what is a very heterogeneous U.S. population (Moen et al., 2021; Calvo et al., 2018; Cahill et al., 2015; Warner and Brown, 2011; Warner et al., 2010).

The chapter begins by reviewing the literature on individual resources that affect work at older ages, beginning with the relationship between health and retirement. As noted in Chapter 1, the committee limited its consideration of the relationship between health and work at older ages to the role of health in determining retirement timing and does not provide a review the effects of work and retirement on health. This is followed by a review of the relationship between education and work at older ages.

Since financial security plays an important role in shaping preferences and expectations for work and constraints on work opportunities at older ages, the next sections focus on contributors to financial security at older ages: financial literacy, employer pensions, Social Security, wealth and income, and debt. The discussion then turns to family and household characteristics, including the roles of caregiving, joint retirement, and family structure. It also covers the role of social capital in finding employment at older ages.

We conclude this chapter by identifying cross-cutting themes, underscoring new scientific findings, and pinpointing gaps in knowledge to shape a research agenda on work and retirement for the 21st century.

INDIVIDUAL LEVEL RESOURCES

Health

Studies using large-scale national surveys of older workers find that healthy older people are more apt to keep working (Cahill et al., 2006; Kim and DeVaney 2005; Henretta et al., 1992), even as doing so appears to have protective health effects (Fitzpatrick and Moore, 2018; Berkman et al., 2014). This makes it difficult to tease out the causal linkage between the two. Still, the evidence shows that the onset of a health condition can trigger labor market exits of older workers (Zajacova et al., 2014; Warner and Brown, 2011), suggesting that health is causally linked to working at older ages.

Many older workers experience health issues that constrain opportunities for working longer. Among individuals ages 55 to 64, seven in 10 have been diagnosed with at least one of six common chronic conditions: diabetes, cardiovascular disease, chronic obstructive pulmonary disease, asthma, cancer, or arthritis—and nearly four in 10 have at least two of these conditions (National Center for Health Statistics, 2009). More than 40 percent of people ages 50 to 64 report difficulty with at least one of nine physical functions because of a health problem, while four percent report difficulty performing activities of daily living (e.g., bathing, dressing, eating, moving from bed to chair, going to the toilet) or instrumental acts of daily living (e.g., preparing meals, shopping for groceries, making telephone calls, assisting with medications; Martin, Freedman et al., 2010).

Large health disparities by race, ethnicity, and socioeconomic status (SES) (National Center for Health Statistics, 2015) result in an unequal incidence of health problems across groups of older workers. These disparities are likely to be compounded by the effects of COVID-19 (see Box 4-1). Black and Hispanic individuals near retirement age have lower baseline health status and experience larger declines in health as they age, when compared to their White counterparts (Sudano and Baker, 2006). Although much has been made of the Hispanic health “paradox”—the lower mortality rates of Hispanic people despite their generally lower SES—it is not clear that this advantage extends to other dimensions of health (Hayward et al., 2014). A life course perspective is invaluable for understanding health disparities by race and ethnicity at older ages, which reflect the cumulative effect of early-life health, social, and economic disadvantages (Brown, 2009).

Health disparities among older Americans by education are also large—the share of college graduates reporting fair or poor health at age 70 is reached by high school graduates by ages 40 to 55 and exceeded by high school dropouts before age 40 (Zajacova et al., 2014). While there are differences in educational attainment by race, research indicates that race and education are both key determinants of health and work (Farmer and Ferraro, 2005). Health deteriorates more rapidly with age for workers engaged in manual work, suggesting that some of the effect of education on later adult health likely occurs through selection into occupation (Case and Deaton, 2005). There are also well-documented disparities in health by gender, with women having longer life expectancies but higher morbidity (Bird and Rieker, 2008). These disparities reflect gender differences in health behaviors and life course

paths, although they may also be partially attributable to biological differences between men and women (Adler and Rehkopf, 2008).

Evidence suggests that health at midlife has been declining in recent years, while mortality has been rising (National Academies of Sciences, Engineering, and Medicine, 2021). The share of men and women ages 55 to 64 reporting themselves to be in fair or poor health rose between the mid-1990s and the mid-2010s, as did the share reporting chronic pain (Hudomiet et al., 2022). There was also an increase in diabetes and overweight status over this period, although smoking decreased. Similarly, there are increases in self-reports of functional limitation, depression, and work limitations over the past two decades (Waidmann et al., 2020). Among non-White Hispanic people, differences in health status across education groups have become more pronounced over time (Case and Deaton, 2017).

These racial-ethnic, gender, and education-based disparities in overall health are manifest in disparities in labor force participation at older ages. While many studies document an association between poor health and retirement (Jason et al., 2017; O’Donnell et al., 2015), the work-nonwork-health interface in later adulthood unfolds over time in complex and unequal ways. Though some types of occupations can reduce activity limitations and improve health—for example, working, as opposed to retirement, delays the onset of cognitive impairment (Carr et al., 2021)—poor health can also constrain opportunities to remain in the labor force. Race and educational disparities in disability pathways, captured in large-scale national surveys that follow older Americans over a period of time (the National Academies, 2018), reveal gender, race, and educational disparities in the timing of exits because of functional limitations (disability), with less-educated Black men in their 50s and early 60s particularly at risk (Moen et al., 2021, 2020; Warner and Brown 2011).

