Annual Review of Economics · 2024 · Michael P. Keane、Timothy Neal
中文摘要
我们综述弱工具变量（IV）文献，旨在为应用研究者提供简明建议。这些文献主要关注工具变量较弱时，两阶段最小二乘（2SLS）双侧t检验的实际检验水平膨胀问题。一个常见的工具变量强度合格标准是第一阶段F统计量为10，在这一水平下，实际检验水平的膨胀幅度较小。然而，2SLS还存在其他重要问题，即使工具变量强度高得多，这些问题依然存在。特别是，在2SLS估计值接近普通最小二乘（OLS）估计值的样本中，2SLS标准误往往会失真地偏小。这种检验功效不对称意味着，当OLS偏误为正时，t检验检出虚假正向效应的功效会被夸大。Anderson-Rubin（AR）检验避免了这一问题，因此，即使工具变量较强，也应使用AR检验替代t检验。我们以2011年至2023年发表于《美国经济评论》的IV论文为例，说明这一问题在实践中的重要性。使用AR检验常常会逆转t检验的结果。特别是，那些接近OLS估计值且在t检验下统计显著的IV估计值，在AR检验下往往不具有统计显著性。我们还表明，当第一阶段F统计量处于10—20区间时，OLS估计值比2SLS估计值更接近真值的概率很高。因此，我们主张在应用研究中采用更高的工具变量强度标准。
Abstract
We survey the weak instrumental variables (IV) literature with the aim of giving simple advice to applied researchers. This literature focuses heavily on the problem of size inflation in two-stage least squares (2SLS) two-tailed t -tests that arises if instruments are weak. A common standard for acceptable instrument strength is a first-stage F of 10, which renders this size inflation modest. However, 2SLS suffers from other important problems that exist at much higher levels of instrument strength. In particular, 2SLS standard errors tend to be artificially small in samples where the 2SLS estimate is close to ordinary least squares (OLS). This power asymmetry means the t -test has inflated power to detect false positive effects when the OLS bias is positive. The Anderson-Rubin (AR) test avoids this problem and should be used in lieu of the t -test even with strong instruments. We illustrate the practical importance of this issue in IV papers published in the American Economic Review from 2011 to 2023. Use of the AR test often reverses t -test results. In particular, IV estimates that are close to OLS and significant according to the t -test are often insignificant according to AR. We also show that for first-stage F in the 10–20 range there is a high probability that OLS estimates will be closer to the truth than 2SLS. Hence we advocate a higher standard of instrument strength in applied work.