‘Expenses don’t earn interest’: Self-insurance and the section 11(a) deduction after Meiring Citrus 

By Francois Grobler & Dr Hendri Herbst

Introduction

A citrus grower paid R10 million under an agreement described as an insurance policy, for an indemnity limit of R12 million. Only R400,000 of that buys a transfer of risk, and then in respect of only R2.4 million of the cover. The balance, R9.6 million, was credited to an account that remained, in substance, the grower’s own funds: it earned interest for its benefit, could be pledged as security, and was repayable on 30 days’ notice. The grower claimed the full R10 million as a deductible premium. Four years later, no claim has been made, it recovers R11,304,932.01.

That arrangement, and others cast in similar terms, is what the Tax Court and, on appeal, a full bench of the Western Cape High Court were required to scrutinise. Whether the disparity between the label and the substance carries consequences for the deductibility of the premium is the question common to the three judgments we will consider below, namely:

• Taxpayer Boerdery v Commissioner for the South African Revenue Service (‘Boerdery‘) (IT 45979) [2024] ZATC 5; 87 SATC 447, Tax Court, Johannesburg, 20 March 2024;
• Pear (Pty) Ltd v Commissioner for the South African Revenue Service (‘Pear‘) (IT 46080) [2024] ZATC 19, Tax Court, Cape Town, 5 December 2024; and
• Commissioner for the South African Revenue Service v Meiring Citrus (Pty) Ltd (‘Meiring‘) (A161/2025; IT 46080), Western Cape High Court, 26 June 2026.

Pear and Meiring are the same dispute at different levels, Pear being the Tax Court judgment of Janisch AJ, and Meiring the successful appeal of the Commissioner for the South African Revenue Service (‘SARS’) against it. Boerdery is a separate matter which informed the discussion in both these judgments.

The self-insurance products were marketed to farmers, but the reasoning is of general application. Any enterprise that sets funds aside to carry its own risk, whether through an experience-account policy, a cell captive or a captive insurer, should take note, as should those who design and sell these structures.

The nature and aims of self-insurance

Self-insurance is the retention and internal funding of risk. Rather than transferring a risk to an insurer in exchange for a premium, the enterprise decides to carry the risk itself and to meet any loss from its own resources, whether from general reserves or from a fund set aside for the purpose. Its usual commercial rationale is to avoid the cost loading in a third-party premium, to retain the underwriting margin an insurer would otherwise earn, and to finance predictable, high-frequency and low-severity losses more efficiently over time.

The distinction matters because, as the court in Meiring observed, self-insurance is ‘not insurance but an antithesis of insurance’, the essence of insurance, being the transfer and spreading of risk, being absent. The court identified five distinguishing elements of a contract of insurance, namely (i) an insurable interest; (ii) a risk of loss through designated perils; (iii) the insurer’s assumption of that risk; (iv) that assumption forming part of a general scheme to distribute losses among a large group bearing similar risks; and (v) a rateable contribution to a general insurance fund, called a premium. Where the insured party’s own funds meet the insured party’s own loss, none of those features is present.

Two of the self-insurance products added a tax objective to the ordinary self-insurance rationale. On the taxpayers’ own evidence, the arrangements were entered into while the taxpayer was ‘looking for expenses’ to reduce taxable income, the product having been marketed at insurer roadshows to accountants with farming clients. The design was deliberate in a further respect: the policy ran for six months from year-end so as to fall within paragraph (aa) of the proviso to section 23H(1) of the Income Tax Act, No. 58 of 1962 (‘the ITA’), which disapplies apportionment where the services are rendered within six months after year-end.

The common structure of the products

Although the insurers and figures differ, the products in question in the various cases share a common structure. The taxpayer paid a large ‘premium’, the bulk of which was credited to an ‘experience account’ held by the insurer for the taxpayer’s benefit. A small charge, described as an underwriting charge or ‘insurer’s margin’, was retained for the limited risk the insurer genuinely assumed, and the account earned ‘notional interest’. Claims were met first from the account balance, and on expiry or cancellation the balance and interest were refundable.

The proportions bear consideration. In Meiring, of the R10 million paid, only R400,000 was a true premium, buying R2.4 million of transferred risk within a R12 million indemnity limit, with the remaining R9.6 million being self-insured. In Boerdery, an annual ‘premium’ of R35,000,037 bought cover of R41.5 million for 2018, roughly 85% of cover, alongside a 2.25% margin and an investment management fee of 65 basis points.

