Date (Creation)

2021-01-01

Citation identifier

gisco-services / https://gisco-services.ec.europa.eu/census/2021/INSPIRE/Data/CENSUS_INS21ES_A_FR_2021_0000

Point of contact

Organisation name	Individual name	Electronic mail address	Role
Insee: National Institute of Statistics and Economic Studies, Demographic and Social Statistics Directorate Demography Department		muriel.barlet@insee.fr	Owner

Keywords

• EUROSTAT metadata

GEMET - INSPIRE themes, version 1.0

• Population distribution — demography

Access constraints

Other restrictions

Other constraints

Public access to spatial data sets and services would adversely affect the confidentiality of the proceedings of public authorities, where such confidentiality is provided for by law.

Access constraints

Other restrictions

Other constraints

No conditions apply to access and use

Use constraints

Other restrictions

Other constraints

No conditions apply to access and use

Spatial representation type

Vector

Distance

1000  metres

Language

English

Topic category

• Society

Begin date

2021-01-01

Reference system identifier

ETRS89-extended / LAEA Europe

Distribution format

Name	Version
Geopackage file(.gpkg)
GML	3.2.1
Text	CSV
Text	SDMX

OnLine resource

Protocol	Linkage	Name
WMTS	https://gisco-services.ec.europa.eu/cmaps/service?REQUEST=GetCapabilities&SERVICE=WMTS	ViewService (WMTS) of the Census 2021 data

OnLine resource

Protocol	Linkage	Name
ATOM	https://gisco-services.ec.europa.eu/census/2021/INSPIRE/Data/FR_PD_3035_CSV.zip	The compressed resource (CSV) file contains data and metadata

OnLine resource

Protocol	Linkage	Name
ATOM	https://gisco-services.ec.europa.eu/census/2021/INSPIRE/Data/FR_PD_3035_GML.zip	The compressed resource (GML) file contains spatial data and metadata

OnLine resource

Protocol	Linkage	Name
ATOM	https://gisco-services.ec.europa.eu/census/2021/INSPIRE/Data/FR_PD_3035_GPKG.zip	The compressed resource (GPKG) file contains spatial data and metadata

OnLine resource

Protocol	Linkage	Name
ATOM	https://gisco-services.ec.europa.eu/census/2021/INSPIRE/Data/FR_PD_3035_SDMX.zip	The compressed resource (SDMX) file contains spatial data and metadata

OnLine resource

Protocol	Linkage	Name
ATOM	https://gisco-services.ec.europa.eu/census/2021/INSPIRE/PD.atom	Downloadservice (ATOM-Feed) of all the various packages

Hierarchy level

Dataset

Statement

The methodology of the population census was renewed in 2004 and resulted in an Insee méthodes «Pour comprendre le recensement de la population» in 2005.

The quality of population estimates in the census also led to an https://www.insee.fr/fr/statistiques/fichier/version-html/4796233/imet136.pdf Insee méthodes in 2020.

The documentation also consists of thematic advice sheets, technical notes and sheets on data processing after collection. They present the information you need to know in order to make the best use of the census results. They are available on insee.fr (cf. 10.7).

Dissemination of the 2021 population geocoded to a 1 km² grid

To provide population data on a 1km² grid, the French INS (Insee) had to face two challenges:

- geocoding of dwellings in municipalities with fewer than 10,000 inhabitants

- cell-level estimates for municipalities with more than 10,000 inhabitants

Moreover, early estimates were conducted to provide Eurostat with 2021 population in December 2022, whereas 2021 population will be disseminated in France in December 2023.

Population geocoding in municipalities of fewer than 10,000 residents

Addresses of dwellings in municipalities of 10,000 residents or more are listed in a geocoded register. This is also the case for communal establishment addresses.

However, in municipalities with fewer than 10,000 inhabitants, addresses of dwellings are not geocoded in the current production process of the Census.

Two geocoding methods are used for these municipalities:

- geocoding by matching on street address: this process assigns coordinates to addresses from the annual census survey by comparing them with an address register (on the basis of the municipality code, the lane and the lane number), in which coordinates are taken from the land register;

- geocoding by matching on individuals characteristics: this process consists in using the characteristics of individuals who reside at a certain address (date of birth, municipality of birth and sex) to match addresses from the annual census survey with those from the geocoded tax files. If the two addresses contain individuals with the same combination of sex, date of birth, and municipality of birth, it is very likely that they are the same so they are matched.

