A Snapshot, Not a Prediction

James Butler

YouGov has released the results of its MRP (multilevel regression and post-stratification) model. It shows the Tories taking 359 seats, Labour reduced to 211 – losing significant heartland constituencies to the Conservatives – and Boris Johnson winning his party’s first substantial majority, of 68, since Margaret Thatcher’s in 1987 of 102. The political impact of such a result is hard to overstate: it would mean a reworked Conservative Party in parliament, singing in unison from Johnson’s hymn sheet, clearing the way for his Brexit deal to pass without a hitch – along with everything else in the manifesto.

As a prediction of the final result, the model is wrong. That’s not a polemical claim: YouGov is at pains to point out in all its literature that the model is intended as ‘a snapshot, not a prediction’. MRP works by

using an extremely large number of interviews to model people’s voting preferences based upon their demographics (age, gender, education, past vote and similar factors) and the local political circumstances (such as whether they live in a Conservative or Labour seat? Is a pro-Brexit area? Is there an incumbent MP?)

It does not take account of specific local factors: a strong independent candidate, say, as with Dominic Grieve in Beaconsfield; or a newly unpredictable local dynamic, as in Labour-held Kensington, which faces a strong campaign from the ex-Tory Lib Dem candidate, Sam Gyimah. It is also hard to tell whether vocally pro-Brexit Labour MPs in Leave-voting seats, such as Caroline Flint in Don Valley, will do better at rebuffing Tory advances – though candidate loyalty is rare in British politics.

Caveats are needed because the MRP method has taken on a near-mythic quality among journalists since the 2017 election, when YouGov’s model – applied to fieldwork undertaken much closer to polling day – accurately predicted a hung parliament, unlike most conventional polls. It actually underestimated the final Conservative tally by 16 seats; the other notable 2017 MRP model, Lord Ashcroft’s, predicted a Tory landslide. The MRP’s headline figures this time are closer to those predicted by conventional polls, as their methods and assumptions have been updated. This is, in part, normal scientific re-evaluation, but it also reflects a pervasive anxiety among pollsters, who been wide of the mark in two general elections running. Much of the current conventional modelling uses weighting to correct errors or bias in the samples – but as the scope of the weighting grows, so does the scope to introduce new bias and error.

The anxiety permeates rather deeper among Labour activists, though alloyed with determination and hope. Some believe that the YouGov fieldwork took place too early to capture Labour’s recent, if gradual, rise in the polls. Others point to the late surge in voter registration – 3.8 million by the deadline, 67 per cent of them under 35 – as a sign of a newly engaged youth vote, likely to favour Labour. If figures from the last election are a guide, about a third of those will be already listed voters re-registering, just to be sure; but it may still be an indicator of high turnout, and it is an article of faith among Labour activists that higher turnout tends to favour the left. YouGov’s MRP model assumes a relatively conventional turnout.

It can be hard to see the boundary between warranted scepticism about the YouGov figures and wild-eyed grasping at straws; however sceptical one might be, the model ought to be sobering for anyone who dreads another five years of Tory rule. The brutalities of first-past-the-post become plain: outside a few Tory/Lib Dem marginals (Richmond Park, St Ives, Cheltenham), a vote for the Liberal Democrats will simply rubber-stamp the Tory victory – and YouGov suggests a Liberal net gain of only one seat. Green votes in marginal seats with Labour incumbents – such as Stroud – risk defenestrating MPs committed to a second referendum and a Green New Deal, and replacing them with Tory MPs signed up to Johnson’s Brexit pledge. One may meditate on the virtues of electoral reform, one may wish it otherwise – but it is the way it is.

YouGov’s snapshot will not have been greeted with jubilation at CCHQ, however. Above all, they fear complacency among their voters: the expectation of a comfortable majority as an excuse for indolence on polling day, especially in the cold and damp of mid-December; or a working-class Brexit supporter forgoing the guilt and shame of voting Tory in the belief that Johnson has it sewn up anyway. Thus Dominic Cummings rose from his dormancy yesterday to exhort the faithful that the race was closer than it appears, and warn – less with a dog whistle than a foghorn – of the dangers of Labour importing foreign voters, or ‘the likes of Goldman Sachs’ writing cheques to fund an anti-Brexit insurgency.

There are deeper reasons for Tories to worry about the headlines that already proclaim Johnson triumphant. Having siphoned back most of the Brexit Party vote, the Conservatives seem to be approaching their electoral ceiling; they have been more successful in hegemonising the leave vote than other parties have the other side. But YouGov’s model also suggests that should the gap close by a few percentage points, the majority might easily crumble back to a hung parliament. There are 67 seats in which Labour and Tory candidates are separated by five points or less.

However forceful YouGov’s insistence otherwise, its model was widely reported as a prediction. It will now cause political feedback of its own in the election cycle. This is amplified by the dominant model of political journalism: journalists who think of themselves as insiders are usually keener to discuss poll movements and voter behaviour than they are to explore the different implications of the political choices now laid before the electorate. Our atrophied public sphere aside, the model will also prompt all parties to shift their campaigning towards defending vulnerable seats.

Labour members long sceptical of the argument that the party would sweep all before it simply by embracing an outright pro-remain position will feel vindicated. The regional manifestos unveiled last night will give campaigners resources to argue for the concrete transformation that the party is promising. It will also lean hard on making the election more and more about the NHS. Attempts to translate the political and constitutional question of Brexit into wider issues of distribution and investment can only ever partially succeed: it can’t afford to let its second referendum offer slip into the form favoured by ultra-remain partisans, as a de facto policy to overturn Brexit. The offer can only be that remainers will have a chance to lay their arguments before the country again; the option to leave must still be plausible, and its implementation guaranteed if it wins.

There are reasons for optimism of the will among Labour activists: the party has half a million members, and is the largest left-wing party in Europe. It has a manifesto that offers profound and real change. The polls are showing a steady, but not sharp, Labour gain. The Tory vote has probably hit its ceiling. And there are still 13 days left in the campaign.