Contents - Previous - Next


This is the old United Nations University website. Visit the new site at http://unu.edu


10. Time-allocation research: the costs and benefits of alternative methods


ALLEN JOHNSON
Department of Anthropology, University of California, Los Angeles, California, USA

INTRODUCTION

Many social scientists are converging in their focus on detailed accurate descriptions of everyday behaviour. In the past, theory has tended to concentrate on activities outside the home (i.e. in the workplace, school, hospital, or church), and has paid correspondingly little attention to behaviour within the home. Although such focused theoretical socialscience research is valuable, it may miss important areas of study. First, human activities that are not currently at the centre of theoretical or policy debates may be overlooked. For example, until recently, activities taking place within the household (including housework, food preparation and distribution, and child care), as well as the division of labour by age and sex, suffered such neglect. Second, activities that may provide a context in which to understand the focal variables of the research are overlooked. For example, when a married woman with children takes on gainful employment, every aspect of home life and all members of the household are bound to be significantly affected by changes in the allocation of her time and resources. Knowing the amount of time a woman spends at work, or at any other single task, is therefore not sufficient to understand the effect of her working on the household. We need to understand how gainful employment affects her total allocation of time, how she partitions her time, how she organizes and sequences tasks, and how other family members reallocate and reorganize their time (or fail to do so).

Social-science research has increasingly turned to research methods which can describe broadly defined human activity patterns, in particular time allocation. A relatively new focus of social research, time allocation is currently measured with a wide range of specific methods, each having its own strengths and limitations. In this paper, I will first discuss the basic goals of research on time use, and then briefly indicate the "trade-offs" of the several quite distinct research techniques that have been used to measure time allocation. Finally, I will suggest certain rules of thumb for ensuring that findings from research on time allocation can be used most effectively by project planners and policymakers.

TIME-ALLOCATION RESEARCH: AN OVERVIEW

Time-allocation data are important for comparing patterns of human behaviour across diverse settings. Time is a "currency" we know how to measure accurately by the clock, irrespective of our research subjects or the settings in which we study them. Given that time is such a natural, universal, and objective scale of measurement (if one is using a clock), it is perhaps surprising that time allocation is not universally studied in comparative behavioural research. But social scientists do not carelessly neglect promising research methods. The reasons for the limited use of time-allocation research include shortcomings of existing data and current data-collection methods. The specifics of those shortcomings indicate the directions such research must take in the future.

The Unfulfilled Promise of Time-allocation Research

Time-allocation research, the effort to gather reliable and valid data on how people spend their time in the ordinary course of their lives, has grown with astonishing rapidity over the past decade (As, 1978; Andorka et al., 1983; Gross, 1984). Social scientists in several disciplines have made systematic, detailed behavioural descriptions of time use. The recent dramatic increase in computing power has made it practical to organize and process large datasets with thousands of individual observations, and has no doubt contributed to the growing interest in collecting time-allocation data. Yet little attention has been paid to standardizing techniques. Researchers devise their own individualized methods of measuring time allocation, making it difficult for other researchers and policy-makers to understand what they have done or to use the data for comparisons across cultures, economic systems, social categories, or political boundaries.

Within limits, a proliferation of research strategies is a sign of healthy growth and creativity in this relatively new field. Successful research requires techniques to be creatively fine-tuned to different settings. Furthermore, experience with a variety of methods is necessary to assess their relative costs and benefits. This methodological variety, however, has hindered our ability to compare results between studies and thus strengthen existing theories and make appropriate policy decisions. Findings from numerous individual studies of time allocation have been published, but it is difficult to draw comparisons between them, owing to differences in how the data were collected, and how the results were reported. Resolving the following specific problems in timeallocation research would significantly enhance the comparability, and thus the usefulness, of these studies.

