7
The Labor Market for Older Workers
As discussed in earlier chapters, the population of the United States is aging. When combined with the lower labor force participation of older adults, this aging of the population suggests that the size of the labor force will grow more slowly than that of the population. As a result, the U.S. dependency ratio—the ratio of nonworkers to workers in the economy—will rise. One policy response has been to try to boost the employment rates of older adults, which could lower the dependency ratio, increase tax revenues, and decrease public expenditures on health insurance, retirement benefits, and income support.1
This has led to a number of policy reforms that have the goal of encouraging employment (or increasing the number of work hours) of older adults who would otherwise retire, such as reducing Social Security benefits received upon early retirement at age 62; gradually increasing the age at which adults become eligible for full benefits from age 65 to 67 (Munnell et al., 2004; American Academy of Actuaries, 2002); and changing the taxation of benefits among Social Security enrollees in several ways that include reducing the marginal tax rate on earnings in excess of the earnings cap, increasing the amount of earnings that are exempt from taxation, and broadening the age range that is exempt from the earnings test (Friedberg, 2000). As the possibility of insolvency of the Social Security program approaches, additional changes or reforms to the system may be considered. However, these policy reforms may be less effective than hoped if other factors or changes in the economy reduce labor demand for older workers or reduce the labor supply of older workers.
Thus, this chapter considers the evidence on a large number of influences on the demand and the supply of older workers—mainly on demand-side influences. On the supply side, the chapter considers the impact of the age structure of the workforce, that is, the larger relative cohort size of the older population owing to population aging. On the demand side, the following influences on the labor market for older workers are considered: automation; other technological change; globalization; immigration; and age discrimination against and stereotyping of older workers (considered in more detail in Chapter 6). The chapter also considers hypothesized influences on the balance between supply and demand for older workers, including: the “lump of labor” fallacy; labor mobility and the geography of labor demand; supply-demand mismatch; changes in the demand for and supply of skills; and the business cycle and economic shocks and crises. Figure 7-1 lists the factors this chapter discusses, and identifies them as primarily supply factors, primarily demand factors, or a reflection of both supply and demand (and hence the balance between them).
___________________
1 The dependency ratio is sometimes defined with respect to working ages (15–64) vs. nonworking ages (0–14 and 65+). But, of course, work at older ages is changing, so defining this for fixed ages seems inappropriate, and we think of the dependency ratio as a measure that can decline if the employment of older workers is increased.
Not all the influences on labor supply or labor demand for older workers necessarily lower labor demand for or the labor supply of older workers, although many might. From a policy perspective (see Chapter 8), there may be a good rationale for trying to mitigate the influence of behaviors or economic changes that inhibit labor demand for or labor supply of older workers.2 This may be particularly important because, if current reforms fail to increase employment at older ages, the resulting fiscal strains may compel policy makers to enact even stronger or harsher policies to encourage older adults to delay retirement, including reducing Social Security benefits overall or further increasing the age of eligibility for full benefits. Changes such as these could negatively affect some older workers, specifically those with disabilities or in declining health who are more likely to have difficulty extending their work life into older ages. In other words, supply-side incentives to work, such as reducing Social Security benefits prior to full retirement, may be undermined by demand-side barriers to work due to age discrimination.
More generally, this chapter is about labor supply and labor demand. The standard economist’s view of supply and demand is that they equilibrate to clear markets. There are many reasons why this perspective may be too simple when applied to labor markets, including rising interest in imperfect competition in labor markets (e.g., Boeri and van Ours, 2013) and other sources of friction like the costs of mobility or retraining (Becker, 1964). This chapter focuses on factors that may influence labor supply or labor demand for older workers, often in ways that may lead to lower employment of older individuals. The factors considered range broadly, covering some that may have shorter-term effects, like the impact of the business cycle or other economic crises on older workers, and
___________________
2 Population aging and the fiscal challenges it poses are not the only rationale for policies to reduce demand-side or supply-side barriers to the employment of older workers. Removing these barriers may also simply help older individuals remain employed and productive, with benefits for both the economy and for themselves, including their physical and mental health (see Chapters 3 and 4).
some that may have longer-term effects, like age discrimination, the shifting age structure, and the weak incentives for older workers to retrain or move to stronger labor markets.
We first discuss the supply of older workers, before turning to factors that primarily affect the demand for older workers, although in quite a few cases there can be effects on both sides of the market. We then turn to considerations regarding labor market equilibrium and what they might imply for older workers.
THE LABOR SUPPLY OF OLDER WORKERS
Population aging affects the age structure of the workforce (or the potential workforce, when we consider the decision of whether to work or not). How does a rising share of those at older ages, a relatively larger cohort of older workers, affect older workers’ employment?
Most of the research on this question comes from the perspective of labor supply, studying the effects of the Baby Boom and less-notable variations in the sizes of birth cohorts that generated sizable shifts in the age distribution of the population. Much of this work conducted in the United States has examined the effects of own cohort size on wages (e.g., Macunovich, 1999; Easterlin, 1980; Welch, 1979), and sometimes on employment or unemployment (e.g., Korenman and Neumark, 2000) among youths born into large cohorts first entering the labor market. In general, these studies, as well as those in other countries (such as Morin [2015], for Canada), find that these youths fare worse than those from smaller birth cohorts, at least initially; they earn lower wages and, as a result, have lower employment rates. These effects of a cohort’s relative size are interpreted as “relative supply” or “cohort crowding” effects, in which a large cohort shifts out labor supply, leading to lower wages, and hence reductions in employment or labor force participation (due to the reservation wage effect). This implies that workers in different age cohorts are only imperfectly substitutable for each other, and some of this research (e.g., Morin, 2015) suggests that as the age difference between cohorts increases, the degree of substitutability between them declines.
