US elections

The event family prediction markets were built for — and the one where the gap between a market price and a poll is most worth reading.

How often: Federal elections every two years, with primaries, special elections and nominations in between.
Who decides, and on what: The certified result. Contracts differ on whether they settle on projection, certification or inauguration — the wording matters more here than almost anywhere.

Open markets on US elections

Live prediction markets whose question turns on this event. Prices update with the rest of the site.

How accurate have these markets been?

Across 194 resolved markets of this kind, the price seven days before resolution scored a Brier of 0.093, against 0.205 for simply forecasting the base rate every time — so the market added information. It carried no meaningful directional bias.

What the market said before resolution against how often it happened, by probability band
The market said	It happened	Cases
0–20 %	3 %	111
20–40 %	too few	22
40–60 %	too few	16
60–80 %	too few	19
80–100 %	too few	26

Measured on our own corpus of resolved markets, using the price seven days before resolution. A band shows a rate only from 30 cases — fewer than that rewards luck. A historical frequency is not a probability for the next one.

The market is not a poll with a dollar sign

A poll measures stated intention now, with a margin of error. A market prices the eventual result, incorporating turnout expectations, historical polling error and the time left for things to change. A poll at 52 % and a market at 70 % are not in contradiction — they are answering different questions.

Where they diverge sharply and the market is liquid, the market is usually the better guess. Where they diverge and the market is thin, it is not.

Certification, projection or inauguration

Election contracts differ on the moment they settle. Some resolve when major outlets project a winner, some on certification, some later still. In a contested result those dates can be weeks apart, and a position can be right about the outcome and wrong about the contract.

This is the single most common source of surprise on election markets, and it is entirely avoidable by reading the rules first.

Common questions

Are election prediction markets more accurate than polls?

In liquid markets they generally aggregate more information and update faster. They also inherit polling error rather than escaping it, because traders read the same polls. Our comparison guide goes through where each one is stronger.

When do election markets settle?

It depends on the contract — some on projection by major outlets, some on certification, some later. In a contested result those can be weeks apart, so read the resolution text before taking a position.

How accurate have these markets been here?

The figures on this page come from our own corpus of resolved markets. Below the publication threshold the page shows the sample size instead of a hit rate.