Our World in Data

What You're Getting

Our World in Data (OWID) aggregates and republishes cleaned datasets on global development, health, energy, environment, economics, and population — primarily for visualization. They source from WHO, World Bank, FAO, UN, and others, adding significant cleaning and harmonization work. Data is available as CSV per chart/indicator. Coverage varies; most series are annual. OWID is best for analysis and exploration, not as a primary pipeline source.

Ingestion Strategy

Bulk CSV download from the OWID GitHub repository (owid/owid-datasets) is the most reliable pipeline approach — each dataset is a CSV in a named folder. For specific indicators embedded in their charts, the catalog at ourworldindata.org/charts links to underlying data CSVs. No API is available; scraping is not recommended.

import requests, pandas as pd, io

# Download a specific OWID dataset from GitHub
url = "https://raw.githubusercontent.com/owid/owid-datasets/master/datasets/CO2%20emissions%20by%20fuel%20%26%20industry%20-%20GCP/CO2%20emissions%20by%20fuel%20%26%20industry%20-%20GCP.csv"
r = requests.get(url)
df = pd.read_csv(io.StringIO(r.text))

Rate Limits & API Details

• No API; GitHub raw file downloads rate-limited at standard GitHub rates
• GitHub repo: github.com/owid/owid-datasets
• Catalog JSON: ourworldindata.org/charts — no machine-readable index; browse manually
• owid-grapher GitHub repo contains the chart config with data references

Schema Stability

OWID datasets are maintained by their research team — structure can change without notice when they revise methodology. Country names use OWID's standard names (not always ISO codes) — build a mapping table. Entity column includes both countries and regions/aggregates. Year column is always a 4-digit integer.

Data Quality Gotchas

• Not a primary source: OWID re-publishes and often lightly transforms source data — trace back to primary source for pipeline use
• Country name harmonization: OWID uses its own country names ("Democratic Republic of Congo" vs "DRC") — map to ISO codes in your staging layer
• Vintage not tracked: OWID datasets are updated in place; historical vintage data is not preserved (use source data directly for point-in-time needs)
• Mixed sources: a single OWID dataset may combine data from multiple sources with different methodologies across time periods

Tool Compatibility

• Python: direct requests + pandas CSV download
• R: direct download
• GitHub API can enumerate datasets in the owid-datasets repo
• Not compatible with standard Airbyte connectors

Benchmarking Approach

Use OWID as a cross-check against your primary source pipelines rather than as a primary source itself. For CO2 emissions: global total should be ~37 Gt CO2 for 2022. Verify country count in dataset matches expected geographic coverage. Alert if the GitHub commit timestamp for a dataset is more than 12 months old (indicates it may be abandoned).