Memoirs of an indifferent trader: Estimating forecast distributions from prediction markets

Published on Jul 1, 2010in Quantitative Economics
· DOI :10.3982/QE6
Joyce E. Berg16
Estimated H-index: 16
(UI: University of Iowa),
John Geweke68
Estimated H-index: 68
(UTS: University of Technology, Sydney),
Thomas A. Rietz21
Estimated H-index: 21
(UI: University of Iowa)
Sources
Abstract
Prediction markets for future events are increasingly common and they often trade several contracts for the same event. This paper considers the distribution of a normative risk-neutral trader who, given any portfolio of contracts traded on the event, would choose not to reallocate that portfolio of contracts even if transactions costs were zero. Because common parametric distributions can conflict with observed prediction market prices, the distribution is given a nonparametric representation together with a prior distribution favoring smooth and concentrated distributions. Posterior modal distributions are found for popular vote shares of the U.S. presidential candidates in the 100 days leading up to the elections of 1992, 1996, 2000, and 2004, using bid and ask prices on multiple contracts from the Iowa Electronic Markets. On some days, the distributions are multimodal or substantially asymmetric. The derived distributions are more concentrated than the historical distribution of popular vote shares in presidential elections, but do not tend to become more concentrated as time to elections diminishes.
References19
Newest
The basic idea of this note is simple. Fundamentally, opinion polls are estimates of the final vote shares of candidates. These estimates are characterised by uncertainty. This uncertainty is the link between prediction market prices and the opinion polls. The simple reason for this is that the price in a prediction market indicates the probability that a given candidate will win a share of the vote greater than a given threshold (usually 50%). As soon as one is able to estimate the distribution...
11 CitationsSource
#1Thomas S. GrucaH-Index: 23
#2Joyce E. BergH-Index: 16
Last. Michael CiprianoH-Index: 5
view all 3 authors...
We compare the forecasts of nineteen movie box office results from real money (Iowa Electronic Market) and play money (Hollywood Stock Exchange) prediction markets. The forecasts were not significantly different, contrary to recent research on incentives and prediction market accuracy. Proponents of play money incentives suggest that (play) wealth concentrates in the hands of knowledgeable traders over time. This should lead to improved accuracy over time. A longitudinal analysis of results (199...
24 CitationsSource
This paper presents an attempt to study and monitor the evolution of research on prediction markets (PM). It provides an extended literature review and classification scheme. The former consists of 155 articles, published between 1990 and 2006. The results show that an increasing volume of PM research has been conducted in a very diverse range of areas. The articles are further classified and the results of this classification are presented, based on a scheme that consists of four main categorie...
117 CitationsSource
#1Richard KnellerH-Index: 44
20 Citations
#1Joyce E. Berg (UI: University of Iowa)H-Index: 16
#2George R. Neumann (UI: University of Iowa)H-Index: 26
Last. Thomas A. Rietz (UI: University of Iowa)H-Index: 21
view all 3 authors...
We conducted prediction markets designed to forecast post-initial public offering (IPO) valuations before a particularly unique IPO: Google. The prediction markets forecast Google's post-IPO market capitalization relatively accurately. While Google's auction-based IPO price was 15.3% below the first-day closing market capitalization, the final prediction market forecast was only 4.0% above it. The forecast also accorded with the level of over-subscription in the IPO auction. Evidence available t...
33 CitationsSource
#1Joyce E. Berg (UI: University of Iowa)H-Index: 16
#2Forrest D. Nelson (UI: University of Iowa)H-Index: 20
Last. Thomas A. Rietz (UI: University of Iowa)H-Index: 21
view all 3 authors...
"Prediction markets" are designed specifically to forecast events such as elections. Though election prediction markets have been being conducted for almost twenty years, to date nearly all of the evidence on efficiency compares election eve forecasts with final pre-election polls and actual outcomes. Here, we present evidence that prediction markets outperform polls for longer horizons. We gather national polls for the 1988 through 2004 U.S. Presidential elections and ask whether either the pol...
