Ring-Fencing of Assessed Losses in South Africa (Section 20A)

If you earn a high salary and run a side trade that loses money, section 20A of the Income Tax Act can "ring-fence" that loss so it does not reduce the tax on your salary. Ring-fencing bites only when your taxable income, before the loss, reaches the level where the top marginal rate applies (R1,817,001 for the 2026 year of assessment), and only where the trade is a listed "suspect trade" or has made a loss in at least three of the last five years. A ring-fenced loss is not lost: it is carried forward and set off against future income from that same trade.

For most people this never applies. The gateway is high, and below it a genuine trading loss can be set off against other income in the normal way. This article explains who is caught, and shows two taxpayers side by side.

The income gateway comes first

Section 20A only starts to operate once your taxable income (worked out before setting off the loss in question) is at or above the point where the maximum marginal rate becomes payable. For the 2026 year of assessment (1 March 2025 to 28 February 2026) that entry point is R1,817,001, the start of the 45% bracket.

If your taxable income is below that figure, section 20A does not apply at all. Your assessed loss from the side trade is set off against your other income under the ordinary rules, and there is nothing further to consider.

A change is on the way for later years. For years of assessment commencing on or after 1 March 2026 (the 2027 year of assessment), SARS states that section 20A "has been amended to apply from a marginal tax rate of 39% as opposed to the maximum marginal rate (45%) of tax." So from the 2027 year the gateway drops to the 39% bracket, pulling in more taxpayers. That lower gateway does not apply to the 2026 year, only from 1 March 2026.

The two triggers: suspect trade or three-of-five

Clearing the income gateway is not enough on its own. Once you are over it, ring-fencing applies only if one of two further tests is met.

The first is the "suspect trade" list in section 20A(2)(b). These are trades Parliament singled out as often being lifestyle or hobby activities. They include, among others, sporting activities, dealing in collectibles, the letting of residential accommodation, vehicles, aircraft or boats (subject to exceptions), animal showing, farming or animal breeding otherwise than on a full-time basis, gambling or betting, and the performing or creative arts. If your loss-making trade is on that list, ring-fencing can apply regardless of how many years it has run.

The second trigger is the "three-out-of-five" test: the trade has made an assessed loss in at least three of the last five years of assessment. A trade that is not on the suspect list can still be caught this way if the losses keep repeating.

The escape: a real business with a reasonable prospect

Section 20A(3) provides an escape. Ring-fencing does not apply where the trade is carried on as a business and there is a reasonable prospect of it deriving taxable income within a reasonable period. SARS weighs listed factors here, such as the commercial manner in which the trade is conducted, the number of employees, the hours spent, and whether the losses look like start-up losses of a genuine venture rather than a subsidised hobby. For certain suspect trades a longer six-out-of-ten-year cool-down applies before this reasonable-prospect escape can be relied on again.

Worked example: two taxpayers, the same R60,000 loss

Both people below run a weekend creative-arts side trade that made an assessed loss of R60,000 in the 2026 year, and both have shown a loss in four of the last five years. The performing or creative arts is a suspect trade, and four of five years also meets the three-out-of-five test. The only thing that differs is their other income.

Person A, taxable income R2,100,000 before the loss. This is above the R1,817,001 gateway, so section 20A is live. The trade is a suspect trade and the three-of-five test is met, so the R60,000 loss is ring-fenced. It cannot reduce the R2,100,000. Tax on R2,100,000 for the 2026 year is:

• R644,489 + 45% of (R2,100,000 − R1,817,000)
• = R644,489 + 45% of R283,000
• = R644,489 + R127,350 = R771,839
• less the primary rebate R17,235 = R754,604

Had the loss been allowed against the salary, taxable income would fall to R2,040,000 and the tax would be R644,489 + 45% of R223,000 = R744,839, less R17,235 = R727,604. The difference is R27,000 (45% of R60,000). Person A does not get that R27,000 now. The R60,000 loss is carried forward and can only be set off against future income of the same creative-arts trade.

Person B, taxable income R450,000 before the loss. This is below the gateway, so section 20A does not apply. The R60,000 loss is set off against other income, bringing taxable income to R390,000. Tax on R390,000 is:

• R77,362 + 31% of (R390,000 − R370,500)
• = R77,362 + 31% of R19,500
• = R77,362 + R6,045 = R83,407
• less the primary rebate R17,235 = R66,172

Without the offset, tax on R450,000 would be R77,362 + 31% of R79,500 = R102,007, less R17,235 = R84,772. So the loss saves Person B R18,600 (31% of R60,000) in this year.

The trade and the loss are identical for both. The outcome differs only because Person A sits above the top-bracket gateway and Person B does not.

How it shows up in practice

Ring-fencing is applied on assessment, on your ITR12, not as a separate election. You return the trade's income and expenses as normal, and where section 20A applies SARS confines the resulting loss to that trade. Because the losses that trigger the three-of-five test build over years, someone approaching the gateway with a repeatedly loss-making side trade should keep records that speak to the section 20A(3) escape, showing the trade is run commercially with a real prospect of profit.

If your side income runs at a profit rather than a loss, section 20A is not in play at all. See the guide to tax on freelance and side income in South Africa for how that profit is taxed, and estimate your position with the income tax calculator.