Office of Management and Budget
9257 New Executive Office Building
725 17th St. NW
Washington, DC 20006
Submitted via https://www.regulations.gov
Re: Request for Comment on Consumer Inflation Measures Produced by Federal Statistical Agencies OMB-2019-0002-0001
Dear Dr. Potok,
I appreciate the opportunity to respond to the Office of Management and Budget’s (OMB’s) “Request for Comment on the Consumer Inflation Measures Produced by Federal Statistical Agencies” published in the Federal Register on May 7, 2019.1 The Economic Policy Institute (EPI) is a nonprofit, nonpartisan think tank created in 1986 to include the needs of low- and middle-income workers in economic policy discussions.
The administration is considering changes to the consumer price index used to update the Official Poverty Measure (OPM, or “poverty threshold”). This would directly affect poverty guidelines produced by the U.S. Department of Health and Human Services (HHS) used to determine eligibility for many government programs and could lead to pressure to change the index used to adjust Social Security benefits. Since such changes should only be made through normal regulatory and legislative processes, this letter should be considered a preliminary response to OMB’s request for comment, not a substitute for formal comment on proposed rulemaking or legislation.
Specifically, OMB appears to be considering a switch from the current Consumer Price Index for All Urban Consumers (CPI-U) to a “chained” version of the same index (“chained CPI” or “C-CPI-U”) for the purpose of adjusting the OPM and related HHS poverty guidelines to keep up with inflation. Since the chained index increases more slowly than the current one, this would have the effect of gradually reducing official poverty rates and the number of people eligible for means-tested government programs tied to the HHS poverty guidelines, including Medicaid and food stamps (the Supplemental Nutrition Assistance Program, or SNAP).
If the OPM had been adjusted using the chained CPI since December 1999 (earliest available data), roughly two million fewer people would have been officially counted as poor in 2017. During this 18-year period, the CPI-U grew by 47 percent while the chained CPI-U grew by 40 percent.2
Many more low-income Americans would have lost access to or receive reduced benefits as a result of a switch to a chained index, since eligibility for many government benefits is based on having an income below a multiple of the poverty threshold. For example, nearly 700,000 children would no longer qualify for free school meals because their family incomes would no longer fall below 130 percent of the OPM; and over a million elderly and disabled Medicare recipients would no longer qualify for low-income subsidies offsetting the cost of their prescription drug benefit because their incomes would no longer fall below 150 percent of the OPM. 3
These are just two illustrative examples. It is difficult to quantify the total number of people who would gradually lose access to benefits if eligibility thresholds were pegged to a slower-growing price index due to the large number of programs at all levels of government with eligibility standards linked to a multiple of the poverty threshold.4 An even greater number of people would be affected if Social Security and other benefit amounts were adjusted using the chained CPI-U.
People of color, women and children would be disproportionately affected. Whereas blacks are 13 percent of the population, they are 18 percent of people who would no longer be officially “poor” if the OPM had been updated using the chained CPI-U since December 1999. Similarly, Hispanics are 18 percent of the population, but 31 percent of those no longer considered poor; women and girls are 51 percent of the population, but 57 percent of those no longer considered poor; and children under 18 are 23 percent of the population, but 35 percent of those no longer considered poor. 5 These groups are also disproportionately represented among those with low and moderate incomes above the poverty line who would lose access to benefits, while seniors and disabled beneficiaries would be most affected by any changes to Social Security, Medicare, and Supplemental Security Income (SSI).
As I discuss below, using the chained CPI-U for policy purposes would increase hardship among beneficiaries of government programs, define away poverty instead of reducing it, and cause the poverty threshold to fall even farther behind a middle-class standard of living. Such a move would likely be presented as a technical correction designed to take into account households’ ability to substitute cheaper goods and services in response to rising prices. However, linking the poverty threshold to the chained CPI-U would do a worse job of identifying people in poverty and near-poverty who could benefit from targeted programs, and using the chained index to adjust benefit amounts would fail to maintain beneficiaries’ standard of living, for the following reasons:
I will devote more attention to the third and fourth points because the first two have been well documented by the Center on Budget and Policy Priority (CBPP) and others, and because OMB’s request for comment asks about price indices and the OPM rather than the HHS poverty guidelines, though the two poverty measures are closely linked.
