Age Related Health Costs and Job Prospects of Older Workers

Transcription

1 First draft Age Related Health Costs and Job Prospects of Older Workers Gary Burtless * THE BROOKINGS INSTITUTION October 22, 2017 * The author is a senior fellow and holds the Whitehead Chair in the Economic Studies program at the Brookings Institution, Washington, D.C. This paper was prepared for the 2017 Working Longer and Retirement Conference, Stanford Institute for Economic Policy Research, November 2-3, I gratefully acknowledge the excellent research assistance of Eric Koepcke and Austin Drukker and the generous research support provided under the Alfred P. Sloan Foundation s Working Longer program. The views are solely my own and do not represent those of Brookings or the Sloan Foundation.

2 Age Related Health Costs and Job Prospects of Older Workers by Gary Burtless MOST AMERICANS WHO WORK or are seeking work think age discrimination in employment is a problem. More than a third think it is a serious problem (Wilson 2006). Among Americans past age 50, one-third believe that they or people they know have been the actual victims of workplace age discrimination (GS Strategy Group 2012). Though many forms of age discrimination are difficult to document, at least one kind of discrimination bias against older job applicants has been confirmed in experimental trials. Findings from resume audit studies suggest that many employers prefer to interview, and presumably to hire, younger rather than older job seekers who have identical qualifications. None of the results from audit studies provide unambiguous evidence on employers motivations for discriminating against older applicants. Some managers may believe older workers are more costly to train, harder to manage, less flexible in responding to workplace change, or more likely to suffer some form of age-related cognitive or physical decline. In short, these managers may assume older job seekers will be less productive than younger applicants who have the same credentials. There is in fact evidence in the personnel and applied psychology literatures that many managers apply negative age stereotypes when evaluating agerelated personnel issues in laboratory settings (Rosen and Jerdee 1977; Ng and Feldman 2012). Some of the stereotypes accepted by managers who discriminate may of course have some foundation in fact. Another possibility is that employers think older workers will be more expensive to employ, even if they are paid the same wage and are just as productive as identically credentialed younger employees who hold the same jobs. Several kinds of employer costs rise as workers get older. Expenses connected to a worker s health, including employer premiums for health insurance, can certainly increase. The increased risk of illness as workers age might also raise the cost of sickness pay for employers offering paid sickness leave. Even employers that do not provide paid leave can face higher costs as workers age if absences resulting from illness increase firms operating costs. Serious health problems can also lead to an early end to a - 1 -

3 worker s career, forcing employers to find or train a replacement workers. If career-ending health episodes occur more frequently for older workers compared with younger ones, it can increase the perceived cost of hiring or retaining older workers. Finally, some employer retirement plans are more costly to fund for older workers than they are for younger workers who earn the same wage. Defined-benefit (DB) pension plans, for example, might require larger annual contributions for middle-age and older workers compared with workers who are under 35. This paper focuses on age-related employee costs connected to the provision of health insurance. Employer-sponsored insurance is by far the most important source of health coverage for working Americans and their dependents. Half of children and approximately 60 percent of nonaged adults obtain health insurance coverage under an employer-sponsored plan. For many employers and for the employer community at large health insurance is the most costly nonwage benefit they provide. In 2014 almost 60 percent of full-time workers between 18 and 64 obtained health insurance under a health plan sponsored by their employer. Employer-sponsored insurance (ESI) is less likely to be offered and taken up by workers on part-time schedules. Nonetheless, about 70 percent of all civilian employees are offered access to employer health plans and slightly more than half participate in them (U.S. BLS 2017a, Table A). If we count workers dependents, about 60 percent of all Americans under age 65 receive health coverage under an employer-sponsored plan. The BLS estimates that 8.3 percent of employee compensation consists of employer contributions for health insurance. This is equivalent to a little more than 12 percent of total money wages (U.S. BLS 2017b, Table A). The cost of paying for such plans is higher for older compared with younger workers. Older people are more likely to experience serious illness and suffer adverse effects from injury or chronic disease. As a result, health insurance plans make bigger annual payouts to older compared with younger employees. Older employees do not bear the full extra costs of these insurance reimbursements, because employee health premiums are not differentiated to reflect a worker s age or expected health care outlays. Nor do aged employees bear the costs indirectly through lower money wages. Such wage differentiation is banned under the Age Discrimination in Employment Act. The gap in health care spending between older and younger working-age Americans is sizeable. In the 2014 Medical Expenditure Panel Survey (MEPS), for example, mean health expenditures per person were nearly four times greater for Americans between 60 and

