Measuring the well-being of the population is essential to understanding the effects of public policies. One common measure of well-being is the value of the consumption of marketed goods and services, such as food, shelter, and recreation. A common theme at the workshop was that a broader concept of well-being should include the consumption of nonmarket goods and services. To do so requires data on how much time is spent producing goods and services not sold in the market. This chapter discusses some of the conceptual issues surrounding how people allocate their time towards different activities (both market and nonmarket), how activities are measured, and how time spent in nonmarket activities is valued.
GENERAL CONCEPTUAL APPROACH
The framework usually used to describe time allocation decisions assumes that people allocate time among alternative uses to satisfy their needs and desires. In economic language, individuals allocate time between leisure and work (both market and nonmarket work) subject to income and time constraints and according to their preferences for goods and the satisfactions provided by different uses of time.
On careful examination, classifying time into market work, nonmarket work, and leisure is a difficult conceptual and operational issue. Market work usually is meant to refer to work done for pay. Yet, as workshop participants discussed, not all time spent at “work” is devoted to job-related activities. Some time is devoted to personal maintenance, such as eating and grooming.
Some time is devoted to nonmarket production, such as searching for a plumber to fix a household problem. Some time is devoted to purely leisure activities, such as talking about yesterday’s football game with a colleague or checking the Internet for movie schedules. Furthermore, as work schedules become more flexible and more market work is conducted at home, the lines between market work, nonmarket work, and leisure may become even more blurred. Although it may be possible to measure time use distinctively enough to separate nonmarket work from market work while at work, many activities can be classified as both work and leisure. And, people sometimes receive enjoyment or satisfaction from doing their jobs, which complicates how work is conceptualized.
Nonmarket work is typically thought of as unpaid time devoted to activities that produce a “commodity,” such as cooking, house cleaning, or mowing the lawn. Nonmarket work and leisure can also be intertwined. For example, gardening is a nonmarket activity that produces a tangible output, such as tomatoes, that can be sold in the market or consumed by the gardener. Gardening may also be a leisure activity, as the gardener may enjoy the process of growing the tomatoes, working with the soil, or enjoying the outdoors. Nonmarket outputs are often difficult to measure, both conceptually and operationally. Although the outputs to cooking and cleaning are easily measured conceptually, such as the number of clean shirts and the number of cooked hamburgers, some household activities do not have easily measured outputs, for example, caring for a relative or reading to one’s child (a point we discuss further below in the section on valuing time).
Most standard approaches to economic valuation assume that leisure time provides satisfaction to an individual that work and other productive activities do not, but several studies challenge this assumption. University of Michigan studies in 1975-1976 and 1982-1982 asked respondents to rate the level of intrinsic enjoyment they received from many activities throughout the day, on a scale from zero to ten (Juster, 1985). Respondents gave high enjoyment rankings not only to activities for which there were interactions with children, but also for paid work activities. They gave low rankings for activities involving household work. In surveys by Robinson and Godbey (1997), time spent at work gave respondents more satisfaction than time spent at many nonmarket and leisure activities, suggesting that time spent at work may not be all disutility—indeed, it may not be “work.”1
Similarly, individuals may receive direct benefits from nonmarket work activities. As noted above, gardening (or fishing) may be a means of produc-
ing food. However, they may also be activities that yield direct satisfaction. In many cases, the value of the fishing is likely to far exceed the value of the fish.
Tom Juster proposed a conceptual definition of activities to distinguish those that are work (both market and nonmarket) from those that are leisure. The proposed distinction is determined by whether the activities provide or produce extrinsic rewards or intrinsic rewards. Extrinsic rewards are the results or products of activities that produce future utility or satisfaction. For example, market work has extrinsic rewards because a worker receives income that is used for future consumption; similarly, doing laundry has extrinsic rewards of producing clean sheets. Intrinsic rewards (also called “process benefits”) result from activities that provide direct utility or satisfaction (Juster, 1999). For example, a drive to look at the autumn leaves is intrinsically rewarding; it is not taken to actually go anywhere (indeed, such an activity almost always goes in a circle).
As workshop participants noted, these distinctions are not always clear. Hiking may be immediately rewarding in body and spirit, but the physical activity may also provide future health benefits. Although people may work primarily for the extrinsic reward of wages and benefits, work may provide intrinsic rewards if, for example, workers enjoy interacting with their coworkers. Reading to a child may have intrinsic rewards in that it provides direct satisfaction. It may also have extrinsic rewards in that it improves a child’s verbal ability and is an “investment” in his or her future.