When President Trump signed into law the Tax Cuts and Jobs Act of 2017 (TCJA), he changed the measure used to index tax brackets and other inflation-indexed amounts in the tax code from the CPI-U to the chained CPI-U. Since the chained CPI-U generally increases at a slower rate than the traditional CPI-U, taxpayers will end up in higher tax brackets and some tax credits will increase at slower rates than they would have under the old system, offsetting some of the enormous revenue loss from the tax cuts in the legislation.
While indexing tax parameters to a chained index has the politically expedient effect of raising tax revenues and reducing tax expenditures over time without requiring elected officials to enact unpopular legislation, there is no economic rationale that I am aware of for adjusting tax parameters to keep up with rising prices rather than rising living standards. Accelerating this form of “bracket creep” whereby even low- and middle-class households eventually end up paying a higher share of their income in taxes may have some practical advantages, but it should not be viewed as a providing theoretical support for adopting a chained price index in other contexts.
Similar proposals have been made to replace the Consumer Price Index for Urban Wage Earners and Clerical Workers (CPI-W) used to adjust Social Security benefits. The co-chairmen of President Obama’s National Commission on Fiscal Responsibility and Reform, Erskine Bowles and Alan Simpson, proposed indexing both tax parameters and Social Security benefits to the chained CPI-U, and the tax changes were later adopted in the TCJA.
OMB has requested comment on the CPI-W without directly signaling that it plans to replace it with the chained CPI-U for adjusting Social Security benefits. Such a move would likely spark a political backlash since it would affect all Social Security beneficiaries, rather than people on the cusp of eligibility for other government programs who may remain unaware that they would have been eligible absent the policy change. However, despite Social Security’s reputation as a political “third rail” and the president’s pledge not to cut Social Security benefits, it is hard to imagine that Social Security would remain an outlier if most other government programs were adjusted to inflation using the chained index.
1. The official poverty threshold is arbitrarily low.
In 1965, economist Mollie Orshansky of the Social Security Administration proposed defining poverty based on a family income below three times a subsistence food budget because a typical low-income family of that era spent a third of its income on food.6 However, a basic food budget in 1955 has little bearing on the overall cost of living today, factoring in housing, utilities, transportation, clothing, and other necessities, as well as government taxes and transfers. CBPP has surveyed the research showing that families whose incomes are somewhat above the official poverty line closely resemble those officially counted as poor and suffer similar hardships, such as food insecurity, difficulty affording rent and utilities.7
The Supplemental Poverty Measure (SPM) addresses some problems with the official poverty measure. It adjusts for geographic price variation; deducts taxes, medical expenses, work expenses, and child support paid to others from the measure of resources available to families; and adds tax credits and in-kind transfers not factored into the official measure. While the SPM does a better job than the OPM of comparing the relative hardship experienced by different groups, such as seniors and working-age adults, it still defines poverty using a low threshold, based on the 33rd percentile in spending on food, clothing, shelter, and utilities (FCSU). Unlike the HHS poverty guidelines, which are closely linked to the official threshold, the SPM threshold is not designed to be used for administrative purposes. However, a version of the SPM threshold could be developed that would be a modest improvement over the official threshold updated using the CPI-U.
