Global Leadership Project

The Global Leadership Project (GLP) is an open database of government leaders worldwide that helps you study who holds power and how leadership patterns vary across countries and institutions. It catalogs leaders’ backgrounds and career paths so scholars, policymakers, and citizens can build evidence-based analyses and reforms.

GLP includes legislators, members of the executive branch, members of the judiciary, and other decision-makers whose power may be formal or informal. If an office directs policy, the project aims to track who holds it and how they got there.

GLP currently covers 162 countries and 70,000+ political leaders across 1,471 political bodies and 2,052 social groups. Coverage grows over time as new leaders and institutions are added in updated releases.

Yes—GLP’s data and tools are open source, with a focus on transparency and analytic rigor. That makes it practical for research, teaching, journalism, and policy work without building a leadership dataset from scratch.

Start with the GLP Data page to download datasets and replication materials hosted on Dataverse and Hugging Face. For representation benchmarks, the Country Rankings page also links to a country-level download (CSV) plus the underlying GLP data release.

GLP data is the leader-level database (who served where, plus background and identity attributes), while the Country Rankings are a derived, country-level benchmark of descriptive representation. Use GLP when you need microdata; use Rankings when you need quick cross-national comparisons across gender, ethnicity, religion, and language.

Descriptive representation asks whether leaders “look like” the population they govern along identity dimensions such as gender, ethnicity, religion, or language. In practice, it compares group shares in a political body to group shares in the population to quantify how closely they match.

The representation gap compares observed representation to what we’d expect under random selection from the population, so values closer to zero mean smaller gaps and more negative values mean larger shortfalls. In short: RepresentationGap = ObservedRepresentation - ExpectedRepresentation.

The current rankings include countries that meet a data-quality threshold (at least 75% completeness for the demographic traits being analyzed). When missingness is limited, the rankings use a principled imputation procedure so countries remain comparable.

For each country and identity dimension, GLP compares the observed shares of groups in national leadership to an expected benchmark under a random-sampling baseline, then aggregates those indices into an overall score. The result is a standardized benchmark you can use to compare countries and track representation gaps over time.

GLP provides an R-based Representation Calculator for quantifying descriptive representation, plus online tools for Country Rankings and interactive visualizations. You can also use the Hugging Face apps to compute representation gaps and explore leader profiles in a browser.

Expected representation is the baseline you’d anticipate if leaders were selected randomly from the population given population shares and body size. Observed representation is the index computed from the real composition of a body (e.g., a cabinet, legislature, or court), which you can compare to the baseline.

GLP is best for questions about who holds power—leadership composition, representation, and elite career paths—within and across countries. It is not a substitute for surveys or policy-position datasets; instead, it complements those sources with leader-level context.

Cite the relevant publication (for example, “The Composition of Descriptive Representation”) and the specific dataset release you downloaded (Dataverse or Hugging Face). The Data and Country Rankings pages provide download links and bibliographic info to make citation straightforward.

Use the contact information on the About page for data questions, corrections, or collaboration, or use the linked GitHub repositories to open issues for tooling and methods. GLP supports reproducibility with public datasets, open-source tools, and replication materials.