The degree to which an activity has intrinsic or extrinsic rewards or costs may also depend on the time of the day of the activity and its sequence in relation to other activities. For example, if someone dreads going to work, a 20-minute walk to work might have lower intrinsic rewards than a 20-minute walk home from work at the end of the day. Likewise, a call from a good friend very early in the morning may have fewer intrinsic rewards than a call from the same friend in the evening. The intrinsic and extrinsic rewards of an activity may also depend upon who else is present for the activity. Having dinner alone versus having dinner with one’s spouse will likely have different intrinsic rewards. Each of these examples underscore the conceptual factors that complicate classification of activities as either intrinsically or extrinsically rewarding.
The discussion on how to define activities has implications for data collection, a point we briefly mention here, but discuss further in Chapter 5. For example, if shared time with family members and loved ones is an important part of the intrinsic rewards of an activity, as some evidence indicates (Bryant and Zick, 1996a; Bryant and Wang, 1990), then it is important to also gather information on who else was present during an activity or to collect data on multiple household members. It might also be useful to understand respondents’ satisfaction level while they participate in the activities to get a better sense of why they engage in the activity, in the hope that intrinsically reward-
ing activities can be better distinguished from extrinsically rewarding activities. Similarly, it is important to collect contextual data about an activity, such as where the activity took place, at what time of day, and who else was there, to understand satisfaction and the intrinsic rewards of activities. Because these issues cut across the fields of psychology, sociology, and economics, understanding them requires an interdisciplinary approach involving all the behavioral disciplines, as discussant Nancy Folbre noted.
HOUSEHOLD TIME USE
The theories that have been developed to understand how individuals and households allocate their time among activities to produce and consume commodities and what affects this allocation depend on clear definitions, conceptually and operationally, of the inputs and outputs of the process. Household production theory assumes that households engage in productive activities that use as inputs market goods and services, the time of household members, and household capital stock. These inputs are used either to produce satisfaction or utility directly or to produce other goods and services that in turn produce future utility or satisfaction.2 The theory covers a wide range of goods and services produced by the household—cleaning, cooking, caring for children and other relatives, recreation, procreation, and education, to name a few.
In a formal model, preferences of the household (or for multiple person households, preferences of the individual members in the household in combination with intrahousehold bargaining) determine what is produced. For example, a household that likes to barbeque and have a place to sit outdoors may decide to build a deck. The household must also decide how to build the deck, by family labor, through a market contractor, or by barter for household services. One important determinant is the cost of outside labor in comparison with the subjective value, or “shadow price,” of household labor. Other determinants are the skills of the householder and capital requirements. Households generally produce their own dishwashing services but seldom produce their own cardiovascular surgery services. Likewise, a household may have the capital and technology available to produce a meal for several people, but probably not for a hundred people. This “make or buy” decision is a complicated function of prices and skills, of which the price of the time and the quality of the output are two of the central determinants.
However, as Robert A. Pollak pointed out in his paper, there are several complications to the theory of household production, related to defining the
commodities and dealing with simultaneous activities. To continue the example of building a deck, household members may get satisfaction out of working with their hands and generally enjoy building the deck themselves. In other words, in addition to receiving extrinsic benefits from the completed project, the builder may also receive intrinsic rewards, or process benefits, from engaging in the activity. Two goods are produced, the deck and the satisfaction from the process of building it. Ignoring the process benefits ignores some of the utility derived by the household. This example shows that while the outputs produced by a firm are usually fairly easy to measure, the outputs produced by a household are often hard to measure, and the intrinsic benefits of household time use (including time for work done for pay) are almost always extremely difficult to assess.
One way time-use data and household production theory have been used together is to try to understand how parental time spent with children affects a child’s outcomes. Understanding how the inputs of this activity (time, capital stock, and market goods) affect the output is a very interesting question. It requires data on the inputs to the process, meaning that time-use data should be collected in addition to data on other inputs used in the process. In her remarks, discussant Jeanne Brooks-Gunn stated that in doing so, it is important to clearly differentiate activities in categories as much as possible. For example, learning activities should be differentiated from playing activities. Understanding how parental time affects child outcomes also requires clearly measured outputs. Yet even with clearly measured outputs and inputs, it may be difficult to separate any effects of parents’ time with their children from unmeasured characteristics of the parents that also contribute to children’s outcomes (Pollak, 1999).
