How We Calculate Inflation & Purchasing Power

What the labels mean

• Published annual average: a value in a verified source snapshot. Currently this covers US CPI-U for every year from 1913 through 2024, on the 1982-84=100 base, not seasonally adjusted. BLS source table, annual-average column. Snapshot checked 2026-09-05; this is not a live feed.
• Linearly interpolated: no annual observation for that year is loaded in our dataset. We estimate between the displayed lower and upper years: CPI = lower CPI + (upper CPI − lower CPI) × (year − lower year) / (upper year − lower year). This is not an observed CPI, and endpoints can themselves be unverified estimates.
• Unverified legacy estimate: a number stored for the requested year, but without a verified observation-level source and period. Being an exact dataset entry does not make it an official observation. This includes US years outside the verified snapshot and the current legacy datasets for the other countries.

Missing from our dataset does not mean missing from official statistics. Linear interpolation can miss wars, deflation and inflation spikes by a substantial margin; we do not claim a fixed error bound or official-calculator accuracy for interpolated results. We do not extrapolate outside bracketing CPI points.

France and Germany: currency units are part of the calculation

France uses the 125 annual inputs published in the INSEE franc–euro converter, 1901–2025, retaining the source precision. The calculation applies the price-index ratio and a separate denomination factor: 100 old francs = 1 French franc; 6.55957 French francs = 1 euro. Older dates default to old francs before 1960, then French francs until 2001, then euros. For example, 3.80 French francs in 1995 ≈ €0.93 in 2025. Early price indices are coarsely rounded; historical household and geographical coverage changes.

Germany uses 150 published Bundesbank purchasing-power coefficients: Mark from 1873, Reichsmark from 1924, Deutsche Mark from 1949 and euro from 2002. Each coefficient already gives 2025 euros per historical unit. An amount in year A is multiplied by coefficient A and divided by coefficient B for a different target year. For example, 200 DEM in 1950 × 3.39 = €678 in 2025. Dividing by the DEM/euro conversion rate again would be an error.

These are purchasing-power comparisons, not statutory currency redemption rates. A multiplier spanning different currency units is not an inflation percentage. German crisis and war periods have major uncertainty; the DEM series does not represent East German marks. No whole-year result is provided for 1922–1923 (hyperinflation) or 1948 (separate RM and DEM half-years). No missing observation is interpolated in either new model.

Earlier unsupported URLs remain accessible with a noindex data-availability explanation, not a fabricated amount. Existing post-2001 franc/mark URLs retain the explicitly named legacy unit rather than silently becoming euros. Wage, goods-price, wealth-rank and fixed-unit money-decline comparisons for France/Germany are withheld until their currency units are verified.

Wages, goods and missing years

Salary, bread, milk, rent and gasoline values are separate legacy estimates. A verified CPI does not verify these prices. We only show a field recorded for the requested year; we never relabel a nearby year, interpolate wages from CPI, or fill a missing commodity from another year. Missing values are omitted and missing-year context is disclosed.

The Historical Rich-O-Meter cannot produce a wage comparison without that year's wage data.

Its tiers are calibrated with the World Inequality Database: your salary, converted to the chosen year's money, is placed among the pre-tax incomes of all adults that year (income thresholds at 14 percentiles, split equally between spouses and including capital income and pensions). Bottom 20% is Struggling, 20–50% Working Class, 50–90% Middle Class, 90–99% Upper Middle Class and the top 1% Wealthy Elite. The rank is linear between the published thresholds and shows WID's own 0–5 data-quality grade. When a year has no distribution but one exists within 10 years, that year's thresholds are applied as multiples of average earnings and the rank is labelled approximate; further away, only the illustrative earnings-ratio band is shown. Because all adults are included, average earnings usually sit between the 50th and 80th percentile.

The Historical Rich-O-Meter now uses its own dataset, separate from the inflation calculator, for all six countries: the United Kingdom (1750–2025), Canada (1917–2025), the United States (1860–2024), Germany (1891–2025), Japan (1899–1938 and 1947–2025) and France (1901–2024). Each country pairs an average-earnings series with a price or purchasing-power series, stored exactly as published, with every annualisation, rebasing and currency conversion recorded per period. Your salary is treated as pay in the dataset's latest complete year, converted to the chosen year's prices and divided by that year's average earnings. It shows the earnings multiple, not a percentile.

Some years are deliberately absent rather than estimated: 1922, 1923 and 1948 for Germany (hyperinflation and the currency reform), 1939–1946 for Japan (no consumer price index in the source), and 1999 for France (the two earnings sources do not meet). Where a definition changes — for example the switch to the US national average wage index in 1951 or to Canadian survey-based earnings in 1983 — the result names the change and the size of the level shift. See sources.

Why calculators can differ

Compare the same country, index series, currency units and reference periods. Annual averages and a specific month's index produce different answers. The Bank of England uses annual averages for past years and the latest available month for the current year, and explains the modelled historical series it uses. The Bank of Canada uses monthly CPI data from 1914. These are different period conventions, not different ratio formulas.

We have not yet imported all available national annual series or a verified current-month pipeline. Those are necessary before claiming comparable accuracy across every supported country and year.