Methodology

How 27 indicators become 6 battery scores — methodology v2.6.0.

Every revision is dated and documented; none is silent.

How a score is built

Each indicator is fetched in its own unit, then mapped onto a 0–100 scale by an explicit rule declared in the registry — the bounds below. A battery score is the weighted sum of its indicators: score = Σ weightᵢ × indicatorᵢ, with weights summing to exactly 1.0. There is no hidden term and no discretionary adjustment: the frontend you are reading never computes a score, it only displays what the pipeline computed.

100 means healthy, 0 means critical. Bounds are calibrated so that within each battery, the worst episode since 1990 lands around 10–25 and the best recent period around 70–85.

Missing data never moves a score. When a source has not published yet, the last known value stays in the calculation — for annual sources that is the normal regime, not a failure. What changes is coverage: the share of weight backed by data that is still within its expected publication cycle. Below 60% coverage a battery is flagged Degraded. We deliberately do not renormalize weights onto the available indicators: a flaky source would make a public score zigzag with no real-world change behind it.

Every version is permanent. The registry carries a semantic version; any change to a weight or a bound requires a version bump and a changelog entry, enforced by CI. Past scores keep the version they were computed under, so a revision never silently rewrites history. Every revision so far is listed under revision history.

The 6 batteries

Environment

Physical and biological state of the planet: climate, oceans, disasters, forests, air, biodiversity. Global aggregate only.

INDICATORWEIGHTNORMALIZATIONCADENCESOURCE
Global temperature anomaly (vs pre-industrial)
°C above pre-industrial (1880-1899 proxy, GISTEMP + 0.26)
25%
Linear (clamped)
1 → 100, 2 → 0 (linear, clamped)
plausible range −0.5 to 3.5
Monthly
overdue after 3 months
NASA GISS GISTEMP v4
Public domain
Global mean sea level rise
mm above the 1993 altimetry baseline
15%
Linear (clamped)
0 → 100, 300 → 0 (linear, clamped)
plausible range −50 to 700
Monthly
overdue after 45 days
NASA-SSH Global Mean Sea Level (PO.DAAC / Earthdata)
Public domain
Extreme events (weighted GDACS alerts)
weighted score over a rolling 90-day window
15%
Logarithmic
250 → 100, 4,000 → 0 (logarithmic)
plausible range 0 to 10,000
Real time
overdue after 40 days
GDACS (EU/UN disaster alerts)
Free with attribution
Global annual tree cover loss
million hectares / year
15%
Linear (clamped)
10 → 100, 35 → 0 (linear, clamped)
plausible range 5 to 60
Annual
overdue after 2.9 years
Global Forest Watch Data API
CC-BY-4.0
Global PM2.5 exposure (population-weighted)
µg/m³ (annual mean)
10%
Linear (clamped)
10 → 100, 40 → 0 (linear, clamped)
plausible range 5 to 60
Annual
overdue after 4.1 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
Red List Index (SDG 15.5.1, world)
index 0-1 (1 = minimal extinction risk)
20%
Linear (clamped)
1 → 100, 0.5 → 0 (linear, clamped)
plausible range 0 to 1
Annual
overdue after 3.0 years
Our World in Data
CC-BY-4.0

Economy

Global macroeconomic health: growth momentum, labour market, price stability and fiscal sustainability. World aggregates.

INDICATORWEIGHTNORMALIZATIONCADENCESOURCE
World GDP growth
% per year (constant USD)
30%
Linear (clamped)
6 → 100, −4 → 0 (linear, clamped)
plausible range −12 to 12
Annual
overdue after 20 months
World Bank Open Data (WDI + WGI)
CC-BY-4.0
World unemployment (ILO model)
% of the labour force
20%
Linear (clamped)
3 → 100, 8 → 0 (linear, clamped)
plausible range 0 to 30
Annual
overdue after 2.0 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
World inflation (CPI)
% per year
20%
Target band
1–3 → 100, falling to 0 at −2 and 10
plausible range −5 to 50
Annual
overdue after 20 months
World Bank Open Data (WDI + WGI)
CC-BY-4.0
World public debt
% of world GDP
30%
Linear (clamped)
50 → 100, 150 → 0 (linear, clamped)
plausible range 10 to 300
Twice a year
overdue after 3.0 years
IMF Data (WEO, API SDMX 2.1)
Free with attribution

Society

Human capital and social cohesion worldwide: education, health, gender equality, and violent death — both interpersonal and organized.

