No open-access location is known for this record yet.
No open-access location is known for this record yet.
Crowd prediction methods offer the promised to collect valuable, widely dispersed information in organizations. To the extent that information is a source of power, crowdsourcing democratizes organizational governance. We report the results of the first large-scale, long-term, experimental test of crowd prediction methods. More than 2,400 participants made forecasts on 261 world events over two forecasting seasons, each lasting more than 9 months. Forecasters in prediction markets made trades about future events in a continuous double auction. Those in prediction polls submitted explicit probability judgments, independently or in teams. Probability values were aggregated statistically. In Study 1, which used full random assignment, prediction markets were more accurate than the unweighted mean of forecasts from prediction polls. However, team prediction polls aggregated with algorithms featuring decay, weighting and recalibration outperformed prediction markets by 12% in terms of Brier score. This pattern persisted in Study 2, and was stable across scoring rules. Prediction polls’ advantage was largest at the start of long-duration questions. Prediction polls with proper scoring, algorithmic aggregation and teaming offer an attractive method for distilling crowd wisdom.
Loading…
Stored only when both papers are in the corpus — 27 citations exist upstream.
Imported from OpenAlex (CC0) — nothing here is generated.