National Poverty Center Gerald R. Ford School of Public Policy, University of Michigan www.npc.umich.edu
“Consumption, Income, and the Well‐Being of Families and Children” This paper was delivered at a National Poverty Center conference. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the view of the National Poverty Center or any sponsoring agency. Poverty in the U.S. Based on NAS Recommendations: Income and Expenditures by
Thesia I. Garner1 and Kathleen S. Short2
April 28, 2006
Discussant: Gary Burtless, Brookings Institution Conference on Consumption, Income, and the Well‐Being of Families and Children Sponsored by the National Poverty Center with Funding from the Office of the Assistant Secretary for Planning and Evaluation, U.S. Department of Health and Human Services Thursday and Friday, May 4‐5, 2006 Hyatt Regency Washington on Capitol Hill, Washington, D.C. JEL Classification: I320 Measurement and Analysis of Poverty
_____________________________________________________________________________________________
1
Thesia I. Garner, Senior Research Economist, Division of Price and Index Number Research, Bureau of Labor
Statistics (BLS), U.S. Department of Labor, Washington, D.C., garner.thesia@bls.gov, (202) 691 – 6576.
2
Kathleen S. Short, Senior Research Economist, Housing and Household Economic Statistics Division, U.S. Census
Bureau, Washington, D.C. 20233, kshort@census.gov, (301) 763 – 8921.
Part of this paper draws on results presented by Garner and Short during the 2005 Joint Statistical Meetings (JSM),
Minneapolis, Minnesota. The comments of Timothy Smeeding, discussant at the JSM session, are appreciated, as
are those of other participants in the session. The comments of colleagues in the BLS Division of Price and Index
Number Research, particularly Ralph Bradley, Rob McClelland, and Uri Kogan, greatly improved this research.
Appreciation is extended to Rob Cage who provided invaluable assistance with regard to the U.S. Consumer Price
Index (CPI), and Laura Paszkiewicz with regard to the U.S. Consumer Expenditure Interview Survey.
Census Bureau disclaimer: This paper is released to inform interested parties of ongoing research and to encourage
discussion of work in progress. The views expressed on statistical, methodological, technical, or operational issues
are those of the authors and not necessarily those of the U.S. Census Bureau.
Bureau of Labor Statistics disclaimer: The views expressed in this research are solely those of the author and do not
necessarily reflect the official positions or policies of the U.S. Bureau of Labor Statistics or the views of other staff
members. The author accepts responsibility for all errors.
Abstract -- This paper presents an alternative to the current official measure, now more than 40 years old. It
describes a new approach to measure basic needs and a consistent measure of family resources that are available to
people and families to meet those needs. Poverty thresholds are based on out-of-pocket spending by families on
basic goods and services. Resources are defined in terms of income and expenditures. A new measure of resources,
out-of-pocket expenditures that include allocations to savings with adjustments for necessary expenses, is compared
to the spending-based threshold. This measure represents the money spent or obligated by the household, its
“outflows,” that reflect a certain level of living. These “outflows” could be supported from current income, savings,
or dis-savings. In contrast to the expenditure-based measure used in this study, a consumption measure would reflect
the implicit value of the goods and services that the household consumes. Consumption too could be financed out of
the same sources as expenditures; however, implicit valuations must be made to determine levels of material wellbeing. A resource measure based on an “inflows” concept reflects money coming into the household that is available
to meet one’s basis needs. Using data for 2001-2003, results from income- and expenditure-based analyses are
compared. An important finding is that increases in expenditures for medical care, utilities, and shelter, captured in
the experimental poverty threshold, suggest a greater increase in the number of families not able to meet basic needs
than is reflected by the official poverty statistics. Income as opposed to expenditure-based resources results in lower
poverty rates. Not surprisingly, subtracting necessary work related expenses and other financial obligations increases
the probability that someone is below the spending-based threshold.
Keywords: Consumer Expenditures, Income Measures, Poverty Thresholds, Poverty Measurement
2
I. Introduction
In 1995, the National Academy Sciences (NAS) Panel on Poverty and Family Assistance, under the
auspices of the Committee on National Statistics (CNSTAT), published a report, Measuring Poverty, A New
Approach (Citro and Michael, 1995). In their report, recommendations were made that the current official measure
of poverty should be revised. The Panel argued that the official poverty measure is outdated given the changes in
U.S. society and in government policies (Citro and Michael, 1995; also see Ruggles, 1990). They noted that the
current measure does not account for the increased labor force participation of mothers and working families in
general. It does not account for differences in health care costs and needs. Family size adjustments in the thresholds
are anomalous in many respects. As living standards and notions of minimum needs change, thresholds too should
change. Resources should reflect changes in government policies that make it possible for families to meet their
basic needs.
Regarding the threshold, the Panel stated that the current poverty threshold “no longer represents the
concept on which it was originally based—namely, food times a food share multiplier [the percentage of after-tax
money income spent on food]—because that share will change (and has changed) with rising living standards.
Rather, the poverty threshold reflects in today’s dollars the line that was set some 30 [now 40] years ago” (Citro and
Michael, 1995, p. 25). The Panel noted that total expenditures of families have increased in real terms, and
spending on nonfood items has risen more rapidly than spending on food. While expenditures on food accounted for
one-third of the after-tax money income in the 1950s, they account for less than one-sixth of the total today. The
Panel stated, “If the original approach were used to develop the poverty thresholds today, their value would be
significantly higher” (Citro and Michael, 1995, p. 30). 1
1
The authors conducted an exercise to determine what the threshold would be if the same procedure were used
today as was used to derive the official threshold 40 years ago. U.S. Consumer Expenditure Survey quarterly income
and expenditure data from 2003Q1 through 2004Q1 were used along with the U.S. average value of the USDA
Thirty Food Plan in June 2003 (USDA, 2003). The Food Plan assumes that all meals and snacks are purchased at
stores and prepared at home. Since the Food Plan is based on the ages of older children and adults and the gender of
family members, the value of the Plan for a specific four-person family was used to determine the value of the food
adequacy standard for this exercise. The family was composed of a husband and wife who are both in the 20-50 year
age range, one child who is 2 years old, and another who is in the 3-5 year age range. After-tax money income data
with imputations for missing values are not available in the CE, so food shares are derived using total before-tax
money income (with imputations for missing data), total expenditures and consumption expenditures. Total
expenditures are defined by the Bureau of Labor Statistics for publication of CE data (USDL, 1997; BLS 2004).
Consumption expenditures are a subset of total expenditures but do not include cash contributions, insurance
premiums for life or other personal insurance, or contributions to retirement accounts or Social Security. The food
share of before tax money income for a four-person consumer unit was 10 percent, the food share of published CE
total expenditures was 14 percent and the share of consumption expenditures was 16 percent. These shares translate
3
The Panel noted that the current measure of resources, compared to the official poverty thresholds is
outdated and not consistent with the threshold measure. The measure of resources used for official poverty
measurement is defined as current before-tax money income. The official measure is based on before-tax income
because when the measure was first developed for poverty measurement, few families at the lower end of the
distribution paid taxes. Now many of these families qualify for the earned income tax credit that did not exist when
the measure was first developed.
