OpenAlex — scholarly-works graph including education research

OpenAlex is OurResearch's index of world research, built as the successor to Microsoft Academic Graph, and the graph is the product: 322,044,938 works joined to 125,675,266 author records, 134,448 institutions, 255,627 sources, 4,516 topics, 10,708 publishers and 45,639 funders. Every work carries its DOI, title, publication date, genre and hosting source - and every one of those entities is a first-class record, not a footnote.

Three link structures turn it into an analysis surface rather than a bibliography. Authorships tie each work to its authors and their institutions, resolved to ROR identifiers and ISO country codes. Topics classify every work along a domain > field > subfield > topic hierarchy - 7,224,638 works sit in the Education subfield alone. And the citation graph is explicit: a cited_by_count on each work, referenced_works as the outgoing edge list.

Ask for the sample cut to your subfields, topics, institutions or countries; the shape you approve is the shape that ships.

Captured rows, exactly as they arrive - one work, plus the graph-scale counters and education rollups captured alongside it:

# one row per work - captured records, shown to fix the shape

id                  W3038568908
doi                 10.1585/pfr.15.2402039
title               Radiation Resistant Camera System for Monitoring
                    Deuterium Plasma Discharges in the Large Helical Device
publication_year    2020        type        article
oa_status           diamond     country     JP
first_author_orcid  0000-0003-0655-7347

# graph-scale counters captured alongside the work rows

all_works           322044938
authors             125675266

# education slices arrive pre-classified - captured rollups

Education (subfield, primary_topic.subfield 3304)     7224638 works
T12217 Higher Education Teaching and Evaluation       110439 works

Nothing was tidied for presentation. The useful part is what surrounds each work row: affiliations already resolved to ROR institutions and country codes, an ORCID on the first author, an OA-status code, and topic assignments carrying IDs you can traverse. Because the education rollups ship pre-counted - 7,224,638 works in the Education subfield, 110,439 in Higher Education Teaching and Evaluation - scoping a feed to education research is a filter decision, not a classification project.

Samples ship cut to the subfields, topics and countries you name, with the same field set throughout.

The thirteen headline fields divide into identity, classification and graph, as tabled above. Three earn special mention.

authorships is where the institutional view comes from: each entry carries an ORCID, resolved institutions with ROR ids and country codes, the raw affiliation string and corresponding-author flags. Institution-level and country-level aggregations need no separate lookup file.

topics rides on OpenAlex's retired concepts layer - legacy concept codes still resolve, but topic filters are canonical, so anything built on the old codes should plan the migration once, on our side of the feed rather than repeatedly on yours.

referenced_works turns 322 million bibliography lines into a traversable citation graph. Paired with cited_by_count, it supports influence, co-citation and collaboration work without anyone opening a PDF.

• Geography: global. Affiliations resolve to ROR institution ids and ISO country codes covering essentially every country with research output, so region and country views aggregate mechanically.
• Temporal: works from the mid-1600s onward, densely covered from the 1990s - the era most education-research questions actually live in. Publication year and full ISO date both ride on each work.
• Granularity: one row per entity - works, authors, institutions, sources, topics, publishers, funders - with rollups delivered as typed aggregates rather than ad-hoc groupings.

The education depth is worth pinning down: 7,224,638 works in the Education subfield, about 41.5 million in the parent Social Sciences field, and 110,439 in the Higher Education Teaching and Evaluation topic alone, as tabled below. Set against the survey censuses on the education services hub - enrollments, completions, finance - this is the other half of the picture: published research about education, attributable to 134,448 named institutions.

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

A graph arrives as tangles: nested authorships, scored topic arrays, edge lists tucked inside each work record. Teams rarely want the nest; they want the shapes underneath it - a works table keyed by ID, an authorship bridge joining works to authors to institutions, a topic dimension, and the citation edge list as its own table. Datadory does that unwinding once, keeps the keys stable across deliveries, and hands over whichever cadence the workload needs. The sample proves the shapes on your subfields before anything recurring starts.

• Research landscape mapping - who publishes what, with whom, across 134,448 institutions and essentially every country, filterable to a single topic.
• Edtech and publishing market sizing - education-research volume by topic, institution and country as demand-side denominators (market sizing playbook).
• Citation-grade evidence - literature reviews and systematic maps anchored to DOIs and explicit citation edges rather than vendor summaries (citation-grade research).
• Model features - citation graphs, topic labels and affiliation networks as training signal (ML model training).
• Trend detection - topic-level counts flag where education-research effort is moving before survey cycles pick it up.
• Institutional benchmarking - output volume and OA mix per named institution, assembled into benchmark sets on stable IDs.

Market researchers and consultants get demand-side denominators for anything sold into education research or academic publishing (market researchers page). Data scientists inherit a citation graph and topic taxonomy that arrive typed, keyed and already migrated (data scientists page). Competitive intelligence teams track research-output moves by institution, country and topic (competitive intelligence page). Investors and quant researchers read publishing volume and collaboration patterns as sector fundamentals (investors and quants). Journalists and academics cite DOIs and explicit counts rather than estimates (journalists and academics page). Sales and growth teams build account lists from named institutions active in their topic (sales and growth teams page).

Taxonomy note - the concepts layer is retired in favor of topics. Historical concept codes still resolve, but new feeds classify on topics; any scoping built on legacy codes gets migrated once at our end, not repeatedly at yours.

Identity note - raw affiliation strings ship alongside ROR-resolved institutions, because the unresolved text is sometimes the only version of a department that exists. Both stay in the feed.

Joining note - OpenAlex work IDs, DOIs, ORCIDs, ROR ids and ISSNs are the crosswalk set. They compose with the survey and statistical records below - IPEDS for output-per-campus views, Eurostat's education tables for national comparisons - instead of overlapping them. Where this graph meets the UIS education statistics comparator, the difference is unit of analysis: published works versus compiled national indicators.