Censuses Of Population Research Paper

2. Rise Of Modern Censuses

The first modern, national census of households and persons was established in the Constitution of the new USA in 1790 and has been continued decennially since then. Its first major purpose was allocating seats in the new national Congress, but it also served other purposes and needs (Anderson 1988). Those purposes and needs increased gradually and the variables collected by the census increased likewise. The idea spread and decennial censuses began in the UK in 1801 and in Belgium in 1846. The International Statistical Institute’s Congresses of 1853, 1872, and 1897 passed recommendations in favor of decennial censuses for all countries, with some basic requirements, aiming at standardization for better inter-national comparability. It also urged open publication and dissemination of results. In 1900 some 68 different countries already conducted and made available population censuses, which included 43 percent of the world’s population. The quality and content of the censuses improved gradually. For example, it is only since 1850 that the data for USA censuses covered all individuals rather than just families.

The greatest international impetus came after 1946 from the Population Division and the Statistical Division of the new United Nations. The stimulation, guidance, and coordination of the UN Statistical Division were crucial for spreading the establishment of censuses and their comparability (UN 1980). The efforts of the UN’s sister agencies (FAO, WHO, and UNESCO) were also important. Beyond the growing coverage of most countries and improved coverage within countries, population censuses have advanced in five other dimensions. First, the quality of the census data has been improved. Second, that improvement has been subjected to measurement by quality checks, often with duplicate observations. Third, both the contents and the quality have been increasingly standardized across countries in order to facilitate international comparability of data. Fourth, the number, diversity, and depth of census variables have been greatly increased. Fifth, the analyses of census variables have been vastly deepened and improved, with the help of modern computing machines and methods.

In addition to censuses of population and housing, many countries also conduct censuses of agriculture, business, manufacturing and mineral industries, etc. These are usually done separately, often by different and specialized staffs and enumerators. The censuses of agriculture may still be linked to population censuses in less developed countries, where most households still also operate small farms. However, in industrialized countries most agricultural products come from large firms of ‘agro-business.’

2.1 Contents Of Population Censuses

Beyond mere head counts, modern population censuses obtain several classes of variables, as shown below, condensed from the UN’s recommended list:

Residences—usual, on the census date, duration, previous, at birth; place of work

Sex, age, marital status, relation to head of family and household

Citizenship, religion, language, national ethnic origin

Children: born alive, living. Births last year: live and dead

Marriage: age, duration, order

Education: attained, literacy, qualifications, attendance

Occupation: status, industry sector, time worked

Income, socioeconomic status, economic dependency

From this recommended list, most countries omit some topics or include them only in some years. On the other hand, some countries add other topics, of course, as desired. With all those variables, and with increasingly complex and detailed analyses, the published outputs of censuses grew into census monographs of many volumes, and then took years to finish (Farley 1995). Therefore statistical offices are continuing to develop cheaper and faster methods of dissemination, with the aid of computers and electronic publishing.

2.2 Timing Of Censuses

Censuses of the population must assign specific locations on specific dates (the census day or reference period) to people. The collection period for enumerating may be one week or a few weeks. However, the reference period is a single census day for items like residence and age; for items like income (or crops) it may be the preceding year. However, humans are highly mobile and they are becoming even more so: with longer and more vacations, longer travel to work, more second homes, more split families, and so on. De facto is the census term for the actual location of persons on census day and de jure for their ‘usual place of residence’ (home). Without this distinction city workers would make the central cities larger in the daytime when their ‘usual place of residence’ locates them in the suburbs. But location problems are more difficult for students living away from home in university cities; for death rates in cancer hospitals far from home; large institutional populations in prisons or in the military; for second vacation homes, etc. These are all examples that census definitions and operations must cover, in the field and in the office. Nomadic populations pose a great variety of special problems in timing and locations, of too great a variety for us to attempt to cover here.

For the census work in most countries, tens or hundreds of thousands of enumerators must be hired and trained for just one or two weeks of fieldwork and a few months in the offices. In China and India they need millions of enumerators. Some experts have doubts about how well they can be trained. However, others contend that the ‘ceremony,’ the patriotic campaign that can be mounted every 10 years, with educational campaigns, and exhortations in the media, increases awareness, participation, and response rates. Still others fear that the campaign arouses the latent and dormant opposition to fact-finding.

Decennial censuses may be regarded as systematic samples of a continuous time dimension: samples of a single census day in the 3,652 days of 10 years, or one year every 10 years for variables like yearly income. Estimates for intercensal and postcensal periods must depend on models and methods, and these are called intercensal and postcensal estimates. Other methods are also available for national and large domain estimates; but what are lacking above all are estimates for small areas and small domains that are also more detailed in time than decennial censuses. These needs have given impetus to methods of small area estimates and small domain estimates (Platek et al. 1987, Purcell and Kish 1986). Their accuracy depends on relations to auxiliary data from registers and sample surveys; also on the stability of variables like total population. But they are not accurate for unstable and fluctuating variables, like unemployment, infectious diseases, and financial variables. Sample surveys have been developed for some of these (see Sect. 3).

