September 10, 2024
By Scott Winship
Sometimes it seems like Americans can’t decide whether we work too much or too little. We hear that because of rising inequality and a lack of good jobs, workers must toil too many hours at wages too low to support a family. By other accounts, the machines— if not robot overlords, then at least their touchscreen-kiosk cousins—are already beginning to take our jobs. Women work more than in the past. But is that because of the declining pay of their husbands, or is it a reflection of greater economic freedom? Older workers can’t afford to retire, but all those men sitting in their parents’ basements playing video games need to get to get off their butts.
What many of these claims have in common—whether they decry how much we must work or how scarce jobs are—is economic declensionism. It’s an increasingly popular view that the economy is failing us, necessitating greater government intervention. Or perhaps the blame lies with policymakers wedded to outmoded ideas. Either way, declensionists of the left and right tell us it’s time for a fundamental rethink.
Assessing the evidence over one hundred years and more, however, suggests that declensionists have it backwards. Both declines and increases in work that have occurred over the long run turn out to mostly reflect the country’s increased affluence rather than an underperforming economy. By misunderstanding trends in work as problems created by rapacious capitalism or out-of-touch elites, we fail to grapple with the choices everyday Americans and denizens of other rich countries have made. The resulting trade-offs have created a variety of quality-of-life improvements while destabilizing other aspects of society. If we wish to remedy problems like increased dependence on safety net benefits, declining marriage, or rising single parenthood, it will not do to indulge myths of economic decline.
A Century of Labor Supply
Before taking a more granular look at how work in America has evolved and why, it’s worth getting the big picture right. Figure 1 goes all the way back to 1890, but for our analyses, it’s useful to begin with 1915. That lets us start from a relatively stable period before the US entered World War I. We can distinguish four broad eras in assessing long-term changes in work—1915 through the 1950s, the 1960s and 1970s, the 1980s and 1990s, and the 2000s and 2010s.
Figure 1. Annual Hours Worked per Adult, 1890 to 2022
The first era is characterized by declines in hours on the job, primarily due to the workweek getting shorter. From 1915 to 1959, labor supply (hours per person) fell 22 percent, a trend punctuated by large fluctuations during two world wars and the Great Depression (Figure 1, Line 1).[4] We can separate the drop into the change due to the probability of working anytime during the year (the employment rate) and the change due to the average annual hours workers spend on the job.[5] The entire decline in labor supply over these 44 years was due to fewer annual hours among people who worked.[6] The employment rate was 56 percent in both years.
Using data from the national census (Figure 1, Line 2), conducted every ten years, we can decompose this post-war decline in annual worker hours further.[7] Starting in 1940, the census began to include a question about weeks worked in the previous year. It also asked about hours worked in the previous week, which I use to approximate the typical weekly hours worked the previous year. Labor supply actually rose from 1939 to 1959, due to strong employment growth (1959 was a peak in the business cycle). But consistent with the trend back to 1915, annual hours per worker declined over this shorter span too (by 9 percent). Most of the drop in worker hours (85 percent) was due to a fall in average weekly hours, with the remainder due to a decline in average weeks worked per year.[8]
The second era covers the 1960s and 1970s, when men’s employment fell and women’s rose. From around 1959 to around 1982, hours per person fluctuate with the business cycle but show little trend. Because there is so much variation corresponding with cyclical factors, it’s important to compare years that are at similar points in the business cycle. With the census data, we can compare 1959 to 1979, both peaks. Labor supply rose by 2 percent over these twenty years.
This increase was fully accounted for by rising female labor supply. Among men, employment, weeks per worker, and hours per week all fell, and men’s labor supply declined by 10 percent as a result. The drop in their employment rates—from 85 percent of men to 79 percent—accounted for two-thirds of that decline. The shortening of the average workweek accounted for another one-fourth of the decline in men’s labor supply.
Among women, labor supply increased 36 percent from 1959 to 1979 (compared with 20 percent from 1939 to 1959). Over 80 percent of that was due to employment rates, which rose from 43 percent of women to 55 percent. Women worked more weeks per year over time, but like men, their average workweek continued to shorten. Combining men and women, the drop in average weekly hours (4 percent) was less than half that from 1939 to 1959.
The 1980s and 1990s comprise the next era of labor supply’s evolution. Worker hours (including the length of the workweek) increased among both men and women, but both the growth in women’s employment and the decline among men decelerated.
We can use richer yearly data for this period, from the Annual Social and Economic Supplement (ASEC) to the Current Population Survey, which includes reliable estimates back to 1967. (See Line 3 of Figure 1. Line 2 confirms the trend using other data sources.[9]) While hours fluctuate with the business cycle, the general trend is upward between the peak years of 1979 and 2000. Labor supply rose 11 percent, ending at the highest level in the postwar era, though still lower than in the pre-Depression years.
As in the previous two decades, the increase in labor supply during the 1980s and 1990s was entirely due to rising hours among women, which went up 30 percent. Women saw increases in both annual hours per worker (up 18 percent) and employment rates (up 11 percent). Rising hours per worker accounted for 62 percent of the increase in female labor supply, accelerating past the 1967-to-1979 growth in work hours. The increase in women’s work hours, in turn, was driven primarily by a rise in weeks worked. The average workweek lengthened too (among women and men). Women’s employment growth decelerated compared with the 1960s and 1970s, with the rate rising from 58 percent in 1979 to 64 percent in 2000.[10]
Men’s labor supply was little changed over this period, falling during recessions from 1979 to 1982 and from 1989 to 1992 but otherwise rising. While women saw increases across the age distribution, men’s labor supply declined for every age group. Men’s hours were flat only because the population shifted from younger, less experienced men to middle-age men working more hours (who nevertheless worked slightly less than their counterparts in the past).
Unlike in the 1960s and 1970s, annual hours per worker rose modestly among men (by 5 percent), driven primarily by an increase in weeks worked. Employment rates fell, but by about half as much as from 1959 to 1979. Though 81 percent of men worked in 1979, just 77 percent did in 2000.
From 2000 to 2022, the most recent era, most of the increase in labor supply over the previous two decades was reversed. We can compare the two business cycle peaks of 2000 and 2019 to assess changes over the period. Hours per person dropped 7 percent during these years. This decline was entirely due to the employment rate falling from 70 percent to 65 percent. Average hours dropped 12 percent just from 2000 to 2010, leaving labor supply lower than in any year since 1983. Hours in 2020 were similarly low. To some extent, these depressed levels reflect the severity of the Great Recession and the COVID-19 recession, but labor supply also fell from 2000 to 2003 when the economy was expanding. Today, hours are roughly where they were in the first years after World War II—23 percent lower than in 1900.
The 2000-to-2019 period saw hours and employment declines for men but also, for the first time, for women. Women’s labor supply fell 2 percent. Their hours per worker rose, primarily due to an increase in weeks per year, but their employment declined from 64 percent to 60 percent. Hours per person fell 11 percent among men, and male employment dropped from 77 percent to 70 percent.
Given this complex picture, it’s no wonder we can’t decide whether Americans should be working more or less. Some of us work more, others less—and sometimes the same people work more or less at different times. We can make sense of these changes by focusing on a few key drivers, each of which relates to rising American affluence.
Productivity Growth
Affluence has reduced labor supply by making leisure more affordable through higher labor productivity. As noted above, the pre-1960s reduction in work overwhelmingly reflected the shortening of the workweek. This was a period that saw the gradual spread of the eight-hour day and, eventually, the institutionalization of the forty-hour, five-day workweek by the Fair Labor Standards Act of 1938.[11] But while that legislation could explain a more-or-less one-time shortening of the workweek—perhaps explaining why weekly hours were 14 percent lower in 1947 (after the war) than in 1929—it cannot account for the longer-term fall.[12]
Behind the success that organized labor and progressive lawmakers enjoyed in reducing the workweek was the remarkable rate at which the economy grew during this period. Gross domestic product per hour (labor productivity) was three times its 1915 level in 1959.[13] The high productivity growth during this period (2.5 percent annually for 44 years) provided an ever-greater bounty from the same amount of work.
