Treatment of Medical Care Expenditures in Poverty Measurement:
The National Academy of Science Panel Proposal

Erin Kalinosky and Beth Kohler

The current U.S. official poverty measure does not explicitly account for medical care expenditures in either the poverty thresholds or the definition of family resources. It is sometimes assumed that in the United States, Medicaid and Medicare meet the medical needs of the poor and the elderly. Not all Americans have health insurance, however, and many of those who are covered by public or private health insurance have out-of-pocket expenditures or uncovered health care needs (Institute for Research on Poverty, p. 25).

Medical care poses a special problem for those concerned with measuring poverty. At least some degree of medical care is arguably a basic need for most (if not all) persons at some point in their lives. Furthermore, medical care needs are "potentially related to destitution, morbidity, even early mortality" (Institute for Research on Poverty, p. 25). Thus, a poverty measurement that does not take account of medical care needs may ignore an important factor in poverty. There is, however, no easy and widely accepted method of incorporating medical care into the poverty measure. The problem has become more complex since the 1960's because Medicaid, Medicare, and private health insurance programs have grown rapidly since then. The growth of public and private health insurance programs led to a concern that the official poverty measure was overstating the extent of poverty among beneficiaries of health plans, because the official measure does not value their medical insurance benefits as resources. Nonetheless, agreement has not yet been reached on the best approach for doing so (Citro and Michael, pp. 223-24).

The National Academy of Sciences (NAS) Panel identifies a number of difficulties in arriving at a method that retains consistency between the definition of thresholds and family resources and is operationally feasible. Although medical benefits do free resources for other purchases to some extent, they are not as fungible as are other in-kind benefits, such as Food Stamps. Because virtually all households spend more on food than they receive in Food Stamps, the use of Food Stamps frees that amount of a household's resources for consumption of other goods. The same cannot be said of medical benefits. Not all families have medical care needs throughout the year. Medical benefits that pay for relatively inexpensive services may free resources for other consumption, but the "extra" benefit received from private or public insurance to cover expensive services does not free other resources to the same degree. Adding the value of medical insurance benefits to income inflates the measured resources of people, even though they cannot use their insurance benefits for consumption of non-medical goods. Moreover, the insurance value of medical care benefits is greater for sick or disabled persons than for healthy or able-bodied persons. If the value of medical insurance benefits is equivalent to what health care costs would be, then adding the value of medical insurance benefits to income has the perverse effect of making the sick or disabled look better off than the healthy or able-bodied (Citro and Michael, p. 224).

According to the Panel, medical care needs are highly variable across the general population, more so than needs for food and shelter. The Panel suggests that it would be necessary to calculate several thresholds for different groups to capture differences in medical care needs accurately, and all the different thresholds would serve to complicate the poverty measure. Furthermore, medical care needs vary among individuals within groups from year to year. People may require more or less than the "average" amount of medical care for people of their age and/or health status. The possibility of erroneous poverty classifications would thus exist, even if many different thresholds to reflect varying medical care needs were applied (Citro and Michael, p. 224).

Finally, the Panel notes that little research has been conducted on medical-out-of-pocket expenditures (MOOP). Some persons or groups with adequate medical insurance still pay some of their medical expenses directly, in the form of insurance premiums, deductibles, copayments, and payments for uncovered services. These expenditures can be high, yet little research has been done concerning how to adjust the thresholds or the definition of family resources to account for these expenses (Citro and Michael, pp. 224-25.)

NAS Panel Proposal

The NAS panel asserts that the combination of nonmedical and medical needs into a single measure of poverty creates a problem. Measuring nonmedical resources retrospectively assesses a family's actual expenditures on basic goods (food and shelter) during a given time period. Measuring the medical component involves assessing a risk that may not actually materialize (Institute for Research on Poverty, p. 25). Thus the Panel recommends separating the measurement of economic poverty from the measurement of medical care needs and the adequacy of resources to meet those needs. Specifically, the Panel makes the following recommendations:

  • Medical care needs should not be built into the poverty thresholds;
  • Medical out-of-pocket expenses, including insurance premiums, should be deducted from family resources; and
  • A separate index of medical risk should be constructed to measure the adequacy of insurance coverage and the ability to pay for medial services.

