How To Understand Polling

A number of polling companies exist, each with slightly different methods, but broadly they are focussed on regularly asking a cross section of the public who are believed to best represent the make-up of the UK what they think of politicians, their policies, and how they plan to vote in the next election. 

While on occasion pollsters' predictions have proven to be off the mark, the patterns that emerge from their data are generally a good indication of public opinion and how the next general election might play out. 

The next election must be called before the end of 2024, and the seismic shifts that have taken place across the political landscape since the last general election in 2019 have made voting habits more volatile, meaning polling is more important than ever. Case in point being the near reversal of Tory-Labour lead position since late 2022.

As people pay closer and closer attention to the polls before the public gets a chance to vote on political parties for real, here's what you need to know about how they work and what to look out for: 

What is polling?

Polling data, which is often made public, can provide a regular picture on which party may or may not be set to win the next general election, and which issues might swing voters' affiliation from one party to another. 

According to Dr Will Jennings a professor of political science at University of Southampton, explained that polling “tries to capture a representative sample of the population using different methods”, and that has been formally conducted around the world since the 1930s. 

“George Gallup famously introduced our opinion polling in America," he told PoliticsHome. “Britain picked it up a few years later and that is the dominant way in which we capture people's intentions to vote.” 

Polling experts PoliticsHome spoke to generally agreed that for most companies, political polling is not what makes them their money, but given the news interest, it is a reliable way to get their name known to attract corporate clients who might commission them for market research, for example.

Political parties pay close attention to polling, and it can influence policy proposals as well as who they might choose as their leaders.  

How does it work? 

Polling companies all have slightly different methodologies, but most now use digital surveys to ask the general public questions on a number of topics each week. 

There are numerous survey formats including voting intention surveys, questions on indivudual political topics, policies or indviduals, and also focus grouping, where people are gathered together in a room or online forum and respond to questions in a more conversational way. 

Adam Drummond, head of political and social research at leading pollster Opinium, explained that most companies have “panels” of up to hundreds of thousands of people who have “signed up to take part in surveys on loads of different subjects”. From that group, a certain number will be asked a set of questions, depending on the demographics wanted for the survey. 

When polling voting intent, Opinium typically surveys around 2,000 people with a focus on political questions. There are “controls” in place to account for the fact that panel sign-ups are likely to attract “people who are really interested in politics” to ensure that the sample is as representative as possible of the general population. Drummond said that "self-selection" is the "main thing we have to try and guard against". 

YouGov, another leading pollster, say they have recruited more than one million adults to their panel over the last two decades. 

Digital polling has made it easier to survey larger groups of people, which in recent years has given rise to the Multilevel Regression and Post-stratification’ model (MRP) from YouGov, measuring voting intention based on tens of thousands of individuals being questioned. 

Once a survey is complete YouGov will statistically weight the data to the national profile of all adults, by looking at age, gender, social class, region and their level of education. When conducting political polling, they will also take into account how people voted at the last general election, in the Brexit referendum and their level of political interest. 

The markers for this weighting come from previous general elections and the EU referendum, Office for National Statistics population estimates, the census, and largescale random probability surveys such as the labour force survey. 

Ipsos, another leading polling company, uses similar demographic markers, but say that their polling design is based on reaching people over the telephone, both on mobile and landline numbers. 

Polling is commissioned regularly regardless of the timing in the electoral cycle, but Jennings suggested that generally voting intention figures will naturally become more reliable in terms of predicting a result the closer it gets to polling day. The next General Election is due to be called before the end of 2024, and Labour has consistently had a double digit lead over the Conservatives since late last year. 

"Polls as we approach election day become increasingly aligned with the result," Jennings said.

The MRP model uses demographic data about individuals, and then their constituency

MRP, the so-called ‘Poll of Polls’ from YouGov has predicted general election outcomes down to a constituency level with a high degree of accuracy in recent years. 

The model tends to ask a greater number of people for their opinions than more traditional polls, and uses data to estimate the relationship between a person's demographic and how they will vote. 

That data is then used "at the local level to predict outcomes based on the concentration of various different types of people who live there," YouGov explains.  

The model can also use local election data to predict the outcome of a subsequent general election. "MRP models first estimate the relationship between a wide variety of characteristics about people and their opinions – in this case, beliefs about their local areas – in what is called a ‘multilevel model’, which allows us to account for specific area (in this case, council) level effects as well as background information about respondents themselves," the polling company explains.

In 2019, the YouGov model predicted that the Conservatives were on course for a majority, with most of the party’s gains to be “expected in the north, urban West Midlands, and former mining areas in the East Midlands and County Durham.” Data predicted that these wins would be concentrated in areas that voted Leave in 2016. 

MPR is favoured for its ability to ascertain patterns based on niche population groups, in order to accurately understand voting intentions of broader demographics. 