Since the Panel’s report, research has been conducted on the experimental measures that they
recommended. Much of this work has been conducted at the Bureau of the Census and Bureau of Labor Statistics
(e.g., Garner, 2005; Garner et al., 1998; Johnson et al., 1997; Short, 2001, 2005; Short and Garner, 2004; Short et al.,
1999). The current research represents a culmination of this previous work, taking into account the concerns of the
research and policy communities regarding the desirable properties of a poverty measure (Iceland, 2005). Properties
deemed desirable by the panel include: consistency in construction, statistical defensibility, understandability, broad
acceptance by the public and operational feasibility (Citro and Michael, 1995, p. 4). The primary aim of this
research is to present a poverty measure that is based on the NAS recommendations with these desirable properties.
Although the Panel produced thresholds and resource measures, they emphasized that their purpose was to propose a
procedure that could be used to produce a poverty measure that is more appropriate for current society. The
procedure calculates a poverty threshold that is, by design, updated on a continuous basis and reflects changes in
levels of basic needs over time. The official measure, in contrast, is updated using changes in overall prices for an
average consumer (using the Consumer Price Index). The NAS proposed measure of family resources reflects the
effect of taxes and transfers and takes account of necessary work expenses such as child care and commuting to
work. The official measure does not take these into account.
The poverty measures presented in this study are based on the recommendations of the Panel, with one
deviation, the introduction of a resource measure based on expenditures, in this case, out-of-pocket spending. The
NAS Panel discussed resources as consumption or expenditures but their focus was a consumption resource measure
(Citro and Michael, 1995, pp. 210-214). Yet, the Panel described a poverty threshold based on spending needs --
into multipliers of about nine, seven, and six. Applying these multipliers to the annual value of the Food Plan
($4893.20) resulted in thresholds of $44,039 using before tax money income, $34,252 using total expenditures, and
$29,359 using consumption expenditures. The official poverty threshold for a two-adult two-child family ($18,660)
in 2003 is about 60 percent of the least generous of these thresholds.
4
median expenditures on food, clothing, and shelter increased by a modest additional amount to allow for other
necessities, without regard to when these items are actually "consumed."
For the thresholds, spending needs are defined in terms of out-of-pocket expenditures 2 using U.S.
Consumer Expenditures Interview Survey (CE) data. The values of most in-kind transfers (e.g., housing subsidies,
energy assistance) are not included in this measure. The one exception is food stamps. Food expenditures made
with food stamps cannot be separated from other food spending in the CE. Out-of-pocket spending underlies the
official thresholds as well; the multiplier used to create the official thresholds is based on food spending out after-tax
money income.
Out-of-pocket expenditures used to derive the thresholds for this study differ from the expenditures
regularly published by the Bureau of Labor Statistics (BLS) in that additional expenditures are counted, such as
payments to reduce one’s mortgage principal (for details, see Rogers and Gray, 1994). Earlier studies conducted by
the current authors used the BLS publication definition of expenditures, following the lead of the Panel. More
recent studies (Garner, 2005; Short, 2005) have compared income-based resources to expenditure outlays-based
thresholds.
The first resource measure presented is after-tax money income plus near-cash benefits that are available to
meet spending needs. Resources are defined as income net of necessary expenses (e.g., taxes and work-related
expenses) and compared to the spending-based thresholds. The resource measure is based on an “inflows” concept
and reflects money coming into the household that is available to meet one’s basis needs. The second resource
measure is out-of-pocket expenditures for consumption with adjustments for other spending including allocations to
savings and subtractions for necessary expenses and other legal obligations. Both resource measures are consistent
with the spending-based threshold presented in this study in that both represent resources that people have available
to meet basic spending needs.
Other researchers have used various definitions of expenditures to define resources for poverty
measurement for the U.S. For example, Federman et al. (1996) first used BLS publication-defined expenditures,
Garner and Short (2004) used BLS defined out-of-pocket spending (as well as a consumption-based measure), and
Bavier (2005) used BLS defined out-of-pocket spending but subtracted necessary expenses. In other studies of
5
poverty measurement (e.g., Charles et al., 2006; Cutler and Katz, 1991; Fisher et al., 2006; Meyer and Sullivan,
2003; and Slesnick, 1993) consumption (based on consumption expenditures) was the preferred resource measure.
The expenditure-based measure used in this study differs from these other measures in the concept used to guide its
construction. This measure is spending-based and thereby represents resources available or obligated by the
household. This spending could be or is being used to support basic needs. Spending "outflows" can be supported
from current income, savings, or borrowing.
An expenditure measure reflects, as does an income measure, one's ability to garner resources to actualize
consumption for a given period of time, regardless of when the goods are actually used up. A challenge in
producing an expenditures-based resource measure is to include all of the expenditures made by the household for
consumption and for other purposes. For poverty measurement, an expenditures-based measure can provide useful
information to policy makers as they try to understand the spending patterns and needs of consumers relative to their
incomes. The expenditure-based measure used in this study is new, and has not been used, to our knowledge, in
studies of U.S. poverty.
The paper begins with a review of the current official measure of poverty followed by a description of the
thresholds and the elements that make up the resource side of the experimental poverty measure. Thresholds and
resources are consistent in concept and in measurement method. The technical and conceptual issues that underlie
the NAS-based approach to poverty measure construction are highlighted. The NAS-based thresholds are compared
to the current official thresholds and examined with regard to underlying concepts and movements. Poverty rates
are produced that compare the measure presented here with the current official measure. Poverty rates for 2001-2003
are presented using the official poverty and experimental poverty measures. Poverty rates over time are presented
for the total population and for specific subgroups of the population that have been most susceptible to hardship
historically. The paper shows that resulting poverty rates are higher with a spending-based resource measure as
compared to an income resource measure.
II. The Current Official Poverty Measure
The current official measure was developed in the early 1960s as an indicator of the number of people with
inadequate income to cover the costs of a minimum food diet and to allow for other needed expenses (see Orshansky
2
Throughout the remainder of this study, spending and expenditures are used interchangeably.
6
1965; Citro and Michael 1995). The official thresholds are based on the share of food spending in after-tax money
income (one-third for families of three or more; other shares were used for smaller families) using data from the
1955 Household Food Consumption Survey, and a food adequacy standard based on the same survey, The food
adequacy standard selected was the USDA economy food plan, issued by the Department for “temporary or
emergency use when funds are low” (Orshansky 1965, p. 20). The food plan did not allow for meals eaten out or
other food eaten away from home. Thus, if family members ate at school or on the job, those meals were to be
prepared at home. Also for this standard it was assumed that “the homemaker is a good manager and has the time
and skill to shop wisely, she must prepare nutritious, palatable meals on a budget” for herself and her family
(Orshansky 1965, p. 24).
Since the first official thresholds were released, thresholds have been updated by changes in prices, holding
the multiplier constant at three and the food adequacy standard constant as the economy food plan, both based on the
1955 Food Survey data. In 1969, the Bureau of the Budget gave “official” status to the following change in the
poverty thresholds: to use the overall Consumer Price Index (CPI) to update the thresholds for price changes instead
of the Economy Food Plan that had been used earlier. Office of Management and Budget Policy Directive No. 14
(OMB, 1978) specifically states that, “The official poverty thresholds do not vary geographically, but they are
updated for inflation using [the] Consumer Price Index (CPI-U).” The CPI-U represents the change in prices of a
fixed market basket of goods and services, with relative prices changing while utility remains constant.3 The
updating mechanism for the official poverty measure is based on the relative change in prices of goods and services
purchased by an average consumer unit (using plutocratic, not democratic, weights 4 ).
