Frequently Asked Questions

V-Dem measures many dimensions of democracy. As it can be challenging to measure latent (not directly observable) concepts, V-Dem relies on multiple different data sources, as well as country-experts to create the dataset. The V-Dem dataset contains ratings from over 4,200 country-experts, 600+ indicators and indices with data for 202 country units.

There is no consensus on what democracy writ-large means beyond a vague notion of rule by the people. Political theorists have emphasized this point for some time, and empiricists would do well to take the lesson to heart (Gallie 1956; Held 2006; Shapiro 2003: 10-34). At the same time, interpretations of democracy do not have an unlimited scope.

A thorough search of the literature on this protean concept reveals seven key principles that inform much of our thinking about democracy: electoral, liberal, majoritarian, consensual, participatory, deliberative, and egalitarian. Each of these principles represents a different way of understanding “rule by the people.” The heart of the differences between these principles is in the fact that alternate schools of thought prioritize different democratic values. Thus, while no single principle embodies all the meanings of democracy, these seven principles, taken together, offer a fairly comprehensive accounting of the concept as employed today.*

The V-Dem project has set out to measure these principles, and the core values which underlie them. We summarize the principles below.

• The electoral principle of democracy embodies the core value of making rulers responsive to citizens through periodic elections, as captured by Dahl’s (1971, 1989) conceptualization of “polyarchy.” Our measure for electoral democracy is called the “V-Dem Electoral Democracy Index.” We consider this measure fundamental to all other measures of democracy: we would not call a regime without elections “democratic” in any sense.
• The liberal principle of democracy embodies the intrinsic value of protecting individual and minority rights against a potential “tyranny of the majority” and state repression. This principle is achieved through constitutionally-protected civil liberties, strong rule of law, and effective checks and balances that limit the use of executive power.
• The participatory principle embodies the values of direct rule and active participation by citizens in all political processes. While participation in elections counts toward this principle, it also emphasizes nonelectoral forms of political participation, such as civil society organizations and other forms of both nonelectoral and electoral mechanisms of direct democracy.
• The deliberative principle enshrines the core value that political decisions in pursuit of the public good should be informed by a process characterized by respectful and reason-based dialogue at all levels, rather than by emotional appeals, solidary attachments, parochial interests, or coercion.
• The egalitarian principle holds that material and immaterial inequalities inhibit the actual use of formal political (electoral) rights and liberties. Ideally, all groups should enjoy equal de jure and de facto capabilities to participate; to serve in positions of political power; to put issues on the agenda; and to influence policymaking. Following the literature in this tradition, gross inequalities of health, education, or income are understood to inhibit the exercise of political power and the de facto enjoyment of political rights.
• The majoritarian principle of democracy reflects the belief that a majority of the people must be capacitated to rule and implement their will in terms of policy.
• The consensual principle of democracy emphasizes that a majority must not disregard political minorities and that there is an inherent value in the representation of groups with divergent interests and view.
  
  The conceptual scheme presented above does not capture all the theoretical distinctions at play in the complex concept of democracy. We have chosen to focus on the core values and institutions that the other principles emphasize in their critique of the electoral conception as a stand-alone system. Each of these principles is logically distinct and—at least for some theorists—independently valuable. Moreover, we suspect that there is a considerable divergence in the realization of the properties associated with these seven principles among the world’s polities. Some countries will be particularly strong on electoral democracy; others will be strong on the egalitarian property, and so forth.
  
  *This consensus only holds insofar as most scholars would agree that some permutation or aggregation of these principles underlie conceptions of democracy. For example, scholars can reasonably argue that the list could consist of seven, six, or five principles; our “principles” may be “properties” or “dimensions;” and “majoritarian” and “consensual” are actually opposite poles of a single dimension. As a result, we intend for this discussion to assure consumers of the data of the comprehensive nature of our inventory of core values of democracy: namely, that it includes almost all the attributes that any user would want to have measured.
  
  -----------------

Gallie, W. (1956), ‘Essentially Contested Concepts’, Proceedings of the Aristotelian Society 56, 167–220.

Held, D. (2006), Models of Democracy, 3 edn, Polity Press, Cambridge.

Shapiro, I. (2003), The State of Democratic Theory, Princeton University Press, Princeton.

Dahl, R. A. (1971), Polyarchy: Participation and Opposition, Yale University Press, New Haven.

Dahl, R. A. (1989), Democracy and its Critics, Yale University Press, New Haven.

If a variable drawn from the V-Dem dataset plays an important role in your project (published or unpublished), please use the applicable citations found in the corresponding codebook.

