Show your working

Every number on this site is reproducible from official data. Here's exactly how each was calculated, which tables were queried, and what filters were applied. Download the data and check it yourself.

1. Data Sources

All NDIS data comes from the NDIS Open Data Portal, published by the National Disability Insurance Agency.

Portal: dataresearch.ndis.gov.au/datasets

2. Core Statistics (The Numbers page)

59,848 autistic children aged 0-8

Participant Numbers and Plan Budgets, Dec 2025

This is the national aggregate row. The table contains rows for each state and service district - the statecd='ALL' and srvcdstrctnm='ALL' row is the national total. Using suppclass='ALL' gives total plan budgets (not broken down by Core/Capacity Building/Capital).

$30,000 average allocated budget

Participant Numbers and Plan Budgets, Dec 2025

$85,000 overall NDIS average budget

Participant Numbers and Plan Budgets, Dec 2025

65% less than average

$1.23 billion total spending on autism 0-8

Payments Data, Dec 2025

2.5% of total NDIS spending

Payments Data, Dec 2025

$20,574 actual spending per child

72% utilisation rate

Utilisation of Plan Budgets, Dec 2025

"Cut 50%" scenario

3.3× adult cost multiplier

Participant Numbers and Plan Budgets, Dec 2025

40% of participants have autism

3. Early Intervention ROI

The site cites an 11.3:1 benefit-cost ratio for autism early intervention, sourced from:

Synergies Economic Consulting / AEIOU Foundation (2013)
"Cost-Benefit Analysis of Providing Early Intervention to Children with Autism"
Submitted to the Productivity Commission as part of NDIS cost modelling.

Disclosure: This study was commissioned by AEIOU Foundation, an early intervention provider with a commercial interest in the result. We cite it because it was submitted to the Productivity Commission and accepted into the NDIS costing evidence base, and because independent international studies (Chasson 2007, Peters-Scheffer 2012) reach similar conclusions through different methodologies. The conservative sensitivity analysis (4.1:1) is also reported for transparency.

• Base case BCR: 11.3:1
• Conservative sensitivity analysis: 4.1:1
• Total economic benefit for cohort of 1,185 children: $1,339.8 million
• PDF: Productivity Commission submission (PDF)

Supporting studies:

• Chasson, Harris & Neely (2007) - $208,500 savings per child over 18 years. Journal of Child and Family Studies, Vol 16, pp 401-413.
• Peters-Scheffer et al. (2012) - €1.1M lifetime savings per individual. Research in Developmental Disabilities, Vol 33(6), pp 1763-1772.
• Synergies/AEIOU (2023 update) - $6.16 return per $1 invested (societal); $4.58 direct NDIS cost saving per dollar.
• AEIOU Foundation collapse (March 2026) - centres closed after 70% funding cuts, 120 children lost services. ABC News, 22 April 2026.

Applied ROI calculation

4. Appeals Section

4.5 Allied Health Evidence — Legislative Change

The site claims allied health evidence moved from mandatory to discretionary consideration. Here is the legislative basis:

The structural change: Under the old framework, allied health reports went directly to the planner who made funding decisions, and that planner was required to "have regard to relevant assessments." Under the new framework, reports the participant proactively provides fall under "may have regard to" — discretionary. Only reports the NDIA specifically requests fall under "must." The assessor is typically not an allied health professional. The budget is then calculated algorithmically from the assessment output.

NDIS Act s 32L (inserted by Schedule 1, Item 36, NDIS Amendment (Getting the NDIS Back on Track No. 1) Act 2024)

AustLII — s 32L NDIS Summary of Legislation Changes Every Australian Counts SWAN Autism

5. International Comparison

Denmark disability services spending (7.1% GDP) vs Australia (<0.5%): Hwang (2024), Australian Journal of Social Issues.

Pre-NDIS Australia lagged: Grattan Institute, "Saving the NDIS" (2024).

OECD database: oecd.org/data/indicators/public-spending-on-incapacity

5.5 Stage 3 Tax Cuts — High Earner Share Derivation

The ~$7-8B/year figure for Stage 3 tax cuts flowing to high earners ($180k+) is derived, not directly cited. Here's the working:

Step 1: The Australia Institute reported the original Stage 3 cuts would deliver $157.5 billion over 10 years to those earning over $180,000/year.
Source: Australia Institute, "95% of Stage 3 Tax Cuts go to high income earners"

Step 2: Labor's January 2024 revision redirected $84 billion over 10 years from high earners to low- and middle-income taxpayers.
Source: Australia Institute, "Stage 3 Tax Changes: A Win for Australians"

Step 3: Remaining high earner share: $157.5B − $84B = ~$73.5B over 10 years = ~$7.35B per year. We round to ~$7-8B to reflect estimation uncertainty.

Caveat: This is an approximation. The $157.5B and $84B figures come from different Australia Institute analyses using slightly different income thresholds and modelling assumptions. The direction and order of magnitude are sound; the precise figure should be treated as indicative, not exact.

6. Electoral Analysis (The Political Cost page)

Important: This is a simplified scenario model, not a prediction. It models one specific pathway (Labor→Greens primary vote switching) using historical preference flows. Real electoral dynamics involve many more variables. The model demonstrates plausible risk, not inevitable outcomes.

