All Collections
Demographic indicators
Modelling the number of missing deaths
Modelling the number of missing deaths

Anticipate the number of deaths at time t

Written by Anthony Cabos
Updated over a week ago


The number of missing deaths is the number of deaths that are not yet enumerated at a time t.

Attention, so back datas provided by INSEE are susceptible to modification in time.


There is a temporal asynchrony between the date of death of a person and the moment when the death is enumerated. However, it is possible to estimate the number of additional deaths to come for the day studied. This asynchrony may be related to several factors:

  • The time taken to register the death, it can be done electronically or by post (longer)

  • Deceased French citizens abroad are registered with a delay

  • The case of missing persons (who are declared dead after a certain time or who may be declared dead but found later)

  • The judicial cases


One way to estimate so-called "missing" deaths is to refer to the previous time period. Our data are obtained monthly via the insee.
We refer to the previous months' data to estimate the number of deaths that were counted in addition.


  • After the January census, we observe this:

  • As of 31 December of the previous year, 99% of deaths were not reported

  • As of 30 December of the previous year, 95% of deaths were unreported

  • As of 29 December of the previous year, 90% of deaths were unreported

  • ...

  • On 1 December of the previous year, 5% of deaths were not reported

  • So the operation carried out in January is as follows:

  • As of 31 January, 99% are considered to be undeclared

  • As of 30 January, 95% are considered undeclared

  • As of 29 January, 90% of deaths are not reported

  • ...

  • On 1 January, 5% of deaths go unreported

This is done over 180 days, after this period the number of deaths may change by less than 0.1%.
For the last month with deaths, the data is aggregated and the so-called "missing" deaths are distributed evenly across all days of the current month.
If the modelling indicates a lower number than the insee publication, then the number published by the insee will be used as a guide.


The March deliverable was published by insee. In February, 20% more deaths were counted.
The modelling in March will perform the following calculation:
nbDecesModel = nbDeces published in March / (1 - 0.2)
March has 31 days, so the modelling will consider that there are nbDecesModel/31 deaths per day.
For February and previous months, the same operation is performed without aggregating to the month. Thus, February 29th will be compared to January 31st, compared to the previous deliverable 98% of the deaths have been published. Thus the model will revalue the number of deaths by 2% for February 29.



Use & Analytics Advanced subscription

Did this answer your question?