This is the old United Nations University website. Visit the new site at http://unu.edu
Household Composition
An enumeration of all household members provides a basis for specifying a number of important household variables. The period of questioning to collect these data is also a convenient time to collect information on other characteristics of household members. The format below can be set up with precoded categories of space for coding to facilitate computer analysis. The definition of "male household head" and "female household head" should be standardized on the basis of data from preliminary ethnographic and survey work. For example, in situations where extended family households are common and older men give authority to their eldest son, resident in household, "male household head" can be used to designate the eldest son, and the older father may be classified as "father of household head." When authority is not so delegated, the elder son may be designated as "son of household head" or "eldest son of household head." These designations must be standardized for a community or region but will vary from one cultural area to another.
TABLE 10.1. Format for Household Composition
Name | Relationship to household head | Sex | Age | Education | Occupation/ Employment |
1) | |||||
2) | |||||
etc. |
From this matrix a number of variables of household composition can be constructed, including:
The last two variables reflect the reproductive history of women in the household and at the same time mark features of current composition.
Who Is a Member of the Household? The Problem of Non-resident Members
For many people in the world today, household membership is no longer a stable, unchanging condition. Both men and women are periodically absent from their home communities following seasonal labor opportunities. Adolescent children may be gone for long periods of time, while remaining significantly tied to the household. In recording household composition, it is very important to elicit data on non-resident household members. One technique for collecting information on total household composition, while maintaining the separation between currently resident and nonresident members, is to elicit the data in stages, as follows:
Decisions about how to handle these data for analytic purposes will, of necessity, vary from one cultural setting to another. It may be useful to establish one or more variables that specify the "degree of stability" of household composition as a distinct variable, which differentiates households within a community or region.
Material Resources
During the basic household interview, a second main category of questioning is the eliciting of data on material resources. In most parts of the world, households range from "low" to "high" in possession of and access to economic resources. Some households are equipped with many technological items, including motor vehicles, heating and cooking equipment, bicycles, television sets, etc. Other households in the same community will appear to have "practically nothing," while the majority of households will have technological and other resources falling in between the two extremes. Differential distribution of resources, especially in nonsocialist countries, is generally found even in communities that appear at first glance to be "all alike in poverty. "
Differences among households in economic resources are generally related to dietary patterns and nutritional status. Often the presence of economic resources indicates a general capability of a household to satisfy the wants and needs of the members. At the same time, the presence of greater numbers of material possessions also can be simply a reflection of time-older households have had time to accumulate more than younger households. Of course, inheritance and other factors (including health and good fortune) also play a role in differential material resources.
Measuring Resources: Material Style of Life
There are a number of ways to measure differential resources of households within a community. One of the most effective, often less difficult than outright questions about income or total wealth, is to develop a scale of material possessions of "material style of life," by identifying lists of principal items that are significant in the local region. (The relevant items will differ, of course, from one region to another.) In order to rank households from "high" to "low" in material possessions, one needs only a sample of relevant material goods that vary in frequency across the sample of households. Here is an example of a set of eight material items that effectively distinguished among households in a rural community in Mexico:
Item |
Frequency (in 57 households) |
Iron (electric or non-electric) | 53 |
Radio | 46 |
Bed | 36 |
Cooking facilities off the floor | 21 |
Sewing machine | 14 |
Wardrobe | 14 |
Stove | 9 |
Television | 7 |
For this particular Mexican community, one finds that the wealthier households (in local terms) have television sets and commercially purchased stoves, as well as the entire list of more common items. At the very poor end of the scale we find persons who perhaps have a non-electric iron and not even a bed or a radio. If we arrange households in their rank order of material style of life, and the items in their order of frequency, the pattern will look something like table 10.2 (see TABLE 10.2 Lifestyle/Material Items Matrix).
Each step or category (from poorest to richest) can be assigned a numerical value, to create a single quantitative expression of material well-being that can be used as a variable in statistical analysis.
