The Entropic
Weather Report
An independent observatory of global transformationExperimental edition / Observation data

Method

Methodology

How the Entropy Index is built.

The Entropy Index runs from 0 to 100. It is the weighted mean of the eight signal scores. Each signal score is the weighted mean of its scored inputs.

The number beside each row of "What moved the needle" is that signal's impact in index points, over the live signals only with their weights renormalised: change since last week × weight share, or in the first week (score − 50) × weight share, comparing the signal with the midpoint.

The label shown always matches the band the value falls in. A boundary value belongs to the upper band.

BandFromTo
Collapse020
Dystopia2040
Uncertain4060
Progress6080
Utopia80100

The scale is labelled Extinction at 0 and Utopia at 100.

Equal weights, held in config/scoring.json (unit: percent). Any change is logged below with a date.

SignalWeight
Public mood12.5%
Social unrest12.5%
Cost of living12.5%
Work and jobs12.5%
Political disruption12.5%
Conflict12.5%
Technology & AI12.5%
Climate12.5%
Total100%

The past is a guide, not the scale. Neither Utopia nor total dystopia has ever been reached, so each scored input has two fixed anchors that describe states of the world beyond anything recorded: a best value that scores 100 and a worst value that scores 0. The score is a straight line between them (a log10 line where the scale is log), capped to 0–100. Higher is always better for the world. A signal is the weighted mean of its scored inputs (see the Weight column under Inputs).

The anchors are editorial judgements, open to revision, and live in config/sources.json. Each stored week records the anchor set version that scored it (version 5 is current); weeks stored before the anchors were introduced were scored by the percentile of the latest value within its own history and keep those scores. Every input has its own unit; the global temperature anomaly is measured in °C above the 1850–1900 pre-industrial level (the file's 1901–2000 base plus an offset of 0.167 °C).

