Network on the Family and the Economy Home Page
Members Index Internet Resources Index Meetings & Conferences
contact info | more about the MacArthur Foundation Research Networks Information

Session VI: Time-Use Surveys: Where should the BLS go from here?

timeuse.gif (2687 bytes)

Michael Horrigan, Bureau of Labor Statistics (BLS)

William Nordhaus, Yale University
Lorna Bailie, Statistics Canada
W. Keith Bryant, Cornell University

Michael Bittman, University of New South Wales

Michael Horrigan: He has no official paper, but does have opinions in this session. Out of this conference, he would like to see:

A list of suggestions to be used by the BLS to inform their next steps in the next two years or so.

What should be the longer-run perspective?

Why are time-use surveys important and what are the implications?

A discussion on the go-between immediacy of the near-run and prospects for the long-run.

William Nordhaus: He is not a time-use survey person, but he finds it fascinating and a terribly important area. Why is it important?

  1. Inadequate data on time-use is the single most important gap in federal statistics. There is a wide variety of implications to this, and it is something that can be easily fixed.
  2. We do not know in a comprehensive way where our children are and what they are doing. We do not know how many hours we are working. What are our elders doing? The median retirement age is 62, and people generally have about another 20 years to live. Much of the population is in this segment, but we do not know what they are doing. Are they doing civic activities, taking care of children, watching TV? In Sweden, retired males spend more time on housework, gardening, TV and reading, up 10 hours per week, an additional 15 seconds per week on organizations and 10 seconds per week on child care. Religious activities went down.
  3. Jim Tobin and he did work on this some time back, and then, as now, time-use data was not very good for non-market activities. This comes up again in environmental accounts. Most of these are non-market activities. We do not have the data to develop these accounts. When the EPA (Environmental Protection Agency) develops models of exposure to air pollution, it is based on the expectation that we live 24 hours a day on top of a pole at an airport. But what is important is how much time is spent inside and outside. There is much money based on modeling without good information about where people really are.
  4. The CPI (Consumer Price Index): Time use is crucial for medical care, especially as we move toward managed care. We ration the amount of time we have to wait. The monetary prices are going down, but the time prices are going up.
  5. Also, there is a shift toward outlets, bias of 10%, with a movement from mom and pop stores to big stores. There is a convenience-time saving to shopping this way. We do not have data on this either.

Some methodology issues:

  1. Technical questions about how to survey. (He does not know anything about it.)
  2. There is the problem of boundaries between activities and the difficulty of capturing them. It is very important in our society, because of the blending between home and work. This makes life tough for a survey.
  3. There is the problem of joint production in economics. Sheep produce both wool and mutton. Many of us traveled and read on the plane flight here. We really are doing two things at once. This needs some careful thinking. We are doing different things — physical, cognitive, etc. He agrees with Thomas Juster, we need the discipline of a 24-hour day in the data. We can classify by final activity. Or, count by what we are doing. Maybe we need to partition the animal (sheep) into a fraction of wool and fraction of mutton. These are very interesting methodological issues.

Where do we go from here?

  1. He is convinced that we need to do lots of educating of our public policy-makers about the importance of this issue.
  2. He strongly supports the BLS in this visionary activity.
  3. Regarding substance, we need replication of early studies (e.g., Michigan studies) to provide historical continuity to our data. We could also add on, but it is crucial to have a comparable benchmark.
  4. Undertake a comprehensive national period time-use study. We need it, but he is not sure where or how to do this. If the BLS agreed, or Congress forced them to do this, it would be great. It will be expensive in absolute terms, but not in the context of what else we do in government.
  5. The methodological issues need serious attention.

Lorna Bailie: From the pragmatic approach, consider the time, cost and response burden. Statistics Canada is currently testing a survey which will enter the field shortly. One reason for doing the survey is to get a look at unpaid work activity and related activity. The key — was this activity helping someone outside the household? This is tricky to do on a telephone survey.

Reasons for doing this: National income accounts, quality of life, and families and children. (as Folbre said yesterday). How do you get from the concepts to the measurement? We need:

comprehensive data
complete data
multiple time diaries
to use the diary in combination with stylized questions
comparable concerts
historical data (as Stafford said).

We must keep in mind that there are limits to the resources. What is the objective? Let’s get the right players in the room and decide what to do. For the BLS, doing the pilot study and collaboration with MacArthur are good. They need to cooperate with others more, be flexible, do sampling and sub-sampling.

