Canada runs the largest federal research and development support program in the country through the tax system, has just expanded it more than at any time in over a decade, and continues to record business investment and productivity outcomes its own central bank has called an emergency. Those facts sit together uncomfortably, and the discomfort is the subject of this article.
Key Takeaway
For the year ended March 31, 2026 the CRA reported $4.9 billion in SR&ED investment tax credits claimed and $4.6 billion allowed, continuing a climb of roughly 50% in credit value since fiscal 2022, and in fiscal 2024-25 the program delivered $4.5 billion to more than 19,000 claimants across over 22,000 claims. Bill C-15 received Royal Assent March 26, 2026, doubling the enhanced expenditure limit from $3 million to $6 million, raising phase-out thresholds from $10-50 million to $15-75 million of taxable capital, restoring capital expenditure eligibility and extending the refundable credit to eligible Canadian public corporations. Arrow's classic analysis of inappropriability justifies subsidising R&D because private actors cannot capture the full social return. Public choice theory explains why a justified subsidy's design drifts toward what is administrable and what organised claimants prefer, rather than toward the spillovers that justified it. The consultations' interest in extending SR&ED to commercialisation activity is where those two frameworks collide most directly.
Two Corrections Before The Argument
This article began as an examination of SR&ED being eroded under an expanding minimum tax regime. Two premises did not survive verification, and we state the corrections openly because the corrected picture is analytically richer than the original.
The Alternative Minimum Tax does not apply to corporations. Canada's AMT operates on individuals and certain trusts. SR&ED's refundable investment tax credit accrues to Canadian-controlled private corporations and, since Bill C-15, eligible Canadian public corporations. We found no mechanism by which the 2024 AMT changes reduce a CCPC's refundable SR&ED credit, and none of the professional commentary reviewed for this article connects them. An AMT interaction can arise for a founder personally, principally on a share disposition interacting with the lifetime capital gains exemption, which this publication addresses elsewhere, but that is a shareholder-level event rather than an erosion of the corporate credit.
The credit is expanding, not eroding. The evidence below shows record claim volumes, roughly 50% growth in credit value since fiscal 2022, and the largest legislative expansion in more than a decade. Any critique of SR&ED must begin from generosity, not austerity.
That reframing produces a better question. If Canada's principal R&D subsidy is at record size and has just been substantially enlarged, and Canadian business investment and productivity remain weak, the interesting issue is not how much is spent but whether the design directs it at the market failure that justifies spending at all.
The Scale Of The Program
The numbers first, because the scale is larger than most business owners assume.
In fiscal year 2024 to 2025, the SR&ED program delivered $4.5 billion in investment tax credits to more than 19,000 claimants through over 22,000 claims processed, and the Canada Revenue Agency describes SR&ED as the largest Government of Canada program supporting research and development in the country. For the year ended March 31, 2026, the CRA reported $4.9 billion in investment tax credits claimed and $4.6 billion allowed, continuing a steady climb of roughly 50% in credit value since fiscal 2022[1].
Finance's own characterisation is consistent: the SR&ED program is Canada's single largest support program for R&D and provides tax incentives to more than 20,000 businesses that conduct R&D[2].
Two structural facts frame everything that follows. The program operates on a two-tier rate structure: the enhanced 35% rate is fully refundable and applies to qualified expenditures of eligible CCPCs, and now eligible Canadian public corporations, up to an annual expenditure limit, while the basic 15% rate is generally non-refundable and applies to others[1]. And refundability is the feature that matters: for most CCPCs the credit is refundable, meaning the government pays it out in cash even when the company owes no tax, and that cash-back feature is what makes SR&ED a genuine funding source for early-stage and pre-profit innovators rather than merely a tax deduction[1].
The gap between $4.9 billion claimed and $4.6 billion allowed deserves note in passing. Roughly six percent of claimed value was not allowed, which is a meaningful denial rate for a program whose administrative burden is already the subject of complaint.
The Expansion, Not The Erosion
The legislative history, which establishes the direction of travel.
