Kalshi disclosed that trading activity in its U.S. midterm election markets for the 2026 cycle exceeded $2.5 billion, a sum almost double the $1.3 billion seen during the 2024 election period. The volume for these midterm elections rose by over $1.2 billion compared to the prior cycle, signaling increased market engagement as outcomes were settled by early Nov. 4. Regarding predictive success, Kalshi noted that candidates favored in its markets secured victories in 479 of the 487 races concluded by that date.
Research from the platform demonstrated a winner-prediction accuracy rate of 98.2% at the close of polls, marking an improvement from 93.8% recorded three months prior. Live updates during election night highlighted trader support for Democratic wins in Texas and South Carolina as data came in, while Kalshi's summary on Nov. 4 observed that traders expected Democrats to hold both congressional chambers before Election Day even though they projected smaller gains than the party ultimately achieved.
Record Trading Volume and Market Shifts
Kalshi's rounded data indicates a volume increase exceeding 92% compared to the prior election period. This metric contrasts two distinct sets of contests: the 2024 cycle featured a presidential race, whereas the current year focused on congressional, state, and local elections.
Kalshi released its first count showing that market favorites secured victories in 479 out of 487 resolved races by 6 a.m. ET on Nov. 4. The scope of this tally encompassed House and Senate seats, gubernatorial positions, as well as contests for attorney general, lieutenant governor, and secretary of state.
Kalshi announced that its platform recorded 356 victories among the favored outcomes for 360 House races, alongside 32 wins in 33 Senate matchups and 33 wins in 34 gubernatorial contests. Across the entire set of results, eight favorites failed to win, resulting in an overall success rate of roughly 98.4%, a figure that was simplified to 98% for the public statement.
At the time the summary was released, certain elections had not yet been decided, specifically the Senate races in Alaska and Mississippi, so the calculated accuracy reflected only the outcomes known by morning. Kalshi noted that its market indicating Democratic control of the House hit a 99% probability threshold at 8:10 p.m. ET on election night, whereas the market for Senate control reached that same level at 9 p.m. ET.
On November 3, Kalshi's real-time data captured shifts in specific markets throughout the election tally. At 6:55 p.m. ET, the platform's feed showed a 97.2% chance for Democratic control of the House. Later, at 8:06 p.m., that same feed indicated Greg Abbott, the Republican incumbent, had lost favor in the Texas governor's race.
Subsequent entries recorded Democrat James Talarico at a 99% probability of winning the Texas Senate contest at 9:35 p.m., followed by Annie Andrews reaching that threshold in South Carolina at 10:10 p.m. The feed also reached the same 99% level for Texas gubernatorial challenger Gina Hinojosa at 11:35 p.m., with Kalshi identifying these figures as market projections distinct from official race calls or contract settlements.
According to Reuters, prediction markets had forecasted Democratic control of Congress but failed to capture the strength of certain candidates; analytics firm BETINT, referenced in that coverage, identified 14 Democratic winners who were market underdogs a week prior and noted that markets and polls performed roughly equally well overall.
A distinct Kalshi Research analysis reviewed races called by 6:26 a.m. ET on Nov. 4, monitoring forecasts spanning three months before voting through poll closure. The researchers indicated that directional accuracy increased from 93.8% to 98.2% during that timeframe, representing a gain of 4.4 percentage points; Brier scores, which gauge the difference between forecast probabilities and actual outcomes, declined from 0.040 to 0.016, with lower scores signifying superior performance.
Nicole Kagan and Rubens Baiocchi authored a working paper demonstrating that counting only winning favorites overlooks market confidence, as a forecast priced at 51% generates the same directional score as one at 95%. The study utilized prices existing before each measurement time and synchronized election-model snapshots with 9 a.m. ET on Election Day. This approach anchored forecasts to voting times since election contracts stay open even after results are known.
Prediction Accuracy and Democratic Gains
Prior to the election, the report documented CFTC oversight of election markets, highlighting remarks by Chairman Michael Selig regarding potential manipulation. Reporting from Oct. 24 referenced Dune research indicating that combined midterm open interest on Kalshi and Polymarket reached $250 million as of Oct. 18, with Kalshi representing $210 million and Polymarket accounting for $40 million.
Following candidate involvement in election-related trading, Representative Don Davis presented the No Betting on Your Own Race Act in October. This measure aims to bar federal candidates, their spouses, dependent children, and authorized campaign committees from engaging in contracts linked to their own elections. The bill outlines potential fines of $10,000 or three times the net financial gain from a transaction, whichever amount is greater.