This session focused on issues involved in defining resources, such as what is included in income in determining medical care economic risk in terms of ability to pay for insurance and for out-of-pocket medical expenses, how to treat assets in addition to income, what constitutes income for the self-employed, and how the elderly and other groups finance medical care.
The presentations covered three topics related to defining resources:
- Incorporating data on assets into measures of financial burdens for health 2.
- Findings from the Health and Retirement Study on changes over time on how the elderly finance medical care 4.
- The financial burden of medical care among the elderly transitioning to long-term care 6.
INCORPORATING DATA ON ASSETS INTO MEASURES OF FINANCIAL BURDENS FOR HEALTH: IMPLICATIONS FOR THE ELDERLY, THE NONELDERLY, AND THE SELF-EMPLOYED
Jessica Banthin (Congressional Budget Office) provided an overview of the background paper she and Didem Bernard prepared for the workshop (see Part III of this volume). She stated that although the data she presents are based on tabulations for people ages 65 and older, a more precise analysis would focus on people who are retired versus nonretired. But their main objective, she said, was to think more carefully about how one would
combine the elderly and nonelderly into a single measure. All of the previous estimates with which she is familiar always look at them separately. In her previous work, different thresholds were applied for the nonelderly and the elderly populations; for the nonelderly, for example, cutoff points of 10 and 20 percent of income spent on medical care were applied (Banthin and Bernard, 2006). She has also done work in which cutoff points of 5 percent were applied to low-income people and 10 percent for others (Banthin, Cunningham, and Bernard, 2008). In a paper that looked at annual burdens for the elderly, she and a colleague actually applied different cutoff points of 20 and 40 percent to the nonelderly and the elderly, respectively because the elderly spend a lot more on health care (Selden and Banthin, 2003).
Why differentiate between the elderly and the nonelderly? Younger families clearly have higher incomes, because generally they are working. They have higher expenses, including work-related expenses and child care, than the Supplemental Poverty Measure recognizes. Also, they are expected to be saving for their future retirement. And of course generally they are in better health, so their medical care needs are lower.
For older families, if they are retired their incomes are lower, they have fewer competing expenses along the lines of work-related and childrearing expenses, and their health needs are generally higher and they are in generally worse health. At the point at which they retire, they have been building up their assets, and now they are expected to draw down their assets in some way or other. That is why the elderly and the nonelderly have always been separated.
Two questions need to be answered. First, how is a reasonable cutoff point or threshold defined for both the elderly and the nonelderly populations that would indicate high medical care risk or high burden? Second, how does one incorporate the accumulated savings of retired families into the measure of resources available for financing health care expenditures?
Banthin addressed the self-employed briefly. No one had really looked at them separately, or some people have but not in the context of financial burden, she said. The question is whether the self-employed have such high levels of assets (including business assets) that they warrant a separate approach in measuring health care burden.
Data and Methods
The Medical Expenditure Panel Survey (MEPS) is used for this analysis, taking advantage of the asset data that it collects. MEPS is designed with 2-year panels that overlap one another, and Banthin and Bernard pooled three panels—panels 10-12, 2005-2008.
They used three panels as the sample size, that is, about 1.5 times an annual sample from MEPS, which most people use. MEPS is a unique
resource for studying these issues because it has information on income, assets, out-of-pocket medical expenditures on health care services, and also out-of-pocket spending on premiums, plus various demographics and family composition information. In working on this paper, they followed the methods of previous papers in constructing a measure of out-of-pocket financial burden. It is not necessarily a risk index; it is a family-level concept, because family members share resources. The numerator is the sum across all family members of out-of-pocket spending on health care services and premiums. The denominator is family income. The burden is the share of family income spent on medical care, and the resulting values are not truncated. There are certainly cases, as seen in the data, in which people are spending more than 100 percent of income in a given year on health care. Those cases are few, and they are plausible in some circumstances. Banthin explained that because they were charged with thinking about assets, she did a variation on her traditional burden measure whereby she added 5 percent of the total net family assets value to family income for elderly families only. No adjustment was made for the nonelderly. The justification is that nonelderly families until retirement are expected to actually be saving toward retirement. Once they reach retirement, they are drawing down. It is at least an approximation of a way of addressing and incorporating the assets into burden measures.
