VAT zero-rating of basic foods in South Africa protects the poor in proportion but targets them badly

Introduction

South Africa gives up a lot of VAT so that brown bread, maize meal and eggs can be sold at a zero rate. The Treasury’s tax expenditure statement puts the cost of the relief on what it calls 22 basic items at R36.5 billion for 2023/24 (National Treasury, 2026a). The usual defence is short. Basically, poor households spend a bigger slice of their budgets on staples, so relief on staples must be pro-poor. It sounds right, and in proportion terms it mostly is, but a tax break can suit the poor in proportion and still send most of its rands to everybody else (Kwebulane and Oyekale, 2024). That gap between the two readings is what this whole debate turns on, and it rarely gets spelled out in public.

This paper asks whether zero-rating basic foods is a sound way to help poor South Africans in 2026. It argues that the relief protects poor households in proportion to their budgets but targets them badly, and that the best case for keeping it rests on what the alternatives can realistically do, not on its own distributional record (Bachas et al., 2024; Goldman et al., 2021). The sections below set out the legal position and the cost (Republic of South Africa, 1991), look at who actually gains, and then ask what informality and price pass-through do to that picture. After that comes the question of alternatives, and then a position.

The legal position and what the relief costs

Section 11(1)(j) and Part B of Schedule 2 to the Value-Added Tax Act 89 of 1991 zero-rate a fixed list of basic foodstuffs, among them brown bread, maize meal, rice, dried beans, lentils, eggs, milk, vegetable cooking oil and fresh fruit and vegetables (Republic of South Africa, 1991). Most other food is taxed at the standard rate, which has been 15 per cent since April 2018, when it moved up from 14 per cent (Kwebulane and Oyekale, 2025). The list has hardly moved, actually. Treasury notes that two food items and sanitary pads were added from 1 April 2019, and its cost estimate still leans on household spending data from the 2010/11 Income and Expenditure Survey (National Treasury, 2026a). A R36.5 billion measure is being costed with survey data that is some fifteen years old, which is not a comfortable place to argue from.

For scale, VAT brought in R457.8 billion in 2024/25, and Treasury now expects R497.6 billion for 2025/26, helped by resilient household consumption (National Treasury, 2026c). Set against those totals the relief on basic foods comes to roughly eight per cent of a year’s collections, a loose comparison because the years do not line up (National Treasury, 2026a; National Treasury, 2026c). That is not trivial for a fiscus under pressure, and it may help explain why VAT keeps turning up whenever the Treasury needs revenue. It also means any change to the list is a revenue question and a household question at the same time, and the two pull in opposite directions.

The 2025 Budget tried to touch both the rate and the list. It proposed VAT increases of half a percentage point in 2025/26 and again in 2026/27, together with expanded zero-rating, and all of it was withdrawn after political and legal contestation (National Treasury, 2025). Part of the lost revenue was replaced through an inflationary rise in the general fuel levy from June 2025 (National Treasury, 2025). SARS now gives the standard rate as 15 per cent and records that the planned increases to 15.5 and 16 per cent were reversed (SARS, 2026). Pairing a higher rate with a longer zero-rated list is the usual patch for the regressivity worry, which suggests Treasury still sees the relief as the default way to soften a rate rise. The 2026 Budget proposals change some zero-rating rules, for gold sold to banks and for services in customs controlled areas, but leave both the food list and the rate alone (National Treasury, 2026b).

Who gains from the relief

The best South African evidence on the question is thinner than the debate suggests. Kwebulane and Oyekale (2024) rank households by per capita expenditure in the 2018 National Income Dynamics Study, split them into deciles, and compute an equity-gain ratio for each zero-rated item by comparing what the poorest four deciles spend, proportionally, with what the richest two spend. Their reading is that most listed items are eaten mainly by lower-income households. Maize meal and rice give the poor a clearly bigger relative gain, while fresh fruit and other vegetables give them a smaller one. That fits what anyone would expect from the basket. It also means the list is a mixed bag, and part of it works more for the middle of the distribution than for the bottom (Kwebulane and Oyekale, 2024).

It helps to be clear about what targeting would mean here. A well-targeted measure sends most of its benefit to the households it is meant for and lets little leak to others. Zero-rating cannot do that by design, because the till does not know who is standing at it, so some leakage is built in. How much leaks depends on how different the baskets of rich and poor households are, which is why item-by-item work such as Kwebulane and Oyekale (2024) matters. The Treasury’s R36.5 billion for 22 items is an aggregate, and the statement does not report how that sum divides across income groups, so nobody outside government can say what fraction reaches the poorest four deciles (National Treasury, 2026a).

