09.06.17 UK Elections

How new voter power identified by the Bureau damaged Theresa May

In a shock general election result, the Conservative Party has lost its majority while Labour has gained 29 seats so far.

The wins and losses confounded all forecasts by pollsters in the run-up, who looked at voting patterns and intentions to predict a Tory majority of anything from the dozens to more than 100. Instead, they lost 12 seats.

However many of these losses fell in line with modelling carried out by the Bureau’s data journalism team the Bureau Local last week. Following a hackday with local reporters in five different cities we identified two large groups of voters who had the power to swing 71 constituencies in England and Wales: people registered to vote for the first time, and people who voted UKIP or Green in 2015 whose party was not standing a candidate this time around. And 20 of those constituencies changed hands.

Of those 71 seats, 35 were Conservative, 32 were Labour and four were Lib Dem. The Tories lost 16 of them, the Lib Dems lost two and Labour held on to all but two.

Across the whole of England and Wales, 44 seats changed hands - meaning our modelling identified nearly half of them.

Read the story and get the full data: How new voters could dent Theresa May's election majority

We found six Conservative seats in which the number of newly-registered voters was greater than the majority held by the incumbent MP: Derby North, Thurrock, Croydon Central, Twickenham, Plymouth Sutton & Devonport, and Crewe & Nantwich.

The Tories lost all but Thurrock: four of them to Labour and one to the Lib Dems. In all apart from Crewe and Nantwich, Labour and the Lib Dems significantly beat the Conservatives, picking up majorities far greater than the previous winner had enjoyed.

There were a further three Tory-controlled seats where new voters were greater than the majority: Colchester, Telford and Bedford. But in these seats there was another factor - the number of former UKIP voters without a UKIP candidate this time around outweighed the new voters. Two of those seats - Colchester and Bedford - were held by the Conservatives, but Telford swung to Labour.

Six seats where new voters outweighed the Conservative majority, where UKIP was also standing
Constituency Winner:majority (2015) No. of new voters Winner:majority (2017) Victor's vote share gain
Derby North Conservative: 41 votes 125 Labour: 2,015 votes 12%
Thurrock Conservative: 536 votes 594 Conservatives: 345 votes 6.2%
Croydon Central Conservative: 165 votes 1880 Labour: 5,652 votes 10.1%
Twickenham Conservative: 2,017 votes 3,112 Lib Dem: 9,762 votes 9.2%
Plymouth Sutton & Devonport Conservative: 523 votes 3,310 Labour: 6,002 votes 7.3%
Crewe & Nantwich Conservative: 3,620 votes 4,726 Labour: 48 votes 9.4%
Three seats where new voters outweighed the Conservative majority, where UKIP was not standing
Constituency Winner:majority (2015) No. of new voters No. of UKIP voters (2015) Winner:majority (2017) Victor's vote share gain
Bedford Conservative: 1,097 2,518 4,434 Labour: 789 votes 6.6%
Colchester Conservative: 5,575 5,823 5,870 Conservative: 5,677 votes 6.9%
Telford Conservative: 730 1,958 7,330 Conservative: 720 votes 9.1%

In ten other constituencies - eight formerly held by Labour and two by Lib Dem - the number of new voters outweighed the majority held by the incumbent. Labour held on to all of those eight, while the Lib Dems lost one to Labour and one to the Tories.

In two of the 71 seats identified by the Bureau, Labour lost to Conservatives. In both these seats - Middlesborough South and Cleveland, and Stoke-on-Trest - there were UKIP candidates in 2015 who picked up thousands of votes, more than the majority held by the incumbent. This time around, there was no UKIP candidate.

Our modelling explained

At the hackday everyone had a chance to dig into a vast dataset of variables related to demographics - from income and education levels to past vote, marginality and past registration. The model also used data on England and Wales from the Office of National Statistics, the Cabinet Office and the British Election Study to estimate the number registered voters seat by seat - something not available in a central database.

Throughout the hackday 65 participants in Glasgow, Cardiff, Bournemouth, Birmingham and London each dug into the data differently in order to find stories for their own communities. Two journalists in London - Matty Edwards and Ed Fairhead - started looking at how our estimated registration numbers compared to the margin in the seats they cover, Croydon Central and Colchester respectively. They found these seats to have an increase in voter registration larger than the 2015 margin. They also looked at the size of past UKIP voters.

The Bureau worked with them and expanded their measures across across the country. We found there to be 71 seats where these 'partyless' voters were larger than the margin in the last election - 19 due to increase in new voters alone.

We then rang up local authorities to get actual registration numbers for over 50 constituencies - in doing do, finding that our model was very close (the error rate was less than 1%) and we used exact numbers in our piece.

In total, our modelling identified nearly 50,000 (49,984) newly-registered voters, more than 370,000 (371,631) UKIP voters in constituencies where UKIP was no longer running, and more than 24,000 (24,041) Green voters in constituencies where the Green Party was no longer running.

This group of 450,000 voters was spread across 71 constituencies in England and Wales.

The Bureau Local has been set up to support data-driven public interest reporting in newsrooms across the UK. We're building a unique network of regional reporters, tech experts and others passionate about local journalism and the role it should play in holding power to account.

For more information about The Bureau Local and its work, visit the project's website. To become part of the network, sign up here.