Focus groups are forums for small groups of people to discuss political topics of the day 

Small gatherings – often between six to eight people – will be used to discuss a certain topic or range of topics, are another way to gauge public opinion. The aim is to determine how people – sometimes of a certain demographic – feel about an issue or political figure. 

Focus groups have moved increasingly online since the pandemic. This makes it easier to assemble people of similar demographics without needing them to be in the same place at the same time.

Lord Robert Hayward, a former Conservative MP and elections expert, told PoliticsHome that focus groups are also good for capturing the public mood on an issue when one question would not quite be sufficient. Pointing towards how public opinion turned on Boris Johnson in the wake of partygate, he said that "you didn't need an actual opinion poll to tell you" that sentiment had shifted. 

"All the words that came up [in focus groups] about him indicated the public frame of mind about Boris in terms of partygate, so it wasn't an opinion poll that was measuring it," Hayward added. 

What can we learn from single issue or topical political polling questions?

As well as voting intention, polling companies will ask questions on the political topics of the day.

Partygate, Liz Truss's disastrous stint in Downing Street, and the Richard Sharp BBC scandal are among the topics that have been polled in recent months.

Hayward believes these topical questions are useful to both the public and political parties, and can be more insightful than voting intention when it comes to judging the reaction to an immediate issue, but the timing of the questions is vital to the value of the data. 

"You can ask those questions when the population at large is paying attention to that and you get relevant information from it," he said. 

"If you asked about partygate or Liz Truss now, most people are not thinking about partygate and they're not thinking about Liz Truss. So any poll that was taken now would be distorted because you would be getting only those who have a specific interest in the subject ." 

However, Jennings warned that the responses to topical questions can sometimes reflect the public's broader feelings on politics more generally, rather than their specific opinions on an issue. 

Citing the example of when Home Secretary Suella Braverman was recently under pressure, he explained: "If someone comes out and says 'X per cent of people say Suella Braverman should resign, you then should realise it's really very little to do with the latest scandal and more to do with the underlying position towards a politician or a party. 

"That is one of the ways in which we should be cautious about these issue questions, because often they're capturing much broader feelings, dispositions among the electorate, rather than actually a real orientation towards a specific policy."

How has polling evolved over time?

Polling operations have changed significantly in recent decades as the traditional methods of door knocking and phone calls have been overtaken by the internet, making it much cheaper to run polls.  

Hayward, who was an MP from 1983-1992, said that this reduction in costs has led to a “surge” in the number of polling companies in the market. 

Looking back to the 1990s, Hayward, reflected that polling was conducted almost entirely on a face to face basis at that time. 

"People turned up on your doorstep," he told PoliticsHome. 

"Nowadays, you actually have everything done online, and that's why everything has become so much cheaper, because previously you had to send somebody out and get the right range of demographics."

Having easier access to a wider range of people digitally has broadened the polling market, which "should be more reliable", Hayward felt. 

"But all new methods need to be tested against numerous general elections to know their reliability for certain, to account for individual issues that can affect a vote," he cautioned. 

He pointed to Brexit as a time-specific issue that altered demographic voting habits in numerous parts of the country between 2016 and 2019, but may not affect voter intent in the same way in future elections. 

"Every so often, you have what's called a 'wave result', whereby you have social change, demographic change," he explained. "Historically, people who lived in Surrey voted Conservative by a certain proportion. Suddenly, because of the great divide of Brexit, people in Surrey changed their voting habits." 

Is polling reliable and can we trust it?

The growth of online polling has made vast surveys the size of the MRP possible. 

Does this mean that there is the possibility of seeing a sample size of 250,000, half a million, or perhaps even more in the future? 

One polling source told PoliticsHome this was doubtful, as the margin for error does not get sufficiently smaller the more the sample size is increased beyond a certain level.

It is generally accepted that while polling is not perfect, it is probably the best indicator that we have got on how people feel about politics and how they may vote. 

“It has it has its flaws, but it is better than any other kind of quantitative way of predicting what’s going to happen in elections,” Drummond said. 

He felt the “baseline” of public opinion provided by regular polling as “useful” regardless of the point in the election cycle. While Hayward warned against taking polls as “gospel” he said there is not a “better indicator” that exists. 

“The reality is that asking people how they're going to vote if there was a general election tomorrow, via whatever method, whether it's telephone, whether it's in face to face, whether it's online, is still the best way of capturing electoral sentiment,” Jennings said.  

“Do I trust the polls? I trust them to give the best picture that we can possibly have at the current moment.

"But we should always use that data cautiously,  to recognise that they can be wrong in lots of different directions.” 

PoliticsHome Newsletters

PoliticsHome provides the most comprehensive coverage of UK politics anywhere on the web, offering high quality original reporting and analysis: Subscribe