Retirement refers both to the act of withdrawing from the workforce and to the period of life that follows this action. By either definition, retirement is central to any discussion of aging. Yet retirement only emerged as a distinct chapter of life during the 20th century—a “third act” that follows the childhood years during which education is acquired and the prime adult years during which people work and accumulate assets.
Longevity increases played a key role in this development. Life expectancy at age 65 in the United States, for example, rose from 11.9 years for men and 13.2 years for women in 1935, the year that Social Security was established, to 17.0 and 19.6 years in 2018, an increase of 5.1 and 6.4 years, respectively (Bell and Miller, 2005). The probability of living to age 65 rose substantially over this period as well.
The age of labor force withdrawal is the other factor that determines the length of retirement, and this too changed dramatically. In the past, it was typical to work at even the oldest ages; the labor force participation rate of U.S. men ages 65 and older exceeded 75 percent in 1880 (Costa, 1998). This rate declined continuously over the next century, falling to below 20 percent by 1990, a pattern also seen in France, Germany, and the United Kingdom. This trend, combined with longer lives, means that most workers today can expect to spend two decades in retirement, and those who end up being long-lived may be retired for three to four decades.
The retirement decision has long been of interest to analysts and policy makers alike. First, it is central to well-being; by working longer, an indi-
vidual can accumulate more assets and has fewer years of retirement consumption to finance but also enjoys fewer years of leisure. For society as a whole, changes in the age of retirement affect taxes paid and transfers received, as well as the size of the labor force and the economy. As longevity increases tend to increase the share of societal resources going to the elderly, increases in the age of retirement can serve to counteract this trend.
In fact, there has been a strong trend toward later retirement for both men and women in the United States and around the world over the past several decades. Many factors may have contributed to this trend, including improvements in longevity and health, increases in education and sector shifts in the economy, and increasing labor force participation by women, which may induce husbands to work longer, due to the complementarity of leisure. Public and private pensions are also likely to have played a central role. Much as the increasing availability and generosity of public and private pensions helped to make earlier retirement possible over the course of the 20th century, changes to pensions in the past few decades that have raised retirement age and strengthened the financial incentives for work at older ages have encouraged people to work longer.
Not only is life’s third act longer than in the past but the line between work and retirement is not always as distinct as suggested by this analogy. Workers often spend a period of time working part time, in self-employment, or at another job after leaving their career employer, or they may retire but subsequently re-enter the labor force. While analysts often abstract from this reality for simplicity and focus on labor force participation or employment, understanding the more complex realities of the retirement transition is necessary for designing retirement policy.
The trends of increasing longevity and longer work lives are in contrast with another trend in the United States, the rising frequency of labor force exit via the disability route before the standard age of retirement. While it may seem puzzling that these trends could coexist, one explanation is that disability insurance (DI) policy parameters, such as the stringency of medical screening, are critical determinants of the use of the disability route and may indeed be more important than health in explaining differences in the receipt of disability benefits across countries or over time. It may also be useful to look beyond population averages at trends in health and work by socioeconomic status (SES), to understand whether there are groups experiencing worsening health and economic opportunities that may be reflected in rising DI use. As public pensions continue to evolve in a way that both encourages and incorporates an expectation of longer work lives, it is important to assess the ability of different segments of the population to respond.
In this chapter, I first highlight key aspects of retirement behavior, including the trends in labor force participation, the complexity of the
retirement transition, and the relationship between trends in disability and work. I then discuss findings from the retirement literature that help one to understand retirement decisions and to explain these trends. I conclude with some thoughts about the frontier of the retirement literature, including directions for future research. The analysis focuses on the United States but also offers some international context on these issues.
RETIREMENT PATTERNS AND TRENDS
Labor Force Participation
In the predecessor to this volume, Quinn and Burkhauser (1994) suggested that the trend toward earlier retirement for men in the United States may have stopped or even reversed. As seen in Figure 8-1, this analysis was prescient. For men ages 60 to 64, participation reached a trough of 53 percent in 1994 and has subsequently risen by 9 points, to 62 percent in 2016. For men ages 65 to 69, the trough occurred about a decade earlier, and participation has since risen by 13 points, to 37 percent. For men ages 55 to 59, participation fell by 4 points over this period, reaching a low of 77 percent in 2016.1 This drop reflects a different labor market trend: declining participation among prime-age men.2
The trends for women, seen in Figure 8-2, show even larger increases in participation. From 1980 to 2016, participation rose by 17 points for women ages 55 to 59 and 60 to 64, reaching 66 and 50 percent, respectively. Participation among women ages 65 to 69 rose by 13 points, to 28 percent in 2016. Unlike the U shape in participation for older men, women’s participation rose continuously over this period, reflecting rising participation by women of all ages. Goldin and Katz (2018) showed that the cohorts of women who are working longer at older ages had higher rates of participation throughout their life cycle; they found that cohort differences in work at older ages are largely a function of earlier changes in human capital accumulation, such as higher educational attainment and greater employment continuity.
The patterns in other developed economies are similar to those in the United States and are often larger in magnitude. Figure 8-3 displays
1 The participation rate by age group may be affected by variation in the size of individual birth cohorts. For example, when there is a large cohort of 60-year-olds, the participation rate at ages 60 to 64 may rise because 60-year-olds work more than 64-year-olds. This concern is mitigated by focusing on the long-term trend in participation rather than short-term fluctuations.
2 The Council of Economic Advisors (2016) documented this trend and explored supply- and demand-side factors that may have contributed to this decline. The authors concluded that the decline in real wages of less skilled workers is an important factor.
the labor force participation rate for men ages 60 to 64 in selected countries from 1980 to the present. In all of the European countries (France, Germany, Italy, Sweden, and the United Kingdom) and in Australia and Canada as well, there is a U-shaped pattern in men’s participation, with an increase of 14 to 21 percentage points from the trough to the 2016 value; Japan and Korea experienced an increase of 10 to 11 points. Despite similarities in trends, participation levels can vary substantially across similar countries. For instance, only 30 percent of men ages 60 to 64 in France were in the labor force in 2016, versus 65 percent in Germany and 75 percent in Sweden. Participation was highest in Japan, at 80 percent.
For older women, there have been large increases in participation in most countries, as seen in Figure 8-4. Many of the countries with the biggest changes are those where men’s participation also rose substantially. Women’s participation increased by 21 to 27 points in Canada, Italy, Sweden, and the United Kingdom, by 38 points in Australia, and by 44 points in Germany. Increases in the other developed economies have been smaller, at 8 to 17 points. Participation levels for women ages 60 to 64 in 2016 varied widely, from about 30 percent in France and Italy to 68 percent in Sweden. The gap in participation between men and women also varies across country, with older men being no more likely to work
than older women in France but about 15 percentage points more likely in most developed countries and more than 25 percentage points more likely in Japan and Korea.
The age pattern in retirement is also of interest. Most U.S. workers retire in their 60s; of those still working at age 60, about two-thirds of both men and women withdraw from the labor force by age 70. Retirement accelerates starting at age 60, with spikes in the retirement hazard (or probability of labor force exit at a given age, conditional on remaining in the labor force until that age) at ages 62 and 65. These spikes correspond to the traditional early and full retirement ages for Social Security and the Medicare eligibility age. Evidence suggests that these programs do contribute to the spikes. Burtless and Moffitt (1986) noted that the age-62 peak in retirement only emerged after the option of claiming benefits at 62 was introduced, and Mastrobuoni (2009) found that the rise in the Social Security full retirement age (FRA) has led to an increase in the average age of retirement that is half as large as that in the FRA. Gruber and Wise (1999) documented that the share of remaining workers who exit at the pension full retirement age is 60 percent or above in Belgium, France, Italy, the Netherlands, Spain, and the United Kingdom, far larger than the spike in the United States.