Despite the generally consistent finding that young workers from large cohorts experience lower wages and employment rates when entering the labor market than those from smaller cohorts, there is reason to expect that relative cohort size will not have similar effects on older workers.3 Several studies have found evidence that these negative effects weaken (Welch, 1979) or dissipate (Wright, 1991) with age. However, these studies are quite dated, and neither focused explicitly on older individuals. Moreover, because earnings-experience profiles flatten by middle age (Heckman et al., 2006), this could lead to a higher degree of substitution between older cohorts and other more-experienced workers, which would reduce the effect of being born into a large birth cohort on wages or labor force participation among older workers.
On the other hand, there are other reasons to expect that relative cohort size may have a sizable impact on labor market outcomes for older workers. Employment rates among individuals in their 50s or 60s are lower than those of adults in their 30s or 40s, and among older workers retirement is quite fluid, as described in Chapter 3. Together, this suggests that, in contrast to workers (especially men) of other ages, older workers may have an elastic labor supply on the extensive margin. If true, the effects of large cohort size on wages—and through wages, on labor force participation or employment—could be sizable for older workers (French and Jones, 2012; Evers et al., 2008). Moreover, if the type of work older workers enter in partial retirement differs substantially from that of their career jobs (i.e., lower-skilled or less-demanding), this could reduce their substitutability with prime-age workers, leading to effects of cohort size on wages among older adults similar to those for young labor market entrants.
Although the relative supply hypothesis about cohort size predicts that there will be negative effects of large relative cohorts on labor force participation and wages, there are two factors that could push in the opposite direction, toward a positive effect. Both of these stem from possible effects relative cohort size could have on labor demand. In particular, we might expect the age structure of the population to affect the composition of consumption and hence labor demand. For example, Reinhardt (2003, Exhibit 1) reports that per-capita health spending for 55–64 year-olds is double that for 25–34 year-olds, and Cohen (2006) documents that the nursing workforce in the United States—which will be in greater demand as the population ages—is itself aging.
___________________
3 Much of this discussion draws from Neumark and Yen (2020), discussed in more detail below.
TABLE 7-1 Percent Retired within Age Group, 2012-2017
| Age Group | Percentage Retired |
|---|---|
| 16–24 | 0.26% |
| 25–49 | 0.81% |
| 50–59 | 6.14% |
| 60–69 | 40.94% |
NOTE: The percentage retired is computed from the employment status question, which captures respondents saying that they are not in the labor force due to retirement.
SOURCE: Data from 2012-2017 March Current Population Survey Annual Social and Economic Supplement Data.
Second, as the proportion of the population that is ages 60–69 increases, the proportion(s) within some other age cohorts must necessarily decrease. If the skills of two age cohorts are substitutable, then when the relative size of one of them declines, the demand for the other may increase. For example, as a large cohort of older workers shifts to partial or bridge retirement, these “post-retirement” workers may move into lower-skilled jobs that were often held by younger workers. In this case, older workers and younger workers would be substitutable, which could increase demand for a large cohort of 60–69 year-olds. Alternatively, if the skills of older workers make them more substitutable for those in the prime/middle-aged cohort, then when the cohort of 60–69 year-olds is large relative to that of this middle cohort, demand for the large 60–69 cohort to remain in peak career positions may increase.
In recent work, Neumark and Yen (2020) explore the effects of age cohort size on the labor force participation and wages of older individuals. They focus on the age group in which labor force participation begins to decline (ages 50–59) and the age range when most people retire (ages 60–69) (see Table 7.1). The latter age group is the age range most likely to be targeted by policy makers interested in incentivizing delayed retirement and working longer, often by introducing reforms to public pension systems (e.g., Gruber and Wise, 2007). Moreover, because labor force participation rates are low at these ages, policy targeted at this age group can have a greater impact on labor force retention (see Chapter 2).4 When estimating the effects of a large older cohort relative to the working-age population as a whole, Neumark and Yen find that when older cohorts are large relative to the younger cohort (ages 16–24), the larger relative older cohort experiences lower labor force participation and wages, consistent with the relative-supply hypothesis. But when the older cohorts are large relative to the middle-range cohort of 25–49 year-olds, older workers experience higher labor force participation, although there is little evidence of a change in wages; results that are more consistent with a relative demand shift.5 As discussed in Chapter 2, over the first half of the twenty-first century, the age distribution of the United States is projected to see rising shares of 50–69 year-olds relative to the adjacent prime/middle-age group, rather than to the particularly small cohort of the youngest workers. As such, the results suggest that this population aging will likely lead to rising labor force participation and employment of older individuals.
FACTORS PRIMARILY AFFECTING THE DEMAND FOR OLDER WORKERS
Automation
On the labor demand size, machines are replacing not only workers performing manual jobs but workers performing jobs that involve cognitive tasks (Autor, Levy, and Murnane, 2003). Recent research has found that humans have a comparative advantage over machines for jobs requiring both hard and soft skills, such as creativity, nonroutineness, and social intelligence (e.g., Frey and Osborne, 2017; Autor et al., 2003).
___________________
4 Neumark and Yen did not include the 50–59 year-olds with 25–49 year-olds, because they found that the behavior of 50–59 year-olds was similar to that of 60–69 year-olds, rather than that of 25–49 year-olds.
5 The demand shift may occur because when prime-age workers are scarce relative to older workers, firms may try to retain or hire older workers. Also, older workers’ extensive-margin labor supply elasticity may be quite high, and older workers often enter into different jobs or employment relationships with more flexible, lower-paying work, which can explain the absence of wage increases.
Other research has documented how automation is replacing mid-skilled workers but not low- or high-skilled ones, resulting in job polarization (Acemoglu, 1999; Katz and Murphy, 1992). Mid-skilled workers—such as those in sales, office, and administrative support, or in craft and repair occupations—perform routine codifiable tasks. Automation has had little impact on low-skilled and service jobs because such jobs require physical dexterity, face-to-face communication, adaptability, and visual recognition. The decline of mid-skilled occupations has reduced options for workers with no more than a high school education, who are concentrated in such jobs.
