Absa Bank Ltd and Another v Commissioner for the South African Revenue Service

The judgment

The court's own file

CONSTITUTIONAL COURT OF SOUTH AFRICA Case CCT 72/24 In the matter between: ABSA BANK LIMITED First Applicant UNITED TOWERS (PTY) LIMITED Second Applicant and COMMISSIONER FOR THE SOUTH AFRICAN REVENUE SERVICE Respondent Neutral citation: Absa Bank Ltd and Another v Commissioner for the South African Revenue Service [2026] ZACC 15 Coram: Mlambo DCJ, Kollapen J, Majiedt J, Mathopo J, Mhlantla J, Musi AJ, Rogers J, Savage AJ, Theron J and Tshiqi J Judgments: Majiedt J (majority):

[1]to

[115]Rogers J (dissenting):

[116]to

[159]Heard on: 23 September 2025 Decided on: 22 April 2026 Summary: Income Tax Act 58 of 1962 — statutory interpretation — sections 80A to L — impermissible avoidance arrangement — Tax benefit — party to an arrangement — knowledge of arrangement — appeal dismissed with costs 2 ORDER On appeal from the Supreme Court of Appeal (hearing an appeal from the High Court of South Africa, Gauteng Division, Pretoria): The appeal is dismissed with costs, including the costs of two counsel. JUDGMENT MAJIEDT J (Mlambo DCJ, Kollapen J, Mathopo J, Mhlantla J, Musi AJ, Savage AJ, Theron J and Tshiqi J concurring):

[1]John Maynard Keynes reportedly said that “[t]he avoidance of taxes is the only intellectual pursuit that still carries any reward”. 1 And former Chancellor of the Exchequer Denis Healey once remarked that “the difference between tax evasion and tax avoidance is the thickness of a prison wall”. 2 This case is about the impermissible tax avoidance arrangements in sections 80A to 80L of the Income Tax Act 3 (ITA), generally known as the “General Anti-Avoidance Rules” (GAAR). This is the first time that courts have had to interpret the GAAR provisions after their amendment in 2006. 4 1 Quoted in Hasan “Organizational Capital, Corporate Tax Avoidance, and Firm Value” (2021) 70 Journal of Corporate Finance 102050 at 1-2. 2 Elliffe “The Thickness of a Prison Wall – When Does Tax Avoidance Become a Criminal Offence?” (2011) 14 New Zealand Business Law Quarterly Review 441 at 441-6. 3 58 of 1962. 4 The amendment was made through section 34(1) of Act 20 of 2006, which is deemed to have come into operation on 2 November 2006 and applies to any arrangement (or any steps therein or parts thereof) entered into on or after that date. MAJIEDT J 3

[2]The case before us is the second round of the parties’ legal skirmish, this Court having earlier, in five cases that were heard together, set aside the Supreme Court of Appeal’s ruling on the question of the High Court’s jurisdiction in terms of section 105 of the Tax Administration Act 5 (TAA). 6 We already granted leave to appeal in those previous proceedings. 7 We are now called upon to grapple with the substantive merits of the case in this appeal against the judgment of the Supreme Court of Appeal in the matter between Absa Bank Limited (Absa) and its subsidiary, United Towers (Pty) Limited (United Towers), collectively referred to as the applicants, and the Commissioner for the South African Revenue Service (SARS), the respondent. Where reference is made to Absa, it also includes United Towers.

[3]The appeal concerns the lawfulness of tax assessments issued by SARS for the 2014 to 2018 financial years through its powers under the ITA and the TAA. SARS is empowered to issue notices under section 80J of the ITA as part of the GAAR regime. Factual background

[4]Between 2011 and 2015, Absa and its wholly owned subsidiary, United Towers, entered into four preference share subscription agreements with PSIC Finance 3 (RF) (Pty) Limited (PSIC3) to the tune of R1.9 billion. Absa received tax- exempt dividends on these preference shares. The Macquarie Group (Macquarie) introduced these transactions to Absa and the latter concluded related agreements with entities in Macquarie. These investments were secured through back-to-back preference share arrangements with PSIC Finance 4 (RF) (Pty) Limited (PSIC4), although, according to Absa, it believed those arrangements to be with Macquarie Securities South Africa Limited (MSSA). The agreements included a right on Absa’s part to put the preference shares to Macquarie in certain circumstances, and an obligation by Macquarie to make up any shortfall in Absa’s anticipated returns on the 5 28 of 2011. 6 United Manganese of Kalahari (Pty) Limited v Commissioner of the South African Revenue Service and four other cases [2025] ZACC 2; 2025 (5) BCLR 530 (CC) (Five Tax Cases). 7 Id at para 328, and para 3 of the order made in CCT 72/24 Absa Bank Limited and United Towers (Pty) Limited v Commissioner for the South African Revenue Service, forming part of the Five Tax Cases. MAJIEDT J 4 shares, including any shortfall arising if the dividends were taxed contrary to expectation.

[5]Absa and United Towers claimed that, unbeknown to them, the funds flowed beyond PSIC3 into— (a) PSIC4; (b) Delta 1 Finance Trust (D1 Trust); and (c) ultimately, back to MSSA.

[6]The D1 Trust lent funds to MSSA through interest-bearing notes as part of a complex funding arrangement designed by Macquarie. The D1 Trust received interest payments from MSSA on these notes, which it then used to acquire Brazilian government bond interest through its purchase of US dollar denominated Brazilian government bonds through buy/sell back transactions with a bank. This bond interest was distributed to PSIC4 as non-taxable income under the South Africa-Brazil double tax agreement 8 and section 25B of the ITA, and was ultimately distributed as dividends. The outflow and inflow of funds back are depicted in the following diagrams. 9 8 Under article 11(4)(b) of the double tax agreement. 9 The diagrams in this judgment are sourced from Pidduck and Swanepoel “‘Semantic gyrations’ – When are Naartijies Oranges? Beneath the Surface of Absa Bank Limited v CSARS” (2022) 33 SA Mercantile Law Journal at 470. MAJIEDT J 5 Figure A: Investment flow out Figure B: Income flow back

[7]On 18 May 2018, SARS issued audit notifications to Absa and United Towers. The notices referred only to PSIC3, preference shares and related dividends. The notifications sought information on PSIC4, the D1 Trust and the Brazilian bond transactions. The applicants responded that they had no knowledge of these entities MAJIEDT J 6 and, given their omission, could not possibly have addressed them in their own documentation. According to SARS, Absa and United Towers invested in preference shares in PSIC3, expecting to receive tax-exempt dividends from these shares. This was part of their investment in the funding arrangement set up by Macquarie.

[8]Following the audit, SARS issued notices in terms of section 80J of the ITA. SARS stated that it would disregard the intervening entities and treat the dividends accruing to Absa and United Towers as taxable interest. The notices relied on the GAAR provisions on the basis that the main purpose of the arrangements was to obtain a tax benefit.

[9]In October 2019, SARS assessed Absa and United Towers for additional taxation, alleging that they were parties to an impermissible avoidance arrangement under the GAAR. SARS re-characterised the tax-exempt dividend income received by Absa as taxable interest income. SARS contended that this structure allowed PSIC4 to receive non-taxable income instead of taxable interest, constituting a tax benefit under the GAAR. A tax benefit, as defined in section 1 of the ITA, occurs where a taxpayer avoids, postpones or reduces their tax liability through an arrangement.

[10]Absa, on the other hand, argued that it was unaware of the transactions beyond its investment in PSIC3 and ancillary agreements with Macquarie. The counter-argument on its part was that its involvement was limited to investing in preference shares in PSIC3. Litigation history High Court

[11]Absa launched a review in the High Court together with a request for a direction under section 105 of the TAA, arguing that the assessments were flawed due to two legal errors: first, that Absa could not be said to have been a party to the MAJIEDT J 7 arrangement, as it was unaware of the full structure, particularly the parts generating the alleged tax benefit; and, second, that Absa did not obtain a “tax benefit” from the arrangement. The High Court found these to be issues of law, suitable for adjudication by the High Court pursuant to a direction under section 105, and set aside the assessments. 10 SARS appealed to the Supreme Court of Appeal. Supreme Court of Appeal

[12]The Supreme Court of Appeal confined its judgment to the question whether the High Court had the requisite jurisdiction to review the assessments. On the tax benefit error, that Court held that the question whether Absa and United Towers obtained a tax benefit by avoiding an anticipated tax liability through the arrangement was a question of fact, not solely a question of law. 11 The High Court had found that SARS’ assessments were based on the premise that the applicants did not obtain a tax benefit themselves, as the benefit occurred at the level of other entities, rendering the assessments legally flawed. However, the Supreme Court of Appeal disagreed, stating that determining whether the applicants avoided an anticipated tax liability required a factual enquiry into the effect of the transactions. 12

[13]The Supreme Court of Appeal noted that the effect, purpose and normality of a transaction are essentially factual questions. Thus, the dispute did not constitute an exceptional circumstance under section 105. The High Court should thus not have granted a direction under that section and should not have entertained the review on this ground. 13

[14]On the party error, the Supreme Court of Appeal ruled that the next enquiry, whether or not Absa and United Towers were parties to the avoidance arrangement, as 10 Absa Bank Limited v Commissioner for the South African Revenue Service 2021 (3) SA 513 (GP) (High Court judgment). 11 Commissioner for the South African Revenue Service v Absa Bank Limited [2023] ZASCA 125; 2024 (1) SA 361 (SCA); 86 SATC 195 (Supreme Court of Appeal judgment). 12 Id at paras 31-5. 13 Id at para 33. MAJIEDT J 8 defined under section 80L of the ITA, also involved disputed facts rather than law. 14 The High Court had accepted that SARS’ assessment letters conceded the applicants’ ignorance of the full arrangement, particularly transactions involving PSIC4 and the D1 Trust, concluding that they could not be parties as a matter of law. The Supreme Court of Appeal rejected this, finding that SARS did not accept the applicants’ lack of knowledge; rather, SARS disputed their contentions in correspondence and affidavits, indicating a factual disagreement about their awareness and participation. The Court emphasised that whether the applicants knew the full ambit of the scheme and whether their participation was for the purpose of securing a tax benefit were factual matters requiring evidence. Thus, the Supreme Court of Appeal held that the High Court erred in characterising this as a purely legal issue, and no exceptional circumstances justified exercising review jurisdiction pursuant to a direction under section 105. 15 This Court’s judgment in the Five Tax Cases

[15]In its earlier judgment, this Court granted leave to appeal, excused peremption and confirmed the High Court’s jurisdiction under section 105, finding that the Supreme Court of Appeal erred in characterising the issues as factual. 16 The remaining issues for determination in this case are whether the assessments should be set aside on review as a result of the alleged legal errors and the appropriate costs order. Issues

[16]There are two central issues on the merits— (a) First, a twofold enquiry with two interrelated questions: (1) whether this is an impermissible avoidance arrangement as contemplated in section 80A; and 14 Id at para 31. 15 Id at paras 25 and 30. 16 See the order in CCT 72/24 in the Five Tax Cases above n 6 at para 385. See also at para 313 thereof. MAJIEDT J 9 (2) relatedly, what constitutes a “party” under section 80L of the ITA, and in particular, whether it requires knowledge of all steps of an avoidance arrangement (the party issue); and (b) Second, whether SARS can invoke the GAAR provisions against a taxpayer who is alleged not to have personally obtained a tax benefit, but had allegedly merely received financial returns from others’ tax benefits (the tax benefit issue). Submissions in this Court Applicants’ main submissions

[17]The parties’ main submissions have already been expounded on to some extent. I add only a few more. The applicants emphasise that this Court has already, in its earlier judgment, held that the Supreme Court of Appeal misdirected itself in holding that the two alleged errors involved disputed facts and were not purely questions of law. They say that SARS can therefore no longer make submissions that the review involves factual disputes as that point is now res judicata (a matter already decided). Thus, the validity of the impugned assessments is to be judged by reference to the factual matrix on which they are expressly based, as conceived by SARS and set out in the letter of assessment. If they are legally unsustainable on their own factual premises, they must be set aside. Therefore, assert the applicants, there are only two legal issues in this matter: first, whether, based on the assertions in the letter of assessment, Absa and United Towers were parties to the impermissible avoidance arrangement for the purposes of the GAAR; and, secondly, whether they received a tax benefit within the meaning of the GAAR.

[18]In relation to the party issue, the applicants contend that SARS wrongly concluded that they were parties to the arrangement, as defined in section 80L of the ITA, despite no evidence that Absa was aware of the full arrangement, particularly the steps involving PSIC4, the D1 Trust and the Brazilian bonds. As a matter of law, it is contended that is not possible for them to be a “party” to an alleged impermissible MAJIEDT J 10 avoidance arrangement as envisaged in the GAAR when they had no knowledge of the parts of the arrangement by which the tax benefit, on which the assessment is based, was achieved. Since such knowledge is essential to make them parties, SARS’ failure to aver this, and its decision to proceed with the assessment despite not being able to aver this, renders the assessment against the applicants fatally defective. It is submitted that the assessment letter acknowledges that Absa’s internal documents make no reference at all to the PSIC4, the D1 Trust or any of the transactions undertaken by them. It also acknowledges that Absa’s internal documents appear to show that Absa understood the transactions as back-to-back preference share investments into MSSA (via PSIC3) to fund MSSA’s broker operations.

[19]Regarding the tax benefit issue, the applicants submit that it is plain that the alleged “tax benefit” identified by SARS occurred at the level of the D1 Trust and PSIC4. The D1 Trust was identified as the entity that avoided an anticipated tax liability, and this is confirmed by SARS’ letter of assessment. The applicants assert that what they had received is a mere financial benefit, unlike other entities which may have received a tax benefit, and for that reason, this issue must be decided against SARS. Their case is that a “tax benefit” must be interpreted as the avoidance of the assessed taxpayer’s own anticipated liability, applying a “but-for” test to determine if the arrangement allowed the taxpayer to escape tax it would otherwise have incurred.