TABLE 10 2 Lifestyle/Material Items Matrix
Household Type | Items Possessed |
|||||||
Iron | Radio | Bed | Cooking | Sewing | Wardrobe | Stove | TV | |
Poorest | no | no | no | no | no | no | no | no |
Next Poorest | yes | no | no | no | no | no | no | no |
Next | yes | yes | no | no | no | no | no | no |
Next | yes | yes | yes | no | no | no | no | no |
Next | yes | yes | yes | yes | no | no | no | no |
Richest | yes | yes | yes | yes | yes | yes | yes | yes |
Animals Owned by Household
For the same community in rural Mexico, the researchers found that possession of animals as economic resources represented a somewhat different "wealth rating" (1).
Therefore, the approximate value of animals owned by each household was developed as a second scale of economic resources.
Annual cash income is often used as measure of material style of life in developed countries. Cash income is often a direct reflection of occupational status in technologically advanced societies, so this measure serves not only as an indication of material resources, but also as an indirect indicator of educational attainments. However, in most developing nations (and many parts of developed nations) information on cash income may be extremely difficult to obtain from individual households, and it may not be a meaningful measure of socioeconomic status. Therefore several other measures of economic resources have been devised. A number of indices of wealth and economic status have proved useful in various regions and nations.
Wealth Rating by Key Informants
In most communities the more well-informed local residents are very knowledgeable about people's relative economic status. Therefore, a key informant can often rank order all or most of the households or families in the community in terms of relative wealth status. For larger communities it may be necessary to obtain ratings from several individuals, in separate sub-communities, with cross-checks at points of overlap.
Procedure
In some instances it is possible to refine the rank order by asking for the differences among individual households within each of the groups. Usually, however, informants will tend to identify four or five categories into which all the households are assigned. In many instances these groupings by relative wealth will not be recognized as different types or "classes" in local terms.
FIG. 10.1. Procedure for Comparing Ratings on "Overlap Households".
If the community is so large that each key informant can rank only a subsection of the entire sample. then it will be necessary to "interdigitate" the separate sub-samples by comparing people's ratings on the "overlap households" that received ratings by more than one informant. Figure 10.1. illustrates such a procedure, where each number is a household.
In many instances it will be useful to have multiple measurements of economic status by using several methods. Pilot testing in the research communities will make clear which measures of economic status will be the most useful and convenient.
Other Aspects of Material Resources
In addition to assessment of wealth, the basic household interview should cover other aspects of material resources related to food and nutrition. Resources for food preparation and preservation can be very significant intervening variables affecting the impact of a nutrition intervention programme. Among the important resources, special attention should be given to collecting information on food preparation and consumption technology, including cooking equipment, fuel, plates, utensils, etc. In some settings, food stores as well as storage facilities should be assessed. Other significant variables related to material resources that are relevant in many parts of the world include distance to various resources, including distances to fields, stores, markets, water sources, cash-earning opportunities, as well as access to transportation.
In the analysis of programme impact the items in the "material resources" section of the basic household interview can be treated as individual variables. However, many of them can be combined, using simple scale construction techniques, into composite measures. Thus, it can be useful to create variables such as:
Beliefs and Attitudes about Food and Nutrition
In addition to the collection of data on actual food intake, the evaluation effort requires information on beliefs and attitudes about food in order to assess the impact of the programme on recipients. Because humans eat food, not nutrients as such, and because they imbue food with a host of symbolic meanings-from ideas about the healthfulness of particular foods to its value as an expression of religious feeling-programme evaluations need to assess "food ideologies" as potentially important intervening or confounding variables. Data about beliefs and values related to food, together with information on actual dietary intake and behavior related to food acquisition, preparation and consumption, are very important for understanding and interpreting the results from clinical, anthropometric, and other types of nutritional status data. Local beliefs about the relationship of food to health maintenance and illness management may be particularly important in their effects on the selection and utilization of food, including food provided by the programme under evaluation.
Some aspects of the cultural significance of food can be ascertained through ethnographic research, rather than collected from every household in the survey. For example, the cultural practice of preparing feasts for events in the religious calendar and the specification of types of festival foods is a descriptive task for ethnography. On the other hand, some aspects of food beliefs and attitudes should be collected as part of the general survey. Of special interest in this respect are questions related to the particular programme being evaluated. For example, evaluation of programmes directed to feeding pregnant women and young children should collect information at the household level on beliefs and practices related to food use in pregnancy, lactation, and early childhood. The rationale for collecting those types of data from all households in the sample is that beliefs and attitudes can vary widely, even within apparently homogeneous communities. Thus, intra-community differences in beliefs may account for significant differences in programme impact from one household to another.