That strong productivity growth (and its continuation through the mid-1960s) facilitated the transition of the economy to an eight-hour day and forty-hour week. Absent productivity growth, steadily falling hours must produce lower earnings for workers or lower profits for employers (or both). But if workers can produce the same amount of value in less time, both they and employers can absorb a scaling-back of work hours. Attributing the decline in weekly hours to labor legislation is a mistake, because without those productivity gains, there would have been insufficient support for hours mandates from either employers or workers. And private employers could and did institute shorter workweeks even without legislation.[14]
From 1959 to 1979, productivity growth fell—to a still-respectable 2.2 percent per year over 20 years—and the workweek shortened again, but this time only by 4 percent. The entire decline in weekly hours came between 1959 and 1969, when productivity rose by 2.6 percent annually. The shortening workweek only accounted for about 15 percent of the drop in labor supply for men under age 60, and among women in the same age range, it offset little of the increase in employment and weeks per year.
Starting in the late 1960s, productivity growth slowed, increasing just 1.6 percent annually over the 50 years from 1969 to 2019. Hours per week rose by 2 percent from 1969 to 1979 and by 3 percent from 1979 to 2000, and weekly hours fell by only 1 percent from 2000 to 2019 (when the employment rate was also falling). Since the 1960s, changes in employment and weeks worked have been much more important than changes in the length of the workweek in affecting labor supply.
In short, one reason we are nowhere near the 15-hour workweek predicted for 2030 by John Maynard Keynes, writing in 1930, is that productivity growth slowed markedly.[15] That created unappealing tradeoffs between a shorter workweek, on the one hand, and lower earnings and profits on the other. Thanks to diminished productivity growth, the workweek is essentially the same length today as 50 years ago.
Federal Social and Education Policy
Policies enacted by the federal government over the past hundred years have also reduced labor supply. The nation’s rising affluence made it possible to provide retirement pensions, disability insurance, and a safety net for low-income seniors, disabled adults, and families with children. Sometimes reducing work was the point of these policies (as in the case of retirement and, for the most part, disability benefits). In other cases, reduced work may have been an unintended consequence (as when able-bodied adults chose to replace earnings with federal assistance).
These policies also played a role, however, in keeping the workweek from shortening after the 1960s. Specifically, they steered the means of achieving greater leisure away from getting weekly hours down while remaining on the job and toward labor force withdrawal.
Retirement Policy
The most important of these social policies was Social Security, which started paying out retirement [SW2] benefits in 1940. The overall employment rates of men and women rose from 1939 to 1959 (by 5 percentage points and 15 points, respectively). For men over 59 years old, however, employment and weeks worked declined during the 1950s while increasing among younger men. (Older women’s employment rose, but was still just 20 percent in 1959, versus 55 percent among older men.)
During the 1960s and 1970s, employment dropped from 55 percent to 39 percent among older men (versus just 4 points among men under 60). It also fell by 2 points among older women despite rising 17 points among younger women. Hours per person fell by one-third among older men during this period and by 11 percent among older women. The introduction of the Supplemental Security Income (SSI) program in 1974, providing benefits to low-income seniors, may have encouraged retirement beyond the impact of Social Security.
We can use the ASEC, to quantify the contribution of retirement to the overall decline in employment. Employment of men at least 16 years old fell from 86 percent in 1967 to 81 percent in 1979. Rising retirement was responsible for 46 percent of that drop.
Between 1979 and 2000, hours among men age 60 or older declined by 14 percent, with essentially the entire drop explained by the fall in employment from 40 percent to 34 percent. Retired men accounted for 69 percent of the overall 4-point decline in male employment during the 1980s and 1990s.[16] Over the extended period from 1967 to 2000, retirement explained 59 percent of the employment decline among men.
Employment, weeks, and weekly hours all increased among older women between 1979 and 2000. That presaged broader trends after 2000, when all three components of labor supply rose among men and women older than 59 years old.
Disability Policy
Disability policy also had an important role in reducing labor supply, particularly among men. The Social Security Act of 1935, the same legislation enacting the Social Security pension program, also created the Social Security Disability Insurance (SSDI) program. From 1967 to 1979, men who indicated they were disabled or ill accounted for 29 percent of the overall drop in men’s employment. Disability explained 59 percent of the 3-point employment decline among prime-working-age men (ages 25 to 54).
Fully 96 percent of the decline in male employment due to disability over these 12 years occurred between 1968 and 1975. Some of that may have been due to the 1974 introduction of SSI, which provides cash transfers for low-income disabled Americans in addition to help for the aged. However, both the share of adults with a disability and the share of workers receiving SSDI benefits had been rising before that.[17] Moreover, the recession of 1974, which produced an average monthly unemployment rate of 8.5 percent in 1975, also likely played a role.[18] The number of nonworking disabled men rises during economic downturns, and SSDI awards in 1975 increased to 7 percent of workers covered by the program—probably a historical peak up to that point, and a rate that would not be seen again until 2010.[19]
Between 1979 and 2000, the sick and disabled accounted for only 5 percent of the decline in overall male employment—much less pivotal than from 1967 to 1979. But disability continued to play an important role among prime-age men, accounting for 50 percent of the employment decline during the 1980s and 1990s. The number of nonworking disabled men 25 to 54 grew by 91 percent from 1979 to 2000 (or 3.3 percent annually), compared with 129 percent from 1967 to 1979 (7.1 percent per year).[20]
Disability explained just 17 percent of the 2000-to-2019 decline in men’s employment. Even among prime-age men, it explained less than a quarter of the decline. Surprisingly, that was similar to the share explained by home and family responsibilities. The number of prime-age nonworking disabled men rose by only 19 percent during these years, or 0.9 percent per year.
The bottom line is that from 1967 to 2019, rising disability explained 41 percent of the decline in employment among prime-age men. From 1967 to 2000, the increase in disability explained over half (53 percent) of the rise in prime-age male joblessness. It was most consequential before 1980.
Welfare Policy
The Social Security Act not only created public pensions and disability insurance, it enacted what came to be called Aid to Families with Dependent Children (AFDC). Originally, AFDC benefits went to mothers who were widowed or abandoned by their husband. But by 1992, over half of single-mother recipients had never been married, and half of all single mothers received AFDC benefits.[21]
That same year, President George H. W. Bush promised in his State of the Union address to ease the process for states to implement waivers to federal welfare policy. Welfare reform featured prominently in the 1992 presidential campaign as well. State waiver requests spiked, tripling over the previous highs from 1984 and 1989 and remaining elevated over the next two years.[22] Many of these waivers sought to encourage work among AFDC recipients. Congress passed President Bill Clinton’s expansion of the Earned Income Tax Credit in 1993, which also encouraged work by sending money to low-income jobholders. This legislative and administrative activity culminated in the 1996 law that dramatically reformed AFDC, creating a new program in its place, Temporary Assistance for Needy Families.
Labor supply among single mothers had risen by 12 percent from 1967 to 1979 and 5 percent from 1979 to 1989. But from 1989 to 2000, it increased 20 percent, fueled by a jump in the single-mother employment rate from 72 percent in 1992 to 85 percent by 2000. Over those eight years, married mothers’ employment increased by less than 2 points, and the employment rate of single women without children fell by a point. By 2019, the employment rate among single mothers was still 81 percent, over 6 points higher than in any year prior to 1992.[23]
Education Policy?
Federal support for higher education also affected labor supply, by removing many young adults from the workforce (at least during the school year). In this case, though, the impact of policy per se is not always evident.
From 1940 to 2019, school enrollment rose from 7 percent of 20- to 24-year-olds to 39 percent.[24] A variety of education policies—from the GI Bills of World War II and the Korean War, to Pell Grants for low-income students, to the federally subsidized student loan and work-study programs, to state aid to public colleges and universities—have surely played a role in raising educational attainment.
At first, however, it’s surprisingly difficult to find the influence of rising school enrollment on labor supply, at least until the 1980s. Between 1939 and 1959, annual weeks worked and weekly hours both fell among men and women under age 25 (and fell more than they did among Americans in age groups above them). But their employment rates increased (and by more than for older people in their twenties). Moreover, labor supply and employment among young men and women increased from 1959 to 1979. Employment did fall among young men between 1967 and 1979, but no more than it did among older men. Moreover, labor supply rose for them and for young women.
