Inflation prediction markets: CPI definitions and revisions

One CPI release, several different numbers

The Consumer Price Index (CPI) measures how the price of a basket of goods and services changes over time, and the US Bureau of Labor Statistics publishes it in more than one form. A seasonally adjusted series strips out patterns that repeat every calendar year so that month-to-month movements are easier to compare; the unadjusted series keeps the raw price level. A single CPI release can carry seasonally adjusted and unadjusted first-published figures, and later revisions can add a different vintage. Add the choice between the first published figure and a later revised one, and a single release date can leave you with several numbers that all get called inflation.

Separately, the monthly change and the annual change are 2 distinct statistics, not 2 views of one. The annual figure compares this month with the same month a year earlier; the monthly figure compares this month with the last. A contract naming only CPI is therefore incomplete until it says which of these it settles on. If you want the wider macro picture around this, the economy overview is the natural starting point.

A worked example: index 100 to 103, and one month inside that year

Illustrative numbers only, not a study of any real release. Suppose an index starts at 100 and ends a year later at 103: the annual change is 3 percent. Inside that path, the previous month was 102.8 and the current month is 103. The monthly change is (103 - 102.8) / 102.8, which equals roughly 0.1945525 percent. Feel free to repeat the arithmetic on a calculator, since the value of the example sits entirely in the calculation.

Both of those figures can honestly be described as inflation over coffee, yet they differ by a factor of about 15. A market resolving on a 3 percent threshold is not the same bet as one resolving on a 0.2 percent threshold, even when the headlines use the same word for both. Before committing any size, write down which statistic the rules name, in what units, and over what window.

Definition choices that decide the payout

Seasonal adjustment is not a one-time subtraction: the BLS recalculates seasonal factors every year with the January data, and the prior 5 years of seasonally adjusted figures can be revised as a result. A number you watched on a screen in the past may not be the number standing in the record today. If the contract settles on a revised series, the figure that decides your payout can differ from the one you saw when you entered.

That single mechanism turns first published and revised into 2 separate bets wearing one headline. Revisions can occur, and a revision alone on a tight threshold can flip the outcome. Read the rule text for the exact series and vintage before assuming 2 CPI markets are interchangeable.

Rounding, 0.19 versus 0.20, and a hypothetical revision

Here is the losing case spelled out. A market asks whether the monthly change is at or above 0.2 percent, and the illustrative value is 0.1945525 percent: under a simple inclusive rule, that resolves DOWN. The annual figure of 3 percent never enters the decision. Had the rule read 0.19 percent or above, the same number would have resolved UP. The statistic and the threshold, not the narrative, settle it.

Now suppose the contract references a revised series and the revised monthly change comes out at 0.21 percent, which is a hypothetical change invented for this walkthrough and not an observation of anything. The first outcome and the final outcome then differ on the same economic reality. 3 things have to be confirmed before entry: which statistic (monthly or annual, adjusted or unadjusted), which vintage (first published or revised), and which threshold comparison (strict or inclusive). If any of the 3 is ambiguous in the rules, you cannot price the bet.

Rounding, headline vs core, timezone, and which month

4 further definitions routinely decide outcomes. Rounding decides whether a print sitting exactly on the boundary counts as up or down. Headline versus core matters because the 2 are distinct measures, so which one the contract names must be stated. Timezone matters because a deadline written in one zone can fall on a different calendar day in another. Month and version matter because February data released in March is not the same series as the annual revision of it.

A market's rules, never its title, fix the source, the deadline and the edge cases. Where the rule page is silent on any of the 4 items above, treat the market as higher-risk and accept that the resolution source decides, not your own reading of the data.

How resolution pays, and why the same view can pay differently

In an ordinary binary market the winning share pays 1 and the losing share pays 0. Rare unknown or 50-50 resolutions can pay 0.50 per share, and that is a payout event, not a universal cancellation or refund: your capital does not automatically come back at your entry price. Check the specific rule text instead of assuming a refund.

Resolution mechanisms differ too. Ordinary prediction markets resolve through UMA, while some short-duration Up/Down markets use a Chainlink time-weighted average price (TWAP), which averages the stream between a start and an end reading. For those, a final reading at or above the start resolves UP, and lower resolves DOWN. Different exchange candles can disagree, so the same price can produce different outcomes on different venues. None of this generalises across every duration or source.

The consequence is practical: one macro view can pay out in several ways depending on which contract carries it. When your main question is how a market reaches its decision rather than what the number will be, the resolution guide covers rule sources, deadlines and edge cases in more detail.

What this example does and does not tell you

Every figure in this guide is authored for teaching: the 102.8-to-103 path, the 0.1945525 percent monthly change and the hypothetical 0.21 percent revision are constructed, not observed, and no specific release month is predicted here. The mechanics described come from published documentation of CPI seasonal adjustment and prediction-market resolution; the arithmetic is there to show how a boundary decision behaves, not to estimate any probability.

If your position depends on interest-rate expectations rather than the price index alone, the Federal Reserve guide is the adjacent topic worth reading before you size anything. Neither that guide nor this one removes the need to check the live rule text and the BLS release schedule for the specific contract you are considering.