4 compared with adults between 25 and 29. Mean spending in the older group was nearly 2.4 times the per capita spending on Americans between 40 and Not all of these spending differences are reflected in employer outlays on their employees health, of course. Insurance plans do not cover the full cost of medical treatments. They reimburse only a fraction of the costs, a share that is determined by a plan s deductibles and employee cost-sharing requirements. Furthermore, under the great majority of ESI plans employees must pay premiums to obtain coverage. These charges partially offset employers cost of providing coverage. Employers do not provide insurance coverage to all working-age adults. The highest health expenditures often focus on people with disabilities and chronic conditions. Many adults with these conditions do not work and are not covered by an ESI plan. Instead their health expenditures are insured under public insurance programs, such as Medicare and Medicaid. The relationship between workers ages and employers health expenses is also affected by the distribution of reimbursable medical expenses among workers dependents. Workers who are past age 50 are less likely to enroll in ESI family plans than are workers who are in their 30s and 40s. Older workers are more likely to enroll in single employee plans and employee-plus-onedependent plans. Since employer plans do not cover as many dependents per enrolled worker in the case of their oldest employees, there may be an offset for the higher costs associated with older employees own health care expenses. On the other hand, the percentage of workers that accepts an employer s offer of ESI coverage tends to increase as workers grow older, at least up through age 65 when workers automatically become entitled to Medicare. In the remainder of this paper I estimate the age-related costs of ESI plans in order to determine whether employee cost differences associated with age can plausibly explain some of the discrimination against older employees and job applicants. I. Discrimination and age-related employment costs Employer bias against older workers is widely suspected and, in the case of older job applicants, has been confirmed in randomized trials. In these trials, commonly called correspondence studies or resume audits, the experimenter responds to help wanted ads with carefully crafted job applications from fictitious job seekers (Neumark 2016). The applications, which typically include a formal resume, mention a variety of the applicant s characteristics. The 1 MEPS, Table 1: Total Health Services-Median and Mean Expenses per Person with Expense and Distribution of Expenses by Source of Payment: United States, 2014 and author s calculations

5 focus, however, is on one key characteristic, namely, the applicant s membership in a group that is the potential target of discrimination. In correspondence studies that examine age discrimination, the key characteristic is the applicant s age. Two applications may be prepared by the experimenter that show identical qualifications for the advertised job. The only difference between the two resumes is the applicant s age. Appropriate fictitious resumes, selected at random, are sent to employers who have published advertisements for a particular kind of job. The experimenter then records the number of favorable responses received from employers for each application. A favorable response might include a call-back from the employer or an invitation to the applicant to visit the firm for a formal interview. If a smaller proportion of older applicants than of younger applicants receives favorable responses from employers, there is a strong presumption that age discrimination is hurting the employment chances of the older fictitious applicants. Most of the experimental resume audits provide evidence of age discrimination in the hiring process (Neumark 2016). Results by Joanna Lahey (2008) are among the best known. She submitted approximately 4,000 fictitious help wanted ad responses for entry-level positions advertised in Boston, MA, and St. Petersburg, FL. All of the fictitious responses carried a woman s name and were in response to advertisements for positions typically held by women. Women described in the responses were ages 35, 45, 50, 55, and 62. If we define applicants age 35 and 45 as younger and applicants age 50, 55, and 62 as older, Lahey found that younger applicants were 42 percent more likely to be offered an interview in Boston and 46 percent more likely to be offered an interview in St. Petersburg. Another correspondence study recorded employer responses to applications from both male and female applicants, half age 32 and the other half age 57 (Bendick et al. 1997). The older applicants were significantly less likely to receive favorable responses from employers. Among employers whose size could be ascertained, the researchers found that large employers were more likely to discriminate against older applicants. It is notable that large employers are also more likely than smaller ones to offer generous benefit plans whose cost increases with workers age (Wiatrowski 2013; U.S. BLS 2017b, Table 8). This may give the discriminating firms stronger financial reasons to favor younger over older job applicants. The findings from resume audit studies confirm that many employers prefer to interview younger rather than older job applicants. It is more difficult to determine whether they prefer to - 4 -

