|Works Discussed in this Essay:
Rebecca Blank, "Are Part-Time Jobs Lousy Jobs?" In Gary Burtless, ed., A Future of Lousy Jobs? The Changing Structure of U.S. Wages (Brookings Institution, 1990).
John E. Bregger and Steven E. Haugen, "BLS Introduces New Range of Alternative Unemployment Measures," Monthly Labor Review, October, 1995.
Polly Callaghan and Heidi Hartmann, Contingent Work: A Chartbook on Part-Time and Temporary Employment (Economic Policy Institute, 1991).
Lawrence Mishel and Jared Bernstein, The State of Working America 1994-95 (M.E. Sharpe, 1994).
John P. Robinson and Ann Bostrom, "The Overestimated Workweek? What Time Diary Measures Suggest," Monthly Labor Review, August, 1994.
Stephen J. Rose, "Declining Job Security and the Professionalization of Opportunity," Research Report No. 95-04, National Commission for Employment Policy, 1995.
Juliet Schor, The Overworked American: The Unexpected Decline of Leisure (Basic Books, 1991).
At least since the 1980s people have said that they work "too hard"—that they are spending too much time on the job, with too little left for family, chores, or leisure. In 1991 this frustration became conventional wisdom thanks to Juliet Schor's best-seller, The Overworked American, which demonstrated that Americans worked an average of 163 more hours in 1990 than they had in 1970—or the equivalent of nearly an extra month of full-time work per year. According to Schor, men were working two and a half more weeks per year; women an average of seven and a half more weeks. These were startling statistics, reversing more than a century of gradual reduction in working time as society became richer and more productive. If Americans were working this much longer, then they were not only overworked by traditional U.S. standards, they were setting new world records.
But critics challenged Schor's data and pointed to a logical flaw in her argument. Today, more people work part-time because they can't find full-time work; more are temping or working as short-term independent contractors. Job insecurity is rampant, and other statistics show that the number of weekly hours on the typical job has actually shrunk steadily since World War II. It seemed implausible that Americans were simultaneously "overworked" and "underemployed," thus prompting the question: Were Schor and all the harried Americans who cheered her book's appearance wrong?
Not necessarily. It's possible, for instance, that we are mixing apples and oranges. The number of contingent jobs and average weekly hours refers to "jobs," not people.
If individuals are moonlighting more—working multiple jobs in any given week—then the average workweek reported by employers can still shrink while the average workweek reported by workers can actually expand. It is also possible that one sector of the workforce is "overworked" while another portion is "underemployed."
But the real story turns out to be even more intriguing and complicated. Based on a new analysis of the data, we have found that Americans are indeed working longer than they once did, if not quite as much as Schor would have us believe. But, more importantly, we have also found that many Americans are both overworked and underemployed. Because of growing job instability, workers face a "feast and famine" cycle: They work as much as they can when work is available to compensate for short workweeks, temporary layoffs, or permanent job loss that may follow. What's more, while American families as a whole are putting in more time, that work isn't producing significant increases in living standards. For the typical two-breadwinner household, having both parents work longer hours may not mean an extra trip to Disney World or nicer clothes for school; more likely, it means keeping up car payments or just covering the costs of food and housing.
At one extreme are workers like Bill Cecil, a 50-year-old United Auto Worker member recently portrayed in the Wall Street Journal. Averaging four hours a day in overtime and volunteering to work seven days a week for most of the year, Cecil clocks an average of 84 hours a week at a Chrysler plant in Trenton, Michigan, where he works as a skilled millwright. In the past two years, Cecil's extraordinary work effort has paid off. He has averaged more than $110,000 a year in gross pay. Sacri ficing, by his own admission, "freedom and time with family," he works as much overtime as the company will let him in order to help send his four kids to college, fill his lunch pail with lobster salad rather than luncheon meat, and underwrite his golf habit, which he indulges whenever a vacation or a layoff permits. While Bill Cecil's case is exceptional, 70 percent of the skilled-trades workers at his engine production facility are working at least some extra hours most weeks.
