The 2026 local elections in short #
Show the background on South Africa’s local elections
This section is for a reader who does not know South Africa’s local elections. A reader who does can go straight to What Ukweli did.
The vote. On Wednesday 4 November 2026, about 14.9 million South Africans voted, a turnout of 52%, the IEC’s stated turnout among 29.2 million registered voters. The voter figures are from the Electoral Commission’s chairperson, Mosotho Moepya, as reported on 8 November 2026: the 14.9 million by SAnews and The Witness, and the turnout and the registered voters by SAnews. Voters chose representatives for 257 municipalities: eight metropolitan municipalities (metros), 44 district municipalities and 205 local municipalities. Together these hold 10,526 council seats (counts from the Electoral Commission’s statement of 16 September 2026). The Electoral Commission of South Africa (IEC), the independent body that runs elections, declared the elections free and fair on Sunday 8 November 2026 (statement by its chairperson, Mosotho Moepya, as reported by SAnews).
Where local government sits. South Africa’s Constitution (section 40) gives the country national, provincial and local spheres of government. Local government is the municipality: the council that runs a town, a city or a district.
What a metro is. A metro is one of the country’s eight largest cities, governed by its own council. Three are in Gauteng: Johannesburg, Tshwane (the city around Pretoria) and Ekurhuleni (east of Johannesburg). Cape Town is in the Western Cape and eThekwini (Durban) is in KwaZulu-Natal. Nelson Mandela Bay (Gqeberha) and Buffalo City (East London) are in the Eastern Cape, and Mangaung (Bloemfontein) is in the Free State (list of metros and provinces; government list of the eight metros).
How people vote. In a metro a voter gets two ballot papers: one to choose a ward councillor and one to choose a party for the metro council. Elsewhere a voter gets three: a ward councillor, a party for the local council and a party for the district council (SAnews voter guide, 26 October 2026). In a metro or local council, half of the seats, rounded up, are ward seats, each won by the candidate with the most votes in that ward, and independents can stand for them. The rest are party seats, shared out in proportion to the votes each party wins. District councils are built differently (Municipal Structures Act, section 22, and Open Cities Lab).
How many parties. 732 parties were registered and 546 qualified to contest (IEC figures, as stated in Ukweli’s Folio 07 and Folio 22). 136,790 candidates were certified on 16 September for the 10,526 seats (the Commission’s statement linked above).
What a council does. The Constitution lists the matters a municipality may run. Ukweli’s table of them has 38 matters, 15 from Schedule 4 and 23 from Schedule 5 (Folio 08 and Folio 22). Matters not on that list, such as policing, are the job of another sphere.
What a manifesto is. A manifesto is a party’s published list of promises, so voters can read them and decide. Ukweli’s unit is one promise: a sentence or part of one that commits the party to something.
Why 2021. The previous local elections were held on 1 November 2021 (SAnews). Ukweli has the 2021 manifestos of the ANC, the DA and the EFF, which is why the 2021 comparison is limited to those three parties.
Why Kenya. After Kenya’s 2022 election, a Commonwealth observer group published a report. One sentence in it says that “a number of Kenyan voters” held the view that “certain manifesto promises were superficial, and without a clearly explained path to implement their promises”. The sentence is on p.5, Chapter 2, “Economics and political apathy”, of Kenya General Election, 9 August 2022: Report of the Commonwealth Observer Group. The sentence reports what voters said. The report cites no survey for it. Ukweli does not test Kenya’s manifestos. It asks the same plain question of South Africa’s: what does the wording of each promise say?
What Ukweli did #
Ukweli has analysed 15 manifestos for the 2026 local elections. It found 2,686 promises and quoted each in the party’s own words, with its page. An AI model does the reading, and a finding is published only when repeated readings agree. Where a sentence is too short or vague to test, the card says “Could not be assessed by Ukweli”. That label is about the reading, not about the party.
This folio does a separate job on the same promises. It counts, for each promise, whether its own wording contains each of the seven things set out below. The 2021 side uses the 498 promises Ukweli read in the ANC, DA and EFF manifestos of 2021, read the same way.