6 retain and promote younger workers already on their payrolls in preference to equally or more qualified older workers, even though this preference is widely suspected by older workers. It is also hard to identify employers specific motives for favoring younger job applicants, although some of the audit studies have been designed to shed some light on this question. None of the results obtained so far provide clear guidance regarding employers reasoning. As noted, some managers may believe older workers are currently or soon will be less productive than younger workers who have the same credentials. The connections between job performance, cognitive aptitude, and physical ability, on the one hand, and age, on the other, have been topics of extensive research in psychology and medicine (Ng and Feldman 2008; Skirbekk 2008; Salthouse 2009; McGee and Wegman 2004). While there is considerable evidence that a number of cognitive abilities and physical functions decline on average with age, it is a matter of controversy whether these declines may be offset, fully or partially, with gains in wisdom or good judgement, possibly arising from experience or greater knowledge. The effects on productivity of these age-related changes are impossible to measure in many occupations, and they are beyond the scope of this study. My focus instead is on a more easily documented effect of aging, namely, its impact on employer costs and in particular on the employer cost of providing health insurance to employees. 2 To the extent that age-related cost differences are real and are not offset by higher worker productivity or lower costs in some other part of the compensation package, employers have a tangible reason to prefer younger and cheaper workers to older and more expensive ones. Health costs and age. A basic reality of health insurance is that personal health spending increases as adults grow older. Figure 1 shows per capita health expenditures in 2014, by age, for two groups of adults who are between 25 and 74 years old. The calculations are based on health consumption patterns observed in the 2014 MEPS household survey file. Spending amounts indicated by the light bars show expenditures among all adults in the noninstitutional population. The spending totals reflect consumption of health care goods and services regardless of the source of payment for the consumption. These include out of pocket outlays by the person or family and reimbursements by public or private group health insurance as well as private insurance obtained in the nongroup market. Because the MEPS sample excludes the long-term 2 Some of these and closely related issues have been examined in earlier research. See Hutchens (1988); Scott, Berger and Garen (1995); and Mermin, Johnson, and Toder (2008)

7 institutionalized, the health spending of some of the most costly adults is missed. The tabulations show a dramatic increase in per capita spending between ages and Between these ages annual expenditures increase by 137 percent or a bit more than $4,600 per year. For purposes of comparison, the economy-wide average wage in 2014 was $46,480, so the difference in annual health spending between Americans in their early 40s and those in their early 60s was about 10 percent of average annual earnings. The solid line in Figure 1 shows per capita spending in 2014 among employed workers, whom we would expect to be healthier on average than nonworkers. Workers mean health expenditures are indeed lower at every age compared with the spending amounts of all noninstitutionalized adults. Moreover, the proportional difference between workers health spending and that of the general population tends to widen at older ages, especially past age 45. Thus, the population at work is increasingly selected from among relatively low spending adults as it grows older. Past age 60, annual personal health expenditures among older workers increase very slowly compared with spending in the general population. Still, workers who are between 60 and 64 have mean expenditures that are twice those of workers between 40 and 44. The annual spending of the older group is more than three times that of workers in their late 20s. These spending gaps are economically meaningful if employers ESI plans cover a large fraction of the rising costs. Figure 2 offers evidence that private insurance, which for the working population consists mainly of employer-sponsored group insurance, is indeed financing a large percentage of the rising health costs associated with aging. The chart shows per capita private insurance reimbursements in 2014, by age, for workers who are covered by private insurance. The tabulations show a sharp rise in average reimbursement payments after age 54. Privately insured workers who are between 55 and 59 receive reimbursements that are $2,670 (or 112 percent) larger than those received by year-old workers. Note that the average reimbursement dips slightly for year-olds compared with year-olds and falls noticeably for workers past 65. The small dip after 60 might be due to health selection effects among the population that remains at work at older ages. The bigger drop after 65 is linked to workers eligibility for Medicare starting at that age. Insured workers in smaller firms may have Medicare as their primary insurance payer, even if they remain covered by their employer s plan. In addition, privately insured workers past 65 may obtain their private insurance through Medigap policies or - 6 -

8 as retired workers under a former employer s retiree health plans. In both cases, Medicare rather than the private policy would be the primary insurer. Thus, the sharp decline in private insurance reimbursements after age 65 provides no assurance to large employers that their older workers would become less expensive after reaching the Medicare eligibility age. 3 For a couple of reasons the tabulations in Figures 1 and 2 do not give direct evidence about the cost to firms of employing an older as opposed to a younger worker who fills the same job and earns the same money wage. Even the estimates in Figure 2, which focus on workers who are covered by a private plan, do not distinguish between the ESI reimbursements that are paid under the employee s own employer plan and those that are derived from a spouse s plan. This difference matters, because the cost of providing ESI insurance is obviously much greater in the case of employers who insure workers and their dependents than it is in cases where a firm s employees choose to enroll in another employer s plan. By focusing on the personal health spending and insurance reimbursements of individuals, the estimates also miss the cost to employers of supplying insurance to employees eligible dependents, including spouses and children. ESI plans typically offer coverage to employees dependents as well as to the employees themselves. The likelihood that an employee will have eligible dependents and that the employee will elect to cover them under an employer s plan varies over the life course. With respect to dependent coverage, employees may actually become cheaper when they grow older, as the number of their eligible dependents shrinks. However, an employee s most likely dependent is a spouse, and the spouse s age and vulnerability to high health expenses usually rise in line with the employee s. Public policy issues. Most Americans, especially those past 50, think workplace discrimination against the aged is a problem. It is less clear whether they believe there are good reasons, on cost grounds, for employers to prefer younger over older workers. One way to reduce employers incentive to discriminate is to restructure or eliminate components of the compensation package that are more expensive to provide to older employees. For example, employers might eliminate paid sickness leave and replace it with a combination of paid 3 Under present law, employers with at least 20 workers must offer to current employees older than 65 the same health benefits they offer to workers under 65. Similarly, they must offer coverage to employees spouses 65 and older the same coverage provided to employee spouses who are younger than 65. Employees in these firms who are simultaneously insured under their employers ESI plan and Medicare receive insurance reimbursements first under the ESI plan. Medicare is the secondary payer; that is, it only provides reimbursement based on the billed amounts not reimbursed by the ESI plan