Labor force participation and retirement patterns differ substantially by education and race. Coile and Levine (2010) reported that the annual probability of retiring at ages 55 to 69 in the United States was a third or more lower for college graduates than for high school dropouts, leading to lower labor force participation rates at ages older than 55 for the less-educated. Flippen and Tienda (2000) noted that the employment rates for Black and Hispanic men ages 51 to 61 in 1992 were 15 and 8 points lower, respectively, than the rate for White men, while the rates for Black and Hispanic women were 2 and 10 points lower, respectively, than the rate for White women. Differences in labor supply of these groups at older ages could be driven by differences in longevity or in the enjoyment derived from work, among other factors. Bound et al. (1995) concluded that differences in health status can explain much of the Black versus White differences in labor force attachment and essentially all of the differences by education. Hayward et al. (1996) concurred that Blacks’ lower participation rates are a function of disability but noted that due to their shorter life expectancy, Blacks spend more of their lives in the labor force than do Whites. They concluded that “retirement is more a White experience than a Black experience.”
Indeed, a focus on labor force participation rates by chronological age ignores differences in life expectancy across racial and education groups and across birth cohorts. Groups that have a longer life expectancy at a given age, say age 65, may be expected to have higher labor force participation rates, reflecting the additional years of retirement consumption to be financed, as well as social norms about the share of adult life spent in work and retirement. Steuerle and Spiro (1999) calculated the labor force participation rate over time for a man with a constant life expectancy (as opposed to a constant chronological age). They found that the decline in their measure between 1940 and 1997 is about 15 percentage points greater than the decline in the traditional age-based measure.
In summary, labor force participation rates have risen for men and women over the past two decades, both in the United States and other developed economies. The pattern for men is a U shape, with the recent increase coming after a century of decline, whereas for women, participation rates have only increased. In the United States, most workers retire during their 60s, and the probability of retirement spikes at ages associated with Social Security and Medicare eligibility. Blacks, Hispanics, and those with less education have lower labor force participation at older ages than do their White and more-educated counterparts. Finally, focusing on participation at a given life expectancy rather than chronological age may alter perceptions of differences in participation over time or across demographic groups.
The Retirement Transition
Whereas researchers often focus on the labor force participation rate due to its ease of measurement, simplicity of interpretation, and longevity in survey data (Hayward et al., 1996), this measure does not capture the complexity of many retirement transitions. The traditional retirement transition is one in which a worker moves from full-time work for a career (or long-term) employer directly to full retirement. Yet many other paths are possible. First, workers may reduce hours but remain with the same employer, a practice known as “phased retirement.” Second, workers may leave their career job for a transitional “bridge job” that may offer the opportunity to perform a different kind of work or to work fewer hours. Third, workers may switch from working for others to self-employment. Finally, an initial labor force exit may subsequently be followed by labor force re-entry or “unretirement.”3
Collectively, the use of these other exit routes is quite high. Cahill et al. (2015) reported that among U.S. workers ages 51 to 56 who were working for a career employer in 1992, 4 in 10 had used a bridge job by 2010. About half of these bridge jobs were part-time jobs. Johnson et al. (2009) found that two-thirds of workers in their early 50s who moved to a new job before retiring also changed occupations. The most common career switch is from a managerial job into a sales or operator position. For those who change jobs, the new job typically features a lower hourly wage and is less likely to offer health insurance benefits, but it is also rated as having less stressful working conditions and higher worker satisfaction when compared to the old job.
Relative to the prevalence of bridge jobs, use of phased retirement is less common in the United States, with Cahill et al. (2015) finding that only 2 to 3 percent of workers reduced the hours on their career job by at least 20 percent over an 18-year period. Lack of access to phased retirement is one potential reason for the low utilization of this route. Hutchens (2010) found that while most white collar employers say that they permit phased retirement, employers are selective in providing employees with the opportunity to enter this arrangement, with older, high-performing workers more likely to be granted access.
3 Other terms have also been used to describe these transitions. “Gradual retirement” is a generic term that refers to a gradual withdrawal from the labor force by any means that results in reduced work effort (Kantarci and Van Soest, 2008); this term could encompass phased retirement, bridge jobs, and transitions to self-employment. In a characterization related to the idea of unretirement, Mutchler et al. (1997, p. S4) distinguished between unidirectional “crisp” exits from the labor force and “blurred” transition patterns that “may include repeated moves in and out of the workforce, and may even include periods during which the roles of worker and retiree are held simultaneously;” they find that the latter is more common than the former.
Table 8-1 details the frequency of part-time work among older individuals in the United States. These values include part-time work on both career and bridge jobs. To allow for some comparison over time, averages are shown for the years 1980 to 1994 and 1995 to 2016, corresponding to the periods in which male labor force participation was falling and rising. For men, part-time work rises with age, with 5 percent of men ages 55 to 59 working part time in the more recent period, versus 10 percent of men ages 65 to 69. Values for women are higher, especially at younger ages: 13 percent of women ages 55 to 59 work part time.
It is also useful to consider the share of workers who are engaged in part-time work. This rises even more dramatically with age, with one-third of male workers and nearly half of female workers ages 65 to 69 working part time. There is no clear pattern of differences across the two time periods, as the share of individuals working part time rises while the share of workers doing so declines over time.
Comparable data for Europe are available from Kantarci and Van Soest (2008). In 2001, the share of workers ages 51 to 65 working part time averaged 9 percent for men and 42 percent for women in their sample of 14
TABLE 8-1 Part-Time Work and Self-Employment in the United States, 1980–2016 (in percentage)
|Part-Time Work||Part-Time Work as Share of All Work|
|Self-Employment||Self-Employment as Share of All Work|
NOTE: Part-time work is defined as working less than 35 hours per week.
SOURCE: Author’s calculations from March Current Population Survey, 1980–2016. Data are weighed using person weights.
countries, with considerable variation across countries. This average value is the same as that for U.S. men and more than twice that for U.S. women. As in the United States, there is no clear trend over time. Partial retirement in Europe may be facilitated by the existence of partial pension schemes, which allow individuals to reduce work hours and draw a share of the full pension. Denmark, Finland, France, Germany, Spain, and Sweden all have, or have had, such schemes, and many other European countries make it possible for workers to combine work and pension receipt in some way (Reday-Mulvey, 2000). Kantarci and Van Soest (2008) noted that the relative lack of mobility in European labor markets, as compared to the U.S. market, may increase the appeal of phased retirement rather than a bridge job for those seeking part-time work.
For some, self-employment may offer yet another option for reducing labor force activity. Self-employment provides greater flexibility in hours, as well as greater autonomy and the potential to work a nonstandard schedule. Zissimopoulos and Karoly (2007) showed that the share of workers who are self-employed rises with age and estimated that this is primarily driven by net switching into self-employment at older ages, rather than by later retirement among the self-employed. Ramnath et al. (2017) found that the self-employment hazard rises at key Social Security eligibility ages and that transitions to self-employment are associated with declines in income (particularly at older ages) and hours worked, supporting the notion that self-employment can serve as a means of gradual retirement.
Table 8-1 reports the prevalence of self-employment by age and sex in the United States. In a reversal of the pattern for part-time work, self-employment is more common among men than women, with 15 percent of men and 7 percent of women ages 55 to 59 engaged in such work. The share of individuals who are self-employed falls with age, while the share of workers rises, reaching 30 percent of male workers and 17 percent of female workers at ages 65 to 69. There is a slight increase over time in the share of workers who are self-employed.
While treating labor force withdrawal as unidirectional may be convenient for analysts, it is fairly common for workers to re-enter the labor force after an initial exit. Maestas (2010) found that 24 percent of those who are not working for pay return to either part-time or full-time work within 6 years, while an additional 16 percent transition from part-time to full-time work. Interestingly, about four-fifths of those who unretire had anticipated making this transition; there is little evidence that people return to work due to negative financial shocks, but some indication that changes in leisure preferences may play a role. As with bridge jobs, the new jobs of those who re-enter often feature lower wages, fewer benefits, and a change of occupation, particularly from manufacturing and managerial or professional work into services, sales, and administrative support positions.
Overall, it is evident that many workers retire gradually, either by reducing hours on their current job, moving to a bridge job, or switching to self-employment; many also return to the labor force after an initial exit. Across all of these paths, it is typical for those who are still working at older ages to be in a different occupation, working fewer hours, and receiving lower pay and benefits as compared to their career job.
Disability and Work
The long-term trends toward increasing longevity and longer work lives run counter to another significant trend in health and work in the United States: the increasing prevalence of workers exiting the labor force due to disability. Disability and the receipt of DI benefits are distinct yet related concepts, both of which are relevant to this discussion.
