Data audit
| file | sha256 | rows (excl header) | cols |
|---|---|---|---|
| data/firm_panel.csv | a1f1af88b1ab05900155991f2468cce09695af978923e5bc3ef9c28d45683b6f | 2140 | 7 |
| data/README.md | 26573e982ab6ca9e7f90d2b89bd2878aa6a670f90da21496bff787b59216b9a2 | codebook | — |
Schema: firm_id (str, F001–F300, 300 firms), year (int 2016–2022), leverage (ratio, 0.02–0.95), cash_ratio (ratio, 0.0046–0.7406), log_sales (log thousand USD, 1.29–8.88), rating_2025 (A/B/C/D, constant within firm), default_next (0/1, empty=unobserved).
- Missing: only default_next, 307 raw rows (300 unique firm-years), all in origin year 2022 (outcome year 2023 after end-2022 cut). No missing outcome in 2016–2021, none observed for 2022.
- Duplicates: 40 exact duplicate rows (80 rows share a duplicated firm_id-year key); all duplicates identical in every column -> drop copies. After de-dup: 2100 rows = 300 firms × 7 years, balanced.
- Anomalies: leverage point mass at 0.02 (57 unique rows in 2016–2022, 54 in the 2016–2021 estimation sample; floor) — retained, noted. Firms can default in several years (synthetic; retained as recorded).
- Information violation: rating_2025 (known 2025, uses realised default history through 2022 and average leverage). Default rate is 0 for A/B/C and 0.227 for D — clear leakage. Decision: drop column.