9 scheduled and unscheduled leave. Workers could still obtain paid leave when they are sick, but sickness would no longer be the sole reason for taking paid unscheduled leave. If younger workers are less likely to be sick than older ones, they could still take as much unscheduled leave as older ones, though for reasons other than sickness. Employers might eliminate or reduce the generosity of employee health insurance. However, if employees place high value on this benefit or if the government mandates its provision, as is now the case for employers with 50 or more employees, it may be impractical to eliminate the benefit or make it substantially less generous. Assuming the benefits continue to be offered by employers, are there public policies that might reduce age-related burdens on employers while preserving appropriate incentives for worker selection and retention? The answer depends on the nature of the age-related cost that is associated with a given employee benefit. In some cases it is feasible for employers to avoid paying age-related compensation costs by hiring younger workers instead of older ones. From the point of view of employers and the workers who get hired this may seem efficient. From a social perspective it appears much less efficient if the age-related costs the employer avoids must still be borne by the workers who fail to get hired, probably with generous help from society at large. For example, because of differences in the cost of providing health insurance to young and old workers, an employer may prefer to fill a vacancy with a less qualified 30-year-old rather than a more qualified 60-year-old job applicant. The employer s preference is rational from a private perspective, but it is less rational (and less efficient) from a social perspective. If the jobless 60-year-old will receive almost the same medical care with a public subsidy as she would have received as a covered employee in the employer s health plan, the employer s hiring decision has reduced the number of older workers who are gainfully employed but has not appreciably reduced the total burden of paying for the older job applicant s health care. Assuming that many employers routinely discriminate against older workers and job applicants in order to avoid paying for age-related costs that will ultimately be borne, in part, by taxpayers, it may be preferable from a social perspective for public policy to reduce the agerelated costs that are borne by employers, possibly by subsidizing some of the age-related health costs faced by employers. This issue is less relevant in most other rich countries, which provide health insurance in a way that does not link a worker s health coverage or its cost to the person s employer. Employers in these countries may face age-related costs that make it more expensive - 8 -

10 to employ an older rather than a younger worker, but the difference in the health insurance expense of the two workers is not one of those costs. II. Data and methods In order to determine the added costs facing firms that employ older rather than younger workers, it is necessary to measure employment costs for a representative sample of workers and to determine how much those costs vary with age, holding constant other critical aspects of the employment relationship, including the wage. In the case of health insurance costs, reliable measurement of employer costs at the level of the employee was not feasible before Publicly available data did not provide analysts with detailed and reliable information on individual-level health care expenditures or the sources of payment for such expenditures. Large private insurance companies probably had access to much of the needed information, but their data were seldom accessible to academic researches. The introduction of the MEPS in 1996 greatly improved the availability of information on individual-level health care spending and the sources of payment for the spending. The MEPS is conducted by the Department of Health and Human Services Agency for Healthcare Research and Quality (AHRQ). It collects comprehensive and detailed information on health care utilization, spending on health care and insurance, and sources of payment for personal health care goods and services. In addition, it gathers demographic, health status, and employment information on family members in the sample. The MEPS research program has three basic components, a survey of representative households, a survey of the medical providers who supply services to these households, and a national survey of public and private employers to gather information on the types and cost of employee health insurance offered (Bernard and Banthin 2007). The first two surveys are the most important parts of the program for the current study. They give detailed information on respondents utilization of medical providers, the cost of health care goods and services supplied by providers, and the sources of payment for the health care expenditures of people in the household sample. Researchers cross-check the reports of household respondents against the responses of providers. As a result, the MEPS data files provide much more accurate information about the cost and sources of payment for medical services than would be possible in a survey that relied solely on the recall of household respondents. The MEPS household survey gathers data covering about 15,000 families containing 35,000 individuals every year

