Questions of identity in the millennium round of Commonwealth censuses.

All Commonwealth census authorities have posed questions about identity in the millennium round of enumerations. The most controversial issue has been ethnicity. No universal definition or classification system has been devised and each of the 71 states and dependencies has tended to pursue the enquiry in virtual isolation from its neighbours. The attempt to describe the population in terms of race and ethnicity has been inherited from the colonial era. More recently the requirements of monitoring affirmative-action programmes in multicultural populations have resulted in the introduction or refinement of questions on these categories. A few states—a small minority of Commonwealth countries—remain hostile to such enquiries. Where the issues of race and identity are not pursued, questions of nationality, language, and religion often fill the gaps left, adding further refinements to the definition of identity. An examination of current questioning about identity in Commonwealth censuses reveals a highly complex picture.

Keywords: Commonwealth; classification; identity; ethnicity; nationality; language; religion; census

Censuses are conducted to promote the regulatory nature of the State by measuring, quantifying, and categorizing its human resources (Scott [57]). Thus information is solicited not only on age and sex, but also on a wide variety of social and economic factors considered relevant to the exercise of control and the fulfilment of government functions. These factors include social identity, questions on which have been posed in many countries from the late eighteenth century onwards. One reason for using them was to allow the surveillance of potentially disaffected groups. British colonial 'divide and rule' policies required the identification and quantification of societies that were perceived to be fragmented. In post-colonial times, when elections were little more than 'ethnic censuses', the manipulation of the delimitation process based on the census served the same process (Horowitz [27], p. 86). Census data were also required for the demarcation of language-based administrative units (India [28]).

In the late twentieth century affirmative-action programmes have targeted specific groups deemed to require special services. Monitoring the progress of these groups has become another census function. One advocate of this use of census data noted that 'information is essential in developing policies which reflect the needs of our society and for the effective delivery of services to particular ethnic communities' (Australia [5], p. 166). By means of the census, the government has identified cleavages in society and adopted classification systems intended to bring order to apparent diversity (Bandypadhyay [6]). In doing so it has been observed that '... the census imposes order and order of a statistical nature. In time the creation of a new ordering of society will reshape that which the census sought merely to describe' (Barrier [8], p. 75).

A substantial body of literature has been devoted to the national censuses and their roles in nation-building, group creation, and identity (e.g., Srivastava [62]; Barrier [8]; Anderson [2]; Maheshwari [35]; Mohanty and Momin [41]; Worton [67]; Anderson and Fienberg [3]; Hannah [25]; Nobles [46]; Rodriguez [55]; Curtis [16]; Kertzer and Arel [32]; Aly and Roth [1]; Makeba Laversuch [36]). In this paper I propose to demonstrate the diversity of approaches to the subject by examining the four main categories of identity—nationality, race or ethnic group, language, and religion—posed in the millennium round of Commonwealth censuses.

Nearly all the countries of the Commonwealth conducted a population census close to the new millennium (1996–2006), usually in the years 2000 and 2001. They ranged in scale from the Census of India which enumerated over 1.0 billion people, to that of Tokelau with a population of 1,500 people. Although there was regional cooperation, notably in the Caribbean and the Pacific, the census questionnaire adopted by each administration was unique to that administration. All countries published census volumes and many also established comprehensive websites. Some reports were limited to basic tables, while others produced substantial monographs on the methods and results of census analyses. In the majority of cases countries maintain permanent census offices that provide ongoing services to data users, including researchers.

Of the 71 censuses conducted, 50 enquired into citizenship or nationality, 56 into religion, 43 into race or ethnic group, and 25 into language (Table 1). Everyone had to answer at least one question about their social identity, although a response to the religious question was not compulsory in some cases. Nevertheless, no Commonwealth-wide statistics could be produced on any of the topics. Furthermore, even for countries where the same question was asked, comparability among them was often limited by the absence of any standard system of categorizing responses. This is an important issue since even the order in which a list of options is offered can have a significant effect on the results (Lang [33]).

Table 1  Subjects of questions posed in the millennium round of Commonwealth censuses

The most legalistic aspect of identity is related to the individual's nationality or citizenship, access to which is controlled by the state (Caplan and Torpey [13]). In its basic form the question seeks to establish who is a national or citizen of the country conducting the census. The question was widely used in the Commonwealth questionnaires, but not in those of either India or the UK. The Indian population is large and highly complex and, since independence, potentially divisive identity questions have been kept to a minimum in an effort to promote national unity (India [29]). In the UK the question is being considered for inclusion in 2011, as part of the parallel investigations into international migration and national identity (Great Britain [24]). In Pakistan and the Seychelles the question was asked as part of the process of monitoring progress in the issue of national identity cards.

