OECD Data

What it is

The OECD publishes its statistics through the Data Explorer, a browser front end over an SDMX web service. Both halves are free, both are anonymous, and the API is the same service the web interface calls — pick a table in the browser, press the Developer API button, and it hands you the query string for the data and for its structure.

The important thing to know before writing any code is that the platform you may remember is gone. On 1 July 2024 the OECD took the OECD.Stat servers offline, and every SDMX endpoint under stats.oecd.org went with them. As of September 2026 the bare host still answers a 301 to the Data Explorer, which is worse than a clean failure — the redirect makes a dead integration look half-alive, while the actual data paths return a plain 404. The replacement base is sdmx.oecd.org/public/rest/. Nothing carries over automatically; the OECD publishes a mapping from old dataset codes to new ones and expects you to rewrite your queries against it.

Pricing

Free, in the way a statistical agency is free rather than the way a freemium API is. There is no key, no account, no paid tier and nothing held back. The one limit is throughput — 60 data downloads an hour, enforced per IP, and exceeding it gets you temporarily blocked. That cap counts CSV downloads from the browser interface too, so a colleague clicking export on the same office connection spends your allowance. Traffic from VPNs and anonymising services is not permitted at all — worth knowing before you point a cloud job at it.

Data & coverage

/dataflow/all listed 1,548 public dataflows in September 2026 — national accounts, consumer prices, trade, labour, health, education, tax, aid flows, house prices, government finance. The statistical universe is the OECD's 38 member countries, plus the five Key Partners (Brazil, China, India, Indonesia and South Africa) and, in specific datasets such as trade in value added and development aid, considerably wider.

Frequency runs monthly at the fastest. The monthly composite leading indicators were current to August 2026 when checked in mid-September; the structural datasets are years behind, and that gap is what you plan around. Responses come as SDMX-ML, SDMX-JSON or CSV, selected with a format parameter, and the CSV variant carrying human-readable labels beside the codes is the one to use.

Integrations

The OECD ships no client library — not for Python, not for anything. The name oecd is unclaimed on PyPI, so at least there is no stranger's package waiting for a guessed install. What people use are generic SDMX clients: sdmx1 and pandasdmx in Python, rsdmx and the community OECD package in R. You are integrating against a standard, not against a vendor.

Limitations

SDMX is the real cost of this data. A REST API returning JSON lets you guess a URL; SDMX does not. A data key is a positional, dot-separated list of dimension values in an order defined by the dataset's own structure, and getting the count wrong fails the whole request — one quarterly national accounts dataflow expects thirteen positions, and supplying twelve returns a 403 reading Not enough key values in query, expecting 13 got 12, with no indication of which dimension is missing. Discovering the right key means fetching and reading a separate structure document first. Budget hours, not minutes, for the first successful query.

It is not market data in any tradable sense. There is exactly one exception, and it is a macro indicator rather than a quote: the Short-Term Economic Statistics "Financial market" dataflow carries a national share price index, rebased to 2015 = 100, monthly and quarterly for 48 reference areas. Beyond that there is no individual security, no ticker, no closing price, no security identifier and no daily observation. If a strategy needs anything priced, this is not a source.

Much of it is old by design. The OECD's own guidance says most datasets are revised once or twice a year. Social expenditure aggregates still stopped at 2024 in September 2026, last refreshed in October 2025 — normal for official statistics, fatal for anything expecting a feed.

It is a rich-country club, not the world. Coverage is deepest for 38 mostly high-income members and thins quickly outside them.

The licence has a hole in it. Reuse, including commercial reuse, is permitted with citation — but the OECD explicitly does not warrant that it owns all rights in all data, and some datasets carry third-party restrictions. Checking the source metadata per dataset is your job, not theirs.

Alternatives

The obvious comparisons are the World Bank's and the IMF's own statistics portals, which cover more countries with shallower indicator sets and are easier to query. For anything priced, tradable or intraday, start from the market data category instead — the overlap with this card is one monthly index and nothing you could trade on.