st: ivreg2

[Kit Baum <baum@bc.edu>]

Sandra asks

 I am using instrumental variables and have a weak instrument,  
according
to the F-test.  I only have one endogenous variable and my sample is
large.  I wonder if there are better statistics to investigate if an
instrument is weak (perhaps the Hahn and Hausman test testing the null
hypothesis that the instrument is strong), and if stata supports these
statistics.  I also would like to ask if stata contains any routine that
would provide estimates robust to weak instruments.


Diagnosis of instruments' weakness is complicated when there are  
multiple endogenous regressors. When there is only one, it is simple:  
if the single first-stage regression fits poorly, you have weak  
instruments, and the IV estimates will suffer. We have recently added  
diagnostics by Stock and Yogo for the weak instruments case. In the  
presence of weak instruments, the bias of the IV estimates cause them  
to approach the (inconsistent) OLS estimates.

There is no such thing as 'estimates robust to weak instruments'. You  
are using instruments to generate consistent estimate of the  
equation, and if the instruments are only weakly correlated to the  
endogenous variable, your results will be poor. The way to render  
them robust is to get better instruments.

Kit

Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/