Other workshop participants stressed that it is important to collect comprehensive information on the inputs used for household production. Discussant Jack Triplett noted that information on the time and other inputs used in producing outputs in the household can also help improve measures of prices and output in the service economy. Triplett used the example of banks that provide ATM services to customers, which save time for the customers. In valuing the output of ATMs, Triplett argues that it is important to have some information on how much time is saved, by whom, and when it is saved, and the valuation of that time. Nancy Folbre emphasized the need for data on inputs and human capital of household members in order to better understand household production processes.
Time-use data are also valuable for furthering understanding of how household members allocate their time and goods. The theory of intrahousehold allocation postulates that who does what in the household is a function of the value of time of individual household members, as well as the household’s wealth and capital stock and the skills, preferences, and bargaining power of household members. For some household time-use activities, one
household member’s time could be substituted for another household members time. For example, one spouse can do the laundry while the other spouse goes to the store for groceries. The combined productivity from both activities is likely to be higher if one spouse does one activity while the other spouse does another. Yet, for some activities, spousal time use may be complementary, that is, there are gains to having both spouses present for the activity, such as the example given above of a husband and wife dining together.
Another interesting application of time-use data is to examine intrahousehold allocation of time and goods. One example involves the allocation of housework among male and female adults in a household. One reason that women do more housework might be because they have a stronger preference for a cleaner house than men; another might be that men have more “power” than women and allocate the unpleasant tasks to women. It may be possible to test these hypotheses with time-use data by looking at the time that one-person male and female households spend cleaning (controlling for other variables and differences in values of time). However, differences in housework by males and females in single-person households might not reflect preference structures; they might instead reflect the attempt by women to build “gender human capital” that can be used in the marriage market: that is, a single woman may cook to make herself more attractive to potential spouses. Or, people in single-person households might be inherently different from those in multiperson households in ways that are correlated with their time use so that the results from single-person households cannot be directly applied to multiperson households. At the very least, the availability of time-use data can provide descriptive measures of how people and households allocate their time and how variations in technology, wage rates, and prices affect those activities.
VALUATION OF HOUSEHOLD PRODUCTION AND TIME USE
A primary conceptual issue in measuring nonmarket output of households is in identifying the nonmarket outputs, that is, identifying the output of a nonmarket activity as a production activity rather than a consumption activity. Some commodities are easily identified, such as a home-cooked meal, which has a clear input and a measurable output. Other commodities and activities are less clearly defined. For example, reading a story to one’s child could be classified as a leisure activity for both the parent and the child. It could also be classified as a production activity, either as a child care activity or an activity to further the child’s development (or both). Viewed as a productive activity, the parent is providing a service that, while not purchased, is nonetheless consumed by the child. The service the parent provides could be sold in the market. But the reading is only one component of the service provided by the parent for her own child; the joint activity also
includes the parent and the child spending time together. It is this additional benefit (output) of the activity that is not easily classified, measured, or valued.
Assuming output can be defined, another conceptual dilemma in measuring the nonmarket output of households is how to value the output produced. One approach is to use a price measure of the output produced. In other words, assign a dollar amount to the output produced by a household that is equal to the amount for which the output could be sold in the market (which would have to be estimated by the price of a proxy good or service that is actually sold in the market). However, as Landefeld and McCulla (1999) note, the producer of the household good is often also the consumer, so there is no transaction. Furthermore, some household outputs could not be sold in the market and therefore have no good proxy measure, as in the parent-child example in the previous paragraph.
The other approach is to value household outputs by adding the costs of inputs used to produce the output. Nonmarket production, like market production, is a process of taking existing goods and services, existing capital stock, and the time and ability of a producer to create an output (product). Valuing the intermediate goods and services and capital stock used to produce a nonmarket commodity is not a difficult task conceptually (although the actual measuring might be). The goods for inputs purchased by households are already included in the national income and product accounts. However, valuing the labor time used to produce a commodity is a difficult task that requires making assumptions about an individual’s value of time spent in nonmarket activities.