[20]Relying on King, 17 Hicklin 18 and Sasol Oil, 19 the applicants argue that the anticipated nature of tax liability is an essential element of the test for whether or not a party received a tax benefit. Therefore, the applicants argue, Absa did not obtain a tax benefit on the established meaning of that term. 17 Commissioner for Inland Revenue v I H B King; Commissioner for Inland Revenue v A H King 1947 (2) SA 196 (A). 18 Hicklin v Secretary for Inland Revenue 1980 (1) SA 481 (A). 19 Sasol Oil Proprietary Limited v Commissioner for the South African Revenue Service [2019] 1 All SA 106 (SCA). MAJIEDT J 11

[21]The applicants then turn to the question whether, even though Absa did not obtain a tax benefit, it can be subjected to a GAAR assessment by virtue of having received a financial benefit. The applicants argue that the correct question to be asked is whether an entity can ever be subjected to a GAAR assessment when it was not the person that avoided an anticipated tax liability. They contend that the answer is no. The applicants submit that there is no averment in the letter of assessment to sustain an argument that if Absa had not made the preference share investment, it would have made an interest-bearing loan to MSSA. SARS’ main submissions

[22]In relation to the impermissible avoidance arrangement, SARS’ case is that a total of 13 entities participated in the scheme and that it was a predetermined arrangement. According to SARS, the purpose of the scheme was to swap a taxable income stream paid by MSSA to the D1 Trust for a tax-free income stream paid by the D1 Trust to PSIC4, PSIC3 and, ultimately, to Absa. SARS submits that, but for this swap, PSIC4 would have had to pay tax on the interest it received from the D1 Trust. As a result of the swap, however, the interest received from the D1 Trust was tax-free. This caused the dividend stream from PSIC4 to PSIC3, and ultimately to Absa, to be enhanced.

[23]SARS invokes section 80G of the ITA in submitting that the scheme is presumed to have been entered into or carried out for the sole or main purpose of obtaining a tax benefit. SARS contends that Absa was a party to the arrangement as envisaged in section 80L of the ITA, in that it took part and participated in the scheme. That participation consists of Absa’s investing in the scheme and deriving enhanced income from it. The receipt of an enhanced income stream from the scheme is also the tax benefit derived by Absa, thus contends SARS.

[24]It is further argued by SARS that, even if it were required to show that Absa derived a tax benefit, the evidence demonstrates that Absa did. According to SARS, but for the artificial structuring of the transactions, Absa’s investment returns would MAJIEDT J 12 have been taxable interest. Instead, through the arrangements, Absa received inflated tax-exempt dividends, which constituted a clear financial advantage at the expense of the fiscus. SARS contends that Absa’s counterfactual test is misconceived: the proper comparison is not with “no transaction”, but with the transaction stripped of its avoidance features. On this test, SARS contends that Absa had plainly obtained a tax benefit. Analysis The GAAR in the context of tax avoidance generally

[25]There is nothing illegal in minimising one’s tax obligations, and avoiding paying tax, provided one does so within the parameters permitted by the law. 20 Tax avoidance has been defined as “stratagems which are prima facie lawful, that is to say, which are lawful unless proscribed by the [ITA]”. 21 A taxpayer is entitled to choose the most tax efficient method of implementing a transaction where the same commercial result can be attained. 22 However, in examining the transaction, a court will give effect to the true nature and substance of the transaction and will not be deceived by its form. 23

[26]Watermeyer CJ, in the seminal judgment of the then Appellate Division in King, 24 outlined a number of examples of how a taxpayer “may [reduce] the amount of his income to something less than it was in the past, or of freeing himself from taxation on some part of his future income”. 25 Some of these include: (a) Selling investments which produce income subject to tax, and in their place make no investments at all. 20 Commissioner of Inland Revenue v Duke of Westminster (1936) AC 1 at 8; Hicklin above n 18. 21 De Koker Silke on South African Income Tax (LexisNexis South Africa, Durban 2023) vol 4 at 19.1. 22 Commissioner for Inland Revenue v Conhage (Pty) Ltd (formerly Tycon (Pty) Ltd) [1999] ZASCA 64; 1999 (4) SA 1149 (SCA) (Conhage) at para 1; Hicklin above n 18 at 311. 23 Conhage above n 22 id, citing Erf 3183/1 Ladysmith (Pty) Ltd v Commissioner for Inland Revenue 1996 (3) SA 942 (A) at 950I-952C. 24 King above n 17. 25 Id at 208. MAJIEDT J 13 (b) Selling shares in companies which pay high dividends and investing in securities which return a lower but safer and more certain income. (c) A taxpayer may even develop such a dislike for taxation that she sells all his investments and lives on her capital or gives it away to the poor in order not to have to pay tax. (d) Reducing his fees or working for nothing. (e) If the taxpayer is a trader, he may reduce his rate of profit or sell his goods at a loss in order to earn a smaller income. (f) By securing deductions from the amount of his gross income, for example by insuring his life. 26

[27]Tax avoidance statutes are one way in which the law sets parameters for tax obligations to counter tax avoidance arrangements that are regarded as undesirable. Unlike specific anti-tax avoidance legislation, the GAAR operates on the basis of conceptual principles used to address tax avoidance, as opposed to addressing specifically defined transactions that may provide taxpayers with loopholes for impermissible tax avoidance. 27 As is the case with legislation generally, anti- avoidance legislation may not lead to absurdities in its interpretation. It was enunciated thus in King: “[I]t cannot be imagined that Parliament intended by the provisions of section 90 [of the then Income Tax Act] to do such an absurd thing as to levy a tax upon persons who carry out such operations as if they had not carried them out. Moreover the problem of deciding what the income of such persons for the tax year would have been if they had not carried out such operations would appear to be insoluble in some cases, if the countless possibilities of what they might otherwise have done with their capital or their labour are borne in mind.” 28 26 Id. 27 SARS Discussion Paper on Tax Avoidance and Section 103 of the Income Tax Act, 1962 (November 2005) (Discussion Paper) at 6. The Discussion Paper is available at https://www.sars.gov.za/wp- content/uploads/Legal/DiscPapers/LAPD-LPrep-DP-2005-01-Discussion-Paper-Tax-Avoidance-Section-103- of-Income-Tax-Act-1962.pdf. 28 King above n 17 at 208. MAJIEDT J 14

[28]This brings me to the GAAR provisions in the present instance. GAAR Historical context: section 103(1)

[29]Historical context matters when the text of a statute must be interpreted. 29 The GAAR, it must be remembered, is a specific form of an anti-avoidance measure. In Bosch, the Supreme Court of Appeal observed that “[i]f the revenue authorities regard any particular form of tax avoidance as undesirable they are free to amend the Act, as occurs annually, to close anything they regard as a loophole”. 30 As far as amendments to the GAAR provisions are concerned, there have been three “GAARs” to date in South African tax law. The first was included in section 90 of the Income Tax Act, 31 while the second was included in the now-repealed section 103(1) of the present ITA. As stated, in its third and current iteration, effected by the 2006 amendment, the GAAR is found in sections 80A to 80L of the ITA and applies to any arrangement entered into after 2 November 2006. Though the first two GAARs have been repealed, they remain of some importance for interpretation purposes. The main focus in this judgment will be on the immediate predecessor, section 103(1), as it read after the 1996 amendment by section 29 32 of the Revenue Laws Amendment Act. 33 29 University of Johannesburg v Auckland Park Theological Seminary [2021] ZACC 13; 2021 (8) BCLR 807 (CC); 2021 (6) SA 1 (CC) at paras 66-7. 30 Commissioner of the South African Revenue Service v Bosch [2014] ZASCA 171; [2015] 1 All SA 1 (SCA); 2015 (2) SA 174 (SCA) at para 40. 31 29 of 1941. 32 It read, in relevant part: “Whenever the Commissioner is satisfied that any transaction, operation or scheme (whether entered into or carried out before or after the commencement of this Act, and including a transaction, operation or scheme involving the alienation of property)— (a) has been entered into or carried out which has the effect of avoiding or postponing liability for the payment of any tax, duty or levy imposed by this Act or any previous Income Tax Act, or reducing the amount thereof; and (b) having regard to the circumstances under which the transaction, operation or scheme was entered into or carried out— (i) was entered into or carried out— (aa) in the case of a transaction, operation or scheme in the context of business, in a manner which would not normally be employed for MAJIEDT J 15

[30]In ITC 1862, the Court described section 103 of the ITA as the exercise of “extraordinary administrative power” that enables the Commissioner to “overturn the express and ordinary consequences of applying the Act”. 34 Under section 103(1) the following requirements had to be met for an impermissible tax avoidance arrangement to be established: (a) a transaction, operation or scheme; (b) resulting in the avoidance, reduction or postponement of tax; (c) entered into or carried out (in the case of a transaction, operation or scheme in the context of business)— (i) in a manner which would not normally be employed for bona fide business purposes, other than obtaining a tax benefit; or (ii) creating rights or obligations which would not normally be created between persons dealing at arm’s length under a transaction, operation or scheme of the nature of the one in question (the abnormality requirement); and (iii) entered into solely or mainly for the purpose of obtaining a tax benefit (the purpose requirement). bona fide business purposes, other than the obtaining of a tax benefit; and (bb) in the case of any other transaction, operation or scheme being a transaction, operation or scheme not falling within the provisions of item (aa) by means or in a manner which would not normally be employed in the entering into or carrying out of a transaction, operation or scheme of the nature of the transaction, operation or scheme in question; (ii) has created rights or obligations which would not normally be created between persons dealing at arm’s length under a transaction, operation or scheme of the nature of the transaction, operation or scheme in question; and (c) was entered into or carried out solely or mainly for the purposes of obtaining a tax benefit; the Commissioner shall determine the liability for any tax, duty or levy imposed by this Act, and the amount thereof, as if the transaction, operation or scheme had not been entered into or carried out, or in such a manner as in the circumstances of the case he deems appropriate for the prevention or diminution of such avoidance, postponement or reduction.” 33 46 of 1996. 34 ITC 1862 75 SATC 34 at para 59, endorsed in ITC 1876 77 SATC 175. MAJIEDT J 16

[31]Some commentators opine that the 1996 amendment was made because tax avoidance schemes had become more widespread since 1985. The amendment introduced a business purpose test and focused on the manner in which the transaction was carried out as opposed to the transaction itself. The amendment initially appeared to have some positive result in making the GAAR more effective. It became more difficult for a taxpayer to structure a tax avoidance scheme that could withstand attack under the provisions of the amended GAAR. The business purpose test also prevented a taxpayer from entering into a scheme that was entered into solely or mainly for the purpose of tax avoidance, because it would not have been entered into for a bona fide business purpose other than to obtain a tax benefit. The particular scheme employed by a taxpayer could be tested to determine whether it was entered into in a manner normally employed for bona fide business purposes, other than to obtain a tax benefit.

[32]Under section 103(1), the Commissioner has to be satisfied that the transaction, operation or scheme was one where these provisions would apply. Once SARS had proved the existence of a scheme that had the effect of avoiding tax and which failed the abnormality test, a presumption applied. This was to the effect that, until the contrary was proved, it was assumed that such transactions were entered into or carried out solely or mainly for the purpose of avoiding, postponing or reducing the amount of any tax payable. All of these requirements had to be met before the Commissioner was entitled to determine the amount of tax liability as if the transaction had not been entered into or carried out. Section 103(1) could only be applied to a transaction as a whole and not to individual steps in the transaction. 35

[33]SARS identified challenges with section 103(1) in the Discussion Paper. 36 It noted that “the [GAAR] has proven to be an inconsistent and at times, ineffective deterrent to the increasingly complex and sophisticated tax ‘products’ that are being marketed by banks, ‘boutique’ structured finance firms, multinational accounting 35 Commissioner for Inland Revenue v Louw 1983 (2) All SA 291 (A); 1983 (3) SA 551 (A) at 561B-F (Louw). 36 Discussion Paper above n 27 at 41. MAJIEDT J 17 firms and law firms”. 37 The difficulties that SARS encountered with section 103(1) were, generally, that— (a) it was not an effective deterrent to tax avoidance; (b) the abnormality requirement failed to stand the test of time, because if a particular transaction was widely used, it became normal through extensive use; (c) the purpose requirement faced the same headwind – obtaining a tax benefit had to be the sole or main purpose of the parties (that is, whether the scheme was entered into by the parties “wholly or mainly for the purposes of obtaining a tax benefit”), but the ease with which taxpayers were able to justify the commercial purpose of transactions left SARS in the difficult position of having to prove that the dominant purpose of the transaction was to obtain a tax benefit; and (d) there were procedural and administrative challenges regarding the scope of the provision.

[34]The various difficulties with section 103(1) led to the 2006 amendment which repealed section 103(1) and introduced, in its place, sections 80A-L. These provisions introduced an “impermissible avoidance arrangement” to overcome some of the weaknesses of its predecessor. It appears that SARS had taken the view at the time that the existing provisions were not an effective deterrent to counter aggressive and increasingly sophisticated and innovative avoidance schemes by taxpayers and their advisers. According to the Discussion Paper, SARS considered similar legislation from Australia, Canada, New Zealand, Spain, the United Kingdom and the United States of America. 38 37 Id at 1. 38 Id at 27. MAJIEDT J 18 Statutory framework

[35]The key GAAR provisions are sections 80A, 80B, 80G and 80L. 39 Section 80A sets out when an arrangement will be an impermissible tax avoidance 39 They read: “80A Impermissible tax avoidance arrangements An avoidance arrangement is an impermissible avoidance arrangement if its sole or main purpose was to obtain a tax benefit and— (a) in the context of business— (i) it was entered into or carried out by means or in a manner which would not normally be employed for bona fide business purposes, other than obtaining a tax benefit; or (ii) it lacks commercial substance, in whole or in part, taking into account the provisions of section 80C; (b) in a context other than business, it was entered into or carried out by means or in a manner which would not normally be employed for a bona fide purpose, other than obtaining a tax benefit; or (c) in any context— (i) it has created rights or obligations that would not normally be created between persons dealing at arm's length; or (ii) it would result directly or indirectly in the misuse or abuse of the provisions of this Act (including the provisions of this Part). 80B Tax consequences of impermissible tax avoidance (1) The Commissioner may determine the tax consequences under this Act of any impermissible avoidance arrangement for any party by— (a) disregarding, combining, or re-characterising any steps in or parts of the impermissible avoidance arrangement; (b) disregarding any accommodating or tax-indifferent party or treating any accommodating or tax-indifferent part and any other party as one and the same person; (c) deeming persons who are connected persons in relation to each other to be one and the same person for purposes of determining the tax treatment of any amount; (d) reallocating any gross income, receipt or accrual of a capital nature, expenditure or rebate amongst the parties; (e) re-characterising any gross income, receipt or accrual of a capital nature or expenditure; or (f) treating the impermissible avoidance arrangement as if it had not been entered into or carried out, or in such other manner as in the circumstances of the case the Commissioner deems appropriate for the prevention or diminution of the relevant tax benefit. (2) Subject to the time limits imposed by sections 99, 100 and 104(5)(b) of the Tax Administration Act, the Commissioner must make compensating adjustments that he or she is satisfied are necessary and appropriate to ensure the consistent treatment of all parties to the impermissible avoidance arrangement. . . . MAJIEDT J 19 scheme and section 80B stipulates what the consequences of an impermissible avoidance arrangement are. Importantly, through a section 80B determination, the Commissioner can effectively create a liability for tax that the Act does not itself substantively impose. 40 Section 80G contains an important presumption of purpose, while section 80L attributes meanings to certain terms for purposes of the GAAR.

