UN Census “Households” and Local Interpretations in Africa Since Independence

Since the 1950s, the UN Statistical Division has encouraged nations to standardize the definitions used in data collection. A key concept in censuses and surveys is the household: This is the unit for which information is collected and analyzed, and is thus an important dimension of data that are the basis for many policies. We aim to understand the tensions between conformity with UN guidelines and national priorities. We analyze the documentation around the UN household definition over this period. Using detailed census and survey documentary data for several African countries, especially Burkina Faso, Senegal, Uganda, and Tanzania, we examine the disparities between national census definitions of “household” and the UN definition. Perspectives from interviews with key informants within national statistical offices demonstrate the variability in the importance accorded to the UN harmonization aims and the problems that arise when these standardized approaches interact with local norms and living arrangements.


Introduction
Collecting and analyzing statistical data on national populations is a key dimension of being a modern state. Regular national data collection exercises with the production of reliable and valid data are one way in which nations signal their membership of a global community (Barrett & Tsui, 1999). Increasing use of metrics to measure "development" and "progress" toward achieving targets such as the Millennium Development Goals means that statistical data are becoming more important although much "data" produced by international organizations such as the World Bank and the International Labour Organization (ILO) are, in fact, guesstimates or extrapolations (Duncan, 2013;Jerven, 2013;Sanga, 2013).
Since its establishment in 1947 under the UN Department of Economic and Social Affairs (DESA), the UN Statistical Division (UNSD) has been concerned with the systematic organization and compilation of country-level statistics and indicators (UNSD, 2013). The UNSD has encouraged postindependence African states to standardize and streamline their data collection and has provided definitions and guidelines to be used in data collection and training (Ching'anda & Ntozi, 1998). Key themes emitting from the UNSD are the development of National Statistical Systems (NSS), regular data collection, and improving and standardizing data collection to facilitate comparability.
A key concept in censuses and surveys is the definition of household; this determines the units for which much data are collected and analyzed, and thus influences the data that are the basis for many policies. However, tensions emerge between UN guidelines about standardization of household definitions and local residence patterns and social organization; these are manifested through different national emphases, such as the reduction of the UN phrase "joint provision of food and other essentials" to eating together out of the same cooking pot, or the addition of being subject to the authority of a single household head.
This article has three aims: (a) to establish the extent to which the UN guidelines influence national data collection and how this has changed over time, (b) to identify key dimensions of the UN household and how these are interpreted and implemented by nation states, and (c) to reflect on national motivations for (non)compliance with UN guidelines. Throughout the article, we consider the role of comparability in the evolution of definitions.

Method
Two research methods and data sources inform this work. We review UN and national documentation on definitions and concepts along with survey and census enumerators' manuals in diverse African countries. We then focus on the definition of the household and the collection and management of household-level data in analyzing the relationship between UN guidelines over the past 50 years and actual practice in two Anglophone (Tanzania and Uganda) and two Francophone (Senegal and Burkina Faso) African countries. Further insight is provided by in-depth interviews in the same countries with individuals in different positions within the hierarchy of National Statistical Offices from retired and contemporary senior personnel to census and survey enumerators (for more detail, see www.householdsurvey.info).

UN Influence on Data Production
Most census and many household survey reports acknowledge technical advice and help provided by outside agencies (e.g., United Nations Population Fund [UNFPA], U.S. Bureau of Census, United Nations Economic Commission for Africa [UNECA]), although it is rarely made explicit how these relationships work in terms of decisions about definitions and their operationalization. References in census documentation indicate the importance of conforming to UN principles and this is a clearly articulated dimension of national modernity.
The Tanzanian 1967 census report stated, The census was conducted according to modern scientific principles as summarized in recent recommendations by the United Nations and its Economic Commission for Africa. (United Republic of Tanzania, 1969, p. viii) The introduction to the 1975 Burkinabe (Upper Volta) census report states, Undertaken within the framework of the African Census Programme, [the census] fulfils the principal requirements of the United Nations Statistical Commission (p. 5) . . . it forms an important step towards the development of a modern system of data collection (p. 6). (République de Haute-Volta, 1978, pp. 5-6, our translation) After a detailed history of colonial censuses, the introduction to the 1960 Ghana census report emphasizes, The 1960 Population census of Ghana is a modern census carried out according to the principles and recommendations laid down by the United Nations. (Republic of Ghana, 1962, p. xi) The Ghanaian administrative report reporting on the 1960 census states, The concepts and classifications used in the Census were largely based on international recommendations. Adaptations were made to suit local conditions. (Republic of Ghana, 1964a, p. 112) Frequent references to the United Nations in Ghana's documentation as far back as 1960 demonstrate the strong awareness of the UN principles and definitions as these guided the evolution of census data collection. But, because of problems in making these definitions workable locally, there are many adaptations. The Ghanaian census bureau's move from collecting data for occupants of "houses" in 1960 and 1970 to "households" in 1984 is evidence of the influence of the international agenda and its concepts.
In general, Anglophone African censuses conducted late in the colonial era and early in independence, although dependent on outside funding and advice, were grounded in detailed knowledge about local conditions and social organization. The guiding principle was to get the most accurate census count (avoiding omissions and double counting through a de facto 1 approach) and using local vocabulary to define enumeration units. The census was primarily presented as a national affair and part of postindependence nation-state building.
