1 Introduction

Much of the focus of research and policy in the United States is on alleviating monetary poverty. However, individuals can also be time poor. That is, they do not have enough discretionary time to engage in leisure, educational, and other activities that improve their well-being. Indeed, if individuals do not have a certain amount of leisure time to participate in customary leisure activities, they may suffer from social exclusion (Bittman 2002). While a few studies on time poverty and time pressure can be found in the time use literature, no recent measures of time poverty in the US exist. Yet, such measures would help us to answer many questions. They would help us to identify those individuals in the population who are vulnerable to time poverty. They would allow us to determine how time poverty affects individuals’ behaviors and health status. They would also tell us whether time poor individuals have adequate time to pursue educational opportunities that would increase their future earnings.

In this paper we define discretionary time using the concepts of necessary and committed time. Discretionary time is the residual number of minutes that an individual has remaining after he or she performs the basic activities of personal care and paid and unpaid work. Using this definition and data from the 2003–2006 American Time Use Survey (ATUS), we develop time poverty thresholds. With these thresholds, we are then able to determine whether an individual is time poor and to compute time poverty rates for various subgroups in the population. We then perform multivariate analyses to identify the correlates of discretionary time and time poverty. Finally, we offer an exploratory regression whereby we examine the association between an individual’s time poverty and health status.

2 Income Poverty

Much of the focus of research and policy in the United States is on alleviating monetary poverty and the literature has been extensively reviewed. Atkinson (1989), Ravallion (1994), and Danziger and Haveman (2001) are all good sources of summaries and discussion of income poverty research. An income poverty standard can be defined and measured using one of three approaches (Ruggles 1990). The first approach focuses on absolute poverty, that is, having less than some objective amount of consumption or income. The second approach focuses on relative poverty, that is, having less than others, where the cutoff may be something like a proportion of the median of the population. The third approach considers subjective poverty, which is the feeling of not having enough, and is based on survey responses regarding the amount of income that is “just sufficient.” The United States takes the first approach in defining its income poverty thresholds, basing its thresholds on the amount of income required to purchase and make meals based in part on the US thrifty food plan (Fisher 1997).

Vickery (1977) was the first to consider that income poverty may be related to available time. Taking the absolute measure approach, she calculated time-adjusted income poverty thresholds for the US based on her estimate of the minimum amount of time that would be needed to complete household tasks, given that an individual had monetary income equal to the US poverty threshold (and so had limited ability to buy household services, etc.). Her definition of this minimum amount of time was based on the average amount of time non-employed homemakers spent on household work according to US time budget data and on other ad hoc assumptions regarding how much time was necessary for certain activities. Vickery calculated different thresholds based on the number of adults and the number of children in the household to account for differences in household time resources and needs. Taking similar approaches, Douthitt (2000) updated Vickery’s time-adjusted income-poverty rates for the United States using the 1985 Time Use Survey and Harvey and Mukhopadhyay (2007) calculated similar rates using 1998 Canadian data.

3 Time Poverty

While adjusting income poverty thresholds to account for time deprivation may be useful, dedicated time poverty thresholds are important in their own right. How should such time poverty thresholds be defined? As with income poverty, time poverty may either be subjectively or objectively measured. A subjective measure would be based on individuals’ perceptions as to whether or not they feel time pressured. An objective measure would be a threshold of the time needed for rest or leisure or other activities after performing personal care, paid work in the marketplace, and unpaid domestic labor. Two studies that have taken the subjective approach and used measures of “perceived” time pressure are Mothersbaugh et al. (1993) and Hamermesh and Lee (2007). Looking at the effects of time stress, Mothersbaugh et al. (1993) showed that perceived time pressure has a negative effect on adherence to recommended dietary practices (RDPs). Hamermesh and Lee (2007) found that higher-income-couple households experience higher time stress than other households and that women feel greater time stress than men. The objective approach, and the approach taken in this paper, compares an individual’s actual usage of time to some absolute or relative standard.