While there is no universal definition of disability, it is generally understood to refer to having a physical or mental health condition that limits an individual’s ability to work for pay. Given this broad definition, it is not surprising that identifying an ideal measure of disability prevalence is challenging. A self-report of a work-limiting disability is directly relevant to labor supply decisions and available in many surveys. However, responses may not be comparable across individuals and may be influenced by the availability of DI benefits or by the individual’s work decision (Waidmann et al., 1995). This criticism is levied at self-reported measures more broadly, including health status, a widely used measure. Objective health measures, such as diagnoses of chronic diseases, skirt these issues but may be only imperfectly correlated with work ability (Bound, 1991). Comparisons over time may be complicated by changes in reported versus true prevalence or changes in the efficacy of treatment. Measures of functioning include self-reports of the ability to perform activities of daily living such as bathing and dressing, instrumental activities of daily living such as housework and cooking, and physical tasks such as walking and climbing stairs. These measures of functioning, as well as reports of pain and measurements of physical ability (for example, grip strength), complement other measures of disability but may be subject to some of the same concerns.
Early work on disability proposed several competing hypotheses about the implications of falling mortality rates for disability at older ages. The “compression of morbidity” theory posits that the age of disability onset will continue to rise as mortality declines reach a natural limit, leading the period of disability to be compressed into a shorter number of years at the end of life (Fries, 1980). By contrast, the “failure of success” view suggests that mortality rates will decline while the incidence of chronic diseases remains constant, resulting in individuals being disabled for a longer period
of time (Gruenberg, 1977).4 In the former scenario, workers who are living longer would have more years free of disability and might work longer as a result. In the latter, advances in life expectancy would not translate into increased capacity to work at older ages; indeed, the share of the cohort in poor health could be rising over time, if additional survivors are disproportionately drawn from the least healthy in the cohort.
The literature has not reached a consensus as to which view prevails. One reason for this uncertainty is the multiplicity of measures of disability and health, as discussed above. As Crimmins (2004) noted, measures do not necessarily all move in the same direction, at the same time, and equally for all age groups. Focusing on disease and mobility functioning, Crimmins and Beltrán-Sánchez (2010) found that the length of life with disease and mobility functioning loss increased between 1998 and 2008. By contrast, Cutler et al. (2014) defined disability in terms of impairments in activities of daily living and instrumental activities of daily living and concluded that disabled life expectancy shrank and disability-free life expectancy rose between 1991 and 2009. Crimmins (2004) reported that the health of older people improved along most dimensions in the preceding two decades.5
Relative to disability, DI receipt is easier to define and measure. Individuals who are disabled may be eligible to receive DI benefits. DI applicants typically must pass a medical screening to verify that their disability meets the established criteria. For example, the U.S. Social Security Act defines disability as “the inability to engage in any substantial gainful activity by reason of any medically determinable physical or mental impairment(s) which can be expected to result in death or which has lasted or can be expected to last for a continuous period of not less than 12 months.” In the United States and many other countries, individuals must have worked and made payroll tax contributions to the DI program for some period before disability onset in order to be eligible for DI benefits (Wise, 2016).
The share of the population ages 60 to 64 that is receiving DI benefits provides a fair approximation of the use of the DI exit route from the labor force (as individuals can only claim DI until the Social Security FRA, now age 66 for those born from 1943 to 1954). In 2016, 14 percent of men and 12 percent of women ages 60 to 64 were receiving DI benefits, reflecting an increase of about 3 percentage points for both groups since the late 1980s. DI participation among the working-age population as a whole dipped briefly in the early 1980s, following a tightening of medical eligibility,
4 An intermediate view is that both mortality and morbidity will continue to decline, with indeterminate effects on the relative length of disabled and disability-free life at older ages (Manton, 1982).
5 See Cutler and Wise (2008) for additional analyses of health trends at older ages. More recently, evidence of rising morbidity and mortality in midlife for U.S. Whites (Case and Deaton, 2015) raises the concern that the heath of the elderly could decline in the future.
but has grown steadily over the past 25 years. Autor and Duggan (2006) attributed this growth to three factors: a 1984 law that loosened medical eligibility, the rise in women’s labor force participation and DI eligibility, and the rise in DI benefits relative to income for low-income workers. Liebman (2015) estimated that DI growth in the 1980s was primarily driven by higher DI incidence rates after the 1984 law, whereas program growth since the early 1990s is largely the result of population aging and rising eligibility and incidence rates among women.6
The rate of DI receipt in the United States today is similar to that in many developed countries. Coile et al. (2016) found that the share of men ages 60 to 64 on the DI rolls is between 12 and 17 percent in nearly all (9 of 11) countries they studied. However, the time trend in the United States is quite different, with the United States being virtually alone in seeing a rise in participation over time. DI participation fell by 40 percent or more from its mid-1990s peak in six of the countries, with smaller decreases in three others. Wise (2012) found that these declines typically followed DI reforms, such as eliminating the ability to enroll in DI for nonhealth reasons and increasing reviews of DI recipients to verify medical eligibility. Koning and Lindeboom (2015) detailed how DI reforms ended the “Dutch disease” of high disability rates, while Banks et al. (2015) provided a similar analysis for the United Kingdom.
To recap, U.S. workers are increasingly likely to exit the labor force via the disability route, with about one in seven now receiving DI benefits before they reach Social Security eligibility. This trend runs counter to the trend in most other countries, where use of DI peaked in the mid-1990s but has fallen since then as a result of DI reforms, although current levels of DI use in the United States are similar to those in other countries. There is no clear evidence that the U.S. trend is driven by an increase in disability. Rather, changes in medical screening and demographic and economic factors may play a larger role.
EXPLAINING RETIREMENT BEHAVIOR AND TRENDS
Data and Methods
Before exploring explanations for recent trends in work at older ages, it is useful to survey what is known about the determinants of retirement,
6Pattison and Waldron (2013) also stressed the role of population aging and rising eligibility for women. Note that while population aging is important for explaining trends in DI receipt for the working-age population as a whole, it explains little to none of the increase in DI receipt among those ages 60 to 64.
beginning with a brief overview of the data and methods central to this work.
The retirement decision has been studied extensively over the past several decades. When the predecessor to this essay was written (Quinn and Burkhauser, 1994), much of the existing work on retirement relied on longitudinal surveys from the 1970s, including the Retirement History Survey and National Longitudinal Survey of Older Men. Although these surveys were critical for the research of that era, they lacked data on women and on more recent cohorts of older workers. A new longitudinal study, the Health and Retirement Study (HRS), had just been launched to fill the gap.
A quarter-century later, the retirement literature has been refreshed with a new series of studies, many of which use HRS. HRS is a nationally representative, longitudinal survey of the U.S. population, ages 50 and older. Respondents have been interviewed biennially since 1992, and the sample is refreshed regularly with new cohorts. The core survey collects extensive information on work, income, wealth, pensions, health, health care, family transfers, and more. This core is supplemented by links to administrative data including Social Security records, private pension data, and Medicare claims data and by the collection of health data such as biomarkers and venous blood.7 HRS is the model for a growing international network of longitudinal aging studies (see the chapter by Lee and Smith in this volume). For example, the Survey of Health, Ageing, and Retirement in Europe (SHARE) includes data for 14 countries, and there are separate studies in China, Costa Rica, England, India, Ireland, Japan, Korea, and Mexico. These studies collect much of the same data as the core HRS, enabling international comparisons as well as within-country studies.8
A second data advance during the past quarter-century is the increase in researcher access to administrative datasets, such as Social Security earnings and benefits records and income tax records. These data offer larger sample sizes than traditional surveys, in some cases covering the full population, and greater accuracy in measuring income and benefits. However, they typically lack information on demographics and other variables of interest,
7 To facilitate use of this extraordinarily rich and complex dataset, the RAND HRS provides researchers with a version of the data in which variables are defined consistently over time and named intuitively. For more information on RAND HRS and related data products, see https://www.rand.org/labor/aging/dataprod.html [April 2018].
8 The current list of these studies (with start dates) includes SHARE (2004), the English Longitudinal Study of Aging (ELSA, 2002), the Irish Longitudinal Study on Aging (TILDA, 2010), the China Health and Retirement Longitudinal Study (CHARLS, 2011), the Japanese Study on Aging and Retirement (JSTAR, 2007), the Korean Longitudinal Study of Aging (KLoSA, 2008), the Longitudinal Aging Study in India (LASI, pilot), the Costa Rican Longevity and Healthy Aging Study (CRELES, 2005), and the Mexican Health and Aging Study (MHAS, 2001). For harmonized data from these studies, see https://g2aging.org/ [April 2018].
except where they have been linked to survey data or other administrative records.
