Lilambd/world-prices

World Prices Long Record: what things cost, month by month, since 1960

🌐 日本語 · 中文 · Español · Português · Français · Deutsch · 한국어 · हिन्दी · العربية

• 日本語 — World Prices：世界の物価長期データ: モノの値段を月ごとに。1960年以降の一次産品価格70品目と指数16種、1970年以降の187か国の総合・食料・エネルギー消費者物価を収録。
• 中文 — World Prices（全球价格长期记录）: 逐月记录万物价格：自 1960 年以来的 70 种大宗商品价格和 16 项指数，以及 187 个国家自 1970 年以来的总体、食品和能源消费价格。
• Español — World Prices: serie histórica de precios mundiales: Lo que cuestan las cosas, mes a mes: 70 precios de materias primas y 16 índices desde 1960, más precios al consumidor generales, de alimentos y de energía para 187 países desde 1970.
• Português — World Prices: série histórica de preços mundiais e inflação: Quanto as coisas custam, mês a mês: 70 preços de commodities e 16 índices desde 1960, mais inflação ao consumidor geral, de alimentos e de energia para 187 países desde 1970.
• Français — World Prices : historique long des prix des matières premières et de l'inflation: Ce que coûtent les choses, mois par mois : 70 prix de matières premières et 16 indices depuis 1960, plus les prix à la consommation (ensemble, alimentation, énergie) de 187 pays depuis 1970.
• Deutsch — World Prices: lange Preisreihen für Rohstoffe und Verbraucherpreise: Was Dinge kosten, Monat für Monat: 70 Rohstoffpreise und 16 Indizes seit 1960, dazu Verbraucherpreise (gesamt, Lebensmittel, Energie) für 187 Länder seit 1970.
• 한국어 — World Prices: 원자재 가격·소비자물가 장기 데이터: 월별 물가 기록: 1960년부터 원자재 70종과 지수 16개, 1970년부터 187개국의 종합·식품·에너지 소비자물가.
• हिन्दी — World Prices दीर्घकालिक मूल्य रिकॉर्ड: चीज़ों की कीमत, महीने-दर-महीने: 1960 से 70 कमोडिटी कीमतें और 16 सूचकांक, साथ ही 1970 से 187 देशों के headline, खाद्य और ऊर्जा उपभोक्ता मूल्य।
• العربية — السجل الطويل للأسعار العالمية (World Prices Long Record): كم تكلّف الأشياء شهرًا بشهر: أسعار 70 سلعة و16 مؤشرًا منذ 1960، إضافةً إلى أسعار المستهلك العامة والغذاء والطاقة لـ 187 دولة منذ 1970.

One flat, long-form table of the prices that move the world, ready to join by country and month:

• 70 commodity prices and 16 indices from 1960-01 to 2026-08, monthly: crude oil (Brent, WTI, Dubai), natural gas (US, Europe, Japan LNG), coal, cocoa, coffee, tea, rice, wheat, maize, soybeans, palm oil, sugar, beef, chicken, shrimp, fertilisers (urea, DAP, potash), metals (copper, aluminium, nickel, zinc, tin, iron ore) and gold, silver, platinum.
• Consumer prices for 187 countries from 1970-01 to 2025-03, monthly: headline, food and energy CPI.
• Yearly panel adding Eurostat household gas and electricity prices and US state energy prices (EIA SEDS).
• Computed per-country features: how much food prices rose 2021-2023, how many months energy and food CPI lag world commodity prices, and how unstable headline inflation was 2015-2024.

Nothing is generated by a model. Values are as published by each source; the computed features are documented in feature_doc.

Data as of 2026-11-02 (commodity prices up to 2026-08; World Bank inflation database monthly up to 2025-03).

Quick start

from datasets import load_dataset
m = load_dataset("Lilambd/world-prices", "monthly", split="train")
oil = m.filter(lambda r: r["feature"] == "px_crude_oil_brent")
jp_food = m.filter(lambda r: r["unit"] == "JPN" and r["feature"] == "cpi_food")

Tables

config	rows	columns
monthly	215,365	unit_kind (world / country), unit (WLD or ISO3), grp, period (YYYY-MM), year, month, feature, value, source
yearly	27,289	unit_kind, unit, grp, year, feature, value, source
features_country	668	iso3, feature, value, year, unit, source
feature_doc	94	feature, unit_kind, ledger, kind, definition, source, caveat

Feature names: px_* are commodity prices in US dollars (units as in the World Bank Pink Sheet, e.g. $/bbl, $/mt, $/kg, $/troy oz); pxi_* are World Bank price indices (2010 = 100); cpi_* are national CPI index levels.

