Cookies remember you so we can ensure to give you the best experience possible. By using this website or closing this message, you are agreeing to our cookies and policies

Do not show this message again
UK Insights

Why the Conservative lead over Labour collapsed

Joel Williams

Head of Methods

GE 22.06.2017 / 11:00


A Kantar Public view on the 2017 General Election

Polling in the 2017 general election was marked by enormous volatility. In particular, most pollsters recorded a rapid erosion of the Conservative lead over Labour as the campaign progressed. Kantar Public, whose final poll was among the most accurate, was no different in this respect.

Taking an average of our first two GB polls (largely carried out in late April), we estimated that the Conservatives had a 48% share of voting intentions compared to just 24% for Labour. How did this 24-point lead shrink to just five points (43% to 38%) in our last poll carried out between the 1st and 7th June? We now know that the lead on election day was even smaller, at 2.4 percentage points.

Since the election, there has been plenty of discussion about the ‘generational divide’ in voting choice but in this report we focus on an even more predictive characteristic: previous voting behaviour. To do this, we segment the poll samples with respect to 2015 general election choice and then show how changes over the course of the campaign in each segment’s voting probability and/or party choice impacted on the headline figures.

Methods used in this report

Over the course of the campaign our poll methods changed in three ways: (i) we started to reallocate UKIP and Green voters to second choice parties if there was no candidate in the constituency, (ii) we gave unregistered individuals a voting probability of zero once the registration deadline had passed, and (iii) in the last three polls, we used other data to impute probable voting intentions where this data was missing.

Each of these changes had a small impact on the headline voting intention shares but - for the purpose of consistency – in this report we omit changes (i) and (iii) while implementing (ii) from the first poll. Consequently, the numbers quoted do not exactly match the published figures although the story is the same.

2015 Conservative voters lost their turnout edge over the course of the campaign

In our first two polls 2015 Conservative voters were more likely than 2015 Labour voters to vote in the upcoming election (an average 88% voting probability compared to 84%). By the end, this gap had narrowed to virtually nothing (96% compared to 95%). Consequently, the 2015 Conservatives lost their turnout advantage over 2015 Labour voters despite an increase in their average voting probability from 88% to 96%.

The 2017 voting probability for other segments also increased very slightly over the course of the campaign - with the exception of 2015 UKIP voters – but movements in this respect did not have a strong influence on the Conservative-Labour lead.

Changes in voting probability over the course of the campaign by 2015 voting behaviour

KPublic _22June

The net effect of these changes in voting probability meant that 2015 Labour voters (and non-voters) increased their influence on the 2017 vote share in the last poll relative to the first two polls. This alone would have been enough to shave two points from the Conservatives’ 24-point lead over Labour at the start of the campaign but clearly changes in party choice were largely responsible for the lead shrinking to just four or five points.

2015 Labour voters ‘came home’

The most important element to the Labour surge was the ‘coming home’ of 2015 Labour voters. In the first two polls, only two thirds (65%) of 2015 Labour voters that were expected to vote intended to vote Labour. Nearly one in six (16%) intended to vote Conservative. In contrast, 2015 Conservative voters were solid: 93% of those that were expected to vote intended to vote Conservative again. By the last poll, this ‘loyalty gap’ had almost disappeared. In the last poll, the Conservative retention rate had slipped slightly from 93% to 88% but the Labour retention rate had increased enormously from 65% to 86%.

These movements in both voting probability and party choice among 2015 Labour and Conservative voters were enough to add nine points to the Labour share of voting intentions and subtract four points from the Conservative share. These changes alone would have been enough to transform a lead of 24 points at the start of the campaign to one of 11 points by the end.

Consistent pro-Labour swings among 2015 third party voters and non-voters

But what about the rest? Over the course of the campaign, Labour added another five points to its share of voting intentions from people who had not voted for either Labour or Conservative in 2015. In contrast, the Conservatives lost three points from this group, meaning that its lead over Labour shrank another seven or eight points to its final figure of four.

Over the course of the campaign, 2015 Liberal Democrats, Greens, UKIP and SNP voters each subtracted approximately one point from the Conservative lead over Labour. 2015 non-voters – the youngest of all segments - did the most damage, taking four points out of the lead as they swung strongly towards Labour over the course of the campaign. In the first two polls, they favoured the Conservatives 43% to 35%; in the last poll they favoured Labour 47% to 26%.

Tactical voting by 2015 Greens?

As noted above, 2015 Greens also shifted strongly in this direction. Although they made up just 2% of 2017 voters, a big swing in their voting intentions was enough to take one point out of the Conservative lead. In the first two polls, two thirds intended to stay loyal to the Greens. By the end, nearly the same proportion intended to vote Labour. It is tempting to think of this as an example of ‘tactical voting’ on the part of some 2015 Green voters: backing a more sympathetic than usual Labour leader in most constituencies but sticking with Caroline Lucas in Green-held Brighton Pavilion.

2015 Conservative voters wavered just enough to have an impact

However, apart from 2015 Labour voters and non-voters, the segment most responsible for the erosion of the Conservative lead was 2015 Conservative voters. We have already mentioned above how movements among 2015 Labour and Conservative voters with respect to voting probability and party choice had the biggest impact on the headline voting intention share. Now we can quantify it: two and a half points of the lost lead was due to 2015 Conservative voters (a) losing their turnout advantage over other segments and (b) becoming ever so slightly more open to Labour over the course of the campaign. Because this segment is the largest single segment, these small swings had a significant impact.


So that’s what happened: the Conservative lead over Labour collapsed due to an accumulation of small changes in voting probability and some large changes in party choice but the biggest single influence was 2015 Labour voters ‘coming home’.

For reference, the table below shows how much each segment added to each party’s share of 2017 voting intentions between the first two polls at the start of the campaign and the last one just before the election.

Change in vote shares between the first two polls and the last of the campaign (note: all percentages show impact on total vote share; they are not column percentages)

KPublic _22June2

Source : Kantar Public

Editor's Notes

For more information, please contact us

Latest Stories

The experts at Kantar look at alcohol consumption habits in the UK

The latest Kantar Worldpanel data shows supermarket sales have increased in value by 3.2% year on year in the run up to Christmas.

The perennial priority of any business is growth. But today that challenge is harder than ever and with a smaller margin for error.

Social media advertising is a good way to create long-term brand impact, boosting metrics such as brand awareness and association.

According to Kantar TNS’s latest Connected Life research, brands still have work to do to build trust in digital channels.

Related Content