Conflict is the sound made by the cracks in a system, the manifestation of contradictory forces coexisting in a single space.
Kenneth Cloke
On the 4th November, South Africans vote for councils in all eight metros. Within fourteen days of the results, the winners will have to form governments. Most of them will do so without a majority, without a finalized legal framework, and with a recent record that voters have already judged. As my articles of the last few years have been pointing out, our politicians are particularly unsuccessful at upgrading their coalition conflict skills. I have had years of private discussions with several of them, with a continued high degree of interest and enthusiasm, and then the organizational paralysis that settles back into the easier answers, the clichés. This is then a frank, practical note to the parties who will sit around those leadership tables.
1. The problem: unnecessary and self-imposed
South Africa’s metros have become some of the most expensive laboratories in the democratic world for a single question: what happens when parties are good at winning votes and bad at sharing power? The numbers are not subtle. Johannesburg has had seven mayors since the 2021 elections, according to Reuters. Nobody has held an outright majority there since 2016.
Tshwane changed hands when the DA’s Cilliers Brink was removed in September 2024 and replaced, 122 votes to 86 in a 214-seat council, by ActionSA’s Nasiphi Moya, backed by the ANC and EFF. Nelson Mandela Bay has been unstable since well before 2016. A former DA mayor there recalled 37 municipal managers since 2009. In 2023 a DA-led coalition fell to a motion moved by the EFF, and in 2026 the ANC-led coalition is still fighting for coherence: two no-confidence motions against Mayor Babalwa Lobishe failed in March, a coalition member resigned from her mayoral committee in May, and a suspended city manager has reportedly cost ratepayers more than R10 million. Nationally, Deputy Minister Dickson Masemola says that since 2000 there have been 82 hung municipalities out of 257. There is no statutory framework for any of this.
The Municipal Structures Amendment Bill, the so-called Coalition Bill, was approved by Cabinet on 29 July 2026, but reporting in October confirms it will not become law before the 4 November elections. The first council meeting will therefore be governed by the old rules: parties have about fourteen days to form a government, and a simple majority can remove a mayor on any given day, for any whim, sulk or reason. Polling suggests the problem is not about to go away. Ipsos has the ANC and DA locked in close races in Johannesburg, Ekurhuleni, Nelson Mandela Bay and Buffalo City. In Nelson Mandela Bay, polling discussed by analyst Melanie Verwoerd has the two parties level at roughly 37% each.
In 2021, turnout across the country was 45.86%, a figure that tells you what voters think of the current product. Let me be direct about the diagnosis, because every party at the table will want to blame someone else. That is part of the problem, as conflict analysis, that is so visibly unnecessary.
The instability in our metros has at least five distinct causes, and each party is responsible for some of them:
Parties find a majority and then, belatedly, look for a programme.
Finance, water and electricity are traded like any other post.
Because a motion of no confidence costs nothing and can be filed without a successor or a plan, it has become a bargaining chip, a revenge tool and, at times, a business model for very small parties.
City managers and senior officials are suspended, replaced or protected according to who holds the mayoral chain. Residents experience this as broken water pipes and potholes, not as politics.
When a coalition quarrels, there is nobody with the standing and the time limit to settle it before it spills into the council chamber, or turns into the slow-cooking dysfunctionalities that voters have to put up with. None of this is unique to South Africa. Which is the good news, because other democracies have already learned what works. Other than lip service, our politicians find it easier to blame and to deflect, than to do the work necessary, staring with skill and comparative analysis.
2. What global practice actually tells us
The comparative literature on coalition governance is large and surprisingly consistent. I will be the first to concede that some of that body of conflict work do not, or may not, be successfully transferred to a South African political environment, so I have limited our discussion to principles that will do so.
Four findings matter for our metros.
