Map
Info
Articles
Support Us

Methodology

A Gini coefficient is usually between 0 and 1. Zero would mean every person held the same wealth. One would mean one person held all of it. Negative wealth can put the result outside that range. The calculation does not force it back. The number describes a distribution. It is not, by itself, a verdict.

What the number is for

Wealth that someone created is not a debt owed to everyone who did not create it. Inequality that comes from building something can be large and still not be an injustice. This page is here so the argument can be about the arithmetic. The counts below are from the dataset generated at 2026-11-21T07:14:22+00:00.

Country wealth

Net personal wealth for each country is the World Inequality Database series mhweal_p0p100_999_i, converted with WID's kl-kl-conversion to current US dollars of the data year at market exchange rates (currency us, exchange x). Purchasing-power parity is a different WID conversion, and it is not the one used for this total. The manifest records the basis as: market exchange rate (WID kl-kl-conversion exchange=x), current USD of the data year.

W_c is the all-ages WID net wealth total (mhweal_p0p100_999_i; a wealth aggregate is the same whichever population definition you use) while the people counts use adults (npopul_p0p100_992_i), matching the adult equal-split shares.

Total net wealth is WID mhweal_p0p100_999_i at market exchange rates. WID returned no data for mhweal_p0p100_992_i or mhweal_p0p100_992_j. No adult total was available to compare. The published total stays mhweal_p0p100_999_i.

The pipeline keeps, for each country, the latest WID year that has net wealth, adult population, and a usable set of wealth shares. On this build, 214 countries use 2024. That is 214 countries in the published file.

Wealth groups and adult population

Wealth shares are WID's adult equal-split shweal series (992_j). The fine groups are p0p20, p20p40, p40p60, p60p80, p80p90, p90p99 and p99p100. The totals year is the latest WID year at or before 2024 with net wealth and the adult population. If that year is missing any of those seven groups, the latest earlier year that has all seven supplies the shares, and those shares are applied to the totals year's wealth and adult population. The country file records both years. Only when no year has the seven groups does the country use p80p100 instead. That figure is a lower bound. On this build, 214 countries use the seven fine groups and 0 use the coarse split. On this build every country has the seven fine groups. None uses the coarse p80p100 split. One country uses wealth shares from an earlier year than the totals: India (shares 2022, totals 2024).

Those share requests use WID's PPP exchange code (exchange=p), because they are proportions. The dollar total they are applied to is the market-rate net wealth above.

For the Gini calculation, population is the adult series npopul_p0p100_992_i. Each group's population is its percentile width times that adult count: 0.2, 0.2, 0.2, 0.2, 0.1, 0.09 and 0.01, or 0.2 for p80p100. The published population is adults (npopul_p0p100_992_i), not the all-ages count (npopul_p0p100_999_i). The median ratio of adults to all ages on this build is 0.7023. The manifest lists the variables as: shweal_{p0p20,p20p40,p40p60,p60p80,p80p90,p90p99,p99p100,p80p100}_992_j, mhweal_p0p100_999_i (kl-kl-conversion, currency=us, exchange=x), npopul_p0p100_992_i (adults), npopul_p0p100_999_i (all ages, old-formula comparison only).

The number shown on the map is the all-ages World Bank total (WDI SP.POP.TOTL, latest year, no API key), published as populationTotal. If World Bank has no figure for a country, the pipeline falls back to WID npopul_p0p100_999_i.

Billionaires, and which country they count toward

Billionaires are the Forbes real-time billionaires list (forbes.com/forbesapi/person/rtb (real-time billionaires)). Current wealth is Forbes finalWorth times one million: US dollars at market value.

A billionaire counts toward their country of residence. When residence is unknown, they count toward citizenship. The field is countryIso2, and countryBasis says which rule applied. On this build that is 2,750 by residence and 584 by citizenship fallback, out of 3,334 billionaires.

The subtraction from the top group uses a pinned Forbes snapshot, not the live list. Current wealth still moves with the live list. A person who is not in the pinned snapshot has base wealth zero. The pinned base snapshot is dated 2026-11-20T20:30:14+00:00. On this build 3,334 billionaires are in that snapshot and 0 are not.

How the Gini is computed

The browser uses the same formula as the pipeline.

