30 economies · 2000–2024 ESTABLISHED DEMOCRACY Panel year 2024

Hard-Metric Democracy Index

Ten published counts and rates. No expert survey. Five pillars, aggregated by geometric mean

PANEL MEAN, 2024

72.4 / 100

arrow_downward −0.1 vs 2023

Highest scoring Norway 90.1 — Very high countable health

Lowest scoring China 28.6 — Very low countable health

Spread, top to bottom 61.5 pts The gap a panel mean hides. Watch it against the mean on the trajectory chart.

Indicators scored 10 of 12 Two source columns restate two others and are published as context rather than scored. Why.

Sits on a source 19% 1,671 of 9,000 cells. The rest is interpolated between anchors. Read this first.

What this index cannot see. Judicial independence, press pluralism and whether an election was actually free are the things that matter most, and none of them are a number anyone publishes — so none of them are here. A high score means the countable things look healthy. It is not a verdict on a country. And 81% of the panel's cells are interpolated between source-backed years rather than measured in them, so a smooth line is often inference, not stability.

Year and pillar weights

2024

2000 2024

Electoral 20%

Power Dispersion 20%

Economic Equity 20%

Information 20%

Due Process 20%

Equal weights are the published default because there is no empirical basis for saying due process is 1.4× electoral health. Move them and watch the table: if a ranking survives every weighting, the weights were never carrying it.

Rankings, 2024

Click a row for the full indicator breakdown

# ▲	Economy ▲	Composite ▲	Pillars	V-Dem # ▲	Classification ▲	On source ▲
1	NOR Norway	90.1	90918610085	10.85	Very high countable health	70%
2	SWE Sweden	89.5	95928210080	20.85	Very high countable health	70%
3	BEL Belgium	88.7	94938710072	50.81	Very high countable health	70%
4	NLD Netherlands	88.4	92878410080	90.77	Very high countable health	70%
5	DEU Germany	87.2	91817610090	80.78	Very high countable health	70%
6	ESP Spain	85.5	86867010088	130.74	Very high countable health	70%
7	POL Poland	83.2	88688010083	180.61	Very high countable health	70%
8	FRA France	82.9	80797610081	60.80	Very high countable health	70%
9	IRL Ireland	82.7	81697910087	40.83	Very high countable health	70%
10	CHE Switzerland	80.9	77867210073	30.84	Very high countable health	70%
11	AUS Australia	80.8	90757410069	70.80	Very high countable health	70%
12	IDN Indonesia	79.2	92695710085	220.33	High countable health	80%
13	ITA Italy	78.4	78726610080	160.69	High countable health	70%
14	GBR United Kingdom	76.8	67716610085	110.75	High countable health	70%
15	JPN Japan	76.3	72537010098	140.74	High countable health	70%
16	CAN Canada	76.0	76687110069	120.74	High countable health	70%
17	SGP Singapore	73.2	86466110088	210.35	High countable health	70%
18	KOR South Korea	73.2	73527310076	170.63	High countable health	80%
19	THA Thailand	72.3	73616610067	230.31	High countable health	100%
20	ARG Argentina	71.8	87854210062	200.56	High countable health	70%
21	MEX Mexico	71.3	78814510065	250.25	High countable health	80%
22	ISR Israel	67.7	8969587851	190.60	High countable health	90%
23	IND India	67.2	7753659654	240.28	High countable health	90%
24	USA United States	65.6	79585010054	100.75	High countable health	90%
25	TUR Turkey	62.4	7158438364	260.11	Mixed countable health	100%
26	BRA Brazil	60.3	88662210061	150.71	Mixed countable health	80%
27	ARE United Arab Emirates	54.6	2050729371	270.08	Mixed countable health	20%
28	RUS Russia	45.4	3727515568	280.06	Low countable health	100%
29	SAU Saudi Arabia	31.9	1020397060	290.05	Very low countable health	20%
30	CHN China	28.6	626365171	300.04	Very low countable health	70%

The On source column is the share of that row's ten indicators anchored to a published figure for that year rather than interpolated from neighbouring years. A row at 20% is four-fifths inference; read its score accordingly.
The grey V-Dem # column is that country's rank on V-Dem's expert-coded Liberal Democracy Index within these same thirty (its 0–1 score beneath it), for the selected year. It is a comparator and never an input: it does not move when the weight sliders move. Where the two ranks disagree is where counting and expert judgement tell different stories — the divergences are tabulated in the methodology.