Finance first announced it would review the SR&ED program in the 2022 federal budget, and had last made significant changes in 2012[2]. Two consultations were launched on January 31, 2024, followed by a further consultation paper released April 25, 2024 seeking feedback on the best use of additional funding allocated in the 2024 federal budget, with comments due May 27, 2024[2]. A second consultation centred on deploying a proposed $600 million investment over four years to modernise and expand the program, and these consultations culminated in proposals announced in the December 2024 Fall Economic Statement with draft legislation released in August 2025[3].
Those proposals were scrapped following the Prime Minister's resignation, and Budget 2025 revived them with additions[3]. On March 26, 2026 the Department of Finance announced legislative approval confirming significant changes as Bill C-15 received Royal Assent, with the stated aims of supercharging investment in R&D, simplifying administration and cutting the claim processing period in half[4].
The substance: the annual enhanced expenditure limit for an associated group of CCPCs rose from $3 million to $6 million, taking maximum refundable credits to $2.1 million from $1.05 million; the taxable capital threshold at which the refundable ITC begins to reduce increased to $15 million from $10 million, and the limit at which it is eliminated rose to $75 million from $50 million[4]. Capital expenditures became eligible again for property acquired after December 15, 2024[5], described elsewhere as eligible for the first time since 2012[6].
And the refundable credit was extended beyond private companies. Previously, access to refundable federal credits up to the expenditure limit was limited to CCPCs, which meant a company going public and seeking investment had to weigh the cost of losing access to this non-diluted capital against gaining the capital it needed to expand; Bill C-15 allows eligible Canadian public corporations to retain access, provided they reside in Canada and are controlled by Canadian residents while listing a class of shares on designated stock exchanges[7].
That last change is genuinely well-targeted and worth crediting. It removes a distortion that penalised a specific, economically valuable decision.
Arrow: Why The Subsidy Is Justified At All
Before critiquing design, the case for the program, which is one of the more robust results in economics.
Kenneth Arrow's 1962 analysis of the allocation of resources for invention established the canonical argument for public support of research. We cite it as a classic work from our own knowledge of the literature rather than from a source retrieved for this article, and readers should consult the original.
Arrow identified three properties of information as an economic good that together produce systematic under-investment by private firms. Information is indivisible, in that the cost of producing it does not scale with the number of users. It is inappropriable, in that once produced it is difficult to prevent others from using it, so the producer captures only part of the value created. And its production is characterised by fundamental uncertainty, in that the output of a research project cannot be known in advance, which makes it difficult to finance through ordinary capital markets.
The inappropriability property carries the argument. If a firm undertaking research captures only a fraction of the social value its discoveries generate, because competitors, suppliers, customers and departing employees absorb the knowledge, then the firm will invest less than is socially optimal. The gap between private and social return is the spillover, and it is the entire justification for subsidy: the state pays for the part of the value the firm cannot capture.
Two implications follow that do work later in this article. The justification is strongest where spillovers are largest, meaning where knowledge escapes the firm most readily. And it weakens as appropriability rises, because research whose returns the firm captures fully needs no subsidy, the private incentive already being aligned with the social one.
Arrow's uncertainty property does independent work, and it maps onto the credit rationing analysis in this publication's treatment of asset-light lending. A research project whose outcome cannot be characterised in advance is precisely the asset a lender cannot underwrite, which is why refundability matters so much: it substitutes for finance that markets will not supply.
The Appropriability Test
Applying Arrow's criterion to SR&ED as designed yields a mixed verdict, which we state as our own analysis.
SR&ED is defined by reference to subsection 248(1) of the Income Tax Act and splits into basic research, applied research and experimental development. Almost all owner-manager claims are for experimental development, the smallest and most accessible of the three categories[8].
Rank those three on Arrow's appropriability criterion and the ordering is clear. Basic research generates the largest spillovers, because it produces general knowledge with no immediate proprietary application and the firm conducting it captures the least. Applied research sits in between. Experimental development, which resolves a specific technological uncertainty in a firm's own process or product, generates the smallest spillovers, because the resulting knowledge is embedded in that firm's operations and is often least visible to competitors.
So the category where the externality argument is weakest is the category that dominates claims. That is not evidence of abuse; experimental development is legitimate SR&ED and the statute says so. But it means the program's realised spending concentrates where Arrow's justification is thinnest, and a policy designed to maximise spillovers would look different from one designed to be claimable.
