Appendix D: What the NDIS Bill 2026 Really Does

A guide to the most serious findings and a correction to the government’s own Impact Analysis

To the Australian Government: why we have sent you this document

Every significant Australian Government policy decision must be supported by an Impact Analysis, assessed by the Office of Impact Analysis (OIA) against the Australian Government Guide to Policy Impact Analysis.

The process exists to make sure decision makers, and the public, can see the problem, the options, the costs and benefits, the risks, and who is affected, all set out honestly and to a published standard (OIA support and guidance).

This is what Disabled People Against Cuts (DPAC) has found when we held the NDIS reforms Impact Analysis to that standard: we have, in effect, corrected the government’s homework, and there turned out to be a good deal more red pen required than the “Adequate” rating suggests.

The government’s original Impact Analysis is here. On the strength of the errors and omissions set out below, and the fact that two of the most important figures were quietly corrected after the rating without the errors being published, DPAC’s position is that the OIA should downgrade its assessment and reassess the Impact Analysis against the corrected information. This document is provided to the OIA and to the Senators conducting the inquiry, and may be circulated to them.

Guide Explaining Serious Problems With the Government’s NDIS

Amendment (Securing the NDIS for Future Generations) Bill 2026

This guide explains the most serious problems with the government’s NDIS Amendment (Securing the NDIS for Future Generations) Bill 2026. It covers what the government’s official report says, what it gets wrong, and what it leaves out entirely.

The government announced the Bill on 12 May 2026 and introduced it to Parliament on 14 May 2026. Every law like this must come with an Impact Analysis: an official report on what the law will do and who it will affect.

The Office of Impact Analysis rated the government’s report only “Adequate”, the second lowest of four possible ratings.

The figures below come from the government’s own report, the ANAO (the national auditor), the Australian Government Actuary, disability organisations, NDIA data and independent economists. Sources are linked throughout.

How to Write an Impact Analysis

Correctly

A marked-up guide to the Impact Analysis for the NDIS Amendment (Securing

the NDIS for Future Generations) Bill 2026

Corrections and insertions shown in red

Prepared by Samantha Connor AM, Disabled People Against Cuts

Page 3

How to Write an Impact Analysis Correctly

A line-by-line teaching guide, using the National Disability Insurance Scheme reforms Impact Analysis (May 2026) as the example.

How to read this document

This document teaches the Australian government how to write a Commonwealth Impact Analysis that meets the standard set by the Office of Impact Analysis. The Office of Impact Analysis is the unit inside the Department of the Prime Minister and Cabinet that sets and assesses the rules for impact analysis across the Australian Government. It uses one real Impact Analysis as the worked example: the Impact Analysis for the National Disability Insurance Scheme reforms, published by the Department of Health, Disability and Ageing in May 2026 to support the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026. The document follows the order of the Impact Analysis itself. For each part, it does four things. First, it states what the Office of Impact Analysis guidance requires, in plain terms. Second, it marks what the Impact Analysis actually did, using the Office of Impact Analysis marking scale of Not Completed, Adequate, or Best Practice. Third, it shows the correct version, or the missing content, so the writer can see what good looks like. Fourth, it inserts, in red, every element the Impact Analysis left out or got wrong, with the correct data and its source. A note on colour. Text shown in red is an insertion or a correction by the marker. It is either content the Impact Analysis omitted, or a figure the Impact Analysis stated incorrectly, replaced with the correct figure and its source. Black text is explanation and teaching. This mirrors the way a tracked change or a red pen works on a marked draft.

The marking scale, in one paragraph

The Office of Impact Analysis assesses each Impact Analysis against seven questions. For each element within a question it can rate the work as meeting adequate practice or best practice, and where a required element is simply absent, that element is not completed. The Office of Impact Analysis publicly rated this Impact Analysis as Adequate, which is the second rating on its scale. Its published comment was that the Impact Analysis “would have benefitted from further quantification of the costs and benefits and a more detailed implementation and evaluation plan.” 1 This document explains, question by question, exactly where the Adequate rating came from, and what a Best Practice version would have contained.

The seven questions every Impact Analysis must answer

The Office of Impact Analysis Impact Analysis Guide is built around seven questions. Every Impact Analysis must answer all seven. They are:

1. What is the policy problem and what data are available to describe it?
2. What are the policy objectives, why is government action needed to achieve them, and how will success be measured?
3. What policy options are being considered?
4. What is the likely net benefit of each option?
5. What consultation was undertaken and how was the feedback used?
6. What is the best option and how will it be implemented?
7. How will the policy be monitored and evaluated?

Around these seven questions sit a set of supplementary guidance notes that tell the writer how to do specific parts of the work: how to run a cost benefit analysis, how to measure regulatory burden, how to value a change in the risk of death, how to analyse who wins and who loses, how to assess effects on competition, small business and community organisations, how to analyse risk, and how to consult. A proper Impact Analysis applies these notes. This one applied some and ignored others. Each is covered in its place below.

The worked example: what this Impact Analysis was for

The Impact Analysis supports the National Disability Insurance Scheme Amendment (Securing the NDIS for Future Generations) Bill 2026. The Government announced the reforms on 12 May 2026 and introduced the Bill on 14 May 2026. 2 The Impact Analysis considers three areas of change and states a preferred option in each: changes to eligibility and to the supports that are funded (preferred option two); changes to plan management (preferred option two); and changes to support coordination (preferred option three). The rest of this guide walks through the Impact Analysis in its own order and marks each part.

