1step,rule,n_before,n_after,n_removed,rationale21,rows read from firm_panel.csv,2140,2140,0,starting point32,"drop exact duplicate rows (same firm_id-year, identical values)",2140,2100,40,assembly error: rows repeated in extract; one copy kept43,exclude column rating_2025 (known only in 2025),2100,2100,0,final-vintage field built from realised defaults through 2022 and sample-average leverage: look-ahead / outcome leakage54,drop firm-years with unobserved default_next,2100,1800,300,outcome year t+1 not observed at data cut (end 2022); missing outcomes all in year(s) [2022]; not imputed65,drop firm-years with missing predictors,1800,1800,0,complete-case requirement for OLS7