Hi all,
I have the following model of 10 years (2010-2020) for banks. I use the lagged dependent variable, and need to make use of two step system GMM.
Dependent variable: nplr
independent variable: EPU
Endogenous controls: size cap aq (bank specific variables)
Exogenous controls: GDP infl (macro-economic variables)
An example of another paper of similar research used the standard second lag of dependent, independent and endogenous controls and instrumenting the exogenous controls. However, I saw another paper using the 2nd to 4th lag of the gmm instruments.
Another paper mentioned the following: "the instruments for the level equation are the lagged differences of the corresponding variables, whereas the instruments for the differenced equation are the lagged levels." Which also quite confuses me.
I am confused when to use eq(level) and eq(diff), in order to reach two-step system gmm. I have been trying many options, but I do not get the right model. My hansen statistics keep on indicating 0.000. Can anybody help getting the right stata code?
xtabond2 nplr l.nplr epu size cap aq gdp infl y11 y12 y13 y14 y15 y16 y17 y18 y19 , gmmstyle(l.nplr epu size cap aq, lag(1 .) eq(diff) collapse) gmmstyle(l.nplr epu size cap aq, lag(1 .) eq(level) collapse) iv(y11 y12 y13 y14 y15 y16 y17 y18 y19 gdp infl, eq(level)) twostep robust
In reading a lot of blogs and papers (e.g. Roodman and slides from Kripfganz) I do not seem to understand when and how to use the level and difference equations and also struggle with determining the lags. Do I need to specify the GMM instruments as both level and differenced? Or only level?
Any tips would be more than welcome!
Thanks a lot in advance
I have the following model of 10 years (2010-2020) for banks. I use the lagged dependent variable, and need to make use of two step system GMM.
Dependent variable: nplr
independent variable: EPU
Endogenous controls: size cap aq (bank specific variables)
Exogenous controls: GDP infl (macro-economic variables)
An example of another paper of similar research used the standard second lag of dependent, independent and endogenous controls and instrumenting the exogenous controls. However, I saw another paper using the 2nd to 4th lag of the gmm instruments.
Another paper mentioned the following: "the instruments for the level equation are the lagged differences of the corresponding variables, whereas the instruments for the differenced equation are the lagged levels." Which also quite confuses me.
I am confused when to use eq(level) and eq(diff), in order to reach two-step system gmm. I have been trying many options, but I do not get the right model. My hansen statistics keep on indicating 0.000. Can anybody help getting the right stata code?
xtabond2 nplr l.nplr epu size cap aq gdp infl y11 y12 y13 y14 y15 y16 y17 y18 y19 , gmmstyle(l.nplr epu size cap aq, lag(1 .) eq(diff) collapse) gmmstyle(l.nplr epu size cap aq, lag(1 .) eq(level) collapse) iv(y11 y12 y13 y14 y15 y16 y17 y18 y19 gdp infl, eq(level)) twostep robust
In reading a lot of blogs and papers (e.g. Roodman and slides from Kripfganz) I do not seem to understand when and how to use the level and difference equations and also struggle with determining the lags. Do I need to specify the GMM instruments as both level and differenced? Or only level?
Any tips would be more than welcome!
Thanks a lot in advance
