Forecasting Efficiency of Prediction Markets: Prediction Markets Platform, 2019

This collection contains the code to run a prediction market, and the replication package for the data that was collected. This is a laboratory study on information aggregation in prediction markets when traders face ambiguity and hold imprecise beliefs. Prediction markets are widely used to aggregate dispersed information, yet many real-world forecasting problems involve rare or unfamiliar events for which probabilities are ill-defined. The study investigates when and why markets succeed or fail to aggregate information in such environments, with particular attention to the role of security design and initial prices set by a market maker.

The primary aim of the study is to test theoretical predictions on information aggregation under ambiguity by comparing standard separable securities (such as Arrow–Debreu contracts) with a newly proposed class of strongly separable securities. The data capture how ambiguity aversion, security type, and the market maker’s initial price jointly affect convergence of prices to intrinsic values.

The dataset covers individual forecasts, trading histories, and outcomes across different market environments, enabling analysis of mispricing, manipulation, and robustness of information aggregation.
The data were collected in incentivized laboratory experiments conducted in February 2019 at Université Paris 1 Panthéon–Sorbonne. A total of 288 subjects were recruited from the university’s experimental database via email invitations and participated in only one session each. Participants were randomly assigned to one of eight treatments in a 2×2×2 factorial design varying belief precision (unique vs. multiple priors), security type (separable vs. strongly separable), and initial market price (0 or 50). Each session involved 18 participants and consisted of 12 rounds of sequential trading between fixed pairs within rounds. Subjects received private signals, made repeated price forecasts under a market scoring rule, and were paid based on their forecasting accuracy, with all payments fully incentivized.

To obtain a free account please register with the UKDA.