Juliet B. Schor

Harvard University and Tilburg University

September 1997


Paper prepared for presentation at Conference on Civic Engagement in American Democracy, September 26-28, 1997, Portland Maine.



In his Ethiel de Sola Pool lecture, Robert Putnam asked what accounts for the "strange disappearance of civic America"? Using the format of a mystery, Putnam considered and rejected nearly all the common explanatory suspects, including some which he had previously suggested himself as likely or plausible candidates. Suburbanization and mobility, racial factors, the welfare state, life cycle and period effects were all dismissed. Higher levels of family instability was accorded t he status of accessory. The culprit, according to Professor Putnam, is television, which has exerted a strong anti-civic influence on generations of Americans who grew up with it. By contrast, Putnam's "long civic generation," those Americans born in the early part of the century, had no exposure to the medium in their formative years, and are markedly more civically involved as a result.


This paper re-examines two explanations which Putnam rejects, namely "the pressures of time and money," and the "changing role of women." With respect to the first, he questions the view that pressures of time have actually increased, and notes that declines among the affluent are more pronounced than among those of lower incomes. With respect to the second, his official view is "not proven," on account of evidence that the declines in civic engagement are greater among housewives than among working women, and because working women join more civic associations than their counterparts without jobs.


As I will argue below, I believe that Professor Putnam's rejections of these hypotheses to be premature. While I do not argue that they are the only or even necessarily major cause of changes in civic engagement, my argument is that we do not yet have the kind of decisive evidence we would need to dismiss them. Given the strong commonsensical and popular notion that declining availability of time has been at least partly responsible for the shrinking pool of volunteer labor, I believe m ore research is needed before abandoning these explanations. In the pages which follow I address the question of what is happening to trends in working hours and free time and how those trends might be affecting civic engagement, paying particular attention to traditional high involvement groups. I end with a discussion of pressures, not of time, but money, which I believe may be operating in ways Professor Putnam has not yet considered.




In the late 1980s and early 1990s, the topic of working hours was thrust into the public spotlight. Having been ignored for years as a "solved problem," the consensus that working hours were unchanged came apart. On one side, two Harvard economists presented estimates from the Current Population Survey which showed that annual hours had risen appreciably. (Leete and Schor 1992, 1994; Schor 1992) "Working harder for less" became a popular mantra in the public discourse. On the other, researchers from the Universities of Michigan and Maryland challenged this view. Relying on time diary data, they claimed that Americans were enjoying more leisure time than ever before. (Robinson 1989a,b, 1994; Juster and Stafford 1991) Subsequent studies ensued, but perhaps because this debate was carried out in the media, rather than in scholarly forums, considerable disagreement about trends in worktime persists.


A. Types of Data on Hours of Work

There are three major sources of data on hours of work. The first is data provided by business establishments to the government. This data has rarely figured in the public debates because it measures hours per job rather than per person . While establishment data are extremely important for some issues (i.e., measurements of productivity per hour), they are less well-suited to trends in individual or household time use, particularly because the importance of multiple jobholding and the underground economy have been growing. The second type of data is household surveys such as the Current Population Survey (CPS), the Census, and the Panel Study of Income Dynamics (PSID). These are large government data sets which provide representative samples of the population. They use recall questions about weekly hours and weeks of work per year such as "how many hours did you work last week?" The third type of data is time-diaries [1]. In these surveys respondents are asked to record, at fifteen minute intervals, how they are using their time over a 24-hour period. The US diary studies vary in methodology (some are recorded at the time of activity, others use recall methods). The diary surveys have been carried out by University survey research centers a t various points in time. They typically use small samples, because of the cost and difficulty of such a detailed and rich approach. Here I will be discussing only the evidence from the household surveys and the diary studies.


B. Findings from the Household Surveys

A number of estimates from the household surveys, such as the CPS and the PSID show clear evidence of rising hours. My own estimates with Laura Leete (Leete and Schor 1992, 1994; also reported in Schor 1992 for a shorter period) show that annual hours of market work as measured by the Current Population Survey have risen substantially for the labor force since 1969, and particularly since 1979. The increase has been particularly marked for what we call the "unconstrained labor force," t hat is employed persons who are not involuntarily underemployed. (See Tables 1 and 2, from Leete and Schor, 1994) Our estimates are derived by combining respondents' estimates of their actual hours last week and the number of weeks they worked in the previous year. We found similar trends using the "usual weekly hours" question which has been available since 1979. (For a defense of our procedures for estimating household hours against Robinson and Godbey's criticisms, see the accompanying note.[2])


A longer series from the CPS, which omits some of the detailed corrections of Leete and Schor's estimates, is provided by Mishel et al (1996). They find that annual hours have increased by 103 hours since 1967. (The increase since 1973 is 142 hours). Notably, they find that hours have continued to rise since 1989, increasing by 26 between 1989 and 1994. (Table 3, from Mishel 1996)


A recent study by Bluestone and Rose (1997) using the Panel Study of Income Dynamics (PSID), reveals similar trends. In response to Robinson's criticisms that the CPS weekly hours series is plagued by overestimation bias (see below), Bluestone and Rose have gone to the PSID. They argue that the PSID is a more accurate measure because respondents are asked to provide more detail about vacation days, leaves, primary and second jobs, and so on. Looking at prime age workers (25-54), Bluestone and Rose find considerable business cycle fluctuation. After controlling for the cyclical effects, they find annual hours have increased on average 3.3 hours per year, for a total rise of 66 hours between 1969 and 1989. (This is similar to Leete and Schor's finding of a 2.75 yearly increase for those years for all labor force participants. The .65 hour difference may be mainly attributable to the exclusion of the 55+ age group from Bluestone and Rose's sample.[3])


Dual-earners families have been a special focus of debate in the time-squeeze discussion. Bluestone and Rose find particularly large increases in hours for families where both husbands and wives are employed--husbands' plus wives' annual hours increased from 2,850 to 3,450. Because there is frequent misinterpretation of this point (including by Bluestone and Rose in their discussion of my estimates), it is worth noting that this rise is not due to rising participation among married women. These estimates show that the average dual-earner family in 1967 worked 600 fewer hours than it did in 1989. Furthermore, once business cycle effects are controlled for, the total increase is even larger, at 640.


