If you want to guess the outcome of the Scottish Independence Referendum, it might be better to consult the bookies than rely on opinion polls.
That’s the conclusion of a leading academic who has been looking at the best way to forecast the outcome of the 18 September poll.
Professor David Bell from the University of Stirling looked at using “prediction markets” – which include the gambling industry – to forecast the outcome of the referendum. He concluded that these markets often produce more accurate results than relying on opinion polls, which sample opinion on the day they were collected, rather than directly trying to forecast outcomes.
In a study for the “Future of the UK and Scotland project” – an ESRC-funded programme which brings together a range of academics looking at Scotland’s constitutional future – Professor Bell says the latest data suggests there is 70 per cent chance of a “No” vote.
He says: “The use of prediction markets in relation to electoral outcomes is well-established. Their accuracy relies on there being a number of well-informed, dispassionate traders/gamblers who wish to profit from their trades/bets. Though these conditions do not always hold, prediction markets have a good record in forecasting future outcomes.”
In terms of the Scottish Referendum, Professor Bell studied the history of odds offered on “No” and “Yes” votes from 23 bookmakers between Sep 2013 and May 2014.
“At the end of the period, (the second week of May), the market estimate is that there is around a 7 in 10 chance of a ‘No’ vote,” he says.
He argues: “First, it is clear that the market estimate of the probability of a No vote is higher than the proportion of voters that may vote No. This is not surprising. Even if the “No” vote in the opinion polls exceeded the “Yes” vote by only a small margin, it would be rational for a high proportion of bets to be placed on a “No” outcome, provided that those betting on this outcome thought that this margin would not disappear before the referendum.
“Second, market outcomes tend to be more stable than the opinion polls. Differences in the way that samples are drawn by polling organisations, sampling errors, and differential delays in the publication of results may explain some of the variability in the opinion polls.
“Third, a succession of polls that showed an increase in support for the Yes campaign during March and April 2014 may have lengthened the odds on a ‘No’ outcome in the referendum. From mid-March to mid-April, an increase in the value of bets on a ‘Yes’ outcome caused the implied probability of a ‘No’ outcome to fall from 0.8 to 0.6. By early May, it had returned to 0.7.”
Lastly, Professor Bell speculates whether some of his earlier research, which shows that ‘Yes’ voters are more likely to take risks than ‘No’ voters, might affect the pattern of betting on the referendum. However, he notes there is no evidence to date that either rejects or supports this hunch.
The Independence Referendum: Predicting the Outcome (PDF, 370KB)