GAAR at a Crossroads – Part III: The ‘Party’ Requirement, Knowledge, and the Objective Turn

By Dr Hendri Herbst & Dewald Pieterse

Introduction

In the first two instalments of this series, we traced the long arc of the South African general anti-avoidance rule (GAAR) and the jurisprudential fault line that the Absa Bank litigation had exposed. Part I described the broader transition from a regime tethered to the taxpayer’s subjective purpose towards one ostensibly governed by the objective structural logic and economic effects of an arrangement, while Part II dissected the mechanics of the ‘tax benefit’ enquiry and the battle between competing counterfactuals. Throughout, we cautioned that the prolonged judicial vacuum following the 2006 enactment of the new GAAR had allowed the South African Revenue Service’s (SARS) more aggressive, objective enforcement theories to develop largely unchecked, and we expressed a reasoned preference for the taxpayer-centric and narrower readings that prioritise legal certainty.

On 22 April 2026, the Constitutional Court delivered its judgment in Absa Bank Limited and Another v Commissioner for the South African Revenue Service [2026] ZACC 15 (Absa Bank). The appeal was dismissed with costs. By a majority of nine to one, the Court endorsed the more objective vision of the GAAR that SARS had advanced, both on the ‘party’ question that we had reserved for this final instalment and on the ‘tax benefit’ question examined in Part II. This article completes the series. It sets out the statutory ‘party’ problem, analyses the majority judgment of Majiedt J and the lone dissent of Rogers J, and then turns to what may prove the judgment’s most consequential and least-noticed feature: its apparent objectification of the ‘sole or main purpose’ test, which, remarkably, was not even a question that was before the Court. We close by considering what an objective purpose enquiry can coherently mean, and whether the once-distinct elements of the GAAR have begun to dissolve into a single, impressionistic standard.

The ‘party’ requirement: the statutory question

The GAAR does not act upon the world at large; it acts upon a ‘party’. Section 80B of the Income Tax Act 58 of 1962 (the Act) empowers the Commissioner to determine the tax consequences of an impermissible avoidance arrangement ‘for any party’, and section 80L defines a ‘party’ as a person who ‘participates or takes part in an arrangement’. The dispute turned on a deceptively simple question: can a taxpayer be a ‘party’ to steps of an arrangement of which it had no knowledge?

Absa’s position was that participation presupposes volition and conscious involvement. On this view, drawn from the ordinary meaning of ‘participates’ and ‘takes part’, one cannot meaningfully take part in something one does not know exists. SARS’s position was that knowledge is irrelevant to the ‘party’ enquiry: if a taxpayer provides the funding for, and derives the economic benefit of, an arrangement, it is a ‘party’ to that arrangement for GAAR purposes, even as to steps of which the taxpayer was wholly ignorant. As we observed in Part I, requiring proof of knowledge, SARS warned, would invite ‘deliberate ignorance’ and ‘wilful blindness’, allowing sophisticated participants to profit from avoidance while preserving plausible deniability. This was the clearest expression yet of the shift away from the taxpayer’s state of mind towards the objective architecture of the transaction.

The majority: participation without omniscience

Writing for the majority, Majiedt J (with eight members of the Court concurring) decisively preferred SARS’s reading. The Court held that the statutory language ‘participates or takes part’ admits of a purposive construction under which a ‘party’ encompasses taxpayers who engage in transactions that, viewed objectively, form part of an avoidance arrangement, ‘even if they lack sight of every internal mechanism’. In a phrase likely to be much quoted, the Court held that ‘[p]articipation does not require omniscience, but it does require a conscious step into the structure from which the avoidance benefit flows’. Rather than actual or constructive knowledge of the avoidance steps, what is required is an objective determination of whether the taxpayer’s conduct forms part of the causal chain constituting the arrangement.