Despite increasing impetus toward harmonized global guidelines and frameworks for statistics (International definition and measurement of standards and levels of living [UN, 1954b], Framework for Social and Demographic Statistics [UNSD, 1975]), it was not until the 1980s and 1990s, when the Human Development Index was first produced, that data produced by censuses and surveys really became international goods: This may have generated greater pressure for individual countries to conform in terms of concepts and definitions. Other non-UN international surveys, such as the World Fertility Survey (late 1970s) had focused attention on the comparative power of harmonized data. Recently the demand for social statistics and indicators has grown significantly (see UNGA, 2014), in part due to the need to monitor progress toward goals agreed at international summits (e.g., 1990World Summit for Children, 1994International Conference on Population & Development, 1995World Summit for Social Development, 1995. Because household surveys and censuses are central to the production of data, much UNSD work has focused on the production of methodological guidelines (DESA, 2005(DESA, , 2008a(DESA, , 2008b.

Comparability
The 1954 UN Handbook of Population Census Methods highlights the key role of comparability: For the purposes of international comparability it is desirable that a de facto enumeration be made; that is, a count of all the persons present in the country at the time of enumeration. Any data on a de jure basis which may be desired should be obtained in addition to the de facto data. (UN, 1954a, p. 37) Most Anglophone censuses followed this de facto approach and some census instructions emphasized the importance of the de facto enumeration over other concerns as highlighted in the Ugandan enumerator's manual for 1969.
A household is defined as a group of persons who normally live and eat together. This is a very loose definition and there may be many cases when you are in doubt as to whether people should be included in the same household or shown as belonging to separate households. It is not possible in these instructions to cover all such cases in detail, and your decision in such cases should be determined by common sense and convenience in the enumeration. It is not a matter of great importance whether or not such persons are included in one household or shown belonging to separate households. The important thing is that every person should be enumerated. (Republic of Uganda, 1974, p. 87, emphasis added) "Comparability" is a dominant theme throughout UN documentation, and it tends to take precedence over other considerations such as local applicability of concepts. In terms of the ways in which these influences affected practices in National Statistical Offices, the emphasis from the United Nations is clear that they should co-operate in the design of standard and uniform procedures for sample surveys to obtain better indicators of levels of living. (United Nations, 1954, cited in United Nations, 1964 and from the same report: Discussion of the value of household inquiries, especially in developing areas, was carried a step further by the Working Group of Experts on Family Living Studies convened by the International Labour Office in 1955. This group recommended inter alia that the international agencies should aid in the development of sound methods of study and encourage international comparability by issuing lists of standard definitions and classifications to be used in household enquiries. (United Nations, 1964, p. 2) The impetus from the United Nations toward comparability and standardization over time is clear and focused on countries in receipt of funding and technical support for statistical data collection. The comparability of statistics is an important part of demographic training and our analyses of discussions with higher level personnel within the Tanzanian statistical office and international organizations make it clear that preoccupations about comparability often overruled other considerations about the validity of demographic data (Randall, Coast, & Leone, 2011). However, although comparability-over time within countries and over space between countries-is frequently talked about, in practice, there are significant deviations from this ideal.
Census enumeration units have become more standardized over the last 50 years and most countries now use "household" rather than another unit such as dwelling or family. For example, Gambia changed from the use of family/yard in 1963, to household in 1983. Malawi moved from using the dwelling unit to a household definition between 1977 and 1987. South Africa moved from using the family to the household between 1985 and 1991.
However, although the term household is now ubiquitous, the way it is locally defined and interpreted still varies considerably.

UN Concept of Household
The UN documentation on household definition is extremely consistent over time. In 1959, there were discussions about two different approaches to household: the housekeeping unit and household-housing unit concepts (United Nations, 1959), reproducing recommendations from the UN document "Principles and Recommendations for National Population Censuses": A private household should preferably be defined as: (a) oneperson household: a person who lives alone in a separate housing unit or who, as a lodger, occupies a separate room or rooms in a part of a housing unit but does not join with any of the other occupants of the housing unit to form part of a multi-person household as defined below; or (b) multi-person household: a group of two or more persons who combine to occupy the whole or part of a housing unit and to provide themselves with food or other essentials for living. The group may pool their incomes and have a common budget to a greater or lesser extent. The group may be composed of related persons only or of unrelated persons or of a combination of both, including boarders but excluding lodgers. (United Nations, 1959, p. 74) The key dimension here is "housekeeping," which revolves around "provide themselves with food." This phrase is itself ambiguous, as demonstrated by the example of two wives of a polygamous man: both obtain the grain for their meals from the family granary that is managed by their joint husband. The granary is filled with grain cultivated on fields "owned" by the husband and his lineage and worked on by his wives and children. The wives take this grain and cook it separately in different kitchens attached to their different houses in the same compound and then feed themselves and their children and each sends food to their husband. In this common case, if "provide themselves with food" refers to the source of food-the communal granary-then both wives, their husband, and any dependent children (and others) will form one household. However if "provide with food" is interpreted as being related to the cooking and eating of the food, thus focusing on the "cooking pot," each wife constitutes a separate household, and a somewhat arbitrary decision is made about which household the husband is assigned to.
Despite this ambiguity, the wording of the UN definition of household for censuses barely changes over the next few decades. In 1980, The concept of "household" is based on the arrangements made by persons, individually or in groups, for providing themselves with food or other essentials for living. (DESA, 1980, p. 50) And in 1997, 1.324. The concept of household is based on the arrangements made by persons, individually or in groups, for providing themselves with food or other essentials for living.
