How not to use polling

If you follow the people they won't thank you for it

Earlier in the month Keir Starmer kicked off an internal Labour row – get used to that phrase –  by telling Laura Kuenssberg he would keep the “two child limit”. This policy prevents families from claiming benefits for more than two children and has taken on symbolic importance for progressives as an especially cruel and counter-productive Tory policy. Scrapping it would take almost 300,000 children out of poverty and cost half a percent of the DWP budget.

I suspect Starmer didn’t mean to make his answer quite so unequivocal. He could have easily said, as other shadow cabinet ministers have done subsequently, that he would like to get rid of a policy he finds objectionable but is unable to make unfunded spending pledges at this stage. By sounding like he actually supported a policy he himself has previously called to be scrapped, and which his deputy has called “obscene and inhumane”, he inevitably triggered a backlash.

Those defending Starmer’s position have repeatedly pointed to a YouGov poll showing 60% of voters support keeping the limit and just 22% want to see it scrapped. This bit of polling is useful – without it progressives might assume everyone views this policy with the same horror they do. It is an important reminder that the “just deserts” conception of fairness is often more powerful than the “equity” one.

But it also shows the limits of single-issue polls. As regular readers will know I’m a big fan of polling. For all its problems it is far more useful than guessing what public opinion might be. It is still, though, badly misused. Even voting intention polls, which are relatively straightforward to read, are widely over-interpreted, with small changes in individual polls trumpeted as a dramatic shift.

Issue polls are much harder to use well. “Who would you vote for in an election tomorrow” is a simple question to which people can give a definitive answer. A policy question like “would you scrap the two-child limit” or “would you nationalise water companies” is far more complex. We don’t know how much respondents know about the policy; or how strongly their opinion is held; or how important the issue is to them; and so on.

Small changes in wording, which frame the issue in different ways, or add information, can dramatically alter responses. YouGov illustrated this earlier this year by asking whether the UK should leave the European Convention on Human Rights in three different ways. As you can see below, they got three very different sets of responses.

Another topical example: the Sunday Telegraph at the weekend argued for a referendum on “net zero”, and indeed you can find polling that shows people want one. But one of the dirty secrets of issue polling is that you’ll find people want a referendum on almost everything – from factory farming to tax avoidance – when the question is posed as a simple either/or. A YouGov poll that asked if people wanted a referendum “before [the government] attempted to meet its climate change target” found only 21% supported the idea.

Some of the misuse of issue polling is just partisan point-scoring. Commentators on the right will highlight polls showing the public are to the right of progressives on social issues like welfare, crime and immigration, but they will ignore those showing strong support for nationalising utilities or increasing public spending. Some on the left do the same the other way round.

But the problem goes deeper than this. Parties have, of course, always reacted to public opinion but it’s happening at an increasingly micro-level, in response to issue polling, forfeiting any consistency or sense of principle.. I don’t think the Labour leadership would have doubled down on the two-child limit without that poll. Equally polling showing their position on taxing private schools does have public support has ensured the policy hasn’t been scrapped despite an assault from the right-wing press.

The Tories are even worse as they openly search for cultural “wedge issues” like Rwanda where they’re in line with voters more than Labour.

These attempts to follow the electorate rather than lead it – or even properly understand it – are one of many reasons why Britain is struggling so badly, with our many and real problems being overlooked in favour of distractions and superficial populism. And it’s not just a policy problem. It doesn’t even work politically. Politicians have more information on what the public think than at any previous point, but they’re more unpopular than ever.

In the rest of the post I’ll look at the three big reasons why this simplistic approach to issue polling doesn’t work and why doing what it seems the public wants is not, in fact, what they want.