ZAMBIA’S election was not supposed to end like this. For much of the campaign, President Hakainde Hichilema looked well-placed to secure a second term. Instead, the country has emerged from the 13 August vote with its democratic credibility in serious doubt.
The most troubling evidence that Zambia’s general election may have been flawed is not an allegation made by the opposition. Rather, it is contained in the election results themselves.
There is much more to say about the broader environment in which the election took place, including the restrictions on opposition party registration and campaigning, the arrest and disappearance of opposition figures, concerns over freedom of expression and the use of new cyber laws. But it is precisely because the results themselves raise such serious questions that we focus on them here.
In a number of constituencies, considerably more votes appear to have been cast in the presidential election than in the parliamentary contest taking place at exactly the same time. In the most extreme cases, the difference is extraordinary.
This matters because such discrepancies are a well-established warning sign of electoral manipulation. Our preliminary analysis suggests that hundreds of thousands of votes for President Hakainde Hichilima may need to be explained.
If that conclusion is sustained when the complete results become available, Zambia will face a difficult reality: an election that was expected to consolidate the country’s democratic credentials may face calls for a rerun.
Hichilema entered the election in a strong position. He could point to significant achievements during his first term, most notably restructuring the enormous debt burden he inherited after Zambia became Africa’s first pandemic-era sovereign defaulter. The economy had stabilised, even if many Zambians continued to struggle with the cost of living.
He also benefited from a fragmented opposition following the death of former president Edgar Lungu and the implosion of the Patriotic Front (PF), which governed Zambia until 2021. That changed with the emergence of Brian Mundubile as the main opposition candidate. A former PF figure, Mundubile was adopted as the presidential candidate of the National Reconciliation Party for Unity and Prosperity (NRPUP) before joining forces with Makebi Zulu and the Pamodzi Alliance to head the broader Tonse-Pamodzi Alliance. This brought together important fragments of the former PF and other opposition forces that had previously been divided.
What had initially looked like a relatively straightforward contest became harder to predict. Despite the growing momentum behind the opposition, election day itself was largely uncontroversial. Southern African Development Community (SADC) observers found the atmosphere calm and peaceful at 99% of the polling stations they visited, while the European Union found voting generally peaceful and orderly.
As so often happens, the situation began to fall apart during the counting process. On 14 August, the Electoral Commission of Zambia (ECZ) suspended the counting and announcement of results nationwide with immediate effect. The Commission cited violence against election officials and security concerns, and the suspension was initially expected to last for up to 24 hours. It was subsequently lifted later the same day after an outcry by the opposition and civil society, but by then confidence in the electoral process had already been shaken.
Although the security concerns were real, a nationwide suspension inevitably generated suspicion because problems were concentrated in particular locations. African Union-Common Market for Eastern and Southern Africa observers concluded that the episode contributed to “increased mistrust” and political tension.
At first, the results that followed offered some reassurance. Transparency International Zambia’s Integrity Index initially found a very close correspondence between presidential results announced by the ECZ and those independently recorded by monitors. That made claims that results had been changed between being read out initially at the polling station and officially by the electoral commission difficult to sustain.
But as more results became available, a different problem emerged. Zambia was holding several elections simultaneously on 13 August. Voters were choosing a president, members of the National Assembly and local councillors during the same visit to the polling station. As parliamentary results became available, it therefore became possible to compare the number of votes cast in the presidential and legislative elections in each constituency.
We compiled the presidential list of results and the first thing that stood out was a high number – at least 27 constituencies – where the ruling United Party for National Development (UNPD) received a super-majority of the vote in the presidential election, with a voter turnout of over 80%. This stands in stark contrast to the national registered-voter-weighted turnout of 57%.
The constituencies with over 80% turnout at presidential level were concentrated in the UPND’s strongholds – North-Western Province (9), Western Province (8), Southern Province (6) – with three constituencies in rural Copperbelt and one in Central Province. Notably, there were none in the opposition’s strongholds or the provinces that they won – Eastern Province, Luapula, Muchinga or Northern Province.
Having seen these extremely high rates of turnout in the presidential election, we used publicly available data to calculate the turnout at parliamentary level which led to a stark picture. In the 16 constituencies we looked at, the variance was almost universally high – with a difference of over 46% in turnout between the presidential and parliamentary races in Solwezi North.
A small difference between these tallies is perfectly understandable as a voter may choose a president but decide that none of the parliamentary candidates deserves their support. But because voters are typically handed all ballot papers, and complete and deposit them before leaving the polling station, large discrepancies between the votes cast in different races require explanation. The difference in the numbers of votes cast in just these 16 seats is over 170, 000 votes – or 15% of the total difference between the two leading presidential candidates.