2Department of Health, Disability and Ageing, About the changes to the NDIS. hhttps://www.health.gov.au/our-work/ndis-legislation-changes/amendments/ndis-amendment- s securing-the-ndis-for-future-generations-bill-2026/about-the-changes-to-the-ndis

Executive summary

What the guidance requires The executive summary should state the problem, the options, the preferred option, the net benefit, and the main risks, in a way that a decision maker can rely on without reading further. Every headline number in the summary must be supported by the analysis inside the document and must be correct.

Marking Verdict: Adequate, with two material errors of fact that must be corrected before the summary can be relied on. The summary sets out the three areas of reform and names a preferred option in each. That meets the adequate standard for structure. It fails on accuracy in two places that matter, because both errors make the case for the reform look stronger than the evidence supports. Both are corrected below in red.

Correction 1: the savings are not mainly about fraud or pricing integrity

The Impact Analysis and the surrounding public messaging frame the savings as being substantially about stopping fraud and tightening pricing integrity. That framing is not supported by the Government’s own tabled modelling. Correction. Of the $38.1 billion in projected savings over four years, only about $0.9 billion, or 2.4 per cent, comes from the measures related to fraud and to making the Minister the decision maker on pricing. The largest single share, about $13.2 billion or 34.6 per cent, comes from reducing community participation and therapy budgets. A further $9.3 billion, about 24.4 per cent, comes from the new functional capacity test. These figures are from the Treasury modelling tabled in late May 2026. 3 This was confirmed on the public record. At Senate Estimates on 5 June 2026, Senator Jordon Steele-John put it to officials that it was “misleading to pick one line item from the ten,” and the evidence was that “for the most part, there are not significant savings to government” from the integrity measures alone. 4

The scale of fraud is also smaller than the framing suggests.

The National Disability Insurance Agency reported 72 active fraud cases with an estimated value of $34.5 million as at 30 June 2024, which is about 0.09 per cent of scheme payments. The often-quoted figure of $3.7 billion in “integrity leakage” for 2024 to 2025 is an internal estimate, not a measured loss; the Australian National Audit Office has put leakage in the range of 6 to 10 per cent, and a properly written Impact Analysis would present it as an estimate with that range, not as an established fact. 5

A worked Impact Analysis would also weigh the evidence that plan management, one of the markets this Bill changes, already prevents fraud cheaply. Disability Intermediaries Australia reported that a survey covering about 30 per cent of plan-managed participants stopped $270.5 million in incorrect, duplicate and suspect invoices before a single dollar reached the Agency, which is about $5.61 stopped for every $1 paid in plan management fees, and roughly 7.6 times the value recovered by the Agency’s own dedicated fraud program. Cutting the market that does this work is not obviously an integrity measure. 6

Correction 2: annualised committed supports are plan budgets, not money spent

The Impact Analysis relies on annualised committed supports as if they were the amount of support people actually receive. They are not. Annualised committed supports are the amounts committed in plans. The amount actually spent is lower, because most participants do not use their whole plan. Correction. Annualised committed supports are committed plan budgets, not payments made. Treating them as actual support received overstates real spending and therefore overstates the saving from cutting them. Aggregate utilisation across the scheme is about 74 per cent, and about 68 per cent once Supported Independent Living is excluded. As at December 2025, about 32 per cent of participants used half or less of their plan, and a further 22 per cent used between 50 and 75 per cent. Utilisation differs sharply by budget type: about 81 per cent for core supports, about 59 per cent for capacity building, and about 56 per cent for capital. An Impact Analysis that measures savings against committed budgets, rather than against actual and forecast spending, will overstate the benefit of every cut. 7 8

What a Best Practice executive summary would have added

A best practice summary would have stated the net cost as well as the net saving. It would have named, up front, the value it places on a change in the risk of death, because the reforms affect supported accommodation, therapy and support intensity. It would have flagged that the largest cost category, the effect on people losing supports, could not be monetised, and said so plainly rather than leaving it out. Those points are developed under Question 4.

Question 1. What is the policy problem and what data are available?

This maps to the Impact Analysis sections on background, the overview and history of the scheme, the policy issues including sustainability, who is affected, what has already been done, and data availability.

1.1 Define the problem: who, what, when, where and how

What the guidance requires. Adequate practice defines the problem in general terms. Best practice defines it by identifying who is affected, what the problem is, when and where it occurs, and how it arises. Best practice also demonstrates, with evidence, the consequences of doing nothing.

Marking. Adequate. The Impact Analysis defines the problem as the growth in scheme cost and describes it in general terms. It does not fully pin down who is affected by the problem as distinct from who is affected by the fix, and it does not clearly separate genuine cost growth from the effect of past administrative decisions.

Insertion. The problem statement should say plainly that a large part of the growth it describes reflects prior decisions about how the scheme was run, not an inherent flaw in the entitlement. As the Justice and Equity Centre put it, “Many of the measures proposed to constraint Scheme growth are only necessary because of prior and ongoing failures in Scheme implementation.” Defining the problem as “the scheme costs too much” rather than “past decisions drove avoidable cost, and some growth is genuine demand” leads directly to the wrong options being chosen.

1.2 Show the magnitude, and the cost of inaction

What the guidance requires. Show the size of the problem with data, and show what happens if nothing is done. The risk guidance note requires the problem to be framed in terms of objective risk, not perceived risk.

