A legislatively-based poverty threshold

Transcription

1 A legislatively-based poverty threshold Richard Bavier Abstract The common view that poverty thresholds are necessarily subjective and arbitrary is challenged. Decisions of elected government appropriately rationalize our judgment about what people need. Needs standards and their supporting evidence underlying a range of federally-funded noncash assistance programs are combined to develop a legislatively-grounded, post-tax poverty threshold. Amounts for particular needs vary by the equivalence scales present in assistance program parameters. Moreover, instead of relying on respondent recollections of the value of noncash transfers received, or on complex modeling of the income value of noncash transfers, thresholds will be varied for recipients on the grounds that noncash transfers reduce the amounts of cash income households require to meet their basic needs. Program-based thresholds and post-tax income raise measured poverty rates by a little more than one-fourth. Arguments that budgetary impacts of these results may be somewhat muted are presented. Nevertheless, more measured poverty is the logical outcome if we start from the premise that the level of assistance elected government acts to provide to those in need represents where our nation draws the poverty line. Keywords: Poverty thresholds, poverty measurement, income 1

2 Introduction In its report, Measuring Poverty, A New Approach, (Citro & Michael, 1995) 1 a National Research Council (NRC) panel observed that setting the level of new poverty thresholds was, the most judgmental aspect of a poverty measure. (xvii) the panel declined to recommend a new threshold level, a judgment it deemed subjective, arbitrary, and ultimately political. (106) However, it did find it within the domain of a panel of scientists to commend as reasonable a range for new thresholds, and the mid-point of this range has been adopted in legislation to establish a modern poverty measure. This would be a misfortune because the way the panel developed its proposed threshold range will be viewed as unreasonable on its face. Spending by couples with two children on food, clothing, and shelter was arrayed from the lowest to the highest, and then the 30 th and 35 th centiles of this distribution were selected as the basis for the boundaries of the proposed threshold range. In this range, incomes are above $50,000 and nearly three-fourths of couples are buying their own homes. 2 The panel s approach violates its own criterion that a new poverty measure have some rationale that has face validity. (38) Now that the current poverty measure has been largely discredited, if we adopt a modern measure so counter-intuitive and complex that public acceptance will be easy for critics to undermine, subsequently it will be difficult to restore public confidence in any poverty measure. If a new poverty measure is to gain public understanding and support, a stronger rationale for drawing the poverty line must be adopted. But is a strong rationale intellectually defensible? Is the Poverty Threshold Really Subjective and Arbitrary? The poverty measurement literature frequently characterizes poverty thresholds as subjective and arbitrary. Of the threshold that would become the official measure, its creator, Social Security Administration economist Mollie Orshansky (1965), wrote, paradoxically, The standard itself is admittedly arbitrary, but not unreasonable. The 1976 publication, The Measure of Poverty, A Report to Congress as Mandated by the Educational Amendments of 1974, by the interagency Poverty Task Force, noted, Thus, all poverty thresholds are necessarily subjective no matter how objective the empirical evidence used to derive them are. In her comprehensive review of issues in the measure of poverty, Patricia Ruggles (1990) noted the essential arbitrariness of the poverty threshold. 1 Page references to this foundational report will appear in parentheses within the text. 2 Among couples with two children and no other household members and classified as complete income reporters in the Consumer Expenditure Survey, in the 30 th to 35 th centiles of the distribution of food, clothing, and shelter, mean pretax money income plus food stamps was $58,552 in 2006 dollars. The 2005 wave of the Panel Study of Income Dynamics also included a full battery of expenditure questions. In the same range of the distribution of spending on food, clothing, and shelter, incomes of couples with two children were even higher. 2

3 The NRC poverty panel explained, In deciding on a new measure to recommend, we used scientific evidence to the extent possible. However, the determination of a particular type of poverty measure and, even more, the determination of a particular poverty threshold are ultimately subjective decisions. (xvi) In the panel s view, expert budgets that involved separate judgments about how much families of different types needed for food, shelter, and other basics,... are inevitably arbitrary and, in our judgment, it is better to have the arbitrariness expressed in a formal equivalence scale. (101)...we believe it best for deriving the official U.S. poverty thresholds to minimize the number of judgments required... (49) Rather than a myriad of arbitrary judgments about what different family types need for food, clothing, shelter, and other basics, the panel deemed it best not to pretend to derive a threshold from supporting empirical evidence. This is a reasonable view from the perspective of professional economists, such as those on the NRC panel. Professional standards require that they avoid presenting opinion as science. Moreover, for their professional purposes, a strong empirical defense of the poverty line is unnecessary. Any marker of economic status, even one that is arbitrary in the sense that it has no convincing intrinsic justification, can serve the purposes of economists well as long as it performs its function of measuring change accurately near the bottom of the distribution. However, in most contexts, it is reasonable and usually necessary, to make judgments based on supporting evidence that, while appropriately strong, falls short of scientific proof. Such an empirical foundation is essential for the poverty measure because it is not, principally, a tool for professional economists but a public statistic. The principal purpose of the poverty measure is to inform the public and its political leadership of the extent, distribution, and depth of economic deprivation and to motivate and regulate the allocation of revenues to address it. For these purposes, it is essential that the public understand the rationale for the poverty line and find it convincing. It is not necessary to argue that the poverty line has the certainty of the results of a scientific experiment, or even that the adopted poverty line has indisputably greater empirical support than any other in the range of the reasonable. But it will not do to explain that a new poverty line has no particular relationship to what families need for food, shelter, and other basics that it is just an arbitrary point within the range that economists judge to be reasonable. From the proposition, Science cannot draw the poverty line, it simply does not follow that an empirical foundation is either inappropriate or unimportant for the poverty threshold, or that it is somehow intellectually preferable that a poverty line have no intrinsic empirical rationale. All that follows from the proposition is that a poverty threshold must not be presented as purely a product of science. This is something of a straw man, because, from Mollie Orshansky until now, researchers have not proposed thresholds that they claimed to be pure products of science. 3

