Student accommodation is a five-body problem. The rent cap is only one of the five.

Everyone arguing about the NSFAS accommodation cap is right. That is one reason the country is not building enough beds.

On 28 August, Parliament's Portfolio Committee on Higher Education held a colloquium on student accommodation. The day before, students had marched from the NSFAS offices to Parliament. The argument, in the committee room and on the street, is about a single number: what the cap should be and whether it should exist at all. Ten universities' student councils want it scrapped.

It is the wrong argument. The cap matters but it is one setting on a machine with five dials, and turning a single dial does not make the machine run. It moves the strain to whoever can least resist it, which at the moment is the student.

The five constraints

Student accommodation will be built at the scale the country needs only when five constraints hold at the same time. Each pulls against the others. Four of them belong to parties. The fifth belongs to engineering and maintenance.

The state wants beds inside an affordability envelope, paid for from a budget that cannot become an open cheque. So it holds the rent down.

The student needs a bed that is safe, clean and close enough to campus to actually study from and cannot pay a cent above the cap. That is quality up and price down at the same time, the hardest ask in the system, and it lands on one person.

The developer wants a return that beats its hurdle, because the same rand of equity could go into a warehouse, or a mall, or a data centre. Short of that return, it walks.

The bank wants to be paid and to know before it lends that it will be. It pushes the other way from the developer: less debt, more cover, a stronger covenant, a smaller and safer deal.

And the building costs what it costs. A bed that is safe, durable and built to last decades has a real cost to build and maintain, and neither of them cares what the cap is set at. It is the fixed point the other four negotiate around.

Every one of these is reasonable, and that is what makes it hard. There is no one to blame and no argument that resolves it. The deadlock breaks only if all five can hold at the same time, and whether such a setting exists is something the numbers can settle.

Whether one exists is arithmetic

Take the rent paid for the bed as the revenue, net of operating costs of about thirty-eight percent and a separate reserve for lifecycle capital on top. Finance it the ordinary way: roughly seventy percent senior debt at about ten and a half percent, the rest equity looking for about thirteen percent after tax. Now hold it to the constraints the capital imposes. First-year net income has to cover the debt service by about 1.2 times. The equity has to clear its hurdle over the holding period and the exit. These are conditions the stack must meet, not a guess at any particular lender's credit policy. Fix all of that, and for a given rent there is a ceiling on what you can spend per bed. Above it, the capital stack no longer closes.

Run it at the cap. At R45,000 the ceiling sits near R218,000 a bed. At the R52,000 metro cap it lifts to about R252,000, and there the binding constraint is the lender's cover. So R52,000 of rent does not finance a R300,000 bed on this stack. It finances roughly R252,000 over a thirty-year payback, and that is as far as it goes: the maximum build cost this capital stack can carry per bed while servicing the debt, clearing the equity hurdle and staying under the cap.

Now the cost of the thing itself. Recent institutional developments put the capital cost of a modern, compliant South African bed at around R300,000 or more. Market build-cost benchmarks sit in the same territory, and even the World Bank's efficient-build baseline was R225,000 in 2020. That is not an assumption chosen to suit the argument. It is where real delivery actually sits, and it sits above the ceiling the rent can carry.

The shortfall at R52,000 rent cap is somewhere between R50,000 and R75,000 capital per bed. At R45,000 it is wider. The machine does not close, and nobody in it is behaving unreasonably. Five reasonable constraints simply do not meet.

But some developers are building at these rents

They are, so the cap plainly does not make development impossible. It makes it selective, and selective is not the same thing as scalable.

A large, sophisticated owner can carry a higher build cost than that ceiling suggests, because its real stack and operating economics differ from the ones assumed here. It might bring cheaper equity, land it already holds or holds cheaply, procurement and design scale, a longer hold, a stronger residual, or capped beds cross-subsidised by market-rent beds inside a bigger portfolio. Change enough of those and R52,000 works. The model does not dispute it. It asks to see the stack. A profitable scheme proves that some combination of advantages is enough; it does not prove the rent supports a bed for a developer coming into the market without them.

