Need help in xtdpd command for twostep GMM

[Chua Mei Shan]

Hi everyone,

I am doing the dynamic capital structure study. For my model estimation, I am using the xtdpdsys for the estimation.

The command and result that i get:

. xtdpdsys bvtd1 size tang prof ndts mtb y08-y17 oilgas-technology, twostep
note: oilgas dropped from div() because of collinearity
note: basicmaterials dropped from div() because of collinearity
note: industrial dropped from div() because of collinearity
note: consumergoods dropped from div() because of collinearity
note: healthcare dropped from div() because of collinearity
note: consumerservices dropped from div() because of collinearity
note: telecommunication dropped from div() because of collinearity
note: utilities dropped from div() because of collinearity
note: financials dropped from div() because of collinearity
note: technology dropped from div() because of collinearity
note: y17 dropped because of collinearity
note: oilgas dropped because of collinearity
System dynamic panel-data estimation Number of obs = 4000
Group variable: code Number of groups = 400
Time variable: year
Obs per group: min = 10
avg = 10
max = 10
Number of instruments = 69 Wald chi2(24) = 1971.86
Prob > chi2 = 0.0000
Two-step results
bvtd1 Coef. Std. Err. z P>z [95% Conf. Interval]
bvtd1
L1. .8197323 .0221754 36.97 0.000 .7762693 .8631953
size -.0058926 .0059311 -0.99 0.320 -.0175174 .0057322
tang .0768351 .022402 3.43 0.001 .0329281 .1207422
prof -.0000512 .0000724 -0.71 0.480 -.000193 .0000907
ndts -1.421603 .3011687 -4.72 0.000 -2.011883 -.8313231
mtb .0006838 .0002075 3.30 0.001 .0002772 .0010905
y08 .0005344 .0054573 0.10 0.922 -.0101617 .0112305
y09 -.0128341 .0052228 -2.46 0.014 -.0230705 -.0025976
y10 -.0142504 .0044459 -3.21 0.001 -.0229641 -.0055366
y11 -.0011741 .0044641 -0.26 0.793 -.0099237 .0075754
y12 -.0005628 .0036553 -0.15 0.878 -.0077271 .0066014
y13 .0044216 .0040627 1.09 0.276 -.0035411 .0123843
y14 .0090864 .0038033 2.39 0.017 .001632 .0165407
y15 -.0001848 .0035237 -0.05 0.958 -.007091 .0067215
y16 -.0014269 .0035585 -0.40 0.688 -.0084014 .0055476
basicmater~s -27375.68 69361.72 -0.39 0.693 -163322.2 108570.8
industrial 4967.675 12075.13 0.41 0.681 -18699.15 28634.5
consumergo~s 3990.209 9896.538 0.40 0.687 -15406.65 23387.07
healthcare -113892.1 286057.2 -0.40 0.691 -674553.9 446769.7
consumerse~s -24376.19 61053.66 -0.40 0.690 -144039.2 95286.78
telecommun~n -12390.92 31793.83 -0.39 0.697 -74705.68 49923.85
utilities 91116.32 229049.2 0.40 0.691 -357811.8 540044.5
financials 12373.62 30862.86 0.40 0.688 -48116.49 72863.72
technology -29757.7 76054.21 -0.39 0.696 -178821.2 119305.8
_cons 4639.713 11893.17 0.39 0.696 -18670.47 27949.89
Warning: gmm two-step standard errors are biased; robust standard
errors are recommended.
Instruments for differenced equation
GMM-type: L(2/.).bvtd1
Standard: D.size D.tang D.prof D.ndts D.mtb D.y08 D.y09 D.y10 D.y11 D.y12	D.y13	D.y14	D.y15	D.y16	D.y17
Instruments for level equation
GMM-type: LD.bvtd1
Standard: _cons
. estat sargan
Sargan test of overidentifying restrictions
H0: overidentifying restrictions are valid
chi2(44) = 49.72555
Prob > chi2 = 0.2559
. estat abond
Arellano-Bond test for zero autocorrelation in first-differenced errors
+-----------------------+
Order z Prob > z
------+----------------
1 -8.6273 0.0000
2 -2.7044 0.0068
+-----------------------+
H0: no autocorrelation
.