From 1979 to 2000 and 2000 to 2019, however, employment declined among both young men and women, while it declined less among 25- to 29-year-old men and rose among late-20s women. Hours per person rose among young women from 1979 to 2000, but not as much as for those 25- to 29-year-olds.
Despite the ambiguity of these trends, we can say that increased enrollment in secondary and postsecondary education accounted for 16 percent of the fall in men’s employment between 1967 and 1979, 13 percent from 1979 to 2000, and 25 percent from 2000 to 2019. It explained 28 percent of the decline in women’s employment between 2000 and 2019.
Much of the fall in the employment of Americans under age 25—about 40 percent, whether measured from 1967 to 2019 or 2000 to 2019—reflects declining employment rates among 16- and 17-year-olds. Falling employment for this youngest group primarily reflects the demise of summer and after-school jobs. Only 20 percent of kids ages 16 and 17 worked at all in 2019, compared with 55 percent in 1967. This means that a sizable part of the decline in employment among adults under age 25 has nothing to do with rising school enrollment, which went up among 16- and 17-year-olds only from 89 percent to 93 percent over these 52 years.[25]
Moreover, increased school enrollment over the decades is a reflection of workers desiring jobs that are more comfortable, secure, and remunerative. Policy has affected enrollment, but probably at the margin since the 1960s. Instead, the rising demand for higher and rarer skills that is typical of rich nations incentivized Americans to invest in their human capital.
Declining Marriage and Fertility
Changes in marriage and fertility have also altered the time that Americans put in at work. Moreover, these changes have had opposite effects on men and women.
From 1939 to 1959, according to the census data, labor supply rose among men and women alike, with employment increasing and worker hours falling. Annual hours per worker fell much more among women than men: male workers’ annual hours fell by 3 percent, compared with a 20 percent decline for women.
Despite that drop, women’s labor supply rose because employment increased. In a typical week of 1910, around 24 to 25 percent of women age 16 and older were employed. That was also true in 1930, but the rate rose to 34 percent by 1960.[26] This increase occurred despite the greater childbearing and -rearing obligations women faced during the baby boom, which predated World War II and peaked in the late 1950s. Women’s attachment to the labor force increased due to the wartime experience of Rosie the Riveter and her coworkers, but their employment rates would not have risen as much after the war if not for the productivity growth that enabled more to engage in part-time and part-year work.
Employment rates rose between 1939 and 1959 across marital and parental statuses, doubling among married women, with an especially large rise among married mothers. During the 1960s and 1970s, employment and hours per worker both rose among all groups but married men.
The decline in employment among married men from 1959 to 1979 was one of two reasons that employment among men ages 25 to 54 fell during that period. The second reason was that marriage became less common, falling from 84 percent of prime-age men in 1960 (a peak in the census data) to 75 percent in 1980. That left relatively more men in a demographic category (the unmarried) with lower employment rates. For example, in 1979, 96 percent of married prime-age men worked, compared with 86 percent of single men.[27]
In fact, falling marriage rates after 1960 served to push male and female labor supply in different directions, because while single men work less than husbands, single women work more than their married counterparts. In 1979, 63 percent of married women ages 25 to 54 worked, while 79 percent of single women did.
Hours per person rose in the 1960s and 1970s among prime-age women with and without children (79 percent and 23 percent, respectively). The increase was particularly strong among married mothers, whose hours rose by 84 percent, but it was also sizable among childless married women (36 percent). Changes in employment rates were primarily responsible for these trends.
From 1979 to 2000, hours and employment rose among both married and single women, though much more so among wives and mothers. (The employment rate of single childless women actually fell slightly.) Hours and employment also increased due to the fall in the share of women married, from 75 percent to 65 percent. That’s because single women continued to work more than married women, though the gap closed to just 7 percentage points.
Despite falling employment rates during the 1980s and 1990s, labor supply actually increased among both married and single men. Had marriage not fallen, labor supply among men ages 25 to 54 would have risen slightly instead of declining by 8 percent.
Finally, during the 2000s and 2010s, married mothers and childless wives saw labor supply increases but employment declines. Single mothers and single childless women experienced declines in both (including slightly larger employment declines than their married counterparts). Hours per person and per worker and employment all fell for married and single fathers and childless men. Marriage rates once again declined (from 64 percent to 56 percent), pushing men’s labor supply down and women’s up.
A variety of consequences of affluence facilitated increased work among wives and mothers. These included the proliferation of home appliances and prepared foods, the greater purchasing power that allowed more families to send two workers into the workforce despite childcare needs and other work expenses, the rising opportunity costs of withholding an earner from the workforce, and the freedom to plan their careers that the birth control pill offered women.
With greater economic opportunities for women and a safety net of growing generosity, there was less need for women to marry, contributing to reduced marriage rates.[28] The pill and legal abortion facilitated the uncoupling of marriage and sex, further contributing to delayed nuptials. For all these reasons, shotgun marriage also declined dramatically, in part because men seeking to evade responsibility could argue that their partner should have used birth control or could get an abortion.[29] [SW3] As marriage became rarer and less stable as an institution, the incentives for women to be attached to the workforce only grew.
Meanwhile, the opposite dynamic affected men. With wives working more and earning more, there was less pressure on husbands to be employed at all costs all the time. Since single men face fewer responsibilities for others, declining marriage also made it easier to replace earnings with disability benefits, food stamps, or other sources of income. With the demise of the traditional breadwinner role, young men also faced diminished incentives to prepare for a life in which they would be relied on to have a firm attachment to the workforce.[30]
Declines in fertility also reinforced all these dynamics. Delayed and reduced childbearing allowed and encouraged single women to invest in their skills and increase their labor supply. With fewer children, wives could work more and longer. Husbands had fewer mouths to feed and less responsibility, and single men could anticipate fewer obligations to others than in the past, reducing their attachment to work.
Finally, these marriage, fertility, and employment trends reinforced each other. Rising affluence led to greater work among women, which reduced marriage, which reduced work among men, which weakened marriage, which increased work among women, and so on.
Changes in Demand for Labor
What of the view that a deterioration of the economy, a collapse in demand for less-skilled workers, or misguided economic policy has played an important role in reducing male labor supply (and increasing female labor supply by pressuring wives to work more)? The evidence is, for the most part, inconsistent with the declensionist critique. The story for the least skilled male workers is somewhat complicated by disability, but even here we can see the influence of the nation’s affluence.
Comparing Advantaged and Disadvantaged Workers
If the economy were failing the most disadvantaged workers, we might expect to see worse trends for the least-educated workers. For the most part, that’s not what the data indicate.
We can drill down on this question beginning with the 1939-to-1959 period. Among adults between the ages of 25 and 54, labor supply and employment rose across educational attainment levels, for both men and women. Increases were largest among women who lacked a bachelor’s degree but were not especially large for women with no more than a high school education.
Between 1959 and 1979, hours and employment fell among prime-age men with no more than a high school education. But the same was true of more-educated men. Similarly, from 1979 to 2000, the small employment dip among prime-age men with no more than a high school degree was only slightly larger than the decline for college graduates. The increases in hours and work among prime-age women over these 40 years were pervasive regardless of educational attainment.
Labor supply and employment fell between 2000 and 2019 for prime-age men and women of all education levels, except that female college graduates saw increases in both. Labor supply was higher the greater women’s educational attainment, but outcomes did not vary much across men with different education levels.
One of the most notable distinctions of the post-2000 period is that employment, weeks per worker, and hours per week all rose among men and women older than 59 years of age. Does this rising labor supply among older Americans indicate a diminished ability to save for retirement? That seems unlikely too. Among women ages 60 and older, employment rose most among those with a bachelor’s degree and least among women with no more than a high school diploma. Consistent with the declensionist view, the employment rates of older men without a college degree did rise while those of college graduates were flat. However, trend aside, in both 2000 and 2019 employment rates were higher the greater educational attainment—for older men and women alike.
Comparing Wives of Advantaged and Disadvantaged Husbands
Perhaps the increases in labor supply among women before 2000 were a reflection of men’s diminished economic standing? A common critique of the economy over the past several decades is the claim that family incomes rose only because wives were compelled to work more to make up for the declining fortunes of their husbands.[31] If true, we should expect to see employment rise more among women with economically disadvantaged spouses than among those whose husbands have better economic outcomes. There’s not much sign of this pattern in the data either.
