Our World in Data — Crop Yields

Our World in Data — Crop Yields is the topic collection where the Our World in Data editorial team - Hannah Ritchie, Max Roser and Pablo Rosado - distils global agricultural statistics into 42 curated charts on how much food the world's cropland actually produces. Around 27 of those charts are single-crop yield series running from almond through banana, barley, bean, cashew nut, cassava, cocoa bean, coffee, corn, cotton, groundnut, lettuce, millet, oil palm fruit, orange, pea, potato, rapeseed, rice, rye, sorghum, soybean, sugar beet, sugar cane, sunflower seed and tomato to wheat. The remainder do analytical work the raw numbers never would: cereal yields plotted against fertilizer use, against GDP per capita and against extreme poverty rates; changes in cereal production, yield and land use tracked against population growth; cropland spared by yield gains; the land demand of vegetable oils; attainable yields; yield gaps; and nitrogen overapplication.

Underneath, the figures come principally from FAOSTAT's production accounts, processed by Our World in Data into one comparable country-year frame. The flagship wheat table alone carries 9,798 country-year rows spanning 1961 through 2024, and every chart keeps the same four-column skeleton.

Get a sample of this dataset and Datadory returns rows shaped exactly like the dictionary below, cut for the crops, countries and years you name.

Three observations exactly as they ship - the opening years of Afghanistan's wheat series beside Zimbabwe's most recent harvest year:

Entity       Code  Year  Wheat yield (tonnes per hectare)
Afghanistan  AFG   1961  1.0220001
Afghanistan  AFG   1962  0.9735001
Zimbabwe     ZWE   2024  4.8629003

Sixty-three years apart, identical shape. Afghanistan's 1961 row records 1.02 tonnes of wheat per hectare - the floor of one of the longest productivity stories in the panel - while Zimbabwe's 2024 row sits at 4.86 tonnes per hectare, nearly five times higher. Because every crop series keeps the same Entity-Code-Year-measure skeleton, both rows load into one table with no reshaping, and any of the other two-dozen-odd crops drops into the same frame with only its measure column renamed.

Four columns carry every chart in the collection. Three are constants - Entity, the ISO alpha-3 Code and Year - and the fourth carries the chart's subject, renamed per chart but always numeric and always in the same position. On the flagship chart it reads "Wheat yield" in tonnes per hectare; elsewhere it becomes yields of maize, rice or soy, cereal-yield averages, or the paired values behind the fertilizer, GDP and poverty comparisons.

Columns belonging to the other forty-one charts fold under additional fields on request: the full per-crop yield set, the cereal aggregates, the yield-versus-fertilizer and yield-versus-income pairings, the cropland-spared and vegetable-oil land-demand series, attainable yields, yield-gap measures, nitrogen overapplication, and the units-plus-lineage metadata sitting behind each measure. Definitions get confirmed against live records when we prepare your sample - which is also where column naming gets locked down for your pipeline.

Geography - global, countries plus world and regional aggregates. Every row identifies its entity by name and an ISO alpha-3 code that stays deliberately empty on aggregate rows, which is what lets a Zimbabwe-versus-world comparison and a forty-country panel come out of the same table.

Temporal - 1961 through 2024 on the FAOSTAT-based charts, with several long-run historical series reaching earlier still for some countries. Six decades of continuous annual observation makes this one of the longest uninterrupted productivity records in the industry slice.

Granularity - strictly country-year, one row per entity per crop or indicator. Individual chart tables run from roughly 3,000 to 10,000 rows, small enough to hold in memory whole and large enough to span every producing country on earth.

Against the wider Datadory catalog - where the average quality score across all 1,744 datasets is 7.81 - this record scores 8/10, carried by its verified four-field dictionary and the reach of the time series.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; the rows arrive identical either way - cleaned, typed and documented against the field dictionary above. The four-column skeleton joins to price, trade or macro panels on Entity, Code and Year without a crosswalk, aggregate rows filter out on the blank Code when you want countries only, and chart-specific measure columns arrive mapped to the names your models expect. Cadence changes are a settings conversation, not a re-integration project.

• Food-security and hunger analysis - six decades of staple yields per country put the Green Revolution, its plateaus and its laggards on one axis, ready to sit beside population and poverty series already in the collection.
• Agricultural lending and insurance risk - country-level productivity trajectories anchor credit committees and underwriting assumptions for input suppliers, processors and farm lenders working across borders.
• Commodity research context - long-run yield trends frame supply-side narratives for grain, oilseed and soft-commodity theses before any position is taken.
• Land-use and sustainability reporting - cropland spared by yield gains, vegetable-oil land demand and nitrogen overapplication give corporate and policy reporting pre-built efficiency metrics rather than raw tonnages.
• Model training - a fixed-schema, sixty-plus-year country x crop x year panel is rare training material, with no schema drift between vintages to debug.
• Teaching and journalism - the yield-versus-fertilizer and yield-versus-poverty pairings are ready-made ledes, with citation-grade provenance preserved behind every figure.

Market researchers and consultants (relevance 3/3) get land-use and yield-gap narratives that drop straight into decks; see market researchers use cases. Journalists, academics and students (3/3) get the chart-first, provenance-preserved package they are built to cite; see journalists academics use cases. Data scientists and ML engineers (2/3) get a cleaned multi-decade yield panel that trains without preprocessing archaeology; see data scientists use cases. Investors and quant researchers (2/3) get fast sanity checks on long-run productivity trends before deeper commodity pulls; see investors quants use cases. Developers and builders (2/3) get stable join keys - Entity, Code, Year - for dashboard integrations; see developers builders use cases. Competitive intel teams (1/3) get sector-wide efficiency context rather than named-competitor tracking; see competitive intel product teams use cases.

For contrast within the same industry: USDA NASS Quick Stats goes below the county line but stops at the US border, FAOSTAT - Detailed Trade Matrix carries flows rather than productivity, and the NASS portal versus Our World in Data comparison scores both head to head.

Three things worth knowing upfront. First, aggregates share the columns with countries: regional and world rows carry a blank Code, so filter before averaging or the world total quietly tops your country league table. Second, the measure column renames itself per chart - "Wheat yield" here becomes a different header on the maize chart - so stacking multiple crops needs a small mapping step, which the sample locks down for you. Third, this collection is scoped to yields and land-use productivity: prices, trade flows and farm finances live elsewhere, so pair it with IndexMundi agricultural commodity prices or the World Bank agriculture and rural development indicators when the question widens. Samples precede any commitment: get a sample of this dataset shaped to your crops, countries and years, dictionary and normalization notes attached.