For v15:

V-Dem Dataset:
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Fabio Angiolillo, Michael Bernhard, Agnes Cornell, M. Steven Fish, Linnea Fox, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Marcus Tannenberg, Eitan Tzelgov, Yi-ting Wang, Felix Wiebrecht, Tore Wig, Steven Wilson and Daniel Ziblatt. 2025. "V-Dem [Country-Year/Country-Date] Dataset v15" Varieties of Democracy (V-Dem) Project. https://doi.org/10.23696/vdemds25

and:

Pemstein, Daniel, Kyle L. Marquardt, Eitan Tzelgov, Yi-ting Wang, Juraj Medzihorsky, Joshua Krusell, Farhad Miri, and Johannes von Römer. 2025. “The V-Dem Measurement Model: Latent Variable Analysis for Cross-National and Cross-Temporal Expert-Coded Data”. V-Dem Working Paper No. 21. 10th edition. University of Gothenburg: Varieties of Democracy Institute.

V-Dem Codebook:
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Fabio Angiolillo, Michael Bernhard, Agnes Cornell, M. Steven Fish, Linnea Fox, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Marcus Tannenberg, Eitan Tzelgov, Yi-ting Wang, Felix Wiebrecht, Tore Wig, and Daniel Ziblatt. 2025. "V-Dem Codebook v15" Varieties of Democracy (V-Dem) Project.

V-Dem Methodology:
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Kyle L. Marquardt, Juraj Medzihorsky, Daniel Pemstein, Linnea Fox, Lisa Gastaldi, Eitan Tzelgov, Yi-ting Wang, and Steven Wilson. 2025. "V-Dem Methodology v15" Varieties of Democracy (V-Dem) Project.

V-Dem Country Coding Units:
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Lisa Gastaldi, Ana Good God, and Sandra Grahn. 2025. "V-Dem Country Coding Units v15" Varieties of Democracy (V-Dem) Project.

V-Dem Organization and Management:
Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Sara Andersson Haug, Susanna Burmeister, Linnea Fox, Lisa Gastaldi, Ana Good God, Sandra Grahn, Melina Liethmann, Natalia Natsika, Evie Papada, Josefine Pernes, and Maria Verkhovtseva. 2025. "V-Dem Organization and Management v15" Varieties of Democracy (V-Dem) Project.

V-Dem is part of and partly funded by DEMSCORE, national research infrastructure grant 2021-00162 from the Swedish Research Council. The simulations and other computational tasks were enabled by resources provided by the National Academic Infrastructure for Supercomputing in Sweden (NAISS), partially funded by the Swedish Research Council through grant agreement no. 2022-06725.

All V-Dem data and graphs generated by the online tools are open source and free for anyone to use. You do not need to seek permission to use the data, but we kindly ask you to cite the data: 

Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, David Altman, Fabio Angiolillo, Michael Bernhard, Agnes Cornell, M. Steven Fish, Linnea Fox, Lisa Gastaldi, Haakon Gjerløw, Adam Glynn, Ana Good God, Sandra Grahn, Allen Hicken, Katrin Kinzelbach, Joshua Krusell, Kyle L. Marquardt, Kelly McMann, Valeriya Mechkova, Juraj Medzihorsky, Natalia Natsika, Anja Neundorf, Pamela Paxton, Daniel Pemstein, Johannes von Römer, Brigitte Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Aksel Sundström, Marcus Tannenberg, Eitan Tzelgov, Yi-ting Wang, Felix Wiebrecht, Tore Wig, Steven Wilson and Daniel Ziblatt. 2025. "V-Dem [Country-Year/Country-Date] Dataset v15" Varieties of Democracy (V-Dem) Project. https://doi.org/10.23696/vdemds25

In addition to the V-Dem dataset, V-Dem also hosts the V-Party dataset, on political parties, and the ERT dataset, on episodes of regime transformations. You can access all available datasets by at the Datasets page on this website. There you will also find a brief description of each dataset.

When you download the V-Dem dataset, you are able to choose between R (.rds), STATA (.dta), CSV, or SPSS (.sav) file formats. Regardless of which dataset and format you choose, you will get a ZIP-file with the dataset, the codebook, cautionary notes, a file listing changes from the previous version, and the suggested citation.

The 'vdemdata' R-Package is a way for users to integrate the V-Dem dataset directly into their scripts, letting users load and explore the most recent “Country-Year: V-Dem Full+Others” dataset directly from R.

