In 2004 the American journalist James Surowiecki wrote a book called ‘The Wisdom of Crowds’, setting out the theory that the average estimate of a mass of ordinary people is often more accurate than the informed view of experts.
Over the years since Surowiecki’s book was published, there have been various poll experiments to see if a random sample of voters can predict the outcome of elections as well, or better, than traditional polls.
The record of the wisdom of crowds approach at elections has been very mixed. But in the Irish referendum last year on gay marriage, all the orthodox voting polls overstated the support for change (which is the predominant pattern in referendums around the world), while the ‘wisdom of crowds’ survey conducted by the leading Irish research organisation Red C got the result exactly right.
Populus decided to test exactly the same approach in the UK’s EU referendum. The question we asked was:
‘Thinking about all the conversations you may have had with friends, family and others about the referendum on UK membership of the EU, we’re interested to see how good you are at predicting the result. What proportion of the voting public do you think will vote for the UK to Remain in the EU? And what proportion do you think will vote to Leave the EU?’
Throughout the referendum campaign there was a significant difference between the online voting polls and those done by phone. This modal difference, however, did not affect the wisdom of crowds responses. Over the weekend of 10-12 June we polled the question simultaneously online and over the telephone. The answer was all but identical: the average of respondent guesses put Remain on 48.6% in the phone poll and 48.7% online; Leave was at 51.3% online and 51.4% in the phone sample.
The final wisdom of crowds poll showed the Leave campaign winning 51/49 – a slightly narrower margin than the final result, but closer to it than six of the seven pre-election predictions from orthodox voting intention polls.
The evidence so far, suggests that the wisdom of crowds may work much better on binary choices, such as referendums, than more complex choices such as general elections. But Populus will continue to experiment with this technique, to see how well it performs in different situations.