Lund et al., (2019) found that such displacement may have particularly strong effects on Hispanic and African American workers, whose rates of potential displacement are about 25 percent. Jobs with higher technology displacement include food preparation, retail sales, office clerks, stock clerks and order fillers, bookkeeping, accounting, auditing clerks, cashiers, secretaries and administrative assistants, and food servers. These jobs have large concentrations of Hispanic workers, Black workers, and women. They also have large concentrations of older adults with lower levels of education. Specifically, adults 51–65 are concentrated in bookkeeping, accounting, auditing clerk, secretaries, and administrative assistant occupations with high technology displacement. As such, the effects of automation on labor demand may vary by race, gender, and skill level of middle-aged and older adults.
Displaced older workers are less likely than other displaced workers to be reemployed (Farber, 2015). Training that could improve displaced workers’ hard and soft skills could promote labor market reentry. However, the success of such training programs may depend on whether those in need have access to them, as well as on the feasibility for older adults to make changes in their lines of work. Some older adults may be precluded from taking advantage of training programs due to poor health or having to provide caregiving to other family members that impose additional barriers to change their line of work.
Research documenting the effects of automation on labor demand is still at a very preliminary stage (see, e.g., Acemoglu and Restrepo, 2018, for theory and evidence). Moreover, there appears to be even less research on the dimensions of automation that might replace some cognitive tasks, for example through machine learning and artificial intelligence.
Effects on Retirement
While automation has its direct effects on labor demand, labor supply responses can influence the impact of automation on the labor market for workers in general and older workers in particular. Likely the most important factor is the possibility of workers retraining for the new jobs that automation offers; conversely, automation may have adverse effects on older workers who are unable to modify their skills or do not find the necessary retraining profitable because of their relatively short time horizon in the labor market. This, in turn, can affect retirement plans and the length of older workers’ effective working life.
Some research has been done to assess the effects of automation on retirement decisions. Angrisani and colleagues (2016) find that using computers is associated with a lower likelihood of retirement for middle-aged and older adults. Angrisani and colleagues (2016) additionally find that workers who are open to new experience (personality trait) and who use computers are more likely to remain in the labor force than others. Lee and Angrisani (2020) find that technology complements individuals who have skills at which human workers hold a comparative advantage over machines. At the same time, technology replaces workers with automatable skills. In this last study, Lee and Angrisani created an index of worker skills, focusing not on educational attainment but on worker creativity, ability to handle nonroutine tasks, and social intelligence (i.e., the skills that render individuals more automation-resistant). They find that workers who score well in these three areas are more likely to postpone retirement.
Maestas (2010) documents “unretirement” decisions, that is, decisions to reenter the workforce after retirement. Altogether, more than one in four retirees studied over the 1990s unretire. Automation may affect unretirement decisions, though there is scarce research on this. Lee (2020) measures the level of protection from automation retirees had for their work skills. She explores how this level of protection from automation—determined by skills such as creativity, ability to handle nonroutine tasks, and social intelligence—influenced unretirement decisions. She finds that workers with higher automation-protection are more likely to unretire, and that those who unretire choose jobs whose task-specific automation-protection matches their skills. Aging individuals are highly likely to make unretirement decisions based on their “soft skills.”
Other Technological Change
The technological change that has been most studied by labor economists is the growth of computer technology and its use in the workplace. Much of their research has focused simply on whether this technological change was responsible for the large growth in economic returns to education, in the context of a broader phenomenon referred to as “skill-biased technical change.” (For a review of the evidence, see Katz, 2000.) As discussed in Burstein et al., (2019), the rise in the skill premium was much larger among middle-aged and younger workers than among older workers. Card and Lemieux (2001) attributed this to changes in labor composition, specifically in the relative supply of more-educated workers among different age cohorts. But Burstein, Morales, and Vogel find this has more to do with changes in equipment productivity, in particular because the computer-use differential between college and non-college-educated workers was higher for young and middle-aged workers.
Globalization
The globalization of economic activity, including trade in goods and services as well as outsourcing, can potentially impact older workers in the United States, just as it can impact workers anywhere else in the world. For example, if the industries in which they are concentrated suffer from globalization, their own employment prospects may worsen, in particular when switching industries is costly, as we might expect it to be for older workers. Ashournia (2018) formalizes this in a model of the effects of international trade shocks when mobility is costly; in the model, mobility is costlier for older workers than others because they have fewer years left to acquire the experience that is valuable in the new industry. If outsourcing is skill-biased—with somewhat lower-skilled jobs being outsourced but the highest-skilled jobs remaining domestic—older workers could be disadvantaged if they do not have the requisite skills, or do not (because of their shorter projected work life) have as much incentive to acquire these skills. Ebenstein and colleagues (2014) present evidence that offshoring (or outsourcing, measured as employment of U.S. multinationals at affiliate locations in low-income countries) disproportionately hurt the wages of workers ages 40 and over.
The impact of aging can also run in the opposite direction. For example, Gu and Stoyanov (2019) present evidence that population aging reduces the comparative advantage of countries in industries intensive in the skill-adaptability of workers. They attribute this to lower ability (or perhaps incentive?) to update skills among older workers. They argue that variation in demographic composition across countries is as important for international trade flows as the variation in standard factors of production, such as physical capital and skilled labor. If true, this could provide a strong incentive for governments (or specific industries) to incentivize the upgrading of older workers’ skills.