The majority drew significant support from the legislature’s choice of words. The deliberate duplication of the synonyms ‘participates or takes part’ was said to underscore a focus on involvement rather than mental state. Decisively, the Court reasoned that had Parliament intended to impose a knowledge requirement, it could readily have adopted the familiar formulation ‘knows or ought reasonably to have known’ found in the Financial Intelligence Centre Act 38 of 2001 and the Prevention and Combating of Corrupt Activities Act 12 of 2004. Its failure to do so, coupled with its choice of the expansive phrase ‘any party’, was treated as a calculated design choice irreconcilable with a requirement of bilateral consensus or full prior knowledge. To insist on such knowledge, the Court held, would render upstream capital providers ‘structurally immune’ to the GAAR through ‘deliberate informational nescience’ and would recreate the very loopholes the 2006 amendment was purportedly enacted to close.

This did not mean that knowledge or volition were rendered wholly irrelevant. The majority accepted that they may feature later – in rebutting the presumption of purpose under section 80G, or in determining a taxpayer’s liability for understatement penalties – but not as a gateway to being a ‘party’. On the facts as they stood, Absa’s role was pivotal: it provided the funding, participated in the returns, and received substantially all of the tax benefits generated. Without its capital, the downstream avoidance could not have occurred. Absa was therefore ‘clearly a “party”’ to the impermissible avoidance arrangement.

In reaching this conclusion the majority drew on comparative authority, invoking the Supreme Court of Canada’s decision in Copthorne Holdings Ltd v Canada 2011 SCC 63; [2011] 3 SCR 721 (which applies that country’s GAAR to a series of transactions viewed as a whole and treats participation as an objective question) and the Australian experience under their GAAR, as recently applied in FCT v Guardian AIT Pty Ltd, ATF Australian Investment Trust [2023] FCAFC 3; (2023) 115 ATR 316 and in Minerva Financial Group Pty Ltd v Commissioner of Taxation [2024] FCAFC 28; (2024) 302 FCR 52. The comparison is instructive but imperfect, as the Australian GAAR rests on an expressly statutory, multi-factor objective ‘dominant purpose’ test in section 177D, whereas our section 80A is, on its face, a purpose test simpliciter. The borrowing thus arguably imports an objective framework that the South African legislature did not enact in the same explicit terms.

The ‘tax benefit’ issue resolved

Having found Absa to be a party, the majority turned to the ‘tax benefit’ question analysed in Part II, and resolved it equally firmly in SARS’s favour. The Court held that there is no legal requirement that the specific taxpayer against whom the GAAR is invoked must personally have secured the tax benefit. Section 80B’s authorisation to determine consequences for ‘any party’ evidences a deliberately wide remedial reach; had Parliament intended to confine liability to the party ‘obtaining a tax benefit’, it would have repeated that phrase from section 80G. The ‘transferable tax benefit’ theory, which lay at the heart of the litigation, was thus accepted in principle.

However, the majority went further, holding that on the objective facts Absa had in any event obtained a tax benefit. The Court endorsed the broader, ‘commercially equivalent’ counterfactual that we examined in Part II and that the Tax Court had embraced in Mr Taxpayer G v C:SARS  (IT 24502) 41 SATC 179. The correct ‘but-for’ test, the Court held, asks whether ‘but for the tax avoidant features and dressing up of the transaction, a tax liability would have occurred’ – not whether a liability would have arisen but for the transaction altogether. Stripped of its avoidance features, the arrangement reduced in substance to a loan by Absa, and Absa’s tax-exempt preference dividends were, in substance, the taxable interest yield of that loan. As Majiedt J put it, ‘if this transaction had not been dressed up with features designed to avoid the imposition of tax, Absa would have incurred the tax liability’. SARS was entitled to disregard the interposed conduits – PSIC3, PSIC4 and the D1 Trust – as special purpose vehicles through which the benefit was funnelled to the first substantive recipient, Absa. That power is rooted in the statutory machinery for stripping out artificial elements – the round-trip financing indicators in sections 80C and 80D, and the treatment of accommodating or tax-indifferent parties in sections 80E and 80F, which permit such entities to be disregarded or collapsed into another party.

The recharacterisation does, however, expose a doctrinal tension. The Act already contains dedicated provisions – sections 8E and 8EA – that recharacterise dividends on debt-like preference shares as taxable income. To treat an exempt preference dividend as ‘in substance’ interest under the GAAR, in a case falling outside those specific rules, risks allowing the general provision to achieve what the specific ones, by their carefully drawn limits, deliberately do not. It is at least arguable that preference shares structured to fall outside sections 8E and 8EA may now face attack under the GAAR on a substance-over-form basis – a prospect difficult to reconcile with the existence of the specific regime.

