Auditing India’s democracy score
India fell from 0.617 to 0.374. Do the underlying facts justify it?
A fact check of V-Dem’s underlying indicators
V-Dem’s decline comes almost entirely from three components: expression, clean elections and association. We traced those scores beneath the headline and compared the post-2014 evidence with the period V-Dem scores more highly.
- −0.390
- Freedom of Expression
- −0.189
- Clean Elections
- −0.172
- Freedom of Association
- 0
- Suffrage · Elected Officials
Clean Elections values shown here are reconstructed from V-Dem’s published EDI formula and the supplied India panel, because the panel carries the index, Expression and Association but not v2xel_frefair. They are marked derived wherever they appear.
The question
Not “does India have problems?” but “did they get worse than before?”
Reasoning we refuse to use
“A bad event happened after 2014, therefore India deteriorated.” An incident is not a comparison. Every claim on this page is measured against what the same indicator looked like before 2014.
The test we apply
Pre-2014 baseline against post-2014 condition. A downgrade counts as factually supported only where comparable evidence shows deterioration relative to the earlier period.
What our audit concludes
Three progressively stronger conclusions.
01First
India’s post-2014 democracy downgrade is misleading where the historical evidence does not support the scale or direction of the fall.
02Second
When comparable pre-2014 restrictions are repeatedly treated more leniently while similar post-2014 events produce sharp penalties, the problem is no longer just an isolated error — it points to a systematically biased scoring framework.
03Third
If such asymmetry persists even where the underlying historical record is well documented, it is reasonable to ask whether the distortion is merely methodological or whether it reflects deliberate ideological selectivity.
We cannot directly observe the intent of individual coders, but intent does not erase the outcome: a score that repeatedly ignores comparable earlier failures while heavily penalising later ones gives readers a distorted picture of how India actually changed.
01 — the score
What V-Dem records
- 2010 · Electoral Democracy
- 0.671
- before the comparison window
- 2014 · Electoral Democracy
- 0.617
- the baseline V-Dem scores higher
- 2023 · Electoral Democracy
- 0.374
- the score under audit
- 2010 → 2014 decline
- -0.054
- 2014 → 2023 decline
- -0.243
The post-2014 decline is about 4.5 times the size of the decline in the four years before it.
That ratio proves nothing on its own. A country’s democratic conditions can genuinely worsen four times faster in one period than another. It is simply the thing that requires an explanation, and the rest of this page looks for one in the underlying indicators.
02 — where the fall comes from
Two of the five components did not move at all
Elected Officials and Universal Suffrage sit at 1.000 in 2010, 2014 and 2023. India’s representatives are still elected, and every adult still votes. The entire factual dispute is therefore concentrated in three areas.
Freedom of Expression
0.889→0.812→0.422
-0.390
2014 → 2023
Clean Electionsderived
0.709→0.673→0.485
-0.189
2014 → 2023
Freedom of Association
0.787→0.769→0.597
-0.172
2014 → 2023
Elected Officials
1.000→1.000→1.000
no change
2014 → 2023
Universal Suffrage
1.000→1.000→1.000
no change
2014 → 2023
03 — the method
Half of the index is a multiplication, not an average
This is not hidden and not a flaw — it is V-Dem’s deliberate “weakest link” concept. But it has a consequence worth seeing plainly: when three components fall at once, they compound.
Additive half — API
0.25 × Association 0.25 × Clean Elections 0.25 × Expression 0.125 × Elected Officials 0.125 × Suffrage
A weighted average. A fall in one component moves this in proportion to its weight.
Multiplicative half — MPI
Association × Clean Elections × Expression × Elected Officials × Suffrage
A product. Three components falling together do not add — they compound. This is V-Dem's deliberate “weakest link” design, not an accident.
The index
EDI = 0.5 × API
+ 0.5 × MPIHalf the headline number is therefore a product of five terms. That is what makes the post-2014 fall steeper than any single component's fall.
Try it yourself
Set any component back to its 2014 value and watch the index move. Restoring Expression alone recovers far more than its additive weight would suggest, because it also sits inside the product.
- Additive · API
- 0.626
- India 2014: 0.814
- Multiplicative · MPI
- 0.122
- India 2014: 0.420
- Electoral Democracy Index
- 0.374
- India 2014: 0.617 · published 2023: 0.374
The product, worked through
0.812 × 0.769 × 0.673 ≈ 0.420
0.422 × 0.597 × 0.485 ≈ 0.122
The three disputed components do not merely add together. For half the index they are multiplied, so the product falls by more than any of them individually.