One set of estimates from the household surveys provides a more ambiguous picture. Coleman and Pencavel (1993a,b), using Census Data from 1940-1980, and the CPS for 1988, claim to find no aggregate trends in hours. However, because Cole man and Pencavel do not provide aggregate estimates, but only estimates broken down by race and gender, direct comparisons with the foregoing studies are not possible. Overall, they find that hours have fallen for persons of low educational attainment and have risen for those of high educational attainment. I discuss their results in more detail below.


B. Findings from the Time Diaries

Results from the household surveys are disputed by researchers who rely on time diaries. Juster and Stafford of the University of Michigan have claimed (1991) that Americans' free time has risen, on the basis of data covering the period 1965-1981. John Robinson, from the University of Maryland, using the Michigan data for 1965 and 1975, and his own diary survey for 1985, also took issue with the claim that hours of work were increasing. Robinson's findings have been published in a monograph for the Office of Technology Assessment (1988), articles in American Demographics (1989a,b), in a Monthly Labor Review article (Robinson and Bostrom 1994), and have now appeared in his jointly co-authored book with Geoff Godbey, Time for Life: The Surprising Ways Americans Use Their Time. Because Professor Putnam has written a foreword for the book, it is of particular relevance to this gathering.


In Time for Life, Robinson and Godbey present diary estimates of trends in paid work. (Table 4, Robinson and Godbey 1997) They find that hours of work have fallen for both men and women between 1965 and 1985. Women's weekly hours declined from 36.8 to 30.8 while men's declined from 46.5 to 39.7. The path of decline is particularly important, however, in light of the sample issues which I address below. In the aggregate, Robinson finds that virtually all the decline in hours comes between 1965 and 1975. For women, there is a 2.5 weekly hour rise after 1975, and for men a 4.2 hour decline. (In the accompanying note, I discuss some of the differences among Robinson's published estimates. For example, an earlier 2.7 hour increase in weekly worktime between 1975 and 1985 now appears as a decline of .7 hours.[4])


Who's right? Has free time increased, as the diaries show, or are hours of work rising, the picture we get from the household surveys? As is usual in debates of this kind, some of the differences are due to the fact that we are not dealing with uniform time periods, uniform measures of work, labor force coverage, and the like. Leete and Schor's estimates include the 18+ age group, Robinson's cover 25-65, and Bluestone and Rose cover 25-54. These differences account for some of the differences in findings. But such large discrepancies are clearly being driven by other factors. Robinson's critique centers on the quality of answers to the weekly hours question in the CPS. The economists (Schor, Leete, Mishel, Rose and Bluestone) emphasize the importance of measuring annual hours and accounting for business cycle and labor market trends. I turn to these issues now, in more detail.


A. A Statistical Artifact?

One of the major sources of disagreement between Robinson and me continues to be the usefulness of the 1965 survey which forms his benchmark year. The importance of the 1965 data is clear from Robinson's 1989 discussion. "Most gains in free time have occurred between 1965 and 1975. Since then, the amount of free time people have has remained fairly stable, with some loss for certain groups." (1989a, pp. 34-5).


The 1965 survey, like the 1975 and its 1981 follow-up, was carried out at the University of Michigan. It was part of a pioneering international comparative project in time-use organized by Alexander Szalai. In order to conform to the requirements of that project, certain sample restrictions were put in place. Thus, the 1965 sample includes only employed heads of households and their spouses, it only includes respondents from SMSAs, and it excludes all farmers. These restrictions in them selves are not a problem if either (1) future samples are similarly restricted or (2) the restrictions are not correlated with trends in time use[5].

Leete's and my analysis of the 1965 sample found that the sample restrictions, while not always significant in uncontrolled comparisons (as for example, the rural/urban comparison that Robinson reports), are correlated with hours of work. Compared to the true population (as measured by the CPS), the 1965 sample is higher-income, it is more white, and it is more employed. These characteristics are all correlated with hours of work. (Income and hours are positively correlated, being white is positively correlated, and obviously, and being employed is naturally positively correlated.) Thus, the fact that Robinson finds a substantial decline in hours of work between 1965 and 1975 is no surprise: it is in part caused by comparing a sample of high hours people with a true representative sample. This is the basis for my view that part of what we are dealing here is a statistical artifact[6].


B. Labor Market and Macroeconomic Conditions

Perhaps the major source of difference between the estimates of the economists and Robinson is in their respective treatments of differences in labor market and macroeconomic conditions. It is well-known among economists that average h ours of work fluctuate with the business cycle, rising in expansions, as firms increase production more rapidly than employment, and falling in recessions as production shrinks. For this reason, trends in hours worked should be calibrated at consistent points in the business cycle, or should be corrected by the inclusion of a cyclical variable in a trend regression. Such corrections are now standard procedure among economists who study trends in hours[7]. By contrast, the time diary studies have never taken these factors into account, despite the fact that cyclical adjustments are possible[8]. In Robinson's work, this problem is not as severe as in other studies, because his endpoints are both expansion years. However, GDP growth in 1965 was almost twice as high as in 1985, so that some cyclical bias does exist. However, his middle year (1975) was a recession[9].


