1step,rule,n_before,n_after,n_removed,rationale21,"regression pairs (y_t, y_{t-1}) with y_{t-1} dated >= 1985-01-01 (pre-1985 quarters and the first, lag-less row dropped)",314,162,152,post-Great-Moderation estimation sample; avoids high-volatility early postwar regime32,drop pairs in which y_t or y_{t-1} is 2020Q2 or 2020Q3,162,159,3,pandemic collapse/rebound (about -28/+35 % SAAR) would dominate OLS and bias the AR coefficient4