All the numbers

Dot: best estimate. Line: 90% range. Tap a row for its sources.

Risk · per

How this works

What a micromort is

A micromort is a one-in-a-million chance of dying. Ronald Howard proposed the unit in the 1970s to compare small risks, and David Spiegelhalter made it popular. Driving 1,000 km in Britain carries about one micromort, as does about seven hours of life for an average 30-year-old American.

How each number is computed

micromorts = 10⁶ × deaths ÷ exposures

Deaths and exposures come from the same population and period, in a source quoted word for word on the row. Where a source publishes a rate, the conversion is shown. Where no count exists, as for a tourist's day on a Bali scooter or a soldier's day in Ukraine, the row is marked as a low-confidence estimate and lists its assumptions.

The 90% range

Each number has a 90% range with two parts, combined on the logit scale so that it stays below certain death:

  • Counting noise is the exact Poisson interval on the number of fatal events. An airliner crash with 260 deaths counts as one event.
  • Model doubt is a factor k, with 90% of the probability within ×/÷k, for uncertain denominators, definitions or applicability. It is lopsided when the direction of the bias is known.
What counts as a death

The numbers cover acute risk inside a stated window, such as during the jump, within 90 days of surgery or during the expedition. Long-term effects are left out: the 311 micromorts for donating a kidney cover the 90 days after the operation and leave out later kidney failure. Each row also says whether it counts deaths caused by the activity or all deaths in the window.

Units
eventa jump, a flight, an operation, a shark bite
durationan hour of climbing, a working day
distance1,000 km by car
perioda year as US president
backgrounda day of being 30

Rows with mismatched denominators stay in the data and out of the comparisons.

The yardstick: days of being 30

An average US 30-year-old faces 3.48 micromorts a day, from the 2024 US life table. Most of that risk is external causes, and it is high for a rich country, so a European 30-year-old's day carries less. Above 1 in 100, the page gives "1 in N" instead of days.

Whose risk this is

Each number describes the people who did the activity, such as licensed skydivers, screened kidney donors or trained marathon runners. Each row describes its population, and your own risk may be higher or lower.

Sources and checks

Almost every row comes from a primary source: a national statistics office, an accident registry, a sport's own fatality reports or a peer-reviewed study. A few well-documented historical counts, such as the Space Shuttle's two losses and the presidents who were assassinated, cite Wikipedia, and the roulette odds follow from the game itself. A row is accepted only if its deaths and exposures appear word for word in the quoted text, from the same population and period.

A validator then recomputes every number and its range from those inputs and rejects any row whose inputs are missing from its quotes. The rows on the scale and their comparisons were reviewed again against their sources. That review caught real errors, such as a roulette citation that described 5-in-6 odds and a presidency figure that counted natural deaths. Software tools helped gather sources and draft rows; the validator and the review check their output. Please report any remaining errors in the comments.

Prior art

The walk up the scale is borrowed from Neal Agarwal's Size of Life. Earlier micromort lists include Wikipedia, micromorts.rip, David Veksler's Actual Risk Dashboard and Blastland & Spiegelhalter's The Norm Chronicles. This page adds a computation from quoted sources, a range and published data for every number.

The data

157 risks, each computed from a quoted primary source, with ranges and caveats. Free to use under CC BY 4.0.

Comments

Corrections and better sources are welcome.