Home / News

Did the Markets Call It? Democrats Win Both Chambers, and the South Carolina Miss

On the eve of Election Day, Kalshi traders gave Democrats a 94.1% chance of winning the House and 75% of winning the Senate. Democrats won both. In South Carolina, a race the same markets priced 70-30 for the Republican went to the Democrat.

At about 2:11 p.m. ET on Monday, Kalshi took what it called a final snapshot before the midterms. Its traders put the Democrats' chance of winning the House at 94.1% and of winning the Senate at 75%, the company said, according to Quartz. An experimental AI-generated summary on Polymarket's site cited a 75.5% implied probability of Democrats winning both chambers. Analytics firm BetInt, which aggregates prices across platforms, put the odds at 95% for the House and 76% for the Senate, Reuters reported.

The Senate number had moved a long way. Kalshi's odds of Democratic Senate control reached 74% by noon ET on Monday, the highest since the market opened. That was up about 11 percentage points from about 63% on Oct. 2. Republicans had led on that market since it opened in 2024, and their odds of Senate control peaked at 60% in July before reversing by mid-September, Quartz reported.

By Wednesday morning the result was clear. Democrats won control of both chambers, according to the Associated Press. AP reported, and Roll Call confirmed, that Democrats flipped at least seven Senate seats: Kansas, Iowa, Texas, Ohio, Maine, North Carolina and South Carolina. They held battleground seats in Georgia and Michigan and flipped more than two dozen House districts.

The consensus was not unanimous

The markets made that call against some skepticism. Tobin Marcus, head of U.S. policy and politics at Wolfe Research, told CNBC on Election Day that his firm shared the view that the "polling is likely overstating Democratic strength." CNBC reported that the stock market was not pricing in Democratic control of both chambers heading into Tuesday. A MarketWatch headline that day said prediction markets "may show 'too much confidence in a good night for Democrats'"; the phrase is a quotation within the headline, and the full article was not available to us. On the headline question, the markets were more confident than the polls, and that confidence held up.

The miss

South Carolina is where the night got more complicated. Reuters reported that Kalshi and Polymarket had implied a 70% chance, as of late Tuesday afternoon, that Republican Darline Graham would win the seat. Graham is the sister of the late Sen. Lindsey Graham, according to Reuters. Democrat Annie Andrews, a pediatrician, won instead. She is the first Democrat to win a Senate seat in South Carolina in 28 years. According to Reuters, Donald Trump carried the state by 18 percentage points two years ago, and neither party considered the race competitive until recent weeks. Reuters called it the markets' "first high-profile miss" of the night.

The companies' standard response, as Reuters reported it, is that events priced at relatively low probabilities happen often and political upsets are no exception. That is correct as far as it goes. A 30% outcome should occur about three times in ten. Kalshi and Polymarket did not respond to Reuters' requests for comment on the night.

Calibration versus accuracy

This is the distinction that matters most. Joshua Clinton, the Vanderbilt political scientist who led a 2025 study of nearly 2,500 markets on the 2024 election, told Reuters that accuracy measures whether a market picked the winner, while calibration measures whether its odds are reliable. Across many markets, contracts priced at a given probability tend to come true about as often as that probability suggests. "But how accurate the price is for any specific market is impossible to know," he said.

Measured that way, one upset at 30% does not discredit the markets. What would discredit them is a pattern. If many races priced at 70% or higher went the other way in the same direction, that would point to systematic bias rather than ordinary chance. The recent record gives some reason to look. Traders gave Flávio Bolsonaro roughly an 80% chance in Brazil's October runoff before Luiz Inácio Lula da Silva won, and Clinton told Reuters that sophisticated traders may have overcorrected for the risk that polls underestimate right-wing candidates. In Wisconsin's August primary, a roughly 95% favorite lost by less than half a point, Fortune reported.

This year's upsets do not point the same way. In Brazil the incumbent, Lula, won against the market favorite. In South Carolina the Democrat won in an election the markets had already priced as favorable to Democrats nationally. On the evidence available so far, that looks more like noise than a persistent tilt. It is still too small a sample to prove either way.

What we will build

A proper verdict requires settled data, not snapshots. Once markets resolve, we plan to publish a race-by-race scorecard. It will compare final pre-close prices on Kalshi and Polymarket for every Senate and governor's race with certified results, along with a calibration chart grouping races by implied probability. Until then, any claim that the markets "called the midterms" should be read as a statement about two national contracts.

The global view

The midterms are the largest test so far of regulated election markets. Reuters reported that more than $1.48 billion was traded on the races. They come as other countries weigh how to treat election betting. The lesson for regulators and investors abroad is the one Carlos Hardenberg of MCP Emerging Markets offered Reuters: the markets are a useful gauge, "but they should not be confused with certainty." South Carolina shows why. A market can price the national outcome correctly and still miss an individual state.

This article is for information only and is not investment, trading or wagering advice.

Sources: Quartz, Reuters, Associated Press, Roll Call, CNBC, MarketWatch, Fortune.