For both geocoding methods, a quality score is calculated, which depends on coordinates quality and matching quality. A selection rule was defined based on the coordinates’ quality score in each method, to determine which coordinates would be selected for each address. If both methods propose bad quality coordinates, they are interpolated using the coordinates of addresses nearby. And in very few cases, the coordinates are randomly imputed in the municipality.

Method for disseminating Census on a grid in municipalities of 10,000 residents and more

The dissemination on a grid is not an issue in two cases, due to exhaustive data:

- municipalities with fewer than 10,000 residents are surveyed exhaustively, in turn, every five years;

- individuals living in a communal establishment or in a tourist residence are also exhaustively enumerated during a five-year cycle.

Moreover, homeless people and people living in mobile homes will not be enumerated on each grid cell, due to the lack of precise geocoding. They will be assigned to a virtual "unallocated" grid cell, at the national level, in accordance with the https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32018R1799&from=EN Regulation (EU) 2018/1799.

On the contrary, municipalities with more than 10,000 inhabitants are surveyed every year, by means of a survey on approximately 8% of their dwellings. Within a 5-year cycle, about 40% of the dwellings of the municipality have been surveyed. As the census of these municipalities is not exhaustive, the challenge is to produce reliable cell-level estimates, despite the low number of sampled dwellings in some grid cells. Indeed, this low level of information in certain grid cells questions the reliability of the usual estimation method at such a fine level.

A hot deck imputation method is used to estimate population in a grid cell in municipalities of 10,000 residents and more. This method consists to impute characteristics of dwellings and individuals to every address in the buildings register from the surveyed addresses and auxiliary information of tax origin, matched beforehand with the buildings register. Imputation is possible because of the high sampling rate (40%) and the existence of a register of localised buildings, listing all the residential addresses of large municipalities.

The aim is that surveyed address keep a weight equal to 1 and allocate the rest of their estimation weight (and the associated characteristics) to non-sampled addresses. To do that, the distance between donor addresses (the addresses surveyed) and recipient addressess is minimised in terms of auxiliary information (the distance is calculated from the average number of persons per dwelling and the structure of the population by sex and by age group), under some constraints (each address must receive its number of dwellings in the buildings register).

More information is available in the Final Methodological Report for Grant Agreement, on methodology for estimating census data on a grid (No. 881993 — 2019-FR-CENS-GEOLOC) and in 2 articles presented at the 2022 edition of the French statistical methods seminar.

• http://jms-insee.fr/jms2022s12_1/ La géolocalisation du recensement de la population dans les communes métropolitaines de moins de 10000 habitants»

• http://jms-insee.fr/jms2022s12_2/ Méthodes de carroyage du recensement de la population dans les communes métropolitaines de 10000 habitants et plus »

Early estimates: method for disseminating 2021 total population one year prior the French Census dissemination

In France, the legal 2021 populations will be disseminated at the end of December 2023. In order to disseminate 2021 total population one year before the French dissemination, early estimates have been conducted.

A census year is created from five annual census surveys for the calculation of legal populations.

In municipalities of 10,000 residents and more, the population is estimated by multiplying the addresses’ survey weights by the ratio of the number of dwellings according to the “median buildings register” (as of January, 1st of the median year of the census cycle) over the number of dwellings estimated during the five years of the cycle.

This is equivalent to multiply the average number of persons per dwelling by the number of dwellings according to the “median buildings register”. To estimate provisional 2021 population were used:

- the number of dwellings as of January, 1st , 2021 is kwown (source: buildings register)

- the average number of persons per dwelling in 2021 is estimated by applying a trend extension to the average number of persons per dwelling in 2019 (last known figure at the time of estimation), based on the trend between 2014 and 2019.

In municipalities with fewer than 10,000 residents, different estimation methods are used according to the survey year of the municipality. The legal 2021 populations, which will be disseminated at the end of December 2023, will use 5 annual census survey: 2018, 2019, 2020, 2022 and 2023 (the 2018 survey is used since there was no survey in 2021, due to sanitary context).

At the end of 2023, the municipal populations as of January, 1st 2021 will be calculated.

- If the municipality was surveyed in 2018, in 2019 or in 2020, the collected data will be extrapolated using the evolution of the number of dwellings observed in tax data combined with an estimate for the variation in the households’ average size.

- If the municipality was surveyed in 2022 or in 2023, a linear interpolation will be made between the last legal population (N-1 census) and the new data collection.

To estimate provisional 2021 population, some adjustments are made:

- for extrapolation, the method applied is the same as usual but tax data used is more recent and less precise;

- a linear interpolation is applied to municipalities surveyed in 2022 from provisional figures;

- for munipalities which will be suveyed in 2023, data from the previous suvey, in 2017, have been extrapolated using the evolution of the number of dwellings observed in tax data combined with an estimate for the variation in the households’ average size.

