Can prediction markets predict?

Prediction markets like Kalshi and Polymarket have become such a big thing that by now, many pundits quote the forecasts from these sources, rather than the conventional consensus forecasts from Bloomberg and the like. But to paraphrase Nate Silver (who is an adviser to Polymarket), prediction markets are a bunch of teenagers in Scotland sitting in their bedrooms betting on events. Ok, he said that about betting odds, but the question is whether prediction markets are an unbiased and reliable forecast of economic events.

And there is good news. It seems that forecasts on Kalshi are indeed unbiased and can convey additional information compared to market-based forecasts or surveys among professional economists. But Kalshi’s forecasts are not necessarily better than these other sources, as a study by Anthony Diercks, Jared Katz, and Jonathan Wright shows.

They compared the forecasting performance of Kalshi on US headline and core inflation, as well as US unemployment data and decisions of the Federal Reserve, with the forecasts from Bloomberg surveys and the Philly Fed Survey of Professional Forecasters (SPF).

The chart below shows the forecast error of Kalshi in comparison to the forecast error of Fed Funds Futures traded in financial markets and the SPF in the six months before a Fed interest rate decision. Because these are forecast errors, lower numbers are better.

The chart clearly shows that Kalshi’s forecasts are significantly better than the Fed Funds Futures traded in money markets. The SPF, on the other hand, is about as accurate as Kalshi’s forecasts. The advantage investors have with Kalshi, however, is that they have continuous adjustments and daily updates, while the SPF is updated only once a quarter. If there are significant events that change the outlook for Fed Fund Rates, Kalshi can provide a good idea of how the odds have changed.

Fed Funds Rate forecast errors

Source: Diercks et al. (2026)

Another test is to compare Kalshi’s forecasts with the Bloomberg consensus of economists for key economic variables like inflation and unemployment. The charts below show that the forecast error of Bloomberg consensus data tends to be somewhat higher for headline inflation, but lower for core inflation and unemployment numbers.

Macro forecast errors

Source: Diercks et al. (2026)

Based on their analysis of macro releases since 2022, the authors conclude that Kalshi predictions clearly outperform conventional forecasts when it comes to inflation and Fed Funds Rate decisions. But they are less reliable when it comes to other measures like core inflation and unemployment that require more expertise to understand, model and predict.

Another advantage of prediction markers is that they provide continuous updates, not monthly or quarterly updates like surveys. Especially in volatile times, this often makes prediction markets the only available data to assess how investor sentiment has changed.