Poverty: Improving the Measure After Thirty Years
A Conference

Thomas Corbett

Meeting at the University of Wisconsin-Madison in April, 1999, a number of academics and public officials from across the country overwhelmingly agreed that the existing official U.S. poverty measure needs to be improved. The present measure does not accurately identify who is poor and who is not; it cannot, therefore, be expected to measure the consequences of recent reforms with any confidence. Conference attendees also agreed that identifying flaws in the existing measure was the easy part; developing a consensus on a new measure remains a political and technical challenge.

Those coming together in Madison were drawn from varied perspectives and situations. There were officials from several federal agencies who have been deeply immersed in the highly technical issues raised in constructing a new measure. There were representatives from research and advocacy organizations who were interested in how a new measure might affect their work or the constituencies they represent. There were public policy students from the LaFollette Institute who had only recently been introduced to the issues.⁽¹⁾ There were state officials and political experts who focused on what a new measure might imply for the existing equilibrium in the debate about policy or the allocation of resources. And there were the academics who research poverty-related questions.

Attendees did review many of the complex conceptual and technical issues involved in constructing a new measure.⁽²⁾ But the gathering primarily directed its attention toward determining where areas of agreement existed and exploring possibilities for advancing improvements to the measure. It is to these aspects of the conference that we pay particular attention here.

The official measure

The present poverty measure was made official by an Office of Management and Budget (OMB) directive in 1969. This directive codified the work that Mollie Orshansky, an analyst with the Social Security Administration, had done earlier in the decade and established a single, authoritative measure by which economic impoverishment would be calculated. Orshansky's poverty thresholds were originally determined as three times the cost of a minimally adequate diet, which in turn was based on the U.S. Department of Agriculture's economy food plan. (The multiplier of three was based on 1950s estimates of the proportion of resources expended on food by low-income households.) Gross cash income is compared with the appropriate threshold, adjusted for family size, to determine poverty status.

This measure worked well in the 1960s. The Orshansky thresholds were remarkably consistent with a poverty measure derived from an alternative approach using 50 percent of median income and also with public perceptions of a "minimally adequate" level of resources for a typical family. But over time, the policy world has changed dramatically. In the 1960s, positive taxes were not an important determinant of well-being for low-income families. The payroll tax was equal to only 3.625 percent on $4,800 of earnings, and the income tax threshold was equal to 93 percent of the poverty line for a family of four. Negative tax policies affecting low-income families, such as the Earned Income Tax Credit (EITC), were not introduced until the mid-1970s. In-kind programs such as Food Stamps and Medicaid were small, experimental initiatives or nonexistent when Orshansky was developing her concept of resource inadequacy.

In recent years, in-kind transfers and negative tax transfers have far outstripped cash welfare transfers as important sources of support for low-income families. Expenditures for food, which accounted for about one-third of family income in the 1950s, now account for as little as one-seventh. Not surprisingly, as the policy environment has changed, the official measure has come under criticism for a variety of shortcomings. It excludes in-kind benefits, including food stamps and housing assistance, when counting family resources. It ignores direct tax payments, such as payroll and income taxes, when measuring family expenditures. By the same token, it also ignores the contribution to family resources provided by refundable tax credits, the EITC being the outstanding example. It disregards regional variation in the cost of living, in particular the cost of housing, in determining a family's consumption needs. It ignores differences in health insurance coverage in determining family resources, and differences in medical spending in determining family consumption needs. It ignores the cost of earning wage income, including child care costs, when calculating the net income available to families containing working members. Finally, the official thresholds have never been updated to reflect changing levels or patterns of consumption by U.S. households.

Steps toward an improved measure

In early 1995, the National Research Council of the National Academy of Sciences published recommendations from the Panel on Poverty and Family Assistance suggesting ways in which the present measure can be revised.⁽³⁾ The report received critical acclaim as a well-considered analysis of the measure's flaws and as a set of reasonable remedies, but public-sector response was muted at best.⁽⁴⁾ As Mark Greenberg of the Center for Law and Social Policy pointed out at the April 1999 conference, 1995 was probably the worst possible time politically to suggest major changes in the measurement of an ideologically charged indicator of societal well-being such as poverty. The transfer of political power in the House of Representatives in the Fall of 1994 distracted political attention and made any changes in the measure difficult at best.

