Methods

What this site is / is not

This site is a static arithmetic view of observed NDIS data and official forecast vintages. It is not a claim about individual participants, providers, workers, or future policy outcomes.

Four-tier labelling

Phase 2 adds slider-driven scenario lines and a hypothetical feedback tier that appears only when feedback strength is above zero.

Definitions

Participant rows use active participants with approved plans through 2022, then active participant plans from 2023. The 2026 participant value is a 31 Mar snapshot.

Provider rows combine several published definitions. Registered and unregistered active provider ABNs do not sum to all active provider ABNs because the all-provider series is deduplicated.

Workforce rows use published headcount values. FTE is present in the contract but not published in the Phase 0 dataset.

Budget actuals use participant plan expenses where annual report financial statements are available. Macro context rows provide the 2026 anchors for population, labour force, GDP, and Commonwealth expenses.

The 2025 age-band break reflects the early-childhood age limit expanding from under 7 to under 9.

French Island NOT used as evidence: motivating anecdote, unverified, excluded from data.

Scenario engine assumptions

Scenario projections run from 2026 to 2036. Participants grow from the 2026 observed junction, providers and workers are fixed ratios of the active participant track, and budget is participants multiplied by average support cost. OWNERS_PER_PROVIDER=1.

Population growth is hardcoded from Budget Paper No.3 projected population rates in macro_findings.md. Nominal GDP growth and Commonwealth expense growth use Budget Table 11.6 rates described in the same findings file; labour-force growth uses the latest staged ABS nearby-year movement.

Overlap estimates (participant-workers, participant-provider-owners, family-linked workers) are modelling assumptions. No public dataset directly identifies these overlaps. Adjust the sliders to test different assumptions.

Cohort model

The cohort model uses post-2025 bands: 0-8 (9), 9-14 (6), 15-18 (4), 19-24 (6), 25-34 (10), 35-44 (10), 45-54 (10), 55-64 (10), 65+ (open). For finite bands, the annual age-out fraction is 1/band width, an approximation of a uniform age distribution inside each band.

Interior bands use P[b,t+1] = P[b,t] - ageout[b,t] + agein[b,t] + entries[b,t]. First-band agein is defined as 0. The 65+ terminal band uses P[last,t] x terminalRetentionRate + agein because there is no higher band to age into. The 15-18 to 19-24 boundary uses the child-to-adult retention slider.

Calibration note: defaults are checked against the 2026 observed total from the 2023 observed baseline mapped from pre-2025 bands, with the 0-6 and 7-14 break treated as an approximation for 0-8 and 9-14.

The cohort model is inferred from age-band snapshots, not longitudinal individual data. Public aggregate tables do not track the same individuals across time. Retention rates are assumptions.

Loop and hypothesis

The loop diagram follows the sequence Participants -> Services & payments -> Providers & workers -> NDIS-linked households -> Service capacity & visibility -> Participant identification & entry -> Participants.

This diagram illustrates a hypothesis. The feedback slider does not claim that providers cause participant growth. It lets users test: IF provider density is associated with higher scheme entry, how much would that change the 2036 picture?

Flagged budget and macro findings

• 2024-25 participant payments are accrual participant plan expenses and differ from cash-basis public references.
• 2013 Commonwealth expenses are below the broad planning sanity band but match Budget Table 11.6 actuals.
• 2026 population uses December 2025 ERP; 2026 labour force and CPI use May 2026; 2026 GDP deflator uses March quarter 2026.
• CPI index cells before 2024 are empty because the current ABS monthly CPI index series starts in 2024.

Full findings: macro_findings.md.

Data version

Current data version: v2026-07-03. Validation report: validation_report.json.

Validation summary: 0 pass, 0 warn, 0 fail.

Source table