1step,rule,n_before,n_after,n_removed,rationale2window,keep quarters >= 1985-01-01,314,163,151,Great-Moderation-era estimation window3lag,first quarter of window used only as lag,163,162,1,AR(1) loses one observation4covid,drop dependent quarters 2020Q2-2020Q4,162,159,3,COVID outliers (-28/+35 % SAAR) would dominate OLS5