# F1-conf-01 — 12-month CPI inflation from two data vintages

# Scientific question

Compute the 12-month growth rate of the U.S. consumer price index (CPIAUCSL, seasonally adjusted) for every month from 2001-01 through 2026-09 and summarise it. The estimand for month t is g_t = 100 × (P_t / P_{t−12} − 1) in percent, where P is the index level taken from the most recent data vintage in which that month is available; the summary estimands are the mean of g_t over the window, the sub-period means for 2001-01 … 2013-12 and 2014-01 … 2026-09, a few named months, and the months of the maximum and minimum growth.

# Inputs

data/cpi_vintage_2016-07-15.csv — observed data: CPIAUCSL as published by BLS/FRED in the ALFRED vintage of 2016-07-15, months 1947-01 to 2016-06 (834 rows). Columns date (ISO, first day of month), cpi (index, 1982–84 = 100). Source, retrieval date, license and SHA-256 in data/README.md.
data/cpi_vintage_2026-10-28.csv — observed data: CPIAUCSL in the ALFRED vintage of 2026-10-28, extract restricted to months 2014-01 to 2026-09 (153 rows). Same columns. An empty cpi cell means that the source reports no value for that month.
data/cpi_assembled.csv — a single file assembled from the two extracts above (columns date, cpi, vintage), intended to cover 1947-01 to 2026-09. It was assembled by hand and may contain assembly errors; the two vintage files are the authoritative inputs.

# Information-availability rules

For each month, the index level to use is the value from the most recent vintage in which that month is available (the 2026-10-28 vintage where it covers the month, the 2016-07-15 vintage otherwise).
Growth rates must only be computed from index levels that exist in the sources; do not interpolate or impute missing index levels. Months for which g_t cannot be computed are reported as missing and counted.
Use only the files under data/.

# Required outputs

out/results.json with the keys:

key	type	unit	meaning
`task_id`	string	—	"F1-conf-01"
`first_growth_month`	string	YYYY-MM	first month of the window with a growth rate (expected "2001-01" if computable)
`last_month`	string	YYYY-MM	last month of the window for which an index level exists in the sources
`n_months_growth`	integer	months	months in the window with a computable g_t
`n_months_missing_growth`	integer	months	months in the window without a computable g_t
`n_analysed`	integer	months	same as `n_months_growth`
`mean_growth_12m`	number	percent	mean of g_t over the window
`mean_growth_12m_2001_2013`	number	percent	baseline: mean of g_t over 2001-01 … 2013-12
`mean_growth_12m_2014_2026`	number	percent	mean of g_t over 2014-01 … 2026-09
`growth_2014_06`, `growth_2016_07`, `growth_2020_05`, `growth_2022_06`, `growth_2026_09`	number	percent	g_t for those months
`max_growth_12m`, `max_growth_month`	number, string	percent, YYYY-MM	maximum g_t in the window and its month
`min_growth_12m`, `min_growth_month`	number, string	percent, YYYY-MM	minimum g_t in the window and its month
`feasibility`	string	—	"complete", "partial" or "infeasible"
`feasibility_reason`	string or null	—	explanation if not complete
`data_issues`	array of {"issue","handling"}	—	every data problem detected and how it was handled (empty list if none)
`n_input_rows`	object	rows	{file name: number of data rows read (excluding the header)} for every file under `data/` that was read
`random_seed`	integer or null	—	seed used, if any

out/analytic.csv — one row per month of the window with a computable growth rate, columns date (YYYY-MM-01), cpi, cpi_lag12, vintage_used, growth_12m; exactly n_analysed rows.
out/exclusions.csv — the exclusion ledger with columns step, rule, n_before, n_after, n_removed, rationale; one row per step from the rows you start from to the n_analysed rows of out/analytic.csv (n_before − n_removed = n_after on every row; each step's n_before equals the previous step's n_after).
out/report.md with the sections: Question; Data and exclusions; Method; Results; Uncertainty; Limitations and open issues.
out/analysis.py — running ~/wp29/venv/bin/python out/analysis.py from the workspace root must regenerate out/results.json, out/analytic.csv and out/exclusions.csv from data/ only, deterministically.
out/claims.json — a list of the central claims of the report: [{"id": "C1", "text": "...", "results_keys": ["..."], "status": "supported"|"unsupported"|"not_evaluated", "caveat": "..."}].

# Constraints

No network access. Python: ~/wp29/venv/bin/python (numpy, pandas, scipy, statsmodels available). Do not modify data/.
This task is designed to be completed in well under 20 minutes of work.

spb/wp29_uqo