Dissertation "Prediction markets versus alternative methods. Empirical tests of accuracy and acceptability" online available [29.01.2010]

The success of prediction markets in the field of election forecasting made them increasingly appealing to organizations and a number of companies started to experiment with them. However, despite widespread initial interest and years of experimental use, there are no major organizations known to use prediction markets as an integral part of their forecasting activities. Prediction markets have not become an established forecasting method yet. The reasons for this are manifold. Aside from election and sports forecasting, the number of empirical studies that analyze prediction markets' performance is limited. In addition, the studies are often small scale or compare the method to weak benchmarks. Since the emergence of the field, no meta-analysis has been published to analyze prediction markets' accuracy. Furthermore, practical experience indicates that cognitive and organizational barriers thwart the implementation of prediction markets within organizations.

This work provides further empirical evidence on the performance of prediction markets and analyzes the method's acceptability among participants. Results from a field experiment showed that prediction markets performed equally well as the Delphi method for long-term forecasting problems. Similar results were derived from a laboratory experiment on a quantitative judgment task: overall, prediction markets performed equally to Delphi, nominal groups, and meetings. Furthermore, they appeared to be particularly valuable for problems where multiple group members had valid information.

However, laboratory experiment participants had comparably unfavorable perceptions of prediction markets, particularly in terms of difficulty of participation, which may lead to low confidence in market results. This seemed to be confirmed by the fact that market results were discounted more often than results from the three benchmark approaches. Yet, in discounting the market results, participants did not improve accuracy; they harmed it. The results suggested that prediction market participants were unable to judge the quality of market results and, thus, should have refrained from revising them.

The empirical evidence from this work supports the value of prediction markets for forecasting. However, it also revealed that prediction markets are afflicted with unfavorable perceptions of participants. This conformed to practical experience indicating barriers to the method's implementation within organizations. Future research aimed at identifying and overcoming these barriers is of crucial importance. There is a need for further empirical studies that analyze performance of prediction markets for different types of problems and in different settings. This should involve the use of prediction markets in conjunction with traditional means of forecasting. In addition, market engineering should search for ways to make market platforms more accessible, particularly to non-experience participants.

Bibliographical references:
Andreas Gräfe
Prediction markets versus alternative methods. Empirical tests of accuracy and acceptability
Karlsruhe: Fakultät für Wirtschaftswissenschaften der Universität Karlsruhe(TH) 2009

  • Contact and Personal Homepage Andreas Gräfe