Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

III. Measurement of key variables

Measurement of key variables
9. Multiple group membership and intra-household resource allocation
10. Time-allocation research: the costs and benefits of alternative methods
11. Use of emic units for time-use recall
12. Data on food consumption by high-risk family members: its utility for identifying target households for food and nutrition programmes
13. Determinants the ability of household members to adapt to social and economic changes

Measurement of key variables


The conceptual frameworks presented in part I identified a series of issues related to the question of how a proposed policy change or new programme might alter intrahousehold processes, and whether these alterations might affect the success of the project and the welfare of individuals involved, directly or indirectly, in the change. The following set of questions was derived from these frameworks as a way of organizing data collection for the analysis of project effects on households and individuals:

  1. Who will participate in the project's activities?
  2. Will the project require or cause a change in household structure, composition, or function?
  3. Will the project change any person's access to productive resources or any person's control over what is produced (including control over income from his/her labour)?
  4. Will the project affect any person's wage rate (returns to labour) or the rate of return to assets under any person's control?
  5. Will the project require changes in the inventory of tasks performed by household members, or in the organization of tasks?
  6. Will the project change the allocation of tasks among members or the time use of members?
  7. Will the project change any person's access to consumption goods (food, health care, education, etc.) which affect individual welfare?

The relevance of these questions to programme planning is discussed in part 1. The need to answer the first question should be self-evident: participants must first be identified before their behaviour can be predicted. The question of household structure is important because of the potential for resistance to or rejection of the project (see Safilios-Rothschild's paper in this section), and because it may indicate how project benefits may be dispersed. Fundamental changes in the household may also cause emotional stress, as discussed by Messer in part I and by Safilios-Rothschild in this section. Moreover, a change in household composition, such as male outmigration or the physical separation of nuclear from extended family units, may increase the work burden on remaining members (by reducing the possibilities for sharing tasks), the income on which they can draw, the resources available to them, and the possibilities for support during an emergency.

The question of individual access to income and capital is key because of the potential harm that projects may do to certain categories of idividuals if this issue is not resolved equitably. Since different individuals have distinct priorities for the uses of income, and since the person who earns the income generally has a greater degree of control over its uses, altering access to income may also have significant consequences for the ways in which income is used (see the papers by Messer, Engle, Behrman, Bennett, and SafiliosRothschild for further discussion).

The task allocation and time-use questions are related. Time burdens may reduce or prevent participation in the project, or may interfere with the performance of other tasks equally important to the welfare of household members. For example, a project which imposes increased demands on a mother's time may reduce the amount of time she can spend in child care, including food preparation and feeding. This reasoning applies to all household members. An employment or schooling project, for instance, takes children's labour time away from the household, increasing the work burden of remaining members, and possibly reducing the total amount of time devoted to particular tasks.

Answering these questions requires information on a variety of specific key variables:

  1. Household structure and composition.
  2. Individual incomes (cash, in-kind).
  3. Individual and household asset ownership.
  4. Tasks performed and their allocation among members.
  5. Uses of individuals' time.
  6. Allocation of consumption goods among household members.

Some of these variables are best measured using a qualitative approach, while others are more suited to quantitative methods. The papers in part 11 addressed these methodological approaches, along with their advantages and limitations for identifying aspects of intra-household processes.

The papers in this section confront the problems of measuring the key variables and propose some solutions. In the first paper, Heywood discusses the difficulty of applying any single definition of the household, especially if it is not specific to the culture being studied. He suggests that, rather than attempt a fixed definition of the household, one should record information on individual members in such a way that the individual can be assigned to any of several differently defined and possibly overlapping household units, depending on the requirements of the specific analysis or project.

The other papers deal with the measurement of specific variables which are particularly critical to intra-household-level analysis. Time and income are the two major categories of resources available to a household. Measurement of income and of time use are therefore central to any study of resource allocation within households. Johnson reviews and discusses the pros and cons of time-use and task-allocation methods, and emphasizes the importance of standardizing measurement techniques. Reflecting the concerns expressed by Scrimshaw, he underlines the importance of validating recail or descriptive data with structured direct observation of time use. Zeitlin, in her paper, suggests that, when direct observation is not possible, there are ways of making recall data more reliable by incorporating local perceptions of time and its measurement. These papers together confirm the importance of defining variables for measurement using both the outsider's (etic) and the local culture's (emic) perspectives, as discussed earlier by Messer.

The measurement of income and asset-ownership is not specifically addressed in these papers. Measurement of household income is treated in a variety of publications, but there are certain principles which are of particular importance to analysis at the intrahousehold level. First, Rosenzweig argues persuasively that income must be measured for the individual earner, and that wage rates as well as total earnings are important for assessing the relative value of household members' time. Because wage rates are a major determinant of resource allocation patterns, it is important to know what proportion of income is earned (and thus affects the value of time in all its uses), and what proportion is obtained from other sources, such as transfers or returns to wealth.