Elderly Versus Nonelderly
If the median is considered the norm for the elderly and nonelderly populations, then the percentile distribution of out-of-pocket burdens for the two groups might suggest that different thresholds are appropriate for defining high burden. Overall, the median financial burden for elderly persons is 10.7 percent of family income compared with 2.9 for the nonelderly. This means that the median individual age 65 or older lives in a family that spends almost 11 percent of family income on medical care. As expected, younger families devote a much smaller share of family income to medical care. These differences between the elderly and the nonelderly are similar across poverty groups at the median; however, they increase when examined at the higher points in the distribution. At the 75th percentile, the elderly are spending almost 21 percent of family income on out-of-pocket medical care, compared with 7 percent of family income for the nonelderly.
As one would expect, examining the distribution of total net assets by family age groups, at the overall median elderly individuals reported $146,000 in family net wealth, and nonelderly individuals reported $20,000. Thus, elderly individuals have about 7 times as much family net wealth as do
nonelderly individuals. Among families living below the poverty level, elderly individuals reported more than $20,000 in family net wealth at the median, and the nonelderly reported zero. Among low-income families (between 100 and 199 percent of poverty), the median family net wealth for an elderly individual was about $77,000 compared with $2,300 for the nonelderly or about 33 percent times as much as that for a nonelderly individual.
Banthin emphasized that the MEPS asset data are comparable across various dimensions with which to assess data reported from the Survey of Income and Program Participation. Both surveys show underreporting of assets compared with the Survey of Consumer Finances, but as survey data go, these assets are reasonable. MEPS has data on both the elderly and the nonelderly and measures different categories of assets, so if there is underreporting, it is measured without bias across the two age groups.
Banthin next discussed high out-of-pocket burdens for the elderly and the nonelderly using the cutoffs of 10 percent or more of family income on medical care. Overall, about 52 percent of elderly and 17 percent of nonelderly individuals have high burdens according to this threshold. Using the 10 percent of family income cutoff would more than triple the number of elderly having high burdens, and this persists across different poverty status groups. About 26 percent of elderly and 7 percent of nonelderly individuals live in families spending 20 percent or more on medical care. Spending actually goes up among the low-income elderly: 70 percent of them are spending more than 10 percent out-of-pocket compared with only 22 percent of the nonelderly. The poor elderly may be eligible for various programs, such as Medicaid and Medicare, so they may have more coverage than the low-income elderly.
Banthin observed that the self-employed population under age 65 does not have substantially higher burdens than their non-self-employed counterparts. They have slightly higher burdens than the non-self-employed (median burden of 3.3 compared with 2.9), but this difference is not large. The self-employed do report higher net family assets, and they also have higher levels of average income. These are net business assets at the family level.
Banthin concluded that further work is needed to refine the method of incorporating assets into income for elderly families. In her view, it is worth considering different thresholds for the elderly and the nonelderly
age groups. The self-employed, however, do not warrant special methods for assessing their medical care risk.
FINDINGS FROM THE HEALTH AND RETIREMENT STUDY ON CHANGES OVER TIME ON HOW THE ELDERLY FINANCE MEDICAL CARE
Michael Hurd (RAND) introduced himself as a coprincipal investigator on the Health and Retirement Study (HRS), with overall charge of the income and asset sections. After a brief background on HRS, Hurd explained how the survey assesses out-of-pocket medical spending and how those data compare with data from MEPS and the Medical Care Beneficiaries Survey (MCBS). He illustrated the application of HRS data to explain economic preparation for retirement with and without health care spending risk.