First, primary time-allocation data, from which behavioural patterns of given populations are compiled, and the specific definitions used to code the primary data into activities are almost never available for examination. Hence, it is impossible to know how actual behaviour was transformed into primary data, and how the primary data were grouped to form the categories in the published tables. For example, two unrelated studies might report that fathers in each case spent 5 per cent of their time in "child care." But one might code "father plays ball with son" as "child care," while another might code the same activity as "recreation." Thus we would not know the specific behaviours each researcher counted as "child care" and we could not conclude that fathers in each of the two cases spent comparable times in child care.

Second, methods of sampling the subjects of time-allocation research are seldom reported, leading to difficulties in assessing the representativeness of the sample. Most time-allocation researchers are well aware of the principles of sampling, but have not reached a consensus on what sampling techniques are most appropriate to this kind of research.

Third, most studies report the general technique by which the data were gathered (e.g. 24-hour recall, diaries filled out by the research subjects themselves, random "spot checks" of directly observed behaviour), but provide little information about how their techniques bias the data or to what degree any given method is capable of replicating the results of the others.

Fourth, for interpreting results it is crucial to know how each technique is related to its ethnographic setting, and how varying methods of data collection influence possibilities for data analysis. For example, in some research settings it is easy and natural to drop in casually in both public and private settings to make momentary observations ("spot checks") of individual activities; in others it is inappropriate or simply impossible. If informant reports must be used instead of direct observations, a substantially different kind of data is collected, and this must be analysed and interpreted accordingly.

For these reasons, although scores of time-allocation datasets are now available from virtually every kind of cultural and economic system around the world, very few comparative studies have been published. Indeed, the few that have been usually involve comparisons between communities that were studied in a single research project, using the same methods (e.g. Gross et al., 1979; Szalai et al., 1972). When comparisons among studies are made (e.g. Minge-Klevana, 1980), they require questionable assumptions (for example, that "work" has been defined comparably in each study), and the lack of standardization leaves their conclusions open to challenge.

To begin to unify the field, at the University of California at Los Angeles we are creating a standardized cross-cultural time-allocation database. (1) Time-allocation studies from all areas of the world and levels of societal complexity are compiled into monographs that provide the original data (on computer diskette), as collected by the field researchers, as well as the same data translated into a standardized coding format to facilitate comparison between studies. The monographs also describe sampling and coding methods, a census of the research population, and an extensive summary of background information on environment, economy, social organization, culture, and the encompassing political system (cf. Johnson and Johnson, 1987).

The monographs and the database should prove to be excellent tools for use in project identification and design. An initial understanding of time-use patterns and constraints could be obtained from time-allocation studies previously done at the proposed project site, or in another similar type of society, helping to define the problem and to assess the feasibility of alternative interventions.

Time-allocation Research: A Theoretical Framework

Time-allocation research has the potential to generate objective measures of human activities that apply without prejudice to all people in all circumstances, regardless of personal differences, geographical location, the nature of the economic system, or of the broader cultural context. Activities of all kinds, whether economic, social, psychological, or political, can be described quantitatively in terms of the time allocated to them, if they may be observed at all. Even though time is a one-dimensional measure that does not take the qualitative diversity of human activities into account, time allocation is extremely useful for comparing human activities in varying contexts, just as in other contexts monetary costs, or the earnings derived from an activity, are useful measures for comparison. Time-allocation studies give an indication of when during the day the activity takes place (e.g. day or evening), the way tasks are organized (e.g. certain activities are always done together), who does the task or activity, and with whom the task is done. This information is essential to effective project planning and implementation.

Time-allocation theory has been based on the assumption that people in general experience their time as scarce and allocate it among alternative activities in order to maximize individual utility (i.e. personal satisfaction). Micro-economic reasoning is thus applied to the analysis of time use, assuming that people are maximizing the value of their time (Becker, 1965, 1981). The application of this reasoning is more complicated in the household context, where members cannot be assumed to have the same priorities or interests (cf. Rosenzweig; Behrman; Safilios-Rothschild, in this volume).