NDIS participants by electorate

NDIS Participants by CED data, Sep 2025

Published by the NDIA at dataresearch.ndis.gov.au. Contains active participant counts per Commonwealth Electoral Division.

Electoral margins

AEC Tally Room, 2025 federal election

Two-candidate-preferred (TCP) results sourced from the AEC's official tally room. Individual seat results:

• Bullwinkel - TCP margin: 1,066 votes
• Solomon - TCP margin: 1,576 votes
• Petrie - TCP margin: 2,730 votes
• Forde - TCP margin: 3,796 votes
• Menzies - TCP margin: 2,358 votes
• Bendigo - TCP margin: 2,983 votes (Labor vs Nationals)
• Deakin - TCP margin: 6,521 votes
• Banks - TCP margin: 4,725 votes
• Aston - TCP margin: 7,531 votes

Affected household member multiplier (×3)

Each NDIS participant is estimated to have 2 additional affected household members (a primary carer and an immediate family member). Not all household members are voting age. The multiplier estimates people directly affected by NDIS plan decisions. This is a conservative estimate - ABS data shows 3.0 million carers nationally for 5.5 million people with disability.

ABS Survey of Disability, Ageing and Carers, 2022

Greens→Labor preference flow rates

AEC preference distribution data, 2025

Per-seat Greens preference flows are published by the AEC as part of the detailed preference distribution for each division. National average: 88.2% of Greens preferences flowed to Labor ahead of the Coalition.

National figure source: The Conversation, "Election flows reveal nearly 90% of Greens preferenced Labor" (2025)

Tipping point formula

How many voters need to switch from Labor to Greens, independents, or other parties on primary vote to flip the seat (modelled here using Greens preference flows as the base case):

Seat classification

• Direct danger: Switches needed < 50% of affected household member pool - plausible from NDIS dissatisfaction alone
• Compounding risk: Switches needed > 50% - NDIS is a contributing factor but needs broader voter dissatisfaction to flip

Assumptions and limitations

Model assumptions:

• All switchers were Labor primary voters
• Model uses Greens preference flows as the base case — in practice, voters may also switch to independents, minor parties, or abstain, each with different preference leak rates
• Turnout is unchanged
• No countervailing shifts from Coalition, One Nation, or independents
• The ×3 affected household multiplier includes non-voters (children, non-citizens)

This is a simplified scenario model, not a prediction. Real electoral dynamics involve many more variables.

7. Limitations and Caveats

• Cohort averages, not individual plans: The NDIS data publishes averages by service district, age band, and disability group. Individual plans can be significantly higher or lower than the cohort average.
• Affected household member multiplier (×3) is an estimate: Not every participant has 2 affected household members. Some have more, some fewer. Not all household members are voting age. The multiplier is conservative relative to ABS carer data (3.0 million carers nationally).
• Preference flows are historical: The 2025 preference flow rates may not hold at the next election, especially if NDIS becomes a bigger or smaller issue.
• The tipping point model is simplified: It assumes vote switches go Lab→Grn only. In reality, some may go Lab→Independent, Lab→abstain, or shift between multiple minor parties.
• Electoral data is from the September 2025 CED dataset: September 2025 CED data was used for the electoral analysis. December 2025 CED data has since become available but was not used because the electoral analysis was performed before its release. Updated analysis using December data may show different participant counts per electorate.
• The ROI figure (11.3:1) is from a 2013 study: It has been updated (6.16:1 in 2023) but the original base-case figure is the most widely cited in policy documents.

7.5 What this analysis does NOT prove

This analysis does not prove:

• That every participant in the 0-8 cohort is underfunded (individual plans vary widely around the average)
• That cutting early intervention will cost exactly X in future adult support (the direction is supported by research; the exact magnitude depends on the cohort and intervention type)
• That low utilisation rates are entirely caused by provider shortages (participant choice, plan timing, and changing needs also contribute)
• That disability voters will definitely switch to any one party (the electoral model is a scenario, not a prediction — voters may go to the Greens, independents, minor parties, or simply not preference Labor)
• That higher adult autism budgets prove early intervention failure (selection effects mean remaining adult participants may have higher needs regardless)
• That NDIS sustainability concerns are entirely fabricated (growth rate projections are legitimate policy questions; the dispute is about whether cutting children is the right response)

We disclose these limitations because transparency is the foundation of this analysis. If the data doesn't prove something, we say so.

8. Reproduce It Yourself

All datasets are freely downloadable from dataresearch.ndis.gov.au/datasets. Open them in Excel, Google Sheets, or any tool that reads CSV files. Apply the filters listed above. You will get the same numbers.

If you find an error, we want to know: powerinnumbers@proton.me

9. About Power in Numbers

Power in Numbers is maintained by a small group of data professionals and NDIS families. We work in data science, policy analysis, and disability advocacy. Some of us are parents of children on the NDIS.

We remain anonymous because some of us are NDIS participants. Publicly challenging the agency that controls your child's funding is not a risk we can take.

Corrections: If you find an error, tell us. We'll fix it and note the correction. Getting this right matters more than being right first.

Contact: powerinnumbers@proton.me