First, coalitions are policed, not trusted:
Lanny Martin and Georg Vanberg’s study Policing the Bargain (2011) shows that parties in coalition do not rely on goodwill; they build monitoring devices, such as detailed agreements and junior partners placed in senior portfolios to watch what the senior partner does. Kaare Strøm, Wolfgang Müller and Daniel Smith (2010) reach the same conclusion about parliamentary control: stable coalitions are the ones with institutionalised oversight inside the executive. The lesson is that suspicion is not a sign of a failing coalition. Unmanaged suspicion is. This puts an effective end to the sullen excuse that a particular party cannot ever be in coalition with people like them, or that “they” cannot be trusted. Know how to do this important part of your craft, and honesty, trust and friendship become extras.
Second, proportionality in the allocation of offices reduces fights over offices:
Gamson’s Law, proposed by William Gamson in 1961 and repeatedly confirmed since, finds that cabinet posts tend to be distributed roughly in proportion to each partner’s seats. Deviations from proportionality are where grievance grows. How to enforce this before our legislation becomes binding we can deal with below.
Third, the most stable systems make destabilizing a government difficult rather than impossible:
Germany’s Basic Law, in Article 67, provides for a “constructive vote of no confidence”: a chancellor can only be removed if the legislature simultaneously elects a successor. This single rule was designed after the Weimar experience to stop negative majorities, coalitions of parties who agree on what they oppose and nothing else, from bringing governments down. Sweden’s 2019 January Agreement and New Zealand’s confidence-and-supply arrangements show a related lesson: smaller parties can give a government reliable support on the budget without sitting in the executive. The “Coalition Bill” is disappointing on certain aspects, but a skilled politician can get most of these practical effects in place, and hold that there, without legislation. Interests, immediate costs to misconduct, identity and face saving dynamics – all there for those who actually want to implement these systems.
Fourth, process is a design choice:
Fisher and Ury’s Getting to Yes (1981) is usually read as a negotiation manual, but its practical core applies directly to coalition formation: separate people from the problem, focus on interests rather than positions, generate options before deciding, and use objective criteria. Since then an impressive body of work has modernized and expanded their pioneering work, and yet we see no trace of it in our coalition politics. Germany’s Koalitionsvertrag (a detailed written coalition agreement) and its standing coalition committee, and the Dutch practice of appointing an informateur to explore the field before the formal formation talks begin, are institutional versions of that discipline. Again, no legislation necessary.
South Africa’s own national Government of National Unity, with its Statement of Intent, has shown both the value of a written compact and the cost of leaving the dispute mechanism unbuilt: reporting has noted that no structure had been established to resolve policy deadlocks. The GNU is an example of an incomplete, well-intentioned conflict instrument that can only reach its potential if at least some of the involved parties are more skilled in coalition conflicts. As we have seen, none of these require a new law. All of them can be agreed by willing parties in the fourteen days after an election. The rest of this note is about how.
3. Eight rules for the first 14 days and the first 100
Rule 1: Appoint a convenor and run the fourteen days as a process
Do not let formation happen in a series of whispered bilaterals that leak to the press. Agree on a neutral convenor, such as a retired judge, a respected academic or a former public protector, with the explicit task of managing the talks. A workable example calendar (that has worked elsewhere) is: days 1 to 3, exploratory meetings and agreement on the ground rules; days 4 to 9, programme and budget priorities; days 10 to 12, offices and portfolios; day 13, signing; day 14, council. The key discipline is sequence: programme first, posts second. A coalition formed in the opposite order is a hiring spree, not a government.
Rule 2: Write a one-page agreement the public can read, and a longer one the partners can enforce
The public document should contain no more than eight commitments for the first 100 days, each with a named owner, a budget line and a date. They should be the things residents actually experience: water reliability, electricity and outages, billing accuracy, refuse and roads. Publish a scorecard against them monthly. This is not utopian. In Tshwane, ActionSA has already run a public service-delivery tracker, and the principle should be extended across every coalition regardless of party. The longer, confidential annex should cover the mechanics: decision-making, budget process, dispute resolution, personnel. Public accountability in a neutral, effective manner.
Rule 3: Distribute offices by Gamson’s Law, and weight them Start from proportionality and then deliberately weight the offices that matter most. Finance, infrastructure and utilities should not be handed over on the basis of who needs a reward. Partners should agree the criteria for those portfolios in writing: qualifications, track record, a declared interest register. Then use Martin and Vanberg’s monitoring device: pair each senior portfolio holder with a partner party’s deputy or shadow, with access to the same information. In metros the legal power to appoint mayoral committee members rests with the mayor, so this needs to be done by written undertaking rather than by statute. But the mayor should sign it.
