The Saphir database is a detailed register of individuals, organised first by municipality and then by IRIS from 1990 onwards. Each record in the database corresponds to an individual described in terms of their socio-demographic characteristics, those of the household to which they belong, and the main characteristics of their main residence. It contains a selection of information – most of which has been harmonised – extracted from supplementary analyses of the population censuses from 1968 to 1999 and from subsequent censuses comparable to the most recent census published.

Total

The concepts used in Saphir are those from population censuses, harmonised over a long period where possible.

Individual

Population residing in mainland France for the 1968, 1975 and 1982 censuses; in mainland France and the overseas departments (excluding Mayotte) for subsequent censuses.

The geographical reference area varies depending on the census year in question: mainland France for the 1968, 1975 and 1982 censuses, and France excluding Mayotte for subsequent censuses

The four historic overseas departments are included in Saphir from the 1990 census onwards.

Certain concepts and definitions have evolved over time. Consequently, classifications, geographical boundaries and variables specific to each census do not always have the same content. Therefore, although Saphir aims for the greatest possible harmonisation, there are still some non-harmonised variables within this database.

The administrative boundaries used in Saphir are therefore not standardised (with the exception of the DCRX and DCLTXX variables). In other words, they reflect the administrative boundaries in force at the time of the census. Municipal mergers and demergers that took place between censuses are therefore not accounted for by Saphir. Great care should therefore be taken when using the data, particularly when the level of analysis is detailed (municipal level) or if the area under consideration is a customised zone constructed from a list of municipalities.
For information, the reference date for the administrative boundaries is:
• 1 March 1968 for the 1968 Census
• 20 February 1975 for the 1975 Census
• 4 March 1982 for the 1982 Census
• 5 March 1990 for the 1990 Census
• 8 March 1999 for the 1999 Census
• 1 January 2009 for the 2007 Census
• 1 January 2014 for the 2012 Census
• 1 January 2019 for the 2017 Census
• 1 January 2025 for the 2023 Census

 The previous place of residence corresponds to ‘time lags’ that vary from one census to the next, which affects comparisons. Up to the 1999 census, the previous place of residence asked of respondents was that as at 1 January of the year of the previous census (for the 1968 census, this was therefore 1 January 1962, for the 1975 census it was 1 January 1968, for the 1982 census 1 January 1975, for the 1990 census 1 January 1982 and for the 1999 census 1 January 1990). For the 2007 census, it is the place of residence on 1 January five years earlier. For the 2012, 2017 and 2023 census, it is the place of residence on 1 January one year earlier.

 Over time, the classifications relating to economic activity and occupation have evolved, in some cases significantly and in ways that are difficult to harmonise at a detailed level. For this reason, the two variables ‘economic activity’ and ‘occupation’ are included in Saphir in a non-harmonised form. They correspond exactly to those found in the original data files.

 Some variables are reported in a standardised manner, but their scope varies from one census to the next. This is the case for variables relating to employment. Indeed, only people aged 15 and over (who have reached that age) are included in the censuses of 1968, 1982, 1990, 1999, 2007, 2012, 2017 and 2023. However, for the 1975 census, the age threshold is 17 years and over (having reached that age). The variables concerned are as follows:

• AEX            -    PRX
• CSX            -    STX
• DPX            -    TAX
• NES4X

Identification of the unemployed
Two variables are used to identify the unemployed: these are the ‘type of activity’ (TAX) and ‘employment status’ (STX) variables. Due to the way their respective categories are defined, the count of the number of unemployed may differ depending on whether one or the other of these variables is used (the difference for mainland France ranges from 12 to 295 individuals, depending on the census). However, it is advisable to give preference to the ‘type of activity’ variable (TAX).

Weighting
Weights vary between censuses depending on the sampling rates in force. There are exceptions to this general principle; here are a few examples:

For 1968 and 1975: the data for Corsica are derived from a 1/20th-scale statistical analysis. It is advisable to treat the results from these censuses with great caution. Certain irregularities were noted in these censuses. The targets for a 1/4th-scale statistical analysis could not be met for Corsica.
For 1982: the results for the municipalities of Corsica should be used with caution; warnings were found in publications from that period. However, the quality is superior to that of the 1968 and 1975 censuses, and the results are derived from a 1/4-scale statistical analysis.
For 1999: certain geographical areas were subject to exhaustive processing. These include, in particular, the two departments of Corsica, a few municipalities (mainly in departments 44 and 49) and certain sub-municipal areas (small Sensitive Urban Areas). The weighting for individuals in these geographical areas is 1.
For the Censuses since 2006: due to the revised methodology (sampling plan and data collection spread over 5 years), the weights vary considerably.
The case of the overseas departments (DOMs): the four overseas departments – Guadeloupe, Martinique, French Guiana and Réunion – were covered in full by the 1990 and 1999 censuses (with individuals assigned a weight of 1).

Changes to the method of calculating the reference person
From the 2016 census onwards, the method for calculating the reference person has changed. It no longer takes gender into account. The criteria taken into account are employment status, whether the person has a spouse, whether they have a child, and age.