From 1939 to 1959, prime-age women with husbands who had no more than a high school education did see a large increase in labor supply (91 percent), but hours increased significantly even among women with college-educated husbands (57 percent). Labor supply and employment rose by similar amounts regardless of whether a husband was employed or not.
Between 1959 and 1979, prime-age married women’s hours again rose regardless of husband education, but the increase was largest among the wives of men with a college degree—double the rise for wives of men with no more than a high school education (123 percent vs. 59 percent). Similarly, the wives of working husbands saw a much larger increase in labor supply than the wives of non-working husbands (70 percent increase versus 18 percent).
The per-worker hours of wives whose husband had no more than a high school education rose by 19 percent from 1979 to 2000, and employment rose 11 percentage points (or 17 percent). Labor supply increased similarly, however, for the wives of college graduates. They saw hours rise 19 percent and employment by 8 percentage points (or 12 percent).
During the 1980s and 1990s, hours per worker rose 15 percent among the wives of non-working men, and employment rates rose by 22 percent. Similarly, wives of men who were out of work for all or part of the previous year despite wanting a job saw their work hours rise by 16 percent and their employment by 11 percent. But wives of men who worked the entire year saw their hours and employment rise by 20 percent and 17 percent.
If we look only at wives with working spouses, those whose husbands had earnings low enough to place them in the bottom fifth of men saw labor supply rise by 24 percent and their employment rise by 9 percent. That compares with 55 percent and 24 percent among wives with husbands in the top fifth. (These trends were a continuation of the pattern from 1967 to 1979.)
Finally, from 2000 to 2019, labor supply fell only for prime-age wives whose husband had no more than a high school education, and it rose the most for the wives of college graduates. Employment fell for wives whose husbands had less than a four-year college education but rose for wives of the highest-educated husbands. Similarly, women with husbands in the top two-fifths of male earnings saw larger increases in labor supply than wives with nonworking husbands or husbands in the bottom fifth. Employment rose 2 percent among women with non-working spouses, fell 3 percent among wives of bottom-fifth earners, and rose 6 percent among wives of top earners. And hours increased more among wives whose husband worked year-round than among those whose husband worked only part of the year for lack of employment. None of these patterns suggest that economic hardship drove the increase in labor supply among prime-age wives.
The Contribution of Difficulty Finding Work to Employment Declines
We can also ask what share of the employment decline in each period was explained by an increase in adults who said they could not find work. From 1967 to 1979, among 25- to 54-year-old men, just 13 percent of the employment drop involved men who cited inability to find a job. That was about the same share accounted for by those enrolled in school.
During the 1980s and 1990s, inability to find work explains only 9 percent of the drop in prime-age male employment. That was less than the portion explained by retirement or even the share taking care of home or family. Even among prime-age men with a high school diploma or less, inability to find work explains only 18 percent of the employment decline from 1979 to 2000.
Inability to find work explained just 14 percent of the decline in prime-age male employment from 2000 to 2019, only slightly more than early retirement did. Among women, inability to find work explained less than 9 percent of the drop.
Over the 52 years from 1967 to 2019, only 12 percent of the decline in employment among men ages 25 to 54 was due to inability to find a job. By contrast, 14 percent was due to an increase in men taking care of home and family, 15 percent was due to rising school enrollment, and 10 percent was due to early retirement. That hardly suggests that a deteriorating economy is behind the long-term decline in male employment.
Disability Benefits Revisited
Earlier I noted that men citing disability account for 41 percent of the decline in prime-age male employment from 1967 to 2019 and that the number of prime-age disabled men who are not working rises when the unemployment rate goes up. That raises the possibility that rising “disability”-related joblessness might hide some of the impact on men of challenging economic circumstances. It may be that because of labor market conditions, men have increasingly opted to receive federal disability benefits and make time for other priorities rather than work at the wages available to them or look for a job.
If so, we might expect the increase in the disabled jobless to be more pronounced among less-skilled men. For these men, disability benefits are larger relative to their potential earnings than is true for men with higher-compensated skills.
On first glance, the evidence from the ASEC on prime-age nonworking disabled men seems unsupportive. Among men with no more than a high school diploma, the proportion who were disabled and didn’t work increased by a factor of 3.8 from 1967 to 2000. That compared with 2.6 among those with a bachelor’s degree, but 6.0 among moderately educated men. (The proportion of least-skilled men who were disabled and didn’t work only rose 7 percent from 2000 to 2019—compared with 30 percent for the most-educated men—so I focus on the earlier period.)
Because so many more nonworking disabled men have no more than a high school education, however, this least-educated group accounts for 68 percent of the overall increase in nonworking disabled men from 1967 to 2000. That despite the fact that the share of prime-age men with no more than a high school education fell from 73 percent to 44 percent over the same period.
If we compare prime-age men in 2000 who had no more than a high school education to those in 1967 who had no more than an 11th grade education, then we’re looking at the bottom 44 percent of men in 2000 and the bottom 38 percent in 1968. In that case, men in the least-educated 40 percent or so account for 74 percent of the rise in disabled nonworkers. Or we can compare men with no more than an 11th grade education in 2000 to those with no more than a 7th grade education in 1968 (both around 10 percent of prime-age men). Men in this least-educated 10 percent account for 25 percent of the rise in disabled nonworkers.
Alternatively, we can say that men in the least-educated 40 percent who are disabled accounted for 37 percent of the overall employment decline among all prime-age men, and those in the least-educated 10 percent accounted for 12 percent of the overall decline. Based on these figures, it is likely that men who are disabled and in the least educated quarter of all prime-age men made up about a quarter of the overall decline in employed men ages 25 to 54 between 1967 and 2000.
What accounts for this increase in nonworking disabled men? First, consider data from the National Health Interview Survey on the share of adults ages 18 to 64 that has an activity limitation due to a chronic condition. That fraction moved with the business cycle between 1967 and 2000, but it mostly fluctuated around a share of about 9 percent.[32] That suggests that the increase in nonworking disabled men is not due to rising disability generally.
In contrast, disabled adults became less likely to work over time. The employment rate among adults with an activity limitation fell from 48 percent in 1967 to 38 percent in 2000 (and was 30 percent in 2018).[33] At the same time, the likelihood that someone with an activity limitation would receive federal disability benefits roughly tripled, so that just over half of those with a limitation were receiving benefits in 2000 (approaching 60 percent in 2018).[34] The share of adults 18 to 64 receiving either SSDI or SSI benefits rose from around 1.5 percent to 5 percent.[35] Consistent with this increase, the share of prime-age men who were disabled and not working increased from 1.6 percent to 5.3 percent.[36]
The evidence on Americans’ changing health status does not suggest that worsening health conditions are behind the increase in disability receipt, which rose among adults in all age groups.[37] Factors increasing disability receipt included mortality declines, the increase in the age at which full Social Security retirement benefits are available (which extends the period during which disability benefits may be necessary), increased numbers of eligible workers, and changes in the composition of the population (such as population aging).[38]
Apart from these changes, disability benefits became easier to access and more generous. The introduction of the SSI program in 1974 and a loosening of SSDI eligibility and review during the 1970s increased benefit receipt.[39] The 1970s also saw SSDI benefit increases. By 1980, they typically replaced 70 percent of someone’s previous earnings. In 1972, SSDI beneficiaries became eligible (with a delay) for Medicare.
Finally, economic hardship could also play a role. Over the years that disability receipt increased, the inflation-adjusted wages of low-earning men generally declined. Consistent data on men’s wages go back to 1973. At their lowest point, both the 10th percentile of hourly compensation among men ages 25 to 54 and the 25th percentile were 12 to 13 percent below 1973 levels.[40] They only recovered in 2001. Disability benefits would have seemed relatively more appealing as pay declined among these groups (and as disability benefits increased in generosity).
