to be among the set of policies that support work. In cases where injuries result in permanent disabilities, which can range from partial to full disability (defined in terms lost earnings capacity), indemnity benefits can be indefinite, although often lump-sum settlements are used. Workers’ compensation laws in the United States are set at the state level, and while the core structure of the program is similar across states, there are very important differences in the generosity of benefits and many other aspects of state programs (see, e.g., Rothkin, 2019). Separate federal workers’ compensation programs cover federal employees and workers in specific high-risk occupations, such as coal miners and longshoremen (Sengupta et al., 2012).

One key piece of evidence to keep in mind in thinking about older workers and workplace injuries is that, overall, workplace injuries are declining across all age groups (Coate, 2019), with the sharpest declines among younger workers—a change that is not attributable to changes in the composition of occupations but rather to occupations becoming safer (Restrepo and Shuford, 2011). Still, potentially important age differences remain.

Data from the Bureau of Labor Statistics (BLS) indicate that, in recent years, overall rates of lost work time due to injuries and illnesses do not differ much by age (whereas younger workers used to have higher rates [Restrepo and Shuford, 2011]). However, the evidence sometimes differs. For example, in a large sample of workers’ compensation claims in Australia, Berecki-Gisolf and colleagues (2012) found increased injury rates through age 60, after which they declined a bit. The most recent evidence for the United States (Savych and Ruser, 2019), from a database of workers’ compensation claims covering a large share of the United States, indicates that injury rates and the rates of injuries with lost work time are not very different for older workers (and are highest for younger workers). But the nature of injuries differs, with older workers more likely to be injured in falls and to have fractures. And BLS data point to a much higher fatality rate for older workers, for example, the rate is five times as high for workers ages 65+ than for those ages 20–24.

The evidence generally indicates that when older workers are injured on the job, their injuries tend to be more severe (Pransky et al., 2005b; Rogers and Wiatrowski, 2005). In BLS data covering all reported injuries, days away from work rise monotonically with age, from a median of about five days for ages 20–24 to 18 days for ages 65 and over (Rogers and Wiatrowski, 2005). Restrepo and Shuford (2011) show that measuring severity by indemnity payments indicates uniformly higher severity with age (with severity highest in the oldest group included, ages 55–64). These age differences have been persistent over time, and the same is true when severity is measured by medical payments. About half of these differences are due to the more severe injuries older workers experience. For example, older workers are more likely to have rotator cuff and knee injuries and lower-back nerve pain, and younger workers are more likely to have sprains and lower-back pain. In a related study using BLS data and focusing on workers ages 65 and over, Wolf (2010) finds that falls, slips, and trips were the largest cause of injury among older workers in this age group.

On the other hand, the frequency of workers’ compensation claims for this 65+ age group is generally lower, owing to lower employment in the more hazardous and injury-prone occupations. Algarni and colleagues (2015) discuss evidence of a higher prevalence of musculoskeletal injuries and slower recovery among older workers, likely associated in part with age-related physical changes. In data from Australia, Smith and Berecki-Gisolf (2014) document that both age and physical demands at work are associated with more musculoskeletal injuries, and the relationship with age was stronger in more physically demanding jobs, although this tends to attenuate after age 45.¹ In data from 2003, the types of injuries that occur also change with age, with a higher share of fractures among those ages 55–64 and especially 65 and older, and a corresponding lower share of sprains, strains, and tears. Biddle and colleagues (2003) report a higher rate of permanent disabilities for older workers injured on the job than for younger workers.

The fatality rate jumps substantially for those ages 65 and older, to nearly three times the rate at other ages (and the rate at other ages varies little); see Rogers and Wiatrowski (2005). Berecki-Gisolf and colleagues (2012) also report that days until return to work rise with age, largely plateauing after age 45, while recurrences of injuries

___________________

¹ One factor the authors emphasize as potentially mitigating evidence of greater likelihood of injury among older workers is the “healthy worker effect,” in which, especially with age, only the healthiest workers tend to select (or get selected into) physically-demanding jobs. It remains on open question whether increases in employment among older workers are likely to reinforce this effect, with healthier workers choosing to continue working, or to undercut it—if more older workers choose to continue working of necessity.

increase into the 50s and then decline somewhat. Looking at Colorado workers’ compensation claims, Schwatka and colleagues (2012) find rising medical and indemnity costs, owing in part to more severe injuries, among construction workers. Some evidence indicates, not surprisingly, that age differences in preexisting health conditions contribute to the greater severity of workplace injuries among older workers, looking at musculoskeletal injuries (Smith et al., 2014).²

The most recent data (from Savych and Ruser, 2019) point to higher indemnity payments with age (although with declines for those aged 65 and older), due to longer duration of temporary disability, greater likelihood of permanent disability payments (and higher amounts), and higher pre-injury wages (which influence indemnity payments via the replacement rate). Medical payments generally increase with age. It’s also true that recovery tends to decline with age, but only slightly, and the percentage of workers without “substantial” return to work increases with age.

There is less research about age differences in treatment for injuries covered by workers’ compensation. This may be relevant, because some types of common treatments, or treatments for common injuries, are controversial. For example, there has been rapid growth in opioids prescribing for workers’ compensation injuries, and opioid prescriptions, including for longer-term use, are frequent.³ But longer-term opioids prescribing is associated with slower return to work (Savych et al., 2019), and opioids prescribing increases with age (Thumula and Liu, 2018). Back injuries are very common for workers and are a major cost of workers’ compensation. Treatment for back injuries—in particular, whether to use surgery—is highly controversial (e.g., Maghout-Juratli et al., 2006). We are not aware of evidence on differences in treatment and efficacy by age.

Workplace injuries among older workers—especially given their greater severity—may lead to retirement and participation in other public disability programs. With regard to retirement, Pransky and colleagues (2005a) studied workers who filed workers’ compensation claims with lost work time, and found that among those ages 55 and over, 11 percent planned to retire earlier due to a workplace injury; this was more likely among those with prior health problems. Scott and colleagues (2018) report similar evidence for Canada, for a subsample of those with permanent impairments (for whom this response seems particularly likely).

Burton and Spieler (2001) identified the problem of increased participation in public disability programs early, in light of what they saw as declining availability of permanent disability benefits. They also pointed out a potential externality problem: state policy to reduce the generosity of these kinds of workers’ compensation benefits could simply shift costs to other, primarily federal, programs. Burton and Spieler also noted that some evidence suggested that workers do turn to SSDI for income support when workers’ compensation benefits are unavailable, based simply on opposite trends in these two programs.⁴ Evidence of injured workers covered by workers’ compensation ending up on SSDI is reported by O’Leary and colleagues (2012), Parent and colleagues (2012), and Sengupta and Baldwin (2015). Aside from the externality problem, this shift could be problematic because workers’ compensation is an experience-rated system, which is intended to create incentives to keep workplaces safe, whereas SSDI is not. The implications for worker well-being are potentially ambiguous. SSDI benefits tend to be lower, although the benefits may be more secure and medical care is provided, as SSDI recipients receive Medicare after 24 months (Burton and Spieler, 2001).

Summary

Older workers do not have higher injury rates or more frequent workers’ compensation claims than younger workers. However, their injuries are more severe, resulting in higher payments, longer periods of benefit receipt, and higher fatality rates. Older workers who are injured on the job and receive workers’ compensation benefits, particularly those with preexisting health problems, are more likely to report that they plan to retire earlier than

___________________

² One conflicting study is Pransky and colleagues (2005b), who study a sample of workers with workplace injuries that entailed lost work time, comparing those ages 55 and over to those younger than 55, matching on gender, injury code, and injury date. Older workers report more severe injuries, and (not surprisingly) had more preexisting comorbidities. But they did not have worse outcomes (looking at time until return to work, change in ability to perform job, post-injury pain, and many other outcomes).

³ Some recent reforms appear to be reducing the prescribing of opioids (e.g., Thumula et al., 2017).

⁴ See Mont and colleagues (2000) and recent work on this question by McInerney and Simon (2012) and Buffie and Baker (2015).

they previously intended. There is some tendency of injured workers receiving workers’ compensation benefits to transition to SSDI.

AGE-SPECIFIC POLICIES THAT SUPPORT WORK

The policies discussed so far may be of particular importance for older workers, but they cover all workers. We now turn to a set of policies that apply exclusively to older workers. Older workers are a group of special concern due to their potential to be negatively impacted by age discrimination in the work force, as discussed in Chapter 6; age-discrimination legislation may mitigate this risk. Similarly, the elimination of mandatory retirement has made it easier for U.S. workers to work in old age, even as such policies remain in force in many other countries. If many workers prefer gradual retirement, as suggested in Chapter 3, it is worth considering whether public policies can support that option. Finally, job training for older workers merits discussion, as the need to help workers retool for available jobs will need to be balanced against the generally lower number of remaining work years when considering the value of training for this group.

Age Discrimination in Employment Act (ADEA)

The federal Age Discrimination in Employment Act (1967) protects people over the age of 40 from discrimination in hiring (including prohibiting age-related advertisements), firing, promotion, layoff compensation, benefits, job assignments, and training. In addition, most states have their own age discrimination laws, which may differ from the federal law, for example by allowing for larger damages or applying to workers at smaller firms.

The ADEA affects organizations that regularly employ 20 or more employees. The current law also protects against mandatory retirement for most occupations (Lahey, 2008). Exemptions to the ADEA include a “bona fide occupational qualification,” or BFOQ, that is directly related to age, for example, in an acting position. In practice, the courts have also allowed age to be considered a BFOQ in cases where public safety may be affected, including occupations such as pilots, air traffic controllers, and bus drivers. The federal law also exempts high-salaried policy-making positions from age discrimination law.