Rands are another matter altogether. The same authors point to earlier Treasury work from 2018 which found that, in money terms, richer households gain more from zero-rating because they simply spend more on food (Kwebulane and Oyekale, 2024). The 2026 Budget Review repeats a related finding from the 2018 VAT Panel, that people in the top four income deciles account for over 75 per cent of VAT revenue even though the tax is broad-based (National Treasury, 2026c). That figure is about who pays VAT, not who gains from zero-rating, so it only points in a direction. Put the two together, though, and the picture is awkward. A rich household saves more rand on a basket of staples, a poor household saves more of its income, and the R36.5 billion total cannot be read as money sent to the poor (National Treasury, 2026a).

The wider evidence says VAT in this region does tend to widen inequality. Using data for 35 sub-Saharan countries over 1995 to 2021, Voto et al. (2024) find that effective VAT, meaning VAT revenue relative to final consumption, raises income inequality in both the short and the long run, with the long-run effect larger. That is a regional average and it says nothing about zero-rating as such, so it should not be pushed too far. The South African household study is closer to the point. After the 2018 rise from 14 to 15 per cent, Kwebulane and Oyekale (2025) estimate welfare losses for households generally, with disadvantaged households hit hardest, and they find that demand for food is inelastic, which means poor families cannot simply buy less when prices move.

Their own numbers carry a warning, though. The descriptive table in the same paper puts food at roughly a third of spending for poor and non-poor households alike, about 32 and 34 per cent, which looks more like a product of how the groups were drawn than a real finding (Kwebulane and Oyekale, 2025). The authors still recommend lengthening the zero-rated list to cover all types of bread, soap, candles and school uniforms, and taxing luxury goods at a higher rate. Maybe so. But a longer list was part of the 2025 package that had to be withdrawn, and every extra item would be costed with the same old survey data (National Treasury, 2025; National Treasury, 2026a).

Informality, pass-through and the limits of the evidence

Two things can make the incidence story look different, and neither has been tested properly for South Africa (Kwebulane and Oyekale, 2024). The first is informality. Bachas et al. (2024) use expenditure surveys from 32 low- and middle-income countries, South Africa’s 2011 survey among them, and treat the type of store where a purchase was made as a stand-in for whether VAT was likely to be paid. Budget shares spent in informal stores fall steeply as income rises, so richer households pay a larger share of their income in consumption taxes than the headline rates suggest. The authors conclude that exempting food is hard to justify on equity grounds in low-income countries. That is a strong claim in its own right, and it deserves a hearing here.

It does not carry over cleanly, however. The same paper reports that the formal part of food spending only starts to fall as income rises in upper-middle-income countries, which is where South Africa sits (Bachas et al., 2024). In a country like this, in other words, the poor still put a bigger share of their budgets into formally bought food, and that is exactly the spending zero-rating reaches. What the informality finding does do is lower the confidence one can place in survey sums that treat every rand of food spending as taxed. If the poorest households buy much of their food informally, the effective value of the relief to them may be smaller than it looks, and the NIDS-based work cannot say either way (Kwebulane and Oyekale, 2024).

The second is pass-through. The South African incidence calculations assume that a zero rate reaches shelf prices in full, which is how Kwebulane and Oyekale (2024) treat it. Shops do not always behave that way. Fuest et al. (2025) follow more than 60,000 German supermarket products through the temporary VAT cut of 2020, using Austrian prices as the comparison, and find that prices fell by 1.3 per cent, which means about 70 per cent of the cut was passed on, with private-label goods passing on more than independent brands. Germany is not South Africa and the figure cannot be carried across. The point is narrower than that. No South African equivalent appears to exist, and if part of the relief stays in margins then some of the R36.5 billion never reaches a household at all (National Treasury, 2026a).

What the alternatives can and cannot do

The obvious alternative is to pay poor households directly. Hartley and Gabriel (2024) link a computable general equilibrium model to a microsimulation of South African households to compare responses to the 2022 food price shock, and find that cash transfers gave the most relief to low-income households without making the real GDP effect any worse, while other measures, including the R1.50 a litre cut in the fuel levy between April and July 2022, did comparatively less for them. They did not model VAT relief, which is a gap for this debate, so their result favours transfers in general terms and does not rule on zero-rating. Goldman et al. (2021) are more specific about design. Using a fiscal incidence model built on the 2014/15 Living Conditions Survey, they find that a household-targeted family poverty grant would cut extreme poverty most efficiently but faces stark implementation challenges, while topping up the child support grant is more efficient than a public works programme.