1.325. The concept of household provided in paragraph 1.324 is known as the housekeeping concept. It does not assume that the number of households and housing units is equal . . . 1.326 . . . Some countries use a concept different than the housekeeping concept described in the previous paragraph, namely, the "household-dwelling" concept, which regards all persons living in a housing unit as belonging to the same household. (According to this concept, there is one household per occupied housing unit.) In the household-dwelling concept, then, the number of occupied housing units and the number of households occupying them are equal and the locations of the housing units and households are identical. However, this concept can obscure information on living arrangements, such as doubling up, that is relevant for evaluating housing needs. The definition of household most often used in national censuses conducted during the 1990 round of censuses incorporates both the housekeeping and household-dwelling concepts. (DESA, 1997, p. 50) In referring to both housekeeping and household-dwelling, the final sentence of this quote highlights the lack of clarity and consistency even in standardized approaches. All the definitions require cohabitation although the UN documentation demonstrates the subtle differences between households defined on coresidence alone and those based on housekeeping where the definition is ultimately is trying to capture an economic unit. However, as elucidated below, the housekeeping concept gets reduced in some contexts (particularly Anglophone East Africa) to cooking and eating together, which is then prioritized in national definitions. In such cases, culturally determined patterns of cooking and coeating come to be the principal defining characteristic of the household rather than having a common budget.
Interviews with statisticians and other individuals along the chain of data production and use in African statistical offices revealed that many see the UN definition of the household (and their own national interpretation of this) as an alien concept that has been developed for statistical and demographic analysis, under demands for comparability, rather than something that represents a fundamental and locally relevant social unit. It is a technical term that needs to be learnt and then applied to generate the "comparable" statistics required.

Local Understanding or Comparable Units?
Although the UN explanations clarify the difference between a housekeeping household and a household-dwelling, translating this into units of data collection may be difficult in the field. Furthermore, data collection in multilingual contexts requires ways for enumerators to explain to respondents the units for which data are being collected (Randall, Coast, Compaore, & Antoine, 2013). Hence, a tension arises between the UN household unit and actual living arrangements (Guyer, 1981;Guyer & Peters, 1987).
Households in early Botswana censuses were based on local knowledge and vocabulary about social organization, with an anthropologist cited as the source for the explanation of the enumeration unit.
The household is the smallest well defined social unit and, in Professor Shapera's words, "It consists basically of a man with his wife or wives, and their unmarried children, but often includes one or more married sons, brothers or even daughters, with their respective families." Every household has its own compound, known as a "lolwapa," consisting of one or more huts and a granary within a courtyard surrounded by a reed fence, a wooden palisade, a low earthen wall or something similar. In most cases therefore the household is an easily recognised physical entity and it formed the basic enumeration unit. It is the compound which is referred to as the "dwelling" and not the individual huts within it, and the people living within the compound are referred to as the household. (Republic of Botswana, 1972, p. 9) This 1972 definition makes no reference at all to housekeeping, provision, or consumption of food. By 1981, Botswana's approach to the household had moved away from local anthropological references and closer to the UN housekeeping approach.
In general those who live in a "lolwapa" or its equivalent should be shown as one household if they eat from the same pot. Otherwise they should be regarded as separate households. (Republic of Botswana, 1983, p. A15) Local vocabulary is retained but the issue of eating from the same cooking pot is introduced, with the interpretation of the UN's "common provision for food" becoming "eating from the same pot." Ghana's earlier censuses also grappled with how to apply a standardized definition of the household: For the unit of enquiry the household was proposed. But owing to difficulties of definition which enumerators were expected to encounter it was decided to record in the census individuals by house or compound and to use the household concept only in the PES [Post-enumeration survey]. It was realised that the house or compound may not necessarily correspond to particular economic or social concepts. (Republic of Ghana, 1964b, p. 112) Later the report does define "household" as used in the postenumeration survey (PES): The definition adopted finally in the PES, "a person or group of persons all living and eating together from the same cooking pot" was in fact, a slightly modified version of the United Nations concept adopted to suit the African social environment. (Republic of Ghana, 1964b, p. 326) Compared with the 1959 UN household definition of "provide themselves with food and other essentials for living," this 1964 Ghanaian definition is actually very different, and potentially a smaller unit-specifically talking about eating out of the same cooking pot-a phraseology that is never used in the UN documentation.
Having undertaken the 1960 and 1970 census using houses as the unit, in 1984, Ghana moved to households and housekeeping.
A household was defined as follows: "a household consists of a person or group of persons who live together in the same house or compound, share the same housekeeping arrangements and are catered for as one unit. . . . For instance two brothers who live in the same house with their wives and children may or may not form separate households depending on their catering arrangements. . . . A usual member of household was considered to be any person who, whether present or absent on Census Night has spent (i.e. lived together in the same house or compound, shared the same housekeeping arrangements and been catered for as one unit with the other members of the household) at least the last 6 months with the household." 2 (Republic of Ghana, 1984, pp. xiii-xiv) This definition (and that of the 2000 census) uses "catering arrangements" rather than cooking pot and is close to the UN definition.
Sometimes, the interpretation of the UN guidelines remains ambiguous. For example, in the (de facto) 1987 Malawian census, A household consisted of one or more persons, related or unrelated, who make common provision for food and who regularly take their food from the same pot and/or share the same grainhouse (nkhokwe) or pool their incomes together for the purpose of purchasing food. (Government of Malawi, 1987, p. 9) The same definition was used in 1998 and in the census report on households and household characteristics it is referred to as "the UN definition" (Government of Malawi, 1998, p. 120). Clearly, the UN vocabulary is seen as an important guiding role. However the inclusion of "and/or" makes this a more inclusive unit than a "cooking pot" interpretation of "provide themselves with food." In most cases, we can discern the ways different countries manage to combine UN guidelines with local social organization through instructions within manuals or comments in reports. In Tanzania, a key-informant interview provided a clear insight into work undertaken to simultaneously integrate local vocabulary with the requirements generated by comparability and conformity with the UN concepts. So when we, at NBS (in mid 1970s) when we sent and we discussed this in meeting and we said well, we now have to look for a word in Kiswahili-there were suggestions more than one-as usual-we said well we have the National Kiswahili Council and we have the Department of Kiswahili at UDSM [University of Dar Es Salaam]. We shall send them the definition of the household as we know it from the UN now we shall ask them to suggest what is it the Kiswahili equivalent that would fit that UN definition, that long thing . . . they also came up with the kaya. Kaya is the arrangement that best suits that definition of the household from the UN. (Senior retired Tanzanian statistician/demographer) In Tanzania, with its national language and a clear desire to follow UN requirements, there was a deliberate move toward a particular word based on a well-articulated piece of research undertaken to specifically identify the best local term. However, in many contexts (including multilingual Tanzania), there are real problems matching the UN concept onto a slightly different concept or word that already exists in a local language (Randall et al., 2013), as shown here for Burkina Faso.