Determining the causal effect of health on retirement is challenging. The first issue is how to measure health. Self-reported health status and work-limiting disability are useful summary measures that are frequently available in survey data. However, these more subjective measures may be subject to justification bias, that is, where people who are out of the labor force characterize their health as poor in order to explain why they are not working (Black et al., 2017). Objective health measures may be subject to reporting error and only imperfectly correlated with work capacity (Baker et al., 2004), and this may vary by occupation or education; using these measures in conjunction with subjective measures may yield clearer results (Bound, 1991). Another major concern is the potential for retirement to affect health (Neuman, 2008). This could create an association between health and retirement that does not reflect the true causal impact of health. Further complicating the relationship between health and retirement timing is the job lock phenomenon, which occurs when an individual continues to work in order to maintain access to health insurance benefits until they reach the age of eligibility for Medicare, regardless of the presence of health conditions (Fisher, Ryan et al., 2016). Individuals in poor health may also retire earlier as a response to their shorter life expectancy.

Using health shocks (unexpected changes in health) to measure the effect of health on retirement may overcome some of these issues by exploiting the arrival of new health information unrelated to the worker’s preferences prior to retirement. Concerns about this approach include the potential for error in measuring changes in health (French and Jones, 2017) and the possibility that workers may respond differently to a temporary vs. a permanent change in health (Blundell et al., 2016). Health shocks are associated with an increase in the probability of retiring, retiring earlier than planned, and revising retirement expectations to anticipate an earlier retirement (Munnell, Webb et al., 2018; McGarry, 2004; McClellan, 1998).

Yet while health is a critical determinant of retirement for some individuals, others exit the labor force for reasons unrelated to health and have substantial work capacity at older ages (Coile et al., 2017; Cutler et al., 2013). The role that health disparities play in explaining the large disparities in work at older ages by race, ethnicity, and education has yet to be fully explored. The fact that there are large disparities in health at older ages across groups, and the fact that health is a critical factor in retirement decisions, together suggest that health inequities may be an important reason why some groups tend to retire earlier than others (Brown, 2009), but further work is needed to determine this definitively.

More work is also needed to explore how the opioid crisis is affecting work at older ages. In theory, if used properly, opioids could enable workers to manage pain and remain in the work force, but they could also interfere with the ability to hold down a job if abused. Employment rates have dropped more in counties with higher numbers of opioid prescriptions, though this fact does not establish the direction of causality (Krueger, 2017). Currie and colleagues (2019) find that prescriptions for working-age people—as predicted based on the level of prescriptions to elderly people in the same area—have a small positive effect on employment for women and no effect for men.

___________________

¹ This section focuses on research on the U.S. Social Security program. Social Security programs in other countries often have different eligibility requirements and therefore likely have different effects on retirement and work.

TABLE 4-1 Average Social Security Wealth at Age 51–56 by Quintile of Wealth Within Race/Ethnicity, 2016 Dollars

SOURCE: Data from Hou and Sanzenbacher, 2020.

may perform tasks that can be easily automated, which forces their early involuntary withdrawal from the labor force (Acemoglu, 1999; Katz and Murphy, 1992). Displacement by automation of mid-skilled jobs has particularly affected Hispanic and Black workers (Lund et al., 2019). The decline of mid-skilled occupations has reduced options for workers with no more than a high school education. This, in turn, has led to lower life-time earnings and, therefore, lower Social Security benefits for such workers.

Due to large disparities in mortality by SES, lower-income individuals also experience shorter life expectancy than higher-income ones (e.g., Deaton and Paxson, 2001; Attanasio and Hoynes, 2000). As a result, such individuals receive Social Security benefits for a shorter period of time than higher-income individuals (Smith et al., 2003). With the growing gap in life expectancy by lifetime income (Chetty et al., 2016), higher-income individuals are increasingly receiving more in Social Security benefits over their lifetimes (the National Academies, 2015).

Social Security income and wealth² have differed by race and ethnicity. Black and Hispanic people historically have had lower average earnings than non-Hispanic White people, and hence they have received lower Social Security benefits. The redistributive component of Social Security has meant that, in 2016, the average Social Security replacement rate (that is, the portion of previous annual wages that Social Security replaced) was 45 percent for Black and Hispanic people and 36 percent for non-Hispanic White people (Munnell, Hou et al., 2018). Still, in 2016 among persons 51 to 56 years of age, average Social Security wealth of non-Hispanic White households was higher at all levels of wealth (Table 4-1) (Hou and Sanzenbacher, 2020).

Black, Hispanic, and women workers close to retirement age have had higher odds of experiencing involuntary job separation and subsequent withdrawal from the labor force and are more likely to have discontinuous work histories, as well as to be in poor health, leading to involuntary job loss in the years prior to retirement (Flippen, 2005; Flippen and Tienda, 2000). This history of job loss, in turn, can lead to lower lifetime earnings and fewer qualifying years of employment, resulting in lower levels of Social Security benefits, and in some cases, a failure to qualify for benefits. Time spent out of the labor force or with low earnings directly reduces the value of and eligibility for Social Security benefits, because these benefits are based on an individual’s highest 35 years of earnings and require individuals to contribute to the program for a minimum of about 10 years to qualify (Hendley and Bilimoria, 1999).

Among Hispanic workers, the foreign-born have lower levels of Social Security benefits, often because they have more truncated labor histories (Johnson et al., 2013). Some immigrants return to their country of origin before achieving Social Security eligibility, or even acquiring an authorized Social Security number, and hence do not collect Social Security benefits (Aguila and Vega, 2017). Most immigrants returning to their country of origin who qualify for Social Security benefits in the United States cannot expect to receive these benefits unless their country of origin has a totalization agreement with the United States that makes Social Security contributions portable between countries (Aguila and Zissimopoulos, 2013). Among male Mexican workers who migrated to the United States and later returned to Mexico, 32.0 percent reported making U.S. Social Security contributions; however, only 5.1 percent received or expected to receive U.S. Social Security benefits in the future (Aguila and Vega, 2017).