Background on HRS
The HRS is a very large survey aimed at people over age 50. It has interviewed about 20,000 persons every 2 years in panel since 1992. So by now it has 10 waves of information on the original sample. New cohorts were added in 1998, 2004, and 2010, filling out the population age 51 and older plus spouses.
Although the initial sample was drawn from the community, respondents are followed into nursing homes. After a few years, it is representative of the entire population, including the nursing home population, depending on mortality, condition, and nursing home status.
The HRS goes to considerable effort to measure income and wealth, including pensions, and those techniques have been refined over time because at older ages wealth is more important than income. HRS data matches the Survey of Consumer Finances quite closely, except at the very top, and that is because of the oversamples of the Survey of Consumer Finances from the high income supplement. It also matches the Current Population Survey (CPS) income data very well, and it is linked to Social Security records.
Additional content in the HRS covers a wide range of topic areas. The main ones of relevance for this session are health conditions, including cognition, the use of health care services, out-of-pocket spending for health care services, and formal and informal care (who gives care, family member or paid help, and out-of pocket cost if paid). These data are linked at the individual level to Medicare data.
Out-of-pocket spending is assessed in the HRS core interview by first asking about the use of services and then if the costs were paid by insurance.
If the answer is no, the respondent is asked what were the out-of-pocket costs, proceeding through the list of services used. It should be noted that the HRS and the MCBS have considerably larger sample sizes than MEPS in this population group because they specialize in the older population.
Hurd showed how the HRS measure of out-of-pocket spending compared with the MCBS and MEPS. Data on annual per person out-of-pocket spending for health care services by the noninstitutionalized population ages 75-79 in the HRS compared with MEPS and the MCBS showed that the median lined up very closely for all three surveys. The HRS, however, had a higher mean than the other two surveys, and that is because of the large values in the HRS at the top of the distribution due to prescription drug costs. The measurement of prescription drug costs is difficult in any survey, and it is particularly difficult in a general social science survey like the HRS, because of the heterogeneity in spending across individuals, and even within an individual over time. This measurement has been improved in HRS 2006 and later, reducing the high values.
Data on annual per person nondrug out-of-pocket spending for health care by the noninstitutionalized population ages 75-79 in the three surveys showed that HRS and MEPS are comparable except at the very top.
However, examining out-of-pocket health care spending among the elderly population, one should also include the institutionalized population. MEPS obtains out-of-pocket spending data in its noninstitutionalized population survey that are to be excluded in the comparisons. Data comparing the annual per person total out-of-pocket spending for health services by both the institutionalized and noninstitutionalized populations ages 75-79 showed the HRS and the MCBS to be very close. Again, this is due to the much higher measurement or assessment of prescription drug cost in the HRS. If prescription drug costs are excluded, then the HRS and the MCBS once again differ, with the MCBS being considerably higher than the HRS.
Hurd’s conclusions about measurement of out-of-pocket spending in health care are that HRS does very well, given its limited resources, compared with the 2003 MEPS and the 2003 MCBS. The higher total costs are due to higher drug care costs in the HRS. For nondrug out-of-pocket spending, the HRS and the MEPS are similar, and the MCBS is considerably higher. Comparing the 2004 HRS with the 2003 MCBS for the noninstitutionalized and institutionalized populations combined, total spending was similar, but nondrug out-of-pocket spending was considerably higher in the MCBS.
Persistence of Spending Over Time and Economic Preparation for Retirement
Hurd explained that the HRS allows analysis of persistence of out-of-pocket spending over time. For example, combining 2-year panels, he found
a lot of stability in spending for single people and also older people. For married people, the stability was lower because they tend to be younger. Nonetheless, the conclusion is that there is a lot of cross-wave stability in spending, which needs to be taken into account in assessing health care spending risk.