Unlike the Civil Rights Act, the ADEA does not allow damages for emotional pain and suffering or for punitive damages. Damages are limited to “make whole” damages and lawyer fees. Although these awards can include requirements to hire, reinstate, promote, provide back pay, and restore benefits, in practice most of what is awarded under the ADEA goes to paying lawyer fees. Additional damages are awarded only in rare cases in which the defendant has willfully violated the law, and these damages are limited to twice the amount of actual damages (Lindemann and Kadue, 2003; O’Meara, 1989). Initial examinations of cases brought under the ADEA showed that, because of these limits on the awards under the ADEA, most plaintiffs were White male middle managers in their 50s who had lost sizable salaries and benefits (O’Meara, 1989). The average amount awarded per ADEA suit has only recently increased and become closer in size to awards for Title VII suits (Causey and Lahey, 2013).

A strict interpretation of the ADEA allows only for cases of disparate treatment, requiring proof of intentional discrimination. However, another interpretation of ADEA allows for disparate-impact cases, in which a policy indirectly impacts a protected group differently than the unprotected group. In ADEA cases, workplace decisions that are based on seniority or wages—for example, firing those with the highest salaries or who have the longest tenure at the firm—are considered to have a disparate impact on older workers. Initially, the Equal Employment Opportunity Commission (EEOC) held the position that an employment practice that had a disparate impact on individuals within the protected age group could not be considered to be due to a “reasonable factor other than age” unless it was justified as a business necessity. However, in Smith v. City of Jackson,⁵ the United States Supreme Court held that the ADEA authorizes recovery for disparate-impact claims of discrimination and that the “reasonable factors other than age” test, rather than a “business necessity” test, is the appropriate standard for determining the legality of practices that disproportionately affect older individuals.

___________________

⁵Smith v. City of Jackson, 544 U.S. 228, 241 (2005).

___________________

⁶Gross v. FBL Financial Services Inc., 557 U.S. (2009).

⁷ See https://www.congress.gov/bill/116th-congress/house-bill/1230/text

⁸Equal Employment Opportunity Commission v. Professional Endodontics, P.C., 4:17-cv-13466 District Court, E. D. Michigan (2018).

people taking bridge jobs or moving to partial retirement, possibly as a result of declining health (Cahill et al., 2006; Johnson, 2014; Johnson et al., 2009; Maestas, 2010). As more workers continue working into older ages, the demand for less physically demanding jobs will likely result in an increase in partial retirement. Second, the most rigorous evidence establishing age discrimination—from correspondence studies such as those discussed in Chapter 6—is focused on discrimination in hiring.

Third, both empirical evidence (Adams, 2004; Neumark and Stock, 1999) and the workings of age discrimination enforcement suggest that, at present, the ADEA is more effective at reducing age discrimination in terminations than in hiring (Neumark, 2009), because in hiring cases it is difficult to identify a class of affected workers, and economic damages are smaller than in termination cases.⁹ Fourth and finally, as a result of their effectiveness in making the termination of older employees more difficult, stronger age discrimination protections can have the unintended consequence of increasing hiring discrimination. In particular, if age discrimination laws less effectively reduce discrimination in hiring than in terminations, they could deter employers from hiring older workers by raising the expected cost of hiring them (Adams, 2004; Bloch, 1994; Posner, 1995). Thus, determining the effects of age discrimination laws on the hiring older workers is of great importance.

Earlier studies that use state-level variation in age-discrimination laws to examine the effect of these laws on the hiring of older adults find no evidence that these laws are effective in improving the hiring of older workers (Adams, 2004; Lahey, 2008). Instead, they find that such laws might have had the opposite effect and reduced hiring, particularly among those ages 65 and over (Adams, 2004; Lahey, 2008). However, these earlier studies either rely on data that measure employment overall and are not well suited to directly measuring changes in hiring (Adams, 2004) or their findings are also consistent with alternative interpretations that could suggest that these laws led to increases in hiring (Lahey, 2008; Neumark, 2009).

More recently, Neumark and Song (2013) used variations across states in older workers’ responses to the changes in Social Security that were phased in between 2003 and 2008 for cohorts born after 1938—a reduction in benefits for those retiring early at age 62 and a gradual increase in the age of full retirement from 65 to 66—to estimate the impact of age discrimination laws. They found that in states that had stronger age discrimination laws (those that allowed for larger damages and/or extended coverage to small firms exempt from federal age discrimination law), older workers were more likely to work longer and claim Social Security benefits later, and that these increases in employment were due to increases in the hiring of older workers. In particular, employment increased for those between the original age of full retirement (age 65) and the new, higher age (by 2008, it reached age 66), the specific age range targeted by these reforms to Social Security. This study points more directly to the ways in which policy can jointly create supply-side incentives to encourage older workers to extend their work lives and demand-side reductions in age discrimination.

Subsequent research using correspondence studies (discussed in Chapter 6) has confirmed and extended the finding that state age discrimination laws that allow plaintiffs to seek increased damages (compared to the federal ADEA) reduce age discrimination in hiring (Neumark et al., 2019; Neumark et al., 2019). In one such study covering retail hiring in all U.S. states, the authors find evidence of lower levels of discrimination against older men and women when state age-discrimination laws allow plaintiffs to file for larger damages (Neumark et al., 2019). Because employers in these states could expect higher costs associated with dismissing older workers, this evidence more strongly rebuts the hypothesis that stronger age discrimination laws may lead employers to reduce their hiring of older workers.

However, research has not consistently found that stronger age discrimination laws increase the hiring of older workers. There is some contrary evidence examining these effects during the economic downturn of the Great Recession (Neumark and Button, 2014). During this period, there is very little evidence that stronger-age discrimination protections improved employment rates or reduced unemployment rates or durations among older workers relative to younger workers, and in some cases the evidence suggests that stronger age discrimination protections were associated with more-adverse effects on older workers (Neumark and Button, 2014). It is possible that severe labor market disruptions that involve significant layoffs, such as the Great Recession, make it harder to discern discrimination in hiring and terminations, which weakens the benefits of stronger state protections for age-discrimination. It is also possible

___________________

⁹ That said, the studies of the effectiveness of the ADEA examine somewhat different environments (e.g., the Great Recession in Neumark and Button, 2014; recessions vs. other periods in Dahl and Knepper, 2020; and earlier state laws and the federal laws in Lahey, 2008).

that the potentially higher costs of terminating older workers loom larger for employers in periods of economic uncertainty, reducing their willingness to hire older workers. However, these results are less a contradiction of other findings suggesting positive effects of age discrimination laws than specific to a highly unusual labor market period.

One interesting observation is that the few studies that have examined differences in the effectiveness of age discrimination laws by gender and race/ethnicity have not found similar effects for either women or racial-ethnic minorities. Specifically, neither Lahey (2008) nor McLaughlin (2020) found effects of ADEA laws for women. Lahey (2008) also did not find effects of ADEA laws for minorities. This could in part be due to the limits on damages that can be sought by plaintiffs under the ADEA. As noted earlier, most early ADEA claims were filed by older White male middle-managers with substantial salary and benefit losses; because older women and minorities are historically less likely achieve these high-paying jobs with extensive benefits, the high cost of litigation relative to the potential award may discourage these groups from filing cases. As another potential explanation of these null findings, Delaney and Lahey (2019) argue that although older women and minorities experience discrimination—the source of which is not always obvious to the target—the ADEA and Title VII employment laws were separate laws and thus limited intersectional claims between age, race, and gender. Nevertheless, this explanation has not received rigorous empirical testing yet.

Summary

In general, the balance of the evidence suggests that stronger state age-discrimination laws improve the hiring and retention of older workers, particularly older White men. Beyond this, there is even clearer evidence that stronger age-discrimination laws do not produce the unintended consequence of creating incentives for employers to reduce their hiring of older workers—with the possible exception of during unusual periods of economic disruption. Thus, one might argue that, at a minimum, these laws do no harm.

Affirmative Action

A potential policy change that would go beyond age discrimination protection would be to extend affirmative action mandates for federal contractors to include older workers. This would require federal contractors to demonstrate that they have developed and enacted measures to ensure that they employ older workers throughout their organization. One rationale for such a policy is that evidence suggests that age discrimination in hiring is a particularly serious problem, and reducing discrimination is an important way to help society adapt to population aging. One might also argue that age discrimination protections should be on par with protections based on race, sex, ethnicity, and religion. This extension of affirmative action could be done by expanding on previously issued executive orders that require federal contractors to take affirmative action to improve representation of other protected groups to cover age as well.¹⁰ There is evidence that affirmative action policies have been successful in boosting employment for other covered groups (for reviews, see Holzer and Neumark, 2000; and Kurtulus, 2012) and, therefore, would likely improve employment prospects for older workers.