4 Characterization of the poverty threshold as subjective and arbitrary is loose usage that masks and distorts the issues and obscures the appropriate path to addressing them. Drawing the poverty line is not self-referential. It is subjective neither in the sense of a private experience, like pain, nor like a taste for spinach. Psychologically, we may each have a different tolerance for observing deprivation before we are discomforted, but when someone argues that the poverty line is too high or too low he does not support his opinion by referring to his own internal states. Neither is the poverty line arbitrary, if by that we mean that generally accepted standards of reason and evidence are irrelevant. If that s the way the threshold were understood, we would not find experts, including the NRC panel, offering objective evidence and argument that their threshold proposals are reasonable but the current threshold is not. The difficulty with drawing the poverty line is not that it is subjective or arbitrary in the conventional meanings of these terms. If the threshold were subjective and arbitrary we couldn t argue about it. But, by their nature, appropriate arguments over where to draw the poverty line cannot compel intellectual assent like a mathematical deduction or a testable hypotheses with replicable results. The central, but not sole, reason for this is the social-evaluative or normative dimension of poverty. In words of Adam Smith (1776/1904) that are quoted often on the topic, By necessaries, I understand not only the commodities which are indispensably necessary for the support of life, but whatever the custom of the country renders it indecent for creditable people, even of the lowest order, to be without. Indecent is an evaluative word, as would be any less archaic term to similarly communicate the dehumanizing aspect of deprivation. It s application depends on more than complete scientific understanding. Indecent also has a necessarily social reference. Evaluative terms, such as delicious, may be subjective, but matters of custom are overtly objective. We don t each decide for ourselves what the custom of the country will be. In Adam Smith s time, what the custom of the country deemed a decent standard of living would be communicated and reinforced by a range of social and religious institutions. Generally, such institutions are less influential and certainly speak with more diversity today, when the term custom itself no longer even has currency in general discourse. Virtually the only institutions capable of serving as the foundation of norms accepted by the nation as a whole are governmental. Elected government formulates and legitimizes what the custom of the country regards as indecent for anyone to be without, and in several ways. It adopts explicit statistical (thresholds) and administrative (guidelines) measures of poverty. It also provides assistance to those deemed to be in need. In this latter function, a determination of how much families need is a forced judgment. To make it, both factual and evaluative issues are engaged and resolved. 4

5 A particularly clear illustration of this sort of government standard of need is the Thrifty Food Plan (TFP), used to set benefit levels for the Department of Agriculture s (USDA) Supplemental Nutrition Assistance Program (SNAP, formerly the Food Stamp Program). Committees in both the House of Representatives (the Committee on Agriculture) and the Senate (the Committee on Agriculture, Nutrition, and Forestry) have responsibility for oversight of SNAP. The legislative process typically includes fact-finding, receipt of expert advice, and the testimony of advocates and critics, then deliberation informed by the record and the views and values of constituents, as well as the members own lights. As with other assistance programs, over the years, advocate and lobby groups have been formed to influence food assistance legislation by producing data and analysis to show the likely effects of proposed changes to the program, and by testifying at committee hearings held to gather information and advice. During the legislative process, minds may change. New facts may be presented, or arguments calling attention to new patterns in the facts may persuade, even if they do not prove. 3 And even when minds are not changed, the strongest evidence and arguments will be put on the record because the legislators judgments will be subject to review in the electoral process. A product of the legislative process, the Food Stamp Act (P.L ), provided that the, face value of coupon allotment issued to certified households be in such amount as the Secretary of Agriculture determined to be the cost of a nutritionally adequate diet, adjusted annually to reflect changes in the prices of food published by the Bureau of Labor Statistics. The cost of a nutritionally adequate diet is established in the indexed Thrifty Food Plan, which, in the 1970s, replaced the Economy Food Plan referenced in Mollie Orshansky s work on the original thresholds. Subsequently, every decade or so, the USDA s Center for Nutrition Policy and Promotion tests whether the TFP is consistent with up-to-date dietary recommendations, food composition data, food habits, and food price information, (Carlson, Lino, Juan, Hanson, & Basiotis, 2007) The results of the latest such recalibration were published in 2007, and employ data from the National Health and Nutrition Examination Survey, the Food Price Database, and contemporary federal nutrition and food portion guidelines. No doubt, there are divisions of opinion among nutrition experts about what constitutes a nutritious diet. Others have argued that the weighted national prices employed in the TFP should be replaced by regional prices, and that underlying assumptions about shopping and food preparation are unrealistic. These differences are not subject to resolution by scientific experiment. 3 The role of evidence, expert advice, and discourse in the legislative process may be contrasted with so-called subjective poverty thresholds developed from responses on opinion surveys. Still, as Garner and Short (2006) note, Personal assessment questions about minimum income and spending support a more populist or democratic framework for living standards measurement than one based strictly on expert judgment. 5

