Prediction as sport: the promise of prediction polls and markets

The aim of this working paper is to explain the principles behind forecasting tools, uncover in what circumstances they work best, and explore how such ideas could be used to make better forecasts that in turn inform better decisions, not just in theory but in practice.

Nesta Working Paper 15/02
Issued: February 2015
JEL Classification: 
Keywords: forecasting, prediction markets, collective wisdom, polling

Abstract

You can in certain circumstances combine the moderately accurate judgments, or forecasts, of many individual people to produce a collective judgment this is more accurate – one frequently more reliable than any lone expert. Two ways of doing this, specifically for forecasting future events, are the focus of this working paper: prediction polls, in which many individuals are asked independently to assign a numerical probability to a future event and those judgments are aggregated; and prediction markets, in which individuals bet against each other about the likelihood of events, and the market price is interpreted as a probability. 

Much evidence suggests that these tools produce predictions that are very accurate, and can improve our ability to forecast a wide range of phenomena – be it geopolitical and macroeconomic events, scientific breakthroughs or the success of an emerging technology, or a company’s sales targets. 

Evidence also suggests that participation in such forecasting tournaments improves, with training and regular practice forecasting events, improves individuals’ ability to forecast dramatically.

The aim of this paper is to explain the principles behind these tools, uncover in what circumstances they work best, and explore how such ideas could be used to make better forecasts that in turn inform better decisions, not just in theory but in practice.

Author

Jeremy Kingsley