# Verification checks (P4)

• Clean re-run of out/analysis.py (deleted results/analytic/exclusions first): results.json byte-identical — PASS.
• analytic.csv rows == n_analysed (1800) — PASS; columns firm_id, year, leverage, cash_ratio, log_sales, default_next; no rating_2025 — PASS.
• Exclusion ledger: n_before − n_removed = n_after on every row; chained; starts at 2140 (n_input_rows), ends at 1800 — PASS. 2140 − 40 − 300 = 1800.
• No duplicate firm-year keys in analytic sample — PASS. Max origin year 2021 (outcome 2022 ≤ data cut) — PASS.
• se_leverage_cluster > 0 (0.0408) — PASS; 95% CI [0.184, 0.344] contains beta 0.264 — PASS; R² in [0,1] — PASS.
• default_rate (baseline, 0.121) equals mean of default_next in analytic.csv — PASS. Magnitude: beta 0.26 per unit leverage; leverage IQR ~0.31 → ~8 pp change, plausible vs 12% base rate.
• Cluster SE (0.0408) > classical SE (0.0358): expected with within-firm correlation.