Trajectories, 2000–2024

Composite score per year against the unweighted panel mean. The vertical axis is scaled to the plotted series, so the visible range changes as countries and weights change.

Trajectory chart needs JavaScript. The ranking table above is complete without it.

Movement, 2000 to 2024

Change in composite over the full panel. A quarter-century is long enough that these are structural moves rather than the noise of one election, but every bar inherits the interpolation caveat: the endpoints are anchored more often than the middle.

Chart needs JavaScript.

Which pillar moved

Panel mean for each pillar, year by year. The composite is a product of these five, so a flat headline with two pillars moving in opposite directions is a different world from a genuinely stable one.

Chart needs JavaScript.

Executive Summary

What 30 economies and 10 years of counting add up to, in plain language. Read under the published equal weighting, not whatever the sliders above are currently set to.

The panel mean has risen by +1.4 points in 10 years — 71.0 in 2014 to 72.4 in 2024 — and that average is the least interesting number here. The gap between the highest and lowest scoring economy grew 3.4 points wider over the same decade, from 58.1 to 61.5. 20 of 30 economies improved, 8 declined, 2 did neither. This is divergence, not a global trend.

What actually drove the change?

One pillar, and inside it largely one indicator. Power Dispersion & Checks contributes +1.6 points of the +1.4-point move; the other four pillars together account for -0.2. The figures below are an exact decomposition — they sum to the headline change with nothing left over.

Power Dispersion +1.64 pts

Pillar mean 59.8 → 66.4. Largest indicator inside it: gender parity +12.0 pts of score, moving the right way in 27 of 30 economies and the wrong way in 0.
Economic Equity +0.37 pts

Pillar mean 62.5 → 64.0. Largest indicator inside it: gini +1.5 pts of score, moving the right way in 19 of 30 economies and the wrong way in 5.
Information -0.28 pts

Pillar mean 96.5 → 94.2. Largest indicator inside it: shutdowns -3.5 pts of score, moving the right way in 2 of 30 economies and the wrong way in 4.
Due Process -0.28 pts

Pillar mean 75.2 → 73.9. Largest indicator inside it: pre-trial -2.9 pts of score, moving the right way in 10 of 30 economies and the wrong way in 18.
Electoral -0.01 pts

Pillar mean 73.6 → 74.1. Largest indicator inside it: turnout +4.3 pts of score, moving the right way in 15 of 30 economies and the wrong way in 12.

Read the top line sceptically. Women in the national legislature is a seat count, and a seat count cannot tell an elected chamber from an appointed one, or a quota from a contest. It moves this index because it is published and comparable, which is the whole basis on which anything is in here — not because a rise in it proves a country became more democratic.

Which countries moved most, and why?

Six of 30, three each way. Each is shown with the two pillars that carry most of its move and the single indicator inside them that moved furthest.

Thailand 57.9 → 72.3 +14.4 pts

8 places up to rank 19, crossing from “Mixed countable health” into “High countable health”. Electoral +6.8 pts (transfer integrity 10.0 → 65.0%); Power Dispersion +4.9 pts (gender parity 4.8 → 19.4%).
United Arab Emirates 45.4 → 54.6 +9.2 pts

1 place up to rank 27, crossing from “Low countable health” into “Mixed countable health”. Power Dispersion +8.5 pts (gender parity 21.2 → 50.0%); Electoral +0.6 pts (turnout 32.5 → 35.2%).
Japan 70.6 → 76.3 +5.7 pts

4 places up to rank 15. Power Dispersion +4.2 pts (concentration 4,700 → 3,500 HHI); Electoral +1.1 pts (disproportionality 16.1 → 11.2).
Russia 57.9 → 45.4 -12.6 pts

2 places down to rank 28, crossing from “Mixed countable health” into “Low countable health”. Information -6.1 pts (shutdowns 1.3 → 92.0 hrs/cap); Power Dispersion -3.6 pts (concentration 5,860 → 8,100 HHI).
China 36.6 → 28.6 -8.1 pts

1 place down to rank 30, crossing from “Low countable health” into “Very low countable health”. Electoral -6.7 pts (transfer integrity 40.0 → 10.0%); Information -2.0 pts (shutdowns 72.5 → 115 hrs/cap).
Israel 74.4 → 67.7 -6.7 pts

6 places down to rank 22. Information -3.5 pts (journalists held 0 → 6.3 per 10M); Due Process -3.4 pts (pre-trial 38.0 → 52.0%).

