Recently, Keane and Neal (2023, 2024) showed that 2SLS t-tests suffer from a “power asymmetry”: 2SLS standard errors are spuriously small (large) when the 2SLS estimate is close to (far from) the OLS ...