how to make xtdpdsys and xtabond2 equivalent?

[Giweta Opolska]

I am trying to receive the same results using xtdpdsys and xtabond2 unsuccessfully.
Could you please guide me how to do it?

Here is the xtdpdsys code that I try to replicate with xtabond2:
xtdpdsys SR Prod Imp year*, vce(robust) end(UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn)
xtdpdsys SR UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn Prod Imp year*, vce(robust)
xtdpdsys SR Prod Imp year*, vce(robust) pre(UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn)


Thank you.

[Sebastian Kripfganz]

The corresponding command syntax for xtabond2 is as follows:

xtabond2 L(0/1).SR UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn Prod Imp year*, gmm(SR UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn, lag(2 .)) iv(Prod Imp year*, eq(diff)) h(2) robust

xtabond2 L(0/1).SR UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn Prod Imp year*, gmm(SR, lag(2 .)) iv(UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn Prod Imp year*, eq(diff)) h(2) robust

xtabond2 L(0/1).SR UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn Prod Imp year*, gmm(SR, lag(2 .)) gmm(UTSO UDSO TPA MarkOpen IRA Bal price VTP Ex GRPd PubOwn, lag(1 .)) iv(Prod Imp year*, eq(diff)) h(2) robust

[Giweta Opolska]

Thank you, Sebastian, for your prompt reply and help. It works perfectly now and the results from both codes are the same.

[kolpo kotha]

Hi, when I tried to used xtabond2 code following result is produced. Please show me the way to solve it.

. xtabond2 L(0/1).GDPPCG GDPPCi0 GI UNEM INFL URBAN EDUCE HEALE shadow LFPART GDSPGDP INV PIT CIT GTGS GOVTS PIT_SA PIT_SHADOW PIT_SHADOW_SA CIT
> _SA CIT_GLOB CIT_GLOB_SA GTGS_SA year1990- year2015, gmm(GDPPCG, lag(2 .)) gmm(GDPPCi0 GI UNEM INFL URBAN EDUCE HEALE shadow LFPART GDSPGDP INV
> PIT CIT GTGS GOVTS PIT_SA PIT_SHADOW PIT_SHADOW_SA CIT_SA CIT_GLOB CIT_GLOB_SA GTGS_SA, lag(1 .)) iv(year1990- year2015, eq(diff)) h(2) robust
Favoring space over speed. To switch, type or click on mata: mata set matafavor speed, perm.
quadcross(): 3900 unable to allocate working space
xtabond2_mata(): - function returned error
<istmt>: - function returned error

[Jorge Castro]

Hi All!, I use this code:

xi: xtdpdsys pcAS z2 pre_mean alt lnC_m_r_f__o lnF_E_A_S lnFungicidas lnHerbicidas lnI_a_n lnMaterial_propagacion i_a_* , lags(1) pre(lagprecio lagOtros) artests(2) vce(robust) twostep

note: pre_mean omitted from div() because of collinearity.
note: alt omitted from div() because of collinearity.
note: i_a_1 omitted from div() because of collinearity.
note: i_a_1 omitted because of collinearity.
note: i_a_5 omitted because of collinearity.

System dynamic panel-data estimation Number of obs = 18,624
Group variable: id Number of groups = 5,586
Time variable: AÑO
Obs per group:
min = 1
avg = 3.334049
max = 5

Number of instruments = 54 Wald chi2(16) = 2364.78
Prob > chi2 = 0.0000
Two-step results
----------------------------------------------------------------------------------------
| WC-robust
pcAS | Coefficient std. err. z P>|z| [95% conf. interval]
-----------------------+----------------------------------------------------------------
pcAS |
L1. | .8552736 .0443983 19.26 0.000 .7682544 .9422927
|
lagprecio | .0002514 .0001079 2.33 0.020 .00004 .0004629
lagOtros | -.0002913 .0000889 -3.28 0.001 -.0004657 -.000117
z2 | .7151745 .3648443 1.96 0.050 .0000928 1.430256
pre_mean | -.0137378 .0099912 -1.37 0.169 -.0333201 .0058446
alt | -.002204 .0009387 -2.35 0.019 -.0040438 -.0003643
lnC_m_r_f__o | -.3431311 .2760784 -1.24 0.214 -.8842348 .1979725
lnF_E_A_S | .2523251 .298972 0.84 0.399 -.3336493 .8382994
lnFungicidas | .8170511 .3499768 2.33 0.020 .1311091 1.502993
lnHerbicidas | .1152862 .2551923 0.45 0.651 -.3848815 .6154539
lnI_a_n | -.1575809 .3806501 -0.41 0.679 -.9036414 .5884795
lnMaterial_propagacion | .0577852 .0796335 0.73 0.468 -.0982936 .213864
i_a_2 | .522413 .1882741 2.77 0.006 .1534025 .8914235
i_a_3 | .1684625 .1445746 1.17 0.244 -.1148984 .4518234
i_a_4 | .3020224 .0866089 3.49 0.000 .1322722 .4717726
i_a_6 | -.1024728 .0774617 -1.32 0.186 -.2542951 .0493494
_cons | -.6070229 5.827687 -0.10 0.917 -12.02908 10.81503
----------------------------------------------------------------------------------------
Instruments for differenced equation
GMM-type: L(2/.).pcAS L(1/.).lagprecio L(1/.).lagOtros
Standard: D.z2 D.lnC_m_r_f__o D.lnF_E_A_S D.lnFungicidas D.lnHerbicidas
D.lnI_a_n D.lnMaterial_propagacion D.i_a_2 D.i_a_3 D.i_a_4 D.i_a_5
D.i_a_6
Instruments for level equation
GMM-type: LD.pcAS D.lagprecio D.lagOtros
Standard: _cons

.