In 2017, the OPM for a two-adult-two-child family was $24,858. In contrast, the SPM was higher for renters and homeowners with mortgages ($27,005 and $27,085, respectively), though it was lower for homeowners without mortgages ($23,261) who are a minority of the low-income population. In contrast, EPI’s basic Family Budget for the same size family in 2018, which factors in additional necessities such as transportation and childcare, is almost double that amount even in the lowest-cost county ($55,257 in Willacy County, Texas). It is considerably higher in most parts of the country.8
2. Seniors and low-income households experience faster inflation than the general population.9
The SPM threshold is updated using a 5-year moving average of FCSU expenditures.10 Thus, unlike the official measure, which is tied to the cost of food over a half century ago and only adjusted for price increases with the CPI-U, the SPM threshold roughly keeps pace with the cost of some basic necessities and a rising standard of living, even if the threshold is too low to begin with. Between 2005 and 2017, the SPM threshold increased by 30 percent, whereas the CPI-U increased by 25 percent and the chained CPI-U increased by 22 percent.11
Even an index designed only to keep up with the price of basic necessities (food, shelter, clothing, energy, and medical care), not a rising standard of living, will tend to outpace the overall price index because it is harder for people to find substitutes for necessities like food and rent. Between 1984 and 2014, such an index grew at an average annual rate of 2.91 percent, versus 2.65 percent for the CPI-U.12 Similarly, researchers at the Federal Reserve Bank of Chicago have estimated inflation rates experienced by different groups between 1983 and 2013, with results suggesting that less educated, elderly, disabled, and low-income groups faced steeper price increases than the average consumer.13
Another measure of the cost of living, the CPI-E, focuses on the consumption basket of seniors. Like the “basic necessities” index, it only adjusts expenditure shares, such as the higher share of their income seniors spend on out-of-pocket health expenses, as opposed to looking at specific goods and services seniors spend money on, such as durable medical equipment, and the prices they actually pay. During the same 2005-2017 period discussed earlier, the CPI-E grew by 27 percent, versus 25 percent for the CPI-U and 22 percent for the chained index. In earlier periods, when out-of-pocket health costs were rising more rapidly, the divergence between the CPI-E and the CPI-U was larger.
These estimates rely on varying the income shares spent on broad consumption categories such as food and shelter. Research that tracked actual spending by households using scanner data found that prices rose 33 percent for households with incomes below $20,000 and 25 percent for households with income above $100,000 between 2004 and 2013. The available data did not include health care at all, and underweighted spending on housing and transportation. Fully including these expenses, which loom large for many low-income and senior households, would almost certainly have shown greater disparities between low and high-income households.14
3. The claim that a chained index is the most accurate way to adjust dollar values to keep pace with the cost of living may rest on unrealistic assumptions about consumer preferences and markets even for the general population.
A chained price index is intended to serve as a middle ground between assuming consumers are perfectly indifferent when they substitute one good for another in response to a price change, and the opposite extreme—assuming consumers can only be made whole if they are fully reimbursed for the price increase based on their original consumption bundle.
Even leaving aside the issue of different consumption bundles for different groups discussed earlier, and even if the “true” cost of living index (COLI) lay somewhere in between these extremes for a representative consumer whose purchases resemble the population average, it does not necessarily follow that an arbitrary middle ground is necessarily a better estimate of the true COLI than either bound.
Imagine, for example, a consumer who buys two quantities of good A and two quantities of good B at one dollar apiece. When the price of good A doubles to two dollars, the consumer is observed buying one quantity of good A and three quantities of good B. If we assume the consumer is perfectly indifferent between the first (2A + 2B) and second (1A + 3B) bundles, a “Paasche” (lower-bound) estimate would show an increase in the cost of living for those goods of 25 percent, since the first bundle cost $4 and the second cost $5. If, on the other hand, we assume the consumer prefers the original bundle and would not have switched if the price had not at least doubled, a “Laspeyres” (upper-bound) price index would show an increase in the cost of living of 50 percent, since it cost $4 to buy the original bundle at the original price and $6 to buy the original bundle at the later price. Using the simplest (Fisher) version of the chained price index—since all produce similar results—the chained COLI would show an increase of 37 percent, because the square root of the multiple of 150 percent and 125 percent is around 137 percent.15
In fact, we do not know whether the consumer would willingly substitute the second bundle for the first in exchange for an increase in their budget of 37 percent. The correct amount could just as easily be 26 percent or 49 percent, depending on the consumer’s reluctance to purchase more of B and less of A. In the case where the consumer is nearly indifferent between the two bundles, the Paasche index would be closer to the true COLI than the chained index. If, on the other hand, the two goods are not close substitutes, the Laspeyres index would actually be closer to the true COLI than the chained index. Because the CPI-U already accounts for substitution between close substitutes using a geometric mean price index formula within item categories (e.g. different types of bread), and partially accounts for substitution between more dissimilar goods and services (e.g. bread vs. muffins) by reweighting expenditure categories every two years, the more relevant question seems to be whether the chained CPI-U underestimates the true COLI.