Marking. Adequate. The Impact Analysis provides cost figures and growth rates. It frames inaction almost entirely as continued cost growth. It does not weigh the objective risks created by the fix itself, which the risk guidance requires.

Insertion: the correct baseline figures. As at the end of March 2026 the scheme had 774,456 participants, and payments were $50.2 billion over the twelve months to March 2026, with an average annualised package of $66,800, rising to $84,030 for participants in Supported Independent Living. These are the figures against which any saving should be measured.

Victoria, Queensland, the Australian Capital Territory and South Australia.

None of this is described as an impact on affected businesses. 11

1.4 Explain existing policies, and why they are not enough

What the guidance requires. Adequate practice explains how existing measures have tried to address the problem. Best practice demonstrates that changing implementation, for example better decision making or stronger enforcement, would not be enough on its own, so that a legislative change is justified.

Marking. Not Completed on the key point. The Impact Analysis does not demonstrate that better administration would fail. This matters, because a central criticism is that many changes are unnecessary.

Insertion. The Justice and Equity Centre advises that the shift to functional capacity “does not in fact require legislative change,” and that the funding of increases that are not necessary or not appropriate “demonstrates issues with the quality of decision-making within the NDIA,” which points to better administration rather than new law. A best practice Impact Analysis would have to answer that argument with evidence, and this one does not. 12

1.5 Identify the data, its quality, the gaps, and the effect of the gaps

What the guidance requires. Adequate practice describes the available data, assesses its quality, and identifies gaps and why they exist. Best practice also discusses how those gaps affect confidence in the whole analysis.

Marking. Adequate. The Impact Analysis lists data sources and notes some gaps. It does not carry the effect of the gaps through into its confidence in the savings, which is what best practice requires and what the Office of Impact Analysis was pointing to when it asked for further quantification.

Insertion. The single most important data problem is that the assessment tool that will decide eligibility does not yet exist. The Government confirms that the thresholds and the assessment for functional capacity will be designed by a Technical Advisory Group that only begins work in mid-2026, and that consultation on the new planning framework, functional capacity assessment, Supported Independent Living commissioning, support coordination commissioning and differentiated pricing will only begin in the second half of 2026, which is after the Bill. An Impact Analysis cannot state a confident saving from a test that has not been written. This gap should be stated as a limit on the confidence of every eligibility-related figure in the document. 13

Question 2. What are the objectives, why is government action needed, and how is success measured?

This maps to the Impact Analysis section on the need for government intervention, including its section on measuring success.

2.1 Establish why government should act

What the guidance requires. Adequate practice gives sound reasons for government to act. Best practice justifies direct intervention with substantiated evidence and shows government can act successfully.

Marking. Adequate. The Impact Analysis argues that the scheme’s growth threatens its own future, which is a sound reason to act. It does not establish that this particular intervention can succeed, given that the tools it relies on do not yet exist.

2.2 State objectives that are specific, measurable, achievable, relevant and time bound

What the guidance requires. Adequate practice states objectives that address the problem. Best practice states objectives that are specific, measurable, achievable, relevant and time bound.

Marking. Adequate. The objectives address cost growth but are not expressed as measurable targets against a baseline, so success cannot be tested later.

Insertion. A measurable objective would read, for example: “reduce the average annualised package from $66,800 while holding the proportion of participants reporting unmet need at or below its current level, measured annually.” Without a target of that kind, the evaluation under Question 7 has nothing to test against. 14

2.3 Identify barriers to success

Marking. Not Completed. The Impact Analysis does not squarely identify the main barrier to its own success, which is that the alternative systems people are meant to move to do not yet exist.

Insertion. The clearest barrier is the dependence on foundational supports and on state and territory systems that are not yet operational. People with Disability Australia warns that the Bill “removes vital supports from hundreds of thousands of people at a time when no equivalent services currently exist,” and “rests on an assumption that state and territory governments will fill this gap, despite clear indications that they are not willing or able to provide comparable supports within the required timeframe.”

A best practice Impact Analysis states this barrier and explains how it will be overcome before the cuts take effect, not after.

2.4 Success factors and their likelihood

Marking. Adequate. Success factors are described but their likelihood is not weighed, which is the best practice requirement.

Question 3. What policy options are being considered? This maps to the Impact Analysis section on policy options across eligibility and supports, plan management, and support coordination.

3.1 Present genuine and viable options, including the status quo and a non regulatory option

What the guidance requires. Adequate practice presents a range of options. Best practice justifies why each option is viable, discusses the options that were discarded, and, importantly, the cost benefit guidance requires at least three options plus the status quo as the base case, and at least one non regulatory option unless a non regulatory option can be shown to be unviable. Marking. Adequate. Each reform area presents options, which meets the adequate bar. The set is weak on the non regulatory alternative and on discarded options. Insertion. The guidance expects a non regulatory option to be presented and tested. Better administration, better decision quality inside the Agency, and improved fraud prevention through the existing plan management market are all non regulatory levers that the evidence says are available. The Impact Analysis should present at least one of these as a genuine option and explain, with evidence, why it was not preferred, rather than moving straight to legislative restriction.

3.2 Demonstrate each option can achieve the objectives

Marking. Adequate. The Impact Analysis explains how each option addresses cost growth. It does not demonstrate the limitations of each option, which is the best practice

• requirement, and it does not test whether an option achieves the objective without creating a larger cost elsewhere.