A second problem is more serious. Typically, time diary studies of working hours divide the population into the employed and nonemployed. This is a reasonable procedure in a world where there is no unemployment or under-employment. However, if unemployment and under-employment are large and/or growing, the failure to differentiate among these states leads to a conflation of voluntary and involuntary leisure time. By extension, it also results in estimates of worktime which are downward biased in a substantive, although not literal sense. This is probably the most important source of difference between the estimates of Leete and Schor and Robinson. During the period under question, there was a pronounced upward trend in unemployment, and more importantly for the estimates of hours among employed persons, in under-employment, defined as involuntary part-time and part-year work. Involuntary part-time and part-year workers are those individuals who identified themselves to the CPS surveyer as having only part-week work but wanting full-time weekly work or as being employed only a fraction of the year but wanting to work all year, or both. Leete and Schor call these people "labor market constrained."


The time diary questionnaires do not contain questions about labor market constraints, so these individuals cannot be identified. However, their inclusion in the average estimates has a large impact on hours, for two reasons. First, their numbers have increased steadily over the period in question. Between 1969 and 1989, the fraction of the labor force which was under-employed rose from 7.16% to 13.9% [10] (Table 5, from Leete and Schor, 1994) Furthermore, their average hours of work have fallen precipitously, from 1,312 in 1969 to 1,066 in 1989, a drop of 246 hours. (Figure 2, from Leete and Schor 1992) Inclusion of these individuals imparts a downward bias to estimates of hours. Comparing the trends for the entire labor force and the unconstrained labor force, Leete and Schor found that the rise in hours was 85 hours greater among the latter (138 compared to 53). Our correction for the growth of the underemployed explains our larger estimates of hours rise than other CPS studies which correct only for business cycle fluctuations. The estimates of Coleman and Pencavel (1993a,b) using Census Data (and for 1988, the CPS) are revealing in this respect. They find that hours of work have fallen for individuals of low educational attainment, and risen for those with high attainment. The growth in underemployment has been considerably more pronounced among those of lower educational attainment level, and also among black males. (See accompanying note for a fuller account of these findings[11]. Among all white women, they find that hours fell dramatically between 1950 and 1960 (-215) and thereafter fell slightly between 1970 and 1988. However their estimates since 1970 are driven by compositional effects pertaining to marital status and child-bearing. In Table 17, they show that annual hours have increased between 1970 and 1988 for white women of all marital statuses and all household structures (with and without children[12]).



Which is the right category for measuring trends in time use? If average hours of work fall because involuntary underemployment is rising, the general conclusion that Americans have more free time, while strictly speaking true, is misleading. The growth of a bi-modal distribution in hours which has emerged since the late 1960s is perhaps the major development in time-use trends which cannot be illuminated by the time diaries. The labor force is increasingly comprised of an overworked segment of full-time workers and a growing group of underemployed whose annual hours are shrinking. Failure to account for this development, rather than the psychological processes they invoke, may well be the most important source of the disconnect between Robinson and Godbey's claim that average American has more free time than ever before, and the pervasive sense of overwork, stress, and a speeded-up pace of life that everyone agrees has taken hold.

C. "Time at Work" versus "Time actually Working"

Another source of difference among the findings concerns the measurement of work. In the household surveys, respondents are asked how many hours they worked last week. It is likely that they respond with an answer which covers "time at work," that is, the hours they are officially on the job[13]. However, as Robinson and Godbey point out, not all hours at work are spent actually working. Employees whose jobs permit it congregate around the water cooler, make personal phone calls, and go to the bathroom during company time[14]. In their estimates, Robinson and Godbey subtract some of this "on-the-job leisure." They also subtract time spent in on the job training.


Which measure is preferable? It depends on the questions one is interested in answering. For measurements of productivity, clearly the time working measure is the right one. For social issues, such as the trend in civic engagement, time on the job is preferable, because it measures time that is formally committed to the employer and not available for participation in civil society[15]. These differences account for some of the divergence in the estimates.


D. Annual versus Weekly Hours

As I have noted, time diaries give us estimates of weekly hours, whereas the CPS and PSID allow us to construct estimates of annual hours[16]. One reason it is important to look at annual hours is that two-thirds of the increase in working time in Leete and Schor's estimates occurred as a result of a growth in weeks of work per year. (Our total labor force measure of weekly hours finds little change over the period.) Among the unconstrained labor force, weeks worked per year rose by 1.4 for men and 6.1 for women. These substantial increases are unlikely to be adequately reflected in the smaller samples of the time diaries, which measure only 24 hour periods, and have not been carried out during the summer, when we suspect much of the increase in women's work has occurred.


F. Different Ending Points

Robinson's data ends in 1985, Coleman and Pencavel's in 1988, Leete and Schor's and Bluestone and Rose's in 1989, and Mishel's in 1994. These differences are important. The household surveys show that the rise in hours of work occurs mainly after the business cycle peak of 1979. The time diaries capture only six of the 15 years since 1979. Part of the time diaries' different story may simply be that they do not cover the period during which hours rose most substantially. (Mishel et al find that annual hours increased by 9.4 per year between 1983-1989). In the 12 years since this data was collected, Americans' lives have gotten significantly less leisured.


E. Overestimation in the Recall Data

The core of Robinson's critique of the CPS is that individuals overestimate their working hours when asked about them in the standard recall format, but give true estimates when they fill out a time diary. (There may be some overestimation in cases where diaries are filled out retrospectively, as roughly 40% of the 1985 sample were, but I set aside this issue.) This view is supported by evidence from at least one beeper study which found the diary method to be the most accurate measure of time use. (Juster and Stafford 1985) If the CPS does overestimate weekly hours, then it is true that Americans think they work more than they do. While plausible, it is important to remember that this conclusion is not necessarily relevant to trends in working hours, which are the focus of the debate. If the extent of overestimation bias is constant over time, overestimation will have no impact on measured trends.