SignalInputBest (100)Worst (0)ScaleReason
Cost of livingFAO Food Price Index90170linearFood prices well below any recent level are the Utopia end; far above the 2022 peak means essentials out of reach.
Cost of livingWorld Bank global consumer price inflation215linearWorld consumer price inflation of 2% a year is the stable end that central banks aim for; 15% a year is double-digit inflation that halves the value of money in about five years (the world annual figure has not been that high since the early 1990s).
ClimateNOAA Mauna Loa weekly CO2280560linear280 ppm is the pre-industrial level; 560 ppm is double it, far past anything we have lived with. 420 ppm scores 50. Anchor set 5 (9 Nov 2026): the earlier best of 350 ppm let today's level, about 428 ppm, score about 63, in the Progress band, which overstated it.
ConflictUCDP GED candidate events (deaths per week)10025,000logLog scale on the mean weekly deaths of the last 8 complete weeks: 100 a week is a world of minor contained conflicts; 25,000 a week is world-war level. Anchor set 5 (9 Nov 2026): from set 5 the score uses the 8-week average, not one week, and the anchors moved from 50 and 250,000 to fit that.
Political disruptionGlobal Economic Policy Uncertainty Index (FRED GEPUCURRENT)80450linearPolicy uncertainty at the calmest levels on record is the stable end; well above the pandemic peak is institutions failing.
Political disruptionWorld Uncertainty Index, GDP-weighted average (FRED WUIGLOBALWEIGHTAVG)5,000150,000logLog scale: a few thousand index points is a calm world with little uncertainty; 150,000 is well above the 2025 peak of about 107,000, uncertainty everywhere at once.
Cost of livingUS consumer price inflation, year on year (FRED CPIAUCSL)015linearPrices flat year on year is the stable end; 15% a year is double-digit inflation that halves the value of money in about five years.
Work and jobsUS initial jobless claims (FRED ICSA)150,000600,000linearClaims below the lowest of recent decades mean anyone can find work; far above the 2008 level means mass layoffs.
Public moodUS consumer sentiment (FRED UMCSENT)11030linearAbove the best readings on record is broad confidence; below the 2022 low is most people expecting things to get worse.
Work and jobsWorld Bank world unemployment rate310linear3% is effectively full employment; 10% of the world's labour force out of work is mass unemployment.
ClimateNASA GISTEMP global surface temperature anomaly (land and ocean)0.52.5linearSame anchors as the NOAA series (anchor set 5, 9 Nov 2026): degrees C above the 1850-1900 pre-industrial level, +0.5 is the best case and +2.5 is warming far past any level we have lived with, so +1.5 scores 50. Sets 3 and 4 used 0 and 4. offset 0.22 converts the file's 1951-1980 base to pre-industrial (the usual estimate of the warming between 1850-1900 and 1951-1980 is about 0.22 C).
ClimateNOAA NCEI global surface temperature anomaly (land and ocean, 1901-2000 baseline)0.52.5linearAnchors are degrees C above the 1850-1900 pre-industrial level (anchor set 5, 9 Nov 2026): +0.5 is the best case and +2.5 is warming far past any level we have lived with, so +1.5 scores 50. Sets 3 and 4 used 0 and 4. offset 0.167 converts the file's 1901-2000 base to pre-industrial (the 1850-1900 mean of this series is -0.167, computed from the raw file on 5 Nov 2026).
Social unrestIMF Reported Social Unrest Index, share of countries above 100 (HEADLINE)053linearAnchor set 5 (9 Nov 2026), no longer provisional: best 0 and worst 53 put the 1986-2026 median of the world share (26.3%) at 50, so a typical month scores 50 and the latest month, 24.43, scores 54. Sets 3 and 4 used 15 and 65 and scored the same month 81. Set from the 488 monthly points read from the IMF API (1986-01 to 2026-08). The value is the percent of reporting countries whose IMF Reported Social Unrest Index (HEADLINE) is above 100 in the month. A share was chosen over the cross-country median because about half of all country-months are exactly 0, so the median sits near zero and hardly moves until unrest is widespread; the share of countries in visible unrest does move. Measured history: median 26.3, quartiles 12.8 and 39.9, 10th and 90th percentiles 6.0 and 48.9, maximum 70.68 (2020-06); 2010-2015 ranged 16.54 to 59.4; since 2016 the low was 21.8 (2026-03). Worst 53 sits between the 90th percentile (48.9) and the 2019-2020 wave of mostly 54 to 66; best 0 is a month with no country in visible unrest.
Public moodEU27 consumer confidence indicator (Eurostat ei_bsco_m, BS-CSMCI)10-35linearprovisional: the balance of opinion from the European Commission consumer survey (seasonally adjusted, EU27). +10 is clearly more optimistic than ordinary times and is the Utopia end; -35 is worse than the energy-price shock low of 2022, most households expecting things to get worse. EU27 only, not global.
Work and jobsEU27 unemployment rate (Eurostat une_rt_m)5.512linearSet from the EU27 seasonally adjusted rate since January 2000 (321 months read from Eurostat): the low was 5.8% in 2024 and the high 11.7% in 2013. 5.5% is just under the lowest month on record, the stable end; 12% is just above the euro-crisis peak, mass unemployment. EU27 only, not global.
Cost of livingIMF world consumer price inflation (DataMapper PCPIPCH, WEOWORLD)220linear2% a year is the stable end that central banks aim for; 20% a year in the world average is hyperinflation-adjacent, far above anything the world aggregate has recorded in the last thirty years. The IMF series is the average of consumer prices, so it runs higher than the World Bank figure in crises; hence the wider worst anchor.
ConflictGeopolitical Risk index, global (Caldara and Iacoviello), scored on Conflict50300linearprovisional: the index is normalised so that its 1985-2019 mean is 100. 50 is half the long-run level, a quiet world, and is the best end; 300 is three times the long-run level and is the worst end (the spikes around 9/11 and the 2003 Iraq invasion went past it).
Political disruptionGeopolitical Risk index, global (Caldara and Iacoviello), scored on Political disruption50300linearprovisional: the index is normalised so that its 1985-2019 mean is 100. 50 is half the long-run level, a quiet world, and is the best end; 300 is three times the long-run level and is the worst end (the spikes around 9/11 and the 2003 Iraq invasion went past it).
Technology & AITone of technology coverage (GDELT GKG: wired.com, theverge.com, arstechnica.com, technologyreview.com)4-4linearprovisional: anchor set 5 (9 Nov 2026) centres the dial, so a mean tone of 0 scores 50: +4 (clearly upbeat coverage) is the Utopia end and -4 (consistently alarmed coverage) the worst end. Sets 3 and 4 used +3 and -5. Neither end is measured yet: the first live run (42 GKG files) found only 19 articles from the four outlets, below the minimum, so the sampling was raised to 24 files a day (168 files a week); replace both ends from the weekly history (scripts/gdelt-tech-history.mjs). A week with fewer than 30 articles from the four outlets across the sampled files gives no reading.

