Endogeneity

Flamed Diving

Hello, does anyone here have much experience with endogenous regressors?

I am running a regression:

[latex]Y = \alpha + \beta{X} + \gamma{W} + \epsilon[/latex]

Where X and Y are mutually causative. In this case, Y is GDP per capita and X is a measure of institutions.

Theoretically, these are meant to endogenous. So, I ran OLS on this regression, calculated the residual and then ran of covariance check between this, X and Y.

As expected, Y is correlated with the residual, but X was not. I tried three alternative measures of X, all from previously published works, but all show no relation to their respective error.

My guess is that I am testing for this incorrectly, right?

Flamed Diving

Ok, just to update this beehive of activity. I ran a Durban-Wu-Hausman test and still could find not reject the hypothesis of consistency in the "endogenised" variables.

I came across the Hausman specification test, but I'm not really sure how to implement it via Stata. I can never quite follow their standard help pages.

from what I recall hausman was pretty trivial to run in stata?

'hausman [var list]' no?

Flamed Diving

Deleted User wrote: »

from what I recall hausman was pretty trivial to run in stata?

'hausman [var list]' no?

No, as far as I can tell, it wants you to "store estimates" and then run the test against two of these stored estimates.

hausman performs Hausman's specification test. To use hausman, perform the following steps.

(1) obtain an estimator that is consistent whether or not the
hypothesis is true;
(2) store the estimation results under name-consistent by using
estimates store;
(3) obtain an estimator that is efficient (and consistent) under the
hypothesis that you are testing, but inconsistent otherwise;
(4) store the estimation results under name-efficient by using
estimates store;
(5) use hausman to perform the test

hausman name-consistent name-efficient [, options]

So, I ran two regressions for the above model. One in OLS and the other in IV, using an instrument.

Oddly enough, both the OLS and IV results were the same, including the SEs. (Solved that). I must admit I have never done this before, bur unfortunately, neither has my supervisor...

Économiste Monétaire

Run first regression

. estimates store FD1

Run second regression

. estimates store FD2

hausman FD2 FD1

That's how I read it. Pretty rusty on this, though.

Flamed Diving

Économiste Monétaire wrote: »

Run first regression

. estimates store FD1

Run second regression

. estimates store FD2

hausman FD2 FD1

That's how I read it. Pretty rusty on this, though.

Yeah, I did that and got:

hausman _est_consistent _est_efficient

estimation result _est_consistent not found
r(111);

I read the help file, and it suggested flipping them the other way round.

hausman _est_efficient _est_consistent

estimation result _est_efficient not found
r(111);

But of course, they are there when I check.

BESman

Have you tried turning Stata off and on again?