The political class is addicted to polling, but asks all the wrong questions

WhatsApp groups across SW1 light up; thumbs paddle at phone screens as columnists and policy advisors scrabble to reverse-engineer an explanation. They look back over the week. Perhaps, someone suggests, it’s because Ed Miliband was on This Morning? Swing, and a miss. Retrofitting an answer tells you nothing. The truth is, in looking backwards, you’ve already missed the boat. 

It’s odd that we live in a political culture saturated with data, yet we’re often surprised by voters. The political class constantly measures opinion but seems to learn very little from what it finds. Polling is treated as punditry, not insight. As the country's mood under Burnhamism shifts, and the dominant narrative becomes listening to those who have long felt ignored, then the way politicians ask questions of the public can’t stay the same either. It must change. 

The fact is, everyone uses polling to prove their hypothesis, but no one uses it to create hypotheses. People start with an assumption, and then commission or find a poll to support it. In my experience, when working with the policy world, people come to Early Studies wanting to prove a point they’ve already made. But what’s the real value in that? Polling is an undeniably powerful tool, but we’re drowning in the wrong kind, used in the wrong way, for the wrong purpose. 

I should be clear that I’m not anti-polling and I'm not anti-message-testing. Both have their place. But there’s an imbalance. A huge investment at the validation stage of polling (“does this slogan land?”, “will this stat support our case?”), and almost none at the discovery stage. There’s so much valuable data there: what people are genuinely worried about, what they’d be embarrassed to admit matters to them, and which local issues shape their daily lives. We’re just not getting at it. 

I discussed the topic with a Lib Dem candidate at the 2024 election, who told me he’d spent a while walking around his constituency asking people what they thought others secretly cared about but might not say out loud. The concerns that came back weren’t immigration, healthcare, or even crime. Overwhelmingly, constituents spoke of embarrassment at closed shop fronts, frustration over local infrastructure developments, and annoyance at the stalling of long-promised estate redevelopment.

This might seem obvious, but when the candidate included these three issues in his town hall agenda, there was a huge, positive reaction. People felt heard. In fact, these issues resonated so much that the rival candidate – adopting his policies and campaigning on those three issues – won out. Such is the dog-eat-dog world of politics. But the point still stands: the research found the real agenda, and the issues it identified won the race. 

With the right kind of polling, there’s a real possibility for politics to become more grounded in lived experience

It’s worth saying that polls are traditionally just seen as measuring public opinion – but they can move it, too. Voting intention figures, released close to an election, can have a real impact. Whether that’s making voters stay home in the face of a sure defeat, or tactically changing votes, published polling is a political force. When that’s combined with an imperfect tool, it begins to look misapplied. We assume polling figures are static representations of the nation’s feelings at a particular moment in time. But put the same question to Britain, phrased five different ways and using five different methodologies, and I’m confident you’d find yourself with five different answers at the end of the process. 

So what does better polling look like?

The first principle is to conduct the research before the policies are written, and not after. Public opinion should be a tool in co-creation, not just for post-mortem.

The second is to use the right methods to generate accurate, truthful results. When people answer questions on behalf of people they know instead of themselves, the pressure to perform or give the “right” answer dissolves. As a result, people answer more honestly. Academics have shown social circle surveying increases accuracy, even in the tricky art of predictions polling. Asking people to consider how they previously felt or expect to feel in the future can be enlightening, and asking people what they are afraid to admit can produce campaign-defining data. These are practical ways to cleanse a data set of the projections and biases that can steer results. 

With the right kind of polling, there’s a real possibility for politics to become more grounded in lived experience. Next time the voting-intention results land, don’t just use them to confirm some preconceived bias. Take a beat and ask, instead, what it isn’t telling you.

Alfred Malmros is co-founder of Early Studies

Related

Third Of Renters Unable To Afford Food, New Research Shows

By Nadine Batchelor-Hunt

20 Oct