Wireva

NowCasting Tool Gives Democrats 90% Chance to Win Senate

A real-time forecasting tool built by Northwestern data scientist Thomas Miller gives Democrats roughly a 90 percent chance of taking control of the U.S. Senate ahead of the November 3 midterms.

This item was produced with AI assistance under the editorial responsibility of Haydamax OÜ.

Democrats have roughly a 90 percent chance of gaining control of the U.S. Senate, according to a new forecasting tool that updates every 10 minutes ahead of the November 3 midterm elections. By early Monday morning, with Election Day one day away, the platform put the odds of a Democratic takeover at about 90 percent, while Republicans had only about a 17.5 percent chance of keeping control in the model's last update before that.

The tool, which its creator calls "NowCasting," pulls data from the prediction markets Kalshi and Polymarket throughout the day. It then uses statistical models to simulate a million versions of the midterms, with outcomes for each of the 35 races in play. "Every 10 minutes, I'm running a gigantic simulation," said Thomas Miller, faculty director of the master's in data science program at Northwestern University, who developed the approach.

Miller said the value of prediction markets lies in how quickly they respond to news. Some research suggests such markets can be more accurate than polls. He cautioned, however, that the price traders pay for a candidate is not the same as a probability. If it costs 60 cents to bet that a Republican candidate will win, that does not equal a 60 percent chance of victory. A large bet from a single trader could theoretically distort the price, and platform fees and other pricing dynamics also affect the numbers. Prices for all candidates in a single race can sum to more than $1, even though their probabilities should total 100 percent.

To correct for these distortions, Miller's approach adjusts the market prices and combines them with statistical modeling used in finance and econometrics. "It's data driven. I'm a data scientist, not a politician, not a pollster," he said. "I'm using the data, and using models that are trusted and that have been used for many years."

Miller has worked with betting-market data since 2020, when he correctly forecast that Joe Biden would win using data from the market PredictIt. In January 2021, in the Georgia runoffs for Senate seats, he correctly predicted that Jon Ossoff and Raphael Warnock would win. In 2024, when some other forecasters predicted a win for Kamala Harris, he predicted that Donald Trump would win the presidential election.

The forecast shifted sharply in the days before the vote. On Friday, Republicans' chances of keeping the Senate had fallen to 17.5 percent, down a couple of percentage points from earlier that day and a steep drop from the day before. On Sunday, Michigan Democratic congresswoman Haley Stevens endorsed Republican Mike Rogers over Abdul El-Sayed in the Senate race. Rogers's chances ticked up slightly on the platform, but only temporarily and nowhere near enough to win.

The forecast comes as the midterms face heavy spending on political advertising. A Fox News opinion column by researcher Lee Carter put political ad spending and reservations at more than $11.6 billion, which the column described as the most expensive election cycle for political advertising in American history. Carter argued that voters are motivated more by opposition to something than by enthusiasm for a candidate, pointing to affordability, socialism, corruption, Trump and turnout as the central themes of the campaign.

Same event, other desks

Story file →