Poll Projections: the Clock Strikes Midnight on Jeremy Corbyn’s Cinderella Story

Update (Thursday 8th June, 4am BST): Final Seat-by-Seat Projection UK General Election 2017

After erasing as much as twenty percentage points from Theresa May and the Tories’ lead over his Labour party, Jeremy Corbyn’s Cinderella run looks set to run out of magic just before the clock strikes midnight on Thursday. CounterPunch projects that while Labour will likely close the national gap to just three points (41% Conservative 38% Labour), Theresa May should nevertheless retain a slim majority of parliamentary seats. The collapse of the UK Independence Party (UKIP) to 4% or less in the wake of Brexit combined with likely and potential Conservative gains in Scotland at the expense of the Scottish National Party (SNP) are of critical importance. Together, they indicate a hung parliament with a chance of Corbyn as Prime Minister just out of reach. Two dozen or fewer seats could change parties, leaving Parliament looking very much the same as it did when the election was called seven weeks ago. A large number of Toss-Up calls in my seat-by-seat projection, however, does hold out some hope that undecided voters breaking Labour and a very strong youth turnout could see May lose her majority.

FiveThirtyEight‘s Nate Silver, in one of his better articles over the last two years, lays out the reasons why normal British polling errors could see Thursday’s vote fall anywhere along the line from a three point Labour win nationally to a massive seventeen point Conservative win closer to initial predictions weeks ago. Various polls in the last several days have come out along almost the entirety of that range. Ipsos MORI’s latest poll [pdf] had a three point Labour lead before its likely voter adjustments. While their strictest likely voter model saw Conservatives rise to a five point lead, their 7-10, self-reported likely voter model still shows a three point Labour lead. ComRes, meanwhile, published numbers [pdf] for its strictest likely voter model with undecided voters pushed for an answer. That “PM Squeeze” model showed a 14 point Tory win. Silver further noted how difficult it has proven to translate national vote share in polls into an accurate prediction of parliamentary seats, a translation obviously far more critical to determining the outcome of a national election in the UK than, say, the outcome of 435 House of Representative races in the United States.

My simple model for CounterPunch outperformed FiveThirtyEight, Huffington Post, RealClearPolitics, the New York Times, Daily Kos, and Princeton Election Consortium ahead of the U.S. election in November, both calling more states correctly and coming closer to the final national voting margin between Clinton and Trump. As in the run-up to the U.S. election, I have kept trendlines for a simple average of the latest poll from all pollsters in the field in the last 10 days. Wherever possible, I have used the figures from each pollster before they stripped Undecided/Don’t Know/Refuse voters from the model. My analysis of the importance of undecided voters here helped me to project, well ahead of the curve in early May, that the Tory lead was slouching toward a single digit. I have also kept a trendline more responsive to the latest polls, clustering them by dates and posting an update to Twitter nearly everyday. The “Last 10” model currently sits at a 4.1% Conservative advantage. The more responsive “Clusters Model” is at 1.8% points, with the trendlines leading me to take the mean, a 3% Tory lead, for my projection two days out. If there are significant changes, I will post final national numbers with a link to a pdf copy of my seat-by-seat projections here just before noon London time (7 a.m. Eastern for North American readers) on Thursday.

I have combined insights from this model with lessons learned from observing Canadian Broadcasting Corporation elections forecaster Éric Grenier over the years. Grenier very accurately forecast seats based on polling for British Columbia’s election in early May. In this UK election, a would-be forecaster has the advantage of two seat by seat models based on tens of thousands of polling responses over the course of the last week and a half. Beyond national, regional, and even sub-regional considerations, I have taken the average of Ashcroft polling and YouGov’s Crunch model. I have then applied small regional adjustments where needed based on the average of regional polling and regional sub-samplings from national polls and have given some weight to particular issues or candidates in a handful of individual parliamentary constituencies. I have also done what I could to factor in the high number of remaining undecided respondents, especially those who voted UKIP or Liberal Democrat in 2015. (Liberal Democrats are projected at an 8.5% national share.)

The resulting convergence names twenty seats most likely to change hands (see chart above) and an additional thirty-nine seats that are too close to call (Toss-Ups) with a projected lead for one candidate or another of less than three percentage points. Fifty-two additional seats are projected as leaning (a three to six point advantage) or likely (between a six and ten point advantage) to remain with the incumbent party. In the event of a far stronger performance by Labour or Conservative, I will be keeping my eye on an additional thirty seats or so that could be bellweathers of an unexpected landslide. The major difference between my model’s outcomes and YouGov’s is in the East of England, the North West, and especially in Yorkshire and Humberside. Where YouGov currently projects a fourteen percent Labour win across the North, the average of subsamples from other pollsters with regional breakdowns shows just a four to five point average Labour win across the North. My model settles upon a seven to eight point Labour win, with 40% of my Toss-Up projections coming from the North West or Yorshire and Humberside. While I have a substantial number of differing seats in my model, Martin Baxter’s Electoral Calculus can be made to show what an overall picture looks like where Conservatives win by just three points nationally and maintain a small majority. Were all the Toss-Ups, including for instance Dewsbury’s showing of just a 0.1% lead for Conservatives, turned into strict projections, Conservatives would gain something like one more seat overall with Labour gaining eight. Attempting such a level of specificity with very little polling data by constituency is a fool’s errand, requiring rather the recognition of inevitable flux at the margins.

As with any election prediction, there is much that could go wrong for my model or that could simply change in the next forty-eight hours. Final polling is in the field as I write with swirling uncertainty in the wake of a second disgusting terrorist attack this election cycle, this time on London Bridge over the weekend. There is a healthy debate amongst pollsters regarding how many enthusiastic young Corbyn voters will turn up to the voting booth, with suggestions ranging from somewhere in the forty percentage point range to over eighty percent. About twelve percent of the voting population remains undecided at this point (down from regular 15-20% Don’t Know/Undecided/Refused numbers throughout the last month). YouGov has argued that as many as a third of that 12% simply will not vote. How the other eight percent break could matter enormously. Scotland’s three or even four-way, often razor tight, races in a dozen or more constituencies could well determine whether Conservatives do in fact retain majority status. Finally, there are the questions around whether pollsters as a whole have fixed problems that persistently underestimated the Conservative vote by about three points on average over the last several national election cycles.

I will be liveblogging commentary on results, beginning with the exit poll at 10pm (5pm eastern) on Thursday here.

If pollsters have fixed their problems and I have accurately interpreted the data, Corbyn will have seriously defied the odds over the course of this election and the expectations of those who have frontally opposed him from his own party and in the supposedly left-of-centre media in Great Britain. Still, absent further last minute magic, I expect Corbyn and Labour to fall short in their quest to end Tory majority rule. Nevertheless, the results should be quite enough to allow Corbyn to stay in leadership for Labour, awaiting a fitting of the Prime Ministerial slipper sometime in the future by the Queen or, perhaps in five years, by the anti-Prince Charming, Charles.

Click #GE2017 Seat-bySeat Projection for a more manageable PDF version of this seat-by-seat projection for 111 Parliamentary Constituencies:

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s