Methodology

1. A death toll is not an attribution

For every entry we separate two claims, because they meet different evidentiary standards:

• The death toll — how many died. Sourced to peer-reviewed research, UN/WHO bodies, truth commissions, demographic databases, or established historians.
• The attribution to capitalism — why those deaths belong here. This is the argument, and it is where critics push. We state it explicitly on every card and flag where it is contested.

We count the conservative end of every range. If we can defend the low number, the high number takes care of itself.

2. What "caused by capitalism" means

The frame is the economic system and its imperatives — not "the United States." Five categories:

1. Structural violence — death from a system that rations food and healthcare by ability to pay, in a world that produces enough of both.
2. Corporate & industrial violence — profit-driven negligence and unsafe labour (workplace deaths, Bhopal, the opioid epidemic).
3. Colonialism & empire — extraction, forced labour, policy famine, and wars of colonial conquest.
4. Settler colonialism — displacement and elimination to seize land (Indigenous North America, Aboriginal Australia, Palestine).
5. Wars & interventions — for resources and market access, to project hegemony, and to crush socialist/communist alternatives.

A note on settler colonialism: for the Americas and Australia we count only the violence- and policy-attributable deaths (massacres, forced removal, frontier killing). The far larger population collapse was overwhelmingly introduced disease — much of it pre-industrial and pre-capitalist — and is deliberately excluded. For Israel/Palestine we cite UN OCHA, B'Tselem and peer-reviewed (Lancet) figures, date every number, and present both the settler-colonial framing and its main counter-framing.

3. How the live counter works

The count is never stored. The browser recomputes it every frame:

total = Σ(fixed historical tolls)
      + Σ(annualDeaths / secondsPerYear) × secondsSinceThatEntry'sAnchor

secondsPerYear = 365.25 × 86400. Every visitor sees the same monotonically increasing number, fully reproducible from data.json. No backend.

Ongoing structural counters anchor at 1990-01-01 — the year Global Burden of Disease data begins. Using current (lower) annual rates across the whole 1990–present span undercounts the higher-mortality early years — malnutrition deaths were ~7.2M/year in 1990 versus ~2.9M now — so the structural figure is deliberately conservative. Pre-1990 chronic structural deaths are not reliably reconstructable and are not counted. Fixed historical tolls carry their own dates and are counted once.

4. The structural figures (93% of the number)

The bulk of the headline is two annual figures plus a correction. This is the most scrutinised part of the site, so it is the most heavily caveated.

Hunger — child & maternal malnutrition

2,937,804 deaths/year (95% UI 2,489,636–3,512,073), GBD 2019, all ages. We use the GBD 2019 figure because it is the most recent one published with an exact, quotable uncertainty interval: GBD 2021 reports this risk only as DALYs and rates, not a free absolute death count, and the "~2.4M" figure circulating from GBD 2021/2023 is the narrower under-5 subset with no published interval. This is a child-malnutrition measure, not all-age "starvation," and it is not the unsourced "9 million/year" advocacy number, which traces to UN/NGO speeches with no underlying dataset.

Healthcare — amenable mortality

8,600,000 deaths/year (UI 8.5–8.8M) across 137 low- and middle-income countries (Lancet 2018), split into ~5.0M from poor-quality care and ~3.6M from non-utilisation of care for treatable conditions. It is based on 2016 data and has no newer re-estimate with this breakdown. The 5.0M poor-quality-care figure is solid; the 3.6M non-utilisation figure is an inference the authors themselves flag as their weakest estimate.

The overlap correction (−1.25M/year)

The two figures overlap: a malnourished child who dies of a treatable disease is counted by both frameworks, and no peer-reviewed study partitions them. Rather than publish an inflated sum, we subtract a transparent, conservative −1.25M/year (estimated overlap 1.0–1.5M). The Lancet 8.6M also already absorbs most vaccine-preventable, TB, malaria, HIV and WASH deaths, so those are not added on top. This deduction is our own adjustment, not a published figure, and it is shown as its own line on the counter.

The same non-double-counting discipline governs two other entries: the Yemen War counts only its ~150,000 direct conflict deaths (its ~227,000 famine and disease deaths are already in the structural counters), and workplace deaths count only the ILO's ~330,000 fatal accidents (the ~2.6M occupational-disease deaths overlap the healthcare and air-pollution figures). In each case we count the non-overlapping subset, not the larger headline number.

5. World Wars: WWI in, WWII out

WWI is counted only for deaths after US entry (April 1917) — a large minority (~35–45%) of military deaths, because the inter-imperialist war's continuation is the attributable part and US entry has a real (if contested) capital-interest argument: J.P. Morgan's Allied loans and a war trade tripled toward the Allies while trade with the Central Powers was blockaded to near zero. Mainstream historians cite unrestricted submarine warfare and the Zimmermann Telegram as the primary triggers, so we flag the causation as a documented minority thesis. The 1918 influenza pandemic and the Armenian genocide (pre-entry) are excluded.

WWII is excluded. The US entered because it was attacked; there is no capital-interest entry argument. And ~85–90% of Holocaust deaths occurred after December 1941, so a "post-US-entry" subset would place the ~6M Holocaust dead — genocide driven by Nazi racial ideology — into capitalism's column. That is indefensible under any honest framing. It remains in data.json with enabled:false so the decision is transparent.

6. Wars & interventions, and attribution tiers

Interventions vary enormously in how directly capitalism caused the deaths. We count the toll but mark the attribution strength, because lumping them together is exactly what an honest critic would attack:

• Direct — the system's military inflicted the deaths (e.g. the strategic bombing of Korea; the Iraq War).
• Enabled — a coup or regime change whose subsequent regime did the killing (e.g. Guatemala 1954, Iran 1953).
• Contributory — proxy or sponsored conflicts where the deaths were largely inflicted by others.