Please refer to V-Dem’s Github page and the packages’s ReadMe for detailed information. The package is perfect for everyone interested in working with V-Dem data regularly and who is familiar with R.

The V-Dem datasets are released in a few different versions: country-year, country-date, and coder-level. The difference is in the unit of analysis. Each variant has its own perks and quirks and allows for different types of analysis.

The Country-Date dataset provides country-indicator data by date. The dataset therefore offers more granular estimates. We recommend this dataset for researchers who want greater precision within each year.

If you are not interested in the events that have occurred within any given year but instead change over multiple years, the Country-Year format is likely more helpful. The Country-Year datasets aggregate country-date data to the country-year level based on a day-weighted mean. As country-year level of analysis is quite common, the country-year data is likely more compatible with other data sources, such as macro-data (e.g., economic data, etc.)

We offer two different country-year datasets: Core and Full+Others. The Core dataset includes V-Dem's five key democracy indices and all the sub-components that make up these indices. The Full+Others includes all V-Dem indices and indicators as well as additional indicators from external sources.

Finally, the Coder-Level dataset includes all C-data - data collected from our country-experts surveys - by coder, meaning that you can see the raw data. In addition to the coders’ responses, we provide reliability estimates for experts calculated by our IRT model. The Coder-Level dataset also includes a few additional variables from the Post-Survey Questionnaire (PSQ). This allows users to inspect the raw anonymized data directly or use it for alternative analysis.

Given the extent and complexity of the V-Dem dataset, the codebook is your best friend. The codebook provides detailed information on each indicator and index, including the survey questions, the possible responses, years covered, scale of the data, and much more. For data collected from external sources, we also include sources.

Most codebook entries are straightforward, but one thing might stand out to many users. The V-Dem dataset is compiled of several different types of data. Generally, V-Dem differentiates between factual data, evaluative data, and indices. The data type of each indicator can be found in the codebook. Each indicator has a capital letter in its title, such as “(A)”. This letter denotes the indicator’s data type.

• A-Data: Factual/observable data (like presidential term length). This data is collected by the V-Dem team.
• A*-Data: Same as A-Data, but Country-Expert coders rate their confidence in the data.
• C-Data: Evaluative data which is based on scores provided by country experts and then modeled.
• A, C-Data: Factual data compiled by Country Experts, cross-checked by V-Dem staff.
• D-Data: Indices composed of other indicators and/or indices (most often C-Data).
• E-Data: Data from external sources.

Our experts give their ratings on an ordinal scale; the output from applying our measurement model to these ratings is on an interval scale. Hence, the ratings as such take on discrete values but the latent scores can be of any numeric value (often between -4 and 4).

No. Comparing absolute index values between different versions of our datasets can be misleading, due to several different factors;

1. Coders add data with every annual update.
2. We currently allow our coders to update and change their ratings back in time if they have come across new and more accurate information. This should lead to a continuous improvement of the data quality over time.
3. There is some attrition of coders as well as recruitment of additional coders for every new release.
4. We strive for full coverage and therefore continuously add new countries and expand our time-series.

For each variable the input of our Bayesian Item Response Theory Measurement model contains the coder scores for that specific question across all countries and years (in reduced form, capturing changes only). Because of all the factors mentioned above the input per variable changes with every annual release. Consequently, the output of the statistical model varies accordingly, and the scores are hence expected to change with new releases.

Each index is an aggregation of several variables. However, an interpretation of index scores is only meaningful in comparison to other index scores across time or countries within the same index (e.g. rank orders).

Some versions of the dataset include less variables than others, which could potentially explain why a specific variable is not included in the dataset that has been downloaded. As a first step, we would therefore recommend you to download the Country-Year Full + Others version of the data set and try again.

Many organizations and websites use V-Dem data for their own analyses, which we encourage. But we can only guarantee the correctness of data that is downloaded directly from the V-Dem website or from DEMSCORE. Please use V-Dem data downloaded from those two sources to make sure it is correct.

If you do not plan to analyze the data with a programmatic language or software, the data can be unruly and challenging to work with in its entirety, given the number of variables, countries, and years. We do not recommend using the full V-Dem dataset on Excel.

If you do not plan to analyze the data with a programmatic language or software, we highly recommend downloading the specific variables of V-Dem data you are interested in through DEMSCORE.

Read more about how to use DEMSCORE to filter a dataset.

Yes! V-Dem Institute is part of a data infrastructure project that offers a unique shortcut to data filtering and merging: DEMSCORE. While DEMSCORE is intended to combine different datasets, it can also be used to filter a single dataset. Through DEMSCORE you can choose and download specific V-Dem indicators and indices of interest and even filter this data by country and year.