It is difficult to disentangle these wage trends from the simultaneously rising attractiveness of disability benefits (and benefits from the Supplemental Nutrition Assistance Program, SNAP, otherwise known as food stamps). The decline in work among those with an activity limitation occurred not only from 1973 to the mid-1990s, when wages at the bottom were falling. It also proceeded from 1969 to 1973, when the real hourly compensation of lower-paid men was likely rising.[41] And it continued from the mid-1990s until 2014, when compensation among the lowest-paid men was increasing.[42]
The economic declensionist interpretation of rising disability receipt is undermined by a number of facts. First is the evidence above showing that inability to find a job has played a minimal role in men’s declining labor supply. Moreover, if nonworking men who, in the past, would have cited inability to find a job have increasingly relied on disability benefits, we would expect to see the number citing disability as a reason for not working to rise as a share of the number citing either disability or lack of work. Instead, that fraction falls during the 1970s and 1980s when comparing similar points in the business cycle, the period during which the wages of the lowest-paid men fell. Elsewhere I have looked at what nonworking men tell surveyors when asked whether they want a job.[43] I concluded that from 1969 to 2014, only about a quarter of the increase in prime-age men who were jobless for a full year was explained by men who wanted a job.
Elsewhere I have argued that the decline in men’s wages may, paradoxically, reflect an affluent economy.[44] One avenue is the increased economic opportunity afforded to women as nations grow richer, and the resulting change in norms around traditional spousal roles. In an earlier era, many men received pay in excess of the value they created on the job due to the widespread norm that a married man should be able to raise a family on one income. That norm came with considerable limitations on women, who often were expected to be homemakers. As more wives came to work longer hours at higher pay, the male-breadwinner norm eroded, and men’s pay stagnated or even declined.
A key data point for this hypothesis is that wage growth among women at the 10th percentile and up outpaced men’s wage growth even at the 90th percentile.[45] Since women’s wages grew robustly during these years, it is difficult to point to a breakdown in the labor market as an explanation for men doing relatively worse than in the past. A second data point is that men’s pay rose more than productivity did for decades during the mid-20th century, suggesting that it was due for a correction.
Note that this explanation is subtly different from the view that men’s pay declined simply due to increased labor competition from women and the wage-dampening effect of increased labor supply. It presumes that many men were, relative to productivity levels, overpaid in the past—that they received economic rents. Still, increased labor supply and labor competition would have added to downward pressure on men’s pay.
In addition, the affluence-caused long-term decline in marriage and fertility discussed in the previous section and the increased labor supply and earnings of wives have also likely reduced men’s wages. With more men single and childless, and for longer, they face less pressure to be breadwinners. That may affect the weight they put on maximizing their wages relative to other considerations such as job security, safety, or convenience. Husbands also face less pressure to the extent that their wives bring home substantial earnings.
Finally, the declensionist argument also ignores a necessary condition for men abandoning work for disability benefits—namely the existence of disability benefits generous enough to substitute for work (when combined with other sources of household income). Having a generous-enough safety net is another consequence of affluence.
To some extent, the growth in the number of nonworking disabled men is analogous to the increase in retired men. In both cases, reduced labor supply is the product of the increasingly generous safety net that affluence makes possible and that allows people to avoid working in pain and discomfort and into old age. By this interpretation, increased receipt of disability benefits reflects our ability as a nation to support sick and injured people who we might not have expected to work if we could have afforded to help them in the past.
Of course, it is also possible that norms around what levels of discomfort and dissatisfaction on the job are reasonable have liberalized. And those who would celebrate the leisure now available to those prime-age men who would have worked through some amount of pain or discomfort in the past should be deeply concerned about the high levels of isolation, unproductiveness, and unhappiness among nonworking men.[46]
Conclusion
The past century of changes in labor supply illustrates the complexity of the economic, social, and political trends that have shaped the contemporary labor market. We live in a nation more affluent than any of its peers (excepting a few unusual cases involving oil reserves and banking centers) and richer than at any time in the past.
America has economic challenges, to be sure. Whether due to slack demand for less-skilled labor, the perverse incentives of our safety net, or a combination of both, we should be very concerned about what Nicholas Eberstadt calls “male worklessness.”[47] Despite earnings and income being at or near all-time highs and poverty being at or near all-time lows, the nation has made little progress increasing the upward mobility of poor children or reducing black-white inequalities.[48] These problems will not be solved by more government transfers, regulation of labor markets, or price controls. They require tougher eligibility standards for disability programs, altering the incentives in other safety net programs through work requirements and time limits, relentless experimentation to determine how to reduce early childhood inequalities, and more innovation in primary and secondary schooling and vocational education.
Economic growth has been uneven across parts of the country, which has helped fuel the populism that has dominated our recent politics. Encouraging local entrepreneurship can help revive activity in communities, while reforms to land use and zoning regulation would help people afford to move to opportunity. Productivity growth—the engine for raising living standards—has been stuck at too-low levels for too long. Shifting government spending away from transferring benefits between Americans and more toward investment in skills, applied research, and infrastructure could boost growth, as could educational reforms focused on building in-demand skills.
These very real challenges aside, the resilience of economic declensionism too often prevents us from seeing that what may look like economic weakness often reflects strength. Many of the broad employment trends described here are common to our affluent peer nations and are likely to have common causes. Where economic growth has had consequences we do not like—say, the weakening of marriage as an institution, the decline of fertility nearly to sub-replacement levels, or the “time crunch” associated with balancing work and family life—the answer is not to rollback living standards or reduce economic opportunity for women or men. Nor is it to blame policymakers or “late-stage capitalism.” Instead, these problems require a willingness to wrestle with the choices we have made.
We face what might be called First World problems—the real, but misunderstood challenges of rich nations. To be sure, there are public policies that could be instituted—or removed—to address them at the margins. But to the extent that many of our challenges reflect the choices everyday Americans have made, they will require nothing less than cultural change rooted in a recognition of long-denied tradeoffs.
Notes
[1] To estimate total hours, I start with estimates from 1948 to 2022 from the Bureau of Labor Statistics Office of Productivity and Technology (“Total U.S. Economy: Hours and Employment” spreadsheet, https://www.bls.gov/productivity/tables/home.htm). I then backcast hours to 1929 using estimates from John W. Kendrick, Postwar Productivity Trends in the United States, 1948-1969(Cambridge, MA: National Bureau of Economic Research, 1973). See Appendix Table A-17. Those estimates are available from 1929 to 1966, and I estimate 1929-1947 figures using the annual proportional change linked to the 1948 estimate from BLS. Finally, I backcast hours to 1890 using estimates from John W. Kendrick, Productivity Trends in the United States (Princeton: Princeton University Press, 1961). See Appendix Table A-X. Those estimates are available from 1890 to 1957. I estimate 1890-1928 figures using the annual proportional change linked to the 1929 estimate backcast as above.
I divide hours by the population aged 16 or older for consistency with the Current Population Survey data described below, which exclude civilians younger than 16 years old before 1967. I obtain population estimates for decennial census years and for 2001-2022 using IPUMS microdata. (See Steven Ruggles, Sarah Flood, Matthew Sobek, Daniel Backman, Annie Chen, Grace Cooper, Stephanie Richards, Renae Rodgers, and Megan Schouweiler. IPUMS USA: Version 15.0 [dataset]. Minneapolis, MN: IPUMS, 2024. https://doi.org/10.18128/D010.V15.0.) I use linear interpolation between census years to estimate population for those intervening years. Data for 1890 is unavailable, so I use linear interpolation between 1880 and 1900.
[2] This series is from microdata for a sample of the decennial census and for the American Community Survey (ACS). See Steven Ruggles, Sarah Flood, Matthew Sobek, Daniel Backman, Annie Chen, Grace Cooper, Stephanie Richards, Renae Rodgers, and Megan Schouweiler. IPUMS USA: Version 15.0 [dataset]. Minneapolis, MN: IPUMS, 2024. https://doi.org/10.18128/D010.V15.0. Prior to 1970, workers at least 14 years old are covered by the census, but I restrict to persons age 16 and older in all years. The census stopped collecting the necessary information after 2000, but the ACS has included it since then. I report census estimates for 1939, 1949, 1959, 1969, 1979, 1989, and 1999, as well as ACS estimates for 2010 and 2022. (While the census asks about the previous calendar year, the ACS asks about the previous 12 months, and respondents are interviewed throughout the year.)