Some proxy wars to crush socialism are included per the framework but carry the weakest attribution, and their cards say so:

• Soviet–Afghan War (1979–89) — most deaths inflicted by Soviet/PDPA forces; the US-funded insurgency is contributory and the motive was mainly geopolitical.
• Greek Civil War — genuinely multi-sided; both sides killed.
• Angola — most deaths fell after Western backing ended, and an oil paradox cuts against a clean "war for capital" story.

If you disagree with any of these, change one line in data.json. The argument is only as honest as its weakest counted entry — so we mark them.

7. What we deliberately exclude

Transparency about what we leave out is what makes the rest credible. These were considered and rejected:

• WWII — the Holocaust-attribution problem above.
• The Iraq sanctions "~500,000 children" figure — traced by peer-reviewed work (Dyson & Cetorelli, BMJ Global Health 2017) to mortality data fabricated by Saddam's government. Sanctions as economic warfare is a sound attribution; that specific number is not, so we will not publish it.
• The "~140M famines since 1870" shortcut — it bundles the Maoist Great Leap Forward famine (15–55M) and Soviet famines, which are communist-caused — the opposite of the thesis — and it double-counts the colonial famines we already list individually. Historical hunger deaths come only from the specific, individually-sourced colonial entries.
• The full Colombian conflict toll (~450k) — ~72% was inflicted by leftist guerrillas and the rest is hard to cleanly apportion; counting insurgent killings as capitalism's deaths is backwards. The 1928 United Fruit Banana Massacre stays in.

8. Documented, not counted: sanctions, fossil fuels, poverty, carceral deaths

Sanctions are real economic warfare, and one peer-reviewed figure is defensible: ~564,000 deaths/year globally from US/EU unilateral sanctions (Rodríguez, Rendón & Weisbrot, Lancet Global Health 2025). But those deaths are the malnutrition and untreated-disease deaths already inside the structural counters, so a separate sanctions line would double-count the same bodies. We feature it as a documented mechanism rather than add it. The per-country figures in circulation — Iraq's "500,000 children," Venezuela's "40,000" — are discredited (fabricated or uncontrolled data) and are not used at all.

Fossil-fuel air pollution kills an estimated ~5.1 million a year (Lelieveld et al., BMJ 2023; full range 1.05M–8.7M). It is large, real, and central to the thesis — but it is a risk-factor attribution that overlaps the healthcare counter heavily: a pollution-caused heart attack that is also treatable is counted by both, and GBD risk-factor tallies are non-additive by construction. Counting it on top would double-count those deaths, so we document it here rather than sum it. We also deliberately avoid the much-cited 8.7M figure (Vohra et al. 2021) as a headline — its confidence interval includes negative values and its exposure-response model is the least validated of the available estimates.

Poverty-attributable deaths (~183,000/year in the US, Brady et al., JAMA Internal Medicine 2023) are documented for the same reason: poverty is the upstream cause that kills through hunger and lack of healthcare — the very deaths already in the structural counters. It is an all-cause attributable-fraction estimate, not a separate bucket, so counting it would double-count. No defensible global figure exists.

Carceral-system deaths — US police killings (~1,250/year, undercounted by ~55% in official records per the Lancet 2021), distinct in-custody deaths (~1,270/year, BJS), and executions (~1,518/year worldwide plus thousands in China, Amnesty) — are real and partly non-overlapping. We document rather than count them because the attribution to capitalism is the weakest on the site: a structural argument (the prison-industrial complex, the criminalization of poverty) rather than a traceable causal chain, and the usable data is US-only. We hold that causal claim to the same standard as every other entry — and it doesn't clear the bar for the counted total.

Commercial products. Tobacco (~7–8M/yr) and alcohol (~2.6M/yr) have strong, even court-adjudicated industry-fraud attribution and low overlap with the healthcare counter — but we leave them documented rather than counted, by choice, to avoid the "people choose to smoke/drink" dispute (the opioid epidemic, with its guilty pleas, is the one commercial case we do count). Ultra-processed food is documented: weak firm-level attribution and heavy overlap.

Also documented, not counted: climate change beyond air pollution (only a 2030–50 projection exists, and it overlaps hunger/health); indentured labour (strong attribution, but no historian publishes an aggregate toll); IMF/austerity deaths (real but reported only as rates, and they operate through the structural counters); and the 1971 Bangladesh genocide (US complicity is documented, but the motive was Cold War realpolitik, not economic — so it fails the attribution test).

9. The communism comparison

The number most often raised as capitalism's moral counterweight is the ~94 million ("~100 million") total from The Black Book of Communism. We cite it — and caveat it the same way we caveat everything else. It is heavily contested: two of the book's own lead historians (Nicolas Werth, Jean-Louis Margolin) publicly rejected the total as inflated, and more than half of it is famine deaths counted as intentional killings.

Note the asymmetry. This site excludes famines caused by communist states and uses conservative figures throughout; the Black Book includes famine deaths and rounds up. Held to the same standard, capitalism's tally is several times larger.

10. Confidence ratings

Each card shows high medium low confidence, reflecting the quality of the death-toll sourcing — not how strongly anyone feels about the politics. Low generally means the toll is a model or back-calculation, or the attribution is contested.

11. How to audit or change a figure

Everything lives in one open file — view the raw data (data.json). Each counted entry keeps at least one reputable source. To exclude something, set enabled:false and add an exclusionReason. To add a category properly, source a defensible study and add a new entry. Figures are sourced to peer-reviewed research, UN/WHO bodies, truth commissions, demographic databases, and established historians — no advocacy blogs, and the conservative end of contested ranges.

This is an argument built on citations. Check them.