Predicting Elections: A ‘Wisdom of Crowds’ Approach

Abstract

Opinion polls are the currency of politics. They are used by media organisations to evaluate the performance of governments, and by governments and political parties to test the policies that shape manifestos and reform agendas. But opinion polls all rely on one thing - asking people how they themselves intend to vote - and, too often, classical opinion research techniques fail to confront the issues that underpin inaccuracy. In the UK and in many other countries around the world, their performance over the past 20 years has ranged from excellent to disastrous.1 The ‘Wisdom of Crowds’ concept turns conventional predictions on their head. It assumes that any crowd that conforms to a core set of principles is capable of delivering a more accurate prediction than the smartest people within it. This paper tests this proposition within the context of actual elections in the UK, showing that the Wisdom of Crowds approach used by ICM Research at the 2010 general election would have produced the most accurate final pre-election prediction. It also shows that a Wisdom approach to regular vote intention tracking produces an interesting complement to classically conducted vote intention polls. Or, if one were to be bold, a competitor to them.