3.3 Give the context, and analyse the distribution of costs and benefits proportionate to the size of the option

What the guidance requires. Adequate practice gives context for each option. Best practice analyses flow-on and distributional costs and benefits in proportion to the size of the option. This is where the distributional analysis guidance note first bites.

Marking. Not Completed. There is no distributional analysis attached to the options, even though the options are very large and fall unevenly.

Insertion. The distributional guidance requires the writer to consider effects across ten cohorts: age, gender, disability type, Indigeneity, region and remoteness, income and wealth, business size, cultural background and immigration status, employment status, and housing stability, and to consider how these intersect. The reforms plainly are assessed unevenly. The Impact Analysis’s own figures show the cut to social, civic and community participation budgets will affect 393,401 people, which is 52 per cent of participants who currently have that funding, and the Impact Analysis itself concedes the cut “will impact some disability groups more significantly.” A distributional section is required for each large option and is absent.

Question 4. What is the likely net benefit of each option?

This maps to the Impact Analysis section on the likely net benefit of the options. It is the single largest gap in the document, so it is marked in the most detail. The Office of Impact Analysis publishes separate guidance notes on cost benefit analysis, on the regulatory burden measurement framework, on distributional analysis, on competition effects, and on valuing statistical life and health. A best practice net benefit section reflects all of them. This one reflects almost none of them.

4.1 Quantify the costs and benefits of each option against the base case

What the guidance requires. Adequate practice describes the main costs and benefits. Best practice quantifies them in dollars where it can, monetises what can reasonably be monetised, and presents a clear comparison of each option against the status quo base case. The cost benefit guidance sets out a nine step method: define the base case, list the options, identify the effects, quantify them, value them in dollars, discount future values to present value using a central real discount rate of 7 per cent, test the result at 3 per cent and 10 per cent, express the result as a net present value or a benefit to cost ratio, and state the assumptions.

Marking. Not Completed. The Impact Analysis does not contain a cost benefit analysis in this sense. It states projected savings to the Commonwealth but does not quantify the costs those savings transfer to participants, families, carers and other service systems, and it does not present a net present value for any option. Savings to one party are not the same thing as a net benefit to the community, and the guidance is explicit on that point.

Insertion. The savings figure is a transfer, not a net benefit. The headline is a reduction in projected Commonwealth spending of about $38.1 billion over four years against the forward estimates. A cost benefit analysis must ask where that money goes and what is lost. Treasury modelling tabled in late May 2026 shows that only $0.9 billion of the four year figure, about 2.4 per cent, comes from minister set pricing and fraud measures. About $13.2 billion, or 34.6 per cent, comes from reducing funding for social, community and civic participation and for therapy. About $9.3 billion, or 24.4 per cent, comes from the new functional capacity test. These are reductions in supports to people, not efficiency gains. The Justice and Equity Centre states plainly that the Bill “seeks to achieve savings primarily by excluding large numbers of participants from the Scheme and by constricting access to supports for those who remain,” and that this “simply shifts these costs to other systems,” including “acute health, mental health, housing and crisis services.” A net benefit calculation

Value Life and Health Effects Where the Option Affects Safety

What the guidance requires. The Office of Impact Analysis publishes a specific guidance note on the value of statistical life. Where an option changes the risk of death or serious injury, best practice monetises that change using the value of a statistical life, currently $5.87 million, or the value of a statistical life year, currently $253,000 in 2025 dollars at a 3 per cent discount rate, adjusted for disability weightings drawn from the Australian Institute of Health and Welfare Burden of Disease study. The guidance is clear that safety and mortality effects are to be valued, not left as narrative.

Marking. Not Completed. The Impact Analysis does not value any mortality or safety effect, even though the reforms remove or reduce supports on which some participants depend for daily survival, and even though submissions raised the risk of death directly.

Insertion. Mortality is a foreseeable effect and the guidance requires it to be valued. The Australian Government Actuary reported 909 deaths of participants in Supported Independent Living in 2022-23, rising to 1,050 in 2023-24. Supported Independent Living is the funded support for people with high daily needs who cannot live safely without paid assistance. The Justice and Equity Centre warns that reduced or withdrawn supports carry, “in the most severe cases, the risk of death.” When an option changes the risk faced by a population with a known and rising death count, the valuing statistical life guidance requires the writer to estimate the change in expected deaths and to value it. Using the value of a statistical life of $5.87 million, a change of even a small number of deaths per year is a cost of tens of millions of dollars per year that must appear in the net benefit calculation.

The Impact Analysis records none of this.

At a minimum it should state the current death count, identify which options plausibly change it and in which direction, and value the change or explain why it cannot be valued. The section does not make any commentary on mortality, despite the guidance and the Agency’s own modelling, that predicts that more disabled people will die in group settings in the future - this modelling shows that more people are dying now than during the peak of the pandemic.

24Australian Government Actuary, advice on NDIS participant mortality reported by the National Disability Insurance Agency (909 deaths in Supported Independent Living in 2022-23, 1,050 in 2023- 24), 14 October 2024. https://dataresearch.ndis.gov.au/media/4468/download?attachment

25Office of Impact Analysis, Value of statistical life guidance note (value of a statistical life $5.87 million; value of a statistical life year $253,000, 2025 dollars). hhttps://oia.pmc.gov.au/resources/guidance-assessing-impacts/value-statistical-life

27 Australian Government Actuary. (2025). June 2025 Financial Sustainability Report: Reviewing Actuary’s letter to the National Disability Insurance Agency Board. Letter from Guy Thorburn, Reviewing Actuary, to Kurt Fearnley AO, Chair, National Disability Insurance Agency, 20 October 2025.