Using results from a paper he coauthored on the economic preparation for retirement, Hurd showed how these data combining income wealth and out-of-pocket spending can be applied to see what difference risk makes in a common assessment of economic status. The objective was to ask whether people continue on a life-cycle spending path shortly after retirement, given the initial level of spending observed in the HRS. Will they be able to afford their life-cycle spending path, or will they run out of wealth? Starting with an initial population, ages 66-69, individuals and couples, their life-cycle spending paths estimated from spending data were followed over the years. Life-cycle spending path was anchored at the initial observed spending level. The paths differed by marital status and education level. The question is whether, as they progress through life and spend and receive income from assets and so on, they will run out of wealth before they die.
Although the focus is economic preparation for retirement, Hurd and his colleagues did stochastic simulations on mortality, which is a very important aspect of this. Poor elderly live substantially shorter lives than the well-to-do elderly, and so they need fewer resources to finance retirement. They account for mortality along the dimensions of taxes, returns to scale and consumption, and the level and risk of out-of-pocket spending for health care.
Hurd proceeded to explain the simulation. First, they estimated serial correlation using the MCBS, and the range was from 0.41 to 0.73, fairly high levels of serial correlation for spending one year apart. The lowest levels were for the youngest married people, and the highest levels were for the oldest single people, who had chronic conditions that caused their spending to be more persistent than younger people. They wanted to take the serial correlation into account, because they were looking at the lifetime risk of running out of money.
They simulated consumption and out-of-pocket spending. A couple or individual is considered adequately prepared if they die with 95 percent or more of their wealth at the time. The researchers simulated stochastic spending using the observed distributions from the HRS in out-of-pocket spending. They put in predictable spending, that is, spending for health care insurance, as part of normal expenditures. So the stochastic part is the part that deviates from average.
Hurd showed, as an example, simulations for couples ages 66-69 and their resources when they are 66-69. They had about 1,100 individuals who were in couples in the initial sample. Their initial wealth was $742,000. On average, these people had $1.2 million in rest-of-lifetime resources.
The distribution by education level was extreme, ranging in the total from $564,000 among couples with less than a high school education, to more than $2 million for those with college degrees and above. Both initial wealth, but also annuities, particularly Social Security, are important, depending on age group. For the less than high school group, wealth was about the same as annuities. For people with college education or more, wealth was substantially more than annuities.
Taxes need to be accounted for when considering spending. For people who have more than a college education, taxes were a very substantial and important part of the calculation. Part of that importance comes from the taxation of tax-advantaged savings. As these savings come out, they have to be taxed. The present value of consumption for some was over $500,000, and total spending was $681,000. If total spending is subtracted from resources, they had about $500,000 left over. This is average and does not tell anything about the distribution.
The simulations showed the chances of people running out of wealth before they die. Average spending for health care was unchanged both with and without health care spending risk. It is the distribution that changes, so people have draws that push them out into the tails. Once they are out in a tail, because of the high serial correlation, they tend to stay out in the tail.
For single persons ages 66-69 with no health care spending risks, 61 percent were adequately prepared. A more interesting question for this group is what difference it makes to have stochastic variation in health care spending—that is, to have health care spending risk, rather than just assured level. If everybody were perfectly insured, the number would be 61.1. Because people are not perfectly insured, the number would be 54.5 percent, or about 7 percentage points lost in adequate preparation for retirement because of health care spending risk.
Hurd pointed out that it may be exaggerated at the population level, but there certainly are groups in the population that are very inadequately prepared for retirement and will have to reduce spending at some point. For example, just 29 percent of single women who lack a high school education are adequately prepared. Married persons show a much higher preparation for retirement. Also, there is much less effect of health care spending risk on economic preparation for retirement.
Based on his analysis of the data, Hurd concluded that health care spending risk has a noticeable effect, but possibly not as great as one might expect. One reason may be that the first-order serial correlation does not adequately capture persistence in spending over many time periods. There are now enough data in the HRS to nonparametrically estimate rest-of-lifetime spending risk. There are data for people age 61 in 1992, who will be 81 in the next wave. There are also data for people age 70 in 1993, who will soon be age 90. So one can map out what 70-year-old people actually
spend over the rest of their lifetime and see what that number actually is, rather than modeling it.