(Elimination of) Mandatory Retirement

In a sense, mandatory retirement is the ultimate form of age discrimination—a policy allowing (if not requiring) the employer to set an age at which an employee must retire, regardless of the employee’s wishes or job performance. A popular theory for the existence of mandatory retirement is that it is a mutually beneficial arrangement for workers and firms (Lazear, 1979). According to this theory, workers are paid less than the value of their marginal product when young and more than this amount when old in order to induce workers whose performance cannot be perfectly monitored to perform at a higher level of effort when young. Mandatory retirement ensures that workers leave the

___________________

¹⁰ The key prior executive orders with regard to hiring and employment include no. 10925 (1961), instructing federal contractors to “take affirmative action to ensure that applicants are treated equally without regard to race, color, religion, sex, or national origin”; no. 11246 (1965), requiring all government contractors to take affirmative action to expand employment opportunities for minorities, establishing the Office of Federal Contract Compliance; and no. 13672 (2014), amending no. 11246 to cover sexual orientation or gender identity.

firm at the previously agreed-upon date. Jolls (1996) posited that the elimination of mandatory retirement would indeed encourage the use of Lazear contracts, a theory tested empirically and affirmed by Neumark and Stock (1999).

Mandatory retirement at age 65 was common in the 1960s and 1970s in the United States, affecting an estimated 40 percent of the male workforce (von Wachter, 2002). While the 1967 ADEA originally protected workers only through age 65, amendments in 1978 and 1986 first outlawed mandatory retirement before age 70 and then abolished the practice. Currently, mandatory retirement exists only in occupations deemed too physically or cognitively challenging for workers to continue beyond a certain age. These occupations include federal law enforcement officers, national park rangers, and firefighters, who generally must retire at age 57 (Costello, 2019); pilots, who must retire by age 65 (previously age 60) and air traffic controllers, who must retire by age 56 (with exceptions up to age 61; Federal Aviation Administration, 2019); employees who enter into contractual agreements at the time of appointment or partnership (for example, lawyers); and certain state officials, particularly judges (Sciocchetti, 1991).

During the period when mandatory retirement at age 65 was in effect, workers covered by this policy had a very high incidence of retirement at age 65. The elimination of mandatory retirement at age 65 is estimated to have increased the labor force participation of workers ages 65 to 69 by five to 20 percent relative to then current levels (Morrison, 1988; von Wachter, 2002).

Despite its virtual elimination in the United States, mandatory retirement is still in place in many high-income countries (Organisation for Economic Co-operation and Development [OECD], 2017a). While Australia, Canada, and New Zealand abolished mandatory retirement along with the United States, the only OECD countries within Europe to have done so are the United Kingdom, Denmark, and Poland.¹¹ This is in spite of a recommendation by the European Parliament that European Union (EU) member states “put a ban on mandatory retirement when reaching the statutory retirement age, so as to enable people who can and wish to do so to choose to continue to work beyond the statutory retirement age or to gradually phase in their retirement” (European Parliament, 2013).

Mandatory retirement still exists in other EU countries, set at age 67 in Sweden, age 68 in Finland, age 70 in France, Iceland, and Portugal, and age 72 in Norway, for example. In these countries, when an employee reaches mandatory retirement age the employer must either create a new employment contract or renew the existing one if the employee is to continue their employment relationship. Since there is limited data on this process, it is unclear how often mandatory retirement ages actually prevent older workers from continuing to work.

Although mandatory retirement policies have largely ended in the United States and several other countries, ending these policies in other countries has remained controversial (Organisation for Economic Co-operation and Development, 2017b). Employers in these countries often argue that mandatory retirement improves the efficiency of their businesses. Therefore it can be difficult to provide objective measures of the performance of workers that can justify dismissing those with poor performance, mandatory retirement offers employers a convenient mechanism for terminating these workers, particularly in countries in which employment protection rules are rigid. Mandatory retirement policies are also sometimes urged based on the (unsubstantiated) belief that they benefit younger workers (but see the discussion of this “lump of labor” fallacy in Chapter 7).

An interesting counterpoint to mandatory retirement is Japan’s Elderly Employment Stabilization Law, which requires employers to provide continuous employment until the pension eligibility age.¹² The law increased employment at older ages for employees of larger firms, for whom the law was binding (Kondo and Shigeoka, 2017).

Partial Retirement

Partial retirement refers to a policy that allows workers to select partial pension receipt in combination with part-time work in order to make a gradual exit from the labor force. It is easy to imagine that many workers might elect this option if available; indeed, the popularity of part-time work and bridge jobs in the United States, discussed in Chapters 2 and 3, points to a preference among many for gradual retirement, even in the absence of a natural

___________________

¹¹ A 2007 European Court of Justice ruling (Palacios de la Villa v. Cortefiel Service SA) affirmed that the EU’s Equal Treatment Framework Directive does not preclude a national law allowing compulsory retirement clauses to be contained in collective agreements under certain conditions.

¹² Prior to the law’s passage, many firms had a long-standing policy of mandatory retirement at age 60, leaving workers to retire, seek new employment, or continue working for the same employer under a new contract, if offered by the firm. Once the pension eligibility age was raised beyond 60, individuals could be left without means of support. The law required firms to offer a new contract for employment until the pension age.

way to combine this with partial pension receipt. Some have suggested that the ADEA could be amended to establish a legal right to phased retirement at one’s current job where feasible (Gotbaum and Wolfe, 2018). The federal government has offered partial (or “phased”) retirement since 2014 (U.S. Office of Personnel Management, 2019).

In the United States, there is no formal partial retirement option. An informal option does exist for certain workers to work while receiving only a portion of their full Social Security benefit. Workers who claim benefits prior to the full retirement age are subject to the Retirement Earnings Test, in which earnings beyond an exempt amount trigger a reduction in benefits. Workers are later compensated for the lost benefits in the form of an increased actuarial adjustment, with their benefit amount increased as though they had originally claimed at a later date. A worker younger than the full retirement age who claims benefits and has earnings between the exempt amount and the level at which benefits are reduced to zero will thus receive a partial benefit. In practice, the option to receive a partial pension does not appear to be well understood, given the ample evidence that workers misperceive the Retirement Earnings Test as a pure tax. For example, workers tend to “bunch” at a level of earnings just below the exempt amount (Gelber et al., 2020).

In the absence of any experience with a partial retirement policy in the United States, its effect can be simulated using a model of retirement behavior (Gustman and Steinmeier, 1983). Results of this exercise suggest that workers will retire earlier if they face a minimum hours constraint—for example, if they are required to either work full-time or retire—than if they are given the option to freely set their preferred number of hours.

Partial retirement policies exist in many European countries. An international comparison of nine OECD countries that have introduced such policies suggests that the policies increase labor force participation (extensive margin) but also decrease the average hours worked (intensive margin; Börsch-Supan et al., 2018). Overall, these policies have either had no significant effect on total labor supply or have decreased it.¹³

Summary

While partial retirement policies may have intuitive appeal, it is unclear whether their introduction would lead to an increase or decrease in work at older ages. Although some workers who would have retired early in the absence of a partial retirement option might choose to extend their working lives on a part-time basis if this option were available, some workers who would have otherwise worked full-time might also elect this option, leading to a reduction in their hours worked (Börsch-Supan et al., 2017, 2018). Whether this policy option ultimately leads to more work at older ages is thus an empirical question.

The Senior Community Employment Program

A final type of policy that may support work at older ages is job training. The Senior Community Employment Program (SCSEP) is a training program for older workers based on community service and work, authorized by Congress under Title V of the Older Americans Act (OAA) of 1965. The OAA promotes a range of services that support economic opportunity and health among older adults, including adult day care, senior centers and activities, legal support, health promotion, disease prevention, and transportation. SCSEP is focused on employment and civic engagement for older adults.

SCSEP provides subsidized, part-time, community service work-based training for low-income older individuals that have poor employment prospects (U.S. Department of Labor, 2020a). The program helps to meet urgent needs in the community that “might otherwise go unmet while empowering participants to become self-sufficient, thus avoiding public assistance as they provide essential community services and gain the necessary confidence and job skills for obtaining unsubsidized employment” (National Council on the Aging, 2001, p. iv). Program participants must be at least age 55, be unemployed, and have a family income of no more than 125 percent of the

___________________

¹³ For evidence from a specific country, see Graf et al., (2011) for evidence on Austria; Huber et al., (2013) and Berg et al. (2020) for evidence on Germany; Ilmakunnas and Ilmakunnas (2006) for evidence on Finland; and Brinch et al., (2015) for evidence on Norway. For European studies demonstrating older workers’ preference for reducing their working hours in a gradual shift toward their retirement over full retirement, see Gielen (2009), Büsch and colleagues (2010), and Cihlar et al. (2014).

federal poverty level. Priority is given to persons who are over age 65, have a disability, have low literacy skills or limited English proficiency, reside in a rural area, or have failed to find employment after using services through the American Job Center System. A variety of organizations are involved in administering SCSEP programs.¹⁴

SCSEP enrollees work with a social worker or case manager on an individual employment plan to assess their skills, interests, capabilities, and needs. They are then placed in community service jobs, such as schools, libraries, social service organizations, or senior-service organizations (Collins, 2019). SCSEP participants receive a stipend based on the typical wage for the job in which they are employed or minimum wage, whichever is higher. The average duration of participation should not exceed 27 months, although this may increase to 36 months in circumstances such as high unemployment or for individuals with severe disability or advanced age (Collins, 2019; Halvorsen and Yulikova, 2020a). The program enrolls only a small fraction of the eligible population (Halvorsen and Yulikova, 2020a).

Since 2016, SCSEP’s performance measures have shifted from a person-centered program (Gonyea and Hudson, 2011; National Council on the Aging, 2001) to emphasize unsubsidized employment and earnings after the program (Collins, 2019). Given these performance measures, the U.S. Department of Labor recently evaluated SCSEP as ineffective (U.S. Department of Labor, 2020b), with some discussion of eliminating the program:

Widening the lens provides a more complex picture. SCSEP participants face many barriers to work, including but not limited to health, social, and economic shocks; age discrimination; caregiving; homelessness or being at-risk of homelessness; limited education; limited English proficiency; chronic health conditions; and/or physically demanding lifetime jobs that are a mismatch with their health (Carolan et al., 2020; Gonzales et al., 2019). In spite of these challenges, participants contribute approximately 40 million hours to communities (U.S. Department of Labor, 2020a) and some find unsubsidized employment.