Rule 4: Make no-confidence constructive by agreement
Before the Coalition Bill becomes law, parties can adopt its logic voluntarily. Every partner signs an undertaking that no motion of no confidence will be tabled or supported unless (a) written notice of at least 21 days has been given to the partners, (b) the dispute process in Rule 5 has been exhausted, and (c) a named successor and a revised programme have been tabled alongside the motion. In addition, agree a quiet period: no motions in the first twelve months of the term and none in the six months before the next election. Parties that break the undertaking should expect to be named, publicly, by the convenor. In politics, reputational cost is usually the only enforcement that works. Think through other mechanisms that you may want to be effective – get that signed while the honeymoon is still ongoing. It will have legal and persuasive power, and it will most certainly have value in public disputes later on, if that is unavoidable.
Rule 5: Build the dispute ladder before you need it
Two tiers are not enough. Use four, each with a deadline. Level one: the officials responsible for the matter, within five working days. Level two: the party principals’ coalition committee, which meets monthly whether or not there is a crisis, as in Germany. Level three: an independent mediator named in the agreement, with ten working days to recommend. Level four: a written statement of the disagreement to all councillors before any public move. Conflicts that are not escalated deliberately escalate themselves. Build in pre-agreed consequences for anticipated misconduct. Let each senior participant spend a compulsory two or three hours on understanding the value and use of modern mediation and even arbitration (or other conflict resolution tools) as tools to serve their public.
Rule 6: Build a firewall around the administration
This is the rule that residents care about most and parties find hardest. City managers, chief financial officers and heads of utilities should be appointed through a competence panel that includes external experts, not a caucus deal, and should be insulated from removal by the mayor’s political fortunes. Suspensions should be exceptional, supported by a preliminary independent investigation, and time-limited, with a hard cap measured in weeks, not years. The cost of getting this wrong is visible in Nelson Mandela Bay, where a suspended city manager became a two-year drain on money and credibility, and in Tshwane, where the status of the city manager has been a recurring flashpoint between partners. An administration that does not know who it works for will work for nobody.
Rule 7: Treat the budget as the one non-negotiable vote
The annual budget and the integrated development plan are the points at which a failure of coalition becomes a failure of government. Under section 139(4) of the Constitution, a council that cannot approve a budget can face provincial intervention. Every partner, and every party that provides confidence-and-supply support from outside the executive, should commit in writing that the budget and the IDP will pass, even when other disagreements remain. That is the New Zealand and Swedish model applied to municipal government, and it gives smaller parties a legitimate, lower-cost way to exert influence without taking posts they cannot staff.
Rule 8: take a day or three and upgrade a selected team of senior leaders and staff involved, preferably non-partisan across the involved spectrum, but a specialized in-house partisan team if nothing else, and upgrade their conflict skills in general, and coalition conflict skills in particular
Close the doors, get elite level coaching on the specific tools you will encounter: escalation and de-escalation, the alternatives to compromise, the mechanisms of identity conflicts and face saving realities, persuasion across those lines, a practical, comparative study on applicable coalition best practices and a few other select indispensable modern conflict skills. Even if only a few people in the room have this upgraded skill they can already influence better outcomes, agreement or cooperation is not necessary.