The dissent: knowledge, and the road not taken

Rogers J alone dissented, and would have upheld the appeal. On the ‘party’ issue, he returned to first principles of language. A ‘party’ is one who ‘participates or takes part’, and ‘[o]ne cannot participate or take part in an arrangement which one does not know exists’. He illustrated the point with the now-memorable analogy of a person who gives another a lift, unaware that the passenger intends to commit a murder at the destination: the driver is not a ‘party’ to the murder, however much the lift may have advanced it. Crucially, Rogers J distinguished this knowledge – the knowledge inherent in the very conduct of taking part – from the separate question of subjective purpose. One may accept that the purpose enquiry has become objective and still insist that the identity of the parties is fixed by who knew of, and intended to take part in, the arrangement.

On the tax benefit issue, Rogers J held that the benefit yielded by the Brazilian interest swap accrued to the D1 Trust and, through the conduit principle, to PSIC4 – not to Absa, which had simply invested in an asset class that is exempt by law. He cautioned against conflating an economic advantage with a tax benefit. Notably, he observed that nothing had prevented SARS from assessing the entities that did obtain the benefit: it could have taxed PSIC4 or the D1 Trust directly, or invoked its power under section 80B(1)(c) to treat connected persons as one and the same and thereby reach entities of substance within the Macquarie group. His conclusions, he stressed, did not impair SARS’s legitimate powers to combat avoidance; they merely subjected those far-reaching powers to limits conforming with the rule of law.

An objective purpose test, smuggled in

Perhaps the most striking feature of the judgment is what it says about a question it was not asked to decide. As we emphasised in Part I, Absa’s review was directed only at the ‘party’ and ‘tax benefit’ errors; the nature of the ‘sole or main purpose’ requirement – the subjective-versus-objective debate that animated much of the first article – was not in issue, and we predicted that the Court would not decisively settle it. Yet, in the course of reasoning towards its conclusions, the majority repeatedly characterised the purpose enquiry as objective. It held that ‘[t]he jurisdictional requirement is the existence of an avoidance arrangement whose sole or main purpose is to obtain a tax benefit and which is abnormal’ and that ‘[t]his is an objective enquiry’. That conclusion was ‘fortified by section 80G’: the requirement that the taxpayer prove that, ‘reasonably considered in light of the relevant facts and circumstances’, obtaining a tax benefit was not the sole or main purpose, was said to be ‘indicative of an objective test, as the focus is not on the taxpayer’s stated intent or purpose, but on the reasonable prevailing facts and circumstances’ – described as ‘a fundamental change’ from section 103(1), which the courts, ‘particularly the Supreme Court of Appeal in Conhage, interpreted to relate to the taxpayer’s subjective purpose’.

The majority went on to speak, in terms, of ‘expanding the purpose requirement to include an objective assessment of the purpose of an avoidance arrangement’, an enquiry said to occur ‘through a reasonable consideration of the arrangement in light of the relevant facts and circumstances in section 80G, and by permitting the GAAR to be applied to steps in and parts of a broader arrangement in section 80H’. Whatever its formal status, this reasoning is in obvious tension with the very statement we relied upon in Part I, namely the Supreme Court of Appeal’s insistence in the same litigation that ‘[w]hat must be determined in every case is the subjective purpose of the taxpayer’. The Constitutional Court has not expressly overruled that proposition, and the point was not argued before it; but a careful reader will struggle to reconcile the apex court’s characterisation of the purpose enquiry as objective with the continued vitality of a subjective test. We submit that the proverbial centre of gravity has shifted, and future litigants should expect SARS to invoke these passages as the death-knell of the subjective purpose enquiry – notwithstanding whether the question was, strictly, left open.