Other program benefits include improvements in social well-being and health. SCSEP participants report improved self-esteem, larger social networks, improved work skills, and a stronger belief that they can find and keep jobs (Aday and Kehoe, 2008; Carolan et al., 2020; Gonzales et al., 2019). Nearly seven out of ten participants report their health has improved because of the program (Gonzales et al., 2019; U.S. Department of Labor, 2020c).¹⁵ These findings underscore the positive relationship between work and health when the person/environmental fit is optimal (Armstrong-Stassen and Ursel, 2009; James et al., 2011; Pak et al., 2018; Slack and Jensen, 2008; Smyer and Pitt-Catsouphes, 2007; Staudinger et al., 2016; Steenstra et al., 2017).

One analysis suggests that the benefits of SCSEP outweigh the costs, with a gain of nearly $3,000 per participant per job obtained (Mikelson, 2017). Further, the demand for workforce programs serving older workers is on the rise with population aging and increases in the desire or need to work longer. This demand will likely grow due to the economic crisis associated with COVID-19 and the need to retool and reskill low-income older adults for emerging jobs post-COVID-19.

Summary

The SCSEP is the only federally funded program to assist low-income older adults with multiple barriers in finding employment. The program provides income support for this vulnerable population and is associated

___________________

¹⁴ Key stakeholders include SCSEP grantees, One-Stop Career Centers, and private-sector adult education and literacy agencies. State agencies and 19 national nonprofit organizations (e.g., AARP, Experience Works, Goodwill Industries International, National Council on the Aging, Center for Workforce Inclusion—formerly Senior Service America, Inc.) make up SCSEP grantees. In most states, the governor selects the State Office on Aging to administer the program.

¹⁵ As Gonzales, Lee, and Harootyan (2019) report, one participant said: “I was depressed. Actually, working over there [SCSEP] helped get me out of my depression. I was happy [laughs], I had a social life and working and I lost weight too . . . physically, mentally, everything. I felt better . . .” (p. 218).

with improvements in social well-being and health but is not currently meeting its statutory goals of fostering economic self-sufficiency and moving low-income seniors into unsubsidized employment.

POLICIES TO SUPPORT THE FINANCIAL SECURITY OF DISABLED AND RETIRED WORKERS

From the perspective of older workers, there are many policies that may affect employment decisions at older ages, only some of which are designed to support work at older ages. The next set of policies we discuss are generally designed to enhance the financial security of disabled and retired workers and protect them against a variety of risks. By providing cash or health insurance benefits upon becoming disabled or reaching a certain age, or by encouraging workers to save for retirement, these policies may enable retirement. Moreover, specific program elements, such as the age of benefit eligibility, may influence the timing of retirement. A number of these programs are focused on low-income individuals, and we pay special attention to them in the discussion below.

Disability Insurance

Disability insurance provides protection against the earnings loss associated with a long-term work-limiting disability. An estimated one-quarter of 20-year-olds are projected to become disabled before reaching retirement age (U.S. Social Security Administration, 2020a). In the United States, workers are covered by SSDI, which has been part of the Social Security program since 1956.

SSDI benefits are available to workers who meet medical eligibility and work requirements. The worker must have sufficient recent and lifetime contributions: five of the prior 10 quarters and at least 40 quarters over the work life, or less if disabled at a young age. An applicant must be determined to have a disability, defined as “the inability to engage in 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.”¹⁶ Applicants must have earnings below the substantial gainful activity level ($1,260 per month in 2020), and beneficiaries whose earnings exceed this level will have their benefits suspended after a trial work period.

The review of an SSDI application can be a lengthy, multistep process. A state Disability Determination office makes the initial decision, but denied applicants have up to four levels of appeals available to them. Although only one-third of applicants have been allowed in the initial determination, nearly two-thirds of original applicants are ultimately awarded benefits (Maestas, Mullen, and Strand, 2013). Once an award is made, benefits are paid (retroactively, if need be) starting five months after disability onset. The benefit calculation is essentially the same as for Social Security retired worker benefits, except that benefits are not reduced for claiming before the full retirement age.¹⁷ Benefits are financed through the Social Security payroll tax. SSDI beneficiaries are also eligible for Medicare after a two-year waiting period. The vast majority of SSDI beneficiaries receive benefits until death or reaching the Social Security full retirement age, when they move to the rolls of that program. Fewer than 10 percent of beneficiaries ever exit SSDI to return to the labor force (Raut, 2017), though a larger share are employed with earnings below the substantial gainful activity level at some point while receiving benefits (Liu and Stapleton, 2011).

The share of nonelderly adults receiving SSDI has risen over time, from 2.2 percent in the late 1970s to 4.6 percent in 2013 (Liebman, 2015). There are a number of possible explanations for the rise in SSDI receipt, including demographic, program, and economic factors.

Population aging is one such factor, since SSDI rates rise with age and the Baby Boom generation has recently aged into the peak ages for SSDI receipt. The rising participation of women in the labor force has increased the fraction of women who are insured for SSDI. There have also been changes in the stringency of medical screening over time, notably a 1984 federal law that instructed examiners to place more weight on applicants’ reported pain and discomfort, relax screening of mental illness, consider multiple nonsevere impairments as constituting a disability even if no single impairment is by itself disabling, and put more weight on medical evidence provided

___________________

¹⁶ Code of Federal Regulations § 404.1505. Basic definition of disability. See https://www.ssa.gov/OP_Home/cfr20/404/404-1505.htm

¹⁷ In the case of younger applicants, fewer years of earnings are used to calculate average earnings.

by the applicants’ own health care providers (Autor and Duggan, 2006). Liebman (2015) assesses the relative importance of these factors in the growth of SSDI from 1985 to 2007, attributing half to the increased award rate (likely related to the 1984 law), one-fifth to population aging, and one-eighth to rising rates of SSDI coverage, primarily among women. The rapid growth over time in awards for mental and musculoskeletal disorders also supports the importance of the 1984 changes in medical screening procedures, as these kinds of awards might be expected to be particularly sensitive to these changes (Autor and Duggan, 2006).¹⁸

In addition to these demographic and legal factors, economic factors may have contributed to the increase. The SSDI replacement rate has risen over time for low-wage workers due the interaction between the benefit formula and rising income inequality (Autor and Duggan, 2006). Autor and Duggan (2003) estimate that the combination of rising replacement rates, reduced medical stringency, and declining demand for less-skilled workers made high school dropouts two to three times more likely to apply for SSDI and nearly twice as likely to exit the labor force in response to demand shocks after 1984. SSDI payments respond counter-cyclically to boom and bust cycles in the coal and oil and gas industries (Charles et al., 2018; Black et al., 2002). Similarly, SSDI applications rise with the unemployment rate (Cutler et al., 2012; Stapleton et al., 1998). While recession-related changes in awards are more muted than changes in applications (Center on Budget and Policy Priorities, 2019), recent evidence suggests that the Great Recession induced nearly one million SSDI applications and 400,000 new awards that would not otherwise have occurred, representing nine percent of all new beneficiaries between 2008 and 2012 (Maestas, Mullen et al., 2018). These studies linking economic conditions to SSDI receipt indicate that the program serves as a safety net for some individuals who meet medical eligibility criteria but would work if jobs were available. Finally, the increase in the Social Security full retirement age from age 65 to age 67 has increased the relative value of SSDI.¹⁹Duggan et al. (2007) estimate that the rise in the Social Security full retirement age led an additional 0.6 percent of men and 0.9 percent of women ages 45 to 64 to be receiving SSDI.

Somewhat surprisingly, the number of SSDI beneficiaries has been falling since 2014, due not only to a rise in people moving off the program due to death or attainment of the full retirement age, but also to a decline in new awards (Congressional Research Service, 2018). Among the hypotheses for this decline are the availability of jobs during the economic expansion between the Great Recession and COVID-19 pandemic, the Baby Boom generation’s beginning to age out of SSDI, the reduced attraction of SSDI’s Medicare benefit following the passage of the Affordable Care Act (ACA), and a decline in the SSDI allowance rate. While determining the relative importance of these factors is difficult, the last one seems important. The total allowance rate at all adjudicative levels fell from 62 percent in 2001 to 48 percent in 2016 (Congressional Research Service, 2018), potentially as a result of changes in the SSDI adjudication process as well as an improving economy (Social Security Technical Panel, 2015, 2019). Negative trends in health at midlife, discussed in Chapter 4, raise the concern that the demand for SSDI benefits may increase in the future (Waidmann et al., 2020).

SSDI receipt varies systematically with a number of individual characteristics, including age. The ratio of beneficiaries to insured workers is 1 percent at ages 30 to 34 vs. 12 percent at ages 55 to 59 and 17 percent at ages 60 to 66 (Center on Budget and Policy Priorities, 2019). There is a strong education gradient in SSDI receipt, with high school dropouts being five to six times as likely to receive benefits at ages 55 to 64 than college graduates (Coile, 2016). While differences in health explain part of this gradient, conditional on health, those with lower education are still about 50 percent more likely to receive SSDI. One factor that could contribute to age and educational differences in SSDI receipt is the medical-vocational grid, which is used in the final step of the SSDI review.²⁰ The

___________________

¹⁸ Some indirect evidence on the role of medical screening comes from international comparisons (Wise, 2012). Among a dozen developed countries, the share of individuals on disability insurance varies a lot across countries and over time, with this variation depending much more on stringency of medical screening than on differences in health.

