A new report from the National Academies of Science, Engineering, and Medicine (NASEM), An Updated Measure of Poverty: (Re)Drawing the Line, has hit Washington with something of a splash. In conservative policy circles, “consternation” might be a better word than “splash.” The AEI’s Kevin Corinth sees it as a path to adding billions of dollars to federal spending. Congressional testimony by economist Bruce Meyer takes NASEM to task for outright partisan bias.

Some readers might be put off by the partisan tone of commentaries like those just cited. However, in their less polemical writing, these and other conservative critics offer a good deal of sound analysis. In fact, as I will explain below, many of their recommendations for improvements in the measurement of poverty are in harmony with the NASEM report. Examples include the need for better treatment of healthcare costs, the inclusion of in-kind benefits in resource measures, and greater use of administrative data rather than surveys.

After some reading, I have come to think that the disconnect between the critics’ political rejection of the NASEM report and their accurate analysis of flaws in current poverty measures has less to do with the conceptual basis of the new proposals and more with the way they should be put to work. That comes more clearly into focus if we distinguish between what we might call the tracking and the treatment functions, or macro and micro functions, of poverty measurement. 

The tracking function has an analytic focus. It is a matter of assessing how many people are poor at a given time and tracing how their number varies in response to changes in policies and economic conditions. The treatment function, in contrast, has an administrative focus. It sets a poverty threshold that can be used to determine who is eligible for specific government programs and what their benefits will be.

There are parallels in the tracking and treatment methods that were developed during the Covid-19 pandemic. By early in 2020, it was clear to public health officials that something big was happening, but slow and expensive testing made it hard to track how and where the SARS-CoV-2 virus was spreading. Later, as tests became faster and more accurate, tracking improved. Wastewater testing made it possible to track the spread of the virus to whole communities even before cases began to show up in hospitals. As time went by, improved testing methods also led to better treatment decisions at the micro level. For example, faster and more accurate home antigen tests enabled effective use of treatments like Paxlovid, which works best if taken soon after symptoms develop.

Poverty measurement, like testing for viruses, also plays essential roles in both tracking and treatment. For maximum effectiveness, what we need is a poverty measure that can be used at both the macro and micro level. The measures now in use are highly flawed in both applications. Both the NASEM report itself and the works of its critics offer useful ideas about where we need to go. The following sections will deal first with the tracking function, then with the treatment function, and then with what needs to be done to devise a poverty measure suitable for both uses.

The tracking function of poverty measurement

As a tracking tool, the purpose of any poverty measure is to improve understanding. Each proposed measure represents the answer to a specific question. To understand poverty fully – what it is, how it has changed, who is poor, and why – we need to ask lots of questions. At this macro level, it is misguided to look for the one best measure of poverty. 

First some basics. All of the poverty measures discussed here consist of two elements, a threshold, or measure of needs, and a measure of resources available to meet those needs. The threshold is based on a basic bundle of goods and services considered essential for a minimum acceptable standard of living. The first step in deriving a threshold is to define the basic bundle and determine its cost. The basic bundle can be defined in absolute terms or relative to median standards of living. If absolute, the cost of the basic bundle can be adjusted from year to year to reflect inflation, and if relative, to reflect changes in the median. Once the cost of the basic bundle is established, poverty thresholds themselves may be adjusted to reflect the cost of essentials not explicitly listed in the basic bundle. Further adjustments in the thresholds may be developed to reflect household size and regional differences. 

Similarly, a poverty measure can include a narrower or broader definition of resources. A narrow definition might consider only a household’s regular inflows of cash. A broader definition might include the cash-equivalent value of in-kind benefits, benefits provided through the tax system, withdrawals from retirement savings, and other sources of purchasing power.

Finally, a poverty measure needs to specify the economic unit to which it applies. In some cases, that may be an individual. In other cases, it may be a family (a group related by kinship, marriage, adoption, or other legal ties) or a household (a group of people who live together and share resources regardless of kinship or legal relationships). Putting it all together, an individual, family, or household is counted as poor if their available resources are less than the applicable poverty threshold.