There is some precedent for both absolute and relative standards. Douthitt (2000) took the absolute standard approach. She calculated time poverty rates by comparing individuals’ actual available time (after deducting market work hours and estimated home production work hours) to the discretionary time that would be available given an absolute amount of time needed to perform personal care and household maintenance. However, this required making ad hoc assumptions about the minimum amounts of time required for particular activities such as personal care and household maintenance. For example, an absolute measure of time poverty requires assumptions as to how much time must be spent talking to or listening to a child although the amount of time that must be spent listening and talking to a child likely depends on the particular child and parent. Similarly, an assumption must be made as to the minimum time required to clean a house, yet the amount of time required for this activity depends on the size of the house and the preferences of the people living in it. While they do not compute thresholds, Goodin et al. (2005) present the concept of discretionary time as being the amount of time potentially available to individuals if they spent only the amount of time strictly needed to maintain a subsistence standard of living.

Bardasi and Wodon (2006), who studied time poverty in Guinea using data for 2002–2003, used a relative standard. They defined time poverty as occurring when individuals do not have enough time for rest and leisure after taking into account the time spent working (whether paid or unpaid). In one set of thresholds they defined time spent in work as time spent in the labor market, in domestic chores, and in collecting water and food. In another set of thresholds they used an expanded definition that included the amount of time spent helping other households and participating in community services. They defined their thresholds relative to the distributions of actual time spent in these necessary activities. In particular, they used two alternative relative time poverty lines, a lower threshold of one and a half times the median of the distribution of time spent working (both paid and unpaid) and a higher threshold of two times the median. Then they calculated time poverty rates for the population as a whole and for various groups of individuals. Another study, Bittman (2002), defined a leisure time poverty threshold, 50% of the population median of leisure time. Most recently, Burchardt (2008) calculated time poverty thresholds for the United Kingdom based on the relative distribution of time use. Like the other recent studies, she included personal care, unpaid work, and paid work in her definition of committed time, calling the residual “free” time. In our paper, we follow these recent studies and take the relative approach, providing alternative thresholds based on different points of the discretionary time distribution.

4 Buying Time and Time Flexibility

Higher income individuals have the option of “buying” time and thus should be less likely to be time poor. They can buy prepared foods and pay for child care, housekeeping, and home maintenance services. By purchasing these goods and services, these individuals can reduce the number of hours they need to spend in committed activities and so increase their time in other activities. However, some activities cannot be bought—an individual cannot pay someone to eat or sleep for him. Higher income individuals also tend to have better-timed and more flexible schedules (Hamermesh 2002). Consequently, higher income households may be less likely to be time poor. Thus we explore the relationship between income and time poverty in this paper.

5 Data

The data we use to compute the time poverty thresholds are time-diary and questionnaire data from the 2003–2006 American Time Use Surveys (ATUS). The ATUS is a continuous survey that the Bureau of Labor Statistics (BLS) began in 2003 and that is conducted by the US Bureau of the Census.Footnote 1 Respondents to the ATUS are drawn from households that have completed their last month of participation in the Current Population Survey (CPS). One person aged 15 or over within each outgoing CPS household is randomly selected to participate, for a total of 60,674 respondents over the period 2003–2006.Footnote 2 As our concern is with the time use of responsible adults in a household, our analysis excludes persons under age 18 who live with at least one parent and who do not have own children living in the household, leaving us with a sample of 57,816 respondents.

The time-diary component to the ATUS describes what individuals were doing over the course of a 24-h period. Individuals’ own descriptions are coded into standardized activities. The survey questionnaire updates the household roster and labor force participation information from the CPS. Consequently, the ATUS data files include not only the respondent’s time diary but also demographic and labor force information for the respondent and the respondent’s household’s members.

We pool the 2003–2006 ATUS data files. Because time use patterns do not change much from year to year,Footnote 3 pooling the 4 years of data allows analysis by various household characteristics. We utilize the ATUS Respondent files, the Roster files, the Activity Summary files, the Replicate Weights files, and the ATUS-CPS files. We use the final weights for each year that are based on the 2006 weighting methodology.Footnote 4 We use SAS 9.1 and SAS 9.2 for generating estimates.

