Corrections made during research

What this evidence tells us.

3 October 2026, internal author review. These are corrections to this research process, not allegations of source errors and not an independent audit.

• Disability assistance: use 49.9% for people with disability needing assistance in 2022. The same ABS page's 52.3% refers to a different, older-person population.

• Unpaid work: 4h53 and 3h52 are averages among participants. Preserve the separate 90%/81% participation rates.

• Wages: align quarterly WPI with quarterly CPI. June-only monthly CPI is not the matching quarter. A mistakenly selected CPI table-10 workbook is retained but unused; table 17 supplies the actual calculation.

• Aged care: the 2026 application-to-commencement series supersedes using the old Home Care Package list as a current measure. It includes personal/provider/system elapsed time and only completed pathways.

• NACCHO: the dated annual report says 146 ACCHOs; the website and other passages have different counts. No membership trend is inferred.

• Housing: calendar-quarter headlines and financial-year appendix values remain separate. The rental-stress denominator wording is flagged rather than turned into an ambiguous new claim.

• Social housing: separate program waitlists may overlap. Use the published consolidated figure instead of adding them.

• APLNG: tax payable is reconciled to the same entity's ATO record; cash income tax and royalties are separate. FY2026 amounts remain preliminary. The company's rounded payment components differ by $1m from its displayed cash-tax total.

• Fuel: country-row FY2025–26 diesel imports sum to 0.1 ML less than national monthly controls. Retain this rounding difference; the national control is the share denominator.

• First Nations land: subtract overlap between native-title and estate areas. State language counts overlap and are not summed.

• Survey differences: ABS describes some numerically different groups as statistically similar. Do not override that judgement from point estimates alone.

• Verification boundaries: downloaded sources are not all fully analysed; selected source controls and named calculations are the reproducible subset. An automated pass is not independent expert or community review.

• APLNG R&D: the FY2024 comparative reconciliation shows a positive $3m tax-expense adjustment. It is not treated as a measured public subsidy.