It's very difficult [in Fulfulde] to find a word like that, but, well, following the definition which has been agreed before the fieldwork, we are forced to explain it. On top of that you add extra things. For example if we say that they have to pool the results of their production, to translate that into Fulfulde . . . [in the field] when we go people often say, well there are the old men, but we, we have our definitions which are there and we are forced to say even though the old man is there that doesn't mean to say that we can't have different households . . . the definitions are there to respond to needs, it's a standard definition that's there. (Burkina Faso; statistician and former Institut National de la Statistique et de la Démographie [INSD] enumerator) This tension is reiterated by a UNFPA advisor But the majority of these surveys they get the definition from [the statistical office] and they try to use it. Unfortunately what happens is, where the respondents have their own perceptions and also the enumerator they have their own perceptions. A lot of the data that we get in this part of the world is indicative of this question, it's not quite perfect information because of this confusion. (UNFPA advisor, East Africa) Where the UN definition is adopted by the statistical office but there are no clear ways of translating it into local languages, then the definition needs to be broken down into its constituent parts and explained to both data collectors and respondents. It seems to be these explanations of the UN definition in recent years that have led the definitions down a route we identify as the "cooking pot" pathway. The simplest interpretation of arrangements . . . for providing themselves with food or other essentials for living is to say "those people who eat out of the same cooking pot," which in some, but not all, contexts is the group who also provisions together to fill that cooking pot. This emerges clearly in Uganda.
A household has got a standard definition. We look at two elements to define a household. The first one-actually the most important-is the eating area. People must be dining together. They may live together but as long as they are not feeding from the same pot, then those ones are different households. (Uganda Bureau of Statistics [UBOS] statistician, Uganda) However, eating out of the same cooking pot is only a very limited interpretation of the UN definition and one that moves away from the original goal of a definition attempting to capture society's basic economic consumption unit. There may be culturally prescribed patterns of cooking and eating together-such as in polygamous populations where every wife cooks in her hut for her children, yet the economic unit of production and consumption is much wider and would put all those wives in one "household." In other contexts, the distribution of a very large household around several cooking pots may be purely practical but would lead to several census households.
Over time, the guidance from the UNSD has become more detailed although not necessarily with clear explanations about how to resolve perennial practical issues. The tensions between the statistical definition of the household and its operationalization in field data collection are well established: While the household concept has not been widely contested as a consumption unit, questions have been raised regarding its meaning as a production unit or income generating unit. The main argument is that persons living in the same housing unit who together make provision for food and other essential items may not necessarily pool their income or make decisions jointly regarding their economic activities. Various situations may arise in different societies. For example, in many African communities an extended family comprising several households may own and cultivate a field together, while cooking and housekeeping arrangements are still made separately by each household level. The consumption unit may also include persons who do not reside with the household although they regularly take their meals in common. The usual concept of household may therefore require considerable adaptation or elaboration in order to be applied consistently in particular societies. (UNSD, 1984, p. 99) The final sentence is informative about the problematic relationship between UN guidelines and local implementation of them because no advice is given about how this adaptation might be achieved while still maintaining comparability. Most countries have ignored this UN awareness of the complexity of African households and oriented their definitions around the minimal group of those who live, cook, and eat together.
The 1997 Principles and Recommendations for Population and Housing Censuses (DESA, 1997) dedicated two pages to explaining the concept of the household, its different dimensions, and how these things should be recorded. The explanation still commences with "[those] who make common provision for food or other essentials for living" (DESA, 1997, p. 65) but then develops a wealth of detail and clarification. Interestingly, a key concept in many national definitions, that of eating together out of the same pot, is not mentioned.

Case Study Countries: Diversity in Evolution of Definitions
Our comparative case study countries were selected because they experienced different colonial histories and postcolonial political ideologies, with contrasting geographic and linguistic settings. Yet all are members of the United Nations and all have invested considerably in statistical development. We aim to establish where there are clear temporal trends across the countries in terms of their relationship with the UN definitions and guidelines, which remained constant throughout the half century with "joint provision of food or essentials of living." Table 1 summarizes the dominant criteria of each national definition over time and is based on detailed extracts from census documentation (see the appendix).
