INSTRUMENTAL VARIABLES REGRESSION WITH WEAK INSTRUMENTS

摘要: This paper develops asymptotic distribution theory for instrumental variable regression when the partial correlation between instruments and a single included endogenous is weak, here modeled as local to zero. Asymptotic representations are provided various statistics, including two-stage least squares (TSLS) limited information maximum- likelihood (LIML) estimators their t-statistics. The distributions found provide good approximations sampling with just 20 observations per instrument. Even in large samples, TSLS can be badly biased, but LIML is, many cases, approximately median unbiased. suggests concrete quantitative guidelines applied work. These help interpret Angrist Krueger's (1991) estimates of returns education: whereas approach OLS estimate 6%, more reliable fewer fall 8% 10%, typical confidence interval (6%, 14%).