Reading CPI without confusing momentum and base effects

Facts: seasonal adjustment answers a specific question

BLS explains that seasonal adjustment removes recurring within-year patterns, including effects associated with seasons, production cycles, holidays and sales. Its methodological discussion says adjusted data are usually preferred for short-term price analysis because they improve comparability across months.[6] The source used here is a May 2022 BLS article preserved by the Federal Reserve Bank of St. Louis, not a newly released inflation bulletin.

The same article describes direct adjustment of component indexes and indirect adjustment of aggregates assembled from components. It also explains annual recalculation of seasonal factors and revision of the previous five years of seasonally adjusted data.[6] These are measurement procedures, not a claim that any particular recent inflation reading has been revised. Always check the vintage of the series before comparing calculations made at different times.

Interpretation: separate three pieces of arithmetic

For our arithmetic framework, define monthly inflation as 100 × (this month’s index / last month’s index − 1), and annual inflation as 100 × (this month’s index / the same month one year earlier − 1). Both compare price levels, but their denominators differ. A year-over-year rate therefore cannot isolate the latest month’s momentum. Use a consistent series and adjustment convention within each calculation.

Consider a hypothetical unadjusted index of 100 a year before January and 104 a year before February. If the current readings are 110 in January and 111 in February, annual inflation falls from 10% to about 6.73%, even though the index rises about 0.91% between the current months. These invented numbers illustrate a base effect: the comparison base rose sharply. They are not observations from any actual CPI release.

A second exercise clarifies annualization. A constant monthly increase of 0.2%, repeated for twelve months, compounds to approximately 2.43%, using 100 × (1.002^12 − 1). That is a conditional run rate, not a forecast and not the observed twelve-month change. Reporting an annualized short window without its window length makes two genuinely different measures look interchangeable.

Watch: compare like with like

Build a worksheet containing the index name, adjustment status, reference month, publication vintage and calculation window. Examine both the recent monthly path and the year-earlier comparison levels. Then ask whether an apparent improvement reflects slower new price increases, a stronger base, or both. Do not mix an adjusted numerator with an unadjusted denominator; the seasonal method is designed to remove recurring patterns from a coherent series.[6]

Our suggested reading order is measurement first, arithmetic second, economic interpretation third. If one month looks unusually strong, test whether a longer window changes the impression. If annual inflation falls, inspect the outgoing comparison month before declaring a lasting improvement. Label these checks as analytical choices, not as additional official inflation indicators.

Uncertainty: adjustment is estimation, not purification

BLS’s pandemic discussion shows why exceptional shocks can interfere with estimation of normal seasonal patterns and require intervention modeling.[6] Adjustment therefore does not make every remaining move persistent or economically meaningful. This explainer supplies no current inflation reading and makes no prediction about the next release; its purpose is to keep different measurement questions separate.