# Data audit (2026-10-30)

| file | SHA-256 | rows (excl. header) | cols | range |
|---|---|---|---|---|
| data/cpi_quarterly.csv | c8652dec1567ec35e3b84ee8fa55545e9bf0ae4990a4d6d05b89a6d22f6801ea (matches README & bundle.json) | 127 | 3 (quarter, cpi_avg, infl_annualised) | 1995-01-01 … 2026-07-01 |
| data/cpi_monthly.csv | 855d069176369d30d02cbbe0cc6456b76d2f3967cdaabc1b362c2461ef9015a2 (matches) | 381 | 2 (month, cpi) | 1995-01-01 … 2026-09-01 |

- Units: cpi index 2002=100 NSA; infl_annualised in percent at annual rate (400·Δln).
- Missing: only infl_annualised for 1995Q1 (by construction, no prior quarter). Monthly file: none.
- Duplicates: none in either file. Spacing regular (quarters 89–92 days, months 28–31 days): no gaps.
- Reproduction: quarterly cpi_avg and infl_annualised recomputed from the monthly file agree to ≤ 5e-7 (rounding to 6 decimals). 127 complete quarters.
- Distribution of infl_annualised: mean 2.13, sd 2.39, min −6.04 (2008Q4), max 12.15 (2022Q2). Strong seasonality: mean by quarter Q1 2.32, Q2 4.18, Q3 1.66, Q4 0.30.
- Anomalies (not errors, genuine episodes): 2008Q4 oil-price collapse (−6.0), 2020Q2 COVID drop, 2022H1 inflation surge (8.2, 12.1). Retained; no outlier treatment (would need post-origin judgement).
- Information-availability: no input contains information beyond quarterly values; the rule is enforced in code by truncating the series to quarters ≤ origin before any estimation/selection. Data are a current vintage (CPI essentially unrevised); no real-time claim made.