Whose purpose? Personifying the arrangement

If the purpose test is now objective, a conceptual difficulty immediately presents itself: whose purpose is being measured? Section 80G(1) speaks of ‘the sole or main purpose of the avoidance arrangement’, and section 80G(2) provides that ‘[t]he purpose of a step in or part of an avoidance arrangement may be different from a purpose attributable to the avoidance arrangement as a whole’. The Act, in other words, already personifies the arrangement, attributing to an inanimate structure a ‘purpose’ of its own. But an arrangement has no mind. Divorced from the intentions of the human and corporate actors who designed and implemented it, the ‘purpose’ of an arrangement can only be an inference drawn from its design, its architecture and, ultimately, its effects.

Herein lies the danger, and it is one our courts have recognised before. In the passage from Gallagher on which the majority itself relied, Corbett JA explained that an objective test in this context is one which ‘has regard rather to the effect of the scheme, objectively viewed’, as opposed to the purpose those carrying out the scheme intend to achieve – and he warned, in the same breath, that the statute ‘draws a clear distinction between the “effect” of a scheme and the purpose thereof … and this virtually rules out an interpretation which seeks to give “purpose” an objective connotation and to equate it, more or less, to “effect”’. The irony is acute: the authority invoked to support the objective turn is the very authority that cautioned against collapsing purpose into effect. If ‘objective purpose’ means no more than ‘what the arrangement was, on its face, structured to achieve’, then purpose and effect have indeed been fused for all practical purposes, and the ‘sole or main purpose’ requirement risks becoming a restatement of the tax-benefit enquiry rather than an independent element. In practice, the test would reduce to asking what a reasonable observer, surveying the structure and its fiscal consequences, would conclude the arrangement was designed to achieve – a question that, for any efficiently structured transaction, tends to answer itself.

The blurring of once-distinct elements

This brings us to the broader concern that the judgment’s objective orientation, combined with its endorsement of a step-by-step methodology, erodes the boundaries between elements that the new GAAR had deliberately kept apart. On its face, section 80A is a structured, sequential enquiry. There must be (i) an avoidance arrangement yielding a tax benefit; (ii) a sole or main purpose of obtaining that benefit; and (iii) at least one ‘tainted element’ – abnormal means or manner, a lack of commercial substance within the meaning of section 80C, the creation of non-arm’s-length rights or obligations, or a misuse or abuse of the Act. Each was intended to be a discrete gate, with its own content and its own boundaries. The taxpayer who failed one might still pass another. The misuse-or-abuse limb in section 80A(c)(ii) is, moreover, overtly purposive: it requires the interpreter to identify and protect the policy of the very provisions said to be abused, introducing yet a further layer of objective, policy-driven evaluation.

The objective approach threatens to dissolve these gates into a single, holistic evaluation. The ‘commercial substance’ enquiry in section 80C is itself defined by reference to effect – whether the arrangement has ‘a significant effect upon either the business risks or net cash flows’ of a party ‘apart from any effect attributable to the tax benefit’. The abnormality enquiry asks whether the arrangement was carried out in a manner ‘which would not normally be employed for bona fide business purposes, other than obtaining a tax benefit’ – a formulation that already imports a notion of purpose into the tainted-element analysis. The ‘commercially equivalent’ counterfactual endorsed for the tax-benefit enquiry, in turn, asks what a normal commercial actor would have done. Once the purpose enquiry is also recast as an objective assessment of commercial normality, and once, by virtue of sections 80G(2) and 80H, each of these questions must be posed afresh at every step of a multi-party structure, the four elements begin to converge on one and the same question: was this step, viewed objectively, commercially real or merely tax-driven?

The result, we submit, is a migration from a regime of reasonably bright lines towards a soft, impressionistic and contextual standard – something closer to a ‘smell test’ than a series of defined statutory requirements, at least in practice if not in theory. That is precisely the uncertainty we cautioned against in Part II in the context of the ‘appropriate hypothesis’. A test that depends on an objective assessment of what is ‘normal’ or ‘commercial’ is inescapably indeterminate, because commerciality is itself relative: what is unremarkable for one party, with one risk appetite and one strategic objective, may be abnormal for another. When the same impressionistic enquiry must be repeated step by step, and when the elements are no longer sealed off from one another, the predictability that taxpayers require in order to structure their affairs is the first casualty.

A reasonable-person overlay?