Page 16

4.3 Measure the regulatory burden using the Regulatory Burden Measurement Framework, for every viable option

What the guidance requires. The regulatory burden guidance requires a regulatory burden estimate for every viable option, expressed as an average annual change in compliance costs across administrative costs, substantive compliance costs and delay costs, and split across the three affected groups: business, community organisations, and individuals. Best practice presents the estimate per option and explains the method.

Marking. Adequate. A regulatory burden table is present, which clears the adequate bar, but it covers only two of the reform areas and its community organisation figure is not credible.

Insertion. The regulatory burden estimate is incomplete and internally doubtful. The table gives an average annual regulatory cost of $32.24 million, made up of $4.44 million to business, $27.8 million to individuals, and zero to community organisations. Two problems. First, the estimate covers only the plan management and support coordination changes. It contains no regulatory burden estimate for the eligibility and functional capacity reforms, which are the largest changes in the Bill and which impose new evidence gathering, reassessment and record keeping obligations directly on individuals. A regulatory burden estimate is required for every viable option, so the eligibility options need their own line and do not have one. Second, a community organisation figure of exactly zero is not credible. Many providers of plan management and support coordination are small not for profit community organisations, and Disability Intermediaries Australia describes a market in which more than 5,600 active providers support five or fewer participants each. The move to a commissioned panel imposes transition, re registration and contracting costs on exactly these organisations. A zero is almost certainly an error or an omission and must be corrected or explained.

Characteristics intersect. Best practice presents a distributional analysis for each option that falls unevenly.

Marking. Not Completed. No distributional analysis is presented against the options, despite the reforms being large and uneven, and despite the Impact Analysis itself conceding uneven effects.

Insertion. The reforms are spread evenly and the guidance requires this to be set out. The Impact Analysis’s own figures, as quoted by People with Disability Australia, show that the cut to social, community and civic participation budgets will affect 393,401 people, which is 52 per cent of participants who currently hold that funding, and the Impact Analysis concedes the change “will impact some disability groups more significantly.” Specific cohort effects raised in submissions and not analysed in the document include:

• First Nations participants. The Justice and Equity Centre states the reforms “are likely to have a disproportionate and adverse impact on certain communities including Aboriginal and Torres Strait Islander participants residing in the Northern Territory”
• Women, girls and gender diverse participants. Women With Disabilities Australia states the reforms “risk deepening existing inequalities” and could “exclude people with fluctuating, psychosocial, chronic, pain-related and trauma-related disabilities particularly women and girls,” and asks that the Bill not proceed “until a comprehensive gender impact analysis has been released”
• People with cognitive disability. The Impact Analysis’s own report notes that 237,253 participants or 65 per cent of participants aged 19 and over have a disability that may affect how they think and may require support to make decisions, which bears directly on the fairness of the new decision heavy processes.

A distributional analysis for each large option across the ten cohorts is required and is absent.

4.5 Assess competition effects and apply the public interest test

What the guidance requires. The competition guidance requires the writer to consider whether an option restricts the number or range of suppliers, limits the ability of suppliers to compete, or reduces their incentive to compete, and, where competition is restricted, to show the restriction is no more than necessary and is in the public interest. Marking. Not Completed. The Bill moves plan management and support coordination to a commissioned provider panel and gives the minister power to set prices, both of which are direct restrictions on competition, and neither is assessed. Insertion. The reforms restrict competition and the restriction is not tested. Moving plan management and support coordination to a commissioned panel means the government, not the market, decides who may supply. That plainly restricts the number of suppliers. Minister set pricing removes price competition. Disability Intermediaries Australia notes that the Impact Analysis “does not mention the commissioned panel once,” contains “no transition pricing,” and does not acknowledge “that the market it is describing today is the same market being legislated out of existence from October 2027.” The competition guidance requires the writer to identify each restriction, to explain why a less restrictive option would not achieve the objective, and to show the restriction is in the public interest. None of this is done. The document should identify the panel model and minister set pricing as competition restrictions and apply the public interest test to each.

of spending, but plan budgets are commitments, not payments. Aggregate utilisation is about 74 per cent, and about 32 per cent of participants used half or less of their plan in the quarter to December 2025. Savings measured against committed budgets will therefore overstate the real reduction in support delivered, because much of the committed money was never going to be spent. This is corrected in the executive summary and must flow through every figure in this section.

Question 5. Who was consulted, and how were their views taken into account?

This maps to the Impact Analysis section on consultation, including its treatment of the social, community and civic participation and capacity building reset and its treatment of fraud.

