Who would win a general election?

This page first posted 5 September 2019

With a general election looking imminent, our standard hypothetical question becomes particularly relevant.

This analysis looks both at the headline prediction and at the details and assumptions to assess the chances of the possible results.

1. Headline Prediction

With current polls (as at 4 September 2019), the Conservatives have a lead of about 8pc over Labour. The size of their lead is a key measure of electoral success. As a rule of thumb, the lead needs to be 6pc just to get a bare majority in the House of Commons, and around 9pc to get a healthy majority. The current lead is inside that range, so that the central prediction is for a moderate majority.

The corresponding Electoral Calculus prediction, as at 5 Sep 2019, is:

Party2017 Votes2017 SeatsPred VotesPred Seats
CON 43.5%318 33.3%350
LAB 41.0%262 25.0%193
LIB 7.6%12 17.9%34
Brexit 0.0%0 12.9%0
Green 1.7%1 4.7%1
SNP 3.1%35 3.6%51
PlaidC 0.5%4 0.6%3
UKIP 1.9%0 0.5%0
Other 0.7%0 1.5%0
DUP 10 9
SF 7 7
Alliance 0 1
NI Other 1 1

Prediction based on opinion polls from 26 Jul 2019 to 31 Aug 2019, sampling 12,603 people.

With a lead of 8pc, our calculations indicate a Conservative majority of 50 seats. But only a small variation in support could change that significantly in either direction. A swing of just 1pc from Conservative to Labour would wipe out the majority, while a similar swing in the opposite direction would boost it even higher.

This UK-wide prediction is also based on recent polling for both Scotland and Northern Ireland by YouGov and LucidTalk respectively.

2. Prediction Methodology

Predicting the next election has been made more complicated by the emergence of the Brexit party and the surge in support for the Liberal Democrats. That raises the key question of how their supporters are spread geographically. If they are evenly spread around the country, then they will not win many seats. But if they are clumped together in areas of strength, then they would win more seats.

To answer this question, Electoral Calculus uses advanced statistical techniques called regression to estimate voters' behaviour in each area of the country and in each parliamentary seat. Although the details can be complicated, the basic idea is straightforward.

We start with opinion poll data which says how each respondent intends to vote, along with their demographic characteristics such as gender, age, occupation status, education and past voting history. The regression technique then analyses all these data to discover how people with particular characteristics tend to vote on average. It will discover, for example, how 35-44 year olds in the north east tend to vote, how women with university degrees tend to vote, and so on.

Once the regression has learnt how people behave it then applies that to individual seats. In each seat we use census data to see how many 35-44 year olds there are, how many women, how many people with university degrees, and so on. The regression then tells us how those groups tend to vote on average, which lets us build up a total picture of the constituency's likely voting patterns.

This method has been extensively tested by Electoral Calculus using historic polls from the 2017, 2015 and 2010 elections. In every case, the regression method was more accurate (or as accurate as) classic polling methodology. You can read that analysis here.

It has also been used in action in the real world for the European elections. Regression methods successfully predicted the Conservative and Brexit parties, along with detailed predictions about the relative strength of the Liberal Democrats and Greens in every region, although the scale of Labour's collapse was under-estimated. For example it correctly predicted that the Liberal Democrats would get more votes in London than the Greens, despite noisy challenges to that statement from local Green supporters. Our track record page for the EU elections has the details.

For the general election, we use a regression which is applied to recent polling data to create a baseline prediction, which captures the geographic variation of party support across the country. For very recent updates, a uniform swing using our strong transition model is applied to the baseline to estimate the most up-to-date support for each party in every seat.

3. Regional patterns

Given the geographic power of the regression method, it also lets us see what is happening in each region of the country. The table below shows how many seats each party gains or loses (relative to the 2017 election result) in each area of the country.

Predicted change in seats in each region

North East3−3000
North West6−9300
West Midlands7−7000
East Midlands6−6000
South West−4−3700
South East0−3300

Prediction based on opinion polls from 26 Jul 2019 to 31 Aug 2019, sampling 12,603 people.

The Conservatives are set to lose many seats in Scotland to the SNP, but they may make good gains from Labour in the Midlands, the North of England and Wales. Labour also loses most of its seats to the SNP and, in addition to the Conservative losses, it also loses seats to the Liberal Democrats in London and the South of England. This pattern is consistent with the Conservatives picking up Leave voters in Wales and north/central England, and the Liberal Democrats picking up Remain voters in southern England. Labour's Brexit strategy of triangulation between Leave and Remain appears to be falling between the two stools.

4. Pollster division

The divergence of pollster opinions continues. Using the most recent poll from each of eight pollsters, we see that there still appears to be two groups: one with a large Conservative lead, and one with a smaller lead. The "big lead" group contains YouGov, Ipsos-MORI, Kantar and Deltapoll. The "small lead" group has ComRes, Opinium, BMG and Survation.

You can see the original analysis of this (23-Jul-2019), but here is an updated version of our divided pollster graphic:

Con/DUP majority of 12

Party'Small Lead'
CON 31.3%322
LAB 25.8%220
LIB 17.2%35
Brexit 14.8%0
Green 4.3%1
SNP 3.6%51

Prediction based on 4 opinion
polls from 9 Aug 2019 to 30 Aug 2019,
sampling 6,568 people, from ComRes,
Opinium, BMG and Survation.

Conservative majority of 112

Party'Big lead' pollsPredicted
CON 35.5%381
LAB 24.1%164
LIB 18.7%32
Brexit 10.9%0
Green 5.1%1
SNP 3.6%51

Prediction based on 4 opinion polls
from 26 Jul 2019 to 31 Aug 2019,
sampling 6,035 people, from YouGov,
Ipsos-MORI, Kantar and Deltapoll.

The 'low' group has a Conservative lead of around 5pc, which isn't quite enough for an outright Conservative majority. On these exact figures, another DUP arrangement would be possible, but their total majority would only be 12 seats. And a small deviation from that result due to campaign swing, poll error or model error could easily obliterate the majority and deliver another hung parliament.

However the 'high' group implies a very different result. If these pollsters are correct then the Conservatives have a lead of 11pc over Labour and they could have a landslide victory with a majority of more than a hundred seats.

Since it is not known which group of pollsters is correct (or least wrong), the uncertainty in the election result is significantly increased.

5. Conclusion

Our central forecast, using modern regression methods and recent polling, suggests a Conservative victory at a general election with a moderate majority. But the pollsters are divided and the actual outcome could easily be anything from a hung parliament to a Conservative landslide.

Other uncertainties include how public opinion will react to the this week's parliamentary manoeuvres, changes in public support during the campaign, polling error and model uncertainty. And the situation could change dramatically if electoral pacts were made between parties.

We estimate the chance of a Conservative majority at 44pc, with Labour much less likely to get an overall majority. This gives the Conservatives a fair chance of victory, but much less than a certainty.

Return to Psephology home page.