1 August 2001
English and French
Symposium on Global Review of 2000 Round of
Population and Housing Censuses:
Mid-Decade Assessment and Future Prospects
Department of Economic and Social Affairs
United Nations Secretariat
Statement from France *
Guy Desplanques **
1. In France, statistical projects are the subject of systematic consultations. To lead these consultations, France has established a National Council for Statistical Information (CNIS), which provides a forum for producers and users of public statistics and for discussion of all statistical survey projects. Where appropriate, CNIS gives its opinions on the justification for and utility of surveys. The Council also discusses the modalities of dissemination of census results. Another body, the Approvals Committee, issues opinions on the methods employed: it gives its views on whether the methodology used is likely to produce the expected results. Lastly, in compliance with the laws governing the public’s right to privacy, statistical projects are referred to the Commission Nationale Informatique et Libertés (CNIL), which is an integral part of the public statistical system and gives its opinions on questionnaires and on the processing and dissemination of data, while maintaining the confidentiality of the data.
2. The decision to conduct a census is currently made by decree: the proposal is therefore not debated in Parliament nor is there any public debate on it. Consultations on a census take place at different stages: during their preparation, the draft questionnaires are submitted to the various administrations and discussed in CNIS. CNIL also gives its opinion. The question of the dissemination of data is also extensively discussed, especially given the need to reconcile the demands of users (local collectivities, companies, researchers) seeking information that is as detailed as possible in geographic terms with the need for confidentiality, which is advocated by CNIL. The National Institute of Statistics and Economic Studies (INSEE) is also sensitive to the need for confidentiality, since the Institute would like to see data-collection exercises take place under proper conditions.
3. Communes are very important partners in French censuses. Since there are no municipal registers in France, censuses first of all establish the legal (official) population of communes. Communes also participate in the collection of data: census takers are recruited by the mayors of communes. In the smaller communes, these are often full-time officers of the commune. This means that the census is closely coordinated with the Ministry of the Interior, which has overall responsibility for the communes. France, moreover, has more than 36,000 communes, half of which have fewer than 400 inhabitants. This dispersion creates conditions not found in any other country and creates difficulties for the organization of the census.
4. In selecting the questions to be asked in a census, INSEE follows a number of quite simple rules. First of all, the operation must be simple in terms of the questions to be put to the population. The documentation is therefore limited: the housing questionnaire is four pages long and the individual census form occupies the two sides of a size A4 sheet of paper. Selection from the demands of the different organizations seeking to introduce new questions is made under this constraint. A second concern is to ensure continuity—in other words, to ensure that no important or sudden changes are made to the questionnaires. In the discussions aimed at resisting pressures to increase the number of questions, the fact that individuals themselves complete their census forms is also important: the questions must be simple and easy to understand. The results must also be easy to process. In France, the special value of a census compared with other elements of the statistical system is that it compiles local data. At the national level, numerous polling surveys are carried out. Administrative sources are concerned mainly with specific fields and the information gathered cannot be applied to the individual level.
5. The outdated nature of census data constitutes a significant problem, particularly since the census dates are not known in advance. Data cannot always be used beneficially in exercises conducted prior to all the arrangements being put in place.
6. This is one of the main justifications for the current rolling census project, which should make available each year comprehensive data that are less than three years old and without any increase in the average annual budget allocated to the census. The project will rely more on administrative sources that are used to identify annual trends, while the rotating method relies on the collection of data in a given area every five years.
7. Increasingly, INSEE is subcontracting some of its tasks, either because these tasks require specific skills or in order to reduce the backlog in its workload and not delay the results. These two reasons are valid for the census. Data-processing bureaux were used for the first time in the 1982 census: the small staff of INSEE and the rules governing recruitment to the public service precluded the work from being done internally within the time limit provided, given the other activities carried out by INSEE.
8. In 1999, INSEE began to optically scan documents. This option was based on financial considerations, taking into account the risks inherent in using a technology that up until then had seldom been used on such a large scale. One of the major problems associated with the use of subcontractors has been respect for confidentiality, since personal census forms are being entrusted to outside firms. It is essential to guarantee the security not only of the premises where the data are processed but also of the documents during their transportation. This question may in future pose significant legal problems. The size of the contract for the optical scanning of data means that it is now necessary to invite tenders from European firms, thereby running the risk that documents may be processed in foreign countries and making it difficult to verify compliance with the norms governing the confidentiality of data.
9. In the 1999 census, for variables that were collected in word form, INSEE made extensive use of automatic coding after the wording was scanned. This procedure represents an important source of savings in terms of both resources and quality, insofar as it ensures greater uniformity of treatment.
