st: IV regress, fixed effects and clustering

[Jeanne Lafortune <jeanne.lafortune@gmail.com>]

Dear all,

Has anybody ever encountered this problem?  I'm running an instrumental 
variable regression:

xi: ivregress 2sls y (x=z) i.countyid, cluster(countyid)

using a data set with a lot of fixed effects (about 3000).  I'm also 
clustering at the same level.  In some cases, STATA encounters a problem 
and does not produce standard errors on the coefficients.  I've tried 
clustering at a different level (using 50 or even 3 clusters) and the 
same result occurs.  On the other hand, if I use no clustering or 
robust, standard errors are produced. 

I've run both the reduced form (reg y z i.countyid, cluster(countyid)) 
and the first stage (reg x z i.countyid, cluster(countyid)) and the 
results appear completely normal and all have standard errors.

Finally, I've also tried small changes to the specification, like adding 
a variable or two and the standards errors in some of these cases reappear.

If anybody has a clue why this may be happening, please let me know.

Best,

Jeanne
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/