There are operational concerns such as using a CADI system, testing in-line coding and whether to use the telephone vs. a face-to-face approach. The telephone worked well for Statistics Canada. It costs one million dollars for the sample of 10,000 people, which seems reasonable.

Michael Bittman (Replaced W. Keith Bryant)

He will be commenting as someone who has been involved in time-use outside of North America. Other countries doing such work are Canada, Norway, Bulgaria, Australia, Laos, and South Africa. More surveys are coming in Third World countries. On the whole, they are large (10,000+) surveys. The Italians surveyed 30,000 diary days. These are undertaken by official statistical bodies. The lesson is that to get these, we need official government bodies to get behind the survey.

The US is a little schizo in this way. You have the finest literature on time-use surveys in the US (e.g., the Michigan study, Robinson’s study). However, it is puzzling how such a powerful nation has not managed to do this work nationally. You have to change this. You need a major survey. Funding is degenerative, samples smaller and methods cheaper. The pilot work we have heard about is very encouraging and of good quality.

Because surveys are being done in other countries, this is an important time. In the 1960s, we got a common classification system. Once we had this, we were able to look back over time with equivalent systems. This also meant we could make comparisons between countries such as the effects of different social policies, or equity in non-market work, etc. Having comparable time-use classifications allows this to be done. He urges the US to do it, but you need a core survey comparable to those outside the US. The next few years are critical because Eurostat has sponsored a project on time-use in Europe. The advantage of doing it at the same time is that you get lots of free goods, and many are studying all these things at one point in time across different nations. The US needs to get moving.

Core demand is comparable with what came before (other countries consulting with Eurostat). They might be doing everyone in the household over age 10, and not by telephone. The US may want to consider this to maximize comparability.

What else would he like to add?

  1. Some countries do not ask about secondary activities (like Canada). Australia does, and they find much more child care and radio consumption. Without it, we miss conversation between couples. This is important if we are interested in non-market work.
  2. He has seen many extra columns on diaries come and go. "For whom is this work don?" may be one column that stays. The United Nations now has a new classification for their time use, including this column.
  3. Add on modules are a very good idea. One can do cross-survey linkages. In Australia, we have full-blown questionnaires, and then lots of modules (child care, etc.).

This is the most important political moment for time-use in the US. Now is the time to get on the train. It is leaving the station, and the US needs to get on it.

Group Discussion and Comments

Robert Michael, University of Chicago: There is an analytic distinction between "joint product" and "multiple product." Joint production is one activity with more than one output.

Consider a two-tier approach: The first tier is a large landscape to get broad national numbers. The second tier gets at more activities and details. Modules are one way to go.

He anticipates that within public policy, there is money illusion. We only look at what we get, and what we get is the dollar as if the dollar is all that matters. We need more facts than the dollar (as with the example of medical care). We need to get the issue of time use on the table for policy deliberations. It is better than not having it out there.

Timothy Smeeding, Syracuse University: He agrees that time is critical. The BLS needs to know that there are allies out there to work with, e.g., the National Science Foundation (NSF), etc. We need to build OUR roads and telescopes. If we do it right and collaborate, we can really make some progress. In 1973, the NSF spent $1.2 billion on time use, now they are spending nothing. We need to go forward and get the appropriations.

Klas Rydenstam, Statistics Sweden: In Europe, time studies have been done in many countries by national statistics agencies for many years, and will continue. They are seeking more comparable data in the future. Eurostat is trying to do this by looking for a common denominator of all surveys. They do not want to change it and lose comparability with previous background data.

Europe has different traditions. Their design differs from what is discussed here. They use multiple diary days. The diary is a self-administered paper and pencil, left behind to be filled. They do not rely on telephone surveys, and they have households as the unit of study. They use multi-diary days and do all household members. The diary recorded the main activity, secondary activity and "with whom they did it" and "for whom they did it."

It is very important to get secondary activities, especially for differences between men’s and women’s days. They have 10-minute fixed time intervals. The number of activities in Sweden 36 for women and 25 for men without counting secondary activities. A diary with fixed interval times gives data which is more detailed as opposed to one with open time slots.

The future of the project is unknown because of budgetary reasons, but he hopes it will be finalized soon.

Thomas Juster, University of Michigan: There are two ways to design the survey. Do you want to use multiple diaries per person for micro analysis or one large diary? The multiple-diary approach gets micro information, but the season and day of the week does matter. If you want child care, you have to do both spouses. In 1975, they did four interviews each quarter of the year and did both spouses on the same day. In this system, analysis dominates the design. This is different from what we may want for the census, lots of people, a great big dependent variable survey. These are very different philosophies. Time use is critical and should be trackable over time. He would like to see both done, because they are not substitutes. You get different things from each.

