The first of what seems likely to be many books about the June 2017 general election was released earlier this week. Betting the House, by Tim Ross and Tom McTague, tells the story of the election through contrasting accounts of the Conservative and Labour campaigns. There’s much more detail about the former, seemingly reflecting both the interests of the authors and the potential for access to political figures keen to settle scores after being involved in such a car crash.
One of the most interesting things is how clearly the failures of political research methods (viz polling and modelling) played a role in the outcome. On loc 1857 they describe how the speed with which sentiment changes within the electorate undermine expectations formed on the basis of political polling:
Pollsters take their raw research findings and then ‘weight’ them against the likelihood of different respondents to turn out on polling day and vote the way they say they will. It was these calculations that led to variations between the pollsters during the campaign. One of the key questions was how likely they believed young people who said they would vote Labour were to make it to a polling station and cast a ballot on 8 June. Then there was the speed of the changes in support for Corbyn. It was a fast-moving phenomenon which took off at a time when campaigning had to be stopped twice because of terrorism. The sharp tightening in the polls came largely after the launch of the Tory manifesto, which proved to be one of the most influential policy documents in recent election history. The effect of that manifesto, and the U-turn that followed, on the Conservative lead was like a gust of wind on a house of cards.
On loc 1874 they give an overview of the modelling activity which the parties engaged in, as well as the data sources upon which they drew to undertake this:
This is where the mysterious wizardry known as ‘modelling’ comes in. There is a wealth of information available to political parties –some free, much paid for –which they can use to identify the swing voters they are most likely to win over. These are the voters who decide elections. They may have voted a different way at the last election but ‘modelling’ helps parties understand which arguments will be most likely to persuade them. The exercise is highly complex, meaning it can go wrong. All mainstream parties model their potential voter types; it is simply sensible market research. They use a combination of data from Facebook, the electoral roll and the credit checking agency Experian. This information reveals what specific voters in specific target seats are like. Reliable polling evidence, broken down by age, social class, gender, education level and other factors, can then suggest how people with particular characteristics might vote.
The underlying assumption here is that past behaviour can be an adequate basis upon which to make predictions about future actions. In many cases, it can be, in spite of the many extrinsic factors which can impinge upon the replication of the past. However, if we are seeing an increase in the speed at which changes in the electorate occur then modelling is going to represent a decreasingly reliable basis upon which to assign limited resources within a campaign. This is something which, as the authors point out on loc 1899, the tactics of Momentum were not vulnerable to:
The one organisation which did not use micro-targeting was Momentum. They simply bombarded their target seats with activists in the hope of persuading people to vote for Corbyn. In doing so, they knocked on doors none of the main parties were bothering with. Even internal Tory modelling experts now question the value of what they were doing. ‘Maybe Corbyn’s plan to build a big groundswell of support, ignoring the seat-by-seat numbers etc., is the right way to go,’ says one. ‘How do you ever factor in for that? This is what happened with Trump, this is what happened with Brexit. People voted who you never expected to vote. How do you work out a way to tackle that?’
This is a complex issue which I’ve barely scratched the surface of. The core question is straight-forward though: are some political tactics more adaptable to intensified social change than others? What does this mean for broader questions of political strategy?
Mark Carrigan 