By Peter McLeod
"You all got it wrong, didn't you?" This, and less polite variations on it, is the kind of thing people say to me when I tell them I'm a pollster. It's been a rich couple of years for conversations like that, what with the 2015 general election, Brexit, and Trump. Despite the fact that the final average of polls in each of those elections was within margin-of-error of the result, plenty of people have decided they've had enough of the polls; if you’re one of them, I doubt I'm going to persuade you otherwise.
On the other hand if you do believe the polls, and you're reading this website, you probably already know that Labour appears on course for an historic defeat. You've seen that the Conservatives could win the most seats in Wales for the first time since the 1850s, that Labour is doing even worse in its own seats than the national polls show, that it looks like it will be the Tories, not Labour, who finally peg the SNP back, that for the past 6 elections the polls 50 days out overestimated the Labour share of the vote. And that Theresa May has the highest rating as most capable prime minister recorded by Ipsos MORI in the 38 years they've been asking the question. You know all this.
So instead, I want to talk about how campaigns use polls. While media coverage is getting a lot more sophisticated, it still tends to fixate on the horse race. Of course, parties pay attention to this and it can make a real difference to how they behave. Throughout the 2010-15 parliament, publicly available polls gave Labour substantial leads over the Tories, enough to convince a lot of people in the party that it was doing and saying the right things. But, as my old GQR colleague has written, our private polling was always more pessimistic than what appeared in the papers. We thought the party needed to shift its position on spending and immigration if it wanted to win a majority, but our bad-news polls could always be set against a weight of public polling that said everything was fine. We all winced when the audience on Question Time a week before the election hammered Ed on Labour's spending record. Any chance to neutralise it had gone years before.
But again, the horse race is only the beginning of what a campaign uses polling for. More than anything, politics is a persuasion business. This goes for the activist knocking on doors as much as for the leader standing at the despatch box, in a private meeting with their parliamentary party or going on Marr. And to persuade people, you need a message. This isn't a line or a paragraph or an essay, it is a story: about how the world is today, what needs to change and why you – not any of the other candidates – are the one to achieve this. Campaign research, then, is all about identifying, testing and honing these messages. Doing it right takes a lot of things, but four are chief among them.
First, targeting. You need to know who you're trying to persuade. Some people will love a particular message, but would never vote for you anyway. You want swing voters, and for that you need a poll with a big enough sample to delve into each party's voters and pick out those who a) aren't convinced about their current choice, and b) haven't ruled out voting for another party. Depending on the circumstances, you might be more interested in your opponent's weak voters who would potentially vote for you, or your own voters at risk of being wrested away by your opponent. But once you've isolated your targets, start filtering out what the rest think. Right now, Theresa May rightly thinks that she should be targeting huge chunks of Labour's vote. That's one reason her attack is so narrowly on Corbyn: she would risk alienating some of her targets if she attacked Labour and its values more broadly. That narrowness is a signal of her ambition.
Second, your subject. You have to find out what your targets care about: you might have a brilliant message on class sizes in primary schools but if the people you want to win are more anxious about the NHS, find something on that instead. The Tories have decided that their winnable voters care about leadership and Brexit. There's no doubt their research is telling them this. So even when they talk about other things, they will still be talking about leadership and Brexit.
Third, what options do you have for what you will actually say? Write it all down and test it. Your tests can range from the macro (do we advocate a second referendum? Quitting the European Court of Justice?) to the really quite micro (does "strength and stability" or "capability and competence" work better?). It all gets put through the wringer – in focus groups, in polls, in A/B testing on social media.
Fourth, how do we do against the other side? Messages must be tested against what other parties and candidates say: if your target likes what you say but likes what others say even more, you'd better change what you're saying. Another trap is to find a message that works wonderfully just so long as no one challenges it. Fail to stress test your message, and this can happen.
By now, most of the heavy lifting should have been done: the parties know who they're trying to reach, which of their buttons to push, how to package their offer and how to fend off attacks. They'll track how they're doing, especially in marginal seats, run quick tests on how to respond to ups and downs in the campaign, and look to tweak their messages for a last late push. But the core message is already brightly polished and ready to spring to life.
I might have given the impression that campaigns and the research that goes into them are a cynical exercise, all about politicians telling voters what they want to hear. I know plenty of people think that – they say so in focus groups. The reality is that our role is not to tell candidates what to say to make people like them, instead, it's to show how they can talk about their own ambitions for the country in a way that people understand and relate to. We like to think that this contributes to the democratic process.
Peter McLeod is vice president of polling firm GQRR. GQRR was pollster to Labour at the 2015 election but does not have a relationship with the current leadership.
The opinions in politics.co.uk's Comment and Analysis section are those of the author and are no reflection of the views of the website or its owners.