04 — component 1 of 3
Freedom of Expression & Alternative Information
v2x_freexp_altinfV-Dem score: 0.889 → 0.812 → 0.422
Freedom of Expression is the single largest contributor to India’s Electoral Democracy decline. V-Dem reduced India’s score from 0.812 in 2014 to 0.422 in 2023, a fall of 0.390 — nearly 48% of the 2014 value.
But this number is not directly measured. It is constructed from nine underlying expert-coded indicators covering media censorship, journalist harassment, media bias, self-censorship, political discussion and academic freedom. We checked those claims against the pre-2014 baseline.
Audit result
Large downgrade not established by the evidence reviewed
Real post-2014 concerns exist, and one — the Election Commission’s finding that the state broadcaster was not neutral in 2019 — survives comparison. But the evidence reviewed does not show deterioration across the nine underlying dimensions relative to the earlier baseline, and does not explain a fall as large as 0.812 → 0.422.
V-Dem’s Freedom of Expression score for India
- 2010 → 2014
- -0.077
- 2014 → 2023
- -0.390
- Share of the 2014 value
- -48.03%
The post-2014 decline is about 5× the preceding decline. That ratio proves nothing on its own — it is the thing requiring an explanation.
How the score is constructed
- Government media censorship
- Journalist harassment
- Media self-censorship
- Media bias
- Critical media
- Range of media perspectives
- Political discussion — men
- Political discussion — women
- Academic and cultural expression
Bayesian factor model
Freedom of Expression & Alternative Information
0.422
V-Dem does not simply count arrests, banned channels or protests. Country experts rate these underlying conditions, and V-Dem combines the expert assessments statistically using a Bayesian measurement model.
What follows from that
The nine underlying indicators
Each is compared against its own pre-2014 baseline, not against an ideal. Select any row for the evidence behind the verdict.
- 7 / 9
- Downgrade not established
- 0 / 9
- Some deterioration supported
- 2 / 9
- Mixed evidence
What kind of evidence did we find?
- Downgrade not established7
- Some deterioration supported0
- Mixed evidence2
V-Dem uses a statistical factor model; these nine indicators are not necessarily equally weighted. This chart summarises our audit findings, not V-Dem’s mathematical weighting. These are fact-check classifications, not replacements for V-Dem’s numerical sub-indicator scores.
Did government interference with news begin after 2014?
Censorship is the sub-indicator we investigated most deeply. Setting the two periods side by side is more useful than listing post-2014 incidents alone.
Pre-2014
- 2008Implemented
J&K local television news and current-affairs ban
- 2010Implemented
Renewed Kashmir local-news restrictions
- 2014Implemented · government-linked allegation
Times Now and India News cable blackout in Uttar Pradesh after negative coverage of the state government
Post-2014
- 2016Ordered but not implemented
NDTV India one-day shutdown order
- 2020Implemented, lifted early
Asianet News and MediaOne Delhi-riot coverage ban
- 2022Implemented
MediaOne permission and security-clearance shutdown
- 2023Judicial reversal
Supreme Court quashes the MediaOne restriction and rejects critical or ‘anti-establishment’ views as sufficient justification
Media bias: what coverage share can and cannot show
2014 — opposition dominated political visibility
CMS Media Lab, early-campaign share of leader coverage
- Arvind KejriwalOpposition28.19%
- Narendra ModiOpposition23.98%
- Rahul GandhiRuling Congress4.76%
2019 — DD News campaign coverage
Election Commission of India analysis, approximate hours
Did India’s media ecosystem become smaller?
The number of channels does not by itself prove viewpoint diversity, but it provides useful context when evaluating claims that alternative information largely disappeared.
Ownership concentration can reduce editorial independence even when the number of channels rises. Do not use channel count as proof that every viewpoint receives equal reach. The dark bar marks the 2013–14 baseline.
What our audit does show
- Freedom of Expression is the largest numerical driver of India's Electoral Democracy decline.
- Several genuine post-2014 concerns exist, and none of them is disputed here.
- The strongest surviving finding is narrow: the Election Commission found the state broadcaster's 2019 campaign coverage inconsistent with neutrality.
- Serious censorship, arrests for political speech, journalist prosecution, syllabus withdrawals by universities, a state book ban and an artist's exile all sit in the pre-2014 baseline.