There is, however, a more measured reading of the judgment which is possible. On this view, the Court has not abolished the subjective enquiry so much as added an objective overlay to it: a reasonable-person dimension through which the taxpayer’s stated purpose is tested rather than displaced. This reading draws support from the legislative history. As Rogers J noted, SARS’s own Revised Proposals retreated from a purely objective purpose test, explaining that it was ‘never the intent … to prevent a taxpayer’s explanation of the reasons for an arrangement from being taken into account’, but rather ‘to ensure that a taxpayer’s statements of intent be rigorously tested against the relevant facts and circumstances’. On that footing, the taxpayer’s ipse dixit survives as evidence, but it must now withstand scrutiny against the objective characteristics of the arrangement, and may be overridden where the two diverge. (Arguably, however, that has always been the case under a subjective test in any event, as a matter of the law of evidence.)

If that is the better view, then what section 80G introduces is effectively a reasonable-person standard of the kind familiar throughout our law – dynamic, fact-sensitive, and capable of adapting to the parties, the market and the sophistication of the actors involved. Such a standard has obvious attractions for an anti-avoidance regime: it resists the formalism that allowed schemes to ‘become normal through extensive use’, and it denies the architects of artificial structures the shelter of a self-serving account of their intentions. But the very elasticity that makes the reasonable-person test attractive to the fiscus is what makes it troubling for the taxpayer. A standard that ‘changes and adapts dynamically’ with the circumstances is, by design, resistant to advance prediction. Whether the objective overlay is a welcome discipline or an open-ended license will depend on how rigorously, and how consistently, the courts hold SARS to the ‘relevant facts and circumstances’ rather than permitting reasonableness to collapse into hindsight.

Does the choice principle survive?

One further aspect of the judgment is striking and warrants specific attention, namely its practical effect on the so-called ‘choice principle’, derived from the classic case IRC v Duke of Westminster [1936] AC 1 and restated, in South African law, by Hefer JA in Commissioner for Inland Revenue v Conhage (Pty) Ltd (formerly Tycon (Pty) Ltd) 1999 (4) SA 1149 (SCA) (Conhage). In its traditional formulation, the principle holds that a taxpayer, confronted with two or more genuine routes to achieve the same commercial end, is entitled to select the more tax-efficient one. The choice principle presupposes, in other words, a chooser – a specific taxpayer exercising a deliberate decision between alternatives that the law itself makes available.

The majority does not expressly displace that principle, and Majiedt J even cites Conhage approvingly at paragraph [25], and the right to arrange one’s affairs so as to attract the least tax is nowhere denied. Yet the majority’s holdings on the ‘party’ element and on knowledge have, we submit, necessarily altered the principle, as the traditional understanding of the principle is irreconcilable with the GAAR extending to parties bearing no knowledge of the impugned features, as we show below.

After Absa Bank, a taxpayer may be assessed under the GAAR in respect of an arrangement it did not design, did not direct, and, on its own version, did not know about. Absa’s only relevant ‘choice’ was to invest in preference shares (an election squarely within the express exemption afforded by section 10(1)(k)(i) as relates to dividend income) as opposed to advancing a loan and receiving taxable interest thereon. The impermissible avoidance architecture (including the Brazilian bond swap and the conduit chain) was Macquarie’s design and occurred unbeknownst to Absa. Yet, because the ‘party’ enquiry is now an objective assessment of structural participation rather than of conscious agency, Absa is drawn into the composite arrangement and the GAAR consequences fall upon it, despite Absa not having exercised any choice related to the impugned features of the structure. If knowledge is not required, therefore, the conceptual link between the taxpayer’s own election and its potential exposure to the GAAR is necessarily severed.

As the principle was not overtly dismantled, it potentially survives for single-party or bilateral planning where the taxpayer itself selects between genuine alternatives to achieve the same commercial outcome, whereas for multi-party composite structures, where one participant’s capital is deployed downstream in ways it neither chose nor controlled, the principle can offer little protection. If the principle holds, a taxpayer electing to pursue a more tax-efficient route may be protected by it, whereas a taxpayer who exercised no choice to pursue tax-efficiency would not. This seems strange. What the policy reason behind such an outcome would be remains unclear – if indeed this outcome was intended to begin with.

