General election 2019: Who the latest polls say will win - and why the polling may be wrong

With just days to go until the UK votes in the third general election in less than five years, we are in the midst of peak polling frenzy.

There are new surveys published almost bi-weekly in an attempt to predict the outcome of this crucial vote - despite the fact it is widely considered one of the most unpredictable elections of recent history.

Read More

Read More
Conservatives urge Labour supporters in Scotland: Vote Tory to save the Union

Sign up to our Politics newsletter

Sign up to our Politics newsletter

Trust in political surveys has deteriorated in the past few years after the research repeatedly failed to accurately predict the outcome of elections.

Prime Minister Boris Johnson

Even so, it seems we can't help but cast at least one eye over the poll headlines.

How do polls work?

Opinion polls are used to find insight into public attitudes towards issues, primarily political, and are used to try and predict the outcomes of elections or referendums.

They are essentially scientific surveys carried out on a cross-section of the public which is specifically chosen to try and give an accurate representation of the whole UK electorate.

Labour leader Jeremy Corbyn at Forest Green Rovers

They can be done by anyone, but those considered to be the most reputable are normally carried out by the bigger pollsters - for example ComRes, YouGov or Ipsos MORI.

To ensure the poll is as accurate as possible, the companies adhere to strict rules about sample size, ensuring a random selection of participants and including margins of error in the results.

Most of the questioning is done via the internet where pollsters have collated large groups of people who form a panel. When they need to dip into a certain poll, it will select a specific quota sample from that larger group and then send them the survey.

Some polling is also conducted over the phone by carrying out random calls to collect demographic details of each caller in order to then match them with a quote sample.

How accurate are they?

It's difficult to produce an accurate poll because political opinions and voter intentions change regularly, especially during an election campaign when all parties are trying to throw mud at each other.

As we know - and have seen in the past two elections - many pollsters get it wrong, even though they are using scientific methods to collect their information.

There are so many varying factors - the size of the sample, the wording of the questions, the time of day people are contacted - but one of the most complicated aspects is the unpredictability of people. Something that it is hard to measure with a simple formula or method.

This has become even more complicated in recent years as more and more voters are less likely to remain loyal to a particular party and are becoming, instead, more volatile.

According to the British Election Survey, up to 40 per cent of voters changed allegiances in the 2015 and 2017 elections. More and more people are becoming so-called floating voters who are considerably harder to predict.

So... why bother?

In the 2017 election, only one poll accurately predicted the eventual hung parliament, which led to Theresa May entering a confidence and supply arrangement with the DUP.

YouGov used the MRP model that predicted the correct results in 93 per cent of constituencies and suggested a hung parliament despite all other polls pointing to a Tory majority.

The Multilevel Regression and Post-stratification model - to use its full name - was designed to produce a more accurate constituency-by-constituency prediction of the results rather than a national average.

This makes it easier to actually predict which seats are likely to change hands, which ultimately impacts the final Commons make up.

What is different about MRP?

YouGov interviewed about 100,000 people about their voting intentions for their 2019 poll, which included an average of 150 votes per constituency.

"The idea behind MRP is that we use the poll data from the preceding seven days to estimate a model that relates interview date, constituency, voter demographics, past voting behaviour, and other respondent profile variables to their current voting intentions," the pollster said.

"This model is then used to estimate the probability that a voter with specified characteristics will vote Conservative, Labour, or some other party."

They used data from the UK Office of National Statistics, the British Election Study, and previous election results to how many of each type of voter would be in each constituency.

They do point out, however, this is an estimate of current voting intentions and not a forecast of how people will vote, noting that people may change their minds.

The sample size in each constituency is also too small to be reliable by themselves which means the polling model also has to take into account patterns in the data across constituencies.

This, however, could lead to errors in seats where voting patterns are not normal. YouGov uses the examples of Beaconsfield - where there is a "high profile independent candidate" or Kensington where there is a "new pattern of local competition".

YouGov's constituency-by-constituency poll, released on 27 November, reiterated the prevailing trend by predicting a Tory lead and an overall majority, indicating that an election held then would see the Conservative Party win 359 seats, 42 more than they took in 2017.

This would leave them with a 68-seat majority, which would be the party's best performance since 1987.

Labour, meanwhile, are set to lose 51 seats, falling from 262 seats in 2017 to 211 now, and taking 32 per cent of the vote, a nine percentage point decrease.

This would be the party's worst performance in terms of seats won since 1983, YouGov said, adding the party are on course to not take any new seats.

Chris Curtis, political research manager at YouGov, said the analysis showed the Tories have a "comfortable majority", with seats coming their way at the expense of Labour in the North and Midlands.

"As expected, the key thing deciding the extent to which each of these seats is moving against Labour are how that seat voted in the European Union referendum," he said.

"In the seats that voted most strongly to Leave in 2016 (60 per cent or more in favour of Brexit), the swing to the Conservatives is over 6 per cent."

What do the latest polls say?

As the general election campaign entered its final days, the Conservatives have maintained a steady lead of around ten points in many opinion polls.

While the misleading forecasts of the 2017 polls means the Tories are taking nothing for granted, they have some grounds for optimism with the gap between the two main parties remaining largely constant.

During the campaign, the average Labour support has risen by about seven to eight points, apparently benefiting from a slide in backing for the Liberal Democrats. Like two years ago, Labour’s line on the polling graph has moved slowly upwards.

Unlike two years ago, however, backing for the Tories has also built, rising by about six to seven points following a collapse in backing for the Brexit Party and Nigel Farage’s decision to pull his candidates out of Conservative-held seats.

The result is that Labour has not trimmed the Tory opinion poll lead significantly.

Despite making some further gains in an Ipsos MORI poll released on 6 December, with the Lib Dems and the Brexit Party slipping further, the Tory lead remained at 12 per cent.

The polling data, conducted with the Evening Standard, showed Labour up four percentage points from November at 32 per cent, with the Conservatives holding firm at 44 per cent.

Behind them, the Lib Dems slipped three points to 13 per cent, with the Greens and the Brexit Party at 3 per cent and 2 per cent respectively.

A YouGov poll released on 7 December showed similar margins, with the Tories on 43 per cent and Labour on 33 per cent.

The Greens and the Brexit Party were both on 3 per cent, while the Lib Dems matched the Ipsos Mori figures on 13 per cent.

If Mr Corbyn’s troops can advance by another three to four points, Labour could deny the Prime Minister the overall majority he says he needs to deliver Brexit – and several factors remain which make the result impossible to call with any confidence:

* How will turnout levels on the day - especially if parts of the country suffer a cold snap - make a difference?

* Will older voters head for their polling stations in greater numbers than younger adults?

* Are the polls again underestimating support for a particular party?

* Could an apparent big lead for the Tories actually encourage “soft Labour” supporters into the Corbyn fold?