Robinson and Bostrom (1994) argue that the extent of bias is not constant, but is growing over time. That is because higher hours individuals engage in more overestimation than those who work less. (There is relatively little overestimation bias in the 20-44 hour range, but high exaggeration in the 55+ category.) But if the size of the bias is growing, it is most likely due to precisely the trend Robinson contests--the rise in working hours. Because more people are working longer hours there is more overestimation. Overestimation may well inflate the magnitude of the increase, but does not account for the basic upward trend itself.


On a more general level, there is little question that Robinson's critique has merit. The recall answers are not nearly as precise as the time diaries. (Although it is also worth remembering that the PSID, which is probably less subject to overestimation bias, shows a rise in working hours, and especially in the 1980s.) But the fact that there is overestimation bias in the household surveys does not imply that one should jettison them. While it is relatively uncontroversial that time diaries are the most accurate measures of individual time use at a point in time, it is by no means clear that they are superior for assessing trends in working hours over time. This is a task they were not designed for, and for which the three available US diary studies have serious drawbacks. Unlike the diary studies, the household surveys offer population samples which are consistent over time and representative. Because trends in hours of work are to a significant extent driven by demographic and socioeconomic factors, such representativeness is crucial to any analysis of trends. The household surveys also include weeks of work per year, which the diary studies do not. Furthermore, the household surveys allow us to account for macroeconomic and labor market variables which have strong effects on hours, which the diary estimates do not. For these reasons, and others I detail below, I continue to be skeptical about the claim that Americans have more free time than ever before.





One commonly noted hypothesis about the decline in civic engagement, and the shrinking pool of volunteer labor is that people no longer have the time to participate because they are working more hours[17]. This explanation suggests that civic time and market time are substitutes for each other. This is likely to be true only under certain conditions, namely when individuals are what we may call "time constrained." (Of course, everyone is "time constrained," in the obvious sense that time it self is limited. By this term, I mean to signal a situation in which there are alternative high (or even higher) productivity uses of time at the margin than those which have been chosen. People without time constraints operate with large amounts of free, or uncommitted time, for which they do not have high productivity alternatives to their current patterns of usage.)


Of course, time availability, and time constraints, do not operate in a simple way. This is why Professor Putnam's evidence that hours of work and civic engagement are positively correlated is true, but not decisive for the argument that time availability matters. The likely explanation for the positive cross-sectional correlation is that we are dealing with selectivity bias, due to the unobserved characteristic of propensity to activity (or busy-ness). As Professor Putnam notes, some people are more active than others. These types work long hours and engage in high levels of civic activity. But this fact implies little about the impact of time availability on civic engagement[18]. The measure we are interested in is the impact of changes in hours of work on civic engagement, holding everything else constant. A simple cross-sectional correlation between hours and engagement does not reveal the sign of that partial derivative, because of the presence of this unobserved population characteristic. We need to know whether, for any given individual, working more reduces civic engagement, ceteris paribus.


Similar selectivity issues may also account for the uncorrected trends in civic engagement among housewives and working women. Professor Putnam notes in his discussion of the impact of women's changing role that the shift into market work may not be random with respect to the propensity to be civically engaged. If more highly engaged women have been more likely to seek market work (a plausible assumption), then the rates of civic engagement can fall for housewives (as they have), and rise for working women. The presence of such a selectivity effect does not negate any possible impact of time pressures on the engagement of employed women. (Thus his not proven verdict.)


A. Theories of Time Allocation

Of course, a negative impact of worktime on civic engagement is only likely to exist where individuals are time constrained, or operating near the region of time constraint. In these cases, the theory of time allocation (Becker 1965) predicts that the shift to more market work raises the level of productivity at the margin that individuals will seek to achieve in all non-market activities. To achieve that higher level, people will eliminate lower-valued activities in both household work and free time activity (including civic engagement). People will shift out of the things they dislike, which yield little in the way of "output," activities which can be done inexpensively by others, etc. Obvious examples of such shifts are the rise in hired help for child care, house cleaning, and other domestic activities, and the shift away from time-intensive leisure activities toward goods-intensive ones, as predicted thirty years ago by Gary Becker and Steffan Linder. Overall, when working hours increase, we expect to see both less time spent in household work and leisure activities, as well as a change in the composition of activities in both spheres.


In terms of civic engagement, a rise in working hours would predict a shift to less time-consuming groups (such as the "paper" organizations which require no time commitments at all), to groups which meet infrequently (like monthly book clubs or wine tasting), to groups which have high payoffs in financial or career terms (such as professional networking groups), and to groups which yield more personal satisfaction. (This may be why "affinity" groups have grown while neighborhood contacts have declined. The marginal "benefit" gained from interaction with a neighbor is likely to be less than the marginal benefit from a person with whom one shares some interest or outlook.[19]) We would expect people to be shifting out of activities which have less personal payoff (such as voting) and more uncertain or longer term benefits. We would also expect people to do less which requires structured time, forward time commitments, and high coordination costs. (I leave it to the experts on trends in civic engagement to verify whether in fact these shifts have occurred.)