There have also been significant advances in the empirical methods used in retirement research over the past quarter-century. First, empirical microeconomics has experienced a “credibility revolution” (Angrist and Pischke, 2010). Many studies in the 1970s and 1980s relied on “naïve” regression analysis using cross-sectional variation that did not adequately address perennial challenges such as omitted variable bias or did not produce results robust to assumptions. By contrast, researchers today realize that identifying a research design that will allow them to estimate a plausibly causal effect of X on Y is paramount. Obtaining experimental evidence through a randomized controlled trial is one means to this end; however, such trials can be expensive and may be difficult to adapt to some questions. A second option is to identify quasi-experimental variation in the key independent variable. Such variation might arise, for example, when a policy changes over time or applies only once people reach a certain age. Empirical techniques including instrumental variables, regression discontinuity design (RDD), and others allow the researcher to exploit this variation effectively. Many of the best recent papers in the retirement literature use such techniques.
Second, structural models have continued to evolve and to be an important part of the retirement literature. Structural models posit that individuals make work decisions so as to maximize lifetime utility (or happiness) as expressed in a specified function. Relative to the literature using cross-sectional or quasi-experimental variation, structural models have the advantage of being able to estimate parameters such as the discount rate (which reflects the strength of the preference for receiving money now rather than in the future) or degree of risk aversion, which can then be used to simulate the effect of policies outside the scope of real-world experience. However, key assumptions are less transparent than in the empirical methods listed above, and the validity of these models relies on specifying the correct form for the utility function. Over time, structural models of retirement have become more complex, incorporating features such as uncertainty in asset returns or in out-of-pocket medical expenses.
Third, advances in behavioral economics have begun to make their way into the retirement literature. While this is most notable in the literature on retirement savings decisions (Choi et al., 2004), insights from this field are beginning to influence the analysis of other questions in the retirement sphere, such as the Social Security claiming decision (Brown et al., 2016).
In short, the retirement research of the past 25 years has relied heavily on HRS but is increasingly making use of “sister studies” from other countries and of large administrative datasets. The empirical literature has come to be dominated by studies that use new techniques aimed at generating
causal estimates, even as structural modeling continues to be used. The new data and methods are highly complementary. For example, using the RDD approach may require administrative data on the full population, in order to compare people in the immediate vicinity of an arbitrary cut-off (such as a January 1 birthdate) who are subject to different policy regimes but are otherwise similar. Calculating financial incentives from Social Security and pensions, whether to be used in a structural model or quasi-experimental estimation, relies on having individual earnings records and detailed information on private pensions. The richness of the current retirement literature is due in no small part to the availability of excellent data.
Determinants of Retirement: Labor Supply
Many factors may influence whether individuals continue to work at older ages. These factors may be divided into those that affect the individual’s decision of whether to supply labor and those that relate to the demand for older workers. While a full examination of this large literature is beyond the scope of this chapter, I briefly review some key findings on labor supply and demand; see Coile (2015) for a longer discussion.
The effect of public and private pensions on retirement has been the subject of much research. One reason for the strong interest is that the long-term decline in labor force participation of older men occurred over the same period that public and private pensions expanded dramatically in the share of covered workers and in generosity. In theory, public and private pensions may influence work by creating income effects if individuals receive benefits in excess of their own contributions. They may also generate substitution effects if work at older ages raises or lowers the expected present value of pension wealth. Further, pensions can create liquidity constraints if individuals would like to retire earlier but are unable to access their pension wealth until the eligibility age. The spikes in the retirement hazard noted above at the early retirement age and FRA for Social Security benefits is consistent with this scenario.
Modeling the financial incentives from public and private pensions properly and convincingly identifying their effect have been an important focus of recent work. Much of the most compelling evidence of income effects comes from historical analyses. Costa (1995) found large effects of Union Army pensions on male labor force participation around 1900. Fetter and Lockwood (forthcoming) also found large income effects, using variation in the state-run Old Age Assistance program of the 1930s. They project that the growth of Social Security can explain more than half of the decline in men’s participation from 1940 to 1960. The “Social Security Notch,” in which benefits were cut sharply for those born in 1917 relative to earlier cohorts, provides a useful natural experiment. Krueger and
Pischke (1992) found that the cut did little to halt the long-term decline in labor force participation, whereas Gelber et al. (2016) obtained larger effects by comparing those born immediately on either side of the Notch. The latter used their estimates to project that growth in Social Security can account for at least half of the decline in older men’s employment from 1950 to 1985.
In modeling substitution effects, it is important to focus not only on the “accrual,” or change in pension wealth, from working one more year but also on forward-looking measures that capture the change in pension wealth from work in future years. Several measures have been proposed, including tax force (Gruber and Wise, 1999), which is the gain or loss in pension wealth from working to a specific age, relative to earnings; peak value (Coile and Gruber, 2007), which is the financial gain from working to the age when pension wealth is maximized; and option value (Stock and Wise, 1990), which is the increase in utility arising from retiring at the optimal future year.
Many studies find that these measures are related to retirement. Gruber and Wise (1999) estimated that 80 to 90 percent of the differences across a dozen countries in the share of men out of the labor force at ages 55 to 69 can be explained by differences in tax force. Friedberg and Webb (2005), Gruber and Wise (2004), and Samwick (1998) found that individuals with a larger incentive to work (PV or OV) retire later. Liebman et al. (2009) showed that Social Security incentives affect work on the intensive margin, using quasi-experimental variation arising from nuances of the Social Security benefit formula. Finally, early studies featuring structural retirement models (Gustman and Steinmeier, 1986; Stock and Wise, 1990) established the value of this approach for evaluating the effect of public and private pensions on retirement, while several recent studies have found that past or proposed future reforms may have large effects (Gustman and Steinmeier, 2009, 2015).
The DI program may affect labor force participation at older ages if medical screening is imperfect and access to the program is not limited to those who are unable to work. Maestas et al. (2013) used variation in DI awards generated by random assignment of DI applicants to disability examiners, as two similar applicants may receive different outcomes depending on the leniency of their examiner. They found that for the one-quarter of DI applicants on the margin of program entry, employment would be about 30 percentage points higher in the absence of the DI program, although additional earnings are fairly modest. French and Song (2014) similarly used random assignment of appeals of rejected DI applicants to judges and obtained comparable results, as did Chen and Van der Klaauw (2008), who used age-based differences in eligibility standards.
The retirement decision is likely to depend on an individual’s total wealth, not on pensions alone. However, estimating the effect of wealth on
retirement is difficult, since individuals decide how much to save and those who are high savers may have other characteristics that affect their retirement decision. Unanticipated shocks to wealth offer one fruitful means to explore this question. Imbens et al. (2001) documented that winning the lottery leads to reductions in labor supply for older workers, while Brown et al. (2010) showed that the receipt of an inheritance increases the probability of retirement, especially when unexpected. Stock market fluctuations also serve as a source of windfall wealth gains and losses, but the evidence for their effect on retirement is more mixed (Hurd et al., 2009; Coile and Levine, 2011a). In structural models, individuals make both work and savings decisions, and thus these models can be used to simulate the effect of wealth changes and borrowing constraints on retirement (French, 2005; Gustman and Steinmeier, 2005).
The role of health and health insurance has long been another focus of the retirement literature. The earlier discussion of disability measures hinted at some of the challenges in estimating the effect of health on retirement, such as the potential for self-reported health measures to be influenced by the work decision and the concern that objective measures only imperfectly capture work ability. Like wealth shocks, negative health events may provide compelling evidence of health’s effect on retirement if the shocks are unanticipated and unrelated to other factors. McClellan (1998) found that workers who experience a heart attack, stroke, or new cancer diagnosis are much more likely to retire, especially when the shock is accompanied by a loss in functioning; Disney et al. (2006) had similar findings for the United Kingdom, while Bound et al. (1999) showed that changes in health are strongly associated with labor force exit.
