Appendix A
American Time-Use Survey
Diane Herz
The proposed time-use survey to be conducted by the Bureau of Labor Statistics (BLS) would draw from the retired sample from the Current Population Survey (CPS). The CPS is a monthly survey conducted by the Census Bureau. The American Time-Use Survey (ATUS) sample would be designed to be representative of the U.S. population age 16 and older. (This recommendation may be changed to 15 and older for comparability with time-use surveys in other countries.) The survey will be designed to produce quarterly estimates of the proportion of time spent in different activities for the population and separately for a set of comparison groups. Potential transportation statistics could be gleaned from this survey, such as the purpose of trips and the time spent traveling.
The time-use survey statistics team at BLS is currently determining the best way to stratify the survey sample and assign respondents to designated days for interviews. The sample is likely to be stratified by a number of household variables, possibly including number of adults in the household; presence of children; or age, education, and race/ethnicity of householder. It may not be feasible to stratify the sample by gender, due to the sample size.
The survey will be designed to generate annual estimates of a wide range of activities for an average week, weekday, and weekend day. Draft activity classification codes have been developed. They are based on the Australian classification system and are comparable with most international time-use coding schemes.
The goal of the survey is to have 24,000 completed interviews per month. Developers are assuming a 70 percent response rate. The CPS interviews about 150,000 individuals in approximately 72,000 households per year. Given that one member of each household is eligible for the proposed time-use survey and given nonresponse over the course of the CPS, the maximum sample size for the time-use survey is 72,000 per year. However, because the CPS oversamples small states and the goal of the survey is to be nationally representative, the maximum available sample size is about 54,000 per year. A subsample of these 54,000 will be drawn for the time-use survey.
The proposed strategy to gather information on an individual’s time use is to use the designated day approach: each individual in the survey will be assigned a day of the week for which he or she will report activities. An attempt to interview the respondent will be made after the designated day. If the respondent cannot be reached for the designated day, the respondent will be reassigned to another designated day. In the case of Friday through Sunday, the respondent will be assigned to the subsequent Friday through Sunday. In the case of a Monday through Thursday, the respondent may be reassigned to another Monday, Tuesday, Wednesday, or Thursday. The exact assignment pattern for those in the Monday through Thursday sample has not been determined.
The period of fielding will probably be between 4 and 8 weeks and may depend on whether an individual is part of the Monday-Thursday sample or part of the other 3 (Friday, Saturday, and Sunday) samples. The distribution of samples across the days of the week has also not yet been determined; the distribution will be defined in order to minimize variances on the estimates of activities on a day, weekend day, and weekday.
The data will be collected continuously, with the samples drawn monthly. Results will be reported quarterly and annually, so that seasonal variation in time-use patterns is illuminated. The data will not be collected on major holidays.
The survey instrument will consist mainly of an activity questionnaire (the time-use diary), which will document activities done by the respondent over the preceding 24-hour period. The survey will use computer-assisted telephone interviewing and respondents will be asked to recall the timing of their activities sequentially. Respondents will also be asked where they were during the activity, whom they were with, and whether they were doing anything else at the same time, in order to record simultaneous activities. Respondents will also be asked what activities were done for pay to
be able to better identify market and nonmarket activities. Publication tables will include information on activities performed as a sole activity and those done in conjunction with other activities.
In addition to the time-use component of the survey, several summary questions will be asked. The current questions in review ask respondents about passive care of children (“looking after” children), care of dependent adults, and extended absences during the previous month.
Other data on respondents will also be collected, including updated (from the CPS) household composition information, updated total family income (categories), the respondent’s labor force status, the labor force status of his or her spouse or partner, updated earnings information for the respondent, and school enrollment. The current plan is to develop a public-use database to be made available for the research community (meeting confidentiality restrictions).
The average length of the interview is projected to be about 25-30 minutes. Completing the diaries is expected to take about 22 minutes of the total. This estimate is based on the pilot test results for the time-use component and on experience with the length of the CPS interview.
Because the sample members for the proposed time-use survey are also CPS sample members, the new data could be linked to the various CPS supplements. The BLS working group also considered several topical modules that could be attached to the time-use survey, such as tool use, child care, elder or adult care, working hours, division of labor within the household, household production, volunteer activities, subjective assessments of activities, and subjective questions about the experience of time.