6 However, the experts who are parties to these disagreements are unlikely to acquiesce to the propositions that the TFP is thereby merely a subjective judgment about a minimally adequate diet or that the processes involved in establishing, updating, and reassessing the plan are arbitrary. Federal programs providing assistance with food, housing, health care, and child care all involve assessments of need for these basics. Moreover, these assessments are supported and updated by empirical evidence intended to establish that the maximum assistance is reasonably sufficient. Such assessments and their supporting evidence, while open to improvement, are likely to be more convincing to the general public than a poverty threshold premised on the NRC panel s view that standards of what people need are necessarily arbitrary and subjective. In the poverty literature, governmental assessments of need feature prominently, although reflection on the legitimizing function of governmental processes less so. Mollie Orshansky s original thresholds were based on the Department of Agriculture s economy and low-cost food plans which, she noted, were referenced by needs standards that states were required to develop as a condition of participation in Aid to Families With Dependent Children (AFDC). More recent scholarship using federal program needs standards to develop alternative poverty thresholds will be reviewed shortly. However, other researchers have expressed concern, not that Congress and the Executive Branch are arbitrary when setting needs standards and maximum program benefits, but that their judgments are influenced by external and invalid reasons of a political or budgetary nature. The NRC panel explained:...if the developers of poverty thresholds rely on program standards that are set by legislation, the standards may change for many reasons other than an evaluation of need (such as the desire to cut program costs). (116) 4 Skepticism about the generosity of taxpayers and their government towards those in need is not uncommon among poverty researchers, who are likely to cite distant history in the principle of less eligibility underlying the English Poor Laws, and more recently, most states practice of setting AFDC benefits well below the needs standards that federal rules required them to estimate. This is not a frivolous concern, and it will be in our minds as we consider evidence supporting the standards of need in federal programs. For the present, we observe that research on alternative thresholds that employ program-based standards of need always results in thresholds substantially higher than the official one, and that the threshold developed here will fall in the middle of the range the NRC proposed. The federal government s programmatic standards for food, shelter, health care, and other basics needs imply a higher poverty threshold than the federal 4 Also see Ruggles (1990)

7 government s current statistical measure of what people need. That is a strong basis on which to argue for a higher official poverty threshold. Analysis Plan Within the literature devoted to poverty measurement, Part I will review recent use of government program standards to set poverty thresholds. In Part II, the paths by which benefits for food, shelter, health care, and child care are set in several federally-funded means-tested programs will be explored and supporting evidence discussed. Legislatively-based needs standards for a range of items will be quantified and their variations described. A multiplier, to calculate total needs from the needs with program-based standards, will be presented. Part III presents the rationale for a simplified, post-tax money income measure that captures the anti-poverty effects of near-cash food, housing, and health care transfers by threshold variations. In Part IV, the program-based needs standards will be combined to create poverty thresholds, reflecting program-specific variations in need among households with different characteristics and living in different locations. Then, poverty rates and distributions using these thresholds and alternate concepts of income will be compared to those using the official measure. Some implications of new thresholds for assistance programs will be addressed in Part V. A concluding Part VI will make the case that the proposed thresholds would rationalize the statistical measure of poverty in a way the general public would find convincing. I. Program Needs Standards in Poverty Measurement Research To create the measure that would become the official poverty threshold, Social Security Administration economist Mollie Orshansky (1963) employed the U.S. Department of Agriculture s Low-Cost and Economy Food Plans. Food was the only item in her poverty market basket because, Despite the Nation s technological and social advance, or perhaps because of it, there is no generally accepted level of adequacy for essentials of living except food. (Orshansky, 1965) For families of three or more, the Economy Food Plan was multiplied by three to create the poverty threshold because Orshansky observed in 1955 food consumption survey data that, on average, families of three or more devoted one-third of their after-tax incomes to food purchases. The Economy Food Plan came to be used as a standard of need in setting maximum Food Stamp Program (FSP) benefits. In her work, Orshansky (1965) noted that the Low-Cost Food Plan was used at the time by state welfare agencies as the basis for food allotments in welfare programs. Often, however, the actual food allowance for families receiving public assistance was less than that in the low-cost food plan Soon after Children of the Poor appeared, Orshansky (1965) already had noted that more recent expenditure survey data suggested that the food spending 7

