Notes (France)

(under construction -- remnants here need to be edited)

Adjustments to raw death counts

  1. Changes to Vallin/Mesle Data
    1. 1949 - the split between upper triangle age 99 and lower triangle age 100 was corrected to match the notes of both Hemery and Depoid.
    2. 1958 - the published data were obviously flawed, placing a disproportionate number of deaths in 100 lower triangle and very few in 100 upper triangle. A correction was taken from the notes of Hemery and Depoid.
    3. 1960-1961 and 1963-1965 - published data by single years of age stop at age 89 or 94. Above this age, figures were taken from the notes of Francoise Depoid.
    4. 1962 (males only) – Vallin’s files contained fictitious data for ages 90+, for ages 90-99 a correction was made to match published data
  2. Distributing deaths of unknown age
  3. Data on deaths of unknown age is available for 1903-1955 only. These deaths were distributed proportionately across the age range.

  4. Splitting Deaths into lexis triangles
    1. For the period 1903-1906 the raw data consist of age-period death totals for ages 100 and above. The tendency to have more deaths at younger ages (hence, in the lower triangle) is balanced by the tendency to have proportionately more deaths in the first months of the year (and hence, in the upper triangle). Therefore deaths in age-period format have been split in half.
    2. For years 1949,1954,1956-1967 the raw data consist of period-cohort death totals for ages 100 and above. Triangle ratios for exact cohorts tend to be above 0.5 (at least above age 80) and to increase sharply with age. In this case, the two tendencies mentioned above conspire to produce much higher ratios. Therefore deaths in period-cohort format have been split according to the ratio of 60/40. The rough nature of this approximation is justified by the fact that, in any case, cohort totals will be exact and the effects of errors in the assignment of age at death will tend to cancel one another when aggregating over age and time.
    3. For years 1968-1987, the deaths listed in the lower triangle of age 109 are apparently part of a 109+ category. There are only 2 years (for females) during this period in which there are 2 or more deaths in this category, so any possible correction does not seem particularly important, and has not been made at this time.
  5. Fictitious Data above age 100

For 1899-1902, 1934, 1935, and 1947, we have only the number of deaths above age 100. These deaths are split among ages and triangles arbitrarily.

Raw data, age 100+:


Type of 100+ category *































note: There are three types of 100+ categories:
1 = age 100, upper triangle and above
2 = exact age 100 and above
3 = age 99, upper triangle and above



Estimating January 1 Population

January 1 population estimates below age 80 are taken from INSEE. For cohorts which have reached age 119 by 1995, deaths are cumulated backwards using the method of extinct generations to obtain population estimates for ages 80 and above. For cohorts aged 105 to 118 in 1995, population counts for ages 80 and above are estimated using the survival ratio method. For cohorts aged 95 to 105 in 1996, the intercensal survival method is used, taking the 1996 survival ratio estimates for these cohorts as the ending population, and the 1990 census counts as the beginning population. All other figures are taken from INSEE.

The method of extinct generations is complicated by the fact that there were several territorial changes during World War I and World War II, and the population figures given by INSEE correspond to different numbers of departments in different time periods. In order to use the method of extinct cohorts, we first adjust all deaths to correspond to the maximum number of departments over the entire time period (90 departments). Next we cumulate deaths backwards using the adjusted deaths to get January 1 population estimates. Finally, these estimates are multiplied by an adjustment factor to convert them back to the actual number of departments in each time period.

Note: When two sets of population estimates are given for the same year, the first corresponds to the estimated population before a territorial change, and the second corresponds to the estimated population after a territorial change.