In the classic theory of time allocation, there are no rigidities, and individuals are free to vary the time spent on activities in a costless and unconstrained manner. This view has been challenged by Schor (1992), who argued that hour s per job are not flexible, and remain largely the prerogative of the employer. If this view is correct, then rising hours of work may be especially important in causing declines in time spent in household activities and civic engagement. This is because individuals can no longer equalize the value of activities at the margin, but may be compelled to work more hours on the job than they would choose, other things equal[20]. Apparently this is increasingly the case. The fraction of people who say they would prefer to work less and earn less has increased steadily since 1985, from roughly 5%, to 15% in 1994, and 17% in 1995[21]. And in some formulations of the question, the fraction of time-constrained individuals is even higher. For example, a Gallup poll from Jul y 1994 found that one-third of all respondents said they would prefer that either they or their spouse reduce hours and income up to 20% in order to gain more time for family. (Health 1994, p.48) In 1991, John Robinson found nearly half the working population reporting they would prefer a four-day workweek with the loss of the fifth day's pay[22].


The view that market hours constitute a special category of time use, with a disproportionate influence on other spheres is supported by recent evidence on the time allocation process. Estimating one of the few simultaneous models of market, household and free time, Peters (1997) finds that hours of paid work exert a strong causal influence on time allocation in the other two spheres, but that the reverse is not true[23]. Thus, Peters finds that increases in market hours lead to reductions in time allocated to both household work and free time.


B. Trends in Worktime among Time Constrained Individuals

If worktime availability has reduced civic engagement, it is most likely that it has occurred among those groups who are both civically engaged and time constrained. Obvious candidates are the highly educated, those in higher socioeconomic categories, parents, and married persons. Indeed, as Professor Putnam notes in his forward to Time for Life, even the diary data show that the "gain" in free time has been virtually nil for middle-aged, college educated, married parents .


Robinson and Godbey provide two measures of time-constraint, namely the extent to which people "always feel rushed," and self-reported levels of stress[24]. They find that women, full-time workers, and parents, and persons of the highest educational attainment status (graduate school) are most likely to "always feel rushed." Furthermore, the feeling of always being rushed rises with hours of work, peaking at 52% for women who work 50 or more hours per week. Robinson and Godbey report substantial increases in the fraction of the population which always feels rushed, from 24% in 1965 to a peak of 38% in 1992[25]. They also find that stress levels (as measured by self-reports of feeling a lot or moderate stress in the past two weeks) show higher levels among women, prime-aged groups, those with higher incomes, and people who either work or are in school. Stress levels also rise strongly with educational attainment. Thus, the high stress segments of the population are those we have identified as "time constrained." (Robinson and Godbey 1997, ch 16) A Gallup-Health Magazine poll on stress found that jobs were the singlemost important identified source of stress. (Health 1994, p.46)


The household surveys show that the working hours of these time-pressured groups have risen substantially. For example, Leete and Schor found that hours of paid employment for mothers rose 346 hours per year. For fathers, hours of paid employment rose only by 14. However, their non-market hours rose by 128, to help offset the 181 hour decline in mothers' household time. (Leete and Schor 1992, Table 9, p.14)[26] If it is correct that increases in market hours have a disproportionate impact on the amount of "free time" (including civic time), then this pattern would help explain the larger declines in engagement among women. (Putnam finds that the relative decline for women is considerably greater than that for men--10-15% per decade for men versus 20-25% for women.)


A second category of time constrained individuals is those who work in demanding jobs. Because more demanding jobs tend to be jobs which require high levels of education, the trends in hours by educational level can serve as a rough proxy. Coleman and Pencavel (1993a,b) find that among those with college degrees, hours of work have risen across the board. White males' hours have risen since 1940 by 89 hours, black males' by 55; white females' by 177; and black females' by 391 hours. Again, the greater increases for women are consistent with the pattern of decline in engagement. (By contrast, hours of work for everyone with fewer than 12 years of schooling have fallen.) Bluestone and Rose find a similar pattern. (For a summary of these findings, see Table 6.) Thus, the available evidence from the household surveys for those groups who are typically time-constrained (and highly civically involved) suggests that the loss of available uncommitted time has been substantial. Whatever one thinks about the aggregate trends in working hours, nearly everyone agrees that these groups are among the most harried and time pressured in our society.


C. Pressures of Money

Putnam's whodunnit raises the possibility not only that time pressures have caused the decline in civic engagement, but that financial pressures have also taken a toll on social capital. Professor Putnam refers here to a decline in civic engagement among lower income groups, who have suffered large relative declines in their financial well-being. He dismisses this argument because the drop-off in civic engagement has been larger among the more affluent. But this finding suggests a different kind of financial pressure--not absolute income deprivation but a worsening relative position.


As I argue in my forthcoming book, When Spending Becomes You: Pitfalls of the New Consumerism, upper middle and middle class American families have experienced a worsening of their position since the beginning of the 1980s . This is not because their absolute levels of income have declined (that fate has mainly been reserved for the bottom 40% of the income distribution). It is because their relative position has fallen. But in a world where what really matters is ho w one is doing relative to some aspirational standard, the worsening of relative situation is an important and powerful event. Powerful enough to impart a sense of financial dissatisfaction, and even anxiety. It can lead households to take on increasing levels of debt (as they have), and put themselves into a constant state of financial insecurity. If "pressures of money" can impede civic engagement, then one would expect that the affluent as well as lower income groups have suffered in recent decades.