Some inputs are context only and are shown but never scored: the Gini index (the median of each country's latest value, not a world figure), the GDELT news tone and political turmoil share (until anchors are set from live readings), the GDELT protest share, UCDP conflict events (they mix minor incidents with major wars) and the Epoch count of notable AI models (a count of models says nothing about good or bad). Social unrest and Technology & AI each have one scored input, with provisional anchors (marked so in the table above) that will be replaced from live history after the first runs. Social unrest is scored from the IMF Reported Social Unrest Index: each month the world value is the percent of reporting countries whose index is above 100, a share rather than the median because about half of all country-months are exactly 0. Technology & AI is scored from the average tone (GDELT V2Tone) of articles in Wired, The Verge, Ars Technica and MIT Technology Review, sampled from GDELT bulk files across the week; it is the tone of coverage in those four outlets, not a measure of technology itself, and a week with fewer than 30 articles gives no reading. When an input has no data its signal shows as unscored, takes no part in the index, and the other weights are renormalised over the live signals. It is not shown as 0.

What the ends of each signal mean:

Every input, with the weight of its signal, its polarity, its baseline and how old it may be before it is left out. Rendered from config/scoring.json, config/sources.json and config/signals.json.

Signal (weight)InputUseWeightCadencePolarityBaselineFreshness
Public mood (12.5%)US consumer sentiment (University of Michigan, US only)Scored1monthlyHigher is betterFixed anchors: 100 at 110, 0 at 30 (straight line)Stale after 75 days
Public mood (12.5%)EU27 consumer confidence indicator (European Commission survey via Eurostat, EU27 only)Scored1monthlyHigher is betterFixed anchors: 100 at 10, 0 at -35 (straight line)Stale after 90 days
Public mood (12.5%)News tone (average tone of world news coverage, GDELT): context until anchors are set from live readingsContext onlyNoneweeklyHigher is betterContext only: not scoredStale after 21 days
Public mood (12.5%)Gallup Global Emotions (annual), used only to calibrateContext onlyNoneannualHigher is betterNo live source yet (not scored)
Social unrest (12.5%)IMF Reported Social Unrest Index: share of countries above 100 (monthly, provisional anchors)Scored1monthlyHigher is worseFixed anchors: 100 at 0, 0 at 53 (straight line)Stale after 130 days
Social unrest (12.5%)Protest share of world news coverage (GDELT): context, no anchors yetContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 21 days
Cost of living (12.5%)FAO Food Price IndexScored1monthlyHigher is worseFixed anchors: 100 at 90, 0 at 170 (straight line)Stale after 75 days
Cost of living (12.5%)US consumer price inflation, year on year (US only)Scored1monthlyHigher is worseFixed anchors: 100 at 0, 0 at 15 (straight line)Stale after 75 days
Cost of living (12.5%)World consumer price inflation (World Bank, annual)Scored0.5annualHigher is worseFixed anchors: 100 at 2, 0 at 15 (straight line)Stale after 900 days
Cost of living (12.5%)World consumer price inflation (IMF World Economic Outlook, annual, completed years)Scored0.5annualHigher is worseFixed anchors: 100 at 2, 0 at 20 (straight line)Stale after 900 days
Cost of living (12.5%)Real wages (ILO)Context onlyNoneannualHigher is betterNo live source yet (not scored)
Cost of living (12.5%)Disposable income, where availableContext onlyNoneannualHigher is betterNo live source yet (not scored)
Work and jobs (12.5%)US initial jobless claims (US only)Scored1weeklyHigher is worseFixed anchors: 100 at 150,000, 0 at 600,000 (straight line)Stale after 21 days
Work and jobs (12.5%)World unemployment rate (World Bank, ILO estimate)Scored0.5annualHigher is worseFixed anchors: 100 at 3, 0 at 10 (straight line)Stale after 900 days
Work and jobs (12.5%)EU27 unemployment rate (Eurostat, seasonally adjusted, EU27 only)Scored1monthlyHigher is worseFixed anchors: 100 at 5.5, 0 at 12 (straight line)Stale after 90 days
Work and jobs (12.5%)Top 1% wealth share (WID)Context onlyNoneannualHigher is worseNo live source yet (not scored)
Work and jobs (12.5%)Gini coefficient (World Bank)Context onlyNoneannualHigher is worseContext only: not scoredStale after 1500 days
Political disruption (12.5%)Global Economic Policy Uncertainty Index (a proxy for political disruption)Scored1monthlyHigher is worseFixed anchors: 100 at 80, 0 at 450 (straight line)Stale after 150 days
Political disruption (12.5%)World Uncertainty Index (text-based, partly independent of GEPU)Scored1quarterlyHigher is worseFixed anchors: 100 at 5,000, 0 at 150,000 (log scale)Stale after 200 days
Political disruption (12.5%)Geopolitical Risk index, global (Caldara and Iacoviello, provisional anchors)Scored1monthlyHigher is worseFixed anchors: 100 at 50, 0 at 300 (straight line)Stale after 75 days
