|
Symposium 2001/49 3 August 2001 English
only
|
Symposium on Global Review of 2000 Round of
Population and Housing Censuses:
Mid-Decade Assessment and Future Prospects
Statistics Division
Department of Economic and Social Affairs
United Nations Secretariat
New York, 7-10 August 2001
Perspectives
on censuses of the new millennium*
K.E. Vaidyanathan **
CONTENTS
A. Strategies for involving
stakeholders in census activity.
B. Data-collection methods:
censuses, surveys and administrative records
C. Adapting new technologies
to census operations.
5. Data management and
dissemination.
D. Maintaining census-related
activity in the intercensal years.
1. Establish the master sample
of the enumeration areas.
3. Undertaking demographic and
social surveys.
4. Establishment of
information system..
5. Preparation for the next
census.
E. Identifying and resolving
problems of census mapping.
F. Post-enumeration surveys:
are they worth it?.
1. Where there is a constitutional requirement to conduct a census every 10 years (for delimitation of constituencies for election) the commitment of the government to conduct a census is present. This has been the case in India, where there has been an unbroken chain of censuses since 1871. Where such a constitutional requirement is not present, there is a need for different stakeholders to press for a census every 10 years.
The potential stakeholders are:
1. Statistical organizations in the country who will benefit from an expansion of staff and equipment arising from a population census. Significant numbers of staff will be able to upgrade their skills through theoretical and hands-on training. The end result of a census is a statistical system that is capable of designing and implementing high-quality censuses and household surveys with little or no outside assistance.
2. Policy and decision makers at the national, state (or province) and local levels. At the national level census data are required as a key input in national accounts, estimation of requirements for different commodities and services and for formulating social programmes and interventions. National governments have been using census data for preparing the Poverty Reduction Strategy Paper (PRSP), which is a requirement for assistance from the World Bank. At the state and local level, in addition to the above, the budget, staff and other resource allocations are often based on census figures of population.
3. Non-governmental organizations (NGOs) within each country. These groups are among the greatest beneficiaries of population censuses. For example, chambers of commerce and industry, syndicates and associations of different trades make use of census data for market research and for informing their clientele about potential business opportunities. National associations in the field of social sciences, population, and family planning use the census data for designing, monitoring and evaluation of their activities, as well as for advocacy purposes. Lately, feminist organizations are using census data to buttress their claims for gender equity and justice. For example, in India feminist organizations are using sex-ratio data from the census as the basis for questioning the legality of amniocentesis. Universities, research institutes and individual researchers are also other stakeholders in census activities for obvious reasons.
4. International organizations like UNDP (United Nations Development Programme), UNFPA (United Nations Population Fund) and the World Bank are major stakeholders. UNDP’s Human Development Report and Human Development Indicators (HDI) draw heavily upon the data and indicators derived from the population census. UNFPA has been a major user of census data for monitoring demographic trends and for preparation of the annual report on state of the world’s population. In addition, UNFPA has been using census data for its strategy development and project formulation missions and for advocacy purposes. The World Bank has been using the census data for the preparation of documents relating to poverty alleviation. Since 1996 the World Bank has been supporting a number of living standards measurement surveys (LSMS) in developing countries, which have drawn heavily upon census data for their sample selection.
2. In countries where censuses are taken regularly, such as India, the different stakeholders are involved in the preparations for the census. For example, in India census data users’ conferences are held to determine the contents of questionnaires and the procedures to be adopted. In addition non-governmental organizations like the Indian Association for the Study of Population (IASP), universities and even individual researchers act as pressure groups to influence the selection and development of the topics included in the census questionnaire and the approaches to measurement issues. In countries where such pressure groups do not exist (such as many African countries) it may be necessary for international organizations and NGOs to step in and act as pressure groups influencing census taking. The census of China in 1982 and many African censuses in the 1960s and 1970s were due to pressure from the UN and financial support from UNFPA.
3. There is a need to establish scientific societies (or associations) in the field of population in many African countries, and the International Union for the Scientific Study of Population (IUSSP) should support the formation of such societies through funding support in the same way that the International Planned Parenthood Federation (IPPF) is supporting several family planning associations. The support may be towards specific activities, such as an annual seminar on population, publication of a journal, establishing a small office for the association or other activities. Ultimately, IUSSP may have to seek funds from funding agencies like UNFPA, DFID (Department for International Development), USAID (United States Agency for International Development) or private foundations to provide necessary support to the national associations so that these associations can act as pressure groups for census and other population activities in these countries
4. The extent to which the clients’ demands are met depends upon the resources available, external funding, technological infrastructure (e.g., Internet facilities) and so forth, and also the capacity of the clients to pay for the information. The traditional methods of disseminating the information, like hard copies of census publications, are fast giving way to diskettes, CD-ROMs, e-mail and even posting of information on the Internet. More and more clients benefit by the use of modern technology.
