WEAKIV10: Stata module to perform weak-instrument-robust tests and confidence intervals for instrumental-variable (IV) estimation of linear, probit and tobit models

Abstract

weakiv calculates weak-instrument-robust tests of the coefficients on the endogenous regressors in instrumental variables (IV) estimation of models with any number of endogenous regressors. weakiv supports estimation of linear IV models by ivregress, ivreg2 and ivreg2h, panel data linear IV estimation (fixed effects and first differences) by xtivreg and xtivreg2, dynamic panel data estimation by xtabond2, and estimation of probit and tobit IV models by ivprobit and ivtobit. In an exactly-identified model with one instrument, weakiv reports the Anderson-Rubin (AR) test statistic. When the IV model is overidentified, weakiv also conducts the conditional likelihood ratio (CLR) test, the Lagrange multiplier K test, the J overidentification test, and a combination of the K and overidentification tests (the K-J test). Minimum Distance/Wald and LM versions of the tests are both supported. weakiv also provides graphing facilities for visual examination and presentation of rejection probabilities and confidence sets based on these tests. For individual endogenous regressors, weakiv will report confidence intervals and graph rejection probabilities that are robust to weak identification. For 2 endogenous regressors, weakiv constructs confidence sets and reports these using 2-D contour plots and 3-D surface plots. For linear IV and panel data model models, weakiv supports all the variance-covariance estimators supported by ivregress, ivreg2 and xtabond2 (robust, cluster-robust, HAC, 2-way clustering, Kiefer and Driscoll-Kraay SEs, etc.). weakiv builds on and extends the command rivtest by Finlay and Magnusson (Stata Journal 9(3), 2009). weakiv requires Stata 11 or above. Users with Stata 10 are recommended to install weakiv10, an older version of weakiv suitable for Stata 10 without many of the extensions available in weakiv.