The current official poverty measure (OPM), which dates from the early 1960s, includes very simple versions of each of these components. It defines the basic bundle in absolute terms as three times the cost of a “thrifty food plan” determined by the U.S. Department of Agriculture. It then converts that to a set of thresholds for family units that vary by family size, with special adjustments for higher cost of living in Alaska and Hawaii. The OPM defines resources as before-tax regular cash income, including wages, salaries, interest, and retirement income; cash benefits such as Temporary Assistance for Needy Families and Supplemental Security Income; and a few other items. Importantly, the OPM does not include tax credits or the cash-equivalent value of in-kind benefits in its resource measure.

The parameters of the OPM were initially calibrated to produce a poverty rate for 1963 of approximately 20 percent. After that, annual inflation adjustments used the CPI-U, which to this day remains the most widely publicized price index. As nominal incomes rose due to economic growth and inflation, and as cash benefits increased, the share of the population living below the threshold fell. By the early 1970s, it had reached 12 percent. Since then, however, despite some cyclical ups and downs, it has changed little.

Today, nearly everyone views the OPM as functionally obsolete. Some see it as overstating the poverty rate, in that its measure of resources ignores in-kind benefits like the Supplemental Nutrition Assistance Program (SNAP) and tax credits like the Earned Income Tax Credit (EITC). Others see it as understating poverty on the grounds that three times the cost of food is no longer enough to meet minimal needs for shelter, healthcare, childcare, transportation, and modern communication services. Almost no one sees the OPM as just right.

In a recent paper in the Journal of Political Economy, Richard V. Burkhauser, Kevin Corinth, James Elwell, and Jeff Larrimore provide an excellent overview of the overstatement perspective. The question they ask is what percentage of American households today lack the resources they need to reach the original three-times-food threshold. Simply put, was Lyndon Johnson’s “War on Poverty,” on its own terms, a success or failure?

To answer that question, they develop what they call an absolute full-income poverty measure (FPM). Their first step in creating the FPM was to include more expense categories in the basic bundle, but calibrate it to make the FPM poverty rate for 1963 equal to the OPM rate for that year. Next, they expanded the resource measure to reflect the growth of in-kind transfer programs and the effects of taxes and tax credits. They also adopt the household, rather than the family, as their basic unit of analysis. Burkhauser et al. estimate that adding the full value of the EITC and other tax credits to resources, along with all in-kind transfer programs, cuts the poverty rate in 2019 to just 4 percent, far below the official 10.6 percent. 

Going further, Burkhauser et al. raise the issue of the appropriate measure of the price level to be used in adjusting poverty thresholds over time. They note that many economists consider that the CPI-U overstates the rate of inflation, at least for the economy as a whole. Instead, they prefer the personal consumption expenditure (PCE) index, which the Fed uses as a guide to monetary policy. Replacing the CPI-U with the PCE reduces the FPM poverty rate for 2019 to just 1.6 percent. It is worth noting, however, that some observers maintain that prices in the basket of goods purchased by poor households tend to rise at a faster rate than the average for all households. In that case, an FPM adjusted for inflation using the PCE would not fully satisfy one of the criteria set by Burkhauser et al., namely, that “the poverty measure should reflect the share of people who lack a minimum level of absolute resources.” (For further discussion of this point, see, e.g., this report to the Office of Management and Budget.)

Burkhauser et al. do not represent 1.6 percent as the “true” poverty rate. As they put it, although the FPM does point to “the near elimination of poverty based on standards from more than half a century ago,” they see that as “an important but insufficient indication of progress.” For a fuller understanding, measuring success or failure of Johnson’s War on Poverty is not enough. A poverty measure for today should give a better picture of the basic needs of today’s households and the resources available to them. 

The Supplemental Poverty Measure (SPM), which the Census Bureau has published since 2011, is the best known attempt to modernize the official measure. The SPM enlarges the OPM’s basic bundle of essential goods to include not only food, but also shelter, utilities, telephone, and internet. It takes a relative approach, setting the threshold as a percentage of median expenditures and updating it not just for inflation, but also for the growth in real median household income. On the resource side, the SPM adds many cash and in-kind benefits, although not as many as the FPM. It further adjusts measured resources by deducting some necessary expenses, such as childcare and out-of-pocket healthcare costs. Finally, the SPM moves away from the family as its unit of analysis toward a household concept.