4. City by city
Johannesburg: the prize, and the most urgent test Johannesburg is the largest budget and the greatest symbolic prize. It is also where the churn and unresolved conflict spirals have been most visible. The city’s core problem is not a shortage of plans. It is, to a very real extent, that no coalition has lasted long enough to complete one. The parties competing hardest there, the ANC and the DA, face a hard truth about each other: neither can govern without somebody, and the likeliest partners are each other, a combination of smaller parties or a patchwork of both. A poll commissioned by the Common Sense group suggested most ANC voters would be comfortable with an ANC–DA arrangement in municipalities, and EWN has reported analysts openly discussing the possibility in Johannesburg. A DA–ANC compact would be controversial inside both parties, and neither should enter one for the sake of headlines. But if it is the only configuration that produces a majority and a programme, both should be willing to negotiate it honestly rather than treat the idea as a betrayal. An important point to notice here: if these two parties, as examples, learn to effectively manage the friction and potential of coalitions, most of their stated objections to working with each other disappear. The practical advice for Johannesburg is as follows. First, whichever parties govern should agree on a short infrastructure compact covering water, electricity distribution and billing, ring-fenced from the daily cut and thrust, with a published performance standard and an independent monitor. Second, accept the plurality convention as a starting point: the party with the most seats gets the first opportunity to form a government and name the mayor, but without an entitlement to do so.
This is the norm in Germany and the Netherlands and it reduces the sense of theft that follows when a mayor is installed by the party that came third. Third, avoid rotating mayors. It has been tried in other places, and it tends to produce a caretaker mentality and a permanent campaign. Finally, make the mayor and deputy mayor jointly accountable for the 100-day scorecard, ideally from different parties. Run this in public, as a mirror, not as an open wound to score X points on.
Tshwane: the lesson of the “sidecar”
Tshwane is the clearest illustration of what happens when the party that holds the chain is not the party that holds the largest bloc. Mr Brink’s description of a mayor who holds the title while another party runs the city is a political jibe, but it describes a real structural risk. Mayor Moya, in turn, says her ANC, ActionSA and EFF coalition has stabilised finances and reduced Eskom debt. The DA disputes the claim and has pointed to the Bronkhorstspruit water crisis and large water-tanker spending. These claims are contested and reasonable people will weigh them differently. What is not contested is that the coalition has been visibly strained, with public disputes between ANC structures and the mayor over procurement and the city manager. For Tshwane, the advice differs by party.
If the mayor comes from a smaller party, agree in writing what the mayor controls: appointment of the mayoral committee, final say over the city manager’s performance agreement, and authority to veto procurement decisions above a threshold. A mayor without these protections is an ornament.
If you accept a smaller party’s mayor for the sake of a majority, do not then try to run the city through your regional structure. That is how coalitions collapse. Use the coalition committee, not the party office.
The 2024 removal is a legitimate grievance, but a grievance is not a strategy. If the DA wants to lead again, it should publish in advance the terms on which it will work with others and what it will do if it is again outvoted. Voters have noticed that every party criticises motions of no confidence when it is on the receiving end. With increased coalition conflict competence will come increased confidence in stating what will, and should, happen, a benefit that will be well-received by voters.
Bronkhorstspruit and Hammanskraal should be on the scorecard of every coalition. Make water quality a jointly owned metric rather than a weapon.
Nelson Mandela Bay: stability by design
Nelson Mandela Bay deserves the most radical thinking, because naïve and / or ineffective conflict thinking has failed there for a decade. Twelve parties sat in the council after 2021, and the ANC and the DA were within a hair of each other. The metro has had a change of mayor after each major quarrel, a parade of acting and suspended city managers, and a history of provincial intervention being threatened or discussed.
In 2026 the ANC-led coalition has been tested not only by opposition motions but by internal ones, including competing motions against the mayor and her deputy, and by a member who left the mayoral committee over the city manager’s suspension. The ordinary cure is not enough.
A bolder one is a five-year stabilization compact, signed in public by every party that wishes to participate in government or in supply of the budget, with the following features:
The ANC and the DA, who between them won most of the vote in 2021 and are polling level now, face the question that neither wants to put on a poster: can they work together in this one city for five years for the sake of the residents? If the answer is no, then they owe voters a clear explanation of who will provide stability instead. If the answer is yes, they should say so before the election and not after it.
Ekurhuleni, eThekwini, Buffalo City and Mangaung
Ekurhuleni and Buffalo City resemble the Johannesburg pattern: ANC and DA races that are tight, with smaller parties holding the balance. The same convenor-led process and the same stabilisation instruments apply. eThekwini is different. The MK Party’s strength changes the arithmetic, and the polling suggests that it has the best chance of a dominant position. Whatever one thinks of MK, the principle of a fair process applies: the largest party should have the first attempt at forming a government, and others should engage on programme rather than refuse to negotiate on principle alone.