Economists often assume the chained CPI-U is the most accurate approximation of the true COLI for a representative consumer, using it as an ideal against which the current index is measured. However, this assumes an unrealistic textbook case where the consumer is indifferent between any combination of goods that lie along a smoothly convex indifference curve, and these preferences are homothetic—that is, not dependent on income.
Even in Microeconomics 101, the homotheticity assumption is treated as a special case. Engel’s Law says that most people will spend a smaller share of their income on a necessity like food (and, implicitly, a larger share on a luxury like travel) as their income rises. However, absent the highly restrictive homotheticity assumption, the chained CPI cannot be viewed as an inherently better measure of the cost of living than the current index. This problem is well known but tends to be brushed aside, since it can be shown that “for small differences in income levels and barring extreme curvatures of the indifference curves” a chained index is a good-enough approximation.16
Aside from some dissenting opinion about whether nonhomotheticity can so easily be ignored,17 there are many other situations that are not well approximated by the textbook model. The model, for example, assumes a representative consumer who purchases a bit of good A and a bit of good B, rather than all of one and none of the other. While this may be a fair description of how some households choose quantities of apples and pears based on their relative prices, many consumer decisions are not like this. A person who travels once a year to visit family may choose between a closer, more expensive, airport and a farther, cheaper, one based on price, but the result will be what is known as a corner solution—an all-or-nothing choice—not a small movement along a smooth indifference curve. In such cases, as in the case of infrequent and large price adjustments, different price indices may diverge more and biases in either direction may be larger.
The point is not to let the perfect be the enemy of the good, but simply to cast a modicum of doubt on the consensus view that the chained CPI-U is inherently a better approximation of the true COLI for a representative consumer than the current CPI-U. It very well might, but the assumptions that underlie this claim are rarely reexamined; and it is often mistakenly assumed that the current index does not account for substitution at all, which lends even more credence to the idea that the chained CPI-U is a technically superior measure.
This is especially true when most of the economic research to date has focused on supporting the claim that the current index overstates inflation. Thus, for example, it is often argued that the delayed introduction of new goods into the price index creates an upward bias, ignoring the opposite effect—a decline in the standard of living caused by goods and services disappearing entirely from the market, not necessarily because they are of poorer quality. Similarly, economists often argue that technological improvements are not fully accounted for in cost of living indices, even though for many consumers the value of increasingly powerful computers has little impact on their standard of living (as software becomes obsolete and replacement software requires more powerful computers) yet has the effect of lowering the CPI. In short, the economics profession has largely weighed in on one side of this debate, and a 45-day comment period is not long enough to right this imbalance.
4. Poverty measures should be indexed to living standards, not prices.
Whether or not the chained CPI-U represents a technical improvement in measuring the cost of living of a representative consumer over the current index, linking a poverty threshold to any price index will cause it to fall farther behind rising living standards. This problem is only compounded by the fact that different consumers buy different things, and even the most orthodox economist will not assume that the benefit one person gains from a cheaper computer makes up for the higher cost of another person’s insulin. Thus, it is generally better to focus on necessities and err on the side of faster upward adjustment.