[36]In section 1 of the ITA, “tax benefit” is defined as including “any avoidance, postponement or reduction of any liability for tax”. Section 80A of the Act stipulates that an avoidance arrangement will be an “impermissible avoidance arrangement” if “its sole or main purpose was to obtain a tax benefit” and it fails what has been described as the “tainted element test” set out in subsections (a) to (c) of section 80A 80G Presumption of purpose (1) An avoidance arrangement is presumed to have been entered into or carried out for the sole or main purpose of obtaining a tax benefit unless and until the party obtaining a tax benefit proves that, reasonably considered in light of the relevant facts and circumstances, obtaining a tax benefit was not the sole or main purpose of the avoidance arrangement. (2) The 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. . . . 80L Definitions For purposes of this Part— ‘arrangement’ means any transaction, operation, scheme, agreement or understanding (whether enforceable or not), including all steps therein or parts thereof, and includes any of the foregoing involving the alienation of property; ‘avoidance arrangement’ means any arrangement that, but for this Part, results in a tax benefit; ‘impermissible avoidance arrangement’ means any avoidance arrangement described in section 80A; ‘party’ means any— (a) person; (b) permanent establishment in the Republic of a person who is not a resident; (c) permanent establishment outside the Republic of a person who is a resident; (d) partnership; or (e) joint venture, who participates or takes part in an arrangement; ‘tax’ includes any tax, levy or duty imposed by this Act or any other Act administered by the Commissioner.” 40 Cronje The Interplay Between Anti-Avoidance Measures and Curbing Tax Evasion and Impermissible Tax Avoidance Arrangements (dissertation submitted in partial fulfilment of LLM degree, University of Pretoria, 2024) at 42. MAJIEDT J 20 (read with sections 80C to 80F). In terms of section 80G(1), once the existence of an impermissible avoidance arrangement has been established, it will be “presumed to have been entered into or carried out for the sole or main purpose of obtaining a tax benefit unless and until the party obtaining a tax benefit provides that, reasonably considered in light of the relevant facts and circumstances, obtaining a tax benefit was not the sole or main purpose of the avoidance arrangement”.

[37]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”. This means that the commercial purpose of the composite transaction cannot be used to disguise a step with a tax avoidance purpose. 41

[38]The requirements of the present GAAR are: (a) A transaction, operation or scheme must be present. (b) The transaction, operation or scheme must result in a “tax benefit”. (c) The sole or main purpose of the transaction, operation or scheme must be to obtain the tax benefit. (d) The arrangement must (in the context of business) be abnormal or lacking in commercial substance or create rights and obligations not normally arising between parties dealing at arm’s length or be abusive of the provisions of the Income Tax Act. Revised Proposals

[39]SARS’ Revised Proposals of September 2006 that accompanied the 2006 amendment shed some light on its objectives and they may be of some utility in the process of interpretation. 42 In the introductory general explanation, it was said that 41 See Kujinga A Comparative Analysis of the Efficacy of the General Anti-Avoidance Rule as a Measure Against Impermissible Income Tax Avoidance in South Africa (thesis submitted in partial fulfilment of LLD degree, University of Pretoria, 2013) at 123. 42 On 3 November 2005, the Minister of Finance launched a Discussion Paper on Tax Avoidance and Section 103 of the ITA. The Paper also included a set of proposed amendments (original proposals) to section 103. Numerous submissions were received and according to SARS’ explanation in the Revised Proposals, the Proposals were based upon the public comments received and extensive discussions with international experts in MAJIEDT J 21 new provisions were included “that are intended to ensure that the new GAAR is broad enough to reach as many forms of impermissible tax avoidance as possible and strong enough to be an effective deterrent against them”. It was explained further that “the revised proposals seek to achieve the proper balance between the need for a strong and effective deterrent and the need for certainty in connection with bona fide business arrangements”. 43

[40]Six major changes in the 2006 amendment were noted in the Revised Proposals. The first was an intention to expand and reinforce the abnormality requirement in section 103(1) through the introduction of a new commercial substance test. It emphasised the test being objective and not subjective. This was to obviate the difficulties with the abnormality requirement that, amongst others, operated under the erroneous premise that tax matters can be neatly divided into good faith transactions and impermissible avoidance arrangements. Thus, an explicit test was introduced which targets avoidance arrangements that lack commercial substance. This is the gist of the present section 80A(a)(ii).

[41]The second change was to reduce the number of factors that had been proposed for use in determining abnormality from eleven to five and to recast those remaining five as indicators of a lack of commercial substance. The non-exhaustive list of five factors are in section 80C(2), which reads: “(2) For purposes of this Part, characteristics of an avoidance arrangement that are indicative of a lack of commercial substance include but are not limited to— (a) a legal or economic effect resulting from the avoidance arrangement as a whole that is inconsistent with, or differs significantly from, the legal form of its individual steps; this field. SARS stated further that in some cases, the original proposals had been retained; in others, they were modified or withdrawn and several new provisions were being introduced. 43 SARS Tax Avoidance and Section 103 of the Income Tax Act, 1962: Revised Proposals (September 2006) (Revised Proposals), available at: https://www.sars.gov.za/wp-content/uploads/Legal/RespDocs/LAPD-LPrep- Resp-2006-03-Response-Document-Revised-Proposal-Tax-Avoidance-section-103.pdf at 2. MAJIEDT J 22 (b) the inclusion or presence of— (i) round trip financing as described in section 80D; or (ii) an accommodating or tax indifferent party as described in section 80E; or (iii) elements that have the effect of offsetting or cancelling each other without substantial change in the economic position of any one or more of the parties; or (c) an inconsistent characterisation of the avoidance arrangement for tax purposes by the parties.”

[42]The third change intended to replace the factor relating to circular flows of cash and assets with one that targets what is commonly known as round trip financing, together with a detailed description of the scope of the new provision. This is an important feature relevant to the case before us. It is explained thus: “In general, the description would encompass any avoidance arrangement in which funds are transferred between or among the parties (‘round tripped amounts’) and those round tripped amounts would both (1) result, directly or indirectly, in a tax benefit (but for the provisions of the GAAR), and (2) significantly reduce, offset or eliminate any credit or economic risk incurred by any party in connection with the avoidance arrangement.” 44 The concept is explained to be analogous to the concept of “round robin financing” in Australia and “circular cash flows” in the United States.

[43]The fourth change was aimed at introducing a second new element targeting avoidance arrangements that would frustrate the purpose of any provision of the Act. Again, this is an important aspect in deciding the central issues in this case. The intention is to introduce a statutory purpose element directed at “schemes that would frustrate the purpose of any provision of the [ITA], including the provisions of the new Part IIA itself”. 45 It is significant that the stated intention with the introduction of 44 Now section 80D(1). 45 Section 80C(a)(ii). MAJIEDT J 23 this statutory purpose element is to enhance the contextual and purposive approach to statutory interpretation.

[44]The fifth change was to introduce a new notice requirement that would apply whenever the Commissioner believes that the provisions of the new GAAR may be applicable in determining the tax liability of a taxpayer. 46 Finally, the sixth change would withdraw the presumption of abnormality included in the original proposals. With these stated objectives as backdrop, the central issues can now be addressed. Does the present arrangement fall foul of the GAAR? General

[45]Avoidance arrangements are deals or structures set up mainly to reduce or avoid paying taxes. Under the GAAR, in terms of sections 80A to 80L of the ITA, an arrangement is impermissible if its main purpose is to dodge taxes and lacks genuine business substance. This matter must be decided through a procedure akin to an exception, that is, that the averments as presented by the applicants must be accepted as correct. 47

[46]There is, however, one important rider to that approach and it is this: what we have here thus far is only an assessment by SARS. It is trite that an assessment is not a pleading. We must remain cognisant of the fact that SARS may still, in its rule 31 statement of case, make averments controverting those presently advanced by the applicants, should further relevant facts become available. And it must be borne in 46 Section 80J. 47 In the Five Tax Cases above n 6, relating to the present matter, it was explained thus in paras 310-11: “The applicants were in reality raising a type of exception: namely that the facts alleged by SARS did not sustain, and were indeed irreconcilable with, the following two conclusions: (a) that the applicants were ‘parties’ to the alleged impermissible tax avoidance arrangement; (b) that, if the applicants were parties, they had received a ‘tax benefit’ and could be subjected to a GAAR assessment. . . . The interpretation of the letters of 19 October 2019, which superseded the section 80J notices, is a matter of law. Although the applicants have not said so, I do not doubt that they would accept that SARS is entitled to the benefit of any reasonable interpretation of which the letters are capable, in much the same way as on exception a court will adopt any reasonable interpretation of the pleading that avoids the ground of exception.” MAJIEDT J 24 mind, as the Supreme Court of Appeal has correctly observed, that SARS has unequivocally contested the correctness of Absa’s statements regarding lack of knowledge, 48 a fact reiterated by this Court in the Five Tax Cases. 49 As was pointed out in this Court’s earlier judgment in the Five Tax Cases, in terms of rule 36(1), it is possible for a rule 31 statement “to include a ground of assessment or a ground for opposing the appeal that was not among the grounds identified in SARS’ section 96(2) notice or in any other notice requiring SARS to identify grounds of assessment”. 50

[47]On this basis, it must be accepted at this stage that the applicants did not know what was happening with the downstream transactions. They must also be accepted as having acted in good faith up to and including their preference share subscriptions with PSIC3; what transpired further fell outside their knowledge. This is borne out by SARS’ assessment letters, as the applicants rightly contend. But, as stated, this is merely an assessment and the litigation process still has some way to go, and SARS must still issue its statement of case.

[48]In interpreting the applicable statutory provisions, fidelity to the statutory text and structure, the legislative intent underlying the 2006 amendments and the constitutional principles that govern the exercise of fiscal authority are required. Interpretation is a unitary exercise, considering text, context and purpose together. 51 As I understand their case, the applicants do not for purposes of the “exception” take issue with SARS’ contentions that what had happened downstream was an impermissible avoidance arrangement. The only questions at this stage are: (a) whether Absa was a party to that impermissible arrangement; and (b) whether it got a tax benefit from the impermissible arrangement. Before deciding these two issues, it is necessary to consider the jurisdictional requirement relating to an impermissible avoidance arrangement. 48 Supreme Court of Appeal judgment above n 11 at para 30. 49 Five Tax Cases above n 6 at para 272. 50 Id at para 18. 51 Chisuse v Director General, Department of Home Affairs [2020] ZACC 20; 2020 (6) SA 14 (CC); 2020 (10) BCLR 1173 (CC) at para 52. MAJIEDT J 25 Jurisdictional requirement in respect of an impermissible avoidance arrangement

[49]The applicants submit that in this analysis regarding the jurisdictional requirement in section 80A, there should also be a consideration of section 80B. I disagree. That section deals with the consequences of a finding that there is in fact such an arrangement. Instead, section 80G is the provision which has a bearing on this question.

[50]The jurisdictional requirement is the existence of an avoidance arrangement whose sole or main purpose is to obtain a tax benefit and which is abnormal. This is an objective enquiry. There is no requirement in the GAAR, and in section 80A in particular, that the taxpayer against whom the GAAR is invoked must personally have secured that benefit. It is sufficient that such an arrangement existed, and, once established, SARS’ remedial powers under section 80B extend to determining the tax consequences for “any party”. Section 80A is the heart of the GAAR, where an “impermissible avoidance arrangement” is defined. Sections 80B-L simply expand on, provide remedies for and deal with related procedural and administrative aspects of, the rule enunciated in section 80A.

[51]The finding that the jurisdictional requirement entails an objective enquiry is fortified by section 80G. That section requires of the party obtaining a tax benefit to “prove that, reasonably considered in the light of relevant facts and circumstances, obtaining a tax benefit was not the sole or main purpose of the avoidance arrangement” (emphasis added). This wording is 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. That is a fundamental change from section 103(1) which the courts by and large, particularly the Supreme Court of Appeal in Conhage, interpreted to relate to the taxpayer’s subjective purpose. 52 There can consequently be no doubt 52 Conhage above n 22 at para 12. See also Secretary for Inland Revenue v Gallagher 1978 (2) SA 463 (A); 40 SATC 39 at 471B-D, where Corbett JA stated: MAJIEDT J 26 that the present instance is the type of impermissible avoidance arrangement that falls under the GAAR provisions. And, as stated, the applicants do not appear to contest that what had happened downstream in this scheme constitutes an impermissible avoidance arrangement. The second, related, question is how “any party” in section 80L is to be interpreted. “Any party”

[52]Section 80L defines this concept. 53 There are two ways of interpreting section 80L. A narrow reading as advanced by the applicants, and endorsed by the second judgment, penned by my Colleague Rogers J, would equate “participation” with volition and intention. On this approach, a taxpayer cannot be said to “take part” in steps that are neither known to nor contemplated by it. This narrow approach seems to me to be overly literal.