11 The household survey collects information from a nationally representative sample of the noninstitutionalized population. Sample members are drawn from the respondents to the National Health Interview Survey. Once enrolled in the MEPS sample, respondents are interviewed in five separate surveys covering a total of two calendar years. The analysis in this paper is based on the overlapping samples that provided information covering the 2010 through 2014 calendar years. The data for each family and worker were organized into annual records showing ESI availability, participation, employee premiums, dependent coverage, and insurance reimbursements separately for each calendar year. The same families and workers could supply information and be included in the analysis for up to two calendar years if they remain in the MEPS panel for two years in The MEPS household survey files give us information on respondents employment, wages, insurance coverage, health premiums, medical spending, and insurance reimbursement. This analysis covers employees who work for pay. The selfemployed are excluded from the analysis, as are adults who do not work for pay. However, because of my focus on employer insurance costs, I gathered information on the health care expenditures and insurance reimbursements of worker dependents who are covered by a worker s ESI plan. This information is attached to covered workers own spending and reimbursement records in order to produce a comprehensive tally of total health care spending in an insured family unit as well as all reimbursement payments under the worker s ESI plan. Some families obtain insurance coverage from multiple insurers. Their insurance coverage may be episodic over the course of a year. This could be because one or more family members had multiple jobs or interruptions in employment that required a change in insurer. To make the analysis tractable, I excluded cases in which shifting insurance coverage made it impossible to determine the exact source of private insurance covering part of a family s medical expenses. If two working members of a family had separate coverage under two ESI plans, I separately tabulated the premiums, health expenditures, and insurance reimbursements under the two plans. These restrictions yield a sample of about 58,000 employee person-years over the five calendar years covered by the analysis (Table 1). This is approximately 11,700 individual respondents per calendar year. On average the respondents records represent the experiences of 107,100,000 adult employees age 25 and older per calendar year. Not all of these employees were offered employer sponsored insurance. Of the workers who were eligible for ESI coverage,

12 many declined the employer s offer of insurance and obtained their insurance from another source or went uninsured. The crucial calculations pertain to ESI reimbursements for health care received by an employee and his or her dependents under a given employer s insurance plan. There are two ways to measure the sum of these reimbursements over a calendar year. The first is as the gross reimbursement payments received by the insured family and the family s health care providers under the employer s ESI plan. The second is as the net ESI reimbursements after subtracting the employee s annual premium contributions for enrollment in the employer plan. The net ESI reimbursement is the best single measure available in the MEPS household survey to capture the variation in employer cost associated with employing older rather than younger workers. The resulting estimates of net ESI reimbursements clearly understate employers full cost of financing ESI plans because they fail to account for the administrative cost of running the plan or for the profit requirements of private firms that manage employer health plans. These costs are not observed in the MEPS household survey. The CBO estimates that only 85 percent of combined employee and employer health premiums is used to pay for medical reimbursement (U.S. CBO 2016). The other 15 percent is absorbed by administration costs and insurers profits. Among smaller firms, enrollees health care claims account for only 81 percent of combined premium payments. Large employers enjoy economies of scale that increase the ratio of reimbursement payments to total premiums. How administrative costs should be allocated across individual employees is an open question. It seems likely these costs are higher in the case of employees who have above-average covered expenses, but employers incur some ESI costs even in the case of employees who do not receive any reimbursements in a given year. The Kaiser Family Foundation estimates that in 2014 the combined employer-employee cost for an average ESI individual insurance plan was slightly more than $6,000 a year (Kaiser/HRET 2017). The cost of an average family plan was about $16,800. If 15 percent of average plan costs were absorbed in administrative expenses, annual reimbursement payouts would have averaged about $5,100 for individual employee plans and $14,300 for family plans. The second and third rows in Table 1 show estimates in the analysis sample of the weighted mean annual ESI reimbursement in three sub-samples of the MEPS analysis sample used in this paper. The largest of the three subsamples consists of employed MEPS respondents 25 and older who meet the broad criteria to be included in the sample. The second subsample