Immigration
Immigration leads to increases in the supply of labor, which can in turn impact the demand for native workers. The effects of immigrants on wages and on the employment prospects of native workers (or more broadly, workers already here) depends on a number of factors:
- the effects of the product market demands of immigrants;
- the effects on the productivity of native workers who might be either substitutable or complementary with immigrant labor (e.g., Peri, 2016);
- internal migration in response to immigration to particular areas (Borjas, 2006); and
- other effects that change the input mix of native labor or other factors, such as the evidence in Foged and Peri (2016) that low-skilled natives move out of manual occupations in response to low-skilled immigrant inflows, or the evidence in Lewis (2011) that capital investment responds to low-skilled immigration.
While immigration is often viewed, in the public debate, as largely an inflow of unskilled workers, in fact U.S. immigration consists of both less-educated and more-educated labor, from different countries. Peri (2016) argues that a balanced composition of immigrants, coupled with these other responses, has led to relatively small effects of immigration on relative and absolute wages.
However, that is not a settled view, and Borjas (2016) argues that at least some of the research concluding that effects of immigration are modest is flawed by ignoring the internal migration that mitigates the local effects of immigration.
In an earlier paper, Borjas (2003) uses a strategy that does not focus on effects in local labor markets, and concludes that when we look at workers with similar education and experience, immigration inflows do result in lower wages.
These two different conclusions likely reflect, in part, heterogeneous effects among workers, as reflected in the conclusions of a National Academies of Sciences, Engineering, and Medicine (2017) report, which notes that when adverse wage and employment effects are found, they are found for the closest substitutes. This means the effects include lower wages for prior immigrants and native-born high school dropouts, lower employment for prior immigrants, and lower hours for teens (National Academies of Sciences, Engineering, and Medicine, 2017, pp. 5–6). Still, as the National Academies of Sciences, Engineering, and Medicine report notes, researchers continue to disagree about the effects for narrow subgroups.
Given this unsettled debate, it is hard to draw any firm conclusions about the effects of immigration on particular groups of workers. Moreover, there is very little focus on workers differentiated by age. Borjas (2003) is an exception, as he studies the effects of immigration on (male) native workers broken out by education and by labor market experience. He finds that the effects are strongly correlated with age, especially because he uses a “potential experience” measure that estimates the number of years since a person completed school, a measure which is more reliable for men than for women, especially in older cohorts when women worked more intermittently. Borjas similarly classified immigrants based on their education and estimated experience, although this overstated their experience in the U.S. labor market. Using the data this way, Borjas can measure the labor supply shocks of immigrants, differentiated by education and experience. One conclusion of interest is that, for a given level of education, immigrants are more substitutable with natives with similar experience, based on occupational similarities. Using the variation within education and experience cells, Borjas finds a negative relationship between immigrant supply and both earnings and employment (fraction of time worked). And he concludes that it is generally not only the low-educated who drive these results.
If we consider the implications of this work for older workers, the conclusions might be relatively sanguine, because, overall, recent immigration has skewed young (see Borjas, 2003, Figure I); his most recent data are for 2000. At the same time, this pattern varies by education, and in the 2000 data the low-skilled immigration shock is more experienced, suggesting—if we accept Borjas’ conclusions—that this immigration may have adversely affected the earnings and employment of older, less-skilled workers. Overall, though, immigrants remain relatively young, although the share ages 50 and over has increased from about 11–17 percent in recent decades, largely because of family-sponsored migration (Carr and Tienda, 2013).
Age Discrimination
Is age discrimination a significant demand-side barrier to the employment of older workers? As discussed in greater detail in Chapter 6, many types of evidence suggest there is labor market discrimination against older workers, lengthening the durations of their unemployment spells. Moreover, this discrimination leads to more separations of older workers from their current employers, lower employment rates, slower wage growth, and reduced expectations of working past their mid-60s (Adams, 2002; Johnson and Neumark, 1997). Indirect evidence of age discrimination also comes from research showing that stronger age discrimination laws (at the federal and state level) have boosted the employment of older workers (Adams, 2004) and strengthened the employment relationship between older workers and firms (Neumark and Stock, 1999). Moreover, stronger state laws have increased the responsiveness of older workers to increases in the Social Security Full Retirement Age—working longer and claiming Social Security benefits later—pointing to important complementarities between supply-side incentives (Social Security reforms) and reducing demand-side barriers (Neumark and Song, 2013).
The most compelling evidence comes from field experiments, “audit” or “correspondence” studies in which researchers send out applications for fictitious job applicants, experimentally manipulating the ages listed on the applications to isolate the effect of age on hiring (see Chapter 6). Although hiring discrimination may seem less relevant to whether or not older people lengthen their work lives, it may be a critical determinant of whether many older adults are able to transition to part-time or shorter-term “partial retirement” or “bridge jobs” toward the end of their careers (Johnson, 2014; Cahill et al., 2006), or return to work after a period of retirement (Maestas, 2010).
These studies generally find strong evidence of discrimination against older workers in hiring, with more and stronger evidence of age discrimination against older women (Neumark et al., 2019; Farber et al., 2017; Lahey, 2008). However, it is more difficult to determine the source of discrimination against older workers. One source may be
aversion to hiring older workers—much as labor economists model discrimination against, say, minorities. Another source of discrimination against older workers may be stereotypes employers hold about their skills or abilities, or what labor economists might refer to instead as “statistical discrimination,” which in this case means employers’ assumptions about the average skills or abilities of older workers that are applied in part to all older workers.6
While economists are interested in the nature of discriminatory behavior, both forms (statistical and taste) of discrimination are illegal under U.S. law. Equal Employment Opportunity Commission regulations state: “An employer may not base hiring decisions on stereotypes and assumptions about a person’s race, color, religion, sex (including pregnancy), national origin, age (40+), disability or genetic information.”7 From a legal perspective, it is not relevant whether the stereotypes are correct (i.e., on average); however, economists would more likely view these actions from an efficiency perspective and be more concerned about whether these stereotypes are supported by evidence (see Chapter 6 for evidence on stereotypes).