[53]The second approach, and the one to which I subscribe, is that of a broader construction which rests on a purposive interpretation. To limit “party” to those with full knowledge would frustrate the purpose of the provisions and would recreate the very loopholes the GAAR was enacted to close. Our law unequivocally sets its face against those who seek to avoid liability through wilful ignorance. 54

[54]On the second approach the enquiry is objective: whether the taxpayer’s conduct forms part of the chain of transactions constituting the arrangement, not whether the taxpayer knew how each downstream step operated. It is no answer, on this interpretation, to say that the taxpayer “could not see inside the box”, if the taxpayer consciously placed itself outside the box. This is exactly what had prompted “By an objective test in this context is evidently meant a test which has regard rather to the effect of the scheme, objectively viewed, as opposed to a ‘subjective test’ which takes as its criterion the purpose which those carrying out the scheme intend to achieve by means of the scheme. . . . Section 103(1) 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’.” 53 Full provision quoted in n 39 above. 54 See R v Myers 1948 (1) SA 375 (A) at 382 and Banda v Van der Spuy [2013] ZASCA 23; 2013 (4) SA 77 (SCA) at para 22. MAJIEDT J 27 the move away from a subjective approach to an objective analysis, by expanding the purpose requirement to include an objective assessment of the purpose of an avoidance arrangement. This objective enquiry occurs 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.

[55]With the GAAR, it is plain that SARS sought to counter a scenario where tax avoidance steps are inserted somewhere between the beginning and end of a legitimate overall arrangement and an argument is then advanced that the arrangement as a whole was concluded mainly for a legitimate business purpose, and not for the main purpose of avoiding tax. SARS wanted to be able to focus on the avoidance steps inserted midway in the arrangement. SARS’ objective was to be entitled to argue that those particular steps constituted an impermissible tax avoidance arrangement, even if they were part of a broader legitimate arrangement.

[56]The two approaches self-evidently have very different consequences. It can be argued that the narrow view secures certainty and prevents the imposition of liability on investors whose role is genuinely incidental. But it risks enabling wilful blindness by institutional actors who benefit from arrangements while disclaiming knowledge of their structure. That would fatally undermine the very reason for the enactment of and, in particular, the significant 2006 amendment to the legislation. On the other hand, the broader approach advances the GAAR’s anti-avoidance purpose. The cognisable risk of overreach by attaching liability to any participant in a transaction later characterised as impermissible, even where the taxpayer’s involvement was commercially ordinary, cautioned against by the applicants, will be dealt with on a case-by-case basis. Axiomatically, this is a fact-based and objective enquiry.

[57]In my view, the statutory language (“participates or takes part”) strongly and persuasively admits of a purposive construction. As we have seen, that is also the stated objective in the Discussion Paper. When read in light of the mischief targeted MAJIEDT J 28 by the GAAR, the stronger position, and the one that I adopt, is that “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. Participation does not require omniscience, but it does require a conscious step into the structure from which the avoidance benefit flows. Thus, instead of requiring actual or constructive knowledge of the avoidance steps, there must be an objective determination whether there has been participation in the causal chain.

[58]The deliberate duplication of the synonyms “participates or takes part” underscores the Legislature’s focus on involvement, not mental state. If it was intended to impose a knowledge requirement, the Legislature could have adopted the familiar phrasing “knows or ought reasonably to have known” used in sections 26B, 29 and 60 of the Financial Intelligence Centre Act, 55 or in sections 28(1)(b)(ii) and 34(1) of the Prevention and Combating of Corrupt Activities Act. 56 This is not to say that knowledge or volition plays no role. It may well feature later in rebutting the presumption in section 80G. And it may also feature later in determining a taxpayer’s liability for understatement penalties in sections 222 to 223 of the TAA. 57 55 38 of 2001. 56 12 of 2004. 57 Section 222 of the TAA reads: “(1) In the event of an ‘understatement’ by a taxpayer, the taxpayer must pay, in addition to the ‘tax’ payable for the relevant tax period, the understatement penalty determined under subsection (2) unless the ‘understatement’ results from a bona fide inadvertent error. (2) The understatement penalty is the amount resulting from applying the highest applicable understatement penalty percentage in accordance with the table in section 223 to each shortfall determined under subsections (3) and (4) in relation to each ‘understatement’. (3) The shortfall is the sum of— (a) the difference between the amount of ‘tax’ properly chargeable for the tax period and the amount of ‘tax’ that would have been chargeable for the tax period if the ‘understatement’ were accepted; (b) the difference between the amount properly refundable for the tax period and the amount that would have been refundable if the ‘understatement’ were accepted; and (c) the difference between the amount of an assessed loss or any other benefit to the taxpayer properly carried forward from the tax period to a succeeding tax period and the amount that would have been carried forward if the MAJIEDT J 29

[59]The inclusion of the phrase “indirectly” militates against the narrower reading, as it confirms an intention to capture derivative or facilitative participation. This is precisely the scheme here, where Absa’s capital injection enabled downstream avoidance. 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 investment, the downstream tax-avoidant transactions would not have been possible.

[60]To accept the applicants’ contentions would create a dangerous precedent that allows investors to profit from tax avoidance simply by remaining ignorant, even wilfully so, of the schemes into which their funds are channelled. Such an approach would incentivise organisations to design tax avoidance schemes as a service to financial institutions and investors. The designing entity would simply tell the investor that they have a transaction structure that can achieve a high rate of return and not to worry about tax liabilities. If “party” were to be interpreted to require knowledge, these schemes would be allowed to proliferate beyond the scope of the GAAR simply by actors purposefully keeping investors ignorant of the details of the downstream transaction’s tax avoidant nature. Therefore, both the statutory text and the purpose of the GAAR strongly support SARS’ view that to be considered a ‘understatement’ were accepted, multiplied by the tax rate determined under subsection (5). (4) (a) If there is a difference under both paragraphs (a) and (b) of subsection (3), the shortfall must be reduced by the amount of any duplication between the paragraphs. (b) Where the ‘understatement’ is the failure to submit a return, the ‘tax’ that resulted from the 'understatement', had the ‘understatement’ been accepted, for purposes of subsection (3), must be regarded as nil. (5) The tax rate applicable to the shortfall determined under subsections (3) and (4) is the maximum tax rate applicable to the taxpayer, ignoring an assessed loss or any other benefit brought forward from a preceding tax period to the tax period. (6) Any penalty imposed under subsection (2) must be reduced by any penalty imposed under section 4(2) of the Employment Tax Incentive Act, 2013, in respect of the same employment tax incentive amount.” Section 223 contains the understatement penalty percentage table. MAJIEDT J 30 “party” to an impermissible tax avoidant scheme, one need not have knowledge of the avoidant nature of the scheme.

[61]As stated, generally speaking, the law does not countenance liability being avoided through deliberate blindness. On the objective facts, this is a scheme with the requisite unity amongst the composite parts, including the further downstream conduits, that tie the several transactions into a deliberate chain. 58

[62]There is a further compelling reason why a wide construction is to be preferred – the use of the word “any”. That is self-evidently a word of wide ambit. This Court said in Kham that “any” is “extremely broad”. 59 The Court cited the dictum of Innes CJ in R v Hugo: “‘[a]ny’ is, upon the face of it, a word of wide and unqualified generality. 60 It may be restricted by the subject-matter or the context, but prima facie it is unlimited”. 61

[63]I have read the judgment of my Colleague, Rogers J. He adopts a different approach and reaches a contrary conclusion to mine. In respect of the present topic, my Colleague states that “it cannot sensibly be said that someone was a ‘party’ . . . if the person did not know that the thing existed or was to be done. . . . One cannot participate or take part in an arrangement which one doesn’t know exists”. 62

[64]This in effect treats the proposition as a self-evident linguistic truism, and yet the statutory scheme in fact treats GAAR “arrangements” as objective anti-avoidance constructs, not bilateral consensual relationships. It mistakenly assumes that “arrangement” carries its ordinary contractual and consensual meaning, but, as stated, 58 Louw above n 35 at 572F. 59 Kham v Electoral Commission [2015] ZACC 37; 2016 (2) BCLR 157 (CC); 2016 (2) SA 338 (CC) at para 39. 60 R v Hugo 1926 AD 268 at 271. 61 Id. 62 See the second judgment at [121]. MAJIEDT J 31 the Legislature expressly decoupled the GAAR from subjective intention in the 2006 reforms.

[65]There is a further problem with this line of reasoning – it is circular, inasmuch as knowledge is required because an “arrangement” supposedly implies understanding, and yet an arrangement implies understanding because knowledge is required. This reasoning does not take into account that Parliament calculatedly chose the term “any party”, a formulation irreconcilable with a requirement of bilateral consensus, and one deliberately broader than “persons who agreed to the arrangement”. It bears repetition that GAAR treats avoidance as a composite scheme whose juridical unity does not depend on agreement.

[66]If the Legislature intended “party” to require knowledge, it could easily have done so. Instead, it used a non-mental-state verb, namely “participates”, and then placed all knowledge-related arguments in a different part of the statute (e.g., the presumption in section 80G, the penalty regime). This stark statutory design choice receives no consideration at all in the second judgment.

[67]The analogy of X giving Y a lift to a place without knowing that Y intends to murder someone there, and that X cannot be a “party” to the murder, is inapt. 63 It imports criminal law culpability into a statutory anti-avoidance regime. Liability for murder turns on mens rea; GAAR participation does not. The GAAR is not concerned with moral fault, but with structural causation. The GAAR hinges on whether a taxpayer’s acts form part of a composite scheme yielding a tax effect that Parliament disfavours.

[68]The effect of this flawed analogy is to transpose a doctrinal universe defined by intention into a statute that explicitly and deliberately moved away from intention. The second judgment’s concession elsewhere relating to subjective purpose appears to 63 Id. MAJIEDT J 32 me to undermine reliance on a culpability analogy to determine whether one “participates”. 64

[69]My colleague’s explication relating to possession crimes in the criminal law and to other fields of the law where “legally relevant conduct usually contains a component of intent” is singularly unpersuasive. 65 Elevating prior mental awareness of the existence of an arrangement to a definitional prerequisite to being a “party” for the concept of “participation” in section 80L, as the second judgment seeks to do, is simply not borne out by the history, text and structure of the GAAR. That approach subverts the explicit purpose of the 2006 amendments to do away with subjective intent, permitting subjectivity to sneak in through the back door.

[70]The point is simply that the GAAR does not allocate moral blame or criminal culpability; it allocates fiscal consequences by reference to structural causation. The question is not whether the taxpayer cognitively apprehended each step, but whether, viewed objectively, its conduct formed a constitutive link in the arrangement.

[71]Moreover, and in any event, the further illustrations also obfuscate matters. The examples provided relating to contracts, wills and company membership concern the validity of voluntary private law acts, not the scope of a regulatory power designed to neutralise avoidance structures. The question asked by section 80L is whether a person “participates or takes part” in an arrangement; it does not ask whether the person understood the arrangement in its entirety. To insist on full prior knowledge would render upstream capital providers structurally immune to the GAAR by deliberate informational nescience, an outcome inconsistent with both the breadth signalled by the phrase “any party” and the objective design of sections 80A and 80G. Considerations of fairness are addressed elsewhere in the statutory scheme, including through rebuttal and penalty provisions; they cannot justify reading into section 80L a 64 See the second judgment at [123]. It is stated there that “[d]etermining the purpose of an arrangement, whether subjectively or objectively, tells one nothing about who the parties to it were”. 65 Id at paras

[124]to [127]. MAJIEDT J 33 knowledge requirement the Legislature conspicuously omitted through the 2006 amendments.

[72]My Colleague cites Lord Denning’s dictum in Newton. 66 But that case must be understood in its proper historical setting. Newton was decided (in 1958) during an era-specific UK doctrine before the existence of modern GAARs. It is inapposite to treat Newton as definitional of “arrangement” under a post-2006 South African GAAR drafted explicitly to overcome the weaknesses of the consensual approach, which SARS criticised extensively in its Discussion Papers. Treating Newton as definitional of the current GAAR is thus anachronistic: it imposes a jurisprudence built around a pre-GAAR contractual and consensual paradigm onto a statutory regime designed to operate independently of consensus, precisely due to the challenges caused by the previous consensual regime.

[73]The second judgment fails to explain why South Africa, having rewritten GAAR to escape these types of limitations, like that in BNZ which my Colleague cites, 67 should be bound by Newton’s pre-GAAR nomenclature. The comparative reliance is selective and ill-fitting; modern Canadian and Australian authorities that reject knowledge-based participation are more suited to our present system under the GAAR.

[74]The second judgment then goes on to express its disquiet about the unfairness “for a taxpayer to suffer adverse tax consequences in respect of an arrangement of which they knew nothing”. 68 This appeal to fairness obscures the deeper structural problem – this approach renders the entire upstream financing layer of a multi-entity avoidance structure immune to the GAAR simply by withholding information from the funder. That incentive structure is precisely what the 2006 reforms aimed to 66 Id at [122]; Newton v Commission of Taxation of Commonwealth of Australia [1958] 2 All ER 759 (PC) (Newton). 67 Commissioner of Inland Revenue v BNZ Investments Ltd [2001] NZCA 184; [2002] 1 NZLR 450 (BNZ) referred to in the second judgment at fn 100. 68 See the second judgment at [128]. MAJIEDT J 34 neutralise. With this approach, ignorance (whether passive or engineered) becomes a shield, enabling financiers and institutional actors to remain immunised against the GAAR by ensuring that downstream avoidance steps are opaque.

[75]The second judgment does acknowledge wilful blindness as conceptually possible, yet because it makes knowledge a gateway requirement to being a “party”, wilful blindness can never matter unless SARS pleaded it upfront. This turns GAAR participation into a pleading battleground from the very start, rather than a structural analysis of parties’ involvement in tax arrangements.

[76]The second judgment then concludes on the basis of the facts set out in the assessment that, while Absa was a party to some of the steps in the scheme, it had no knowledge of some of the other steps occurring. 69 This concern is based on a faulty premise, not to be found in the legislation. Section 80L asks whether the entity “participates” in an arrangement, defined as the composite scheme. It does not ask whether the entity knows each sub-step. Properly understood with GAAR, participation is the objective involvement in a constitutive step: in the present instance it means that without Absa’s capital injection, no downstream steps exist. This does not receive proper attention in the second judgment. The approach propounded in the second judgment effectively creates a rule that the only “parties” are the architects and insiders – precisely the parties least likely to be assessable or solvent.

