# Connecting the evidence to useful public questions

This package supplies research candidates. No record is labelled ready for production ingestion: source reuse, uncertainty and the concrete destination contract still require review. The accepted WA gas journey remains the active product context; this national research does not change its delivery order or mix Queensland APLNG evidence into WA production.

## Three reviewable content candidates

| Citizen question | Evidence records | Intended use | Required presentation boundary |
| --- | --- | --- | --- |
| Why can a country look wealthy while families struggle? | cashflow-2020/2025; emergency-2020/2025; wealth-highest/lowest; real-income series | National household context, linked to housing and tax/service explanations | Show dates and survey coverage; no combined personal budget or implication that old wealth distribution describes 2026 |
| Who finances a resource project, and what money reaches government? | APLNG ownership/EXIM ledger, tax/cash records and ATO crosswalk | A separate Queensland company/project example supporting the national money explanation | USD authorisation distinct from AUD cash; consolidated liability distinct from tax paid; no net-return percentage |
| How dependent are we on foreign diesel suppliers? | diesel-dependency.json; diesel-cover; iea-cover | Supply-country exposure and clear stock definitions | Country is recorded supply source, not beneficial ownership or crude origin; different stock-day definitions remain separate |

Each candidate needs a page contract with a question, exact observation IDs, source/release date, plain-language denominator, visible limitation, comparison period and correction route. Preserve publisher tables separately from typed observations; a raw table's numeric-looking cell may be a confidence interval, suppressed value or category code. The manifest and source hashes allow review of the same release used here.

## One bounded comparison specification: diesel import exposure

**Question:** How much of the recorded FY2025–26 diesel import volume corresponds to a selected reduction in imports from one source country?

**Inputs:** the twelve-month import controls and country rows in `diesel-dependency.json`; selected country; a user-selected illustrative reduction between 0% and 100%. Default reference period stays fixed at FY2025–26. No live claim is implied.

**Arithmetic:** affected gross import volume = country import ML × assumed reduction. Exposure share = affected gross import volume / national gross diesel imports. Show the small published-country rounding difference separately. “Remaining recorded import volume under this assumption” is an arithmetic remainder, not forecast available supply.

**Outputs:** country baseline, assumed reduction, affected ML and percentage of baseline imports. Show sensitivity examples at 10%, 25% and 50% only as user assumptions. Use the same source scope and denominator for every option. Preserve country labels and period; never reallocate suppressed data or combine countries by an inferred political bloc.

**Explicit model boundary:** no prediction of shortage, price, jobs, GDP, replacement imports or days until exhaustion. Those require opening usable stocks, seasonal demand, contractual and product constraints, refinery inputs/capacity, route lead times, substitution and behavioural responses. Stock-cover days use a different denominator and must not be multiplied into this calculation.

**Verification before implementation:** selected country's twelve-month sum reproduces; the national denominator matches; all inputs finite/nonnegative; reduction bounds enforced; zero and 100% reductions produce the expected arithmetic; period/source changes produce a new model version; UI copy labels an illustration. Keep this prospective model separate from `cash-scenario-v1`.

This is a completed requirement specification, not a built or evaluated policy model. It is deliberately answerable with the evidence collected, while making the missing inputs for a true resilience model concrete.

## Review decisions still owned by people

Choose the citizen journey before loading data; review statistical uncertainty and licensing at that destination. Obtain subject-matter review for consequential interpretations and appropriate community review for community-specific presentation. Those reviews have not occurred. Reader comprehension should be tested with the actual bounded claims, not an empty dashboard or a request to trust a national score.