He emphasized that the HRS is a vehicle for assessing health care spending risk over time. First, it should be obvious that one needs panel data. A cross-section is not adequate because of serial correlation. How long a panel depends on the structure of the intertemporal correlation in spending risk. For the first-order mark-off only two waves are needed, but one probably wants more than that.
Hurd pointed out that if one is interested in relating spending to economic resources, one ought to spend as much effort assessing economic resources as health care spending. That is actually a harder job, as there is a lot of measurement error in income and wealth measurements in all data sets. He would not include the CPS in that, but would include the Survey of Consumer Finances and the HRS.
FINANCIAL BURDEN OF MEDICAL CARE AMONG THE ELDERLY IN TRANSITIONING TO LONG-TERM CARE
Eric Stallard (Duke University) focused his presentation on the long-term care population and the financial burden of medical care among the elderly in transitioning to long-term care. Using data from the National Long-Term Care Survey (NLTCS), he defined the chronically ill population, briefly described the survey and the population studied in the NLTCS, and provided numerical results from that survey.
Stallard pointed out that the Health Insurance Portability and Accountability Act (HIPAA) rules for tax-qualified long-term care services and insurance policies define a chronically ill individual as someone who meets either an activity of daily living (ADL) trigger or a cognitive impairment trigger (Internal Revenue Service, 1997). Although chronic illness is important, in terms of expenditures and costs, in his view they are not being represented in many of the measures that are being discussed today.
There are six ADLs that are fundamental to functioning on a daily basis: bathing, dressing, toileting, continence, eating, and transferring (i.e., getting into or out of a bed or a chair). In order to be certified as a chronically ill individual, the HIPAA ADL trigger requires that the individual be unable to perform without substantial assistance from another individual at least 2 out of the 6 ADLs for at least 90 days due to a loss of functional capacity (Internal Revenue Service, 1997).
The HIPAA cognitive impairment trigger requires that an individual needs substantial supervision to protect himself or herself from threats to
health and safety due to severe cognitive impairment, defined as a loss or deterioration in intellectual capacity that is
(a) comparable to (and includes) Alzheimer’s disease and similar forms of irreversible dementia and
(b) measured by clinical evidence and standardized tests that reliably measure impairment in the individual’s short-term or long-term memory; orientation as to people, places, or time; and deductive or abstract reasoning (Internal Revenue Service, 1997).
Individuals who are certified as chronically ill because they meet the ADL and/or cognitive impairment triggers are eligible for tax-free benefits under a long-term care insurance policy, and they can deduct the costs of qualified long-term care services and insurance premiums as itemized medical expenses, subject to certain limitations, when filing their federal income tax returns for that year (Internal Revenue Service, 1997).
The National Long-Term Care Survey
The purpose of the National Long-Term Care Survey was to measure disability and use of long-term care among the noninsured elderly (ages 65+) at multiple points in time beginning in 1982 and every fifth year from 1984 to 2004 (Stallard, 2011). The survey was stopped after the last round of data collected in 2004. Stallard commented that there continues to be passive monitoring of Medicare and Medicaid services and expenditures, on which he has drawn for his presentation.
The total cumulative sample was about 49,000 people over all six surveys. In the 2004 survey, the total sample was nearly 16,000 people, with 6,171 detailed in-person interviews for persons who met various screening criteria and a shorter, mostly telephone interview for the 9,822 persons who screened out.
The disability definitions that were used include ADL and instrumental activities of daily living (IADLs) limitations for at least 90 days, cognitive impairment, and institutionalization. Although the IADLs differed from the ADLs, they are still daily activities and primarily focus on the maintenance of daily life and daily lifestyle, with a very strong cognitive component— doing laundry and light housework, getting around outdoors, going places beyond walking distance, making telephone calls, managing money, preparing meals, shopping for groceries, and taking medications (Stallard, 2011).
The first six ADLs in the NLTCS are the same as those listed in the HIPAA ADL trigger; the seventh one, inside mobility, is effectively equivalent to walking and is not included in the HIPAA trigger.