Political disruption (12.5%)Political turmoil share of world news coverage (GDELT): context until anchors are set from live readingsContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 21 days
Political disruption (12.5%)US daily economic policy uncertainty (FRED USEPUINDXD): context, not scoredContext onlyNonedailyHigher is worseContext only: not scoredStale after 14 days
Political disruption (12.5%)St. Louis Fed Financial Stress Index (FRED STLFSI4): context, not scoredContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 21 days
Political disruption (12.5%)Geopolitical risk for the United States (Caldara and Iacoviello): context, not scoredContext onlyNonemonthlyHigher is worseContext only: not scoredStale after 75 days
Political disruption (12.5%)Share of news coverage on US elections and politics (GDELT): context, not scoredContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 21 days
Political disruption (12.5%)Tone of news coverage on US elections and politics (GDELT): context, not scoredContext onlyNoneweeklyHigher is betterContext only: not scoredStale after 21 days
Conflict (12.5%)Conflict fatalities (ACLED weekly, or UCDP)Scored1weeklyHigher is worseFixed anchors: 100 at 100, 0 at 25,000 (log scale)Stale after 60 days
Conflict (12.5%)Geopolitical Risk index, global (Caldara and Iacoviello, provisional anchors)Scored1monthlyHigher is worseFixed anchors: 100 at 50, 0 at 300 (straight line)Stale after 75 days
Conflict (12.5%)Conflict events (UCDP): context, mixes minor incidents with major warsContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 60 days
Conflict (12.5%)Share of news coverage on Israel, Gaza and Lebanon (GDELT): context, not scoredContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 21 days
Conflict (12.5%)Tone of news coverage on Israel, Gaza and Lebanon (GDELT): context, not scoredContext onlyNoneweeklyHigher is betterContext only: not scoredStale after 21 days
Conflict (12.5%)Share of news coverage on Russia and Ukraine (GDELT): context, not scoredContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 21 days
Conflict (12.5%)Tone of news coverage on Russia and Ukraine (GDELT): context, not scoredContext onlyNoneweeklyHigher is betterContext only: not scoredStale after 21 days
Conflict (12.5%)Geopolitical risk for Israel (Caldara and Iacoviello): context, not scoredContext onlyNonemonthlyHigher is worseContext only: not scoredStale after 75 days
Conflict (12.5%)Geopolitical risk for Russia (Caldara and Iacoviello): context, not scoredContext onlyNonemonthlyHigher is worseContext only: not scoredStale after 75 days
Conflict (12.5%)Geopolitical risk for Ukraine (Caldara and Iacoviello): context, not scoredContext onlyNonemonthlyHigher is worseContext only: not scoredStale after 75 days
Conflict (12.5%)Daily Geopolitical Risk index, 30-day average (Caldara and Iacoviello, preliminary): context, not scoredContext onlyNonedailyHigher is worseContext only: not scoredStale after 21 days
Technology & AI (12.5%)Tone of technology coverage in Wired, The Verge, Ars Technica and MIT Technology Review (GDELT, provisional anchors)Scored1weeklyHigher is betterFixed anchors: 100 at 4, 0 at -4 (straight line)Stale after 21 days
Technology & AI (12.5%)AI capability and breakthroughs (Epoch notable models): context, a count of models says nothing about good or badContext onlyNoneweeklyHigher is betterContext only: not scoredStale after 45 days
Technology & AI (12.5%)Annual AI energy estimates (IEA)Context onlyNoneannualHigher is worseNo live source yet (not scored)
Technology & AI (12.5%)Share of news coverage on China and technology (GDELT): context, not scoredContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 21 days
Technology & AI (12.5%)Tone of news coverage on China and technology (GDELT): context, not scoredContext onlyNoneweeklyHigher is betterContext only: not scoredStale after 21 days
Technology & AI (12.5%)Geopolitical risk for China (Caldara and Iacoviello): context, not scoredContext onlyNonemonthlyHigher is worseContext only: not scoredStale after 75 days
Climate (12.5%)Atmospheric CO₂Scored1monthlyHigher is worseFixed anchors: 100 at 280, 0 at 560 (straight line)Stale after 21 days
Climate (12.5%)Global surface temperature anomaly, land and ocean (NOAA NCEI, 1901-2000 baseline)Scored1monthlyHigher is worseFixed anchors: 100 at 0.5, 0 at 2.5 °C above the pre-industrial level (straight line)Stale after 75 days
Climate (12.5%)Global surface temperature anomaly, land and ocean (NASA GISTEMP, 1951-1980 baseline)Scored1monthlyHigher is worseFixed anchors: 100 at 0.5, 0 at 2.5 °C above the pre-industrial level (straight line)Stale after 75 days
Climate (12.5%)Sea surface temperature anomalyContext onlyNoneweeklyHigher is worseContext only: not scoredStale after 10 days

Weight is the input's share of the mean inside its signal: every input counts 1, except annual inputs, which count 0.5 because they move once a year. It applies from anchor set version 4; weeks stored earlier keep the equal weights they were scored with.