Since the SPM adds items to both thresholds and resources, it could, depending on its exact calibration, have a value either higher or lower than the OPM. In practice, as shown in Figure 1, it has tracked a little higher than the OPM. (For comparison, Burkhauser et al. calculate a variant of their FPM that uses a relative rather than an absolute poverty threshold. The relative FPM, not shown in Figure 1, tracks slightly higher than the SPM in most years.)

That brings us back to our starting point, the NASEM report. Its centerpiece is a recommended revision of the SPM that it calls the principal poverty measure (PPM). The PPM directly addresses several acknowledged shortcomings of the SPM. Some of the most important recommendations include:

  • A further movement toward the household as the unit of analysis. To accomplish that, the PPM would place more emphasis on groups of people who live together and less on biological and legal relationships.
  • A change in the treatment of healthcare costs. The SPM treats out-of-pocket healthcare spending as a deduction from resources but does not treat healthcare as a basic need. The PPM adds the cost of basic health insurance to its basic package of needs, adds the value of subsidized insurance (e.g., Medicaid or employer-provided) to its list of resources, and (like the SPM) deducts any remaining out-of-pocket healthcare spending from total resources.
  • A change in the treatment of the costs of shelter that does a better job of distinguishing between the situation faced by renters and homeowners.
  • Inclusion of childcare as a basic need and subsidies to childcare as a resource.

Although Burkhauser et al. do not directly address the PPM, judging by their criticisms of the SPM, it seems likely that they would regard nearly all of these changes as improvements. Many of the changes recommended in the NASEM report make the PPM more similar to the FPM than is the SPM.

Since the NASEM report makes recommendations regarding methodology for the PPM but does not calculate values, the PPM is not shown in Figure 1. In principle, because it modifies the way that both needs and resources are handled, the PPM could, depending on its exact calibration, produce poverty rates either above or below the SPM.

Measurement for treatment

The OPM, FPM, SPM, and PPM are just a few of the many poverty measures that economists have proposed over the years. When we confine our attention to tracking poverty, each of them adds to our understanding. When we turn to treatment, however, things become more difficult. A big part of the reason is that none of the above measures is really suitable for making micro-level decisions regarding the eligibility of particular households for specific programs. 

For the OPM, that principle is laid out explicitly in U.S. Office of Management and Budget Statistical Policy Directive No. 14, first issued in 1969 and revised in 1978:

The poverty levels used by the Bureau of the Census were developed as rough statistical measures to record changes in the number of persons and families in poverty and their characteristics, over time. While they have relevance to a concept of poverty, these levels were not developed for administrative use in any specific program and nothing in this Directive should be construed as requiring that they should be applied for such a purpose.

Despite the directive, federal and state agencies do use the OPM, or measures derived from it, to determine eligibility for many programs, including Medicaid, SNAP, the Women, Infants, and Children nutrition program, Affordable Care Act premium subsidies, Medicare Part D low-income subsidies, Head Start, and the National School Lunch Program. The exact eligibility thresholds and the rules for calculating resources vary from program to program. The threshold at which eligibility ends or phase-out begins for a given household may be equal to the Census Bureau’s official poverty threshold, a fraction of that threshold, or a multiple of it. The resource measure may be regular cash income as defined by the OPM, or a modification based on various deductions and add-backs. Some programs include asset tests as well as income tests in their measures of resources. The exact rules for computing thresholds and resources vary not only from one program to another, but from state to state.

That brings us to what is, perhaps, the greatest source of alarm among conservative critics of the NASEM proposals. That is the concern that a new poverty measure such as the PPM would be used in a way that raised qualification thresholds while resource measures remained unchanged. It is easy to understand how doing so by administrative action could be seen as a backdoor way of greatly increasing welfare spending without proper legislative scrutiny. 

Kevin Corinth articulates this fear in a recent working paper for the American Enterprise Institute. He notes that while the resource measures that agencies should use in screening applications are usually enshrined in statute, the poverty thresholds are not. If the Office of Management and Budget were to put its official stamp of approval on a new poverty measure such as the SPM or PPM, Corinth maintains that agencies would fall in line and recalculate their thresholds based on the new measure without making any change in the way they measure resources. 

Corinth calculates that if the current OPM-based thresholds were replaced by new thresholds derived from the SPM, federal spending for SNAP and Medicaid alone would increase by $124 billion by 2033. No similar calculation can be made for the PPM, since it has not been finalized, but Corinth presumes that it, too, would greatly increase spending if it were used to recalculate thresholds while standards for resources remained unchanged.