___________________

² Social Security wealth is defined as the present-day value of the stream of future benefits the household can expect to receive over time.

TABLE 4-3 Changes in Median Income by Race and Ethnic Origin and Income Percentile, 1970 to 2016

^* Table indicates, for example, that, in 2016, Blacks at the 90th income percentile for all Blacks received 68 percent of the income that Whites at the 90th percentile for all Whites received.

SOURCE: Data from Kochhar and Cilluffo, 2018.

a disincentive for retirement savings or wealth accumulation: the main safety net programs make ineligible individuals with assets above a given threshold, reducing incentives for low-income earners to save or accumulate above that level (Yoong et al., 2019).

Within the United States there are large disparities in wealth accumulation by race, ethnicity, and gender. Longitudinal analyses from the Health and Retirement Study (1992–2010; Brown, 2016) revealed wealth accumulation patterns among non-Hispanic White people that were consistent with the life cycle hypothesis of Modigliani (1975), which proposes that individuals smooth their consumption over the life course, borrowing during periods in which income is lower and saving when income is higher. Non-Hispanic White people increased their savings at a higher rate during the period approaching retirement, when lifetime income trajectories peak, and then began to spend down their wealth in retirement; however, this theoretical perspective was not empirically supported by the wealth accumulation behaviors of racial and ethnic minorities. Among Mexican Americans, wealth accumulation peaked earlier, around age 60—several years before retirement—before beginning to decline, while non-Hispanic Black people experienced little change in wealth throughout the life course.

These racial-ethnic disparities in wealth may be better explained by the cumulative impact of economic (dis) advantages experienced throughout the life course by racial-ethnic minorities due to the effects of systemic discrimination and resource hoarding by racial-ethnic majorities (Rothstein, 2017). Non-Hispanic White women and women of color have had similar disparities in wealth accumulation compared to men (Brown, 2012).

These disparities in income and wealth are important for understanding financial security in old age, because financial security is highly correlated to income and wealth during working life. It is more tenuous for workers without access to employer-sponsored pensions, savings in the financial sector, or other sources of retirement income. Munnell, Hou, and colleagues (2018) found that, among U.S. households headed by persons 30–59 years of age, the proportion “at risk” for financial insecurity in old age increased from 44 percent in 2007 to 50 percent in 2016.

Hispanic and Black workers in 2016 were more at risk for financial insecurity in old age than non-Hispanic White workers (Table 4-4); in part, because Hispanic and Black people are less likely to participate in an employer’s retirement plan (Butrica and Johnson, 2010) and more likely to lack bank accounts (Blanco et al., 2018). Most of the difference in unbanked rates between non-Hispanic White, non-Hispanic Black, and Hispanic workers are accounted for by lower individual and neighborhood SES (non-Hispanic Black workers) and language barriers (Hispanic workers; Blanco et al., 2018). The effect of these differences on financial security at older ages is compounded by the other disparities noted above, namely, Black and Hispanic people receiving lower

TABLE 4-4 Indicators of Possible Financial Insecurity in Old Age

SOURCES: Data from Munnell, Hou et al., 2018; Federal Deposit Insurance Corporation, 2016.

lifetime Social Security benefits (Yoong et al., 2019), being less likely to invest in high-risk and high yield assets (e.g., stocks), and being more likely to invest in real estate than non-Hispanic White people, who have more diversified investment portfolios (Yoong et al., 2019).

Yoong et al. (2019) found that non-Hispanic White workers have Social Security wealth 40 percent higher than that of Black and Hispanic workers. They also found that Black and Hispanic people were less likely to invest in high-risk and high-yield assets (e.g., stocks) and more likely to invest in housing or other real estate than non-Hispanic White people who have more diversified investment portfolios.

Rising Debt

Debt is rising among older adults and some evidence suggests it is an important contributor to working longer (Butrica and Karamcheva, 2018, 2013; Mann, 2011; Belkar et al., 2007; Lahey et al., 2006). In 1989, approximately 38 percent of adults ages 65+ had debt, and this figure increased to 61 percent in 2016 (Congressional Research Service, 2019). At the same time, the real average debt increased from $29,916 to $86,767 among this age group. Pre-retirees (those ages 55+) similarly experienced an increase in the percentage of adults with debt, though more recently, their average debt load has declined. The percentage of adults ages 55+ who had debt increased from approximately half (54%) to 68 percent between 1992 and 2016 (Copeland, 2018). However, between 2010 and 2016, the average size of their debt decreased from $82,968 to $76,679. This debt is largely from mortgage-related debt and an increase in student loan debt (Congressional Research Service, 2019; Zinshteyn, 2019; Copeland, 2018).

Longitudinal analyses from the Survey of Consumer Finances³ (1989–2016) reveal that indebted older adults are more likely to work, less likely to be retired, and, on average, expect to work longer than those with less debt (Butrica and Karamcheva, 2019). These findings are similar to analyses of the Health and Retirement Study (1992–2008),⁴ which found that those with higher debt in later life were more likely to remain attached to the labor force (Mann, 2011). Among older women, mortgage debt was positively associated with a higher propensity to be currently working and they also had a higher expectation that they would be working at age 65 (Lusardi and Mitchell, 2016). Debt appears to be a more important predictor of remaining in the labor force for women and women with young children, than it is for men (Belkar et al., 2007). While Black and Hispanic workers ages 75+ had higher levels of debt overall compared to White workers, there is little evidence that this disparity leads to racial-ethnic differences in the association between labor force attachment and debt (Mann, 2011). The relatively small literature on debt and labor force participation generally uses race and ethnicity as a covariate rather than a predictor of interest. Research that examines cohort differences, race and ethnicity, gender, education, and type of debt is warranted.

