Traders in prediction markets have consistently outperformed professionals in forecasting inflation, particularly when actual readings diverge significantly from estimates, as highlighted in a study by prediction market Kalshi.
In a comparison of inflation predictions from its platform against Wall Street consensus estimates, Kalshi found that market-based traders displayed greater accuracy than traditional economists and analysts over a span of 25 months, especially during economic instability, according to a report shared with CoinDesk.
Market-based predictions of year-over-year changes in the Consumer Price Index (CPI) recorded a 40% reduction in average error compared to consensus forecasts from February 2023 to mid-2025, the study reveals. This discrepancy became even more pronounced during significant deviations from expectations, with Kalshi’s forecasts outperforming consensus by up to 67% in such instances.
The research, titled “Crisis Alpha: When Do Prediction Markets Outperform Expert Consensus?,” also explored the link between the level of disagreement in forecasts and the probability of a surprise outcome.
When Kalshi’s CPI forecast diverged from consensus by more than 0.1 percentage point a week prior to release, the likelihood of a notable discrepancy in the actual CPI reading surged to approximately 80%, in contrast to a 40% baseline.
Unlike conventional forecasting, which tends to rely on a uniform set of models and assumptions, prediction platforms like Kalshi and Polymarket consolidate forecasts from individual traders motivated by financial incentives to make accurate predictions.
Kalshi has recently expanded its user base through its integration with the prominent crypto wallet Phantom. The company secured $1 billion at an $11 billion valuation earlier this month as interest in prediction markets continues to escalate. In October, Polymarket was reported to be in negotiations to raise funds at a valuation approaching $15 billion.
The authors of the report emphasize that while the sample size for substantial shocks is relatively limited, the findings suggest a promising role for market-based forecasting within broader risk management and policy planning strategies.
“While the sample size of shocks is small (as expected in a world where they are largely unforeseen), the trend is evident—when forecasting becomes most challenging, the market’s information aggregation advantage proves most beneficial,” the study states.
Earlier this year, a data scientist’s research indicated that Polymarket boasts a 90% accuracy rate in predicting events one month in advance, and 94% accuracy just hours before the actual occurrence. However, factors such as confirmation bias, herd mentality, and limited liquidity can result in inflated event probabilities.
The reason prediction markets excel over consensus during stressful times may be attributed to their information aggregation methods. Traditional forecasts often depend on similar data and models across institutions, which can limit their adaptability as economic conditions change, according to the study.
In contrast, prediction market platforms capture the perspectives of a varied group of traders using an array of inputs, ranging from sector-specific trends to alternative datasets, fostering what the study refers to as a “wisdom of the crowd” effect.
Moreover, the incentives differ significantly. Institutional forecasters encounter reputational and organizational limitations that may deter daring predictions. Conversely, traders in prediction markets have financial stakes, gaining rewards or facing penalties solely based on performance.
The ongoing nature of market pricing, which updates in real-time, also helps avoid the delays inherent in consensus estimates, which are usually established several days before data releases.
“Instead of completely replacing traditional forecasting methods, institutional decision-makers might explore incorporating market-based signals as supplementary information sources with unique benefits in times of structural uncertainty,” the study proposes.