- For seven of the nine underlying dimensions, the comparative evidence does not establish a clear post-2014 deterioration; two remain mixed.
- The evidence reviewed does not explain a component fall as large as 0.812 → 0.422.
What our audit does not show
- That India has perfect freedom of expression.
- That journalists are never harassed.
- That government pressure does not exist.
- That all nine V-Dem indicators are wrong.
- That no deterioration occurred in any of these areas — several verdicts turn on comparisons we have not completed, not on evidence of no change.
- That the exact correct score should still be 0.812.
- That V-Dem intentionally manipulated India's score.
Fact-check verdict
Some deterioration is real. The magnitude of the V-Dem decline is not established.
India’s expression environment after 2014 contains genuine problems, and this audit does not dispute them. The clearest is the Election Commission’s own finding that Doordarshan’s 2019 campaign coverage was inconsistent with neutrality — a regulator’s determination about a broadcaster legally bound to be impartial.
But the broader V-Dem score falls by almost half — from 0.812 to 0.422. When the same nine dimensions are compared against the pre-2014 baseline, serious censorship, arrests for political speech, journalist prosecution, self-censorship, university syllabus withdrawals, a state book ban and an artist driven into exile are already visible in the earlier period.
For seven of the nine underlying indicators, our review has not established a clear post-2014 deterioration. Two remain mixed. On journalist harassment the counts do differ — three imprisoned in 2012 against seven in 2022 and 2023 — but we have not case-checked them, and this page does not accept a count as a finding.
The evidence reviewed therefore supports criticism of specific post-2014 developments, but does not adequately explain the magnitude of V-Dem’s Freedom of Expression downgrade.
“Downgrade not established” means the evidence reviewed does not demonstrate deterioration relative to the earlier baseline. It does not mean no negative incidents occurred.
05 — component audit
Clean Elections
v2xel_frefairDowngrade not established2010
0.709
2014
0.673
2023
0.485
2014 → 2023
-0.189
Several measurable election-administration indicators improved after 2014, while genuine concerns existed in both eras. The comparative evidence reviewed does not establish the magnitude of the Clean Elections downgrade.
Underlying indicators · 8 of 8 audited
06 — component audit
Freedom of Association
v2x_frassoc_thickDowngrade not established2010
0.787
2014
0.769
2023
0.597
2014 → 2023
-0.172
Political-party competition remained extremely broad, while civil-society regulation tightened in some respects. The very large overall decline is not explained by the comparative evidence reviewed.
Underlying indicators · 6 of 6 audited
07 — objective indicators
Measures that moved in the opposite direction
These are administrative counts, not expert judgments. They do not prove that every aspect of election quality improved. They do make a broad claim of declining election-management capacity difficult to sustain without identifying countervailing evidence.
Polling stations administered
Election Commission of India
Voter-verified paper audit trail
Election Commission of India
- 2009
- No nationwide paper audit trail
- 2014
- Limited deployment
- 2019
- Deployed at every polling station
17.40 lakh
VVPAT units available in 2019
Did India become less multiparty?
Election Commission data show the opposite trend in the number of parties participating in Lok Sabha elections.
08 — comparable events, both periods
Broadcast restrictions, 2008 to 2023
Interference with broadcasting is real in both periods. What matters for a comparative score is whether it became clearly more frequent or more severe — so each event is marked by whether it was carried out, attempted and dropped, reversed, or only alleged, and by whether a state or the Union government acted.
- 2008Jammu & KashmirWent off airState government
Local television channels prohibited from broadcasting news and current-affairs material for nine days.
- 2008JammuReversed by court or governmentState government
Two local TV channels ordered off air; the High Court subsequently allowed them to resume.
- 2010KashmirWent off airState government
Local cable news and current-affairs broadcasts banned again; channels restricted.
- 2014Uttar PradeshAllegationState government
Times Now and India News disappeared from UP cable networks after negative reporting about the state government. Operators reported calls from ‘higher-ups’ — an allegation, not an established order.
- 2016UnionOrder not implementedUnion government
NDTV India received a one-day shutdown order over Pathankot coverage. The government put the order on hold before any blackout occurred.
- 2020UnionWent off airUnion government
Asianet News and MediaOne received 48-hour bans over Delhi-riot coverage. Both returned early after the restrictions were revoked.
- 2022UnionWent off airUnion government
MediaOne’s uplinking permission revoked after denial of security clearance.