The goal of the earliest time-allocation in the social sciences was to describe human activity patterns in quantitative terms (Sorokin and Berger, 1939). Sorokin collected over 5,000 records of daily activities of 100 Americans, which were analysed for time of day, age, sex, occupation, motivation, and social interaction. Curiously, sociologists appear to have made little use of these early studies:

The groundwork for the quantitative study of time in this country. . . [was] primarily laid by sociologists, such as Bevans ( 1913), Lundberg et al. (1934), Sorokin and Berger (1939), and Reiss (1959). Despite the attention given it by prominent sociologists, however, the empirical potential of time to function as a currency for sociology - representing a "hard" measure of human preferences and values - is mostly an unfulfilled aspiration. Linder (1970: 6) claims that this inability of sociologists to use their own results is due to their failure to recognize the time-scarcity problem. (Robinson, 1977, pp. 4-5)

Time-allocation methods have also been used to some extent by researchers in psychology (Zelkind and Sprug, 1974), agricultural economics (Clark and Haswell, 1970; Pimentel and Pimentel. 1979; Ruthenberg, 1980), and comparative sociology (Szalai et al., 1972; McSweeney, 1979).

PRACTICAL APPLICATIONS OF TIME ALLOCATION RESEARCH

In anthropology, patterns of time use were traditionally reported in the description of the daily round that usually formed part of the ethnographic monograph in which anthropologists reported their field research, but these were non-quantitative descriptions of the typical day and seasonal round. Since early quantitative research in anthropology was primarily in the realm of economics, quantitative time-allocation data were reported mainly as a measurement of labour inputs in production, rather than as a general description of activity patterns. Foster (1948), Lewis (1951), and Erasmus (1955) in their studies of Latin American communities, for example, provide detailed estimates of the amount of time spent in food production and manufacture, but only the broadest outlines of time spent in social and leisure activities.

The use of time-allocation data in the description of economic systems has continued in the study of New Guinea horticulturalists and foragers (Salisbury, 1962; Pospisil, 1963; Waddell, 1972; van Arsdale, 1978; Grossman, 1981, 1984), African foragers (Lee, 1969; Tanaka, 1980) and farmers (Tripp, 1982), Nepalese and Javanese villagers (Nag et al., 1978), and South American forager-horticulturalists (Johnson, 1975, 1978a; Lizot, 1977; Gross et al., 1979). Data from these Amazonian studies have contributed to the continuing debate on the role of limiting factors in lowland South American Indian adaptations, and have inspired researchers to expand this database to include other lowland societies (Baksh, 1984, 1985; Johnson and Baksh, 1987; Paolisso and Sackett, 1985). Theoretical explorations of the anthropological usefulness of formal models from evolutionary ecology, such as optimal foraging theory, and from micro-economics, such as linear programming solutions of the "diet problem," have been made using these data (Behrens, 1981, 1984; Hames and Vickers, 1982; Hawkes et al., 1982; Johnson and Behrens, 1982; several studies in Hames and Vickers, 1983).

Time-allocation data have also been instrumental in the study of technological change. Labour-saving associated with the adoption of the steel axe was studied by Salisbury (1962), Townsend (1969), and Sillitoe (1979) in New Guinea, and Carneiro (1979a, 1979b) in Amazonia. Hames (1979) and Yost and Kelley (1983) have similarly assessed the impact of the shotgun on Amazonian hunting efficiency. Gross and coworkers (1979) have used time-allocation studies to uncover changes in labour investments accompanying greater market involvement among tropical horticulturalists.

Information on time allocation is also essential to determining daily energy expenditure in studies of nutrition (Richards, 1939; Fox, 1953; McCarthy and McArthur, 1960; Hipsley and Kirk, 1965; Norgan et al., 1974; Montgomery and Johnson, 1977; Edmundson, 1977; Dufour, 1983) and energy flow in human ecosystems (Rappaport, 1971; Thomas, 1976).

A number of hypotheses in cultural evolution have been tested in time-allocation studies. Sahlins (1972) used time data on several foraging groups to support his argument that hunters and gatherers meet their food needs with relatively little effort. Boserup (1965) found time studies supported her hypothesis that per capita labour requirements increase with the shift from extensive to intensive cultivation. Reviews by Minge-Klevana (1980) and Ember (1983) have demonstrated how total labour requirements and types of labour change with the shift from an agrarian to an industrialized economy.

