Valid <i>t</i>-ratio Inference for IV

In the single-IV model, researchers commonly rely on t-ratio-based inference, even though literature has quantified its potentially severe large-sample distortions. Building Stock and Yogo (2005), we introduce tF critical value function, leading to a standard error adjustment that is smooth function of first-stage F-statistic. For one-quarter specifications in 61 AER papers, corrected errors are at least 49 136 percent larger than conventional 2SLS 5 1 significance levels, respectively. confidence intervals have shorter expected length those Anderson Rubin (1949), whenever both bounded. (JEL C13, C26)