The Survey Population
The key questions for the survey were two: Who are the elderly, and how homogeneous or heterogeneous are they with respect to these disability measures? The survey was statistically weighted up to match the elderly 2004 U.S. Medicare-enrolled population. The average age was approximately 76 years.
Stallard presented unpublished tabulations of the NLTCS which showed that the mean age by disability status for persons meeting only the HIPAA ADL trigger was 79.5 years for men and 82.0 years for women; for persons meeting only the HIPAA cognitive impairment trigger, the mean age was 82.5 for men and 84.1 for women, and for persons meeting both the ADL and the cognitive impairment triggers, the mean age was 81.7 for men and 86.0 for women. For those who met both triggers at the same time, the average age was actually slightly younger for men than for the cognitive impairment trigger, 81.7 versus 82.5, but older for women, 86.0 versus 84.1. The standard deviations for these measures ranged from 6.7 to 8.6 years.
Among people ages 65 and older, there was substantial variability by age in the proportion meeting either HIPAA trigger. Of the population ages 65 and older, 10.1 percent met at least one HIPAA trigger, but this ranged from just 2.8 percent at ages 65-69 all the way to 58.7 percent at ages 95 and older. When the data were analyzed by considering each trigger separately, the overall percentage meeting the ADL trigger was 8.2 percent, and the overall percentage meeting the cognitive impairment trigger was 6.7 percent. The modest drop from the 10.1 percent that met one or both of the triggers considered jointly indicates that there was substantial overlap between the two types of impairments.
Long-Term Care Intensity and Costs
Stallard next presented data on costs (in 2010 dollars) for people with and without disability, according to their status on the HIPAA ADL and cognitive impairment triggers based on Stallard (2011). Persons with both ADL and cognitive impairments can expect to spend $45,000 per capita annually for nursing home services. Their annual per capita cost of paid community care was estimated at $5,050, of which $1,360 would be paid out-of-pocket, with an average out-of-pocket cost of $16,548 for the 8.2 percent of the group who actually make out-of-pocket payments.
Estimated unisex lifetime costs of long-term care services at ages 65 and older were estimated to be $89,000 (also in 2010 dollars). However, the sex differences were substantial: for men, the estimated cost was $44,000, and for women, $124,000. The overwhelming majority (92 percent for
both sexes combined) of long-term care costs was incurred during episodes of disability severe enough to meet at least one of the HIPAA triggers. The remaining costs (8 percent) were incurred during episodes of mild or moderate disability, which would not meet either of the HIPAA triggers.
He next looked at the Medicare program expenditures, excluding payments for persons with end stage renal disease, payments made while in long-term institutional status, and payments for hospice care based on his unpublished tabulations of the NLTCS. The retained payments included only the components that were used in setting capitation rates for managed care plans.
The average annual overall unisex Medicare program payment (in 2010 dollars) for ages 65 and older was estimated at $9,071. For the “dual eligibles,” which include people who were enrolled in both Medicare and Medicaid, the average was $11,954, and, for the Medicare-only participants (i.e., not enrolled in Medicaid), the average was $8,761. For men, the average annual cost was $9,787, and for women, $8,550. Stratification by age, Medicaid status, and disability showed that the highest average annual unisex cost was $31,940 for dual eligibles ages 75-84 who met both HIPAA triggers.
Marilyn Moon (American Institutes for Research), session discussant, observed that she has been working on some of these issues for many years, and the discussion has changed over time to include two different concepts, risk and burden. The risk versus burden issue is important, because they measure different things. Also, they are aimed at different things, and probably should be applied in different ways. That does not mean that one is better than the other or that one excludes the other, just that it is important to think about what one wants to do and to whom to apply these things and in what circumstances.