Retired inputs

These inputs were listed as scored but never had a data source, so they never entered a score. From anchor set version 4 they are no longer expected, and the confidence of their signals counts only the inputs that exist. Weeks stored earlier still list them.

SignalInputCadencePolarityRetired in anchor setReason
Public moodWeekly sentiment from news and social mediaweeklyHigher is better4No source ever fed it.
Work and jobsJob displacement and layoffsweeklyHigher is worse4No source ever fed it.
Political disruptionEffective Number of Issues (Shannon entropy across political topics)weeklyHigher is worse4No source ever fed it.
Political disruptionAgenda volatility (KL divergence week on week)weeklyHigher is worse4No source ever fed it.
Political disruptionAI governance developmentsevent-drivenHigher is better4No source ever fed it.
ClimateOil and gas consumptionmonthlyHigher is worse4No source ever fed it.

Every input declares whether a higher value is better or worse for the index, so that the score always reads higher is better.

Momentum is the week-on-week change plus a 4-week trend, labelled Improving, Stable or Deteriorating. A trend within ±1 index points is Stable. Change is shown with a marker (▲, ▼ or ▬) and a word, never by colour alone.

Confidence is Low / Moderate / High. It is the share of expected inputs that are fresh and sourced: Moderate from 0.5, High from 0.8.

Solar activity is a natural cycle, but some researchers have asked whether it shapes human behaviour at scale. The evidence is contested, and the eight signals are not scored on it. It is covered here so readers can weigh it for themselves.

Some historical and scientific writers have noted correlations between the 11-year solar cycle and waves of revolution, social protest and armed conflict. We have not found a peer-reviewed study that confirms this, so we treat it as a hypothesis, not a finding.

The Chizhevsky Hypothesis: In the early 20th century the biophysicist Alexander Chizhevsky argued that mass mobilisation, revolutions and wars peak alongside the 11- and 22-year solar activity cycles. [1]

Variations in solar output and geomagnetic conditions have been linked in early studies to regional weather and crop yields [2][3], which, if the link holds, could add to resource scarcity, economic stress and existing political tension.

Observational studies also report associations between geomagnetic storms and cardiovascular events [5][6] and, in one study, higher cortisol [4]. A 2025 scoping review found that 8 of 36 studies showed no effect [7], and the causal pathways are not yet established [8].

Sources

  1. [1] Chizhevsky, A. (2018). From “Mass Movements and Short Periods of Solar Activity,” in The Earth in the Sun’s Embrace. In Russian Cosmism, MIT Press, pp. 41-54. doi:10.7551/mitpress/11378.003.0005
  2. [2] Bollinger, C. J. (1934). Relation of Oklahoma weather and crop yield to the eleven-year sunspot cycle. Bulletin of the American Meteorological Society. doi:10.1175/1520-0477-15.1.28
  3. [3] Bhalme, H. N. and Jadhav, S. K. (1984). The double (Hale) sunspot cycle and floods and droughts in India. Weather. doi:10.1002/j.1477-8696.1984.tb05478.x
  4. [4] Breus, T. K., Boiko, E. R. and Zenchenko, T. A. (2015). Magnetic storms and variations in hormone levels among residents of North Polar area, Svalbard. Life Sciences in Space Research 4, 17-21. doi:10.1016/j.lssr.2014.12.001
  5. [5] He, P. et al. (2025). Potential influence of geomagnetic activity on blood pressure statistical fluctuations at mid-magnetic latitudes. Communications Medicine 5, 143. doi:10.1038/s43856-025-00822-w
  6. [6] Gaisenok, O., Gaisenok, D. and Bogachev, S. (2025). The influence of geomagnetic storms on the risks of developing myocardial infarction, acute coronary syndrome, and stroke: systematic review and meta-analysis. Journal of Medical Physics 50(1), 8-13. doi:10.4103/jmp.jmp_122_24
  7. [7] Belenko, J., Cancel, G. and Mayrovitz, H. N. (2025). Exploring the potential observations between geomagnetic activity and cardiovascular events: a scoping review. Cureus 17(12), e99851. doi:10.7759/cureus.99851
  8. [8] Maghrabi, A. H. and Maghrabi, M. A. (2026). The sun-earth-health connection: a short review of potential mechanisms and implications. International Journal of Biometeorology 70(2), 61. doi:10.1007/s00484-026-03134-3

The Outlook chart on the home page draws the Entropy Index for every week so far. Once 8 weeks of history exist, it adds a dashed line 4 weeks ahead. The rule is an ordinary least-squares straight line fitted to the index values of the last 8 weeks (week number against index), extended 4 weeks and held between 0 and 100.