The Boerdery judgment

Boerdery was the first of these matters. The Tax Court accepted that the premium was ‘expenditure’ ‘actually incurred’, reasoning that cash had been exchanged for contractual rights and that, on the Supreme Court of Appeal’s approach in CSARS v Labat Africa 2013 (2) SA 33 (SCA) (‘Labat’), a movement of assets suffices, even if the diminution is only temporary. It then disallowed the deduction on the ground that the expenditure was ‘of a capital nature’, the premium having secured a right to an interest-bearing, refundable fund, meaning that the expenditure related to the acquisition of an income-producing asset (capital in nature) as opposed to working such an asset (revenue in nature). The court also upheld a 10% understatement penalty and section 89quat(2) interest.

The tension in the reasoning is apparent. The findings pointed to a deposit rather than a premium: the ‘notional interest’ was real interest, the balance was the taxpayer’s asset repayable on notice, and a full refund at the end of a claim-free year is uncommercial for a true premium. That sits awkwardly with a conclusion that the money was expended at all. The court did not decide whether the contract was insurance in the first place, the question the High Court later placed at the centre of the analysis.

The Tax Court judgment in Pear

The Tax Court in Pear found for the taxpayer, but importantly, the finding turned on prescription. Section 99(1)(a) of the Tax Administration Act, No. 28 of 2011 (‘the TAA’) precludes SARS from making an assessment three years after the date of an original assessment by SARS, unless the failure to assess the full amount of tax was due to ‘fraud’, ‘misrepresentation’ or ‘non-disclosure of material facts’ under section 99(2)(a). The court held that claiming a deduction expressed the taxpayer’s legal opinion rather than a fact, so there was no ‘misrepresentation’; that the omission of R1,197.52 of notional interest was too small to be a non-disclosure of a material fact; and that SARS, which bears the onus, had led no evidence linking any non-disclosure to the under-assessment.

On the merits, decided in the alternative, the court followed Boerdery in holding the premium ‘actually incurred’ and distinguished a bank deposit, but parted from Boerdery on capital, holding that the payment protected the income stream and was not ‘of a capital nature’.

Counsel for SARS confirmed that it was not SARS’s case that the contract was a sham or simulation, a concession that shaped the whole approach. And had prescription not applied, the taxpayer would have failed under section 23L(2), for want of expert accounting evidence. The taxpayer prevailed on time-bar, not on entitlement.

The High Court judgment in Meiring

On appeal, the full bench upheld SARS’s appeal on every issue, and in doing so reconciled the two Tax Court lines, adopting the result in Boerdery while reconstituting its reasoning.

Not insurance, but self-insurance

Approaching the contract as a unitary exercise of text, context and purpose (University of Johannesburg v Auckland Park Theological Seminary 2021 (6) SA 1 (CC); Natal Joint Municipal Pension Fund v Endumeni Municipality 2012 (4) SA 593 (SCA); Centriq Insurance Company Ltd v Oosthuizen2019 (3) SA 387 (SCA)) and applying the simulation test in CSARS v NWK Ltd 2011 (2) SA 67 (SCA), the court held that the arrangement was an investment mischaracterised as insurance, and that to the extent of the R9.6 million it was not a contract of insurance at all. It relied on features inconsistent with insurance: the insured, not the insurer, set the premium; claims within the account balance were paid within 72 hours without investigation because the insurer bore no risk on those funds; the balance was refundable on cancellation; no risk analysis was carried out; and the policy could be pledged as security, confirming that the funds were an asset in the taxpayer’s hands. Only the R400,000 charge was a true premium.

In this regard, two observations are notable. First, the court held expressly that the contract was not a sham, and that the parties intended the contract to operate according to its tenor. What the court condemned was the label given to the contract – yet the court still applied the ‘commercial sense’ test in NWK, which the Supreme Court of Appeal has since been steered away from in Roshcon 2014 (4) SA 319 (SCA) and Bosch 2015 (2) SA 174 (SCA), and which sits uneasily beside a finding that the agreement was genuine. Secondly, not even the fact that the counterparty was a licensed insurer rescued the contract, on the view that regulatory status determines who may write insurance, not whether a given contract is insurance.

Expenditure actually incurred

The court held that the R9.6 million was not ‘expenditure’ ‘actually incurred’. A deposit recoverable on demand is a neutral factor, as the advance and repayment of a loan are neutral and Labat requires a parting with value rather than a mere change in the form in which an asset of unchanged value is held. That claims might have reduced the balance did not make the right of recovery contingent: a depositor who authorises its bank to meet its own liabilities does not thereby convert the deposit into expenditure. On this point the court declined to follow Boerdery, while agreeing with its result. The only amount genuinely expended was the R400,000 underwriting fee, which SARS had allowed at objection.

