How CPI Is Calculated: Basket, Weights, and Methodology

Every month, a national statistical agency announces a single number — the Consumer Price Index — that summarises the price changes of thousands of goods and services across an entire economy. That number looks simple, but behind it sits a highly structured, multi-stage process.

The five core stages are:

1. Define the basket — decide which goods and services to track
2. Set the weights — assign a relative importance to each item based on how much households spend on it
3. Collect prices — send field workers or automated scrapers to record actual prices
4. Compute the index — apply a mathematical formula to aggregate price changes
5. Publish — release the headline figure, the core figure, and detailed sub-indexes

This article explains each stage in plain English, with real-world numbers from the US Bureau of Labor Statistics (BLS), the UK Office for National Statistics (ONS), and other major agencies.

The CPI basket is a representative list of goods and services that consumers typically buy. It is not a literal shopping basket — it is a classification system that covers everything from a litre of milk to a university degree.

In the United States, the BLS divides the basket into more than 200 item categories, grouped into eight major classes:

• Food and beverages
• Housing
• Apparel
• Transportation
• Medical care
• Recreation
• Education and communication
• Other goods and services

Each major class is broken down further. Under "Housing," for example, the BLS tracks:

• Rent of primary residence
• Owners' equivalent rent of residences
• Fuel oil
• Electricity
• Natural gas
• Water and sewer maintenance
• Furniture and bedding

In total, the US CPI basket contains roughly 80,000 individual price quotes every month. The UK's basket is smaller — about 700 representative items — but covers a similarly broad range of spending.

Not every item in the basket is equally important. If households spend 30% of their income on housing and 1% on haircuts, a 5% rise in rent should move the overall index far more than a 5% rise in barber prices. Weights encode that logic.

Where the weights come from

Statistical agencies run large-scale household expenditure surveys — diaries and interviews that ask thousands of families what they buy and how much they spend. In the US, the Consumer Expenditure Surveys (CE) produce updated weights every two years. In the UK, the Living Costs and Food Survey feeds into the ONS weights annually.

Typical US CPI weights (2024–2025)

Housing dominates because it is the largest single outlay for most households. Energy is relatively small as a standalone share, but because it feeds into transportation and housing, its movements can still swing the overall index.

Urban vs. rural weights

The US CPI-U (Urban) covers about 93% of the population. A separate CPI-W (Wage Earners and Clerical Workers) uses a slightly different basket tilted toward the spending patterns of hourly workers. Social Security COLAs are still tied to CPI-W, though most policy discussion focuses on CPI-U.

A CPI is only as good as the price data feeding it. Statistical agencies use three main methods:

1. In-person field collection

In the US, about 400 BLS field workers visit roughly 23,000 retail and service establishments across 75 urban areas each month. They record the price of specific items — for example, a 12-ounce can of a particular brand of soup, or a specific men's haircut at a named salon. The item must be identical from month to month, so any change in the product itself is flagged for quality adjustment.

2. Web scraping and electronic data

For items with standardised pricing — airline tickets, groceries sold online, petrol, utility tariffs — agencies increasingly pull data directly from websites and corporate APIs. The UK ONS now collects a significant share of its prices electronically. This is faster and cheaper, but it works best for products where the specification is clear and unambiguous.

3. Administrative data

Some prices come from government records rather than market observation. Medicare reimbursements, council tax bands, and regulated utility rates are examples. These are highly accurate but may lag market prices by a month or more.

Once prices and weights are in hand, the agency computes the index using a mathematical formula. The most common starting point is the Laspeyres formula, which fixes the basket at a base-period set of quantities.

In plain English: multiply each item's current price by its base-period weight, add them all up, divide by the same calculation using base-period prices, and multiply by 100. If the basket cost £100 in the base year and costs £108 today, the index is 108.

Why Laspeyres has a bias

The Laspeyres formula tends to overstate inflation because it assumes consumers keep buying the same quantities even when relative prices change. In reality, if beef becomes expensive, shoppers buy more chicken. Because the formula does not capture that substitution, it records a larger price increase than consumers actually experience.

The Fisher ideal and chained CPI

To reduce this bias, the US introduced the Chained CPI (C-CPI-U) in 2002. It updates the basket every month rather than every two years, capturing substitution effects more quickly. The chained index typically runs 0.25–0.30 percentage points lower than the standard CPI-U over a full year.

Some economists prefer the Fisher ideal index, a geometric average of Laspeyres and its mirror image (Paasche). It is theoretically superior but computationally demanding, so it is used more in academic research than in official statistics.

If a laptop costs the same this year as last year but has twice the memory and a faster processor, its price has effectively fallen — even though the sticker price did not change. Quality adjustment is the process of stripping out those non-price changes so the index measures pure inflation, not product improvement.

Common quality-adjustment methods

• Hedonic regression — statistically decompose a product's price into its characteristics (screen size, battery life, RAM) and estimate the value of each
• Overlap pricing — when an old model is discontinued, price the new and old models side by side for a short overlap period to isolate the pure price difference
• Option cost / cost-to-produce — estimate how much it would cost to build the old product using current inputs
• Direct comparison — if the product change is trivial, assume the price difference is the pure price change

Substitution bias

There is a second, distinct bias: consumers switch to cheaper alternatives when relative prices change. If Gala apples rise and Granny Smiths fall, shoppers buy more Granny Smiths. The standard CPI does not fully capture this at the item level until the next expenditure survey updates the weights. The chained CPI mitigates it by updating weights more frequently.

The journey from an individual price quote to the national CPI number involves several layers of aggregation:

1. Elementary aggregates — prices for the same item in the same city are averaged (usually a geometric mean)
2. Item-area indexes — these elementary aggregates are combined into indexes for specific items in specific regions
3. Category indexes — item-area indexes are weighted together to form indexes for the 200+ CPI categories (e.g., "new vehicles" or "medical care services")
4. Major-group indexes — categories roll up into the eight major groups (food, housing, etc.)
5. All-items index — the major groups are weighted together to produce the headline CPI

At each step, the agency uses the expenditure weights to ensure that a price change in a heavily weighted category (like rent) moves the overall index more than a price change in a lightly weighted one (like apparel).

CPI indexes are anchored to a base period — a specific month or year when the index is set to 100. Over time, the base period becomes distant and the absolute index level grows unwieldy. The US CPI is currently indexed to 1982–84 = 100, which means the base-period basket cost $100 and the same (quality-adjusted) basket today costs several hundred dollars.

Rebasing

Rebasing resets the index to 100 at a more recent date. It does not change the inflation rate — it merely rescales the numbers. Many countries rebase every five or ten years. The US rebased to 1982–84 in 1988 and has not rebased since, though the BLS updates weights every two years within that base.

Chain linking

Chain linking is a technique that updates the basket more frequently than a full rebase. The index is computed as a series of short links — each link covers a period (say, one or two years) with fixed weights, and the links are multiplied together to form a continuous series. The US Chained CPI uses monthly chain links, which is why it better captures substitution.