How to deal with endogeneity in fixed effect model?

[Lema Kargar]

Hi, I am trying to estimate my model using fixed effect . but before that i did a unit root test and all my variables are stationary only in first difference. I also did a correlation test, which shows that one of my control variable is highly correlated to my explanatory variable, so in order to estimate my model using

areg lngdppercapita dl.lnmigrationinflow dl.lnpopulation dl.lncapitalstock dl.lnhumancapital y*,absorb(countryid)vce(cluster countryid)

I was wondering if i should estimate the model using the first difference of the variables and also lagged values, because i want to deal with reverse causality present in my model.is this correct?

and also what is the solution to the serial correlation of the control with explanatory variable?
Regards

lema

[Phil Bromiley]

You didn't get a quick answer. You'll increase your chances of a useful answer by following the FAQ on asking questions.

With fixed effects, I'd normally go to the panel data tools like xtreg and xtivreg (assuming an endogeneity problem). xtivreg2 (user written) provides additional statistics not automatically available in xtivreg. areg fits fixed effects models, but xtreg is more flexible.

A control being correlated with an explanatory variable is not inherently a problem. It depends on your sample size, etc. If the control is important, then omitting it is quite problematic.

[Lema Kargar]

Thank you very much for your reply!

[Lema Kargar]

Hi Phil, a question again regarding the endogeneity. I want to use the lagged values of my explanatory variables as instruments to deal with endogeneity. is that possible and how can I proceed? can you provide me with the stata code? thanks a lot.