A related question that has arisen more recently is the effect of retirement on health. This effect is of interest to policy makers, as policies that encourage later retirement could have unintended negative consequences for health, if health effects are negative. The empirical landscape is complex, but overall seems to suggest that retirement is good for physical health (Bound and Waidmann, 2007) and mental health (Charles, 2004; Johnston and Lee, 2009), but retirement is also associated with cognitive decline (Rohwedder and Willis, 2010) and an increased risk of mortality (Fitzpatrick and Moore, 2017). Notably, many of these studies use the increase in retirement at public pension eligibility ages to identify these effects.
Health insurance access is likely to be more important in the United States than in other countries, given that most U.S. workers obtain insurance coverage for themselves and their families through their employer, while coverage is typically not dependent on employment in other developed countries. Employees who are offered retiree health insurance by their employer may more easily be able to retire before age 65, when Medicare
becomes available. Using data from firms with varying benefits, Nyce et al. (2013) showed that there is a large increase in retirement at ages 62 and 63 for employees who have such coverage, particularly when employers cover more of the premiums. A literature review on this topic concluded that health insurance is a “central determinant of retirement decisions” (Gruber and Madrian, 2004). However, the 2010 Affordable Care Act may reduce the importance of this factor by providing ways to obtain insurance outside the employment relationship.
Among the most important advances in the retirement literature over the past 25 years is the increasing attention paid to women’s retirement decisions and to family issues. As noted above, the longitudinal datasets of the 1970s focused primarily on male workers, complicating the effort to study women’s retirement. In addition, analysts often treated the husband’s retirement status as an exogenous explanatory variable in studies of women’s retirement, while excluding the wife’s retirement status in studies of men’s retirement (see Weaver, 1994, for a review of the early literature on women).
With the rise of two-career older couples, the lack of research on women and asymmetric treatment of husbands and wives in retirement models became increasingly anachronistic. The fact that both spouses retire within one year of each other in one-third of working couples (Coile, 2004) suggests that many couples approach retirement as a joint decision. The coordination in retirement timing could reflect several factors, including similarity of preferences, shared financial resources, and complementarity of leisure. Several authors estimated structural models of joint retirement and found complementarity of leisure to be a key driver of joint retirement (Gustman and Steinmeier, 2000; Maestas, 2001). Other authors (Baker, 2002; Coile, 2004) illustrated the presence of joint decision making by showing that the financial incentives facing one spouse have “spillover” effects on the other’s retirement decision. On a different note, with women more likely to serve as caregivers for elderly family members, studies on the effect of caregiving on retirement (Fahle and McGarry, 2018) are another welcome addition to the literature.
Determinants of Retirement: Labor Demand
Although analysts have focused much of their attention on factors that affect the decision to supply labor at older ages, it is critical to also examine the demand for older workers. One important consideration is how productivity and wages change with age. Standard economic theory suggests that workers will be paid according to their contribution to the firm’s output. Yet Skirbekk (2004) concluded from a review of the literature that productivity peaks in mid-career whereas earnings rise until late in the
work life. One explanation for this divergence comes from Lazear (1983), who posited that firms offer a steep age-earnings profile, paying workers less than their value when young, with the promise of high wages when old in order to encourage them to put forth effort on the job. A defined benefit (DB) pension is part of the worker’s deferred compensation and also provides an incentive for workers to retire (despite high wages) by having negative accruals in pension wealth after the plan’s retirement age, which effectively reduces pay.
A mismatch between wages and productivity is one reason why older workers might experience age discrimination. Testing for age discrimination is challenging, as it is difficult to know whether age is the reason that a worker was not hired. The best evidence comes from field experiments, in which fictitious resumes are created for workers who vary in age but are otherwise comparable. Neumark et al. (2015) reported that there is age discrimination against older women, but less evidence of discrimination against men. Lahey (2008a) also found age discrimination against women and concluded that it is likely statistical discrimination, in which employers use the average characteristics of a group as a proxy for the unknown characteristics of the applicant.
Analysts have explored the effectiveness of laws to protect older workers, such as the Age Discrimination in Employment Act, obtaining mixed results. Neumark and Stock (1999) found the law boosted the employment of those ages 60 and older, while Lahey (2008b) concluded that employment of older workers is lower when there is also a state age discrimination law.
The business cycle may affect the demand for older workers, resulting in more transitions to retirement when the labor market is weak. Coile and Levine (2007) found this to be true for U.S. workers, particularly once they reach the age of eligibility for Social Security. Dorn and Sousa-Poza (2010) reported that involuntary retirement due to economic conditions is widespread in Europe, especially in countries with strict employment protections. Late-career job loss has many detrimental effects, including reduced employment and earnings (Chan and Stevens, 2001), lower retiree income (Coile and Levine, 2011b), and higher mortality (Sullivan and von Wachter, 2009).
Longer Work Lives
The preceding discussion points to several factors that may have contributed to the rise in work at older ages. A first candidate is improving longevity and health. The known association between poor health and early retirement lends credence to this possibility. However, for most workers, health is not the main cause of retirement. Most workers retire between
ages 60 and 70, as discussed earlier, while health declines only slowly over this range. Moreover, comparing how much older individuals today work relative either to those in the past or to slightly younger individuals in similar health suggests that there is significant health capacity to work at older ages (Coile et al., 2017). Further, as the decline in mortality rates over time has been continuous, it seems unlikely to have precipitated the sudden reversal of the decline in men’s participation (see Figure 8-1). Thus, while better health may have supported longer work lives, there is little evidence that it is a primary driver.
Another plausible candidate is the rise in education and shift away from physically demanding jobs. As noted above, those with more education have higher labor force participation at older ages. Yet, as with health, the educational attainment of the near-elderly population has improved continuously over time. Moreover, the male labor force participation rate for each education group has followed the same U-shaped trend as the overall rate, pointing to a need to look beyond the composition shift in education.9
Another potential factor is rising participation among women, which might encourage men to work longer due to complementarity of leisure between the spouses. Isolating this effect is challenging because both men and women’s participation may rise over time due to other factors. Schirle (2008) overcame this constraint by using women’s labor force participation at age 40 to predict their participation at older ages. She estimated that the rise in women’s participation can explain one-quarter of the rise in men’s participation in the United States, one-half in Canada, and one-third in the United Kingdom. Changes in U.S. men’s education and age were estimated to explain about one-third of the rise, leaving 40 percent unexplained.
Changes in employer-provided benefits, including a shift from DB to defined contribution (DC) pensions and a reduction in retiree health insurance, also may have played a role. In DB plans, pension wealth typically grows until the plan’s early or normal retirement age and declines thereafter. DC plans, by contrast, lack age-specific retirement incentives. Friedberg and Webb (2005) estimated that the absence of those incentives leads workers
9 This statement reflects the author’s calculations from the March Current Population Survey, not shown due to space constraints. Although the series are somewhat variable due to small sample sizes, among men ages 60 to 64 there is an increase in participation between 1995 and 2016 of 5 percentage points for high school dropouts, 6 percentage points for high school graduates, and 9 percentage points for male college graduates. The equivalent statistics for women are gains of 3, 7, and 9 percentage points, respectively. Thus, although college graduates have the largest increases, participation has risen substantially in other groups as well. In terms of the shift in educational attainment over time, there is a 20 percentage point increase in the share of men ages 60 to 64 who are college graduates between 1980 and 2016 and a 35 point decrease in the share that are high school dropouts. The analogous changes for women are very similar in magnitude.
with a DC plan to retire nearly 2 years later. A simple calculation suggests that the 25-point drop in the share of private-sector workers with a DB plan from 1980 to 201410 could have led to a roughly 6-month increase in the average retirement age, one-quarter of the total increase reported by Munnell (2015).
Public pension reforms in the United States and other developed countries also may have led to later retirement. The most significant changes in the United States are an increase in the FRA from 65 to 66, an increase in the benefit of delaying claiming after the FRA, and the elimination of the Social Security earnings test after the FRA. Mastrobuoni (2009) estimated that the FRA increase led to a 6-month increase in the mean retirement age. Most studies of the earnings test find its removal had little effect on retirement, though Gelber et al. (2017) concluded that it affects employment at ages 63 to 64. Many other developed countries have made reforms to their public pensions, raising eligibility ages and reducing benefits due to fiscal concerns. In Italy, Japan, and the United Kingdom, increases in eligibility ages were followed by sharp increases in employment at the affected ages (Banks et al., 2017; Brugiavini et al., 2016; Oshio et al., 2016).
