US CPI, the consumer price index

What the index measures 

The consumer price index measures the change in the prices paid by consumers for a representative basket of goods and services. It is a price index, not a cost of living index, and the Bureau of Labor Statistics is explicit about the distinction: the index tracks what a fixed pattern of consumption costs from period to period, and it does not attempt to measure what it costs to maintain a constant standard of living as consumption patterns, product quality and household circumstances change.

The headline series covers urban consumers, a population definition that accounts for the large majority of the country but not all of it. A second series covering urban wage earners and clerical workers is published alongside it and is the one used to index certain benefit payments, which is why two different official inflation rates for the same month can both be correct and both be quoted.

How the number is built 

Two separate collection efforts feed the index. Field staff and automated collection gather individual prices for specific items at specific outlets across metropolitan areas, and a separate expenditure survey of households establishes how much of the average budget each category absorbs. Prices provide the movement, expenditure provides the weight, and the index is the weighted aggregate of the two.

The largest single weight in the basket is shelter, and it is not measured the way an intuition would suggest. Rented accommodation is measured by rents actually paid. Owner-occupied housing is measured by owners' equivalent rent, an estimate of what an owner would pay to rent the dwelling they live in, because the purchase of a house is treated as an investment rather than as consumption. This single methodological choice gives the index a substantial component that is derived rather than observed, and that moves slowly because it is built from rental contracts which reset infrequently.

Quality change is handled by hedonic adjustment. When a product is replaced by a version with different characteristics, the statistician estimates how much of the price difference is attributable to the change in characteristics and removes it, so that the index reflects the price of a constant quality item. The technique is standard, published and contested at the margins, and it is one of the reasons measured inflation and observed shelf prices can diverge.

Headline and core 

Two aggregates are published from the same collection. Headline CPI covers the whole basket. Core CPI excludes food and energy. The exclusion is not a claim that food and energy do not matter to households; it is a claim that their prices are dominated by supply shocks, weather and global commodity cycles, and that those components therefore carry a great deal of month to month noise relative to the underlying trend a central bank is trying to observe.

Analysts also decompose the core into goods and services, and further into shelter and core services excluding shelter, on the reasoning that these components respond to quite different forces: goods prices to global supply chains and the exchange rate, shelter to the housing market and to the slow reset of rental contracts, and remaining services to domestic labour costs. Those decompositions are analytical constructions, assembled by practitioners from published sub-indices, rather than official aggregates.

The monthly rate and the annual rate 

Every release carries two rates that answer different questions. The month on month rate is the change from the previous month, seasonally adjusted, and it describes what happened recently. The year on year rate is the change from the same month a year earlier, conventionally quoted unadjusted, and it describes the cumulative effect of the preceding twelve monthly changes.

The relationship between them produces the base effect, which is one of the most frequently misread features of any inflation release. Because the annual rate drops the month from a year ago and adds the current month, it can fall while current prices are rising, or rise while current prices are flat, purely because of the size of the month that has just left the window. The annual rate is as much a statement about last year as about this month.

Seasonal adjustment and revision 

The monthly figure is seasonally adjusted, which means a statistical model has removed the regular within-year pattern so that consecutive months can be compared. The adjustment factors are re-estimated annually against the accumulated history, and that re-estimation changes previously published seasonally adjusted months without any new price data being collected. The unadjusted index itself is not revised in the ordinary course, so the CPI has the unusual property that the headline monthly rate can change while the underlying index does not.

The basket's weights are also updated periodically to reflect newer expenditure surveys. A reweighting does not restate history, but it changes how future monthly price movements aggregate, so a component that was previously a minor influence on the headline can become a material one without its own price behaviour changing at all.

Cadence and publication 

The index is published monthly, in the middle of the month following the reference month, in the morning Eastern Time, under an embargo arrangement that releases the figures to all wires at the same instant. The release schedule for the whole year is published in advance by the Bureau of Labor Statistics. The accompanying report carries the full component detail, and the sub-indices practitioners read are available at the same moment as the headline, not later.

Two related releases sit near it on the calendar and are frequently confused with it. The producer price index measures prices received by domestic producers rather than paid by consumers, and is published within days of CPI. The import and export price indices measure prices at the border. Neither is a component of CPI, and the common description of PPI as a leading indicator of CPI is a rough empirical claim about correlation, not a definitional relationship.

How a release is conventionally read 

The dominant convention reads the core monthly rate first, on the reasoning that it is the least contaminated by transitory supply effects and the most closely related to the trend a central bank is assessing, and reads the headline and the annual rate second. A further convention weighs the composition of the print, distinguishing a rise concentrated in a single volatile category from one distributed across many.