231 CitationsSource
#1Andrew Leigh (ANU: Australian National University)H-Index: 44
#2Justin Wolfers (NBER: National Bureau of Economic Research)H-Index: 62
We review the efficacy of three approaches to forecasting elections: econometric models that project outcomes on the basis of the state of the economy; public opinion polls; and election betting (prediction markets). We assess the efficacy of each in light of the 2004 Australian election. This election is particularly interesting both because of innovations in each forecasting technology, and also because the increased majority achieved by the Coalition surprised most pundits. While the evidence...
72 CitationsSource
#1Charles F. Manski (NU: Northwestern University)H-Index: 80
Abstract Prediction markets are futures markets in which prices are used to predict future events. I present the first formal analysis of price determination supposing traders have heterogeneous beliefs, deriving the equilibrium when traders are risk-neutral price takers.
231 CitationsSource
While most empirical analysis of prediction markets treats prices of binary options as predictions of the probability of future events, Manski (2004) has recently argued that there is little existing theory supporting this practice. We provide relevant analytic foundations, describing sufficient conditions under which prediction markets prices correspond with mean beliefs. Beyond these specific sufficient conditions, we show that for a broad class of models prediction market prices are usually c...
181 Citations
#1John Geweke (UI: University of Iowa)H-Index: 68
Preface. 1. Introduction. 1.1 Two Examples. 1.1.1 Public School Class Sizes. 1.1.2 Value at Risk. 1.2 Observables, Unobservables, and Objects of Interest. 1.3 Conditioning and Updating. 1.4 Simulators. 1.5 Modeling. 1.6 Decisionmaking. 2. Elements of Bayesian Inference. 2.1 Basics. 2.2 Sufficiency, Ancillarity, and Nuisance Parameters. 2.2.1 Sufficiency. 2.2.2 Ancillarity. 2.2.3 Nuisance Parameters. 2.3 Conjugate Prior Distributions. 2.4 Bayesian Decision Theory and Point Estimation. 2.5 Credibl...
507 Citations
Cited By6
Newest
#1Thomas S. Gruca (UI: University of Iowa)H-Index: 23
#2Thomas A. RietzH-Index: 21
1 CitationsSource
#1Ming-Chien Sung (University of Southampton)H-Index: 11
#2David McDonald (University of Southampton)H-Index: 2
Last. Eng-Tuck Cheah (University of Southampton)H-Index: 7
view all 5 authors...
Abstract We examine the impact of price trends on the accuracy of forecasts from prediction markets. In particular, we study an electronic betting exchange market and construct independent variables from market price (odds) time series from 6058 individual markets (a dataset consisting of over 8.4 million price points). Using a conditional logit model, we find that a systematic relationship exists between trends in odds and the accuracy of odds-implied event probabilities; the relationship is co...
Source
Source
#1Jason Wei Jian Ng (Monash University)H-Index: 4
#2Catherine S. Forbes (Monash University)H-Index: 14
Last. Brendan McCabe (University of Liverpool)H-Index: 15
view all 4 authors...
The object of this paper is to produce non-parametric maximum likelihood estimates of forecast distributions in a general non-Gaussian, non-linear state space setting. The transition densities that define the evolution of the dynamic state process are represented in parametric form, but the conditional distribution of the non-Gaussian variable is estimated non-parametrically. The filtered and prediction distributions are estimated via a computationally efficient algorithm that exploits the funct...
12 CitationsSource
Non-Gaussian time series variables are prevalent in the economic and finance spheres, with state space models often employed to analyze such variables and, ultimately, to produce forecasts. A review of the relevant literature reveals that existing methods are characterized by a reliance on (potentially incorrect) parametric assumptions and are often computationally expensive. The primary aim of this thesis is to develop a non-parametric approach to forecasting - within the state space framework ...