Modest but Adequate: The National Academy of Sciences Proposal to Change the Poverty Measure

Beth Kohler and Kristen Voskuil

The National Academy of Sciences Panel's project began as a request from the Joint Economic Committee of Congress and the House Subcommittee on Census, Statistics and Postal Personnel to study theories, methods of measurement, and data needs of a poverty measure. The U.S. Department of Health and Human Services later expanded the panel's task to include reviewing similar issues implicit in the creation of standards for welfare payments to needy families with children. The proposal that resulted from these requests is a significant departure from the current method used by the Census Bureau. The NAS proposal makes a dramatic move toward a relative measure of poverty by linking the poverty threshold to average expenditures rather than an expert-determined income level. The new proposal also changes the definition of income from cash income before taxes to "money and near-money disposable income" (Citro and Michael, 1995, 9). The NAS panel also proposes that the Census Bureau change the survey used to measure the prevalence of poverty in the United States from the March supplement to the Current Population Survey's (CPS) March supplement to the Survey of Income and Program Participation (SIPP).

The Threshold

Under the NAS measure, the base threshold is applied to a reference family of two adults and two children. Median consumer expenditures for food, clothing and shelter for families of that type are determined from the U.S. Consumer Expenditure Survey and averaged for the last three years. The panel recommends that a percentage of a median be selected so that in the first year of the measure the same number of people fall below the threshold as do under the current system. A small multiplier of 1.15 to 1.25 would be added to the threshold amount to allow for the purchase of additional needs such as household and school supplies or personal care products. The same thresholds would apply to the elderly and the rest of the population, and two cohabiting adults would be treated as one family rather than two, if they live in the same space.

Adjustments by Family Size and Geographic Area

The reference family threshold would be adjusted for other family types by considering the consumption needs of a child under 18 to be 70 percent of an adult's needs. The increased ability of larger families to take advantage of bulk buying or shared housing and clothing (economies of scale) would be accounted for in the measure by calculating the number of adults plus 0.7 times the number of children and raising this to a power of 0.65 to 0.75. The panel proposed the scale economy factor of 0.65 to 0.75 to take into account the idea that a family of two adults does not consume two times as much as a single adult (i.e., some goods can be shared between the two). For example, if the poverty threshold for a family of two adults and two children is $10,000, the threshold for a family of one adult and two children would be $8,000 at a power of 0.65 and $7,700 at a power of 0.75. (1)

The basic threshold would also be adjusted for geographic differences in costs of housing using decennial census data and a methodology of the U.S. Department of Housing and Urban Development index to estimate average rents. Housing costs would vary by the nine census regions and by metropolitan areas grouped by size within each region. In effect, this allowance for geographic variation would create 54 separate thresholds, which introduces both greater complexity and greater sensitivity.

Updating the Threshold

The threshold would be updated annually on the basis of the average changes of the past three years of median expenditures on food, clothing and shelter. The panel proposed using the average of expenditure data of the past three years so the poverty level was not as tied to the ups and downs of the economy.

The design of the threshold was by no means a product of consensus by the NAS panel. Panel member John F. Cogan wrote a formal dissent for the end of the panel's published report, "Measuring Poverty: a New Approach." Cogan argues that updating the threshold based on consumption patterns produces a relative, not absolute, standard that would raise the amount allowed by the threshold at a rate that would be unacceptable to policymakers and the public. Furthermore, he argues that the panel's decision on which goods are "basic needs" was too subjective and not based on statistical analyses of consumption data.

The remainder of the panel argued elsewhere in the report that defining and measuring poverty is inherently value laden. The other panel members, believed that linking poverty measurement to expenditure data would account for the fact that the standard of living has increased in the United States and that what we view as minimally adequate should increase to reflect that. They also stated that while basing the threshold on food, clothing, and shelter was a value judgment, they felt that most people would agree that all people require them. They did concede that updating the threshold on changes in expenditures rather than price changes would likely raise the threshold at a faster rate. It is unknown at this point the extent to which the threshold would rise annually. Preliminary estimates actually resulted in lower increases in threshold amounts under the NAS method than by the current method, but figures are still being produced.

Family Resources

The current measure does not distinguish among families of the same size and income with different amounts of disposable income. For examples, a family with two young children in child care has the same threshold as a family with two older children with no child care expenses. The panel could have addressed this by developing several different thresholds but instead chose to adjust the calculation of "available resources" to reflect a family's disposable income.

The panel proposes a definition of family resources that includes some thins but not others. Included resources are:

  • gross money income from all sources (same as the current measure)
  • the value of near money, nonmedical in-kind benefits such as Food Stamps subsidized housing, school lunches, and home energy assistance.

Subtracted from resources are the following:

  • out-of pocket medical expenditures (including health insurance premiums)
  • income taxes and Social Security payroll taxes
  • child care expenditures (up to the earnings of the parent with lower earnings, or with a cap adjusted annually for inflation)
  • work-related expenses (a flat amount per week)
  • child support payments from the payer;

Several issues arise in considering the proposal: the value of noncash income, homeownership, nondiscretionary expenditures, and medical expenditures.