While a two-index poverty measure provides a clean measure of nonmedical needs against non-medical resources, it does not entirely solve the problem of how to treat medical needs in measuring poverty. It may be relatively easy to determine a threshold of medical risk, defining a package of basic medical services based on current information--for example, the benefit package identified in the Health Security Act proposed by the Clinton administration, or the largest federal employee health benefits plan (Institute for Research on Poverty, p. 26). (1) Updating the basic package, however, presents a problem--standards for medical treatment have changed over the last few decades, and the pace of medical innovation is increasing while the cost of medical care continues to rise. As the Panel notes, updating the poverty measure for prices alone may be even less appropriate for medical care than for other goods (Institute for Research on Poverty, p. 26).

For the nonmedical measure of poverty, the Panel defines the poverty threshold based on consumption patterns of "basic necessities" (food, shelter and clothing), and then defines family resources as only those resources that can be used to consume the basic necessities represented in the threshold. As medical care expenditures cannot be used toward consumption of the basic necessities built into the threshold, they are deemed "nondiscretionary," and are deducted from family resources. There is a certain rationale behind the Panel's treatment of medical care expenditures as a subtraction from resources. Their measure would not have the perverse effect of making families who spend money on health insurance or necessary medical expenditures seem worse off. However, as Cogan points out in his dissent from the Panel, research in health care economics suggests that medical care is a "normal good"--one that people consume more of as their incomes increase. Cogan thus argues that medical care expenditures, including insurance premiums, are discretionary (Citro and Michael, pp. 886-387). The problem lies in trying to determine whether there is a basic level of medical care that is necessary for all persons, and if so, what constitutes a basic and necessary level of spending on medical care needs. That is, at what point do expenditures on medical care become discretionary? Should experimental or costly new procedures, or procedures that may be in part cosmetic (such as orthodontics), be deducted as MOOP? There may also be one further problem with the Panel's proposal to deduct MOOP from resources: in its initial calculations, the Panel may have underestimated the degree to which that methodology would lead to a rise in poverty rates.

By recommending that medical care expenses be deducted from family resources, the panel suggests that medical care, as a nondiscretionary expense, is a sort of necessity (albeit one without which many poor persons live). The Panel justifies their choice of food, clothing, and shelter as basic necessities on the fact that major government programs provide assistance in these areas. Yet the government also provides medical care assistance to the poor, disabled, and elderly, suggesting that it too is a basic necessity. While the Panel's definition of basic necessities is intuitive and in line with much of the research on poverty, the selection of only these three goods may still be somewhat arbitrary.

The Panel asserts that because medical care needs and spending on MOOP vary highly across the population, they should be excluded from both the poverty threshold and countable family resources Bavier (1998) tested the panel's assertion by examining the variation in spending on MOOP and spending on shelter plus utilities by consumer units with incomes less than twice their poverty threshold (i.e., "low-income" consumer units). He used 1992-93 data from the Consumer Expenditure Survey that overlaps the reference period of the March 1993 Current Population Survey data used most extensively in the Panel's report. He concluded that the data does not support the panel's claim that expenditures on MOOP vary more than that of spending of shelter. (2) This pattern was consistent for all the groups Bavier tested: "couples with two children"; "householder age 65 or older"; and "head or spouse ill, disabled, or unable to work" (Bavier, 1998).

Health care, however, often has an "all-or-nothing" quality. Persons who have insurance may incur some MOOP expenses for uncovered services. Yet persons who have no insurance coverage may be poor enough that they are constrained in MOOP spending altogether, precisely because they lack financial resources. That is, we cannot assume that expenditures on MOOP (including insurance premiums) accurately represent the optimal utilization of medical care needs.

The panel's recommendation is a first step in the direction of incorporating medical care needs into the poverty measure, although meshing medical and nonmedical needs into a single measure presents a number of problems to be resolved.

