xtabond2 dynamic panel with binary endogenous variable

[Cooper Felix]

How do I specify the command if I have a binary endogenous variable? Is there any problem for the following specification? Thanks!

xtabond2 f(1/0).dep_var endogenous_dummy, gmmstyle(dep_var) gmmstyle(endogenous_dummy, equation(level)) robust twostep

[Sebastian Kripfganz]

Such a binary variable is not necessarily treated different compared to a continuous variable. Depending on your data, lagged differenced may or may not be weak instruments for lagged levels of the dummy (and vice versa). There is no general rule in that regard.

[Cooper Felix]

Such a binary variable is not necessarily treated different compared to a continuous variable. Depending on your data, lagged differenced may or may not be weak instruments for lagged levels of the dummy (and vice versa). There is no general rule in that regard.

Thanks for your comments. Indeed, as Dr. Wooldridge mentioned over here (https://www.statalist.org/forums/for...ummy-variables), differencing a binary variable shouldn't be treated differently.

Here is another post discussing the same issue: "as long as there is temporal variation in a dummy variable (i.e. a switch from 0 to 1 or 1 to 0 over time), then there is no problem in including it in a first differences (FD) model. He talks nothing about the interpretation of the variable."

[Cooper Felix]

Such a binary variable is not necessarily treated different compared to a continuous variable. Depending on your data, lagged differenced may or may not be weak instruments for lagged levels of the dummy (and vice versa). There is no general rule in that regard.

Sebastian,

Can I apply xtabond2 to a model with a limited dependent variable? If not, can you suggest other commands that might be useful? Thank!

[Sebastian Kripfganz]

You could. It then becomes a linear probability model.