Dr Patrick English is the Director of Political Analytics at YouGov.
It’s time to talk seriously about the range of possible outcomes implied by current polling.
For this piece, I will focus exclusively on MRP projections, which take large-scale samples of vote intention from nationally representative samples and project them down to the constituency level to help us better understand what is happening seat-to-seat, for each party standing there.
MRP polls have, to date, been produced by each of YouGov, Focaldata, Savanta, Survation, More in Common, IPSOS, and Electoral Calculus. At least one more organisation is expected to join this particular list before election day.
The table below briefly summarises where each pollster thinks the race currently stands for both the Conservatives and Labour.
Pollster | Con Seats | Lab Seats |
YouGov | 108 | 425 |
Focaldata | 110 | 450 |
Savanta | 53 | 516 |
More in Common | 155 | 406 |
IPSOS | 115 | 453 |
Survation | 71 | 487 |
Electoral Calculus | 66 | 476 |
Given the above range of stated outcomes for both main parties, some election commentary has suggested that pollsters, and MRP pollsters in particular, disagree about the outcome of the coming election on July 4th. This, however, is not particularly the case.
Two truths can describe current MRP polling:
- All MRP models project a substantial Labour victory
- There is a healthy range of estimates for the size of the Labour majority
While the projected Conservative seat count comes in from anything between 53 (Savanta) and 155 (More in Common), and the estimated Labour number goes from 406 (More in Common) to 516 (Savanta), the story is the same between all seven: a (near) record-breaking Labour win, and a record-low Conservative seat tally.
All models are united in their anticipation of Keir Starmer walking into Number 10 on July 5th with a whopping majority, they just differ as to quite how big that majority will be.
The reason for this is that we are facing an election with an extraordinary rate of change. The two-party swing between Labour and the Conservatives is estimated to be anything up to around 17+ points, handily beating the post-war record of 12 points in 1945.
In one election cycle, we are going from Labour’s worst defeat since 1935 to what the vast majority of pollster expect to be their best-ever win in their history.
The record-breaking nature of the coming swing far eclipses anything for which we have any historical precedent to help us model. We as pollsters simply have no idea how a swing like that will fall out in terms of seats at a British general election – we have nothing to help or guide our expectations.
The 1945 result is the closest we have to a parallel to what we expect to happen on July 4th. That historic election, held on the 5th of July, saw Clement Atlee’s Labour defeat Winston Churchill’s Conservatives by a vote share margin of 50 per cent to 36 per cent, winning 393 seats to the Conservatives’ 189.
The 1945 election provides us with one of the only examples we have in British politics of ‘proportional swing’ – whereby a party, in this case the Conservatives, sees its support drop most where it started strongest.
Proportional swing is very dangerous territory for any party to find themselves in, and it means that far more seats are vulnerable to loss than they are under uniform swing situations. The reason is quite simple.
Let’s assume Party A has 58 per cent of the vote in Seat X, and 40 per cent of the vote in Seat Y. Party B has 35 per cent of the vote in Seat X, and 32 per cent of the vote in Seat Y.
Let’s also say that the two-party national swing is 10 points against Party A, and that it applies ‘uniformly’ across all seats.
That would leave Party A on 38 per cent in Seat X, and 30 per cent in Seat Y, with Party B on 37 per cent and 42 per cent respectively. Party B therefore takes Seat X, but not Seat Y.
Uniform national swing is the most consistent and accurate account of how vote shares tend to change across constituencies in British politics: if a given party is down 5 points nationally, they tend to be 5 points down (or thereabouts) in most seats, and a simple ‘UNS model’ can tell you how many seats that will mean they win (using their previous result as a baseline).
But now let’s imagine that the national swing is still 10 points, but that it is a ‘proportional’ swing right across the country. Party A is dropping from 40 per cent nationally to 30 per cent, while Party B is increasing from 32 per cent nationally to 42 per cent.
If the swing for both parties is completely proportional, that implies that Party A is losing ¼ of their overall vote in every seat (rather than simply -10 points) and that Party B is growing their vote by a third (rather than simply +10 points).
Returning to Seat X and Seat Y, this would imply that Party A dropped down to 43.5 per cent in the former and 30 per cent in the latter. Party B would be up to 47 per cent in the former and 43 per cent nationally. In this instance, Party B takes both Seat X and Seat Y.
We do not know the extent to which the swing will be proportional on June 4th. In some elections – specifically in 1997 – the two-party swing was somewhere in between uniform and proportional. In others, for example, 1945 and in the Scottish results in 2015, it was extremely proportional.
If the swing is generally more uniform than is currently expected, we would be looking at Labour being much closer to 400 rather than 450 without changing a single individual vote share estimate.
But if the swing is more proportional than that, and gets toward 1945 levels, then the Conservatives winning 50 seats is a distinct and strong possibility.
And in many ways, that explains much of the difference between the MRP projections; we are each anticipating a different shape to the swing across the country, and a different pattern of results therein.
But as well as assumptions or expectations about swing – for which we have little to no historical evidence to go on – all kinds of model choices and user decisions are made which can guide the estimates in one direction or another.
At this stage, with so many unknowns and so much about this election and the magnitude of the changes coming that could mislead or misdirect pollsters, it is wise to treat the full range of MRP projections as plausible and realiastic outcomes.
The post Patrick English: How the proportionality of swing can turn a big Conservative defeat into a wipeout appeared first on Conservative Home.
Click this link for the original source of this article.
Author: Dr Patrick English
This content is courtesy of, and owned and copyrighted by, http://www.conservativehome.com and its author. This content is made available by use of the public RSS feed offered by the host site and is used for educational purposes only. If you are the author or represent the host site and would like this content removed now and in the future, please contact USSANews.com using the email address in the Contact page found in the website menu.