While it is difficult to apportion precise shares to the various factors, available evidence suggests that changes in education, women’s growing role in the economy, the shift from DB to DC pensions, and changes to public pensions all likely played some role in the trend of longer work lives. Although there have been considerable advances in mortality and health over time, the evidence of their contribution is less clear.
Reconciling Trends in Health and Work
The rise in prevalence of DI receipt in the United States runs counter to the trends of longer work lives, rising life expectancy, and (arguably) improving health at older ages. Research on the DI program helps to explain this discordance. Changes in population health have not been identified as a major driver of rising DI receipt. Rather, the 1984 law liberalizing medical eligibility is a key contributor, particularly for men during the late 1980s. Rising women’s participation and population aging contributed to a rise in the share of nonelderly adults on DI but do not explain increases for men within a narrow age range.
The 1984 law was followed by significant shifts in the composition of the DI population. Awards rose sharply for musculosketal and mental conditions, which are often viewed as more difficult to verify, but were flat or declining for cancer and heart disease (Autor and Duggan, 2006). This pat-
10 Employment Benefit Research Institute, see https://www.ebri.org/publications/benfaq/index.cfm?fa=retfaqt14fig1 [April 2018].
tern is consistent with greater leniency in the evaluation of applicants with the former conditions, even as population health held steady or improved. Alternatively, it could reflect an increase over time in the incidence of musculosketal and mental conditions or that labor market conditions for low-skill workers have changed so as to make it more attractive for those who have these conditions to apply (Liebman, 2015). Consistent with the latter theory, Autor and Duggan (2003) found that low-skill workers facing a negative shock to labor demand were more likely to exit the labor force to DI after the 1984 law liberalizing eligibility.
Overall, research from the United States and other countries suggests that reforms to the medical screening process and other program parameters have substantial effects on DI participation. These reforms interact with health and work, in that they may facilitate entry onto the program for those who have certain medical conditions or those with poor employment prospects. Thus, trends over time in health or work can magnify or diminish the initial effect of a reform.
DISCUSSION AND DIRECTIONS FOR FUTURE RESEARCH
The past few decades have been marked by large increases in labor force participation at older ages and an associated rise in the age of retirement. This trend is occurring against a backdrop of increasing life expectancy, yet improving health does not appear to be the primary driver of the trend. Rather, increasing education, women’s greater participation in the economy, the shift from DB to DC pensions, and reforms to public pensions all appear to play a role. There is an increasing use of the DI route to exit the labor force, which also seems to be primarily explained not by changes in population health but by changing medical eligibility criteria and economic factors.
The retirement literature has advanced considerably over this period, making use of new data and benefiting from the “credibility revolution” in empirical economics to bolster knowledge on topics of longstanding interest, such as the effect of Social Security on retirement, as well as to explore new topics, such as joint retirement decision making by couples. Yet while much has been learned, important unanswered questions remain.
One critical question is whether there are forces that could impede the current trend toward longer work lives. Age discrimination is a potential factor, as older workers will be challenged to extend their working lives if employers are reluctant to hire them or if they are treated differently in pay and promotion decisions. The existing literature on this topic provides compelling evidence that older women are less likely to receive a call-back for a job interview. It has proven more difficult to examine how this translates into differences in actual hires and wage offers and also to see if older
workers on the job are treated differently than younger workers, in part because interactions between employers and employees are often not empirically observable. New data and methods (such as the use of “big data”) may provide an opportunity to push further in investigating this topic.
A related research issue is that while the retirement literature has long focused largely on supply-side factors that may affect retirement decisions, labor-demand-side topics merit more attention. Past work has sought to understand age profiles in wages and productivity, including how DB pensions may be part of an informal long-term employment contract; more recently, research has focused on how the business cycle affects retirement. At present, there is a pressing need to explore labor demand in the context of ongoing shifts in the economy. Trends such as sector shifts, technological change and automation, and the rise of alternative work arrangements (such as use of temporary and contract workers) may change the relationship between age and productivity, posing new challenges or opportunities for older workers.
A topic at the intersection of labor supply and demand that also merits further attention is the conditions of work. A job is defined not only by the tasks a worker performs in exchange for a wage and benefits package but also by many nonmonetary attributes. These might include, for example, whether the worker is able to set his or her own hours, shift to a part-time schedule, work remotely, receive training and opportunities for professional growth, or enjoy paid time off. Job attributes such as these (or their lack) may influence workers’ decisions to remain on the job at older ages. Those attributes that contribute to a worker’s ability to manage health problems could be more important for workers in poor health. Historically there has been little information on the conditions of work in the United States, but the 2015 American Working Conditions Survey provides such data and is just beginning to be used by researchers (Maestas et al., 2017). More work is needed to understand how job attributes are valued by older workers and affect their labor supply and whether firms are adapting jobs to match these preferences.
Finally, there is an increasing awareness of the importance of looking beyond population averages at the experience of groups of different SES. Those with less education are known to retire earlier, have shorter life expectancies, and be more likely to use DI. Trends over time by SES within the older population are sometimes more difficult to discern, but with respect to life expectancy, there is a clear trend toward rising inequality (National Academies of Sciences, Engineering, and Medicine, 2017). More research on trends in health and work at older ages by SES is needed, as well as research examining whether workers’ responsiveness to labor supply factors such as pension incentives varies by SES and whether labor demand differs by SES. Given the long-run fiscal challenges facing public pensions
and other programs for the elderly, such programs are likely to continue to evolve in a way that both encourages and includes an expectation of longer work lives. If various segments of the population are less able to respond to incentives to work longer, reforms may have unanticipated impacts on the well-being of older individuals.
Angrist, J.D., and Pischke, J.-S. (2010). The credibility revolution in empirical economics: How better research design is taking the con out of econometrics. Journal of Economic Perspectives, 24(2), 3–30.
Autor, D.H., and Duggan, M.G. (2003). The rise in the disability rolls and the decline in unemployment. Quarterly Journal of Economics, 118(1), 157–206.
Autor, D.H., and Duggan, M.G. (2006). Disability rolls: A fiscal crisis unfolding. Journal of Economic Perspectives, 20(3), 71–96.
Baker, M. (2002). The retirement behavior of married couples: Evidence from the spouse’s allowance. Journal of Human Resources, 37(Winter), 1–34.
Banks, J., Blundell, R., and Emmerson, C. (2015). Disability benefit receipt and reform: Reconciling trends in the United Kingdom. Journal of Economic Perspectives, 29(2), 173–190.
Banks, J., Emmerson, C., and Tetlow, G. (2017). Long-run trends in the economic activity of older people in the UK. Unpublished paper. Available: http://www.nber.org/chapters/c14051.pdf [April 2018].
Bell, F.C., and Miller, M.L. (2005). Life Tables for the United States Social Security Area, 1900–2100 (Social Security Administration Actuarial Study No. 120, SSA Publication No. 11-11536, August). Washington, DC: Social Security Administration.
Bound, J. (1991). Self-reported objective measures of health in retirement models. Journal of Human Resources, 26, 103–138.
Bound, J., and Waidmann, T. (2007). Estimating the Health Effects of Retirement (Michigan Retirement Research Center Working Paper No. 2007-168). Ann Arbor: Michigan Retirement Research Center.
Bound, J., Schoenbaum, M., and Waidmann, T. (1995). Race and education differences in disability status and labor force attachment in the Health and Retirement Study. Journal of Human Resources, 30(Supp.), S227–S267.
Bound, J., Schoenbaum, M., Stinebrickner, T.R., and Waidmann, T. (1999). The dynamic effects of health on the labor force transitions of older workers. Labour Economics, 6(2), 179–202.
Brown, J.R., Coile, C.C., and Weisbenner, S.J. (2010). The effect of inheritance receipt on retirement. Review of Economics and Statistics, 92(2), 425–434.
Brown, J.R., Kapteyn, A., and Mitchell, O.S. (2016). Framing and claiming: How information-framing affects expected Social Security claiming behavior. Journal of Risk and Insurance, 83(1), 139–162.
