How Accurate Are Prediction Markets on Inflation? Kalshi’s CPI Track Record

By Arthur Angelo, PredictionScout Research · Updated Oct 15, 2026, 7:47 AM ET

Key findings

• Usually right, rarely confident. The monthly favorite was exact 12 of 22 times (55%). That is about what its prices promised: the favorite cost 45¢ on average, which adds up to roughly 10 expected hits. CPI markets spread their bets across several neighboring numbers, so a favorite at 40–55¢ is normal.
• Close calls count. When the favorite missed, it was usually by a hair: 19 of 22 monthly prints landed on the favorite or the number right next to it (0.1 point away).
• The yearly number is harder. The yearly rate was called exactly 9 of 22 times and within 0.1 point 19 times. Its favorite was cheaper too (41¢ on average), so traders knew it was less certain.
• No clear lean. Of the monthly misses, 4 came in hotter than the favorite and 6 cooler.
• Last 12 releases: 10 of 12 monthly prints called exactly.
• Big miss: March 2025. A day before the Apr 10, 2025 report, the favorite was 0.1% at 41¢. BLS reported -0.1% or lower, which traders had priced at just 4¢.
• Big miss: June 2026. A day before the Jul 14, 2026 report, the favorite was -0.2% at 53¢. BLS reported -0.3% or lower, which traders had priced at just 8¢.

Favorite vs. actual, release by release

Every monthly CPI release

“Favorite” is the single print with the highest implied chance 24 hours before the 8:30 AM ET release. “Price on actual” is the chance traders gave the number that came out. Highlighted rows missed by more than 0.1 point.

Every yearly CPI release

Older releases: counting every Kalshi CPI market with a full price ladder (back to Jan 12, 2023), the monthly favorite was exact 23 of 44 times and within 0.1 point 40 times; the yearly favorite 17 of 42 and 36. Those earlier markets traded far less, so the headline uses January 2025 onward. All of them are in the CSV.

What is CPI, and why do people bet on it?

The Consumer Price Index is the government’s main measure of inflation: how much a typical basket of goods and services costs compared with before. The Bureau of Labor Statistics publishes it once a month at 8:30 AM Eastern. Two numbers get the headlines: how much prices rose in the last month (seasonally adjusted) and how much they rose over the past year.

The Federal Reserve watches these numbers when it sets interest rates, so a surprise moves stock and bond markets within seconds. On Kalshi, a federally regulated exchange, anyone can buy a contract that pays $1 if CPI comes in above a certain level. The price of that contract is the crowd’s estimate of the chance it happens.

Related: how accurate prediction markets are on Fed decisions and our full accuracy scorecard.

Download the data

Release-level CSVLive price ladder CSV

The release file has one row per CPI report and measure (monthly and yearly): the actual print as Kalshi settled it, today’s BLS value, the market’s favorite and its price 24 hours before, the price on the actual print, and contracts traded. The ladder file has every contract for the next release with bid, ask and the price we used. Free to reuse under CC BY 4.0.

Methodology

• Markets. Kalshi’s monthly CPI series (KXCPI, seasonally adjusted one-month change in CPI-U) and yearly CPI series (KXCPIYOY, not seasonally adjusted 12-month change), including older CPI- and CPIYOY- events, from Kalshi’s public historical and live market API. Each release has a ladder of “Above X%” contracts in 0.1-point steps.
• Timing. Prices are taken 24 hours before trading closed (trading closes a few minutes before the 8:30 AM ET release), from hourly candlesticks: the last hour with a price at or before that moment, looking back up to 48 hours because Kalshi skips hours with no change.
• Price. Midpoint of the closing bid and ask when the spread is 10¢ or less, otherwise the last trade.
• Chance of each print. The difference between neighboring “Above” contracts (negative differences from out-of-order quotes count as zero). The lowest bracket is “this or lower” and the highest is “this or higher”. The favorite is the print with the highest chance.
• Actual print. Taken from how Kalshi settled the ladder, which follows the first BLS release. We also show the value in today’s BLS data. Seasonally adjusted monthly numbers are revised every February, so a few older rows differ (marked “revised”). Kalshi settled on the original number.
• Exact match means the actual print was the favorite print. Within 0.1 point includes exact matches and prints one step away.
• Gaps. A few far-out contracts had no trades in the window; we leave them out, which merges two tail brackets. That happened in 1 monthly and 5 yearly releases in the headline window, always in the tails.
• Limits. One exchange, about two dozen releases per measure, and 0.1-point brackets that make an “exact” call hard by design. Read the hit rate together with the favorite’s price.
• Updates. Live prices refresh every morning. New results are added after each BLS release once Kalshi settles the markets. Last run: Oct 15, 2026, 7:47 AM ET.

Data sources: prices and volumes come from the public APIs of Kalshi and Polymarket, with CPI data from the Bureau of Labor Statistics. All trademarks belong to their owners. PredictionScout is independent and not affiliated with any exchange. Nothing here is investment, betting, or legal advice.

More research: PredictionScout Research · Live odds