On Tue, 8 Mar 2016, stefania innocenti wrote:

Hello,
I am trying to use the ivregress function to estimate a second stage model which looks like the following:

LogGDP=GEO+RULE+OPENNESS

I would like to instrument Rule (RULE) with Settler mortality (SETTLER) and Openness (OPENNESS) with logFrankelRomer (FR). I thus have one instrument per each endogenous variable.

Thus the second stage look like:
RULE=GEO+logFR+SETTLER
OPENNESS=GEO+logFR+SETTLER

Can I do this simultaneously with the ivregress function?

Which package is this?

Maybe I got something wrong with the syntax but if I do:
ivregress(logGDP ~RULE+GEO+OPENNESS, c(RULE~SETTLER+FR+GEO, OPENNESS~SETTLER+FR+GEO), mydata), the function bugs and I cannot proceed.

Of course, I could run the two first stages separately, but the coefficient of the first stage are then wrong.

The command ivreg(logGDP ~ RULE + GEO +OPEN| GEO + SETTLER+FR) gives me correct second stage coefficients (I checked with what I obtained doing the 2sls by hand).

That's package "AER", I presume?

But how do I retrieve the first stage coefficients for RULE=GEO+logFR+SETTLER and OPENNESS=GEO+logFR+SETTLER in this case?

The ivreg() function from "AER" does not store the first stage results but just returns the IV estimates from the second stage. If you want the first stage you need to run lm() manually:

lm(cbind(RULE, OPENNESS) ~ GEO + logFR + SETTLER, ...)

Any advice is more than appreciated.
Thanks a lot in advance.
Best regards,
Stefania

______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.


______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

Reply via email to