6 Methodology

The resource that we are comparing across households is discretionary time. Having discretionary time improves individuals’ well-being by allowing them to take part in leisure activities as well as exercise, educational activities, and other activities that provide utility directly or improve individuals’ human capital. We define an individual’s daily discretionary minutes by subtracting minutes of necessary and committed time from 1440 (total daily minutes). Necessary activities are those activities that must be performed by an individual for him or herself, i.e., personal care. This includes sleep, grooming, health-related self-care, and other personal and/or private activities. Committed activities are those activities that must be performed given previous life choices (e.g., whether to marry, to divorce, to have and raise children, and to be employed). Such activities include time spent in household work, time spent in child care, time spent caring for household adults, and time spent in market work or related activities (see Table 1). Although the individual has the ability to make these choices over time, in the short-run these commitments are essentially fixed. Time spent in travel related to these necessary and committed activities is also included. This conservative list of necessary and committed activities is similar to Burchardt’s (2008).Footnote 5

Table 1 Necessary and committed activities by major activity group
Full size table

Using this measure of discretionary time we designate time poverty thresholds. We present thresholds at 50, 60, and 70% of the median of the population discretionary time distribution similar to Bittman (2002), Blackden and Wodon (2006), and Burchardt (2008).Footnote 6 Because income poverty thresholds have historically been defined and presented for different household composition groups, and because household composition is expected to affect an individual’s discretionary time, we also present thresholds for various household types.

7 Results

7.1 Time Poverty Thresholds and Rates

Table 2 Footnote 7 shows alternative time poverty thresholds for the total population (row 1) and for subgroups defined by household composition (remaining rows) as well as each group’s representation in the population. The alternative thresholds are defined according to 50, 60, or 70% of the median. As expected, individuals in households with children have considerably less discretionary time than those without children, and discretionary time decreases with the number of children in the household. However, the relationship between discretionary time and the number of adults is not clear. Perhaps this is because the presence of an additional adult, while potentially making another 24 h available to the household, does not necessarily mean that the additional adult offers any of their time to alleviate the responsibilities of the respondent. In addition, we did not distinguish between spouses and other household adults. As a result, the second adult in the household could be the respondent’s adult son or daughter or elderly parent, which may help explain the lack of a clear result between discretionary time and the number of adults.

Table 2 Time poverty thresholds for the total population and for household composition groups
Full size table

For one-adult households, the difference in median discretionary time between households without children and households with one child is 175 min (583 min − 408 min), almost 3 h, and the difference between households with no children and those with 2 or more children is 193 min (583 min – 390 min), 3.2 h. This means that the single person with one or more children has, on average, about 3 h of necessary and committed tasks on a given day more than the single person without children. The difference between median discretionary time for the single parent with one child and the two-adult household with one child is only 10 min (418 min – 408 min). Thus, a second adult in a household affords, at the median, only 10 additional minutes of discretionary time. For households with two or more children, individuals in the two-adult households actually have less discretionary time (a median of 382 min) than one-adult households have (390 min). Perhaps this is due to differences in preferences across the groups about how much child care should be performed, with two-parent households deciding to spend more time. For individuals in households with three or more adult households, the third adult in a household does appear to increase the discretionary time of respondents.

Using the 60% threshold defined for the total population, 289.8 min (4.8 h), Table 3 shows cell percentages by group for those individuals whose discretionary time falls below this threshold (those designated time poor) and those whose discretionary time is at or above this threshold (those designated not time poor). These are columns 2 and 3 in the table. We focus on the 60% of median results because this is the percentage of the median used by Burchardt (2008). Time poverty rates using 50 and 70% of the median are presented in Appendix 2. Column 4 presents the time poverty rates for the various groups. For example, 4.94% of all non-employed individuals in single-person households with income below the poverty threshold are also in a state of time poverty. Note that several cell percentages and rates are suppressed due to small cell sizes. We used the BLS standard of 60 observations to define a minimum cell size for an estimate. Panel A of the table shows results for those not employed and Panel B shows results for those employed. Time poverty rates are much higher for those who are employed than those who are not employed, as expected. Time poverty rates are also higher for individuals living in households with more children, as expected. Although our prior was that those who are income poor are more likely to be time poor also, due to their inability to “buy” time by outsourcing activities like child care and housecleaning, the data do not show a clear relation between income poverty and time poverty.

