st: Rivtest p-values

[Jessica Ellen Leight <jeleight@mit.edu>]

Hi,

I am using rivtest to do weak instrument-robust inference following the use of
the command ivreg2 with clustered standard errors.  I have found that for some
equations, the p-values for the Anderson-Rubin test generated by rivtest are
actually lower than those reported in ivreg2, such that a previously
insignificant result becomes insignificant (and this is reflected in the
confidence intervals as well.)   I was puzzled by this as I expected the
opposite.  Is there any general rule about comparing p-values for conventional
IV inference and those for weak-instrument robust inference, or something I
should infer by virtue of the fact that the p-values drop?

Best,
Jessica Leight
*
*   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/