1.3 Time-allocation surveys
1.3.2 Time-allocation methodology
1.3.2.1 Data collection
Three broad techniques are employed for collecting time-allocation data: diary keeping, observation and recall. With the diary method, the respondent records a log of activities and the duration of each, on the basis of instructions and forms provided by the researchers. This method has been widely used in developed countries (Berk and Berk, 1979; Walker and Woods, 1976; Szalai, 1972a). It minimizes recall bias (Robinson, 1977: 9) but has proved impractical in developing countries (Birdsall, 1980: 165). Evenson et al. (1980: 297) tried this method in the pre-test of the Laguna Multipurpose Questionnaire, but rejected it because respondents found it too
The observation method requires the interviewer to stay with members of the household throughout the period. For developing countries, where work is less structured, this method assures greater data accuracy. Joint activities and the intensity with which an activity is undertaken can be more accurately recorded. The observation method is also advantageous when respondents have a limited sense of time (Birdsall, 1980: 167). However, it introduces bias insofar as a person's activities, especially those of children, are influenced by the presence of an observer. Furthermore, it is very costly to implement and fewer households can be covered (Robinson, 1977: 9; Wigna et al., 1980: 4-5, Juster, 1985b: 27). Juster (1985b: 27) noted that the direct observation method involved a relatively high refusal rate, and that the refusals were not random.
A modification of the observation method is the use of random spot checks or the instant record technique. Erasmus (1955) introduced this method when he studied the activities of Mayo Indians. In this approach, hours were selected in advance and households were visited during the hours specified. The activity recorded was that which the subject was doing at the moment he/she was first seen (Erasmus, 1955: 323-324; Johnson, 1975: 303; Acharya and Bennett, 1983: 65; Reynolds, 1991: 84-86). Different methods have been used to record activities. Johnson described the activities of the Machiguenga Indians in longhand in notebooks at the time of the visit and later coded the different activities. Acharya and Bennett, by contrast, used a pre-coded form containing a detailed set of activities listed vertically with space to record the name and person number of each household member across the top. During the visit, the interviewer observed what each household member was doing and recorded the observation by marking the corresponding box on the activity list under that person's name.
An advantage of the random spot check is that it consumes very little time and does not allow the observed individual to alter his or her behaviour. However, it also has disadvantages. The most serious is that activities involving movement or
travelling are missed (Reynolds, 1991: 84-86). Similar to direct observation, individuals are usually observed in particular sites such as the village or homes; thus many important activities that occur in other place are missed. The other disadvantages are that the interpretation of activities depends on the enumerator, sequences of tasks are unlikely to be captured, and crises of health or family problems which may have affected an individual's activities are also unlikely to be noted.
With the recall method, an interviewer asks the subject to narrate what he or she did during a certain period. The interviewer either prompts the subject regarding specific activities, a list of which is available to the interviewer, or asks the respondent to relate the activities in his/her own words and in sequence. Alternatively, the respondent is asked how much time has been devoted to certain activities or how often he or she has been engaged in various activities. Andorka (1987: 151) refers to this as 'stylized questions'. Recall data are cheaper to collect, since interviewers spend less time per household. The limitation of this approach is that people forget (Converse, 1968: 45). However, White (1976: 434-435) found that the Javanese respondents in his study could correctly recall time to within a quarter of an hour. There were several features in everyday life, such as the invariable time of sunrise and sunset, school time and the five obligatory daily Moslem prayers that helped respondents in remembering the time devoted to different activities.
Secondary and concurrent activities are not usually reported in the recall method of data collection; yet these can be critical to analysis (Birdsall, 1980: 168). White (1976: 436), for example, divided the recorded time between simultaneous activities. However, the arbitrary splitting of hours was not realistic. White found that the households which reported the most time in child care appeared overall to spend the least time in cooking. The number of hours devoted to cooking while
baby-sitting probably were hardly more than time devoted to cooking so that the reduction in cooking time was not practical.
Reynolds (1991) used the observation, random spot check and 24 hour recall in collecting time-allocation data from women, girls and boys in the Zambesi Valley (Zimbabwe). A comparison of these methods showed that each technique lends a particular perspective to people's activities. The most striking feature was that the order of magnitude within groups remained consistent across all three techniques. That is, the observation method recorded the highest number of activities related to work for women, boys and girls, the random spot check recorded the highest number of activities related to self-care and the 24 hour recall recorded the highest number of activities related to leisure. It was also found that respondents not only reported the lowest number of activities related to work but also underestimated the time spent on these activities when asked to recall their time use (Reynolds, 1991: 86-87).
These findings have important implications in the current study because time- allocation data were collected through recall from Laguna households. There are certainly inaccuracies in recall data when compared with the other methods of data collection, but the flaw in the Laguna data may not be as serious as indicated from the results of the data collected through recall in the Zambesi study because of the differences in the conditions prevailing in the location of the two studies. In the Zambesi study, few people had watches, clocks, radios or other means of telling time (Reynolds, 1991: 69), whereas in the Laguna study, nearly all of the households had radios which helped the respondents in recalling the number of hours spent on different activities. In addition, a list of activities was used in the Laguna surveys and respondents were asked the amount of time spent on the activities. In this manner, respondents were prompted to recall the activities which they might have inadvertently forgotten.