Dashed line: a straight-line projection of the last 8 weeks of the index, 4 weeks ahead. A projection of the trend, not a prediction of events. It uses no information beyond past index values: no events, no signal-level models. With fewer than 8 weeks of history no line is drawn and the page says how many weeks exist.

SourceRoleCadenceAttribution
FAO Food Price IndexscoredmonthlyFood and Agriculture Organization of the United Nations (FAO), Food Price Index. FAO terms were not read in the source check; confirm before publishing derived values.
World Bank global consumer price inflationscoredannualThe World Bank: Inflation, consumer prices (annual %), World Development Indicators (world aggregate). World Bank open data, CC BY 4.0.
World Bank Gini indexcontextannualThe World Bank: Gini index (SI.POV.GINI), World Development Indicators (world aggregate). World Bank open data, CC BY 4.0.
NOAA Mauna Loa weekly CO2scoredweeklyNOAA Global Monitoring Laboratory, Mauna Loa atmospheric CO2 (gml.noaa.gov/ccgg/trends/). Made freely available to the public by NOAA; citation requested. Recent values are preliminary.
Climate Reanalyzer OISST world sea-surface temperaturecontextdailyNOAA OISST v2.1 via Climate Reanalyzer, Climate Change Institute, University of Maine (climatereanalyzer.org). Licence not confirmed in the source check; data derive from NOAA OISST v2.1.
UCDP GED candidate events (events per week)contextweeklyUppsala Conflict Data Program (UCDP), Georeferenced Event Dataset candidate events, ucdp.uu.se, CC BY 4.0. The file name carries the release version (v26_01_26_09 = January to September 2026): update the URL on each release.
UCDP GED candidate events (deaths per week)scoredweeklyUppsala Conflict Data Program (UCDP), Georeferenced Event Dataset candidate events, ucdp.uu.se, CC BY 4.0. Shares the download of ucdp-events.
Epoch AI notable AI models (models per week)contextweeklyEpoch AI, Notable AI Models dataset (epoch.ai/data). Licence not confirmed in the source check; confirm before publishing.
Wikimedia most viewed pagescontextmonthlyWikimedia Foundation, pageviews API (wikimedia.org). Wikimedia REST API; attribution and a descriptive User-Agent are expected. Social layer: context only, never scored.
Mastodon trending hashtagscontextdailymastodon.social public trends API. Per-instance terms of service were not checked. Social layer: context only, never scored.
Global Economic Policy Uncertainty Index (FRED GEPUCURRENT)scoredmonthlyBaker, Bloom and Davis (2016), Measuring Economic Policy Uncertainty, via FRED, Federal Reserve Bank of St. Louis. FRED: public domain, citation requested
World Uncertainty Index, GDP-weighted average (FRED WUIGLOBALWEIGHTAVG)scoredquarterlyAhir, Bloom and Furceri, The World Uncertainty Index, via FRED, Federal Reserve Bank of St. Louis. FRED: public domain, citation requested. The index is by Ahir, Bloom and Furceri; check their terms before publishing derived values.
US consumer price inflation, year on year (FRED CPIAUCSL)scoredmonthlyUS Bureau of Labor Statistics, Consumer Price Index for All Urban Consumers (CPIAUCSL), via FRED, Federal Reserve Bank of St. Louis. FRED: public domain, citation requested. US only, not global. The year on year change is computed from the index.
US initial jobless claims (FRED ICSA)scoredweeklyUS Employment and Training Administration, Initial Claims (ICSA), via FRED. Public domain, citation requested
US consumer sentiment (FRED UMCSENT)scoredmonthlyUniversity of Michigan, Surveys of Consumers, via FRED, Federal Reserve Bank of St. Louis. FRED: citation requested. US consumers only, not global.
World Bank world unemployment ratescoredannualThe World Bank: Unemployment, total (% of total labor force) (modeled ILO estimate), SL.UEM.TOTL.ZS, World Development Indicators (world aggregate). World Bank open data, CC BY 4.0.
NASA GISTEMP global surface temperature anomaly (land and ocean)scoredmonthlyNASA Goddard Institute for Space Studies (GISS), GISTEMP v4, global land-ocean temperature index (data.giss.nasa.gov/gistemp/). NASA GISS data are released for public use; the GISTEMP terms and baseline were not read in the source check, confirm before publishing derived values.
NOAA NCEI global surface temperature anomaly (land and ocean, 1901-2000 baseline)scoredmonthlyNOAA National Centers for Environmental Information (NCEI), Climate at a Glance: Global Time Series. NOAA NCEI climate data are public; the Climate at a Glance terms were not read in the source check, confirm before publishing derived values.
IMF Reported Social Unrest Index, share of countries above 100 (HEADLINE)scoredmonthlyBarrett, Appendino, Nguyen and de Leon Miranda (2022), Measuring social unrest using media reports, Journal of Development Economics 158, 102924. IMF Western Hemisphere Department, Reported Social Unrest Index (api.imf.org). © International Monetary Fund. All rights reserved; terms at imf.org/external/terms.htm. Confirm they permit derived values before publishing.
EU27 consumer confidence indicator (Eurostat ei_bsco_m, BS-CSMCI)scoredmonthlyEurostat, Consumer confidence indicator (ei_bsco_m, BS-CSMCI), European Commission business and consumer surveys. Eurostat data are free to reuse with acknowledgement of the source (Commission Decision 2011/833/EU). EU27 only, not global.
EU27 unemployment rate (Eurostat une_rt_m)scoredmonthlyEurostat, Unemployment by sex and age, monthly data (une_rt_m), EU27 from 2020. Eurostat data are free to reuse with acknowledgement of the source (Commission Decision 2011/833/EU). EU27 only, not global.