The documentation for the 1975 Burkina Faso (then Upper Volta) census identifies compounds and households with a somewhat confusing overlapping use of "family" (République de Haute-Volta, 1978). Households were enumerated within compounds and a household could only include one married man with his wives and unmarried dependants. There is no mention of provision of food or housekeeping. In all subsequent censuses (1985,1996,2006), Burkina Faso follows some aspects of the UN definition closely by requiring co-residence and common provision of food and daily life, but it also requires the recognition of a household head and that a household cannot contain two married couples. Both these conditions diverge from the UN guidelines, and the latter hinders comparability with other countries. The report on the Burkina 1985 census specifically recognizes: As it is defined here [i.e. one married couple per household], the household loses its meaning as an economic and production unit. The notion of household has principally been used in the census as a control variable to ensure exhaustive enumeration. (INSD, 1989, p. 284) The 1985 census report tables on données collectives [group data] present the characteristics of compounds and compound heads, with just one table on "households per compound." However, in 1996, although the definitions remain identical, and despite the 1985 proviso above, the données collectives focus on characteristics of households and their heads and the report section starts with "The household constitutes the basic socio-economic unit" (INSD, 2000, p. 271). Tanzanian censuses are de facto, and the 1967 census followed the UN definition closely. In the 1970s, those living geographically close but in separate houses could be part of the same household if they eat together, thus prioritizing the cooking pot. In 1988 and 2002, there is no mention of eating together, but shared living costs are the criterion for household membership. It is unclear how "shared living costs" could possibly be operationalized in a de facto census and the enumerators we interviewed interpreted the guidelines as meaning eating together.
Uganda's de facto censuses utilize a restricted interpretation of "joint provision of food or essentials of living" summarized as "eating together." This cooking pot dimension is prioritized by all our Ugandan key informants.
Senegal is totally different from both Anglophone Tanzania and Uganda, and other Francophone countries, with clear definitions that prioritize what are seen to be Senegalese characteristics. Pilon and Vignikin (2006) show that in all comparative analyses (census or surveys) Senegal has substantially larger households than any other African country; this is probably a consequence of their approach to definitions. In 1976, Senegal avoided the problems posed by the household by avoiding the concept altogether and de jure enumerating compounds (concessions) and their constituent family nuclei (noyaux). In 1988 and 2002, Senegal abandoned this approach for the household (ménage), defined as living together in the same compound, eating together, and under the authority of one head of household: subsequent clarifications indicate that living together takes precedence over eating together. Clear examples of the enumeration units were given by providing local words (see the appendix), which almost certainly means that during enumeration these local words were used in preference to any precise definition.
Each country has taken a different route to reconciling local conditions with the UN definition. Whereas Tanzania made huge efforts to conform to the United Nations, Uganda concentrated on a particular interpretation of UN "joint provision" by focusing on the cooking pot. Senegal has largely remained detached from UN recommendations apart from changing from concessions and noyaux to "households." By retaining the importance of the household head and the use of local words, it is clear that they are prioritizing local organization rather than international comparability. Burkina Faso complied with much UN wording but added local priorities about married couples.

Census: Household Structure and Relationships Within Them
Early census data were collected on people within "households" because that was the most effective way of enumerating the whole population (Republic of Uganda 1974, appendix).
Relationships within households, and household structure, have become an increasingly important dimension of household data collection because of their importance for understanding support and welfare. UN documentation provides detailed guidelines for household composition data collection but national censuses oscillate between following UN guidance and trying to cope with making these data meaningful locally. The practical problems of recording standardized relationships within the statistical household are well established: Traditional kinship systems especially in developing countries, may permit several interpretations of "mother," "brother," "sister," "wife," "widow" and other kin and therefore, special knowledge is required in order to translate data based on these relationships into internationally comparable form. (United Nations, 1964, p. 33) UN advice acknowledges challenges of application in the field while exhorting the need for comparability.
After identification of the head or other reference member of the household, each of the remaining members of the household should be distinguished in relation to that person, as appropriate, as one of the following: (a) spouse, (b) child, (c) spouse of child, (d) grandchild or great-grandchild, (e) parent (or parent of spouse), (f) other relative, (g) domestic employee or (h) other person not related to the head or other reference member. Where this classification is considered too detailed for successful collection of the information, categories (e) and (f) may be consolidated as Other relative and (g) and (h) can be consolidated as Other unrelated person. (DESA, 1997, p. 66) Collecting data on household structure via relationship with the household head assumes that household head is a valid concept and that individuals are members of that household through their relationship with that person. This may generate problems of coherence in de facto data collection when the household head is absent or for de jure approaches, when the recognized household head has been absent for longer than the residential cutoff (often 6 months). By only allowing a limited number of relationships with the household head, locally coherent household structures can be obscured.
This limited classification may work well in populations where most households are small and constituted of nuclear families or their close derivatives. Its ability to represent the diversity and complexity of many African households, let alone contribute to understanding how support is provided for the socially or physically vulnerable (a stated aim of the data collection), is limited.
Whereas the importance of understanding changing household composition was recognized in early UN documentation, approaches to this issue have not really been revisited in Africa in the light of new computer technology and the possible inappropriateness of the categories. Relationships could now be recorded in more meaningful ways to include relationships between different household members while retaining comparability by simultaneously recording relationships to household head. The potential comparative analysis of household structure from the different codes in recent censuses (Table 2) is very limited.
Not every country follows UN guidelines. Ghanaian reports on the 1970 census repeated earlier concerns that the UN relationships did not match well onto African usage (Republic of Ghana, 1975, p

. xi):
The conventional relationship titles which are so deeply rooted in African society had to be avoided so as to make analysis of the household pattern meaningful. (Republic of Ghana, 1964b, p. 327) This led to the development of a large number of detailed codes such as "mother's brother's son/daughter." In 1970, because of the de facto enumeration, temporary heads (99) had a different code to head (11). There were 14 relationship codes for people in the house who were relatives of the head-and a code for those who were relatives of the head's spouse (Republic of Ghana, 1975, p. xiv).