¹⁹ With a Full Retirement Age (FRA) of 65, a worker claiming Social Security retired-worker benefits at age 62 receives a monthly benefit equal to 80 percent of their Primary Insurance Amount (PIA). Once the FRA rises to age 67, the same age-62 claimant receives 70 percent of PIA. A successful SSDI applicant receives 100 percent of PIA. Thus the relative benefit of being an SSDI recipient vs. retired worker beneficiary is rising with the FRA.

²⁰ If an applicant has met the work requirements and is found to have a condition that is sufficiently severe that is not included in the Listing of Impairments (or deemed to be of equal severity) and who is found to be unable to engage in his or her previous employment but able to undertake some form of work is evaluated using this grid. Age, education, language, and previous work skills are used in the evaluation, making it easier for older or less skilled workers to be awarded benefits than for younger or more skilled workers with the same medical condition.

worker’s age, education, English language proficiency, and work skills factor into the award decision for disabled applicants who do not have a qualifying condition and retain some residual functional capacity (Warshawsky and Marchand, 2015). The probability of an award among applicants reaching the final step jumps by an estimated 9 percentage points at age 50 and by 25 percentage points at age 55 (Chen and van der Klaauw, 2008).

Rates of SSDI receipt are substantially higher for Black people than for White people (Autor and Duggan, 2006), potentially reflecting the effect of racial disparities in health and education. There is also striking geographic variation in rates of SSDI receipt. The sources of this variance are not yet well understood, but demographic and labor market factors appear to play a role (Gettens et al., 2016; Michaud et al., 2019).

Disability insurance benefits may affect labor supply decisions. In an ideal world with perfect medical screening, benefits would be strictly targeted to those who are unable to work, negating the possibility of work disincentives. In reality, this kind of targeting is impossible, both because disability occurs on a spectrum and because determining the work capacity of applicants is quite difficult. In the face of imperfect medical screening, the availability of SSDI benefits may induce some individuals who have work capacity above the substantial gainful activity level to apply for benefits. Some analysts have argued that the SSDI screening process is broken (Autor and Duggan, 2006), although the recent decline in award rates might call for new work on this question.

Estimating the effect of SSDI benefits on labor supply is inherently challenging, because the analyst does not observe the counterfactual—how much the SSDI recipient would have worked in the absence of the program. Comparing outcomes of SSDI recipients to the population as a whole is clearly unworkable due to the poorer health of beneficiaries. An early solution to this quandary was to assume that the earnings of unsuccessful applicants after rejection serves as an upper-bound estimate of the potential work capacity of those with successful SSDI applications, because the former group is likely in better health (Bound, 1989). Subsequent studies raised and debated potential problems with this approach (Bound, 1991; Lahir et al., 2008; Parsons, 1991).

More compelling recent evidence identifies plausibly exogenous sources of SSDI receipt. Maestas, Mullen, and Strand (2013) exploit variation in the allowance rates of disability insurance examiners at the first step of the determination process. They find that among those applicants on the margin of program entry, employment would have been nearly 30 percentage points higher in the absence of SSDI. French and Song (2012) similarly make use of variation in the assignment of SSDI cases to administrative law judges, while Chen and van der Klaauw (2008) use the variation in receipt arising from the medical-vocational grid; the latter two studies obtain estimates in the range of the former’s results.

The long-term rise in the SSDI rolls has generated interest in existing or prospective policies that might increase labor force participation and reduce SSDI receipt among disabled individuals who retain some work capacity. One such policy is state vocational rehabilitation programs, which provide employment-related services to individuals with disabling conditions. These services are associated with higher labor force participation and reduced take-up of SSDI (Dean et al., 2014; Loprest, 2007; O’Neill et al., 2015). Individuals who face a longer wait time for these services are more likely to end up on SSDI, suggesting that they play a causal role in SSDI deterrence (Hyde et al., 2014).

Some analysts have called for firms to be required to provide short-term disability insurance for a period of time before individuals could apply for SSDI (Autor and Duggan, 2010). Firms may be in the best position to offer accommodation and rehabilitation to a newly disabled worker, and requiring them to bear some share of the cost of disability benefits may induce them to do so. A reform of this type in the Netherlands led to higher labor force participation and reduced use of long-term disability benefits (van Sonsbeek and Gradus, 2013). Others have suggested that contributions to the SSDI program could be experience-rated, where firms whose workers have been more likely to use SSDI face a higher payroll rate (Burkhauser and Daly, 2011).

Finally, the Social Security Administration and other federal agencies have conducted a number of large-scale demonstration projects to test various interventions that aim to support employment and reduce SSDI receipt among people with disabilities. The best-known of these is the Ticket to Work program, which provided SSDI beneficiaries with greater access to job placement, job training, and other services. The Ticket to Work program increased use of these services but had no consistent effect on SSDI use (Stapleton et al., 2014). A pilot program allowing SSDI recipients to keep part of their benefits if they earned more than the substantial gainful activity level did not have significant effects on employment (Weathers and Hemmeter, 2011). In general, these demonstration

projects have had fairly small effects on employment and earnings, relative to what might be needed to replace SSDI, with somewhat bigger impacts for programs targeted at younger disabled persons (Wittenburg et al., 2013).

Summary

In sum, the SSDI program provides important protection against the risk of earnings loss due to a work-limiting disability. The long period of growth and very recent decline in the SSDI rolls is likely due to many factors. Changes in the program, including a 1984 law to liberalize medical screening, seem quite important, but demographic factors also play a role. There is ample evidence that economic factors matter also. Long-term structural decline in the demand for low-skilled workers appears to increase SSDI applications, and recessions or industry-specific busts lead to an increase in both applications and awards.

There is a strong relationship between age and SSDI receipt, and there is a steep education gradient that is not fully explained by differences in health. Recent evidence suggests that employment rates of marginal SSDI applicants would be about 30 percentage points higher in the absence of SSDI. Interventions designed to support employment and reduce SSDI use among disabled individuals have a mixed record of success, although there is some evidence that providing access to vocational rehabilitation services and employer-provided short-term disability benefits may support work.

Social Security

Established by the Social Security Act of 1935, the Social Security program provides benefits to retired workers and their dependents and survivors. Workers and their employers make payroll tax contributions on earnings (up to a taxable maximum) and those with 40 quarters of contributions receive retired worker benefits. Benefits are calculated based on the worker’s highest 35 years of wage-indexed earnings. The benefit formula is progressive, such that the replacement rate (benefit as a share of average earnings) is higher for lower-wage workers, although the benefit amount is lower. The mean replacement rate is 44 percent among all workers, 69 percent for workers in the lowest quintile of lifetime earnings, and 31 percent for workers in the highest quintile (Khan et al., 2017). The average replacement rate in the United States is lower than that in most advanced economies (Ruffing, 2013). The average monthly benefit for workers claiming benefits in 2018 was $1,481 (U.S. Social Security Administration, 2019a).

Retired worker benefits are first available at age 62 but may be claimed as late as age 70. The benefit calculation is centered on the Full Retirement Age (FRA), historically age 65 but rising to age 67 for those born in 1960 or later. Claiming benefits before the FRA results in a lower benefit due to an actuarial adjustment, while claiming after the FRA results in a higher benefit due to the Delayed Retirement Credit. A worker whose FRA is 66 receives 75 percent of the FRA-level benefit by claiming at age 62 and 132 percent of this benefit by claiming at age 70. Workers may claim benefits and continue working, but a Retirement Earnings Test applies for workers below the FRA; this test reduces benefits by $1 for each $2 in earnings beyond an exempt amount ($18,240 in 2020). While the benefit amount is later adjusted to compensate for lost months of benefit receipt, there is ample evidence that workers perceive the Retirement Earnings Test to be a tax (Gelber, Jones, and Sacks, 2020).

Dependent spouses can receive a benefit of 50 percent of the retired worker benefit, although spouses entitled to a retired worker benefit receive their own benefit plus a “top-up” to the spousal benefit (if larger), but not both. Women’s rising labor force participation has diminished the importance of dependent benefits over time, a trend that is projected to continue. More than 80 percent of married women in the late Baby Boom (born 1956–1965) and Generation X (1966–1975) cohorts will receive benefits based solely on their own earnings (Iams, 2016). Surviving spouses receive 100 percent of the retired worker benefit; dependent children of retired or deceased workers receive benefits also.

Social Security protects workers against a number of risks. First, since benefits last until death, the program protects workers against the risk of outliving their retirement resources. Second, benefits are indexed to the Consumer Price Index, providing protection against inflation risk. Third, survivor benefits protect the worker’s family in the event of the worker’s death. Finally, to the extent that some workers may fail to save optimally for retirement due to heuristics and biases in savings decisions (Benartzi and Thaler, 2007), Social Security ensures a base level of consumption during retirement. Notably, Social Security benefits are themselves at some risk due

to the program’s long-term financial deficit. The combined Social Security and Disability Insurance Trust Fund is currently projected to be depleted in 2035, and at that point, tax revenue would be sufficient to pay 79 percent of scheduled benefits (U.S Social Security Administration, 2020c).

