When the result of the broadcasters’ exit poll was revealed at 10 o’clock on 4 July, the lack of surprise was almost palpable. The headline projection – a Labour landside – had been echoed consistently by virtually every major opinion poll carried out during the election campaign, and indeed for several months before that. The only real doubt as we awaited the chimes of Big Ben that Thursday night was the exit poll projection regarding seat numbers. When the figures flashed up, it suggested Labour would win 410 seats and the Conservatives 131. Over the following 24 hours, the game switched to comparing those numbers with the actual result as the votes were counted and seats declared. The poll proved remarkably accurate, missing the Labour number by just two and the Tories’ by 10. For the smaller parties and those in Scotland, the disparities were slightly greater, mainly due to there being fewer data points. But overall it had been a triumph both for the polling industry and for the BBC, ITV and Sky, who jointly commission the project. As the back-slapping subsided, however, the questions began. And they were not new.
Influencing elections
There have long been concerns about opinion polling during elections. Central to the issue is the fundamental conundrum: do polls published during an election campaign influence the outcome rather than merely forecasting it? Applying that premise specifically to the 2024 general election with its clear-cut result gives rise to some obvious further questions. Chief among them, were Conservative supporters disinclined to turn out because the polls all suggested an inevitable and irreversible defeat for their Party? Put it another way, were the opinion polls indirectly responsible for voter suppression which increased the scale of the Tory defeat?
This coin, of course, has two sides. It’s equally reasonable to ask whether, given the consistency with which polls projected a significant Starmer victory, some Labour supporters simply didn’t bother to vote because their favoured result seemed to be guaranteed. And at a seat-by-seat level, did people change the way they intended to vote having looked at the polls and calculated that switching their allegiance could produce a more optimal result in their home constituency?
Aside from the impact on voters, the effect of polls on media coverage and on the very nature of political discourse during elections are often cited as further reasons for restrictions. The 2024 election saw more polls commissioned and published by more organisations and outlets than in any previous campaign. The media, and social media in particular, was dominated on a daily basis by the latest numbers. Accordingly, whenever politicians were interviewed, journalists could not help but ask about the polls. The cumulative effect was that discussion and coverage of actual policies, ideas and manifesto commitments was too often crowded out. The election discourse became about who’d gained or lost the odd point here or there, and how that might translate into seats. Surely the fourth estate and the political classes should be focused on the bigger picture, and curtailing publication of polls during campaigns would help achieve that?
Even if we agree on the pertinence of the questions, the search for a consensus on causes and remedies proves consistently elusive.
The premise that opinion polls may affect people’s voting decisions is hardly controversial. ‘Floating voters’, those whose support is not rusted on to any particular party, will by definition be influenced by a range of factors when deciding how to vote. Given the ubiquitous coverage of opinion polls in the run up to election day, it would be bizarre if that wasn’t one of them. This is supported by a number of academic studies.
After the 2015 general election, at which the pollsters lamentably failed to forecast the Conservative majority – one of the many errors which, incidentally, saw your author elected to parliament – the polling industry took a long, hard look at itself. Professor Jouni Kuha from the Department of Statistics at the London School of Economics was one of those tasked with the inquiry. Kuha concluded that opinion polls during election campaigns are not independent of the political process, but could influence voters’ behaviour by affecting their expectations of the outcome.
The extent to which this phenomenon can accurately be measured, however, is open to question. Asking every voter to assess what impact the polls have on their decision-making process would be both ineffective and impractical. ‘Easy’, we might say, ‘just use sampling methodology’. But guess what the 2015 inquiry concluded had been the main cause of error in the pollsters’ forecasts? Indeed. Failures of sampling methodology.
Perhaps we can learn lessons from other parliamentary democracies. There’s a long history of opinion polls being restricted, curtailed or banned in various countries during elections because of their potential to skew the results. Twelve years ago, academics at the University of Hong Kong carried out a survey to determine which countries operated a blackout period for the publication of poll results during elections. Of the 83 replies they received, 38 countries reported bans of one kind or another. This was later written about by David Cowling, the BBC’s brilliant psephologist from whose expertise I benefitted enormously during my time as a journalist. In the aftermath of the 2015 polling failure, Cowling reflected that the ‘Hong Kong list’ of 38 jurisdictions where restrictions were in place included Bhutan, Brazil, Canada, France, Greece, Mexico, Norway, Poland and Venezuela. ‘It is difficult to characterise the list as comprising countries where democracy is fragile and where authoritarian regimes prevail. So, many mature democracies find no problem with restricting pre-election polls‘, he said.
Arguments against
Cowling, however, was not persuaded that polls should be banned during election campaigns in the UK. He’s clearly not alone. Such measures have been frequently discussed but never seriously proposed, and certainly never implemented. For the moment, and for the foreseeable future, the status quo prevails. The Parliamentary Office of Science and Technology recently produced a report on the extant position. (Presciently, it did so just three weeks before Rishi Sunak called the ‘snap’ election of 2024, which should not actually have been regarded as a surprise – see my previous blog here). The report noted that the Representation of the People Act 1983 prohibits the publication of the results of polls conducted on election day whilst voting is taking place. But other than that, there are no further statutory restrictions. Even this limited measure is now more than 40 years old, and so takes no account of the huge expansion in the number of polls, polling organisations and outlets in the digital age. Subsequent regulation and guidance have sought to encourage best practice among the polling industry and the media, with some success. But teeth are conspicuous by their absence.
There’s no single reason why the various ideas for tightening the rules have never proceeded beyond the drawing board. One of the main arguments against restricting the publication of ‘official’ polls during elections is that the resulting vacuum will simply be filled by the ‘black market’, making things worse, not better. Unregulated pollsters will fly under the radar and publish dodgy numbers, or reputable organisations will carry out supposedly private surveys for their clients which end up being conveniently ‘leaked’.
No amount of restrictions will dampen the demand for polls because so many stakeholders find them incredibly useful. Political parties use the figures to adapt their messages and target voters; lobby groups use them to identify the most effective avenues to advance their chosen messages; and, of course, the media has a voracious appetite for material. While the demand persists, there’s always a bloke who’ll sell you stuff from his boot in the pub car park, even if it’s dangerous or wrong. LibDem grandee Mark Pack, author of Polling Unpacked, puts it neatly: ‘…banning good information leaves the field open to bad information’.
There’s a broader philosophical argument against further restrictions. Why shouldn’t voters – and therefore election outcomes – be influenced by opinion polls? They’re influenced by so many other factors, many of them unintended. Who are we to tell voters they’re not allowed to take polls into account when deciding how to vote? This is especially so now, given the increasing trend for voters to shift allegiance. Tribal loyalties are less important than ever, and people are looking for data to help inform their voting decisions like never before. As long as the polls are fair, accurate and not a source of misinformation, what’s the problem?
Quite often in this debate the waters get muddied. Those who argue for greater restrictions are often doing so to solve a different problem; that of the inaccuracies of polls rather than their influence on voters. That’s important of course, but it’s a different issue. In addition, polling methodology has improved hugely since the bad experience of 2015. The properly regulated ones are more accurate than they’ve ever been. And regulation is the key here. The polling industry is not perfect and never will be, but its house is broadly in order. Any attempt to restrict its activities could have unintended consequences. But more importantly, it would mean withholding facts from voters at a time when access to accurate, reliable information is at a greater premium than ever.
Peter Heaton-Jones.
Peter Heaton-Jones was Member of Parliament for North Devon 2015-2019. He now works as a public affairs consultant.
The Constitution Society is committed to the promotion of informed debate and is politically impartial. Any views expressed in this article are the personal views of the author and not those of The Constitution Society.