A robust test for weak instruments in Stata
Abstract. We introduce a routine, weakivtest, that implements the test for weak
instruments by Montiel Olea and Pflueger (2013, Journal of Business and
Economic Statistics 31: 358–369). weakivtest allows for errors
that are not conditionally homoskedastic and serially uncorrelated. It extends
the Stock and Yogo (2005, Testing for weak instruments in linear IV regression.
In Identification and Inference for Econometric Models: Essays in Honor of
Thomas Rothenberg, ed. D. W. K. Andrews and J. J. Stock, 80–108.
[Cambridge University Press]) weak-instrument tests available in ivreg2
and in the ivregress postestimation command estat firststage.
weakivtest tests the null hypothesis that instruments are weak or that
the estimator’s Nagar (1959, Econometrica 27: 575–595) bias is
large relative to a benchmark for both two-stage least-squares estimation and
limited-information maximum likelihood with one endogenous regressor. The
routine can accommodate Eicker–Huber–White heteroskedasticity
robust estimates, Newey and West (1987, Econometrica 55: 703–708)
heteroskedasticity- and autocorrelation-consistent estimates, and clustered
variance estimates.
View all articles by these authors:
Carolin E. Pflueger, Su Wang
View all articles with these keywords:
weakivtest, F statistic, heteroskedasticity, autocorrelation, clustered, weak instruments, testing
Download citation: BibTeX RIS
Download citation and abstract: BibTeX RIS
|