Nevertheless, with support from the Annie E. Casey Foundation, the Institute for Research on Poverty and the Brookings Institution, with assistance from Wendell Primus at the Center on Budget and Policy Priorities, have been working to move the poverty measure agenda along. Since April of 1997, this project has held a number of workshops and conferences designed to stimulate work toward a new measure and facilitate cooperation toward that end.

In 1997, the Office of Management and Budget convened a federal Interagency Technical Working Group to "improve the measurement of income and poverty." The group has met at least five times and has created subgroups to work on specific technical issues.

A good deal of noteworthy research has been conducted over the past several years by analysts in the Bureau of Labor Statistics and the Bureau of the Census, which has responsibility for official thresholds and annual poverty statistics. Patricia Doyle, Thesia Garner, David Johnson, Martina Shea, Stephanie Shipp, and Kathleen Short have been particularly active. In addition, poverty panel member David Betson of the University of Notre Dame has written on valuing home ownership, equivalence scales, and estimates of out-of-pocket medical expenditures.

In June of 1999, the Bureau of the Census plans to issue a report on experimental measures of poverty that incorporate the NRC panel recommendations, as modified by subsequent research.

Why changing the measure is problematic

If changing the poverty measure were easily done, it would have been done by now. First, we must agree on what it really means to be poor. This is not as straightforward as it may appear. At the conference, Robert Haveman and Melissa Mullikin pointed out basic differences in the way poverty can be conceptualized.⁽⁵⁾ The standard approach measures ones economic position by comparing "command over resources" against some measure of a minimal level of well-being. But one can plausibly argue for alternative concepts, such as a family's annual consumption or its capacity to be self-reliant. In short, even our basic notion of poverty is not without controversy.

Second, we expect a poverty measure to accomplish several tasks. A number of panelists at the conference commented on the demands and expectations we place upon this single measure of well-being. Depending upon perspective and situation, some see the measure as merely a way to separate the impoverished from those not in dire economic need. From this perspective, the measure must adequately separate those without even minimal resources from those who have enough and enable us to accurately identify the needy. Others see the poverty measure as a way to determine who is in need and should therefore receive social assistance of various kinds. Still others would use poverty rates as a way to distribute social resources across groups and jurisdictions, as an important dimension of any resource allocation formula. Still others focus on its role as an indicator of societal well-being. In this role, it can be used to assess performance (e.g., the consequences of welfare reform), or as an advocacy tool. Finally, some see the poverty measure as an important tool for assessing how we are doing as a society over time, and how various subgroups are faring relative to one another. In this sense, the poverty measure becomes an instrument for creating a critical social indicator. As Patricia Ruggles commented at the conference, we may simply be asking too much of a single measure.⁽⁶⁾

Third, we must confront the political reality that there will be winners and losers. Papers by Gary Burtless, by David Betson and Jennifer Warlick, and observations by other conference participants laid out the various choices and tradeoffs that a new measure entails.⁽⁷⁾ Altering the relative odds of being identified as poor has profound real and symbolic consequences. Primary among these are:

• Compositional changes. Changing the measure alters the relative vulnerability to poverty of various groups. Claims on resources or public attention, in part, may depend on the perceived vulnerability of the group, whether children, or the elderly, or the working poor, or single mothers. How will advocacy groups respond if the relative vulnerability of the people they represent changes under a new poverty measure?
• Geographical distribution. Changing the measure can alter the geographical distribution of the poor. Under some scenarios, poverty increases on both coasts, whereas the midsection of the country does relatively well. This can be seen as good or bad news, depending on whether the relative change is viewed as a reflection of extant policies or as a predictor of future federal resources.
• Historical understandings. A new measure may well reshape our understanding of the record of U.S. poverty. The conventional wisdom holds that the economic position of children has been worsening and that of the elderly has been improving. What if a new measure modifies those assumptions? What would that do to our understanding of which public policies have worked and which have not?
• Resource allocation consequences. There is nothing that necessarily ties a new statistical measure of poverty to the allocation of public resources. Still, many would reasonably worry that a change in the measure would eventually have real consequences for who is helped and who isn't.
• Recasting the political debate. The political consequences of a new poverty measure might be large; they are certainly unpredictable. For example, suppose the poverty rate changes as a result of a new measure. A rise might be viewed as a transparent attempt to increase spending on the poor or, conversely, to demonstrate that prior spending has been ineffective. A drop in the rate might be viewed as a threat to continued efforts to deal with poverty--or as evidence that public initiatives are working and should be expanded.

