# Contract — F1-conf-01

Question. 12-month growth of CPIAUCSL (SA) for each month 2001-01..2026-09 (309 months), and summaries. Estimand. g_t = 100*(P_t/P_{t-12} - 1), with P taken from the most recent vintage covering the month: 2026-10-28 vintage for 2014-01..2026-09, 2016-07-15 vintage for months before 2014-01. Lags for 2001-01 reach back to 2000-01 (2016 vintage). Mixing vintages across t and t-12 (e.g. 2014-06 vs 2013-06) is what the rule prescribes; it is meaningful under the assumption that revisions between vintages (seasonal factor updates) are small. Summaries. mean over window; means 2001-01..2013-12 and 2014-01..2026-09; g for 2014-06, 2016-07, 2020-05, 2022-06, 2026-09; max/min and months. Outputs. out/results.json (keys per TASK.md table), out/analytic.csv (date,cpi,cpi_lag12,vintage_used,growth_12m; n_analysed rows), out/exclusions.csv, out/report.md (6 sections), out/analysis.py, out/claims.json. Info rules. Latest vintage wins; no interpolation/imputation; missing g_t counted; only data/ files. Invalidity. Using the hand-assembled file as source if it has errors; imputing missing levels; using older vintage where newer exists.