We would resist the strong version of this critique. Firm-level process innovation has real diffusion effects through labour mobility, supplier relationships and imitation, and the boundary between development and applied research is genuinely blurry in practice. The honest conclusion is that SR&ED subsidises a spectrum of activity with widely varying spillover content at a uniform rate, which is a design choice carrying a cost.
Public Choice: Why The Design Drifts
The framework that explains the gap between the justification and the design.
Public choice theory, associated principally with Buchanan and Tullock and with Olson's analysis of collective action, applies the assumption of self-interested behaviour to political and administrative actors rather than treating them as disinterested maximisers of social welfare. We cite this literature from our own knowledge as a body of canonical work rather than from retrieved sources.
Two of its results bear directly on SR&ED. The first is the asymmetry of concentrated benefits and diffuse costs: a policy conferring large benefits on an identifiable group while spreading its cost thinly across taxpayers generates intense organised advocacy on one side and near-silence on the other. The beneficiaries of a $4.9 billion credit know precisely what it is worth to them; the individual taxpayer funding it does not know the program exists.
The second is that program design responds to the preferences of those who participate in consultations, and participation is itself a function of stake. Finance's consultations sought feedback from stakeholders on potential changes[2], which is proper process. But the stakeholders with the resources and motivation to make detailed submissions on the design of an R&D tax credit are predominantly claimants, their advisors, and industry associations representing claimants. Constituencies who would benefit from spillovers they cannot identify in advance do not make submissions.
The prediction that follows is not corruption but drift: over time a program's design moves toward what its organised constituency prefers, which is greater generosity, broader eligibility and easier claiming, and away from features that maximise spillovers but constrain claims. Every change enacted in Bill C-15 increases generosity or breadth. None narrows eligibility toward higher-spillover activity.
That is a description rather than an accusation. Each individual change is defensible on its own terms and several are good policy. The public choice point concerns the systematic direction of a sequence of individually reasonable changes.
The Commercialization Question
The proposal where the two frameworks collide most sharply, and the most important item in the consultation record.
Finance's April 2024 consultation asked stakeholders to explore potential changes proposed by stakeholders, including allowing small and medium-sized public corporations to access the enhanced SR&ED refundable tax credit, making the general SR&ED credit refundable, and expanding the SR&ED definition to allow for some aspects of commercialisation activity[2].
The first was enacted. The third is the one worth examining, and our assessment is that it runs against the intellectual foundation of the program.
Commercialisation is the process by which a firm converts knowledge into captured revenue. It is, definitionally, the activity through which appropriability is achieved. Arrow's argument for subsidy rests on inappropriability: the state funds the portion of value the firm cannot capture. Subsidising the firm's capture of value inverts that logic, because the private return to successful commercialisation is precisely what motivates firms to pursue it without assistance.
There are serious counterarguments and they deserve statement. Canada's documented weakness is frequently characterised as strong research output with poor conversion into scaled domestic firms, so a targeted subsidy at the conversion stage addresses the binding constraint rather than the theoretically purest one. Commercialisation is also capital-intensive and faces the same financing gap Arrow's uncertainty property predicts. And a policy maximising measured spillovers while producing no Canadian firms of scale has arguably failed on the terms that matter politically.
Our view is that both positions are coherent and that choosing between them is a genuine policy judgment rather than a technical question. What we would insist on is that it be recognised as a change of objective. Extending SR&ED to commercialisation converts a market-failure correction into an industrial subsidy. That may be the right decision, and it should be defended as such rather than presented as a continuation of the existing rationale.
The Industry The Complexity Built
A public choice observation about the intermediary layer, offered as analysis.
SR&ED has generated a substantial advisory industry, and the advisory literature is candid about its value proposition. One firm puts it directly: the companies that capture the full benefit will not be the ones with the most R&D activity, but the ones with advisory teams that know how to find it, document it and defend it, and an integrated CPA and engineering advisory team consistently identifies eligible work that financial-only advisors miss[6].
Read that as an economic statement rather than a marketing one and it is striking. It says the determinant of benefit is advisory capability rather than research activity. If accurate, the program partly allocates public funds according to which firms bought better help, which is not a spillover-maximising allocation rule.