9 of 30 economies changed tier over the decade. A tier change is a threshold being crossed, not a transformation: a country a tenth of a point either side of 65 is the same country.

Does counting agree with the experts?

Mostly, and the disagreements are the point. Against V-Dem's expert-coded Liberal Democracy Index for 2024, the rank correlation is ρ = 0.82 across the 30. High enough that this index is measuring democracy rather than something else; far from 1.0, so it is not reproducing V-Dem by a longer route.

Poland (11 places) and Indonesia (10 places) place better on countable evidence than expert coders will allow — the mechanics look intact, the judgement of how they are used does not. United States (-14) and Brazil (-11) place worse: the counts penalise things — disproportional seat allocation, incarceration, two-party concentration — that experts weigh lightly against a functioning liberal order. Neither ranking is the correction to the other.

How much of this is measured, and how much is inferred?

Across the whole 2000–2024 panel, 18% of scored cells sit on a published source; the rest are interpolated between anchor years. In 2024 itself the figure is 73%, because recent years are the ones with the most reporting behind them.

Coverage is not spread evenly, and that is the more serious problem. Saudi Arabia rests on 7% source-backed cells — its trajectory is mostly a line drawn between distant points. Closed states publish least, so the scores this index is least sure of are exactly the ones a reader most wants to trust. Per-country coverage is tabulated in the methodology.

Bottom line

On countable evidence the last decade is divergence, not decline: a mean that drifted +1.4 points while the distance between the ends of the table grew 3.4 points wider. The improvement is concentrated in power dispersion and is largely one published seat-share statistic; the deterioration is concentrated in information and due process, where the falls (shutdowns worst among them) are the kind of thing states do deliberately and then report.

None of this is a verdict on any country. This index scores what is counted, and the things that matter most — whether courts are independent, whether an election was free, whether a newspaper can print — are not counted by anyone. A high score means the countable things look healthy. Where the counts and the experts disagree, treat both as evidence and neither as the answer.

What this index measures, and what it refuses to

V-Dem, Freedom House and the EIU all build their democracy scores from expert questionnaires: country specialists answer ordinal questions and the answers are aggregated. That method sees things no counter can, and it carries an irreducible coder-subjectivity term that no inter-rater statistic removes.

This index takes the opposite trade. Every input is a count, a rate or a ratio somebody published — votes over voting-age population, seat shares through Gallagher's formula, women's seats over total seats, detainees over prisoners, prisoners over population. Nothing here is an opinion about a country, and no country's classification, regime type or region touches its score.

The cost of that discipline

The things hardest to count are often the things that matter most. Judicial independence, press pluralism, media capture, whether an election was actually free — none appear below, because none is a number anyone publishes. An index that only counts what is counted will systematically miss a country that keeps its forms while hollowing out their substance, which is the most common way democracies decline.

Read a high score as the countable things look healthy, not as this is a good democracy. Where this index and a survey-based one disagree, the disagreement is the finding — this one is not the tiebreaker.

Five pillars, ten scored indicators

Every indicator is a rate, ratio or per-capita quantity, so population size drops out

Electoral Health & Representation

Do people vote, and do the votes turn into seats?

Turnout — %, higher better
Disproportionality — index, lower better
Transfer integrity — %, higher better

Power Dispersion & Checks

Is legislative power concentrated, and who holds it?

Concentration — HHI, lower better
Gender parity — %, higher better

Economic Equity & Parity

How unequally is national income distributed?

Gini — index, lower better

Information & Telemetry Freedom

Does the state switch off the network or jail reporters?

Shutdowns — hrs/cap, lower better
Journalists held — per 10M, lower better

Due Process & Rule of Law

How many are held, and how many are held without trial?