There are two general methods for valuing time spent in nonmarket and leisure activities: the opportunity cost approach and the market cost approach. The opportunity cost approach is based on the view that individuals allocate their time so that the value of an additional hour spent in market work activity is equal to the value of an additional hour spent in nonmarket or leisure activities. To “buy” one more hour of time in nonmarket activities, one must give up the pay from one additional hour of work for pay. This approach uses an individual’s post-tax hourly wage rate as a measure of the value of an hour of market and nonmarket time. The opportunity cost approach is conceptually and operationally appealing, and it is consistent with economic theory. Moreover, wage rates of employed individuals are fairly easy to obtain from survey data.
Workshop participants noted some complications to this approach, however. First, people who do not work for pay have no wages to measure. For them, the “reservation wage” or wage rate at which the person is indifferent to working or not working may be the relevant wage rate to use. However, the reservation wage rate is not observed and therefore, not easily measured. It is also not at all clear how to measure the opportunity cost of time for children,
who obviously do not work for pay. Another complication of the opportunity cost approach is that it ignores any intrinsic reward or process benefits gained from activities. Nonmarket time spent taking out the garbage may have process costs while market work may have process benefits. If the wage rate is used to value the time spent in both types of activities, it implicitly assumes the process benefits for both activities (and across all other activities) are the same. A more practical concern with the opportunity cost approach is that it also assumes that people can allocate exactly the number of hours they wish to work. In reality, hours of work are often fixed by custom or law, and in many salaried jobs the marginal pay for an additional hour of work (above 40 hours of work) is generally zero, at least in the short run. Finally, wage rates do not account for any fringe benefits a worker may receive, which may distort the wage rate as a measure of one’s opportunity cost of time.
The market cost approach to valuing time uses the wage rate of a substitute provider of the activity. For example, to value an hour spent building a deck, the market wage rate of a carpenter would be used. In its simplest form, this method has the shortcoming of assuming that there are no quality differences in a deck built by a carpenter and a deck built by the householder. To get around this issue, the wage rate of a generalist is used rather than the wage rate of a specialist: in other words, the wage rate of a handyman is used instead of the wage rate of a specialist carpenter. For the market cost approach, as in the opportunity cost approach, any process benefits are ignored. The net effect may be to understate the value of a home-produced good because it does not account for any process benefits the household obtains from the activity (Landefeld and McCulla, 1999).
These conceptual issues have been understood for many years, and no easy resolutions have been proposed. It is clear that further efforts to define and measure activities, to value activities, and to develop better models of how individuals and households allocate their time are needed. Time-use data, with clear definitions of activities, identification of simultaneous activities (see below), contextual data on who else is present during the activity, measures of inputs used in household production processes, and information on wage rates of the respondents, will aid in the resolution or refinement of these conceptual issues.
At any time, an individual may be engaged in more than one activity. It is possible to cook dinner and watch a child at the same time (or at least be “on call” if the child needs help). Similarly, one can watch television and fold laundry at the same time, or read a newspaper while riding on the train. Even if a perfect classification system for activities that could separate work from leisure was developed, it would still be difficult to separate activities con-
ducted simultaneously, or jointly. Ignoring these simultaneous activities will miss many daily activities. Therefore, it is important to consider the different approaches to measuring time spent in simultaneous activities. Some of the alternatives are reviewed in Nordhaus (1999).
One method for measuring time spent in joint activities is to count an hour spent jointly doing two activities as two hours. That is, jointly folding laundry and watching television for an hour would be measured as an hour of folding laundry and an hour of watching television. This approach is unsatisfactory, however, because it does not satisfy the constraint that a day has 24 hours. Moreover, it is likely that the number of hours of activities would increase as surveys became more detailed. The method also presumes the amount of laundry folded per minute while watching television is the same as the amount of laundry folded per minute of time spent solely folding laundry.