The public choice implication concerns incentives around simplification. An intermediary industry whose revenue derives from navigating complexity has no interest in complexity being removed. Bill C-15's stated aims include simplifying administration and halving processing time[4], which is welcome, and it is worth observing that the constituency best organised to comment on how simplification should be implemented is the one that would lose most from it succeeding.
We should be fair here. Competent SR&ED advisors perform genuine work, identifying eligible activity that engineers do not recognise as claimable and building documentation that withstands review. The critique targets a program design that makes such intermediation necessary, not the intermediaries responding rationally to it. And this publication's own commercial interests sit adjacent to that observation, which readers should weigh.
Refundability And Who It Serves
The feature most worth defending on Arrow's terms, and the consultation proposal that would have extended it.
The non-refundable credit's value depends entirely on tax payable: any unused portion is available up to 20 years later to reduce tax payable in that year[9]. For a pre-profit company that carryforward is close to worthless in present-value terms, which is why refundability converts the credit from a deferred tax attribute into working capital.
The cash economics for a claiming CCPC are material. A $400,000 R&D salary spend generates approximately $140,000 in federal credit at 35%, with the provincial credit and proxy method overhead commonly producing a combined refund in the $175,000 to $210,000 range, arriving roughly six to nine months after fiscal year end; for owner-managed CCPCs with genuine technical staff, the refundable federal credit plus provincial stack typically returns $0.40 to $0.50 of cash per dollar of qualifying expenditure[8].
This is where Arrow's uncertainty property and refundability connect. A research project's output cannot be characterised in advance, so it is exactly the activity private credit markets will not finance, a problem this publication examines in the context of asset-light lending. A refundable credit supplies capital that markets decline to supply, for a reason Arrow identified in 1962. On that logic, refundability is the single most theoretically defensible feature of the program.
Which makes the consultation proposal to make the general SR&ED credit refundable[2] more interesting than it first appears. It was not enacted. Extending refundability to large and foreign-controlled corporations would be expensive and would direct cash to firms with access to capital markets, so the case is weaker on financing-gap grounds than for small CCPCs. But it would also remove a distortion under which identical research attracts different support depending on the ownership structure of the entity conducting it, which is not a distinction Arrow's framework recognises.
The Productivity Puzzle
The uncomfortable juxtaposition this article exists to state plainly.
SR&ED credit value has grown roughly 50% since fiscal 2022 and reached $4.9 billion claimed for the year ended March 31, 2026[1]. Over a broadly comparable period, as this publication has documented, the Bank of Canada's Senior Deputy Governor described Canadian productivity as an emergency, noting business sector productivity was more or less unchanged over seven years, and C.D. Howe subsequently reported real machinery and equipment investment falling further.
We are careful about what this does and does not establish. It is not evidence that SR&ED does not work, for at least three reasons: the counterfactual is unobservable and R&D intensity might have fallen further without it; the lag between research spending and productivity effects is long and variable; and aggregate productivity is driven by many factors, of which R&D subsidy is one.
But it does place a burden of explanation. A program of this scale, growing at this rate, in a country with this productivity record, has to answer why the relationship is not more visible. The three candidate explanations below are the ones we consider most serious.
Three Explanations Worth Weighing
Additionality is lower than assumed. A subsidy generates social value only to the extent it causes research that would not otherwise occur. Where it funds activity a firm would have undertaken regardless, it is a transfer rather than an incentive. Because claims are assessed after the fact against work already performed, and because the advisory value proposition is explicitly about identifying eligible work already embedded in operations[6], the program's structure does not distinguish induced research from reclassified research. This is our inference rather than a measured finding, and additionality is genuinely difficult to estimate.
The composition is wrong for spillovers. As argued above, claims concentrate in experimental development, the category where appropriability is highest and spillovers smallest. A uniform rate across a spectrum of spillover content directs the most money to the least externality-generating activity, because that activity is the most common.
The binding constraint is elsewhere. R&D subsidy addresses the cost of research. If Canada's constraint is the scaling of firms that already innovate, through market access, growth capital or management capability, then subsidising research more heavily relieves a constraint that was not binding. This is the strongest argument for the commercialisation extension and, we think, the most likely of the three to be substantially correct.