13 consists of employees in the larger sample who were offered ESI at their place of employment. The smallest subsample consists of employees who took up the offer and enrolled in their employer s insurance plan. In the third sub-sample, which consists solely of ESI-insured workers, the average gross ESI reimbursement for a calendar year is $5,261 per worker. Because reimbursement amounts are calculated for five separate calendar years, 2010 to 2014, they are converted into 2014 dollars before averaging. The average net ESI reimbursement is $3,181 which implies the average employee premium payment is $2,080. This is higher than the Kaiser/HRET estimate of average premiums for ESI single-employee plans but below the estimate of average premiums for ESI family plans. While the MEPS-reported employee premium may appear low compared with the Kaiser/HRET estimates, some of the MEPS respondents are reporting premiums for insurance coverage that does not last a full year. The ESI reimbursements reported in the MEPS have shortcomings. Researchers have found that the MEPS survey fails to capture all of the medical expenditures of Americans covered by ESI plans. Aizcorbe et al. (2012) compared ESI-covered medical expenditures reported in the MEPS with health care expenditures reported for an identically selected population in the MarketScan Research Databases. The researchers found that the MEPS expenditure totals for 2005 were 10 percent lower on average than the comparable totals in the far larger sample of ESI-covered people in the MarketScan database. While the authors conclude that MEPS respondents underreported episodes of care in all ranges of the spending distribution, a disproportionate share of the MEPS spending shortfall is traceable to missing observations at the extreme upper tail of the distribution. The MEPS survey uncovers annual expenditure totals for some individual cases in which spending exceeds $300,000 a year. These high-spending cases, though rare, are relatively more common in the MarketScan database. The sensitivity of health spending estimates to cases with high outlays is obvious in the MEPS data. Figure 3 shows the cumulative probability distribution of gross ESI reimbursements for ESI-insured employees who were between 55 and 64 years old in the calendar years. The chart shows the cumulative share of total ESI reimbursements disbursed on this population, with recipients ranked from lowest to highest spenders. Cumulative ESI reimbursements to the 75 percent of ESI enrollees with the lowest spending accounted for 18.4 percent of total reimbursements received by this population. The bottom three-quarters of spenders received an average of $1,760 in gross ESI reimbursements per year, while the top

14 quarter of spenders received average annual reimbursements of $23,470. Another line in the chart shows ESI reimbursements to the bottom 95 percent, who accounted for 55.9 percent of total reimbursements. By implication, the top 5 percent of spenders received 44.1 percent of total reimbursements, an average of $63,470 per year. The top 1 percent of spenders received an average reimbursement of about $130,400 accounting for 18.1 percent of all ESI reimbursements. The concentration of spending is important to bear in mind for two reasons. First, if the MEPS sample underrepresents high spenders, even slightly, the resulting omission can have a sizeable impact on estimates of the average reimbursement payment. Further, even if the overall MEPS sample achieved perfect representation of high spending cases, the rarity of such cases would mean they are almost certainly over-represented in a few MEPS subsamples while under-represented in others. The pattern of over-representation and under-representation can have sizeable effects on relative spending patterns across cells. Median spending levels are little affected by the problem of outliers. However, employers and health insurers costs are driven by the average, not the median, covered health care spending of insured workers. Among all ESI-insured workers in the MEPS sample, about 8 percent of ESI-insured workers between 55 and 64 are in the top 5 percent of ESI-insured spenders. In comparison, less than 3 percent of ESI-insured workers between 25 and 34 are among the top 5 percent of spenders. Although instances of very large ESI reimbursements are comparatively rare for both the young and the old, the fact that they are much more common among the old makes older workers much more expensive to insure. The importance of high-spending cases to health insurance costs makes it helpful to analyze costs using the largest possible sample. For that reason, I have combined information on ESI reimbursements across five calendar years rather than analyze costs within only a single year. The disadvantage of the procedure is that about two-thirds of the unique families analyzed supply information for two successive calendar years. III. Results The cost to employers of providing health insurance to their employees depends on the percentage of their workforce that is eligible to enroll in their plan, the fraction of eligible workers that enrolls, and the cost of plan administration plus reimbursement payments for covered care less any premiums collected from insured workers. The MEPS data file permits us to identify employees who are offered enrollment in an ESI plan by their employers. For

15 employees in the MEPS sample who are at least 25 years old, the probability that workers will be offered enrollment in ESI is only modestly affected by the worker s age once we account for an employee s weekly wage. Workers with very low weekly wages, either because of low hourly pay or short weekly hours, have very low ESI offer rates. Only 22 percent of employees in the bottom one-tenth of the weekly wage distribution are eligible to enroll in an employer s ESI plan. In contrast, more than 90 percent of wage and salary workers in the top three-tenths of the wage distribution are offered eligibility for an employer plan. Employer offers of ESI. Figure 4 shows estimates of ESI offer rates at successive ages for workers in three weekly earnings groups: Workers who are in in the bottom 30 percent of weekly earners, workers in the middle 40 percent of earners, and workers in the top 30 percent of earners. It is plain in the chart that an overwhelming share of employees past age 25 who earn average or above-average pay also receive an offer of employer-provided health coverage. For middle and high earners, workers ages appear to have very little influence on the probability their employers will offer them an ESI plan. Only at ages past 65 do we see a fall-off in the ESI offer rate. This is almost certainly because workers in this age group, who are entitled to Medicare, can afford to choose employers that do not offer ESI. Among workers in the bottom pay group there is a stronger relationship between workers ages and the offer of an ESI plan. Workers between 40 and 64 are more likely to be offered ESI than workers who are younger or older. Except among wage earners at the bottom of the wage distribution, there is little evidence of noticeable difference between the sexes in the likelihood an employee will receive an ESI offer. For employees in the middle and at the top of the weekly earnings distribution, the probability a worker will receive an ESI offer is virtually the same at each age for women as for men. Among wage earners in the bottom 30 percent of the earnings distribution, however, female employees are less likely to receive an ESI offer (Figure 5). This may be because a larger fraction of women who earn low weekly pay are on part-time schedules and hence may be excluded from participating in their employers ESI plan. The differences between men and women may also be traceable to distinctive patterns of the two sexes industrial attachment or job tenure. The MEPS data file permits us to identify an employee s industry and job tenure and the employer s establishment size. Ideally, we would like to know the employer s overall size as well as the size of the establishment in which the employee works. This information is not ascertained in the MEPS interview, but respondents are

