why xtabond2 omitted several coefficient estimates?

[Fang Dong]

Hi,
I am wondering why xtabond2 omitted several coefficient estimates and reported "0" for those coefficients.
Any suggestions are sincerely appreciated.
Thank you.

[Fang Dong]

Does anyone can help me? If you may need anything else, please just let me know. Thank you very much.

[Sebastian Kripfganz]

Without seeing the command line you used and the corresponding estimation output, it is hardly possible to give a meaningful answer. In general, omitted coefficients are not identified by the specified model. This could be because the respective regressors are time invariant and no instruments are specified for the level model, or simply because of multicollinearity. Often this situation arises when you specify a full set of (time) dummy variables.

[Fang Dong]

Hi Prof. Kripfganz,
Thank you very much for replying my question with suggestions. I fully agree with you for your comments. Please see below my commands. Essentially, I am using the dynamic panel data model to study the macroeconomic trilemma. I created purposely several interaction terms and I doubled there may be problems (such as multicollinearity etc.) there besides some other potential problems. But the regular fixed and random effects models can estimate those interaction terms' coefficients without any problem. I look forward to hearing from you again for your further input and guidance. Very deeply appreciated.

* Macro\Monthly

cd "F:\FALL 2021\Research\Macro"
log using "Monthly data output with Japan i 02282022 1.log", replace

//use "F:\FALL 2021\Research\Macro\Monthly data-54.dta", clear
use "F:\FALL 2021\Research\Macro\Monthly data 02282022 balanced upto 2019.dta", clear

xtset Country_New monthly, monthly

//rename CBPRate i

xtfisher i, lags(0)
xtfisher i, lags(1)

gen c_i=d.i
gen c_USi=d.USCBPRate

save "Monthly data 02222022 new with Japan.dta", replace

xi i.year

xtreg d.i l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, fe
outreg2 using "results 02282022.xls", replace ctitle(fixed effects-1)
estimate store fe1

xtreg d.i l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, re
outreg2 using "results 02282022.xls", append ctitle(random effects-1)
estimate store re1

hausman fe1 re1

xtreg d.i l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, fe
outreg2 using "results 02282022.xls", append ctitle(fixed effects-2)
estimate store fe2

xtreg d.i l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, re
outreg2 using "results 02282022.xls", append ctitle(random effects-2)
estimate store re2

hausman fe2 re2

xtreg d.i l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, fe
outreg2 using "results 02282022.xls", append ctitle(fixed effects-2)
estimate store fe3

xtreg d.i l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, re
outreg2 using "results 02282022.xls", append ctitle(random effects-2)
estimate store re3

hausman fe3 re3

xtreg d.i l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, fe
outreg2 using "results 02282022.xls", append ctitle(fixed effects-2)
estimate store fe4

xtreg d.i l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, re
outreg2 using "results 02282022.xls", append ctitle(random effects-2)
estimate store re4

hausman fe4 re4



//Two-step GMM Estimator
xtabond2 d.i ld.i l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, gmm(ld.i, lag(2 .) collapse) iv(l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, eq(level)) robust twostep artests(2) h(2)
outreg2 using "results 02282022.xls", append ctitle(Dynamic-1)
//testparm _I*

//Two-step GMM Estimator, orthog
xtabond2 d.i ld.i l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, gmm(ld.i, lag(2 .) collapse) iv(l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, eq(level)) robust twostep artests(2) orthog h(2)
outreg2 using "results 02282022.xls", append ctitle(Dynamic-1 orthog)
//testparm _I*

//Two-step GMM Estimator
xtabond2 d.i ld.i l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, gmm(ld.i, lag(2 .) collapse) iv(l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, eq(level)) robust twostep artests(2) h(2)
outreg2 using "results 02282022.xls", append ctitle(Dynamic-2)
//testparm _I*

//Two-step GMM Estimator, orthog
xtabond2 d.i ld.i l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, gmm(ld.i, lag(2 .) collapse) iv(l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#cd.USCBPRate VIXCLS_PC1, eq(level)) robust twostep artests(2) orthog h(2)
outreg2 using "results 02282022.xls", append ctitle(Dynamic-2 orthog)
//testparm _I*

//Two-step GMM Estimator
xtabond2 d.i ld.i l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, gmm(ld.i, lag(2 .) collapse) iv(l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, eq(level)) robust twostep artests(2) h(2)
outreg2 using "results 02282022.xls", append ctitle(Dynamic-3)
//testparm _I*

//Two-step GMM Estimator, orthog
xtabond2 d.i ld.i l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, gmm(ld.i, lag(2 .) collapse) iv(l.i economicgrowthforecast inflationforecast i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, eq(level)) robust twostep artests(2) orthog h(2)
outreg2 using "results 02282022.xls", append ctitle(Dynamic-3 orthog)
//testparm _I*

//Two-step GMM Estimator
xtabond2 d.i ld.i l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, gmm(ld.i, lag(2 .) collapse) iv(l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, eq(level)) robust twostep artests(2) h(2)
outreg2 using "results 02282022.xls", append ctitle(Dynamic-4)
//testparm _I*

//Two-step GMM Estimator, orthog
xtabond2 d.i ld.i l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, gmm(ld.i, lag(2 .) collapse) iv(l.i economicgrowthforecast inflationforecast i.crisis#i.fixed#i.nocc#c.grow_FRSGDP#cd.USCBPRate VIXCLS_PC1, eq(level)) robust twostep artests(2) orthog h(2)
outreg2 using "results 02282022.xls", append ctitle(Dynamic-4 orthog)
//testparm _I*

save "Monthly data 02222022 new with Japan.dta", replace

log off
log close

[Sebastian Kripfganz]

I cannot see any immediate reason why some coefficients might be omitted. I would suggest that you start with a simplified model and then step by step add more complexity. This might show you where a problem occurs. With my xtdpdgmm command (an alternative to xtabond2), you could start by replicating the FE results - see slides 88 and 89 of my 2019 London Stata Conference presentation - as the maintained model, and then amend the instruments to your liking.

[Fang Dong]

Thank you very much Prof. Kripfganz. I will follow your suggestions to try and let you know if I have any further questions. Thank you again.