THE BALANCE BETWEEN LABOR SUPPLY AND LABOR DEMAND
The “Lump of Labor” Fallacy
In considering what public policies might encourage or discourage the continued employment of older workers—such as mandatory retirement, retirement policy generally, or policies to reduce age discrimination—it is important to dispel the fallacy that there is a given “lump sum” amount of paid work to be done in the country at any given time. Such an assumption would imply that enabling older people to work longer would damage employment prospects for younger workers. Economists call this the “lump of labor” fallacy. This fallacy has been used to argue against immigration, to argue that automation necessarily reduces jobs, to argue that restricting work hours will lower unemployment, and, most importantly in our context, to argue that we should compel older workers to retire in order to create jobs for younger workers (Kemmerling, 2016). The flaw in this fallacy is ignoring the fact that when more people work, the economy expands, more jobs are created, and so on.8
Gruber and colleagues (2010) list many examples of policy makers in different countries trying to ease older workers out of the labor market to create more jobs for younger workers.9 For example, Kemmerling (2016) shows that in European countries where voters think there is a trade-off between the employment of older and younger workers, voters are more hostile to reforms that lead to longer working lives. Gruber and colleagues (2010) describe results from a set of papers in a volume they edited that addresses this question. They first show that employment rates of older and younger worker covary positively, because they both reflect underlying macroeconomic conditions. These positive relationships weaken but remain with controls for the economic conditions. They then ask the question more specifically in the context of retirement policy, focusing on within-country policy changes that changed retirement ages. The answer is clear. To quote the authors:
There is no evidence that reducing the employment of older persons provides more job opportunities for younger persons. And, there is no evidence that increasing the labor force participation of older persons reduces the job opportunities of younger persons (Gruber et al., 2010, pp. 13–14).10
___________________
6Neumark and colleagues (2019) present a good deal of evidence against their results being driven by statistical discrimination.
7 See http://www1.eeoc.gov//laws/practices/index.cfm?renderforprint=1
8 As Autor (2015) points out, for example: “While intuitively appealing, this idea is demonstrably false. In 1900, for example, 41 percent of the United States workforce was in agriculture. By 2000, that share had fallen to 2 percent, after the Green Revolution revolutionized crop yields. But the employment-to-population ratio rose over the twentieth century as women moved from home to market, and the unemployment rate fluctuated cyclically, with no long-term increase” (pp. 237–238).
9 For example, they quote the French Prime Minister, in 1981: “when it is time to retire, leave the labor force in order to provide jobs for your sons and daughters” (p. 2).
10 Some recent work (Bianchi et al., 2019) finds that an unanticipated pension reform in Italy led to sudden retirement delays, which blocked internal upward progression of other workers in the same firm in which the retirement delay occurred. This work does not address the lump of labor fallacy, because it does not account for any increased demand for goods and services from older workers working longer. Indeed, the paper is presented as providing evidence on how internal labor markets work.
Similarly, Munnell and Wu (2012) review existing evidence and present new evidence for the United States. They conclude that there is no evidence for the “lump of labor” theory.
Labor Mobility and the Geography of Labor Demand
Economic outcomes have always varied across U.S. regions, owing to numerous factors including differences in transportation, natural resources, locations of universities and other sources of concentration of highly-educated workers, and agglomeration economies from early growth, as well as public policy. At the same time, there was evidence of long-term forces moving toward an economic convergence of regions (Barro and Sala-i-Martin, 1991), owing to the migration of labor toward high-income regions and of capital to lower-wage areas.
However, as documented by Couillard and colleagues (2021) and Austin and colleagues (2018), which also draw on earlier work, this convergence has slowed or even reversed, as a result of two forces. First, it appears that with changes in economic structure, the forces that led to job growth and increased labor demand in underperforming areas have weakened. Most notably, they have been weakened by the continuing decline in manufacturing, coupled with strong economic performance in particular regions—especially so-called “superstar cities” (Gyourko et al., 2013)—with concentrations of high-skilled labor, perhaps owing in part to the growth of the knowledge economy (Davis and Dingell, 2019). Second, there has been a decline in migration to high-income regions and less migration generally, perhaps owing in part to rising house price differentials associated with inelastic housing supply in some of the growing regions (Gyourko et al., 2013). As evidence, Foote and colleagues (2019) document that whereas in the past, in response to large negative shocks to local labor markets (mass layoffs), the largest contributor to reductions in the labor force was through out-migration, but during and after the Great Recession growth in nonparticipation in the labor market became the primary source of adjustment.11 The combination of these forces has led to large and now persistent differences in rates of employment and nonemployment across regions.
This literature has tied emerging employment differences across regions to the rising nonemployment (combined unemployment and not in the labor force) of men ages 25–54, the upper bound of which includes older adults. As an example of the variation across regions, Austin, Glaeser, and Summers (2018) note that in 2016 the nonemployment rate of 25–54 year-old men was higher than 35 percent in Flint, Michigan, but was only five percent in Alexandria, Virginia. Austin, Glaeser, and Summers note that major social problems, including high levels of nonemployment, disability, opioid deaths, and increases in mortality, are concentrated together in the states that are east of the Mississippi River and extend south to north from Mississippi to Michigan, excluding the Atlantic coastal states. They argue that these regional disadvantages have calcified over time, which presents policy makers with a choice: either let these differences continue to fester or adopt place-based policies to stimulate economic activity in these depressed areas.12 Their argument marks a break with that of some urban and regional economists (and, in some cases, with their own earlier views) in which migration between these and better-performing areas could address these differences (e.g., Glaeser, 2009).