Increasing overtime is becoming more commonplace throughout the manufacturing industry. For the first four out of five post-World War II business cycles, average weekly hours of work for production and nonsupervisory workers in manufacturing remained roughly constant, varying only slightly between 40.1 and 40.4 hours. However, during the current business cycle, from 1989 to 1996, the average workweek has jumped to 41 hours—with average overtime reaching a post-World War II peak of 4.7 hours per week in 1994.
A Fortune magazine poll of Fortune 500 CEOs in 1990 found a similar tendency toward more work among executives. Sixty-two percent of CEOs reported their executives were working longer hours than they had ten years before. They reported that nearly nine out of ten of their high-level executives normally put in more than 50 hours a week while three-fifths of middle managers did the same.
Moonlighting is also on the rise. In 1979, 4.9 percent of U.S. workers reported working more than one job during the same workweek. By 1995, the percentage was up to 6.4 percent. Virtually all of this increase has occurred among women, who now represent nearly half of all multiple job holders. According to a recent survey sponsored by the Washington Post, the Kaiser Family Foundation, and Harvard University, two out of five families report they have sent an additional family member into the paid labor force or had an existing working member take on an additional job—simply because the family needed extra money.
Working more makes sense from both the employers' and the employees' perspectives. Manufacturing firms like Chrysler do not hesitate to schedule large amounts of overtime when product demand outstrips supply, even if it means paying time and a half, double time, or triple time during holidays, because it is still less expensive than covering the high fixed costs of recruitment, training, and possibly the underwriting of future severance pay associated with hiring new workers. For salaried white-collar employees who are exempt from hours regulations, the arithmetic is even simpler—the extra hours often cost the company nothing at all.
Fortune 500 CEOs and their executives say they need to put in overtime just to keep up with global competition and compensate for internal restructuring or middle-level management downsizing. But why would blue-collar workers so willingly give up leisure or family time? Schor has identified one factor, which she calls "capitalism's squirrel cage"—an "insidious cycle of work and spend" where people work long hours to support a material lifestyle always a bit beyond their reach. But that suggests the increased work hours are buying a rising standard of affluence, which is somewhat misleading. Indeed, a more compelling reason for extra work is the slowdown in wage growth during the past two decades. Between 1947 and 1973, real hourly wages for production and nonsupervisory employees rose by 79 percent. Since 1973 hourly wages have actually declined by more than 13 percent. For many workers, working longer hours is the only way to compensate for lower hourly wages.
Of course when pollsters ask people, "Would you like to work less?", most say "yes." But when pollsters include a caveat—that fewer working hours would mean less take-home pay—the answer changes sharply. Over the last 20 years, surveys with this appropriately worded question have been answered with great consistency: Approximately 60 percent say they prefer their current work schedule and pay. Of those who express a desire to change their working time, more people, by about three to one, express the desire to work longer rather than shorter hours.
Union negotiators in the U.S. know this, which is why they so rarely make reducing work time a priority in collective bargaining. In fact, many workers complain bitterly whenever management prerogative or union contract restrict overtime hours. In a "real experiment" on this issue, New York's state government in 1984 began allowing their workers to take voluntary reductions in work schedules without affecting their career statuses. The plan was flexible and permitted workers to move on and off "V-time." Since its inception, however, the program has never enrolled more than 2 percent of the workforce.
This expressed desire for more hours is consistent with the trend toward more contingent work. At the same time that many workers are looking to expand their number of working hours, the economy has shifted steadily from manufacturing to sectors like retail trade and services, where part-time work is more common. One estimate for 1995 places the total number of contingent workers (part-time, temporary, and contract workers) at close to 35 million—28 percent of the civilian labor force. Of these, 18 percent of the workforce or 23 million workers were part-time, working 35 hours or less per week. Smaller in absolute numbers, but growing much faster, is the temporary workforce, which between 1982 and 1995 more than tripled to 1.4 million workers. Manpower, Inc. now boasts it is the largest employer in America, submitting more W-2 forms to the Internal Revenue Service each year than any other firm. The number of contract and self-employed workers is also growing rapidly, indeed explosively. The U.S. General Accounting Office has reported that the number of individuals who are self-employed or working under personal contract was growing at more than 13 percent a year in the late 1980s. By 1988, 9.5 million Americans worked for themselves either full-time or as a supplement to regular or part-time employment.