[77]In the premises, there can be little doubt that Absa was a party as envisaged in the provision. The next aspect for consideration is whether Absa derived a tax benefit from the arrangement. “Tax benefit”

[78]Section 80B(1) empowers SARS to determine the tax consequences for “any party”. The deliberate breadth of this phrase evidences legislative intent for a wide 69 Id. MAJIEDT J 35 remedial reach. Had Parliament intended to restrict liability to the party “obtaining a tax benefit”, it would have repeated the phrase from section 80G. So, even if Absa did not in fact receive a tax benefit, it would have been covered by this wide remedial reach. But, for the reasons that follow, I hold that, in any event, Absa did in fact receive a tax benefit.

[79]In this assessment, a transaction must be assessed “stripped of its avoidance features”, meaning it must be viewed as if the artificial elements serving no genuine commercial purpose were removed. The enquiry is therefore directed at the economic substance of the arrangement rather than its formal structure. On this view, Absa’s receipt of tax-exempt preference dividends was functionally equivalent to earning taxable interest (the dividend form merely masking what was, in substance, a loan yield). This re-characterisation confirms that Absa derived a tax benefit within the meaning of section 80A, as the exemption flowed solely from structural manipulation rather than commercial reality.

[80]On this approach, the GAAR analysis places substance over form, ensuring that the statute targets the fiscal effect of avoidance features rather than the taxpayer’s chosen labels. It thereby preserves the GAAR’s remedial purpose, aligning legal outcomes with economic reality and preventing formal compliance from defeating substantive taxation. A proper assessment of a scheme to test it against the GAAR provisions must avoid a scenario where an untenable outcome is produced in multi- layered schemes, with some of the actors downstream possibly being in a position to deny liability, leaving the GAAR with very little effect. A broader reading closes this loophole by including all objective participants within SARS’ remedial jurisdiction, while allowing factual differentiation of liability on a case-by-case basis.

[81]The role that the presumption in section 80G plays must be properly understood. That section addresses the evidentiary burden, not the scope of liability. The “party obtaining a tax benefit” merely identifies who bears the onus of rebuttal. Sections 80A, 80B and 80G must follow a coherent sequence, where section 80A MAJIEDT J 36 defines impermissible tax avoidance, section 80B describes the consequences of such avoidance and section 80G governs proof. One must guard against a reading of these sections that would fragment the statutory scheme and create internal inconsistency.

[82]SARS is correct that, even if there was a requirement to show that Absa derived a tax benefit, the evidence demonstrates that it did. But for the artificial structuring of the transactions, with the further downstream entities plainly being mere conduits, Absa’s investment returns would have been taxable interest or a smaller, tax- exempt dividend. Instead, through the arrangements, Absa received inflated tax- exempt dividends, which constituted a clear financial advantage at the expense of the fiscus. What brings this arrangement under the GAAR, that is, renders it an impermissible tax avoidance arrangement, are the roles played by the D1 Trust and PSIC4. I agree with SARS that the counterfactual advanced by the applicants is misconceived: the proper comparison is not with “no transaction”, but with the transaction stripped of its avoidance features. On this test, Absa plainly obtained a tax benefit.

[83]The GAAR specifically empowers SARS to identify tax avoidant schemes, locate intermediary entities that have no legitimate business purpose beyond tax avoidance and ignore those entities when tracing the recipient of the tax benefit. That is exactly what happened in this case – SARS identified the avoidant nature of this scheme and declared that PSIC3, PSIC4 and the D1 Trust solely operated to funnel funds through themselves to eliminate a tax liability. This created a tax benefit that flowed to Absa, which was the first legitimate entity to receive the funds following the wrongful elimination of the tax liability.

[84]The correct “but-for” test to determine if a tax benefit arose requires assessing whether, but for the tax avoidant features and dressing up of the transaction, a tax liability would have occurred. Enquiring whether a tax benefit occurred but for the transaction itself, as Absa has argued, is the incorrect test, and ignores the realities that MAJIEDT J 37 both tax liabilities and benefits arise within the scheme rather than the scheme only being used to eliminate pre-existing tax liabilities.

[85]Could SARS have taxed one of the conduits in the chain in this arrangement? I think not – it would be insensible to expect SARS to do so, because the conduits simply received the tax benefit in order to pass it on to its ultimate destination, Absa. The funds it had invested were utilised for subscription in the PSIC4 shares and that was in turn applied to make capital contributions to the D1 Trust. The latter, on its part, applied the proceeds of these capital contributions to make interest-bearing loans to MSSA.

[86]The second judgment suggests that SARS could have taxed the actual tax beneficiaries, PSIC4 and/or the D1 Trust. 70 I disagree. First, to reiterate, stripped of the arrangement’s anti-avoidance features, the tax benefit accrued to Absa and United Towers. SARS in its assessment letter made this abundantly clear. It summarised the tax benefit to these two entities. 71 It quoted the definition of a “tax benefit”, set out applicable case law in that regard and then explained its understanding of how the arrangement worked. 72 There, SARS also pertinently described PSIC4 as a mere conduit.

[87]Second, this question must be determined on the facts. I hold (and repeat for emphasis) that the arrangement, on the facts, worked thus. A total of 13 entities operated in this arrangement. Absa subscribed for preference shares in PSIC3, which used the proceeds to subscribe for preference shares in PSIC4. The latter, in turn, used the proceeds to make a capital contribution to the D1 Trust and them, in turn, used these contributions to make interest-bearing loans to MSSA. Interest on these loans was paid by MSSA to the D1 Trust, which used this interest income to acquire a tax-free income stream derived from Brazilian government bonds. The D1 Trust then 70 Id at

[136]and [141]. 71 SARS letter of assessment at para 11. 72 Id at paras 68-78. MAJIEDT J 38 distributed this tax-free income stream to PSIC4, which in turn used it to pay preference dividends to PSIC3, which used these dividends to pay preference dividends to Absa.

[88]Plainly then, in effect, the purpose was to swap a taxable income stream paid by MSSA to the D1 Trust for a tax-free income stream paid by the D1 Trust to PSIC4, PSIC3 and, ultimately, to Absa. Absent this swap, PSIC4 would have been liable for tax on the interest received from the D1 Trust, but the swap caused the interest received to be tax-free. Ultimately then, this enhanced the dividend stream from PSIC4 to PSIC3 and then, in the end, to Absa. Applying the GAAR to what I hold was an arrangement to avoid tax, it is Absa (and United Towers) who actually gained the tax benefit and not PSIC4 or the D1 Trust.

[89]The second judgment deals with the tax benefit in a different way. It relies on several propositions to hold that Absa did not receive a tax benefit. First, it seeks to answer the question whether there was a tax benefit, by contrasting what actually happened with a plausible counterfactual. 73 It then states that “[t]he plausible counterfactual is self-evident . . .simply think away the Brazilian swap. . . . Everything else can stand without step (g) [the Brazilian interest swap].”

[90]This counterfactual appears to conflate two different questions: (a) whether an arrangement yields a tax benefit; and (b) whether Absa is the beneficiary when the avoidance features are disregarded.

[91]If the swap is removed and the structure collapses into an economic loan, the tax character of Absa’s return on its capital shifts from exempt dividends to taxable interest. The second judgment does not engage with this re-characterisation analysis; it treats the dividend exemption as unshakeable, even within a GAAR reconstruction. It also seeks to counter this reasoning by asserting that “Absa’s position was, quite 73 See the second judgment at [135]. MAJIEDT J 39 simply, unrelated to the tax treatment of other elements of the arrangement” and that “removing the Brazilian swap does not convert Absa’s return on its preference shares from exempt dividends to taxable interest”. 74 This is a misconception on at least two levels, the nature of the counterfactual enquiry and the locus of the tax benefit under the GAAR.

[92]First, the comparison is inapt. It is not, as suggested by the second judgment, between two levels of exempt dividends in Absa’s hands, but between the actual arrangement and the arrangement stripped of its avoidance features. Once the Brazilian swap and associated conduit steps are removed, the economic substance of Absa’s investment is no longer that of a genuine preference share yielding exempt dividends, but of a risk-neutral, yield-protected funding instrument producing a return functionally indistinguishable from interest. The form of Absa’s receipt cannot be determinative where that form is itself the product of the avoidance structure. To frame the counterfactual at the level of PSIC3’s dividend-paying capacity is to accept the avoidance architecture as given, rather than to interrogate whether that architecture is what generated the exemption in the first place. This is what the GAAR in its present amended form requires us to do.

[93]Furthermore, the existence of a gross-up guarantee does not sever the causal link between the downstream tax benefit and Absa’s tax position. The guarantee is part of the same composite arrangement that neutralised risk, fixed Absa’s yield and ensured that Absa would receive a tax-exempt return regardless of the fiscal consequences elsewhere in the structure. That insulation is precisely what reveals the arrangement’s lack of commercial substance. The GAAR is concerned with the allocation of tax consequences by reference to objective effect, not with whether the taxpayer was economically indifferent to how the benefit was achieved. To permit Absa to rely on contractual insulation to deny receipt of a tax benefit would allow the very hallmark of impermissible avoidance, that is, risk-free, tax-engineered returns, to defeat the operation and purpose of the GAAR. 74 Id at [138]. MAJIEDT J 40

[94]Moreover, the counterfactual proposed (removing only the Brazilian interest swap) begs the question – the GAAR removes all avoidance-oriented features, not only the one step that SARS emphasised. Under the correct counterfactual, the structure reduces to what is economically a loan, and Absa’s yield correspondingly takes the form of taxable interest. The second judgment’s selective removal narrows what can count as a “benefit” and risks inverting the GAAR methodology.

[95]My Colleague cautions against confusing a tax benefit with an economic advantage. 75 In principle that is of course correct, but the question is not whether Absa avoided tax in a narrow sense, but whether the dividend exemption is itself a product of the arrangement’s tax-driven restructuring. It is unavailing for a taxpayer to dress up its return on capital as a tax-exempted dividend when in truth and in fact that return is the product of avoidance planning.

[96]If one strips away avoidance features, it converts what Absa received into interest, and then Absa did obtain a tax benefit. The avoidance steps caused the return to be dressed up as a dividend. The second judgment skirts this point by fixing the dividend label as immutable.

[97]The second judgment reasons that section 80G(1) “necessarily conveys that the party SARS is entitled to tax is the party which got the tax benefit”. It states that it “would be absurd to impose a burden of proof on a non-party”. 76 But this conflates the allocation of a tax burden with the scope of liability. Section 80G creates a presumption about purpose, not about who can be assessed. Nothing in the text limits section 80B’s remedial scope to the beneficiaries of the tax benefit.

[98]The second judgment’s argument requires reading the phrase “party obtaining a tax benefit” into section 80B – effectively redrafting the statute. The Legislature 75 Id at [137]. 76 Id at [141]. MAJIEDT J 41 could have aligned section 80B with section 80G, but it did not. Rather than explain that structural choice, the second judgment treats misalignment as inadvertence.

[99]It is stated in the second judgment that “[t]he current GAAR and the revisions of the repealed section 103(1) of the ITA are not materially different”. 77 This underplays the significance of the 2006 reforms, which expressly addressed the shortcomings of section 103(1), including its failure to capture multi-party structured finance and its over-reliance on subjective purpose and doctrinal narrowness. The shift to objectivity, series analysis and purposive statutory interpretation fundamentally altered the GAAR’s methodology. To state that the statutes are “not materially different” evades the Legislature’s explicit repositioning of the GAAR as substance-oriented, objective and system-protective. The second judgment effectively resurrects the pre-2006 jurisprudence which the Legislature intended to override.

[100]The second judgment engages in an extensive discussion of the BNZ judgment. But it bears emphasis that BNZ must be understood against the contextual backdrop that reliance was placed on statutory wording (“arrangement or understanding”) very different from section 80L at issue here. Moreover, Peterson, also cited by my Colleague, when properly read, rejects the idea that lack of knowledge prevents GAAR consequences for someone who obtained a tax advantage. The second judgment acknowledges this but tries to minimise it by stressing that New Zealand legislation also contained the wording “person affected”.

[101]Even more fundamentally, as stated, the second judgment compares South Africa’s GAAR to New Zealand’s earlier, more restrictive version rather than to modern Canadian and Australian models which better resemble our sections 80A-80L. This creates a comparative imbalance.

[102]Lastly, the second judgment observes that my holding that a GAAR assessment may be issued against a person who (a) was unaware of the impermissible tax 77 Id at [144]. MAJIEDT J 42 avoidance and (b) obtained an economic advantage but not a tax benefit from it. It says that this appears to be “unprecedented internationally”. 78 Not so – this conclusion is reached only by omitting OECD 79 -style GAARs. The Canadian decision in Copthorne Holdings, for instance, imposes GAAR consequences on taxpayers who did not know all steps, so long as their transactions form part of the series producing the tax advantage. 80

[103]Australia’s Part IVA (post-2013) similarly disregards subjective awareness and focuses on objective contribution to a scheme. Plainly then, the second judgment misstates the novelty of the approach adopted in this judgment.

[104]The conclusion is reached in the second judgment that SARS was not bereft of a remedy here, as it could, in terms of section 80B(1)(c), exercise the power to deem persons who are “connected persons” in relation to each other “to be one and the same person”. In this way, SARS could in all likelihood have reached other entities of substance within the Macquarie. 81 But this evades two key structural issues: (a) GAAR is designed precisely to prevent avoidance structures from insulating the true economic recipient via conduits. Merely taxing PSIC4 does not neutralise Absa’s enhanced yield unless Absa’s return is also reconstructed. (b) Section 80B(1)(c) allows SARS to treat connected persons as one and the same person. The dissent acknowledges this but does not confront the implication that the GAAR explicitly contemplates that the assessment may reach beyond the technical holder of the income.