___________________

³ The Survey of Consumer Finances is a triennial cross-sectional survey of U.S. families sponsored by the Federal Reserve Board in cooperation with the Department of the Treasury. It includes information about families’ balance sheets, pensions, income, and demographic characteristics. Though the survey is cross-sectional, there were two periods during which longitudinal data were collected. The 1983 panel of respondents were reinterviewed in 1986 and 1989, while the 2007 panel was reinterviewed in 2009.

⁴ The Health and Retirement Survey is a longitudinal survey of older adults (ages 51 and over) conducted biennially since 1992 by the University of Michigan.

___________________

⁵ Health insurance coverage rates for adults generally exclude 18-year-olds because eligibility for children’s coverage through Medicaid and state health insurance programs extends to those under age 19.

life course—women, minorities, immigrants, those with less education—have less control or choice over transitions into or out of the workforce or between jobs (Choi et al., 2016; Zajacova et al., 2014; Warner and Brown, 2011; Brown and Warner, 2008; Willson, 2003; Choi, 2000). Thus, pathways to work that provide meaning and purpose are also constrained for less advantaged workers at older ages.

FAMILY AND HOUSEHOLD STRUCTURE

Family roles and responsibilities, as well as evolving household structures, shape work and retirement trajectories at older ages. This section reviews how unpaid work in the form of informal caregiving constrains opportunities for, and choices to, engage in paid work in later adulthood, particularly for women, racial and ethnic minorities, and individuals with limited education. We then describe and analyze complex processes of joint retirement decisions. We end this section by reviewing how family and household structures have evolved significantly over time, shifting from marriage as a normative and universal practice in the 20th century to the current diversity of family types in the 21st century, and discussing the implications of these changes for work and retirement research.

Informal Caregiving

According to a recent survey of caregivers in the United States conducted by AARP (2020), approximately 53 million adults in the United States provide unpaid care for others, and six out of 10 of these caregivers are also employed. Over a third (35%) of caregivers are ages 50–64 AARP (2020); however, this study did not separately examine the experiences of these older caregivers. Informal caregivers assist others with activities of daily living and are not paid for providing their assistance. Informal caregiving is highly gendered; the majority of informal caregivers, regardless of age, are women. These caregivers are most likely to assist one recipient (often a relative), though approximately 18 percent provide care to multiple recipients.

Informal caregivers who also are employed are engaged in a variety of employment contexts (AARP, 2020). Slightly over half (54%) are paid hourly in their main job, while nearly four out of 10 report being salaried. A smaller percentage report that they are self-employed or own their own business. Men who provide care are more likely than women to be salaried. White, Black, and Hispanic informal caregivers are more often employed in hourly paid jobs, while Asian American informal care providers are more likely to be salaried. Long-distance caregivers, caregivers to a spouse, and high or medium-intensity care providers are more likely to indicate that their supervisor is aware of their caregiving responsibilities. Among individuals who are not self-employed, approximately half of informal caregivers reported their employer offers paid sick days (58%), flexible work hours (56%), and unpaid family leave (53%), while only a quarter reported programs to help with caregiving responsibilities (26%; information, referral, employee assistance programs) or telecommuting (25%). In 2020, White caregivers report more paid family leave, while Black caregivers are more likely to have paid sick days. Since 2015, more employers have provided paid sick days and paid family leave, possibly in response to the growing attention to caregiving in the United States (AARP, 2020).

Although there is great diversity of informal care work and paid work arrangements, longitudinal studies in both the United States and abroad suggest that informal older caregivers are more likely to reduce their hours at work, be forced into retirement, retire earlier, and they are less likely to return to work after retirement (Fahle and McGarry, 2018; Gonzales et al., 2017; Szinovacz and Davey, 2005; Dentinger and Clarkberg, 2002; Pavalko and Artis, 1997). When compared to older men, older women were more likely to permanently leave the labor market, more likely to be working part-time due to informal care, and more likely to take time off of work each week due to the demands of care, while controlling for occupations and industries that are gendered, job tenure, and union membership (Smith et al., 2020). As the intensity of providing care increases, the odds of returning to work decrease (Gonzales et al., 2017). Taken together, these findings suggest that the roles of “worker” and “caregiver” are not complementary, but compete with each, forcing many caregivers to leave the labor force (Fabius et al., 2020; Smith et al., 2020; Kamakura, 2009; Southerton and Tomlinson, 2005; Moen et al., 1995).

Contrary to the studies above, there is a small body of research suggesting that informal care is associated with delaying retirement or going back to work after retirement. Caregiving is costly. Informal caregivers spend

approximately $7,000 per year on out-of-pocket expenses related to care (Rainville et al., 2016). These out-of-pocket expenses are felt more acutely among low-income families, Hispanic people, long-distance caregivers, those caring for an adult who is age 50 or over, and when the care recipient has dementia. Because older Black caregivers are more likely than older White caregivers to provide care in excess of 40 hours per week, care for an older adult with dementia, and live below the federal poverty limit (Fabius et al., 2020), these costs would have a disproportionate impact on them. A cross-sectional study found that among caregivers considered to have “high financial strain,” 24 percent had put off retirement or decided to never retire, compared to only 12 percent and four percent with moderate or low financial strain, respectively (AARP, 2020). Similarly, qualitative research on grandparents who gain full custody of their grandchildren finds that these grandparents often report that they returned to work and/or delayed retirement as a result of these custody arrangements (Gonzales et al., 2019; Hayslip, et al., 2009).