The argument that middle and upper middle income families have experienced a worsening relative position comes from two trends. The first is the worsening of the income distribution (which began in the mid-1970s and has continued to the present. See Mishel et al 1996, Table 1.6, p. 51) Here two trends are relevant. First, the top 5% of the distribution has experienced a large increase in their share of income. This resulted in a shift within the top 20%, such that the bottom 15% were now worse off in comparison to the people with whom they make upward comparisons (the top 5%). This has resulted in substantial numbers of households making more than $75,000 or even $100,000 a year saying they spend nearly all their incomes on the " basic necessities of life." (Merck Family Fund poll 1995)


Overall, however, the top 20% did gain, relative to the bottom 20%. Thus, four-fifths of the population found itself in a worse comparative position. This matters because the top 20% has come to occupy a special place in the formation o f consumption aspirations. I believe they are increasingly looked to by all lower-income groups. This is in large part an effect of television, which typically portrays upper income individuals and households. (See Schor 1998 for evidence on the impact of television on aspirations and spending patterns). But it is also a product of the fact that workplace contacts (which tend to be more economically diverse) have become more important. Through the 1980s and into the 1990s, American consumers found themselves frustrated by the fact that their requirements for attaining the good life were being upscaled far more than their ability to pay for it. By 1995, only a third of the population reported that it was not in debt (excluding mortgage debt), and about a third of those were either in heavy or moderate debt. (Merck Family Fund 1995).


Thus, one hypothesis (and it is just that), is that the pressures of time and money, exerted through a widening of the aspirational gap, and exacerbated by the impact of television and other popular media, have effected a change in the behavior of American households. They work longer hours, live with higher levels of financial insecurity, are less willing to pay for public consumption, and feel more deprived on a personal level. They have become more oriented to their own private consumption needs. And in the process, it is possible that they are less civically minded.




This paper has not addressed the debate over the actual trends in civic engagement, a topic which is beyond my area of expertise. It takes the decline in engagement as given, and explores one possible explanation for that trend--a growth in "pressures of time and money," as Professor Putnam terms it. In contrast to recent characterizations of the American population as enjoying more leisure time now than at any time in the past, I have argued that hours of work have risen, and especially for the groups whose measured trends in engagement are most sharply downward--women, those of higher educational attainment, and those in higher income brackets. Underlying the rise in time pressure for these time-constrained groups is a bifurcation of the labor force into a group of overworked and a growing segment of underemployed, who experience a surfeit of involuntary leisure.


Whatever the real trends in time use there are some developments about which all time researchers seem to agree. The U.S. has become a society characterized by time scarcity, "time famine" or "time squeeze." Since the 1960s, increasing numbers of Americans experience their lives as hectic, rushed, and pressured. This subjective sense, whether it is due to a real increase in hours of paid work, or simply a change in the way Americans perceive time, has not been challenged[27]. The possibility that this subjective reality is transforming patterns of civic engagement is worth further exploration.




Becker, Gary, 1965, "A Theory of the Allocation of Time," Economic Journal, 75:493-517.

 Bluestone, Barry and Stephen Rose, 1997, "Overworked and Underemployed: Unraveling an Economic Enigma," The American Prospect 31( March-April): 58-69.

 Bosworth, Barry and Gary Burtless, 1992, "Effects of Tax Reform on Labor Supply, Investment, and Saving," Journal of Economic Perspectives, 6(1):3-25.

 Coleman, Mary T. and John Pencavel, 1993a, "Changes in Hours of Work of Male Employees Since 1940," Industrial and Labor Relations Review, 46(2):262-83.

 ----, 1993b, "Trends in Market Work Behavior of Women Since 1940," Industrial and Labor Relations Review, 46(4):653-76.

 Fuchs, Victor, 1986, "His and Hers: Differences in Work and Income, 1959-1979," Journal of Labor Economics 4:S245-72.

 Health Magazine, Stress Poll, October 1994.

 Juster, F. Thomas and Frank P. Stafford, 1985, Time, Goods and Well-Being, (Ann Arbor, Mi.: Survey Research Center, Institute for Social Research, University of Michigan).

 ---- and ---, 1991, "The Allocation of Time: Empirical Findings, Behavioral Models, and Problems of Measurement," Journal of Economic Literature, 29(2):471-522.

 Hymowitz, Carol, 1991, "Trading Fat Paychecks for Free Time," Wall Street Journal, August 5, 1991, p. B1.

 Leete, Laura and Juliet B. Schor, 1992, "The Great American Time Squeeze," (Washington, D.C.: Economic Policy Institute, February.

 --- and ---, 1994, "Assessing the Time Squeeze Hypothesis: Estimates of Market and Non-market Hours in the United States, 1969-1989," Industrial Relations, 33(1):25-43.

 Linder, Peter, 199x, Voiding, forthcoming.

 Merck Family Fund, 1995 Yearning for Balance Poll, (Takoma Park, Md.).

 Mishel, Larry et al, 1996, The State of Working America 1996-97, (Washington: EPI).

 Office of Technology Assessment, 1988, "Technology and the American Economic Transition: Choices for the Future," (Washington: Congress of the United States).

 PEW Global Stewardship Initiative Survey, February 1994, (Washington, D.C), mimeo.

 Peters, Pascale, 1997, "Differences in time allocation of Dutch men and women in 1995," unpublished paper, Tilburg University, Department of Leisure Studies, Netherlands.

 Putnam, Robert D., 1995, "Bowling Alone: America's Declining Social Capital," Journal of Democracy, Winter.

 ---, 1996, "The Strange Disappearance of Civic America," The American Prospect, Winter, 34-48.

 Robinson, John P., 1989a, "Time for Work," American Demographics, (April):68.

 ----, 1989b, "Time's Up," American Demographics, 13 (July):33-35.

 ---- and Ann Bostrom, "The Overestimated Workweek? What Time Diary Measures Suggest," Monthly Labor Review, August: 11-23.

 ---- and Geoffrey Godbey, 1997, Time for Life: The Surprising Ways Americans Use Their Time (State College, Pa: Penn State Press)

 --- and ---, 1996, "The Great American Slowdown," American Demographics, 42-48, June.

 Schor, Juliet B., 1998, When Spending Becomes You: Pitfalls of the New Consumerism, (New York: Basic Books), forthcoming.

