Story audio is generated using AI
The ANC was the largest party in 117 of 175 administrations in hung councils since 2021, but it joined government in only 71 of those administrations and held the mayoralty in 55, according to the Mistra Coalitions Barometer 2026.
The report, which examines coalition governments in 80 municipalities, says the ANC was the only party to hold seats in every council in the dataset.
“In the scope and period covered by the Coalitions Barometer dataset — hung councils and coalition governments since 2021 — the ANC was ubiquitous,” the report says.
“It was the only party that held seats in every council, and it was the plurality party in 117 of the 175 administrations.”
But the report found that the ANC converted its position as the largest party into government less often than the DA and the IFP did when they held a plurality.
“A second striking fact is that, despite (or perhaps because of) its size, the ANC converted its position into office at a lower rate than expected,” the report says.
“In the councils where it held a plurality, it joined the government less often than the DA or the IFP did from the same position.”
The report sets out the conversion from being the largest party to joining government and then securing the mayoralty.
Of the 117 administrations where the ANC was the largest party, it joined government in 71, representing 61%, and held the mayoralty in 55, representing 47%.
The DA was the largest party in 38 administrations. It joined government in 24, or 63%, and held the mayoralty in all 24.
The IFP was the largest party in 26 administrations. It joined government in 19, or 73%, and held the mayoralty in 18, or 69%.
“The ANC lost ground at both steps, converting a plurality into membership, and membership into the mayoralty, less often than the other two,” the report says.
The ANC also did not always take the mayoralty when it was the largest member of a coalition.
“When it did join the government, even though it was almost always as the largest member, the mayoralty went to a smaller partner in more than a fifth of cases (16 of 71),” the report says.
“The combined result is that the ANC led the government in fewer than half of the councils where it held a plurality (55 of 117), compared to roughly two-thirds for the DA and the IFP.”
One clear trend was that the ANC’s presence in government grew: it governed in 45.6% of the coalition-governed councils in 2021 and in 68.4% in 2025
The report records 90 administrations in which the ANC was in government.
Its participation in government increased during the period covered by the dataset.
“One clear trend was that the ANC’s presence in government grew: it governed in 45.6% of the coalition-governed councils in 2021 and in 68.4% in 2025, weighting each council by the time the ANC spent in government there,” the report says.
The ANC’s most common coalition partner, measured by days spent governing together, was the PA, with 18,616 shared days.
The EFF was next, with 12,638 days, followed by the African Independent Congress (AIC), with 10,296 days.
The ANC also governed with smaller and local parties.
“If ‘local parties’ are defined as parties that sat in only one municipality in the country, the ANC governed with them for a total of 18,360 days across 26 administrations, which made them collectively the ANC’s second-largest coalition partner after the PA,” the report says.
“In general, the ANC seems to have recruited pragmatically, according to local conditions, with a clear preference for smaller parties.”
The report says the ANC never co-governed with the DA in a local coalition.
“It never co-governed with the DA in any local coalition, only in the national GNU and the KwaZulu-Natal government of provincial unity.”
The report also highlights the role played by smaller parties in coalition governments.
“There were 103 different parties in the Coalitions Barometer dataset that held at least one seat somewhere in the country; 64 of these parties were a coalition member at some point; and 14 anchored a government as its largest member.
“South African local politics is heavily shaped by the country’s many local parties, residents’ associations, one-councillor formations and so on,” it says.
“Often, these smaller parties are the decisive actors in determining which of the larger parties’ coalitions are able to govern and survive.”
The PA, for example, held seats in 75 administrations but was never the largest member of a government.
Despite this, it held the mayoralty nine times.
“There is no other political party with a national profile that has managed to convert junior membership into a mayoralty as often as the PA,” the report says.
The report says the PA’s nine mayoralties were all in ANC-anchored governments.
“Most of the PA’s success in obtaining leadership positions came from its close relationship with the ANC: its nine mayoralties were all in ANC-anchored governments.”
Mistra also found that coalitions excluding the largest party were less stable.
“Coalitions that governed without the plurality party collapsed at about four times the rate of those that included it.”
Of the 55 administrations that governed without the plurality party, 37 collapsed, compared with 21 of the 120 administrations that included the plurality party.
The report says the finding applied regardless of which party was the plurality party.
“The ANC is the most frequent plurality party in the dataset, so it is reasonable to wonder whether the real risk factor is opposition from the ANC.
“However, the identity of the plurality party turns out to be irrelevant: coalitions formed against an ANC plurality were no more and no less fragile than the anti-plurality coalitions that excluded the DA or the IFP.
“The key factor is the distribution of seats within the council, not the identity of the plurality party itself.”
The report found that when the ANC did lead a coalition government, those administrations tended to last longer than the average administration in the dataset.
“Coalition administrations led by the ANC had a median lifespan of 1,056 days compared with 651 for the dataset as a whole (counting every kind of ending).”
