Opinion polls are surveys of public opinion used to estimate the voting intentions of the electorate. They are a prominent feature of British political life, delivering insights into party support and leader approval ratings, courtesy of a host of polling companies.

In the lead up to a general election and during the campaign, they dominate media coverage as a barometer of the political mood.

Following a series of high-profile “misses” in the 2015 general election, 2016 Brexit referendum and the 2017 election, the British polling industry generally forecast the correct outcome of the 2024 general election.

Months of opinion polling pointed to a large Labour landslide which ultimately came to pass.

A YouGov MRP poll correctly predicted 92% of constituencies. But the organisation’s review of its performance during the general election acknowledged that its prediction of vote share was less accurate. Across the industry, polling companies generally overestimated Labour’s level of support.

Yougov said that “like almost all other polls we overestimated the level of Labour support”.

How opinion polls work

A standard opinion poll surveys a sample of between 1,000 and 2,000 people.

Polling companies then “weight” this raw data, adjusting it for demographic factors like age, gender, social class and past voting behaviour to ensure the sample accurately reflects the country at large. This is to ensure the adjusted sample is representative.

What is an MRP poll?

An MRP poll is a type of opinion poll that uses a statistical technique called Multilevel Regression and Post-stratification to estimate election results at a constituency level.

The “multilevel” aspect means the model accounts for both individual characteristics and geographic differences at the same time.

The model is then applied to real demographic data for each constituency, estimating how the actual population mix in each seat would likely vote. This is the “post-stratification” stage. The result is projected vote shares for all 650 constituencies and therefore a potentially more reliable seat forecast than other models.

MRP polling gained major attention in the 2017 general election, when YouGov’s model correctly predicted the prospect of a hung parliament, contradicting most traditional polls at the time, which expected a Conservative majority.

Why can opinion polls get it wrong?

Polling is not an exact science.

As a poll only surveys a small sample of the population, there is always a statistical margin of error.

Beyond simple sampling error, polls can be inaccurate if the respondents are not representative of the wider public or are untruthful about their intentions. This has previously been cited as the “shy Tory” factor, where voters are reluctant to admit their support for the Conservatives. The extent of this phenomenon is debated by experts.

Pollsters also struggle with turnout. Their results will struggle to account for the unpredictability associated with who will actually show up on election day.

Historic polling ‘misses’

The current opinion polling landscape has been shaped by a series of high-profile “misses”.

One of the most famous polling failures came at the 1992 general election, when the Conservatives under John Major secured a parliamentary majority despite most polls pointing to a Labour victory.

More recently, in 2015, polling averages suggested a hung parliament. Instead, the Conservatives under David Cameron won an outright majority. The scale and uniformity of the polling error led to the establishment of an inquiry by the British Polling Council to examine what had gone wrong.

In the 2016 Brexit referendum, the majority of polling suggested a narrow Remain victory.

Surveys during the 2017 general election campaign largely projected a comfortable Conservative majority – with some even suggesting a Tory landslide. Instead, a stronger-than-expected Labour performance resulted in a hung parliament.

Should opinion polls be banned during a general election campaign?

The influence of polls on the political process is a subject of ongoing debate, with some calling for restrictions on their publication during the election campaign period.

An opinion polling ban during a campaign period is sometimes described as “election silence”.

A 2012 global survey by the World Association for Public Opinion Research (WAPOR) and collaborators found that, of 83 countries and regions responding, 38 reported the existence of a blackout period banning the publication of opinion poll results in the run-up to elections. The length of these bans ranged from 21 days in South Korea; 15 days in Argentina, Greece and Italy; 7 days in Columbia, Cyprus, Peru and Romania; to three days in Canada, the Czech Republic, and Mexico.

The case for a ban

Critics argue that the constant focus on the latest polling figures can crowd out substantive debate about policy. Around election time, media coverage can become concentrated around marginal movements in the polls or certain “shock polls” (likely outliers) which predict a politically momentous outcome.

But criticism of polling during election campaigns centres on its potential impact on voter behaviour. It has been suggested that polls indicating a large projected majority for one party may depress turnout, either by encouraging complacency among that party’s supporters or by fostering fatalism among its opponents.