The 15 manifestos
| Party | Promises | Ukweli page | Party’s own manifesto |
|---|
Promise counts are the numbers on Ukweli’s pages, re-derived from its ledger on 8 Nov 2026 (total 2,686). They are the sizes of the documents, not a measure of anything else. A manifesto is in Ukweli’s set when Ukweli can obtain it from the party’s own publication; that is a rule about documents, never a judgement about a party. Party names: ACDP is the African Christian Democratic Party; AMM the Afrika Mayibuye Movement; ANC the African National Congress; DA the Democratic Alliance; EFF the Economic Freedom Fighters; FF Plus the Freedom Front Plus; ICM the Independent Citizens Movement; IFP the Inkatha Freedom Party; NCC the National Coloured Congress; SACP the South African Communist Party; UDM the United Democratic Movement. ActionSA and GOOD are the full names. BOSA and Al Jama-ah are written as each party writes them.
The SACP did not publish a standalone manifesto document that Ukweli found; its manifesto text is printed in its newsletter, Umsebenzi Online (August 2026 issue), on pp. 7 to 11, and that is the text Ukweli read. Party logos belong to the parties and are shown for identification only. Ukweli is not affiliated with any party.
What this folio does not do or show #
- It does not score, grade, rank or endorse any party, does not say who should govern, and names no party on any finding.
- It does not say a promise is good, bad, honest, wise or achievable, or whether a party will keep it.
- It makes no statement about what any party knows or understands, including what a council may do. A promise outside a council’s powers is not treated as false, and a “no” means the wording does not say it, not that the party has not.
- It does not say how any party did at the polls, or whether voters read the manifestos.
- It does not test whether South African promises are “superficial”. The Commonwealth sentence is about Kenyan voters’ view of Kenyan promises. The folio counts whether the wording contains things a reader could check later: a number, a date, an office, a place, a source of money, a report.
- It draws no trend. 2021 and 2026 are two points in time for three parties, and the party mix differs between the groupings.
Folio 23 counts how many promises in five parties’ manifestos match the conditions ActionSA and the DA have named for coalition talks in Johannesburg, Tshwane and Ekurhuleni.
The seven measures #
Each is a question put to the wording of one promise. The answer is yes or no, and “no” means the wording does not say it. Measures 1 to 6 are shown in ten rows in Graphs 1 to 3; measure 7 is Graph 4.
| Measure | Row | Counts when | Does not count |
|---|---|---|---|
| 1. Is it checkable? | Contains a number | The wording states a number. This is what “quantifiable” means in this folio. | |
| Sets a numeric target | The wording gives a number that is the size or level to reach, such as “build 500 houses” or “cut waiting times by 20%”. | A deadline, a year or a section number. | |
| Fixes a deadline | The wording sets a date or a time limit. | ||
| 2. Does it start from a known position? | States where things stand now | The wording states where things stand now, such as “currently 40% of homes have…”. | A year range. |
| 3. Who answers for it? | Names an accountable office | The wording names an accountable office. | |
| 4. Where? | Names a place | The wording names a specific place: a city, town, township, suburb, informal settlement or named facility. | “Each ward” or “every township”. |
| 5. How is it paid for and reported? | Says where the money comes from | The wording names a source: a named grant, borrowing, own revenue, ring-fenced funds, reallocating the budget, “funded by…”. | An amount alone, a wage, a benefit or a price. |
| Commits to report or review | The wording commits to report on, publish results of, audit, review, monitor or track delivery of this promise or the programme it belongs to. | ||
| 6. Who else is involved? | Names another body | The wording names a specific organisation or sphere other than the municipality, such as Eskom, a national or provincial department, the police service or a named programme. Traditional councils, authorities and leadership structures count as another body. | “The private sector” or “communities”; a bare “respect traditional leaders”. |
| Asks or calls on another body | The party will ask, lobby, petition, demand, negotiate with or work with another body to get it done. | ||
| 7. Powers and limits (Graph 4) | Which sphere the matter sits in | The promise falls within what the Constitution lets a council do, or sits with another sphere. It is counted against the Constitution’s lists (Folio 08) and says nothing about what anyone knows. |
The grouping of ten rows under six questions, plus powers and limits as the seventh, is Ukweli’s way of presenting the measures. The first three rows (number, deadline, office) are counted by fixed rules with no AI model in them; the rest are read.
How to read the graphs #
- Each row of Graphs 1 to 3 is one measure. A dot is a percentage of that year’s promises. Where a row has two dots, the blue dot is 2021 and the orange dot is 2026.
- Position is a count of wording, not a quality grade. Further right is not better, and the rows are in a fixed order, not sorted by value.
- 2021 is the ANC, DA and EFF only. Their 2026 manifestos are shown beside them so that like is compared with like. The fifteen-party graph is separate and has no 2021 partner.
- The numbers appear when you hover over a row, or tap it, or move to it with the keyboard. Every graph also has a table view under it with the same counts.