These are not mutually exclusive, and none is a case for cutting the program. They are arguments about design, and the honest position is that the empirical work required to distinguish between them is not something this article can perform.
The Audit Turn
From macro critique to the operational reality facing a claimant, which has shifted materially.
Eligibility turns on whether the work is scientific or technological in nature and whether it meets the SR&ED criteria of experimental development, applied research or basic research as defined in subsection 248(1) of the Income Tax Act. The CRA's modern administrative position, following Northwest Hydraulic Consultants, CW Agencies and the 2021 SR&ED Eligibility Guidelines, tests experimental development through a five-question framework[8].
The critical administrative constraint is the filing deadline: 18 months after the corporation's tax year-end, so for a December 31, 2024 year-end the deadline is June 30, 2026, and the CRA does not grant extensions to this deadline except in narrow statutory cases[9]. This is a hard bar. A meritorious claim filed late is generally lost regardless of quality, which makes the deadline the first thing to verify in any SR&ED discussion.
And the enforcement emphasis has moved. One advisory source states that the CRA has significantly increased scrutiny on AI and machine learning claims, and that documentation quality is now the primary audit trigger[6]. We report that as a practitioner observation rather than an official CRA position, and it is consistent with a program simultaneously expanding its generosity and, per Bill C-15's stated aims, seeking to halve processing times[4]. Faster processing alongside larger credits implies review effort concentrating where risk is highest rather than spreading uniformly.
The AI Documentation Problem
Why machine learning work attracts scrutiny, and what that implies for documentation. This section applies the eligibility framework and is our analysis rather than a report of CRA policy.
SR&ED requires technological uncertainty that cannot be resolved by routine engineering or standard practice, and systematic investigation through experiment or analysis to resolve it. Much AI and machine learning work in Canadian businesses, however commercially valuable and technically demanding, does not meet that description: selecting an established architecture, fine-tuning on proprietary data, tuning hyperparameters and integrating an API are activities where the technological approach is known and the uncertainty concerns outcomes rather than method.
The distinction that matters is between uncertainty about whether a known approach will work on our data, which is ordinary applied engineering, and uncertainty about whether any known approach can achieve the required result, which may be technological uncertainty. The first is a business risk. The second is the statutory concept.
This creates a documentation problem specific to machine learning work. The evidence distinguishing the two lives in artefacts teams generate anyway and rarely retain in usable form: experiment tracking showing hypotheses tested and rejected, records of approaches abandoned and why, baseline comparisons against standard practice, and contemporaneous reasoning about why known methods were insufficient. A team that retained only the successful final model has destroyed the record of the systematic investigation that would have supported the claim.
The practical instruction is that failed experiments are the evidence. A claim narrative describing a smooth path to a working model describes routine development. One documenting hypotheses, failures and iteration describes experimentation, and the second is both more claimable and, usually, more truthful.
The Pre-Claim Approval Programme
A material administrative development that changes the risk profile of a claim.
Starting April 2026, the CRA offers a voluntary pre-claim approval program under which companies can submit project descriptions for upfront technical approval[6].
The significance is that it moves the eligibility determination from after the expenditure to before it. Under the conventional sequence a company spends, claims, and discovers on review whether the CRA agrees the work qualified, which means eligibility risk is borne after the money is committed. Upfront approval reverses that ordering for participating projects.
For a company whose R&D decisions are sensitive to the credit, that is a genuine improvement in decision quality: knowing before committing whether a project is likely to qualify converts a contingent recovery into a plannable one. It also, on Arrow's logic, should improve additionality, because a subsidy can only induce research if its availability is known at the point the research decision is made. A credit discovered afterward cannot have caused the spending.
We have not established the program's scope, eligibility conditions, timelines, or whether an approval binds the CRA on subsequent review, all of which determine how useful it actually is. A company considering it should obtain current details directly from the CRA.
Documentation Architecture
The operational bridge, structured around what the eligibility test actually asks.
Capture uncertainty at the outset, not the outcome at the end. The claimable event is the existence of a technological uncertainty that could not be resolved by standard practice. That has to be recorded when identified, because reconstructing it after resolution is both unconvincing and difficult.
