In the UK, our government is currently looking at how the census should be conducted in the 21st century. But to understand how best to develop the census for the future, it’s worth looking back at how the concept has changed throughout human civilisation, in order to deal with the new challenges that history has presented us with.
Political groupings have always been empowered by the size of their membership, but as states become larger and their populace more dispersed, this size is not necessarily known. Records were collated at local level in renaissance Europe but these were unwieldy for the central aggregation government required and could be incoherent, idiosyncratic and out of date. Settlers in the New World were at first enumerated exactly because they were small in number and clustered in tightknit communities, but continuing settlement put paid to early systematic records.
To regain control of their administrative systems, nations fell back on a method used by Ancient Greece and the Roman Empire. Classical civilisations needed to have accurate records of their population in order to know their electorate, military strength and taxation prospects. Various methods were used such as the biblically mentioned ‘numbering’ whereby male citizens were required to return to their town of birth for registration. A more quaint and simple method of enumeration is reported by Myssiakoulis when a city state required each person (male and female) to add a rock to a pile which could then be counted to know the total population. From these historical practices we still use the Latin word 'census'.
The modern concern, conceived in the eighteenth century, was also concerned with the ideas of measuring military power (in European nations), and the apportionment of electoral representation (such as laid down in the US constitution). But another purpose had emerged with a more subtle concern about the understanding of the population and how it was changing: demography, the life conditions of communities of people. For example, Malthus wrote with concern about the problem of unchecked population growth and the pressure this would put on living conditions in nations and communities within them.
Historic censuses had cheerfully ignored swathes of the population, those who could not vote, fight or pay taxes, but demography required a complete account of everyone in the nation. This meant not just enumerating the people but categorising them by age, sex and location, and other things besides, yielding geo-demographic data which could be put to many purposes. This enhanced data requirement gave rise to enumeration by questionnaire rather than simple registration and counting, and thus the role of enumerators being literate enough to collect this information so that it was comparable across the nation.
Such a role for enumerators evolved the strategy of de facto (literally as they are found) household enumeration and central co-ordination by registrars-general in nascent statistical offices. Thus statistical compendia could be produced for provinces, with vast tabulations of household characteristics describing the state of the nation at one point in time. These data could be used to make actuarial calculations about the population and develop demographic accounts supported by periodic renewal of the data every ten years.
The power of data about people was becoming clear in its ability to systematically identify similarities and contrasts whose utility for public policy was taken up in new statistical societies. In Britain, the Royal Statistical Society lobbied for a Census Act to enshrine the decennial, universal enumeration in law as early as the 19th century, at a time when the British Empire conducted censuses all around the world. But it had to wait until 1920 for the formal recognition and the statute for compulsory response by the householder, subject to a fine, and this legal compulsion means that nation states still form the authority for modern census practice.
As the interest in data on society grew, the number of questions asked in census operations expanded, starting to include more social concerns about facilities and circumstances of the household. Contemporary practice recognises this by describing the activity as 'population and housing censuses' which also draws out one of the statistical issues around census - although they try to enumerate the population, they survey households, not individuals.
Data about households is perhaps more helpful as it is in households that poverty exists, and in households that we know the family demography of a nation. But here the tradition of 'de facto' enumeration presented a problem: people were not necessarily in their home on census day, particularly if their occupation was mercantile, seafaring or military. This created distortions in some populations where rural districts could be undercounted by those away on business and cities could swell correspondingly. More perverse was the allocation of the crew of coastal barges (common before modern road networks) to the parish where the barge was offshore on census night.
Although the United Nations still recommended de facto enumeration, most developed countries adopted a more natural measure of 'de jure' residence which means ‘in law’ but is generally administered by a definition of ‘usual’ rather than 'permanent’ residence. This places individuals in their family home, or in the case of students, at their term time address, and will have rules for all types of exceptional circumstances so that each individual can be uniquely decided. The fundamental of census is enumeration is not so much that everyone is surveyed but that each person is recorded exactly once, in the household where they live their life.
Enumerating every household required a register of addresses to be certain that each household would be visited (countries without address data for rural areas have now utilised satellite technology to simultaneously geocode at the point of enumeration). But once such an address register was available, it became possible to conduct sample surveys of the population, a methodology which had been viewed as unworkably biased at the start of the 20th century.
Sampling populations allowed the development of statistical systems using more detailed surveys on specific topics made representative by weights derived from census data. Thus census more obviously had two purposes: to estimate and benchmark the basic geo-demography of the population, and to produce more detailed information on the aggregated attributes of individuals and households. However, the availability of specialised surveys to produce more timely data on the attributes of the population challenged the primacy of census activity.
These themes are explored in greater detail in an article published by the author and other contributors here.