Apart from the main emphasis on problems in ecology, economics, and economic change (see Carlstein, 1982, for an overview), time-allocation research in anthropology has concentrated on socialization. Here researchers have found time allocation to be a valuable measure in charting the amounts of time caretakers of various kinds (e.g. mother, father, sibling, babysitter, etc.) spend with children, and in describing the varying ways in which children of different ages actually spend their time (Munroe and Munroe, 1971; Munroe et al., 1983; Rogoff, 1978). Whiting and Edwards (1974) used such data to show that differences in the tasks assigned to different children affected their social behaviour, for example that boys assigned child-care responsibilities will become less aggressive in play than other boys.

It will be apparent from even this brief overview that time-allocation research has a place in many kinds of policy-related research. Indeed, whenever patterns of time use are relevant, as in estimating how hard people work (say, for nutrition studies) or how their activity patterns change following an intervention, time-allocation research should be built into project design. Neumann and Bwibo (1987) provide an excellent recent example of how such research fits into a complex, multifaceted project design.

COSTS AND BENEFITS OF TIME-ALLOCATION RESEARCH METHODS

The ideal approach to the collection of time-allocation data would be to get a sort of "god's-eye view" of what each subject in the research population was doing all the time. This ideal is both technically impossible and ethically untenable. A research subject's right to privacy implies that whenever the research is felt to be inordinately intrusive the subject should exercise the power to stop it. Thus, any realistic method of measuring time allocation will, from the outset, be a compromise between the ideal of complete unimpeded observation of research subjects and the practical limits set by available resources for research and the subjects' willingness to be observed. Since there are a number of different ways to compromise the ideal in favour of practical reality, several distinct research strategies have evolved in time-allocation research each with its own rationale and constituency.

A Basic Dimension of Contrast: Direct Observation v. Subject Self-report

Many researchers prefer to use the informant as an observer of his or her own behaviour. Several techniques depend on the subject recording his own behaviour throughout the day, or recalling behaviour over a given time period. This resolves most ethical issues, because the informant decides which behaviours to report and which to hide. It also saves a great proportion of the researcher's time, since the research subjects themselves do the work of observation and recording.

The major drawback to informant self-reports is that, in general, there is a surprisingly low correspondence between informants' reports of their own behaviour and that behaviour as measured by outside observers. Bernard and colleagues (1984) have found that, in virtually all cases where both kinds of data were collected simultaneously, informant self-reports have been weakly correlated, if at all, with direct observations made by scientific researchers. For example, in a study of rural women's time use, McSweeney (1979) found that when the informant-recall technique was used, women failed to report 44 per cent of their work as recorded by direct observations. The main reason for such discrepancies is probably not deliberate deception by informants but rather that people tend to remember their own behaviour selectively, in terms of cultural models of "appropriate" or "significant" behaviour (cf. D'Andrade, 1974; Johnson, 1978b). It follows that the closer in time the informant is to the actual behaviour being described, the more accurate will be the description.

Methods of Direct Observation

Two primary methods of direct observation are commonly used. The first comes closest to giving a "god's-eye view," whereas the second saves much labour and cost by introducing sampling methods.

Following the Subject

The most straightforward approach to studying how people spend time is to follow them around all the time. When this is done (e.g. Lewis, 1951; Lee, 1979) it is necessary to focus on a specific individual or on a close-knit group such as a household or a migratory camp. Depending on the level of accuracy desired, a researcher can find it difficult to describe even the behaviour of a solitary individual, so the major shortcoming of this method is that the results are limited to very small numbers of individuals and short periods of time (e.g. one or two weeks). The representativeness of such data, as a sample of a community and of the varying seasons of the year, is highly doubtful. Such studies are normally done by a single researcher. Many details of behaviour are lost, since no observer can focus on details of behaviour for more than a few minutes at a time without requiring rest and diversion. If a group is the subject of observation, as soon as individuals undertake separate tasks (the normal situation), the observer must select which one to follow.