IMF world consumer price inflation (DataMapper PCPIPCH, WEOWORLD)scoredannualInternational Monetary Fund, World Economic Outlook, Inflation, average consumer prices (PCPIPCH), world aggregate, via the DataMapper API. © International Monetary Fund. Terms at imf.org/external/terms.htm.
Geopolitical Risk index, global (Caldara and Iacoviello), scored on ConflictscoredmonthlyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Confirm terms before publishing derived values. Shares the download of gpr-global.
Geopolitical Risk index, global (Caldara and Iacoviello), scored on Political disruptionscoredmonthlyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Confirm terms before publishing derived values. Shares the download of gpr-global.
Tone of technology coverage (GDELT GKG: wired.com, theverge.com, arstechnica.com, technologyreview.com)scoredweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee
GDELT protest coverage sharecontextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee
GDELT news tonecontextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee
GDELT political turmoil coverage sharecontextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee
US daily economic policy uncertainty (FRED USEPUINDXD)contextdailyBaker, Bloom and Davis (2016), Measuring Economic Policy Uncertainty, Economic Policy Uncertainty Index for United States (USEPUINDXD), via FRED, Federal Reserve Bank of St. Louis. FRED series USEPUINDXD: confirm series terms before publishing derived values. FRED hosts the series; the index is by Baker, Bloom and Davis. Context only, never scored.
St. Louis Fed Financial Stress Index (FRED STLFSI4)contextweeklyFederal Reserve Bank of St. Louis, St. Louis Fed Financial Stress Index (STLFSI4), retrieved from FRED. Series STLFSI4, Federal Reserve Bank of St. Louis: confirm series terms before publishing derived values. Context only, never scored.
Share of news coverage on US elections and politics (GDELT GKG theme and location filter)contextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee. Context only, never scored. The GKG theme and country names in the filter were not verified against the GDELT codebook.
Tone of news coverage on US elections and politics (GDELT GKG theme and location filter)contextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee. Shares the download of gdelt-theme-us-politics-share. Context only, never scored.
Share of news coverage on Israel, Gaza and Lebanon (GDELT GKG theme and location filter)contextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee. Shares the download of gdelt-theme-us-politics-share. Context only, never scored. The GKG theme and country names in the filter were not verified against the GDELT codebook.
Tone of news coverage on Israel, Gaza and Lebanon (GDELT GKG theme and location filter)contextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee. Shares the download of gdelt-theme-us-politics-share. Context only, never scored.
Share of news coverage on Russia and Ukraine (GDELT GKG theme and location filter)contextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee. Shares the download of gdelt-theme-us-politics-share. Context only, never scored. The GKG theme and country names in the filter were not verified against the GDELT codebook.
Tone of news coverage on Russia and Ukraine (GDELT GKG theme and location filter)contextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee. Shares the download of gdelt-theme-us-politics-share. Context only, never scored.
Share of news coverage on China and technology (GDELT GKG theme and location filter)contextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee. Shares the download of gdelt-theme-us-politics-share. Context only, never scored. The GKG theme and country names in the filter were not verified against the GDELT codebook.
Tone of news coverage on China and technology (GDELT GKG theme and location filter)contextweeklya citation to the GDELT Project and a link to this website (gdeltproject.org). GDELT: available for unlimited and unrestricted use for any academic, commercial, or governmental use of any kind without fee. Shares the download of gdelt-theme-us-politics-share. Context only, never scored.
Geopolitical Risk index, global (Caldara and Iacoviello)contextmonthlyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Context only, never scored; confirm terms before using any value in a scored signal.
Geopolitical risk for Israel (Caldara and Iacoviello)contextmonthlyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Context only, never scored; confirm terms before using any value in a scored signal.
Geopolitical risk for Russia (Caldara and Iacoviello)contextmonthlyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Context only, never scored; confirm terms before using any value in a scored signal.
Geopolitical risk for Ukraine (Caldara and Iacoviello)contextmonthlyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Context only, never scored; confirm terms before using any value in a scored signal.
Geopolitical risk for the United States (Caldara and Iacoviello)contextmonthlyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Context only, never scored; confirm terms before using any value in a scored signal.
Geopolitical risk for China (Caldara and Iacoviello)contextmonthlyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Context only, never scored; confirm terms before using any value in a scored signal.
Daily Geopolitical Risk index, 30-day average (Caldara and Iacoviello, preliminary)contextdailyCite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading. No licence stated on the page; the authors ask for citation and the download date. Context only, never scored; confirm terms before using any value in a scored signal.

