Prediction market traders consistently beat professionals in forecasting inflation, especially when readings deviate from estimates by a larger amount, according to a study by prediction market Kalshi.
By comparing inflation forecasts on his platform to Wall Street consensus estimates, Kalshi found that market-based traders were more accurate than conventional economists and analysts over a 25-month period, particularly during periods of economic volatility, according to a report shared with CoinDesk.
Market-based estimates of year-on-year changes in the consumer price index (CPI) showed a 40% lower average error than consensus forecasts between February 2023 and mid-2025, the study found. The difference was more pronounced, as the figure deviated strongly from expectations. In these cases, Kalshi’s forecasts outperformed the consensus by as much as 67%.
The study, called “Crisis Alpha: When Do Prediction Markets Outperform Expert Consensus?”, also examined the relationship between the size of forecast disagreement and the likelihood of a surprise.
When Kalshi’s CPI estimate deviated from the consensus by more than 0.1 percentage point a week before the release, the chance of a significant deviation in the actual CPI reading rose to about 80%, compared to a baseline of 40%.
Unlike traditional forecasting, which often reflects a common set of models and assumptions, prediction markets such as Kalshi and Polymarket aggregate forecasts from individual traders with financial incentives to predict outcomes accurately.
Kalshi’s user base has recently grown with the integration of the prediction market into the major crypto wallet Phantom. The company raised $1 billion at an $11 billion valuation earlier this month as betting on prediction markets continues to grow. In October, Polymarket was said to be in talks to raise funds at a valuation as high as $15 billion.
The report’s authors note that while the sample of large shocks is relatively small, the data point to a potential role for market-based forecasts as part of broader risk and policy planning tools.
“Although the sample size of shocks is small (as it should be in a world where they are largely unexpected), the pattern is clear – when the forecasting environment becomes most challenging, the information aggregation of markets becomes most valuable,” the study reads.
Earlier this year, research by a data scientist showed that Polymarket is 90% accurate at predicting how events will play out a month out, and 94% just hours before the actual event occurs. Still, origin bias, herd mentality, and low liquidity can lead to overestimated event probabilities.
Why prediction markets outperform consensus in times of stress may come down to how they gather information. Traditional forecasts often rely on similar data and models across institutions, which can limit their responsiveness when economic conditions change, the study suggests.
Prediction market platforms, by contrast, reflect the views of a diverse set of traders who draw on a range of inputs, from sector-specific trends to alternative data sets, creating what the study describes as a “wisdom of the crowd” effect.
The incentives are also different. Institutional forecasters face reputational and organizational constraints that can discourage bold predictions. However, traders in prediction markets have money at stake and are rewarded or punished solely based on performance.
The continuous nature of market prices, which are updated in real time, also avoids the lag inherent in consensus estimates, which are typically set several days before data releases.
“Rather than wholesale replacement of traditional forecasting methods, institutional decision makers may consider incorporating market-based signals as complementary sources of information of particular value during periods of structural uncertainty,” the study suggests.
