Wednesday 11 February 2015

Who's best at predicting the 2015 General Election? 'The academic experts' or 'the public'?


In the chart below, see how the public betting market on a 'hung parliament' seems to be converging on the top academic forecasters' predictions from Electionforecast.co.uk.  These academics have consistently predicted more chance of a hung parliament than the public (89% probability compared to 79% as of yesterday), and the public now seem to be coming into line.  Are the 'expert academics' winning this tug-of-war of prediction over 'Joe Public'? 

 
The conflict between academic prediction and public prediction is now the key theme of this blog.  The next article will discuss how different parts of the betting industry are claiming to represent the public forecast, and which companies can be most trusted to do so: particularly spread firms like Sporting Index or Spreadex, fixed odds firms like Ladbrokes, or the leading betting exchange, Betfair.  Over 20 years, I've worked for all three types of betting companies in the UK.  Now I'm looking at the 2015 General Election as an academic researcher, for a team led by Dr Matthew Goodwin. 

The disparity between expert and public predictions, hinges on the public faith in a Tory revival between now and the 7th May, as Ed Miliband's unproven Prime Ministerial leadership credentials are tested, along with Labour competency on key issues such as the economy.  However, the polling fortunes of the Conservatives are still not improving fast enough, if at all, to deliver them an overall majority of 326 seats.  At the same time, both Conservative and Labour prospects of an overall majority have been badly dented by the rise of UKIP and the SNP, whose support is proving more difficult to erode than the main parties expected.

On Betfair, punters have consistently given a 15%+ chance to a Tory overall majority, whereas Electionforecast have never ascribed more than a 5% chance to that outcome.

Traditionally, past results have shown a revival in the incumbent government's polling fortunes in the final months and weeks of the campaign, and a tendency for voters to 'revert to the mean' (or do what they did before).  Neither of these two trends are reflected in snapshot polling as published daily in the media.  Whilst both are factored into the top expert academic models and public prediction markets, they are more so by the public, who are invariably more bullish about Conservative chances across all markets. 

Have 'the experts' sufficiently accounted for a Tory revival with their model? Are the expert academic models unduly biased towards Labour, or 'the public markets' towards the Conservatives?  If the academics backed their own model with their own cash, they would still be betting on a hung parliament on Betfair.com. 


 
Albert Tapper (@mincer) is a former trader of political markets at the spread betting firm Sporting Index.  He now works on a leading academic project analysing the 2015 General Election, led by Dr Matthew Goodwin (University of  Nottingham).