In a number of countries with a large expatriate population, identifying the citizenry was a particular concern. Thus in Singapore, where a fifth of the population is non-resident, the government chose to tabulate most results for the resident population only, thereby preserving the balance between the three main citizen communities (Singapore [58]). In the case of dependent territories, the census sought to establish whether the inhabitants were 'belongers' or 'islanders' (Turks and Caicos Islands [66]; Saint Helena [56]). In some of these territories, the residents are offered the choice of a local 'national status', for example, as Anguillians or Gibraltarians in addition to their legal status as British citizens (Anguilla [4]; Gibraltar [22]). The legal basis of such a status might be acquired 'by birth, by parentage, by naturalisation or by reason of having lived here for so long that they have become one by association' (Turks and Caicos Islands [66], p. 58).

The Kenyan census posed the question along slightly different lines, asking: 'What is this person's tribe?' and if not Kenyan: 'What is this person's nationality?' (Kenya [31], p. 358). Thus Kenyan Asians and Kenyan Europeans became effectively two more tribes, while a person of Asian descent holding a British passport became British and was grouped with European nationals! Other countries adopted similar combined questions, thereby saving space on the census schedule. It is worth noting that the conflation of nationality, race, and ethnicity had been widespread in the nineteenth-century census terminologies. The more precise modern legal term contributed comparatively little to the quest for what the individual felt was his or her identity.

The most controversial topic is that relating to race or ethnic group, a topic inherited from colonial anthropological studies (Stepan [63]). Sometimes the terminology differed little from that used in colonial times (Hirschman [26]; Christopher [15]). This question was asked by more than half the censuses and involved some complex racial and ethnic classifications. Furthermore, the censuses in Malaysia and the UK both adopted three different classifications for different parts of the country. Most systems had a racial basis and such terms as the following were widely adopted: White, Black, African, Asian, Indian, Chinese, and even Mixed. Dimensions of a more cultural nature were reflected in the addition of place-specific descriptions, such as: Portuguese, Mennonite, Syrian/Lebanese, Irish Traveller (Table 2). Many of these groups included people whose origins were diverse but who had been linked for census purposes. Thus in Fiji, 'Fijians' included Melanesians and Polynesians while 'Indians' shared only a common origin in the subcontinent (Fiji [20]). The countries of the Caribbean region had adopted a fairly standard classification for the 1946 post-war emergency census and have continued to use it, with minor individual modifications ever since. Elsewhere comparability was limited, as was demonstrated by a comparative study of the Australian and New Zealand systems (Treliving [65]). Furthermore, even where the classification did remain static, the subjective nature of self-identification could lead to change through time (Platt et al. [53]). The reported changes in the ethnic balance within the relatively stable population of Dominica is an example of where the use of time series must be treated with caution (Table 3). The debate associated with the concept of 'blackness' has a bearing here because the census is concerned with self-appraisal rather than biology (Zuberi [68]; Brunsma and Rochquemore [11]).

Table 2  Examples of census questions on ethnic group in the millennium round of Commonwealth censuses

Table 3  Percentage distributions of population of Dominica by ethnic group 1946–2001

Where ethnicity is the major element in national politics, the balance between communities is carefully monitored and controlled. Few went as far as the South African government between 1950 and 1994 in policing racial boundaries through legislation, but race or ethnic identification was and is regarded as essential for the pursuit of national policies in a number of countries. Hence, the revised classifications adopted in Malaysia since 1991 sought to realign with that of the mainland the classifications used in Sarawak and Sabah and Labuan dating from the colonial period. The object was to recognize who was an indigenous, 'Bumiputera' for purposes of affirmative action and national development (Malaysia [37], p. xxvi).

While similar pieces of legislation in a number of countries, such as the UK, Canada, and South Africa, were enacted to ensure that monitoring took place, it was sometimes difficult to ensure that people from such groups identified themselves correctly for the purposes of the legislation. In Canada 'visible minorities' were identified as requiring assistance (Canada [12]). However, it was noted that people identifying themselves as White, 'might be correct within self-description' but that 'information was required as they had been designated visible minority groups for Employment Equity purposes' (Boyd et al. [10], p. 44). In Australia the question was reduced to the identification of people of Aboriginal and Torres Islander origin only. Similarly in India only members of Scheduled Castes and Scheduled Tribes were identified separately in the census (Sundar [64]). In contrast, in New Zealand, after a basic ethnic–race question, New Zealand Maori were required to give additional information on their tribe (iwi). In South Africa, in the interests of equity, classifications of the previous apartheid-era groups were preserved as a means of monitoring progress. Thus in order to remove any ambiguity, it was specified that the classifications should identify 'A group with common characteristics (in terms of descent and history), particularly in relation to how they were (or would have been) classified before the 1994 elections', thereby ensuring that the children born since the end of apartheid were classified in like manner (South Africa [60], p. 12). Furthermore, the small numbers of those who opted for the 'Others' box on the census questionnaire were reclassified according to the apartheid system (South Africa [59], p. 4).