Social Security plays a central role in the retirement income of older Americans. Among persons age 65 and above, 86 percent live in a family that receives Social Security income (Waid, 2016). These values are slightly higher for White people (89%) and lower for Black (82%), Hispanic (75%), and Asian people (69%). Social Security is distributed far more equally across the income distribution than are other sources of retirement income. The probability of having Social Security income is quite similar across income quintiles. By contrast, families in the highest quintile of income are at least four times as likely to have income from pensions and retirement savings and twice as likely to have income from other assets than are families in the lowest quintile (Waid, 2016). Overall, Social Security represents about half of income for households ages 65 and above, though the share is higher in low-income households (Bee and Mitchell, 2017).

Including the value of Social Security benefits in calculations of retirement wealth dramatically reduces inequality across race and ethnic groups. Excluding Social Security, wealth at ages 51 to 56 for middle-wealth households born between 1960 and 1964 is $25,000 for Black people, $35,000 for Hispanic people, and $177,000 for White people. This results in Black:White and Hispanic:White wealth ratios of 14 percent and 20 percent, respectively. Including the value of Social Security in retirement wealth raises these values to 46 percent and 49 percent, respectively (Hou and Sanzenbacher, 2020). While Social Security wealth is distributed far more equally than other sources of wealth, average Social Security wealth is higher for White people than for Hispanic people and Black people due to differences in earnings histories, as discussed in Chapter 4.

Effects of Social Security on Retirement

With respect to the older workforce, Social Security is important in two respects. First, Social Security provides resources that make retirement possible, at least for those workers lacking other resources with which to finance retirement. Second, Social Security may affect the timing of retirement, as the availability of benefits at age 62 as well as the way in which benefits change if workers claim later may affect retirement decisions.

To use the language of economic theory, these two mechanisms correspond roughly to wealth effects and accrual effects. Social Security wealth can be calculated as the present-day value of the stream of Social Security benefits that the worker can expect to receive over their lifetime, where future years of benefits are discounted (worth somewhat less from today’s perspective) to account for mortality risk and the preference for money today over money in the future. A recent estimate put Social Security wealth (as measured at ages 51 to 56) for individuals born between 1960 and 1964 at about $200,000 for White people in the middle of the Social Security wealth distribution and at about $150,000 for Black people and Hispanic people (Hou and Sanzenbacher, 2020). A higher value of Social Security wealth is expected to lead to earlier retirement, as more wealth enables households to consume more of the things they value, including leisure time.

The effects of working longer on Social Security wealth are complex. The worker (if age 62 or older) foregoes one year of benefits by delaying retirement and claiming but receives a higher future benefit due to the actuarial adjustment. The additional year of earnings may factor into the benefit computation, replacing a lower earnings year. The worker and employer pay additional payroll taxes.²¹ The net of all of the effects of working longer can be to increase or decrease Social Security wealth. Two useful measures of the financial incentive to work longer are the accrual, or change in Social Security wealth associated with working one year longer, and the tax rate, which is the (negative of the) accrual scaled by earnings—the latter can be interpreted like a standard tax rate, reflecting the share of earnings lost by working one year longer. A third measure, peak value, is similar to the accrual but measures the change in Social Security wealth arising from working to the age when wealth is maximized (Coile and Gruber, 2007).

___________________

²¹ Economic theory suggests that contributions by employers are fully “passed through” to workers in the form of lower wages, indicating that they are also paid by workers. Recent evidence indicates that this is the case when there is a clear link between a higher contribution and a future higher benefit, but less true when there is no such link (Bozio et al., 2019).

also varies by living arrangements and income. Some states provide additional income to account for differences in living expenses. Individuals who receive SSI are eligible for other welfare programs including Medicaid and the Supplemental Nutrition Assistance Program (commonly known as food stamps; U.S. Social Security Administration, 2020b; Daly and Burkhauser, 2003).

Among adults age 65 and older, SSI coverage decreased from more than nine percent of these adults in 1973 to less than five percent in 2019. This reduced coverage was a result of means-testing rules on income and assets that have not changed much over time and have not been adjusted for inflation. In 2019, 11 percent of older adults with SSI also received Social Security benefits. SSI has increased its coverage of nonelderly individuals with disabilities. Among adult SSI recipients age 65 and older, 42 percent have income (including SSI benefits) below the poverty line (Center on Budget and Policy Priorities, 2020), and 71 percent are women (Sweeney and Fremstad, 2005).

There is some research exploring the effects of SSI on health and preretirement labor. SSI improved the health of older recipients within three years of program introduction (Taubman and Sickles, 1983), while higher SSI benefits are associated with lower disability rates (Herd et al., 2008). Perhaps surprisingly, only 56 percent of eligible individuals claim SSI, and claiming is associated with being in poverty and lacking any other sources of income (McGarry, 1996). Low-income older adults close to the eligibility threshold are less likely to apply. SSI recipients who are enrolled in multiple safety net programs have fewer incentives to work, given cumulative marginal tax rates and the complexity of satisfying multiple program requirements (Daly and Burkhauser, 2003). More generous SSI benefits reduce employment and the work hours of likely SSI participants among low-income adults 62 to 64 years old (Neumark and Powers, 2005). Some low-income individuals claim SSI as an early retirement option before claiming Social Security benefits (Neumark and Powers, 2000).

Supplemental income programs such as SSI have been introduced around the world to alleviate poverty among older adults. Most programs are universal or have few means-testing rules and cover a large proportion of older adults with generous benefits (Willmore, 2007). Research on South Africa, Brazil, Mexico, and other Latin American countries has found that supplemental income programs reduced household poverty and inequality (Ali et al., 2010; Barrientos et al., 2003).

Summary

In the United States, the increasingly strict rules for SSI have made it a welfare program for those in severe poverty. Over time, it has had less influence on reducing poverty or affecting the labor supply of low-income older adults because of reduced benefits and lack of program participation. To our knowledge, no studies have shown how SSI has functioned as a safety net during economic recessions. One study suggests that for low-income older adults who are able to work, the rules of SSI and other safety net programs function more as a poverty trap than as a means to improve recipient well-being (Daly and Burkhauser, 2003). Nonetheless, SSI is a relevant safety net program for nonelderly individuals with disabilities, both children and adults, as well as for low-income elderly.

Retirement Savings Policies

Government policies that support saving for retirement may increase the likelihood that workers nearing traditional retirement ages have adequate resources to retire.

Adequacy of Retirement Saving

There is an active debate in the literature over whether people are saving enough for retirement. Using a life-cycle model to estimate the optimal level of saving for households, Scholz, Seshadri, and Khitatrakun (2006) conclude that less than 20 percent of retirement-age households are saving too little. Following a similar approach, Scott and colleagues (2020) find that it may be optimal for very low earners to save little or not at all, due to the high replacement rate of Social Security for such workers. A different approach to determining savings adequacy is proposed by Munnell and colleagues (2006). They project households’ retirement income based on assets and entitlements to Social Security and pensions, and then calculate whether income is sufficient for households to maintain their preretirement

standard of living; they assume that 65 to 85 percent of preretirement income is needed, depending on family type and income level, as older households do not save for retirement and pay less in taxes. In the most recent update, they estimate that 50 percent of households are at risk of falling at least 10 percent below this benchmark and that this figure could rise to 55 percent due to COVID-19–related unemployment (Munnell et al., 2020).

Analysts have scrutinized changes in consumption at retirement as another possible indicator of retirement saving adequacy. A simple life-cycle model suggests that people will prefer constant consumption over time. If households experience an unexpected drop in consumption at retirement, this may indicate that they are surprised to find that they have inadequate savings. The literature on this question is generally reassuring when it comes to the population as a whole. Consumption drops are largely anticipated (Hurd and Rohwedder, 2003) and concentrated in food and work-related expenditures, while food intake remains constant (Hurst, 2008). While up to one-quarter of households experience a drop at consumption in retirement, this is concentrated among households experiencing involuntary retirement due to health shocks (Hurst, 2008); low-income older adults are also more likely to have such a drop (Bernheim et al., 2001). Hurst (2004) finds that 20 percent of individuals did not plan adequately for retirement. Aguila and colleagues (2011) find that the lowest income quartile dropped their nondurable consumption by 36 percent when the head of household retired.

Another issue that may shed light on retirement savings adequacy is the inability of low-income households to smooth their consumption between Social Security payments. Stephens (2003, 2006) finds that recipients’ expenditures increased during the first weeks after check receipt and then decreased until the arrival of the next payment. Shapiro (2005) and Mastrobuoni and Weinberg (2009) document a similar pattern for caloric intake. Studies of other countries have found that more frequent benefit payments increase food availability, health care use, and health.²² Similar U.S.-based studies could reveal whether varying the frequency of Social Security, SSI, or other benefit payments would affect the well-being of resource-constrained households.

Older adults may spend larger payments quickly because of their lack of options, such as a bank account, for storing money safely (Blanco et al., 2018). Older unbanked adults are more likely to have lower income, to be non-White, to have fewer savings, and to face higher monetary costs for everyday transactions (Carbó et al., 2005; Thaler 1999). Gaining access to bank accounts resulted in mental health benefits for older low-income Hispanics who lacked access to the financial sector (Aguila et al., 2016).

Employer-sponsored Pension Plans

We turn now to the role of the government in promoting and regulating retirement saving. Retirement saving may occur in employer-sponsored pension plans or in other designated accounts. The government’s role includes setting pension plan regulations, which may affect an employer’s willingness to offer a plan, the type of plan offered, and other plan provisions, as well as providing tax incentives for saving.