16 asked whether their employer has other locations besides the one where the employee works. Basic statistics about these variables are presented in Table 1. The effects of these employee-specific and employer-specific factors on ESI offer rates can be ascertained in weighted logit analysis. Table 2 displays these results. The β parameters confirm that a worker s age has only modest effects on the likelihood of receiving an ESI offer, except in the oldest age group. In contrast, the employee s weekly earnings rank has a consistent, large, and statistically significant impact. Workers at the top of the wage distribution have a greater probability of receiving an ESI offer compared with workers in the middle and especially at the bottom. Note that, unlike the estimates displayed in Figures 4 and 5, the odds ratios displayed in the right-hand column of Table 2 control for other factors in addition to workers age and gender. One of these factors is the industry in which an employee works. Controlling for other factors in the specification, offer rates are significantly higher in manufacturing and public administration; they are significantly lower in construction, transportation, leisure and hospitality, and professional business, among other industries. The findings with regard to employer size are not surprising. Smaller establishments and employers that have only a single location are less likely to offer their employees an ESI plan. Employees job tenure also has the expected effect. Employees with under a year s tenure are much less likely to be offered employer-sponsored health coverage, even controlling for other aspects of the worker s employment. This result may explain the small apparent effect of a worker s age on the offer of employer health insurance. Many young workers have begun to work with their employer only recently, and consequently may not meet an eligibility test for coverage under the employer s plan. As employees job tenures lengthen, employees are more likely to be offered enrollment in an ESI plan. This pattern is similar both for young and old employees of firms of equal size and in the same industry. A key difference between young and old employees, therefore, may be that the young are more likely to be in the early years of their job tenure. Employee enrollment in ESI. An employer s ESI costs are affected by their employees take-up of the offered benefit. Are older employees more likely than younger ones to enroll in an employer plan? The raw statistics certainly suggest this is the case. Figure 6 shows ESI takeup rates, by employee age, among the workers in the MEPS sample who are offered an ESI plan. The take-up rate reaches a peak among year-old workers, precisely the ones likely to face the highest spending burdens for medical care. After age 65 there is a sharp fall-off in the take-up

17 rate. Many employees past 65 who are offered an ESI plan at their place of work may have access to lower cost insurance under Medicare or a previous employer s retiree health plan. Between ages and 55-59, however, the take-up rate of ESI climbs from 77 percent to 83 percent, which suggests employers with an older workforce may face higher costs as a result of higher enrollments. Table 3 presents weighted logit estimates of ESI take-up among MEPS employees who are offered an ESI plan. The explanatory indicator variables in the analysis are the same as those predicting ESI offers in Table 2. As in Table 2, the results in Table 3 imply that wage levels are much more important than age in forecasting outcomes. Controlling for the other factors included in the specification, workers earning the lowest wages are much less likely than wellpaid workers to take-up an employer s offer of ESI. Many low-paid workers may have cheaper insurance alternatives than the plan offered by their employer. The alternatives might include public insurance or dependent coverage under the ESI plan of another family member. In addition, the generosity of an employer s ESI plan may be linked to the average wage paid to its workforce. Low-wage employers on average probably offer less generous insurance, and this may discourage some low-paid employees from enrolling in the employer s plan. In contrast to the effects of workers wages, the effects of their age on take-up are less consistent. For reasons already mentioned, the workers who are least likely to enroll in an employer s plan are those past 65. However, after controlling for other factors that influence take-up, employees between 25 and 29 are the ones with the highest probability of enrolling in an employer plan. Enrollment rates among year-old employees are about the same as those among year-olds. Among workers who are offered insurance and less than 65, those who are between 40 and 49 have the lowest ESI take-up rates. So, while it is true that enrollment rates tend to rise with age among workers past 50, it is hardly the case that workers in this age group have exceptionally high take-up rates once the effects of wages, job tenure, and employer characteristics are taken into account. The number of employees dependents covered by an employee s plan does not rise noticeably with age. Figure 7 shows the proportion of ESI participants enrolled in single employee plans, in employee-plus-one-dependent plans, and in family plans, by age. The tabulations cover only employees in the MEPS sample who participate in their employers health plans. The age group between 40 and 44 is the one with the highest combined enrollment in