For individuals living in a communal establishment or in a tourist residence, homeless people and people living in mobile homes, the 2019 census results are used.

More information is available in the Report for Grant Agreement on early estimates (No. 07112.2017.007-2017.442). Data collection period

Collection begins on the third Thursday in January, except in La Réunion and in Mayotte where it begins two weeks later and in March-April in Saint-Pierre-et-Miquelon.

It lasts a little over four weeks in municipalities with fewer than 10,000 inhabitants and in communities, and a little over five weeks in municipalities with 10,000 inhabitants or more.

The other overseas territotycarry out general population censuses every five years by way of derogation (Article 157).

The last full enumeration in Mayotte took place in 2017. Mayotte integrated the annual census surveys in 2021 and will be subject to a joint dissemination to the metropolitan territories and other DOMs from the 2023 census.

Collection mode

• Face to face by interviewer

• By internet

Survey unit

• Dwelling

• Individual

Sampling method

For each annual survey, the municipalities with 10,000 inhabitants or more follow a two-phase sampling plan in the Address Survey Base (ASB) created from the annual data of the Localised buildings register (RIL).

First phase of sampling:

In metropolitan France, the residential buildings (or addressed entities) were divided into five rotation groups by balanced random draw. The drawing of the sample of buildings from a given annual census survey mobilises only one of the five rotation groups, by rotation. The constitution of the buildings rotation groups constitutes the first phase of the draw.

In the French overseas departments, each large municipality was partitioned into several sets of contiguous buildings known as «îlots». In each municipality, the «îlots» were divided into five rotation groups. The constitution of the rotation groups constitutes in this context the first phase of the draw.

Second draw phase:

Addressed entities belonging to the strata of large buildings, new buildings and tourist establishments are exhaustively enumerated. The buildings belonging to the stratum of known small addresses are selected in such a way as to obtain 40% of the dwellings surveyed in the annual rotation group.

Sample size

Over a 5-year cycle, 25 million dwellings and 47 million people are enumerated (for metropolitan France and French overseas departments).

Data collection documents

The individual form covers the characteristics of the individual: sex, age, marital status, place of birth, nationality, place of study, diplomas, place of previous residence (one year before), employment, social category, economic activity sector.

The housing form describes the characteristics of:

• The household: composition, size, family ties, number of cars, etc.;

• The dwelling: type, year of completion, size, type of occupation, public housing, heating, bathroom, sewage disposal, year of moving in, parking.

The forms (individual , individual for communities, housing for metropolitan France and the overseas territories) as well as the explanatory notes are available on the website «le recensement et moi».Each of the ingredients involved in elaborating the results of the census is validated progressively as it is implemented. In addition to these data checks (collected or administrative), different operations verify the processing. For example:

→ In municipalities of 10,000 residents and more, which are surveyed every year by means of a survey, the quality of the survey depends in part on the exhaustivity of the localised buildings register (RIL). This register is used for collection and for population estimates. The RILis updated every year, both by Insee and the municipalities. Moreover, national survey operations measuring the quality of the RIL have been organised on several occasions since 2004 (every two years on average) to check the match between the RIL and the ground. In practice, they have conducted exhaustive "combing" of a certain amount of districts (IRIS), thus ensuring good coverage of the survey address base.

→ In municipalities of fewer than 10,000 residents, the exhaustivity of dwellings is checked on the ground before each survey.

→ To ensure collection quality, all players receive proper training (enumerators, municipal coordinators, supervisors and INSEE personnel). Next, as soon as the questionnaires are available, different indicators are checked in the INSEE offices and a score is assigned to each municipality in the census. If the score is insufficient, INSEE operators conduct checks on the ground. These checks can lead to correcting the questionnaires, in collaboration with the municipalities concerned, thus contributing to improving the collection and verification protocols.

→ The non-response is very low (less than 4%) and methods for dealing with non-response are applied.

→ The quality of data entry is checked each year: a sample of questionnaires (between 5,000 and 6,000) is double-entered, one by the usual service provider and the other by a competitor. The INSEE analyses differences between the two entries for each question and deduces error rates that can be assigned to the main service provider. This double entry of data is conducted in the course of the entry campaign, ultimately enabling correction of protocols for the following batches (all questionnaires are distributed each year into twenty batches).