The income measure for official poverty measurement is before-tax money income. This measure includes
the following sources of income: earnings, unemployment compensation, workers’ compensation, Social Security,
Supplemental Security Income, public assistance, veterans’ payments, survivor benefits, pension or retirement
income, interest, dividends, rents, royalties, income from estates, trusts, educational assistance, alimony, child
3
The CPI market basket changes on a periodic basis. In addition to this regular change in the market basket, other
changes have been made in the production of the index over the years (see the BLS web site for references).
4
Plutocratic weights are based on the value of goods and services across all consumer units for each commodity
represented in the index. For example, the weight for food is total spending on food for all consumer units divided
by the value of all goods and services for all consumer units. These weights are then combined with prices to
produce the index. Democratic weights reflect the distribution of consumer units in the population. For example,
consumer unit specific price indexes would be created based on the value of goods and services of each consumer
unit and the prices that the consumer unit faces. To obtain an overall index for the population, the consumer unit
7
support, assistance from outside the household, and other miscellaneous sources. Non-cash benefits (such as food
stamps and housing subsidies) do not count, nor do capital gains or losses. If a person lives with family members,
their incomes are added together and compared to thresholds that reflect their family size and composition.
III. Concepts, Methods, and Data
A. Poverty Thresholds
Poverty is most often defined in terms of one’s ability to meet his/her basic or minimum needs for survival
or participation in society. In the U.S., we have most often been concerned with the costs, at least officially, of some
minimum or basic bundle of goods and services that can be used to meet one’s needs, and the income or resources
available to meet those needs. A poverty threshold based on “costs,” such as the current official measure, is
expressed in terms of the dollar spending necessary to pay for a basic bundle of goods and services. For a spendingbased threshold, expenditures would most appropriately be used to derive the thresholds. For this study,
expenditures for a selected set of goods and services are used to construct a spending-based poverty threshold.
Others have constructed spending-based thresholds or standards using alternative approaches. As noted
previously, the official measure is based on food adequacy and the expenditures necessary to pay for adequate food
and to meet other spending needs. Other researchers have constructed spending-based thresholds by building basic
needs budgets, specifying adequacy standards for each of the major groups of goods and services that families
consume and adding an appropriate amount for direct taxes (for example, see Renwick and Bergman 1993;
Renwick, 1998). Spending-based self-sufficiency standards have been constructed by building basic family budgets
using local prices and expenditures, and expert judgment (for example, see Bernstein et al., 2000; Pearce, 2005a and
2005b). Personal assessments of what is needed financially to provide for one’s minimum level of living too have
been used to produce thresholds based on minimum spending needs. Minimum spending questions are used for the
assessments and thus the thresholds are often referred to as minimum spending question (MSQ) thresholds. (see
Garner and Short, 2003 and 2004; Morissette and Poulin , 1991).
One of the advantages of defining the poverty thresholds in the way proposed in this study, following the
NAS panel’s basic recommendations, is that we know exactly how the thresholds are constructed and how each of
the elements in the thresholds are measured. While we can surmise what is contained in the current official
specific price indexes would be averaged using demographic population weights to reflect the value of goods and
8
thresholds, there remains controversy about the exact measurement process used by Orshansky in the construction of
those thresholds (for example, Betson, 2000 and Bavier, 2001). Another important advantage of this procedure is
that the thresholds are automatically updated over time by actual spending on these basic selected items.
Setting the threshold has several steps. These include selecting a reference family, choosing an appropriate
equivalence scale, and identifying the goods and services included in the thresholds. The reference family used here
has two adults and two children. According to the Panel, the criteria used to select the reference family type was
that family would “fall near the center of the family size distribution rather than at one of the extremes…also, it is
preferable for the reference family to be one that accounts for a relatively large proportion of the population because
its spending patterns observed in a sample survey will be the basis for the poverty threshold…” (Citro and Michael,
1995, p. 101). Short and Garner (2002) found that about 9 percent of all families were two-adult/two-child families.
Of families with children, those with two adults and two children were the largest group. Since children make up a
large portion of the poverty population it is reasonable that the reference family represent spending patterns for that
group.
Once the reference family is chosen, median expenditures for a select group of goods and services are
calculated. This group of commodities includes food, clothing, shelter (including utilities), and a small additional
amount to allow for other needs (e.g., household supplies, personal care, and non-work-related transportation).
Also, medical needs are accounted for by including medical out-of-pocket expenses in the poverty thresholds. Thus,
medical care is treated as a basic need, along with food, clothing, shelter, and utilities.5
For this study, the threshold based on an out-of-pocket definition of expenditures includes food, clothing,
shelter, utilities, and medical care, and is referred to as the “experimental” threshold. The experimental threshold
includes the following:
•
Out-of-pocket spending on:
ƒ
Food
ƒ
Clothing
ƒ
Utilities (includes telephone)
ƒ
Medical care
services and price experience of all consumer units.
9
ƒ
For renters, shelter expenditures
ƒ
For homeowners, non-vacation shelter expenditures that include:
ƒ
Mortgage interest payments
ƒ
Repayments of mortgage principal
ƒ
Prepayment penalties
ƒ
Property taxes
ƒ
Maintenance, repairs, insurance and other related expenditures.
Percentile values for expenditures, based on percentages of the median, drive the poverty thresholds. The
use of the percentage links updates in the threshold to changes in expenditures at the median. The percentages
selected correspond to the reference family’s expenditures between the 30th and 35th percentiles of the distribution of
the sum of food, clothing, shelter, utilities, and medical care expenditures. Our applications have used the midpoint
of the recommended range to set the value of the thresholds. Once the percentage of median expenditures on a basic
bundle had been estimated, multipliers are applied to the basic bundle to add a small additional amount to allow for
other needs. A range of multipliers represents the basic bundle plus personal care, one-half transportation, education,
and reading materials costs. 6
Finally, an equivalence scale is applied to the reference family threshold to obtain thresholds for families of
other sizes and composition. We use a three-parameter scale that allowed for a different adjustment for single
parents (Betson, 1996). The three-parameter scale had been used in several BLS and Census Bureau studies
(Johnson et al., 1995; Short et al., 1999; Short 2001). The three-parameter scale is shown below.
One and two adults: scale = ( adults )0.7
(1a)
5
This represents a deviation from the original recommendations of the NAS panel that medical out-of-pocket
expenses be subtracted from income as a necessary expense. For earlier research with medical care included in the
thresholds, see: Banthin, et al. (2001), Bavier (2001), Short (2001), Short and Garner (2002).
6
Transportation expenditures were defined by the Panel to include vehicle finance charges, expenses for gasoline
and motor oil, maintenance and repairs, vehicle insurance, public transportation (including air fares), and vehicle
rentals, licenses and other charges. In addition, transportation included the total purchase price (minus the trade-in
value) on new and used vehicles.
Personal care includes products for hair, oral hygiene, and shaving, cosmetics and bath products, electric personal
care appliances, other personal care products, and personal care services.
Education includes tuition, fees, textbooks, supplies and equipment for public and private nursery schools,
elementary, and high schools, colleges, and universities, and others schools.
10
Single parents: scale = ( adults + 0.8* firstchild + 0.5* otherchildren )
All other families: scale = ( adults + 0.5* children )
0.7
0.7
.