What the numbers say

• August 2026: crude oil average $84.4/bbl, gold $4,411/troy oz, Thai 5% rice $471/mt, arabica coffee $7.97/kg, cocoa $5.95/kg.
• Food prices from the 2021 average to the 2023 average rose most in Venezuela (×12.5), Lebanon (×12.1), Argentina (×4.3) and Turkey (×3.1).
• In 2016-2025 the most volatile monthly series were US natural gas, European natural gas and Kolkata tea; the most stable were EU sugar, timber and tobacco, which behave like administered prices.

Gaps and known limits

• CPI values are index levels and base years differ by country: compare changes, not levels, across countries.
• The World Bank inflation database is monthly only up to 2025-03; later months are not filled or estimated.
• Commodity prices are world benchmark prices, not local retail prices. For local market prices see Lilambd/world-markets.
• Japan's detailed BOJ and Statistics Bureau series and EIA US fuel series exist in the source model but are not in this first release.

Use it in production

For AI agents: 14 more Lilambd open-data APIs (airports, regions, markets, security, travel and more), pay per call with x402 (from $0.005 USDC on Base, no signup), free sample with sample=1, also as an MCP server: https://api.lilambd.com/llms.txt

Need a monthly feed, a price or country that is not here, or a finished answer ("how exposed is this country's food basket to world prices?")? Write to hello@lilambd.com.

Related: Lilambd/world-hazard, Lilambd/world-earth, Lilambd/world-signals.

Quick start: datasets, pandas, DuckDB

import pandas as pd  # hf:// paths need `pip install huggingface_hub`
BASE = "hf://datasets/Lilambd/world-prices/data/"
m = pd.read_csv(BASE + "monthly.csv")
oil = m[m["feature"] == "px_crude_oil_average"].set_index("period")["value"]
print(oil.tail())
jp_food = m[(m["unit"] == "JPN") & (m["feature"] == "cpi_food")]
print(jp_food[["period", "value"]].tail())
doc = pd.read_csv(BASE + "feature_doc.csv")  # what each feature means

-- DuckDB 0.10+ reads hf:// paths directly
-- yearly mean of the crude oil average price
SELECT year, avg(value) AS usd_per_bbl
FROM 'hf://datasets/Lilambd/world-prices/data/monthly.csv'
WHERE feature = 'px_crude_oil_average'
GROUP BY year ORDER BY year;

SELECT * FROM 'hf://datasets/Lilambd/world-prices/data/features_country.csv' WHERE iso3 = 'JPN';

A runnable notebook is in notebooks/quickstart.ipynb.

Use cases

• Long monthly commodity series (since 1960) for forecasting and regime studies.
• Pass-through analysis: how world food and energy prices reach national CPI.
• Country features on food-price shocks and inflation volatility for risk models.

Citation

@misc{lilambd2026_world_prices,
  author       = {Lilambd},
  title        = {Lilambd World Prices Long Record},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/Lilambd/world-prices}},
  url          = {https://huggingface.co/datasets/Lilambd/world-prices}
}

Please also credit the original sources listed in this card.

Sources and licence

• World Bank Commodity Price Data (Pink Sheet), monthly and annual. CC BY 4.0.
• World Bank Global Database of Inflation (Ha, Kose and Ohnsorge 2023). CC BY 4.0.
• Eurostat HICP and energy prices (nrg_pc_202 to 205). CC BY 4.0 under the Eurostat reuse policy.
• US Energy Information Administration, State Energy Data System. Public domain.

The derived tables are released under CC BY 4.0. Please cite "Lilambd World Prices Long Record" and the World Bank, Eurostat and EIA.

Maintained by Lilambd. Questions: hello@lilambd.com

Changelog

• 2026-11-03 — CPI duplicates removed from monthly; yearly CPI rebuilt. In the World Bank inflation database, 31 country × indicator combinations (29 cpi_energy, plus cpi_food for Singapore and Thailand) came as two different source series. Both had been loaded under the same key, so 4,669 (unit, period, feature) keys had two conflicting values. The two series are not the same index on another base year: their ratio is not constant. For each combination we now keep the series with the most months (ties broken by the later end month, then by source order) and drop the other. monthly goes from 220,053 to 215,365 rows (−4,688; 19 of these months existed only in the dropped series). Each (unit, period, feature) now appears once. Countries affected: BHR, BOL, BRA, BWA, CMR, DOM, ECU, EGY, GAB, IRQ, JOR, KWT, LBN, LSO, MOZ, MYS, NAM, OMN, PHL, PRY, QAT, SEN, SGP, TCD, TGO, THA, TUN, UGA, URY. The yearly CPI values for these combinations were recomputed from the corrected monthly data as the mean of full 12-month years. This adds the 392 years that had been missing because of the duplicates and removes Brazil 1999 cpi_energy, which had averaged the two series. yearly goes from 26,898 to 27,289 rows.