A large proportion of the contingent workforce has chosen voluntarily to work part-time, as temporaries, or as independent contractors. Still, involuntary part-time employment is growing much faster than the voluntary variety. In 1973, 19 percent of total part-time employment was accounted for by individuals who wanted full-time jobs but could not find them. By 1993, this proportion was up to 29 percent. The incidence of involuntary part-time work is especially high among men. In 1985, one in four part-time women reported their part-time status was involuntary; nearly half of all part-time men did so.
For the labor force as a whole, these numbers begin to add up. Since 1994, the Bureau of Labor Statistics (BLS) has been compiling a new set of alternative measures of unemployment and underemployment—what the Labor Department calls "labor resource underutilization." In addition to the official unemployment rate, the BLS adds three types of "underutilized" workers: (1) those who have stopped looking for work only because they have become discouraged by their apparent job prospects; (2) those who are "marginally attached" to the civilian labor force; and (3) those who are working part-time only because they cannot find full-time jobs. The "marginally attached" include those who want and are available for a job and have recently searched for work, but have left the official labor force because of such constraints as child care or transportation problems.
The official unemployment rate in 1995 was 5.6 percent with an average of 7.4 million failing to find work each month. Adding discouraged workers to the total brings the "underemployment" rate up to 5.9 percent. Adding the "marginally attached" ups the rate to 6.8 percent. Finally, adding in the involuntarily part-time brings the rate to 10.1 percent. In what was a good year for the economy and employment growth, 1995, the total number of unemployed and underemployed workers reached nearly 13.5 million—one in ten of the total labor force.
All of these trends contribute to the decline in the average workweek reported by employers since at least World War II. As the chart on this page indicates (see "The Shrinking Workweek"), from 1947 to 1958 the average workweek was nearly 40 hours, the "full-time" standard for much of this century. In the most recent business cycle, the average workweek fell below 35 hours, the cutoff normally used to define a "part-time" job. Ironically, in what is supposed to be an "overworked" nation, the typical job is now part-time! Again, we should ask, "overworked," "underemployed," or perhaps both?
WHOSE NUMBERS SHOULD WE BELIEVE?
Whether we believe that Americans are overworked or underemployed depends, in part, on whether we believe the work time data. Many economists question Juliet Schor's findings and it's not hard to understand why: The idea that Americans are, on average, spending the equivalent of an extra month a year in paid work seems almost unbelievable.
But is it? According to one recent study, Schor's basic finding holds up, but her estimates of overwork appear somewhat exaggerated. Using data from the Current Population Survey, Larry Mishel and Jared Bernstein of the Economic Policy Institute have re-estimated annual work hours for various years. Their research confirms the general proposition of increased annual working hours, but for a comparable period (1973 to 1992) their estimate is only three-fifths as large as Schor's. They calculate that in 1973, the average workweek (for both employed and self-employed workers toiling in the public as well as the private sector) was 38.4 hours. The average work year was 43.2 weeks, yielding an annual estimate of 1,659 hours of work. By 1992, the average workweek had climbed by 0.6 hours while the average work year had increased to 45.2 weeks. Hence, annual average hours had risen to 1,759, an increase of 100 hours or 6 percent—but 63 hours less than Schor's estimate.
|Since World War II, the average workweek, as reported by employers, has declined.|
|Years||Average Week* In Hours|
|*Total private-sector employment.
Source: Council of Economic Advisers, Economic Report of the President, 1987, Table B-41, and Council of Ecomoice Advisers, Economic Report of the President, 1996, Table B-43
Yet even these more reasonable figures raise questions. Note that the steady decline in the average workweek reported by employers as shown in the "Shrinking Workweek" chart suggests that for average hours per job to decline while average hours per worker increases, there would have to be enormous increases in moonlighting. This seems implausible, because even with the recent increase in moonlighting, only 8 million workers out of a workforce of more than 125 million report holding more than one job.
