Methodology: How the Death Calculator Model Works

This page documents, in full, how death-calculator.com turns a date of birth, sex, country, and a handful of lifestyle answers into a life-expectancy estimate, a median age at death, a death-clock countdown, and a set of survival percentiles. For the interactive version with your own numbers, use the calculator.

The mortality model: Gompertz–Makeham

The calculator is built on the Gompertz–Makeham law of mortality, a well-established actuarial model combining an age-independent hazard with an age-dependent hazard that rises exponentially with age:

h(x) = A + B · exp(C · (x − 30))

Here A is the Makeham term — a constant, age-independent hazard — and B · exp(C · (x − 30)) is the Gompertz term, which grows exponentially as age increases. The shape parameters are fixed, and a single overall hazard scale is solved numerically in your browser so that the model's implied life expectancy at birth matches your selected country's documented baseline. No part of this calibration happens on a server; it runs client-side, using bisection, every time you submit the form.

From hazard to survival function

The hazard function is integrated into a survival function, S(t) = exp(−∫₀ᵗ h(u) du), using fixed-step numerical integration. Every number the calculator shows you — life expectancy, median age at death, percentiles, survival probabilities — is derived from this one survival function, so the figures are internally consistent with each other rather than computed by separate, unrelated formulas.

• Life expectancy is computed as ∫₀^∞ S(t) dt, the area under the survival curve — the mean remaining lifetime.
• Median age at death is found by inverting S numerically via binary search to locate the age at which survival probability drops to 50%.
• Percentiles (P10, P25, P50, P75, P90) are found the same way at different survival thresholds. Because they come from one monotonic survival function, their ordering is guaranteed: P10 ≤ P25 ≤ P50 ≤ P75 ≤ P90.
• Survival probabilities to a specific target age are read directly off S at that age, conditioned on having already reached your current age.

Country and sex calibration

Each country in the model's dataset carries a male and female life expectancy at birth from a named, dated source (see data sources). The model is calibrated separately for each sex using that country's figures. Where a source does not publish sex-specific figures, a documented fallback — the mean of male and female — is used, and this substitution is disclosed in the result rather than silently applied.

Lifestyle, diet, sleep and social adjustments

Smoking, alcohol, physical activity, diet quality, sleep duration, and social connection are applied as multiplicative adjustments to the calibrated hazard scale, each one reflecting a population-level association reported in published research rather than an arbitrary weighting:

• Smoking: current smoker ×1.45; former smoker ×1.12; never = reference.
• Alcohol: 0–14 drinks/week ×0.97; above 14 drinks/week, the multiplier increases with the amount in excess of that threshold.
• Physical activity: low ×1.10; moderate = reference; high ×0.93.
• Diet quality: poor ×1.10; average = reference; good ×0.95; excellent ×0.90.
• Sleep duration: short ×1.08; normal = reference; long ×1.04.
• Social connection: low ×1.08; moderate = reference; high ×0.94.

BMI is calculated and displayed for information only. It is deliberately excluded from the hazard calculation, because BMI's individual-level predictive value for mortality is limited, and using it as a hard multiplier would imply a level of individual precision the evidence does not support.

What the model does not do

It does not know your genetics, medical history, access to healthcare, environmental exposures, or the possibility of accidents and rare events — all of which materially affect individual lifespan and none of which a population mortality model can observe. It does not predict a date or time of death for any individual. It is a statistical tool for population-level estimation, not a diagnostic or actuarial underwriting tool, and should not be used for medical, financial, insurance, or legal decisions. See the full disclaimer.

Versioning

The model and its underlying data are both versioned independently, and both version numbers are shown on the results page and in the site footer, so a change to either is visible rather than silent. See the editorial policy for how updates and corrections are handled.