Nearly one in 10 caregivers (of all ages) identifies as lesbian, gay, bisexual, or transgender (AARP, 2017); sexual and gender minorities are more likely to become caregivers than the general population, are more likely to be caring for another individual who identifies as a sexual or gender minority, are less likely to have support from their biological family members, report higher levels of workplace discrimination due to caregiving, and also encounter fewer culturally competent support services (AARP, 2020, 2017; Meyer et al., 2018; Knauer, 2016).

In sum, informal caregiving is associated with earlier retirement and a lower probability of returning to the labor force after retirement. However, a small body of literature suggests that the costs of providing such care may also lead some caregivers to extend employment at older ages. More research is needed to fully understand the heterogeneity of caregiving experiences among older informal caregivers, the ways in which caregiving contexts shape and constrain work opportunities and the quality of work available to caregivers, and work characteristics that enable or constrain caregivers’ ability to engage in both informal caregiving and paid employment. Intervention studies examining the effects of flexible work arrangements as well as state and federal policies and practices on paid employment among caregivers are also warranted.

Joint Retirement

Among two-earner couples, the retirement behavior of each partner may influence that of the other. Nearly one-third of couples retire jointly, or within 12 months of each other (Ho and Raymo, 2009). There are several reasons why this could occur. First, a couple may face a shared financial situation, so the need to work longer (or ability to retire early) will be common to both members. Second, assortative mating may occur. The tendency of people with similar levels of education to marry each other is well documented (Domingue et al., 2014), and the similarities between partners could extend to characteristics like retirement preferences. Finally, complementarity of leisure—the notion that leisure time is more valuable when spent with one’s spouse—may be important.

Two types of studies have examined couples’ retirement behavior. The first estimates structural models of retirement in which each partner is an independent decision maker facing shared finances (see Casanova, 2010, for a review). Several of these studies conclude that complementarity of leisure is critically important in explaining joint retirement (Gustman and Steinmeier, 2004; Maestas, 2001). A second set of studies examines whether public and private pension provisions affecting one spouse have spillover effects onto the retirement decision of the other spouse, indicating that one partner’s retirement behavior influences the other’s. Studies in both the United States and the United Kingdom find evidence of such spillover effects and attribute them primarily to complementarity of leisure (Banks et al., 2007; Coile, 2004).

In sum, there is some tendency for both members of married/partnered couples to retire together. This could arise for several reasons in an economic model: (1) assortative mating; (2) shared household budget constraints; (3) complementarity of leisure. Previous work has established the importance of the last factor. Changing marriage/divorce patterns can affect work at older ages.

Household Structure

Marriage was a normative and nearly universal experience of U.S. adults in the 20th century (Carr, 2020). Household structure has become far more diverse in the 21st century due to a combination of extended longevity;

disparities in longevity along the lines of gender, race, and ethnicity; and changes in personal and structural constraints. Older women, Black workers, Hispanic workers, and individuals of lower SES are less likely engage in marriage or remained married in later life. Older women are more likely to be single than older men, partly because wives tend to outlive their husbands and are less likely to remarry in later life (U.S. Census Bureau, 2015; England and McClintock, 2009). White people are more likely to be married than Black people, partly due to the destabilizing effects of systemic racism such as the War on Drugs, mass incarceration, and reverberating effects of the Great Recession (Carr, 2020; Hayslip et al., 2019). Women of color are more likely to live alone. Being single, whether through divorce, never marrying, or becoming widowed, means an individual cannot pool resources to cover costs with a spouse or partner, and may not benefit from social insurance programs that privilege marriage (such as Social Security). Living alone in mid- and later-life increases the likelihood of experiencing acute financial strain to cover basic costs of living such as rent/mortgage, student loans, credit card debt, food, medicine, and transportation.

Household structures thus may influence the timing of retirement. In the United States, single women (never married, widowed, divorced) reported lower levels of retirement confidence when compared to married women (Copeland, 2020). In addition to working longer for financial reasons, single workers may lack meaningful relationships at home and derive greater benefit from the social connections with work colleagues (Smeaton and McKay, 2003). A cross-sectional study from the Netherlands examined retirement preferences between single and married workers (Eismann et al., 2019). They classified spouses into three categories: “pulling” (spouses who prefer the worker to retire), “neutral” (spouses who did not have a preference for the worker’s retirement age) or “pushing” (spouses who wanted the worker to remain in the labor force). Results indicate that single workers preferred to work longer when compared to workers with a “pulling” spouse but preferred earlier retirement when compared to workers with “pushing” spouses. There was no statistical difference between single workers and workers with a neutral spouse. Research is needed to understand the push and pull factors to work and retirement by household structure given this emerging diversity by gender, race, ethnicity, and SES in the 21st century.

SOCIAL CAPITAL

Acquiring, maintaining, and advancing one’s career is a social process (Smith, 2017; Granovetter, 1995, 1973). Yet research establishing the mechanisms through which social networks, and social capital embedded within them, play a significant role for acquiring employment is mixed (Gayen et al., 2019; Smith, 2017). Scant research has focused on the social processes of older job seekers (Gayen et al., 2019) and even less on older racial and ethnic minorities and low-income older adults. There is evidence that suggests that without social networks, the odds of acquiring employment are reduced, particularly among low-income job seekers (Fernandez and Fernandez-Mateo, 2006; Newman, 1999; Newman and Lennon, 1995) although there are ongoing debates about the causal nature of this relationship (Smith, 2017, 2010; Mouw, 2003).

