Where can I get oil and gas production data by country?

The question reads like a sourcing errand, and half the internet answers it like one - a page of portals, each with its own footnote. It is really a choice between three different quantities wearing one flag.

Compiled national statistics put one number per country per period on a harmonized basis - the figure behind rankings, market sizings and any chart where Saudi Arabia and Brazil have to share an axis. Country-reported monthly boards publish what each government itself submitted on a shared questionnaire: production, trade, stocks and demand, graded for submission quality. Bottom-up rollups keep the field, lease and well records beneath the national total, every row carrying an operator, a coordinate and a status.

They are not substitutes. Compiled series buy comparability and sell away depth; boards buy timeliness and self-reporting and stop at the border; the asset layer buys the individual unit and leaves the summing to you. Pick deliberately - everything below assumes you did.

Four records carry genuine cross-country coverage, and they divide the work by definition rather than by anything else.

EIA International Energy Statistics (quality 9) holds petroleum and natural gas series for 218 countries and territories - crude including lease condensate across production, proved reserves, imports and exports, plus dry natural gas production, with world, regional and OPEC/EU aggregations tagged beside the country rows - at annual, quarterly and monthly frequencies, its newest verified point dated April 2026. JODI-Oil World Database (quality 9) standardizes what 114 participating governments themselves submit - more than 90 percent of global oil supply and demand - monthly from January 2002, ten flows across four crude and product groups in five units, a colour-coded assessment grade riding every figure. JODI-Gas World Database (quality 8) mirrors the model for gas: 88 economies covering around 90 percent of world gas supply, fourteen flows from January 2009. Eurostat Energy Database (quality 8) applies one balance taxonomy to every EU member state, EFTA and candidate country - indigenous production, imports, exports, stock changes - with annual oil balances from 1990 and monthly gas detail from 2008.

A world ranking wants the first; a supply-demand model wants the middle pair; European benchmarking wants the fourth. For calibration: only 534 of the 1,744 records in Datadory's catalog score 9 on the rubric - this 18-record upstream slice holds four of them. The table below lines the comparability tier up.

Three shapes cover the whole question, exactly as samples arrive:

# country-period snapshot (EIA International Energy Statistics) - monthly producer ranking, April 2026
iso  name           ranking  value
USA  United States  1        24342
RUS  Russia         2        10478
SAU  Saudi Arabia   3        7619
CAN  Canada         4        6392
# value: thousand barrels per day, crude incl. lease condensate

# country-reported observation (JODI-Oil World Database) - United Arab Emirates, January 2025
REF_AREA        : AE
TIME_PERIOD     : 2025-01
ENERGY_PRODUCT  : CRUDEOIL          # also NGL, OTHERCRUDE, TOTCRUDE
FLOW_BREAKDOWN  : CLOSTLV           # INDPROD carries indigenous production
UNIT_MEASURE    : CONVBBL           # also KBBL, KBD, KL, KTONS
OBS_VALUE       : 7596.0000
ASSESSMENT_CODE : 3                 # colour grade: 1 blue, 2 yellow, 3 white

# one extraction unit (GOGET) - the row standing behind a national total
unit_id   : L100000321014    fuel_type : oil
unit_name : 10103FF/Cold Lake Oil Asset (Alberta, Canada)
status    : operating        operator  : Strathcona Resources
latitude  : 54.534           longitude : -110.394    onshore_offshore : onshore

Read the anatomy rather than the digits. The EIA snapshot drops straight into a briefing - ordered, dated, unit declared - so the United States-Russia-Saudi scale argument ends before it starts. JODI's row grades itself: the assessment code announces whether a figure was reported or estimated before anyone builds on it, and the identifying stack rides the row instead of hiding in a filename. The GOGET line resolves a national number to a place, an owner and a status, which is what turns “country X produced Y” into something a portfolio can actually screen.

Yes, and this is where upstream work separates from a generic energy-statistics portal. GOGET inventories roughly 7,673 field-level units across 95 countries in its March 2026 vintage - 6,481 active extraction areas, 229 in development, 455 discoveries - each carrying operator, ownership stakes with percentages, status, discovery year, FID year, basin, field outlines and production converted onto one million-barrels-per-year basis. Inclusion has a floor: at least a million barrels of oil equivalent produced a year, or 25 million boe in reserves - a league table, not a phone book.