8 multiplier creating her original threshold may have been too low. She did not disavow her approach, although a later account noted that the White House had chosen the lower of the two thresholds she had developed. Far from stressing the normative authority that the government s Economy Food Plan lent her threshold standard, Orshansky (1965) noted, The standard itself is admittedly arbitrary, but not unreasonable. For her, the empirical evidence underlying the food plans justified its use. The plans represent a translation of the criteria of nutritional adequacy set forth by the National Research Council into quantities and types of food compatible with preference by United States families, as revealed in food consumption studies. (Orshansky, 1965) During the 1980s, attention focused on the measure of economic resources tested against the poverty threshold. In response to Congress and the Office of Management and Budget, the Census Bureau developed a series of experimental poverty statistics that reflected the effect of direct taxes and noncash benefits on economic resources, but the charge did not extend to changes in the thresholds to which the broader income measure was compared. However, by the early 1990s, several alternative thresholds were developed, each relying to some degree on legislated standards of need. Patricia Ruggles comprehensive discussion of the measure of poverty includes two alternative thresholds of her own design to illustrate that program standards of need would produce higher poverty rates than the official measure. Although she asserts the essential arbitrariness of the poverty threshold, (Ruggles, p.24.) the author also maintains that, on their face, some consumption levels are too low, and some are too high, to be serious candidates for the poverty line. In other words, even though there is no one right bundle of consumption needs for the poor that all experts would agree on, we do know enough to eliminate a very large number of clearly wrong answers. In this sense, an expert-determined market basket need not be seen as essentially arbitrary, even conceding that an exact scientific determination of needs is not really possible. (Ruggles, p.49.) One of Ruggles alternatives, termed the updated multiplier, suggests that, if it was reasonable to set the poverty thresholds at three times a minimally adequate diet in the 1960s because average families spent one third of their after-tax income on food, then in 1990, when average families spend a little less than onefifth of their budgets on food, the threshold should be five times a minimally adequate diet. The other alternative, termed the housing consumption measure, is based on the Fair Market Rent (FMRs), at that time set at the 45 th centile of local market rents and used in federal rental assistance programs. To produce a total threshold from a national average FMR, Ruggles first observes that Section 8 rental assistance requires recipient households to devote 30 percent of their countable income to meet shelter needs. By implication the program regards households that can cover the FMR with 30 percent of their 8

9 income to need no assistance. So, dividing the FMR by.3 produces the total standard of need implicit in the Section 8 rules. At 154 percent of the official thresholds for a family of four, the housing consumption threshold falls just below Ruggles updated multiplier threshold. In calculating her alternative thresholds, Ruggles assumes that the FMR and TFP are equivalent concepts representing necessary levels of spending for shelter and food. However, the TFP is a floor. Under normal circumstances, we would be concerned that a household spending much less than the TFP was not eating a nutritious diet. As will be explained in more detail in the next section, the FMR is a ceiling rather than a floor. Section 8 will subsidize rents up to the FMR. However, if a family is renting a standard quality unit of appropriate size for less than the FMR for shelter, the Department of Housing and Urban Development (HUD) regards that family as adequately housed. The FMR ceiling is set so high because the supply of affordable standard quality dwellings falls below demand. Ruggles understands that the FMR is not the lowest possible rent for a unit of a given size, but takes it to be a reasonable market rent rather than a ceiling. We will argue that the implicit standard of need in Section 8 is neither the lowest possible rent nor the FMR, but the median of market rents below the FMR. In their 1992 volume, The Forgotten Americans, setting out a comprehensive program to help low-income workers, John E. Schwarz and Thomas J. Volgy suggest an economy budget for a couple with two children, based on the TFP, the FMR, and actual or estimated spending on other categories of needs. Their 1990 pre-tax threshold stands at 155 percent of the official threshold; post-tax, the economy budget is 132 percent of the official threshold. The authors do not argue for the adequacy or normativity of the TFP or FMR, but emphasize the inconsistency of the statistical measure of poverty and the same government s program-based standards of need. One often hears that the Federal government s left hand doesn t know what the right hand is doing. (Schwarz & Volgy, p.36) Like Ruggles, these authors recognize that, near the median of all rentals, the full FMR may seem too high as a measure of housing need. They explain, Use of the 45 th percentile of units seems appropriate both because some of the lowestcost units are in rent-controlled apartments with long waiting lists (often they are rented by families with above-average incomes) and many other lowest-cost units are in dangerous neighborhoods that have high rates of violent crimes, areas that families may wish to avoid. (Schwarz & Volgy, 1992.Table 4, note 2) Trudi J. Renwick and Barbara R. Bergmann (1993) produce a Basic Needs Budget (BNB) to show that accounting for necessary work expenses, including transportation, child care, and taxes, increases poverty rates among working female family heads. While the authors note that their BNB is derived almost entirely from U.S. government-set standards, they do not argue either that such 9

