Hi Statalisters,
I would like some input on my choices between xtabond2 and xtdpdgmm, as they differ and as I am unsure if I am on the right path. Excuse the confusing variable names they may not be final.
The specification is to the best of my knowledge in line with the literature. Specifically, I follow : Lags 2 and lags 3 of the levels of firm performance variable rdint_log, governance variables pressure_sens pressure_res inter1 inter2 owner_share and control variables ln_age log_employees. are employed as GMM-type instruments for the first differences equation. Lags 1 of the first differences of firm performance, corporate governance, and control variables are used as GMM-type instruments for the levels equation. gii_score is a country level variable which I also may want to interact with the governance variables.
inter1 and inter2 are owner_share * pressure_res owner_share * pressure_sens interactions respectively. I would highly appreciate some guidance, in terms of pointing out anything clearly off in my code or some other concern - as I am confused about the estimations, or rather the postestimation tests which I think fail in my model.
Since I face the issue of omitted time variables, and also the constant, I resorted to xtdpdgmm as suggested by Dr Kripfganz here on Statalist.
I tried to replicate - altough I am not sure if this is entirely equal to a system gmm, the results are not matching :
Running this code, I notice the missing of tests for serial correlation and Hansen tests, compared to xtabond 2. For example, when I try estat serial, I get the following no matter how I sort the data: .
I would like some input on my choices between xtabond2 and xtdpdgmm, as they differ and as I am unsure if I am on the right path. Excuse the confusing variable names they may not be final.
The specification is to the best of my knowledge in line with the literature. Specifically, I follow : Lags 2 and lags 3 of the levels of firm performance variable rdint_log, governance variables pressure_sens pressure_res inter1 inter2 owner_share and control variables ln_age log_employees. are employed as GMM-type instruments for the first differences equation. Lags 1 of the first differences of firm performance, corporate governance, and control variables are used as GMM-type instruments for the levels equation. gii_score is a country level variable which I also may want to interact with the governance variables.
inter1 and inter2 are owner_share * pressure_res owner_share * pressure_sens interactions respectively. I would highly appreciate some guidance, in terms of pointing out anything clearly off in my code or some other concern - as I am confused about the estimations, or rather the postestimation tests which I think fail in my model.
Code:
xtabond2 rdint_log L.rdint_log pressure_sens pressure_res inter1 inter2 owner_share ln_age log_employees i.year gii_score, gmm(rdint_log pressure_sens pressure_res owner_share inter1 inter2 log_employees, lag(2 3) collapse equation (diff)) gmm (pressure_sens pressure_res owner_share rdint_log, lag(1 1) collapse equation(level)) iv(i.year ln_age gii_score, equation(level)) two ro
Code:
Dynamic panel-data estimation, two-step system GMM
------------------------------------------------------------------------------
Group variable: id Number of obs = 3836
Time variable : year Number of groups = 664
Number of instruments = 27 Obs per group: min = 1
Wald chi2(17) = 4099.12 avg = 5.78
Prob > chi2 = 0.000 max = 7
-------------------------------------------------------------------------------
| Corrected
rdint_log | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------+----------------------------------------------------------------
rdint_log |
L1. | -.0231576 .0348396 -0.66 0.506 -.091442 .0451267
|
pressure_sens | -.6548435 .4478926 -1.46 0.144 -1.532697 .2230098
pressure_res | -2.142851 .8957042 -2.39 0.017 -3.898399 -.387303
inter1 | 8.663443 3.873895 2.24 0.025 1.070747 16.25614
inter2 | .0250038 1.146118 0.02 0.983 -2.221346 2.271354
owner_share | .0541628 .7377693 0.07 0.941 -1.391838 1.500164
ln_age | -.0232487 .116177 -0.20 0.841 -.2509514 .2044539
log_employees | -.2985976 .0797236 -3.75 0.000 -.4548531 -.1423422
|
year |
2010 | 0 (empty)
2011 | -3.030497 .8781278 -3.45 0.001 -4.751596 -1.309398
2012 | -3.181193 .9038798 -3.52 0.000 -4.952764 -1.409621
2013 | -3.164345 .899639 -3.52 0.000 -4.927605 -1.401085
2014 | -.1484917 .9110414 -0.16 0.871 -1.9341 1.637117
2015 | -3.003705 .8902785 -3.37 0.001 -4.748619 -1.258791
2016 | -3.13521 .9119268 -3.44 0.001 -4.922553 -1.347866
2017 | -3.111658 .9127353 -3.41 0.001 -4.900586 -1.322729
|
gii_score | .0351893 .0123263 2.85 0.004 .0110303 .0593484
_cons | 0 (omitted)
-------------------------------------------------------------------------------
Instruments for first differences equation
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(2/3).(rdint_log pressure_sens pressure_res owner_share inter1 inter2
log_employees) collapsed
Instruments for levels equation
Standard
2010b.year 2011.year 2012.year 2013.year 2014.year 2015.year 2016.year
2017.year ln_age gii_score
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
DL.(pressure_sens pressure_res owner_share rdint_log) collapsed
------------------------------------------------------------------------------
Arellano-Bond test for AR(1) in first differences: z = -12.60 Pr > z = 0.000
Arellano-Bond test for AR(2) in first differences: z = 5.25 Pr > z = 0.000
------------------------------------------------------------------------------
Sargan test of overid. restrictions: chi2(9) = 104.00 Prob > chi2 = 0.000
(Not robust, but not weakened by many instruments.)