Variables for weeks worked in the previous calendar year that are broken into categories are available in all years, while actual weeks worked are missing from 1960, 1970, and 2010. I estimate weeks worked for 1960 and 1970 by estimating the mean weeks within each category for 1950 and 1980, separately for men and women, and then linearly interpolating mean weeks within each category for 1960 and 1970. I then assign everyone their category mean for those years. I estimate weeks worked for 2010 similarly, interpolating between 2000 and 2022.
Variables for hours worked in the previous week that are broken into categories are available in all years from 1940 to 1990. Actual hours worked the previous week are missing in 1960 and 1970. I impute those via interpolation, as for weeks. From 1980 forward, actual hours usually worked the previous year are available in all years.
The share of people who were employed is just the percentage who worked at least one week during the previous year.
For the 1979 to 2022 estimates, I multiply each person’s weeks worked last year times their usual hours worked per week last year. I then average their hours last year across people (including those who did no work last year). For the 1939 to 1969 estimates, I estimate the share employed last year times average weeks per worker last year times average hours last week. The last of these three terms is conditional on having nonzero hours last week, except that everyone who did not work the previous year is assigned weekly hours of zero. An alternative is to average (weeks worked the previous year times hours worked the previous week) after dropping people who worked in the previous year but had no hours the previous week. Another is to compute this same average conditional on working the previous week. However, the estimates from these alternative calculations did not track as well the 1979 and 1989 estimates using “usual” hours worked last year. I confirmed that the 1939-to-1989 series using my calculation maps nearly seamlessly onto the 1979-to-2022 series using “usual” hours worked.
Note that these estimates include members of the Armed Forces as well as the institutionalized population, though my checks indicated that omitting both groups made little difference.
[3] These estimates are from the Current Population Survey (CPS) Social and Economic Supplement (ASEC). Estimates for 1961 to 1962 and from 1967 to 1974 are from files provided by the Unicon Research Corporation, a company no longer in existence. (See Unicon Research Corporation, “Current Population Surveys, March 1962-2014,” [producer and distributor of CPS Utilities], 2014. Archived website available at https://web.archive.org/web/20150212044706/http://www.unicon.com/.) Estimates for 1963 to 1966 are from files created by Robert Mare and Christopher Winship, available on the website of the National Bureau of Economic Research (NBER). (See https://www.nber.org/research/data/nbers-mare-winship-data-1964-1992.) Estimates for 1975 to 2022 are from files offered by the Minnesota Population Center at the University of Minnesota. (See Sarah Flood, Miriam King, Renae Rodgers, Steven Ruggles, J. Robert Warren, Daniel Backman, Annie Chen, Grace Cooper, Stephanie Richards, Megan Schouweiler, and Michael Westberry. IPUMS CPS: Version 11.0 [dataset]. Minneapolis, MN: IPUMS, 2023. https://doi.org/10.18128/D030.V11.0.) I have determined that the Unicon files are more accurate and complete than the IPUMS files in the earlier years of the ASEC and that the Mare-Winship files are more accurate for 1963 to 1966. I am less confident of the pre-1967 estimates generally, particularly from 1962-66. The Census Bureau only began producing its own public-use files with the 1968 survey. A good overview of the issues is provided in Section 5, Appendix W of Unicon’s Data Dictionary. This used to be publicly available but is not archived. For a copy, see https://www.sugarsync.com/pf/D697698_06273781_150336. The 1962-66 estimates are referenced only in Figure 1.
The estimates for 1961 to 1974 multiply the share of people who worked during the calendar year preceding the survey by the average number of weeks that those employed people worked, and then by the average hours worked by people who were employed and at work in the survey’s reference week (for most respondents, the week containing March 12). The third of these terms is conditional on having worked during the reference week. Everyone who did no work the previous year is assigned zero hours. Ideally, the average hours estimate would be based on the average hours per week the previous calendar year, but no such measure is available in the ASEC until 1976. However, I confirmed that when I calculated such estimates for 1975-2022 (not used in my analyses) they were very similar in level and trend to the 1975-2022 estimates that I do use in my analyses.
Those 1975-2022 estimates are the average across people of (weeks worked last year * usual hours per week worked last year). The estimates are for the civilian noninstitutionalized population of Americans aged 16 and older.
[4] Published estimates of work hours in the late nineteenth and early twentieth century are rough, often applying to employees in firms, often only in one or a few sectors, sometimes confined to fulltime workers, regularly excluding agriculture, and generally excluding the Armed Forces. What constitutes “work” during a period when many families both lived and labored at home, on farms, also complicates matters. My series relies on estimates of the number of workers and the hours they labor in the national economy, as part of efforts to understand national economic output, productivity, and income. See endnote 1.
[5] Hours are from sources in endnote 1. Workers are estimated as follows. For 1947 to 2023, I obtain the number of employed persons age 16 and older from the “Labor Force Statistics from the Current Population Survey” Database on the Bureau of Labor Statistics website (https://www.bls.gov/cps/data.htm). I obtain the 1946 estimate by taking employed persons age 14 and older from the same source, but multiplying it by the 1947 ratio of employed persons age 16 and older to employed persons 14 and older.
Next, I obtain the number of active military personnel. For 1946, I average 12 monthly estimates from US Census Bureau Current Population Reports, Series P-57, No. 60, Table 3. For 1947 through 1957, I use estimates from Simona E. Cociuba, Edward C. Prescott, and Alexander Ueberfeldt, “US Hours at Work,” Economics Letters 169, 87-90 (2018), data file updated October 2021, https://docs.google.com/spreadsheets/d/11Xsewm7otywHSEt1ZnwDk4YafV-d5Te8/edit?gid=1715229130#gid=1715229130. They cite the original source as “monthly issues of the Current Population Reports, P-57, Labor Force, published by the US Bureau of the Census, Table 3 or 1.” These are 12-month averages, except 1951 averages only July through December. Estimates for 1958 through 1999 are also from Cociuba, Prescott, and Ueberfeldt. (Data also available at https://old.datahub.io/dataset/united-stats-military-and-civilian-personnel-by-service-agency-by-state-country-2008-2016.) These estimates are as of September 30 of each year. Estimates for 2000 and 2001 are from https://dwp.dmdc.osd.mil/dwp/app/dod-data-reports/workforce-reports, following Cociuba, Prescott, and Ueberfeldt. Estimates are as of end of third quarter. Finally, estimates for 2002-2023 are from https://dwp.dmdc.osd.mil/dwp/app/dod-data-reports/workforce-reports, following Cociuba, Prescott, and Ueberfeldt. I use September estimates.
Worker estimates for 1929-45 are back-casted from my 1946 estimate using Kendrick (1973). Estimates for 1890-1928 are back-casted from 1929 using Kendrick (1961). The underlying measure for 1890-1945 is “persons engaged” (fulltime-equivalent workers, including active military personnel, plus proprietors and unpaid workers). The assumption is that workers and military personnel would have the same trend over these years as this measure.
Other estimates suggest that workers put in longer hours earlier in the century, so that the drop by 1929 was substantially larger than my figures indicate. See, for instance, Michael Huberman and Chris Minns, “The Times They are Not Changin’: Days and Hours of Work in Old and New Worlds, 1870-2000,” Explorations in Economic History 44 (2007): 538-67, https://personal.lse.ac.uk/minns/huberman_minns_eeh_2007.pdf. These estimates, however, are not reflective of the entire US economy.
[6] For these decompositions, I determine the change in logged annual hours per person, the change in logged workers per person, and the change in logged hours per worker, and I divide the latter two differences by the first (all using natural logs). The difference of logged measures approximates the percentage change in the original measures. Moreover, when a measure, such as hours per person, is the product of two other measures, transforming them all into logs turns it into the sum of the two other measures. That allows one to apportion the change in logs between the change in the logs of each of the two other measures. It is only approximate, as the difference of logged measures approximates the percentage change in the original measure less well the larger the change. Moreover, it is simply an accounting exercise. If, for instance, change in hours per worker affected the change in workers per person, then this approach would overstate the causal importance of changes in workers per person.
[7] See endnote 2 for methodological details.