The graphs #
Graph 1. The ANC, DA and EFF together, 2021 against 2026 (498 and 875 promises)
The ten measures for the ANC, DA and EFF together. Blue is 2021 and orange is 2026.
Party mix: one of the three 2021 manifestos has a list of councillor conduct rules, and all 40 of the 2021 promises that name an office are in it. That lifts the three-party 2021 figure for naming an office. Graph 2 leaves that manifesto out.
Table view: the counts behind Graph 1
| Measure | 2021 | 2026 |
|---|---|---|
| Contains a number | 9.2% (46) | 7.3% (64) |
| Sets a numeric target | 4.0% (20) | 3.4% (30) |
| Fixes a deadline | 3.0% (15) | 2.5% (22) |
| States where things stand now | 0.8% (4) | 0.3% (3) |
| Names an accountable office | 8.0% (40) | 2.3% (20) |
| Names a place | 0.2% (1) | 2.1% (18) |
| Says where the money comes from | 1.2% (6) | 3.9% (34) |
| Commits to report or review | 2.4% (12) | 6.7% (59) |
| Names another body | 5.2% (26) | 8.8% (77) |
| Asks or calls on another body | 4.2% (21) | 9.0% (79) |
| Promises read | 498 | 875 |
Graph 2. The ANC and DA together, 2021 against 2026 (233 and 581 promises)
The same ten measures for two of the three parties, so that the manifesto with the conduct list in Graph 1 is out of both years. Blue is 2021 and orange is 2026.
The counts here are small for several rows (see the table view), so a move of a few promises shows as a large move in the dots.
Table view: the counts behind Graph 2
| Measure | 2021 | 2026 |
|---|---|---|
| Contains a number | 4.7% (11) | 7.7% (45) |
| Sets a numeric target | 0.9% (2) | 2.9% (17) |
| Fixes a deadline | 0.9% (2) | 3.1% (18) |
| States where things stand now | 0.4% (1) | 0.3% (2) |
| Names an accountable office | 0.0% (0) | 3.1% (18) |
| Names a place | 0.0% (0) | 0.9% (5) |
| Says where the money comes from | 1.7% (4) | 5.2% (30) |
| Commits to report or review | 1.7% (4) | 9.3% (54) |
| Names another body | 9.4% (22) | 11.0% (64) |
| Asks or calls on another body | 7.7% (18) | 10.5% (61) |
| Promises read | 233 | 581 |
Graph 3. All fifteen 2026 manifestos together (2,686 promises)
The same ten measures for every promise Ukweli read in 2026.
The pool counts promises, so a larger manifesto weighs more than a smaller one. The fifteen manifestos range from 41 to 462 promises.
Table view: the counts behind Graph 3
| Measure | 2026, all fifteen |
|---|---|
| Contains a number | 4.2% (113) |
| Sets a numeric target | 1.5% (41) |
| Fixes a deadline | 1.5% (40) |
| States where things stand now | 0.2% (6) |
| Names an accountable office | 3.6% (97) |
| Names a place | 1.5% (40) |
| Says where the money comes from | 3.3% (88) |
| Commits to report or review | 5.6% (151) |
| Names another body | 6.4% (173) |
| Asks or calls on another body | 7.5% (201) |
| Promises read | 2,686 |
Graph 4. Powers and limits, 2026, all fifteen manifestos (2,686 promises)
Each bar is the share of all 2,686 promises that fall in one group. The groups are in a fixed order, the order Ukweli uses in Folios 20 to 22, not in order of size. A longer bar is not better.
2021 is not shown: the 2021 promises have not been tested against the Constitution’s lists, so there is nothing to set beside 2026. The last bar, drawn in grey, is not a finding about a promise: it holds the 263 promises that mix different powers and the 377 that Ukweli could not assess (the same split as Folios 22 and 23).
Table view: the counts behind Graph 4, and what each group means
| Group | What it means | Share of all 2,686 | Share of the 2,046 assessed |
|---|---|---|---|
| A council’s own services | The matter is on one of the municipal lists. | 15.2% (407) | 19.9% |
| How the council runs itself | Staff, budgets, procurement, public participation and other administration. | 30.2% (810) | 39.6% |
| Shared with other spheres of government | The power is split between spheres and neither leads. | 0.04% (1) | 0.05% |
| Led by another sphere, council has a role | Another sphere of government leads, and the law gives the council a defined role. | 7.5% (202) | 9.9% |
| Outside a council’s powers | The matter is not on the municipal lists. It is not about how the council runs itself, is not shared between spheres, and is not led by another sphere with a defined role for the council. | 13.8% (370) | 18.1% |
| The party’s own conduct | What a party or its councillors will do, and general pledges that name no matter on a council’s lists. Ukweli judged it not a function of a council. | 9.5% (256) | 12.5% |
| Could not be assessed, or mixes different powers | The sentence is too short or vague to test, or Ukweli’s checks did not settle it, or it promises several things under different powers. | 23.8% (640) | not applicable |
| All promises | 2,686 | 2,046 |
What the graphs show together #
- Almost no promise says where things stand now. Six of the 2,686 promises in 2026 do (0.2%). In the 2021 manifestos of the three parties it is 0.8% of 498.