10 official standards are well supported empirically or due rational assent as the judgment of elected government. Indeed, they reject the TFP in favor of the Department of Agriculture s Low-Cost Food Plan, on the grounds that research had shown few households spending at the level of the TFP were actually obtaining an adequate diet. Renwick and Bergmann set a shelter standard at the 25 th centile of rents for twobedroom units. They do not offer reasons for departing from Ruggles, who included the full FMR at roughly the 45 th centile of rents. However, because Renwick and Bergmann cite Ruggles approvingly on other points, it seems likely that they had concluded that using the full FMRs would overstate shelter needs. Daniel H. Weinberg and Enrique J. Lamas (1993) develop several alternative thresholds and income measures to illustrate distributional effects. They adopt the TFP as the measure of food need, and the 25 th and 35 th centiles in the distribution of nonsubsidized rentals as their shelter need measure. These authors do not mention the FMR or why it was not chosen, but indicate awareness of the effect of housing supply constraints: Arbitrarily, the 25th percentile in each region was chosen as the indicator of a minimal housing budget (as roughly one-eighth of all persons are poor, this allows some choice). The 35 th percentile was also chosen for analysis. In How Much is Enough? Basic Family Budgets for Working Families, Jared Bernstein, Chauna Brocht, and Maggie Spade-Aguilar (2000) make a valuable contribution to the poverty measurement literature by surveying recent basic needs budget proposals and their rationales before developing one of their own. They reject the TFP for the Low-Cost Food Plan on the grounds that: 1) families spending at the level of the TFP usually don t eat diets meeting nutritional standards; 2) many families run out of food stamp money before the end of the month; 3) about half of all families below 200 percent of the poverty line worry about running out of food. The authors adopt the FMR for their housing standard, noting the constraint on the supply of low-cost housing and HUD s quality standards. They do not reflect on the floor/ceiling issue, despite observing that FMRs represent, the maximum allowable for a family to rent the unit under Section 8. Bernstein, Brocht, and Spade-Aguilar utilize Medical Expenditure Panel Survey data and commercial insurance price quotes to calculate a weighted average of the costs of group and nongroup coverage to reflect the fact that low-wage workers often are not offered health insurance by their employers. Estimates of mean out-of-pocket costs then are added. Because their focus is working families, child care costs are included and emphasized. Data from a detailed, state level survey of providers are judged to be the best available. A combination of survey and other data are used to estimate needs for telephone 10

11 service, apparel and personal care, household supplies, and a miscellaneous residual. To place Bernstein, Brocht, and Spade-Aguilar in the literature using government measures to calculate standards of need, the authors principal interest in comparing needs to earnings from low-wage work should be kept in mind. Typically, researchers who wish to estimate post-tax poverty adjust pre-tax income for the effect of taxes before comparing the result to poverty thresholds that do not include amounts for taxes. Bernstein, Brocht, and Spade-Aguilar include estimates of direct taxes in their illustrative family budget for Baltimore families. Similarly, although they commend the poverty measure proposed in the 1995 NRC report, they include amounts for health care, child care, and work expenses such as transportation in their family needs budget rather than adjusting income as the NRC panel recommended. The authors goal is to measure the adequacy of low wages to meet all basic needs, which must be included on the other side of the equation in the threshold. The research and development summarized here, along with references to program standards in the NRC report that will be discussed in subsequent sections, display several patterns. Department of Agriculture food plans are used widely to establish food needs, with some researchers, including Mollie Orshansky herself, criticizing the least expensive plan as insufficient. The most common rationale is that, while, in theory, the TFP provides a nutritionally adequate diet, in fact low-income families spending at this level do not purchase the TFP items. Whether we are to conclude that the amounts associated with the TFP are impractically low because the items and prices do not reflect the selection and prices available to the low-income population is an open question. Shelter needs, typically the largest single category in a family s budget, were included in the multiplier in Mollie Orshansky s thresholds. Among the literature summarized here, the FMRs are used most widely. However, as noted and explained in more detail below, FMRs are a ceiling on rental assistance subsidies, rather than a floor on shelter needs like the TFP is a floor on food needs. Renwick and Bergmann and Weinberg and Lamas choose lower levels, and other authors who use the full FMR cautiously offer a variety of arguments as to why it does not overstate shelter need. No common approach to health care needs is evident. None of the work summarized takes the NRC panel s approach of eliminating health care from the measure of economic need in favor of a separate measure. Similarly, none of these authors suggests that, just as the TFP expresses a government measure of food needs, so the health care reimbursable under a program such as Medicaid expresses a government measure of health care needs. That will be argued here shortly. 11

