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

Meta-analysis of polls points to a seven point win for the Conservative party

Luke Taylor

Head of Social and Political Attitudes

GE 08.06.2017 / 16:00


A Kantar Public view on the 2017 General Election

The final polls show a wide range of possible outcomes – from a one point Conservative lead to a thirteen point Conservative lead.

The difference in these projections can broadly be explained by the different turnout weighting applied by pollsters, and in particular the impact which these weights have on the demographic profile of likely voters.  Reviewing the published data tables pollsters have made available (nine have published sufficient information to be included in this analysis), there are two key areas where the demographic profile of likely voters differ substantially:

• Age– the proportion of young people (18-34) ranges from 15% of likely voters to 28%

• Previous voting behaviour– some pollsters are making the assumption that very few voters will be new voters that did not vote in 2015, whereas other polls have a far higher proportion of these individuals (ranging from 4% to 15% of likely voters). It should be noted that not all published polls have a cross-break showing the number of previous non-voters, however, most do show the proportion of 2017 likely voters that voted for the four major parties in 2015 (Conservatives, Labour, Lib Dems or UKIP). We can therefore calculate the proportion of voters that did not vote for one of these four major parties in 2015 and use this as a proxy for the proportion of non-voters in each sample.

In this general election age and previous voting behaviour are strongly related to voting intention; this means that the different profiles of likely voters are having a substantial impact on the overall poll results.

In fact looking at the Conservative lead over Labour, a model with just these two variables – age & previous non-voting - accounts for 88% of the variation in poll results (R2 = 0.8783 and p=0.0018). Out of these two variables, age is the most strongly related to the Conservative lead over Labour; for each percentage point increase in the proportion of those aged 18-34 in a poll (controlling for the level of non-voters), the Conservative lead is cut by 0.62 percentage points (p=0.00). The impact of previous non-voting is more muted but still significant; for each percentage point increase in previous non-voters (the proxy measure described earlier) the Conservative lead is cut by 0.35 percentage points (p=0.03). The relationship between these three variables is shown in the following chart.

Electinoday _1

We can use data from the 2015 BES to use this model in order to predict the outcome of the election - on the assumption that the voter profile will be identical to the 2015 election – this suggests a seven percentage point win for the Conservative party.

Source : Kantar Public

Editor's Notes

For more information, please contact us.

Latest Stories

Kantar Millward Brown’s 2018 AdReaction study finds that three quarters also find multimedia campaigns unmemorable.

What do VR and AR offer in terms of enhancing the customer shopping experience?

Countries whose rating of China increased the most were all developed countries; awareness of the Belt and Road Initiative rose significantly from three years ago; Chinese food and high-speed railway are most mentioned cultural and high-tech elements.

According to the latest smartphone OS data from Kantar Worldpanel ComTech, Apple achieved its highest share in more than three years in the month of November.

Kantar Worldpanel data on grocery market share in the UK for the 12 weeks to December 31 2017.

Related Content