[105]The alternative offered (tax the empty shells and their connected group) risks leaving the upstream funder’s outcome untouched even where the economic return is 78 Id at [155]. 79 Organisation for Economic Co-operation and Development. 80 Copthorne Holdings Ltd v Canada 2011 SCC 63; [2011] 3 SCR 721 (Copthorne Holdings). 81 See the second judgment at [158]. MAJIEDT J 43 tax-shaped. A further premise running through the second judgment, albeit not explicit, is that the GAAR should preserve the integrity of each bilateral transaction unless the taxpayer knowingly crosses into avoidance terrain. This assumption appears in the repeated insistence that Absa’s preference share subscription stood on its own legs and that downstream steps were entirely separate. Yet the GAAR was enacted precisely to prevent taxpayers from compartmentalising complex structures into insulated bilateral components. Section 80A’s composite-purpose enquiry, combined with section 80L’s definition of “arrangement”, rejects the idea that the legality or commerciality of one step immunises it from inclusion in an overarching avoidance structure.

[106]By implicitly reintroducing transactional fragmentation, my Colleague reinstates a pre-2006 analytical frame that Parliament deliberately abandoned. The question under the GAAR is not whether each leg of the transaction is commercially explicable, but whether the composite arrangement, viewed holistically, yields an impermissible tax advantage and whether the taxpayer’s step is causally indispensable to that outcome. The dissent never engages with this compositional logic, and without doing so, its analysis cannot account for the GAAR’s structural focus.

[107]Finally, the second judgment relies on an equivalence that is doctrinally and textually unsupported, that ignorance equates to non-participation. This assumption allows my Colleague to conclude that Absa, unaware of steps (b)-(j), cannot be a party to the arrangement that produced the tax benefit. But that equivalence collapses once one recognises that the GAAR’s statutory test for participation is functional, not epistemic. Section 80L requires that the taxpayer “participates or takes part”, not that the taxpayer knows or intends the downstream avoidance consequences. If Parliament sought to make knowledge a prerequisite, it would have adopted familiar formulations such as “knows or ought reasonably to have known”. Instead, the statute assesses participation by reference to objective causation: did the taxpayer’s act form a constitutive element of the structure generating the avoidance result? Ignorance does not negate objective participation any more than ignorance of a downstream effect MAJIEDT J 44 negates causation in regulatory fields such as competition law, environmental law or anti-money-laundering. By equating non-knowledge with non-participation, the second judgment imports a mental-state qualifier that the statute deliberately avoids and that would, if accepted, defeat the GAAR’s purpose by creating a ready-made template for plausible deniability through deliberate informational silos.

[108]In summary then, the returns from MSSA were shielded from income tax in South Africa by swapping the taxable income stream for a tax-exempt Brazilian bond income stream, and a further conversion into local dividend income (also tax-exempt). Thus, in the final analysis, Absa received an enhanced return on its initial investment solely by reason of the fact that its funds had been used in an impermissible tax avoidance arrangement. There is one last substantive aspect that bears consideration in this analysis – the comparative international law. International law

[109]As stated, the Discussion Paper alludes to similar legislation in certain other countries having been considered when the 2006 amendment was prepared. 82 Comparative jurisprudence supports this broad interpretation of “party” under section 80L. The Supreme Court of Canada in Copthorne Holdings held that the Canadian GAAR must be applied to the series of transactions as a whole, and that even steps not individually abusive may constitute avoidance when viewed in broader sequence. 83 That Court stressed that the test is objective: whether the taxpayer’s participation formed part of a series that produced a tax benefit, not whether the taxpayer subjectively appreciated the mechanics of the scheme.

[110]This approach mirrors the mischief which South Africa’s GAAR was designed to prevent – sophisticated actors fragmenting arrangements to preserve plausible deniability. Just as Copthorne Holdings rejected a piecemeal view of isolated 82 Discussion Paper above n 27. 83 Copthorne Holdings above n 80 at para 43. MAJIEDT J 45 transactions, so too should section 80L reject a narrow focus on what Absa knew of downstream steps. Once Absa knowingly entered one leg of a structure that, viewed objectively, yielded an avoidance outcome, it “participated” within the meaning of the Act.

[111]Similarly, Australian jurisprudence under Part IVA of its Income Tax Assessment Act of 1936 offers persuasive support for a broad reading of participation in tax avoidance arrangements. The 2013 amendments to Part IVA, enacted to counter the “do nothing” defence and to narrow the scope of hypothetical alternatives available to taxpayers, underscore the legislature’s determination to prevent avoidance through formalism or deliberate blindness. The recent decisions in FCT v Guardian AIT Pty Ltd, ATF Australian Investment Trust 84 and Minerva Financial Group Pty Ltd 85 provide the first judicial interpretations of these reforms. In Guardian, the Full Federal Court engaged directly with section 177CB and found that the structuring and timing of the scheme were primarily tax-driven, concluding that the dominant purpose was to obtain a tax benefit. This confirms that the test is an objective one, in which courts will disregard tax-motivated hypothetical postulates and focus on the commercial substance of the arrangement.

[112]By contrast, in Minerva, while a tax benefit was present, the Court was satisfied that the dominant purpose was commercial (preparing for an initial public offering and streamlining operations), thus allowing the taxpayer’s defence. Taken together, Guardian and Minerva illustrate that Part IVA requires more than mere subjective awareness of downstream steps; it imposes liability wherever participation forms part of a scheme that, viewed objectively, yields an avoidance result. The analytical emphasis on objective indications of purpose, rather than taxpayers’ professed ignorance, aligns closely with the South African GAAR’s purposive thrust. Just as Guardian narrowed the space for avoidance defences based on alternative 84 Commissioner of Taxation v Guardian AIT Pty Ltd, ATF Australian Investment Trust [2023] FCAFC 3; (2023) 115 ATR 316 (Guardian). 85 Minerva Financial Group Pty Ltd v Commissioner of Taxation [2024] FCAFC 28; (2024) 302 FCR 52 (Minerva). MAJIEDT J 46 hypotheticals, so too should section 80L of the ITA be construed to prevent taxpayers from disclaiming knowledge of steps while reaping the tax benefits of a broader scheme. Conclusion

[113]In sum then, on the objective facts as they stand at this juncture, this is an impermissible avoidance arrangement. There is no legal requirement that the specific taxpayer must have obtained the tax benefit, although on the objective facts Absa did in fact obtain one. The tax benefit was passed on from the D1 Trust through the PSIC4 transaction to the party which put up the funds for the arrangement and which ultimately received the financial benefit of the arrangement, Absa. It is a fallacy to argue that SARS ought to be restricted to applying the GAAR provisions only to the parties that directly avoided the tax, but by careful design were mere special purpose vehicles which were interposed in the arrangement as mere conduits. Monies merely flowed through these conduits. Put differently, if this transaction had not been dressed up with features designed to avoid the imposition of tax, Absa would have incurred the tax liability. Absent the tax avoidant features of the arrangement, it would, in effect, have been a loan by Absa. Absa was clearly a “party” to the impermissible avoidance arrangement.

[114]For all these reasons, the appeal must be dismissed and the order of the Supreme Court of Appeal must be confirmed, but for different reasons. Costs must follow the outcome. Order

[115]I make the following order: The appeal is dismissed with costs, including the costs of two counsel. ROGERS J 47 ROGERS J: Introduction

[116]I have had the benefit of reading the judgment of my Colleague, Majiedt J (first judgment). I disagree with the first judgment on both issues that arise in this case – the party issue and the tax benefit issue. The first judgment contains a summary of the transactions. A fuller description can be found in Five Tax Cases. 86 I shall, as was done in the latter judgment, refer to these components as steps (a) to (m). For convenience, I refer only to Absa, but what I say applies equally to United Towers. I adopt the abbreviations in Five Tax Cases. SARS’ key allegations in the Absa assessment letter are quoted in that case. 87

[117]Section 80A, in referring to the “sole or main purpose” of an avoidance arrangement, reflects a shift from a subjective determination of purpose to a more objective test. How complete that shift is can be left for another day. 88 The fact that the taxpayer may subjectively have had a different purpose does not preclude a 86 Five Tax Cases above n 6 at para 253. 87 Id at para 256. 88 The shift from a subjective to an objective test was emphasised in two documents issued by SARS in advance of the introduction of the new GAAR: see Discussion Paper above n 27 at para 10 p 48 and at para 10.3 p 56 and SARS Tax Avoidance and Section 103 of the Income Tax Act, 1962, An Interim Response (March 2006) (Interim Response) at para 5.1 pp 17-19. In a third document, Revised Proposals above n 43, which contained SARS’ proposed new GAAR substantially in the form later enacted, SARS retreated somewhat from the assertion that a purely objective test was intended – see at para 10.1 p 21: “The revised proposals would leave the current ‘sole or main’ purpose test unchanged. They would also modify the original proposal requiring the purpose of an arrangement to be determined ‘objectively’. Some commentators expressed concern that the proposed ‘objective’ purpose requirement might preclude the court from considering a taxpayer’s ipse dixit. It was never the intent of the original proposals to prevent a taxpayer’s explanation of the reasons for an arrangement from being taken into account. Rather, it was intended to ensure that a taxpayer’s statements of intent be rigorously tested against the relevant facts and circumstances. The revised proposals are intended to better reflect that intent and reinforce existing precedent in this regard.” In its Draft Comprehensive Guide to the General Anti-Avoidance Rule (2010) at paras 4.1 and 4.3 pp 20-1, SARS said that in the new GAAR the standard was “more objective” than it was under section 103(1). The learned authors of Silke regard this “coyness” as misplaced, and doubt whether “objective” admits of degrees of comparison— see De Koker and Williams Silke on South African Income Tax (Silke) Service 44 (2011) Volume 4 at 19-79. As I have said, the question need not be decided in this case. ROGERS J 48 finding that the sole or main purpose of an arrangement was to obtain a tax benefit. Cases where a taxpayer’s subjective purpose differs from the objective purpose of an arrangement may be rare, since the objective characteristics of an arrangement would be important material from which to infer the taxpayer’s subjective purpose. 89 For purposes of this judgment, I proceed on the assumption that the purpose test is wholly objective. The party issue

[118]With this as background, I start with the party issue. In terms of section 80B(1), the Commissioner’s powers to determine the tax consequences of an impermissible avoidance arrangement are confined to the tax consequences “for any party”. Since SARS has purported to determine tax consequences for Absa, the question is whether the assessment letter contains allegations disclosing that Absa was a “party” to the alleged impermissible avoidance arrangement.

[119]To answer that question, one first needs to know what SARS alleges the “arrangement” was and why SARS alleges that it was “impermissible”. The answer is clear. It is the totality of the steps summarised in Five Tax Cases. 90 In order to reach its taxing powers over Absa in terms of section 80B, SARS is bound to allege that step (a) – Absa’s subscribing for preference shares in PSIC3 – was part of an arrangement that included the remaining steps, most importantly step (g). Step (g) – the Brazilian interest swap – is critically important, because according to the assessment letter this is the step that justifies characterising the arrangement, objectively, as “impermissible”. More particularly, the Brazilian interest swap was 89 In SARS’ Interim Response id at p 17 and fn 70, SARS noted that, in response to the Discussion Paper id, some commentators had observed that the shift to an objective test would, to a large extent, “do no more than reinforce the approach the courts have generally taken under current law”, since “in most reported cases, courts have considered all the circumstances very carefully and, should these indicate the tax avoidance motive, the taxpayer’s arguments to the contrary have been overlooked”. The Revised Proposals id reflects some ambivalence as to the completeness of the shift. See also Roberts “Recognising Tax Avoidance: An Analysis of Pt IVA of the Income Tax Assessment Act 1936 (Cth)” (2006) 21 Australian Tax Forum 223 at 255: “It is hard to imagine a scheme which points to the conclusion that a participant had a dominant tax purpose, when actually he or she did not.” 90 Above n 6 at para 253. ROGERS J 49 allegedly inserted so as to permit reliance on Article 11(4)(b) of the double taxation agreement (DTA) between South Africa and Brazil and thereby obtain a tax benefit.

[120]Does the assessment letter contain allegations that Absa was a “party” to the arrangement thus defined? A “party” is defined in section 80L as a person who “participates or takes part in an arrangement”. I can discern no difference between “participates in” and “takes part in” – each expression is the primary definition of the other. And both of these expressions are in turn the primary and ordinary meaning of being a “party” to something.

[121]It cannot sensibly be said that someone was a “party” to something, or participated in it, or took part in it, if the person did not know that the thing existed or was to be done. If X gives Y a lift to a place without knowing that Y intends to murder someone there, X is not a “party” to the murder – X cannot be said to have “participated” or “taken part” in the murder. This is so even though Y may have advanced his nefarious purposes by getting a lift from X.

[122]So it is with participation in an arrangement. One cannot participate or take part in an arrangement which one does not know exists. This seems to me to be such an axiomatic proposition as hardly to call for substantiating analysis. As Lord Denning said in Newton: 91 “[T]he word ‘arrangement’ is apt to describe something less than a binding contract or agreement, something in the nature of an understanding between two or more persons – a plan arranged between them which may not be enforceable at law.” 92 (Emphasis added.)

[123]This has nothing to do, and should not be confused, with the shift from a subjective to an objective approach in getting at the sole or main purpose of an arrangement. Determining the purpose of an arrangement, whether subjectively or 91 Newton above n 66. 92 Id at 763g-h. ROGERS J 50 objectively, tells one nothing about who the parties to it were. The parties to an arrangement are those who know about it and intend for it to take place. Under the current GAAR, it matters not (so I have assumed) that one or more of those parties subjectively had a purpose different from the sole or main purpose as determined by the objective characteristics of the arrangement.

[124]The first judgment criticises my example of X giving Y a lift as turning on an inapt borrowing from criminal law, namely of mens rea (guilty mind). The first judgment says that this is inconsistent with my acceptance that the new GAAR has moved away from a subjective test for purpose. 93 What I have said in the preceding paragraphs shows that this criticism misses the point. I accept that being a party to an impermissible arrangement does not require that the alleged participant should know that the arrangement has the effect or purpose of avoiding tax. The knowledge of which I speak in relation to the party requirement is the knowledge inherent in the relevant conduct, the conduct of being a “party to” an impermissible avoidance arrangement.