Michigan never got details for what people did in their market-work regarding on-the-job learning, flexible, etc. We need to develop codes for what happens at work. This is important, and no one has done it.

Robert A. Pollak, Washington University: He seconds most of what others have said. The issue of purpose is important both at the broad general level and the micro level. What you collect and how you collect it (the underlying theory). We need to think of purpose first, not after the data is collected. Index is moving toward a welfare-index that includes leisure. There is lots of interest in looking at non-market time. One data is output-based, a different project. If we are not going to do both, then we need to think about which one we would rather do. We should not dichotomize into work and non-market work categories because this is misleading. Rather, we should look at what is actually happening in both markets.

The CPI issues are interesting and difficult, e.g., add waiting time. There are different types of "waiting time" we would like to pick up. We need a model of "search and shopping." There are lots of prices out there. It is important to collect data on time use and on goods in order to make sense out of the CPI.

Reed Larson, University of Illinois: He was impressed by the WESTAT study, but also struck by the methodological. He wants to stress the need for the "point sampling" method. It is good for getting around memory problems and missing short activities. One can ask more follow-up questions, i.e., 30 or 40 questions about an activity such as clarification if paid, child involved, etc. before a participant gets frustrated. With one sampling strategy, it is inefficient to sample across the day, because they are too representative, repetitive, and not independent. But we may be interested in linkages, representativeness; then it is better to sample across days or the week as they do in his sample. Also, you do not get the attrition that you get in some places by using the telephone. Once people had the point-sampling method technology equipment, they got an approximate 80-85% response rate to the beeps in his study. You can use either low technology like beepers and/or high technology like pocket computers to ask participants about their time.

Point sampling does have its disadvantages, but it has many advantages. He wants people to start to think about it as an option for certain purposes.

Paula England, University of Arizona: From the gender perspective: Do women indeed work the "double shift" today? John Robinson’s data does not show employed woman working much more than unemployed women. This does not pass the Grandmother Test for her. Women are indicating other activities, like leisure, rather than child care as the primary activity. We can read between the lines. Most child care is reported as a secondary activity, and women probably do more of it as secondary activity. The BLS needs to ask about primary and secondary activities, and analysts cannot throw away this information if we want to get gender equity in total hours of work or to say where our children are and when they are being watched by a household member. With the "for whom" question, we find that women do a lot of public-good provision, i.e. well-behaved children are public goods.

Duncan Ironmonger, University of Melbourne: In an earlier joint production example, wool and meat are the outputs, and the sheep is the input. One way to solve the joint production problem is to count both. For national accounts of household production, we need to collect data on outputs and the prices of those outputs. We need to count the time use and the value of the inputs and outputs too.

The US has to get into the field by 1999 to be with other countries. The US needs to get on the train now. It may be useful to do it as a continuous operation. Maybe have a five-year program or a continuous operation like the US Household Expenditure data collection. It may have lower costs due to less start-up costs. once trained everyone, cheaper than doing it every 5 years. Complete national time accounts very much needed where valuation is not necessary.

Katharine Abraham, BLS: Budgeting concerns: We have missed the funding window for 1999, so we cannot get anything done in 1999. We are now working on activities for the year 2000. There are long time lags getting things through the budget process. Realistically, funding for federal programs will get tighter and tighter in the next couple of years, so we will need to prioritize. She thinks caps are real. The question we will have to answer is not how would we like to collect data, but which is a higher priority of what we are currently doing. So, help us — what can the BLS stop doing in order to do this work instead?

Andrew Harvey, Saint Mary’s University: We wanted everything from everyone, but we tried a split-sample which worked well. We do not need all the information from everyone. There are many innovative ways to work with data. Do random sample episodes. The time-points method is a possible way to go if you do not want detail on more than 7-8 activities. You cannot recreate a day with it.

If one is traveling and preparing for teaching, which one is the primary activity? You can have two primary and two secondary activities as they have done successfully in the 1981 study. They used codes to specify and analyze on both dimensions. It is not hard to get.

Their time-use web site address is:

Clyde Tucker, BLS: Some methodological points: It is hard to get people to keep their own diaries and get all the detail.

The main thing is that interviewers must know why you are collecting the data. It may help them to code some of the data by coding someone else’s interview before they go out and do their own. This will help them understand what to look for.

Session ended at scheduled time.

back to top