Table 3 Time poverty by employment status, household composition, and household income
Full size table

In order to obtain a clearer picture, Table 4 provides an aggregate look at time poverty and its relation to income poverty in the United States. Individuals are identified as income poor if their income is less than the income poverty threshold. Although the actual percentage of the population that is both income poor and time poor is quite small (2.18%), 18.60% of individuals who are income poor are also time poor. However, individuals who are not income poor are even more likely to be time poor at 21.06%. The time poverty rate for those not reporting income is similar to that for individuals who are income poor at 18.16%, although the two rates are statistically different at the 90% confidence level. Note that these summary rates combine the two groups with income above the poverty threshold but below 185% of the poverty threshold with those with income above 185% of the poverty threshold.

Table 4 Time poverty by income poverty: cell percentages and rates
Full size table

Time poverty rates by household composition are provided in Table 5. Here we see that households without children have the lowest time poverty rates and, in some cases, rates that are half that of other households with the same number of adults. Interestingly, there is not a great deal of difference between the time poverty rates for households with one child or with two or more children, for a given number of adults. Of particular interest is the fact that there is little difference in time poverty rates between a person in a household with one child and one adult (28.69%), and a person in a household with one child and two adults (27.15%), and between a household with two or more children and one adult (30.57%) and with two or more children and two adults (31.85%). This indicates that an additional household adult on average does not add much to the time resources of the household. The fact that we did not distinguish between a spouse and another adult who could be living in the household, say, an adult child, may be influencing these results. However, this could also be due to differences in preferences regarding the amount of child care and housework that are “required” to be performed across one- and two- adult households. It could also be due to a consumption aspect of child care where two-adult households consume more than one-adult households (e.g., family time).

Table 5 Time poverty by household type: cell percentages and rates
Full size table

7.2 Multivariate Regression Analysis: Correlates of Discretionary Time and Time Poverty

Although the results in Tables 2, 3, 4, and 5 suggest that household composition and employment status are important correlates of discretionary time and time poverty, multivariate analyses better ascertain their importance. Table 6 shows Ordinary Least Squares (OLS) estimates from a discretionary time regression and maximum likelihood estimates from a time poverty probit. Similar to the descriptive results presented above, the estimated associations between the presence of an additional household adult and discretionary time and time poverty are quite minimal. The presence of an additional adult in a household is associated with only 6 min more discretionary time and a lower probability of being time poor of less than 0.01. An additional child, on the other hand, appears to be more important, with an additional child being associated with 35 fewer discretionary minutes and a greater probability of being time poor by 0.04. Being employed has the largest association, with an employed person having 188 min (over 3 h) less discretionary time and a 0.03 higher probability of being time poor than a non-employed person. Finally, consistent with the descriptive evidence, these multivariate regression results show that income is not generally associated with a person’s discretionary time or the likelihood of being time poor when other factors are controlled.

Table 6 Multivariate analyses: discretionary time continuous regression and time poverty probit
Full size table

7.3 Illustrative Example: Association of Time Poverty with Health Status

We present our time poverty thresholds with the expectation that they will be used to enlighten policy. We therefore provide a simple illustration as to how they might be used. In 2006 the Eating and Health (EH) Module was added to the ATUS. Among other things, this module included an individual’s self-reported health status. While not a true outcome variable because it is measured at the same time as a person’s time poverty status, we present an exploratory regression to ascertain whether time poverty is associated with an individual’s health status. Table 7 shows the results of a simple probit where the dependent variable is an indicator variable for whether or not a person reports being in good health and the explanatory variables are an indicator for time poverty and the usual other controls.Footnote 8 Gender, age, education, and income all have the expected associations. However, while we might expect the association between time poverty and good health to be negative, as a person who does not have enough time may not exercise or eat right, the data show that, controlling for several personal and household characteristics, it is positive. A time poor individual has a 0.03 higher probability of being in good health than someone who is not time poor. This result is likely due to the fact that employed individuals have, on average, better health status than those who are not employed, and are also more likely to be time poor. While this result is not definitive and a more thorough analysis that accounts for endogeneity and unobserved factors is beyond the scope of this paper, it highlights the potential relevance of time poverty to individual well-being. Additional research on the relation between time poverty and general health as well as other measures of well-being such as body mass index (BMI) would provide health policy- and program-relevant insight. Research on the relation between time poverty and food assistance program participation would help food assistance and nutrition programs identify and target individuals who are eligible but too time poor to participate.