[125]My example of X and Y happened to come from criminal law, but the same principle is encountered widely whenever the law makes it relevant to ascertain whether a particular form of conduct occurred. To stay for the moment with the criminal analogy, the knowledge that an arrangement has the sole or main purpose of avoiding tax could be regarded as the counterpart, in criminal law, of knowledge that one’s conduct is unlawful, which lies at the heart of mens rea. This does not mean that every aspect of a person’s state of mind goes to mens rea. Sometimes the conduct which the law criminalises makes knowledge or intention a definitional element of the actus reus (unlawful act). 94

[126]The most obvious examples are possession crimes, where knowingly having something under one’s control with a particular state of mind (to derive a benefit, to 93 See the first judgment at

[66]to [67]. 94 See generally, Snyman (updated by Hoctor) Criminal Law 7 ed (LexisNexis, Durban 2020) (Snyman) at 62-5. ROGERS J 51 control as owner etc) is part of the actus reus of possession, not mens rea. 95 Statutory crimes of participation, for example participating in a criminal gang, 96 must likewise necessarily require some element of knowledge or intention before one can say that the definitional element of the actus reus, participation, has occurred. This distinction is found in the common law doctrine of common purpose as well: in the absence of proof of prior agreement between the alleged participants, the doctrine requires proof – before one gets to mens rea – that the person was present at the scene, was aware of the crime that the others were committing there, and manifested an intention to make common cause by performing some act of association. 97

[127]Beyond the field of criminal law, legally relevant conduct usually contains a component of intent. A person cannot normally be said to have concluded a contract if they did not perform an act of signification with the intention of concluding a contract. A person cannot normally be said to have executed a will if they signed a document which they did not know purported to be their will. A person cannot be said to have become a member of a company if they did not know that the company existed or that their name was added to its register of members. And so, a person cannot participate in an impermissible avoidance arrangement if they do not know of the arrangement’s existence.

[128]The shift to an objective approach on purpose renders it all the more important to be careful about identifying the persons who can properly be said to have been “parties” to the arrangement. It seems unfair for a taxpayer to suffer adverse tax consequences in respect of an arrangement of which they knew nothing. There is nothing in the SARS discussion papers preceding the enactment of the new GAAR 98 95 S v Adams [1986] ZASCA 82; 1986 (4) SA 882 (A) at 891G-I; S v Jacobs [1989] ZASCA 127; 1989 (1) SA 652 (A) at 659D-G and 661C-D; and S v Cameron [2005] ZASCA 40; 2005 (2) SACR 179 (SCA); [2005] 3 All SA 18 (SCA) at para 10. See also Snyman id at 57. 96 Section 9(1) of the Prevention of Organised Crime Act 121 of 1998. 97 S v Mgedezi [1988] ZASCA 135; 1989 (1) SA 687 (A) at 705I-706C and Thebus v S [2003] ZACC 12; 2003 (6) SA 505 (CC); 2003 (10) BCLR 1100 (CC) at paras 20-1. 98 Discussion Paper above n 27. ROGERS J 52 to suggest that its provisions were intended to include, as “parties”, persons who were unaware of the anti-avoidance transactions.

[129]To know about an arrangement and to intend for it to take place may well include what is referred to in the first judgment as “wilful blindness”, which in our law could be regarded as a species of dolus eventualis (that is, where a person subjectively foresees the possibility that his conduct may have an unlawful consequence, but goes ahead in reckless disregard of that possibility). 99 However, the question does not arise here, because the assessment letter does not allege wilful blindness and does not allege facts from which an inference of wilful blindness could properly be drawn. SARS has made its position clear. It does not matter, says SARS, whether Absa knew about the steps in the arrangement beyond step (a). SARS alleges that it can assess Absa even though Absa did not have knowledge, and that includes knowledge in all its forms.

[130]I observe, in passing, that there might be good reason for a group such as Macquarie not to disclose details of its own transactions to a client such as Absa. The financial structuring devised by such a group “downstream” of the client’s “upstream” investment may have proprietary value which the group would not want to fall into the hands of competitors. 100 Keeping such information confidential does not in itself point to anything nefarious. Indeed, it would not be unusual for such a group to obtain legal opinion on the tax effects of the financial structuring, including vulnerability to anti-avoidance legislation. 99 In England, see Manifest Shipping Company Limited v Uni-Polaris Shipping Company Limited and Others [2001] UKHL 1; [2003] AC 469; [2001] 1 All ER 743 (HL) at paras 23-6, dealing with “blind eye knowledge” (or “Nelsonian blindness” as it is sometimes styled – deliberately putting the telescope to the blind eye: see for example Economides v Commercial Union Assurance Co plc [1997] EWCA Civ 1754; [1997] 3 All ER 636 (CA) at 648f-g and 653e-f). 100 Compare BNZ above n 67, which involved an “upstream” preference share investment by a Bank of New Zealand (BNZ) subsidiary and “downstream” transactions by Fay Richwhite, a merchant banking group that had made the preference share investment available to BNZ. The New Zealand Court of Appeal noted in para 13: “Importantly, the Judge found that Fay Richwhite had good commercial reasons for keeping this structure secret as far as and for as long as possible. It would not have wished to see that effort picked up and used by competitors, including BNZ.” ROGERS J 53

[131]So, to be a “party” to an arrangement requires that the taxpayer should know about it and intend for it to take place. This does not mean that the taxpayer need have knowledge of the tax consequences of the steps comprising the arrangement or of the fact that one or more of the steps will result in a tax benefit or may be impermissible tax avoidance. One can be a party to an arrangement without knowing about its legal consequences. In the present context, that means that the taxpayer must know of the steps comprising the arrangement, even if the taxpayer is ignorant of the tax consequences which might render the arrangement an “impermissible” one.

[132]On the basis of the facts stated in the assessment letter, to what “arrangement” was Absa a party? It was a party to step (a) – the subscription for shares in PSIC3. Relatedly, it was a party to steps (l) and (m) – the commercial protections Absa enjoyed in terms of its right to put the shares to MSSA 101 and the guarantee from MGL. 102 If there had been a back-to-back subscription by PSIC3 for preference shares in MSSA, and the use by MSSA of the money to fund its broker operations, those further steps would have been matters of which Absa had knowledge. 103 But those further steps did not take place. Instead, steps (b) to (j) occurred. SARS has assessed on the basis that Absa can be regarded as having been a party to an arrangement that included those steps, even though Absa did not know of them. In my view, that is an error of law. The resultant injustice of the legal error is that SARS has sought to tax a party, Absa, in respect of impermissible transactions of which it had no knowledge. The tax benefit issue

[133]If Absa was not a “party” to the impermissible avoidance arrangement, that would be sufficient to uphold the review against the assessment. However, I shall deal with the tax benefit issue, since it was dealt with by the High Court and has been 101 Macquarie Securities South Africa Limited. 102 Macquarie Group Limited. 103 See para 18 of the assessment letter, quoted in Five Tax Cases above n 6 at para 256. ROGERS J 54 addressed in the first judgment. I shall do so on the supposition that, contrary to what I have said thus far, Absa was a “party” to the impermissible avoidance arrangement.

[134]What “tax benefit” did the arrangement yield? The assessment letter leaves this in no doubt: it was the benefit yielded by the Brazilian interest swap. According to the assessment letter, the swap was undertaken to ensure that interest that would otherwise be taxed in the hands of PSIC4 would be exempt from interest in terms of Article 11(4)(e) of the DTA. It is a curious feature of the assessment that, according to SARS, the arrangement did not in truth yield this benefit, because the exemption in Article 11(4)(b) was disapplied by virtue of Article 11(9). SARS says, however, that practically the same outcome was reached because of expenses PSIC4 was entitled to deduct in consequence of earning the Brazilian interest. Whichever of these two tax benefits one focuses on, the tax benefit was obtained by D1 Trust and, through the conduit principle in section 25B(1) of the ITA, PSIC4.

[135]To determine whether there has been a “tax benefit”, one must contrast what actually happened with a plausible counterfactual. Inherent in the concept of a “tax benefit” is a contrast between the tax treatment of what was actually done and the tax treatment of what would plausibly have been done if the impermissible component of the arrangement had not been undertaken. In the present case, the plausible counterfactual is self-evident, flowing from the very nature of the alleged tax benefit. The impermissible feature was the insertion, into other transactions, of step (g), the Brazilian interest swap. Although it is not always appropriate simply to think away an impermissible feature, in this case it is the obvious course of action, since everything else can stand without step (g). In the absence of the Brazilian interest swap, the interest which the D1 Trust earned from MSSA, and which it distributed to PSIC4 in accordance with the conduit principle, would have been taxable in PSIC4’s hands.

[136]The parties that thus avoided a liability for tax were the D1 Trust and PSIC4. The Commissioner’s remedial powers under section 80B must, in my view, be directed at denying the tax benefit to the party that got it. And needless to say, ROGERS J 55 recharacterisation under section 80B is not a method of determining whether a tax benefit was obtained. Recharacterisation is a remedial power which the Commissioner has once the arrangement on its own terms has been found to have resulted in a tax benefit. As noted, the latter enquiry involves a comparison between what was actually done and the plausible counterfactual.

[137]Absa did not obtain a tax benefit, even if it was a party to the whole arrangement. Absa did not avoid any tax. It made a preference share investment. In terms of the ITA, the investment yielded exempt dividends. We do not know, from the assessment letter, to what extent the tax benefit obtained by the D1 Trust and PSIC4 affected the quantum of the dividends PSIC4 was willing to pay PSIC3 and which PSIC3 was thus willing to pay Absa. The assessment letter suggests that the Macquarie group extracted profit from the arrangement through the identification of Macquarie entities as residual beneficiaries of the D1 Trust.

[138]One may nevertheless assume that some portion of the tax benefit obtained by the D1 Trust and PSIC4 resulted in an economic benefit to PSIC3 and thus to Absa. However, one should not confuse tax benefits and economic advantages. It often happens that others benefit economically from a taxpayer’s avoidance of tax, and that is so whether the avoidance is permissible or impermissible. A listed company with many shareholders may be able to declare higher dividends because it has arranged its affairs in a tax-efficient way. A tax-efficient employer may be able to pay its staff higher salaries. The GAAR is concerned with tax benefits, not economic advantages.

[139]The furthest one can go, in the present case, is to say that, but for the tax benefit obtained by the D1 Trust and PSIC4, PSIC3 would not have been able to pay as high an exempt dividend to Absa as the one actually paid. The contrast at that level is between different levels of dividends that would in either case be exempt from income tax. But it goes further, because Absa’s economic benefit from the arrangement did not depend on the D1 Trust and PSIC4 getting a tax benefit and on PSIC3 being able to pay the full promised dividend. If for any reason PSIC3 was not ROGERS J 56 able to pay the full promised dividend, Absa had a grossing-up guarantee from MGL. Absa’s position was, quite simply, unrelated to the tax treatment of other elements of the arrangement. Contrary to what the first judgment states, 104 removing the Brazilian swap does not convert Absa’s return on its preference shares from exempt dividends to taxable interest.

[140]On the basis that Absa did not obtain a tax benefit from the arrangement, was SARS entitled nevertheless to impose tax on it pursuant to section 80B? It is true that under that section the Commissioner may determine the tax consequences of the arrangement “for any party” in a number of different ways. In my view, however, the purpose of the remedial power is to target the tax benefit, by ensuring that the party which got it cannot keep it. I accept that an arrangement may yield tax benefits, as distinct from economic advantage, to multiple parties. What the Commissioner may not permissibly do, in my view, is to allow the tax benefit to stand in the hands of the party that got it and instead tax another party which did not.

[141]This purpose is, in my view, inherent in a remedial power to combat impermissible tax avoidance. This conclusion is clinched by the formulation of section 80G(1). In terms of that subsection, an avoidance arrangement “is presumed to have been entered into or carried out for the sole or main purpose of obtaining a tax benefit unless and until the party obtaining a tax benefit” proves otherwise (my emphasis). This formulation necessarily conveys that the party SARS is entitled to tax is the party that got the tax benefit, and that in resultant litigation it is that party that will be the litigant. It would be absurd to impose a burden of proof on a non-party.

[142]In my view, the first judgment gives no satisfactory answer to this point. It is not permissible to interpret other provisions of the GAAR as if section 80G(1) does not exist, and then to conclude that the relevant phrase in section 80G(1) merely “addresses the evidentiary burden”. 105 The GAAR must be interpreted with regard to 104 See the first judgment at [87]. 105 See the first judgment at [80]. ROGERS J 57 all its provisions as part of the context. In that respect, section 80G(1) sheds unmistakable light on the purpose of the remedial power conferred by section 80B.

[143]The only possible response is to postulate that in section 80G(1) the lawmaker made a mistake and should have said “unless and until the taxpayer” (my emphasis). However, replacing the lawmaker’s actual and unambiguous wording with different wording is an extreme remedy which can be adopted only where the lawmaker’s true intent is clear beyond doubt and the actual wording would result in absurdity. That is not so here.

[144]My view also conforms with our tax-avoidance history, where it has always been the party obtaining the tax benefit that has been targeted by assessments. In that respect, the current GAAR and the provisions of the repealed section 103(1) of the ITA are not materially different. 106 The repealed section did not explicitly state that only the party obtaining the tax benefit could be taxed, but this was always taken for granted. Comparative perspectives

[145]My approach on the party issue accords with the way in which the English courts have applied their tax-avoidance jurisprudence to a series of transactions. A sequence of transactions can be regarded as a series, and the anti-avoidance jurisprudence applied to the series as a whole, if the individual transactions comprising it “are linked or glued together through ‘firm’ arrangements or understandings that each component will be completed”, in other words where “it is ‘well understood’ that the entire sequence will be carried to completion”. 107 This does 106 Compare Silke above n 88 in Service 58 (2016) Volume 4 at 19-81, where, with reference to section 80B, the authors observe: “The outer limits of the similar general power in the now-repealed section 103(1) – for it is noteworthy that no limits whatever were specified save that the Commissioner had to be satisfied that his determination of the tax liability of the taxpayer and of any other party to the scheme was appropriate for the prevention or diminution of the avoiding or postponing of tax liability – have never been tested in the South African courts.” 107 Krishna The Fundamentals of Canadian Income Tax 7 ed (Carswell, Toronto 2002) at 892-3, principally with reference to Inland Revenue Commissioners v Burmah Oil Co Ltd [1981] UKHL TC 54; [1982] STC 30 (HL); ROGERS J 58 not require the transactions to have been linked by an enforceable contract, but there must be an understanding or intention that each transaction will unfold as a preordained sequence. SARS does not claim that Absa was a party to an understanding or intention in respect of the transactions comprising steps (b) to (j).