As can be seen the AR(2) result show second order serial correlation (pvvalue< 0.05) which is inconsistent with GMM theory.

After reading thru the articles, prior researches applied for deeper lag in order to find the instruments for differences and level equation.

Here is the command that i entered into the stata and the result that i obtained:

. xtdpd bvtd1 l.bvtd1 size tang prof ndts mtb y08-y17 oilgas-technology, twostep dgmmiv(bvtd1, lagrange (2 .)) lgmmiv(bvtd1, lag (5))

> div(l5.size l5.tang l5.prof l5.ndts l5.mtb y08-y17 oilgas-technology, nodifference) liv(L5D.size L5D.tang L5D.prof L5D.ndts L5D.mtb)

> artests(2)

note: y08 dropped from div() because of collinearity

note: y09 dropped from div() because of collinearity

note: y10 dropped from div() because of collinearity

note: y11 dropped from div() because of collinearity

note: y12 dropped from div() because of collinearity

note: y17 dropped from div() because of collinearity

note: oilgas dropped from div() because of collinearity

note: y08 dropped because of collinearity

note: y09 dropped because of collinearity

note: y10 dropped because of collinearity

note: y11 dropped because of collinearity

note: y12 dropped because of collinearity

note: y17 dropped because of collinearity

note: oilgas dropped because of collinearity

note: D.basicmaterials dropped because of collinearity

note: D.industrial dropped because of collinearity

note: D.consumergoods dropped because of collinearity

note: D.healthcare dropped because of collinearity

note: D.consumerservices dropped because of collinearity

note: D.telecommunication dropped because of collinearity

note: D.utilities dropped because of collinearity

note: D.financials dropped because of collinearity

note: D.technology dropped because of collinearity

Dynamic panel-data estimation Number of obs = 2000

Group variable: code Number of groups = 400

Time variable: year

Obs per group: min = 5

avg = 5

max = 5

Number of instruments = 64 Wald chi2(10) = 1126.02

Prob > chi2 = 0.0000

Two-step results

bvtd1 Coef. Std. Err. z P>z [95% Conf. Interval]

bvtd1

L1. .7541001 .0275762 27.35 0.000 .7000518 .8081484

size .0348102 .0073834 4.71 0.000 .0203389 .0492815

tang .0488008 .0406117 1.20 0.230 -.0307967 .1283983

prof -.0032621 .0038996 -0.84 0.403 -.0109052 .0043811

ndts -.0497678 .5253582 -0.09 0.925 -1.079451 .9799153

mtb -.0003713 .0002597 -1.43 0.153 -.0008804 .0001377

y13 .0060966 .0034824 1.75 0.080 -.0007288 .012922

y14 .0121045 .0034664 3.49 0.000 .0053105 .0188985

y15 .0047397 .0030176 1.57 0.116 -.0011747 .0106541

y16 .0015274 .0031382 0.49 0.626 -.0046234 .0076782

_cons -.4027725 .0980327 -4.11 0.000 -.5949131 -.2106319

Warning: gmm two-step standard errors are biased; robust standard

errors are recommended.

Instruments for differenced equation

GMM-type: L(2/.).bvtd1

Standard: L5.size L5.tang L5.prof L5.ndts L5.mtb y13 y14 y15 y16 basicmaterials industrial consumergoods healthcare

consumerservices telecommunication utilities financials technology

Instruments for level equation

GMM-type: L5D.bvtd1

Standard: L5D.size L5D.tang L5D.prof L5D.ndts L5D.mtb _cons

. estat sargan

Sargan test of overidentifying restrictions

H0: overidentifying restrictions are valid

chi2(53) = 53.12699

Prob > chi2 = 0.4693

. estat abond

Arellano-Bond test for zero autocorrelation in first-differenced errors

+-----------------------+

Order z Prob > z

------+----------------

1 -5.8309 0.0000

2 -1.7698 0.0768

+-----------------------+

H0: no autocorrelation

My question is: is the command that i entered in the correct order?

[Sebastian Kripfganz]

Welcome to the Statalist forum. Please always use CODE delimiters for your Stata output here on Statalist.

Discussion continued in new topic: https://www.statalist.org/forums/for...r-two-step-gmm
(Generally, please do not open multiple topics on the same question.)