Valuing Non-Cash Income

The proposal to include non-cash income in the measurement of resources raises two main questions: which forms of non-cash income to include and how to assign value to this income.

The panel chose to include "near money, nonmedical in-kind benefits such as food stamps, subsidized housing, school lunches, and home energy assistance," (Citro and Michael 1995, 209), benefits on which the Census Bureau already collects adequate data. Most of these programs have a cash-equivalence value and, since program participants would have to spend cash if the in-kind benefits were not available, they free up cash resources for other consumption.

Still, assigning value to non-cash income is problematic. The Census Bureau uses the market value of the benefits. However, at a Census Bureau Conference on the Measurement of Noncash Benefits, others argued that the utility gained from these transfers is probably less than the market value, so using the market value overstates the worth of the transfers (Citro and Michael 1995, 222). Therefore, they argue, using the recipient value (the amount a recipient believes the benefit is worth) is a better measure of the value of the non-cash income. However, as Ruggles points out, "although the recipient may have spent little or nothing on the good in question if it had not been provided for free, that does not mean that it has little or no value to the recipient if it is provided free" (Ruggles 140). Another problem with using recipient value for non-cash income is the difficulty in accurately estimating that value.

The panel decided to recommend using the market value of noncash income, except in the area of public housing, because obtaining either market or recipient value is very difficult (Citro and Michael 1995, 223). The panel suggested further research in valuing housing subsidies, and the Census Bureau is currently conducting such research using CPS, SIPP, and other data (Institute for Research on Poverty 1998, 32).


One source of income the panel chose not to include is that which stems from home ownership. Many people argue that not including the value a person receives from home ownership in a poverty measure overestimates the incidence of poverty (because homeowners with small or no mortgages may have lower housing expenses than renters and therefore have more resources available for other forms of consumption). However, the panel decided that calculating the value of home ownership (through imputing rent or otherwise) is too difficult, given the current data constraints. It suggests research be done on improving the methods for such calculation and that perhaps income derived from home ownership should be included in the poverty measure in the future.

Subtracting Taxes and Other Nondiscretionary Expenditures

The proposal suggests deducting income and Social Security payroll taxes from resources because these payments are nondiscretionary and therefore not available for consumption. The panel also proposes deducting weekly out-of-pocket child care expenses, a flat work expense per week (an estimated $14 in 1992), medical out-of-pocket expenditures (including health insurance premiums), and child support payments. The panel suggests using SIPP data to estimate the value of each of these expenses, with the exception of medical out-of-pocket expenditures.

Medical Care Expenditures

Perhaps the greatest controversy regarding resources surrounds the treatment of medical care costs. The Census Bureau indicates poverty rates are "extremely sensitive to the way in which medical benefits are calculated" (Institute for Research on Poverty 1998, 25). The Census Bureau experimented with adding medical benefits (both government- and employer-provided) to a family's resources. Weinberg and Nelson point out that estimating the value of medical benefits is problematic because "coverage of high medical expenses for someone who is sick does nothing to improve his or her poverty status" (Weinberg and Nelson 1999, 4). Because medical benefits are less fungible than other non-cash benefits, and thus do not free up resources for consumption in the same way other benefits do, including them in income does not reflect their real value to the recipient. Furthermore, because medical expenses vary greatly throughout the population, the value of medical benefits might vary according to an individual's health status.

The panel argued that, owing to questions about the valuation of medical benefits and their fungibility, the best way to deal with medical benefits is to deduct out-of-pocket medical expenditures from a family's available resources. (2) Additionally, it suggests developing a separate risk index, which would measure possible medical needs. The index should not be included in the actual measure because these medical needs may not actually develop (Citro and Michael 1995, 235).

Implementation of the New Measure

The NAS panel believes that the Census Bureau should use the SIPP to estimate the level of poverty in the United States rather than the March CPS. The panel has argued that the main purpose of the March CPS is to estimate the amount of unemployment in the United States, and it has only limited questions about income, program usage, or household expenditures. The survey's focus on job participation makes it unsatisfactory to measure poverty if the NAS definition of poverty was adopted as the new standard. In contrast, the SIPP is an income survey and contains more detailed information necessary to implement the proposed measure, although since it attempts to follow samples for four years, it does face the problem of sample attrition.


Citro, Constance F. and Robert T. Michael (eds.), 1995. Measuring Poverty: A New Approach. Washington, DC: National Academy Press.

Institute for Research on Poverty, "Revising the Poverty Measure," Focus Volume 19, Number 2, Spring 1998.

Ruggles, Patricia, 1997. Drawing the Line: Alternative Poverty Measures and Their Implications for Public Policy. Washington, DC: Urban Institute Press.

Weinberg, Daniel H. and Charles T. Nelson, 1999 "Changing the Way the United States Measures Income and Poverty: A Progress Report." Working Paper, U.S. Bureau of the Census.


1. (P+AK) to the power of F. Where P is the # of adults, A = # of children, K = .7, and F is .65-.75.

2. Under this methodology, a family with high health care needs that cannot afford to pay for health care would be treated equivalent for poverty measurement purposes to a family that incurs few medical expenditures because of low health care needs.