5.1 Show consultation was proportionate to the size of the impact What the guidance requires. The consultation guidance expects consultation to be proportionate to the size and reach of the problem. For a reform of this scale, best practice means a public exposure draft, a consultation period of at least 30 to 60 days, and evidence that the consultation genuinely informed the decision rather than being run after it. Marking. Not Completed. The reform was announced on 12 May 2026 and the Bill was introduced on 14 May 2026. The decision was therefore taken before any public consultation on the Bill, and there was no exposure draft. Insertion. Consultation after the decision is not consultation for the purpose of the guidance. Best practice consultation happens early enough to change the decision. Here the design decision was announced two days before the Bill was introduced, and the government’s own material confirms that consultation on the substance, on the new planning framework, the functional capacity assessment, Supported Independent Living commissioning, support coordination commissioning, differentiated pricing and market reforms, will “begin in the second half of 2026,” that is, after the Bill. The Justice and Equity Centre states that “the timeframe needs to enable genuine consultation with the disability community, which has not occurred,” and People with Disability Australia describes the approach as “a ‘cut first, design later’ model of legislative reform that prioritises immediate cost reduction over considered, evidence-based policy development.” A best practice Impact Analysis would either show a

5.2 If consulting after the decision, meet the exceptional case standard

What the guidance requires. The guidance allows, in exceptional cases, for consultation to occur after a decision where there is a demonstrated need for extreme confidentiality, for example to prevent market gaming or fraud before a measure starts. Where this exception is relied on, the writer must state it and justify it.

Marking. Not Completed. The Impact Analysis does not claim the exceptional case exception, and no case for extreme confidentiality is made out for reforms of this kind.

Insertion. The confidentiality exception does not fit these reforms. The exception exists for measures where advance notice would let people rearrange their affairs to defeat the measure, such as some tax changes. Changes to eligibility criteria, functional capacity tests and support budgets for a known population of participants do not carry that risk in a way that would justify bypassing consultation. If the government wishes to rely on the exception it must say so and justify it. It has not, so the ordinary standard applies, and on the ordinary standard the consultation is inadequate.

Section Heading

The Justice and Equity Centre records that the approach “has attracted sustained and vocal criticism from the disability community.” A best practice Impact Analysis summarises this opposition honestly and explains what, if anything, in the design changed as a result. The document should record the main contrary positions, including that the Bill shifts power from Parliament to the minister, that it converts a rights-based scheme into a discretionary one, and that alternative systems do not yet exist, and should state the government’s response to each.

• e Social, community and civic participation and capacity building reset Marking. Not Completed. The reset of social, community and civic participation and capacity building daily activity budgets from 1 October 2026 is described as an administrative change, but its distributional effect is not consulted on or analysed. Insertion. This reset is one of the largest single effects in the reform and needs its own consultation and distributional treatment. On the Impact Analysis’s own figures it affects 393,401 people. Resetting these budgets before the replacement supports exist is the clearest example of the “cut first, design later” concern, and the people affected were not consulted before the decision.

• e Fraud Marking. Not Completed. The consultation section relies on a fraud narrative to justify the urgency and confidentiality of the reforms, but the fraud figures are overstated and are corrected in the executive summary. Insertion. Fraud does not carry the weight placed on it. As set out in the executive summary, minister set pricing and fraud measures together account for about 2.4 per cent of the four year savings. The Fraud Fusion Taskforce reported 72 active fraud cases and $34.5 million in suspected fraud at 30 June 2024, about 0.09 per cent of scheme spending. Disability Intermediaries Australia reports that plan managers, whom the reforms replace, stopped $270.5 million in incorrect, duplicate and suspect invoices in a single member survey, returning $5.61 for every $1 in fees, about 7.6 times the return of the Agency’s own fraud detection program. Removing that layer to fight fraud is likely to reduce fraud prevention, not

increase it. The fraud narrative cannot bear the confidentiality and urgency claims it is being used to support.

Question 6. What is the best option and how will it be implemented?

The maps to the Impact Analysis section on the preferred options and implementation.

6.1 Choose the option with the greatest net benefit, and justify the choice against the analysis What the guidance requires. Best practice selects the option with the greatest net benefit to the community as shown by the analysis and explains the choice by reference to that analysis. Marking. Not Completed. A preferred option is stated for each reform area, but because there is no net benefit analysis under Question 4, the choice cannot be, and is not, justified against one. Insertion. The preferred options are stated without the analysis that is meant to support them. The document prefers option 2 for eligibility and supports, option 2 for plan management, and option 3 for support coordination. A best practice Impact Analysis would show that each preferred option has the highest net present value of the options considered. Because Question 4 contains no net present value for any option, the preference is asserted rather than demonstrated. Once the transferred costs, the mortality and safety effects, the regulatory burden on small providers, the distributional effects and the foregone economic contribution are added, it is not established that the preferred options carry the greatest net benefit, and it is possible that a less restrictive option, such as improved administration and decision quality inside the Agency, carries a higher one.

6.2 Set out an implementation plan with clear responsibilities and timing

What the guidance requires. Best practice sets out who does what and by when, and shows that the entities responsible can deliver.

The published assessment noted that this Impact Analysis “would have benefitted from … a more detailed implementation and evaluation plan.”

as “a significant expansion of ADM use in the NDIS without the necessary parliamentary scrutiny,” and which many contributors link directly to Robodebt; and the risk that debts are raised automatically on a failure to keep records, which the Justice and Equity Centre notes has no equivalent elsewhere in Australian law and no clear review path. A best practice risk section would rate each of these and set out a management plan. The document does neither.

Question 7. How will the reform be monitored and evaluated? This maps to the Impact Analysis section on evaluation.

