US annual CPI can fall from its previous reading while exceeding a forecast. Those are different comparisons, not a contradiction. Identify the series, period and reference point before calling a release “good for the dollar.” The report itself supplies neither a currency direction nor the price at which an order can be filled.

Why can your prices differ from the average?

Energy rises 20% and other prices 2% in both custom baskets. Different initial shares change fixed-quantity costs.

Scroll sideways to see the full diagram

Why can your prices differ from the average?A starts energy 10, other 90; later 12, 91.8, total 103.8. B starts 30, 70; later 36, 71.4, total 107.4. Both initially cost 100. Changes 3.8%, 7.4%. Cost bars share a zero-based scale.Same prices, different fixed basketsEnergy:+20% • Other prices:+2%EnergyOther spending0255075100A before100A after103.8B before100B after107.4A: 10×1.20+90×1.02=103.80B: 30×1.20+70×1.02=107.40A cost:+3.8% • B cost:+7.4%
Our two baskets use identical categories and initially cost 100 hypothetical units. Quantities stay fixed; prices change identically over one period. No substitution, new products, quality or seasonal adjustments. This models fixed-quantity costs, not official CPI, utility or actual household spending. BLS explains that CPI describes a relevant population rather than every household. Numbers and baskets are invented.
Basket composition changes the experience of rising prices.

7.4%−3.8%=3.6 percentage points follows from energy shares, not different price readings. It does not prove official CPI wrong. Check scope, period and SA/NSA treatment before relating a report to Fed decisions.

What CPI really is, and who publishes it

The US Consumer Price Index measures changes in prices paid by urban consumers for a defined basket of goods and services. The Bureau of Labor Statistics publishes monthly and annual changes. Check whether a series is seasonally adjusted: a year-over-year comparison does not by itself eliminate every seasonal or base effect, and neither horizon universally matters more to markets.

Headline CPI covers all items; core excludes food and energy. This does not make excluded spending irrelevant to households or remaining prices solely interest-rate driven. There is no permanent core-first ranking of FX reactions. Record monthly or annual comparison and seasonal-adjustment status for each series; those are data identifiers, not a hierarchy of truth.

BLS recalculates seasonal factors annually and can revise the previous five years of seasonally adjusted indexes. Preserve the vintage known at publication in a historical test, rather than silently substituting later revisions.

Why CPI steers expectations for the Fed

The Federal Reserve’s statutory goals include maximum employment and price stability. Its longer-run inflation objective is 2 percent measured by the annual change in the headline PCE price index, rather than CPI. Policy decisions weigh the outlook and risks to both goals, so one CPI result does not mechanically determine whether rates rise or fall.

CPI is one source for assessing price pressures, not a mechanical vote for a particular Fed action. Policy also weighs employment, outlook and risk. Compare the same measure with its dated forecast and keep contradictory components rather than assigning one label to the report. See Fed decisions for the separate policy process.

CPI and PCE: scope and methodology

The Fed’s 2% longer-run objective concerns annual headline PCE price inflation, not core CPI. BEA identifies formula, weight, scope and other differences between PCE and CPI. PCE includes spending by and on behalf of the personal sector; CPI concerns household out-of-pocket spending. One CPI value cannot simply be copied into the next PCE estimate.

CPI usually becomes available before PCE for the same reference month, giving analysts information for estimating parts of the later report. That timing does not establish a universal ranking of currency reactions: surprises, revisions, other components and market context can matter in either release. See the PCE price index for their methodological differences.

One report can contain opposing comparisons

All following figures are hypothetical. Headline CPI year-on-year, NSA: previous 3.2%, forecast 2.9%, actual 3.0%. Headline month-on-month, SA: previous and forecast 0.2%, actual 0.3%. Core month-on-month, SA: previous and forecast 0.3%, actual 0.2%. Each forecast was recorded before publication for the same field. The yearly reading fell 0.2 percentage point but exceeded forecast by 0.1; headline monthly exceeded by 0.1 while core monthly missed by 0.1.

This is a mixed report. A percentage-point forecast difference is not extra monthly price growth, and one calendar supplier’s forecast is not an official BLS projection or every participant’s expectation. A hawkish interpretation requires more than the sign of one difference and cannot guarantee USD appreciation.

In a separate unadjusted index model, 100 → 103.2 → 106.296 over two years gives inflation 3.2% then 3.0%. Inflation slows while prices still rise. Another model holds the current index at 103 in January and February but moves last year’s base from 100 to 101: annual inflation falls from 3.00% to about 1.98%, despite zero current monthly change. This base effect is not a 1.02% fall in current prices; the examples are not reconstructed official series.

If USD actually appreciates against EUR, EUR/USD falls; against JPY, USD/JPY rises. A higher CPI alone guarantees neither. A simultaneous Treasury-yield move is another observation, not independent confirmation of a future trade outcome.

Why a CPI day is first of all about risk control

Fast conditions can produce a fill different from a requested level. An ordinary stop is not a guaranteed stop; limits also have different fill constraints. Check the provider, product and order terms rather than assuming every protective order fills at the first available price. IG’s discussion distinguishes order types and its own tolerance rules; those rules are not global broker terms. A fixed waiting period or small volume cannot eliminate execution risk.

Decide your exposure before the release and allow for execution risk. If a wider stop is appropriate, recalculate volume from its distance, pip value and costs so that it fits the planned risk budget; a wider stop and smaller lot are not automatically sufficient. Check the BLS release schedule and convert 8:30 a.m. Eastern to your local time, including daylight-saving differences. The economic calendar guide explains that routine.

Your next step

  1. Identify the event and data. Check the BLS date and 08:30 Eastern time, convert for that date, and retain reference month, field, SA/NSA, forecast source and timestamp.
  2. Save the original vintage. Record every relevant actual-versus-forecast difference and the released document. Later forecasts or revisions were not available at that moment.
  3. Define observation windows. For example, choose 5, 15 and 60 minutes beforehand and save a consistent price type, spread, other news and related-yield observations. These are observation conventions, not safe entry times. A few releases do not establish a profitable relationship.