What do we do with Undecided Voters? 






24th May 2024

With under six weeks to go until the polls open on July 4th, Labour’s lead still varies noticeably across pollsters. In the last week alone it has ranged from sixteen to twenty-six percentage points. What might help explain some of this divergence? It all comes down to how different pollsters treat a crucial group of respondents: those who don’t yet know how they will vote.

When pollsters disagree, it is tempting to think that one of them must be wrong. But sometimes, at least at this stage in the campaign, differences can be partly explained by the fact that different pollsters are actually answering subtly different questions. Some pollsters, like Deltapoll, take the approach that polls can only ever provide a snapshot of voting intention a certain point in time. These pollsters treat undecided voters as truly undecided, and remove them from headline results. As a consequence, their results are made up exclusively of people who know how they intend to vote.

The alternative, favoured by some other pollsters, is to try to accurately predict the outcome of a future general election. Come July 4th, many currently undecided voters will have made up their minds. Some polls attempt to take this into account and try to assign undecided voters to the party they are likely to end up voting for. There are different ways to go about this. Some ask ‘squeeze’ questions to undecided voters, prompting them again to choose a party, and allocate them based on their answer. Another approach, which ends up having a similar effect, is to remove undecideds from the data before weighting, to ensure that the sample of voters is representative of how the electorate voted last time. Others will try and use statistical modelling to predict how those who say Don’t know will vote, based on a range of different factors.

Analysis of published voting intention figures from different pollsters suggests this methodological choice really does make a difference. The average Labour lead across polls conducted between the Local Elections and the announcement of the General Election was twenty-one percentage points. Among pollsters who take out Don’t know, this figure rises to twenty-two percentage points, whereas other pollsters give Labour an average lead of just seventeen percentage points.

Why do those pollsters tend to have a lower Labour lead? One reason is that most of the people currently answering “don’t know” in polls voted Conservative in the 2019 General Election, and it has been assumed that they are likely to ultimately do so again on July 4th. By leaving these respondents out entirely, other pollsters tend to have larger leads for Labour.

As the General Election looms, however, the impact of undecided voters on the polls is likely to decline. Many of these voters will begin paying more attention to politics as the campaign progresses, manifestoes are published, and debates take place. Eventually, it is likely that many of them will make up their minds and begin telling pollsters who they intend to vote for. When that happens, the number of undecideds in pollsters’ samples will decline and so too may the effect of their contrasting methodologies.

Ultimately, if a respondent says they do not know who they are going to vote for right at the start of an election campaign, they may well end up making their mind and voting for one party or another. In contrast, if a respondent says they do not know who they are going to vote for right at the end of an election campaign, there is a much greater chance they will not vote.

The data for this analysis was taken from polls by selected pollsters conducted between the Local Elections and the announcement of the General Election. 

This article was first published in our newsletter. Sign up to receive regular updates with more exclusive polling insights and analysis: https://deltapoll.co.uk/

Ruby CooperAuthor: Ruby Cooper