Since the content of questions about beliefs and attitudes should be culturally specific, to a large extent they have to be designed after the initial ethnographic work. This is particularly true when fixed-choice or precoded items are to be used. For example, in those many areas where the legacy of humoral medicine appears in the form of concerns about the hot and cold qualities of foods, medicines and diseases, questions about safe foods and practices in relation to this system would be important. However, preliminary ethnographic work is necessary to determine the presence or general significance of this type of belief system in the community.
Measuring Beliefs
Recognizing that there are often major disparities between what people say they believe (their verbal responses) and their actions, it is nonetheless very useful to elicit stated beliefs. Social scientists have developed a number of methods to collect this type of data. One of the most useful for food-related beliefs is based on a scaling technique, which requires individuals to rank or rate items (foods) in terms of a series of dimensions (e.g., a three-point, five-point, or seven-point scale). By using pictures and a board on which they can be physically placed, it is possible to use this technique with non-literate respondents. Variations of the technique have been used in many groups around the world, including populations with little exposure to written or even pictorial representation. To illustrate:
- Present the respondent with a set of cards depicting the foods on which you want to get responses (see Figure 1).
- Present the respondent with a board, with clear slots or markings (Figure 2).
- Suggest a dimension of value that you want to measure. E.g., "Here, at this end {right) are foods that are good for a sick baby. At this end (left) are foods that are bad for a sick baby. Here in the middle are foods that are neither good nor bad. (Pick up the card) Can you tell me where you would place this food? Is it good to give a sick baby?, bad?, or neutral?" Continue with each of the relevant foods.
By the researcher's assigning numerical values to the slots, the answers can be represented quantitatively, so that comparisons can be made among foods and between individuals in relation to other variables, such as programme use. If the number of foods to be scored is kept small, it is possible to inquire about a number of dimensions, because respondents tend to find this rating task relatively interesting compared to answering typical interview questions.
The Household as an Organization
Households can be regarded as small-scale organizations that have a large number of tasks to accomplish. Across cultures there are certain basic similarities in the composition of these tasks, although the extent to which other organizational units are also charged with responsiblity for accomplishing them varies widely. For example, in industrialized societies responsibilities for food preparation are increasingly given over to commercial establishments, as pre-prepared foods, restaurants and canteens assume a larger role in providing significant proportions of the individuals' food.
Between and within cultures there are also important differences in the way in which households, as organizations, arrange or allocate work. Differences in task allocation result from many factors, including household composition, the nature of available resources, cultural expectations, and a complex of individual physical and psychological characteristics. In turn, variations in task allocation or the organization of work can be expected to have significant consequences for health and nutritional status. Some households are more efficient. more active, and more capable of accomplishing tasks than others. Thus, variations in household management or organization can be regarded as one factor that helps to account for differential nutritional status within the same environment.
By extension, it can be argued that a food supplementation programme can bring about changes in nutritional status (or fail to bring about changes) because of its effects on household task allocation. Figure 10.2 (see FIG. 10.2. Potential Mechanism for Changes In Nutritional Status) below illustrates this potential mechanism.
In agricultural communities where household task organization has been studied, it is common for adult females to work at least 10 to 12 hours per day or more. Also, in most communities children work a number of hours a day. Any change that requires alterations in the household work source is likely to shift tasks to other household members. Thus, food programmes that change male adult work patterns or encourage child school attendance can, under some circumstances, increase women's work loads, even as they also increase food availability. On the other hand, increased food availability could change food preparation patterns, leading to changes in the total amount of time spent in food-related activities, in contrast to other types of domestic tasks.