Alternatives to the NAS Panel Recommendations

Upon the proposal's release, the panel faced immediate criticism for its treatment of medical care expenses. Researchers offered alternative proposals to address issues they found lacking in the proposal. Wolfe developed a proposal that takes into account a person's health care needs rather than actual expenditures. In his defense of the panel's proposal, Betson describes an approach he attributes to Aaron and Burtless--a two-step proposal that measures medical and nonmedical needs separately. Finally, Banthin offers her own proposal, which treats health insurance as a necessary good--similar to food, clothing, and shelter--that should be incorporated into a poverty threshold.

Accounting for Health Care Needs

Wolfe argues that the NAS Panel's treatment of medical care expenditures is flawed because MOOP may respond to factors other than health care needs (Institute for Research on Poverty, p. 29). For example, a poor person may need medical attention but lack access to adequate health care or have other nondiscretionary expenses they deem more important. Therefore, while the person has no out-of pocket expenditures for medical care, he or she may not be effectively utilizing health care and thus may be sacrificing good health. Wolfe argues that health care needs, rather than expenditures, should be included in a measure of poverty. She proposes to calculate these needs and deduct from a family's available resources the amount the family would pay for the health care that is necessary to meet these needs.

Wolfe's proposal identifies a minimum bundle that "consists of those services for which a well-informed person of moderate income would be willing to pay the full costs" (Institute for Research on Poverty, p. 29) to maintain a lifestyle unimpeded by health concerns. The bundle should be calculated using data from actual health plans. For example, the minimum plan offered to federal government employees or a minimum package charged by HMO's could be used to estimate the minimum bundle. The bundle should be adjusted for various attributes that might affect health care needs--age, chronic illness, or disability--and updated as the composition of health care expenditures changes. Wolfe recommends deducting the portion of the bundle that would be paid by the family from that family's available resources. This net income should then be compared to the family's poverty threshold to determine whether the family is poor. Wolfe contends that this measure would more accurately identify those people who are poor.

The Aaron/Burtless Two Step Method

In his defense of the Panel's proposal, Betson outlines a proposal he attributes to Aaron and Burtless. Under the two-step method, families must meet two qualifications to be classified as nonpoor. First, they must have sufficient resources to meet their nonmedical needs. Second, they must also be able to meet their medical needs. If a family's resources (cash income plus nonmedical, in-kind benefits) are less than the nonmedical needs of the family (plus a small allowance for MOOP (3) that would be needed if the family had a "minimally adequate health insurance policy"), that family is "Cash poor" (Betson, p. 14). (4) In other words, this proposal builds MOOP into the threshold, rather than deducting MOOP from resources, as the Panel suggested.

The second step would be to determine if the same family has adequate health insurance coverage. If the family does not have a minimally adequate health insurance plan and cannot afford to purchase such a plan, then the family is "Medically poor" (Betson, p. 14). A family that does not have adequate health insurance or does not have health insurance at all is first classified as underinsured. If an underinsured family cannot meet both its nonmedical ("cash") needs and its need for adequate health insurance (by purchasing a health plan), then the family is medically poor (Betson, p. 14). (5) The family would be classified as non-poor only if it could meet both sets of needs.

Betson acknowledges the difficulty in determining the "adequacy" of a health care plan and assumes, if the proposal were implemented, "adequate" health insurance would probably be replaced by "any" health insurance. Therefore, anyone with a health insurance plan would be considered medically nonpoor (although high MOOP expenditures could make him or her cash poor). He also estimates that this two-step measure would yield a larger population that would be classified as poor, with more working age families who lack health insurance classified as poor, and fewer elderly people considered poor. (Betson, p. 16).

A Proposal to Incorporate the Need for Health Insurance Coverage

Banthin argues government spending on programs such as Medicare and Medicaid supports the assertion that many Americans consider medical care to be a basic necessity. She explains, "Explicitly accounting for basic medical needs along with other basic needs such as food, clothing, and shelter into a revised poverty definition has the advantage of identifying the neediest persons in our society by a single combined measure" (Banthin, 3/23/99). She identifies variation in medical care needs across families as one reason for the difficulty of including medical care in a poverty measure (although, as discussed earlier, this is disputed by Bavier). She also identifies another important factor, however,--the uncertainty that this variation yields. Other basic necessities may face variation, but it is variation mostly by geographic area or by choice. For example, housing costs vary, but most of the variation is by area, and poor families within a particular region face similar housing costs (at least, this is assumed by the Panel's proposal). Furthermore, some people may choose to purchase more food than is deemed "necessary," but few families of four would face a hundred thousand dollar food bill (or, at the other extreme, no food expenses at all). On the other hand, a family might face astronomical medical care expenses if a member of the family became seriously ill, or might face no health care costs at all. Determining a family's current status is easy, but predicting whether the family might face high medical expenses sometime in the future is considerably more difficult.

