testing the validy of the instrument in xtivreg

[Danielle leblanc]

Hi,

I am using xtivreg , where xtivreg y (x=z)


The issue is that how can i tell that the instrument z is a valid instrument? I don't see an indicator from the generated results that tests the instrument.

[Joro Kolev]

With only one instrument you cannot test overidentifying restrictions.

With the option -first- you can check the strength of your instrument.

[Jeff Wooldridge]

If the instrument varies over time, you can include, say, the lead value in the equation. xtivreg (preferably with the FE option) assumes instruments are strictly exogenous, and this can be test.

xtset id year
xtivregress y (x = z) w1 ... wK, F.z i.year, fe vce(cluster id)

Use the t statistical on the variable F.z. Under the null of strict exogeneity, it should be insignificant.