The problem may be with the very survey data upon which Schor, Mishel, and Bernstein all rely. The estimates of hours worked come from the March Current Population Survey (CPS) for each year, which the U.S. Census Bureau and the Department of Labor compile annually. Among several dozen questions about labor market activity, the CPS asks respondents to report "hours worked last week" and "usual weekly hours of work last year." Individuals have only a few seconds to answer these questions. In making what may be a wild guess, particularly for those people whose hours vary substantially from week to week, the individuals frequently guess high. And the more harried and rushed they feel, the higher they guess. Could you account for the actual number of hours you spent working last week?
A more accurate measure of hours worked comes from special studies that target the work time issue by asking respondents to keep a 24-hour time diary of everything they do over a one- to two-day period. Such time diary surveys were first carried out by the University of Michigan Survey Research Center in 1965 and 1975, and then again by the University of Maryland in 1985. The accuracy of work time estimates derived from this survey approach is presumably superior to CPS measures for two reasons. First, the exercise's sole purpose is studying the use of time; second, respondents do not have to plum their memories for what they did a week ago or try to calculate instantly how many weeks they worked all of last year.
Sure enough, a comparison of CPS-estimated hours of work and diary entries suggests that people overestimate how much they work—and that the overestimates get bigger the more hours they put in. According to John Robinson of the University of Maryland and Ann Bostrom of Georgia Tech University, who studied the two sets of surveys, among those estimating 20 to 44 weekly hours, the CPS-type estimates were only slightly higher than the diary entries. But among workers claiming to "usually" work more than 55 hours per week, the gap was 10 hours or more per week. Robinson and Bostrom concluded that "the diary data suggest that only rare individuals put in more than a 55- to 60-hour workweek, with those estimating 60 or more hours on the job averaging closer to 53-hour weeks." Moreover, using the diary studies for 1965, 1975, and 1985, Robinson and Bostrom found a systematic increase in the size of the estimate gap over time. The gap rose from just one hour in 1965 to four hours in 1975 to six hours in 1985, which is more than enough to account for the alleged "overwork" that Schor and Mishel and Bernstein claim to have found.
When Robinson and Bostrom analyzed diaries for 1965, 1975, and 1985 more carefully, they found only small changes in hours worked among those who normally work 20 hours or more per week. Between 1965 and 1985, men's average hours declined by 0.7 hours per week from 47.1 to 46.4 hours, while working women's hours increased by the same amount (0.7) from 39.9 to 40.6 hours. If these numbers are believed, then the source of increased hours worked that Schor observed must be new entrants to the labor force—again, many of them women—and part-timers who have increased their part-time hours. Of course, whether this should be counted as "overwork" or not is a matter of deeply divided opinion.
SCHORING UP THE FINDINGS
What can we make of such sharply different findings? To answer this question, we decided to pursue still another approach, using yet another type of survey instrument. So far, all of the research on working hours has relied on data snapshots at different points in time using either the CPS or diary information. An alternative approach is to use longitudinal data—in other words, information about the same people gathered year after year—to track working hours. Using this information, we can follow the work time pattern of, say, a particular age group over several years—as one could with CPS data—or follow the same workers over time. Here we do both in order to provide completely new estimates of work time. We use the Panel Study of Income Dynamics (PSID), a data set of families that the University of Michigan Survey Research Center has been following since 1968. The long-running nature of the PSID permits a comparison of working time during two ten-year periods—the 1970s (1969-1979) and the 1980s (1979-1989). (These periods had similar growth rates in real output per person and in job creation, and each encompassed two complete business cycles. Hence, the comparison is a reasonable one to make.) We also combine the two decades of data to follow a particular age group (in this case, prime age workers with job experience) in order to derive typical trends in annual work hours for men and women, whites and blacks, and for segments of the population with differing amounts of schooling.