Brugiavini, A., Pasini, G., and Weber, G. (2016). Employment at Older Ages: Evidence from Italy. Unpublished paper. Available: http://www.nber.org/chapters/c14046.pdf [April 2018].
Burtless, G., and Moffitt, R. (1986). Social Security, earnings tests, and age at retirement. Public Finance Quarterly, 14, 3–27.
Cahill, K.E., Giandrea, M.G., and Quinn, J.F. (2015). Retirement patterns and the macroeconomy, 1992–2010: The prevalence and determinants of bridge jobs, phased retirement, and reentry among three recent cohorts of older Americans. Gerontologist, 55(3), 384–403.
Case, A., and Deaton, A. (2015). Rising morbidity and mortality in midlife among White non-Hispanic Americans in the 21st century. Proceedings of the National Academy of Sciences of the United States of America, 112(49), 15078–15083.
Chan, S., and Stevens, A.H. (2001). Job loss and employment patterns of older workers. Journal of Labor Economics, 19(2), 484–521.
Charles, K. (2004). Is retirement depressing? Labor force inactivity and psychological wellbeing in later life. Research in Labor Economics, 23, 269–299.
Chen, S., and van der Klaauw, W. (2008). The work disincentive effects of the Disability Insurance Program in the 1990s. Journal of Econometrics, 142(2), 757–784.
Choi, J.J., Laibson, D., Madrian, B.C., and Metrick, A. (2004). For better or for worse: Default effects and 401(k) savings behavior. In D.A. Wise (Ed.), Perspectives on the Economics of Aging (pp. 81–126). Chicago: University of Chicago Press.
Coile, C.C. (2004). Retirement incentives and couples’ retirement decisions. B.E. Press Journal of Economic Analysis & Policy, 4(1), 1–30.
Coile, C.C. (2015). Economic determinants of workers’ retirement decisions. Journal of Economic Surveys, 29(4), 830–853.
Coile, C.C., and Gruber, J. (2007). Future Social Security entitlements and the retirement decision. Review of Economics and Statistics, 89(2), 234–246.
Coile, C.C., and Levine, P.B. (2007). Labor market shocks and retirement: Do government programs matter? Journal of Public Economics, 91(10), 1902–1919.
Coile, C.C., and Levine, P.B. (2010). Reconsidering Retirement: How Losses and Layoffs Affect Older Workers. Washington, DC: Brookings Institution Press.
Coile, C.C., and Levine. P.B. (2011a). The market crash and mass layoffs: How the current economic crisis may affect retirement. The B.E. Press Journal of Economic Analysis & Policy, 11(1), 1–22.
Coile, C.C., and Levine, P.B. (2011b). Recessions, retirement, and Social Security. American Economic Review, 101(3), 23–28.
Coile, C., Milligan, K.S., and Wise, D.A. (2016). Introduction and summary. In D.A. Wise (Ed.), Social Security Programs and Retirement Around the World: Disability Insurance Programs and Retirement (pp. 1–44). Chicago: University of Chicago Press.
Coile, C., Milligan, K.S., and Wise, D.A. (2017). Health capacity to work at older ages: Evidence from the U.S. In D.A. Wise (Ed.), Social Security Programs and Retirement Around the World: The Capacity to Work at Older Ages (pp. 359–394). Chicago: University of Chicago Press.
Council of Economic Advisors. (2016). The Long-Term Decline in Prime-Age Male Labor Force Participation. Washington, DC: Executive Office of the President of the United States. Available: https://obamawhitehouse.archives.gov/sites/default/files/page/files/20160620_cea_primeage_male_lfp.pdf [March 2018].
Costa, D. (1995). Pensions and retirement: Evidence from Union Army veteran. Quarterly Journal of Economics, 110(2), 297–319.
Costa, D. (1998). The Evolution of Retirement: An American Economic History, 1880–1990. Chicago: University of Chicago Press.
Crimmins, E.M. (2004). Trends in the health of the elderly. Annual Review of Public Health, 25, 79–98.
Crimmins, E.M., and Beltrán-Sánchez, H. (2010). Mortality and morbidity trends: Is there compression of morbidity? Journals of Gerontology, Series B: Social Sciences 66B(1), 75–86.
Cutler, D.M., and Wise, D.A. (2008). Health at Older Ages: The Causes and Consequences of Declining Disability Among the Elderly. Chicago: University of Chicago Press.
Cutler, D.M., Ghosh, K., and Landrum, M.B. (2014). Evidence for significant compression of morbidity in the elderly U.S. population. In D.A. Wise (Ed.), Discoveries in the Economics of Aging (pp. 21–51). Chicago: University of Chicago Press.
Disney, R., Emmerson, C., and Wakefield, M. (2006). Ill health and retirement in Britain: A panel data-based analysis. Journal of Health Economics, 25(4), 621–649.
Dorn, D., and Sousa-Poza, A. (2010). “Voluntary” and “involuntary” early retirement: An international analysis. Applied Economics, 42(4), 427–438.
Fahle, S., and McGarry, K. (2018). Women working longer: Labor market implications of providing family care. In C. Goldin and L.F. Katz (Eds.), Women Working Longer: Increased Employment at Older Ages (pp. 157–181). Chicago: University of Chicago Press.
Fetter, D.K., and Lockwood, L.M. (forthcoming). Government old-age support and labor supply: Evidence from the Old Age Assistance Program. Submitted to American Economic Review.
Fitzpatrick, M., and Moore, T. (2017). The Mortality Effects of Retirement: Evidence from Social Security Eligibility at Age 62 (NBER Working Paper 24127). Cambridge, MA: National Bureau of Economic Research.
Flippen, C., and Tienda, M. (2000). Pathways to retirement: Patterns of labor force participation and labor market exit among the pre-retirement population by race, Hispanic origin, and sex. Journals of Gerontology Series B: Social Sciences, 55B(1), S14–S27.
French, E. (2005). The effects of health, wealth, and wages on labor supply and retirement behavior. Review of Economic Studies, 72(2), 395–427.
French, E., and Song, J. (2014). The effect of disability insurance receipt on labor supply. American Economic Journal: Economic Policy, 6(2), 291–337.
Friedberg, L., and Webb, A. (2005). Retirement and evolution of the pension structure. Journal of Human Resources, 40(2), 281–308.
Fries, J.F. (1980). Aging, natural death, and the compression of morbidity. New England Journal of Medicine, 303, 130–135.
Gelber, A.M., Isen, A., and Song, J. (2016). The Effect of Pension Income on Elderly Earnings: Evidence form Social Security and Full Population Data (NBER Working Paper). Unpublished manuscript. Available: https://gspp.berkeley.edu/assets/uploads/research/pdf/gelberisensong051716.pdf [April 2018].
Gelber, A.M., Jones, D., Sacks, D.W., and Song, J. (2017). Using Kinked Budget Sets to Estimate Extensive Margin Responses: Method and Evidence from the Social Security Retirement Earnings Test (NBER Working Paper 23362). Cambridge, MA: National Bureau of Economic Research.
Goldin, C., and Katz, L.F. (2018). Women working longer: Facts and some explanations. In C. Goldin and L.F. Katz (Eds.), Women Working Longer: Increased Employment at Older Ages (pp. 11–53). Chicago: University of Chicago Press.
Gruber, J., and Madrian, B.C. (2004). Health insurance, labor supply, and job mobility: A critical review of the literature. In C. McLaughlin, (Ed.), Health Policy and the Uninsured (pp. 97–178). Washington, DC: Urban Institute Press.
Gruber, J., and Wise, D.A. (1999). Social Security and Retirement Around the World. Chicago: University of Chicago Press.
Gruber, J., and Wise, D.A. (2004). Social Security Programs and Retirement Around the World: Microestimation. Chicago: University of Chicago Press.
Gruenberg, E.M. (1977). The failures of success. Milbank Memorial Fund Quarterly: Health and Society, 1977, 3–24.
Gustman, A.L., and Steinmeier, T.L. (1986). A structural retirement model. Econometrica, 54(3), 555–584.
Gustman, A.L., and Steinmeier, T.L. (2000). Retirement in dual-career families: A structural model. Journal of Labor Economics, 18(3), 503–545.
Gustman, A.L., and Steinmeier, T.L. (2005). The Social Security early entitlement age in a structural model of retirement and wealth. Journal of Public Economics, 89(2–3), 441–463.