One type of employer-sponsored pension is the defined benefit plan. Defined benefit plans were first regulated under the 1974 Employee Retirement Income Security Act, which established minimum funding requirements and set up the Pension Benefit Guarantee Corporation to insure benefits against plan insolvency.²³ In a defined benefit plan, the employee receives a regular benefit payment from retirement until death. While details vary across plans, the benefit is generally calculated using a formula based on the worker’s years of service and average earnings near the end of a career. A certain tenure at the firm is usually required to be vested in the plan. Most plans have both early and normal retirement ages, and the benefit amount is often adjusted for claiming age in a way that creates strong financial incentives to work to these ages and retire and claim thereafter—for example, increasing for each year the worker delays between the early and normal retirement ages (Stock and Wise, 1990).

___________________

²² Evidence from Mexico and Kenya suggests that one-time or less frequent payments are more likely to be used for durable goods while more frequent payments increase food availability and health care use (Haushofer and Shapiro, 2016; Aguila, Kapteyn et al., 2017). More frequent payments to low-income older adults in Mexico also improved health and well-being more than less frequent, larger payments did (Aguila and Smith, 2020), potentially because of the difficulty of smoothing consumption between payments.

²³ Multiemployer plans are a type of plan protected by the PBCG. Some of these plans are significantly underfunded—currently one million employees (10% of those in multiemployer plans) are in plans that could run out of money within 15–20 years. The PBCG estimates that its multiemployer insurance program will run out of money within 10 years (Munnell et al., 2017).

arising from higher (or lower) than usual stock market returns or unexpected inheritances. There is only weak evidence of an effect of wealth on retirement in the former case (Hurd et al., 2009) but stronger evidence in the latter case (Brown et al., 2010).

Effects on Saving

The U.S. federal government provides $250 billion per year in tax incentives for retirement saving through defined contribution plans and IRAs (Tax Policy Center, 2020). Using the model of a rational, forward-looking consumer, these tax incentives may affect saving through several pathways. First, defined contribution plans and IRAs increase the rate of return to saving. They do so by enabling individuals to invest funds on a pre-tax basis (to invest a full dollar of earned income, rather than the fraction remaining after income tax) and by allowing investment earnings to compound tax-free before being taxed at withdrawal. In theory, the impact of an increased rate of return on saving behavior is ambiguous. Consumers may choose to save more since a dollar of foregone spending today yields more spending in the future when the rate of return is higher (substitution effect), but they also may choose to spend more today and save less because the higher return has made them better off (income effect). Saving will increase if the former effect dominates. A second pathway, also consistent with this model, is that exposure to a retirement savings plan may increase financial literacy, through financial education offered by the firm or peer effects (Duflo and Saez, 2004; Lusardi, 2009), which could lead to higher retirement saving.

Running counter to the standard model is the insight that saving for retirement is “both a difficult cognitive problem and a difficult self-control problem” (Mullainathan and Thaler, 2001). It is now well understood that there are many biases present in human decision-making that can affect saving and other decisions. For instance, individuals approach tradeoffs between present and future in a way that is dynamically inconsistent, reflecting a bias toward the present (Laibson, 1997). Procrastination leads to a tendency to stick with the status quo (Samuelson and Zeckhauser, 1988). Exponential growth bias refers to the tendency of individuals to underestimate the gains from compound interest (Stango and Zinman, 2009). Under mental accounting, labeling funds as being for current vs. future spending can impact their use (Shefrin and Thaler, 2004). The effect of retirement plans on saving may be quite different from what the standard model predicts if behavioral biases are important. For example, auto-enrollment in a pension plan or auto-escalation of contributions, either of which is easily undone by a rational consumer, will have a greater impact if status quo bias is important (Thaler and Bernartzi, 2004). Introducing tax-advantaged retirement accounts may lead to new saving, rather than the diversion of existing assets into the new accounts, if mental accounting is present.

Turning to the evidence, the early literature on 401(k)s and IRAs and saving focused on whether retirement account balances constitute new savings or a reallocation of savings that would have occurred elsewhere. This question has been intensely debated (Poterba et al., 1996; Engen et al., 1996). An arbiter examining evidence on both sides conservatively estimates that at least 26 cents of every dollar of IRA contributions represents new saving while 401(k) contributions “largely” represent new saving (Hubbard and Skinner, 1996).

More recently, the literature has focused on the role of defaults in defined contribution plans. Employers with such plans decide whether employees are automatically enrolled in the plan and often set a default for the contribution rate, asset allocation, and post-retirement distribution. An employee who is automatically enrolled may opt out of the plan or change other defaults, but this requires an action on their part. A large literature has established that such defaults have a very substantial effect on savings decisions (U.S. Government Accountability Office, 2009; Beshears et al., 2007). For example, in a firm adopting auto-enrollment, the plan participation rate at 3 to 15 months after hire was 86 percent in the group hired after auto-enrollment, as compared to 37 percent at a similar point post-hire among those hired under an opt-in regime (Madrian and Shea, 2001). Effects were larger among employees from groups that were historically less likely to participate, including younger, lower-income, Black, and Hispanic employees.

The literature has also begun to explore whether auto-enrollment promotes new saving or a reallocation of other resources into retirement accounts. In the Danish context, a one percentage point increase in mandatory savings contributions is associated with a 0.8 percentage point increase in total savings (Chetty et al., 2014). Interestingly, 85 percent of Danish individuals are passive savers who are influenced by defaults but not by financial incentives to save, while those who do respond to financial incentives were saving already, implying that financial incentives

are not particularly effective in increasing savings among those who are unprepared for retirement. Recent evidence from the United States as to whether there is any crowd-out of default-induced savings is more mixed. Workers who are automatically enrolled in their current job’s pension plan contribute less to a following job’s opt-in pension (Choukhmane, 2019), but auto-enrollment is not associated with an increase in debt (Beshears et al., 2019).

In the wake of this influential research, there is now an understanding that defaults are not neutral but, in fact, are likely to either encourage or inhibit retirement saving. The 2006 Pension Protection Act encouraged firms to adopt auto-enrollment by providing a safe harbor from antidiscrimination tests (which require firms to certify that the plan does not discriminate in favor of highly-compensated employees), so long as the plan has adequate default contribution levels, a qualified default investment fund such as a “target-date” retirement fund, and an adequate employer match. The Pension Protection Act also protected firms selecting a qualified default investment fund from fiduciary liability in the event of investment loss. Between 2005 and 2008, the share of firms using a money market fund as a default investment dropped from 25 to two percent, while the share selecting a target-date fund rose from 42 to 87 percent (U.S. Government Accountability Office, 2009). Given the difference in the expected rate of return by fund type, such changes can meaningfully impact account balances at retirement.

Evidence on state-established auto-IRA programs thus far primarily comes from Oregon. Among eligible employees in firms with at least 50 workers, 62 percent are participating in the plan (Belbase and Sanzenbacher, 2018). Of those not participating, the most commonly reported reasons are not being able to afford to save (30%) or having another retirement saving plan (19%). Over 90 percent of participants keep the default contribution rate of five percent. In the program’s second year, 20 percent of those with account balances made a withdrawal from their account. While these “leakages” inhibit the accumulation of retirement assets, they also suggest that auto-IRA balances may serve as a valuable source of precautionary saving (Quinby et al., 2019). In the first year after California adopted an auto-IRA program, new employer-sponsored plans were nine percent of the total plans in the state, well above the national average; Illinois also experienced an increase in new plans in the year after implementation (Scott, 2021).

Models from outside the United States may also prove informative. In the United Kingdom, a 2008 law required all private-sector employers to auto-enroll their employees in a plan. Several years after implementation, enrollment rates for employees in medium and large firms were nearly 90 percent, similar to rates in large U.S. firms with auto-enrollment, and nearly 70 percent in small firms (Cribb and Emmerson, 2019). Exploiting variation in the timing of the roll-out, the auto-enrollment mandate is found to raise participation in the smallest firms (under 30 employees) by 44 percentage points. In other countries, such as Australia, a mandatory retirement savings program is one of the pillars of the retirement system. In the Australian case, employers contribute 12 percent of earnings, which are invested in Superannuation Funds at the employee’s direction (Agnew, 2013).

Policies for Low-income Savers

As discussed in Chapter 4, there are substantial differences in wealth by race and socioeconomic status. One key factor driving these differences is the disparity in the probability of being offered a pension. The share of wage and salary workers offered a pension by their employer is 65 percent for White workers, as compared with 56 percent for Black workers and 38 percent for Hispanic workers; conditional on offer, White workers are about 5 percentage points more likely to participate than non-White workers (Butrica and Johnson, 2010). Similarly, in a comparison of higher and lower wage workers, differences in the probability of offer are much larger than differences in conditional participation (Wu and Rutledge, 2014). Differences in financial literacy may contribute to the wealth gap. Finally, providing financial incentives for retirement savings in the form of a tax deduction favors high-income workers, who face higher marginal tax rates; an alternative would be to offer tax credits, which do not depend on tax rate.

A few programs have aimed to promote retirement savings for low-income workers who lack access to employer-provided pensions. MyRA, which started in 2014 and ended in 2018, targeted employees in firms without employer-provided pensions. This program failed because of low participation and high management costs (Powell and Ferraro, 2018); low participation is unsurprising given that enrollment was voluntary (Munnell, 2015). The Saver’s Credit program provides a nonrefundable federal income tax credit of up to 50 percent of contributions

to a 401(k) or IRA by low-income filers. The nonrefundability of the credit limits eligibility, because many low-income filers have no tax liability (Koenig and Harvey, 2005). Duflo and colleagues (2006) find that the saving response to the Saver’s Credit is smaller than what is obtained in a randomized controlled trial offering a match to IRA contributions, potentially due to the opacity of the credit and framing.

