Prediction markets can’t get away from sports right now.
That’s where the fight is. That’s where the attention is. That’s where the money is. That’s where the debate — “is this gambling, or finance, or gambling disguised as finance?” — has landed for the moment.
But prediction markets weren’t originally built for sports.
The original pitch was simpler: When people have money on the line, the price is a clean signal of the “wisdom of the crowds.” Not perfect. Not magical. But more honest than surveys, hot takes, or expert consensus.
Two academic papers suggest that idea might actually hold up, at least according to the early data.
One paper looks at Polymarket’s quarterly earnings markets and finds the bettors on Polymarket did a better job than Wall Street analysts at calling who would beat earnings. The authors of that study, “Financial Prediction Markets: A New Measure of Earnings Expectations,” are Roberto Gómez-Cram, Yunhan Guo, and Howard Kung of London Business School, along with Theis Ingerslev Jensen of Yale University.
The other, an NBER working paper titled “Kalshi and the Rise of Macro Markets,” looks at Kalshi’s macro markets, such as inflation, jobs, and what the Fed is going to do, and argues the prices can work as a real-time expectations read that stacks up well against the usual forecasts, and in at least one spot appears to do better. That paper is by Anthony M. Diercks of the Federal Reserve System, Jared Dean Katz of Northwestern’s Kellogg School of Management, and Jonathan H. Wright of Johns Hopkins (and NBER).
Polymarket’s earning markets
For the Polymarket paper, the researchers used the company’s “earnings beat” contracts, markets tied to whether a company reports earnings per share above a benchmark number. Then they checked how often the market got it right.
In a sample window (Sept. 15 to Nov. 15, 2025), they report the market was about 68% accurate a week before earnings, climbing to 77% accurate the day before.
Their analyst comparison came in at 62%.
They also looked at whether the market signal lines up with the real-world reaction. In other words, does it help explain what happens to stocks around earnings? Their answer is basically yes: The market information adds something beyond analyst expectations alone.
The authors note this is a small, early dataset, but it’s clearly a notable result.
Kalshi and macro markets
The second paper is about CPI, unemployment, GDP, and Fed funds. Its basic point is that Kalshi’s markets can give you a live read on those outcomes and because the contracts often come in ranges, you can see a fuller picture than a single number.
They show the markets reacting quickly to news, including a July 2025 example where rate-cut odds moved after Fed governor remarks, then shifted again after jobs data.
Then they compare it to the stuff people already use.
For Fed rate expectations about five months out, they say Kalshi’s errors look very similar to the New York Fed’s Survey of Market Expectations.
Against Bloomberg’s consensus forecasts, they say Kalshi is generally in the same neighborhood for core CPI and unemployment. But for headline CPI, they report Kalshi does better and they flag that result as statistically meaningful.
Common threads
For a gambling-forward audience, the interesting part isn’t “academics approve of prediction markets.” Who cares?
The interesting part is what both papers are really pointing at: Prices contain information. And in these early tests — one on earnings, one on macro economic factors — that information looks pretty good.
It doesn’t end the argument. It doesn’t mean these markets can’t be gamed. It doesn’t mean the early results won’t change once there’s more data.
But it does widen the story.
Because while everyone’s yelling about whether prediction markets should be allowed to touch the NFL, these papers are testing the claim prediction markets have made since day one: that markets, with money behind them, certainly seem to get you closer to the truth.