→ The population estimates are subjected to consistency and likelihood checks, and then assessed (for example, to identify atypical trends).Methods used for data collection and population estimation are designed to ensure the accuracy of the census results. However, there are few elements that can affect this accuracy:

• Rolling census and survey: the french census is based on a 5 years cycle. Each year, there is a census survey concerning 8% of the dwellings in large municipalities (10,000 inhabitants and more) and 100% of the dwellings in 1/5 th of the small municipalities (less than 10,000 inhabitants). During a 5 years cycle, 100% of the inhabitants of small municipalities have been interrogated and 40% of the inhabitants in large municipalities.

• Absence of 2021 survey: in 2021, the census survey was cancelled because of the pandemic. The method to estimate the population has been adapted.

• Non-response treatments

• Quality of geocoding

• Imputation method in municipalities with 10,000 inhabitants or more to estimate population in a 1km² grid

• Early estimates: early estimates have been made in order to disseminate the 2021 total population one year prior the French Census dissemination.

• Integer numbers: integer numbers are disseminated in the grid data, according from European Regulation (EC) No 2018/1799. Some calibration is done in order to obtain the same rounded total population as in the data hypercubes. Moreover, the additivity between the grid cells is preserved: for example, for a given grid cell, the sum of the number of females and the number of males is equal to the population of the grid cell. However, the total of each breakdown (number of males, number of females, number of individuals under 15 years old, etc.) is not the same in grid data and in hypercubes data.

At the national level, Article 6 of Law No. 51-711 of 7 June 1951, as amended, on the obligation, coordination and secrecy of statistics determines what statistical secrecy is, its limits and the conditions for its application.

Generally speaking, as regards access to public data, confidentiality obligations relating to the protection of privacy or business secrecy and the protection of personal data are guaranteed by law (Article 1 of the Law for a Digital Republic).

A Statistical Confidentiality Committee ensures that these statutory guarantees are maintained.

The answers to the annual census survey questionnaire are protected by statistical secrecy and are intended for INSEE. Their use and access are strictly controlled and limited to the preparation of statistics or research work. In particular, surnames and first names are not kept beyond 31 December of the year following the survey.

The General Regulation 2016/679 of 27 April 2016 on data protection (RGPD) and Law No. 78-17 of 6 January 1978 on information technology, files and freedoms apply to this survey.

Enumerated people may exercise a right of access, rectification or limitation of processing for data concerning them during the period of conservation of identification data.

Rules on the dissemination of the results of the population census are set out in the https://www.legifrance.gouv.fr/loda/id/JORFTEXT000000465454/ Order of 19 July 2007 which specifically indicates geographical levels (and size) for distribution, the list of sensitive variables and the possible forms of products (maps, tables, detailed files).

Four variables are defined as sensitives: place of birth, country of citizenship, year of arrival in France, place of usual residence one year prior the census.The issue of confidentiality was applied to the variables "Place of birth" (for people born abroad) and "Place of usual residence one year prior to the census" (for people who have moved from from outside France) as their diffusion is strictly regulated in France by the decree of 19 July 2007 on the dissemination of the results of the population census.

The French INS applied a method developed by itself, called multilevel grids method. This method is a perturbative method (no suppression of confidential values), but in a deterministic way.- Three types of risks are defined:→ Grid cells with less than 11 households are treated.→ Grid cells with at least 80 % of individuals with the same sensitive modality are also treated.→ Moreover, a geographic differencing problems treatment (grid cells versus municipalities) has been done.Risky grid cells are grouped (with any additional grid cells) to form groups of more than 11 households. In each group, the number of people for each variable is broken down in proportion to the population of the grid cell.- Process:→ Step 1: make groups of grid cellsRisky grid cells are gathered with other grid cells to make groups of grid cells above 11 households.→ Step 2: pertubation stepEach subtotal at the group level (for example the number ofmales) is broken down in the grid cells of the group, proportionally to the population of each grid cell.Moreover, grid cells detected to solve geographic differencing issues or risky grid cells because ofthe attribute disclosure are added in a close group to be treated.- Remark: with this method, the additivity between the grid cells is preserved. Moreover, the spatial structures are preserved.

Metadata

File identifier

CENSUS_INS21ES_A_FR_2021_0000 XML

Metadata language

English

Character set

UTF8

Hierarchy level

Dataset

Date stamp

2024-05-31T09:00:00

Metadata author

Organisation name	Individual name	Electronic mail address	Role
Insee: National Institute of Statistics and Economic Studies, Demographic and Social Statistics Directorate Demography Department		muriel.barlet@insee.fr	Point of contact