Gustman, A.L., and Steinmeier, T.L. (2009). How changes in Social Security affect recent retirement trends. Research on Aging, 31(2), 261–290.
Gustman, A.L., and Steinmeier, T.L. (2015). Effects of Social Security policies on benefit claiming, retirement, and saving. Journal of Public Economics, 129(C), 51–62.
Hayward, M.D., Friedman, S., and Chen, H. (1996). Race inequities in men’s retirement. Journals of Gerontology: Social Sciences, 51B(1), S1–S10.
Hurd, M.D., Reti, M., and Rohwedder, S. (2009). The effect of large capital gains or losses on retirement. In D.A. Wise (Ed.), Developments in the Economics of Aging (pp. 127–163). Chicago: University of Chicago Press.
Hutchens, R. (2010). Worker characteristics, job characteristics, and opportunities for phased retirement. Labour Economics, 17(6), 1010–1021.
Imbens, G.W., Rubin, D.B., and Sacerdote, B. (2001). Estimating the effect of unearned income on labor earnings, savings, and consumption: Evidence from a survey of lottery players. American Economic Review, 91(4), 778–794.
Johnson, R.W., Kawachi, J., and Lewis, E.K. (2009). Older Workers on the Move: Re-careering in Later Life (AARP Public Policy Institute No. 2009-09, April). Washington, DC: AARP Public Policy Institute.
Johnston, D.W., and Lee, W.-S. (2009). Retiring to the good life? The short-term effects of retirement on health. Economic Letters, 103(1), 8–11.
Kantarci, T., and Van Soest, A. (2008). Gradual retirement: Preferences and limitations. De Economist, 156(2), 113–144.
Koning, P., and Lindeboom, M. (2015). The rise and fall of disability insurance enrollment in the Netherlands. Journal of Economic Perspectives, 29(2), 151–172.
Krueger, A.B., and Pischke, J.-S. (1992). The effect of Social Security on labor supply: A cohort analysis of the notch generation. Journal of Labor Economics, 10(4), 412–437.
Lahey, J.N. (2008a). Age, women, and hiring: An experimental study. Journal of Human Resources, 43(1), 30–56.
Lahey, J.N. (2008b). State age protection laws and the age discrimination in employment act. Journal of Law and Economics, 51(3), 433–460.
Lazear, E.P. (1983). Pensions as severance pay. In Z. Brodie and J.B. Shoven (Eds.), Financial Aspects of the United States Pension System (pp. 57–90). Chicago: University of Chicago Press.
Liebman, J.B. (2015). Understanding the increase in disability insurance receipt in the United States. Journal of Economic Perspectives, 29(2), 123–150.
Liebman, J.B., Luttmer, E., and Seif, D. (2009). Labor supply responses to marginal social security benefits: Evidence from discontinuities. Journal of Public Economics, 93(11–12), 1208–1223.
Maestas, N. (2001). Labor, Love, and Leisure: Complementarity and the Timing of Retirement by Working Couples. Berkeley, CA: University of California, Berkeley.
Maestas, N. (2010). Back to work: Expectations and realizations of work after retirement. Journal of Human Resources, 45(3), 718–748.
Maestas, N., Mullen, K.J., and Strand, A. (2013). Does Disability Insurance receipt discourage work? Using examiner assignment to estimate causal effects of SSDI receipt. American Economic Review, 103(5), 1797–1829.
Maestas, N., Mullen, K.J., Powell, D., von Wachter, T., and Wenger, J.B. (2017). The Value of Working Conditions in the United States. Unpublished paper. Available: http://crr.bc.edu/wp-content/uploads/2017/08/7b.-Kathleen-Mullen.pdf [April 2018].
Manton, K.G. (1982). Changing concepts of morbidity and mortality in the elderly population. Milbank Memorial Fund Quarterly: Health and Society, 60, 183–244.
Mastrobuoni, G. (2009). Labor supply effects of the recent Social Security benefit cuts: Empirical estimates using cohort discontinuities. Journal of Public Economics, 93(11–12), 1224–1233.
McClellan, M.B. (1998). Health events, health insurance, and labor supply: Evidence from the Health and Retirement Survey. In D.A. Wise (Ed.), Frontiers in the Economics of Aging (pp. 301–350). Chicago: University of Chicago Press.
Munnell, A.H. (2015). The Average Retirement Age: An Update (Issue Brief No. 15-4). Boston: Center for Retirement Research, Boston College.
Mutchler, J.E., Burr, J.A., Pienta, A.M., and Massagli, M.P. (1997). Pathways to labor force exit: Work transitions and work instability. Journals of Gerontology, Series B: Social Sciences, 52B(1), S4–S12.
National Academies of Sciences, Engineering, and Medicine. (2017). The Growing Gap in Life Expectancy by Income: Implications for Federal Programs and Policy Responses. Washington, DC: The National Academies Press.
Neumark, D., and Stock, W.A. (1999). Age discrimination laws and labor market efficiency. Journal of Political Economy, 107, 1081–1125.
Neumark, D., Burn, I., and Button, P. (2015). Is It Harder for Older Workers to Find Jobs? New and Improved Evidence from a Field Experiment (NBER Working Paper 21669), Cambridge, MA: National Bureau of Economic Research.
Nyce, S., Schieber, S., Shoven, J.B., Slavov, S., and Wise, D.A. (2013). Does retiree health insurance encourage early retirement? Journal of Public Economics, 104(C), 40–51.
Pattison, D., and Waldron, H. (2013). Growth in new disabled-worker entitlements, 1970–2008. Social Security Bulletin, 73(4).
Oshio, T., Usui, E., and Shimizutani, S. (2016). Labor Force Participation of the Elderly in Japan. Unpublished paper. Available: http://www.nber.org/chapters/c14047.pdf [April 2018].
Quinn, J.F., and Burkhauser, R.V. (1994). Retirement and labor force behavior of the elderly. In L.G. Martin and S.H. Preston (Eds.), Demography of Aging (pp. 50–101). Washington, DC: National Academy Press.
Ramnath, S., Shoven, J.B., and Slavov, S.N. (2017). Pathways to Retirement through Self-Employment (NBER Working Paper 23551, June). Cambridge, MA: National Bureau of Economic Research.
Reday-Mulvay, G. (2000). Gradual retirement in Europe. Journal of Aging & Social Policy, 11(2–3), 49–60.
Rohwedder, S., and Willis, R.J. (2010). Mental retirement. Journal of Economic Perspectives, 24(1), 119–138.
Samwick, A.A. (1998). New evidence on pensions, Social Security, and the timing of retirement. Journal of Public Economics, 70, 207–236.
Schirle, T. (2008). Why have the labor force participation rates of older men increased since the mid-1990s? Journal of Labor Economics, 26(4), 549–594.
Skirbekk, V. (2004). Age and individual productivity: A literature survey. Vienna Yearbook of Population Research, 2, 133–153.
Steuerle, E., and Spiro, C. (1999). Adjusting for Life Expectancy in Measures of Labor Force Participation (Urban Institute Straight Talk on Social Security and Retirement Policy Brief No. 10). Washington, DC: The Urban Institute.
Stock, J., and Wise, D. (1990). Pensions, the option value of work, and retirement. Econometrica, 58, 1151–1180.
Sullivan, D., and von Wachter, T. (2009). Job displacement and mortality: An analysis using administrative data. Quarterly Journal of Economics, 124(3), 1265–1306.
Waidmann, T., Bound, J., and Schoenbaum, M. (1995). The illusion of failure: Trends in the self-reported health of the U.S. elderly. Milbank Quarterly, 73(2), 253–288.
Weaver, D.A. (1994). The work and retirement decisions of older women: A literature review. Social Security Bulletin, 57(1), 3–24.
Wise, D.A. (2012). Social Security Programs and Retirement Around the World: Historical Trends in Mortality and Health, Employment, and Disability Insurance Participation and Reforms. Chicago: University of Chicago Press.
Wise, D.A. (2016). Social Security Programs and Retirement Around the World: Disability Insurance Programs and Retirement. Chicago: University of Chicago Press.
Zissimopoulos, J.M., and Karoly, L.A. (2007). Transitions to self-employment at older ages: The role of wealth, health, health insurance, and other factors. Labour Economics, 14, 269–295.