If you do not plan to analyze the data with a programmatic language or software, we highly recommend using Demscore. It is easy, convenient, and prevents potential mistakes.

Country experts are individuals with expertise on topics relating to democracy within a specific country. They are the ones that fill in our annual surveys. To ensure coders’ safety and independence, their identities are anonymous. V-Dem never confirms nor denies the identities of Country Experts in any form.

Many key features of democracy are not directly observable. Our underlying assumption is that, in such contexts, experts are able to synthesize different sources of information to provide accurate estimates of these “latent” (not directly observable) concepts. V-Dem therefore asks its global network of almost 4000 Country Experts to use an online interface to code these concepts, and thereby obtains detailed, local knowledge from qualified experts familiar with political developments in a given country.

For more information about expert recruitment, please refer to the V-Dem methodology document. V-Dem WP 53 provides a more detailed discussion of the use of expertise when coding political phenomena, while this R&P article discusses variation in expert reliability.

V-Dem asks experts to code indicators that are difficult to directly observe, while we ask research assistants to code indicators that are not (we refer to the latter as "factual"). The key distinction between these coding procedures is that coding directly observable indicators requires no particular conceptual or case expertise. Instead, coding these indicators only requires the ability to follow a detailed protocol and find the relevant materials.

We believe that both types of indicators are important for measuring different aspects of democratic institutions and, indeed, many of our high-level indices include both types of data.

For a more detailed discussion of this distinction, please see this PS article and V-Dem Working Papers 53 and 145.

Most V-Dem indicators have five or more expert coders for each country and year. In general, the ratings the experts provide are values on a five-point ordinal scale, with low scores representing worse conditions and high scores representing better. Because the concepts V-Dem asks experts to code are not directly observable, we expect experts to provide different values in many - if not most - cases.

V-Dem has developed innovative methods for aggregating these different expert ratings in a way that produces valid and reliable estimates of difficult-to-observe concepts. Specifically, our state-of-the-art measurement model (MM) uses advanced statistical techniques to aggregate their ratings while accounting for possible noise in the rating process, while also producing estimates of uncertainty. In general, when experts disagree more there is greater uncertainty about our estimates.

See this document for a brief overview of the V-Dem MM. For a more detailed overview, please refer to V-Dem Working Paper 21. These three research articles also discuss the advantages and disadvantages of the V-Dem MM: PA, PSRM, JPR.

There is no credible evidence for this claim, and substantial evidence to the contrary. This PS article and V-Dem Working Paper 140 provides a detailed discussion of our research into this issue.

Our Expert Coders / Country experts are generally living in or are from the country they are coding. There are several reasons that lie behind the decision not to reveal the identity of our Country Experts:

• Following national and EU laws and regulations (GDPR), it is prohibited to share personal identifying information;
• There are a number of repressive countries in the world where the participation in V-Dem may be dangerous to Country Experts and/or their relatives;
• It is impossible to predict which countries may become repressive in the future and thereby make participation in the V-Dem surveys dangerous;
• V-Dem data is used in evaluations and assessments internationally in ways that could affect a country’s status. Thus, there are incentives for certain countries and other actors to try to affect ratings;

Hence, we preserve Country Expert confidentiality by a strict set of security policies. V-Dem never confirms nor denies the identities of Country Experts in any form.

Before diving into the specifics of our aggregation methods, it is first important to understand how the data looks when collected. The data is collected on the country-date level, however in practice there is often only one value for each year, in these cases each value is coded on the “default-date” which is the last day of the year (December 31). To frame when change happens in a given year, we carry over December 31st observations to January 1st of the following year, so we have a reference of the value at the beginning of the year.

So, how does the V-Dem data get from the country-date to the country-year format? The exact aggregation process depends on the data type. The standard method is day-weighted average. A day-weighted average assigns weights to values according to the duration of the year they cover. Each value covers the time frame until the next rating.

Some data is not aggregated, and instead the last observation of each year becomes the value for the country-year dataset. For example, in the case of the name of a country’s head of government (v2exnamhog). Ulf Kristersson was appointed Sweden’s head of government in October 2022; therefore, Ulf Kristersson is the value for 2022 in the country-year format.

In other cases, the highest observed value for a year becomes the country-year value. For example, election type (v2eltype). The 2022 Swedish parliamentary election took place on September 11, 2022, meaning v2eltype_0 is 1 on that date and 0 in all others. Therefore v2eltype_0 is 1 for 2022 in the country-year dataset.