Other deviations from UN guidelines are observed in Kenya. In 1962, the census form instructs "relationship to head of household." In 1969, this has become "relationship" and in 1979 "what is the relationship of this person to the head of household or other members of the household?" The enumerator instructions state that  Kenya, 1981, p. 19) These confusing instructions are followed by indications that they should write things like "daughter of 4" using line numbers. In 1989, Kenyan relationship codes reverted back to a precise repetition of the UN guidelines to record relationship to the head of household (Republic of Kenya, 1989). Uganda's 1991 census questionnaire also asked for relationship to head or other member of household. It is not clear why these more flexible and informative approaches have been abandoned, but it might be the influence of international standardization and comparability.
Yet again, Senegal demonstrates its independent approach to data collection. In 1976, people were enumerated in compounds and family nuclei within the compound. It is explained thus: The familial nucleus (noyau) is the smallest possible family cell that can exist and is constituted by the spouses (or one spouse) and their directly descended unmarried descendants who must live in the same compound. By extension, the same family nucleus can include a husband with several wives and their unmarried children living in the same compound and direct ascendants (mother of the head of the family nucleus), brothers and sisters, and close unmarried kin (nephews, nieces, uncles, etc.) on condition that they live with the head of the familial nucleus and do not have their own unmarried children in the compound. This is identical to the unit called "household" in the Burkinabe censuses.
Although this notion was abandoned for the subsequent two censuses, it was reinstated in the 2013 census where the enumerator's manual states, The familial nucleus corresponds to the "biological" family. It is made up of the parents (or one of the parents) and their unmarried/ unpartnered direct descendants (biological children). Thus a household can be made up of one or several familial nuclei. Note that a polygamous household which includes unmarried children makes up one single nucleus if all the members live and take their meals together in the same compound. A nucleus can also include direct ascendants, brothers and sisters, uncles and aunts, grandchildren, unmarried nephews and nieces who are supported by the head of the nucleus.
Enumerating household members depends on the principle of the closest kin link. The household head is the first person to record on the questionnaire. Then you enumerate close kin of the household head before moving onto distant kin and those with no kinship link with him, keeping track, where possible each person's membership of a specific familial nucleus. (République du Sénégal, 2013, p. 51, our translation) Key informant interviews indicate that many Senegalese researchers and statisticians consider that this idea of nucleus and the relationships within it captures the essence of African familial and household structure.
It [familial nucleus] is an excellent thing, especially for censuses. It helps us avoid many errors, because the concept of noyau allows you, when you are in a household, which is usually polygamous, to be certain that you have first identified all the biological children for each wife and all the other people who have no biological links with the household head. And there is the advantage that, when you are with a polygamous couple you can, for each wife, identify her biological children and the children who are related by more distant kinship links. So if you do that you can be sure that you haven't omitted a single person, because when you are interested in the biological family what is certain is that there is a strong chance no-one will be left out . . . . It's a way of checking, but also for analysis, it allows you in some way to have a good understanding of the exact composition of the household. But particularly for data collection this approach ensures exhaustivity. (Statistician: Agence Nationale de la Statistique et de la Démographie [ANSD]) Senegal has also used "noyaux" in surveys and has thus approached the problems of enumerating complex African families and their residential and economic arrangements through local management of data collection rather than following the UN guidelines.

Households in Nationally Representative Surveys: The Influence of Censuses
Censuses and surveys have rather different purposes; complete enumeration for censuses versus in-depth information for surveys. However, because the establishment of statistical offices to conduct censuses generally preceded sample survey development, alongside pressure for comparability, the influence of established UN census design and concepts on survey definitions is clear: The problems of definition encountered are common to population enumeration in any context; therefore, it is suggested that, where the difficulties have been faced and a satisfactory definition of a household has been evolved for purposes of population census, it will usually be desirable to adopt that for sample surveys also. In most cases this will be the international standard definition of private household, developed to promote international comparability in population census results. (United Nations, 1964, p. 10) The merits of standardization were promoted by international organizations at an early stage: Discussion of the value of household inquiries, especially in developing areas, was carried a step further by the Working Group of Experts on Family Living Studies convened by the International Labour Office in 1955. This group recommended inter alia that the international agencies should aid in the development of sound methods of study and encourage international comparability by issuing lists of standard definitions and classifications to be used in household enquiries. (United Nations, 1964, p. 2) Comparisons of the definitions used in most recent household surveys in our sample countries suggest that this standardization and harmonization has not been achieved despite the production of national and international documents specifically outlining harmonized concepts (e.g., National Bureau of Statistics [NBS], 2005;UBOS, 2012).
The 1964 United Nations Statistics Office document did recognize that global diversity might present problems for standardized approaches: -problems of application of household definition in "underdeveloped" countries "where variations from the so-called 'normal' family structure are present" . . .
-in such circumstances, application of the recommended international definition of a household requires care. (United Nations, 1964, p. 12) One assumes that "normal family structure" referred to nuclear families, themselves now considerably eroded in many contexts where they might once have been considered to be "normal" (Cherlin, 2012).
UN documentation recognizes the problems the standardization enterprise has set in Africa, especially with respect to survey data collection, as surveys cover diverse issues for which different definitions may be more appropriate. However, it remains unclear whether it is expected that respondents should be reconfigured to make them fit with the definition, as in Uganda, or the definition should be used flexibly to be able to match local conditions, as in Senegal.

Conclusion
A number of themes emerge from this investigation of the role of UN guidelines in determining the collection of census data in Africa.
There is a temporal transformation with early postindependence censuses preoccupied with a complete enumeration of the population and avoiding double counting and omissions. The units of data collection reflected this, and utilization of local terminology or very strict de facto approaches were strategies used to do the best "counting" operation possible. Being able to undertake a competent and exhaustive census was seen as an essential part of the modern state. Issues of comparability or sophisticated analysis of the nature and structure of the units for which data were collected were largely ignored.