To the best of our knowledge, this literature has not focused on older workers per se. But to the extent that mobility declines with age after ages 25–29 (see Figure 7-2), it seems likely that these geographic disparities in economic opportunity could particularly harm the employment prospects of older workers. Since older adults are less geographically mobile, this suggests that place-based policies along the lines suggested by Austin, Glaeser, and Summers (2018) might be most successfully targeted at creating jobs for older workers, and this in turn raises the question of what place-based policies would do this. Here a strong cautionary note is in order: Many, if not most, place-based policies have shown limited if any effectiveness at creating jobs (for a review, see Neumark and Simpson (2015). However, in the United States, these efforts have typically been narrowly targeted (especially enterprise zones) and there is continuing debate about whether alternative place-based policies at a broader geographic level may be more effective (Bartik, 2020). At the same time, to the extent that these policies focus on training or retraining workers for new jobs, their effectiveness for older workers may be limited (as discussed more below).
___________________
11 Moreover, touching on a topic this chapter returns to later, among workers ages 55+, mass layoffs lead to increases in county-level disability insurance applications.
12 Place-based policies target incentives or benefits to areas based on the socioeconomic characteristics of those residing in or near the areas, as opposed to “people-based” policies that target them based on individual or family characteristics.
NOTE: These are one-year mobility rates.
SOURCE: Data from U.S. Census Bureau (https://www.census.gov/data/tables/2019/demo/geographic-mobility/cps-2019.html) Table 1.
Mismatch
When there is reallocation of labor demand, either across geographic areas or by skills required by employers, the possibility of “mismatch” arises. Such a mismatch can be between the skills people have and the ones employers want, or between the areas where labor demand is strong and those where workers tend to reside.13 The mismatch that results is referred to as “structural unemployment,” as opposed to the purely cyclical form of unemployment that stems from fluctuations in labor demand and labor supply (depending on one’s perspective on macroeconomic models). This issue has typically become a concern after large economic changes, for example due to the deindustrialization of jobs that became prominent in the 1970s and with the continuing decline in manufacturing (Kollmeyer and Pichler, 2013; Lilien, 1982). Another example is the sharp changes in employment and unemployment, accompanied by the possibility of restricted mobility because of housing price declines, after the Great Recession. Concerning the Great Recession, Pissarides (2013) suggests that structural unemployment emerged in the United States, based on evidence of rising vacancies at a given level of unemployment (or an outward shift in the Beveridge curve). However, Valletta (2013) presents evidence suggesting that “housing lock” did not contribute to long unemployment durations during this period, and Neumark and Valletta (2012) summarize research suggesting that the Great Recession did not create skill mismatch.
As with other areas of research and evidence, research on mismatch to date has not focused on older workers. However, there are indications that problems associated with job loss and reemployment can be more severe for older workers, likely because of the lower geographic mobility documented and discussed earlier and the weaker incentives older workers have to invest in retraining. In fact, these same weak incentives, owing to a shorter period to recoup the gains, may also account for the lower mobility of older workers. For example, Neumark and Button (2014) show that unemployment durations after the Great Recession increased substantially more for older workers, men and women ages 55+. Looking over a longer period (1981–2001), Farber (2004) shows that the reemployment rates of displaced workers (those who lose jobs involuntarily for economic reasons) are substantially lower for workers ages 55 and over, and correspondingly, long-term earnings losses are higher for older workers (ages 51–60, in Davis and von Wachter, 2011), although job displacement rates are lower for older workers.
___________________
13 There is also a large body of work on “spatial mismatch” within urban areas, but this usually refers to mismatch between supply and demand between urban and suburban areas (e.g., Kain, 1968).
A different type of mismatch, which has received less attention from economists, pertains to the preferences of older workers for particular job characteristics, in contrast to the characteristics of the jobs employers offer (see the discussion in Chapter 5). This mismatch may stem in part from the health problems of older workers, for whom flexible work-time and workplace flexibility may be more desirable and may contribute to better health as well (Vanajan et al., 2020). Similarly, allowing for phased rather than complete retirement might attract a continuing labor supply from older workers (Siegenthaler and Brenner, 2001). In this case, employers may be constrained by laws that prevent the offering of fringe benefits only to older part-time workers rather than to all part-time workers, and by laws that limit paying out partial retirement benefits to employees working some hours (Johnson, 2011).
One question is whether government employment services can help match older workers with jobs requiring their skills. For example, the United States has American Job Centers that offer services to job seekers, including training referrals, career counseling, job listings, and similar employment-related services. To the best of our knowledge, there is no research on the effectiveness of these centers nor any material on efforts devoted to older workers, although the CareerOneStop website associated with the program that runs these centers does provide specific advice to older workers,14 and some centers appear to have a focus on older workers.15 The U.S. Department of Labor also has a Senior Community Service Employment Program (SCSEP), which is described as “a community service and work-based job training program for older Americans.”16 This program is authorized by the Older Americans Act and provides training for low-income, unemployed seniors ages 55 and over. SCSEP is discussed in greater detail in Chapter 8. As discussed there, the Department of Labor recently characterized the program as ineffective, but there appears to be limited evaluation of this program, without comparing participants to a control group of nonparticipants (Kogan et al., 2012).17 Moreover, to the best of our knowledge, research on SCSEP has not focused explicitly on the role of reducing mismatch between older workers’ skills and job preferences, on the one hand, and employer demands on the other.
An interesting model is Japan’s Silver Human Resource Centers, which help place retirees in jobs that are part-time and temporary in nature, and typically provide community-based services (Flynn, 2014; Bass and Oka, 1995). We are unaware of evaluations of this program or the existence of evaluations of similar programs in other countries. Developing an inventory of such programs and evaluations of them (if any exist) would be useful, as would an evaluation of the U.S. SCSEP.
Changes in Skill Demands and the Supply of Skills
Technological and market changes often require workers to learn new skills. Lately these have included computerization in white-collar work, computer-aided design/computer-aided manufacturing, robotics, green technologies, and more. As discussed earlier, however, worker retraining is an investment on the part of the worker, entailing both direct training costs (whether paid by workers or employers) and the opportunity cost of lost earnings. Older workers may have less incentive to invest in this retraining, and employers may have less incentive to invest in those workers if their likely remaining work life is relatively short.