7.1 Set out a monitoring and evaluation plan tied to the objectives What the guidance requires. Best practice sets out how the reform will be monitored and evaluated, ties the evaluation to the measurable objectives from Question 2, states what data will be collected, states who is responsible, and sets a review date. This is consistent with the Commonwealth Evaluation Policy and the Australian Centre for Evaluation guidance. The published assessment noted that this Impact Analysis “would have benefitted from … a more detailed implementation and evaluation plan.” Marking. Not Completed. The evaluation content is thin, is not tied to measurable targets, and does not describe how the effects on people who lose supports will be tracked. Insertion. The evaluation cannot work without the measurable objective it is missing. Because Question 2 does not set a measurable target, there is nothing for the evaluation to test against. A best practice plan would state, for example, that the Agency will track the proportion of exited participants who obtain equivalent support elsewhere, the rate of unmet need among remaining participants, the rate of automated decisions later overturned on review, and the death rate in Supported Independent Living, and would set a firm date for an independent review with the data to be published. None of these measures is committed to. In particular, the reform removes supports from a large group and assumes other systems will absorb them, yet there is no plan to collect data on what happens to those people. That is the single most important thing to measure and it is not measured.

                                                   Page 25

7.2 Commit to publishing the data and to correcting course

What the guidance requires. Best practice commits to publishing the monitoring data and to acting on it, including reversing or adjusting the reform if the evidence shows it is causing harm.

Marking. Not Completed. There is no commitment to publish the underlying modelling or the monitoring data, and no course correction mechanism.

Insertion. Transparency after the fact is as absent as transparency before it. People with Disability Australia asks the government to “publish detailed, transparent economic modelling.” A best practice evaluation plan would commit to publishing that modelling, to publishing the monitoring data annually, and to a defined trigger for pausing or adjusting the reforms if the data shows people are being harmed. The document contains no such commitment. 55

Overall assessment

Against the seven questions, this Impact Analysis meets the adequate standard on the questions of policy problem, objectives and options, and fails to complete the most important question, the likely net benefit, along with consultation, implementation and evaluation. The published rating of “Adequate,” the second lowest on the four point scale, is consistent with that pattern. The document reaches the adequate bar by describing costs and benefits in words. It does not reach a higher bar because it does not quantify the net benefit, does not value the mortality and safety effects the guidance requires it to value, does not measure the regulatory burden of its largest reform, does not analyse how the costs fall across cohorts, does not assess the competition effects of moving to a commissioned panel, does not consult before deciding, and does not set a measurable target that an evaluation could test. Two of its factual foundations, the weight given to fraud and the treatment of committed supports as if they were payments, are incorrect and are corrected above.

Appendix 1. NDIS Review recommendations contraindicated

by the 2026 Amendment Bill

The NDIS Review, titled Working Together to Deliver the NDIS, reported in December 2023 with 26 recommendations and 139 supporting actions. The government presents the Bill as implementing the Review. The table below shows where the Bill narrows, contradicts or omits what the Review recommended. Source references are to Bill sections (S) and to submissions (SUB) and parliamentary material (APH) before the Senate Community Affairs Legislation Committee.

Foundational supports

Cuts scheduled before supports Premature or contradicted S43, S30, S63

supports operational before NDIS eligibility is tightened operational or independently evaluated; Queensland out

Rec 7 Navigator role integrated with the NDIS Navigators not in the Bill Omitted S46

Rec 9 Investment in social inclusion and community participation funding 50 per cent cut to social, community and civic participation funding Contradicted S43, S44, S54; SUB-318

Recs 12 to 14 Direct commissioning of support coordination and plan management sits in access, planning and rules rather than as a discrete schedule Commissioning model adopted; with narrowing Implemented, APH-6

Rec 15 Housing and living supports Supported Independent implemented Supported Living commissioning announced; details deferred Independent Living commissioning

Recs 18 to 21 Fraud and integrity framework Schedule 2 powers plus mandatory registration Implemented APH-6; S14

Rec 22 Risk proportionate registration extended to high risk categories Mandatory registration Partially implemented APH-6

Rec 26 Joint Commonwealth state and territory governance architecture Commonwealth can set rules without state or territory consent; Schedule 3 governance

Appendix 2. Disability Royal Commission recommendations

contraindicated by the 2026 Amendment Bill

The Royal Commission into Violence, Abuse, Neglect and Exploitation of People with Disability reported on 29 September 2023 with 222 recommendations. The Commonwealth response was released on 31 July 2024. The table below shows where the Bill runs against those recommendations.

• Rec 12.1
  • Co design of
  • Disability
  • Not met
  • S42, S48; `SUB-
  • NDIS reforms
  • representative
  • organisations and
  • preferred
  • submitters not
  • consulted in
  • drafting;
  • 11
  • business day
  • window

Appendix 3. Bill coverage matrix

Does the Impact Analysis assess everything the Bill does? What the guidance requires. An Impact Analysis must assess the impact of the regulatory change it accompanies. Where a Bill makes many changes, the analysis should cover the changes that have a real effect on people, business or the community, not only the few that are politically prominent. Marking. Not Completed. The Bill amends the National Disability Insurance Scheme Act across 366 separate items. The Impact Analysis assesses a small share of them. On a clause by clause review, 37 items are assessed in the document, 25 are mentioned but their impact is not measured, and 304 are not assessed at all. Insertion. The coverage is far narrower than the Bill. Grouped by subject, the 366 items are as follows, with the count that carries serious impacts on participants or providers noted.