(1b)
(1c)
The economy of scales factor was set at 0.70; the Panel recommended a range of 0.65 to 0.75.
The general formula for deriving the reference family threshold, using food, clothing, shelter, utilities, and
medical care is shown in the equation below.
(1 − smedical )
(1.15* PL * M ) + (1.25* PH * M )
)+
2
( smedical )
(2)
( PL * M ) + ( PH * M )
)
2
where
s medical = medical share of threshold value
PL = lower percentage of median costs
PH = higher percentage of median costs
M = median expenditures for reference family.
The multipliers of 1.15 and 1.25 are applied to the non-medical part of the threshold, (1-smedical), since they
are based on the relationship between the sum of food, clothing, shelter, and utilities expenditures and expenditure
for smaller and larger other bundles of needed goods and services as noted in the 1995 report. The three-parameter
equivalence scale is also only applied to the non-medical part of the threshold. This is because the medical care
needs of children are not expected to be less than those of adults and because there are few inherent scale economies
in medical care consumption with increasing family size.
Medical care expenditure risk indexes were created that account for variations in family expenditures as
related to family size, and the age, health status, and health insurance coverage of family members. 7 The medical
Reading materials includes subscriptions for newspapers, magazines, and books through book clubs, purchase of
single copy newspapers, and magazines, newsletters, books, encyclopedias, and other reference books.
7
One of the goals of the current research was to use the CE Interview Survey for the production of the thresholds
without auxiliary data from another survey. Since the CE does not include information on the health status of
members, only the remaining factors were used in the production of the CE-based medical risk indexes. An
additional challenge in using the CE to produce the medical indexes is that the CE does not collect health insurance
information for each member; data are collected on whether the consumer unit as a whole has various private health
11
risk indexes are calculated as the ratio of median medical out-of-pocket expenditures for different groups, varied by
the factors just noted, compared to the median expenditures of the reference family. Only the medical part of the
threshold is adjusted by the medical risk index.
This procedure to calculate thresholds repeated each year employs a quasi-relative approach for annually
updating the thresholds. This approach allows for “…changes in real consumption but in a conservative manner”
(Citro and Michael, 1995, p. 154). Such thresholds are more reflective of current needs than the official threshold.
B. Income-Based Resources
The next step in constructing a poverty measure is to calculate the resources that families have to meet the
needs specified in the thresholds. The resource measure presented here represents an attempt to construct a poverty
measure that is internally consistent. Cash and near-cash income, adjusted for required spending (e.g., income taxes
paid) that reduces a family’s income to meet its basic spending needs, is compared to a spending-based threshold.
The main theme in this section is that, while the selection of a poverty threshold is a policy choice, once
that selection is made, the measure of resources employed for comparison must be consistent. “It is important that
family resources are defined consistently with the threshold concept in any poverty measure. The current [official]
measure violates this principle, as has some recent work to investigate alternatives.” (Citro and Michael, 1995, p. 9).
The current official definition of poverty finds a family to be poor if total family pre-tax money income is
below that family’s poverty threshold, defined to be a particular dollar amount depending upon family size and
composition. This measure was based on the availability of data in the 1960s, so that, while the thresholds were
conceptually based on after tax income, for example, no estimates of after tax income were available at that time.
Income-based resources in this paper are the sum of money income from all sources plus the value of nearmoney benefits that help the family meet spending needs, less necessary expenses that must be paid. This alternative
concept of family income is referred to as “discretionary income” -- income that can be used to meet a family’s
basic needs (food, clothing, shelter, utilities, medical care plus a little bit more) after subtracting necessary expenses
such as taxes and work-related expenses.
Constructing an experimental measure of poverty starts with gross money income and adds values for nearcash transfers from federal programs aimed at helping poor families meet their basic spending needs. Experimental
insurance policies. However, data are collected concerning the total number of people who are covered by Medicaid
12
poverty thresholds as noted above include all expenditures for food, including purchases with food stamps. Since the
thresholds include all food spending, it is appropriate to include food stamps in a measure of resources. Food
stamps are the only near-cash benefit that is added to income to meet spending needs in this measure because no
other near-cash benefits are included in the spending threshold.
Other important benefits are often included as near cash income such as housing subsidies and school lunch
subsidies. 8 As mentioned earlier, spending on food in the thresholds only includes the out-of-pocket spending for
school lunch or breakfast, not the value of food consumed. The CE thresholds do not include subsidies for school
lunch by children at school. For this reason we do not add the value of subsidies to school lunches here. The CE
thresholds also do not include subsidies for housing 9 and thus are not included in resources. Federal housing
assistance consists of a number of programs administered primarily by the Department of Housing and Urban
Development (HUD). The programs generally reduce tenants’ rent payments to a fixed percentage of their income
after certain deductions, currently 30 percent. If the threshold is based on reported out-of-pocket spending then the
housing subsidy is not added to income since it does not help the family meet their spending requirement.
After calculating the income component of the resource measure, the next step is to subtract expenses that
must be paid before determining how much money is available to purchase basic necessities. Families must first pay
taxes and expenses required to work. 10 Taxes are subtracted from income since taxes must be paid and are not
available to meet basic needs. Federal and state income taxes and all payroll taxes paid are accounted for in this
resource definition (O’Hara, 2004).
Further, earning a wage may entail incurring expenses, such as travel to work and purchase of uniforms or
tools as well as care provided for children while parents are at work. These expenses are viewed as necessary and
are paid by families before expenditures are made on goods and services (see Short 2004 for a detailed
examination). The remaining monies represent the ability of the family to purchase, on an annual basis, what they
need. While borrowing and dis-saving are options that may be used to meet spending needs in the short run, this
and Medicare.
See, for example, the Census Bureau’s alternative income definitions (Census 2006).
9
Earlier studies included imputed values for housing subsidies in the thresholds (Garner and Rozaklis, 1999 and
2001).
10
Child support monies that are paid out should be deducted from income since they are included as income by the
receiving family. In Census Bureau income statistics using the CPS this is not done because data regarding the
amount of child support paid by one household to another is not collected, while the amount received by another
household is collected and added into income. The consequence of this is that child support transfers are doubly
8
13
measure suggests an amount that is required over the year to get along. Shortfalls in available resources on an annual
basis suggest circumstances that are untenable in the longer term.
The income measure used in this study includes realized capital gains. This is done as part of the tax model
calculations. The inclusion of realized capital gains in income is inconsistent with the basic income concept
recommended by the Canberra Group (2001) and International Labour Organization (ILO). However, the ILO
reports that the inclusion of realized capital gains is consistent with the Haig-Simons approach. For the Haig-Simons
approach, income is defined as the sum of consumption expenditure and change in net worth in a period (see
Simons, 1938, in Atkinson and Stiglitz, 1980). 11 The ILO recommended that income not include reductions in net
worth, “as a rule”, in income (p. 11) but noted that when “receipts from the sale of assets, withdrawals from savings
and loans obtained … are important for the analysis of the financing of consumption expenditure, they should, as
much as possible, be collected along with the income receipts” (pp. 11, 22). 12
Finally, the value of the flow of services of owned homes is not included in income resources. Garner
(2005) explains that for a spending-based threshold, a consistent treatment of homeownership in the resource
measure would not include values for net implicit rents from owner-occupied housing. Net worth, considered by
some in poverty measurement (Wolfe, 1990; Short and Ruggles, forthcoming), is also not included in the incomebased resource measure. It is important to note that expenditure-based resources may include spending from both
income and assets.