- Two graphs, two directions. For the ANC, DA and EFF together, containing a number, fixing a deadline, naming an office and setting a numeric target are each lower in 2026 than in 2021 (Graph 1). For the ANC and DA alone, the same four are each higher (Graph 2). The folio draws no direction from either graph.
- Naming another body, and calling on one, are higher in 2026 than in 2021 in both graphs. So are naming a place, saying where the money comes from, and committing to report or review. Several of these rest on small counts in 2021 (see the table views).
- Powers and limits. 15.2% of all 2026 promises fall within a council’s own services and 30.2% concern how the council runs itself. 13.8% fall outside a council’s powers and 7.5% are led by another sphere with a defined role for the council. 23.8% could not be assessed or mix different powers.
Our view #
Ukweli’s view, kept apart from the counts above. Not a finding.
Our answer: for most promises in the 2026 manifestos, the wording does not give a reader a path they could check later.
The question behind this folio came from Kenya. Some Kenyan voters said that certain manifesto promises were “superficial, and without a clearly explained path to implement their promises”. Ukweli cannot say how any voter experienced a manifesto. It can ask a narrower question of South Africa’s 2026 manifestos: does the wording of the promise give a reader a path they could check later?
Of the 2,686 promises, 2,274 (84.7%) say none of six things that make up a path: a deadline, an accountable office, a numeric target, a source of money, a commitment to report or review, or a named place. 370 say one of them. 42 (1.6%) say two or more, and none says more than three.
We do not call the promises superficial. That word is a judgement about their worth, and this folio makes none. We say something plainer: the path is mostly not written in the promise. A manifesto is short, and a party may hold the detail in a budget or a plan. A voter who reads the manifesto has only the sentence.
Against 2021 the findings do not move one way. For the ANC, DA and EFF together, 82.5% (411 of 498) of the 2021 promises said none of the six, and 81.1% (710 of 875) of the 2026 promises say none. For the ANC and DA alone, the share moved from 94.8% to 78.5%, but in 2021 only 12 of the 233 promises said any of the six, so a small count carries that move. Graphs 1 and 2 show the same split, and we draw no trend.
What the wording does carry is more often a pointer to someone else than a date. Naming another body (6.4%) and asking or calling on one (7.5%) are more common than fixing a deadline (1.5%) or naming a place (1.5%). Where things stand now is stated in 0.2% of promises.
Powers add to this. Only 15.2% of the promises fall within a council’s own services, and 30.2% concern how the council runs itself. Another 21.3% are outside a council’s powers or led by another sphere. A promise outside a council’s powers is not a false promise. It is one where the voter’s question is who else must act.
A voter does not need these counts to use this view. Five plain questions put to any candidate turn a sentence into a path:
- By when?
- Who answers for it?
- Where does the money come from?
- How will we know it happened?
- Where do we start from?
This is our view, not a finding. It rests on the same counts and carries the same limits: the readers are one AI model, a “no” means the wording does not say it, and nothing here is a statement about what any party knows or intends. The choice of the six is Ukweli’s; the method is in How we counted, and what limits this.
How we counted, and what limits this #
- The unit. One promise: a sentence or part of one that commits the party to something. 2,686 in the 2026 manifestos of fifteen parties, 498 in the 2021 manifestos of the ANC, DA and EFF.
- Number, deadline, named office. These three are counted by fixed rules with no AI model in them, so anyone can re-run them. Checked against 527 promises read by hand, the number rule was right on 55 of the 58 it flagged. It found 14 of the 15 real cases. The deadline rule was right on 29 of 30 and found 8 of 8. The named-office rule was right on 30 of 31 and found 21 of 23. Those checks rest on few real cases, so treat them as rough.
- The other seven labels. Place, funding mechanism, numeric target, report or review, starting position, another body, and asking or calling on another body were each read by two AI readers working alone from the same written guide. A further reading settled each disagreement against that guide. The readers are the same AI model, so their agreement shows consistency, not independence.