Jelín's paper stresses the importance of the timing and reliability of income as determinants of its use. It is important to know whether income is received weekly, daily, or seasonally, and whether it comes from steady employment or from irregular, occasional work. Messer and Engle in their conceptual frameworks, and Bennett and SafiliosRothschild in their papers, contend that both the size of an individual's economic contribution to the household and the degree of his/her economic independence are significant determinants of that person's access to household resources and his/her control over the consumption and allocation decisions of the household. This perception is based on the conflict-resolution model of household decision-making. It is not only earned income which affects an individual's degree of economic independence. Unearned income may also be attributable to a single individual in the household, as when a son sends remittances to one parent, or when one member of a couple receives transfers or access to productive resources from his/her natal family.

For intra-household analysis, income and assets should be associated with the individual responsible for them, if possible. Measurement of income is notoriously difficult, both because it is a sensitive subject about which respondents may be unwilling to talk, and because in many cases they may simply not know how much income they earn. For purposes of assessing the relative contributions of individuals within the household, approximate amounts are enough; absolute accuracy is not necessary.

Allocation of consumption goods is another category of information needed to assess the effects of development projects on individual household members. In this section, Pinstrup-Andersen and Garcia evaluate several measures of food consumption, and conclude that household-level food consumption information is a very poor proxy for individual consumption levels. Their results underscore the importance of individual-level data for the assessment of intra-household patterns. Food is in some ways the best indicator of allocation of consumption goods: consumption of health-care and educational services is conditional on a variety of factors, including service availability and perceived need. But all households at every economic level consume food, and food represents a critical element in human capital formation in addition to being a consumption item.

The final paper in this section deals with the measurement of a variable which is much harder to define, but no less important. We know that development interventions introduce changes in the economic and social environment that cause households to adapt in a variety of ways' reallocating tasks, responsibilities, roles, and access to consumption. In order to assess the probable effects of an intervention, planners would like to predict just how the reallocation will occur, who will gain and lose, and in what ways. Safilios-Rothschild addresses the question of how to predict the degree and direction of adaptation in households to a given project-induced change. She focuses on one particular aspect of household adaptation, the relative power of men and women, and suggests that households will accept the changes induced by development programmes readily if the changes do not visibly alter the distribution of power.

In this argument Safilios-Rothschild disagrees with Bennett and with Engle, who suggest that the visibility of a woman's economic contribution is precisely what enhances her decision-making power. However, she agrees with Rosenzweig that individual time-use may adapt more readily than a person's sense of power. This may be an indication of the phenomenon discussed by Engle, that behaviour changes more readily than attitudes, and that the traditional knowledge-attitudes-practices model should really start with behaviour change.

The papers in this section demonstrate that a focus on intra-household allocation, in both implementation and research, poses unique measurement questions, many of which have yet to be resolved. Some problems, such as defining the household, are simply intractable; the solution lies in redefining the objective. In the case of measuring food consumption and income, short-cut methods are not available; there is no way around the need to expend additional effort to get the information required for intra-household analysis. A meassage inherent in all these papers is that responsible planning involves addressing the measurement questions directly; the need for intra-household analysis, and its pay-off in terms of programme success, merit the extra effort.

9. Multiple group membership and intra-household resource allocation

Papua New Guinea Institute of Medical Research, Madang, Papua New Guinea


The field of anthropology is responsible for much of what we know about how families and households function in different cultural settings (see Messer's discussion of intrahousehold dynamics in the context of anthropology in part 1). However, defining the household is far from straightforward (Messer, 1983). Is a household a reference group defined by those who live together? Under one roof? Who sleep there? Who eat at the same table? Who contribute money (income) to the common coffer? Or is the household defined in terms of members' kinship ties? Often a member of a household is also a member of several other reference groups, depending on the task at hand. Can these multiple group memberships be taken into account in studying the household unit?

Household composition, structure, and function vary among and within cultures and within individual households over time. In many cultures, the functions we associate with "the household" in a Western context are divided among several reference groups. This means that defining the household, no easy task in any setting because of the fluidity of boundaries over time, also depends on the particular household function of interest. Thus, recognizing that there is no single definition of "household" which will serve all purposes, this paper addresses the particular problem of assigning, to a given individual, membership in a variety of reference groups, each of which may be considered a household for some purposes. Examples are taken from our work and experience in the subsistence cultures of Papua New Guinea (PNG).

