How the Ledger counts

The Demise Ledger looks like a live feed. It is not one, and the distinction is the single most important thing on this page. No mortality surveillance1 system on earth reports deaths in real time. Certification2, coding3 and aggregation4 take months to years, and global datasets typically run two to four years behind the present. Anything claiming to show you a live global death count is showing you arithmetic.

This is that arithmetic, written down.

The core model

Every counter starts from one number: an annual global death count for a cause, rounded from published estimates. That figure is divided by the number of seconds in a year to give a rate per second, and the counter advances at that rate.

The Ledger's all-cause6 baseline is 61,500,000 deaths a year, a round figure in the recent range. It works out to 1.9489 deaths per second, the figure shown as the Demise Rate. That baseline is discussed in detail in Why 1.95 deaths per second?, including why it sits where it does in a range that recent years have pushed around considerably.

Why the main counter stutters

The headline counter does not tick smoothly twice a second. Deaths are added in irregular bursts, because they are treated as a Poisson process7, the standard statistical treatment for independent events occurring at a constant average rate.

In practice the site draws each gap between deaths from an exponential distribution8, using the inverse-transform method9: with a random number U greater than 0 and less than 1, the wait until the next death is −ln(U) ÷ λ, where ln is the natural logarithm10 and λ is the rate per second. Over a long visit the average settles on the underlying rate. Over a short one it clusters and gaps, the way real independent events do.

This is a deliberate choice over a smooth counter. A metronome would imply that mortality is evenly distributed in time, which is exactly the wrong intuition. The category counters beneath it do accumulate smoothly, because they are showing a running total rather than individual arrivals.

Today and this year

The Deaths Globally Today and This Year counters are computed directly from your device's clock: elapsed seconds since local midnight, or since 1 January, multiplied by the rate. They are recalculated continuously rather than accumulated, which means they stay correct if you leave the tab in the background, and they reset when the date rolls over.

Because they use your local midnight, two people in different time zones will see different "today" figures at the same moment. That is intentional. "Today" is a local idea.

The rates in full

Every figure the dashboard uses is listed below, with what it works out to per second.

Pathological11 causes

Accidental causes

These twelve categories total about 48.4 million deaths a year against an all-cause baseline of 61.5 million. They are not meant to sum to the total: they are a selection, and the remaining roughly 13 million deaths a year fall under causes the first two columns do not break out, including maternal17 and neonatal18 conditions, nutritional deficiencies19, self-harm, interpersonal violence20 and the large remainder of ill-defined21 causes that every mortality dataset carries.

The unlisted toll

Four of those omitted lines now have their own counters. They are a further selection, not a partition of the 13 million, and they are not added to the 48.4 million. Where a line has no published annual total that can be cited cleanly, the dashboard explains it and does not tick it. The essay is The unlisted toll.

Cirrhosis, drug use disorders, nutritional deficiencies and ill-defined causes are named on the left-hand panel and do not tick. Collective violence and natural disasters are named on the right-hand panel and do not tick. A war year and a disaster year are not a smooth rate.

Bizarre and rare

These twelve use the printed annual figure below. Several are estimates, national series, or folklore. This Year follows the printed figure. The accelerated counter multiplies it by 160. The second line on a card is a scope or a real recent count, not a made-up last year. The sorting is in How rare is a rare death?.

The age distribution

The age bars distribute the since-arrival total across five age bands22 using fixed approximate shares: 8.3% at ages 0–4, 1.7% at 5–14, 10% at 15–49, 30% at 50–69, and 50% at 70 and over. These are rounded approximations of the global pattern, not a live breakdown, and they do not vary by cause. That rounding flatters nobody, since a road death and a dementia death obviously do not share an age profile.

The bars are drawn in proportion to the largest band, so the 70+ bar fills the track and every other bar is scaled against it. The percentage printed next to each label is the actual share. Where the deaths fall sets those shares beside a published age series, and says where the rounding is close and where it is loose.