C. Expenditure-Based Resources
The expenditure measure used to define resources represents what is spent and what could be spent to meet
basic needs as defined by the threshold. The expenditure measure of resources reflects spending for consumption
goods and services financed out of current income, assets, or borrowing; interest associated with non-secured
counted in household income and official poverty statistics. These expenses and reported receipts were about $22
billion in 2001.
11
As noted in the ILO report, “This [Haig-Simons] approach does not impose any requirement of regular recurrence
nor requires the notion of not reducing net worth. No restriction to only those receipts available for current
consumption expenditure is made. The implication is that income should include all receipts, recurrent or otherwise,
regular or irregular and even those resulting in a reduction of net worth” (ILO, 2003, p. 9).
12
See Citro and Michael (1995, pp. 214-215) and Short and Ruggles (forthcoming) regarding the use of net worth
and poverty measure.
14
borrowing and other necessary expenses are subtracted. Allocations to savings and financial accounts and
expenditures for capital improvements to property that could be used instead for consumption are also added to
expenditures. Compulsory transfers (e.g., child support and alimony) are subtracted from expenditures as spending
for these cannot be used to meet the threshold spending needs of the household. Also, as is done for the income
resource measure, work-related expenses, including those for child care, are subtracted. Additional subtractions
include interest paid on credit cards and related debt used to support the purchase of consumption goods and
services, child support and alimony payments, and other legal obligations.
Three measures of expenditures are presented in order that the impact of each set of subtractions and
additions can be assessed. The first measure is out-of-pocket expenditures as defined by the BLS plus allocations to
financial accounts, the second subtracts from this amount the same necessary expenses as is done for the income
resource measure, the third subtracts from this second measure interest on home equity loans and lines of credit,
vehicle finance charges, interest on other consumer credit, child support and alimony.
D. Data
This paper uses several surveys to construct alternative poverty measures. First, the Consumer Expenditure
Survey (CE) quarterly interview data, for 1998 quarter two through 2003 quarter one, are used to construct
alternative poverty thresholds following the procedure recommended by the NAS panel. 13 CE Interview data are
made available on a quarterly cycle. Data collected in an interview refer to expenditures made during the three
months prior to the interview month. It is assumed that data from each reference quarter is independent of the data
from other quarters, just as it is assumed for CE publications. Thus, although annual thresholds are produced, these
are based on quarterly reports of expenditures. Three years of quarterly data are used to produce each threshold. For
example, for the 2003 threshold, 36 months of data from 2000 second quarter through 2003 first quarter are used.
Data from earlier years in the three-year cycle are updated to the threshold year using the annual All Items
13
When research began to produce the thresholds used in this study, it was expected that the Census Bureau would
be publishing the thresholds. Given the CE data release dates and Census Bureau publication dates, the data chosen
to produce the thresholds are from the three most recent years minus one. However, shortly before the thresholds
were completed, it was determined that the NAS-based thresholds would not be published; thus the three most
recent years of data could have been used as in the earlier BLS and Census Bureau studies previous cited. If the
earlier approach has been used, the 2003 threshold would have been based on data from 2001quarter 2 through 2004
quarter 1 and the thresholds would have been more reflective of more recent spending patterns. The experimental
15
Consumer Price Index, U.S. City Average (CPI-U). 14 Thresholds are produced for the reference family. The
reference family includes two related adults and two children, also related to the adults.
To measure family income or, as more broadly defined, family income-based resources, the analysis uses
the Current Population Survey Annual Social and Economic Supplement (CPS ASEC) for the income years 2001 to
2003. Payroll taxes, federal and state income taxes, and earned income tax credits (EIC) are based on a model
developed for the CPS ASEC. All but the EIC are also subtracted from income; EIC amounts are added. Information
from the Survey of Income and Program Participation (SIPP) 2001 panel is used to value work and child care
expenses, both are subtracted from income, in the experimental measure. Income-based resource measures are
produced for families and unrelated individuals.
The calculation of the income-based resources starts with current money income as defined and measured
in the CPS ASEC. Current money income is used to calculate official poverty statistics and includes cash income
received on a regular basis, such as income from earnings, any cash transfers, and property income. It includes
money income received in the previous calendar year by the family residing together as of February, March, or April
(the interview date) of the current year. Before-tax income, regularly received, does not include net realized capital
gains, gifts, lump sum inheritances, or insurance payments.
The CPS collects no information on taxes paid, so a tax model is required. As part of the tax calculator, net
realized capital gains are simulated and added to income in the computation of adjusted gross income (AGI). It
should be noted, this is done only for tax filers because non-filers have no such income by definition. Also, net
simulated realized capital gains are added to income even though they are not regularly received.
For work-related expenses (other than child care) a fixed amount per week is used for all workers. The
weekly amount is estimated to be 85 percent of median work expenses reported in the SIPP. Reported mileage
commuted to work is used with the IRS mileage allowance for taxes to compute these work expenses. Total
expenses were obtained by multiplying this fixed amount by the number of weeks respondents reported working in
the year. Accounting for child care expenses while parents worked in the CPS also uses estimated weekly median
expenses as reported in the SIPP. Only those families who indicated that someone paid for child care while they
threshold using the most recent three years of data is $21,635. This is somewhat lower than the threshold for 2003
used in this study ($23,170).
14
In another exercise, medians were produced based on an updating of the expenditures of threshold components
using their corresponding CPI-Us; there was essentially no difference in the medians in threshold year dollars with
this approach or when the overall CPI-U was used for the sum of expenditures.
16
worked were assigned these amounts. For these families weekly amounts were multiplied by number of weeks
worked by the “secondary” worker, or the worker with the lowest earnings.
For the expenditure-based resource measures, the analysis uses quarterly CE Interview data collected in
2001quarter two through 2004 quarter one. Expenditures for consumer units, as opposed to families, are collected
and thus are used for the analysis. However, experimental poverty rates are produced for people rather than
consumer units. Expenditures for the resource measure are quarterly; thus expenditure-based resources are
compared to the expenditure thresholds divided by four. For the CE Interview, collected expenditures refer to those
made in the three months prior to the interview month. This means, for example, that quarterly expenditures
collected during the 2001 quarter two through 2002 quarter one time period are compared to the 2001 threshold
divided by four. 15,16 The study time period begins with 2001 quarter two since several variables of interest (e.g.,
cash contributions and occupational expenses) are asked in each quarter of the Interview beginning with this quarter.
Before this time, questions regarding these variables were asked only once, in the fifth interview. The last year of
data, 2003, corresponds to the most recently produced experimental thresholds.
Poverty rates using three different expenditure-resource measures are compared to the experimental
thresholds. The first is used to provide a baseline for comparison to the other two that result from additions and
subtractions to expenditures. The base line measure is similar to that used by Garner and Short (2004) in their
comparison of expenditure outlays to official poverty and minimum spending and income thresholds, although
expenditures collected only in the fifth interview were not included in their measure of outlays. 17
The second and third expenditure-resource measures are attempts to produce measures that are more
consistent with the threshold. For these two measures, necessary expenditures that cannot be used to meet the
threshold spending needs are subtracted. For the second expenditure-based resource measure, payroll taxes, and
work-related and child care expenses are deducted from expenditure outlays. These are the same subtractions as
those made for the income-resource measure with the exception of income taxes. Income taxes are not included in
15
The expenditure months in scope for each threshold year could have been used. However, since the thresholds are
estimated using data collected in the first quarter of the following year, the same time reference period is used for
calculating poverty rates.