Retain the failures. Abandoned approaches, negative results and rejected hypotheses are the evidence of systematic investigation. Version control history, experiment logs and issue trackers usually contain this already; the discipline is retention and retrievability rather than new record-keeping.
Record why standard practice was insufficient. The test asks whether the uncertainty could be resolved by routine engineering. A contemporaneous note explaining which known approaches were considered and why they were inadequate addresses that question directly.
Separate technical narrative from financial claim preparation. The technical record must be generated contemporaneously by the people doing the work. A narrative written months later by someone who was not present is the profile most exposed on review.
Track time to project, not to cost centre. Salary allocation to specific SR&ED projects is a frequent point of failure, and general engineering time codes will not support an allocation.
Diarise the 18-month deadline institutionally. It is hard, extensions are narrow, and it runs from tax year-end rather than from any project milestone.
Reconcile against the enacted expansion. Given Bill C-15's effective dates, a claim prepared before enactment may have applied a superseded limit, which this publication addresses in detail elsewhere.
A Worked Case: Two Identical Projects
Two Canadian companies undertake technically comparable machine learning work over the same period. The comparison is constructed to isolate the documentation variable rather than reported from specific engagements.
Company A assigns three engineers to improve a forecasting model. They try several architectures, discard two, encounter a data sparsity problem that no standard technique resolves, develop a modified approach, and ship. Their repository records only the final implementation; intermediate branches were deleted on merge, and nobody wrote down why the discarded approaches failed. At claim time, an advisor interviews the team eleven months later and constructs a narrative from recollection.
Company B does the same work with one difference: an engineer maintains a short running log recording each hypothesis, the result, and the reason for abandoning it, and the team retained baseline comparisons against the standard method. Nothing in this cost material time.
On review, Company A must persuade the CRA that technological uncertainty existed, using evidence generated after the uncertainty was resolved and reconstructed from memory. Company B produces contemporaneous evidence that hypotheses were formed, tested and rejected, which is the definition of systematic investigation.
The underlying research was identical. The claimable position was not, the difference cost roughly nothing to create, and it cannot be created retroactively. Given that documentation quality is reported as the primary audit trigger[6], this is the highest-return process change available to a claiming company.
What To Do
Verify the filing deadline first. Eighteen months from tax year-end, with extensions available only in narrow statutory cases.
Instrument the research process, not the claim process. Contemporaneous capture of uncertainty, failed approaches, and reasons standard practice was insufficient.
Treat failed experiments as assets. They are the evidence of systematic investigation, and deleting them destroys the claim's foundation.
Investigate the pre-claim approval program. Available from April 2026 and, if it suits your circumstances, it moves eligibility risk to before the spending decision.
Confirm which expenditure limit your last claim applied. Bill C-15's effective dates reach into completed and filed years.
Be honest internally about additionality. If the credit is not influencing what research you do, it is a recovery rather than an incentive, which is worth knowing when assessing what your advisory spend actually buys.
Read the advisory value proposition carefully. If benefit depends on advisory capability rather than research activity, that is a statement about the program worth factoring into how you resource the claim.
The Limits Of This Analysis
Several caveats matter. This article corrected two premises it began with, that AMT erodes CCPC SR&ED credits and that the program is contracting; the AMT correction reflects our understanding that Canada's AMT applies to individuals and trusts rather than corporations, which should be confirmed with a tax advisor for any specific situation. Arrow (1962) and the public choice literature are cited as canonical works from our own knowledge rather than from sources retrieved for this article, and our application of them to SR&ED is our own argument rather than a finding of that literature. The productivity juxtaposition establishes a burden of explanation, not causation, and we have not attempted the empirical work that would distinguish the three candidate explanations. Program statistics are reported from a secondary compilation citing CRA departmental reports rather than from those reports directly. The AI scrutiny observation and the pre-claim approval program details derive from a single advisory source and should be verified with the CRA. This article does not address provincial credits, the proxy versus traditional overhead methods, contract R&D and third-party payment rules, partnership allocations, or the interaction with government assistance. Nothing here is tax advice; engage a qualified Canadian tax professional before filing or amending.
Frequently Asked Questions
Does the Alternative Minimum Tax reduce my company's SR&ED credit?
Is SR&ED being cut back?
What is the economic case for the program?
