Learning from Prediction Markets: The Transmission of Information and Noise to Traditional Assets
Prediction markets are widely promoted as efficient mechanisms for aggregating dispersed information and generating signals for the broader economy. We document information spillover from Polymarket to traditional financial markets during the 2024 U.S. presidential election. Changes in Polymarket-implied probabilities predict next-day returns of ‘Trump trades’ across asset classes, followed by partial reversals, indicating that both information and noise are transmitted. Using the blockchain-based transaction data, we classify Polymarket traders as informed or uninformed at the wallet level and show that the price impact of uninformed trading contributes to the next-day predictability, but, unlike the price impact of informed trading, does not persist and fades over the following days. As the transmitted noise cannot reflect the sequential arrival of fundamental information, it identifies an active cross-market learning channel through which both information and noise propagate. We further document an ‘information hangover’ effect: recent informed trading amplifies the traditional market’s subsequent reliance on Polymarket signals, which in turn facilitates noise spillovers.
EFA 2026 Ghent, Georgetown Politics in Finance Conference, UGA Finance Conference, RAPS/RCFS European Conference 2026