There are many definitions and dimensions to social networks and social capital (Son et al., 2008; van Deth, 2003; Portes, 1998; Coleman, 1988). Social support, for example, is a type of social capital that helps one “get by or cope” (Briggs, 1998, p. 178) while social leverage is a form of social capital that enables individuals “to get ahead or change one’s opportunity set through access to job information . . . or a recommendation for a scholarship or loan” (Briggs, 1998, p. 178). Social support and leverage may offer different types of support among low-income populations. Free or cheap childcare by friends and relatives to assist a job seeker may be just as important as job information and personal referrals (Dominguez and Watkins, 2003). Thus, Smith (2017) argues both of these social relations may be equally important to formal labor force participation.

Weak versus strong social ties may also play an important role when searching for employment (Granovetter, 1995, 1981; Marsden and Campbell, 1984). From the perspective of social stratification, individuals within homogenous social networks tend to have strong social ties but redundant information about economic opportunities, thereby limiting new information concerning job opportunities; whereas weak social ties are a product of heterogeneous social networks that promote nonredundant information that may foster new information concerning jobs (Granovetter, 1981). There is some evidence to support Granovetter’s (1983, 1973) argument that weak social ties—friend’s friends or acquaintances—are more important than strong social ties (Hällsten et al., 2017; Snijders et al., 2010; Yakubovich, 2005; Calvo-Armengol and Jackson, 2004).

Yet, Gayen and colleagues (2019) found that older adults transitioning out of unemployment did so by accessing strong social ties, as opposed to weak ones. Granovetter (1995) later suggested that individuals from low SES may not benefit from weak social ties when compared to those from high SES because friends of friends within poor social networks may not have information and access to job opportunities. Others argue that there is a ceiling effect among high-status job seekers (Lin and Dumin, 1986; Lin et al., 1981) and that low-status job seekers may benefit more from weak social ties. The evidence supporting these perspectives is mixed (Smith, 2017).

Stemming from ecological systems theory, social theorists have expanded the lens of studying social networks to the role and function of neighborhood organizations and government. Neighborhood organizations, it is argued, are important to social networking and social capital (Smith, 2017; Small, 2010; Putnam, 2000). Membership to civic and noncivic organizations within neighborhoods foster social ties, and over time, trust, reciprocity, and norms help to build social capital. Policies from the state, federal, and supraorganizational bodies of authority, shape networks and social capital (Small, 2010). This perspective underscores the importance of building the capacity of institutions to facilitate meaningful productive roles for older adults, an important concept within the productive aging literature (Morrow-Howell et al., 2001).

Civic organizations at the national, state, and local levels can help build social capital within neighborhoods and within groups, while also promoting the health of older adults (Morrow-Howell et al., 2001). These social and health resources could then be used to acquire employment in later life for older adults who need and/or want to work. Carr and Kail (2013) found that older adults who engaged in paid work and formal volunteering prior to full retirement were more likely to continue to volunteer during retirement and subsequently return to work compared to those who did not volunteer. They speculated that social resources associated with volunteering had facilitated the transition out of retirement to paid work. Gonzales and Nowell (2017) expanded this literature by examining social bonding (close social ties among family members), social bridging (looser social ties among friends and neighbors that “bridge” people horizontally within a social stratum, p. 1102) and social linking (social ties that connect individuals to others in higher and lower social stratum) within the context of retirees returning to work. They found that retirees who helped friends, neighbors, or relatives, as well as those engaged as formal volunteers with religious, educational, health-related, or other charitable organizations, significantly increased their odds of returning to work, after controlling for factors associated with un-retirement (e.g., health, education, lifetime occupational status, wealth). They also found a dose response, where high-intensity volunteering, either formal or informal, significantly increased the odds of returning to work when compared to individuals who did not volunteer or only engaged in low-intensity volunteering. These findings offer some support that social bridging, such as informal volunteering, and social linking, such as formal volunteering, are important resources to obtaining employment in later life.

Smith (2017) argues social compositions and social structures vary by race, ethnicity, gender, and nativity, and may contribute to longstanding labor force participation inequities. Age is another important dimension, as older adults are more likely to access their social capital to find employment, whereas younger people are more likely to rely on their human capital (Gayen et al., 2010). Ajrouch et al., (2016) also argue social capital might be more important than human capital for older adults seeking employment, particularly for low socioeconomic populations. Additional research that clarifies the social mechanisms predictive of formal labor market participation in later life is warranted given scant and mixed evidence among working individuals generally, and the limited research on older adults more specifically.

CROSS-CUTTING THEMES OF INEQUITY: A LIFE COURSE PERSPECTIVE

Life Course Disparities

The move away from a standardized normative life course, standardized career paths, and conventional retirements (Carr et al., 2021; Calvo et al., 2018; Warner et al., 2010; Kohli, 2007) in tandem with a transforming global economy and shifting policies in both Europe and the United States (including the demise—

in the United States at least—of mandatory retirement ages) means that older individuals and couples find they must customize their own employment and retirement decisions (Hofäcker et al., 2016; Ebbinghaus and Hofäcker, 2013; Cahill et al., 2015, 2006). Decision-making involves more than simply deciding on the timing of exits, and can involve decisions around whether to move to more limited engagements, take on bridge jobs, or, if retired, to seek reemployment (Kail and Warner, 2013; Tang et al., 2013; Warner et al., 2010; Moen, 2007). Such choices are made in the face of current and perceived future demands, resources and risks, as well as motivations and goals (Kojola and Moen, 2016; Wang et al., 2011). But demands, resources, and risks are factors in motion. Unexpected exigencies, such as an acute health condition, job loss, the sudden need to provide family care, enhanced job demands, layoffs, an economic downturn, or a health pandemic—all can shift the calculus around continued work engagement, retirement, disability or family-care exits or reemployment.