16
In contrast, Bavier (2005) used expenditure outlays adjusted for NAS necessary expenses with data for consumer
units for which four quarters of data were available and thereby produced annual poverty rates.
17
Garner and Short (2004) restricted expenditure outlays to those collected in Interview quarters from October 1995
through January1996. Expenditures not included were those for finance charges (excluding those on mortgages and
financed vehicles), occupational expenses, and cash contributions to organizations and non-CU members. This was
17
the CE definition of outlays and thus did not need to be deducted. The third measure additionally subtracts child
support and alimony payments, vehicle finance charges, other finance charges (excluding those for mortgages and
vehicles), and interest on home equity loans and lines of credit. All of these represent legal obligations of the
consumer unit and therefore expenditures for these cannot be diverted to meet the basic threshold needs of the
consumer unit. Added to expenditure resources are changes in financial accounts (i.e., positive changes to savings
accounts, checking accounts, and U.S. savings bonds), and expenditures for capital improvements to property (i.e.,
owned home, vacation home, and other property). For this study, it is assumed that increases in saving accounts,
checking accounts, and the purchase of U.S. savings bonds represent allocations that the consumer unit made rather
than returns to investments and thus could have been used for the purchase of goods and services instead. Given the
time period, interest rates for these financial accounts were low so we think that the assumption is reasonable.
All the additions and subtractions from expenditure outlays are based on reports from the CE Interview
except for work related expenses. Work-related expenses are subtracted for each working adult in the consumer
unit. 18 Work-related expenses are determined by multiplying the number of weeks worked, as reported in the CE,
by the same fixed amount as was done for the income-based resource measure. 19
Not all CE child care
expenditures are subtracted from outlays to obtain the third expenditure-resource measure. Child care expenditures
are only subtracted for a consumer unit with a dependent child who is less than 13 years of age (based on the U.S.
tax code of child dependency for child care) and in which there are two or more working adults or a single parent
consumer unit in which the parent is working. These values were capped at the level allowed for dependent care
credit in the federal tax code.
IV. Results
The thresholds for the three years are presented along with poverty rates using income- and expenditure
outlays-based measures. The thresholds are first examined in relationship to the movement of other economic
statistical series. This is followed by a detailed examination of how the experimental threshold differs from the
official threshold with particular reference to the CPI that is used to adjust the official poverty thresholds. Poverty
done in order that all consumer units would have the opportunity to have the same expenditures as every other
consumer unit each quarter.
18
Bavier (2005), when using expenditure outlays to compare to NAS-based thresholds, subtracted work related
expenses for the reference person or spouse only, not all working adults.
19
The amount deducted per week per working adult in 2001 was $17.94 and $18.96 for 2002 and 2003.
18
rates are presented for the total population and selected population subgroups using the income-based resource
measure and the three expenditure-based measures.
A. Thresholds
The experimental thresholds and the official poverty line for a family with two adults and two children are
presented in Table 1 and Chart 1. The experimental thresholds increase over the 2001 to 2003 period, just as the
official threshold increases. However, the rate of increase for the experimental thresholds is greater. The official
threshold in 2001 is $17,960 and in 2003 it is $18,660, a 3.9 percent increase. In contrast, the experimental
threshold for 2001 is $21,177 and $23,170 by 2003, a 9.4 percent increase. The experimental poverty threshold
increased almost three times as fast as the official poverty threshold. The official poverty line is about 85 percent of
the experimental threshold in 2001 but only 81 percent in 2003. 20
Table 2 compares the official and experimental poverty thresholds to published CPS average household
before-tax money income and published CE average consumer unit total expenditures 21 (available from the BLS,
2006 and Census, 2006, web sites) to determine if movements in the poverty thresholds reflect general movements
in the economy. Before-tax money income for all households increased by only 1.5 percent from 2001 to 2003; total
expenditures increased by 3.3 percent for all consumer units. These results indicate that income and expenditures
were not increasing as fast as the poverty thresholds but total expenditures were moving at about the same rate as the
official measure.
When comparing the experimental with the official thresholds, it is important to understand the differences
in the two thresholds. The official threshold is based on the assumption that one-third of all spending is for food.
As noted in the introduction, the official threshold is based on the cost of a minimum diet (defined in 1955)
multiplied by the food share as a percentage of after-tax money using 1955 Household Food Consumption Survey
data. The food share was based on the spending and income of all families with three or more people. With the
experimental threshold there is no explicit assumption regarding shares of spending, only that the threshold accounts
20
To compare the relative differences in spending needs determined by the experimental threshold and other
spending-based thresholds, thresholds for 1995 from Garner (2005) are considered. The 1995 experimental threshold
is about 97 percent of the minimum spending threshold (MSQ) threshold presented by Garner and Short (2004) for
that same year. It is 95 percent of the value of the budget-based threshold constructed by Renwick (1998) for a
husband-wife family with two children and one working parent, and 78 percent of the Renwick threshold for this
family but with both parents working.
19
for food, clothing, 22 shelter, utilities, medical care, and other basic goods and services. The experimental thresholds
are produced each year using the most recent three years (minus one year) of spending of families composed of two
adults and two children. The official poverty threshold is updated each year to current dollars using the CPI. 23
Changes in the official poverty threshold each year are based on movements in the CPI, while changes in
the experimental threshold are based on movements in consumer spending. The consumer price index is based on
price changes and expenditure weights for aggregation. To better understand how the official and experimental
thresholds compare, expenditure shares underlying the CPI and expenditure thresholds are presented for 2001-2003.
Indexes for all items and for food, clothing, shelter, utilities and medical care are presented to show how the overall
CPI moves relative to the CPI of selected subcomponents. Implicit shares in the experimental threshold are also
presented to highlight differences between the official poverty measure and the experimental threshold, and to
provide insight regarding why the official and experimental thresholds move differently.
Shares of expenditures implicit in the experimental thresholds for the reference family are presented in
Table 3, and are based on the underlying distribution of expenditures used to derive the thresholds. The thresholds
are based on percentages of the median, 78 and 83 percent, corresponding to the 30-35th percentiles of the
distribution based on the sum of expenditures for the threshold bundle. The expenditure shares implicit in the
thresholds are based on shares in this lower part of the expenditure distribution. Tables 4a and 4b include shares
based on aggregated annual expenditure estimates for the CPI. The shares in Table 4a are based on the 1998 CPI
market basket structure (see Greenlees and Mason, 1996) while Table 4b is based on the definition of expenditures
used for the experimental threshold. The shares in both Table 4a and 4b are derived from the same underlying
21
Published CE expenditure data are based on integrations of expenditures collected using the Diary and Interview
instruments.
22
The NAS Panel referred to Apparel as Clothing in the 1995 report. In this study, we do this as well. Thus,
Apparel in the CPI-U is referred to as Clothing.