When it comes to money, one comparison is useful. Treasury has set aside R36.9 billion for the SRD grant in 2026/27, extended to 31 March 2027 at R370 a month, which is about what zero-rating of the basic items cost in 2023/24 (National Treasury, 2026d; National Treasury, 2026a). The years differ, so this is a comparison of scale and nothing finer. It does show that the relief costs roughly as much as a grant reaching around eight million people. Swapping one for the other is not just a matter of moving the money, though. Households that lose the zero rate pay more on every loaf, and the replacement has to land with the same households, on time, and at a size that beats the extra tax. Nothing in the 2026 spending chapter points to a permanent income support that would take over once the SRD expires (National Treasury, 2026d).

Coverage is the weak point. In a survey of low-income Tshwane households, about half earned less than R3,500 a month and roughly 47 per cent had no access to social grants, with what access there was concentrated in child and elderly grants (Mazenda and Mushayanyama, 2024). A working-age adult with no children is poorly served by that structure, and zero-rating does not ask who is in the household. Zimbabwe is the cautionary case from the other direction. Limiting zero-rating there to exports and exempting only a few essential items is expected to raise VAT revenue by 0.88 per cent of GDP but also to lift the poverty headcount by 1.4 percentage points, and the authors say compensation has to reach beyond current cash transfer beneficiaries, which needs a social registry the country lacks (Sharma et al., 2025). South Africa has the payment machinery and 26.5 million grant recipients (National Treasury, 2026d), but the Tshwane figures show that having the machinery is not the same as having reached everyone who is poor.

Against all of that, zero-rating has a few things going for it that are easy to undervalue. It needs no application, no registry and no payment system, which is precisely what Zimbabwe lacks, and it reaches working-age adults without children who fall through the grant structure (Sharma et al., 2025; Mazenda and Mushayanyama, 2024). It is also already written into the Act, so keeping it costs no new legislation and no new institution (Republic of South Africa, 1991). Still, easy to run is a different virtue from well aimed, and Goldman et al. (2021) are a reminder that the better-aimed options bring implementation problems of their own.

Zero-rating also sits next to other tax expenditures that deserve the same scrutiny. Riedel and Zinke (2025) find that the marginal tax rate schedule is the main source of redistribution in the personal income tax, that tax expenditures often favour higher-income individuals, and that below-inflation adjustments to thresholds raised post-tax inequality and weakened redistribution. The point for this paper is a plain one. If the aim is to make the tax system kinder to the poor, the income tax brackets and deductions may be a better place to look than the bread list, and they do not carry the pass-through doubts discussed above (Riedel and Zinke, 2025; Fuest et al., 2025).

Conclusion

The evidence supports a more modest position than either side of the usual argument takes. Zero-rating basic foods does shield poor households in proportion to their budgets, and for staples such as maize meal and rice the proportional gain looks clearest (Kwebulane and Oyekale, 2024). It targets them badly, since richer households save more rand and the R36.5 billion cost is spread across everyone who buys the listed foods (National Treasury, 2026a). The informality and pass-through findings suggest the real benefit to the poorest could be smaller than the household-survey sums imply, though nobody has measured either for South Africa (Bachas et al., 2024; Fuest et al., 2025).

On balance the relief should stay for now, on a list no longer than today’s, until something better is in place. Something better means a permanent and wider income support that actually reaches working-age adults, which the SRD grant ends without becoming (National Treasury, 2026d; Goldman et al., 2021). Expanding the list, as one study recommends and the 2025 package proposed, looks premature while the existing list has not been properly costed (Kwebulane and Oyekale, 2025; National Treasury, 2025). The evidence base itself needs work. A fresh incidence study using current surveys and observed shelf prices, with the compensation questions built in as the Zimbabwe exercise attempts, would tell South Africa more than another round of argument about whether the relief is pro-poor (Sharma et al., 2025; National Treasury, 2026a).