Clearly, Corinth is onto a real problem. The whole conceptual basis of the SPM and PPM is to add relevant items in a balanced manner to both sides of the poverty equation, so that they more accurately reflect the balance between the cost of basic needs and the resources available to meet them. Changing one side of the equation while leaving the other untouched makes little sense.

Logically, then, what we need is an approach to poverty measurement that is both conceptually balanced and operationally suitable for use at both the macro and micro levels. Corinth himself acknowledges that at one point, when he notes that “changes could be made to both the SPM resource measure and SPM thresholds.” However, he does not follow up with specific recommendations for doing so. To fill the gap, I offer some ideas of my own in the next section.

Toward a balanced approach to micro-level poverty measurement

To transform the PPM from a tracking tool into one suitable for determining individual households’ eligibility for specific public assistance programs would require several steps.

1. Finalize a package of basic needs. For the PPM, those would include food, clothing, telephone, internet, housing needs based on fair market rents, basic health insurance, and childcare. The NASEM report recommends calibrating costs based on a percentage of median expenditures, but conceptually, they could instead be set in absolute terms, either based on costs in a given year or averaged over a fixed period leading up to the date of implementation of the new approach.

2. Convert the package of needs into thresholds. Thresholds would vary according to the size and composition of the household unit. They could also vary geographically, although there is room for debate on how fine the calibration should be. Thresholds would be updated to reflect changes in the cost of living as measured by changes in median expenditures or (in the absolute version) changes in price levels.

3. Finalize a list of resources. These would include cash income plus noncash benefits that households can use to meet food, clothing, and telecommunication needs; plus childcare subsidies, health insurance benefits and subsidies, rent subsidies, and (for homeowners) imputed rental income; minus tax payments net of refundable tax credits; minus work expenses, nonpremium out-of-pocket medical expenses, homeowner costs if applicable, and child support paid to another household. 

4. Centralize collection of data on resources. Determining eligibility of individual households for specific programs would require assembling data from many sources. It would be highly beneficial to centralize the reporting of total resources as much as possible, so that all resource data would be available from a central PPM database. The IRS could provide data on cash income other than public benefits (wages, salaries, interest, dividends, etc.) and payments made to households through refundable tax credits such as the EITC. The federal or state agencies that administer SNAP, Medicaid and other programs could provide household-by-household data directly to the PPM database. Employers could report the cash-equivalent value of employer-provided health benefits along with earnings and taxable benefits. 

5. Devise a uniform format for reporting eligible deductions from resources. Individual households applying for benefits from specific programs would be responsible for reporting applicable deductions from resources, such as work expenses, out-of-pocket medical expenses, homeowner costs and so on. A uniform format should be developed for reporting those deductions along with uniform standards for documenting them so that the information could be submitted to multiple benefit programs without undue administrative burden.

6. Use net total resources as the basis for all program eligibility. Decisions on eligibility for individual programs should use net total resources, as determined by steps (4) and (5), to determine eligibility for individual programs. 

7. Harmonization of program phase-outs. It would be possible to implement steps (1) through (6) without making major changes to the phase-in and phaseout rules for individual programs. However, once a household-by-household measure of net total resources became available, it would be highly desirable to use it to compute benefit phaseouts from all programs in a harmonized fashion. At present, for example, a household just above the poverty line might face a phaseout rate of 24 percent for SNAP and 21 percent for EITC, giving it an effective marginal tax rate of 45 percent, not counting other programs or income and payroll taxes. As explained in this previous commentary, such high effective marginal tax rates impose severe work disincentives, especially on families that are just making the transition from poverty to self-sufficiency. Replacing the phase-outs for individual programs with a single harmonized “tax” on total resources as computed by the PPM formula could significantly mitigate work disincentives. 

Implementing all of these steps would clearly be a major administrative and legislative undertaking. However, the result would be a public assistance system that was ultimately simpler, less prone to error, and less administratively burdensome both for government agencies at all levels and for poor households.