23
Over the 1993-2003 period of this study, the Consumer Price Index for All Urban Consumers (CPI-U) was used to
adjust the official poverty thresholds from year to year. The CPI-U is based on the spending and prices faced by
consumers living in metropolitan statistical areas (MSA’s) and urban places of 2,500 inhabitants or more. Non-farm
consumers living in rural areas within MSA’s are also included. In contrast, the experimental thresholds are based
on the spending and prices faced by consumers living in both urban and rural areas. The CPI-U measures the
average change in the prices paid by urban consumer for a fixed market basket of goods and services. As noted in
the BLS Handbook of Methods (USDL, 1997, p. 170), “The CPI uses a fixed market basket to hold the base-period
living standard constant. The CPI equals the ratio of the cost of the base-period basket at this month’s prices to the
actual cost of the base-period basket in the base period. It is an index of price change only and does not reflect
changes in buying patterns that consumers probably would make to adjust to relative price changes.” The weight of
an item in the CPI is derived from expenditures from the CE. These are known as quantity weights. These weights
are used along with the prices for goods and services to calculate the index.
20
data. 24 There are some differences between the CPI categories of expenditures and the category definitions used for
the experimental poverty thresholds. For example, the CPI includes the value of owner-occupied shelter services
(rental equivalence) while the experimental threshold shelter is based on out-of-pocket expenditures. The
experimental threshold definition structure includes telephone services with other utilities (Table 4b); these are not
included in utilities in the CPI structure and therefore are not included in Table 4a. Medical care in the CPI structure
includes adult day care while this is not included in medical care in the experimental threshold definition structure. 25
The results presented in Table 3 reveal that food accounted for less than a third (28.0 percent) of the
experimental threshold in 2001 followed by shelter at 30.1 percent. By 2003 this relationship had reversed with
shelter accounting for 30.5 percent of the experimental threshold and food for 26.7 percent of the total. Utilities
come in third (not counting “other”) with between 13 and 14 percent of the total, followed by medical care and
clothing. Expenditures for other basic goods and services account for about 15.6 percent of the experimental
threshold from 2001 to 2003.
24
Only the 1999-2003 data are official production data used for the CPI-U. Data for the earlier years were not
specifically used for the production of the official CPI-U but are used in this study to show the shares for the
commodity groups defined using the CPI-U market structure and the experimental threshold structure. The shares
are based on aggregated CE data and are presented for each elementary item using the 1998 CPI market basket
structure (there are 211 elementary items). The data file used to create the shares contained aggregate annual
expenditure estimates for the CPI-U population for each year 1993-2004 (the last year of data are not used in this
study). The 1999 to 2004 data are official production data. That is, the composite estimated and raked annual
expenditure data are those that were used to construct the biennial aggregation weights for the CPI-U periods 19992000, 2001-2002, and 2003-2004. Prior to 1999, the BLS did not composite estimate and rake expenditure data
annually. According to Cage (2006), “only the 1993-1995 data were used as an official CPI-U expenditure
reference period, and back then the data were aggregated together (over the three years) prior to composite
estimation.” Hence, the 1993 to 1998 data are not composite estimated and raked. Cage notes that this probably
does not have significant effect. He did state however that “the 1993 to 1998 data may have slightly higher variance
than the 1999 [and] forward data. But then again, the sample size of the CE was increased in 1999 so the difference
in variance would be expected.”
25
Details regarding the differences between the CPI-U groupings, using the 1998 market structure, and the
experimental threshold groupings are available from the authors. However, some of these differences follow. (1)
CPI food includes the value of meals as pay; the experimental threshold measure for food does not. (2) CPI clothing
(apparel) includes clothing for men, boys, girls, women, and infants and toddlers, footwear, jewelry and watches
while the experimental threshold clothing includes these plus material for making clothes, sewing notions and
patterns, shoe repair and other shoe service, coin-operated apparel laundry and dry cleaning, alteration and repair of
apparel and accessories, clothing rental, watch and jewelry repair, apparel laundry and dry cleaning not coinoperated, and clothing storage. (3) CPI shelter includes the rent of primary residence including rent as pay, lodging
away from home, owners’ equivalent rent of primary residence, and tenants’ and household insurance. The
experimental threshold shelter includes mortgage principal and interest of primary residence, prepayment penalties,
property taxes of primary residence, shelter rent (not on vacation, away at school, or rent as pay), tenants’ and
household insurance, property management for primary residence, maintenance and repairs, and some materials for
remodeling. (4) CPI utilities are those for primary and vacation rental and owned properties and include natural gas,
electricity, fuel oil, other fuels, and water and other public services (not including septic tank cleaning). The
experimental threshold utilities include only the utilities for primary residences and rentals and include those in the
21
In Tables 4a and 4b, shelter accounts for approximately the same share of total CPI expenditures as it does
in the experimental thresholds, approximately 30 percent. In contrast, food accounts for about 14.5 percent of total
CPI expenditures (Table 4a), almost half the share implicit in the experimental thresholds. The expenditure share for
utilities in the CPI is also about a third of the experimental threshold share for utilities. In the CPI expenditures,
clothing accounts for about 4.7 percent while medical care accounts for a slightly higher percent of the total at about
5.5 percent (Table 4a). The largest share of expenditures in the CPI (about 40 percent 26 ) is represented by goods and
services not represented by the basic needs in the thresholds; food, clothing, shelter, utilities, and medical care. The
shares in Table 4b, based on the definitions used for the experimental thresholds, reveal higher shares of
expenditures for clothing and utilities than when using the CPI aggregation structure.
The results presented in Tables 3, 4a, and 4b reveal how the underlying structure of the CPI, which is used
to adjust the official poverty threshold each year, and the experimental thresholds differ. The official poverty
threshold is influenced more by goods and services “other” than food, clothing, shelter, utilities, and medical care
than is the experimental threshold. Food and shelter account for about 40 percent of CPI expenditures while they
account for about 60 percent of expenditures in the experimental threshold.
Rates of increase in the CPI and underlying threshold expenditures are presented in Tables 5 and 6. Table 5
includes the CPI for all items and for the items included in the experimental threshold. 27 Table 6 includes the
implicit average annual expenditures in the experimental threshold for the reference family. Expenditures are
presented in threshold year dollars. The experimental threshold rose faster than the all items CPI from 2001 to 2003
(3.9 percent versus 9.4 percent), and thus rose faster than the official poverty threshold, as noted earlier. Food
expenditures increased at about the same rate in the thresholds (4.2 percent) as did the CPI for food (4.0 percent).
Implicit clothing expenditures in the thresholds rose by 7.5 percent in the thresholds but the CPI for clothing fell 5.0
percent. Shelter expenditures in the thresholds increased by 11 percent compared to an increase in the shelter CPI of
6.2 percent. Implicit expenditures for utilities also increased at a faster rate in the thresholds than the CPI for
CPI with the addition of septic tank cleaning and telephone services. (5) Medical care is defined the same in both
with one exception: adult day care is only included in the CPI medical care category.
26
Approximately 17.9 percent of the CPI-U aggregate is for transportation, 6.3 percent for recreation, 5.9 percent for
education and communications, 5.0 percent for household operations, and 4.2 percent for other goods and services
(authors’ own calculations using aggregates provided by Cage (2006).
27
A CPI for “other” is not presented as there is no CPI for “other” and the authors did not calculate a composite
index to represent the other commodity groups.
22
utilities (11.8 percent compared to 2.9 percent). The greatest increase in threshold expenditures from 2001 to 2003
was for medical care, increasing 23.3 percent. The increase for the CPI was only 8.9 percent.