However, to some extent lower rates of retraining may be driven by other factors, such as stereotypes about the ability of older workers to learn new skills (Posthuma and Campion, 2009). Both conjectures are consistent with evidence of lower participation in training with increasing age in the European Union and many individual countries (Picchio, 2015). Picchio suggests that some evidence is consistent with older adults taking longer to acquire new skills and having worse training performance (Charness and Czaja, 2006; Kubek et al., 1996). Piccho (2015) further suggests that some types of training programs (self-paced, job-related, and integrated with work; Zwick, 2012) can offset these difficulties, and also that subsidies may be needed to incentivize training for older workers.
___________________
14 See https://www.careeronestop.org/ResourcesFor/OlderWorker/older-worker.aspx
15 See, e.g., https://www.careeronestop.org/LocalHelp/EmploymentAndTraining/find-older-worker-programs.aspx?location=80401&radius=25&persist=true
16 See https://www.dol.gov/agencies/eta/seniors
17 The Workforce Investment Act (now Workforce Innovation and Opportunity Act) offers career counseling, training, etc. However, an earlier report (U.S. Government Accountability Office, 2003) suggested that performance measures for the Workforce Investment Act may have discouraged a focus on older workers (mainly because older workers tended, in part from taking part-time jobs, to be less likely to show earnings gains, leaving SCSEP as, functionally, the main such program for older workers.
There is other evidence on the effects of retraining for older workers who need to retrain. Jacobson and colleagues (2005a) studied the impact of community college on displaced workers in Washington State. Although their definition of “older” is quite broad—age 35 and up18—they estimate similar effects on long-run earnings regardless of age. For older men, a year of community college schooling boosts earnings by about 7 percent, and 10 percent for older women. These estimates are similar for younger men and women, leading the authors to conclude that “at least among those who participate in retraining, older workers acquire new skills about as effectively as do younger workers” (Jacobson et al., 2005a, p. 406). However, underscoring the effects of the opportunity cost of foregone earnings, and the shorter prospective work life,19 Jacobson and colleagues estimate that the internal rate of return to the investment for older displaced workers is about half of what it is for comparable younger displaced workers. This calculation raises legitimate questions about the net social benefit from retraining older displaced workers, although the authors point out that the internal rate of return calculation is more favorable for courses teaching more quantitatively oriented material, or courses in health occupations or the trades.20
There is also evidence on training older workers in the European context, in this case training currently employed workers. Picchio and van Ours (2013) estimate the effects of retraining provided by current employers on the future employment of workers, possibly via retention at the current job, although the authors could not measure that. They find positive effects for prime-aged as well as older workers (aged 26–34, 35–49, and 50–64). Similarly, Dauth and Toomet (2016) study a German program that subsidizes training for workers at least 45 years old employed in small- and medium-sized firms, and find positive impacts on remaining in paid employment in the two-year post-treatment period, with larger impacts for those ages 56 and older. Presumably, the idea of this program is to subsidize employers to train older workers to keep them employable, and the subsidy may be needed because employers might not otherwise see the training of older workers as profitable because of limited future tenure with the firm. In contrast, Schwerdt and colleagues (2012) find no gains from an adult education voucher program in Switzerland, although this program did not target older workers per se. Instead, it was available to those ages 20–60, and take-up (compliance, since it was a randomized controlled trial) was similar for those above age 30. The authors also find that there was crowding-out of firm-financed training.
Impacts of the Business Cycle and Economic Shocks and Crises
Earlier, we discussed the potential for mismatch between older workers’ skills and the skills employers demand, or for mismatch between where jobs are growing and where older workers live. This kind of mismatch can also be generated from adverse business cycle shocks that are followed by reallocation of jobs to other regions, or to other industries with different skill needs. At the same time, recent empirical evidence from the Great Recession suggests that, for that episode at least, the recession did not contribute to skill mismatch (Abraham, 2015); this evidence does not, however, address older workers specifically.
Retirement plans may be disrupted by a late-career job loss. The risk of job loss facing American workers is considerable: from 1990 to 2011, an average of seven percent of all jobs ended each quarter due to employer-initiated separations, representing about nine million layoffs per quarter (Davis and von Wachter, 2011).21 Older workers have historically experienced lower rates of layoff than younger workers, due partly to the protective effect of job tenure, but this advantage has largely evaporated in recent years (Farber, 2015). Compared to younger displaced workers, older workers who experience a job loss take longer to find a new job, experience larger declines in earnings, and are more likely to exit the labor force (Farber, 2015). Older workers who experience a late-career
___________________
18 The older men and women average about age 44 vs. about 29 for the younger men and women, in the Jacobson and colleagues (2005a) study.
19 This argument is laid out formally in Jacobson et al. (2005b).
20Cummins (2015) discusses community college outreach to displaced older workers, emphasizing (based on observations of key informants and other literature) that it is important to reach these workers quickly, to provide them with a realistic assessment of the labor market and how well their skills do or do not match current employer needs (career counseling), and to help with job search. However, this study does not assess effectiveness. Catalfamo (2018) describes a qualitative study of a “Second Career” program at a Canadian college, highlighting both the challenges in balancing school-life demands, but also positive outcomes in terms of skills, confidence, and more.
21Farber (2015) presents measures of job loss based on the Current Population Survey’s Displaced Workers Survey, whereas this estimate is based on the Job Openings and Labor Turnover Survey. The two sources do not provide strictly comparable measures.
job loss are substantially less likely to be employed years after the loss and have lower earnings as compared to similar nondisplaced workers (Couch, 1998; Chan and Stevens, 2001, 1999).