INDICATORWEIGHTNORMALIZATIONCADENCESOURCE
Education Index (HDI component)
index 0-1
30%
Linear (clamped)
0.85 → 100, 0.4 → 0 (linear, clamped)
plausible range 0 to 1
Annual
overdue after 3.6 years
UNDP Human Development Reports
CC-BY-3.0-IGO
Life expectancy at birth (world)
years
20%
Linear (clamped)
85 → 100, 55 → 0 (linear, clamped)
plausible range 40 to 90
Annual
overdue after 2.9 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
Infant mortality (world)
deaths under age 1 per 1,000 live births
Deliberate partial overlap with life_expectancy, of which infant mortality is a component — documented in the methodology.
15%
Logarithmic
5 → 100, 100 → 0 (logarithmic)
plausible range 1 to 160
Annual
overdue after 2.9 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
Gender Inequality Index (GII, world)
index 0-1
20%
Linear (clamped)
0.05 → 100, 0.7 → 0 (linear, clamped)
plausible range 0 to 1
Annual
overdue after 3.6 years
UNDP Human Development Reports
CC-BY-3.0-IGO
Intentional homicides (world)
victims per 100,000 people
8%
Linear (clamped)
1 → 100, 12 → 0 (linear, clamped)
plausible range 0.5 to 30
Annual
overdue after 4.1 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
Deaths in organized violence (world)
deaths per 100,000 people
8%
Logarithmic
0.15 → 100, 8 → 0 (logarithmic)
plausible range 0 to 30
Annual
overdue after 2.6 years
Uppsala Conflict Data Program (UCDP)
CC-BY-4.0

Innovation

Global innovation capacity and technology diffusion: R&D effort, connectivity and digital infrastructure.

INDICATORWEIGHTNORMALIZATIONCADENCESOURCE
World R&D expenditure
% of world GDP
40%
Linear (clamped)
3.5 → 100, 1.5 → 0 (linear, clamped)
plausible range 0.5 to 5
Annual
overdue after 4.1 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
Internet access
% of world population using the internet
35%
Linear (clamped)
95 → 100, 30 → 0 (linear, clamped)
plausible range 0 to 100
Annual
overdue after 2.0 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
Digital infrastructure (fixed + mobile composite)
composite index 0-100
Composite of fixed broadband and mobile subscriptions per 100 people, combined by an explicit formula rather than taken from a single published series. Usage and installed capacity are tracked separately: their divergence is itself a signal.
25%
Passthrough
already a 0–100 index (×1)
plausible range 0 to 100
Annual
overdue after 2.0 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0

Governance

State of the world's political institutions: electoral democracy, political stability, control of corruption, voice and accountability. Every indicator is a population-weighted world aggregate — what the average human being actually experiences.

INDICATORWEIGHTNORMALIZATIONCADENCESOURCE
Electoral democracy (V-Dem polyarchy, population-weighted)
index 0-1
35%
Linear (clamped)
0.55 → 100, 0.15 → 0 (linear, clamped)
plausible range 0 to 1
Annual
overdue after 2.3 years
V-Dem (via the OWID mirror)
CC-BY-SA-4.0
Political stability and absence of violence (WGI PV, population-weighted)
WGI estimate (≈ -2.5 to +2.5)
25%
Linear (clamped)
0 → 100, −1 → 0 (linear, clamped)
plausible range −2.5 to 2.5
Annual
overdue after 2.7 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
Control of corruption (WGI CC, population-weighted)
WGI estimate (≈ -2.5 to +2.5)
20%
Linear (clamped)
0 → 100, −1 → 0 (linear, clamped)
plausible range −2.5 to 2.5
Annual
overdue after 2.7 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0
Voice and accountability (WGI VA, population-weighted)
WGI estimate (≈ -2.5 to +2.5)
20%
Linear (clamped)
0 → 100, −1 → 0 (linear, clamped)
plausible range −2.5 to 2.5
Annual
overdue after 2.7 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0

Resources

Resource pressure and the energy transition: energy sufficiency, decarbonisation of electricity, critical mineral reserves and freshwater.

INDICATORWEIGHTNORMALIZATIONCADENCESOURCE
Energy supply per capita (world)
kWh per person per year
Deliberate editorial stance: too much energy is itself a pressure signal, so this indicator is scored as a band rather than a ladder. The consequence is assumed — the score peaked before ~2005.
30%
Target band
12,000–17,500 → 100, falling to 0 at 5,000 and 35,000
plausible range 3,000 to 50,000
Annual
overdue after 21 months
Our World in Data
CC-BY-4.0
Renewable share of world electricity
% of generation (12-month rolling average)
35%
Linear (clamped)
90 → 100, 15 → 0 (linear, clamped)
plausible range 0 to 100
Monthly
overdue after 4 months
Ember electricity data
CC-BY-4.0
Critical mineral reserves — years remaining (basket median)
years (median reserves/production: Cu, Li, Co, Ni, rare earths)
15%
Linear (clamped)
100 → 100, 15 → 0 (linear, clamped)
plausible range 5 to 500
Annual
overdue after 2.2 years
USGS Mineral Commodity Summaries
Public domain
Freshwater withdrawals per capita (world)
m³ per person per year
20%
Linear (clamped)
350 → 100, 800 → 0 (linear, clamped)
plausible range 100 to 2,000
Annual
overdue after 8.2 years
World Bank Open Data (WDI + WGI)
CC-BY-4.0

