The V-Dem Dataset is a leading comparative measure of democracy in the world today. Built by the V-Dem Institute at the University of Gothenburg, it covers roughly 202 countries and historical polities, stretching from 1789 to the present — version 16, published in March 2026, carries data through 2025. The Full+Others release contains 531 V-Dem indicators and 251 indices (plus 62 indicators drawn from other sources); the Core release distills five high-level democracy indices — electoral, liberal, participatory, deliberative, egalitarian — with 93 sub-indices and 179 indicators. The portal describes the series as “the world’s most comprehensive and detailed democracy ratings” — its own description.
Unlike the survey datasets in this directory, V-Dem does not interview citizens or observe regimes. Thousands of country experts rate each indicator for each country-year, and their judgments are aggregated through a Bayesian measurement model that produces point estimates alongside uncertainty intervals and coder-reliability scores. Those intervals are the point: V-Dem treats democracy as something measured with error, and publishes the error.
Its annual V-Dem Democracy Report draws on this data to track democratic backsliding and autocratization worldwide — making V-Dem the quantitative backbone of cross-country regime comparison.
Identity
- Producer: V-Dem Institute, University of Gothenburg (institution record in this directory).
- Version currency: v16, published March 2026 (read from the official data portal, September 2026). Prior version v15 (March 2025) remains available in the dataset archive.
- Coverage period: 1789–present; v16 covers data through 2025.
- Geographic coverage: roughly 202 countries and historical polities.
- Unit of observation: country-year (also country-date and coder-level releases).
- Update frequency: annual releases each March — values lag current events by months.
What the numbers are — and what they aren’t
Every value in V-Dem is an expert judgment about a country-year, aggregated by a statistical model. It does not directly observe regime behavior, count violations, or survey citizens. It does not measure what ordinary people experience day to day; it does not count arrests, censorship events, or deaths; it does not measure economic performance or state capacity.
For China specifically, its scores reflect the judgments of country experts working with incomplete information about a closed political system — structured expert perceptions, not field measurements. Treat wide uncertainty intervals on contested items as part of the result, not as a defect to ignore.
Methodology
Structured expert surveys aggregated through a Bayesian (IRT-style) measurement model producing point estimates plus uncertainty intervals and coder-reliability scores. The portal’s reference documents include Methodology, Structure of Aggregation, and Cautionary Notes — consult those, not this page, for the current model specification.
Access
- Access point: https://v-dem.net/data/the-v-dem-dataset/ (prior versions: https://v-dem.net/data/dataset-archive/)
- Access terms: open — free download in STATA, CSV, R, and SPSS formats; the portal notes an email/consent-form step for some downloads. A vdemdata R package on GitHub mirrors the latest release.
- Citation: Coppedge et al., “V-Dem Country-Year Dataset v16”, V-Dem Institute; DOI 10.23696/vdemds26 (third-party citation; not verified against the official DOI listing).
Research uses
The primary comparative-democracy measure for cross-country regime comparisons, democratic-backsliding analysis, and autocratization tracking. V-Dem feeds the annual V-Dem Democracy Report, and Our World in Data mirrors per-indicator CSVs (e.g., rule-of-law and political-corruption indices). For comparative-transition research it is the dataset behind the regime classifications used by democratization scholars.
Note on a common confusion: “V-Dem” is distinct from the Economist Intelligence Unit’s Democracy Index and from International IDEA’s Global State of Democracy Indices — three separate measures; do not merge them.
Limitations
- Expert-coding subjectivity is the central limitation: coder disagreement, differential item functioning across regions, and reliance on a small pool of country experts for closed systems. For China, experts must code a system that restricts independent verification; V-Dem’s own “Cautionary Notes” document addresses this.
- Historical back-projection (pre-1900) rests on historiography, not observation.
- Release lag: annual March releases mean values trail current events by months.
Authoritative documentation
- Official V-Dem dataset portal: https://v-dem.net/data/the-v-dem-dataset/
- Dataset archive (prior versions): https://v-dem.net/data/dataset-archive/
- V-Dem reference documents: Methodology; Structure of Aggregation; Cautionary Notes (on the portal)
- Our World in Data origins mirror (per-indicator CSVs)