While the PSID does not provide the full detail nor perhaps the precision of hours estimates culled from the diary method, its data on hours worked is superior to that of the CPS. First, PSID asks respondents to detail their work experience by recalling how many days they were on vacation, on sick leave, on strike, or on leave due to other family members' illness. It then asks respondents to answer questions about regular hours of work per week and weeks worked on his or her main job. Then it poses the same questions concerning up to three other jobs respondents held during the year. Finally, all of this information is combined to yield an estimate of annual hours. Obviously, this approach suffers from recall problems, much as the CPS does, but the detail on each job presumably permits a better estimate.
The first part of our analysis is based on computing the average hours of work in each year from 1967 through 1989 for prime age workers (ages 25-54). In this case, we use the PSID as a series of cross sections where the sample individuals in each year vary as younger individuals enter the prime age group and aging workers leave it. We limit our sample in each year to those who reported hours of work, eliminating those from consideration who were out of the labor force in a given year. We generated separate estimates for men and women, and broke the findings down by race and by education. The graph on page 63 ("Individuals Working Harder, Again") provides the results for all prime age workers.
There is clear evidence of variation related to the business cycle. Average hours dip sharply in 1970-71, in 1975, and then again during the steep 1981-82 recession. But overwhelming the business cycle is a U-shaped trend in hours of work. Average hours appear to decline through the early 1980s and then begin a sharp recovery throughout the decade. If we compare 1979 and 1989, the last two business cycle peaks, there does indeed appear to be an increase of 79 hours per year for the average worker. But over a longer period, this increase marks not so much a startling increase as a return to levels that prevailed in the late 1960s.
To obtain a more accurate estimate of the trend in hours, we ran a statistical exercise to control for the business cycle. Having done this, we find a small, but statistically significant, overall upward trend in annual hours for prime age workers as a group. The trend amounts to only 3.3 hours per year. Hence, over a 20-year period, we find a 66-hour increase in annual work—the equivalent of 1.5 weeks of full-time work per year. This is well below Schor's estimate of 163 hours and a third below that what Mishel and Bernstein found. But, importantly, the trend is decidedly upward, in contrast to the essentially flat line Robinson and Bostrom found for the 1965-1985 period using the diary method.
Among men, working hours declined slightly, after we control for the business cycle. But for women, hours increased significantly. Indeed, our estimate of 18.8 additional hours per year translates into a 20-year total somewhat greater than even Schor's estimate. We also find significant differences in the hours trajectories by race. Reflecting trends well documented elsewhere, our estimate of a decline of 7.7 hours per year for black men translates into an average work year in the late 1980s more than 150 hours shorter than in the late 1960s. In 1989, we estimate that black men averaged only 1,950 hours per year compared with just under 2,300 hours for white men. Higher unemployment rates are responsible for part of this difference. Shorter workweeks explain the remainder. This suggests that the continuing earnings gap between white and black men is only partly accounted for by differences in wage rates—the traditional measure of labor market "success." A large amount of the gap is also due to differences in hours worked. Wage rates matter, but what is really killing black men in the labor market is their inability to find full-time, full-year jobs as readily as their white counterparts.
The racial gap in hours worked among women shows an intriguing time pattern. On an annual basis, there appears to have been virtually no gap in work hours in 1967. The gap then widened significantly, so that by the mid-1970s black women were working almost 200 hours more per year than white women. White women caught up again, and by 1989 white and black women were working virtually the same amount. To close the gap, white women's cycle—adjusted hours had to rise substantially faster than that of black women. This is precisely what happened. Over 20 years, white women's annual hours increased by the equivalent of 10.3 weeks of full-time work, nearly double the 5.4 weeks for black women.
As a general rule, then, there has been a slight reduction in men's work hours and a large increase in women's hours. Given these trends, we can ask what has happened to family work effort as America has undergone the transition from the prototypical "Ozzie and Harriet" division of labor of the 1950s to the dual-career family of the 1980s and 1990s.
To investigate the trend in family work effort, we have estimated the combined hours of work for "prime age" families in which both husband and wife are working. The long-term trend is shown in the graph above ("More Total Hours . . ."). There is a clear and nearly unbroken trend toward much greater work effort, interrupted only modestly by the recessions of 1971, 1974-75, and 1980-1982. By 1988, prime age working couples were putting in an average of 3,450 hours per year in combined employment, up from 2,850 two decades before. (Data for family hours and earnings in our version of the PSID were incomplete for 1989, so we use 1988 as the end point for this analysis.)