5. In the large majority of countries there is no alternative to the population census as the means for establishing the frame for conducting surveys and fulfilling administrative needs such as delimiting constituencies for elections. In countries like the Netherlands, where the population-registration system is efficient, a population-registration system may appear redundant, but in the absence of the population census there is no basis for determining whether the population-registration system is accurate. It is also doubtful if the range of questions asked in the population census can be included in the population-registration system. In countries like Bosnia and Herzegovina and Somalia, where large-scale population movements have taken place as a result of war, there is an urgent need for new census data. Since no alternative sources of data are available, the previous census data are being used with adjustments in spite of the fact that these data are becoming obsolete. For instance, in Bosnia and Herzegovina the sample for the LSMS survey is based on the pre-war survey results with updating of selected municipalities. In Somalia the current estimations of population were derived by the author using the pre-civil-war census and survey data with various assumptions concerning fertility, mortality and migration during the war and post-war period.
6. Yet another application of the census is the “master sample” which is a large number of clusters—usually several hundred—which are selected by the statistical agency after the census and for which updated household listings are maintained in order to select from the pool of households the sample of households to be interviewed in each survey. The advantage of the master sample is that there is no need to select a new set of clusters for each survey or to conduct a household-listing operation each time. Travel costs can be reduced if interviewers can be selected from the same area.
7. Where census are taken every 10 years, there is still the necessity to conduct sample surveys to obtain information on current demographic and social statistics. Indeed, this is what every developing county is doing. India, for example, has annual rounds of national sample surveys to obtain current data and to complement the information obtained from the decennial census. In many developing countries administrative records are incomplete and inaccurate and cannot be regarded as a substitute for censuses; moreover, the information provided by the administrative records is not necessarily the same as that obtained from the census. Administrative records can serve different purposes, namely, to monitor the performance of the provider of the administration; only incidentally do they provide statistical information.
8. The technology of censuses has come a long way from the stage when manual counting of census slips was done (e.g., in the Indian censuses up to 1961) to the use of the unit-record equipment, then the large first-generation computers, and now to the use of personal computers and laptop computers. The march of technology cannot be stopped and census takers need to keep pace with these changes. The following are areas where new technologies could be adapted to census operations in the planning, execution, tabulation and dissemination stages.
9. Since information pertaining to the census should reach the entire population, media coverage is important. The census administrator should use all available technologies to inform the public about the census. This would include radio, the Internet, e-mails and so forth.
10. Assuming that the conceptual structure of the census questionnaire (including modules on fertility, labour or migration) is well thought out, technology can play a great role in designing the proper format of the questionnaire. A good format minimizes the error during the interviewing and subsequent data entry, thus improving data quality and timeliness of data. In my view only one questionnaire should be used for each household. In some countries, such as India, the census questionnaire has two parts—a household questionnaire and an individual slip. Unless proper identity codes are included in all questionnaires, separate questionnaires create the risk of improper matching. A grid may be required in cases where there is more than one unit of analysis in a household—for example, several eligible women for the fertility module, several employed persons in the labour module and so forth. The EXCEL software is very helpful in proper formatting of the questionnaires. Other word-processing and graphic software packages can also be used for producing the questionnaire layout. Revisions of the drafts can be made more easily with the help of this software. The computerized approach to translations allows the overwriting of parts of the questionnaires in the local languages, leaving skip codes, response codes and the general format unchanged.
11. The logistics of the population census can be mind-boggling at all stages of using the questionnaires—their printing, distribution, collection for data processing, data entry and processing and storage. All these activities require thorough planning. Equally important is the selection of the interviewers, their training and deployment and monitoring of their fieldwork. Communication between the census administrator and field units and between the interviewers and supervisors can be improved through the use of cell phones and e-mail communication. Videoconferencing can be used for training interviewers in different locations, ensuring uniformity of training. The data collected in the field can be entered into the computer through the use of laptop computers and transmitted to the data-processing unit through e-mail. Use of new technologies involves additional costs, and one should consider the trade-offs between cost and time when deciding whether to use such technologies.
12. Commercially produced data-entry packages have fast superseded customized data-entry programs adopted in earlier censuses. The Sudan census in 1993 produced all the results within two years after the census because of the use of IMPS (Integrated Microcomputer Processing System) software developed by the US Bureau of Census. Other similar software like Blaise, ISSA and EPI-INFO can be used, but each has its limitations. Software developments are occurring rapidly and there will be other software packages available soon to meet the needs of the new round of censuses. For example, the new software CSPro (Census and Survey Processing System) combines the features of IMPS and ISSA. Concurrent data entry makes it possible to use it in a pilot census or in the first month of the population census to gain insights into the quality of the work, enabling the improvement of the training and ensuring better control of the fieldwork. However, the use of this software has not been tested in a census. Eventually, with advances in computer technology, enumerators will be able to enter data directly into a portable laptop computer during the interview, thus eliminating paper questionnaires completely. This will require a totally new approach to questionnaire design, organization of the fieldwork, the system of quality control and management of data.