Hansen test of overid. restrictions: chi2(9) = 66.44 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(5) = 38.48 Prob > chi2 = 0.000
Difference (null H = exogenous): chi2(4) = 27.96 Prob > chi2 = 0.000
gmm(pressure_sens pressure_res owner_share rdint_log, collapse eq(level) lag(1 1))
Hansen test excluding group: chi2(5) = 38.48 Prob > chi2 = 0.000
Difference (null H = exogenous): chi2(4) = 27.96 Prob > chi2 = 0.000
iv(2010b.year 2011.year 2012.year 2013.year 2014.year 2015.year 2016.year 2017.year ln_age gii_score, eq
> (level))
Hansen test excluding group: chi2(1) = 15.37 Prob > chi2 = 0.000
Difference (null H = exogenous): chi2(8) = 51.07 Prob > chi2 = 0.000
I tried to replicate - altough I am not sure if this is entirely equal to a system gmm, the results are not matching :
Code:
xtdpdgmm L.rdint_log rdint_log pressure_sens pressure_res inter1 inter2 owner_share ln_age log_employees gii_score, noserial gmmiv(L.rdint_log pressure_sens pressure_res owner_share inter1 inter2 log_employees, collapse model (difference)) iv(gii_score ln_age, difference model(difference)) twostep vce(robust) teffects overid
Code:
Group variable: id Number of obs = 3836
Time variable: year Number of groups = 664
Moment conditions: linear = 57 Obs per group: min = 1
nonlinear = 5 avg = 5.777108
total = 62 max = 7
(Std. Err. adjusted for 664 clusters in id)
-------------------------------------------------------------------------------
| WC-Robust
L.rdint_log | Coef. Std. Err. z P>|z| [95% Conf. Interval]
--------------+----------------------------------------------------------------
rdint_log | -.2300437 .2514488 -0.91 0.360 -.7228742 .2627868
pressure_sens | .4139911 .4576866 0.90 0.366 -.4830582 1.31104
pressure_res | 1.787731 .8391795 2.13 0.033 .1429689 3.432492
inter1 | -9.637961 3.956848 -2.44 0.015 -17.39324 -1.882682
inter2 | -.4744583 1.130968 -0.42 0.675 -2.691115 1.742198
owner_share | .8176477 .7949369 1.03 0.304 -.7404001 2.375695
ln_age | .1277035 .2034326 0.63 0.530 -.271017 .526424
log_employees | .1106846 .1047606 1.06 0.291 -.0946425 .3160116
gii_score | -.0275028 .0247245 -1.11 0.266 -.0759619 .0209564
|
year |
2012 | -.4822724 .1729211 -2.79 0.005 -.8211916 -.1433533
2013 | -.5913051 .1852836 -3.19 0.001 -.9544543 -.2281558
2014 | .3287016 .6799291 0.48 0.629 -1.003935 1.661338
2015 | 2.087275 .3400099 6.14 0.000 1.420868 2.753682
2016 | -.3082466 .1692142 -1.82 0.069 -.6399003 .0234071
2017 | -.0683088 .1565524 -0.44 0.663 -.3751459 .2385283
|
_cons | -4.436074 1.619953 -2.74 0.006 -7.611124 -1.261024
-------------------------------------------------------------------------------
Instruments corresponding to the linear moment conditions:
1, model(diff):
L1.L.rdint_log L2.L.rdint_log L3.L.rdint_log L4.L.rdint_log L5.L.rdint_log
L6.L.rdint_log L1.pressure_sens L2.pressure_sens L3.pressure_sens
L4.pressure_sens L5.pressure_sens L6.pressure_sens L7.pressure_sens
L1.pressure_res L2.pressure_res L3.pressure_res L4.pressure_res
L5.pressure_res L6.pressure_res L7.pressure_res L1.owner_share
L2.owner_share L3.owner_share L4.owner_share L5.owner_share L6.owner_share
L7.owner_share L1.inter1 L2.inter1 L3.inter1 L4.inter1 L5.inter1 L6.inter1
L7.inter1 L1.inter2 L2.inter2 L3.inter2 L4.inter2 L5.inter2 L6.inter2
L7.inter2 L1.log_employees L2.log_employees L3.log_employees
L4.log_employees L5.log_employees L6.log_employees L7.log_employees
2, model(diff):
D.gii_score D.ln_age
3, model(level):
2012bn.year 2013.year 2014.year 2015.year 2016.year 2017.year
4, model(level):
_cons
Code:
estat serial not sorted r(5);
Comment