In the 1970s and 1980s, the movement toward referring to UN guidelines in the census documentation of individual countries, the use of "household," and the reiteration of UN notions of household became more important. However, a major problem arose, and persists, because of ambiguity in the UN definition of household and the fact that it referred to three criteria: residence, housekeeping, and provision of food. Different countries have emphasized these criteria in different ways, and in some, the provision of food has been reduced to those eating out of the same cooking pot (e.g., Uganda); this is a minimalist interpretation of the UN requirements of "provision of food" which may, when applied strictly, split basic socioeconomic units into smaller components because of social diversity in eating behaviors. Other countries integrate concepts into their definitions, which are not mentioned in the UN but are seen as locally important like "answer to one household head." In Burkina Faso, the continued restriction of census households to one married couple is a further elaboration away from the UN guidelines.
Senegal stands out as a nation with confidence in its own statistical collection and approaches. Even after adopting the UN terminology of household (ménage), the instructions retained local language terms, an approach abandoned elsewhere in Africa after the 1960s and which probably explains the consistently larger household sizes observed in Senegalese census and survey data compared with other countries The structure of Senegalese data collection retains this independence with many surveys retaining the idea of "noyaux" and the census returning to it.
This article aimed to provide a historical account and analysis of the changing relationships between individual African countries, their national data collection activities, and the United Nations through the lens of the definition of household. It is beyond the scope of the research reported on here to determine the extent of repercussions of these definitional evolutions on policy. Furthermore, as policies should, in theory, be based on a triangulation of diverse data sources, as well as be driven by political strategies, it would be of great concern if definitions alone could be shown to have directly influenced policies. Nevertheless, the indicators generated from census and survey data can be powerful tools in orienting development initiatives. Cross-national comparisons are increasingly used in the tracking of development goals. If analysts and policy makers are not clear about the different ways in which the same terminology household is operationalized and applied in different settings, then the quality of such comparisons is undermined.
Census households matter beyond facilitating the enumeration of individuals because, increasingly, research and policies do not focus on the characteristics of individuals but on the characteristics of the social units within which they live. This focus is demonstrated in the Republic of Kenya Strategic Plan "The household is central to the development process. Not only is the household a production unit but it is also a consumption, social and demographic unit" (Kenya, 2003, p. 59), with similar phrases echoed elsewhere. If interventions are focused on the "household" units for which policy makers have data, the characteristics of those households become important. Thus, wherever the definition influences the characteristics of the household, there will be repercussions for the conclusions drawn from analyses of those characteristics. Two examples illustrate this: The first constitutes analyses that focus on the characteristics of the household head and its association with, or influence on, the well-being or characteristics of different members: Such analyses could provide information on the sorts of households where children are more or less likely to be schooled, or those identified as economically or socially vulnerable such as female-headed households. In the latter case, definitions that depend primarily on cooking pot organization will often lead to high frequency of femaleheaded households in regions where polygamy is widespread and where polygamous wives cook separately such as Maasai areas of Kenya and Tanzania (Kenya, 2005;Randall & Coast, 2015). A second example draws on analyses that focus on the size and structure of households: Analyses of census data such as Ruggles and Heggeness's (2008) work on intergenerational coresidence in developing countries are particularly sensitive to household definitions. Burkina Faso was not part of that study, but the fact that in census data Burkinabe elderly married couples are, by definition, separated into different households whether they eat or live together with offspring, would have given an erroneous picture of intergenerational coresidence. Analysis of both household size and proportion of female-headed households for our study countries did demonstrate quite considerable difference between the results generated by the census definition and those of surveys (Golaz, Antoine, Randall, & Coast, 2012). A focus on the living arrangements of older people is becoming an important dimension of evaluating welfare and, as highlighted in a recent review of the SAGE study in Ghana "the social opportunities and challenges of the evolving living arrangements in Ghana will be worthy of consideration in national policy discussions" (Biritwum et al., 2013, p. 11). Policy makers are interested in what analysis of living arrangements reveals, and these analyses are sensitive to household definitions.
A number of tensions emerge: between the requirement for comparability over time and space (recognized by all documentation in all countries) and the need to accommodate changing social contexts (urbanization etc.) and diverse forms of social organization; the desire to do the best data collection possible, which in the census means enumerating everyone once and once only; and the national recognition that this may not be achievable through using the UN definition of household. The fundamental tension seems to be that of applying a concept of household that remains largely Eurocentric and organized around complete enumeration where most households are composed of small nuclear families or fragments of them, to the rather different and dynamic living arrangements in much of Africa. Different nations have chosen different approaches. In some cases, the UN definition is adopted (and slightly modified) and these households become the somewhat alien statistical category just used for data collection (Uganda), whereas others, like Senegal have forged their own, more independent pathway. The increased power of computeraided data collection, entry, and processing means that simultaneously collecting and organizing both standardized, internationally comparable data, and locally defined and relevant data on the ways in which people live should be possible, and is desirable for policy makers and researchers alike. A household is defined as a group of persons who normally live and eat together. This is a very loose definition and there may be many cases when you are in doubt as to whether people should be included in the same household or shown as belonging to separate households. It is not possible in these instructions to cover all such cases in detail, and your decision in such cases should be determined by common sense and convenience in the enumeration. It is not a matter of great importance whether or not such persons are included in one household or shown belonging to separate households. The important thing is that every person should be enumerated. Difficult cases generally occur in towns rather than rural areas, and here, common sense should always be followed (Republic of Uganda, 1974, p. 87 A household is a group of persons who normally live and eat together. Very often the household will be a family living in the same house or compound and eating together. A household will normally consist of a man, his wife and children and sometimes relatives and maids. The following constitutes a household: (i) A household may consist of one person who lives and eats on his or her own. (ii) A household may consist of several persons who are not related to each other. What matters is that they live together in the same house or compound and eat together. (iii) If a man has two or more wives and they and their children live and eat together, they form one household. If the wives and their children live and eat separately, they will form more than one household. (iv) If two or more groups of persons, each of which has its own separate eating and housekeeping arrangements, live in the same dwelling, treat them as separate households (UBOS, 2001).