Reference list

Bachas, P., Gadenne, L. and Jensen, A. (2024) Informality, consumption taxes, and redistribution. Review of Economic Studies, 91(5), pp. 2604-2634. doi:https://doi.org/10.1093/restud/rdad095

Fuest, C., Neumeier, F. and Stöhlker, D. (2025) The pass-through of temporary VAT rate cuts: evidence from German supermarket retail. International Tax and Public Finance, 32(1), pp. 51-97. doi:https://doi.org/10.1007/s10797-023-09824-7

Goldman, M., Bassier, I., Budlender, J., Mzankomo, L., Woolard, I. and Leibbrandt, M. (2021) Simulation of options to replace the special COVID-19 Social Relief of Distress grant and close the poverty gap at the food poverty line. WIDER Working Paper 165/2021. Helsinki: UNU-WIDER. doi:https://doi.org/10.35188/UNU-WIDER/2021/105-1

Hartley, F. and Gabriel, S. (2024) Distributional impacts of global food price shocks in South Africa: case of Russian-Ukraine conflict. WIDER Working Paper 81/2024. Helsinki: UNU-WIDER. doi:https://doi.org/10.35188/UNU-WIDER/2024/544-8

Kwebulane, B.L. and Oyekale, A.S. (2024) Effects of value added tax zero-rated food products on households’ welfare in South Africa. Research in Agriculture, Livestock and Fisheries, 11(3), pp. 401-415. doi:https://doi.org/10.3329/ralf.v11i3.78947

Kwebulane, B.L. and Oyekale, A.S. (2025) Effects of value added tax increase on households’ welfare in South Africa. Acta Economica, 23(42), pp. 49-77. doi:https://doi.org/10.63356/ace.2025.003

Mazenda, A. and Mushayanyama, T. (2024) Food insecurity and coping strategies of low-income households in Tshwane, South Africa. Poverty & Public Policy, 16(4), pp. 388-406. doi:https://doi.org/10.1002/pop4.417

Riedel, N. and Zinke, I. (2025) On the redistributive impact of the personal income tax: evidence from South Africa. WIDER Working Paper 17/2025. Helsinki: UNU-WIDER. doi:https://doi.org/10.35188/UNU-WIDER/2025/574-5

Sharma, D., Chingozha, T., Goldman, M., Iyer, I. and Steenbergen, V. (2025) Fiscal and distributional implications of VAT reforms in Zimbabwe. Washington, DC: World Bank. doi:https://doi.org/10.1596/42911

Voto, T.P., Ngepah, N. and Gerendawele, N.H. (2024) Value added tax, ethnic fragmentation and income inequality in sub-Saharan Africa. South African Journal of Economic and Management Sciences, 27(1), a5793. doi:https://doi.org/10.4102/sajems.v27i1.5793

Legislation and case law

National Treasury (2025) May 2025 Budget overview. Pretoria: National Treasury. Available at: https://www.treasury.gov.za/documents/National%20Budget/2025May/review/May%202025%20Budget%20Overview.pdf (Accessed: 18 October 2026).

National Treasury (2026a) 2026 Budget Review, Annexure B: Tax expenditure statement. Pretoria: National Treasury. Available at: https://www.treasury.gov.za/documents/national%20budget/2026/review/Annexure%20B.pdf (Accessed: 18 October 2026).

National Treasury (2026b) 2026 Budget Review, Annexure C: Additional tax policy and administrative adjustments. Pretoria: National Treasury. Available at: https://www.treasury.gov.za/documents/national%20budget/2026/review/Annexure%20C.pdf (Accessed: 18 October 2026).

National Treasury (2026c) 2026 Budget Review, Chapter 4: Revenue trends and tax proposals. Pretoria: National Treasury. Available at: https://www.treasury.gov.za/documents/national%20budget/2026/review/Chapter%204.pdf (Accessed: 18 October 2026).

National Treasury (2026d) 2026 Budget Review, Chapter 5: Consolidated spending plans. Pretoria: National Treasury. Available at: https://www.treasury.gov.za/documents/national%20budget/2026/review/Chapter%205.pdf (Accessed: 18 October 2026).

Republic of South Africa (1991) Value-Added Tax Act 89 of 1991, section 11(1)(j) and Schedule 2 Part B. Available at: https://www.sars.gov.za/legal-counsel/primary-legislation/ (Accessed: 18 October 2026).

South African Revenue Service (SARS) (2026) Value-added tax. Available at: https://www.sars.gov.za/types-of-tax/value-added-tax/ (Accessed: 18 October 2026).