In a recent posting on the Census poverty measurement working papers web page, David Betson provided a forceful restatement of the National Research Council's rationale for proposing a poverty measure with no amounts included for health care.1 No doubt there will be much more debate before we know which of the NRC's proposals, including a recommended separate health care measure, will be adopted as the official measure of poverty. The current Census program of experimental poverty statistics based on combinations and variations of the NRC proposals should inform that debate. Other postings on the working papers web page present reasons and data in support of adding to the experimental data program a threshold variation that includes something for medical out-of-pocket expenses. Instead of that issue, what follows will address a question raised explicitly for the first time in Betson's paper: Do the current poverty thresholds implicitly include any amounts for health care?
The question may have practical consequences as well as historical interest. Many experts have supported a strategy of selecting the dollar level for a new poverty measure such that, in the year of transition from the old measure, the old and new thresholds produce the same overall poverty rate. The NRC panel's report, Measuring Poverty, A New Approach, 2 illustrates the distributional effects of its proposals by setting the level of the new thresholds where they produce the same overall poverty rate as the official rate for 1992. The Census experimental poverty reports have adopted a similar approach of "benchmarking" the experimental variations so that they all yield an overall poverty rate comparable to the official rate and to each other.3 And, in an August 2, 2000 open letter to the Director of the Office of Management and Budget and the Director of the Bureau of the Census, 42 poverty experts recommended that the new thresholds, "be set at a level that at least initially produces a total count of the number of people in poverty that does not vary substantially from the poverty count obtained using the existing official poverty measure."
If the current thresholds include something for health needs, whether or not the implicit amount is sufficient, then, all else being equal, a new poverty measure, such as the NRC proposal, that included nothing for health needs but yielded the same overall poverty rate would represent an increase in the level of economic well-being we regard as the poverty line. So the question of whether the current measure implicitly includes any amount for health care is relevant to whether the preferred "benchmarking" strategy for introducing a new measure with nothing for health care would amount to an increase in the poverty line.
David Betson's recent paper lays out the case that the current official thresholds include nothing for health care as completely as I have seen it. Betson adduces: a) recent personal communications between Mollie Orshansky, the creator of the current measure, and Gordon Fischer, its foremost historian; b) an April 26, 1967 proposal to index the official thresholds using the CPI for all items except medical; c) a 1987 letter from Wilbur Cohen indicating that the official thresholds include nothing for health care. Without questioning anyone's recollections, I will try to explain why I, and many other people, think otherwise.
First, and most important, Mollie Orshansky's seminal articles on the creation of the poverty thresholds give no indication that health care was excluded. Her several accounts of her calculations and reasons for them at no point state that the poverty budget does not include health needs.4 Such an omission in the record would be serious because food is the only need included explicitly in the original poverty budget. A food consumption survey had found that families of three or more spent around one-third of their after-tax income on food. The other two-thirds was all that was available for all other purposes, including health care. The new thresholds were set at three times an economy food budget. Shelter, clothing, and all other needs were included only implicitly. So others would know that the original poverty thresholds were not meant to include some other necessity, such as health care, only if that were stipulated in the documentation.
If the intent were for the thresholds to cover all necessities except health care, one consequence would have been that, unless the poor paid nothing for health care, food would represent a smaller share of the complete budget of poor families, including health care costs, than in the budget of average families. But early documentation of the poverty threshold concept cites standard economic theory and data showing that poorer families devote a greater than average share of their budgets to food.
If it was assumed that poor families spend nothing on health care because medical needs were met through charity, Hill-Burton uncompensated care, or the new Medicare and Medicaid programs, that assumption was not correct. Medicare and Medicaid were not available to all poor persons. And it is apparent from 1960-61 Consumer Expenditure Survey data that units with the lowest incomes did indeed have health care costs. All urban families and single consumers devoted about 7 percent of their expenditures to health care, while those with annual incomes below $1,000 reported 10 percent for this purpose.5 Units with incomes below $1,000 and heads between 25 and 55 years of age reported that about 8 percent of their expenditures were for health care.6
The presence of medical needs in the official thresholds is confirmed by the 1969 decision to index the thresholds to the CPI for all items, rather than to continue to index with price changes in a low cost market basket of food. Betson's paper notes that, at one point, it was proposed that the thresholds should be indexed with the CPI for all items minus medical care. That would have been appropriate if the thresholds included nothing for medical care. In fact, however, it was the CPI for all items including medical care that was adopted as the new index. It is apparent from Israel Putnam's history of the thresholds through the mid-1970s that increased medical costs played a part in that decision.7
The Bureau of the Budget, SSA, Agriculture and the Office of Economic Opportunity (Williams and Sutton) agreed in the spring of 1969 to revise the poverty thresholds as follows:
Whatever the amount implicit in the original thresholds for health needs, the decision to index with the CPI for all items (including retrospectively recalculating poverty rates for earlier years) meant that the official thresholds were increased each year along with a CPI market basket that included health care. If, instead, the CPI for all items minus medical had been used as the index, the thresholds would have risen more slowly over time. By 1999, they would be about 5 percent lower for that reason. (CPI-U all items: 12/60 = 29.8, 12/99 = 168.3. CPI-U all items minus medical: 12/60 = 30.4, 12/99 = 163.8. 1982-84=100)
Finally, expert opinion has consistently inferred from the multiplier approach used to set the original thresholds that something for medical care was included. For example, among the methods for valuing noncash income included in an earlier Census experimental poverty data series was the "poverty budget share." This approach was, "particularly useful for assessing the effect of medical care benefits on estimates of poverty because of the problematic nature of assigning values to these benefits." The poverty budget share approach rested on an assumption about the market basket underlying the official thresholds.8
The derivation of the current poverty level, which is based on well-specified food needs, implies certain levels of need for other commodities as well (i.e., some amount of expenditure for each good is budgeted into the poverty level) even though no well-specified levels of minimum housing or medical care needs exist.
A proposal for separate threshold levels for health care and for other needs presented to a December 1985 conference on noncash benefits by Lawrence Summers and David Ellwood included this observation.
Our recommendation that medical expenses be excluded from income might lead to a call for a revision in the poverty line since some medical care costs are implicitly included in the budget studies ... used to form the multiplier and set the poverty line.
The NRC panel's 1995 report, Measuring Poverty, A New Approach, notes (p.68)
The original thresholds implicitly allowed for some out-of-pocket medical care expenditures in the multiplier, but not for the fact that such costs differ substantially by people's health status and other characteristics.
When they were developed in the early 1960s, the official poverty thresholds implicitly included (through the multiplier) an allowance for some out-of-pocket medical care expenses.