Xtabond2 and dummy variables‏

[Sebastian Wang]

Hi,
I have to estimate a dynamic model, with individual heterogeneity, and I'm therefore using the system GMM estimator with -Xtabond2-.
Simplified the model looks like:
y= l.y + l.x + l.z + l.p + i.year

My sample can be split into two different groups, so what I would like to do, is to use a dummy variable so I get estimates for each group that are easily comparable. My problem is then, how I should correctly specify the interaction term in -Xtabond2-. I tried estimating it as:

y (l.c.y l.c.x l.c.z l.c.p i.year)#dummy, gmm(l.y, lag(1 5) collapse) gmm(l.c.x l.c.z l.c.p, lag(2 5) collapse) iv(i.year dummy) twostep robust 

This is with the assumption that l.y is predetermined, the other lagged variables are possibly endogenous and the time dummies and the group dummy are exogenous. I'm not sure if it is correct to write the variables individually or if I should rather specify the whole terms including the dummy? My results seem to be very sensitive to which specification I use. I looked in Roodman's paper on how to use the -xtabond2-, but unfortunately that didn't give me an answer either.

I'm very grateful for any help.

Best,
Sebastian Wang

[Sebastian Wang]

Nobody that has an idea on this? I'm glad to elaborate if anything is unclear.

[swati saini]

Hi..I am facing a similar problem.


I am running a regression of income inequality on iprs and other control variables.Since I have a dynamic panel specification, I am using xtabond2 command.My regression eq looks like:

Income Ineq = b1 +b2. IPRs +b3.IPRs.TRIPS +b4.TRIPS.

TRIPS is a dummy variable that takes up the value of 1 for years after 1995 and 0 for years before 1995. ( I am testing for a structural break in 1995 when WTO came into existence.)
IPRs is intellectual property rights index that takes up a value ranging between 1 to 5.

While estimating this eq, I consider IPRs as an endog. variable which is instrumented by its lags while using xtabond2 command.My problem is how should I instrument IPRs.TRIPS ?

Suppose my data looks like:

country	year	gini_net	ipr	TRIPS	IPR.trips
Australia	1	28.1872	2.49	0	0
Australia	2	29.0448	2.49	0	0
Australia	3	28.7599	3.28	0	0
Australia	4	29.9817	4.33	1	4.33
Australia	5	31.1922	4.33	1	4.33
Australia	6	31.7795	4.33	1	4.33

Can i instrument IPRs.TRIPS using its lags ?

Suppose if I use one period lag as instruments, this means my observation for 4th year : IPR.TRIPs = 4.33 will be instrumented by 0 .( levels are instrumented by differnces of lags in system-gmm) which seems illogical. How should I proceed? Any idea about this?

[Cooper Felix]

Hi..I am facing a similar problem.


I am running a regression of income inequality on iprs and other control variables.Since I have a dynamic panel specification, I am using xtabond2 command.My regression eq looks like:

Income Ineq = b1 +b2. IPRs +b3.IPRs.TRIPS +b4.TRIPS.

TRIPS is a dummy variable that takes up the value of 1 for years after 1995 and 0 for years before 1995. ( I am testing for a structural break in 1995 when WTO came into existence.)
IPRs is intellectual property rights index that takes up a value ranging between 1 to 5.

While estimating this eq, I consider IPRs as an endog. variable which is instrumented by its lags while using xtabond2 command.My problem is how should I instrument IPRs.TRIPS ?

Suppose my data looks like:

country	year	gini_net	ipr	TRIPS	IPR.trips
Australia	1	28.1872	2.49	0	0
Australia	2	29.0448	2.49	0	0
Australia	3	28.7599	3.28	0	0
Australia	4	29.9817	4.33	1	4.33
Australia	5	31.1922	4.33	1	4.33
Australia	6	31.7795	4.33	1	4.33

Can i instrument IPRs.TRIPS using its lags ?

Suppose if I use one period lag as instruments, this means my observation for 4th year : IPR.TRIPs = 4.33 will be instrumented by 0 .( levels are instrumented by differnces of lags in system-gmm) which seems illogical. How should I proceed? Any idea about this?

Have you figured out the answer?

[Sebastian Kripfganz]

Answered here: xtabond2 dynamic panel with binary endogenous variable