Pre-trial — %, lower better
Incarceration — per 100k, lower better

Indicator	Pillar	Formula	Scores 0 at	Scores 100 at	Source
Voter turnout (% of voting-age population)	Electoral	`votes cast / voting-age population`	20.0	95.0	International IDEA Voter Turnout Database
Against VAP rather than registered voters, so a country cannot raise its turnout by shrinking its electoral roll. Compulsory-voting states (Belgium, Australia, Brazil) score high here for a legal reason rather than a civic one; this index does not adjust for that, and the turnout column should be read with it in mind.
Gallagher disproportionality index	Electoral	`sqrt( 0.5 * sum (v_i - s_i)^2 )`	35.0	1.0	Michael Gallagher, Election Indices
How far the seat allocation departs from the vote. Near 0 under proportional systems, high under first-past-the-post — it measures the electoral system as much as the conduct of the election.
Constitutional transfer integrity	Electoral	`share of executive transfers made under constitutional rule`	0.0	100.0	Comparative Constitutions Project; succession records
Coups, annulled results and irregular successions pull this down; it is the one indicator here that is a coded judgement about events rather than a published count, and 24 of 30 countries sit at 100, so it discriminates only at the bottom.
Legislative concentration (HHI)	Power Dispersion	`sum of squared seat-share percentages (0-10,000)`	10000.0	1200.0	National electoral registries; IPU Parline
10,000 is a single party holding every seat. The Effective Number of Parties is the same statistic inverted (10000 / HHI) and is published as context rather than scored a second time.
Women in the national legislature	Power Dispersion	`seats held by women / total seats`	0.0	50.0	IPU Parline, Women in National Parliaments
Scored against 50% as parity rather than as a maximum. This indicator cannot tell an elected chamber from an appointed one: the UAE Federal National Council is half women by appointment and scores 100 here. That is a real limitation, and the geometric aggregation is what stops it from carrying a country's composite.
Income Gini coefficient	Economic Equity	`standard Lorenz-curve inequality index (0-100)`	60.0	22.0	World Bank Poverty & Inequality Platform; WID
The single scored indicator in its pillar. The Palma ratio is the same Lorenz curve summarised differently (rho = +0.99 here) and is published as context rather than scored, which would have double-weighted income inequality.
Internet disruption (person-hours per capita)	Information	`sum(outage hours x affected population) / national population`	120.0	0.0	OONI; NetBlocks; SFLC; Cloudflare Radar
State-ordered outages only. 21 of 30 countries sit at zero.
Journalists imprisoned or killed, per 10M people	Information	`documented cases / population, per 10 million`	15.0	0.0	Committee to Protect Journalists, annual census
A floor detector, not a discriminator: it separates states that jail reporters from states that do not, and says nothing about the 21 countries at zero.
Pre-trial detention rate	Due Process	`unconvicted detainees / total prison population`	80.0	8.0	World Prison Brief; UNODC; NCRB
Detention before trial as a share of everyone detained — judicial backlog made countable.
Incarceration rate	Due Process	`prisoners / population, per 100,000`	750.0	30.0	World Prison Brief
Rewards states that publish low numbers, including those that detain outside the prison system or do not report at all. Reads as a carceral-burden measure for countries with credible statistics and as an artefact for those without.

Bounds are fixed, not sample-derived. Normalising against the sample would restate every historical score the moment a country was added, and would make the index able to say only “worse than these thirty” rather than “bad in absolute terms”. Values outside the bounds clamp; the diagnostics below report how often that bites.

Aggregation: why a geometric mean

composite = exp( ∑_p w_p · ln( max(pillar_p, 1) ) ) with ∑ w_p = 1

Indicators average arithmetically within a pillar. The five pillars then combine geometrically, not arithmetically, because an arithmetic mean makes pillars perfectly substitutable — a country can buy back a pillar score of zero with a hundred somewhere else. Democracy does not work that way; the pillars are complements.

This is not hypothetical. Under an arithmetic mean the United Arab Emirates — a federal absolute monarchy with no elected national legislature — scored 61.4, because a legislative-concentration score of 0 was averaged against an appointed 50% female chamber scoring 100, and the resulting 50 was then averaged away. Geometrically it scores 54.6. The switch is the standard fix — it is why the Human Development Index changed in 2010 — and it barely disturbs the league table: the top twenty move by at most one place.

