Building Employer-Employee Panels From a Knowledge Graph and Company Microdata: Code and Data, 2007-2023

Gray, Steven and Kemeny, Tom and Nathan, Max and Ozgen, Ceren and Pialli, Guido and Reades, Jon and Rosso, Anna and Sere, Mateo and Siddiqui, Sameera and Valero, Anna (2026). Building Employer-Employee Panels From a Knowledge Graph and Company Microdata: Code and Data, 2007-2023. [Data Collection]. Colchester, Essex: UK Data Service. 10.5255/UKDA-SN-858769

This is the code and data repository for Gray et al (2026): GRAPH-EE: building employer-employee panels from a knowledge graph and company microdata, CEP Discussion Paper 2156, London: Centre for Economic Performance.

This research is funded by UKRI Grant ES/W010232/1 as part of the project "Diversity and UK Firm Performance". It was reviewed and approved by the UCL Research Ethics Board, Application 22883/00. The usual disclaimers apply.

The repo contains public release data from the project. It also contains the codebase and public release data layers needed to reproduce the UK worker-firm data stack and analysis in the paper.

The public release files in this repository contain information from OpenCorporates, which is made available here under the Open Database License (ODbL), from PATSTAT Global and from Diffbot, shared with permission.

Note: Claude Code was used to tidy parts of the codebase, and to produce initial versions of the filetree and data documentation. Final versions of all outputs have been written through and verified by the authors.

Data description (abstract)

Linked employer-employee data provide a resource for researchers and policymakers across the social sciences, but access to such data is typically highly restricted and data quality varies substantially across countries.

Leveraging a vast knowledge graph of the global public internet linked to UK company microdata and global patents data, GRAPH-EE is a rich complement to administrative data sources. This version of the platform covers over 10,500 companies and 800,000 workers active in the UK during 2007-2023. We show the potential to extend our approach to multiple other countries.

This is the code and data repository for GRAPH-EE (Gray et al 2026). The project builds novel worker-firm panel data from a mix of online web data, company microdata and public administrative sources. The collection contains:
1) Public release version of the company-year panel, company location lookup, and company / area-level Diffbot skills. Public release input data required in the build.
2) Codebase / reproduce the full worker-firm dataframes, as well as the public release data
3) Documentation to implement the workflow, including guidance on how to obtain other public / proprietary data layers
4) A copy of the working paper.

Data creators:

Creator Name	Affiliation	ORCID (as URL)
Gray Steven	University College London	https://orcid.org/0000-0001-7673-9993
Kemeny Tom	University of Toronto	https://orcid.org/0000-0003-4984-9104
Nathan Max	University College London and CEP	https://orcid.org/0000-0002-4617-4300
Ozgen Ceren	University of Birmingham	https://orcid.org/0000-0002-7242-9610
Pialli Guido	University of Turin	https://orcid.org/0000-0003-4482-168X
Reades Jon	University College London	https://orcid.org/0000-0002-1443-9263
Rosso Anna	University of Milan	https://orcid.org/0000-0001-7467-0734
Sere Mateo	University of Montevideo	https://orcid.org/0000-0003-4712-8426
Siddiqui Sameera	University College London	https://orcid.org/0009-0009-8180-7319
Valero Anna	London School of Economics and Political Science	https://orcid.org/0000-0003-2864-402X

Sponsors:

ESRC, SPF

Grant reference:

ES/W010232/1

Topic classification:

Science and technology
Economics
Trade, industry and markets
Demography (population, vital statistics and censuses)
Labour and employment

Keywords:

DATA, LABOUR AND EMPLOYMENT, PRODUCTIVITY, INNOVATION, SKILLED WORKERS

Project title:

GRAPH-EE: building employer-employee panels from a knowledge graph and company microdata: code and data

Alternative title:

GRAPH-EE

Grant holders:

Max Nathan, Tom Kemeny, Ceren Ozgen, Anna Sivropoulos-Valero, Jonathan Reades, Steve Gray, Anna Rosso

Project dates:

From	To
30 June 2022	30 March 2026

Date published:

30 Sep 2026 11:13

Last modified:

30 Sep 2026 11:14

Coverage and Methodology

Temporal coverage:

From	To
1 January 2007	31 December 2023

Collection period:

Date from:	Date to:
1 April 2024	31 May 2024

Geographical area:

United Kingdom

Country:

United Kingdom

Spatial unit:

Administrative > Countries
Administrative > Local Authority Districts
Other
Postal > Postcode (Unit)
Administrative > Regions

Data collection method:

We combine information from a world-leading knowledge graph – essentially, a database storing linked entities and their relationships – with open company data to build GRAPH-EE, a unique employer-employee data platform for the UK. We draw rich worker and organisation data from Diffbot, a commercial knowledge graph of the public internet. We exploit the fact that the UK companies register is open data, with company identifiers in Diffbot’s graph. We use this to precisely link companies and workers in Diffbot to companies in the UK register, to rich company financial data from Orbis Historical and to patents data from PATSTAT Global and Orbis IP. The resulting core data comprises over 10,500 larger UK companies active between 2007 and 2023, and over 800,000 workers in these companies. We focus on larger companies in this release because we need information on company headcounts, and larger firms are legally required to provide full, audited accounts.

Observation unit:

Organization, Geographic unit, Time unit

Kind of data:

Numeric, Text

Type of data:

Business microdata , Census data , Cohort and longitudinal studies, Experimental data , Geospatial data , Time series data

Resource language:

English

Access and Administration