The conference participants were in agreement that any new measure could not be separated from real or imagined political concerns and that it must be carefully thought through.

Assessing current sentiment

As a final action of this conference, attendees participated in a straw vote to determine areas of agreement and areas where confusion or conflict remained (Table 1). The vote should not be taken too seriously. Although the attendees present during the vote were drawn from several important constituencies--the federal government, academia, the political arena, and advocacy groups--they were in no way representative of all the constituencies that would be involved in any process of change. Still the vote may indicate what a well-informed group of decision makers, and those in a position to influence them, think about key issues.

Ten issues were posed to the group.⁽⁸⁾ Five of them were concerned with how available resources would be calculated, three with the setting of poverty thresholds, and two with strategies for moving to a new measure. On some issues, there was clear agreement; on others, considerable disagreement. On still other issues, it was clear that the attendees wished to have more information before making a decision.

Changing the resource definitions

With respect to resources, conference participants were nearly unanimous in their belief that a new poverty measure should incorporate in-kind public benefits that are a cash equivalent, such as food stamps. There was a similar consensus on incorporating positive (e.g., income and payroll) taxes and negative taxes (e.g., the EITC) in the calculation of available resources to be applied against the poverty threshold. In both cases, 95 percent of those voting supported the change. There also was substantial agreement that nondiscretionary expenses associated with work should be deducted from available resources. Some 85 percent of those voting on the issue and 75 percent of all those participating in the voting exercise agreed with this suggested modification to the poverty measure.⁽⁹⁾

There was much less agreement on other suggested changes to the calculation of resources. Arguments have been made that an income stream should be computed from the equity in property (i.e., home ownership). It also has been argued that resources should be adjusted downward for out-of-pocket medical expenditures. These are very complex topics that affect certain groups quite differently. Accounting for medical expenditures affects the well-being of the elderly disproportionately. Moreover, there is not even a conceptual agreement on the best technical approach. For instance, how does one separate discretionary from nondiscretionary medical outlays? Only 27 percent (16 percent of all participating in the voting process) agreed with incorporating actual out-of-pocket medical expenditures in the calculation of resources.

Changing the thresholds

Conference attendees first considered whether to change the way in which equivalence scales, adjustments for households of different sizes, are calculated. They next voted on whether the thresholds ought to differ in response to variations in the cost of living, particularly housing costs. Finally, they voted on whether the thresholds should be updated by changes in society's income (or resource) levels over time, or whether adjustments should reflect only price changes. Support for the former would constitute a relative poverty scale that would be sensitive to alterations in the distribution of income; the latter would continue the current practice where the thresholds represent the same purchasing power as when first developed in the 1960s (an absolute scale).

There was support for updating the equivalence scale. All of those actually voting and two-thirds of those participating in the voting exercise supported this measure. The complex issues of adjusting the thresholds over time and across jurisdictions (states) generated considerable controversy. In both cases, there were sharp divisions among actual voters and many abstained. Geographic variation in the thresholds, for example, would recast the existing distribution of the poor across states and communities. Immediately, some states (and politicians) would look better or worse, and the current equilibrium in the distribution of resources might be upset. Though adjusting thresholds for varied regional living costs makes conceptual sense, it is not an easy decision politically.

Finally, the group explored potential strategies of moving from the old to a new poverty measure. They first examined a proposal suggested in earlier work by Gary Burtless, Tom Corbett, and Wendell Primus. These authors argued that the parameters of a new poverty measure should be calibrated in such a way that the aggregate rate of the new measure and the existing measure would be identical in the initial year of use. The rationale for this artificial constraint is to defuse reservations about any implicit political agenda. There proved to be little support for this expedient strategy. Only one-third of those voting and one-fourth of those participating in the voting process endorsed the idea. On the other hand, there was support for renaming the new measure. Those endorsing this approach presumably saw an improved measure existing alongside the current poverty measure, eventually replacing it as acceptance grew.

