V-Dem’s Democracy Reports: what ten years of grading shows

Plain-language summary of the longitudinal CID analysis. CID ID: CID-0046 (provisional).

What This Report Is

Every year the V-Dem Institute, an academic project based at the University of Gothenburg, publishes a Democracy Report that rates every country in the world on how democratic it is. V-Dem is the group that, in 2021, announced that India had stopped being a democracy and become an “electoral autocracy” — a label you have probably seen quoted many times since, often without anyone saying where it came from. We pulled every edition from 2017 through 2026 and graded each one.

We did not grade whether India really is an electoral autocracy. We graded V-Dem’s method. The question is the same one we ask of everyone: could someone outside V-Dem check the work?

What We Looked At

We scored ten annual editions against the Citation Integrity Dashboard rubric. The rubric ignores whether we agree with V-Dem’s conclusions. It asks structural questions. Are the key terms defined? Can a reader verify the individual numbers? Is the evidence independent of the people making the case? Does the report admit its own limits and its own uncertainty?

What We Found

Every edition scores Adequate, near the very top of that band. Among the institutional reports we have graded, that makes V-Dem the highest — only Pew’s one-off 2021 survey of religion in India, the corpus’s single Research-Grade study, scores higher. V-Dem beats Amnesty’s India reports by more than three points. It beats Freedom House, the group it most resembles, by almost two.

The reason is the exact opposite of what we found everywhere else. Amnesty, USCIRF, and the Freedom House country chapters all scored low for the same four reasons: no codebook, no clear rule for which cases get counted, no check on whether different people coding the same thing agree, and no way for a reader to verify a claim inside the document. V-Dem closes all four.

It publishes a codebook that defines every one of its 200-plus measures and tells coders exactly how to score them. It covers every country in the world — 202 of them, going back to 1789 — on the same fixed framework. It has about five independent experts rate each measure, then runs their ratings through a statistical model that checks how much the coders agree and reports a margin of error on every number. And it releases the entire underlying dataset — 31 million data points — for anyone to download and re-check. A reader who doubts the “electoral autocracy” label can do what no Amnesty reader can: pull the actual scores, read the margins of error, and test the call themselves.

What changes over the ten years is not the method — it is the news and the packaging. The early editions used a rough “compare to ten years ago” rule to spot countries going backward; from 2021 on, V-Dem switched to a more careful, peer-reviewed method. The reports also got far more heavily sourced over time, citing six different outside sources in 2017 and ninety-nine by 2026. But the engine underneath — the codebook, the expert coding, the statistical model, the public data — does not change. That is why the grade barely moves.

V-Dem also does something we rarely see: it argues against its own headline. In the 2024 edition it warns readers that because it weights big countries more heavily, the global decline it reports is driven largely by India — and that reading too much into that “may give misleading impressions.” It even shows the numbers with India taken out. A report that volunteers the strongest objection to its own most-quoted finding is doing exactly what the rubric rewards.

The Bottom Line

This is not a claim that V-Dem is right about India. It is a claim about how the report is built. Of everything we have graded, this is the document that most fully shows its work. The “electoral autocracy” label may or may not be the right call — but unlike almost every other source that repeats it, V-Dem hands you the tools to check it.

One honest caveat. Most of that rigor lives in V-Dem’s codebook, dataset, and statistical model — not in the annual PDF itself, which is a shorter, plainer summary that points to those things rather than reprinting them. We chose to credit the full apparatus, because the report genuinely rests on it and names it. If you graded only the PDF on its own, V-Dem would still land in Adequate, but in the middle of the band, right alongside Freedom House. Either way, it is the strongest document in the set — the only question is by how much.