12 Another common feature in the work summarized here is the largely unreflective way that the TFP and FMR are accepted as appropriate measures. To the extent that the rationale for these standards is mentioned, it is the supporting empirical evidence. The authors do not suggest that the processes of elected government legitimize the resolution of disagreements among reasonable but not compelling arguments such as those about the poverty line. II. Needs Standards Underlying Federal Means-tested Assistance Food Needs The Thrifty Food Plan underlying Supplemental Nutrition Assistance. The days of paper food stamps are over. These days, eligible households receive benefits in the form of electronic debit cards. However, the maximum allotment available to households that have no countable income is based primarily on legislation from the 1960s and 1970s. 5 In the 1970s, the Thrifty Food Plan (TFP), described earlier, replaced the Economy Food Plan as the basis of the maximum food stamp allotment. In addition, although previously the maximum allotment had been described as, such amount as will provide households with an opportunity more nearly to obtain a low-cost nutritionally adequate diet, the 1971 amendments to the 1964 Food Stamp Act (P.L ) provided that the, face value of coupon allotment issued to certified households be in such amount as the Secretary of Agriculture determined to be the cost of a nutritionally adequate diet, adjusted annually to reflect changes in the prices of food published by the Bureau of Labor Statistics. The predecessor Economy Food Plan had been described as sufficient to provide an adequate diet only for short-term or emergency use, and the TFP was pegged at the same dollar amount. So skepticism has been voiced over whether the TFP, even adjusted for changes in food prices, really is sufficient to provide an adequate diet over a longer period. To address this issue, about once each decade the Department of Agriculture s Center for Nutrition Policy and Promotion revises the TFP to reflect a minimal cost diet based on up-to-date dietary recommendations, food composition data, food habits, and food price information. (Carlson, Lino, Juan, Hanson, & Basiotis, 2007) The results of the latest such recalibration were published in 2007, and employ data from the National Health and Nutrition Examination Survey, the Food Price Database, and contemporary federal nutrition and food portion guidelines. It may be argued that these recalibrations do not build a new minimal cost nutritionally diet each time from actual food consumption data, but, instead, are designed to test whether the current TFP, the basis of SNAP benefits, is sufficient to purchase a diet meeting current nutrition and portion standards. However, given that the TFP is adjusted for food price inflation between these CNPP validations, we would not expect it to fall behind the cost of a nutritionally adequate diet 5 Draws on various documents related to SNAP Legislation, retrieved January 15, 2009 from 12

13 unless changing nutrition standards required relatively more expensive foods or the price of nutritious food rose faster than the price of all food. Efforts to provide evidence that the TFP represents adequate nutrition and realistic prices have been substantial. Nonetheless, the electoral and legislative processes provide a forum for scrutiny and challenge to the TFP, based perhaps on insufficient geographic price variation or a suspected stinginess with respect to quality or variety. In that connection, it is important to note that such a challenge, while hardly a simple matter, is likely to be more focused, empirically informed, and productive than debates over exactly where within the NRC panel s proposed range a new poverty threshold representing all, but unspecified, needs should be set. Other food assistance programs. When suggesting the TFP underlying SNAP benefits as a needs standard for food in a poverty threshold, the presence of other food assistance programs with their own implicit needs standards must be addressed. Principally, some SNAP beneficiaries also may be eligible for benefits from the National School Lunch Program (NSLP) and the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC). Maximum benefits under these programs also represent an implicit standard of need for food, and the issue is whether these needs standards should be combined with the TFP to complete the measure of food need in a new poverty threshold. In the case of the NSLP, they probably should not. There is no suggestion in documentation of the TFP that it is not meant to include the cost of lunches eaten by children attending school. On the other hand, pregnant and lactating women and infants under one year of age were not included in the National Health and Nutrition Examination Survey subsample used to develop the latest TFP. (Carlson, Lino, Juan, Hanson, & Basiotis, 2007) In calculating food needs in a program-based threshold, we will adopt the practice followed in setting the TFP, described in a note on Table C-1 in Nord, Andrews, & Carlson (2007), and assume that the nutrition needs of children in their first year of life will be met by breast-feeding, and that the additional nutrition needs of the breast-feeding mother will be reflected as onehalf the amount for feeding a toddler. Calculation of food needs in the poverty threshold. A standard for food needs based on The Thrifty Food Plan requires the simplest calculation of any component of our thresholds. The TFP specifies amounts for 17 age-gender categories, and uses only a single weighted national price. For this exercise, TFP amounts for December 2006 were employed, in which the age-gender categories were reduced to 12. (Nord, Andrews, and Carlson, 2007) To calculate a household s 2006 food needs, weekly TFP amounts were associated with each household member, then the amounts were summed over the household and 13