Geopolitical Risk (context only)

The Geopolitical Risk (GPR) index is shown on the Conflict, Political disruption and Technology & AI pages. It is context only and does not change any score, except the global index, which is also scored on Conflict and on Political disruption with provisional anchors (the index is normalised so that its 1985–2019 mean is 100; 50 scores 100 and 300 scores 0). Cite as: Caldara, Dario and Matteo Iacoviello (2022), 'Measuring Geopolitical Risk,' American Economic Review, April, 112(4), pp.1194-1225. Data downloaded from https://www.matteoiacoviello.com/gpr.htm on the download date given with each reading.

When the anchors change, earlier weeks are scored again on the current anchor set, so the change on the previous week, the momentum, the biggest movers and the Outlook chart compare like with like. Each earlier week is rescored from its saved raw inputs. A source added after an earlier week was stored is filled in for that week from its later download, using only readings dated on or before that week, so a new source never shows up as a change. A week with no saved raw inputs keeps the score it was stored with. The stored snapshot files themselves are never rewritten, and each still records the anchor set that first scored it.

Date (UTC)Change
2026-10-31First entry. Eight signals, Entropy Index bands and equal weights as configured in config/scoring.json.
2026-11-02Added the Outlook chart and its least-squares projection over the last 8 weeks.
2026-11-05Replaced percentile scoring with fixed anchors (anchor set version 1) for scored inputs.
2026-11-06Added a second source for two signals, introduced in anchor set version 2: the World Uncertainty Index (Political disruption) and US consumer price inflation (Cost of living).
2026-11-07Added a scored input for Social unrest (the IMF Reported Social Unrest Index) and for Technology & AI (tone of coverage in four technology outlets), introduced in anchor set version 3 with provisional anchors.
2026-11-09Anchor set version 4. Added scored inputs: EU27 consumer confidence (Public mood), World Bank and IMF world inflation (Cost of living), EU27 unemployment (Work and jobs), the global Geopolitical Risk index (Conflict and Political disruption) and NASA GISTEMP temperature (Climate). Scored inputs now carry a weight inside their signal: annual inputs count 0.5, the rest 1. Retired six scored inputs that never had a source: weekly sentiment, layoffs, Effective Number of Issues, agenda volatility, AI governance and oil and gas consumption.
2026-11-09Anchor set version 5. Social unrest (IMF unrest index): best 0, worst 53, no longer provisional, because the 1986-2026 median of the world share is 26.3%, so a typical month now scores 50 (the latest month, 24.43, scores 54; it scored 81 before). Technology & AI (tone of coverage): best +4, worst -4, still provisional, so a tone of 0 scores 50. Climate: both temperature series now run from +0.5 °C (best) to +2.5 °C (worst) above pre-industrial, so 1.5 °C scores 50, and atmospheric CO2 from 280 ppm (pre-industrial) to 560 ppm (double it). Conflict: UCDP deaths are scored on the mean of the last 8 complete Monday-to-Sunday weeks, with log anchors of 100 and 25,000 deaths a week, to stop one quiet or bloody week swinging the dial. A tone that rounds to zero is now shown as 0.00, never -0.00. Earlier weeks are no longer compared on the scores they were stored with: see "How earlier weeks are compared".
2026-11-09Earlier weeks are filled in from sources added since they were stored (readings dated on or before that week only), so the weekly change no longer counts a new source as movement.