The only certain thing in this tight election is that all the polls are wrong

The point is - I don’t know which way they will go and neither do any of the polls.

Much was made of YouGov’s recent MRP poll predicting a Conservative majority of 68, which moved the money markets. MRP is a new methodology that in 2017 was ‘the only poll to correctly predict a hung Parliament’. Yet YouGov’s prediction on the night before the election was: “Tories lead by seven points and set to increase majority”. Well, that was the result from another poll using a conventional methodology. So, if I had conducted two polls with different predicted outcomes I would be able, like YouGov, to retrospectively choose the one closest to the result and trumpet how clever I was.

The MRP survey is nonetheless an impressive beast, based on a sample of over 100,000 people. YouGov then apply their own ‘magic sauce’ called ‘Multilevel Regression and Post-stratification’. No wonder they shorten it to ‘MRP’.

However, MRP’s sample size is still too small to offer predictions at constituency level. And that is the problem, because to be as precise as a ’68 Conservative majority’ that is exactly what they have to do. So, let’s take Finchley and Golders Green. Mike Freer, the Conservative candidate, is a popular, hard-working local MP. But to be 18 points ahead of the well-known Luciana Berger, as YouGov have him, stretches all credulity.

MRP is now in vogue with other pollsters such as Best for Britain and Datapraxis whose own version (based on YouGov polling of 500,000 people) in the Sunday Times has a Conservative majority of 38. But for the uninitiated, all pollsters undertake what is called ‘herding’ where they amend their results to fit in light of other polls. For example, in 2016 a ComRes poll predicted the Referendum outcome, but they changed it to a Remain win.

The LSE has just released an analysis of the YouGov MRP poll with the ‘poll of polls’ and we see this ‘coming together’ of the results. In his analysis, the difference between the two results is no more than eight additional seats to the Conservatives.

The problem with predictions

My prediction for this election is that anyone who thinks that they can predict it is a mug. I say this with the track record of correctly predicting 8/10 recent elections including the Scottish referendum (“the Yes vote would get less than 45%” – it was 44.7%), Brexit “vote leave will get 52%” – it was 51.9%) and the 2015 election (“Conservatives will win in single figures” – they won by 6 MPs).

The problem in 2019 for all polls is Labour’s Leave/Remain policy ploughing up the traditional political field so that the permutations are now too numerous to forecast. In October another YouGov poll showed that a majority of people over 30 were going to vote Lib Dems or Green rather than Labour. This is unheard of and would result in many Conservative candidates coming ‘through the middle’ to win.

However, a few weeks later and the Lib Dem vote is crumbling. The Conservatives greatest fear was the Remain vote coalescing around Labour and that seems to be what is happening. According to Best for Britain there are 57 seats where tactical voting by pro-EU candidates could alter the result. There is even an app, Swap My Vote, to help facilitate this.

Another factor is the impact of the Brexit Party. On the face of it, its decline in the national vote share looks substantial since the election was called. But this is a consequence of Farage standing down candidates in 317 constituencies. The impact of the Brexit Party splitting the Leave vote remains potent in Labour constituencies that voted Leave.

Even if pollsters methodologies are super accurate it still depends on people telling them the truth. In the past pollsters blamed ‘shy Tories’ for why they got the results wrong. I have it on good authority that Momentum has briefed its members to say they are voting Conservative when polled to confuse their enemy.

So, everything is up in the air. People are reacting cynically to Dominic Cummings’ declaration that the election is too close to call. However, I believe his is the only prediction worth heeding until we see the results of “the only poll that matters”: the election this Thursday.