Could online search offer a glimpse of the election result?

The Google election: Search analysis tips Tories for election win

The Google election: Search analysis tips Tories for election win

The Conservatives are on course to be the largest party in a hung parliament, according to an analysis of online search terms.

The new technique, by Mavens of London, combines traditional polling data with information about people's online searches for their local candidates.

It suggests the Tories would remain the largest party in a hung parliament, losing just two seats to end up with 301. Labour would fail to improve on their 256 seats from the 2010 general election and in fact lose another four seats.

The Liberal Democrats would lose over half their 56 seats, leaving them with just 19.

The SNP would be the big winners, going from six to 52 seats, with the new technique backing suggestions by traditional pollsters that the nationalists are set to clean up north of the border.

The analysis suggests Ukip's rise may have been overstated, with the eurosceptic party winning just one extra seat, leaving it with a total of three. The Greens would remain with just one seat.

The results mean neither Labour nor the Tories would be able to form a majority, even in coalition with the Liberal Democrats. Labour would need an agreement with the SNP in addition to a coalition with the Lib Dems, while the Tories would need to group together several smaller parties, probably including the DUP.

Party candidate share of search over the last 12 months

The Mavens technique hypothesises that online searches for local candidates are a strong indicator of intent to swing. A similar technique was used to successfully predict the outcome of the 2012 US election, but the firm had to construct a new model for Britain's more complex 650-constituency political system.

Researchers found national search results over-represented the Conservatives due to interest in high profile Cabinet figures and provided little insight about local political campaigns.

Instead, a master list of search terms indicating interest in parties, candidates and issues at a constituency level was created. Where candidate searches over-lapped with other well-known individuals – such as Mark Webber, a name shared by a well-known racing driver and a Ukip candidate – the search term was adjusted to remove ambiguities, for instance by only recording searches for 'Mark Webber Ukip'.

High profile figures, such as Cabinet ministers, had search volume results restricted so that only searches from the areas close to their constituency were counted. In the case of Nick Clegg, for instance, results were restricted to searches from the city of Sheffield.

Researchers then assessed voter loyalty to political parties as a baseline to predict the swing across major regions of England, Scotland, Wales and Northern Ireland and offset the results against local interest in candidates. Traditional polling data was then aggregated as a supplement to the digital information.

How Britain will vote if the online search hypothesis is accurate

 

Initial results found 592 of the UK's 650 constituencies exhibited behaviour which indicated a candidate was far enough ahead of their rivals to predict a winner. Seats which could not be forecast were predicted to remain with their incumbent.

The technique's focus on digital behaviour obviously limits its view of the election outcome, although the fact that it includes around two million relevant searches per month means it represents perhaps the largest sampling of potential voter interest currently available.

The technique might have been expected to over-represent liberal or left wing parties because of the prevalence of young, urban voters online, but it appears the use of search data, as opposed to social media, may have limited the impact of demographic imbalances online.