Conclusion

In a recent commentary for the Foundation for Research on Equal Opportunity, Michael Tanner points to the potentially far-reaching significance of proposed revisions to the way poverty is measured. “Congress should use this opportunity to debate even bigger questions,” he writes, such as “what is poverty, and how should policymakers measure it?” Ultimately, he continues, “attempts to develop a statistical measure of poverty venture beyond science, reflecting value judgments and biases. Such measures cannot explain everything about how the poor really live, how easily they can improve their situation, or how policymakers can best help them.”

The “beyond science” caveat is worth keeping in mind for all discussions of poverty measurement. A case in point is the issue of whether to use an absolute or relative approach in defining needs. It is not that one is right and the other wrong. Rather, they reflect fundamentally different philosophies as to what poverty is. For example, as noted earlier, Burkhauser et al. compute both absolute and relative versions of their full-income poverty measure. The absolute version is the right one for answering the historical question they pose about the success or failure of the original War on Poverty, but for purposes of policy today, the choice is not so clear. Some might see an absolute measure, when used in a micro context, as producing too many false negatives, that is, failures to help those truly in need. Others might see the relative approach as producing too many false positives, spending hard-earned taxpayer funds on people who could get by on their own if they made the effort. The choice is more a matter of values than of science.

The choice of a price index for adjusting poverty measures over time also involves values as well as science. Should the index be one based on average consumption patterns for all households, such as the PCE or chained CPI-U, or should it be a special index based on the consumption patterns of low-income households? Should the index be descriptive, that is, based on observed consumption patterns for the group in question? Or should it be prescriptive, that is, based on a subjective estimate of “basics needed to live and work in the modern economy” as is the approach taken by the ALICE Essentials Index?

In closing, I would like to call attention to four additional reasons why conservative critics of existing poverty policy should welcome the proposed PPM, even in its unfinished state, as a major step in the right direction. 

The PPM is inherently less prone to error. Bruce Meyer is concerned that “the NAS[EM]-proposed changes to poverty measurement would produce a measure of poverty that does a worse job identifying the most disadvantaged, calling poor those who are better off and not including others suffering more deprivation.” In fact, many features of the PPM make it inherently less prone than either the OPM or the SPM to both false positives and false negatives. The most obvious is its move toward a full-income definition of resources. That avoids one of the most glaring flaws in the OPM, namely, the identification of families as poor who in fact receive sufficient resources in the form of in-kind transfers or tax credits. The PPM also addresses some of the flaws of the SPM that Meyer singles out in his testimony, most notably in the treatment of healthcare.  Furthermore, by placing greater emphasis on administrative data sources and less on surveys, the PPM would mitigate underreporting of income and benefits, which Burkhauser et al. identify as a key weakness of the SPM. (See NASEM Recommendations 6.2 and 6.3).

The PPM offers a pathway toward consistency and standardization in poverty policy. In his critique of the PPM, Tanner suggests that “Congress should decouple program eligibility from any single poverty measure, and adopt a broader definition of poverty that examines the totality of the circumstances that low-income people face and their potential to rise above them.” I see that kind of decoupling as exactly the wrong approach. Our existing welfare system is already a  clumsy accretion of mutually inconsistent means-tested programs – as many as 80 of them, by one count. It is massively wasteful and daunting to navigate. In large part that is precisely because each component represents a different view of the “totality of circumstances” of the poor as seen by different policymakers at different times. What we need is not decoupling, but standardization and consistency. The proposed system-wide redefinition of poverty offers a perfect opportunity to make real progress in that direction.

The PPM would be more family-friendly. One pro-family feature of the PPM is its recognition of childcare costs on both the needs and the resources sides of the poverty equation. In addition, by moving toward households (defined by resource-sharing) rather than families (defined by legal relationships), the PPM would mitigate the marriage penalties that are built into some of today’s OPM-based poverty programs.

Properly implemented, the PPM would be more work-friendly. As noted above, the benefit cliffs, disincentive deserts, and high effective marginal tax rates of existing OPM-based poverty programs create formidable work disincentives. Moving toward a harmonized phaseout system based on the PPM’s full-income approach to resources could greatly reduce work disincentives, especially for households just above the poverty line that are struggling to take the last steps toward self-sufficiency.

In short, it is wrong to view the proposed PPM as part of a progressive plot to raid the government budget for the benefit of the undeserving, as some conservative critics seem to have done. Rather, both conservatives and progressives should embrace the PPM as a promising step forward and direct their efforts toward making sure it is properly implemented.