Banthin contends "the widespread presence of health insurance coverage has thus transformed the need for basic medical care into a need for an adequate health insurance plan that provides access to medical care when necessary." She suggests considering adequate health insurance coverage, rather than medical care itself, to be a basic need. She compares the need for protection against high medical costs to the need for food, clothing, and shelter: it cannot be deferred. Therefore, because the risk of needing medical care is always present (although it may vary to a certain degree from family to family), guarding against that risk is equally as necessary as having food, clothing, and shelter. Conclusion: These medical care needs should be incorporated into the poverty thresholds.

Further research is required to determine which methods are most feasible for categorizing medical insurance needs and incorporating these needs into poverty thresholds. Banthin suggests calculating a "benchmark insurance plan" (4/6/99) that offers a minimum level of coverage. Thresholds should include the amount a family would pay for the minimum plan (including premiums and MOOP), adjusting for health status and family size. To be consistent, the difference between a family's actual MOOP and the predicted MOOP (from the benchmark plan) should be deducted from the family's resources. An alternative approach (also supported by Banthin) suggests calculating an average MOOP, adjusted for insurance status, age, and family size, and incorporating this average expenditure into the thresholds. This approach would create separate thresholds for each category.


Obviously, incorporating medical care into a poverty measure--whether through adjusting for out-of-pocket expenditures or the value of health insurance--is a difficult, yet necessary task. Health care expenditures have grown rapidly over the past twenty years and have affected the available resources of countless families. Because medical benefits are not as fungible as other in-kind benefits, and because medical care expenditures may vary across the population, valuing and including medical care in a poverty measure is a complex procedure. Yet few researchers believe it should be omitted completely from the measure. The panel's proposal to deduct out-of-pocket expenditures from available resources uses the amount a family actually spends on health care in calculating their poverty status. Other proposals incorporate a family's health care needs into the measure, or develop a separate measure for medical care needs. Because poverty rates are extremely sensitive to the treatment of medical care expenditures, it is important to examine various alternative proposals before arriving at a decision as to how to include medical care in the poverty measure.


Banthin, Jessica, E-mail correspondence, 3/23/99, 4/6/99.

Bavier, Richard, "Medical Needs and the Poverty Thresholds," Working Paper, U.S. Census Bureau. 1998.

Betson, David, "Did the Smiths Really Not Keep Up with the Jones? Or Is It True That If You Have Your Health You Have Everything?" Working Paper, July 1998.

Citro, Constance, and Robert Michael, Measuring Poverty, A New Approach. Washington, DC: National Academy Press. 1995.

Institute for Research on Poverty, University of Wisconsin-Madison, "Revising the Poverty Measure," Focus, Volume 19, Number 2. Spring 1998.


1. Of course, there is the issue of distinguishing "discretionary" expenditures and "experimental" procedures from "necessary" expenditures and "standard" procedures.

2. For example, Bavier found that the dollar amount of variation in MOOP expenditures for all consumer units between the 90th and 10th percentiles was $3,530, compared to a $7,887 variation for expenditures on shelter plus utilities.

3. In this case, MOOP (or RMOOP in the formula below) includes only deductibles, copayments, and coinsurance, not uncovered services or health insurance premiums.

4. Here Betson explains if Y=income, NCC=nonmedical needs of a family, and RMOOP=out of pocket payments for medical expenses, a family is cash poor if YNCC+RMOOP.

5. Here, Y=income, PIV=an adequate health care policy, NCC=nonmedical needs of a family, and RMOOP=out-of-pocket payments for medical expenses. A family is medically poor if they are underinsured and Y<NCC+RMOOP+PIV.