 ---, 1992, The Overworked American: The Unexpected Decline of Leisure, (New York: Basic Books).

 ---, 1994, "Short of Time: American Families and the Structure of Jobs," unpublished paper, Harvard University.

 Shank, Susan E., 1986, "Preferred Hours of Work and Corresponding Earnings," Monthly Labor Review, November: 40-44.

 Shaw, Kathryn, 1992, "Labor Supply, Human Capital Investment and the Changing Distribution of Family Income," unpublished paper, (Carnegie Mellon University).




 [Not Provided]


 [Not Provided]


 [Not Provided]


 [Not Provided]


 [Not Provided]

Table 6

Rising Hours Among Time-Constrained Groups

Market Hours

Leete and Schor 1969-89





Bluestone and Rose 1967-89

Dual-earner families (combined hours)


At least college degree


Some college


High school degree


High school degree


High school dropout


Coleman and Penceval 1940-88

College Graduates, white females


College Graduates, black females


College Graduates, white males


College Graduates, black males








1) In the time diary studies, respondents are asked to record, for every 15 minute interval, how they spent their time. Analysts then choose a particular definition of "work," or "leisure," in order to produce the kinds of measures that are used in debates on time allocation. Time budgets are always weekly, either because the diaries are filled out for an entire week, or because a synthetic week is created from daily diaries. In a classic diary study, respondents receive the diary before it is to be filled out. In Robinson's 1985 survey, some respondents were asked about yesterday and asked to recall in 15 minute intervals how they used their time.

2) Our basic methodology, following that of Victor Fuchs (1986), was the following. To avoid the shortcomings we were concerned about in the time diary data (see text for discussion), we used CPS data to construct annual hours of paid work. Beyond the quality of the data, this is no controversy about these estimates. Robinson and Godbey's comments (1997, p.50 concern the way we calculated household hours.)


For household hours we created predicted estimates from the 1975 Michigan Time Diary Survey and the CPS. (Time diaries provide the only existing estimates of household hours.) Using the 1975 survey, we constructed an econometric model to predict average amounts of weekly housework for various demographic and socioeconomic groups. Taking the coefficients from that model, we applied them to the CPS in order to construct trends in household work. We reasoned that because trends in housework are largely driven by demographic and socioeconomic trends, and because the CPS is the more accurate reflector of demographic and socioeconomic trends, this was the preferable method. On this point it may be worth noting that like Robinson and Godbey, we too find significant declines in women's hours of housework and rises in male hours, so that their suggestion to the contrary is somewhat puzzling.


Robinson and Godbey take exception to the fact that we used only the 1975 survey. (Actually, we used both the 1975 and the 1981 survey, as a reading of our appendix will show.) We did not use the 1965 sample because it was missing various demographic categories, and was not representative of the entire population. We did not use Robinson's 1985 sample because, although we asked him for it, he did not make it available to us. (It hardly seems fair of him to complain about that now.) We are well aware that these choices limited the information that we had, but in our judgment, this was the best choice in a bad situation.

3) There are differences in the trends once they are broken down by gender, however. Bluestone and Rose report (but do not present) that in their estimates male hours declined slightly, after controlling for the business cycle. This may be partly due to the large declines in hours (7.7 per year) registered for Black men, a finding which accords with that of Coleman and Pencavel (see below). (But note also that Leete and Schor find a 20 hour decline in market hours for men before underemployment and unemployment are taken into account.) By contrast, Bluestone and Rose find even larger increases in female hours than Leete and Schor, recording a rise of 18.8 hours per year, or 376 over the period.

4) I am not sure why these figures differ from Robinson's earlier estimates, as all calculations are based on the same time-budget surveys. In two papers for the OTA (a mimeo entitled preliminary in 1986, and the other published in 1988, Table 2.3 p.69), significant differences with the 1997 figures appear. (I do not reproduce the 1986 estimates here because a "Do not Cite" caveat is contained in that paper. However in that paper, the rise in women's hours is even more pronounced.) Levels of worktime are significantly higher in the 1988 table, and there is a rise in both male and female hours between 1975 and 1985. (The 1988 table does not include data from the 1965 sample, presumably because there are missing categories in the 1965 sample.) These differences are particularly puzzling in light of the fact that the estimates in Time for Life exclude respondents aged 65 and over, which should raise, not lower average hours of work.


Weekly Hours of Paid Work






Change in Worktime







Women (1988, OTA)






Women (1997, TFL)












Men (1988, OTA)






Men (1997, TFL)








Similar discrepancies exist between the 1989 American Demographics article on free time and Time for Life.


Weekly Hours of Free Time for Men














Change Free Time








Men (1989, AD)





Men (1997, TFL)






One possible explanation of these differences is differences in the handling of the missing categories of the 1965 sample.

5) Because Professor Robinson did not include documentation in his published articles and did not respond to my requests for documentation on his methods, I am not sure how he has dealt with these sample selection issues. Time for Life includes some methodological discussion, but not enough for me to understand what he has done, beyond including rural respondents in 1975 and 1985, and stratifying the sample by employed and non-employed. It is my understanding from our conversations that for his earlier calculations, he compared the samples with no corrections.

6) I have not figured out where Robinson's figures for non-employed men in 1965 have come from, given that these households were excluded from the sample.

7) Leete and Schor, Mishel, Bluestone and Rose, and Bosworth and Burtless have all corrected for output fluctuations in their estimates. For example, using CPS data and correcting with a GDP gap variable, Bosworth and Burtless find a strong positive trend in hours since 1981 for males, and since 1969 for women. They find that between 1967-1989 annual hours of work rose for both men and women aged 25-64. (Burtless and Bosworth 1992. See Table 3, p.11 and note 7.)


