Why Kantar is showing a smaller Labour lead than other pollsters

The latest poll from Kantar gives Labour (only) a four point lead over the Conservatives, noticeably lower than the Labour lead in recent polls from other firms.

Here is Kantar’s explanation for the difference:

A four-point Labour lead is smaller than most other recent voting intention polls. Other polls from the last two weeks have shown a larger Labour lead (ranging from +7 to +13 points). We think this disparity is due primarily to a methodological difference between our poll and other polls. In particular, we do not provide the option to answer ‘don’t know’ to the party choice question, only ‘prefer not to say’ and ‘would not vote’. Most other pollsters also have a ‘don’t know’ option and 2019 Conservative voters have been selecting this option at an unusually high rate compared to 2019 Labour voters.

It is our understanding that these other pollsters treat any responses of ‘don’t know’, ‘prefer not to say’ or ‘would not vote’ as non-voters so far as the next election is concerned. Because 2019 Conservative voters have been selecting ‘don’t know’ at a higher rate than 2019 Labour voters, proportionally fewer of them are ‘surviving’ as next-election voters. This adds materially to the Labour voting intention share.

In our view it is unlikely that such a large proportion of Conservative 2019 voters would choose not to vote in the next General Election, or that they will turn out at a lower rate than 2019 Labour voters. For context, analysis of the 2019 post-election British Election Study shows that c.90% of both 2017 Conservative voters and 2017 Labour voters went on to vote in the 2019 General Election.

In contrast, in our poll, we see no difference in such survival rates between 2019 Conservative and Labour voters. This, on its own, means a smaller Labour lead than other pollsters are showing. The ‘loyalty’ levels among those 2019 Conservative and Labour voters who have selected a ‘next election’ party is very similar to the levels found by other pollsters.

UPDATE: Opinium’s new methodology (as of February 2022) also takes a more hostile approach to ‘don’t knows’, as explained here. Although, as with financial markets, past performance isn’t a great guide to future performance, it’s worth noting that Opinium and Kantar were two of the most accurate pollsters at the 2019 general election.

NOW AVAILABLE TO ORDER: My next book tells you all about how to make sense of political polling – Polling Unpacked: the history, uses and abuses of political opinion polls.