In many cultures, there is an interplay between the members of a household and the resources they bring in. The resources are generated through group as well as individual processes, and each individual may be a member of several different groups. Food production, food consumption, health-care behaviour, and cash income generation are group processes that may be performed within very different reference groups, even within a given cultural and economic setting. The groups, whose memberships cut across the boundaries of the primary household unit, may make demands on the individual's time for joint processes. In return, the individual gets his or her share of the resources (food, income, etc.) generated by the group effort. These resources are then brought into the household, that is, the primary group to which the individual belongs.' Resources can be time, cash, or consumable goods.

Development policies seeking to affect these resources cannot ignore multiple group membership. Policy-makers and programme planners as well as survey researchers must take this into account in order to identify correctly target individuals for programmes, or respondents for research. The most effective programme designs incorporate an understanding of intra-household dynamics and the individual members' relationships. Since (1) the household is not a static unit, (2) different groups may be formed according to specific tasks being performed, and (3) an individual member can be part of several different (perhaps overlapping) units depending on the task at hand, the household may be defined differently depending on the aims of the project or investigation.

For instance, nutritionists and programme planners focusing on food and nutrient consumption often define the household in terms of "eating units" or "food budget units" (Messer, 1983). We used this approach in our study of changes in food intake patterns over 25 years in a village in the Simbu Province of PNG (Harvey and Heywood, 1983). We defined a household as the group of people who ate together in the same house, since the aim of the study was a comparison of individual food intakes with earlier data on individual food consumption patterns.


In societies which are at least partly dependent on subsistence production, household members who eat together are not always part of the same unit(s) that produced the food for the household. Projects seeking to effect changes in consumption and production, then, cannot assume that the same unit (identical members) is always responsible for both food production and consumption. Although in many parts of the highlands of PNG the consumption group would also be a production unit, there are variations. For instance, in Southern Highlands Province, Huli men grow and cook their own food while women and children eat food produced by the women (Brown, 1978).

Elsewhere in PNG, the pattern differs. The Garia in Southern Madang Province (Lawrence, 1984) engage in shifting cultivation, clearing and cultivating new sites each year. Plots are prepared by groups working with a garden leader, generally a middle-aged man, experienced in co-ordinating the activities of his fellow workers, and recognized as a performer of successful agricultural rituals. At the beginning of the agricultural year, he may seek out associates to work with him, or he may announce publicly that he intends to clear bush in a particular area. Those who desire to work with this leader then approach him in the next few days to form a team.

Members of the Garia food production unit are not grouped by kinship ties. Agnates (i.e. members of the paternal family line) rarely work together in the same garden. A nuclear family also rarely works together: husband, wife, and adolescent children may even plant on separate sites. Moreover, an individual may work with more than one garden leader. The result is that a single nuclear family, through its individual members, may be associated for the period of the agricultural cycle, approximately 18 months, with distant kin and non-kin in up to six food-producing gardens. Among the Garia, then, a household as defined by kinship, resource-sharing, or common dwelling, might be represented by its individual members in a number of food-production units. The total amount of food produced by the household is therefore derived from several gardens through the participation of its members.

There is yet a different pattern of food production in Western Madang Province (Rappaport, 1984). Here, the Tsembaga men and women co-operate in making gardens, and usually work in pairs: a woman may make more than one garden with different men, including her husband, her own or her husband's unmarried brothers, or her widowed father. Likewise, a man may cultivate several plots with his wife, his or her unmarried sisters, or his widowed mother. As among the Garia, the total food consumed by a Tsembaga household is produced through the participation of individual household members in several, though smaller, food-producing units. It is clear that defining the household in such a subsistence setting is not straightforward: the food-producing and consuming units are not identical. A project seeking to affect food consumption must address the various roles which an individual has as a member of multiple groups.

Decision-making which ultimately affects the level of household resources is also determined in part by household and multiple group membership. Clearing and initial preparation of the land are generally group activities in which the adult members of a number of households participate. Heavy work like felling trees is usually carried out by the male members. The type and amount of ground that can be cleared is determined by the size of the group and the characteristics of its members, particularly of the men. Thus, many of the decisions affecting household food production, and consequently the amount of food within the household, will be made by the production group, a different group from the primary dwelling or consumption unit. If the male household head is absent, there may be a very marked reduction of the total resources available at the household level. In contrast, food consumption decisions, given the available quantities of food, tend to be made within the household of an individual wife and her children.


