# CPI Prediction Board Public-Good Plan

Date: 2026-06-12

## Purpose

The board should use prediction-market mechanics for learning, not for gambling. A simple bucket market makes uncertainty visible:

- what range is the market expecting?
- what evidence moved the distribution?
- what did official CPI later show?
- which model or data contributor improved the forecast?
- where did policy, corporate pricing, or household reality diverge?

## Initial boards

1. MIC official CPI: headline national CPI for the survey month.
2. BOJ underlying CPI: core indicators after official CPI release.
3. JGB curve response: 10Y yield or yield-curve regime after CPI surprise.
4. Yen import-pressure lane: USD/JPY and resource import-pressure buckets.
5. Pacific Rim resource basket: LNG, oil, shipping, copper, lithium, grains, rice, seafood, fertilizer, and other supply-chain drivers.

## Resolution principles

- Use named official sources for resolved public statistics.
- Publish the source URL, release date, survey month, and settlement rule before the board opens.
- Later revisions should be tracked but should not rewrite the first-resolution result unless the rule explicitly says so.
- If a source changes methodology, freeze affected boards and document the transition.
- Keep simulated and real-money statuses visibly separate.

## Forecast-composition engine

The CPI basket proposal engine should compare:

- official CPI weight structure
- live price proxies
- shrinkflation and quality changes
- substitution effects such as food switching
- regional differences
- wage and rent pressure
- energy and FX pass-through
- policy subsidies and institutional factors

The goal is not to replace official CPI. The goal is to show which composition better predicts future official data, market reactions, and household experience.

## Local prototype behavior

The current `app.js` board:

- starts with demo research credits
- stores only local browser state
- moves bucket probabilities when the user commits simulated credits
- has no backend, no account, no wallet, and no settlement

## Product milestones

1. Static public page.
2. Historical dataset import.
3. Official release calendar import.
4. Model-score leaderboard for research contributors.
5. Basket proposal and backtest module.
6. Private expert review workflow.
7. Counsel-reviewed token and prediction-market memo.
8. Regulator-ready system spec if Japan licensing becomes viable.