From the foregoing discussion, it is apparent that data on household organization are important for assessing the impact of a food programme for several reasons:
Task Allocation and Performance
Household organization is a highly complex and often subtle phenomenon, which makes it difficult to measure. A major component of organization involves the allocation and performance of tasks. Households vary in the numbers and types of tasks that are carried out, as well as in the efficiency and quality with which they are accomplished. Within any community there are households that appear to maintain high levels of activity in many spheres (economic. social) while other households engage in fewer activities. There are differences, as well, in task performance. Even within the most egalitarian of communities, there are recognized differences in quality of performance, as some individuals are acknowledged to be outstanding farmers, artisans, etc. While issues of quality are quite difficult to discern, some aspects of time use and types of activities can be measured as part of the basic household survey. More in-depth analysis requires specialized sub-samples.
Measuring Household Organization
Three basic methods can be used to measure time use and activities of household members:
- direct observation
- interviewing {including 24-hour recall of activities)
- activity diaries
The selection of one method rather than another is based on a series of considerations, including resources available for the evaluation and the degree of precision required. But characteristics of the community also affect the decision. In some cultures, observers are readily tolerated, while detailed interviewing is regarded as tedious or impolite. In others, just the reverse situation may obtain. Before a final decision is made about which basic method to use, ethnographic work must be carried out.
From interviewing both men and women, an initial list of the types of tasks performed by households can be drawn up. This should include an assessment of what the interviewees regard as "typical" or "usual" (e.g., "women wash clothes once a week"). It may also be useful to elicit information on what people regard as desirable or ideal behaviour (e.g., "a good housewife whitewashes the house every six months"). These two assessments provide a background of the cultural expectations related to household organization.
Since all three methods - direct observation, repeated 24-hour recall of activities, and activity diaries - are relatively costly research activities, it is unlikely that they can be carried out on the full survey sample. Earlier we referred to the value of using sub-samples, selected from the full survey sample, for specialized data collection. With respect to household organization, the ideal sequence would be to collect and partially analyse household task performance data from a sub-sample before administering the basic household interview to the entire sample. Based on this work, a series of key questions can be incorporated into the general survey. The data collection activities outlined below are intended to be used with small sub-samples.
The Purpose of an Activity Record
Whatever method - observation, interview, or diary - is selected, the purpose of an activity record is to provide a body of data from which relevant behavioral categories can be coded. For example, it can be hypothesized that the presence of a food programme changes the amount of time spent in food preparation. The amount of time spent in food preparation can be calculated from the behavioural record. A comparison of households participating in the programme with households that do not then provides the conditions for testing the hypothesis. Similarly, time devoted to child care, income-generating activities or cleaning and household maintenance can be assessed and compared. For some activities, such as young child feeding, frequency (number of times per day) may be more significant than total time. For others the mere presence or absence of the task in the daily record may be important. The more complete the behavioural record, the greater is the flexibility to assess a range of organizational tasks in relation to nutritional status and programme participation.
Direct Observation with Timing of Activities
A complex, behavioural record can be obtained through direct observation, with timing of specific activities. Because the recording of activities, along with the time it takes to accomplish them, is a tiring task that cannot be maintained for long periods of time, a decision must be made about how to handle recording of the data. Two alternatives are:
- to record activities on a schedule (e.g. 10 minutes every half-hour for three hours, followed by a two-hour break)
- to record fully a pre-determined set of specific activities regardless of how long it takes (e.g., detailed recording of meal preparation and all activities related to infant feeding.
In either case. the procedure begins with the selection of a sub-sample of households to be observed. Observations by one field researcher require a minimum of one day per household, although some researchers recommend two days of observation since the first is somewhat distorted simply by the presence of the observer. At the second visit, activities are more likely to be closer to normal routines. If the first alternative above is selected, pre-prepared data sheets should be set up in terms of the timing units. When the second method is selected, the observational tasks can be made easier if sub-categories of an activity are prerecorded on data sheets, based on preliminary observations and interviews.
Random Short-term Observation
When day-long observation is not feasible, some researchers have had good success with systems of random "spot checks" of individual households.
Procedure:
The output of this method produces a "group profile" of time spent in various activities. Also, different activity patterns by sex and age are produced. However, differences among households are not revealed by this method.
In order to use the short-term observation method to gain data on individual households, the method can be revised as follows:
After the individual time slots are selected, they are discarded. Thus, the time slot sampling is "random without replacement". Once the slip for household No. 3 at 11 a.m. has been selected, it is not to be repeated.
