Do Decentralized Markets Beat Centralized Forecasters?
Evidence
from the 2024 Presidential Elections
Authors
Dr. Lambis Dionysopoulosa, Dr. Andrew Urquhartb
Publication details
ISSUE 171 2025
Keywords
Prediction Markets; Polymarket; US Elections; Polls
Abstract
This paper examines whether decentralized blockchain-based prediction markets enhance forecasting accuracy compared to traditional prediction markets and statistical models. Through a comparative analysis of three prediction markets — Polymarket (decentralized), PredictIt (centralized), and FiveThirtyEight (poll aggregator) — during the 2024 U.S. presidential election, we find a statistically significant hierarchy: decentralized markets outperform centralized counterparts, which in turn surpass poll aggregation. Decentralized platforms’ censorship-resistant design, global participation, and reduced information suppression may explain their superior accuracy. Our results position decentralized prediction markets as robust tools for synthesizing crowdsourced information in real time, particularly in contexts requiring resilience to censorship and dynamic data aggregation.
Author Details
Dr. Lambis Dionysopoulosa. ICMA Centre, Henley Business School, University of Reading, Reading, United Kingdom, c.dionysopoulos@pgr.reading.ac.uk
Dr. Andrew Urquhartb. Birmingham Business School, University of Birmingham, Birmingham, United Kingdom, a.urquhart@bham.ac.uk