Country Definitions of Household and Guidelines
En général un ménage comprend un homme, son épouse ou ses épouses, ses enfants non mariés, d'autres parents et domestiques non mariés qui vivent ensemble. NB: As for 1996 (INSD, 2006, pp. 11-12) Note. EA = enumeration area; INSD = Institut National de la Statistique et de la Démographie; UBOS = Uganda Bureau of Statistics. a All footnotes are author translations: In theory, normal households are made up of a man married to one or more wives, the unmarried children from that union and, if applicable, other unmarried children of earlier unions, ascendants, descendants, extended kin, and others. b A compound consists of a hut or a group of huts or other sorts of dwelling delimited (or not) by a wall. The family nucleus is the smallest possible family unit. It is made up of wives (or a wife) and their direct, unmarried descendants, that is to say parents and unmarried children. These people must live together in the same compound. By extension, the same family nucleus can include the husband, several spouses, and their children provided they live in the same compound. Furthermore the same nucleus can include direct ascendants (mother of the head of the family nucleus), brothers and sisters, and close unmarried kin (nephews, nieces, etc.) provided they live with the head of the family nucleus and themselves have no unmarried children in the compound. c The basic socioeconomic unit in which different members, related or not, live together in the same house or compound, pool their resources, and jointly meet their needs for food and other items of daily living. Some examples of households: • Every married man, with his wife and unmarried children constitutes a separate household.
• Each child of a man, with his wife(ves) and children make up a separate household even if they cohabit in the same house or compound and jointly meet their needs for food and other items of daily living. • Any person, male or female, who lives alone and provides for himself or herself forms a separate household, and so on. d The household is the group of people, related or not, who live in the same compound, eat their daily meal together, and recognize the authority of a single person called the household head. This concept is equivalent to "ndieul" in Wolof, "ngank" in Serer, and "hirande" in Toucouleur. If someone lives in the compound and eats with a household in that compound, you should record him or her in that household. If he or she lives outside the compound but eats with a household, he or she should not be counted in that household. Furthermore, if a person lives alone and eats alone, he or she should be recorded as a distinct single person household. The compound is a group of buildings enclosed (or not) by a wall or fence . . . by residence we mean "usually living in a place for a certain length of time." For the 1988 census, this place is the compound and the duration of residence is 6 months. e The basic socioeconomic unit in which members can be related or not. They live together in the same house or compound and jointly meet their needs for food and the majority of other items of daily living. They usually recognize the authority of one of the household members who is the household head, whatever their sex. NB: In compounds or houses occupied by parents and their married children, you must count the parents as a separate household from the household of their married children: Each married child, his wife(ves), and unmarried children make up a different household. However, if just one parent depends on their married child, they belong to his household. f A group of people, related or not, who live together under the same roof and jointly or partly pool resources to meet their basic needs, notably housing and food. These household members usually eat together and all recognize the authority of a single household head. In our national languages, the terms njël in Wolof, ngank in Serer, hirande in Pulaar, and siitik in Joola are accurate translations of the concept of household.
Ernestina Coast is an associate professor of Population Studies in the Social Policy Department at the London School of Economics and Political Science. She has a background in both demography and anthropology, and her research focuses on sub-Saharan Africa.
Philippe Antoine is a demographer and emeritus head of research at IRD. Over 40 years he has worked on family transformations and intergenerational relations in different African capital cities.
Natacha Compaore is trained in Sociology. Between 2005-2012 she worked on projects researching reproductive and sexual health, HIV/ AIDS and intergenerational relations. She is currently employed by an NGO working towards achieving food security in Burkina Faso.
Fatou-Binetou Dial has a PhD in Family Sociology from the University of Paris X, Nanterre. She is currently a researcher at the cultural anthropology laboratory of IFAN in Cheikh Anta Diop University, Dakar. She works primarily on marriage and ageing in Senegal.
Alexandra Fanghanel is a lecturer in Criminology at the University of Bedfordshire. From 2010-2012 she worked as a post-doctoral research associate at University College London on the HH-MM project which informs this paper.
Sadio Ba Gning has a PhD in Sociology. She is a lecturer in sociology at the Gaston Berger University in St Louis, Senegal. Her research focuses around gender issues and intergenerational relations.
Bilampoa Gnoumou Thiombiano has a PhD in Demography and is a lecturer/researcher at the Institut Supérieur des Sciences de la Population (ISSP) in Ouagadougou University. She researches family dynamics, marriage and marital disruptions, gender relations, education and child health in Africa.
Valérie Golaz is a research fellow at INED, based at LPED, Aix-Marseille University, France. Her research interests lie in access to land and intergenerational relations in East Africa. She was based at Makerere University Centre for Population and Applied Statistics, during the course of this research programme (2009)(2010)(2011)(2012)(2013).
Stephen Ojiambo Wandera is an assistant lecturer at the Department of Population Studies and a PhD student registered at Makerere University working on access to healthcare among older Ugandans. Previous research includes gender-based violence and reproductive health outcomes; health status; and access to healthcare.