Could Sadiq Khan Lose in London?

His advisers say the race is closer than it looks. Do their claims stack up?

Sadiq Khan is worried, and he wants you to know this. Although polling puts him well ahead in his race for a third term, the London Mayor has been heavily briefing their anxieties about a supposed “perfect storm” of negative factors which could yet cost Labour control of the capital. A few weeks ago, they shared their presentation with George Eaton of the New Statesman, and thereby sent up a distress flare to the broader progressive commentariat. Expectations management is a basic part of campaigning, so claims of weakness from an apparently dominant candidate should not be taken at face value. Khan would be unlikely to play up his weaknesses so publicly if his team did not see a political advantage in doing so. But things political campaigns say can occasionally be true, and at least one pollster who has tested opinion on the race has raised similar concerns about Khan’s position being weaker than it looks, while politically well connected pundits say Team Khan’s anxieties are sincere.

It is easy to see where these anxieties come from. Early spring polls of the 2021 London Mayoral race did indeed substantially over-estimate Sadiq Khan’s support, and under-estimated support for his Conservative opponent Shaun Bailey. The four polls taken in March and early April of 2021 gave Khan an average first preference lead lead of 23.5 points (50.5% vs 27%) - the final result was far closer, with Khan winning first preferences by just 5 points (40% to 35%). It is no wonder Khan’s advisors don’t see his current 20 point plus poll lead as secure - the polls only have to be off by the same margin this time as last time for the contest to be a dead heat. There are poll based arguments to make the other way, which I’ll come back to, but I think we can grant them the opening premise - the race could be closer than polls suggest. It was before.

The polls substantially over-estimated Khan’s strength in 2021 (but didn’t in 2016)

Graphs show average shares for Labour (Sadiq Khan) and Conservatives (Zac Goldsmith, 2016 and Shaun Bailey, 2021) in March-early April polls with all parties’ candidates (“all”) and with just Labour and Conservative candidates (“runoff”). Polls used 2016: Opinium 3-7 March; YouGov 8-10 March; ComRes 14-17 March; Opinium 30th March - 3rd April; ComRes 30th March - 3rd April. Polls used 2021: Redfield & Wilton 6-8 March; Opinium 17-20 March; YouGov 29 March-1st April; Opinium 7-10 April

Team Khan argue there are four factors which could further tip the balance of a close race away from Sadiq Khan and towards his Conservative opponent Susan Hall - the change in the electoral system used to choose a mayor; the introduction of compulsory voter ID; the “third term itch”; and a weakening position for an unpopular incumbent facing a restive electorate. Having looked at each, I think Team Khan’s claims about the electoral cost of rule changes and voter ID are overstated, while the “third term itch” claim doesn’t make much sense at all. If Khan faces a close race, it won’t be due to changes in the rules or quirks of third election runs, but because his polling has worsened substantially since 2021. But Khan may yet prove a lucky candidate - his weakening position coincides with a massive Conservative to Labour swing nationwide, a swing which should be enough to carry him to victory. Susan Hall would have to outperform Conservative national polling by a record margin to be in contention - and she does not at present look like a candidate with the kind of star power to pull off such a feat.

(1) The electoral system

The electoral system used to elect London’s mayor is shifting from a “supplemental vote” system where Londoners have an opportunity to cast two ballots, one for their favourite candidate and one for the candidate they would prefer in a run-off, to a “first past the post” electoral system where voters indicate a single preference only. Khan’s advisors have briefed concerns this damages his chances, as many progressives prefer to cast a first preference ballot for another party such as the Greens or the Liberal Democrats but then back Labour with their second preference. The results from 2021 support this idea - Khan’s lead grew from 5 points on first preferences (40% to 35%) to 10 points in the run-off (55% to 45%).

Khan’s aides told the New Statesman they were worried most Londoners were not aware of the changed voting system, and as a result Khan’s support would be depressed due to voters unaware of the change continuing to back smaller parties with no hope of victory. There are certainly going to be many Londoners unaware of the rule change - only political obsessives track electoral rules closely. But the rule change isn’t likely to matter in the way Team Khan claim it does. To understand why, we have to think about ballots.

The New Stateman’s George Eaton reports the problem as follows “Unlike in previous contests, voters will no longer be able to rank two candidates in order of preference”. This is an incorrect description of the old voting system used in London.¹ The old ballot paper did not ask voters to rank their preferences between candidates. Instead, their ballot papers had two columns next to each candidate’s name (see below). Voters made a single choice for a “first preference” candidate from the left hand column, and then a single choice for a “supplemental preference” candidate from the right hand column.

For the changed electoral system to matter, it requires that the change impacts on voters’ ability to express their intentions correctly. It is easy to see how that could happen if the change had been from ranking candidates on a single ballot to picking just one candidate. The ballot paper would look exactly the same, and voters might therefore put in multiple ranked preferences and inadvertently spoil their ballots. If Khan drew support from people whose first preferences went elsewhere, such a change could impact him.