Economy	Geometric (published)	Arithmetic (rejected)	Compensation denied	Weakest pillar
Indonesia	79.2	80.8	−1.6	Economic Equity (57)
Japan	76.3	78.4	−2.1	Power Dispersion (53)
United Kingdom	76.8	77.8	−1.0	Economic Equity (66)
Singapore	73.2	76.0	−2.9	Power Dispersion (46)
Argentina	71.8	75.1	−3.3	Economic Equity (42)
South Korea	73.2	74.8	−1.6	Power Dispersion (52)
Mexico	71.3	73.7	−2.4	Economic Equity (45)
Thailand	72.3	73.4	−1.1	Power Dispersion (61)
Israel	67.7	69.0	−1.3	Due Process (51)
India	67.2	69.0	−1.8	Power Dispersion (53)
United States	65.6	67.9	−2.4	Economic Equity (50)
Brazil	60.3	67.5	−7.2	Economic Equity (22)
Turkey	62.4	63.8	−1.5	Economic Equity (43)
United Arab Emirates	54.6	61.4	−6.8	Electoral (20)
Russia	45.4	47.6	−2.3	Power Dispersion (27)
Saudi Arabia	31.9	39.9	−8.0	Electoral (10)
China	28.6	37.9	−9.3	Electoral (6)

Only rows where the two methods differ by more than a point. The gap is exactly the compensation an arithmetic mean would have granted for an unbalanced profile. Pillar scores are floored at 1.0 before the log, so a genuine zero darkens a composite hard without collapsing it and discarding every other measurement in the row.

Two indicators were the same indicator

Spearman ρ across the panel, 2024

The source dataset carried twelve scored indicators. Two pairs turned out to be one statistic counted twice:

Effective Number of Parties and legislative HHI are algebraically identical — ENP = 10000 / HHI, exactly, for seat shares. They had been keyed independently and disagreed with each other by a median of 6.6% (24% for Brazil), which is a self-inconsistency rather than two readings. Scoring both put party fragmentation into the composite at 15% weight through two different pillars. HHI is now scored; ENP is derived from it and published as context, which also removes the disagreement.
Gini and the Palma ratio summarise one Lorenz curve and correlate at ρ = +0.99 here. Gini is scored; Palma is published as context. Scoring both would have double-weighted income inequality inside a pillar that is already 20% of the index.

The table below is left on the page permanently rather than deleted once the problem was fixed. The only reason anyone found those two pairs is that somebody computed this; leaving it visible is what makes the next redundancy findable.

Indicator A	Indicator B	ρ	Reading
Disproportionality	Concentration	+0.855	Strongly related; plausibly one underlying cause
Shutdowns	Journalists held	+0.786	Strongly related; plausibly one underlying cause
Transfer integrity	Shutdowns	-0.667	Related, distinct enough to score separately
Gini	Incarceration	+0.636	Related, distinct enough to score separately
Transfer integrity	Journalists held	-0.596	Related, distinct enough to score separately
Gini	Journalists held	+0.553	Related, distinct enough to score separately
Transfer integrity	Concentration	-0.529	Related, distinct enough to score separately
Concentration	Journalists held	+0.516	Related, distinct enough to score separately
Gini	Shutdowns	+0.503	Related, distinct enough to score separately
Transfer integrity	Incarceration	-0.493	Independent

Highest remaining pair after the fix is disproportionality against concentration (+0.855), which is expected: both are downstream of the electoral system. They measure different consequences of it — how votes map to seats, and how seats are held — so both are scored.

Where this parts company with V-Dem

V-Dem Liberal Democracy Index (v2x_libdem), 2024

This index is built in opposition to expert-coded indices, and an argument like that is worth nothing if the reader cannot see the other side of it. The grey column in the rankings is each country's rank on V-Dem's Liberal Democracy Index within these same thirty — not V-Dem's global rank over 179 countries, which would not be comparable with a rank out of 30. V-Dem is never an input here: no expert score touches the composite, and the grey column does not move when the weight sliders do.

Across the panel the two rankings agree at Spearman ρ = +0.819 for 2024. That is the number to hold onto: high enough that counting recovers most of what country experts see, and far enough from 1.0 that the residual is real. The residual is the table below.

