Methodology
How the Coffee Inflation Tracker is built
The tracker follows commodity, shelf and cafe prices across 42 countries. This page lists every source, choice and exclusion behind it, so a fact-checker can retrace any number. Questions or corrections: hello@baristabench.com.
Primary series only, provenance on every row.
Commodity prices come from the World Bank's monthly Commodity Price Data (arabica and robusta, USD/kg, back to 1960). US retail comes from the Bureau of Labor Statistics: the average price of ground roast coffee per pound (series APU0000717311) and the CPI coffee item (CUUR0000SEFP01). European retail comes from Eurostat's HICP at item level, coffee and coffee substitutes, for every reporting country. Deflators are US CPI all items and each country's all-items HICP. Every observation in the downloadable dataset carries its source URL and retrieval date.
2020 = 100, and why.
All indexed series are rebased so their 2020 monthly average equals 100. 2020 is the last calendar year before the current commodity run-up began, every monthly layer covers it in full, and a reader's natural question is what happened since then. Countries whose price series does not cover all twelve months of 2020 are excluded from indexed outputs rather than rescaled from partial years.
Missing data is excluded, never estimated.
BLS did not publish the average-price series in every month of the 2000s; those months are simply absent, and index charts use the complete CPI series instead, with the average price as the dollar anchor. The same rule holds everywhere: no interpolation, no estimates, no filling from third-party databases.
A classification change, handled in the open.
European price statistics moved to a new product classification (COICOP 2018) in January 2026 and the classic Eurostat dataset was frozen at December 2025. This tracker uses the successor dataset, which recalculates the full history from 1996 under the new classification, so there is no break in the series. The coffee item is now coffee and coffee substitutes, marginally broader than the old pure-coffee item.
Pass-through and asymmetry, stated as co-movement.
The lag is the number of months, from zero to twenty-four, at which the correlation between twelve-month changes in the arabica price and twelve-month changes in the retail index is highest. The asymmetry compares two regression slopes: retail change on lagged commodity change, fitted separately for months where the lagged commodity move was positive versus negative. These are descriptions of how the series move together. They are consistent with slow, asymmetric pricing, and they do not prove intent by any roaster or cafe.
Commodity cycles via a fixed, dumb rule.
Cycle turning points are found with a zigzag: a peak or trough is confirmed once the three-month-smoothed arabica price reverses by 35 percent. The retail response is measured over the same window shifted by the pass-through lag, and the give-back figure is the share of the preceding retail climb surrendered by the end of the following crash, computed on index levels. No fitting, no discretion.
Country rankings that survive hyperinflation.
The headline ranking is the relative price change: the coffee index divided by the country's all-items index. Raw percentage rankings put Türkiye first by a mile, but that is a general-inflation story, not a coffee story, and the page says so where the raw number appears.
The cafe layer is labelled for what it is.
There is no official statistic for cafe prices, so the tracker archives a dated snapshot of Numbeo's crowd-sourced cappuccino price by country each quarter, starting July 2026. Crowd-sourced data is indicative, not official; it is included because a longitudinal archive of it does not otherwise exist, and it is never mixed into the official series.
Reproducible on request.
The pipeline is one command per stage (pull, normalize, analyze), every download is logged to a manifest with URL, timestamp and HTTP status, and 24 automated cross-checks verify the series against second publishers (FRED's mirrors of the BLS series, Eurostat's mirror of the ECB exchange rates) and against the sources' own published figures before anything ships.
Back to the tracker, or download the indexed dataset and country table.