But when I try to estimate the same regression with xtabond2, I don't get the same coefficients but these are similar.

This is the code with xtabond2:
xi: xtabond2 L(0/1).pcAS z2 lagprecio lagOtros pre_mean alt lnC_m_r_f__o lnF_E_A_S lnFungicidas lnHerbicidas lnI_a_n lnMaterial_propagacion i_a_2 i_a_3 i_a_4 i_a_6, gmm(l.pcAS lagprecio lagOtros, lag(1 .) ) iv(z2 lnC_m_r_f__o lnF_E_A_S lnFungicidas lnHerbicidas lnI_a_n lnMaterial_propagacion i_a_2 i_a_3 i_a_4 i_a_5 i_a_6, eq(diff) ) twostep robust h(2)

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 = 18624
Time variable : AÑO Number of groups = 5586
Number of instruments = 54 Obs per group: min = 1
Wald chi2(16) = 11004.57 avg = 3.33
Prob > chi2 = 0.000 max = 5
----------------------------------------------------------------------------------------
| Corrected
pcAS | Coefficient std. err. z P>|z| [95% conf. interval]
-----------------------+----------------------------------------------------------------
pcAS |
L1. | .8682682 .0389188 22.31 0.000 .7919888 .9445476
|
z2 | .751538 .3542788 2.12 0.034 .0571644 1.445912
lagprecio | .0002435 .0001042 2.34 0.019 .0000393 .0004478
lagOtros | -.0002373 .0000787 -3.02 0.003 -.0003915 -.0000832
pre_mean | -.0122261 .0092467 -1.32 0.186 -.0303493 .005897
alt | -.0017944 .0008031 -2.23 0.025 -.0033685 -.0002202
lnC_m_r_f__o | -.1566321 .2695845 -0.58 0.561 -.685008 .3717439
lnF_E_A_S | .3404469 .2604232 1.31 0.191 -.1699731 .8508669
lnFungicidas | .8017626 .2731656 2.94 0.003 .2663678 1.337157
lnHerbicidas | -.1259277 .2210927 -0.57 0.569 -.5592614 .3074059
lnI_a_n | -.0983911 .3257134 -0.30 0.763 -.7367777 .5399954
lnMaterial_propagacion | .0772578 .0650679 1.19 0.235 -.050273 .2047886
i_a_2 | .523123 .1491537 3.51 0.000 .2307872 .8154588
i_a_3 | .2013174 .1222973 1.65 0.100 -.038381 .4410157
i_a_4 | .3628124 .0791912 4.58 0.000 .2076005 .5180243
i_a_6 | -.0931929 .0733073 -1.27 0.204 -.2368725 .0504868
_cons | -2.64894 4.559091 -0.58 0.561 -11.58459 6.286714
----------------------------------------------------------------------------------------
Instruments for orthogonal deviations equation
Standard
FOD.(z2 lnC_m_r_f__o lnF_E_A_S lnFungicidas lnHerbicidas lnI_a_n
lnMaterial_propagacion i_a_2 i_a_3 i_a_4 i_a_5 i_a_6)
GMM-type (missing=0, separate instruments for each period unless collapsed)
L(1/5).(L.pcAS lagprecio lagOtros)
Instruments for levels equation
Standard
_cons
GMM-type (missing=0, separate instruments for each period unless collapsed)
D.(L.pcAS lagprecio lagOtros)

Can someone explain to me why these differences happen, I will be very thankful.

[Sebastian Kripfganz]

As far as I can see, the two specifications should be equivalent. You could double check the results with my xtdpdgmm command:

xtdpdgmm L(0/1).pcAS z2 lagprecio lagOtros pre_mean alt lnC_m_r_f__o lnF_E_A_S lnFungicidas lnHerbicidas lnI_a_n lnMaterial_propagacion i_a_*, gmm(L.pcAS lagprecio lagOtros, lag(1 .) m(diff)) gmm(L.pcAS lagprecio lagOtros, lag(0 0) diff m(level)) iv(z2 lnC_m_r_f__o lnF_E_A_S lnFungicidas lnHerbicidas lnI_a_n lnMaterial_propagacion i_a_*, diff m(diff)) twostep vce(robust) wmatrix(ind)

• Kripfganz, S. (2019). Generalized method of moments estimation of linear dynamic panel data models. Proceedings of the 2019 London Stata Conference.