Data Files Explanation

 

The Human Mortality Database (HMD) includes the following types of data:

  • Live birth counts,
  • Death counts,
  • Population size on January 1st,
  • Population exposed to risk of death (period & cohort),
  • Death rates (period & cohort), and
  • Life tables (period & cohort).

Format of the data

  • All HMD data files are tab-delimited text (ASCII) files.
  • Files are organized by sex, age, and time.
  • Population size is given for one-year1 and five-year2 age groups.3
  • Deaths are given in the following formats:3
    • Lexis triangles (i.e., by age1, birth cohort, and calendar year)
    • 1x1 (i.e., by age1 and calendar year)
    • 5x1 (i.e., by 5-year2 age group and calendar year).
  • Exposure-to-risk, death rates, and life tables are given in similar formats of age and time. In all cases, period data are indexed by year of death, whereas cohort data (if available) are indexed by year of birth:
    • 1x1 (i.e., by age1 and year)
    • 1x5 (i.e., by age1 and 5-year time interval)
    • 1x10 (i.e., by age1 and 10-year time interval)
    • 5x1 (i.e., by 5-year2 age group and year)
    • 5x5 (i.e., by 5-year2 age group and 5-year time interval)
    • 5x10 (i.e., by 5-year2 age group and 10-year time interval).

1One-year age groups (or "by age") means 0, 1, 2,..., 109, 110+.
2Five-year age groups means 0, 1-4, 5-9, 10-14,..., 105-109, 110+. Age groups are defined in terms of completed age, so "5-9" extends from exact age 5 to just before the 10th birthday (sometimes written elsewhere --as "5-10").
3Some of these numbers are estimates (of population size or numbers of deaths), not actual counts, and therefore may be expressed as non-integers.


The following columns are included in each life table:

Year Year or range of years (for both period & cohort data)
Age Age group for n-year interval from exact age x to just before exact age x+n, where n=1, 4, 5, or ∞ (open age interval)
l(x) Number of survivors at exact age x, assuming
l(0) = 100,000
d(x) Number of deaths between exact ages x and x+n
q(x) Probability of death between exact ages x and x+n
L(x) Number of person-years lived between exact ages x and x+n
T(x) Number of person-years remaining after exact age x
e(x) Life expectancy at exact age x (in years)

See the Methods Protocol (pg. 35-44) for more details about life table calculations.

Important Notes:

  • Missing data are denoted by a single dot ("."). Currently, there are two situations in which data are unavailable. First, cohort exposure and death rates are missing for ages attained outside the period covered by the HMD (see below for more details). Second, the national statistical organization may not publish death counts for certain years because of war or other reasons (e.g., Belgium 1914-18). When death counts are missing, all quantities that depend on those data (i.e., population estimates, exposure-to-risk, death rates, period and cohort life tables) are also missing.

  • Deaths, population estimates, death rates, and life tables are provided by single years of age up to 109, with an open age interval for 110+. However, these data are sometimes the product of aggregate raw data (e.g., 5-year age groups, open age intervals), which have been split into single years of age using the methods described in the Methods Protocol. The original raw data that were extracted from published or unpublished sources are available from the HMD Input Database.

  • For populations with territorial changes, two sets of population estimates are given for years in which a territorial change occurred. The first set of estimates (identified as year "19xx -") refers to the population just before the territorial change, while the second set (identified as year "19xx +") refers to the population just after the change. For example, in France, the data for "1914 -" cover the previous territory (i.e., as of December 31, 1913), while the data for "1914 +" reflect the territorial boundaries as of January 1, 1914.

  • Cohort death rates (i.e., by year of birth) are provided if there are at least 30 consecutive calendar years of data for that cohort. For example, the mortality series for Sweden begins in 1751, therefore we can show death rates for the 1675 birth cohort for ages 76 and older. The cohort death rates at younger ages are shown as missing (denoted by "."). Similarly, if the mortality series ends in 2002, we can show death rates for the 1972 cohort up to age 29 because by December 31, 2002, everyone in that cohort has reached exact age 30. Yet, mortality data for age 30 will remain incomplete until December 31, 2003.

  • Cohort life tables are presented for a population if there is at least one cohort observed from birth until extinction (i.e., the date by which all cohort members are assumed to have died). In that case, life tables are provided for all extinct cohorts and for some almost-extinct cohorts as well (see the Methods Protocol, pg. 42-44).

  • For period life tables, the central death rate m(x) is used to compute probabilities of death q(x). Although not given here, the values of mx below age 80 are by definition equal to the observed population death rate M(x) shown on each country page. At older ages, however, the number of deaths and the exposure-to-risk eventually become quite small, and thus observed death rates display considerable random variation. Therefore, we smooth the M(x) values for ages 80 and older and use these smoothed values to compute q(x) above a certain age (based on the number of observed deaths). For details, see the Methods Protocol (pp. 35-37). This procedure helps to avoid certain difficulties in period life table calculations at older ages that may be caused by: 1) extremely high death rates resulting from exposure being smaller than the number of deaths, 2) death rates of zero resulting from no deaths at an age where exposure is non-zero, and 3) undefined death rates at all ages where exposure is zero. For cohort life table calculations, such difficulties are not present.

Raw Data

The raw data and an explanation of the format of those data can be found on the Input Database page.