But “same ballot, different rules” isn’t what is happening. The change is “same rules, different ballot”. Before, Londoners faced a ballot paper with two columns, and were asked to indicate a single preference in each column. There was in fact a change in 2021 when, due to a record number of candidates, the ballot paper had two lists of names and two columns next to each name, a change which may well have contributed to a large surge in rejected ballots (from 50k in 2016 to 114k in 2021) as it was not clear whether “column” referred to the two lists of candidate names, or the two lists of check boxes.²

Now, the second column - the “supplemental vote” - is being ditched. Voters will be faced with a single list of names, a single set of boxes to check, and an instruction to make a single choice. This is a change that is, if anything, likely to reduce the rate of spoiled ballots as it makes the process of voting considerably simpler. While some voters may still try and mark two preferences, the disappearance of a second set of boxes and the change in instructions on the ballot should be enough to alert most to the change in what they are being asked to do. What is more, if people do notice and respond to this change, this will likely help Labour - many progressive voters who prefer Labour to the Conservatives will no longer back smaller parties who cannot win when they know they can only make one choice.

Not all of these votes will come back to Labour. There are a lot of disgruntled progressives in London who may be inclined to back the Greens to show their unhappiness with Khan (though in doing so they are hurting the man who introduced London’s strict emissions rules, and helping the woman who wants to scrap them). But fragmentation under first past the post can also hurt Khan’s Conservative opponent. The big story of the past year politically has been the surge in support for ReformUK. The radical right Reform take votes primarily from the Conservatives in national polling, and there’s no reason to think things will be different in London. Reform didn’t even stand a candidate last time, but they are on the ballot this time and their candidate Howard Cox is averaging around 5% in polling. With four fifths of Reform votes coming from the Conservatives, even a modest showing by Cox will hit Conservative prospects.

(2) Voter ID

This is the first London Mayoral election since the introduction of compulsory voter ID, and Khan’s advisors worry that the ID requirement will disproportionately impact groups over-represented among London Labour voters, in particular the young and ethnic minorities. The slides shown to the New Statesman claimed “our internal polling shows 15 per cent of Londoners don’t have the necessary voter ID”, leading to some dire predictions of electoral disruption: “Labour estimates that this measure alone could cost Khan five points, enough to overturn his first-round victory against Bailey in 2021.”

While it is plausible that Khan supporters are more likely to come from groups lacking ID, an impact on this scale isn’t credible. Estimates of the share of voters lacking any relevant form of ID from other sources range from 2-4% in recent research to 7.5% in older research looking at a more restricted list of ID.³ Team Khan’s overall figure, sourced to unreleased “internal polling”, is twice as high as the highest estimate from published research. While Londoners may well be more likely to lack ID documents, it stretches credulity to imagine the numbers lacking any of the relevant forms of ID are five times or more higher in London than elsewhere in the country.

The claims about the electoral impact of voter ID are also hard to stand up. Research by former BBC Political Research editor David Cowling suggests that less than 0.5% of those who arrived at polling stations to cast ballots in the 2023 local elections were turned away due to a lack of ID documentation, and over 60 per cent of those turned away initially returned later and successfully voted. The share of voters who were turned away for lacking ID and did not return was less than 0.2% of total votes. These figures were substantially higher in more ethnically diverse local authorities, with more than 1% of voters turned away initially in boroughs such as Slough, Sandwell and Blackburn. But voters in such districts were just as prone to return later. Even in districts with higher initial refusals, the share of voters who were turned away and did not return never reaches 0.7%.⁴

While London Mayoral elections have a very diverse electorate, and higher turnout than local elections - meaning more low engagement voters who forget to bring ID may participate - it is hard to believe that the rate of voter refusal would be twenty times higher than the average recorded in 2023 local elections, and seven times higher than that recorded in even the most diverse and deprived boroughs last year. Nor is it plausible to imagine that all of those who are refused would have gone on to support Labour - older white voters with low education levels are also more likely to lack ID, and break heavily for the Conservatives.⁵ Even if turned away voters leaned heavily Labour, the share of voters turned away and not returning would have to be massively above that observed anywhere else in the country for the impact to be on the scale claimed to the New Statesman.

(3) Third term itch

Team Khan also raised what we might call a “third term itch”, pointing to Blair in 2005, Ken Livingstone in 2008 and Michael Bloomberg in New York City in 2009 as evidence that seeking a third term is a tougher proposition than re-election. This isn’t a sensible claim even about these three candidates, and makes no sense at all when the net is cast wider. All three cited cases involved dramatic shifts between the second third terms that do not apply to Khan. Blair’s third campaign was hampered by the backlash triggered by his decision to join the Iraq invasion. Ken Livingstone faced a far stronger opponent - Boris Johnson - in 2008 than he had faced in either previous contest. Bloomberg, elected twice as a Republican, ran for his third term as an independent in a heavily Democratic leaning city a year after Democrat Barack Obama won a landslide Presidential victory.

