inconsistency of iv at first stage regression with and without county FE

[Tariq Abdullah]

I'm working with instrumental variables and following the published paper of Dr. Card [Card, David. 1995. “Aspects of Labour Economics: Essays in Honour of John Vanderkamp.”] , it shows how can a college presence at a county level can be instrumented for the education level of students present in that county. Dr. card's coding example is like this one.

use https://github.com/scunning1975/mixtape/raw/master/card.dta, clear

* OLS estimate of schooling (educ) on log wages

reg lwage educ exper black south married smsa

* 2SLS estimate of schooling (educ) on log wages using "college in the county" as an instrument for schooling

ivregress 2sls lwage (educ=nearc4) exper black south married smsa, first

* First stage regression of schooling (educ) on all covariates and the college and the county variable

reg educ nearc4 exper black south married smsa

* F test on the excludability of college in the county from the first stage regression.

test nearc4

In my research setup, I'm also trying to instrument a treatment at the county level. Following Dr. Card's example ( College in the county with Dr. David Card example data) I tried to use the following equation to test my strength in the first stage:

reg endogneous_var male ismarried wasmarried age age2 black asian hispanic lths hsdegree somecollege i.county instrument_var , cluster(county)

* F test on the excludability of IV in the county from the first stage regression.

test instrument_var

This gives my F -test to be above 17. So, this shows my first stage is significant and my instrument is strong.

However, when I use the following command , I get my instrumental variable insignificant in first stage regression as it gives me a p-value of .160.

ivregress 2sls dep_var (endogneous_var=instrument_var) male ismarried wasmarried age age2 black asian hispanic lths hsdegree somecollege i. county , cluster(county) first

Now, when I use the following equation without the county fixed effects , then I again get my instrumental variable significant at first stage with p-value 0.000.

ivregress 2sls dep_var (endogneous_var=instrument_var) male ismarried wasmarried age age2 black asian hispanic lths hsdegree somecollege , cluster(county) firs

In the equation of Dr. Card example he didn't use the county fixed effects. However, his data is at individual level and instrument at county level. In my case, the data and instrument both are at county level. So, should I use the ivregress command without the county fixed effects ? Or Am I doing something wrong.

In my regression like the following command when I cluster at county and use state fixed effects like the following the IV is significant by giving p-value 0.000

ivregress 2sls dep_var (endogneous_var=instrument_var) male ismarried wasmarried age age2 black asian hispanic lths hsdegree somecollege i.statefip , cluster(county) first

Should I use without county fixed effects while clustering cluster at county level when I'm using ivregress command or should I use state fixed effects while clustering at county level?

Or is this the indication that my instrument is not significant ?

[Tariq Abdullah]

There was a mistake with my coding, and therefore I ran into the issue. I cautiously looked back into my coding. After a couple of hours, I figure out what was wrong. Now, it exactly gives me the value I was getting with separate regress command.