Adjusting for business cycle effects, we calculate that for all husband-wife working couples, family work effort increased by more than 32 hours per year for each year of the 1970s and 1980s. Hence, in the span of just two decades, working husband-wife couples increased their annual market work input by a cycle-adjusted 684 hours or 4 months of full-time work. The typical dual-earner couple at the end of the 1980s was spending an additional day and half on the job every week. If individuals are not more overworked than before, families certainly are.
Increases in family work effort differ significantly depending on race and education. The increase in working hours among white working couples was 60 percent larger than the increase for black couples—a reflection of both the sharp decline in black men's hours and the large increase in white female work effort. More-educated working couples also increased their work effort more than those with less schooling. Those in which the husbands had at least undergraduate college degrees increased their combined work effort by nearly 730 hours compared to an increase of only 490 hours for couples headed by high school dropouts. The "overeducated" are the ones most "overworked."
Has this enormous increase in work effort paid off in terms of increased family earnings? The graph on the right-hand page (". . . and for What?") shows our comparison of hours worked and earnings growth.
For prime age working couples as a group, combined real earnings rose by 18.5 percent between 1973 and 1988. (This represents an increase from $43,851 to $51,955 in 1989 dollars.) Most of this modest increase, however, did not come from improved wages, but from increased work effort. The 18.5 percent increase in real earnings was purchased with a 16.3 percent increase in hours worked. Over the entire 15-year period, the combined average husband-wife hourly wage increased by only 1.8 percent—the equivalent of a real hourly wage increase of less than 30 cents over the entire period, or 2 cents each year!
As such, Schor's "squirrel cage" does not appear to be far off the mark. American mythology holds that long hours will pay off in a steadily increasing standard of living; in other words, sacrificing time with family can pay for a dishwasher or microwave and, down the road, a more expensive college for one's children. Yet from a purely material perspective, all the extra hours from the "average" working family have yielded only a very modest improvement in the amount of goods and services they can buy.
But even this story is too sanguine for most families. When we break down the hours and earnings data by education group the tale gets even more depressing. Most Americans are not working harder so they can afford a fancier minivan; they're just trying to make payments on their old car or cover the rent. When you remove from the equation families headed by a worker with at least a college degree, it turns out that the enormous increase in work effort over the past 20 years has allowed families to maintain their old standard of living—but almost nothing more. For families headed by high school dropouts, the situation is the most dismal. Between 1973 and 1988, such families increased their annual work effort by nearly 12 percent yet ended up with 8 percent less annual income. For families headed by high school graduates or some college, work effort was up by 16 to 17.4 percent, producing less than a 4 percent increase in total earnings. These families are trapped in an Alice in Wonderland world, running faster and faster just to stay in the same place. For all of these families, the "family" hourly wage has fallen precipitously, by as much as 17 percent in the case of the high school dropout.
Of course, more work still pays off for one group: families headed by a college graduate. These families increased their work effort by about the same percentage as those headed by high school graduates or those with some college, yet their material consumption standard increased by nearly a full third between 1973 and 1988. Unfortunately, such well-educated families comprise less than a third of all American dual-income families.
FEASTING BEFORE THE FAMINE
To this point, we have been concerned with trends in hours worked and earnings for particular demographic groups. We now shift our attention to an equally important issue. What can we say about the year-to-year variation in work hours for individual workers? This is of obvious importance given the debate over the apparent growth in job insecurity. If a worker is insecure about his job, then it is possible he may voluntarily work as much overtime as he can in order to cushion the blow of depressed income from future joblessness. Or, for that matter, he may work extra hours because he has to pay off credit card debts that accumulated in the last bout of underemployment.
To study this issue, we again separate the PSID into two ten-year time frames corresponding to the 1970s (1969-1979) and the 1980s (1979-1989). To focus on prime age workers, we restrict our analysis of each decade to individuals who began at ages 24 to 48 and ended the decade at ages 34 to 58. This prime age range provides a sample of those who, for the most part, are old enough to have completed their formal schooling and not old enough to have begun cutting back their work hours in anticipation of retirement.