2.3 Coverage By Censuses

All censuses have problems of coverage; and even the best suffer a few percentages of noncoverage (or missing units). Furthermore, typically the units (households, persons) are not missed evenly, or at random; they are typically missed more often among the lower socioeconomic classes, some ethnic groups, and areas. These uneven coverages persist despite efforts to find and to measure them. However, the size and bias of this noncoverage may often be smaller than the even greater distortions due to the uses of obsolete decennial censuses for 10 years. All national census results also consist of combinations of regions or provinces or states, also of districts or counties or communes, and so on down to small domains. The populations of the provinces may differ greatly in culture and standards and the national results represent averages of the provincial levels. For some provinces different languages and questionnaires may be needed. Further, deliberate omissions may exclude some areas because: (a) they are vast, ‘uninhabited’ mountains, deserts, jungles, or small islands; or (b) they may be areas under hostile populations or rebellions. Changes of boundaries that have occurred, either losses or acquisitions, after wars and treaties cause problems for publishing and reading census monographs, tables, and graphs.

2.4 Costs Of Censuses

The fine spatial detail of complete censuses is bought with high total costs. Because the total costs of censuses are high, they are usually only held every 10 years; and only a few countries had five-year censuses. The growing number of topics covered by censuses have also increased their total high costs and these prevent more frequent census taking, despite demands for annual data with spatial detail. At the same time the large size of the census is also an obstacle to obtaining greater richness of data (with more variables) with spatial details. This high cost is one cause of the sporadic political controversies about censuses in many countries. But even more serious are the conflicts due to the policy implications of census results, especially those that are used for apportionment between provinces and between ethnic groups. Apportionment of seats in Congress (Parliament), of taxation, of armies, of welfare funds—the very reasons for collecting census data—can also lead to conflicts, and even (rarely) to suppression of data. The collection and open publication of reliable and trusted statistical data is (like liberty) ‘the fruit of constant vigilance.’

The costs of censuses must vary with the methods used to collect and process them. Door-to-door personal interviews have been the standard throughout history. However, now they have been modified in places with telephoned and mailed-in questionnaires, and also with mailed-out-mailed-in postal methods. Processing and editing have been speeded with mechanical and electronic methods; but the savings seem to be swallowed by increased complexities of analyses. The comparison of total costs across times, and between countries are complicated by population sizes, exchange rates, and living standards; but guesses are made that the total cost of censuses per capita population equal about 1 2 to 1 hour of the countries’ median hourly wage for every 10 years.

3. Sampling And Censuses

Two other methods for collecting population data have emerged as both complementary to and competitive with complete census taking: sampling methods and administrative registers. Sampling methods have come to be used in several distinct ways in different countries.

3.1 Census Samples

The addition of great numbers of desirable topics increased the costs of censuses, and since 1950 a solution to this problem was increasingly found in the use of census samples. In addition to the complete count using the ‘short form’ containing a few basic questions, a sample of the population completes a ‘long form’ with many variables. This sampling reduces both the total cost of the census and the burden on the majority of respondents in 2000. The rates and methods of sampling differ between countries. The sampling units may be enumeration districts (areas— EDs or EAs), or households, or persons. Selection of persons has lower sampling variance, but selecting EDs may have better enumerative controls and greater flexibility. The sample population may only get the long form or they may get duplicates of both the short and the long forms. Census samples generate large samples, whether 5 or 10 or 20 percent of the complete count, and thus they can provide detailed spatial data, but with reduced costs. They are still much larger than ordinary survey samples, which have selection rates of 1/100, 1/1,000 or 1/10,000 of complete counts.

3.2 Postenumerative Surveys, Quality Checks, Coverage Checks

Samples may be taken for quality checks with measurements that may use more costly and (hope-fully) improved methods. These measurements may be in addition to, or instead of, the regular census data collection and the sampled units may be persons, households, or enumeration districts. Coverage checks are attempts to find the sort of persons missed by the regular census count. The surveys are often called postenumerative because they are usually done after the census collection, but they may be done simultaneously.

3.3 Sample Censuses

These have been conducted by a few countries, and sometimes are called ‘minicensuses,’ perhaps based on one-percent samples and with fewer topics than the sample surveys described below. The greatest expressed need is for annual data to supplement the decennial censuses, which are increasingly judged to be inadequate because of obsolescence. Typically, census monographs containing the richer data of the long form become available five or more years after the census year, although the simple counts may be released quickly (Farley 1995).

3.4 Sample Surveys

Generally, sample surveys differ greatly from complete censuses; they are much smaller; they are much cheaper, but they also lack detail in spatial (geographic) and other domains. But periodic samples yield much needed temporal detail, and also have other advantages. They can yield much richer data, especially because the content and the methods can be focused on specific contexts. Also because that greater concentration can be achieved for the questionnaires, for the interviewers and for the respondents as well.