Voters who believe the result to be a foregone conclusion may conclude that their individual ballot carries little weight. Conversely, close polls can sometimes stimulate turnout by heightening perceptions of stakes and competitiveness.

The polls pointing to a Labour landslide in 2024 could well have contributed to the election’s low turnout. The polls predicting a hung parliament in 2015 might have actually motivated Conservative voters; this picture played into the Tory campaign’s warnings of a “coalition of chaos” comprising non-Conservative parties.

Polling can also influence tactical voting calculations under the first past the post system.

Constituency-level polling, and sometimes even national polling extrapolated locally, is frequently invoked by political parties in campaign literature to argue that they are best positioned to defeat a particular opponent in a given seat. Where such claims rely on limited or methodologically weak data, they can mislead voters and distort strategic decision-making.

The case against a ban

Proponents of polling during election campaigns argue that any prohibition on publishing surveys would represent an unjustified restriction on the public’s right to receive information.

In liberal democracies, they argue, voters should be trusted to evaluate information critically rather than be shielded from it by the state. A ban would therefore raise constitutional as well as practical concerns.

It is also argued that a ban would be ineffective in practice. Political parties, financial institutions and large corporations would almost certainly continue commissioning private polling for strategic purposes. The results could easily circulate informally as leaks, rumours or selective briefings.

Such an environment, therefore, could actually worsen the information ecosystem. Instead of transparent methodologies and published sample sizes from professional pollsters, voters would be left to navigate unverified claims and partisan spin.

Supporters further maintain that polling is simply one among many sources of information available to voters. For voters inclined to vote tactically, reliable polling can be an essential tool in determining which candidate is best placed to defeat a less-preferred option. Tactical voting organisations commission and base their analysis on large polls.

In 2018, the House of Lords committee on polling and digital media did not express support for “election silence”.

It concluded: “We are not convinced of the case for introducing a ban on the undertaking and publication of voting intention polls in the run-up to elections.

“In the future, if polls continue to be a poor predictor of the eventual outcomes of elections, and if the media reporting of such polls continues to influence public and political discourse in a misleading way, then arguments by supporters of a ban would be strengthened.”

The 2015 general election polling miss

The 2015 general election arguably represents the most significant polling failure in the UK since 1992. The near-unanimous failure of the polling industry to predict the correct result prompted a major independent inquiry that led to reforms in how British polling companies conduct their research.

Throughout the election campaign, a steady stream of polls from all the major companies appeared to show the Conservative and Labour parties as effectively neck-and-neck.

The final Guardian/ICM campaign poll placed both parties on 35% of the vote. The Guardian’s poll projection, an average of all the polls made public, placed the Conservatives on 274 commons seats and Labour on 271 – a hung parliament.

After all, the final opinion polls for the nine British Polling Council members had all indicated a statistical dead heat.

In the end, the Conservative Party under David Cameron’s leadership won 36.9% of the vote, while the Labour Party, led by Ed Miliband, won just 30.4%. This 6.5-point lead delivered a surprise, if small, overall majority for the Conservatives in the commons.

Peter Kellner, chairman of YouGov, suggested that “shy Tories” deliberately misreporting their vote intention could have contributed to the polling miss.

Kellner wrote: “This year, as in 1992, the Tories have a weak image. They are widely thought to be out of touch and for the rich. But, at the margin, there may be some people who both have a poor view of the party but nevertheless think it will run the economy better than Labour.

“They are ‘shy Tories’ not because they are unwilling to admit their choice of party to a stranger but because they really would like to support someone else but, faced with a ballot paper in the privacy of the polling booth, simply can’t.”

In the aftermath, the British Polling Council commissioned an independent inquiry, chaired by professor Patrick Sturgis of the University of Southampton, to determine what went wrong and to “make recommendations for future polling.”

The BPC stated: “The final opinion polls before the election were clearly not as accurate as we would like, and the fact that all the pollsters underestimated the Conservative lead over Labour suggests that the methods that were used should be subject to careful, independent investigation.”

The Sturgis inquiry published its final report in March 2016 and concluded that the primary cause of the failure was unrepresentative samples. The methods used by the polling companies were consistently recruiting too many Labour supporters and too few Conservative supporters into their surveys, the inquiry found.

