F statistics for excluded instruments after re ec2sls

[Maria Kravtsova]

Dear Statalisters,

I am running Baltagi's ec2sls random effects model. xtivreg investment (favoritism = elf) hdi education_expenditures population_growth year2009 year2011 year2012 year2013 year2016 y
> ear2008 year2010 year2014 year2015, first re ec2sls. My instrumental variable is time invariant. I would like to test if a have a weak instrument problem. I run xtoverid, noi and I get the following message:

Warning - endogenous variable(s) collinear with instruments
Vars now exogenous: __000015 __00001B __00001E __00001H __00001K
Warning - collinearities detected
Vars dropped: __000019 __00001G __00000W __00000Z __000012 __000018

If I don't include dummy variables I don't get this message. I checked if my year dummies might be overlapping but it was not the case. Could you please give me an advise what could be the problem and how to calculate F-statistics for excluded instruments.

[Phil Bromiley]

You didn't get a quick response. You'll increase your chances of a useful response by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

It is extremely difficult to diagnose colinearity problems without being able to see the data. I also notice that you have a very strange order for your year variables - you'll find it easier if you have a normal sequence for years. I don't know if Hausman works with xtivreg but it is worth checking. Alternatively, you might be able to use a Mundlak approach to test whether fe and re are the same. Perhaps before running the test, I'd check visually if they look close.