Sources and licences

Every source is open, automatable and legally redistributable. Anything that failed one of those three tests was cut rather than approximated — the Democracy Index (proprietary), Gallup World Poll (proprietary) and the Transparency International CPI (CC BY-ND, which forbids derivatives) are not in any formula.

SOURCELICENCEACCESSPUBLISHES EVERY
World Bank Open Data (WDI + WGI)
api.worldbank.org/v2
CC-BY-4.0Open12 months
IMF Data (WEO, API SDMX 2.1)
api.imf.org/external/sdmx/2.1
Free with attributionOpen6 months
Our World in Data
catalog.ourworldindata.org
CC-BY-4.0Open12 months
NASA GISS GISTEMP v4
data.giss.nasa.gov/gistemp
Public domainOpen45 days
NASA-SSH Global Mean Sea Level (PO.DAAC / Earthdata)
archive.podaac.earthdata.nasa.gov/podaac-ops-cumulus-protected/NASA_SSH_GMSL_INDICATOR
Public domainAPI key7 days
GDACS (EU/UN disaster alerts)
www.gdacs.org/gdacsapi
Free with attributionOpen2 days
Global Forest Watch Data API
data-api.globalforestwatch.org
CC-BY-4.0API key12 months
Ember electricity data
api.ember-energy.org
CC-BY-4.0API key45 days
V-Dem (via the OWID mirror)
ourworldindata.org/grapher/electoral-democracy-index
CC-BY-SA-4.0Open13 months
UNDP Human Development Reports
hdr.undp.org/data-center
CC-BY-3.0-IGOOpen13 months
USGS Mineral Commodity Summaries
www.usgs.gov/centers/national-minerals-information-center
Public domainOpen13 months
Uppsala Conflict Data Program (UCDP)
ucdp.uu.se/downloads
CC-BY-4.0Open13 months

Share-alike notice — V-Dem (via the OWID mirror)

Share-alike propagates to derivatives. Since 2.5.0 every score this project publishes is under CC BY-SA 4.0 — not only Governance — so this source no longer sets a battery apart. Source data is unaffected: each source keeps its own licence, and the twelve do not share one. Five are CC BY 4.0, three public domain, one CC BY-SA 4.0, one CC BY 3.0 IGO, and two (IMF, GDACS) are free with attribution without a standard licence.

Licence notice — Uppsala Conflict Data Program (UCDP)

CC BY 4.0 requires the citation, so it is published here rather than kept in a code comment: Shawn Davies, Therése Pettersson, Magnus Öberg, Organized violence 1989-2025, and violent political protests, Journal of Peace Research, 2026. Three datasets are combined — battle-related deaths, non-state conflict and one-sided violence — read from UCDP's versioned CSV archives. The UCDP API is not used: since version 26.1 it requires an access token issued by e-mail, and this project takes no source it cannot reach without a private credential.

Editorial choices we stand behind

Normalization bounds are judgement calls. These 6 are the ones most open to attack — all of them deliberate, and all of them defensible only if we write them down in the open.

Energy per capita is scored as a band, not a ladder

More energy is not automatically better. The Resources battery treats 12,000-17,500 kWh per person per year as the healthy band, anchored on the ETH Zurich 2000-watt society. The consequence is deliberate: this score peaked before ~2005 and has been falling since. We are saying that today's growth in energy use is itself a pressure signal.

The Red List Index is rescaled to 0.5-1.0, not passed through

A straight passthrough would score biodiversity around 72 and move roughly one point per decade — a reassuring number for what the UN calls an extinction crisis. Rescaling from 0.5 makes the trend legible, at the cost of a lower bound that is our editorial judgement rather than a published threshold.

The governance bounds are relative, and demanding

The Worldwide Governance Indicators are built so that the average country sits near 0 each year. We score the population-weighted world against a ceiling of 0.0 — meaning 100 would require the average human to live in a country of average governance. That has never happened in 29 years of data. We keep the target rather than flatter the score.

Earthquakes count in Environment

GDACS mixes climatic and geophysical hazards. Earthquakes say nothing about human-driven environmental decline, but their civilizational impact is real and the battery is called Environment, not Climate. Filtering them out would be the more contestable choice.