14 adjusted for economies of scale, following USDA practice. The result was multiplied by 52 to create an annual measure. The results, through households with nine members, follow: Table 1. Mean 2006 poverty threshold food component by household size Numer of persons Mean 1 1, , , , , , , , ,100 Shelter Needs The Fair Market Rents underlying Section 8 rental assistance. The Section 8 Housing Choice Voucher program is the largest federal program subsidizing shelter costs for low-income households, reaching over two million. The maximum Housing Assistance Payment (HAP) in behalf of voucher recipients is set at percent of what the federal Department of Housing and Urban Development (HUD) terms the Fair Market Rent (FMR). Every year, HUD estimates FMRs for 530 metropolitan areas and 2,045 nonmetropolitan county areas. (Office of Policy Development and Research, 2007) Decennial census records from two-bedroom renter households that moved recently are arrayed by gross rent, including all tenant-paid utilities, except telephones, cable or satellite television service, and internet service. Then FMRs are set at the 40 th centile of the distribution in each area. The resulting FMRs are updated for price changes since the last decennial census, then varied for different numbers of bedrooms based on cost ratios found in the decennial census data. 6 Other analysts have taken the FMRs to be conceptually equivalent to TFP, the basis of SNAP benefit for households with no income, but they are not. (Ruggles, 1990; Schwartz & Volgy, 1992; Bernstein, Brock, & Spade-Aguilar, 2000) The TFP is a floor, a national standard for a nutritious diet at a minimal cost. 6 Eventually, data from the American Community Survey (ACS) that collects the same array of data formerly on the decennial census long-form, but on an ongoing basis from a large sample of households, will be used to update the distributional data as well. 14

15 (Carlson, Lino, Juan, Hanson, & Basiotis, 2007) Strictly speaking, TFP is an average floor, because, depending on local variation in price, the items and quantities in the TFP may cost more or less than the national weighted average price allows. Nonetheless, under ordinary circumstances, spending much less than the TFP amounts for food can be a sign of an inadequate diet. By contrast, the FMRs represent ceilings that explicitly reflect supply constraints in the availability of affordable rental units that meet Section 8 quality standards. 7 HUD explains that it, sets FMRs to assure that a sufficient supply of rental housing is available to program participants. To accomplish this objective, FMRs must be both high enough to permit a selection of units and neighborhoods and low enough to serve as many low-income families as possible. (Office of Policy Development and Research, 2007) Under Section 8, any standard quality rental of the appropriate size meets a household s shelter needs, and roughly 40 percent of such units lie below the FMR. The FMR marks the maximum price Section 8 usually will subsidize. Spending less than the FMR ceiling for rental housing is not a sign of inadequate housing in the sense that spending less than the TFP floor for food is a sign of an inadequate diet. To constitute a program-based measure, a poverty shelter component must be based on the range of costs Section 8 will subsidize, a range between the least expensive standard quality rental of the appropriate size and the FMR. We know that wherever we set the shelter needs level within this range, there is very little chance that a particular household will actually find a rental unit priced at exactly this level. To find available rentals of standard quality, nearly all households will pay more or pay less than the shelter need standard in their poverty threshold. The TFP is no different in this respect. Because of variations in price and other unaccommodated variations in food need, such as health status, nearly every household s food need will be a little lower or a little higher than the appropriate TFP amount. In fact, with any statistical poverty threshold, there will be some error in measurement of practically every individual household s need. The goal is to minimize the combined effect of these errors in measuring the poverty rate of groups and the characteristics of the poor. 8 If we set the threshold shelter need component at the median of standard quality rental units below the FMR, we ll include too much in the poverty thresholds (and 7 Regulations at 24 CFR set criteria that units must meet to be eligible for Section 8 subsidy, pertaining to: sanitary facilities; food preparation and refuse disposal; space and security; thermal environment; illumination and electricity; structure and materials; interior air quality; water supply; lead-based paint; access; site and neighborhood; sanitary condition; smoke detectors. 8 By contrast, individual programs applying the needs standards employed here can ask applicants for more information than statistical household surveys usually can, and thereby come closer to measuring the needs of individual applicants. An example is the excess shelter deduction of SNAP. 15

16 so tend to overstate poverty) for the half with rents below the median, and include too little (and so tend to understate poverty) for those among the half with rents above the median who actually cannot find more affordable housing. While any point below the FMRs will involve about the same number of households misclassified either as poor or not poor, the median will tend to minimize net poverty misclassification. 9 In addition to location, FMRs vary by the number of bedrooms in the unit. Then HAPs vary according to family composition, which determines the number of bedrooms deemed necessary. HUD regulations (24 CFR ) prohibit more than two persons to share a bedroom, and require that children of the opposite sex, other than very young children, not share a bedroom. Local public housing agencies have flexibility to apply additional restrictions. The shelter needs component of our poverty thresholds will approximate the federal variations for household composition. Set close to the median for all rents and updated regularly for price changes, FMRs present a prima facie case for adequacy as the basis for a shelter need standard. However, given the prevalence of affordable housing supply constraints, we should ask whether other evidence supports the conclusion that FMRs are high enough that the median of lower rents represents an adequate shelter component in a poverty threshold. Analysis of Section 8 voucher success rates offers an entry into this question. If FMRs were set too low to permit voucher holders to find rentals of standard quality, we would expect to find many vouchers unused. In a sense, many are. In the mid-1980s, the success rate, or the rate at which voucher holders succeeded in using the vouchers to enter leases, was barely more than twothirds. This success rate increased to more than four-fifths by the early 1990s, but dropped again to 69 percent at the beginning of this decade. (Such temporarily unused vouchers may be reassigned.) A number of factors may have contributed to the drop in the latest measured success rates, including a policy that reduced FMRs from the 45 th to 40 th centiles in One policy response has been to give local Public Housing Agencies permission, on a caseby-case basis, to set FMRs at the 50 th centile. (Office of Policy Development and Research, 2001) The discussion over how much the FMRs contribute to low success rates is too specific and complex to treat adequately here. We will just note that this is the kind of question about benefit sufficiency, and, indirectly, about a standard of need, that can be engaged productively if poverty thresholds are built up from needs standards implicit in program parameters. 9 Median FMRs are not far below the full FMRs, and not highly dispersed. In the 2006 ACS, the mean ratio of median to full FMRs is.77 and the median is.79. The standard deviation of the ratios was.07, with half falling between.74 and.81. Only 10 percent were less than.66 or greater than

