How Polymarket Election Markets Work — and Where They Break

A Polymarket election market is a set of binary contracts, not a single odds board. For a US presidential election there is usually one market per state or district asking whether a candidate wins that jurisdiction, plus markets for the national winner, the popular vote, the Senate, the House and individual races. Each market resolves independently against its written rules. The national winner market is effectively a portfolio of state outcomes, which is why its price can diverge from a simple poll average.

Prices are probabilities only in a weak sense: they are the midpoint of a bid-ask spread, set by traders with different information and risk limits. In 2024 the main presidential market favored Donald Trump while many polls described a tie. Trump won, but one result does not prove the market right or the polls wrong; the score depends on the price at the moment it is measured.

Structure: state markets, the electoral college and separate chambers

Most US presidential markets use state-level binary contracts. A "Pennsylvania winner" market pays $1 if the named candidate is certified as the winner there. The national market pays $1 if the candidate wins the presidency under its rules. Because most states award all their electoral votes to the statewide winner — Maine and Nebraska split theirs by district — the national price is a weighted sum of state probabilities, not a forecast of the popular vote.

That structure creates a correlation problem. Suppose a candidate is 70% in Pennsylvania (19 electoral votes) and 50% in Michigan (15). Multiplying 0.70 by 0.50 gives 35%, but that assumes independence. State outcomes share a national swing, so the true joint probability is higher when the candidate is strong nationally and lower when the states are driven by different local factors. Averaging state markets without modeling that correlation can misprice the national contract.

Senate markets are seat-by-seat, with the chamber control contract resolving on the resulting majority. House markets are district-by-district or a national seat-count range. Nomination markets are a separate phase and resolve before the general election. Each has its own liquidity, rules and resolution source, even on the same event page.

Buy 100 Pennsylvania YES shares at $0.70 and you pay $70. If the candidate wins, the position pays $100, a $30 profit on $70, or 42.9%. If the candidate loses, you lose the full $70. The break-even probability is 70% before costs. Add the politics taker fee at that price — 100 × 0.04 × 0.70 × 0.30 = $0.84 — and the all-in cost is $70.84, so the break-even rises to 70.84%.

Resolution sources and the rules that control

Every market names a resolution source and the edge cases the resolver should consider. Election markets commonly rely on race calls from a major wire service — the Associated Press is the usual reference — and, where the rules require it, state certification and the congressional count. The AP calls races from vote counts and waits until it believes a lead is beyond the possibility of a recount reversal.

Prediction markets on Polymarket resolve through UMA's Optimistic Oracle. A proposer posts a bond, typically $750, and names the outcome. A two-hour challenge window follows. If nobody disputes, the market resolves in about two hours. A dispute opens a second proposal round; a second dispute escalates to UMA's Data Verification Mechanism, where token holders vote after a 24-48 hour debate period. A disputed market can take four to six days. If the outcome is genuinely indeterminate, a market can resolve 50/50, and each side redeems for $0.50.

The market title is not the contract. The rules are. Read the fine print on "win the presidency" versus "win the election," on faithless electors, on a candidate who withdraws before inauguration, and on what happens if a recount or legal challenge is still running when the market's end date passes. Those clauses decide payouts in the tail, and the tail is where election markets are most likely to surprise.

Liquidity cycles

Election markets are seasonal. Months out, books are thin and a few thousand dollars can move a price. Liquidity arrives with the nomination, the conventions, the debates and the final two weeks. On election night, volume and volatility peak, then collapse as states are called. A market that looks deep in October may be impossible to exit in July.

The scale is event-driven. Reuters reported that by November 4, 2024, about $1.97 billion of the $3.1 billion in volume on Polymarket's presidential winner contract was on Trump or Harris. The pool is concentrated in the final stretch.

Thin markets also distort the signal. The displayed midpoint is the average of the best bid and ask, and a wide spread means the "probability" is a range, not a point. If the best bid is 0.42 and the best ask is 0.50, the market displays 46%, but a buyer pays 50% and a seller receives 42%.

Base rates versus polls

Polling averages are snapshots with sampling error, house effects and likely-voter assumptions. They are also correlated across states, because a national swing moves many races at once. Base rates — incumbent party performance, redistricting, fundraising, seat exposure, turnout patterns — are slower-moving and often more stable than a single poll. Markets blend both, plus positioning and news.

The evidence is not that markets are always right. Berg, Nelson and Rietz (2008) found that election prediction markets beat polls at longer horizons, while both converge near the election. That is an average over many elections, not a guarantee in one. A market can be well calibrated in aggregate and still be wrong in a specific race, especially when the price is thin or the rules are ambiguous.

A practical routine is to compare the market price with a poll-based model and a base-rate model. If they disagree by more than the spread plus your uncertainty, there may be a trade; if they disagree because the market is thin, the edge may be illiquidity.

Risks and documented controversies

Regulatory history. In 2022 the CFTC ordered Polymarket to pay a $1.4 million civil penalty, wind down markets that did not comply with the Commodity Exchange Act, and cease and desist from the violations. Polymarket subsequently restricted US users and later returned through a separate CFTC-regulated venue. The episode is a reminder that access can change by jurisdiction and that the legal status of event contracts is not settled everywhere.

Manipulation and concentration. Large positions are visible onchain, and a thin market can be moved by a relatively small amount of capital. A 2026 study of Polymarket's 2024 presidential markets found that naive volume aggregation overstated turnover — $391 million of corrected Trump-market turnover in October against $958 million of naive volume — and that moving the October forecast by five percentage points cost about $9.1 million rather than $15.6 million. The same study found capital entering on both sides of the race, which it describes as consistent with heterogeneous beliefs rather than one-sided manipulation.

Resolution disputes. Election markets can be disputed even when the outcome is clear to most observers, because the rules may require a call, a certification or a concession that arrives late. A disputed market can take days and, in rare cases, resolve at 50/50. Traders who need liquidity before resolution should assume the position is locked.

Liquidity and correlation. State markets are not independent bets. A portfolio of ten state contracts may be one national bet with ten fees. Wide spreads and thin tails make the effective cost higher than the headline price suggests. And a market that is liquid on election eve may have no bid the morning after.

Information risk. Election markets attract informed traders. Suspicious trades are investigated, but a profitable trade is not proof of wrongdoing; treat unusual flow as a reason to check the rules, not as evidence of a crime.

What the market can and cannot tell you

An election market aggregates dispersed information continuously, and that is genuinely useful. It is not a poll, a statistical model or a legal prediction. The price is a tradable opinion, the rules are the contract, and liquidity is the constraint. A trader who remembers those three things will avoid the most expensive mistake in election markets: confusing a probability with a promise.