity model, which will include a few countries in the sample. The gravity model is often used in cross-country analyses of trade and investment flows. A dummy variable will be used to specifically look at China's determinants.
In this case, will it be sufficient to use a dummy variable to identify the true determinants of FDI in China in a dynamic cross-country panel model?
Is it appropriate to use cross-country analysis to address my real intention, which is to explore the determinants of inward FDI in China?
Code:
xtabond2 lifdic l.lifdic lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis, gmm(l.lifdic, lag(2 2)) iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis) robust twostep ***robust test: (developing+devloped countries) xtabond2 lifdic l.lifdic lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis developed, gmm(l.lifdic, lag(2 2)) iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis developed) robust twostep xtabond2 lifdic l.lifdic lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis developing, gmm(l.lifdic, lag(2 2)) iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis developing) robust twostep
PHP Code:
***System GMM - using xtabond2
. xtabond2 lifdic l.lifdic lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis, gmm(l.li
> fdic, lag(2 2)) iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis) robust twostep
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: idc Number of obs = 1428
Time variable : time Number of groups = 141
Number of instruments = 32 Obs per group: min = 1
Wald chi2(11) = 38771.08 avg = 10.13
Prob > chi2 = 0.000 max = 13
------------------------------------------------------------------------------
| Corrected
lifdic | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lifdic |
L1. | .5934971 .1088609 5.45 0.000 .3801337 .8068606
|
lcgdp | 12.88194 3.40175 3.79 0.000 6.214631 19.54925
lcgdppc | 7.239806 3.24163 2.23 0.026 .8863284 13.59328
ldistance | -.2553822 .1843251 -1.39 0.166 -.6166528 .1058884
lexportpc | .2121053 .0932632 2.27 0.023 .0293127 .3948979
cexchrate | 2.181785 .5674756 3.84 0.000 1.069554 3.294017
cinf | -.2170425 .0574007 -3.78 0.000 -.3295458 -.1045392
cenrollsenh | .0674857 .0208496 3.24 0.001 .0266212 .1083502
crail | -16.16331 4.352948 -3.71 0.000 -24.69494 -7.631691
ctwage | -19.12807 4.80573 -3.98 0.000 -28.54713 -9.70901
crisis | -.0652163 .0948914 -0.69 0.492 -.2511999 .1207673
_cons | 0 (omitted)
------------------------------------------------------------------------------
Instruments for first differences equation
Standard
D.(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail
ctwage crisis)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L2.L.lifdic
Instruments for levels equation
Standard
lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage
crisis
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.L.lifdic
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -5.36 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = -0.29 Pr > z = 0.769
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(20) = 59.04 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(20) = 37.48 Prob > chi2 = 0.010
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(9) = 15.33 Prob > chi2 = 0.082
Difference (null H = exogenous): chi2(11) = 22.15 Prob > chi2 = 0.023
iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis)
Hansen test excluding group: chi2(10) = 10.62 Prob > chi2 = 0.388
Difference (null H = exogenous): chi2(10) = 26.85 Prob > chi2 = 0.003
.