[146]The present case bears some similarity with the New Zealand BNZ case. 108 The Bank of New Zealand (BNZ) had funded its wholly -owned subsidiary, BNZ Investments Limited (BNZI), which invested in preference shares in companies provided for that purpose by Capital Markets Limited (CML), an entity within a merchant banking group (the “upstream” transactions). There were various transactions “downstream” of BNZI’s preference share investments, of which BNZ was not aware. The downstream arrangements contained elements that the Commissioner of Inland Revenue regarded as impermissible tax avoidance. The Commissioner sought to invoke New Zealand’s anti-avoidance provisions against BNZ.

[147]The trial court and the New Zealand Court of Appeal (NZCA) held that the upstream and downstream transactions could not be regarded as a single arrangement. The concept of an arrangement presupposed two or more participants who arrive at an understanding. This need not be an enforceable contract. Nevertheless, an arrangement “involves a consensus, a meeting of minds between parties involving an W T Ramsay Ltd v Inland Revenue Commissioners [1982] AC 300; and the speech of Lord Brightman in Furniss (Inspector of Taxes) v Dawson [1983] UKHL 4; [1984] AC 474; [1984] 1 All ER 530 (HL). See also Macniven v Westmoreland Investments Limited [2001] UKHL 6; [2003] 1 AC 311; [2001] 1 All ER 865 (HL) at para 2 per Lord Nicholls: “Ramsay brought out three points in particular. First, when it is sought to attach a tax consequence to a transaction, the task of the courts is to ascertain the legal nature of the transaction. If that emerges from a series or combination of transactions, intended to operate as such, it is that series or combination which may be regarded. Courts are entitled to look at a pre-arranged tax avoidance scheme as a whole. It matters not whether the parties’ intention to proceed with a scheme through all its stages takes the form of a contractual obligation or is expressed only as an expectation without contractual force.” (Emphasis in original.) See also, in Canada, OSFC Holdings Ltd. v Canada (C.A.) 2001 FCA 260 (OSFC) at paras 18-24. The actual result in that case, by which a fourth transaction – not part of the preceding three preordained transactions – was treated as part of the overall scheme, was the result of an extension brought about by section 248(10) of Canada’s Income Tax Act, RSC 1985: see OFSC at paras 25-39. 108 BNZ above n 67. See also the judgment of the court of first instance reported as BNZ Investments Ltd v Commissioner of Inland Revenue [2000] 19 NZTC 15 732 (HC). ROGERS J 59 expectation on the part of each that the other will act in a particular way” and “consensus as to what is to be done”. Richardson P, in delivering the majority judgment in the NZCA, said: “51. The justification for construing the concept of arrangement in that way is that it would be inequitable for a taxpayer who enters into an apparently unobjectionable transaction to be deprived of its rights thereunder merely because, unknown to the taxpayer, the other party intended to meet its obligations under that transaction, or in fact did so, in a legally objectionable way. . . . 52. In order to avail the Commissioner, the consensus – the meeting of minds – necessary to constitute an arrangement under section 99 must encompass explicitly or implicitly the dimension which actually amounts to tax avoidance; albeit the taxpayer does not have to know that such dimension amounts to tax avoidance.”

[148]As later cases show, this should not necessarily have been the end of the matter, because section 99(3) of New Zealand’s Income Tax Act, 109 unlike our GAAR, permits the Commissioner to adjust the assessable income “of any person affected by” the arrangement. In other words, the assessed person need not have been a “party” to the arrangement. Even if the upstream and downstream transactions were not part of a single arrangement, BNZ was arguably a person “affected” by the downstream arrangements.

[149]That would still have left the question whether BNZ, as a non-participant but arguably a “person affected”, obtained a “tax advantage” as contemplated in section 99(3). In that regard, Blanchard J in a concurring judgment said: “The adjustment can be made against both a party to the arrangement and a person affected, who is not necessarily a party. But it can be made only where a tax advantage has been obtained ‘under that arrangement’. The Commissioner therefore cannot make an adjustment as against someone who is not a party merely because 109 65 of 1976. ROGERS J 60 that person has received a payment subsequent to the operation of an arrangement but outside the arrangement.” 110

[150]In BNZ the Commissioner had advanced his case on the basis that BNZ’s subscription in preference shares was part of the relevant arrangement. By contrast, in Peterson, 111 which concerned a film scheme, the Commissioner’s case did not depend on Mr Peterson’s investment in the film partnership having been part of the relevant arrangement; he was taxed not as a participant in the arrangement but as a person affected by it. 112 The crucial question in Peterson, according to the NZCA, was thus whether “the Commissioner was entitled to adjust the taxable income of the taxpayer who was not a party to the arrangement and had no knowledge of it”. 113 This point was decided in favour of the Commissioner. In response to Mr Peterson’s reliance on the requirement of a “meeting of the minds”, the NZCA said that this statement in BNZ was directed at the issue of what constituted an “arrangement”, and “not for the implications for a ‘person affected’ who is not a party to the arrangement once an arrangement exists”. 114

[151]Mr Peterson’s appeal to the Privy Council succeeded by a majority, but on other grounds. 115 Lord Millett, with Baroness Hale and Lord Brown concurring, said that section 99 required the Commissioner to prove (a) that the identified arrangement had the purpose or effect of avoiding tax; (b) that “whether or not the taxpayer was a party to the ‘arrangement’, he was affected by it”; and (c) that “he obtained a tax advantage from it”. 116 If the Commissioner could satisfy these conditions, he could adjust the assessable income of any person affected by the arrangement “in order to 110 BNZ above n 67 at para 175. 111 Commissioner of Inland Revenue v Peterson [2003] NZCA 27; [2003] 2 NZLR 77. 112 Id at paras 27, 29 and 30. 113 Id at para 38. 114 Id at para 40. 115 Peterson v Commissioner of Inland Revenue [2005] NZPC 1; [2005] UKPC 5; [2006] 3 NZLR 433 (PC). 116 Id at para 33. ROGERS J 61 deny him the tax advantage which he has derived from it”. 117 The taxpayer need not have been a party to the ‘arrangement’ and need not have been privy to its details, and to that extent they agreed with the dissenting judgment of Thomas J in BNZ. 118

[152]In their dissenting judgment, Lords Bingham and Scott noted that an “arrangement” could be something “as loose and informal as a ‘plan’ or an ‘understanding’” and that the anti-tax avoidance net caught “not only parties to the ‘arrangement’ but also any person affected by the arrangement whether or not a party to it”. 119 They agreed with Richardson P in BNZ that an arrangement presupposed participants reaching a consensus, but emphasised that the tax advantage vulnerable to nullification under section 99 could be “a tax advantage enjoyed by someone who is not part of that consensus, not ‘. . . a party thereto’”. 120

[153]Lord Millett contrasted the Peterson case with the BNZ case, saying that in Peterson “the investors did not merely obtain an economic advantage from the ‘arrangement’ (as in that case [BNZ]); they obtained a tax advantage”. 121 In other words, BNZ – the equivalent of our Absa – got an economic advantage but not a tax advantage, so the Privy Council considered.

[154]These decisions thus appear to me to support the propositions, first, that Absa could not have been a party to the “downstream” arrangements without knowing of them and intending that they should happen; and, second, that while Absa arguably obtained an economic advantage from the “downstream” arrangements, it did not get a tax benefit. In terms of the New Zealand provision, in contrast to ours, the first point would not be dispositive, since the taxpayer need not have been a “party” to the impermissible arrangement as long as it was “affected” by it. The two regimes, 117 Id. 118 Id at para 34. The same point was made by the New Zealand Supreme Court in Ben Nevis Forestry v CIR [2008] NZSC 15; [2009] 2 NZLR 289 at paras 165-8. 119 Peterson id at para 59. 120 Id. 121 Id at para 34. ROGERS J 62 however, have the second point in common: the enhanced preference share dividend, as a result of impermissible tax avoidance that occurred downstream, is an economic advantage but not a tax benefit.

[155]In view of differences in formulation, there may be limited value in contrasting our GAAR with anti-avoidance provisions in other jurisdictions. I nevertheless note that, on my reading of the GAAR provisions in Australia, New Zealand, Canada and the United Kingdom, the remedial powers in the case of tax-avoidance are confined to cancelling the tax benefit in the hands of the person that got it. 122 The first judgment holds that a GAAR assessment may be issued against a person who (a) was unaware of the impermissible tax avoidance and (b) obtained an economic advantage but not a tax benefit from it. As far as I can ascertain, that is unprecedented internationally. 122 See the following provisions: (a) Australia: Section 177F(1) of the Income Tax Assessment Act 1936 refers throughout to nullifying the tax benefit obtained by the relevant taxpayer. It was on this basis that the High Court of Australia, in Federal Commissioner of Taxation v Peabody [1994] HCA 43; (1994) 123 ALR 451; (1994) 181 CLR 359 at paras 33-5, held that Mrs Peabody was not liable to be assessed under the Australian provision – the tax benefit had been obtained by a company, Loftway, and not by her. (b) New Zealand: Section BG(2) read with section GA(2) of the Income Tax Act 2007 empowers the Commissioner to take appropriate action so as “to counteract a tax advantage obtained by the person from or under the arrangement”. (c) Canada: Section 245(2) of the Income Tax Act RSC 1985 states that where a transaction is an avoidance transaction— “the tax consequences to a person shall be determined as is reasonable in the circumstances in order to deny a tax benefit that, but for this section, would result, directly or indirectly, from that transaction or from a series of transactions that includes that transaction.” (d) United Kingdom: Section 209(1) of the Finance Act 2013 states that the tax advantages from an abusive arrangement “are to be counteracted” by the making of adjustments. Section 209AA(1) requires Revenue and Customs to give notice to a person where the tax officer considers “that a tax advantage might have arisen to the person from tax arrangements that are abusive” (my emphasis). These provisions can be contrasted with section 210, which provides that if such counteracting action has been taken in respect of a person that received the tax advantage, “consequential adjustments” may be made in respect of any other person, “whether or not a party to the tax arrangement” (section 210(4)(b)), provided that such consequential adjustments shall not have the effect of increasing the latter person’s liability for tax (section 210(5)). ROGERS J 63 Conclusion

[156]The conclusions I have reached do not impair SARS’ legitimate powers to combat impermissible tax-avoidance. On the contrary, my conclusions subject SARS’ far-reaching taxing powers to appropriate limits conforming with the rule of law. 123 Assuming for the moment that the Brazilian interest swap caused the arrangement as a whole to fall foul of section 80A, SARS was able to tax the D1 Trust and PSIC4, the parties that obtained the tax benefit. Given the definition of “arrangement” in section 80L, nothing stopped SARS from identifying, for example, steps (b) to (j) – or some further subset of those steps – as the relevant “arrangement” for purposes of the GAAR. 124

[157]This is as it should be, since this would impose the tax liability on entities within the Macquarie group. If a GAAR assessment had been issued against PSIC4 (by treating the interest distributed to it by the D1 Trust as taxable), this would have rightly reduced PSIC4’s after-tax income. Such a GAAR assessment would not have had any economic effect on Absa, because Absa was contractually entitled to the full dividend, either from PSIC3 or, through a guarantee, from MGL. 123 That a GAAR regime implicates the rule of law was recognised by SARS in the Discussion Paper above n 27 at 45: “The third and perhaps most basic issue from both a practical and conceptual standpoint concerns the ‘uneasy tension’ between a GAAR and the basic notion of the rule of law.” One aspect is that a GAAR permits the tax gatherer to depart from the legal consequences enacted by Parliament and tax legislation. Another aspect concerns certainty and predictability. Much has been written on this subject. See among others, Tooma Legislating against Tax Avoidance (International Bureau of Fiscal Documentation, Amsterdam 2008) at 34-5; Lindsay “Tax Avoidance and Two Aspects of the Rule of Law” in De Cogan et al (eds) Tax, Public Finance and the Rule of Law (Hart Publishing, Oxford 2025); and Prebble and Prebble “Does the Use of General Anti- Avoidance Rules to Combat Tax Avoidance Breach Principles of the Rule of Law? A Comparative Study” (2010) 55 Saint Louis University Law Journal 22. More recent writers on the subject often cite Cooper “Conflicts, Challenges and Choices – The Rule of Law and Anti-Avoidance Rules” in Cooper (ed) Tax Avoidance and the Rule of Law (International Bureau of Fiscal Documentation, Amsterdam 1997). In the paper by Prebble and Prebble, the authors write: “General anti-avoidance rules demonstrate that the rule of law is not an unqualified good. As with all principles, the rule of law can be outweighed by competing considerations. General anti-avoidance rules give an example of what those competing considerations might be.” 124 An “arrangement” is defined as “any transaction, operation, scheme, agreement or understanding (whether enforceable or not), including all steps therein or parts thereof, and includes any of the foregoing involving the alienation of property” (emphasis added). One of the intended results of the new GAAR was to clarify that the anti-avoidance provisions “may be applied to steps within a larger scheme (and that a general business purpose for a larger scheme is not sufficient to shield each and every step in that scheme from review)”. See Discussion Paper above n 27 at para 10 p 48 and para 10.4 p 56; Interim Response above n 88 at para 6 p 40; and Revised Proposals above n 43 at para 10.2 p 21. ROGERS J 64

[158]SARS’ counsel in oral argument suggested that the D1 Trust and PSIC4 were mere conduits and empty shells, so that taxing them would be to no avail. The assessment letter doesn’t say this, but if it be so, it would not be the end of the matter. In terms of section 80B(1)(c), the Commissioner’s remedial powers include the power to deem persons who are “connected persons” in relation to each other “to be one and the same person”. In this way, SARS could in all likelihood have reached other entities of substance within the Macquarie group, such as MSSA and MGL.

[159]I would thus uphold the appeal, set aside the Supreme Court of Appeal’s order and restore the High Court’s order. This should carry costs here and in the Supreme Court of Appeal, including the costs of two counsel. For the Applicants: For the Respondent: M Janisch SC, K Hofmeyr SC, L Mnqandi and C Kruyer instructed by Bowman Gilfillan Incorporated. A R Sholto-Douglas SC and M B E Mbikiwa instructed by the Office of the State Attorney, Cape Town.