In PNG, we have attempted to incorporate multiple group membership into a general system of recording information on household members. This coding system is currently in use in the larger population studies being carried out by the Papua New Guinea Institute for Medical Research (PNGIMR). It is informed by anthropologic data which indicates that the larger populations are organized into clans, subclans, and families.(2) This pattern of settlement varies among geographic areas. In the Tari Basin (Riley, 1979), the settlement is dispersed; in Eastern Highlands Province, it is in villages that correspond to subclans. A seven-digit census number incorporating the anthropologically derived relevant subgroup memberships is assigned to each individual for use in large-scale surveys. The first two digits of the code indicate the clan, the third the subclan, and the fourth and fifth identify the heads of household within the subclan. The sixth and seventh digits represent the individual, as follows:

The system allows easy identification of the major groups to whom individuals belong through the census number, which is unique to each individual. As an individual's status changes (e.g. through marriage), his or her census number changes, but previous numbers are also recorded, allowing cross-checking and identification of changes of status over time. This system has been used in the Tari Basin for approximately 15 years, and, over time, it has become clear that too much information was being stored in the census number. This number has been retained, but now it indicates only the current status of the individual. Old numbers are kept on file and can be used for cross-referencing and tracing changes in status over time. A separate, unique ID number is now used for each individual.

This coding method has been successfully modified for use in other geographic areas with different settlement patterns, although the basic principle remains the same. Such a system could also easily be modified for other purposes by adding digits to represent membership in other production or consumption units. For example, the first wife of the head of household 9 might also be identified as a member of a particular group of market women, or as a participant in her brother's gardening unit. An ethnographic study would be needed to identify the reference groups of interest; this would be used to define relevant groups. The resulting coding system could be applied in a larger-scale survey.

This system is one approach that has been successfully used in large-scale, longitudinal, epidemiologic, and demographic studies in several populations in PNG. Other approaches may be more useful in other populations. This does, however, in conjunction with other information, allow membership of some groups within the community to be recorded efficiently. Other communities with other bases of group membership will require different ethnographically derived systems for recording membership. The principles, none the less, remain the same.

It is obviously impossible to try to devise one system for recording household members which will fit every need. In large-scale studies with multiple objectives (even when they appear closely related, such as household food consumption, production and expenditure studies), it is particularly important to be able to derive multiple definitions of group membership from the information recorded in the individual's record so that macro- and micro-studies can be co-ordinated. This will facilitate the most efficient two-way flow of information between large- and small-scale studies on the same overlapping populations and samples.


The examples of work done in PNG illustrate the importance of identifying multiple group membership for studying the intra-household allocation of resources and other intra-household dynamics. Such studies can clearly benefit from combining the methods of anthropology with those of other relevant disciplines (cf. Messer; Engle; Scrimshaw; Safilios-Rothschild, in this volume). Ethnographic investigations are necessary to determine the extent and nature of multiple group membership, and to generate initial and specific hypotheses about its nature and effects on intra-household processes. Testing the hypotheses in a broader context and in a wider population will often depend on operationalizing definitions of group membership in such a way that information can be collected on large numbers of individuals by people who have little or no training in ethnography. The specific group memberships to be recorded will depend on the household process being investigated. It will not usually be possible to collect and code information on all memberships related to all intra-household allocative processes; it will be necessary to select the specific processes for study, and to identify the effects of specific memberships on them. So long as the studies are set up to look at specific effects and to test specific hypotheses, the data coding and analysis should be relatively straightforward. The important point is that the design of macro-studies should be influenced by the results of micro-studies, and vice versa. This is essential if the effects of very diverse social, economic, and biological phenomena on intra-household resource allocation are to be understood.

A similar co-ordination between the two types of studies is important in understanding the allocation of resources within households. Anthropologic insights about the nature of groups to which an individual belongs, and the demands he or she makes on the group and vice versa, should help identify the information to be collected on the larger population in order to make realistic determinations of multiple group membership. Large-scale studies of their effect on intra-household processes should then be possible. A system of recording members, which allows precise multiple definitions of the many groups to which the various members of the household belong, may permit a single study to be used for an even wider and more diverse set of aims and objectives.


  1. For some activities, individuals keep control of the (monetary) resources they bring into the household; in others, the resources are pooled within the primary group.
  2. Clans and subclans are localized, territorial patrilineal groups and, depending on size and area, they may form all or part of a village. Alternatively, the settlement pattern may be dispersed.


Brown, P. 1978. Highland Peoples of New Guinea. Cambridge University Press, Cambridge.

Harvey, P., and P. Heywood. 1983. Twenty-five Years of Dietary Change in Simbu Province, Papua New Guinea. Ecol. Fd. Nutr., 13: 27-35.

Lawrence, P. 1984. The Garia: An Ethnography of a Traditional Cosmic System in Papua New Guinea. Melbourne University Press, Melbourne.

Messer, E. 1983. The Household Focus in Nutritional Anthropology: An Overview. Fd. Nutr. Bull.. 5(4): 2-12.

Rappaport, R. 1984. Pigs for the Ancestors: Ritual in the Ecology of a New Guinea People. New, enlarged edition. Yale University Press, New Haven, Conn.

Riley, I.D. 1979. Pneumonia in Papua New Guinea. M.D. thesis. University of Sydney.

Contents - Previous - Next