Of a capital nature

In the alternative, the court held that if the amount had been incurred it would have been ‘of a capital nature’. Applying BP Southern Africa v CSARS [2007] ZASCA 7; 69 SATC 79 and New State Areas v CIR 1946 AD 610, the enquiry was not what risks the policy purported to cover but what the taxpayer acquired for its money, namely a credit recoverable on termination, the right to interest on it, and a right capable of being pledged. That is an income-producing asset, the fruit-and-tree distinction in CIR v Visser 1937 TPD 77 applying so that the funds were the tree and the interest the fruit, in agreement with Boerdery.

That distinction, between a ground of assessment, which SARS may not novate, and the taxpayer’s onus, which it must discharge whether or not SARS argues the point, is notable. It permits an appellate court to decide an element that SARS never developed, on the footing that the taxpayer always bore it. The practical consequence is that a taxpayer cannot safely prepare only for the elements SARS has chosen to run.

SARS’s alternative grounds: sections 23L and 23(e)

Two further grounds, on which SARS relied, were decided only at first instance. Section 23L(2) denies a deduction for a premium incurred in terms of a policy ‘to the extent that the premium is not taken into account as an expense for the purposes of financial reporting pursuant to IFRS in either the current year of assessment or a future year of assessment’. In Pear this would have defeated the deduction, the court treating the correct IFRS treatment as requiring expert evidence which neither party led, so that the taxpayer failed to discharge its onus of proof. Section 23(e), which disallows the deduction of ‘income carried to any reserve fund or capitalized in any way’, was also raised, though the court’s prima facie view was that a premium did not fit that description. Both were abandoned by SARS on appeal, so they remain undecided, and section 23L(2) in particular is an under-appreciated line of attack.

Two qualifications should be noted. Section 23L is not a universal answer for SARS, as ‘policy’ is defined in section 23L(1) as a policy of insurance or reinsurance other than a long-term policy, such that if, as in Meiring, the contract is not insurance at all, the section cannot apply. Furthermore, section 23L(3) supplies a measure of symmetry where a deduction is denied under section 23L(2), later benefits being included in income only to the extent that they exceed the premiums disallowed. There is no equivalent relief where the deduction fails under section 11(a).

Prescription

The High Court reversed the Tax Court’s finding on prescription. The taxpayer’s view that the premium was deductible was an opinion, but it rested on facts about the arrangement that had never been placed before SARS: what had been produced in answer to the verification request was an application form and a debit-order authority, not the policy. Withholding those facts was a ‘misrepresentation’ under section 99(2)(a)(ii). The omission of the notional interest was, separately, a ‘non-disclosure of material facts’, as materiality attaches to a fact and its relevance to taxability, not to the quantum, and ‘expenses do not earn interest’, so disclosure of even R1,197.52 would have exposed the character of the payment. For establishing causation, the nexus could be viewed on the uncontested documentary record without oral SARS evidence, CSARS v Spur Group [2021] ZASCA 145 being distinguished on that point and Technology Corp Management v De Sousa 2024 (5) SA 57 (SCA) applied, aided by adverse credibility findings against the taxpayer’s accountant.

The Tax Court had reasoned that causation must be established item by item. The High Court rejected that, holding that the two grounds were ‘inextricably linked’, and once an assessment is validly reopened SARS may correct every component of it. A single material non-disclosure can therefore unlock an entire prescribed year.

The causation holding is the most contestable part of the judgment. On the same record the Tax Court had reached the opposite view, and Technology Corp Management was a company-law matter under section 163 of the Companies Act, 2008, in which the party who declined to testify did not bear the onus. Substituting an inference from documents for the direct evidence the Tax Court thought necessary is defensible, but open to argument, and would likely be the principal issue on any further appeal.

Understatement penalty

The court reinstated the 10% understatement penalty the Tax Court had set aside. Section 222(1) of the TAA imposes a penalty for an ‘understatement’ unless it results from a bona fide inadvertent error, calculated as the shortfall multiplied by the percentage in the table in section 223(1). The lowest behaviour category, a ‘substantial understatement’, attracts 10% in the standard case and, crucially, depends on no culpable conduct at all, being defined in section 221 as prejudice exceeding the greater of 5% of the tax properly chargeable or R1 million. That threshold was plainly met. Where an arrangement of this size fails, the penalty follows on the numbers alone.