Decision-making is often couched in terms of weighing various options and making optimal choices. But choice around labor market continuities and exits are often constrained, with freedom of choice differentially distributed across older adult populations (Moen, 2016; Sohier, 2019). Control over deciding when, where, and how much to work or retire is limited for those in historically disadvantaged social positions (women, minorities, immigrants, the less educated—see Brown and Warner, 2008; Choi et al., 2016; Warner and Brown, 2011; Zajacova et al., 2014). Moreover, decisions about work and retirement in later adulthood are made without knowledge of the consequences of doing so, as well as possible future circumstances, particularly with regard to individual and family members’ future health and income, but also possibilities in terms of, for example, future employer buyouts or reemployment.

These decisions can be viewed from a model of decision-making behavior called prospect theory, which applies particularly to risky or uncertain choices. “Risky” choices are defined as decisions made “without advance knowledge of their consequences” (Kahneman and Tversky, 1984, p. 341). Kahneman and Tversky find that people typically conceptualize outcomes in terms of gains or losses relative to a given reference point, rather than as final assets. Losses usually loom larger than gains; an individual is more distressed at the prospect of a loss than pleased by a potential gain (“loss aversion”). Given declines in social protections like secure DB pensions and increased longevity, retirement can be seen as a risky choice for some older adults who fear they may outlive their savings. People tend to be risk-averse, but if staying in their current jobs might mean a possible loss, risk seeking is observed (Kahneman and Tversky, 1979). Given rising intensity and precarity of work (Kelly and Moen, 2020; Kalleberg, 2018), others might see exiting in the form of retirement, disability, or because of caregiving demands, as a potential gain over remaining in a state of chronic pain or stress, especially coupled with some, even minimal, guaranteed income in the form of Social Security benefits. Those in poor health may view remaining in their current jobs as a potential further loss of health and well-being, compared to the relief of full-time retirement leisure.

Race and ethnicity may also shape the timing of exits, with White people both better positioned in the labor market and having more resources (health, education) than Black people, Hispanic people, or other minority groups (Moen et al., 2021; Lahey, 2018; Stainback et al., 2018; Zajacova et al., 2014; Warner and Brown, 2011; Willson, 2003), favoring greater time in paid work.

Certain working conditions can also be beneficial or harmful for health (Kelly and Moen, 2020; Staudinger et al., 2016). Noxious environments and physically demanding jobs obviously take their toll, leading to early retirements when financially possible. But psychosocial conditions also matter; jobs with high demands, low control, and little support lead to poor health outcomes for workers of all ages and life stages (Karasek and Theorell, 1990). Alternatively, those who feel satisfied with their jobs, coworkers, and bosses are apt to avoid risks around the uncertainties of leaving, while those who feel they cannot sustain the level of work required and who cannot scale back or move to a more supportive manager will take on the risks associated with exiting. Greater flexibility and control, together with supportive supervisors, can be conducive to deciding to remain with one’s employer longer (Moen et al., 2016a).

Recall losses usually loom larger than gains; for healthy college-educated workers, retirement can be seen as the loss of interesting, meaningful activity as well as a paycheck. For those older workers remaining in the workforce past age 65, the health and personal development aspects of work ran parallel to that of the financial

motivations (Reynolds and Wenger, 2012). By contrast, for workers in poor health and with little education, exits concomitant with Social Security eligibility may render retirement a welcome risk and even gain in the face of the difficulties of working and the uncertainties around both jobs and earnings.

It is worth highlighting that the population 65 years old or older is becoming more racially diverse. Older Hispanic adults in the United States are one of the fastest-growing population groups in the United States (U.S. Census Bureau, 2018; Flores and Radford, 2017). Hispanics were 6.9 percent of the U.S. population ages 65 and older in 2010, and they are projected to comprise 18.6 percent of the U.S population that is ages 65 and older in 2050 (U.S. Census Bureau, 2018). Among Hispanics 65 years old or older, 56 percent are foreign-born (Johnson et al., 2013). Poverty rates among minority older adults 65 years old or older in the United States is higher among Blacks and Hispanics than among non-Hispanic Whites. Poverty rates among older adults are 16.2 percent for Blacks, 17.7 percent for Hispanics, and 6.6 percent for non-Hispanic Whites. Foreign-born older Hispanics poverty rate is 20.1 percent, compared to 14.7 percent for Hispanic U.S.-born (Flores and Radford, 2017).

According to the life course theory, early SES and social interactions affect adulthood social conditions that in turn affect health at older ages (e.g., Berkman, 2009). According to Flippen and Tienda (2002), race, ethnicity, and gender are ascribed characteristics associated with education, occupation, and marital status. All of these factors are associated with employment stability, income and wealth trajectories, saving for retirement, and health status, which in turn affect retirement behavior and income security in old age. Adults that face limited work opportunities during their prime working years continue to face a considerable disadvantage when reaching retirement (Flippen and Tienda, 2002).

Racial-ethnic disparities in economic outcomes begin to appear at young ages and expand throughout the life course. Although differences in education can only partially explain these disparities in economic outcomes at older ages, older Black and Hispanic people have completed less education than have older White people. Black and Hispanic people are less likely to have completed a bachelor’s degree; in 2014, only 11 percent of Hispanic older adults 65 years old or older had completed college, compared to 15 percent of Black older adults and 26 percent of non-Hispanic White older adults.