And even the selective building is slowing. As construction and running costs climb against a fixed rent, the people who actually build and fund these schemes are saying the same thing on the record: projects paused or cancelled, financing harder to raise, lenders re-pricing the risk and stepping back. Supply is falling while demand rises. The fact that profitable schemes exist does not tell you the settings are producing enough beds. What is visible tells you the opposite: the deals that close are the exceptions, and there are fewer of them each year.

The market has shown that these rents can support institutional development. It has not shown that they can support enough of it.

The part a property firm has to say plainly

Some private providers do overcharge. Collusion in the sector is real enough that the Competition Commission has been looking into it, and that scrutiny is warranted.

But it is a separate problem. The gap here is not what a landlord pockets. It is the distance between what the rent cap can finance and what an honest, efficiently-built, safely-run bed costs before a cent of profit is added. Strip every rand of profiteering out of the market and the gap barely moves, because the problem is structural. The cost of building safely is fixed, and a rent set below it does not get you a cheaper bed. It gets you no bed at all, and the shortfall lands on the student, as debt, or as a mattress on a residence floor in February.

Why the cap fight is unwinnable

The five constraints are simultaneous. You can satisfy any four by breaking the fifth.

Raise the rent above the cap and the machine closes financially, but the state and the student pay for it. Cut the bank's cover and it closes, right up until first-year revenue disappoints and the lender is short. Cut the developer's return and it closes, until the equity goes to the data centre instead. Build the bed cheaper and it closes, until you have engineered out the fire escape along with the student's safety. Leave the cap where it is and it closes on paper, while in the real world the beds are not built and the student bears the brunt.

Every lever that eases one party tightens another. What counts is not whether a lever moves the number, but whether it moves the binding constraint. Change the rent and the number moves; but when the lender's cover is what binds, you have only slid along the frontier, and the gap is exactly where it was. The rent cap is one dial. Turning it adds nothing to the system. It only decides who absorbs the shortfall.

The lever that loosens without tightening

That points to the one intervention that changes the answer instead of relocating the strain: public capital and concessional finance, placed inside the stack rather than on top of the rent.

A grant reduces the amount that has to be financed at all. Concessional or development-finance debt lowers the blended cost of the money and stretches its tenor. None of this touches the rent, so Treasury and the student are left alone. It strengthens the bank's cover rather than weakening it. It makes the developer's required return reachable inside the cap rather than cutting it. It takes nothing from the building. What changes is the amount and the cost of the capital, not the price of the bed.

And the difference matters. Subsidising the rent pays for the symptom every year; buying down the capital changes the economics of the asset once. You can watch it happen in the model. The public version shows the problem, that at the cap, on private capital alone, the machine will not close. Add a grant and/or a concessional tranche and the supportable cost climbs out of the low R200,000s toward the real R300,000 with the rent never moving. That is the machine closing, and it is not a thought experiment: it is the structure the department's own Student Housing Infrastructure Programme and the Development Bank already use, where the capital exists to do it.

The dials, and what each one costs

• Rent: relieves the developer and the bank; breaks Treasury and the student.
• Payback period: helps at the margin, then flattens out below cost. No length closes the gap on its own, and a longer one only lengthens the commitment.
• Gearing: relieves the equity; the bank resists it, because more debt is less cover.
• Cost of debt: relieves everyone, but the market prices commercial debt. You cannot wish it down, only replace it with concessional money.
• Build cost: relieves everyone financially, down to a hard floor, below which the bed is unsafe.
• Capital grant and concessional debt: loosen the binding constraint without pushing the rent up. They change the amount and cost of the capital, not the price of the bed.

What the debate is actually asking

Strip the politics out and the question underneath is not what the cap should be. At the colloquium, the department itself asked for the cap to be reconceptualised and for a proper investigation into what accommodation costs. Universities and ownership bodies called for an evidence-based cap, built from the real costs of utilities, maintenance, security and compliance. Nobody in the room was asking for a bigger number. They were asking for an honest one, and that is a question arithmetic can answer.

The beds get built when the machine closes. The machine closes when the strain stops landing on the student.

That is why the useful question is not what the cap should be. It is what the rent cap can build, and what capital structure closes the gap between the two.

We built the arithmetic. Every dial is on one screen — move them yourself.