Why South Africa’s coalition politics is a failure

Thembelani! Thembeeeeelaaani!

The call came at the worst possible hour, which was the only hour it ever came: the moment I stepped out of the basin, clean for the first time since sunrise. Bathing came last in a village boy’s day, reached only when everything else was done, and reaching it was its own small achievement. It came from across the yard, and it meant one thing. Banana was in the maize.

Banana was ours. The maize was Eva’s. Eva was not her name, but she ended every sentence with uyeva, do you hear, and to us children, the word had become the woman.

So I went: clean legs back through the blackjacks, maize higher than my head, the cow entirely unbothered, my one decent shirt collecting seeds I would be picking out for a week.

Some evenings Banana broke out of the kraal; some afternoons she came down off the mountain ahead of the herd and did it in daylight. The pattern never changed, and neither did the call. Eva never shouted for somebody. She never shouted at the cow. She shouted a name, because she knew whose cow it was and whose gate had failed, and that knowing was the whole system: a name, a household, a boy who had to go.

A hung council is a maize field with a cow in it and nobody to shout for.

In the Johannesburg council chamber, a budget vote seldom sounds like an argument about money. Nobody rises to say that the tariff increase falls on the wrong households, or that the repairs and maintenance line is a fraction of what the pipes and substations require.

The speeches are about betrayal: who walked out of which agreement, who was promised which seat, who is a sell-out and who is a puppet. Then the budget is defeated, or deferred, and the city drifts toward the one deadline in municipal law that forgives nothing. The maize is eaten. Nobody owns the cow.

The voter’s injury

The case I want to make is not about stability. Stability is a poor foundation for any political argument; a one-party state is stable. The case is about accountability, and accountability has a precondition that political scientists call clarity of responsibility.

A voter can reward or punish performance only if she can identify who produced it. The ballot is a promise; the audit is the receipt.

When the government that signs the receipt was assembled in a hotel boardroom three weeks after the election, from parties that campaigned against one another, the promise and the receipt no longer belong to the same signature. The voter has been asked to grade a paper she did not set.

This is the injury. Not that coalitions are slow or that mayors change too often, though both are true, but that voting has been disconnected from governing, and the parties made the disconnection, not the voter. Since 2021, Johannesburg has had six mayors.

The 2021 election produced a record 66 hung councils, four of them metros. Nobody voted for any of that.

What Cape Town actually taught

Cape Town is usually cited as the coalition that worked. It is better understood as the coalition that was absorbed.

In 2006, the DA formed a government with six smaller partners. By 2010 the largest of them, the Independent Democrats, had dissolved into the DA; by 2011 the rest had shrunk to irrelevance and the DA won the city outright.

Whatever the partners contributed to five years of government, the voter credited one party and one leader. That is not a Cape Town peculiarity; voters everywhere attribute performance to the face at the front.

The lesson the system teaches small parties is brutal, and it is the key to everything that followed in Gauteng. Inside a coalition you are invisible.

Cooperation earns you nothing at the next election because the credit flows upward to the leader. The only route to survival is differentiation, and the cheapest differentiation available to a party with four seats is disruption. Sabotage is not a character defect in coalition partners. It is the rational response to a system that punishes loyalty with extinction and rewards defiance with a headline.

‘Opposition’ is the wrong word

We describe these councils in a vocabulary borrowed from Westminster, and it misleads us. Opposition presupposes a government. In a hung council there is no government to oppose, only a temporary arrangement that every party, including those inside it, is already campaigning against. 

Each party is at once a veto player and a candidate. Opposition stops being a position and becomes a posture, and the posture never ends because the next motion of no confidence is always a week away.

The budget is where this does its damage. The Municipal Finance Management Act requires the council to approve a budget before the financial year starts; if it does not, the province may intervene under section 139 of the Constitution.

It is therefore the one council vote with a hard deadline and a hard consequence, which makes it the most valuable hostage in the building. Whatever the substantive dispute in a given year, the incentive is unmistakeable: blocking the budget is the one move guaranteed to produce a crisis, a camera and a concession. The point is to be seen.

A system that manufactures permanent campaigns will get permanent campaigning. That sentence should be carved above the door of every council chamber in Gauteng.

Move the coalition

The instinctive remedy is to abolish proportional representation in councils and elect every seat by ward, first past the post. The appeal is real. Parties too small to win on their own must join forces before the election or risk winning nothing.

The deal is therefore made in public, before a single ballot is cast, and voters get to accept or reject the coalition itself. The Multi-Party Charter of 2024 attempted this and fell apart because under proportional rules, there was no penalty for walking away. First past the post supplies the penalty. Coalitions before the ballot, not after it.

I state the appeal fairly because I am about to reject the form of it.

Section 157 of the Constitution requires that the local electoral system produce, in general, proportional representation. Abolishing that needs a two-thirds majority, which needs the ANC, and that is the least of the problems.

Our wards were drawn on to apartheid spatial planning. A pure ward system in this country would produce ethnically territorial councils almost by construction, with minorities inside every ward invisible.

Proportional representation at local level was not a technocratic preference in 1996; it was part of the settlement. And there is a test every advocate of first past the post must pass: run the 2021 Johannesburg ward results through it. The ANC won more Johannesburg wards than any other party. The system would have handed the city an ANC majority. If you cannot live with that outcome, your argument was never about accountability.

There is a further objection from my own trade. Msunduzi, and most of the municipalities with disclaimed and adverse audit opinions, are majority councils. A clear signature on the receipt does not make the receipt good. First past the post removes the coalition excuse; it does not supply the consequence. Accountability does not guarantee competence. It only makes incompetence punishable, and that is the most a voting system can be asked to do.

Where the voter already looks

So the reform should go where the sabotage bites, which is the executive, not the legislature. Cape Town’s real lesson is that voters think in terms of the leader.

Give them the ballot that matches how they already think. Let residents elect the executive mayor directly, on a separate ballot, in two rounds if no candidate wins a majority, while the council remains proportional. A mayor with a personal mandate cannot be traded between parties in a boardroom, and a party that wants the mayoralty must win it from residents rather than from rivals.

Pair it with two rules that remove the hostages.

First, a constructive motion of no confidence, as in Germany: a council may remove a mayor only by electing a successor in the same vote, so a motion becomes a decision about who governs rather than a weekly instrument of extortion.

Second, a budget reversion rule: if a council fails to pass a budget before the financial year, last year’s budget runs on, adjusted for inflation, until one is passed. Blocking then produces nothing, and a frozen budget freezes the mayor’s programme too, so neither side wins by stalemate.

A directly elected mayor can, of course, face a hostile council: two mandates from one city. But a quarrel between two identifiable mandates is better than government whose mandate cannot be identified at all. The object is not to remove conflict from the chamber; it is to make clear who owns each side of it.

Most of this sits in the Municipal Structures Act and the MFMA, which is where Parliament is already working; whether direct election of mayors could be introduced without constitutional amendment is a question for constitutional lawyers.

The Coalitions Bill now before it compels written coalition agreements, sets a one percent threshold and shifts hung councils to collective executives. It is unlikely to pass before 4 November, and in any case it regulates conduct after the ballot. Everything proposed here changes the incentives before it.

Eva could shout my name because Banana belonged to somebody. That is the whole argument. A city in which no one owns the government is a maize field where the cow is eating and the old woman at the fence has no name to call, and has learnt to stop calling.

We have built that city three times over in Gauteng since 2016, and we have blamed the cows. It is time to ask who built the kraal. Uyeva? DM