Fed and CPI markets: what the ladder says before a release

Macro traders use Kalshi to trade the number itself: the next Fed decision, the CPI print, payrolls, jobless claims, GDP. The market trades every day until the release, so its prices are a running forecast you can compare with consensus and grade against the official figure.

#How the markets are structured

• Fed decision (KXFEDDECISION): one contract per outcome (cut more than 25bp, cut 25bp, hold, hike 25bp, hike more than 25bp). Exactly one settles YES, on the FOMC statement.
• Fed funds rate (KXFED): nested "above X%" strikes on the upper bound after a meeting.
• CPI (KXCPI, KXCPIYOY, KXCPICORE): "above X%" strikes on the published one-decimal figure, settled on the BLS release at 8:30am ET. The month in the ticker is the reference month, so September's CPI releases in October.
• Jobs (KXPAYROLLS, KXU3), claims (KXJOBLESSCLAIMS), GDP (KXGDP) and gas prices (KXAAAGASD and weekly and monthly versions) follow the same pattern.

Settlement uses the first published value. Later revisions do not count.

#Step 1: The distribution for the next release

For decision markets, take each outcome's mid price and normalize the outcomes so they sum to 100%. A hypothetical meeting: hold 84%, cut 25bp 14%, other outcomes 2%.

For "above X%" markets, each strike's price is the probability the print exceeds it. The difference between neighboring strikes is the probability of landing between them, and the strike where the probability crosses 50% is the median.

One decimal matters. A CPI print of exactly 0.3% settles "above 0.3%" as NO.

#Step 2: Compare with consensus

Economists' consensus estimates are published on economic calendars before each release. Put the consensus next to the market's median, and note the date and source of the consensus. A market median a tenth above consensus is a disagreement. It becomes interesting only if you check who was closer over past releases.

#Step 3: The track record

For the last twelve releases, take the market's implied median the day before each release and compare it with the print. Report the average error, whether the market leaned high or low, and how often the print fell inside the market's 10–90 range.

Twelve observations is a case study. Report the counts, avoid significance claims, and look at the misses individually. One surprise print can dominate the average.

#Step 4: The path into the release

Daily prices on the two or three strikes nearest the median show how the market's view moved: after the prior release, after Fed speeches, after related data. A chart of the cut and hold probabilities over the eight weeks before a meeting tells that story directly.

#Pitfalls

• Reference month versus release date. Don't line up a CPI event with the wrong month's print.
• Far-dated events listed first. Markets for meetings years out trade alongside the next one. Pick the event that closes first.
• Thin early markets. Weeks before a release, spreads can be wide. Note the spread next to any probability you report.
• Look-ahead. A consensus figure revised after your snapshot time is not what the market saw.

#Run it in Jordan

"What does Kalshi think the Fed will do next?" is one of the starting questions in Jordan. Jordan finds the next meeting's markets, loads them onto a sheet, normalizes the outcomes with formulas and charts the distribution. For CPI and jobs, it can collect past releases priced the day before each one and line them up with the published prints. Related reading: Prediction market prices as probabilities.

#Frequently asked questions

#Do Kalshi CPI markets settle on revised data?

No. They settle on the first published value in the BLS release. Revisions after expiration do not change the result.

#Why do the Fed decision prices not add up to exactly 100%?

Each outcome is quoted with a bid-ask spread, so asks sum to more than $1 and bids to less. Normalizing mid prices gives a distribution that sums to 100%.

#Is the market better than the economists' consensus?

It depends on the release and the period, and a dozen observations a year cannot settle the question. Jordan can lay out the comparison release by release so you can judge it yourself.