A question of procedure

The conference also clarified some strategic divisions over how to change the poverty measure. Should revision be an explicitly political process that directly involves legislation by Congress, because of the controversy that swirls around poverty programs and issues? Or should it be treated in the same way as revisions to equally important and potentially controversial measures, such as inflation or unemployment statistics, which occur as the result of technical review by the responsible statistical agencies? Over the years, many government statistical measures have been regularly updated. Only the poverty measure has remained frozen in time.

No agreement on how to proceed was reached, but several points were made regarding future strategy:

• The policy community must agree on one alternative to the current measure or acceptance in the broader community will be difficult. If multiple alternatives are proposed, confusion will result and the benefits of an official measure will be lost.⁽¹⁰⁾
• The process of establishing a new measure can be done through a new OMB directive, but must also involve much work in key political circles to address concern about what a change might mean.
• As part of this selling process, the political and practical consequences of moving to a new measure must be fully explicated. If a new measure alters the compositional or geographic distribution of poverty, the extent of the change must be measured and dealt with openly.
• Mechanisms for routinely improving the poverty measure must be thought through. There have always been certain functions and statistical measures that we believe to be too important to be left fully in the political arena. We insulate some of these technical tasks from political interference. The measurement of poverty may be one of these.

Fixing other flaws noted by the NAS panel will not be easy, because of conceptual controversy, data problems, and political concerns. Experts and interested parties remain divided over many of the issues first raised by the NAS panel, or in the debate generated by the panel's deliberations: whether and how to reflect medical expenditures, the costs and benefits of home ownership, and differences in the regional cost of living.

Despite the differences among conference attendees, there was overwhelming agreement that the poverty measure must be changed. We regularly reexamine and improve key economic indicators such as national output, the incidence of joblessness, or price levels. A national measure of economic hardship is no less important to understanding how we are doing as a society. Yet our official measure is flawed by virtually any standard of assessment. That recognition is the starting point for renewed efforts to continue the work so well begun by the NRC panel in 1995. This effort will be further reinforced by the publication of several variants of the experimental measure in the forthcoming Census Bureau report.

1. A class of students from the LaFollette Institute for Public Affairs, under the direction of Tom Kaplan, prepared background materials for the conference and assisted in many important ways. These background papers will appear in the IRP Special Report on the conference. Articles not otherwise identified below were presented at the conference, which was held at the University of Wisconsin-Madison on April 15-17, 1999.

2. See Focus 19, no. 2 (Spring 1998 for a full discussion of the problems with the official poverty measure.

3. C. Citro and R. Michael, Measuring Poverty: A New Approach (Washington D.C.: National Academy Press, 1995).

4. See G. Fisher, "An Overview of Developments since 1995 Relating to a Possible New U.S. Poverty Measure," unpublished paper (April, 1999). Mr. Fisher is an analyst with the U.S. Department of Health and Human Services, Office of the Assistant Secretary of Planning and Evaluation. The paper represents his own views and not those of the Department.

5. Robert Haveman and Melissa Mullikin, "Alternatives to the Official Poverty Measure: Perspectives and Assessment."

6. Patricia Ruggles is the Deputy Assistant Secretary for Human Services Policy in the Office of the Assistant Secretary for Planning and Evaluation in the federal Department of Health and Human Services.

7. Gary Burtless, "Political Consequences of an Improved Poverty Measure," and David Betson and Jennifer Warlick, "Reshaping the Historical Record with a Comprehensive Measure of Poverty."

8. An eleventh issue was posed to conference attendees in a slightly different format and is not included in the table. They were asked whether the poverty measure (old or revised) should continue to be based on the Current Population Survey (CPS), or whether the Survey of Income and Program Participation (SIPP) should be used instead. The vote was: SIPP as currently constituted, 0; a revised and improved SIPP, 32; CPS, as now, 0; abstain or not voting, 12.

9. On each issue the number voting in favor or against differed. Some actually voted to abstain whereas others did not vote. These may represent different views with abstentions clearly representing a position that the voter desires more information and the latter something closer to confusion. Not all attendees were equally conversant with the issues. For convenience, we aggregated the abstentions and nonvoters.

10. This point was made by Ron Haskins, Staff Director of the Human Resources Subcommittee, U.S. House Ways and Means Committee.