Toby Bevan
The Mother of All Polls: beyond prediction and into influence

Election 2019

Is YouGov’s MRP poll accurate and might it influence the course of this General Election? PB Consulting’s Toby Bevan gives his thoughts in the most anticipated poll of the election so far.

If there is one thing voters will have learned over the last four years, it is that you should never trust the polls. Pollsters have wrongly predicted the results of three of the most significant democratic processes held in recent times: Brexit, Trump and the 2017 UK General Election.

In each scenario, commentators and analysts have been forced to eat their own words, scrambling to find some reason or rhyme to seismic results that have irrevocably changed UK and US political landscapes.

It is now widely agreed that the Brexit referendum and Donald Trump’s ascendancy both came about as a result of the winning camps being able to target a forgotten voter cohort; a selection of the electorate that had felt marginalised by mainstream politics for some time. Pollsters missed their mark because their data didn’t account for the fact that this group of people even existed.

The mainstream argument for 2017’s shock result is that it came about as a result one very impressive campaign by the Labour Party, and a very poor showing from Theresa May. Yet, there was one poll, YouGov’s MRP poll, which did prophesise May’s self-inflicted disaster, at a time every other poll was suggesting the Conservatives would secure an overwhelming majority. Since then, this model of polling has been held in high regard and since the General Election was confirmed at the end of October, political commentators have waited with baited breadth for its release.

What is an MRP poll?
MRP polling, which stands for Multilevel Regression and Post-stratification, is a recently developed technique that aims to give more detailed prediction than a standard opinion poll. By using a much larger sample than a standard poll – around 50,000 people – MRP aims to identify lots of different types of voters – from the low-income working class voter who voted to leave the European Union, to university students whose politics are socially influenced; anti-Brexit and overtly liberal.

By correlating opinion to voter type, pollsters can then project this information onto each individual constituency in the UK giving a far more accurate understanding of the voter landscape. Imagine, if you will, the UK electorate is a bowl of soup. A simple opinion poll would take a spoonful and presume the rest of the soup tastes exactly the same. An MRP poll does nothing of the sort. It recognises that there are multiple behavioural and local factors that dictate the way someone votes. It understands the complexity of voting patterns and predicts accordingly.

The first MRP poll was released last week by The Times. It predicts that Boris Johnson is on course for a comfortable majority, suggesting he will win 359 seats, Labour 211, the SNP 43 and the Liberal Democrats 13. That result would give Mr Johnson a healthy 68-seat majority, with the Tories making significant gains in the Midlands and the North of England at Labour’s expense. With that said, the projected margins of victory is below 5% in at least 30 of the seats predicted to be Conservative. If Mr Johnson’s poll lead drops of 11% to 7%, he could be denied a majority.

It is important to remember that this poll model is not magic. It is only as good as the analysis so if pollsters inaccurately identify voter types, or fail to predict their voter behaviour, the results can be misleading. Furthermore, the data for this poll was recorded before the main party manifestos were released. The leaders’ debate had not happened nor had the Labour antisemitism row resurfaced. A week is a long time in politics and elections can turn on a knife edge.

Might this poll influence the election?
There is always a great sense of irony when a poll goes beyond its job of predicting elections and steps into the realms of having an influence on their final outcome. In this sense, the very act of publishing the MRP model might influence the result, thus ultimately proving it inaccurate. Voters have a tendency to change their voting behaviour if they think an election is going to be close or a full-gone conclusion.
This year, if remain-voting Conservatives have reason to believe the result will be close, they are likely to be spooked back into voting Conservatives by the prospect of Corbyn taking the keys to Number 10.

Many believe that this kind of behaviour played a key role in the surprise result of 2017. Back then, many soft Conservative voters felt the choice was between either a large Tory majority or a small one, and thus voted with more freedom. This time around, Boris Johnson and his team have been very clear to spell out a simple choice for voters; either vote Conservative or face a ‘coalition of chaos.’

Given all this, it is of no surprise that the Tories have been quick to play down the results of this poll. Boris Johnson’s chief aide and Vote Leave mastermind Dominic Cummings, has warned against complacency, writing on his blog: ‘Trust me, as someone who has worked on lots of campaigns, things are MUCH tighter than they seem and there is a very real possibility of a hung parliament.’ This messaging is set to continue given the memories of 2017 are still so raw.

At the same time and quite possibly off the back of the MRP poll, the Labour Party have announced they will re-shape their campaign strategy – particularly in Leave-voting areas – to try and turn the tide. If this is the case, then the nature of polling might not be the only thing changing. So too might be their purpose; shifting from predictors to influencers.