There are also plenty of examples of candidates sailing to a third victory without difficulty - Ed Koch in New York City won a landslide in his 1985 third election, so did Tom Bradley in Los Angeles in 1989, while Richard Daley in Chicago won six successive elections before retiring undefeated in 2011. Margaret Thatcher’s won her third general election in 1987 by a huge margin, and both Mario Cuomo and his son Andrew Cuomo won third terms as New York state governor by landslide margins. There is, in short, no law of political gravity that says third campaigns are harder. Neither Andy Burnham in Greater Manchester nor Steve Rotherham in Merseyside - both running for third terms at the same time as Khan - look likely to face greater trouble from their electorates than in their previous contests.

(4) Incumbency and low approval

While third terms may not be inherently harder to win, Khan’s third campaign may be harder because he faces an unhappy electorate eager for change. As the long term incumbent in London he is less able to benefit from this restive mood and more likely to become its target. The polling supports this argument - Khan’s approval in the most recent YouGov poll is -16, 20 points below his approval in the pre-election poll from the same pollster in May 2021 (though the question asked changed slightly, complicating comparison).

The biggest drops in Khan’s approval numbers come among Labour leaning groups

Source: YouGov/QMUL poll May 2021 and YouGov/QMUL poll Feb 2024

Even more worrying for the Mayor, the biggest drops in approval ratings come among Labour leaning groups - his net approvals are down 46 points among Remainers, 34 points among BME voters and 31 points about 2019 Labour voters. Approval among groups such as Leavers and 2019 Conservatives are stable - they were very negative about Khan in 2021 and remain so today. But the Mayor no longer attracts much enthusiasm from his natural constituencies, all groups who are overwhelmingly negative about the national Conservative government. With approval slumping among natural supporters, and hostility high among natural opponents, it is no wonder Khan advisors worry that Khan could struggle if his natural supporters stay home.

And they might. Self-reported turnout intentions in YouGov polling show the share of Londoners saying they are absolutely certain to vote was 39 per cent in February 2024 - down 14 points on the 53 per cent figure reported on the eve of the 2021 election, with larger drops among Labour leaning groups such as young people, ethnic minorities and inner London residents. While turnout intentions are likely to rise as election day approaches, this does suggest London Labour have a major job on their hands mobilising core supporters who are currently not very enthused by their candidate and nor very inclined to show up on polling day.

Reasons to be cheerful? National polling and the London lean

Sadiq Khan isn’t wrong to worry about the coming campaign. While his aides’ anxieties about the electoral system, voter ID and the “third term itch” look overblown, but there are real warning signs in the polling, with both Khan approvals and turnout intentions well down among natural Labour supporters. But London does not exist in a vacuum. Khan won re-election in 2021 in a very blue political context - this was the middle of the “vaccine bounce”, Boris Johnson’s Conservatives were polling in the low 40s, well ahead of Labour, and on the same day Khan narrowly won re-election Labour lost hundreds of council seats and the Hartlepool by-election. Khan’s third campaign takes place in a radically different environment - the Conservatives have lost half their 2019 vote, Labour have surged to a 20 point lead, and a landslide win for Khan’s party looks on the cards in the coming general election.

Labour national polling and London mayoral first preference vote performance 2000-2021

Polling figures are monthly averages for April in each election year, derived from Mark Pack’s PollBase dataset: https://www.markpack.org.uk/opinion-polls/ Figures for 2000 aggregate the vote for Ken Livingstone, standing as an independent, and official Labour candidate Frank Dobson

London voting patterns don’t match the national context perfectly, but there is some relationship. Labour Mayoral candidates have since 2000 done better in a pro-Labour national environment, as the above chart illustrates.⁶ Indeed, there is some evidence of a slight Labour lean in London voting relative to national polling, which makes sense given the capital’s demographics. Currently, Sadiq Khan is outperforming Labour by five points - this is in line with the long term average performance by London candidates, and the same as Khan’s advantage over national polling in 2021. Khan may still go on to do worse, but he would have to do very badly indeed relative to national polling to be in jeopardy. Since 2004 no Labour Mayoral candidate has under-performed national Labour polling by more than 1%. With Labour polling well above 40% nationally, even an unprecedented under-performance by Khan would likely leave him strong enough to prevail.