It is notable that a number of countries did not ask a question about race or ethnicity. In some cases this was because another identifier, usually language or religion, had been substituted for the question. Elsewhere, particularly in Africa, the absence of the question was a symbolic rejection of the colonial past. In Namibia, for example, the apartheid racial classification was eschewed and instead the census asked about language, thus cutting across traditional ethnic lines (Namibia [43]). In other countries, such as Mauritius, the question was rejected in order to foster national unity (Nirsimloo-Anenden [45]). In Nigeria census-taking is particularly contentious because population numbers affect the regional balance of power. That is why the millennium enumeration was postponed until 2006 and the proposed questions on ethnicity and religion were excluded—thus avoiding the use of material that could be perceived to be separatist as a means of achieving a generally acceptable result. In other cases (such as Bangladesh and the Falkland Islands) the ethnicity or race of the population was regarded as sufficiently homogeneous to warrant not undertaking any such investigation into what constituted a small minority. On the other hand, on Norfolk Island an enquiry into the extent of 'Pitcairn descent' was undertaken, though no other ethnic information was sought.

An added complication in the enquiry into ethnicity was the acceptance, even encouragement, of multiple entries on the census returns. New Zealand census questionnaires had, from the early twentieth century, sought to deal with people of mixed ethnicities by seeking the proportion of each ethnicity represented by an individual (New Zealand [44]). In 1986 this practice was abandoned in favour of asking for more than one option to be chosen from the list of proffered ethnicities (Lang [33]). In 1991 some 9.0 per cent of the population marked two or more options. The complex problems of interpreting multiple entries have been examined in the USA for the 2000 census (Perlmann and Waters [52]). Most censuses retained the instruction to mark only one option.

Language was the least frequently asked of the four indicators of identity, probably because of the pervasiveness of the English language within much of the Commonwealth. However, language was one of the major areas of enquiry for colonial administrations in Africa and Asia and so it remains, although the language and ethnic-group questions were often interchangeable. Within the UK it was the survival of endangered peripheral languages that attracted attention. Thus in relevant parts of the country, the questionnaires included questions on the ability to speak and understand the Welsh, Scottish Gaelic, Irish, Manx Gaelic, Guernsey Norman French, and Jersey French languages. Except in Jersey no further enquiry was made and in England no question at all was posed.

Questions on literacy were often of little assistance because literacy 'in any language' was the usual (footnoted) requirement. However, those only literate in Turkish written in Arabic script were regarded as illiterate by the census authorities in Northern Cyprus (Northern Cyprus [48]). Sometimes literacy in specific, often official, languages was investigated (Sri Lanka [61]; Malta [38]; Mozambique [42]). The Gambia census questionnaire asked whether a person could read the Latin and Arabic scripts, but not the specific languages involved.

Interest in language shifts was evident in some countries. For example, in Belize information was requested not only on the individual's ability in English and Spanish, but also on the language spoken at home, and the language spoken at home as a child. In Mauritius the enquiry went further and asked for the language of the individual's ancestors, as well as that usually spoken in the home. This was a way of avoiding an ethnic question, while still identifying the population of Indian origin (Mauritius [39]). In recognition of the linguistically complex society in Papua New Guinea, the census questionnaire enquired into which language was spoken at home, and which was used in the marketplace. However, when home, business, and official languages differ, multi-lingualism is a remarkably difficult ability to assess in the census.

The political aspect of language use was evident in some countries. In Canada the relative popularity of French and English has exercised minds since Confederation. In India the major linguistic groups protected their rights within linguistically based states. In a few other countries the promotion of the national language was monitored and progress noted. The use of Malay in Malaysia or Swahili in Tanzania was not monitored in the censuses. In South Africa the nine official languages were enumerated, but others, whether of small indigenous groups or of other immigrants, were ignored (Maartens [34]). The phrasing of the language question might not provide a complete picture. A 1997 survey in India concluded 'contrary to the census myth that English is the language of a microscopic minority, the poll indicates that almost one in three Indians claims to understand English, although less than 20 per cent are confident of speaking it' (Ostler [50], p. 576).

The censuses varied in what they sought and revealed about the emergence of distinctive forms of Creole on the Caribbean and Indian Ocean islands and Pidgin in the Pacific Ocean. Creole was recognized as a separate, and indeed majority, language in Mauritius, with only small numbers reporting French and English as the 'home' language. The matter was not investigated in neighbouring Seychelles. In the Pacific, in such countries as the Solomon Islands and Papua New Guinea, Pidgin was recognized as a separate language, while in others, such as Tonga and Fiji, it was ignored on the assumption that the population used the English taught in the schools. In the Caribbean, Creole was not recognized as a separate language; it appears to have been generally regarded as slang or a variant of English (Christie [14]; Pyne-Timothy [54]).