The Global Registry of Fossil Fuels adds the carbon column: 46,741 asset-level records across 89 countries, national series for 139 countries covering about 99 percent of world production, every volume already translated to CO2-equivalent. The Global Methane Emitters Tracker ties roughly 3,475 satellite-detected plumes to 18,023 named assets in 166 countries - modelled estimates, printed as such on every row.

Underneath sit the regulators. NSTA UKCS Open Data keeps 13,378 UKCS wellbore records reaching back to the 1960s plus day-by-day production histories for 182 producing wellbores and 40 fields, the sampled series opening in October 1995. The Railroad Commission of Texas ships lease-month production from 1993 beside drilling permits going back to 1976, split across its districts and counties. New York reports every producing well once a year from 2001, each row carrying its 14-digit API well number.

The reconciliation is the point. Aggregate the asset rows by country code and they land close enough to the official national series to argue with - screen at asset grain, publish at country grain, and both halves of the report hold.

Depth decides whether a record supports trend work or only a snapshot, and the spread here is wide enough to choose badly. The compiled EIA series carry multi-decade annual histories - several running back through the 1980s - at all three frequencies, which makes them the default backcast engine. Eurostat's balances open at 1990 for annual oil and 2008 for monthly gas. JODI-Oil starts January 2002 and JODI-Gas January 2009, both running to within a month of reference - anything older routes to the compiled shelf.

At the asset layer the registers are ledgers, not archives: GOGET reaches discovery years into the nineteenth century while its production rows carry their own data year (2023 in the sampled rows), and the Registry's observed years cluster around 2018-2023. The regulators go deepest of all - NSTA wellbores open in the 1960s with per-field series spanning decades, and Texas pairs half a century of permits with three decades of lease-months.

One further discipline matters for anything reproducible: pinning. Deliveries carry vintage labels as standard, and the whole compiled EIA corpus can be taken frozen - the PUDL Raw EIA Bulk Archive pins it at version 81.0.0, dated August 16, 2026 - so a panel rebuilt next quarter reproduces the numbers it was built on.

When two records disagree about one country, the cause is definition or timing, never mystery. Compiled series harmonize product boundaries across borders; country-reported boards publish what the government submitted on its own conventions, and where crude ends and lease condensate begins is a jurisdictional decision. Units multiply the problem honestly - thousand barrels, barrels per day, kilolitres, kilotonnes, converted barrels, cubic metres, terajoules - so convert once, centrally, before anything gets summed.

Three habits keep a country panel honest. Filter aggregates: world, regional, OPEC and EU rollups sit beside sovereign states in the same tables, and an unfiltered ranking pits the Netherlands against an entire region - Datadory pre-tags aggregate rows so the predicate is one clause long. Respect missingness: an economy that did not report a board cell arrives as an explicit dash with a status flag, never a silent blank. Pin vintages: figures revise, so every row carries its edition label and a revision lands as a new vintage instead of quietly overwriting the number a model trained on.

Market researchers and consultants size producing-country footprints from one ranked, comparable column instead of forty statistical yearbooks on forty layouts (more on that workflow). Investors and quants treat production as a regressable panel - monthly where the boards carry it, stable country keys throughout, units inline - and read drilling activity against it (the quant stack). Competitive intelligence teams resolve who operates what through GOGET's ownership stakes, then watch which of the 229 in-development units actually move. Policy and energy-security analysts compute import dependence directly from production, imports and exports that already agree on period and unit. Carbon accountants take the same national volumes pre-converted to CO2e, with methane attribution waiting at asset level. Journalists and academics quote national figures safely because each traces to a named compiler with a published methodology behind it.

The panel is unusually cooperative downstream. ISO country codes join to reserves, trade and price series on exact matches rather than name matching. Unit columns travel on every row, so conversion happens once, in one place, before anything gets summed. Aggregate rows arrive pre-tagged, unreported board cells keep their explicit status flags, and vintage labels turn revisions into editions. A monitoring loop can run hourly against the monthly boards while an annual planning cycle takes the compiled series on a slower drumbeat - same keys either way, so moving between them is configuration rather than migration.

Start with a sample: name the economies, the products and flows, the years, and the extract comes back shaped like your question - field dictionary attached.