st: Dummies for Instruments using "IVQREG"

[Jim Bang <bangjamest@sau.edu>]

Statalist,
I am estimating expenditures for a model with potentially endogenous
regressors at various quantiles using IVQREG.
I seem to have trouble obtaining standard error estimates for my model
because I am including dummies for region and occupation, and I am
using dummy variables for instruments in my first stage estimates.
The initial 2SLS results compute fine. In the grid search, I get the
following warning:

Warning:  variance matrix is nonsymmetric or highly singular

The routine does not report standard errors; only coefficients. Does
anyone know of a possible bug in the program that might make the SE's
unstable when there are dummy variables in the instrument list and/or
controls. I get a similar error when I am estimating the model with
non-dummy instruments, if there is a substantial number of dummy
variables in the list of exogenous controls.
Best,
Jim
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/