Inflation slowed. Does that mean prices fell?

How it works

CPI change over a period is calculated from two values of the same index: (new index / old index − 1) × 100%. The monthly rate compares adjacent months; the annual rate compares the current month with the same month a year earlier. When an unusually strong or weak month leaves the annual comparison base, the annual rate can move substantially even if the current monthly change is small.

BLS commonly presents the 1-month change on a seasonally adjusted basis and the 12-month change on a not seasonally adjusted basis. Do not mechanically add or subtract these series. A slower positive rate is disinflation; a decline in the index itself is deflation over the selected period. Index levels for different regions or populations are not direct comparisons of living-cost levels. All-items CPI is a weighted aggregate, so a simple average of category rates is wrong. Published relative importance can support an approximate contribution check, but it is not a permanent fixed weight; use the BLS-calculated index for the exact all-items rate.

A hypothetical example

All figures are hypothetical. The index is 100.0 one year ago, 102.5 one month ago and 103.0 now. Current monthly change: (103.0 / 102.5 − 1) × 100% ≈ +0.49%. Annual change: (103.0 / 100.0 − 1) × 100% = +3.0%. One month later, suppose the index rises to 103.1 while the new annual base is 100.8. The new monthly rate is about +0.10%, and the annual rate is about +2.28%. The annual rate slows to roughly 2.3% even though the index rises from 103.0 to 103.1. A separate simplified contribution exercise: category A has hypothetical importance of 80% and rises 0.2%; category B has 20% and rises 1.2%. A first-order approximation gives 0.80 × 0.2% + 0.20 × 1.2% = 0.40%. The simple average of 0.70% is wrong. This illustrates weighting; it does not replace the official index calculation.

Try it yourself

For historical practice, use the 14 October 2026 CPI release at 08:30 ET for September 2026. For the next release, open the current BLS schedule; this date belongs to an archived exercise. Before seeing the actual, record the reference month, adjustment basis and two comparison bases. After release, recompute m/m and y/y from index levels when available, then identify the changing-base effect separately. For two selected categories, estimate weighted contributions using published relative importance, label the result as an approximation, and compare it with the official all-items rate. Do not infer an asset-price direction from one CPI change.

Using the hypothetical levels 100.0, 102.5, 103.0 and 103.1 with the new annual base 100.8, calculate current and next m/m and y/y. Then estimate the aggregate rate for A and B with 80/20 importance and +0.2%/+1.2% changes. Where is the base effect, and why is the simple category average wrong?