Election prediction markets: polls, contracts and results

3 numbers that look alike but answer different questions

A poll that reports 52% support, a contract trading near 52 cents, and a 52% win probability are 3 different objects. The poll is a vote-intention estimate produced under a specific sample, question wording and weighting. The contract price is what someone will pay or accept for a share whose payout is defined by written resolution rules. The win probability is an output of a model that converts inputs, including turnout and seat allocation, into a chance of winning. Only the first is a survey measurement; only the second is a tradable instrument; the third is an inference you build yourself.

AAPOR best practices concern transparent methods, question wording, sampling, fieldwork dates, nonresponse and methodological disclosure. Vote-share estimates are not probabilities of victory. Turnout and sampling assumptions matter, and hypothetical scenarios must not be presented as a validated election model. So when a headline says a candidate is at 52%, you have an intention reading at one point under one method, not a win chance and not a contract payoff.

What a contract actually pays on

An ordinary binary market pays 1 per winning share and 0 per losing share. A rare unknown or 50-50 resolution can pay 0.50 each; that is not a universal cancellation or refund. The rules specify the source, deadline and edge cases, and the title alone is insufficient. You have to read which source decides the outcome, on which date the decision is final, and what happens in ambiguous scenarios before the market means anything for you.

An event groups one or more markets, and events, conditions and outcome token identifiers differ. Tags overlap instead of forming an exclusive hierarchy. Check each individual market's active, closed and acceptingOrders flags rather than copying the parent event status onto every contract in the group. Read each market’s flags and resolution rules separately instead of inferring them from the parent event.

Resolution mechanics also differ by market design. UMA resolves ordinary prediction markets, while Chainlink TWAP resolves Up/Down markets. An Up/Down market uses the same asset stream and a time-weighted average price at start and end: final at or above start resolves UP, lower resolves DOWN. Different exchange candles can differ, so verify the actual market rules and do not generalize every duration or source.

Projection, count, certification and appeal are separate stages

Election coverage mixes 4 stages that should stay separate: projections before official results, the count of ballots, certification of the result, and appeals as a separate stage. A contract that resolves on certification behaves differently from one that resolves on a projection, even when both reference the same election night. Write down which stage your contract names and which exact source and date the rules point to.

The same discipline applies across countries. Peru is used here only as a country example of search-interest data related to vote intention; no actual candidate or outcome is implied. Search interest from one country does not transfer mechanically into win probability, seat count or runoff qualification elsewhere, and automatic country clone pages or attached political positions have no place in that reading.

Turnout, seats and runoff are different categorical outputs, not one estimate

Vote intention, win probability, turnout, seat count and runoff qualification are categorically different outputs. Each requires its own conversion step: intention to ballots cast, ballots cast to valid votes, valid votes to seat allocation, seat allocation to government formation or a runoff procedure. A single survey question cannot answer all of them, because they rest on different assumptions about who votes and how those votes convert.

This matters for execution. Suppose a candidate sits at a high intention level but the seat system or a runoff threshold makes that level insufficient to govern. A contract on winning the presidency can fall while a contract on topping the first round rises, both correctly. If you trade only the headline number, you are trading the wrong output, and the losing case is exactly the scenario where intention and qualification diverge.

Worked example: from poll number to contract decision

The case below is explicitly illustrative and is not an empirical study. It walks one reader through the comparison of a poll figure and a contract before deciding whether to trade. Suppose a poll reports 52% vote intention for a candidate, and a contract on that candidate winning trades at 0.55. Nothing in the poll verifies the 0.55; the contract instead depends on the rules and on a conversion from intention to seats or votes.

Reading limits and what you cannot conclude

From a single 52% intention figure you cannot infer a win probability, a turnout level or a seat count. You also cannot infer that a contract at 0.55 is mispriced. The poll has its own uncertainty from sampling, wording and fieldwork dates, and the contract has its own execution costs, liquidity and resolution ambiguity. A positive expected value computed on a guessed probability is not a reason to take the trade.

The losing case is concrete: you read 52% intention, treat it as a 52% win chance, buy a contract that actually resolves on a certification stage you never checked, and lose because the rules named a stage the poll never measured. The fix is procedural rather than predictive: identify the exact question, match it to the right categorical output, and read the resolution rules before sizing anything.

Where to go next in this cluster

This hub routes the different questions to the detailed guides. If your question is about survey numbers and how they are collected, start with public opinion polling methodology. If your question is about thresholds, runoff procedures and seat conversion, start with election rules and thresholds explained. If your question is about how inflation, rates and growth affect political markets, start with macroeconomics for prediction markets. If your question is about which source decides an outcome and when, start with resolution and settlement sources explained.

Before any trade, write one sentence naming the output you need: vote intention, win probability, turnout, seat count or runoff. If the contract's rules answer a different sentence, treat it as a separate market rather than as confirmation of your polling view. That single check keeps the poll, the contract and the result aligned.