18 family plans and employee-plus-one-dependent plans. Combined enrollments in these two plans represent 56 percent of total enrollments in the age group. Enrollments in these costly plans gradually decline to just 36 percent of total enrollments among ESI-covered employees who are between 60 and 64. As we shall see, the total cost of providing insurance coverage to older workers is higher than it is to provide coverage to workers in their 40s, but the reason is not the large number of dependents covered by the older workers policies. Employers reimbursement costs. The association between employers net reimbursement costs and the employee-policyholder s age is displayed in Figure 8. The numbers reflect the estimated gross ESI reimbursements less employee premium payments per ESI-insured employee in each age group. Per-employee reimbursement costs reach a peak of $5,540 for workers between 60 and 64. Costs for all three of the age groups past 55 are well above those of any of the age groups under 55. The net reimbursement cost associated with an average employee between 55 and 59 is almost $1,260 (40 percent) higher than that of employees between 45 and 49. The average net reimbursement for year-old insured employees is $2,300 (70 percent) higher than that for year-old employees. The differences may be due to characteristics of workers employment situation that make older employees more costly to insure. For example, the older workers may disproportionately be employed in industries providing more generous insurance. Holding constant workers earnings, job tenure, and industrial attachment, are older workers more expensive to insure than younger ones? The weighted least squares regression results in Table 4 offer an answer. The regression is estimated with a MEPS sample that contains all employees 25 and older who are enrolled in their employers ESI plans. The specification includes controls for workers gender, age, position in the weekly wage distribution, industrial attachment, establishment size, and job tenure. ESIinsured workers who are in the lowest earnings group receive significantly lower ESI reimbursements than workers earning higher ages, holding all the other factors constant. A bigger influence on ESI reimbursements, however, is the employee s age. Insured workers in the youngest age group receive net reimbursements that are about $1,900 per year below those received by insured workers in the reference group, which contains insured workers between 45 and 49. Insured workers who are 55 and older receive significantly larger net reimbursements than younger workers, even controlling for the other factors in the specification. The estimated effects of employee age on net ESI reimbursements are very similar to the age-related spending

19 differences shown in Figure 8. Thus, inclusion of a variety of statistical controls, including industry, employer size, and workers position in the earnings distribution does not affect the basic conclusion that workers past 55 and their dependents are significantly more expensive for employers to insure compared with workers who are under 55. The estimates just reported show employers net cost of insuring older workers conditional on workers decision to participate in the employer s health plan. A more useful estimate of the extra cost of employing older workers would account for differences in the probability that younger and older workers will enroll in an employer s plan if ESI coverage is offered. Figure 9 shows employers average net reimbursement costs for employees who are offered enrollment in the employer s ESI plan. The numbers in the chart show the estimated net ESI reimbursements per employee offered eligibility for employer-provided insurance. Note that I assume the net reimbursement amount per employee who is offered insurance, but who declines to take it, is $0. (For employees who accept the employer offer, the average net reimbursement costs are indicated in Figure 8.) Of course, employers incur some administrative costs in making ESI offers to employees who decline insurance, but these costs are presumably low relative to the costs of actually providing insurance. The tabulations displayed in Figure 9 show sizeable differences in the cost of offering insurance to older compared with younger workers, though the differences are smaller than the ones displayed in Figure 8. Workers offered insurance who are between 55 and 59 cost $1,220 (50 percent) more per employee than workers between 45 and 49. Workers between 60 and 64 cost $2.060 (82 percent) more in net ESI reimbursement payments compared with similar workers in the younger age group. The weighted regression results displayed in Table 5 show employers expected costs of offering insurance to workers of different age controlling for the effects of workers wage rates, job tenure, and industry and their employers establishment sizes. Workers rank in the wage distribution has a striking impact on expected reimbursement costs. Employees who are in the bottom deciles of the weekly earnings distribution are far less costly than those in the middle and at the top of the distribution, even controlling for age and other characteristics of the worker and employer. The estimates also show, however, that the offer of insurance coverage is significantly less expensive in the case of workers in the youngest age group and significantly more expensive in the case of workers past 55 compared with the expected cost of offering insurance to year-old employees. The estimated effects are very