The economic effects of these differences in education on employment and income cumulate over the life course, affecting access to public and private retirement benefits and wealth accumulation, resulting in large racial-ethnic disparities in financial resources and security at older ages. Although almost 90 percent of older non-Hispanic White workers receive Social Security benefits, only 81 percent of older Black workers, and 75 percent of older Hispanic workers do so. Non-Hispanic White workers are also more likely (39%) to receive employer provided pensions than Hispanic workers (22%). An important component of wealth for older adults is housing, and here too, non-Hispanic White people enjoy a large advantage in homeownership, even as they also enjoy the benefits of possessing a more diverse asset portfolio that is less reliant on housing value (Johnson et al., 2013). As a result, only four percent of non-Hispanic White workers had zero or negative net wealth, compared to 19 percent of Black workers and 23 percent of Hispanic workers. These racial-ethnic disparities in financial security differentially constrain retirement and work options at older ages, leaving racial-ethnic minorities less able to realize their preferences for work and retirement.

Life course research, in conjunction with cumulative (dis)advantages and inequality, has expanded our conceptual understanding of work and retirement trajectories among diverse older adults. Given the challenges of adopting an intersectional approach, critical race theory (which holds that legal and social institutions are designed in ways to enforce and perpetuate social, economic, and political inequalities between White people and non-White people) and perspectives on scarcity may yield important insights on how individuals and households make complicated decisions in a context of resource scarcity (Collins, 2019; Calasanti and King, 2015; Collins and Bilge, 2015; Mani et al., 2013; Mullainathan and Shafir, 2013; Shah et al., 2012; Delgado and Stefancic, 2000; Crenshaw, 1991). Further, much of our longitudinal and/or population data have limited sample sizes on race and ethnicity or gender and sexual minorities. There is great heterogeneity within these subgroups, and exploring this heterogeneity across time within life course, critical race, and scarcity perspectives, may prove useful to developing sensitive policies and practices at the national, state, and employer levels.

RESEARCH IMPLICATIONS

The science on work and retirement trajectories at the individual and social levels is complex. Social research has clarified longstanding research questions, yet there is much that we do not know. Below are the most promising areas for future research.

• Health is a key determinant of retirement for some workers, while others retire for nonhealth-related reasons and retain substantial work capacity during the early years of retirement. Health disparities by race-ethnicity and education are a leading candidate to explain group-level differences in work at older ages, but more research is needed to establish this connection. Similarly, more research into the work capacity of different types of older workers would be of great value, given that policy discussions frequently include proposed

• reforms aimed at encouraging longer work lives. In addition, more research on the causal effects of mental and cognitive health on retirement is needed, as much of the existing literature focuses on physical health or on the health effects of retirement.
• The differences in labor force participation at older ages by education group are strikingly large. However, the existing literature has not established the extent to which this reflects a causal effect of education on retirement, the underlying mechanisms for this association, or the role that rising education has played in the recent increase in labor force participation at older ages. More work on all of these questions is urgently needed, because differences in work at older ages translate into meaningful differences in retiree well-being. Importantly, this work needs to take account of the fact that education-based differences in labor force participation rates begin long before traditional retirement ages.
• Research is needed to understand barriers to saving for retirement, particularly for low-income, women, and minorities including the role of (1) financial literacy, (2) access to employer-sponsored pensions, (3) other mechanisms to save for retirement, and (4) social insurance programs that may reduce incentives to save and accumulate wealth. In addition, we need to develop and test culturally competent incentive mechanisms and financial literacy programs to diverse populations.
• Changes in the retirement landscape—notably the shift from DB to DC pension plans—have put much more responsibility on workers to make financial decisions that affect their retirement security. Research suggests that financial literacy can affect financial decision-making, and also that pension plan provisions such as default contribution rates or investment allocations strongly influence decisions. In a sense, these represent two options for changing retirement savings behavior, and additional research may help to clarify their relative efficacy.
• The association of informal caregiving and labor force attachment is dynamic and complex due to workplace policies and job characteristics that can support caregiving, resources and characteristics of the care provider, the type of dyadic relationship between care provider and receiver, as well as the demands of the caregiving role (Clancy et al., 2020). Research is needed to understand how informal caregiving interacts with household and employment characteristics over time by gender, race, ethnicity, education, and SES. Longitudinal population studies that explicitly ask about employment policies and practices among informal caregivers would enable research on employment and retirement trajectories among a diverse U.S. population.
• Household structures have become far more varied than in the past. Research that provides greater insight into the health, economic, social, and psychological factors to work and retire by household structure, and quality of household ties, is warranted.
• What social processes are most effective for older adults to work and retire? How might social support and leverage enable low-income older adults, caregivers, or individuals with health limitations to obtain and maintain work should they want and/or need to work? Are weak or strong social ties most effective for low-income older job seekers? How might the social composition of an older worker evolve as they transition between work and retirement, and possibly back to work? What federal, state, and community policies and practices enhance social capital among older adults to achieve their preferences for work and retirement?
• A common theme that emerged in the extant research on individual and social resources is the lack of data and knowledge on Black, Indigenous, and People of Color as it relates to work and retirement. Further research is needed to explore heterogeneity and disparities across race and ethnicity, as well as within racial and ethnic groups. Emerging theories from the fields of critical gerontology, sociology, law, public health, and social work are critical race theory, standpoint theory, intersectionality, health equity, and systemic racism, which have helped to guide some empirical knowledge base on health, education, financial assets, and financial capacity. These theoretical perspectives will likely help to center the lives of racial and ethnic minorities in work and retirement inequities research, which can then inform policies and practices to promote choices and opportunities for work and retirement in later life for a growing diverse and aging population.
• Research using a life course perspective has expanded our conceptual understanding of work and retirement trajectories among diverse older adults and provided a lens through which the processes that generate cumulative (dis)advantages and inequality among older adults can be observed and elaborated. Further elaboration of the processes through which early life experiences shape late-life outcomes is needed.