Table 7 Multivariate analysis: good health probit
Full size table

8 Sensitivity Analyses

To determine the robustness of our results, we defined alternative thresholds for various subgroups in the population because even within a subgroup an individual will have different necessary and committed time requirements and different resources at his/her disposal. In Table 2 we presented a set of alternative thresholds defined by household composition group. Time poverty rates for different subgroups in the population that are based on these alternative thresholds are presented in Appendices 3 and 4. Although total time availability and commitments should be relatively similar for individuals in a particular type of household, financial resources and thus the ability to buy time will differ across individuals in this group. Thus, one can garner some information from an individual’s position relative to these thresholds about his or her inability to buy time. For example, the time poverty rate for households with an employed respondent, second adult, and two or more children is 39% for those below the income poverty threshold (Appendix 3) and 36.5% for those with higher incomes (above 185% poverty threshold), suggesting that the ability to buy time reduces the rate for these households by 2.5 percentage points. The associated time poverty rates can be used to identify relative deprivation within a population subgroup.

Our second alternative set of thresholds is based on medians for groups defined by household composition and age of youngest child (see Appendix 5 for associated time poverty rates). The thought is that some modification must be made for child age because younger children require more care. Our third set of alternative thresholds is based on groups defined by household resources, that is, the number of adults in the household (potential time resource) and the household income level (money resource) (see Appendix 5). The fourth alternative set of thresholds we construct is based on medians for non-employed persons by household composition and income level so as to approximate Vickery’s (1977) idea of a time poverty threshold (see Appendix 5). Our final alternative set of thresholds is based on medians calculated by group defined by household composition, income, and employment status (employed or not). Associated time poverty rates are also found in Appendix 5.

Interestingly, for the population thresholds presented in this paper and for four alternative sets described above, the summary time poverty rates are very similar. For example, assuming 60% of median thresholds, the overall time poverty rates range from 20.12 to 20.35%. This indicates that small changes in the way thresholds are defined produce qualitatively similar results at the total population level.

The set of thresholds which applies the medians for non-employed individuals to all individuals produces considerably higher time poverty rates than the others. According to the threshold based on 60% of the median, the overall time poverty rate is 33.45%. This is as expected because those individuals who are employed have the added time requirement of market work that those who are not employed do not have.

9 Discussion

The time poverty thresholds and rates that we present in this paper confirm the idea that individuals in households with children have more demands on their time than those in households without children. However, this paper goes further and demonstrates and measures the extent to which this is true. Controlling for other factors, an additional household child reduces a person’s daily discretionary time by 35 min. This paper also shows the surprising result that, while one might expect the necessary and committed activities required of an individual to be lessened by the presence of an additional adult in a household, the reduction, controlling for other factors, is minimal. Employment is also a big driver of time poverty, with employed persons having over 3 h less discretionary time per day and a 0.91 higher probability of being time poor, controlling for other factors. Surprisingly, income poverty is not clearly related to time poverty.

Traditionally, policymakers have used income poverty measures to identify individuals and households lacking sufficient material resources. However, time is also a scarce resource that affects individuals’ well-being. For example, a lack of time to exercise or eat right may lead to obesity and its associated health problems. Recognition of the importance of time poverty could further improve analysis and policy-making in a wide variety of public assistance and public welfare contexts. If used jointly with existing monetary poverty thresholds, we expect that our time poverty thresholds and rates will offer policy researchers a clearer picture of individuals’ well-being.