Conservative national polling and Conservative mayoral first preference vote performance 2000-2021

Polling figures are monthly averages for April in each election year, derived from Mark Pack’s PollBase dataset: https://www.markpack.org.uk/opinion-polls/

Just as Labour’s national strength should bouy up Khan, the Conservative national slump is a problem for Tory candidate Susan Hall. On average Conservative performances in the first round of past London Mayoral contests have roughly matched their national polling - Tory weakness in the capital overall is perhaps exaggerated as they remain very strong in some of the outer boroughs. The average masks some big differences though - the best Tory showing was in 2012, when Boris Johnson outperformed his party by eleven points to secure his second term. In 2021, Shaun Bailey equalled the record for Conservative under-performance - his 35% was seven points below his party’s national polling.

Hall, like Bailey, is regarded as a weak candidate lacking the charisma and cross-over appeal of Johnson, the only Conservative London Mayor to date. Yet she is so far actually matching Johnson’s 2008 performance by out-performing Conservative national polls by 2 points. Running slightly ahead of his party was enough for Johnson in the exceptionally blue year of 2008, but is nowhere near enough for Hall. Indeed, even if she were to match Johnson’s record performance of 2012, she would still be on only 35% support, unlikely to be enough to win even if Khan under-performs.

If you can’t be good be lucky. Sadiq Khan looks weaker now as a candidate than before, but has been lucky both in the timing of this election and in his opponents’ choice of candidate. His weakening poll numbers have coincided with a massive Conservative to Labour swing nationwide. While we only have a small set of past elections to examine, Tory victory in the current national climate would require a record under-performance by Khan relative to Labour and a record over-performance by Hall relative to the Conservatives. While a poor Khan performance is plausible, it is hard to credibly argue the case for Susan Hall as an electoral superstar able to outperform her party by double digits. She may do well enough to confound critics, but defeating Khan looks like a bridge too far.

There is, however, another mayoral contest taking place next month featuring a high profile Conservative incumbent with a proven track record of out-performing his party, in a hotly contested part of the country. Andy Street has twice out-performed Conservative national polling by seven points to narrowly prevail in two close West Midlands Mayoral elections. Street, like Khan, is running once again for re-election. But a seven point overperformance will be nowhere near enough to win given dire Conservative national poll numbers, so Street will either have to dramatically improve on past showings or face defeat. This contest ought to be a fascinating case study in the degree to which a high profile, independently minded candidate can escape the reputation of an unpopular party. Yet while there have been fifteen polls already of a London contest which doesn’t look close, there have so far been no reputable polls at all testing the waters in this crucial, hard fought contest in one of England’s largest metro areas.⁷ Perhaps it is time for Britain’s pollsters and pundits to stop asking “can Khan win?” and start asking “can Street win?”

1. It is unclear whether the error is Eaton’s own or comes from the briefing he was given, but it is consequential for the claim that the electoral system matters for Khan’s prospects.

2. Election analyst Lewis Baston has argued this ballot design primarily hurt Labour - “Among the London Assembly constituencies, the rate of rejection was highest in City & East (6.7 per cent) and lowest in Bexley & Bromley (2.8 per cent). We shall await further detail, but this suggests that the rejected ballots came disproportionally from deprived areas and those with large minority populations. Had these votes been able to be counted, it seems likely that Sadiq Khan’s first-round lead would have been a bit larger.” If so, then the shift to a first past the post ballot should help Khan recover many of these lost 2021 votes.

3. See this excellent House of Commons library briefing for more detail - https://researchbriefings.files.parliament.uk/documents/CBP-9187/CBP-9187.pdf

4. ID requirements could also depress turnout if people who lack the relevant ID, and would otherwise have voted, decide to stay at home. This is a hard effect to estimate, but one indirect means is to look at whether turnout declines when ID is introduced. There is not much evidence so far of a substantial or systematic fall in turnout in local elections after ID requirements were introduced (though such a fall could still happen in higher turnout elections which attract less engaged voters). See for example this analysis by the Royal Holloway Democracy and Elections Centre Team: https://hollodec.github.io/2023/05/26/voterid.html

5. The loss of elderly voters seems to have changed some Conservatives’ views of a reform they had insisted was about preventing electoral fraud. Jacob Rees-Mogg told the National Conservative conference in May 2023: “"Parties that try and gerrymander end up finding their clever scheme comes back to bite them, as dare I say we found by insisting on voter ID for elections….We found the people who didn't have ID were elderly and they by and large voted Conservative, so we made it hard for our own voters and we upset a system that worked perfectly well."

6. The 2000 contest, when Ken Livingstone ran as an independent after being defeated in the process to select an official Labour candidate, presents some methodological challenges. As the vast majority of Londoners would know of Livingstone - the former head of the Greater London Council and long serving London Labour MP - as a Labour figure, I have opted to aggregate his first preference votes with those of the official Labour candidate Frank Dobson.

7. The Centre for Cities have done a poll looking at visibility of metro mayors and opinions about the jobs they do, but it doesn’t seem to have asked any vote intention questions: https://www.centreforcities.org/publication/place-over-politics/