§Methodology & data

Methodology & data

Last reviewed: 19 August 2026. This page explains where every number on Costely comes from, which numbers are measured and which we work out ourselves, and where each measure stops being reliable.

The rule the whole site follows

A figure is either reproduced as published by an official statistical office, or it is a Costely calculation built on such figures. Every table says which is which, in the row itself and again in the source line beneath. No number on this site comes from an estimate, a survey of readers, a crowdsourced database or a model of our own invention.

Sources, country pages

• Price level index — World Bank, World Development Indicators, indicator PA.NUS.PRVT.PLI, United States = 100. Licence CC BY 4.0.
• Category basket, 25 groups — World Bank, International Comparison Program, benchmark round 2021, World = 100. Licence CC BY 4.0. The ICP runs in benchmark cycles rather than annually; 2021 is the latest published round, released in May 2024, and the next round is 2024 with results expected around 2027.
• Food price levels in Europe — Eurostat, price level indices by expenditure category (prc_ppp_ind), EU27 = 100, reference year 2024, nine food groups. Shown only when both countries of a pair are covered by Eurostat.
• Household energy prices in Europe — Eurostat, bi-annual energy prices for household consumers (nrg_pc_204 electricity, nrg_pc_202 gas), euro per kilowatt-hour including all taxes and levies.
• Pay in Europe — Eurostat, Structure of Earnings Survey: median monthly earnings, and mean monthly earnings for nine occupation groups (earn_ses22_21). The survey runs every four years; 2022 is the latest round.
• Homicide rate — UNODC, distributed through the World Bank as VC.IHR.PSRC.P5, latest year available per country.
• Climate — NASA POWER, monthly temperature and rainfall for capital cities.
• Visa-free travel — Passport Index (Arton Capital), MIT licence.
• Wages in Canada — Statistics Canada.

Sources, United States pages

• Regional price parities — Bureau of Economic Analysis, table SARPP, United States = 100.
• Grocery prices — Bureau of Labor Statistics, Average Price Data, U.S. city average. State columns are the national price multiplied by the state goods parity, and are labelled as Costely estimates.
• Pay by occupation — BLS Occupational Employment and Wage Statistics, annual median wage, 40 occupations across all states.
• Private-sector pay — BLS Quarterly Census of Employment and Wages, average weekly wage converted to a month.
• Housing and property tax — U.S. Census Bureau, American Community Survey five-year estimates: B25064, B25077, B25103, B19013, B25088, B25071, B25132.
• Energy prices — U.S. Energy Information Administration: residential electricity and natural gas as published; petrol converted by us from the State Energy Data System.
• Safety — FBI Crime Data Explorer, homicide rate per 100,000.

What we calculate, and how

• Safety score — clamp(105 − 20 × ln(1 + rate), 0, 100) on the homicide rate. A transformation of ours, never an official rating.
• Difference columns — (B − A) / A, rounded to whole per cent.
• Same lifestyle, new price — 1,000 × price level of the dearer country / price level of the cheaper one. The 1,000 is a round illustrative amount, not a converted sum.
• Budget planner — your own figures repriced line by line with the ICP ratio of the matching category. Filling with local averages uses World Bank household spending per person, scaled with the OECD-modified equivalence scale (first adult 1.0, further adults 0.5, children 0.3) and split across lines with our own shares.
• State grocery prices — national measured price × state goods parity. Because one national price runs through one parity, the gap between two states is identical on every row.
• Effective property tax rate — median tax paid / median home value. Home value against income — median value / median household income.
• Petrol per gallon in the states — converted from dollars per million Btu at the EIA factor of 5.053 million Btu per 42-gallon barrel.

Why the years differ

Each series has its own release cycle, and we do not paper over the gaps. The world category basket is a 2021 benchmark; European food price levels are 2024; European pay is 2022; U.S. occupation wages are May 2025; European energy prices are half-yearly and a few months old. Comparing two numbers from different vintages means comparing slightly different moments, and where that matters we say so on the page.

Coverage rules

Eurostat permits redistribution of its data for the EU, EFTA and candidate countries, so European blocks appear only for those, and only when both countries of a pair are covered. A pair with one country outside that group keeps the World Bank basket, which is worldwide but older. The two bases — EU27 = 100 and World = 100 — are never mixed in one table.

Known limits

• Price levels describe whole economies or whole categories, not your shopping basket, and not one city.
• European pay figures are means; large salaries pull them up. The U.S. occupation figures are medians. The two are not interchangeable.
• Europe publishes pay for nine broad occupation groups only. The 830 named occupations of the U.S. survey have no European equivalent.
• The BLS suppresses occupations with too few workers in a state; those cells show a dash rather than a guess.
• Reporting to the FBI is voluntary, so agency coverage varies by state and year.
• We publish no derived euro prices for European food. We tested the method — anchoring one country and scaling by a category price level — against measured electricity prices and found a median error of 16 per cent and a worst case of 46 per cent. It was rejected on those grounds.

Updates and corrections

Datasets are refreshed from the publishers on the schedule each publisher keeps; every table prints its own reference period and the date we retrieved it. If a figure here contradicts its source, write to us with the page address and we will correct it and say so.