What is dramatised

Four parts of the site are theatre. They are listed here rather than buried, because a site that cites the WHO in its footer owes you a clear boundary.

The Accelerated ×160 counters

In the Bizarre & Rare section, the counters labelled Accelerated ×160 run at one hundred and sixty times the real rate. At true rates these events are so rare that nothing would ever visibly happen: lightning would register roughly once every two hours and twelve minutes, and spontaneous combustion, already not a real cause of death, once a year. The multiplier exists so the section does something during a visit.

The This Year figure on each of those cards uses the printed annual figure, with no multiplier. Where the two disagree by a factor of 160, that is why. On a folklore card, that printed figure is not a measurement.

The second line on a rare card

The rare cards no longer print a decorative "Last Year". The second line is either the scope of the figure (a country, a floor, an estimate) or a real recent count of the same thing. On the aircraft card the tick is the five-year mean, about 200, and the second line is 394 onboard deaths in 2025. Those two numbers are supposed to disagree. Do not cite the coconut line, the hippo line, the selfie line, or spontaneous human combustion as a measured total.

The Grim Reaper Telemetry Ticker

The terminal feed writes invented lines, such as "a heart stops in Hamburg, Germany", about every six seconds. Each line's cause is drawn in proportion to the annual figures on the twelve disease and accident cards. Heart disease appears often. Electrocution almost never does. The rare cards are not narrated here, and neither are neonatal deaths, maternal deaths, self-harm or homicide. No line in that feed is a real death, a real person or a real event. The city names are set dressing. A ceremonial note is written at 100, at 500 and at 1,000 departures since the page was opened. The count beside the cursor is that same session total. It is a mood device, and it is the part of the site most likely to be mistaken for a data feed, which is why it is called out here.

And the calculator

"How Will You Demise?" applies rough weights to a handful of tick-boxes and picks a weighted random outcome. The "Ledger confidence" percentage it reports is that outcome's share of the weight pool, an internal artefact23 of the toy, not a probability that anything will happen to you. It is not a risk assessment, it is not a health service, and it is not medical advice.

Sources

The baseline rates derive from the patterns published in:

• WHO Global Health Estimates, the World Health Organization's cause-of-death series, the standard reference for global mortality by cause.
• Global Burden of Disease, the Institute for Health Metrics and Evaluation's study at the University of Washington. It produces the most detailed cause, age and country breakdowns available.
• UN World Population Prospects, for total deaths and crude death rates24, which underpin the all-cause baseline.

The country atlas

The Atlas of Lethal Claims is a 2023 extract, not a sketch. Cause shares are the WHO Global Health Estimates for 2023, all ages and both sexes, rounded half up to one decimal place. A cause is shown when that share is at least 0.1%, and at most five causes are shown in each column. The candidate list is fixed, and cancers are one group in every country, so a name means the same cause everywhere. Life expectancy and the crude death rate are the UN World Population Prospects for 2023, via the World Bank, rounded the same way.

Drug overdose is kept with illnesses, under drug use disorders, because that is where this WHO file puts it. Firearm deaths are inside self-harm or interpersonal violence, by intent. The atlas lists every country in that WHO file. The places with no card are explained in Countries with no card. The awkward terms are defined in the glossary.

Known limitations

• The rates are static. Mortality shifts year to year, and sharply during a pandemic. The Ledger holds a single set of annual figures and does not update itself.
• Category boundaries are fuzzy. Someone with diabetes who dies of a heart attack can be coded either way depending on the certifying practice. Grouped categories like "Cardiovascular" hide a lot of that judgement.
• Coverage is uneven. Many countries lack complete civil registration25, so global totals rest partly on modelled26 estimates rather than counted deaths, which means the precision implied by a ticking counter is not real precision.
• The rare figures are the weakest. Several have no defensible source at all; see How rare is a rare death?.

Corrections

If a figure here is wrong, please send the correction with a source. Confirmed corrections are applied and noted on this page. Nothing on this site is worth defending against better evidence.

Last reviewed: 5 October 2026