17 Other housing assistance benefits. As with food assistance, the shelter needs standards implicit in the FMRs do not cover all possible household arrangements, and many other smaller federally-funded programs provide lowincome households with assistance in meeting shelter needs. The largest is HUD s Public and Indian Housing, that funds the construction and improvement of units and subsidizes the rents of low-income households. The predecessor of Section 8, Public Housing has similar rules for eligibility and household contributions to rent. (Spar, 2006) However, its complex funding arrangements and questions about economic efficiency that led to Section 8 s creation disqualify Public Housing for our purpose of measuring a shelter need standard. The Low-income Home Energy Assistance Program provides $2 billion each year to states to help low-income households offset the costs of heating and necessary cooling. However, the FMRs explicitly include costs of most utilities and heating in their gross rent distributions. On the other hand, the FMRs explicitly exclude the costs of telephones and internet connections. The 1996 Telecommunications Act s Universal Service Fund, under which carriers and other users subsidize the cost of basic phone service for low-income households, is evidence that Congress regards phone service as something like a basic need. For this reason, the poverty thresholds developed here will include amounts for basic telephone service. The final issue with a shelter need component considered here is whether there should be separate amounts for households buying or owning their homes. The NRC panel recognized the theoretical validity of distinguishing between the shelter needs of renters and of owners with no or small mortgage payments. (244-47) Such a distinction can be quantified by imputing to owners income equal to the rental equivalence of the flow of services they receive from their homes. Typically, this is done by asking owners to estimate the amount they would pay to rent their home, then netting necessary costs, including mortgage payments, real estate taxes, and maintenance. The literature does not suggest that homeownership is itself a basic need, as shelter is. Accordingly, most suggestions are to add income to homeowners or lower thresholds where the costs of homeownership are lower than the implicit shelter component in the thresholds. Lower poverty thresholds for those with lower shelter costs is apt to be grasped more intuitively by laymen than the notion of imputed rental equivalence income. However, threshold variations for households with low homeownership costs departs from our general approach of developing poverty thresholds from needs standards implicit in means-tested programs providing for basic needs. In addition, the burden of collecting data necessary to estimate or impute homeownership costs reasonably well would not be insignificant. Consequently, our thresholds will not vary for households with low homeownership costs, but the open issue is acknowledged. 17

18 Calculation of shelter needs in the poverty threshold. To estimate household needs for shelter, microdata from the 1.3 million household 2006 American Community Survey (ACS) public use file was employed. Separately for household sizes of one through four, ACS rental housing units of standard quality were selected. 10 For each unit, gross rent plus utilities were summed. Then, for each state, units were sorted by this measure, and all those under the 40 th centile of gross rent were selected. The median gross rent plus utilities for units under the FMR 40 th centile ceiling was calculated for each unit size in each state to stand for shelter need in our thresholds. Typically, median rent under the FMRs are about four-fifths of the full FMRs. HUD calculates thousands of local FMRs, and also excludes units regarded as renting for less than market value. Our data source, the ACS, does not identify households receiving rental assistance from HUD, the largest category excluded as spending below market rent. To make a similar adjustment, data from the national samples of the American Housing Survey for 2003 and 2005 were pooled to calculate the share of each state s rental units that reported rental assistance. On the assumption that the market-level rent of half these subsidized units would be above the median of market rents under the FMR, and half below, one half of the percentage of each state s rental units represented by rental assistance units were dropped from the bottom of the ACS distribution before calculating the 40 th centile of the remaining units. For example, if rental assistance units represented 6 percent of all rental units in a state, then the bottom 3 percent of the distribution of rental units was excluded from the median FMR calculation. This adjustment will result in a slight overestimate of median FMRs, insofar as the measured median will be influenced by inclusion of some rental assistance units above the actual median. Other units besides rental assistance households may be excluded from the HUD FMR calculations on the grounds that they rent for less than market value, so our estimate probably will understate the median slightly. The table below displays the poverty threshold shelter components, by numbers of bedrooms, for each state. Following HUD s practice, gross rent for a fifth or sixth bedroom was set at 115 and 130 percent of the four-bedroom level, respectively. 10 To identify standard quality rental units when setting the FMRs, HUD selects on the following attributes: Occupied rental units paying cash rent; Specified renter on 10 acres or less; With full plumbing; With full kitchen; Unit more than 2 years old, and Meals not included in rent. Office of Policy Development and Research (2007). 18