The midterm markets on the eve of the vote: 67 races, two exchanges and what the prices imply about correlation

The day before the 2026 midterms, we took a snapshot of every election market on both major platforms. At 4 p.m. ET on 2 November 2026, we recorded the price of each Senate and governor's race on Kalshi and Polymarket, the two chamber-control markets and Kalshi's trading volume in each race. This is what the markets believed going into Election Day, recorded before any votes were counted.

The full dataset is available as a CSV file.

Key findings

1. Democrats were favoured in most races. On Kalshi they led in 20 of 34 Senate races and 21 of 33 governor's races.
2. If every favourite won, Democrats would hold 54 Senate seats. Adding their 34 seats not up for election to the 20 races they led gives 54.
3. The two exchanges agreed closely. In competitive races, the typical gap between Kalshi and Polymarket was 1.5 points, and the largest was about 3 points.
4. Volume was concentrated. The Texas Senate race alone accounted for about a quarter of all contracts traded on Kalshi's Senate and governor's races; the top five races accounted for almost half.
5. The Senate control price implies correlated races. If the races were independent, the race prices would imply about an 88% chance of Democratic control. The market said 74–76.5%.

The data

We collected:

• 34 Senate races listed on Kalshi (all 35 seats on the ballot except Louisiana's, which Kalshi did not list), and 33 governor's races;
• the price of the Democratic candidate in each race on Kalshi, from hourly candles, and on Polymarket, from its price history, at 4 p.m. ET (21:00 UTC);
• Kalshi's Senate and House control markets and its seat-count markets;
• the number of contracts traded in each race on Kalshi up to the end of 2 November.

Where Polymarket listed a race with the Republican as the main outcome, we used one minus that price. Polymarket did not list two of the 67 races in a form we could match.

How the races were rated

We grouped races by the Democratic candidate's price on Kalshi, using bands similar to those election forecasters use.

Source: Kalshi market data, 4 p.m. ET on 2 November 2026.

The Senate map was balanced at the extremes: 11 safe seats for each party. The difference was in the middle. Democrats led in five "lean" races (Texas, Maine, Alaska, Ohio and Kansas) and three "likely" ones (Michigan, Minnesota and New Hampshire), while Republicans led in only three competitive races (South Carolina, Mississippi and Florida). The single toss-up was Iowa.

The competitive Senate races

Source: Kalshi and Polymarket public market data, 4 p.m. ET on 2 November 2026.

The competitive governor's races

Source: Kalshi and Polymarket public market data, 4 p.m. ET on 2 November 2026. "—" where we did not record a matching Polymarket price.

The governor's map leaned further toward Democrats than the Senate map. Fourteen governor's races were safe Democratic against five safe Republican; of the fourteen competitive races, Democrats led in seven. Texas and Florida were the two big Republican-held states where Democrats were genuine but clear underdogs.

Kalshi and Polymarket agreed

Across the 28 races and chamber markets where Kalshi priced the Democrat between 5% and 95%, the median gap between the two exchanges was 1.5 points and the mean 1.46 points. No race was more than about three points apart. In only one race, Iowa, did the two platforms disagree about who was favoured, and that was a coin flip: 51% on Kalshi, 49.5% on Polymarket.

That is consistent with our earlier study of price gaps, which found a median gap of about one point across hundreds of matched outcomes. When the same contract trades on two regulated exchanges, arbitrage keeps the prices close.

There was a small pattern. Polymarket was slightly more bullish on Democrats in the chamber markets (Senate 76.5% against 74%; House 95.5% against 93.8%) and in Texas, Maine, Ohio and Alaska. Kalshi was slightly more bullish on Democrats in Kansas, Iowa, South Carolina, Mississippi and Florida. The differences are too small to read much into.

Where the money was

Kalshi traders had exchanged about 264 million contracts on the 67 races by the end of 2 November. The distribution was very uneven.

Source: Kalshi market data.

Texas accounted for 26% of all contracts traded on Senate and governor's races; the top five races for 49%. Races rated safe by our bands, 41 of the 67, accounted for only about 15% of volume. Traders put their money where the uncertainty was.

What the Senate price says about correlation

The most interesting number in the snapshot is the gap between the race prices and the Senate control price.

Democrats needed 17 of the 34 races Kalshi listed to reach 51 seats (they held 34 seats not up for election, and the Vice President, J.D. Vance, breaks ties for Republicans). Adding up the Democratic price in each race gives an expected 18.35 wins, or about 52.4 seats.

If the races were independent, like separate coin flips, the chance of 17 or more Democratic wins would be about 88%. The Senate control market said 74% on Kalshi and 76.5% on Polymarket.

The difference is correlation. Races are not independent: a national swing, a polling error or a turnout surprise moves many races in the same direction at once. If the polls were underestimating Republicans in Texas, they were probably underestimating them in Maine and Ohio too. Traders in the control market priced that risk; the independent calculation ignores it.

Kalshi's seat-count market shows the same thing from another angle. It priced "more than 52 Democratic seats" at 54%, but also gave about 18% to outcomes of 49 seats or fewer, a much fatter tail than independent races would produce.

How the markets moved in the final week

The snapshot is the end point of a late shift. A week earlier, on 27 October, Kalshi priced Democratic Senate control at about 64%. It reached 71% on 30 October, 73% on 1 November and 74% on 2 November. The race markets moved with it: the Texas Senate price for the Democrat rose from 70% on 29 October to 74%, and the Florida governor's price for the Democrat from 18% on 22 October to 36%.

The House market barely moved over the same week, from about 91% to 94%. Traders had treated a Democratic House as likely for months; the late money went into the Senate.

The House seat-count market

Kalshi's market on the number of Democratic House seats gave the highest single price, 30%, to "more than 249 seats". The next most likely buckets were 238–241 and 242–245 seats, at about 13% each, and 234–237 at about 11%. Democrats won 215 seats in 2024, so the market's central expectation was a gain of roughly 25 to 35 seats. Outcomes that would leave Republicans in control, below 218 Democratic seats, together traded at about 5%.

Method and limitations

• Prices are last trades in the hour ending at 21:00 UTC on Kalshi, and the last recorded price at or before that time on Polymarket. A last trade can differ from the mid-price in thin markets.
• Volume is in contracts, each paying $1 if correct. It is not the same as dollars wagered, since contracts trade below $1.
• The independence calculation treats each Kalshi race price as a probability and assumes no correlation. It is a benchmark, not a forecast.
• Third parties. Where an independent candidate had their own contract, we used only the Democratic and Republican contracts.

What to watch

We will compare these prices with the results once the races are called, in a scorecard of how accurate the markets were. In the meantime, our election-night guide explains how to follow the prices as the votes come in, and our Election Day 2026 guide has the poll-closing times.

Download the data

The CSV file has one row per market: the race, the candidates, the Kalshi and Polymarket prices at 4 p.m. ET on 2 November, the gap between them and Kalshi's contracts traded through 2 November.

Data: Kalshi and Polymarket public APIs, collected for the snapshot at 4 p.m. ET on 2 November 2026. Not investment advice.