The Journal of Prediction Markets is an academic peer reviewed journal, publishing articles both commissioned and submitted, survey articles, case studies and book reviews on every aspect of Prediction Markets.The September issue of JPM is now out, covering 5 articles.
Prediction Markets as a Tool for Management of Political Risk
Author: Bergfjord, O.
Abstract: Recently, several prediction markets for various events have been launched. The literature so far has focused on the predictive power of such markets. This paper considers such markets as tools for management of political risk. It outlines a model for use and pricing of such assets, and discusses the various benefits of a well-functioning, liquid prediction market for political decisions.
The Relative Importance of Strength and Weight in Processing New Information in the College Football Betting Market
Authors: Durham, Greg; Santhanakrishnan, Mukunthan
Abstract: Griffin and Tversky (1992) suggest that individuals, when formulating posterior probabilities based on the available evidence, tend to overreact to a new piece of evidence's strength while underreacting to the relative importance of its weight. We test this prediction using the college football betting market, a market that is commonly employed in tests for efficiency and rationality. Using average points in excess of the spread and streak against the spread as measures for strength and weight, respectively, we find that bettors overreact to strength and underreact to weight. These results are consistent with the predictions of Griffin and Tversky, as well as with the findings of Sorescu and Subrahmanyam (2006) and Barberis, Shleifer, and Vishny (1998) in financial market settings. Our work also provides insight into how behavioral biases might affect price-formation processes in other markets.
Bookmaker and Pari-Mutuel Betting: Is a (Reverse) Favourite-Longshot Bias Built-In?
Authors: Koch, Alexander K.; Shing, Hui-Fai
Abstract: A widely documented empirical regularity in gambling markets is that bets on high probability events (a race won by a "favourite") have higher expected returns than bets on low probability events (a "longshot" wins). Such favourite-longshot (FL) biases however appear to be more severe and persistent in bookmaker markets than in pari-mutuel markets; the latter sometimes exhibit no bias or a reverse FL bias. Our results help understand these differences: the odds grid in bookmaker markets leads to a built-in FL bias, whereas that used in pari-mutuel betting pushes these markets toward a reverse FL bias.
Event Studies in Real- and Play-Money Prediction Markets
Authors: Slamka, Christian; Soukhoroukova, Arina; Spann, Martin
Abstract: Event study methodology is a powerful procedure to quantify the impact of events and managerial decisions such as new product announcements on the value of a publicly traded company. However, for many events, appropriate financial data may not be available, either because suitable securities are not traded on financial markets or confounding effects limit the insights from financial data. In such instances, prediction markets could potentially provide the necessary data for an event study. Prediction markets are electronic markets where participants can trade stocks whose prices reflect the outcome of future events, e.g. election outcomes, sports results, new product sales or internal project deadlines. One key distinction between different prediction market applications is whether they require real money investments or play-money investment with non-monetary incentives for traders. Thus, the goal of this paper is to compare prediction markets' ability to conduct event studies with respect to these two different incentive schemes. We empirically test the applicability of event study methodology in real-money vs. play-money prediction markets with two data sets. We show that event studies with prediction markets deliver robust and valid results with both incentive schemes.
Long-Term Forecasting with Prediction Markets – A Field Experiment on Applicability and Expert Confidence
Authors: Graefe, Andreas; Weinhardt, Christof
Abstract: While prediction markets have become increasingly popular to forecast the near-term future, the literature provides little evidence on how they perform for long-term problems. For assessing the long-term, decision-makers traditionally rely on experts, although empirical research disputes the value of expert advice. Reporting on findings from a field experiment in which we implemented two prediction markets in parallel to a Delphi study, this paper addresses two questions. First, we analyze the applicability of prediction markets for long-term problems whose outcome cannot be judged for a long time. Second, by comparing trading behavior of an expert and a student market, we analyze whether there is evidence that supports the assumption that experts possess superior knowledge. Our results show that prediction markets provide similar results as the well-established Delphi method. We conclude that prediction markets appear to be applicable for long-term forecasting. Furthermore, we observe differences in the confidence of experts and non-experts. Our findings indicate that, in contrast to students, experts reveal their information well-considered based on what they think they know. Finally, we discuss how such analyses of market participants' confidence provide valuable information to decision-makers and may be used to improve on traditional forecasting methods.