Instrument Validity and Weak IV Concepts

Notes

Lecture 11: Checking Instrument Validity 11-1:Weak IV Dainn Wie wie-dainn@grips.ac.jp A Bivariate IV Model • Checking instrument validity: We need to check if instrument variable employed s atisfy two conditions: relevance and exogeneity conditions. • The validity of your instrument should be well argued in the paper. • In case of relevance condition, showing weak IV test (first-stage F) is essential. • IV should not be simply significant factor affecting X. X and IV should have strong relationship. • Let’s consider simple bivariate model. A Bivariate IV Model (2) A Bivariate IV Model (3) Weak Instrument Problems • Staiger and Stock, Econometrica, 1997 • Dark line: Weak IV • Dashed line: strong instrument. Weak IV Test: Homoskedasticity (1) When we have homoskedasticity Weak IV Test: Homoskedasticity (2) The standard method so far… • 1 endogenous regressor: First-stage F • More than 1 endogenous regressor: Cragg-Donald (1993) F-Statistic • Then compare the F-statistics with • Rule of thumb “10” (Staiger and Stock, 1997) • Stock and Yogo (2005) critical values • Stock and Yogo’s critical value: Find a critical value such that the relative bias of th e 2SLS with respect to the bias o...