what it costs to keep last year's life
What does it cost to live last year's life? A replication-cost price
measure for bottom-40%-income US households: freeze the consumption experience
a household actually bought last year — including formerly-included features —
and price the cheapest way to reproduce it on today's menu. Published as a
spread over official CPI series (the "Restoration Gap"), never as a rival
index. Public site: site/ (deploys to restorationgap.org).
Status: pre-launch. Methods pre-registration: Oct 2026 · first print: targeted Q1 2027. No index values are published yet; everything numeric in this repo is verified event data or test fixtures, not index output.
Firms increasingly raise effective prices by unbundling included features into fees (airline bags), moving the old standard behind a premium tier (ad-free streaming), and redesigning menus so the cheapest option is a thinner one. Matched-specification CPI answers a different (compensated) question, and average-spending weights read household coping as savings. This measure holds the experience fixed and prices its restoration, line item by line item — an explicit upper bound on constant-experience cost growth, reconciled to the official series it differs from in every release. It is not a "true inflation" number, and says so everywhere.
| Path | What it is |
|---|---|
_~Restoration Index Spec v*.md |
The methodology specification (formal object, worked examples, decisions, hazards) |
_~Restoration Index Project Plan v*.md |
Phased plan: coverage map, sector rulings, gates G0–G3, workstreams, next steps |
docs/codebook-v0.md |
Operating rules: equivalence protocol, trigger rules, route hierarchy, conventions — freezes as the public pre-registration |
docs/g0-plausibility-band.md |
Pre-committed pass/kill band for the 2008–09 retrospective (committed before any inputs/results, by design) |
docs/ (other) |
Discussant shortlist, DOT docket comment draft, budget memo |
data/ledger-events.json |
The Inclusion Ledger event registry (verified unbundling/tiering events, with schema) |
data/retro-2008-inputs.md |
Phase 0 inputs: the 2008–09 bag-fee chronology + designated negative-print episode |
data/incidence-proxies.md · data/cex-b40-quintiles.csv |
Utilization-weight register · bottom-two-quintile CE expenditures (byte-verified) |
pipeline/ |
Menu-snapshot capture (capture.py), targets, weights builder — the evidence archive tooling |
engine/ |
The replication-cost computation engine; its tests are the spec's worked examples |
site/ |
The public site (plain static HTML, no build step) |
python3 -m unittest discover engine -v # engine tests = executable spec examples
python3 pipeline/capture.py # snapshot posted menus (see pipeline/README.md)
python3 pipeline/build_weights.py # rebuild CE quintile weights from sourcePre-registered methods frozen before live collection; a retrospective validation gate (G0) with a band committed before results existed; external review before the first print and again before any composite; negative controls (groceries ≈ 0, banking < 0) shipped alongside the headline; an Exclusion Register that counts what the measure refuses to price; open code; hash-committed evidence snapshots; a public correction log.
Building in the open: this single repository is the project's public home — working documents, decision records, review dispositions, and internal drafts included, deliberately. Treat internal-sounding docs (discussant shortlist, budget memo) as process transparency, not published positions.
License: code (engine/, pipeline/, site/ markup) under MIT;
documents, data, and site text under CC BY 4.0.
Contact: see the site's About page.