8) The work of Juster and Stafford is particularly notable in this respect. They compared 1965 (an expansion) and 1981 (a serious recession), and then claimed that free time was on the rise.


10) These figures underestimate the extent of underemployment because they do not exclude involuntary retirees, whose numbers have also increased, discouraged workers, and prison inmates.


11) Coleman and Pencavel find that annual hours of work have fallen for both white and black males of low educational attainment, and by contrast have risen for those with 16 of more years of schooling. (Other educational categories are not provided, nor do the detailed breakdowns include 1960 and 1970 estimates.) The decline is particularly pronounced for black males with fewer than 12 years of schooling, from 1953 hours in 1950 to 1738 hours in 1988 (or 215 hours). This is the group, of course, which has experienced the largest increase in labor market constraints. By contrast, black male college graduates have increased their annual hours considerably since 1950, from 1971 to 2070, or 99 hours.


There are similar differences by educational attainment level. Coleman and Pencavel find that the average white woman with fewer than 12 years of schooling experienced a reduction in annual hours of work between 1950 and 1988 of 317 hours. College educated white women also had a reduction in total hours between 1950 and 1988, of 44 hours. However, between 1980 and 1988, hours of work rose for this group, by 131 hours. Coleman and Pencavel do not present estimates for 1960 and 1970, however, based on their aggregate estimates, it is highly likely that the full-period 44 hour decline is comprised of an initial sharp drop in hours during the baby-boom years of the 1950s, with a growth in hours beginning in the 1970s.

Among black women, the trends replicate those of men. Lesser educated women had a reduction of 230 per year, while those with college degrees increased their hours by 207.

12) Respondents who work at home presumably also include those hours.

13) The issue of bathroom time has become quite serious in the U.S. Among factory workers, constraints on bathroom use have become widespread. See Peter Linder's forthcoming book, Voiding. (199X)

14) Of course there is ambiguity about time spent in work-related voluntary associations and how those are recorded or recalled by individuals. For employees with enough job autonomy to attend work-related associations, high hours recorded as "time at work" are obviously not an impediment to participation.

15) Robinson and Godbey claim that Leete and I did not need to look at annual hours because the BLS surveys workers throughout the year, but this is a mistake. We, like almost all economists who study these questions, used the March CPS because it provides the required detail we need for our estimates.


16) Given Professor Putnam's endorsement of the time diary data, it may be worth noting that they give an ambiguous picture about trends in civic engagement. Taking all time devoted to categories 60-69 (Organizational Activities), we find a total of 20 minutes per day spent in 1965, 24 in 1975 and 18 in 1985. This pattern is not obviously consistent with the long term trends presented in Putnam (1995), which are more or less a consistent decline. If I am correct that the 1965-1975 results are partly driven by a statistical artifact, they become easier to explain. The 1965 sample, because it is has higher than average working hours, has lower than average civic engagement. So the rise in organizational activity time is also an artifact. (The fact that 1975 is a recession year could also affect the time devoted to civic engagement.) The drop in engagement between 1975 and 1985 is real, and is partly due to rising hours of market work. (Author's calculations from Robinson and Godbey 1997, Appendix A.)

17) It does imply, as Professor Putnam points out, that long hours of work do not prevent high levels of engagement.

18) Of course the real anomaly here is the growth of television watching, because television is an activity which yields relatively low process benefits, at least in terms of what viewers tell surveyers. (Juster and Stafford 1985). But if television is an addictive activity, as it may well be, or if it is one of the few activities that is pleasurable after a long, tiring and stressful day at work, then the rise of viewing is less puzzling.

19) In a true Beckerian world, these type of people should not exist. They would merely act on their preferences, with no structural rigidities to impede them.

20) 1985 figure from Shank (1986); 1994 figure from PEW Global Stewardship Initiative Survey, February 1994 (data provided to author); 1995 figure from Merck Family Fund Yearning for Balance Poll, February 1995 (data provided to author).

21) Hilton Hotels Values Survey, reported in Hymowitz, (1991).

22) Peters uses Dutch time diary data for 1995, and looks separately at men and women, ages 25-44, who are either married or co-habitating.

23) The trends by educational level are not monotonic. Persons of low educational level also display high levels of always feeling rushed, perhaps because overtime and multiple jobholding are high among non-college educated workers.


24)The "always feel rushed" measure peaked in 1992, and dropped 6 points by 1995. Robinson and Godbey 1997 report other measures of a possible slowdown in time pressure, which is consistent with the downshifting trend reported by Schor (1998).

25) Leete and Schor found that for young parents (defined as ages 18-39), the total rise for women was 241 hours. Among single parents, the vast majority of whom are mothers, the rise in total worktime was 222 hours. In addition to mean increases, the distribution across annual hours is changing. As women's commitment to jobs has increased, the fraction who work very long hours is rising. In 1967 only 3.2% of all women worked more than 2100 hours a year, but by 1987, 11.2% did. (Shaw 1992, p. 14).

26) Robinson and Godbey do not offer a comprehensive explanation for why time pressure has increased, beyond a brief discussion of the standard theory of time allocation. But that theory, as put forward by Becker and Linder, predicts that declining wages should lead to less not more time pressure. Thus, if Robinson and Godbey are correct, a substantial fraction of the population, namely those whose wages have fallen since the mid-1970s, should have experienced a decline in time pressure. Furthermore, among those whose wages have been stagnant, time pressure should have remained constant. Only individuals who experienced upward wage movements should have become more harried. However, Robinson and Godbey find a uniform increase in time pressure, which is more consistent with my contention that hours have increased, than with the standard story.