10. In France, the interval between two censuses varies. INSEE is required to consult with budget officials before selecting a date for the census. This is not helpful to the preparations for the census, since there is little advance notice as to when the next census will take place. The average intercensal interval (seven to nine years) is not conducive to the proper exploitation of the data compiled. On the one hand, it is so long that little thought is given to the census that will follow the one currently under preparation. It is long in terms of the mobility of staff, a policy that has been favoured by INSEE for some years now: nearly all supervisory staff would have changed in the intercensal period. It is too long also in terms of computer applications: materials, infrastructure and software evolve rapidly and it is necessary to start again from scratch or very nearly with each new census. It should be noted, however, that for the past 25 years INSEE has been using software for the processing of census data in the tabulation phase, in the absence of software that might be more efficient in all phases.
11. On the other hand, the interval is too short to introduce major changes in areas such as the modalities of data collection and the elaboration of questionnaires, especially if these changes would have outside implications. Little time is available for testing to be carried out. In 1999, for example, a question was added which cannot now be retained since the reply cannot be processed. However, it is by no means certain that a longer period of preparation would always prevent this type of difficulty.
12. Given that censuses represent one of the rare sources of data for long-term analysis, it is sometimes felt that not enough care is devoted to archiving because of the lack of resources and also because the importance of this essential task has not been recognized.
13. Up until 1990, cartography was first and foremost a production tool used to prepare maps that facilitated the work of census takers. The work involved in preparing these maps was therefore limited, taking place as it did just before the census. It was done largely by hand and relied on maps supplied by the property tax services.
14. With the introduction of geographic data systems and the development of map-making software, the need for detailed geographic information rapidly increased, particularly since decentralization gave expanded powers to local collectivities.
15. INSEE therefore turned to digitized maps, at least in the larger communes (generally communes with more than 10,000 inhabitants, which cover half of the population of France). This was done for the 1999 census with the tool being used mainly for the organization of the census and dissemination of the results. Map-making for census purposes is currently evolving to take account of other needs: geocoding of files containing addresses, thereby permitting data to be derived from sources other than the district census.
16. This trend should continue in order to ensure greater harmonization between the publicly managed geographic systems (in their design and degree of geographic detail). Under these conditions, managing the geographic infrastructures is now a task separate from carrying out the census, even though it has remained within the population department. In particular, management will be an ongoing activity. In future, however, the census will not be a mere client of cartography. It will also be a supplier using the information obtained in its data-collection campaigns.
17. On two occasions in the past (1962 and 1990), INSEE has conducted a post-census survey. The survey was initially designed to measure the rate of omissions and double counting, but was also used to measure the quality of the replies obtained through the deposit and withdrawal method, as compared with data collection by a census taker. As regards their comprehensiveness, these two types of survey had rather similar rates. In 1990, the rate of omission measured was 1.8 per cent, while 0.7 per cent of persons were wrongly counted twice.
18. At a practical level, the conduct of post-census surveys creates significant problems. Since there is no database of the housing stock in France, the sampling frame can cover only one area. In 1990, INSEE used the area sampling frame that had been established for the European workforce survey and its survey covered approximately 20,000 dwellings. Such a sample already represents a large survey and requires substantial resources that are difficult to make available at the precise time when the staff of the regional offices of INSEE are heavily involved in completing the compilation of census data. Problems also arise because of the movement that occurs in the interval between the census and the survey: in one month, nearly 0.7 per cent of dwellings change occupants. Moreover, there is no assurance that the survey will cover the entire population, even with experienced census takers: problems of physical access to buildings and problems of dual residence also arise.
19. The value of the survey lies not only in the fact that it establishes a rate of omission or of net omission for all of France. In addition, it provides rates that are later differentiated according to zone so that the data can be corrected, if necessary. An area sampling, such as the one conducted in 1999, does not yield results for districts or populations of negligible size. Such a situation was never envisaged in France. Under these conditions, the data cannot be corrected and the survey’s objective is merely methodological.
20. Controversy has arisen over the communication of the results of a post-census survey. While, on the one hand, publication of errors committed may be viewed as a sign of transparency and provide information to explain why certain data do not reflect the reality, on the other hand, it may give rise to feelings of mistrust of the statistics. When the results of the 1999 census were being presented, the announcement of a certain inaccuracy in relation to the 1990 survey instead facilitated the communication of the census results and of the related sociodemographic analyses.
21. The census update project again raises but in a different manner the question of post-census surveys. It underscores the need to integrate into the system a process to measure quality and results, at least in the large communes, in which the biases that result from omissions or double counting will be corrected by the very methodology used.
* This document was reproduced without formal editing
** National Institute of Statistics and Economic Studies (INSEE), France. The views expressed in the paper are those of the author and do not imply the expression of any opinion on the part of the United Nations Secretariat.
 There is currently no legislation providing for the conduct of the census. This situation will change with the census update project, which is the subject of draft legislation currently before the French National Assembly.