TABLE 10.3. Household Profile of Time Spent in Various Activities
Household A |
Household B |
|
9 a.m. | Mother: Preparing food, house-hold
maintenance and tending child Father: agricultural work 16 yr daughter: household 11 yr son: agricultural work, tending animals 8 yr daughter: agricultural work, tending animals |
Mother: Preparing food, house-hold
maintenance Father: leisure 16 yr son: leisure (social) 14 yr daughter: handicraft maintenance |
11 a.m. | Mother: household maintenance,
tending child Father eating Children eating |
Mother: eating Father: leisure (social) Children: selling handicraft |
3 p.m. | Mother: Preparing food, child care Daughters: same Father: agricultural work Son: agricultural work, tending animals |
Mother: eating, leisure (social) Father: agricultural work Son: leisure Daughter: handicrafts |
6 p.m. | Mother: household mainte nance Father: repair equipment 16 yr daughter: child care 11 yr son: leisure (social) 8 yr daughter: leisure (social) |
Mother: leisure (social) Father: getting firewood Son: leisure Daughter: food preparation |
Adult leisure/eating (raw score) | 2 | 8 |
Mother leisure/eating | 0 | 3 |
Some comments about the short-term observation method of data gathering are:
- The spot checks do not provide data concerning the total time spans of particular activities.
- The data are best conceptualized as frequencies of particular activities by persons in households.
- The results from two different households might look like table 10.3.
In the (hypothetical) example presented in table 10.3. it should be clear that Household A has a higher "work ratio;" or conversely, Household B has a higher "leisure ratio." In both there are three adults (age 16 and over), which yields a total of 4 x 3 = 12 observations. Thus, the "leisure score" of Household A is 2112 or 0.17, while the "leisure score" of Household B is 8/12 or 0.67. The corresponding "mother's leisure score" is even more striking: 0.00 vs. 0.75. Experienced field workers have frequently observed these kinds of differences among households, although attempts to quantify them have rarely been made.
In the example above, only four observations were allocated to each household. Increased numbers of observations per household would make possible more fine-grained and "robust" analysis of inter-household variations. Also, relatively brief observations can be combined with short interviews and conversation concerning other activities, particularly those immediately preceding the observation. The method could also be targeted on mothers' behaviours in feeding and caring for infants. Such a focus would require a special sampling of households with 0 to one-year-old infants at the time of the survey.
Interviewing for Household Task Allocations
Interviewing methods are usually best structured as 24-hour recall with the female head of household, or some other household member. Ideally, one would wish that the respondent would be able to give approximate times of each activity. An alternative to "24-hour recall" is possible if all interviews are conducted in the evenings, asking respondents to recount all activities of the day just completed.
The sequence of events can be listed in time charts, with space for additional commentary. In some communities there are fairly regular time markers (e.g., church bells, radio news programmes, etc.) that help to preserve realistic time frames. Also, in some societies people maintain fairly clear notions of the hours of the day, perhaps in part because of interest in watches and clocks as technological prestige items. The procedure for interviewing for household task allocation is as follows:
Interviews should be sufficiently structured in format so that comparable data are gathered by different field researchers in the community. Generalized interviewing on other topics, therefore. should only be added to the protocol after the structured time-activity data for each person has been recorded.
The greater the number of call-backs to each sampled household, the more valid and reliable will be the overall results. However, it should be noted that care in the initial sample selection, plus rigorous control of the data-gathering procedures, contribute more to the strength of conclusions than does the total number of observations.
Keeping Work and Activity Diaries
In an increasing number of communities and regions throughout the world, there are sufficient levels of literacy that members of households can be asked to keep their own records of activities in "diaries." Such diary keeping by selected households is easier when the usual rounds are not highly complex for various household members.
If the diary method is attempted, field researchers should carefully assess their ethnographic data in order to establish categories and reasonably clear protocols for the households to follow. Short-term observations and spot interviews can be used to enhance data quality control.
The diary method permits the researchers to expand the total sample size without great additional personnel costs, though the regular monitoring and pick-up of the diaries can be more time-consuming than it appears at first glance. Collecting diary sheets at regular intervals, plus regular "de-briefing" and encouragement, can greatly increase the quality of the records.