- 2023Supreme CourtReversed by court or governmentJudiciary
The Court rejected a channel’s critical or ‘anti-establishment’ views as a legitimate basis to restrict press freedom.
FCRA did not begin in 2020
The regulation of foreign funding to civil society is central to the Association score. Its structure changed most in 2010.
Structural regime change
FCRA 2010
- Replaced the earlier regime with what the Supreme Court later called a more stringent dispensation
- Five-year validity and periodic renewal of registration
- Substantial suspension, cancellation and compliance machinery
- Statutory administrative-expense ceiling of 50%
- Regulation of organisations considered political in nature
- Expanded monitoring and reporting
Further tightening of the same architecture
FCRA 2020
- Complete prohibition on transferring foreign contribution to another person or NGO
- Administrative-expense ceiling cut from 50% to 20%
- Receipt required through a designated SBI New Delhi Main Branch account
- Additional identity requirements
- Parts of the existing regime made more stringent
2020 made the regime stricter. But the baseline matters: the core restrictive architecture being evaluated already changed substantially in 2010, before the period under audit.
- 2012
- 4,138
- registrations cancelled for failure to submit annual returns
- FY2014-15 to FY2016-17
- 10,000+
- cancelled, mainly for non-filing of mandatory returns
- since 2014, to early 2018
- 14,938
- cancellations reported by government
This does not make later cancellations irrelevant. It shows that mass regulatory enforcement against FCRA organisations was not a post-2014 invention.
09 — the audit matrix
All 23 underlying indicators
All 23 now carry a completed or partial comparison against the pre-2014 baseline. A verdict of “downgrade not established” is a statement about the evidence reviewed, not a finding that nothing happened.
- Downgrade not established
- Mixed evidence
- Some deterioration supported
- Observable improvement
- No material deterioration
- Audit pending
Expression
Clean Elections
Association
Select any tile to see the comparison behind it.
10 — who produces these values
Expert judgment, aggregated by a statistical model
V-Dem is an academic democracy-measurement project, not a media organisation, and this section describes its method fairly. For many subjective indicators, country experts answer detailed questions — V-Dem generally aims for at least five per country-indicator, with roughly two-thirds intended to be nationals or residents. Contemporary expert identities are confidential. Ordinal answers are aggregated through Bayesian measurement models that adjust for differences in scale use and estimated reliability. Final values are model-generated latent estimates rather than direct administrative statistics.
- 01Observable events and expert knowledge
- 02Country experts
- 03Ordinal assessments
- 04Bayesian measurement model
- 05Underlying V-Dem indicators
- 06Bayesian factor aggregation
- 07Expression · Association · Clean Elections
- 08Additive + multiplicative EDI formula
- 09Electoral Democracy score
Why this design exists
Scores can change retrospectively
11 — our finding
What the evidence does, and does not, prove
India’s V-Dem Electoral Democracy score fell from 0.617 in 2014 to 0.374 in 2023. Elected officials and universal suffrage did not cause the decline. The fall is generated by Freedom of Expression, Clean Elections and Freedom of Association.
Our audit finds genuine post-2014 concerns in all three areas. However, comparable problems also existed before 2014, while several objective indicators — including election-management capacity, nationwide VVPAT deployment and the number of parties contesting Lok Sabha elections — moved in the opposite direction.
The evidence reviewed therefore does not adequately establish the magnitude of the post-2014 downgrade.
This is not proof that every V-Dem assessment is wrong, nor proof of intentional manipulation. It is evidence that headline democracy scores should not be accepted without examining the historical baseline and the underlying inputs.
Three questions every democracy ranking should answer
- 01
What exactly changed?
- 02
Was the same problem already present in the comparison year?
- 03
Does the magnitude of the score change match the magnitude of the factual change?
12 — check it yourself
Don’t trust our conclusion. Check the evidence.
Every source used in this audit is exposed to the reader. The purpose of this page is not to replace one authority with another. It is to make the assumptions underneath a headline democracy score visible enough that anyone can challenge them — including us.
Publisher, title and document type are recorded for every source. Links are deliberately not fabricated — where a canonical URL has not been verified, none is shown.
Verdict vocabulary used throughout: Downgrade not established · Mixed evidence · Some deterioration supported · Observable improvement · No material deterioration · Audit pending. “Downgrade not established” is a statement about the evidence reviewed. It is not a finding that V-Dem was wrong, and certainly not that it was dishonest.