Medical Spending, Health Insurance, and Measurement of American Poverty

Critics of U.S. poverty measurement have long complained that the official poverty definition has serious defects. These deficiencies are most apparent in its treatment of health spending needs. Unfortunately, there are no simple approaches to incorporating medical spending in poverty measurement that command wide support among economists and policy analysts. This paper examines the effects of three methods of including household spending on health care in the measurement of poverty. The first is the method embodied in the current poverty statistics. The second is based on the proposal of a National Academy of Sciences panel. The third adds an estimate of “reasonable” medical spending to the modified poverty thresholds proposed by the NAS panel, which include poverty budgets for food, clothing, and shelter. Two conclusions stand out in our analysis. First, the inclusion of medical spending in the poverty definition has a large effect on the level and composition of poverty, providing a very different picture than the one produced using the official poverty guidelines. Groups that are heavy users of medical care, such as the aged and disabled, appear to suffer relatively worse poverty when explicit account is taken of the burden of medical spending. This is true whether medical spending is subtracted from family resourcesas proposed by the NAS panelor approximations of “reasonable” spending levels are added to the poverty thresholds. Under either of these procedures, groups with high out-of-pocket expenditures on health care appear to suffer worse poverty rates than revealed by the official poverty statistics. Second, levels and composition of poverty are comparatively unaffected by the decision to add “reasonable” medical spending to poverty thresholds rather than subtract actual medical spending from family resources. By judiciously selecting estimates of “reasonable” health spending, analysts can derive estimates of poverty thresholds that nearly duplicate the level and pattern of poverty found when actual medical spending is subtracted from family resources. The choice between these two methods of measuring poverty depends on the user’s theoretical preferences, since both approaches to including health spending can produce virtually identical pictures of the nation’s poor. Medical Spending, Health Insurance, and Measurement of American Poverty Controversy has swirled around the measurement of U.S. poverty for at least three decades. Unlike other economic indicators, such as the gross domestic product or the unemployment rate, the poverty rate arouses such intense controversy that government statisticians have been unable to make fundamental improvements in its calculation. The consumer price index is the only other economic indicator that receives a comparable degree of public scrutiny. Political controversy surrounding the measurement of price change has not prevented the Bureau of Labor Statistics from implementing major improvements in measuring inflation over the past two decades, however. Indeed, the political controversy over price measurement probably hastened a technical revision process in the late 1990s that might otherwise have stretched out over several years. One of the most controversial aspects of poverty measurement is the appropriate treatment of personal spending on health care. Patterns of medical care use and of paying for health care have changed significantly since the current poverty measure was developed in the 1960s. In addition, the total resources devoted to health care consumption have also risen steeply, in part because modern medical practice delivers a much improved level of care. The way the government measures poverty has not changed to reflect these developments, however. The current measure of poverty takes no explicit account of consumer medical spending or of the subsidized health insurance that families receive as a result of participating in employer-sponsored or government insurance plans. Critics are divided on how health insurance and medical expenses should be included in poverty measurement. In 1960 medical spending accounted for just 5 percent of national income, but by 1999 this fraction had risen to 13 percent. Medical care now represents a large fraction of all consumption, and many observers believe it has become a necessity at least as important as food and shelter. They believe the poverty definition should accurately reflect this development. If poverty measurement took full 2 account of households’ expenditures on medical care, the poverty rates of the disabled and aged would be particularly affected because of their heavy spending on care. On the other hand, relatively little of the health spending increase was financed directly out of household budgets. Between 1960 and 1999, the proportion of health spending paid out of public budgets more than doubled, and the fraction financed through third-party payments from private health insurers rose almost 60 percent. The actual percentage of health care costs paid as out-of-pocket payments by households fell from 55 percent to 18 percent between 1960 and 1999 (Health Care Financing Administration, 2000). In spite of the dramatic increase in medical care consumption, a smaller percentage of household expenditures is now devoted to health care than was the case in 1960. Many critics of the current poverty measure believe that the consumer value of subsidized health insurance should be included when counting the income available to American households. Depending on how the subsidy is included in income, the resources of many households could be substantially increased and poverty rates reduced. On the other hand, U.S. health insurance coverage is very uneven. More than one in seven Americans, or 42 million people, lacked health insurance coverage during all 12 months of 1999. This paper examines the effects of three basic methods of including household spending on health care in the measurement of poverty. The first is the method embodied in the official poverty statistics. The other two are based, directly or indirectly, on the recommendations of the National Academy of Sciences Panel on Poverty and Family Assistance (Citro and Michael, 1995). That panel argued that the nation’s poverty statistics should be revamped to reflect a new measure of family need and an improved measure of family resources. Its recommendations for treating health insurance and medical spending have not won wide acceptance in the research community, but they offer a starting point for analysis. In the 1960–61 Consumer Expenditure Survey, the share of household expenditures devoted to health care consumption was 6.7 percent; in the 1999 Survey, the share devoted to health care was just 5.3 percent (Jacobs and Shipp, 1990, p. 21; and <ftp://ftp.bls.gov/pub/special.requests/ce/standard/y9399/multiyr.txt> [downloaded on 16 March 2001]). After attending a two-day conference on poverty measurement, 44 social scientist specialists and public policy students were asked to evaluate the recommendations of the NAS panel. Only 40 percent of voting 3 In Section I, we review the definition of poverty and describe alternative approaches to treating household medical spending in an assessment of family needs and resources. We describe the theoretical approach proposed by the NAS poverty statistics panel and outline an alternative to this approach that has been suggested since publication of the panel’s report. In Section II, we describe the alternatives implemented in this paper and outline our methods for calculating household medical spending and health care needs. Section III presents and discusses our statistical results. The paper ends with a brief discussion of conclusions. I. MEDICAL SPENDING AND POVERTY Most social scientists who have studied poverty believe that the official U.S. poverty definition does a poor job of distinguishing between the nation’s poor and nonpoor. The official measure is deficient in a number of respects, a fact that has long been recognized by specialists. These defects can pose problems both for policymaking and for social science. For example, trends in the number of people who are officially classified as poor are often used to decide whether public policies have been effective in reducing poverty. If poverty is mismeasured, this kind of assessment can produce seriously misleading results. Official Poverty Definition The Census Bureau’s current estimate of the official poverty rate is based on poverty thresholds and definitions of countable income developed in the early 1960s by the Social Security Administration and modified by the Council of Economic Advisers. The official poverty thresholds were originally developed by determining the minimum cost of an adequate diet and then multiplying a family’s minimum food budget by a multiplier believed to cover other consumer necessities. This multiplier, in participants—and just 27 percent of all participants at the conference—approved of the panel’s recommendation for treating household medical spending. This is a far lower level of agreement than reported for other elements of the panel’s proposal (Corbett, 1999, p. 53). See also Bavier (2000) and the response by Betson (2000). 4 turn, was derived from a 1955 food consumption survey which showed that families on average spent about one-third of their budgets on food. Part of the remaining two-thirds of spending was devoted to purchasing medical products and services, so in one sense the poverty thresholds reflect Americans’ medical consumption behavior in the mid-1950s. The poverty thresholds vary by family size, under the assumption that large families require more income than small ones to enjoy the same standard of living. To determine whether a family is poor, its resources are compared with the poverty threshold. The family resource measure used by the Census Bureau is gross money income. It includes before-tax ca