[8] For this decomposition, rather than using the change in the average of annual hours across workers, I compute the change in (average weeks per year times average hours per week). This ensures that in my decompositions, the change in the log of annual hours is equal to the change in the log of weeks per year plus the change in the log of hours per week.
[9] See endnotes 2 and 3 for methodological details, respectively, of the census and American Community Survey analyses and of the ASEC analyses.
[10] The employment rate in the ASEC for 1979 differs from that in the census because the census includes members of the Armed Forces and the institutionalized.
[11] The earlier reduction in the workday may be traced to the Adamson Act of 1916, which established overtime pay after eight hours for interstate railroad workers.
[12] This estimate of the decline in hours per week divides average annual hours per worker by 52. Assuming that workers are employed 52 weeks affects the level of hours. The trend is very similar to the trend in average weekly hours from the monthly CPS data described in endnote 5, which is not based on an assumption about how many weeks per year workers are employed.
[13] For 1959 (and later years discussed later in this section), the estimate begins with real GDP from the Bureau of Economic Analysis National Income and Product Accounts Table 1.1.6 (http://www.bea.gov/national/nipaweb/Index.asp). Real GDP is divided by hours, from the Bureau of Labor Statistics Office of Productivity and Technology (“Total U.S. Economy: Hours and Employment” spreadsheet, https://www.bls.gov/productivity/tables/home.htm). For 1915, I first backcast real GDP per hour back to 1929 from 1947 (starting with a 1947 estimate from the BEA and BLS sources), using the change in real GDP per hour from Kendrick (1973), Appendix Table A-17. I then compute real GDP per hour for 1915 by obtaining real GDP from Kendrick (1961), Appendix Table A-III and hours from Appendix Table A-X. I adjust the 1915 estimate by backcasting from 1929 real GDP per hour (using the Kendrick, 1973 series just described), using the change in real GDP per hour from 1915 to 1929 from Kendrick (1961).
[14] The Ford Motor Company cut the workday from nine hours to eight in 1914. (See “Gives $10,000,000 to 26,000 Employees,” New York Times, January 6, 1914, p. 1, https://www.nytimes.com/1914/01/06/archives/gives-10000000-to-26000-employes-ford-to-run-automobile-plant-24.html?searchResultPosition=1. In 1926, it adopted the “five-day week.” See “Ford Establishes 5-Day Week After Test; Expects Spur to Labor Will Bring 6-Day Pay,” New York Times, September 26, 1926, p. 1, https://www.nytimes.com/1926/09/26/archives/ford-establishes-a-5day-week-after-test-expects-spur-to-labor-will.html?searchResultPosition=36.
[15] John Maynard Keynes, “Economic Possibilities for Our Grandchildren,” in Essays in Persuasion (New York: W. W. Norton and Co., 1963).
[16] All estimates accounting for men’s employment declines that span the early 1990s exclude the 1992-to-1993 change. In the ASEC data, the number of men reporting disability as the reason for not working increases by 19 percent in 1993 (using the 1994 data). That is unmatched by any increase in disabled men between any other years. Nor do any of the other categories of nonworking men increase unusually between 1992 and 1993 (or any other year). Moreover, the number of disabled men remains elevated thereafter. I suspect that the issue is that beginning with the 1994 survey, the weights used to inflate sample estimates to match national totals were adjusted to account for the undercounting of members of disadvantaged groups in the 1990 census. Using unweighted estimates does not produce the same permanent post-1992 elevation. If I am correct, the post-1992 estimates are more accurate than the earlier ones, but a consequence of better representing the disadvantaged is that the number of nonworking adults with an illness or disability increases, making the earlier estimates incomparable to the later ones. To adjust for this survey change, I look at the changes in employment from 1979 to 1992 and from 1993 to 2000, both of which are internally consistent series.
[17] Estimates for the disabled share of the population are from the National Health Interview Survey (NHIS). See Lynn A. Blewett, Julia A. Rivera Drew, Miriam L. King, Kari C.W. Williams, Annie Chen, Stephanie Richards, and Michael Westberry. IPUMS Health Surveys: National Health Interview Survey, Version 7.3 [dataset]. Minneapolis, MN: IPUMS, 2023. https://doi.org/10.18128/D070.V7.3. I compute the share that had a chronic condition that prevented them from performing a major activity or limited the amount or kind of a major activity they could do. Estimates for the share of covered workers receiving SSDI benefits is from William J. Nelson, Jr., “Disability Trends in the United States: A National and Regional Perspective,” Social Security Bulletin 57(3): 27-41, 1994, Table 2. https://www.ssa.gov/policy/docs/ssb/v57n3/v57n3p27.pdf.
[18] See the Labor Force Statistics Database on the Bureau of Labor Statistics website (https://www.bls.gov/cps/data.htm).
[19] Social Security Administration, “Annual Statistical Supplement to the Social Security Bulletin, 2024,” Table 6.C7, https://www.ssa.gov/policy/docs/statcomps/supplement/2024/index.html. For data, see https://www.ssa.gov/policy/docs/statcomps/supplement/2024/6c.xlsx. SSDI beneficiaries as a share of insured workers was higher in 1975 than in any year going back to 1957. See Nelson, Jr., “Disability Trends in the United States: A National and Regional Perspective.”
[20] Per endnote 16, the 1979-to-2000 period is actually a 20-year period (excluding 1992-1993) rather than a 21-year period.
[21] For the rate of AFDC receipt, I multiply the average monthly number of families receiving AFDC by the share of AFDC families with no father present, and then divide by the number of female-headed families with children. The number of families receiving AFDC comes from US Department of Health and Human Services, Administration for Children and Families, “Statistics,” May 23, 2002, https://web.archive.org/web/20021202123900/http://www.acf.hhs.gov/news/stats/3697.htm. The percentage of AFDC families with no father present is from US House of Representatives, 1996 Green Book, Table 8-28. The number of female-headed families with children is from US Census Bureau, “Historical Poverty Tables: People and Families—1959 to 2019,” Table 4, https://www.census.gov/data/tables/time-series/demo/income-poverty/historical-poverty-people.html.
For the share of single-mother AFDC beneficiaries never married, see United States General Accounting Office, “Families on Welfare, Sharp Rise in Never-Married Women Reflects Societal Trend,” Report to the Chairman, Subcommittee on Human Resources, Committee on Ways and Means, House of Representatives, May 1994, https://www.gao.gov/assets/hehs-94-92.pdf.
[22] Shelly Arsneault, “Welfare Policy Innovation and Diffusion: Section 1115 Waivers and the Federal System,” State and Local Government Review 32(1): 49-60 (Winter 2000).
[23] On welfare reform’s effect on the employment of single mothers, see Scott Winship, “Yes, the ’96 Welfare Reform Helped Reduce Child Poverty,” National Review Online, September 7, 2016, https://www.nationalreview.com/2016/09/welfare-reform-child-poverty-reduced-1996-prwora/.
[24] National Center for Education Statistics, Digest of Education Statistics: 2023, Table 103.20, https://nces.ed.gov/programs/digest/d23/tables/dt23_103.20.asp?current=yes.
[25] Ibid.
[26] My calculations using the IPUMS Online Data Analysis System, https://usa.ipums.org/usa/sda/ (Ruggles et al., 2024). Note that unlike the other calculations in this report, these refer to employment the previous week (the reference week of the census), not the previous year.
[27] Author’s analysis of census data.
[28] US Congress Joint Economic Committee, “The Demise of the Happy Two-Parent Home,” Social Capital Project SCP Report No. 3-20, 2020, https://www.jec.senate.gov/public/index.cfm/republicans/2020/7/the-demise-of-the-happy-two-parent-home.
[29] George A. Akerlof, Janet L. Yellen, and Michael L. Katz, “An Analysis of Out-of-Wedlock Childbearing in the United States,” Quarterly Journal of Economics 111(2): 277-317, 1996.
[30] Richard V. Reeves, Of Boys and Men: Why the Modern Male is Struggling, Why It Matters, and What to Do about It (Washington DC: Brookings Institution, 2022).
[31] Heather Boushey and Kavya Vaghul, “Women Have Made the Difference for Family Economic Security,” Washington Center for Equitable Growth, April 4, 2016, https://equitablegrowth.org/women-have-made-the-difference-for-family-economic-security/.