Incomplete and poorly-kept diaries can be examined statistically to estimate the effects of inadequate recording. Such quality control methods, through statistical analysis, can make possible the judicious use of relatively uneven diary records.
Specialized Observations: Mother-Child Interaction
Up to this point, we have focused primarily on gathering data about all the people in a household. In some cases it may be more time-efficient to focus particularly on the mother and on mother-infant pairs. The key role of maternal performance and behaviour has been well documented in a number of studies. On the other hand, an evaluation study should never focus on maternal behaviour until after the ethnographic reconnaissance and related general data-gathering has given a fairly precise picture of the situation and roles of females in households. This will allow the concentration on maternal behaviour to be finely tuned to data on other aspects of household organization.
Interviews can produce useful information about differential behaviours of parents and children, but far more valid and reliable data can be obtained from direct observation.
Direct observation of mother-child interactions requires the presence of one or more observers in the same room with the mother-child dyed or else the use of videotape cameras. Use of videotape preserves a more accurate and complete set of raw data for analysis, but has the following disadvantages:
Whether one uses a videotape recording or depends on an observer recording detailed notes, researchers must decide on the frequencies of observation, length of time of each unit of observation. and total numbers of such observations. Ideally one would like to have at least three or four separate observations, varying the time of day.
Mother-infant observation could, under some circumstances, be combined with data collection on other variables such as diet and food use. This combined data collection process would be more feasible if the researchers worked as two-person teams, at least in the more intensive portions of observation.
We suggest the following pattern as a general research procedure, from which individual projects would deviate to varying degrees depending on available resources and the specific contexts of local conditions.
- total duration of feeding times per infant,
- total time of holding the infant by caretaker (mother),
- duration/total time of all positive, nurturant behaviour toward infant,
- activity level of infant (total number of different activities; ratio of active to inactive time),
- number of different caretakers of infant,
- time left unattended,
- total time (or number of times) crying or fussing,
- number of times infant vocalizes (all types),
- others
These behavioural variables can be grouped or clustered by means of factor analysis, cluster analysis, or other statistical methods into groups of related, co-occurring items. The clusters can then be examined in relation to nutritional variables and other data for correlations and other patterns suggested in the evaluation hypotheses.
During the past decade a major change has occurred in ethnographic field work strategies. Both the general conditions of research and heightened concern with validity and reliability of household data have encouraged (sometimes necessitated) greater use of local community persons as researchers. In addition, heightened local community interest in various research-and-development projects has increased the need for direct community participation in the planning and implementation of various projects, both theoretical and applied. In some instances, local community participation complicates research and evaluation efforts, particularly if local level factionalism and politics enter into research team recruitment and training. However, effective utilization of local expertise concerning cultural and social features can increase the quality and quantity of household data and thus facilitate more effective evaluations.
The training of household interviews,, and careful monitoring of their work, is essential. One effective training method is role playing, in which interviewers administer the schedule to each other, with critique provided by the project supervisor. Completed interviews should be reviewed in detail, especially during the early stages of data collection, but continuing throughout the data collection period. Among the important aspects to monitor is the correspondence between the answers respondents are giving to questions and the intent of questions.
In the basic household interview, missing data are very troublesome. Pretesting should eliminate those items that will not or cannot be answered by most respondents, so that for the most part, missing data in the final form of the interview schedule may be regarded as a problem of interviewer technique Interviewers will need help to develop skills for probing and handling difficult questions.
There are several strategies that can be used to address the difficult problems of reliability and validity with household data of the types required here. Deliberate falsification of information occurs, particularly in socially sensitive areas. (Topics of sensitivity will vary from one cultural setting to another, so that these potentially difficult areas have to be discovered through ethnographic work). Generally, however, misinformation is the result of misinterpretation of the meaning of questions. The importance of integrating ethnographic work, pretesting and discussing the results with local key informants cannot be stressed too strongly as the major check on quality control for socio-cultural data.
Another control for reliability is to use multiple indicators rather than relying on the response to a single item. When several items can be combined into a constructed variable (a scale or index), statistical procedures to test for scale reliability or coherence can be applied, providing a further check on data quality.