Why is extending SR&ED to commercialisation contentious?
What is the biggest documentation risk right now?
What does the pre-claim approval program change?
References
- SRED.ca. (2026, June). The State of SR&ED 2026: Canada's R&D Tax Credit by the Numbers, on fiscal 2024-25 and year-ended March 31, 2026 statistics, the two-tier rate structure and refundability. Compiled from the CRA Departmental Results Report 2024-25, CRA Departmental Plan 2026-27, CRA SR&ED Annual Program Statistics, Budget 2025, the 2024 Fall Economic Statement and Bill C-15. Note: published by a SR&ED advisory business. sred.ca/state-of-sred
- KPMG Canada. (2024, April 26). SR&ED Consultation: Finance Asks for More Feedback, TaxNewsFlash-Canada No. 2024-20, on the April 25, 2024 consultation paper, the May 27, 2024 deadline, the proposals on public corporation access, general credit refundability and commercialisation, the 2022 budget review announcement and the 2012 baseline. assets.kpmg.com/.../ca-sred-consultation-finance-asks-for-more-feedback.pdf
- Mintz. (2025, November 13). Innovate, Baby, Innovate? Key Enhancements to Canada's SR&ED Program in Budget 2025, on the $600 million four-year modernisation consultation, the scrapping and revival of proposals, and the shared associated-group limit. mintz.com/insights-center/viewpoints/2906
- BDO Canada. (2026, April). SR&ED Tax Credit Enhancements and Updates, on the March 26, 2026 Royal Assent of Bill C-15, the stated aims of supercharging investment, simplifying administration and halving processing, and the revised limits and thresholds. bdo.ca/insights/sr-ed-program-enhancements-and-updates-draft-legislation-released
- SRJ Chartered Accountants. (2026, August). Canadian SR&ED Program: Complete 2026 Guide to R&D Tax Credits, on the rate structure, eligible expenditure categories and capital expenditures for property acquired after December 15, 2024. srjca.com/blog/comprehensive-guide-to-the-canadian-sred-program
- Henderson & Associates. (2026, May 8). SR&ED Tax Credit Advisory in Canada, on the advisory value proposition, increased CRA scrutiny of AI and machine learning claims, documentation as the primary audit trigger, capital expenditures eligible for the first time since 2012, and the April 2026 voluntary pre-claim approval program. Note: published by a SR&ED advisory firm. hendersonassociates.ca/sred-resources/sred-tax-credit-advisory-in-canada
- Welch LLP. (2026, April 9). 2026 Changes in SR&ED: Largest Expansion in Decades, on the August 15, 2025 draft proposals and the ECPC access change and its rationale. welchllp.com/insights/knowledge/2026-changes-in-sred-largest-expansion-in-decades
- Insight Accounting CPA. (2026, May). SR&ED Tax Credit in Canada (2026): Eligibility, Claim Process, and Real Refund Amounts, on subsection 248(1), the five-question test following Northwest Hydraulic Consultants, CW Agencies and the 2021 Eligibility Guidelines, the dominance of experimental development claims, and the cash economics of a claim. insightscpa.ca/sred-tax-credit-canada-2026-eligibility-claim-amounts
- SRJ Chartered Accountants. (2026). Canadian SR&ED Program Guide, on the 18-month filing deadline, the narrow extension cases and the 20-year carryforward of non-refundable credits. srjca.com/blog/comprehensive-guide-to-the-canadian-sred-program
- Arrow, K. J. (1962). Economic Welfare and the Allocation of Resources for Invention. In The Rate and Direction of Inventive Activity, Princeton University Press for the NBER. Cited as a canonical work from the economics literature rather than from a source retrieved for this article.
- Buchanan, J. M., & Tullock, G. (1962). The Calculus of Consent; Olson, M. (1965). The Logic of Collective Action. Cited as canonical works in the public choice literature from our own knowledge rather than from sources retrieved for this article.
This article discusses tax policy, program statistics and economic theory and is provided for general informational purposes. It is not tax advice. Program statistics derive from a secondary compilation of CRA reports. The AI scrutiny and pre-claim approval observations come from advisory commentary and should be verified with the CRA. Engage a qualified Canadian tax professional before filing or amending a claim.