When it was introduced in the 1960s, the OPM threshold was approximately equal to half the median household income, an alternative benchmark less sensitive to changes in the cost of food relative to other prices. Relative poverty measures based on half (or sometimes 60 percent) of the median household income are used by the Organisation for Economic Co-operation and Development (OECD) and other international agencies. While the two measures were close in the 1960s, the official poverty measure has since fallen behind the OECD benchmark. As a result, while the poverty rate in 2016 was 17.8 percent based on a threshold set at 50 percent of the median income after taxes and transfers, the official poverty rate was 12.7 percent.18 The fact that official poverty rates for age groups other than seniors have been flat or even trended upward since their inception even as real per capita GDP rose by a factor of 2.4 reflects poorly on the official poverty measure, our society, or both.19
Social Security has remained relevant for over 80 years because initial benefits received by retirees and other beneficiaries are indexed to average wages, though subsequent adjustments are tied to a consumer price index. In contrast, the poverty threshold will fall farther and farther behind a middle-class income over time since it is only adjusted to keep pace with price increases, not living standards.
Relative poverty matters as much as absolute poverty. While an absolute poverty standard may, in theory, determine whether a person’s survival needs are being met, social inclusion is based on a person’s relative standard of living. Thus, for example, a child with only one set of old clothes would feel ostracized in modern-day America, but might feel at home in medieval Europe or an isolated Amazonian village. Similarly, crowded living conditions that were the norm in an early 20th century tenement housing would now disqualify families from being foster parents.
The widespread ownership of electronic devices and modern appliances is often cited as evidence that poor Americans are not really poor. However, American children increasingly need Internet access to complete their homework, something unimaginable to people in poorer times or countries—even to many Americans a decade ago. Internet access is important not just for social inclusion but in order for children to become productive workers in a modern economy. Similarly, air conditioners were once luxury items, but have increasingly become life-or-death necessities, especially for seniors and children, as the population has shifted to hotter and more densely-populated areas, low-cost dwellings are not built to take advantage of shade and breezes, and temperatures have grown more extreme due to climate change.
Even seemingly frivolous purchases, such as handheld game consoles, often serve as inexpensive substitutes for forms of entertainment that have become less accessible to harried low-income families—including such “free” amenities as convenient access to parks and recreational facilities that families in high-cost neighborhoods take for granted. These devices can serve to distract children for parents who cannot afford childcare or who endure long daily commutes to school and work as rents in more accessible neighborhoods have grown out of reach. The fact that many devices have little resale value only a few years after they are on the market, as higher-income consumers discard them en masse to replace them with newer models, suggests that forgoing ownership of these devices might do little to help low-income families afford basic necessities like food or rent. Simply put, the reason many low-income families endure food insecurity at the end of the month has less to do with PlayStations and much more to do with erratic work schedules and low earnings.
Moreover, while technological progress slows the growth of consumer price indices, it increases the need for more powerful technology just to keep up. The computing power of a modern smartphone versus an old IBM mainframe may have improved our standard of living, but the fact that many of us carry such phones does not make us fabulously wealthy, and what would have once been considered a powerful computer is now used to stay in contact with family members as more mothers are in the workforce and as pay phones have disappeared.
If we want a cost-of-living index for poor people and other beneficiaries of government programs, we need to concern ourselves with what poor people and other beneficiaries actually buy, and what they might need to buy, to maintain a decent—if basic—standard of living that does not fall even farther behind a middle-class lifestyle.
2. Authors’ analysis of 1999-2017 U.S. Bureau of Labor Statistics’ (BLS) CPI-U and C-CPI-U accessed through https://www.bls.gov/data/ and 2018 U.S. Census Bureau Current Population Survey microdata, accessed through IPUMS-CPS, University of Minnesota, www.ipums.org.