‘Representative sampling’ was first proposed by F. N. Kiaer in 1895 in Norway, for replacing censuses. The fundamental concept requires a careful, scientific method for selecting a sample from the population and then estimating the population variables from the sample statistics. Many statisticians, academic and practical, in many journals and in the International Statistical Institute for about 50 years, opposed sampling as a substitute for complete enumeration. Meanwhile probability samples came into use for social surveys, notably for unemployment rates and then for the famous quarterly surveys of the labor force, which became the Current Population Surveys of the US Bureau of the Census (1978). All countries in the European Union and many others now have quarterly or monthly surveys of the national labor force. These surveys, mostly of 5,000 to 100,000 households, are financed because governments need current monthly or quarterly data on (un)employment and its periodic fluctuations. National statistical offices generally conduct many other samples every year, some of them only once, some repeatedly, and some periodically. Some of the censuses are also produced by sampling methods and the Food and Agriculture Agency (FAO) has been in the forefront of promoting agricultural censuses by means of sample surveys.

4. Administrative Registers

Administrative registers related to population counts, sometimes called civil or population registers, may be of different kinds, purpose, and quality in different countries and at various times. Vital registers of births and deaths may be the most common examples, but others exist for voter registration, or recently auto licenses, telephone numbers, etc. The most relevant registers date from 1749 in Sweden and the Nordic countries, where the official church started keeping records of the parish populations, which continues today. Some statisticians propose replacing decennial censuses with data from administrative registers of different kinds, or with the combined social resources available in specific countries. The data are inexpensive, almost without cost, because some other agency, with other aims and resources, pays for them. Furthermore, they are often detailed spatially, and also detailed temporally, timely, and they are kept up-to-date. They have been used in several Scandinavian countries to replace their population counts. However for two chief reasons, most countries cannot use administrative registers to replace censuses. First, for lack of the conditions that facilitate the successes of Scandinavian registers: a highly educated and cooperative society, which is highly motivated to register promptly their own changes of residence. Second, no registers, not even the Scandinavian registers, have ever obtained the richer data (beyond a few lines) that modern censuses need. In addition, some statisticians fear the misuse of registers by careless, or malevolent, governments and societies.

5. Future Developments

Modern censuses face some gravely conflicting demands: higher quality and better coverage of the population, but with reduction of mounting costs. We need spatial detail, but also temporal detail and timeliness. Censuses yield more spatial detail, but sample surveys are superior in all of the three other aspects. The greatest need is for more timeliness, because decennial data do not satisfy modern needs, especially for annual data with spatial details. At the other extreme the monthly surveys provide adequate national statistics and satisfy the other requirements, but they yield no spatial detail for local needs. This conflict gave rise to designs of ‘cumulative representative samples’ and then to ‘rolling samples,’ in the last few years. The National Household Health Survey in the USA yields separate samples of about 1,000 households each week, and they are cumulated into 52,000 households yearly, with health data on about 150,000 persons. Thus the yearly cumulations yield data for even moderately rare diseases and infirmities. Furthermore, the American Community Survey of the US Census Bureau is conducting a large pilot study for a ‘rolling sample’ by AD 2003, which may replace the long form of the US Census of 2010 (Alexander 1996, Kish 1998). Many statisticians in AD 2000 felt that the greatest need generally is for regular, standardized annual surveys (sample or census) that are harmonized internationally, and also large enough to yield reliable data for small domains. It is likely that the censuses of the twenty-first century will be quite different from those of the twentieth century.

Bibliography:

1. Alexander C 1996 A rolling example survey for yearly and decennial uses. Bulletin of the International Statistical Institute. Helsinki, 52nd session
2. Alterman H 1969 Counting People, The Census in History. Harcourt, Brace and World, New York
3. Anderson M 1988 The American Census: A Social History. Yale University Press, New Haven, CT
4. Farley R 1995 State of the Union, Vols I and II. Russell Sage, New York
5. Goyer D S 1992 Handbook of National Population Censuses: Europe. Greenwood Press, NY
6. Kish L 1998 Space time variations and rolling samples. Journal of Official Statistics 13: 31–46
7. Kish L, Verma F 1986 Complete censuses and samples. Journal of Official Statistics 2: 381–95
8. Platek R, Rao J N K, Sarndal C E, Singh M P 1987 Small Area Statistics. Wiley, New York
9. Purcell N J, Kish L 1987 Estimation for small domains. Biometrics 35: 365–84
10. Shryock H S, Siegel J S 1973 The Methods and Materials of Demography. US Bureau of the Census, Washington, DC
11. Taeuber C 1986 Census. In: International Encyclopedia of the Social Sciences. Macmillan, New York
12. United Nations 1980 Principals and recommendations for population and housing censuses. Series M No. 67 (Sales E.80.XV11.8)
13. US Bureau of the Census 1978 The current population survey: Design and methodology. USCB, Technical Paper 40, Washington