In order to obtain a representative sample of individuals in a community, as well as representative samples of times of day and seasons of the year, sampling procedures must be used. Note that in a modification of the "following the subject technique," Peet has introduced a degree of representativeness in his research in Nepal by following each of his research subjects for a 24-hour period once a month (Nag et al., 1978). However, true random sampling requires larger numbers of observations than are practical using this method.

Spot Checks

A solution to the problem of representativeness is to sample behaviour over time, making periodic spot checks of the behaviour of many individuals (Erasmus, 1955; Johnson, 1975). The spot-check technique uses a random pattern of visits or "checks" to determine what the members of the study population are doing at a given moment. By randomizing the observations by person and time, a representative sample, based on hundreds or thousands of brief observations, allows a statistically accurate picture of patterns of time use in the community to emerge.

In recent years this has become the technique of choice in cross-cultural behavioural studies (e.g. Gross et al., 1979; Hames, 1979; Tripp, 1982; Grossman, 1984; Baksh and Paolisso, 1986). This technique has seldom been attempted in complex urban societies, perhaps because the method of randomly dropping in on research subjects for brief visits is not as feasible in privacy-oriented modern groups as it is in the small communities anthropologists study. On the other hand, the willingness of modern individuals to accept random visits from an anthropological researcher- that is, a participant observer known personally to the research subject, as opposed to an unknown pollster - has not really been tested.

Generally, this technique is reliable and requires relatively little field time. Where the community is a neighbourhood, village, or other cluster of residences, it is possible to visit a great many people in the course of an hour, and observe and record their activities. If care is taken to randomize the visits so that the time of day and the order of visits is unpredictable, then rather accurate and complete descriptions of time-allocation patterns for all members of the community for periods of one year or more are obtained with a few hours of work each week, leaving the field-worker free to pursue the other goals of participant observation research.

The main shortcomings of this technique are, first, that the night hours are not covered. Usually, the hours between 8 p.m. and 6 a.m. must be described through interviews and general ethnographic observations, as when the researcher spends a few nights in a family dwelling. Second, direct observation certainly introduces observer effects on the behaviour of the research subjects. This latter problem varies in intensity from one ethnographic setting to another. For example, among the Machiguenga Indians of the Peruvian Amazon (Johnson, 1975), casual, short visits to see what someone is doing are not only allowed, but are a regular part of daily life. Spot checks fit in the most natural way. We can imagine settings where the disruption created by such visits might be massive, as in a male researcher trying to visit a group of Muslim wives, or an outside visitor calling at a Chinese village home, where everyone immediately scurries to make the guest feel welcome. None the less, such failures of the method have not actually been reported, and we may imagine more difficulties than actually would arise in field situations.

Methods Depending on Self-report

Three styles of time-allocation research depend on informant self-reports. Specific methods are even more variable, but the diversity in costs and benefits of self-report methods are illustrated by these three styles of research.

Global Self-reports

In this technique, informants are asked only once to describe the overall pattern of their time allocation. They may be asked to estimate the hours per week they devote to a checklist of activities, or to assign their time to activities as a percentage of the whole. For example, the University of California periodically asks its faculty to describe their time use: how many hours per week are spent in teaching, how many in research, administrative work, and so on. These data are then compiled and reported as averages, which are then compared to patterns of time use in previous years or used to draw contrasts between different campuses of the university.

Such data must be viewed with the greatest caution. Global patterns of time use are very difficult for informants to report with accuracy, even (or perhaps especially) about themselves. Variations over the course of a year will be missed, and informants are likely to forget about the amount of time spent in idle chatter, daydreaming unrelated to work, and other less-esteemed activities.