3. See footnote 2.
4. For more examples in this vein, the Center on Budget and Policy Priorities (CBPP) has documented the effect of switching to the chained CPI-U on eligibility for a range of government programs after ten years. Since the effects are cumulative, more people would lose access to benefits or receive lower benefits if the chained CPI-U had been in effect longer than ten years (Aviva Aron-Dine, Matt Broaddus, Zo? Neuberger and Arloc Sherman. 2019. “Administration’s Poverty Line Proposal Would Cut Health, Food Assistance for Millions Over Time,” CBPP, June 18 https://bit.ly/2WOsqyz).
5. See footnote 2.
6. Gordon M. Fisher. 2008. “Remembering Mollie Orshansky—The Developer of the Poverty Thresholds,” Social Security Bulletin, Vol. 68, No. 3. https://bit.ly/2IK40kE
7. Arloc Sherman and Paul N. Van De Water. 2019. “Reducing Cost-of-Living Adjustment Would Make Poverty Line a Less Accurate Measure of Basic Needs,” Center on Budget and Policy Priorities, June 11. (https://bit.ly/2XXWjhb)
8. EPI Family Budget Calculator, accessed June 2019 at http://www.zhifa32.icu/resources/budget/; and Elise Gould, Zane Mokhiber, and Kathleen Bryant. 2018. “The Economic Policy Institute’s Family Budget Calculator: Technical Documentation,” March 5 (updated March 13). (http://www.zhifa32.icu/publication/family-budget-calculator-documentation/).
9. See Sherman and Van De Water (2019) for an overview of this topic (https://bit.ly/2XXWjhb).
10. Liana Fox. 2018. The Supplemental Poverty Measure: 2017, U.S. Census Bureau, September (https://www.census.gov/content/dam/Census/library/publications/2018/demo/p60-265.pdf).
11. Authors’ analysis of BLS SPM, CPI-U and C-CPI data accessed through https://www.bls.gov/pir/spmhome.htm and https://www.bls.gov/data/. The SPM threshold is a weighted average by housing tenure (homeowners with and without mortgages and renters).
12. Jonathan Church. 2015. “The cost of ‘basic necessities’ has risen slightly more than inflation over the last 30 years,” Beyond the Numbers, Bureau of Labor Statistics, June (https://bit.ly/2KsjXiF).
13. Authors’ analysis of the Chicago Fed Income Based Economic Index (IBEX), accessed June 2019 at https://www.chicagofed.org/research/data/ibex/ibex-inflation.
14. Greg Kaplan and Sam Schulhofer-Wohl. 2017. “Inflation at the Household Level,” Journal of Monetary Economics 91 (August), pp. 19-38 (https://bit.ly/2WYuGrX).
15. For a useful brief description of the chained CPI, see Alicia H. Munnell and William M. Hisey. 2011. “Implications of a ‘Chained’ CPI,” Center for Retirement Research at Boston College Issue Brief, September (https://crr.bc.edu/briefs/implications-of-a-chained-cpi/).
16. Marshall Reinsdorf and Jack E. Triplett. 2009. “A Review of Reviews: Ninety Years of Professional Thinking About the Consumer Price Index,” in Price Index Concepts and Measurement, National Bureau of Economic Research Conference on Research in Income and Wealth, Chicago: University of Chicago Press (W. Erwin Diewert, John S. Greenlees, and Charles R. Hulten eds.) (https://www.nber.org/books/diew08-1/).
17. Jesus C. Dumagan and Timothy D. Mount. 1997. “Re-examining the Cost-of-Living Index and the Biases of Price Indices,” Department of Commerce Working Paper, August 7 (https://bit.ly/2KsrI8c).
18. OECD and U.S. Census Bureau statistics accessed online at https://stats.oecd.org and https://www.census.gov/library/publications/2017/demo/p60-259.html on June 18, 2019.
19. Authors’ analysis of U.S. Census Bureau poverty rates and Bureau of Economic Analysis GDP data accessed June 2019 via https://fred.stlouisfed.org.
See more work by Monique Morrissey