Xtabond2 for system GMM. Please help me for coding

[Khati Zolfaghari]

Hello to everyone, I am searching in terms of the effect of the uncertainty on the saving and in one part of my robustness check I want to do a Xtabond2 . I read the construction of doing xtabond2 from David Roodman . but I become confused. I have a panel data from 1996 to 2017. and :

saving (i,t)= b0+b1saving (i,t-1)+b2 uncertainty (i,t-1)+ b3 X(i,t-1) +vt + vi+ e(i,t)

and x(i,t-1) is a vector of controls, which in the baseline model includes only human capital and per capita income and economic growth.

in the Xtabond2 I want to address a solution to the possible endogeneity problem between economic uncertainty and the saving by instrumenting them with suitable lagged variables. To obtain efficient findings in the System GMM estimations, I need evidence for the validity of the first-order autocorrelation in the residuals, but second-order autocorrelation must be rejected. Then We run the Sargan test to avoid possible over-identification problems.

I did in this way, based on the things that I read, but I need your assistance:

 xtabond2 saving L.saving WUI  ln_gdpper HC economicgrowth_ i.Time i.ifscode , gmm(L.saving ,lag(1 2)) gmm( WUI     ,lag(1 1)) iv( WUI i.Time ) small two

But I get error, however I am not an expert in this field. I hope I receive your assistance .

Thank you very much in advance.

Regards,

[Sebastian Kripfganz]

Without seeing the precise error message, there is not much we can do to help. Could it be that the underscore at the end of the variable name economicgrowth_ should not be there?

[Khati Zolfaghari]

@Sebastian Kripfganz thank you so much for your reply. this is my error. However I do not know this is the correct way I do in the command?

.  xtabond2 saving L.saving WUI  ln_gdpper HC economicgrowth_ i.Time i.ifscode ,
>  gmm(L.saving ,lag(1 2)) gmm( WUI     ,lag(1 1)) iv( WUI i.Time ) small two
Favoring space over speed. To switch, type or click on mata: mata set matafavor 
> speed, perm.
1996b.Time dropped due to collinearity
2015.Time dropped due to collinearity
111b.ifscode dropped due to collinearity
466.ifscode dropped due to collinearity
474.ifscode dropped due to collinearity
626.ifscode dropped due to collinearity
668.ifscode dropped due to collinearity
676.ifscode dropped due to collinearity
Equation not identified. Regessors outnumber instruments.
r(481);

Thank you so much .

regards,

[Sebastian Kripfganz]

As far as I can tell, you did not specify any instruments for the variables ln_gdpper HC economicgrowth_ and for the dummy variables i.ifscode. So, as the error message states, you have not enough instruments for the large number of regressors? Are you sure you need to include i.ifscode? The command automatically accounts for panel-specific fixed effects.

[Khati Zolfaghari]

@ Sebastian Kripfganz thank you so much for your reply. they are my control variables. so I should also add any instrument for them ?

regards,

[Sebastian Kripfganz]

they are my control variables. so I should also add any instrument for them ?

Yes. I recommend to have a look at Roodman's documentation again about how to specify instruments for strictly exogenous variables.

Alternatively, the following presentation might be useful:

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

[Khati Zolfaghari]

@ Sebastian Kripfganz thank you so much for your reply and for sharing this link.

regards,