Dear all,
I that panel with 9 years. My dependent variable is highly autocorrelated. I check this by using "ar()" option within the system gmm command. Ar(1)-Ar(7) test are all rejected implying high autocorrelation of dependent variable.
The issue is that no instrument work. At first, AR(2) test and Hansen-Sargan test are not satisfied (they are both rejected). To circumvent the issue with AR(2) test (which reject the Arellano and Bond test of no autocorellation) I have to add l2.Y and even l3.Y, because my dependent variable is so highly autocorrelated. In addition, I used further lag to avoid second order autocorrelation. circumvent this issue using these "techniques". However, Sargan test does not produce good results.
my independent variable are not endogenous, so they do not need to be instrumented.
Has anyone faced this issue?
I that panel with 9 years. My dependent variable is highly autocorrelated. I check this by using "ar()" option within the system gmm command. Ar(1)-Ar(7) test are all rejected implying high autocorrelation of dependent variable.
Code:
xtabond2 Y l.Y x m c l i.year /* */ robust twostep small ar(7) /* */ gmm(L.Y , lag(4 5)) /* */ iv(yr2-yr9 x m c l)
my independent variable are not endogenous, so they do not need to be instrumented.
Has anyone faced this issue?