  • Divide each group's share by the sum of the shares, so the shares add to 1. If the raw sum is more than 0.02 away from 1, the country is named in the manifest. On this build no country's raw shares are that far from 1.
  • Group wealth is the normalised share times total net wealth. A negative share stays negative. That group wealth is not reset to zero.
  • Billionaires counted in the country are taken out of the top group. The top group is p99p100, or p80p100 on the fallback. Their base-snapshot wealth is subtracted, and one adult is subtracted for each of them. If the top group's wealth would become negative, it is set to zero and the country is flagged. On this build that floor binds for one country: Monaco.
  • Each billionaire is then their own group: one person, at their current Forbes wealth.
  • Total wealth is the sum of these group wealths. When the top-group floor does not bind, that equals WID net wealth minus base-snapshot billionaire wealth plus current billionaire wealth.
  • Sort the groups by average wealth, poorest first. The Lorenz curve starts at (0, 0). The Gini is one minus the sum, along that curve, of each step's population share times the sum of the wealth shares at the two ends of the step. When the totals are positive, the curve ends at (1, 1).
  • If total wealth is not positive, or the people left are not positive, that country has no Gini.

Net debt in the bottom 20% (student loans, mortgages) pushes the measured wealth Gini up, which is why Scandinavian welfare states such as Sweden score high on wealth Gini. On this build Sweden's bottom 20% share of net wealth is -11.15% and its wealth Gini is 0.947. Negative group wealth stays in the calculation.

The result is not forced back into 0 to 1. On this build every published Gini is inside 0 to 1.

People inside a WID group are treated as equals. Inequality inside the group is left out, so this Gini is a lower bound on the Gini of the individuals. A Pareto tail for the top group is planned. It is not applied.

The map still draws five bars. Those bars add the fine groups back into fifths of the adult population and are labelled aggregated for display. The Gini uses the fine groups, not the bars.

Moving someone

Moving a billionaire removes that one-person group from the country they leave and adds it to the country they join. Total wealth and the adult count of each country change by that person's current wealth and by one person. The WID groups stay as they were, including the top group after the base-snapshot subtraction. Editing someone's current wealth changes only that one-person group. Both Ginis are computed again.

Photographs

A portrait is shown only when photos.json lists a Wikimedia Commons file under a free licence that allows commercial use and derivatives (CC BY, including the version Commons reports, such as CC BY-SA 4.0, plus CC0, public domain, and the short name Attribution). The image is the original file on upload.wikimedia.org, scaled in the page with CSS object-fit: contain. The card credits the author, the licence and its version, and links the Commons file page. This build accepts 618 such photographs. Everyone else is drawn as initials from their name. No photograph is loaded from Google, LinkedIn, Forbes, or the old image server.

Refresh

Forbes is fetched on every pipeline run. The base snapshot is not. WID responses are cached on disk and refreshed at most once every 24 hours. If a refresh fails, a cache younger than 168 hours can still be used. A new dataset replaces the published one only when the quality gate passes. The run fails if unmapped Forbes rows exceed 1%, if any Gini is missing or outside 0 to 1.1, if any country's raw shares are more than 0.05 away from 1, if the billionaire count drops by more than 10% versus the previous published run, or if any country's Gini moves by more than 0.05. Below those limits the same checks are warnings in the manifest. The two comparisons with the previous run are skipped on the first publish, when the schema or Gini method changed, and when the pipeline is run with GATE_RESET_BASELINE=1. The run still requires more than 2,500 Forbes records and at least 150 countries, with the wealth groups present and no NaN. A raw share sum more than 0.02 from 1, and still within the fail limit, is a warning, and the shares are still normalised. A Gini above 1 and at or under 1.1 is a warning and is left as calculated.

Sources

  • World Inequality Database — net personal wealth, adult equal-split wealth shares, adult population.
  • Forbes real-time billionaires — the billionaire list, current market-value wealth, and the pinned base snapshot.
  • Wikimedia Commons — the only photographs, under the licence on each file.
  • Natural Earth — country outlines, carried here as TopoJSON (countries.topo.json, with the Natural Earth fields ADMIN, ISO_A2 and ISO_A3). The basemap tiles under those outlines are Mapbox.
Contact Us
PrivacyMethodology
Terms of Service