The Basket of Goods, the Laspeyres Formula, and the Geometric Mean: How CPI, RPI, and HICP Measure International Inflation

In January 1999, the US Bureau of Labor Statistics fundamentally altered how it calculates the Consumer Price Index by introducing the geometric mean for most lower-level item categories. This mathematical shift immediately reduced the reported annual inflation rate by roughly 0.2 percentage points, changing how cost-of-living adjustments were paid out to millions of Americans. The change acknowledged a basic consumer reality: when the price of beef spikes, shoppers buy chicken.[1][5]

The measurement of inflation relies on a "basket of goods," a representative sample of over 80,000 items consumers buy daily. However, the formula used to aggregate the prices of these items dictates the final inflation figure. In the past, statistical agencies relied heavily on the Laspeyres index, a formula that measures the cost of a fixed basket of goods over time. As the International Monetary Fund notes in its 2020 Consumer Price Index Manual, the Laspeyres approach provides a clear baseline but suffers from "substitution bias" because it assumes consumers continue purchasing the exact same quantities even as relative prices change.[2]

To correct this bias, statisticians introduced the geometric mean. Unlike the arithmetic mean, which simply averages price changes, the geometric mean multiplies the prices and takes the nth root. This mathematical property implicitly assumes that consumers will substitute away from goods that have become relatively more expensive, keeping their total expenditure share constant. The Bureau of Labor Statistics found that incorporating this formula at the lowest level of aggregation—such as choosing between different types of apples—better reflects actual purchasing behavior.[1][5]

The divergence in formulas explains why different countries, and even different indices within the same country, report varying inflation rates for identical economic conditions. In the United Kingdom, the Retail Prices Index (RPI), introduced in 1947, has historically used the Carli formula, an arithmetic mean of price relatives. The Office for Budget Responsibility reported in October 2024 that the mathematical difference between the Carli formula used in the RPI and the Jevons formula (a geometric mean) used in the Consumer Prices Index (CPI) creates a persistent wedge.[4]

"The formula effect is the largest single contributor to the long-run difference between RPI and CPI inflation," the Office for Budget Responsibility stated, quantifying this mathematical wedge at 0.8 to 0.9 percentage points annually. Because the Carli formula fails the time-reversal test—meaning that if prices rise and then fall back to their original level, the index shows a net increase—it systematically overstates inflation compared to the geometric mean. This wedge compounds to nearly a 9% difference in measured price levels over a 10-year period.[4]

Across the English Channel, the European Central Bank relies on the Harmonized Index of Consumer Prices (HICP) to set monetary policy for the eurozone. The conceptual foundations of the HICP strictly prohibit the use of the Carli formula due to its upward bias. Instead, European statistical agencies use either the Jevons formula (geometric mean) or the Dutot formula (ratio of arithmetic means) for elementary aggregates.[3][6]

"The choice of the index number formula at the elementary level is of crucial importance," the European Central Bank noted in its working paper on HICP foundations, emphasizing that harmonization requires all member states to abandon biased arithmetic estimators. By standardizing the geometric mean across 20 member states, the ECB ensures that a 2% inflation reading in Germany is mathematically comparable to a 2% reading in France.[3]

The transition away from fixed-basket arithmetic formulas has profound financial consequences. In the US, the shift to the geometric mean lowered federal outlays for Social Security, which is indexed to the CPI. In the UK, the government is phasing out the RPI entirely by 2030, aligning the issuance of inflation-linked government bonds with the CPIH, a metric that incorporates the geometric mean and housing costs. The inflation rate a nation reports is not just a reflection of store prices, but a product of the specific mathematical lens used to aggregate them.[4][7]