general election: Game of seats as forecasters disagree on tally

As the UK political system has fragmented, electoral forecasts have taken on new significance.

They allow us to crunch bewildering amounts of data and gauge what kind of government will emerge after the general election in May.

While forecasters agree a hung parliament is likely, they are split on which party will emerge with the largest number of seats.

The notion of "uniform national swing" - a forecasting method that assumes national vote shares will apply in each constituency - has fallen out of favour. It is seen as less reliable now that the traditional two-way battle between Labour and the Conservatives has become more a series of regional contests - the Scottish nationalists are predicted to take many Labour votes north of the border, while Ukip is eating into Conservative support in a number of English costal towns.

Some 12 forecasting teams recently agreed that the most likely outcome was that no single party would win enough seats (326) for a majority. But the forecasters, who met at the London School of Economics last month, were split about whether Labour or the Conservatives would end up with most.

Statistical models of forecasting grew more popular after Nate Silver, a journalist, successfully predicted the outcome for each state in the 2012 US presidential election. For the UK general election, Mr Silver's website, FiveThirtyEight, hosts forecasts from academics, Chris Hanretty, Benjamin Lauderdale and Nick Vivyan, who publish on electionforecast.co.uk and whose data are used on FT.com. Using seat-level polls and raw data from YouGov, a polling company, as well as forecasts based on national polls, their model allows them to project probabilities for each of the UK's 650 constituencies.

"The model uses all sources of information available," Mr Hanretty says.

"We're not going to get things exactly right. In fact, even if we're right on average, we know we'll get some seats wrong. But this is true of all forecasting models, and if someone tells you they're going to be exactly right, they're drastically underestimating the uncertainty of these forecasts."

In the UK, forecasters are using a variety of sources but all agree there will be a hung parliament. Most use data on voting intentions, others take local by-election results or data on economic indicators and leadership ratings.

The 12 forecasting models presented at the conference were based on a different set of assumptions and data source. The model used by Polling Observatory used voting intention statistics from opinion polls and the concept of regression to the mean - this contends parties above their historical average are likely to fall and those below the average rise as the election gets closer.

Stephen Fisher, a sociologist from the University of Oxford, uses a model that factors in historical bias against the Conservative party in the opinion polls. Prof Fischer's model has the highest forecast for the party, with 296 seats.

The highest forecast for Labour came from Mary Stegmaier and Laron Williams of the University of Missouri. They saw data on current voting intentions, satisfaction with the economy and partisan identification with the two main parties. The model projects Labour will win 312 seats.

Forecasting relies on extrapolating from historical trends, and if these break down, or the data are wrong, the results are misleading.

Professor Will Jennings, from Southampton university says, "It's important to understand all the aspects of the model and that these are assumptions. If there's something systematically wrong with the opinion polls, or . . . constituency polls, all of the forecasts could be out."

© The Financial Times Limited 2015. All rights reserved.
FT and Financial Times are trademarks of the Financial Times Ltd.
Not to be redistributed, copied or modified in any way.
Euro2day.gr is solely responsible for providing this translation and the Financial Times Limited does not accept any liability for the accuracy or quality of the translation