Burden, she said, is the general ability to meet standard expenses over time, averaged out somehow, thinking about what one should be planning for, about how to measure what in public policy is reasonable for people to bear. That is particularly important for poverty and for poverty measures and has always been tough, because in fact nobody has the average burden, and everybody is either below or above it for the most part. That is one of the reasons why poverty discussions did not ever decide how to deal with medical expenditures very well; people were not fully satisfied that this was the measure that public policy was after.
Now, some 20 some years later, people are talking about this issue more in terms of the risk of dealing with the unusual or catastrophic expenses associated with health care and how to measure how well people are doing
with risk as a society. Risk affects both poor and nonpoor individuals and therefore needs to be thought of as something different from burden. Risk affects not only expenditures, but also resources.
If someone experiences some massive risk in the middle of life, chances are his or her resources going forward are going to be substantially lower. Data about the number of people who file for bankruptcy protection show this, for example, because of medical care expenses throughout their lifetime. Or, in the case of Medicaid, a person essentially spends down and then is poor for the rest of his or her life. If this problem was difficult 20 some years ago, when most of the focus was on burden, when risk is added fully into the discussion, it becomes even more complicated.
The situation is a researcher’s dream because, as Michael Hurd pointed out, one has to get lots of things right to do all of this—the resources, the expenditures, and the risks. One has to think about things through time, as well as at one point in time. And all of this is very challenging to deal with. Both burden and risk are likely to be of interest, but one needs to think about when to use what, when, and whether or not one is doing the right things in terms of measuring them, and whether or not people are using the wrong measure.
In health reform, for example, the study panel will be dealing with both burden and risk. With respect to burden, what is reasonable to ask low- and moderate-income people to pay toward their own health care? That is a burden issue, to consider what kinds of general protections to offer to individuals, when all of the discussion focuses on 6 percent or 10 percent of income or whatever comes into play. That is an important issue, and one to be concerned about particularly around the issue of low income and poverty.
In terms of dealing with risk, the quality of health care coverage, of health insurance, is important. The goal is not only lowering the average burden on people, but also taking into account untoward risks and not just allowing one new sneaky way of cream skimming that is going on in the marketplace.
Moon cautioned that once a measure exists, it is irresistible not to use it for comparisons, and that can cause problems. She gave as an example her experience 25 years ago hearing people say that older Americans are not really poor because they have Medicare and Medicaid, so that problem is solved. But digging down below the surface and looking not only at the value of Medicare and Medicaid benefits on the resource side but also at out-of-pocket costs on the expenditure side, one recognized that the issue is more complex.
It also matters now, with respect to comparisons between the elderly and the nonelderly going forward and, in terms of health reform, what changes to make to Medicare compared with provisions in the ACA for
the population under age 65, and whether the policy is fair in the same way.
Another major comparison is across geography in terms of cost of living and the costs of health care. That area has a lot of unknowns: Is the average right, is the lowest level of spending right, is the top level of spending right, and what drives those differences in health care spending in many cases?
It is important to think about comparisons over time. If risk is to be used to measure the value of health reform in providing various protections, then measures need to capture changes through time. Comparing different kinds of health care needs is also important. Is risk really more important for people who have untoward, acute care consequences—a car accident, a surgery gone wrong—than for chronic conditions for which the burdens are high, increasing, and persistent?
In closing she had a question for Jessica Banthin: When talking about treating resources differently for the elderly and the nonelderly, what is the right age cutoff? Is age 75 the new 65? Also, there is the issue of when people spend down their assets. She gave as an example the case of some of her friends with children in college, who are not talking about spending down, but about working and accumulating assets.
FLOOR DISCUSSION AND COMMENTS
Banthin responded to Moon’s question that she would have preferred to do the adjustment based on retirement status. That is the proper way, she said, because eventually the cutoff should be age 67, not 65. When people decide to stop working is when they have stopped saving and have moved into retirement, when they are presumably spending down. She thought that would be the most accurate cutoff point.