Economy	HMDI #	V-Dem #	Gap	What the gap is
United States	24	10 (0.75)	-14	Experts judge it better than it counts — carceral, turnout or disproportionality figures drag a country whose institutions V-Dem rates highly.
Poland	7	18 (0.61)	+11	Counts better than experts judge it — turnout, proportionality and parity look healthier than the qualities nothing here measures.
Brazil	26	15 (0.71)	-11	Experts judge it better than it counts — carceral, turnout or disproportionality figures drag a country whose institutions V-Dem rates highly.
Indonesia	12	22 (0.33)	+10	Counts better than experts judge it — turnout, proportionality and parity look healthier than the qualities nothing here measures.
Spain	6	13 (0.74)	+7	Counts better than experts judge it — turnout, proportionality and parity look healthier than the qualities nothing here measures.
Switzerland	10	3 (0.84)	-7	Experts judge it better than it counts — carceral, turnout or disproportionality figures drag a country whose institutions V-Dem rates highly.
Netherlands	4	9 (0.77)	+5	Counts better than experts judge it — turnout, proportionality and parity look healthier than the qualities nothing here measures.
Ireland	9	4 (0.83)	-5	Experts judge it better than it counts — carceral, turnout or disproportionality figures drag a country whose institutions V-Dem rates highly.
Australia	11	7 (0.80)	-4	Experts judge it better than it counts — carceral, turnout or disproportionality figures drag a country whose institutions V-Dem rates highly.
Canada	16	12 (0.74)	-4	Experts judge it better than it counts — carceral, turnout or disproportionality figures drag a country whose institutions V-Dem rates highly.

Positive gap = better on hard counts than on expert coding; negative = the reverse. Sorted by absolute gap, so the rows that most need explaining are at the top. Ties on the V-Dem score share the lower rank. Source: Varieties of Democracy project, via Our World in Data — V-Dem v15 (data through 2024), scale 0-1, higher is more liberal-democratic. Regenerate with scripts/build_vdem_reference.py.

Which indicators actually discriminate

Normalised score spread across the panel, 2024

An indicator on which most of the panel scores exactly 100 is not measuring most of the panel; it is a floor detector, separating states that do a thing from states that do not. That is a legitimate job. A reader is entitled to know which indicators are doing it, because they inflate every clean country's score by the same amount and discriminate only at the bottom.

Indicator	Mean	SD	At 100	At 0	Clamped	Reads as
Journalists imprisoned or killed, per 10M people	95.8	10.7	21	0	0	Floor detector — silent on the 21 at the ceiling
Income Gini coefficient	64.0	16.2	0	0	0	Discriminates across the panel
Incarceration rate	80.7	16.9	0	0	0	Discriminates across the panel
Pre-trial detention rate	67.0	18.7	0	0	0	Discriminates across the panel
Internet disruption (person-hours per capita)	92.6	21.7	21	0	0	Floor detector — silent on the 21 at the ceiling
Women in the national legislature	63.0	21.9	2	0	1	Discriminates across the panel
Voter turnout (% of voting-age population)	63.7	23.4	1	1	2	Discriminates across the panel
Constitutional transfer integrity	85.2	27.6	21	0	0	Floor detector — silent on the 21 at the ceiling
Legislative concentration (HHI)	69.9	29.3	1	3	1	Discriminates across the panel
Gallagher disproportionality index	73.4	29.4	0	3	0	Discriminates across the panel

Clamped counts panel-wide values (all 25 years) pinned by a scoring bound rather than by their own magnitude — where it is non-zero, the bound rather than the country decided part of that column.

Provenance: what is measured and what is inferred

1,671 of 9,000 cells sit on a source

This panel is not twelve annual series

Twelve indicators for thirty economies over twenty-five years is 9,000 cells. Nobody keyed 9,000 cells. What was keyed is 1,671 anchor points — a value attached to a year for which a real source exists: an election result, an IPU Parline snapshot, a World Prison Brief report, a World Bank PIP release, a CPJ census — plus a number of series asserted flat across the whole period. Everything between two anchors is straight-line interpolation; everything outside the outermost anchors is held flat.

That is a reasonable way to carry sources that do not publish annually. An election result exists in election years and not between them. It is not the same thing as measurement, and the difference has to survive into the published figures rather than being flattened into a table where every number looks equally solid.

So every row carries its anchor share, the ranking table shows it, and the chart below shows it by year. Where the density is thin, a smooth trajectory is the interpolation drawing a straight line — not evidence of a stable decade.