More general evidence on IRAs suggests that these programs have not been particularly successful in closing the racial wealth gap. Among older adults (age 65 or older) in 2012, 46 percent of non-Hispanic White people but only 12 percent of Black people and 14 percent of Hispanic people held an IRA. For older adults between the 10th and 50th percentiles of wealth, IRAs represented six percent of total wealth for non-Hispanic White people and one percent for Black and Hispanic people. For older adults between the 50th and 90th percentiles of wealth, IRAs represented 20 percent of total wealth for non-Hispanic White people, four percent for Black people, and five percent for Hispanic people (Johnson et al., 2013).

Health Insurance Policies

Government policies that facilitate access to health insurance at older ages are potentially important in the employment decisions of older workers. The United States has the distinction of being the only large, wealthy country without universal health coverage. As discussed in Chapter 4, a majority of nonelderly Americans receive health insurance through their own or a family member’s employer. The share of workers who have retiree health insurance—coverage from their employer after leaving the job—has declined over time. Workers with access to retiree health insurance retire substantially earlier than other workers. Government policies that provide access to health insurance to older individuals may facilitate retirement by addressing one possible motivation for continued work at older ages.

Medicare

Medicare is central in any discussion of health insurance and older Americans. Established in 1965 through Title XVIII of the Social Security Act, Medicare provides health insurance to people age 65 or older and to SSDI beneficiaries (after a two-year waiting period).²⁴ The Medicare program has several components: Part A, which covers inpatient hospital services; Part B, which provides physician and outpatient services; Part D, which offers prescription drug coverage; and Part C, or Medicare Advantage, which allows beneficiaries to enroll in a private insurance plan for their Part A, B, and D benefits.²⁵ Most Medicare beneficiaries pay no premiums for Part A benefits but have premiums for Parts B and D as well as deductibles and co-insurance payments (Kaiser Family Foundation, 2015). The cost-sharing in Medicare leaves beneficiaries with a substantial risk of out-of-pocket medical expenditures (Goldman and Zissimopoulos, 2003). Nearly nine in ten beneficiaries have some protection against these expenses, through enrollment in a Medicare Advantage plan or a secondary insurance plan such as an employer-sponsored plan, Medicaid, or a “Medigap” policy (Kaiser Family Foundation, 2015). Nearly 11 million of Medicare’s 61 million beneficiaries are dually eligible for Medicare and Medicaid, based on having low income and assets (Grabowski et al., 2017). Medicare expenditures on this group are twice as high as for Medicare beneficiaries as a whole, due in large part to their poorer health.

The probability of retirement has historically been higher at age 65 than at surrounding ages (Coile, 2004). This phenomenon could reflect the role of Medicare eligibility in retirement decisions, but there are several other possible explanations. The Social Security FRA was age 65 for people born before 1938, potentially creating a social norm of retirement at this age (Behaghel and Blau, 2012). Earlier cohorts faced financial disincentives to work past age 65 due to a less-than-fair actuarial adjustment for delayed claiming (Coile, 2020). Defined benefit pension plans typically feature strong disincentives for work beyond the plan’s early or normal retirement age, often ages like 55, 60, and 65 (Stock and Wise, 1990).

___________________

²⁴ Medicare also covers people with end-stage renal disease and Amyotrophic Lateral Schlerosis or Lou Gehrig’s disease.

²⁵ For more details about the benefits provided under each part of Medicare, as well as other background information on the program, see Kaiser Family Foundation (2015).

___________________

²⁶ Premiums may also differ based on tobacco usage, at a maximum 1.5-to-1 ratio (Centers for Medicare and Medicaid Services, 2020).

allow individuals to buy in to a former employer’s plan for 18 months (in case of the federal mandate; state laws may vary). They have been shown to raise the retirement rate of men by 30 percent (Gruber and Madrian, 1995), which is at the low end of the estimated effect of retiree health insurance on retirement (Gruber and Madrian, 2004). This is a large effect, given that retirees are fully responsible for COBRA premiums and that coverage is available for a limited time only. The size of the effect could reflect the value retirees place on maintaining coverage for preexisting health conditions, as prior to the ACA these could be excluded from coverage under a new health insurance plan (Currie and Madrian, 1999).

Summary

In sum, the evidence that government policies affect retirement behavior at older ages is surprisingly mixed, given the strong and consistent effect of employer-provided retiree health insurance on retirement. It is unclear if the spike in retirement at age 65 can be attributed to Medicare, though a recent study suggests that the availability of prescription drug coverage through Part D increases retirement at that age. There is little evidence so far that the ACA has altered the employment decisions of older workers, but some evidence that the earlier Massachusetts health reform did so. Continuity-of-coverage mandates have also been shown to have a large effect on retirement in the pre-ACA era, though their role could be supplanted by the ACA.

CONCLUSIONS

In this chapter, we have surveyed a broad range of public policies that may support work at older ages, encompassing both non-age-specific and age-specific policies, as well as an extensive set of policies that may enhance the financial security of workers in retirement. The existing literature suggests that many of these policies affect retirement decisions. In other cases, the literature has yet to reach a conclusion as to the employment effects of these policies. We first review the key findings from each section of this chapter, before presenting our recommendations for future research.

Findings: Non-age-specific Policies

• The ADA protects workers of all ages from disability-based discrimination but is of potentially greater relevance for older workers due to their increased risk of being disabled. Research on the employment effects of the ADA has yielded inconclusive results.
• The FMLA provides access to 12 weeks of unpaid leave for workers dealing with their own or a family member’s health issues, while a growing number of states offer paid family leave for use in similar circumstances. Existing research on these programs has largely focused on their use following the birth of a child.
• Workers’ compensation provides access to medical care and compensation for lost wages following a workplace injury. While there is no clear relationship between age and the probability of a workplace injury, older workers tend to experience more severe injuries and are less likely to return to work.

Findings: Age-specific Policies

• The ADEA protects workers age 40 and over from discrimination in the workplace on the basis of age. Some states provide stronger age-based protections. The ADEA led to changes in human resources practices and stronger relationships between older workers and their employers. Stronger state-based age discrimination laws do not reduce the hiring of older workers.
• Mandatory retirement has been abolished in the United States since 1986 (except in a few occupations), though it is still in effect in many high-income countries. The elimination of mandatory retirement after the ADEA’s passage led to an increase in work after age 65.
• Partial retirement programs allow older workers to combine partial benefit receipt with part-time work; these programs exist in some European countries, but not in the United States. Research indicates that these

• policies either reduce labor supply at older ages (as some individuals shift from full- to part-time work) or have no effect on labor supply.
• The Senior Community Employment Program is a community service and work-based training program for low-income older workers with poor employment prospects. While the program has had limited success in helping participants achieve economic self-sufficiency, it increases health and well-being while providing income for recipients and service hours to their communities.

Findings: Policies to Support Financial Security

• SSDI provides income support to individuals experiencing a long-term work-limiting disability. Growth in SSDI over time has been driven by changes in medical stringency, population aging, rising female labor force participation, and economic factors. Use of SSDI is higher among the less educated. There is non-trivial residual work capacity among SSDI recipients on the margin of program entry. Interventions that aim to boost employment and reduce SSDI use among disabled individuals have had only modest effects.
• Social Security is a critical source of income for older Americans that is distributed far more equally than other sources of retirement income (or wealth). Program provisions shape the financial incentive to work at older ages, and workers are highly responsive to these incentives. Changes to the program over the past several decades have strengthened the incentive to work at ages 65–69.
• SSI is a cash benefit program for very low-income elderly adults as well as disabled adults and children. The share of older individuals receiving SSI has declined over time as means-tested income and asset limits have not been updated for inflation. SSI improves health and reduces the labor supply of older recipients.
• Workers face increasing responsibility to plan for their own retirement, due to the ongoing shift from defined benefit to defined contribution pension plans. Defined contribution plan defaults for participation, contribution, and asset allocation tend to be “sticky,” exerting a powerful impact on retirement saving. Evidence suggests that a substantial share of retirement account balances represent new saving, rather than assets shifted from nonretirement accounts. Whether most households have adequate retirement savings remains a hotly debated question. While most households are able to maintain their previous level of consumption at retirement, lower-income households experience consumption declines as well as difficulty smoothing consumption between benefit checks. Low-income workers are less likely to have IRAs and defined contribution pensions, and special programs targeting this group have only a mixed record in promoting saving.
• Most Americans obtain health insurance through their own or a spouse’s employer. As discussed in Chapter 4, having access to employer-provided retiree health insurance is strongly associated with earlier retirement. While access to coverage through Medicare at age 65 and through the ACA would be expected to increase the probability of retirement, the evidence on this point is surprisingly mixed.

Implications for Research

Impact of Age-specific Policies

More research is needed on how non-age-specific policies that support work are affecting older workers. While all of these policies have been extensively studied, research on how they affect older workers is quite limited. Researchers could make use of differences in state disability laws to explore whether disability protections affect hiring or other employment outcomes for older workers, as has been done for state age discrimination laws. For the FMLA and state family leave policies, research could exploit the introduction of these policies or their applicability by firm size to explore whether these policies help older workers manage their own or a family member’s health crisis. For the Workers’ Compensation program, more research is needed into how to deter injuries and their impact on return to work, how to hold down Workers’ Compensation costs for older workers, and whether the risk of more serious injury deters employment of older workers.

This page intentionally left blank.