How to read the general election polls (and how not to)

The 2024 general election campaign has seen more polling than ever before – based on a crude tally last week, there had been around 20% more polls released so far this campaign than at a similar point in 2019.

This flurry of polling activity has led to some scepticism about the role of polling and some confusion about why different polls apparently show different levels of support for the parties.

As someone who works in research, it will come as no surprise that I think there is a value to robust opinion polling. Misinformation loves a vacuum, and if we didn’t have polling data, we would be left trying to navigate the competing claims of politicians, all of whom would no doubt claim that they’re hearing strong support for their message ‘on the doorstep’.

For those trying to work out how to vote – particularly the 1 in 4 in Scotland who say they plan to vote tactically – robust data on how others might be leaning can provide useful context for informed decision-making. And for politicians, journalists, academics and anyone else trying to understand the dynamics of public opinion, having good quality data is essential.

Read more:

Analysis | The battle for Scotland is too close to call

However, I also sympathise with those who are finding the current polling landscape somewhat overwhelming and confusing. Even for someone who works in the field, the volume of polling this time round has been hard to keep up with and digest.

Of course, one option at this stage is to ignore the polls completely, on the basis that the only poll that counts, as the well-worn saying goes, is on July 4. However, for those of us who are interested in what the polls are showing, how should we be ‘reading’ them – what should we look out for, and what might we want to ignore?

First, it is best to focus on the overall trends and patterns across polls – don’t place too much weight on any one poll, or on small changes between individual polls.

In general, if the level of support for a party shifts by less than four percentage points, this is described as within the ‘margin of error’. This is statistical speak for the degree of inbuilt uncertainty that accompanies all polls, simply because they are based on samples and not the whole population – samples, even representative ones, are not perfect.

Read more:

Unspun | Could ‘outlier' with score to settle cause election upset for the SNP

So if one poll says a party is on 40% and the next says they are on 43%, this is within the margin of error and you shouldn’t put much stock in it. However, if multiple polls are showing the same trend or pattern – as, for example, has been seen with the increase in Reform support at a UK level since the start of the election campaign – this is more likely to be capturing something shifting in reality.

Second, beware of polls that are not polls. This includes self-selecting Twitter/X ‘polls’ (which are only representative of the poster’s specific bubble) and ‘polls’ conducted on the doorstep by political parties as they are canvassing (which are extremely self-selecting, since many people will not talk to the candidates of parties they are not considering voting for).

Polls conducted while canvassing are extremely self-selecting by nature (Image: Newsquest)
Both are often cited on social media and election leaflets, but neither are likely to be a useful guide to how either the general public as a whole, or those in a particular constituency, plan to vote. If in doubt, check whether a company is signed up with the British Polling Council, which requires transparency about methods from all member companies.

Finally, as we enter the final week of campaigning, it is important to remember that all polls, including MRPs, are snapshots of a moment in time. They tell you what samples of voters are saying to polling companies right now about whether and how they think they will vote. This is always important to keep in mind – but it is arguably even more important for this particular general election, given that 42% of Scottish voters in Ipsos’ June poll told us they might change their minds about how they vote.

Get Scotland's top politics newsletter straight to your inbox.

In our recent MRP model, although our ‘headline’ estimate was for 34 Labour, 15 SNP, 3 Conservative and 5 Liberal Democrat seats in Scotland, we also identified 12 seats that currently look ‘too close to call’ and 14 others where voters are only leaning one way or another by a relatively modest margin. If people switch – or decide to stay at home – at the last minute, this may mean that the result on July 5 looks different to even those polls conducted nearest to election day.

That is not intended as an excuse for polling companies ‘getting it wrong’ – there will of course be time to unpick who was more or less accurate after the election. Rather, it is to emphasise that the outcome of any election – even one where one party seems a long way ahead – is uncertain, and that voting behaviour can change right up to the moment people put pencil to ballot paper.

Rachel Ormston is a Research Director at Ipsos Scotland