Page 31

Review and appeal rights

8 Not assessed; narrows merits review

Automated decision making

6 Not assessed; expands computerised decisions, no transparency, contradicts Robodebt Royal Commission recommendations and new Commonwealth Ombudsman recommendations

Plan management

3 Assessed as a reform option, does not detail job or business loss impact

Support coordination

1 Assessed as a reform option, does not detail job or business loss impact

The most significant unassessed provisions are drawn together in Appendix 3A. They are not minor. They include the power to raise debts and to garnishee, the power to make computerised decisions, the extension of the reassessment decision period and the narrowing of review rights, and the information gathering and privacy powers. They contradict the recommendations of the Robodebt Royal Commission, and raise issues about potential exposure to litigation, as do the recent issues around the Aged Care Legislation. Each of these changes the position of a participant in a way the guidance would expect an Impact Analysis to assess.

Appendix 3A. Significant provisions not assessed in the Impact Analysis

The clause by clause review identified 112 provisions with a material effect on participants or providers that the Impact Analysis does not assess. Grouped by effect:

Category of effect Number of provisions Why it matters

Information gathering and 25 New powers to require, share privacy and match information about participants Registration and provider 18 New registration obligations regulation and costs, including on small providers

Debt, garnishee and recovery

16

Debts may be raised, including on a failure to keep records, with garnishee powers and unclear review paths

Reassessment and plan

14

Reassessment decision period variation extended; variation powers widened

Functional capacity and access

11

The new “without assistive technology or supports” test that drives exits

Compliance, offences and penalties

12

New offences and penalties

Funding reductions and plan budgets

7

The budget reset and category cuts

Automated or computerised decisions

6

Decisions may be made by a computer program, raising the Robodebt risk

Review and appeal rights

3

Merits review narrowed, including by repealing section 48(4)

The Justice and Equity Centre highlights several of these directly. On debt: “We are not aware of any other framework in Australian law that automatically gives rise to a debt upon a failure to keep records,” and “there is currently no process under the NDIS Act for a participant to seek review of that decision.” On reassessment: the Bill “extends the timeframe for the NDIA to decide a reassessment request from 21 to 90 days and, by repealing section 48(4), limits participants’ rights to review.” On automated decisions: “a significant expansion of ADM use in the NDIS without the necessary parliamentary scrutiny.” Each of these belongs in the Impact Analysis and is absent.

Page 33

Appendix 4. Reviews and inquiries the Impact Analysis should have addressed

What the guidance requires. An Impact Analysis should draw on the relevant evidence base, including major reviews and inquiries that bear on the problem and the options. Best practice states what each found and how the reform responds to it.

Marking. Not Completed. Several major reviews and inquiries bear directly on these reforms and are not engaged with in the document.

Insertion. (At the very least), the following should have been fully addressed.

Review or inquiry Date What it found, in Why it belongs in

                                                  brief                     this Impact Analysis

NDIS Review, Working December 2023 26 recommendations The Bill is presented together to Deliver the and 139 actions on as implementing it; The NDIS eligibility, planning, Appendix 1 shows foundational supports where it does not. and provider regulation Recommendation 6 requires foundational supports to be operational before eligibility is tightened

Royal Commission into 29 September 2023; 222 recommendations The reforms run Violence, Abuse, government response on safety, inclusion against several Neglect and 31 July 2024 and independent living recommendations, set Exploitation of People out in Appendix 2. The with Disability Commission documented the harms of the pre-NDIS system that submissions warn the reforms may return to

Royal Commission into 7 July 2023 56 recommendations; The Bill expands the Robodebt Scheme found automated debt automated decision raising caused serious making and automatic harm debt raising, which is the mechanism Robodebt condemned. People with Disability Australia draws the link directly

Senate Community Affairs Legislation Committee inquiry into the Bill

Australian National Audit Office reports on NDIS fraud and compliance (2024–25) Found the Agency and the Quality and Safeguards Commission only “partly effective” at monitoring compliance. Bears directly on the fraud narrative used to justify urgency, and on whether the integrity problem is one of law or of administration.

Two points follow. First, the NDIS Review recommendation that foundational supports be operational before eligibility is tightened is the pivot of the whole reform, and the reform does the opposite. Second, the Robodebt Royal Commission is the most recent Australian finding on exactly the mechanism the Bill expands, and it is not mentioned. 57 58 59 60 61

A note on method This guide marks the Impact Analysis against the seven questions in the Office of Impact Analysis assessment tool and against the separate guidance notes the Office publishes on cost benefit analysis, regulatory burden, distributional analysis, competition, valuing statistical life and health, consultation, and risk. Every correction in red is sourced to a primary document: the tabled Treasury modelling, the National Disability Insurance Agency quarterly data, the Australian Government Actuary, Senate Estimates, the parliamentary submissions of disability organisations, and the government’s own published material.

57Independent Review of the NDIS, Working Together to Deliver the NDIS, final report, December 2023. https://www.ndisreview.gov.au/resources/reports/working-together-deliver-ndis/ 58Royal Commission into Violence, Abuse, Neglect and Exploitation of People with Disability, final report 29 September 2023; Australian Government response 31 July 2024. https://www.health.gov.au/our-work/disability-royal-commission-response 59Royal Commission into the Robodebt Scheme, final report, 7 July 2023. https://robodebt.royalcommission.gov.au/publications/report

Figures drawn from commissioned or think tank work are marked as such. The purpose is to show what a complete and accurate Impact Analysis of this Bill would have contained and demonstrate further issues with the government’s impact assessment to OIA on the proposed NDIS reforms, which constitute an argument for reassessment, especially post community consultation and during the current NDIS Bill inquiry. Prepared by Samantha Connor AM, Disabled People Against Cuts.