In studying African elections over the past two decades – including for (Nic Cheeseman and Brian Klaas’s book) How to Rig an Election – we have generally treated a gap of more than five percentage points between simultaneous presidential and parliamentary turnout as a warning sign that deserves investigation. Once that difference exceeds 10%, alternative explanations become progressively harder to sustain.
In Zambia, some of the gaps appear to be far larger than that. In several constituencies, they exceed 10%. In the most extreme cases, the difference appears to be over 40%.
The constituency-level evidence is important because the discrepancies are not simply concentrated in parliamentary races that nobody seriously contested. Some appear where there was a meaningful contest for the National Assembly, making it difficult to explain why so many voters would participate in the presidential election but not the parliamentary one.
Taken together, these cases are consistent with what we called in the book “amateur hour” election rigging: the kind of hurried manipulation that occurs when ruling-party figures panic, as in Kenya in 2007. Unlike more carefully planned rigging, it is poorly coordinated and highly visible, leaving obvious inconsistencies in the results that can be especially damaging to an election’s credibility.
The turnout figures alone cannot tell us who was responsible, when manipulation occurred, or exactly how it was carried out. Reports are emerging of monitors being forced out of polling stations, of opposition candidates alleging that votes were swapped, and that domestic election observers were not even given access to all of the necessary results forms.
According to the final results announced by the ECZ, Hichilema received about 61% of the vote against about 38% for Mundubile – a difference of roughly 1.1 million votes. Our unverified estimate based on publicly circulated figures is that the suspicious difference between presidential and parliamentary turnout may account for somewhere between 350,000 and 550,000 votes.
This suggests that this form of vote inflation did not determine the outcome of the polls, but it is important to note that it is only one way of identifying manipulation. It cannot detect other ways in which results might have been altered. The wider context also matters. The European Union concluded that the contest took place on an “uneven playing field”, with restrictions on most presidential candidates’ ability to campaign, significant incumbency advantages and heavily biased state-owned media.
SADC was more positive overall but nevertheless recommended that campaign rules apply equally to the president and his rivals, that official duties be clearly separated from campaigning, and that the state broadcaster provide equitable coverage.
This means that Zambia may ultimately need to rerun the presidential election. But a fresh vote is not a simple solution. Recent experience in Malawi and Kenya shows that what happens between the first election and the second can be decisive.
Malawi shows what is possible when leadership and institutions change. Before the rerun, Jane Ansah was replaced as head of the Malawi Electoral Commission by High Court judge Chifundo Kachale, helping to restore confidence in the body overseeing the vote. The resulting June 2020 election was widely regarded as credible and led to a transfer of power to opposition candidate Lazarus Chakwera, who had joined forces with former rival Saulos Chilima after they competed separately in 2019.
The lesson for Zambia is therefore not simply that flawed elections can be rerun, but that reruns only restore confidence if the conditions that undermined the original election also change. This is particularly important because SADC observers have already recommended reforming the appointment of ECZ commissioners so that the process commands confidence across the political spectrum.
None of this proves that Hichilema would have lost a clean election. But if these discrepancies are confirmed, Zambia may need a fresh vote – and reforms to ensure that it does not reproduce the problems of the first.
***Cheeseman is Professor of Democracy at the University of Birmingham, founder of www.democracyinafrica.org and author of ‘How to Rig an Election’. Dr Beardsworth is a political science lecturer at the University of the Witwatersrand. The article was first published by theafricareport.com on 18 August 2026.
****************************************************************************************************
Rejoinder: Zambia’s rlection was flawed and may need to be rerun
By Mcqueen Zaza
In their article of 18 August, Prof. Nic Cheeseman and Dr Nicole Beardsworth suggested that Zambia’s 13 August election was flawed and may need to be rerun. This is a serious charge, and deserves a serious answer. Fortunately, most of that answer is already contained in their own article.
Consider what the authors themselves concede. Election day was calm and orderly, with observers from the Southern African Development Community (SADC) reporting a peaceful atmosphere at 99% of the polling stations they visited. Independent verification by Transparency International Zambia (TI-Z) found what the authors call a “very close correspondence” between the results announced by the electoral commission and those recorded by TI-Z’s own monitors, a match of 98.9%.
