Home ›
Docs › How Accurate Are Prediction Markets?
How Accurate Are Prediction Markets?
Prediction markets have a strong academic track record — but accuracy depends heavily on liquidity, aggregation method, and the type of event being forecast.
The evidence on accuracy
Multiple studies have examined prediction market calibration — whether a 60% market corresponds to a 60% event frequency. The short answer: for liquid, near-term events, they're well-calibrated. The key findings:
- Iowa Electronic Markets correctly called 9 of 10 US presidential popular vote winners from 1988–2020
- Tetlock's superforecasting research shows that aggregated crowd forecasts exceed individual experts on 1–2 year horizon questions
- Sports prediction markets show good calibration for pre-game lines but can be slow to update to in-game developments
Why aggregation improves accuracy
Any single market can be influenced by a large trader, delayed by slow liquidity, or biased by the preferences of its participant pool. Aggregating across multiple independent markets averages out these idiosyncratic factors.
This is the principle behind Meridian Edge's consensus — combining signals from multiple markets produces a probability estimate that's more stable and better calibrated than any single source.
Where prediction markets struggle
- Low-liquidity events: Thin markets are easily moved by a single order and may not reflect genuine probability assessments
- Long-horizon events: Markets more than 6 months out have poor calibration due to limited information and participant interest
- Black swan events: By definition, truly unprecedented events are under-priced by all models
Using divergence as a quality signal
When multiple prediction markets disagree significantly on the same event, it often signals that one market hasn't incorporated new information yet — not that the event is inherently unpredictable. Meridian Edge flags this as divergence, allowing you to investigate the reason before relying on the consensus probability.
Frequently asked questions
How accurate are prediction markets?
Research consistently shows prediction markets exceed individual expert forecasters and simple polls for near-term events. Events priced at 70% resolve in the predicted direction roughly 70% of the time — a property called calibration. Long-horizon events are less well-calibrated due to lower liquidity.
Are prediction markets more accurate than polls?
For election forecasting, prediction markets have generally exceeded single polls, though both have had notable failures. Markets update faster to new information than polls, which run on a fixed schedule.
What factors affect prediction market accuracy?
Liquidity is the biggest factor — thinly traded markets can have wide bid-ask spreads and slow price discovery. Aggregating across multiple markets (as Meridian Edge does) improves accuracy by averaging out noise from any single low-liquidity source.
Access Consensus Data
Query live aggregated prediction market probabilities via API.
View Plans →
Need more? Starter ($29/mo) — 1,000 calls/day + divergence signals