During a labor market downturn, the risk of layoff increases and the probability of finding new work decreases. As a result, retirement transitions rise with the unemployment rate, although this effect is primarily evident among workers who have reached (or nearly reached) the age of eligibility for Social Security benefits (Marmora and Ritter, 2015; Coile and Levine, 2007). Higher unemployment is similarly associated with earlier retirement in Sweden and the United Kingdom (Disney et al., 2015; Hallberg, 2011). U.S. workers who experience a weak labor market in their late 50s and early 60s claim Social Security benefits earlier and have lower retirement income (Coile and Levine, 2011). They also experience higher mortality, which is plausibly due to the reductions in employment, health insurance coverage, and health care utilization in the years following a job loss (Coile et al., 2014).
These factors are likely magnified during the COVID-19 crisis, which led to an economic downturn that was to some extent government-engineered. In the early part of the pandemic, employment rates fell fastest and unemployment rates rose fastest for those ages 65 and older, in contrast to past (more typical) recessions (Bui et al., 2020). This was the result not only of layoffs and retirements from career jobs that occurred during the downturn, but also of the targeted economic shutdowns, which may have made it harder for older workers to find “bridge” or “partial retirement” jobs on the path to permanent retirement, particularly if these jobs were more likely to involve face-to-face contact (Montenovo et al., 2020). Moreover, the higher morbidity and mortality risk that older workers exhibited during the on-going pandemic might give employers pause when they consider hiring or re-hiring older workers, particularly if they may be liable for healthcare costs or work-related infections. As of the time this report was being completed, the end of the COVID-19 pandemic remained elusive, so the full effects of this event on retirement and work among older adults remain to be documented and examined by future researchers.
SUMMARY
We have surveyed an extensive list of factors that can influence the labor supply of older workers, labor demand for older workers, and the balance between them. Many of the findings of existing research on these factors suggest concerns about older workers’ employment as the population ages, but there is also some evidence pointing to more sanguine perspectives. In some cases, the research base needs to be strengthened considerably.
Summary Findings
We summarize the findings from each subsection of this chapter, as follows.
- While a rising share of the population at older ages has typically been viewed as a source of lower employment rates for older workers, the evidence is more nuanced, with some evidence that population aging can increase labor demand for older workers, especially when the older cohort is large relative to the cohort just below it in age.
- We do not know much about the effects of automation on labor demand for older workers. There is some evidence that some types of automation, at least, increase demand for older workers, especially for those who adapt to the new technologies. But other types may have the opposite effect. Most work on the effects of automation is speculative at this stage, with only limited empirical evidence on recent changes.
- While computerization has increased demand for more-skilled workers, this change was less in evidence for older workers.
- Globalization leads to changes in the industrial and occupational structure of employment. The lower mobility of older workers can leave them at a disadvantage when labor demand for the jobs they have occupied declines. In addition, in the United States (and other advanced economies), globalization may increase the premium for learning new skills. Evidence suggests that countries with older workforces tend to lose out, perhaps because of less skill adaptability. On the other hand, globalization could perhaps spur governments to invest more in retraining older workers.
- There is compelling evidence that older workers suffer from age discrimination in hiring, which constrains labor demand for older workers.
- There is rather strong evidence against the “lump of labor” fallacy. Countries that have reformed retirement systems to encourage people to work to later ages have not seen concomitant declines in the employment of younger workers.
- Growing regional disparities in economic outcomes in the United States have left behind rising numbers of people in areas with low employment rates, high unemployment rates, and related social ills, and the role of geographic mobility in reducing these disparities has weakened. These issues are likely quite salient for older individuals, given their low mobility, although research has not focused on older individuals per se.
- Mismatch between workers’ skills and the skills employers demand can arise from economic dislocations, and there is evidence that this mismatch can be worse for older workers, in part because of weaker incentives to invest in training or in moving. There are some government programs in the United States that try to help displaced older workers find new jobs, but their effectiveness has not been evaluated. More work on such programs in other countries, such as Japan, would be useful. Governments may also be able to do more to allow more flexibility in the work arrangements of older workers, to better enable employers to offer working conditions more amenable to older employees.
- Changing the supply of skills among older workers to respond to changes in the composition of jobs is a challenge, because incentives for workers or firms to invest in retraining may be weak. There is evidence that some kinds of retraining—like community college education in the United States and firm training subsidies in Germany—boost earnings and employment, although the social returns may be relatively low if workers are relatively close to retirement.
- Adverse economic shocks tend to result in less job loss for older workers but larger and more persistent earnings declines and more exit from the labor force when these job losses occur, which can spur early retirement. Early evidence from the COVID-19 crisis points to particularly strong adverse outcomes for older workers in contrast to more “normal” recessions, and there are factors unique to the pandemic that may imply adverse longer-run effects as well.
Implications for Research
The above findings, and the gaps in knowledge that this chapter has identified, prompt the following implications (needs) for future research.
- We need more research on the effects of automation on workers generally, and a deeper understanding of the implications for older workers and how policy might respond to mitigate adverse impacts.
- We need more research to understand the effects of globalization on older workers and how globalization might require governments to adapt to encourage employment at older ages.
- Researchers thus far have not figured out how to generate rigorous evidence on age discrimination along dimensions other than hiring; this is a critical challenge for research.
- Research on the labor market effects of immigration on older workers would be useful, although we might be doubtful whether this research will be any more decisive than similar, existing evidence that does not focus on older workers.
- Policies to stimulate labor demand in distressed regions may help all workers in these regions, and particularly less-mobile older workers. But we need to know more about what effect even successful place-based policies (and many have been unsuccessful) have had on older workers.
- More research would be useful to measure the effectiveness of programs in other countries, such as Japan, that help displaced older workers find new jobs.
- We need evaluations of initiatives to encourage the retraining of older workers, including cost/benefit analyses.
- It is important to obtain early evidence on the effects of the COVID-19 crisis on the employment and retirement of older workers.