Chart needs JavaScript; the table below carries the same figures.

Anchor coverage by economy

Thinnest evidence first. These are the rows to trust least.

Economy	Anchored cells	Share
Saudi Arabia	18 / 250	7.2%
United Arab Emirates	20 / 250	8.0%
China	35 / 250	14.0%
Belgium	40 / 250	16.0%
France	43 / 250	17.2%
Italy	43 / 250	17.2%
Ireland	43 / 250	17.2%
Singapore	43 / 250	17.2%
Germany	44 / 250	17.6%
Switzerland	44 / 250	17.6%
Argentina	44 / 250	17.6%
Sweden	44 / 250	17.6%
Norway	44 / 250	17.6%
United States	45 / 250	18.0%
Japan	46 / 250	18.4%
United Kingdom	46 / 250	18.4%
Canada	46 / 250	18.4%
Indonesia	46 / 250	18.4%
Brazil	47 / 250	18.8%
Netherlands	47 / 250	18.8%
Poland	48 / 250	19.2%
Mexico	49 / 250	19.6%
Australia	49 / 250	19.6%
South Korea	50 / 250	20.0%
Spain	50 / 250	20.0%
India	53 / 250	21.2%
Russia	56 / 250	22.4%
Israel	56 / 250	22.4%
Thailand	58 / 250	23.2%
Turkey	62 / 250	24.8%

Bar scaled to a 25% ceiling, not 100% — no economy in the panel exceeds a quarter. Anchors are the input of record and are interpolated into the published panel; changing a figure means changing an anchor, never the generated series.

Known limitations

Each of these is a real defect, not a disclaimer

Compulsory voting is scored as civic participation. Belgium, Australia and Brazil post high turnout for a legal reason. No adjustment is made, so the turnout column reads partly as a map of compulsory-voting law.
Gender representation cannot tell an elected chamber from an appointed one. The UAE Federal National Council is half women by appointment and scores 100 on that indicator. Geometric aggregation stops it carrying the composite, but the indicator itself is blind here.
Incarceration and pre-trial rates reward states that publish low numbers. A state that detains outside the prison system, or reports partially, scores as though it detains little. The pillar reads as carceral burden where statistics are credible and as an artefact where they are not — which is exactly backwards from where it is most needed.
More parties is treated as more plural. Legislative concentration scores a two-party system poorly by construction. That is a defensible reading of power dispersion and a contestable reading of democratic health; a stable two-party system is not obviously less democratic than a fragmented multi-party one.
Transfer integrity is the one coded judgement here. It is a scored reading of historical events rather than a published count, and 24 of 30 countries sit at 100, so it discriminates only at the bottom of the panel.
Zero turnout is recorded as a zero, not as an absence. Years in which a country held no national election carry a 0 that scores as catastrophic participation rather than as a structurally different situation.
The sample is the thirty largest economies. That is a selection on GDP, not on population, region or regime, and every panel-wide average inherits it. It is not a world average.

Data sources

Where each anchor value comes from

Every anchor point is keyed from one of the publishers below. The indicator table above names the source for each of the ten scored indicators; these are the underlying datasets.

Turnout International IDEA — Voter Turnout Database
National election turnout as a share of voting-age population.
Disproportionality Michael Gallagher — Election Indices
Gallagher index of vote-to-seat disproportionality, plus effective number of parties.
Constitutions Comparative Constitutions Project
Constitutional text and succession records behind the transfer-integrity reading.
Legislatures IPU Parline
Chamber composition and women in national parliaments, with national electoral registries.
Inequality World Bank Poverty & Inequality Platform
Gini and Palma, cross-checked against the World Inequality Database.
Connectivity OONI
Shutdown hours corroborated across NetBlocks, SFLC.in and Cloudflare Radar.
Press Committee to Protect Journalists
Annual prison census and killed-journalist database, per 10M population.
Detention World Prison Brief
Incarceration and pre-trial rates, with UNODC and, for India, NCRB.

Panel generated 2026-09-02. Figures licensed CC BY 4.0.

—

Composite

—

Rank

—

On source

—

V-Dem rank

—

Indicators

Indicator	Value	Score	Src

Context, not scored

Measure	Value

Both restate a scored indicator — see the collinearity note.