At the same time, every party in eThekwini, MK included, should accept the same rules on city manager independence and budget support. Mangaung, a smaller coalition metro, should use the Free State government’s oversight capacity as a partner rather than as a threat. Cape Town, where the DA has governed with a majority, is not a realistic coalition problem at this stage, I would think, speculation aside. It is worth noting, however, that the lesson of Cape Town for the rest of the country is not that the DA is uniquely competent, but that continuity of political leadership and administrative leadership is what produces results. A coalition that cannot offer continuity should offer something just as valuable: predictability.
5. Where reasonable people disagree Some of this advice is contested, and it should be. Let’s look at a few general questions and responses.
Critics say that a party with 35% is not entitled to lead if a 55% bloc can agree on a programme. That is true, which is why the convention gives the largest party the first attempt, not the last word. ANC–DA cooperation Opponents on both sides regard it as a betrayal of their voters, and the parties are ideologically distant. Supporters say that in a hung metro a stable arrangement between the two largest parties may be the least bad option. Voters should be told before they vote, not after. For parties co-operating in the GNU, this squeamishness is unhelpful so close to the elections.
The South African Local Government Association has argued for measures to reduce the number of very small parties. Defenders of proportional representation argue that thresholds undermine representation. I need not take a side in the debate, but I can note that the tiered partner structure in Nelson Mandela Bay achieves some of the benefit without changing the electoral system.
Some will argue it protects bad mayors. It does make removal harder, which is why it needs a robust dispute process and a clear route for removal on grounds of misconduct, and why the Bill’s own time limits on motions are still a work in progress. Look at the conflict dynamics also – we often see mayors needing to be removed because the rest of the coalition is not working, or because some party(ies) benefit from instability and chaos. This framework is designed to exactly remove or lessen that foundation, which should need to less reason to remove mayors in the first place.
Politicians may reasonably resist outsiders. The counter-argument is that the alternative to a mediator is a court, a provincial intervention or a collapse. Modern mediation is a hard skill, not a compromise machine, and our politicians need to start using it as it is intended, to the benefit of all its constituents.
6. A closing word to the parties
What residents of Gqeberha, Tshwane and Johannesburg want is not, in the first instance, a better coalition. They want water that comes out of a tap and a city that does not change its mind every eight months. The parties that learn to share power on those terms will govern. The parties that continue to treat coalition as a transaction will be on the opposition benches within two years, and their leaders know it.
And so it should be: one uncomfortable fact that flows from our understanding that our politicians are underperforming in the area of coalitions is that there should be consequences. We should stop blaming the other side for being dishonest / corrupt / insert your favourite excuse, and voters should end this narrative driven by politicians. If a coalition fails it is a systems failure, not a party or individual failure. Fourteen days from an election is a short time, but it is enough to agree on a convenor, a one-page programme, a dispute ladder, a firewall around the administration and an undertaking on the budget. Nothing in this solutions requires a statute. It requires the discipline to put the city ahead of the caucus. That is not a naive ask. It is the oldest lesson in coalition politics, and it has been confirmed in every functioning democracy that has had to learn it.
Coalitions, even in the South African political conflict environment, are rocket fuel for progress and prosperity, it is exactly what we need to leave behind us the decades of single party arrogance and failures. There is a strange disconnect in some political circles that holds that coalitions can only work if everyone cooperates. This is outdated and doomed to the failure we see. Modern conflict competence takes our opposition s we find them, and then, from there, work towards the results we want. Our conflicts can and should be assets, not the fatal limitations our politicians try to convince us of. We do not need less conflict, we need better conflict.
Sources and further reading
My own work on the topic
Current reporting and policy (as of 30 October 2026)
Comparative and conceptual
Our blog at www.conflict-conversations.co.za has a rich variety of articles expanding on specific aspects of the article, and conflict in general.
Andre Vlok can be contacted at andre@conflict1.co.za for any further information.
(c) Andre Vlok
October 2026