Religion is the most widely asked identity question in Commonwealth censuses, though, as noted in Ghana 'no attempt was made to find out if respondents actually practised the faith they professed' (Ghana [21], p. ix). Nevertheless, the Pakistani census commissioner noted that 'The Pakistani are a deeply religious people' (Pakistan [51], p. 39). The religious questions were probably included out of curiosity in some cases, and unlike other questions, a response was not always compulsory. In Botswana the delegates to the national dissemination seminar on the 1991 census requested that language and religion be included 'even though the two might not have any direct benefit or contribution to socio-economic developments in the country' (Botswana [9], p. 423). Questions on both were included in the 2001 census. Where religion is regarded as the basic identifier of the nation, as in Bangladesh for example, it is monitored carefully (Table 4).

Table 4  Percentage distribution of population of Bangladesh by religion 1901–2001

The classification systems used to collect information on religion varied substantially. In the UK census the question on religion was phrased differently in the constituent parts of the country (Great Britain [23]). The English questionnaire was the least demanding of detail, thereby avoiding any implications for the constitutional controversy over the possible disestablishment of the national Church of England (McLean and Linsley [40]). A general category of 'Christian' covered all denominations. In Scotland on the other hand, adherents of the Church of Scotland and Roman Catholics needed to identify themselves separately. In Northern Ireland an even wider choice of denomination was offered. Furthermore, in both Scotland and Northern Ireland there were questions about the religion within which the individual had been brought up. In Northern Ireland the balance between the Protestant and Roman Catholic communities was the subject of rigorous statistical enquiry (Northern Ireland [49]).

It should be noted that the majority of questionnaires offered 'None' as the first option for religious affiliation, suggesting the expectation of a post-religious society. The inclusion of the option 'Others' permitted 390,000 people in England and Wales to identify themselves as 'Jedi', as the result of an e-mail campaign urging people to support this 'religion', which had been created by the 'Star Wars' films. The Jedi phenomenon was repeated around the world—even in Norfolk Island three people identified themselves as Jedi (Norfolk Island [47]).

In some countries, for example New Zealand and South Africa, attention was directed towards identifying support for the indigenous churches, which retained significant elements of indigenous belief systems. Those reporting affiliation to Christian denominations tended to be analysed in some detail, but this was not the case for other religions. Muslims were treated as a homogeneous block in all censuses, although cleavages within Muslim societies might be as significant as those distinguished within Christian communities.

The picture that emerges is one of considerable diversity in the questions asked about identity, which is scarcely surprising given the absence of any central direction of census-taking in the colonial period, and the complex national experiences that followed it. The individual state was able to control the questions asked and the systems of categorization adopted, and could then interpret the information gathered to promote its own viewpoint. If the census was an element in the creation and development of an identity, then it was at the level of the modern post-colonial state.

Motives for including identity questions in the census questionnaire have ranged from local legislative requirements for electoral or equity monitoring purposes, to popular and academic curiosity about the composition of local society. Questions on nationality or citizenship were used widely, as a means of numbering the national population, controlling participation in the political process, and excluding 'others' in recognition of the state's primary responsibility to its own nationals. A question about ethnic identity, the most controversial of the identity issues, was a feature of many censuses, but its form was peculiar to each state, or even sub-national community. It also probably yielded the least reliable of the statistics on identity, since it was rarely open to verification or indeed posed against a universally agreed set of definitions. Nevertheless, significant cleavages within the nation were explored and statistical material gathered for future use, more particularly in monitoring various affirmative-action and equity programmes or projecting possible election results. Questions on religion and 'home' language—key aspects of national cohesion and indirect indicators of identity—were widely asked, not surprisingly in view of their functions in nation-building.

In some cases it is the questions that were not asked which are of significance. For example, none of the following identities were investigated: English, Roma, Yoruba, and Quebecois. The same applies to Wahabi and Shia religious affiliations. Evidently these ethnicities and affiliations were regarded as being too politicized or too personal for objective enquiry and therefore as phenomena more suitable for opinion polls or to be demonstrated in elections. Similarly most issues of caste in India were regarded as too contentious to be investigated by the kind of probes used pre-independence. Yet these are core concerns of the populations concerned and their absence from a census often reflects a government's unwillingness to recognize the problems confronting its society.

The questions on identity posed in the Commonwealth censuses provide a vast amount of data that can be used by governments, academics, and others to monitor or promote particular developments. In the process of categorizing identity and involving the population in revealing their affiliations, the censuses may help to promote or modify contemporary views of the nation. Nevertheless, most of the Commonwealth censuses present only a part of the total picture, and not the general illumination sought and attained in a limited number of cases. The latter show the potential that exists for extending the usefulness of censuses in future.[1]