. ***robust Check (1)(developing+devloped countries)
. xtabond2 lifdic l.lifdic lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis developed
> , gmm(l.lifdic, lag(2 2)) iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis devel
> oped) robust twostep
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: idc Number of obs = 1428
Time variable : time Number of groups = 141
Number of instruments = 33 Obs per group: min = 1
Wald chi2(12) = 39625.67 avg = 10.13
Prob > chi2 = 0.000 max = 13
------------------------------------------------------------------------------
| Corrected
lifdic | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lifdic |
L1. | .5602429 .1191553 4.70 0.000 .3267027 .793783
|
lcgdp | 12.29317 3.47719 3.54 0.000 5.478001 19.10833
lcgdppc | 6.612397 3.370993 1.96 0.050 .0053733 13.21942
ldistance | -.330394 .192357 -1.72 0.086 -.7074067 .0466187
lexportpc | .1865692 .0878448 2.12 0.034 .0143966 .3587418
cexchrate | 2.024548 .6118818 3.31 0.001 .8252815 3.223814
cinf | -.2033565 .0602156 -3.38 0.001 -.3213768 -.0853361
cenrollsenh | .065221 .0208237 3.13 0.002 .0244073 .1060347
crail | -15.26296 4.499788 -3.39 0.001 -24.08238 -6.44354
ctwage | -18.0594 5.060348 -3.57 0.000 -27.9775 -8.141302
crisis | -.0708723 .0946918 -0.75 0.454 -.2564649 .1147202
developed | .6804534 .2994321 2.27 0.023 .0935772 1.26733
_cons | 0 (omitted)
------------------------------------------------------------------------------
Instruments for first differences equation
Standard
D.(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail
ctwage crisis developed)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L2.L.lifdic
Instruments for levels equation
Standard
lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage
crisis developed
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.L.lifdic
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -4.97 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = -0.37 Pr > z = 0.715
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(20) = 58.75 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(20) = 36.59 Prob > chi2 = 0.013
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(9) = 14.55 Prob > chi2 = 0.104
Difference (null H = exogenous): chi2(11) = 22.05 Prob > chi2 = 0.024
iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis developed)
Hansen test excluding group: chi2(9) = 9.98 Prob > chi2 = 0.352
Difference (null H = exogenous): chi2(11) = 26.61 Prob > chi2 = 0.005
. xtabond2 lifdic l.lifdic lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis developin
> g, gmm(l.lifdic, lag(2 2)) iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis deve
> loping) robust twostep
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: idc Number of obs = 999
Time variable : time Number of groups = 105
Number of instruments = 32 Obs per group: min = 1
Wald chi2(11) = 192.65 avg = 9.51
Prob > chi2 = 0.000 max = 13
------------------------------------------------------------------------------
| Corrected
lifdic | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
lifdic |
L1. | .6407642 .1246456 5.14 0.000 .3964633 .8850652
|
lcgdp | 10.64585 30.47404 0.35 0.727 -49.08217 70.37387
lcgdppc | 6.427089 29.19794 0.22 0.826 -50.79982 63.654
ldistance | -.4653956 .2116647 -2.20 0.028 -.8802508 -.0505404
lexportpc | .0585626 .0743212 0.79 0.431 -.0871042 .2042295
cexchrate | 2.208828 .7372744 3.00 0.003 .7637965 3.653859
cinf | -.1672561 .0664066 -2.52 0.012 -.2974106 -.0371016
cenrollsenh | .0854052 .0235581 3.63 0.000 .0392321 .1315784
crail | -14.75972 5.015519 -2.94 0.003 -24.58995 -4.929482
ctwage | -16.17256 6.072314 -2.66 0.008 -28.07408 -4.271046
crisis | -.0673141 .1221707 -0.55 0.582 -.3067643 .1721361
developing | 24.48937 618.2806 0.04 0.968 -1187.318 1236.297
_cons | 0 (omitted)
------------------------------------------------------------------------------
Instruments for first differences equation
Standard
D.(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail
ctwage crisis developing)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L2.L.lifdic
Instruments for levels equation
Standard
lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage
crisis developing
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.L.lifdic
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -4.87 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = 0.18 Pr > z = 0.858
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(19) = 38.95 Prob > chi2 = 0.004
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(19) = 29.00 Prob > chi2 = 0.066
(Robust, but weakened by many instruments.)
Difference-in-Hansen tests of exogeneity of instrument subsets:
GMM instruments for levels
Hansen test excluding group: chi2(8) = 12.98 Prob > chi2 = 0.112
Difference (null H = exogenous): chi2(11) = 16.02 Prob > chi2 = 0.141
iv(lcgdp lcgdppc ldistance lexportpc cexchrate cinf cenrollsenh crail ctwage crisis developing)
Hansen test excluding group: chi2(9) = 13.99 Prob > chi2 = 0.123
Difference (null H = exogenous): chi2(10) = 15.01 Prob > chi2 = 0.132
