step,rule,n_before,n_after,n_removed,rationale
0,rows in last-origin vintage file,314,314,0,raw input (vintage 2025-12-31)
1,form AR(1) pairs (first quarter has no lag),314,313,1,1947Q2 has no lagged value in file
2,keep t >= 1985Q1,313,163,150,post-Great-Moderation estimation sample (a priori)
3,"drop t in {2020Q2,2020Q3,2020Q4} (pandemic outlier as y_t or y_{t-1})",163,160,3,"extreme pandemic outliers (-28%, +35%) would dominate OLS"