The levy relief was swamped by the oil price
The Budget Review included a downside "scenario B": escalating geopolitical tension, higher oil and gas prices, a weaker rand, delayed rate cuts and slower spending. Treasury's own debt model said that world would need a long-run primary surplus of 3.7% of GDP just to keep debt stable, against 2.9% in the baseline. That gap of about 0.8% of GDP a year is Treasury's own price tag for this shock.
The government's first response was to cut the fuel levy by R3 a litre from April, and in May to take the diesel levy to zero. The chart shows why that hardly registered at the pump. October's adjustment then added another R2.84 to diesel and R3.33 to petrol.
Inland (Gauteng, zone 9C) prices, rand per litre: 0.05% sulphur diesel by component (diesel is regulated at wholesale, so pump prices vary) and the regulated 95 unleaded pump price. Department of Mineral and Petroleum Resources via Econdata; October 2026 from the DMPR announcement, with its diesel split assumed unchanged except for the basic fuel price.
In May the diesel levy was zero, and diesel still cost R12.64 a litre more than in March.
The primary surplus is on track. The ground under it has moved.
Treasury's director-general, Duncan Pieterse, wrote in September that government will deliver a primary surplus of about R131 billion this year, its fourth in a row, and that revenue is running ahead of the Budget thanks to company tax. That is the February number to the rand, and it deserves credit.
Two things have moved against him since then.
1. An oil shock raises consumer prices, but it shrinks the tax base
Consumer inflation rose to 4.4%. But South Africa imports its oil, so a dearer barrel lowers the price of what we produce relative to what we buy. In the second quarter the GDP deflator fell 1.6% and nominal GDP fell 1.9%. Nominal GDP was 4.5% higher than a year earlier, against the 5.6% the Budget assumed for 2026/27. Nominal GDP is both the tax base and the denominator of the debt ratio. With debt near 78% of GDP, each percentage point of missing nominal growth adds roughly 0.8 points to the debt ratio.
2. Borrowing costs are up, and a lot of debt is about to be refinanced
The 10-year yield is about 0.8 points above its level at Budget time. Short-term debt and floating-rate notes, about R930 billion, reprice within a year. That adds about R7.5 billion a year in interest. New bonds to cover R1.4 trillion of gross borrowing over three years add about R10 billion more by 2028/29. By then that is roughly R17–18 billion a year. My rough sum, about the cost of the fuel levy relief, every year.
Debt redemptions by fiscal year, R billion. Read from the vector data of Figure 3.1, National Treasury 2026 Budget Review; matches Table 7.2 for 2024/25–2028/29 and the stated R283.8 billion average for 2026/27–2033/34.
Credibility pays back slowly, a few billion at a time as debt rolls over. Losing it costs fast: the whole yield curve reprices at once.
Poor households felt it in the taxi before they felt it at the till
Petrol and diesel hardly feature in a poor household's budget. In the CPI basket, the poorest fifth of households spend 0.2% of their budget on fuel for their own vehicles and 6.9% on taxi and bus fares. For the richest fifth it is the other way round: 6.4% on fuel and 0.8% on fares. Food takes 38% of the poorest budget and 10% of the richest.
So far fares have taken the hit. Food inflation actually fell this year, because meat and grain prices dropped.
Annual inflation, %. Stats SA CPI, all urban areas, via Econdata. Fares are COICOP class 07.3 (mostly road: taxi and bus).
Who got the R17.2 billion? I followed the relief through the three channels I can measure: cheaper fuel for a household's own car, cheaper fares, and cheaper food. Then I asked how much reached each fifth of households.
My estimate, R billion, April 2026–March 2027. Household spending: SARB national accounts (Q2 2026, annualised). How it splits across fifths and what each fifth buys: Stats SA CPI expenditure-quintile weights (2012, the latest with this detail). Fare and food effects from the pass-through estimates below. The remaining ≈R5 billion went to firms, government and exports, or was passed on through prices I can't trace.
About R1 billion, one rand in every seventeen, reached the poorest 40% of households. The same R17.2 billion paid through the child support grant would have been R1,364 per child: more than two months' extra grant.
Is more still to come on food?
I estimated how food prices and fares respond to diesel, using monthly data since 2008. Within a year, a 10% rise in diesel has typically meant:
- Fares up about 3.3%. This is a strong, clean result, and it falls back when diesel falls.
- Food up about 0.6%. This is a weaker result. It is driven by the 2008 and 2022 commodity spikes, so treat it as an upper bound.
- Vegetables up about 2.2%, and they don't come down again when diesel falls. Prices rise like a rocket and fall like a feather.
Inland diesel is up 72% since March. On these estimates, a food price push of a few percent is still working through, hidden so far by cheaper meat and grain.
What I'll be watching
The left column is what the February Budget promised. On the night I'll fill in the right column and say whether the Minister held the line.
| Number | February Budget | What I'm looking for | October MTBPS |
|---|---|---|---|
| Primary surplus, 2026/27 | R131.2bn (1.6% of GDP) | Kept close to R131bn. A shortfall explained by a one-off is fine; a quiet slide is not. | 21 Oct |
| Debt peak | 78.9% of GDP in 2025/26, falling to 77.3% in 2026/27 | Does the peak move a year later, or above 80%? Weaker nominal GDP pushes it that way. | 21 Oct |
| Growth, 2026 | 1.6% real; 5.6% nominal in 2026/27 | An honest cut. The SARB is at 1.2%; nominal GDP is running near 4.5%. | 21 Oct |
| Gross borrowing, 2027/28 | R568.7bn, including R284.2bn of redemptions | How much higher, and how much is pushed into short-term Treasury bills that reprice fast. | 21 Oct |
| Help for households | No relief budgeted; R17.2bn of levy relief spent April–June | If there is more help, it should go through grants or fares, not the levy, and be paid for inside the spending envelope. | 21 Oct |
| Fiscal anchor | A "principles-led" anchor promised for this MTBPS | A clear escape clause for big shocks, and a clear statement that this one doesn't trigger it. | 21 Oct |
| Reform money | Guarantees R661bn (up R148bn in a year, mostly Transnet) | Electricity transmission funded through guarantees and private partners. Guarantees are debt in waiting, so watch the total. | 21 Oct |
My view going in. The Minister should hold the primary surplus. A supply shock can't be fixed with spending: the Reserve Bank, defending a new 3% target, would raise rates to offset it, and bond markets would add a risk premium on top. Help for the poor should be targeted, temporary and paid for inside the budget, which the revenue overrun makes possible. Electricity reform is cheap in rands and expensive in political will. Where it does need money, it will come as guarantees, and those deserve the same scrutiny as debt.
Michael Avery recently recalled Churchill's regret that finance was proud while industry was not content. He is right that respectability is not a growth strategy. It is the thing that keeps an oil shock from turning into a debt crisis.