13. Every change in technology—for example, improved data-entry programs, portable hardware and computer-assisted interviewing—will have implications for data management and dissemination. For example, with new data-entry programs the data manager has to ensure that the household data files prepared by different data-entry operators are included without duplication or omission and convert the large number of household-based files into a few thematic files. A master version of the files may be maintained in ASCII, but it may be necessary to convert them into other formats to facilitate their use by data users who use different software packages, such as SPSS, SAS, STATA and so forth. It should be ensured that the different thematic files can be matched and merged for further analysis. The data manager should take note of the current trends in handling this task.
14. The traditional method of dissemination of data through hard copies of census data will need to be maintained, since the large majority of users still do not have the access to computers or the Internet. At the same time developing countries need to adopt the new technologies of dissemination, such as the Internet, e-mail and others. They should also use software for mapping and graphics for better presentation of the information.
15. The following census-related activities may be undertaken during the intercensal period.
16. A master sample is a large number of clusters of households which are selected by the statistical agency following the census and for which updated household listings are maintained. From these samples of households, subsamples of households are selected for each survey. The master sample should be available for each analytical domain or for subgroups of population—rural, urban, provinces or regions—for which information is required. These master samples should be updated every two years or prior to the implementation of a survey.
17. Soon after the census only the essential tabulations required by the analysts and policy makers are produced in the form of an abstract. This will typically include the frequency tables of the various variables covered in the census according to the selected socio-economic groups and analytical domains. Further in-depth analysis of the data is necessary to establish relationships between different variables and for studying changes during the intercensal period.
18. Following the census the statistics agency could undertake sample surveys to obtain information on current demographic and social indicators. Examples of such surveys are the World Fertility Survey, Demographic and Health Surveys and Living Standards Measurement Study.
19. Using census data as the basis, an information system can be established providing key socio-economic indicators. This must be updated annually using information available from sample surveys, administrative records and estimations based on the census.
20. Preparations for the next census normally begin two years prior to the date of the census. Preparations include the house-listing operation, preparation of census maps, preparation of the questionnaire and pre-testing it, selection and training of interviewers and organizing all the logistics of the census. In the past census organizations were similar to the proverbial phoenix which arises from the ashes. This was the case with the census organization in India up to the 1950s. However, it was found necessary to have a permanent census commissioner and registrar general with a core staff on a permanent basis. At the state level the office of the director of census operations is also maintained with a core staff of census statisticians and computer personnel. However, the large body of staff—enumerators, supervisors, data-entry operators and so forth—are recruited prior to the census and disbanded soon after the census. The census organization in India is separate from the central statistical organization and the national sample survey. When census taking is a responsibility of the statistical agency, the staff of the agency serves as census planners and administrators. This is the only practical way continuity and institutional memory of the earlier censuses can be retained and utilized in later censuses. In some countries (like Iraq) census records are stored for a long time as reference materials for administrative purposes (e.g., for purchasing property in Baghdad a person should be enumerated in the previous census). There are countries which keep their records for 10 years and then discard them. With the advance of computers census records can be stored on CD-ROMs, diskettes, Bernoulli boxes and so forth, and they can be retrieved at any time.
21. This is the weakest area in the preparation for the census in every developing country and needs to be addressed. There are two kinds of problems: maps may not be available or they may be missing due to war, mishandling, etc. Secondly, maps may be available but incorrect due to changes in boundaries, construction of new dwelling units, factories and so forth, or changes in street names, street numbers or numbers of buildings. This is particularly significant in urban areas which undergo change due to annexation of rural areas or re-drawing of boundaries between the areas within the locality. It is necessary to strengthen the cadastral and mapping authorities to carry out revisions of maps on a regular basis at least once every five years. These maps should be secured by the census authorities during the house-listing operation and further updated based on the information obtained during the house-listing operation.
22. In most developing countries the post-enumeration survey (PES), in the manner it is done at present, serves as means of self-assurance rather than as a means for correcting the data. Countries are able to boast that their censuses are 99.5 per cent complete, and so on, but hardly anything is done to correct the data. This is not the case in developed countries like the United States and Canada, where corrections to census data are carried out using PES results. There is a need to change the concept of the PES from the means to correct the data to the means of informing the analyst about the pitfalls in the data, so that analysts can apply their own corrections. The PES should be focused on identifying the coverage and content errors in key variables, like rural/urban population, employment and unemployment, gender, children ever born/surviving, occupation, school enrolment and so forth. This would help the analyst to look at the data more closely rather than drawing erroneous conclusions. There is a need to thoroughly re-examine the concept and methodology of the PES to make it useful to identify problems in the census that are often overlooked and consequently have an impact on the census results