Above all, in this example, it is very clear to the informant that these data will be used in the state government to make the case for university funding, including academic salaries, in future years. It is not surprising, therefore, that professors in the University of California system regularly report themselves as working on university business more than 60 hours per week throughout the year. Some no doubt do work this many hours per week, especially at times of peak load, but we are entitled to be sceptical until some kind of behavioural confirmation has been supplied (i.e. by direct observation). It is noteworthy that these data are used in a number of university publications and documents without questions ever being raised concerning their reliability.

Ethnographic research experience in a community can help compensate for some of the distortion inherent in global estimates, because the knowledgeable ethnographer can identify areas in which informants are not adequate reporters and then can devise strategies to obtain compensatory data. For example, in their study of rural women's work, Deere and de Leal (1979) found that women's global self-reports virtually ignored agricultural labour. Knowing that women were indeed part of the agricultural workforce from their previous ethnographic field-work, the researchers developed a specific questionnaire concerning who performed what tasks in each agricultural plot in the sample, and did obtain good evidence of the real extent of women's labour in agriculture.

In this vein, Rogers (1984) has suggested that asking householders to list the tasks for which they normally are responsible in their family division of labour may be a useful shorthand method to estimate patterns of time allocation. Although such a method, used alone, would be vulnerable to the same criticisms as global self-reports (i.e. that we are obtaining a description of the informants' cognitive maps of 'appropriate behaviour" rather than a description of actual behaviour), this does suggest the possibility of combining such shorthand methods with behavioural methods to develop ways of estimating time-allocation patterns. For example, in a region where a certain similarity in households may be observed (a rice-growing region or a coastline of fishing communities), random spot checks might be used during repeated visits to a selected range of villages to establish common patterns of time use. Such a baseline study could develop a set of reliable predictors of time allocation. Task allocation might be one of these, as might occupation, household structure, ethnicity, and many other variables which can be determined through household surveys. Then relatively simple interviews of large numbers of households could be used to determine their relevant characteristics, in conjunction with the time-allocation study, to estimate probable patterns of time use throughout the entire region at much lower cost than a full-scale regional time-allocation study.

Twenty-four-hour Recall Interviews

A definite gain in accuracy is achieved by not expecting informants to know the global patterns of their behaviour, but instead asking them to report their activities over the previous 24-hour period. Informants can generally remember bedtimes, mealtimes, and the approximate times spent in major activities such as going to work or watching TV. By careful probing, an interviewer can help the informant go back over the previous day thoroughly, retrieving smaller, forgotten pieces of behaviour.

The limits to this method are set by the human propensity to forget routine activities and the many small activities that momentarily interrupt the more salient activities. Hence, conversations with family and friends, helping a child tie his shoes, eating a snack while watching television, and so on may well be lost to the research. Where the researcher is a stranger to the informant, there is also an unwillingness to admit to embarrassing or prohibited activities. Certain activities may be over- or underestimated, again reflecting the cultural bias of the informant or his or her wish to appear in a certain light to the researcher.

But, done with careful probing, perhaps with a checklist to remind both the researcher and the subject of behaviours that may have been forgotten, this method produces reliable data, as far as we know, and it is a widely practiced technique (e.g. Sanjek, 1972; White, 1976). It is probably most successful when an interviewer who is already known and trusted by the informant collects the data, rather than asking informants to fill out a questionnaire by themselves. Recall instruments can be carefully constructed to preserve simultaneity and organizational or chronological relationships between activities (cf. Schlossman, 1986, for details). These instruments have been successfully used and can provide reliable and accurate results even by telephone (Schlossman, 1986; Zeitlin et al., in preparation).

Informant Diaries

In this method, informants are asked to keep records of their own activities (e.g. Bergman, 1980). The reliability of the data can be very high or very low, depending entirely on how well-prepared the informants are, and how committed to the research goals they are. Obviously, if the informants view the task with suspicion or boredom, the results will be poor; but if they can be motivated to keep careful records of their activities at frequent intervals throughout the day, this method overcomes many of the lacunae due to forgetfulness that afflict the global self-report and even the 24-hour recall methods.


Contents - Previous - Next