To measure inter-year variation in work hours for these prime age workers, we have developed a special measure we call "Hi-Lo." This statistic measures the proportion of individuals in a group who, during a decade, experience at least one year in which they work more than 2,400 hours and at least one year of 1,750 hours or less. The "Hi" value is equivalent to an average workweek of approximately 46 hours or more. The "Lo" value is equivalent to less than 35 hours per week. These cutoffs correspond to common definitions of "overtime" work and "part-time" work.
According to our analysis, among all prime age males, nearly three out of ten workers (28 percent) had at least one year of substantial "overtime" and at least one year of significant "underemployment" during the 1980s. Compared to the 1970s, the proportion of such individuals experiencing such hours variation was up by nearly 8 percent.
For black men, the incidence of Hi-Lo variation is substantially higher than among white men, with 37 percent of black men experiencing this variety of "feast and famine" employment history. Those who have completed a high school diploma or college degree appear to experience less hours variation than those who drop out of high school or do not complete college.
But by the far the strongest indicators of the feast-or-famine syndrome emerge when we break the Hi-Lo numbers down according to earnings levels and the number of job changes. Among men in the lowest 20 percent of the earnings ladder, four out of ten experience Hi-Lo hours variation—more than double those in the top 20 percent. Those who have low earnings even when they are working full-time are the most likely to experience a feast-and-famine work life. Not surprisingly, those who change employers more often face the highest rates of Hi-Lo activity. More than half of prime age men who change employers at least four times in a decade face years of "overtime" and years of "underemployment." For women, the "overworked-underemployed" phenomenon expanded as well between the 1970s and 1980s. In the earlier decade, only 12 percent experienced such Hi-Lo work histories. In the 1980s, nearly 21 percent of women spent their lives on the work time roller coaster.
High school dropouts have seen a substantial rise in Hi-Lo activity between the two decades. But so have college graduates and the top-quintile earners. One might conjecture from these findings that those with the fewest skills and those in the ranks of middle managers have been particular victims of downsizing. Future research with the PSID should be able to provide more evidence to test this hypothesis.
Taken together, the results for men and women suggest that increased job instability has led to increased hours variability for men and increased hours and variability of work for women. While the data presented here cannot prove that male job instability causes men to work more overtime when it is available and at the same time increases women's workforce participation, the results are fully consistent with such a thesis. In short, we can conjecture that "underemployment" of men may be leading to "overwork" for families. Because Dad's work prospects are more uncertain than ever, Mom is working harder than ever before.
In the end, then, it turns out that both Schor and her critics were partially right. There is compelling evidence of both overwork and underemployment not only across the workforce, but for individual workers (particularly men) who may face bouts of full-time work interspersed with years in which part-time hours are the rule. In both the 1970s and the 1980s, more than one-quarter of men experienced a decade in which they worked at least one year of "overwork" (more than 46 hours per week) and at least one year of "underemployment" (less than 35 hours per week). How much of this is voluntary cannot be judged, but the finding is consistent with other research that shows growth in job instability and income insecurity. Adding to this evidence is our finding that those workers who change jobs more than four times in a decade are more than three times as likely to face bouts of "overwork" and "underemployment" as those who have at most one job change during the same period.
The reason for this overwork, ironically, turns out to be underemployment. Men are working overtime to compensate for expected job loss in the future. Women have expanded their work effort to cover for what otherwise would be a sharp reduction in family living standards.
What does this foreshadow for family and community? Americans will not find a better balance between work and leisure, between earning a living and spending time with loved ones, between wage earning and "civic engagement," until the economy provides long-term employment security and rising wages. If past is prologue, the last 25 years of U.S. labor market history should not make us sanguine about the possibilities.
There is serious political talk, now and then, about legislating shorter weeks. But no matter how much we may complain about being overworked and no matter how much we worry about latchkey kids, few American workers will support political action unless it is tied to a much broader set of policies aimed at improving material living standards along with more leisure.