Sources & Data Policy

Pew Research Center, Religion & Public Life

Gold-standard country-level estimates for 201 countries — adherents and shares for Christians, Muslims, Hindus, Buddhists, Jews, other religions and the unaffiliated, for 2010 and 2020, built from 2,700+ censuses and surveys. Separate 2015 projections to 2050.

Used for

Primary population figures — our preferred estimates lead with Pew 2020.

Access

CSV dataset behind a free Pew account; the same data is republished machine-readably by Our World in Data, which is our automated path.

Updates

Landscape studies roughly every decade; topical reports every few years.

Reliability

The most widely cited, methodologically transparent global source. Counts self-identification, which runs lower than affiliation-based databases for some groups.

Global Change Data Lab

Pew's 2010/2020 religious-composition estimates re-published as clean, no-login CSV/JSON with ISO country codes, backed by a versioned open ETL on GitHub.

Used for

The automated pipeline's primary demographics feed.

Access

Stable no-auth endpoints — append .csv or .metadata.json to the grapher URL. This is what our monthly job actually fetches.

Updates

Updated when Pew publishes; endpoint stable and safe to poll monthly.

Reliability

Best machine-readable proxy for Pew data. Endpoint verified live before each refresh.

xAI

AI-generated encyclopedia (6M+ articles) including a detailed Timeline of Religion and per-religion articles. One of the two reference works our history timeline is cross-referenced against, and the tie-breaker where references disagree.

Used for

Timeline events and encyclopedic detail; date tie-breaker per site policy.

Access

No official API (JS-rendered pages only) — reviewed by hand each refresh cycle; never auto-fetched.

Updates

Irregular AI regeneration, no public changelog.

Reliability

AI-generated content without revision history — rich in detail but not independently citable, so every Grokipedia-attested claim is marked and cross-checked against Wikipedia and academic dating.

Wikimedia Foundation

Timeline of Religion, History of Religion, Major Religious Groups and List of Religious Populations articles, fetched programmatically with revision ids via the MediaWiki API.

Used for

Timeline events, population tables, cross-referencing.

Access

MediaWiki Action API (action=parse) and REST API — no auth; every fetch records the revision id shown here after each refresh.

Updates

Continuous, crowd-edited; we snapshot monthly and diff.

Reliability

Crowd-maintained: tables cite Pew/WRD/CIA but can churn. All Wikipedia-derived diffs land in a human-reviewed pull request, never auto-merge.

Nilesh Nilkanth Oak (self-published research; nileshoak.com)

Day-precise absolute datings of the Ramayana (c. 12240–12208 BCE) and Mahabharata (war 16 October – 2 November 5561 BCE) derived from the epics' own astronomical references — the Arundhati–Vasistha observation, precessed season-month statements, pole-star and nakshatra positions. Powers the plotted Oak chronology inside the Hindu Mythic Timeline.

Used for

The Oak chronology track on the Hindu Mythic Timeline only — never for the main historical timeline.

Access

Compiled by hand from his books (When Did the Mahabharata War Happen?, 2011; The Historic Rama, 2014; Bhishma Nirvana, 2019), his published timeline summaries, and lecture records.

Updates

New books/lectures occasionally; positions stable.

Reliability

A contested research program, not scholarly consensus: mainstream indology treats the epics' astronomy as non-datable poetic material, and other traditional daters (3139 BCE Puranic, 3067 BCE planetarium school) reject his dates. Presented on the site as his claims, plotted alongside the rival dates.

Encyclopædia Britannica

Editorially reviewed population figures per religion, largely drawing on the World Religion Database tradition of counting.

Used for

Cross-checking population estimates.

Access

Manual review each refresh cycle.

Updates

Irregular editorial updates.

Reliability

Professionally edited; tends toward affiliation-based (higher) counts than Pew for some groups — a useful upper-bound cross-check.

Brill / Boston University CGC (Johnson & Zurlo)

The most granular academic dataset — every religion, every country, 1900–2050, with thousands of Christian denominations. Paywalled; we cite figures republished in Wikipedia, Britannica and the free annual Status of Global Christianity table.

Used for

Cross-checking, denominational branch figures.

Access

Paid institutional subscription — cited only, never scraped.

Updates

Annual reconciliation.

Reliability

Academically respected; counts affiliation rather than self-identification, so its totals run higher than Pew's for some groups. We show both rather than picking.

Association of Religion Data Archives / Penn State

Free downloadable historical series: adherent counts by religion per state, 1945–2010 in five-year intervals — historical depth Pew doesn't cover.

Used for

Historical population series for the timeline.

Access

Free CSV downloads behind an interactive terms flow; the series is static, so it's vendored once rather than refetched.

Updates

Static historical dataset (ends 2010).

Reliability

Excellent for 20th-century trends; not a source for current-year figures.

ReligionFacts.com

Comparative at-a-glance chart of beliefs, practices, texts, deities, afterlife views and worship across dozens of traditions — the template for our Big Chart's qualitative dimensions (its adherent-count cells are empty in the served page, so counts come from elsewhere).

Used for

Qualitative comparative dimensions on the Big Chart.

Access

Manual review.

Updates

Infrequent.

Reliability

Tertiary reference without published methodology — its comparative dimensions are cross-checked against Grokipedia and Wikipedia before use; never a figure of record.

Wikimedia Foundation

Structured facts per religion entity — member counts (property P2124), inception dates, founders. Queried monthly as a drift detector against our committed figures.

Used for

Cross-check and entity metadata.

Access

SPARQL endpoint, no auth. Verified query returns Christianity 2.4B, Hinduism 1.2B, Buddhism 520M — but coverage is patchy (Islam lacks a member-count statement), so it can only ever be a cross-check.

Updates

Continuous, crowd-edited.

Reliability

Incomplete and uneven for adherent counts; strong for entity metadata like founding dates and relationships.

CIA (factbook.json mirror)

Per-country religion percentage breakdowns with estimate years, as public-domain JSON.

Used for

Per-country cross-checks.

Access

Raw GitHub JSON per country, no auth — fetched monthly for country-level cross-checks.

Updates

CIA updates rolling-weekly; mirror regenerated periodically (we record the mirrored commit).

Reliability

Broad coverage but heterogeneous estimate years and free-text categories — good per-country sanity check, unsuitable for global aggregation.

World Bank

Total population per country and for the world, per year — the denominator under every percentage on the site.

Used for

Population denominators.

Access

No-auth JSON API, fetched monthly; every response exposes a lastupdated stamp we record.

Updates

Refreshed several times a year.

Reliability

The most robust no-auth denominator endpoint; sourced largely from UN World Population Prospects.

UN DESA, Population Division

The canonical demographic denominators — historical population and projections for every country and the world.

Used for

Population denominators.

Access

REST JSON Data Portal API (occasionally token-gated, so the pipeline falls back to the World Bank mirror of the same series).

Updates

Revised roughly every two years.

Reliability

Authoritative but API access intermittently changes; used to validate the World Bank numbers.

Population Connection

Educational summary of world religious composition, itself derived from Pew figures.

Used for

Cross-checking presentation of headline figures.

Access

Manual review.

Updates

Occasional.

Reliability

Derivative of Pew — useful only to confirm how the headline figures are commonly presented.