Temperature is scored against 2.0 °C, and will need recalibrating

The Paris upper limit is the most defensible public anchor, but on the current trajectory this indicator reaches 0 around 2045 while the world keeps warming. A recalibration will be legitimate then — and it will be a documented methodology version, never a silent adjustment.

Every published score carries one licence: CC BY-SA 4.0

V-Dem's share-alike clause propagates to the Governance battery score, and the composite includes Governance. Rather than run two licensing regimes inside one dataset, every score this project publishes — the 27 normalized indicators, the 6 batteries, the composite — is released under CC BY-SA 4.0. One sentence stays true everywhere; the accepted cost is that share-alike binds derivative reuse. Source data keeps each source's own licence, listed per source on this page.

Known limits

World aggregates only. Every score describes the planet, not a country. Country-level scoring is a separate project, not a toggle.

Almost everything here moves once a year. Only 5 of the 27 indicators republish more often: Global temperature anomaly (vs pre-industrial) (monthly), Global mean sea level rise (monthly), Extreme events (weighted GDACS alerts) (real time), World public debt (twice a year), Renewable share of world electricity (monthly). The other 22 describe a calendar year and arrive months — sometimes years — after it ends. Society, Innovation, Governance are built entirely on annual series. The slowest, Freshwater withdrawals per capita (world), may lag 8.2 years before we flag it overdue.

Gender inequality is not income inequality. The World Bank Gini series has no world aggregate at all, so the Society battery measures gender inequality and says so, rather than implying a coverage it does not have.

The WGI scale is relative by construction. Governance measures the gap between the average human's country and the average country, not an absolute level of good government.

What the pipeline actually collected last night, source by source, is on the status page.

Revision history

Every version of the registry since it became the single source of truth, newest first. Each one is a permanent snapshot: past scores keep the version they were computed under. Most revisions change what is published — a label, a licence, a staleness threshold — without moving a single score, and they say so.

The date is when the revision was written, not when it started counting: a version only applies to scores once a collection run has recomputed under it. When the two differ, the banner at the top of this page says so.

  1. v2.6.016 Aug 2026Computed the scores on this page

    Deaths in armed conflict now have an indicator: UCDP's three organized-violence datasets, scored per 100,000 people, sharing Society's violence weight with the homicide rate. Society and the composite both fall.

  2. v2.5.112 Aug 2026

    A published note about source licences was wrong; it is corrected, along with three stale comments. Scores unchanged.

  3. v2.5.012 Aug 2026

    Every published score now carries a single licence, CC BY-SA 4.0. Scores unchanged.

  4. v2.4.011 Aug 2026

    The five editorial calibration choices moved into the registry, and licence names are spelled out in full. Scores unchanged.

  5. v2.3.011 Aug 2026

    The composite index and the homepage display labels are now declared in the registry. Battery scores unchanged.

  6. v2.2.011 Aug 2026

    Every published string in the registry switched to English. Scores unchanged.

  7. v2.1.411 Aug 2026

    The global temperature unit now names the baseline actually used. No bound or value changed.

  8. v2.1.311 Aug 2026

    Two more staleness thresholds widened to match real publication lags. No bound or weight changed.

  9. v2.1.211 Aug 2026

    Staleness thresholds recalibrated on the first full six-battery run, so the flag marks a broken publishing cycle rather than a structurally slow source. No bound or weight changed.

  10. v2.1.111 Aug 2026

    Two implementation fixes found by live collector tests: the extreme-events staleness window, and the UNDP world geography code.

  11. v2.1.011 Aug 2026

    Live verification of all eight sources. Sea level moved from a frozen NOAA feed to NASA-SSH, the innovation index indicator was dropped for lack of a comparable series, and several weights and bounds were revised — this revision does change scores.

  12. v2.0.311 Aug 2026

    Deforestation now reads a pre-aggregated country dataset instead of a raster that requires a geometry per request. No bound or weight changed.

  13. v2.0.211 Aug 2026

    Two staleness thresholds corrected by the first real run. No bound, weight or indicator changed.

  14. v2.0.111 Aug 2026

    Collector fixes from live API checks, including rebuilding the world public-debt aggregate from the official IMF SDMX API. No bound, weight or indicator changed.

  15. v2.0.011 Aug 2026

    The owner adopted the review's five arbitrations, including the public rule for calibrating bounds across batteries.

  16. v2.0.0-draft.111 Aug 2026

    The editorial review landed: health split into life expectancy and infant mortality, sea level became a cumulative level rather than a rate, and air quality moved to the World Bank series.

  17. v2.0.0-draft11 Aug 2026

    The registry was rebuilt as the single source of truth for 26 indicators and 6 batteries, replacing three divergent V1 definitions. Bounds and weights provisional.