Another approach is to count only one of the simultaneous activities, which involves a determination (by the respondent or by the analyst) of which activity is the primary activity or the one that is more important. Most surveys of time use have asked respondents to designate a primary activity and a secondary activity. Usually, respondents are asked to report what they are doing at a certain time and whether they were doing anything else at the same time. The primary activity is then classified as what the respondent first reported, and the secondary activity is what the respondent reported doing in addition to the primary one. In accounting for hours spent in a day, only the time spent in the primary activity is counted; the sum of secondary activities may be counted and tabulated separately. Counting only primary activities in producing statistics on daily totals of time use in different activities would not count many meaningful activities that are likely to be reported as secondary activities. For example, a family may eat a meal together and spend the meal discussing the day’s activities. Most likely, the primary activity the respondent lists would be eating, and the secondary activity would be visiting with family members. However, as Julie DaVanzo noted, the actual eating part may take only a few minutes, while the visiting takes more time and may be the more interesting activity to total and report. Essentially, the problem is that the respondent or the analyst must designate one activity as the primary activity and the other activity as the secondary activity. Inevitably, analysis will correctly or incorrectly focus on those activities designated as primary activities when the secondary activity may also be of interest.
A third approach is to create compound activities—that is, to define a joint activity as a distinct activity. For example, an hour spent visiting with the family while eating dinner would be a separate activity from visiting and from eating. While this approach is conceptually appealing, it may lead to an enormous number of activities (although, the number of compound activities used in the classification scheme could be limited to a few of particular relevance for a survey). It would also complicate tabulations to the extent that
the compound activity combined two types of higher level classifications (such as work and leisure).
A fourth approach may apply to certain activities that are recorded by a respondent as simultaneous, but are in fact distinct and sequential. These distinct activities may be aggregated over in the survey and can be called “multiplexed” activities. (This term refers to a process of transmitting several messages or signals simultaneously on the same channel.) That is, one is either talking or listening but not doing both at once. Or to continue a previous example, one is either folding laundry or watching television, but not actually doing both at the same time. If one could just slice up time finely enough (which is not an easy task), the multiple activities would become short bursts of solo activities. A survey could ask respondents to report what they were doing in 5-minute or even 1-minute intervals to distinguish two separate activities that may be reported as simultaneous activities if 15 minute intervals were used instead. So, for example, 15 minutes of folding laundry and watching television might be divided into 10 minutes of folding laundry and 5 minutes of watching television. This approach may work for some studies that require such detail. However, this method puts a huge burden on respondents to recall very fine intervals of time use. Most time diaries allow respondents to designate their own intervals for this reason. Furthermore, there are some activities for which this approach would not work because they are intrinsically simultaneous (such as reading while traveling on a train) and cannot be separated at however fine a slice of time. Riding the train does not exclude one from also reading, conversing, or sleeping. Both activities can truly be conducted simultaneously.
Another approach, proposed by Horrigan et al. (1999), allocates joint production on the basis of the proportion of the time that a group spends on the solo activities. This method computes the total amount of time a demographically defined group of people spend their time, on average, and assigns an hour of time spent by an individual jointly doing the activity on the basis of the proportion of the population totals. For example, if teenage girls spend 10 hours a week only on the phone and 20 hours a week only watching television, then 9 hours jointly spent talking on the phone while watching television would be allocated as 3 hours on the phone and 6 hours watching television. While this might be appropriate if one is aggregating time over multiplexed activities, participants pointed out that there is no justification for this division if the activities are truly simultaneous. It also is not appropriate if time spent jointly in two activities is different from the solo time spent in the two activities. In other words, the output of a half hour spent talking on the telephone while watching television may not be equivalent to the output of a half hour spent only talking on the phone, for example, if the quality of the conversation is lessened while multitasking. This approach would treat the two half hours the same.
A final approach, called the value theoretic measure of time use and described by William Nordhaus, divides time spent in joint activities by the value of the outputs produced by the time. For example, if a person is jointly cooking and babysitting and the value of babysitting is $5 per hour and the value of cooking is $15 per hour, then an hour of simultaneously babysitting and cooking would be allocated as 15 minutes of babysitting and 45 minutes of cooking. This approach would use either the value of the products for extrinsically valuable activities or the subjective value of the time for intrinsically valuable activities. This approach might be a useful framework for allocating simultaneous time use, particularly when integrated with the national income and product accounts. However, it poses serious practical obstacles because of the need to measure the outputs or values of alternative time uses.
Discussion at the workshop made it clear that finding a practical and theoretically satisfactory approach to measuring simultaneous activities is a tough issue. Existing surveys indicate that attention to simultaneous activities is important. Some activities, such as child care, often show up as secondary activities and would be lost if only primary activities are recorded and reported. Further research is needed on this issue. Whatever approach is taken, it is clearly important to allow for multiple activities in time-use surveys.