[32] See Lynn A. Blewett, Julia A. Rivera Drew, Miriam L. King, Kari C.W. Williams, Annie Chen, Stephanie Richards, and Michael Westberry, 2023. For 1963 to 1996, I estimate the share that had a chronic condition that prevented them from performing a major activity or limited the amount or kind of a major activity they could do. That variable is unavailable after 1996, so from 1997 to 2018 I estimate the share that had any activity limitation. The latter series produces higher levels of disability, but the decline in disability from 1997 to 2000 very much resembles the decline from 1993 to 1996 in the earlier series.
[33] Ibid.
[34] I divide my 1967 estimate of the share of adults 18-64 receiving SSDI (see the following end note) by the share of adults in that age range with a chronic condition that prevented performance of a major activity or limited the amount or kind of a major activity that could be done. (See note 31.) I divide my 2000 estimate of the share of adults 18-64 receiving SSDI, SSI, or both by the share in that age range with any activity limitation. (See note 31.) I adjust the denominator to account for the higher disability level produced by the post-1996 variable in the NHIS. I take half the difference between the 1996 and 1997 shares and subtract it from the 2000 share. (I do the same series of calculations for 2018.) Finally, dividing the 2000 estimate by the 1967 estimate produces a ratio of 3.4.
[35] Table 2 of Nelson, Jr., “Disability Trends in the United States: A National and Regional Perspective” shows the share of insured workers that received SSDI benefits for disabled workers for 1967. Table 1 shows the share of the 18- to 64-year-old resident population receiving SSDI benefits for disabled workers, but only back to 1975. The ratio of these two shares in 1975 is 0.63, while for the business cycle peak of 1979, the ratio is 0.75. I multiply the share of insured workers receiving disabled worker benefits in 1967 by 0.66 and obtain an estimate of 1.5 percent. (SSI did not exist in 1967.)
For 2000, I obtain the number of people 18 to 64 years old receiving SSDI benefits for disabled workers, SSI benefits for disabled workers, or both. See Social Security Administration, Annual Statistical Report on the Social Security Disability Insurance Program, 2022, Table 66, 2023, https://www.ssa.gov/policy/docs/statcomps/di_asr/index.html. I divide by the resident population ages 18 to 64 from the Census Bureau intercensal estimates program, https://www2.census.gov/programs-surveys/popest/datasets/2000-2010/intercensal/national/us-est00int-alldata.csv.
Omitted here are veterans’ disability benefits, which would raise these percentages slightly. Roughly 8 percent of veterans of all ages received veteran disability benefits in both 1967 and 2000. (See Sarah K. Burns, Kristen M. Guerrera, David E. Hunter, and Brian Q. Rieksts, “Trends in VBA Disability Compensation Spending,” Institute for Defense Analyses, 2016, Figure 2, https://apps.dtic.mil/sti/trecms/pdf/AD1015430.pdf.) When combined with estimates of the number of veterans to get the number of veterans receiving disability benefits, divided by the resident population of 18- to 64-year-olds, and then added to the share receiving SSDI or SSI, the share receiving federal disability benefits increases by 1-2 percentage points. (For number of veterans see Scott Sigmund Gartner and Hugh Rockoff, “Veterans living in the United States and Puerto Rico, by age: 1865–1999,” Table Ed229 in Historical Statistics of the United States, Earliest Times to the Present: Millennial Edition, edited by Susan B. Carter, Scott Sigmund Gartner, Michael R. Haines, Alan L. Olmstead, Richard Sutch, and Gavin Wright. New York: Cambridge University Press, 2006. http://dx.doi.org/10.1017/ISBN-9780511132971.Ed229-48210.1017/ISBN-9780511132971.Ed229-482.) However, the number of veterans receiving disability benefits is not confined to the 18- to 64-year-old population, so its impact on my estimates would be less what my estimates suggest.
[36] Author’s analyses using NHIS data.
For similar evidence, see Richard V. Burkhauser and Mary C. Daly, The Declining Work and Welfare of People with Disabilities (Washington, DC: AEI Press, 2011).
[37] Scott Winship, “How to Fix Disability Insurance,” National Affairs, 2015, https://www.nationalaffairs.com/publications/detail/how-to-fix-disability-insurance. Scott Winship, “Declining Prime-Age Male Labor Force Participation: Why Demand- and Health-Based Explanations are Inadequate,” Mercatus Working Paper, Mercatus Center at George Mason University, 2017, https://www.mercatus.org/research/working-papers/declining-prime-age-male-labor-force-participation.
[38] Social Security Administration, “Trends in Social Security Disability Insurance,” Briefing Paper No. 2019-01, 2019, https://www.ssa.gov/policy/docs/briefing-papers/bp2019-01.html. Winship, “How to Fix Disability Insurance.”
[39] Winship, “How to Fix Disability Insurance.”
[40] Based on analyses in Scott Winship, “Understanding Trends in Worker Pay over the Past 50 Years,” American Enterprise Institute, 2024. For 1973 to 1978, I use the data from the May supplement to the Current Population Survey, provided by the Unicon Corporation. For 1979 to 1993, I use the data from the Outgoing Rotation Groups file provided by the Unicon Corporation. For 1994 to 2022, I use the Basic Monthly Survey data from IPUMS. See Sarah Flood et al., 2023. Hourly earnings are the reported amount for workers paid by the hour or, for other workers or if otherwise unavailable, usual weekly earnings divided by usual weekly hours. I restrict the sample to workers in the private nonfarm sector (consistently excluding the self-employed across all years). I adjust the estimates for inflation using the implicit price deflator for personal consumption expenditures, from US Bureau of Economic Analysis, “Table 1.1.9. Implicit Price Deflators for Gross Domestic Product,”
April 25, 2024, https://apps.bea.gov/iTable/?reqid=19&step=3&isuri=1&1921=survey&1903=13. I multiply these wage estimates by the ratio of aggregate compensation to aggregate wage and salary income from US Bureau of Economic Analysis, “Table 2.1 Personal Income and Its Disposition,” https://www.bea.gov/itable/national-gdp-and-personal-income.
[41] The 20th percentile of real annual earnings among full-year male workers ages 25 to 54 dipped in 1970 but otherwise increased between 1967 and 1973. Author’s analyses using ASEC data. Using full-year workers brings the annual earnings trend closer to the hourly wage trend, since it controls for differences in weeks worked (though not for hours worked).
[42] Author’s analyses using NHIS data and CPS data (see endnotes 32 and 41).
[43] Scott Winship, “What’s behind Declining Male Labor Force Participation: Fewer Good Jobs, or Fewer Men Seeking Them?” Mercatus Research, Mercatus Center at George Mason University, 2017, https://www.mercatus.org/research/research-papers/whats-behind-declining-male-labor-force-participation.
[44] Scott Winship, “Understanding Trends in Worker Pay over the Past 50 Years” and Scott Winship, “Bringing Home the Bacon: Have Trends in Men’s Pay Weakened the Traditional Family?” American Enterprise Institute, 2022, https://www.aei.org/research-products/report/bringing-home-the-bacon-have-trends-in-mens-pay-weakened-the-traditional-family/.
[45] Ibid.
[46] US Congress Joint Economic Committee, “Inactive, Disconnected, and Ailing: A Portrait of Prime-Age Men Out of the Labor Force,” Social Capital Project SCP Report No. 3-18, 2018, https://www.jec.senate.gov/public/index.cfm/republicans/2018/9/inactive-disconnected-and-ailing-a-portrait-of-prime-age-men-out-of-the-labor-force. Nicholas Eberstadt, Men Without Work: America’s Invisible Crisis (Templeton Press, 2016).
[47] Nicholas Eberstadt, “Education and Men without Work,” National Affairs, 2020, https://www.nationalaffairs.com/publications/detail/education-and-men-without-work.
[48] Scott Winship, “Economic Opportunity and Social Mobility,” National Affairs, 2024, https://www.nationalaffairs.com/publications/detail/economic-opportunity-and-social-mobility. Scott Winship, “The Dual Promise of the American Dream—and Why It’s Worth Defending,” National Review Online, 2023, https://www.nationalreview.com/2023/01/the-dual-promise-of-the-american-dream-and-why-its-worth-defending/.