How SARS approaches these arrangements

A consistent pattern emerges from the three judgments. SARS assesses at face value first and verifies afterwards; and a verification closed without adjustment is not an assurance. Once SARS is satisfied that an assessment prejudices the fiscus, section 92 of the TAA obliges it to correct that prejudice. SARS is prepared to reopen a prescribed year and, once one item is in issue, to revisit the assessment as a whole. It audits several years at once and seemingly investigates an industry at a time rather than a single taxpayer, the same product having generated queries to farmers across a district. SARS apparently also expects the substantive documents, the policy wording, and the account statements, not only an application form, when it questions an unusual deduction.

Defending a self-insurance structure: what must now be shown

Meiring does not outlaw self-insurance, captives or cells. What it requires is that the taxpayer discharge its own onus on each element of section 11(a), regardless of the label applied to the expenditure. These are arguably the points on which the tax treatment of such an arrangement will now stand or fall:

• Whose patrimony absorbs the loss? If a claim is met from a fund the payer may recover, the insurer’s patrimony is untouched and no risk has passed. Cover attaching only above a self-funded layer transfers only the excess, and only the premium referable to that excess is a premium at all.
• Who priced the risk, and on what? A clause permitting the insured to set its own premium is fatal. The file should show an underwriting assessment: loss history, exposure data, a risk analysis, and a rate fixed by the insurer on that assessment. The absence of any risk analysis, and non-disclosure of prior losses, counted heavily against the taxpayer.
• Is the premium rationally related to expected loss, or to the limit of indemnity? A premium approaching the sum insured signals a deposit, not a transfer of risk. The rate must be actuarially supported.
• Is there a right of recovery? Profit commissions and no-claims bonuses are not objectionable in themselves, but must be genuinely contingent on the underwriting result and determined after the period of cover. A balance recoverable at will on short notice, carrying interest, is a deposit.
• Are claims adjusted? A term paying claims within the self-funded layer inside 72 hours without investigation, while claims above it may be assessed, is an admission on the face of the document that the insurer bears no risk below the line.
• Can the arrangement be dealt with as an asset? A right to pledge, cede or encumber the ‘policy’ shows that what the payer holds is an asset in its own hands, which is fatal on both the incurral and the capital enquiries.
• Is the risk pooled or distributed? This is the hard element for captives and first-party cells. Under the Insurance Act, No. 18 of 2017, a cell captive insurer may write first-party or third-party risks, and section 25 prohibits combining them in one cell structure. Structures relying on brother-sister risk, genuine third-party business or real reinsurance retention are materially better placed.
• Is the purpose documented, and commercial? Evidence that the taxpayer was ‘looking for expenses’ is corrosive. Board minutes, a risk register, and a policy period driven by the risk rather than by the six-month proviso to section 23H(1) are the contemporaneous record a court will look for.
• What does IFRS say? A policyholder that recognises the recoverable balance as a financial asset rather than an expense, no premium is taken into account in profit or loss for IFRS purposes, and section 23L(2) defeats the deduction on that ground alone. Settle the accounting before the return is filed.
• Has SARS been told? Producing the policy wording and account statements, rather than an application form, preserves the three-year bar in section 99(1)(a) and removes the grounds in section 99(2)(a) on which SARS succeeded.

Two consequences deserve planning attention. The first is symmetry. If, as Meiring holds, the deposit was never expenditure, then on the same reasoning its later refund is not gross income but a return of capital, and only the interest is income. Meiring Citrus nonetheless returned the whole R11,304,932.01 as income in 2022, escaping tax only because it was in an assessed-loss position. A taxpayer denied the deduction should consider promptly whether the later receipt is properly excluded, and whether relief can still be obtained for that year before it too prescribes, and section 23L(3) will not assist.

The second is that interest credited to an experience account accrues as it is credited, whether or not it is drawn. It should be declared annually, and the accounting and tax treatment of the balance kept consistent from the outset.

Key takeaways

The judgments may have involved farmers, but the principles reach any enterprise that funds its own risk. Three fundamental takeaways for all taxpayers would be:

• The label does not determine the tax treatment, nor does the counterparty’s license as a licensed insurer. A payment described as a ‘premium’ that funds an amount recoverable by the payer is exposed on three fronts – whether it was expended at all, whether it was of a capital nature, and whether it was laid out in the production of income.
• Recoverability is the decisive fact. Where the amount remains the payer’s, is refundable, earns a return or may be pledged, it was in substance never spent, and if it was, what it bought was in any event a capital asset.
• A material non-disclosure forfeits the three-year bar and reopens the whole assessment, and a substantial understatement then attracts 10% penalties without any finding of fault.