- Checks on the readers. A third reading of a random sample of 200 of the 3,184 promises, again from the guide alone, agreed with the final label on 1,389 of 1,400 judgements (the 11 that differ are 0.8%). A person at Ukweli then read 60 of those 200 promises with the labels beside the wording and changed none of the labels. The other 140 were not read by a person.
- Rulings written into the labels. Rand amounts, wages and benefits are not counted as where the money comes from. Traditional councils, authorities and leadership structures count as another body; a bare “respect traditional leaders” does not. After the third reading showed this rule had not been applied evenly, nine promises were relabelled under it and the table was re-run.
- What the readers can miss. The readers measure what they were shown. A promise that no reader marked is not in the count, and the method cannot say how many such promises exist. A “no” means the wording does not say it.
- The six-part path count in Our view. For each promise, Ukweli counted how many of six rows its wording says: deadline, accountable office, numeric target, source of money, report or review, named place. The other four rows (number, starting position, another body, asking another body) are not in it. The choice of the six is Ukweli’s, made because they say when, by whom, how much, how paid for, how reported or where. The count uses the same labels as the table, so it adds up to the same row totals (457 hits across 412 promises).
- Powers and limits. Each promise was tested against the Constitution’s lists of municipal matters (Folio 08) in a separate pass from the quotation of the promise, so a verdict cannot change the quoted words. Graph 4 uses all 2,686 promises as its base and shows the ones not assessed; Folio 21 used the assessed promises only, and the table view gives both.
- Party mix. The pooled groups count promises, so a larger manifesto weighs more. In 2021, all 40 promises that name an office sit in one manifesto’s list of councillor conduct rules, which is why Graph 2 is shown. Counts of 1 to 4 in a row, such as the starting position, can move the dots a long way.
- Two points in time. 2021 and 2026 are two points for three parties. Other parties’ 2021 manifestos are not in Ukweli’s set.
- Not tested here. Whether any party will keep a promise; whether a promise is wise or affordable; what any party knows; how the manifestos did at the polls; whether voters read them; Kenya’s manifestos.
- Election background. The vote figures are the IEC chairperson’s as reported by SAnews and The Witness on 8 November 2026; the IEC’s own results dashboard was not read for them. The count of parties that qualified to contest (546) is as stated in Folios 07 and 22. Half of seats being ward seats is from Open Cities Lab; the Municipal Structures Act was not opened for this folio. The length of a council’s term is left out because the sources opened for this folio gave different figures.
Sources #
- Electoral Commission of South Africa, “Electoral Commission Certifies 136 790 Candidates to Contest 2026 Local Government Elections”, media statement, gov.za, 16 September 2026.
- SAnews, “IEC declares LGE 2026 free and fair”, 8 November 2026.
- The Witness, “2026 local election results: IEC confirms 14.9 million voters cast ballots”, 8 November 2026.
- Local Government: Municipal Structures Act 117 of 1998, section 22 (composition of councils), as published in 1998; amendments not checked.
- SAnews, “Ready to Vote? Know what to expect on Election Day”, 26 October 2026.
- Open Cities Lab, “How Your Vote Counts”.
- SAnews, “SA to head to the polls on 1 November 2021”.
- Constitution of the Republic of South Africa, 1996, section 40(1).
- Wikipedia, Metropolitan municipality (South Africa), list of the eight metros and their provinces; South African Government, Local government, list of the eight metros.
- Commonwealth Observer Group, Kenya General Election, 9 August 2022: Report of the Commonwealth Observer Group, p.5, Chapter 2; Commonwealth news page for the report.
- Ukweli, Folio 08, Folio 07 and Folio 22, for the table of council powers, the 2026 campaign and the party counts.
- The fifteen 2026 manifestos and the ANC, DA and EFF 2021 manifestos, as read by Ukweli; each promise is on the elections page in the party’s own words.
The standing block #
Read every promise in the party’s own words →
Read the register and download the data →
Integrity Watch is a verified register, neutral by design. Every entry is confirmed against a reputable source, and if it cannot be confirmed, it is not listed. Ukweli is a content integrity platform built by The Orkestra.
Get the next folio
The next Ukweli folio, by email. Nothing else, and you can leave at any time.
The register itself stays free and public, and the full dataset is yours to download whether you subscribe or not.
Know the truth before you publish.
Ukweli checks documents for fabricated and unverifiable citations, sources, quotes and claims, before they leave your hands. Integrify a document · Get the next folio