What the ruling means

It is notable that the dispute reached the Court by an unusual route – a review of the section 80J notice rather than the ordinary objection-and-appeal process – and the majority decided the legal questions on a footing ‘akin to an exception’, taking the facts in SARS’s assessment as they stood ‘at this juncture’. The Court itself observed that this ‘is merely an assessment and the litigation process still has some way to go’, and references to a Rule 31 statement and a live penalties dispute suggest that the factual contest, and related proceedings, may yet continue. The holdings are therefore authoritative on the law but provisional on the facts.

Notwithstanding, the practical significance of Absa Bank is difficult to overstate. A financier that derives a lawful return from a structure may now be assessed under the GAAR for the avoidance conduct of other participants, even where it neither knew of nor controlled the impugned steps. As we anticipated in Part II, this fundamentally alters the risk profile of structured finance. Diligence can no longer be confined to a party’s own tax position; it must extend to the architecture of the entire arrangement into which a party’s capital is deployed, because the ‘conscious step into the structure’ is now enough to make one a party to all of it.

We have not concealed our preference, across this series, for the narrower and more certain readings. Notwithstanding, the majority has gone the other way. With respect, we remain of the view that the chief casualty of this approach is legal certainty, particularly acute against the backdrop of South Africa’s narrow tax base and the long interregnum during which the market structured its affairs on the strength of Secretary for Inland Revenue v Gallagher 1978 (2) SA 463 (A), Conhage and the entity-specific logic of Hicklin v SIR (1921) 12 TC 368. For taxpayers who relied in good faith on those settled principles, the retrospective application of a more objective interpretation is, as we put it in Part I, less a clarification of the law than the closing of a long-obscured trap. Equally, the majority’s anxiety about wilful blindness by sophisticated institutional actors is not without force, and the dissent’s answer – that SARS could and should have pursued the entities that actually obtained the benefit – will not always be available where those entities are empty shells. The disagreement is perhaps, at bottom, one about where the burden of avoidance’s collateral consequences should fall.

Two further statutory features must be considered. On the one hand, section 80B(2) obliges the Commissioner to make compensating adjustments to ensure consistent treatment of all parties – a fairness valve that should, in principle, prevent the same income being taxed twice across a structure, though its practical operation across a cross-border, multi-party arrangement is anything but straightforward. On the other, a party drawn into the net without knowledge faces the full force of the regime: typically including understatement penalties and, under section 80K, interest from which the Commissioner’s ordinary discretion to grant relief is expressly withdrawn.

Conclusion

When we began this series, the new GAAR stood at a jurisprudential crossroads, its true nature untested at the apex court for nearly two decades. With Absa Bank, the Constitutional Court has chosen its direction. On the two questions it decided, the GAAR has been recalibrated towards objective effect: a taxpayer may be a ‘party’ to an arrangement without knowing its every step, and may be assessed on a tax benefit it did not itself directly secure, measured against a counterfactual stripped of the transaction’s avoidance features. And on a question that it was not asked to decide – the nature of the ‘sole or main purpose’ test – the majority’s reasoning lends powerful momentum to the objective reading we doubted in Part I. Moreover, the choice principle may survive in name, but its contents have, for all practical purposes, been radically altered.

What remains is a regime whose elements are harder to keep apart and whose application turns increasingly on an objective, reasonable-person-style assessment of what an arrangement, viewed as a whole, was structured to achieve. Whether that mode of assessment is a disciplined enquiry into the ‘relevant facts and circumstances’ or a license for impressionistic, hindsight-driven taxation will be settled only in the cases to come. For now, the lesson for taxpayers and their advisers is sobering but clear: the shield once offered by the choice principle, a party’s own exempt status, by its ignorance of a counterparty’s structuring, or by a sincerely held commercial purpose, has been substantially withdrawn. If Parliament regards the balance struck in Absa Bank as wrong, it remains free, as the courts have repeatedly reminded us, to amend the Act. Until it does, those who deploy capital into structured arrangements would do well to assume that the GAAR now looks not to what they knew, nor even to what they intended, but to what the arrangement objectively achieved, and the impression this may create from the perspective of an outside observer.