Hurd commented on Banthin’s numbers about the different levels of spending on health care for the elderly and the nonelderly. One thing to keep in mind—and this shows up very clearly in the Consumer Expenditure Survey as well as in HRS spending data—is the budget shares that go to health care spending. Of course, the shares do increase with age, and that is quite reasonable, he said. There is no reason they should be the same for a 45-year-old as an 85-year-old; it is much more productive to spend on health care at age 85.
At the same time, other components of the budget correspondingly decline, which is also very reasonable. For example, spending on private transportation declines from around 15 percent of the budget down to 6 percent. So it is not prima facie a problem that the older population spends a greater fraction on health care; it is because it is more productive. Whether it should be more than 10 percent or more than 20 percent is not known.
Barbara Wolfe asked Hurd: If you compare the distribution of prime age individuals, say 25-45 years old, do you know what the distribution of the HRS looks like? What proportion, for example, in the lowest quintile, are actually in the HRS? Has anyone ever tried to do that kind of comparison? How useful would some of the numbers presented be, with respect to the entire age distribution? The HRS is a really rich data set, but if it misses the lowest tail, then it is less promising.
Hurd responded that the poverty rate in the HRS is very close to the CPS poverty rate, within half a percentage point. He explained that they reweight to CPS totals along a number of dimensions; and that has been studied a lot, and there is no known substantial bias in HRS recruitment. The baseline response rate was 80 percent. Very large differential nonresponse by some variable is therefore needed in order to get a lot of bias when the response rate is that high.
Wolfe wanted to emphasize that, in the work that she and colleagues have done at the University of Wisconsin, they looked at the risk of having income go below the poverty line or move into near poverty if someone has a health or cognitive effect. They found that the risk is strong for those who are very close to the poverty line, but it is not relevant for people who are at 400 times or even 300 times the poverty line. So the risk is important for a small group.
Although she is sure there is a lot of overlap between people with less than a high school education and those with low income, if he did that by initial level, maybe in the first year they are in the survey, or some average, he will probably find higher risk for people with low income than he calculated using education.
Hurd explained that they thought about doing the analysis by income or wealth but decided not to because of the classification error on income and wealth. They do as much as they can, but they have to admit that it is inaccurate. As a number of the presentations today have shown, people with low incomes have a high ratio of spending to income.
Michael O’Grady (NORC at the University of Chicago) asked Jessica Banthin when she moved to adjusted income and used 5 percent of assets, was there anything special about 5 percent or she just needed to pick something to move forward? Also, because part of what has been discussed or at least implied is that different assets are more or less fungible or available, or even originally planned to be used for things like health care spending, what are the options in terms of thinking about different kinds of assets using MEPS?
Banthin explained that, regarding the adjusted income and 5 percent of assets, that was just a rough approximation. She wanted to move forward and give a simple approach to provoke this conversation. She repeated that a more careful analysis would have separated by retirement status, not age.
She picked 5 percent, however, because she talked to some people about what is reasonable. Some financial planners would tell people to draw down 4 percent a year, so 5 percent is close to that. They probably would not tell people to liquidate their investment in their house. So she used total net assets to make it easy. But a more refined analysis might separate pension assets from other types of assets and might account for having a spouse and so forth, leading to a much more complex projection. Also, there are different ways one might expect people to spend down. There is also significant underreporting here in both income and assets.
James Ziliak asked both Hurd and Banthin if they included Social Security wealth in their definition of assets. He questioned Hurd’s statement that 61 percent of people ages 66-69 are adequately prepared for retirement. The number he recalls is more like 80 percent, based on analysis by Shultz and Sheshardi, who are using the full HRS. He asked whether this 20 percentage point difference of adequacy is a different calculation, or whether it is something that is happening with using a smaller subset of the HRS.
Hurd responded that neither he nor Banthin included Social Security as a wealth measure. Both included Social Security as an income measure, which is the proper way to do it, in his view. The 61 percent figure for adequacy is for single persons only. The figure for married persons is in the 80s. Adding the two together, one gets about 72 percent. Schultz and Sheshardi have come down somewhat in their number; they are a little bit higher, but not that much higher.