And by their own arithmetic, their “unverified estimate” of questionable votes, even taken entirely at face value and subtracted wholesale from the president’s column, would not change the outcome. President Hakainde Hichilema would still have won in the first round, comfortably above the 50% threshold. They write, in their own words, that “this form of vote inflation did not determine the outcome of the polls”. The article then runs under a headline calling for a rerun. The headline is not supported by the paragraphs beneath it.
What remains is a single statistical observation. Namely, that in some constituencies, more votes were cast for president than for members of parliament. The authors treat this as a warning sign of manipulation.
There is a far simpler explanation, and it is one Dr Beardsworth knows intimately, because she is the first named author of the research that documented it.
In March 2026, the Zambia Election Research Network published its first survey findings through the University of Cape Town’s Institute for Democracy, Citizenship and Public Policy in Africa, analysed and written up by Dr Beardsworth and three colleagues. That survey found that only 33% of Zambians planned to vote for their sitting MP, that 24% wanted the same party to field a different candidate, and that a mere 15% intended to vote for their incumbent councillor.
The same survey put the president’s approval at 79%, and concluded that he was heading for a comfortable re-election, while warning of what the authors called “major electoral vulnerability at sub-national levels”.
In plain terms, Zambians told these researchers, months before the vote, that they would turn out for their president while punishing their parliamentarians. On 13 August, that is exactly what they did. The gap the authors now present as evidence of manipulation is the gap their own survey predicted.
If the authors are confident in their datasets, they should publish it in full so that it can be checked, line by line, against the Gen 20 forms.
Political scientists have a name for this behaviour. It is called roll-off. It is common in every democracy that holds simultaneous elections.
The geography of the gaps confirms the point rather than undermining it. Dissatisfaction with MPs was most pronounced in north-western and western provinces, precisely where the president’s approval and net favourability were at some of the highest, and where the party’s adoption process generated the most visible anger. A voter in Solwezi who queued patiently to reelect her president and then declined to reward an MP she considered a failure was not evidence of fraud. She was exercising discernment. That the authors’ framework interprets her judgement as a crime statistic says more about the framework than about Zambia.
Results in Zambian elections are recorded on Gen 20 forms, completed and posted publicly at each polling station in the presence of party agents and observers before anything travels anywhere. A parallel projection, built purely from those station-level forms on the morning of 14 August, put the race at 60.6% to 38.7%. The final declared result was almost identical.
If hundreds of thousands of votes had been injected somewhere between the polling station and the national declaration, the final tally would have diverged sharply from that early station-level picture. It did not move. And five days on, no one has produced a single polling station anywhere in Zambia where the publicly posted result differs from the figure the commission announced.
Against this stands an analysis the authors describe, repeatedly and candidly, as “unverified”, based on “publicly circulated figures”, covering 16 constituencies selected by the authors themselves, and compiled, while the parliamentary count was still incomplete.
Indeed, as of this week, the commission’s own results portal still showed constituencies awaiting parliamentary results altogether. Comparing a finished presidential count against an unfinished parliamentary one will manufacture turnout gaps by simple arithmetic.
If the authors are confident in their dataset, they should publish it in full so that it can be checked, line by line, against the Gen 20 forms. We would welcome that exercise without reservation.
Zambia’s Constitution provides a clear route for any candidate who disputes a result. They can file a petition before the constitutional court, within defined timelines, tested against evidence. That process is open, and it will be respected. What the constitution does not provide for is the annulment of eight million people’s votes by op-ed.
The electoral commission budgeted 1.7 billion Kwacha (about US$91m) for this election, an exercise involving almost nine million registered voters, more than 13,000 polling stations and close to 90,000 staff. A rerun would conservatively consume the price of several hundred fully equipped ambulances in a country still building out rural emergency care. Those proposing one on the strength of an unverified spreadsheet should be asked to say so to the communities that would pay for it.
A last observation. When elections are stolen, people do not pour into the streets to celebrate. They protest, or they stay home behind locked doors.
What Zambia saw on declaration day were joyous scenes at the University of Zambia (UNZA) and Community House, and then something more telling still. The next morning, Lusaka was simply normal. Markets opened, people went to work, life carried on, because Zambians were confident their voices had been heard.
Compare that with scenes in Harare (Zimbabwe), Dar es Salaam (Tanzania), or Kampala (Uganda) after recent elections there, and you will understand the quiet trust Zambians place in their own process. That trust was earned in 2021, when this system delivered a transfer of power the authors rightly celebrated, and it was renewed on 13 August. Zambia has voted.
Zambia has moved forward. We invite our critics to check the forms.
Zaza is a lawyer, civil rights activist and member of Action Governance Forum (AGF), a non-profit organisation.