The report argued that “consistent lying” on the part of those sampled (the shy Tory explanation) was unlikely.

The report stated: “In sum, while it is very difficult to definitively rule out the possibility that respondents lied to pollsters before and after the election, circumstantial evidence from contemporaneous surveys suggests that it is unlikely to have been a contributory factor in the polling miss.”

Similarly, the report rejected claims of a significant “late swing” to the Conservatives that pollsters had failed to detect.

The inquiry did suggest that differential turnout modelling may have contributed modestly to the error. Differences in turnout rates between supporters of different parties that were not addressed by pollsters, the report found, may have slightly amplified inaccuracies in vote share projections.

Finally, the report identified evidence of a “herding” phenomenon. It found that some polling organisations appeared to make design or reporting decisions that reduced the variation between their results and those of competitors.

The BPC initially vowed to issue a follow-up report before the next scheduled general election outlining how member companies had revised their sampling and weighting procedures in light of the 2015 failure. In practice, while no single unified blueprint emerged, the immediate aftermath of the Sturgis inquiry did prompt methodological changes across the industry.

The House of Lords committee on polling and digital media, reporting in 2018, called for the remit of the British Polling Council to be expanded to take on a more substantial oversight function.

The ‘Conservative Britain Alliance’ controversy (2024)

In January 2024, YouGov published a MRP poll predicting a Conservative wipeout at the forthcoming general election. The poll was commissioned by a secretive group called the Conservative Britain Alliance.

The poll, which surveyed over 14,000 people at an estimated cost of £70,000, was published on the front page of the Daily Telegraph and predicted a 120-seat Labour landslide. It appeared as the House of Commons prepared for votes on legislation to operationalise the government’s Rwanda deportations scheme, known as the Safety of Rwanda (Asylum and Immigration) Bill.

The poll and surrounding campaign was widely viewed as an attempt by certain figures on the Conservative right to force Rishi Sunak out of office and install a leader with a more hardline stance on immigration.

The survey in question was notable for its use of a specific “hypothetical thought experiment.” After asking a standard “best prime minister” question involving Rishi Sunak and Keir Starmer, respondents were presented with a scenario in which the Conservatives had replaced Sunak with a new, unnamed leader who implemented five specific policies:

• Taking a much tougher approach on people coming to the UK illegally on small boats from France
• Introducing tough new measures on crime and anti-social behaviour
• New measures which significantly reduced legal migration
• Bringing in new tax cuts for working people
• Successfully getting waiting times for operations with the NHS falling

The results suggested that while Starmer led Sunak, a “hypothetical” leader with this platform would be more competitive.

Critics and pollsters, including YouGov itself, noted that such hypothetical polling is complex to interpret. As the leader was not a specific person, respondents could project their own ideal characteristics onto them.

Significantly, the so-called Conservative Britain Alliance which commissioned the poll had no website and no public profile. The British Polling Council ultimately clarified that YouGov had complied with transparency rules by naming the client and the contact person. Lord (David) Frost, a former cabinet minister and prominent critic of Sunak, was named as the group’s contact.

In a Telegraph piece accompanying the findings, Lord Frost said the prediction of a Conservative wipeout showed that Sunak needed to “be as tough as it takes on immigration”.

The original Telegraph article argued that Reform UK, a party with a hardline stance on immigration, could be the difference between Labour securing a substantial majority and merely emerging as the largest party in a hung parliament.

YouGov, who carried out the poll, challenged this reporting in an official statement. The organisation said the Telegraph calculation appeared to be based on “simply on adding the Conservative and Reform UK vote shares together in each constituency, which is not a reliable way of measuring their impact.”

In late January 2024, former cabinet minister Simon Clarke publicly called for Sunak to resign as Conservative leader. He argued that “Rishi Sunak is leading the Conservatives into an election where we will be massacred”.

Sunak ultimately remained in post as Conservative leader and led his party into the 2024 general election.

Additional resources

• The British Polling Council: https://www.britishpollingcouncil.org/
• Politico – Poll of Polls: https://www.politico.eu/europe-poll-of-polls/united-kingdom/
• YouGov: https://yougov.com/en-gb
• Ipsos: https://www.ipsos.com/en-uk