The experimental threshold changes each year based on the buying habits and spending of families with
two adults and two children whose expenditures are at the lower end of the expenditure distribution (between the
30th and 35th percentiles of the distribution for the sum of food, clothing, shelter, utilities and medical care
expenditures). The official poverty threshold changes each year based on the average change in prices paid by urban
consumers for a specific market basket of goods and services. The results in Tables 4 through 6 have been used to
highlight differences in the two thresholds. The official poverty threshold is adjusted by the CPI each year so
movements in the CPI were used to explore differences in the official and experimental thresholds. The
experimental threshold increased more rapidly during the 2001-2003 time period than did the official poverty
threshold. Even though shelter accounts for about the same share of the CPI-U expenditures and the experimental
threshold, shelter expenditures increased at a faster rate in the experimental thresholds. Utilities increased at a much
faster rate in the experimental thresholds and represent about three times the expenditures in the CPI.
To summarize, these results reveal that the spending for basic needs as represented by the experimental
threshold is increasing at a faster rather than the CPI. This is reflected by differences in commodity composition and
movements in spending over time.
B. Resources
Change in poverty thresholds would be expected to change trends in poverty statistics over time. However,
these statistics, when measured using income are also affected by changes in income and other tax and transfer
policies. Expenditure-based resource measures can also be affected by changes in policies. These changes too may
change trends in expenditure-based poverty rates over time. An examination of the elements in the resource
measures sheds light on changes in overall resources.
Table 7a and Chart 2 include the necessary expenses that are subtracted from both cash income and out-ofpocket expenditures to derive at the resource measures used for experimental poverty measurement for 2001 to
2003. These include payroll taxes, work expenses, and child care expenses. Additional expenses are subtracted from
before-tax money income to derive the experimental poverty measure. These include federal and state income taxes.
Additions to income include the cash value of food stamps, net realized capital gains, and the earned income tax
23
credit (see Short 2005). The value of food stamps is implicitly included in expenditures since these are used for
spending. Payroll taxes increase in the CPS and CE samples from 2001 to 2003. Since, over time we are
subtracting larger and larger amounts for payroll taxes, the experimental measures are expected to result in higher
poverty rates over time. Other expenses do not change as much and are thus not expected to affect trends in poverty.
Average annual payroll taxes are slightly higher in the CPS than in the CE, both around $3,000 by 2003. Work
expenses are marginally lower in the CPS than CE (both about $1,000). This might be expected since a consumer
unit is defined to be more inclusive than a family in that it may include unrelated individuals. By 2003, average
child care expenses are almost indistinguishable for the CPS and CE. The averages presented are computed over all
families and consumer units in the study samples and therefore do not represent the averages for only those families
and consumer who incurred the expenses.
Additions to out-of-pocket expenditures are shown in Table 7b and Chart 3 along with additional
subtractions. Positive changes in financial accounts vary from about $700 in 2001 to almost $900 in 2003.
Consumers spent an annual average of $765 in 2001 on capital improvements to property; they increased this
spending to $1,048 by 2003. The NAS Panel recommended that child support payments be subtracted from an
experimental resource measure. Since the CPS does not include data on child support payments, they were not
subtracted to derive the income-based resource measure. However, these payments are collected in the CE and thus
can be subtracted for the expenditure-based measure. Annualized average child support payments decreased over
the 2001 to 2003 period from $195 to $165. Alimony payments are also subtracted as legal obligations;
expenditures for payments of alimony cannot be used to meet the basic needs of the consumer unit making those
payments. These payments were on average quite low, less than $50.
The other subtractions from out-of-pocket expenditures are those associated with the use of credit, other
than that for owned real estate property. Since the interest on mortgages for owned home that are financed is
included in the spending-based threshold, this interest is included in expenditures, not subtracted from expenditures.
Subtractions of expenditures associated with the use of consumer unit include those for vehicle finance charges,
finance charges other than those for mortgages and vehicles, and interest on home equity loans and lines of credit.
The average total of these charges ranges from $672 in 2003 to $827 in 2002. Clearly credit is important to
consumers in financing their levels of living.
24
The result of the additions and subtractions on average annual resources are shown in Table 7c and Chart 4,
along with other statistics previously presented including before-tax money income, CE published total
expenditures, and the poverty thresholds. The income-based resource measure is $14,926 less than the before-tax
money income measure used for official poverty measurement in 2001; this difference increases to $16,430 by 2003.
There are only marginal differences between published CE total expenditures and the base expenditures, defined as
outlays plus some additions, used for this study. However, once necessary expenses are subtracted from the base
expenditures, differences in expenditure resource measures arise. The base expenditure resource measure for 2001
is $39,215 and by 2003 it rises to $41,130. The NAS necessary expense subtractions result in the second
expenditure resource measure; the second expenditure-based resource measure is from $3,700 to $4,000 lower than
the first measure in 2001 as compared to 2003. The average annual expenditure for the last measure, the one that is
most consistent with the spending-based threshold concept, is even lower than the others. In all, subtractions from
the base expenditure measure result in reductions in spending available to meet basic needs by about $4,900 in 2003.
The income-based resources are leveling out between 2002 and 2003 resulting in less of a difference between that
measure and the third and last expenditure-based measure. The difference is about $8,800 in 2001 but $6,400 by
2003. These averages suggest that the income-based measure will result in lower poverty rates than will the
expenditure-based measure.
C. Experimental Poverty Rates
To determine poverty status, total family or consumer unit resources are compared to the spending needs
thresholds. If resources are below the amount needed, then all individuals in the family or consumer unit are
classified as poor. The official poverty measure shows the percent of people in families with before tax cash income
below official poverty thresholds. The experimental measure uses poverty thresholds that provide a more current
estimate of the cost of an explicitly defined set of basic goods and reflects out-of-pocket spending needs. Taking
account of spending on basic goods, and the additions and subtractions to base resources, provides information
about individuals who, while not poor using the official measure, may have difficulty meeting basic needs.
Table 8 and Chart 5 show poverty rates for the period of time from 2001 to 2003. The rates shown using
income measures are based on annual data while those based on expenditures are based on quarterly data. The
figures show that the official poverty rate rose across the period from 2001 to 2000 from 11.7 to 12.5 percent. The
25
experimental poverty measure using income-based resources displays a similar though more pronounced pattern
with a rate of increase around 14 percent as compared to 6.8 percent for the official measure. Poverty rates based on
expenditure resources also exhibit greater increases than the official measure but not as great as the income-based
measure. Differences in trends, between the official and experimental measures, are due to increases in the
experimental thresholds as well as increased payroll taxes and work expenses that families and consumer units face.
Poverty rates based on income resources are slightly lower than those based on expenditures with no
subtractions. Once necessary expenses and other obligations are subtracted from base expenditures, poverty rates are
8 to 9 percentage points higher than the rates based on income resources.
Finally, Table 9 shows trends in poverty measures for two basic demographic subgroups, age and race
categories. These groups are of interest because children and Blacks have historically high poverty rates. While
levels vary in expected ways, there are also some differences in trends for these groups. Trends in poverty rates for
children and for Blacks roughly parallel the official measure. For these groups, poverty rates in 2003 are higher than
in 2001. For the elderly and persons of other races, poverty rates in 2002 are lower than those in 2001 when the
expenditure-based resource measure is used.
All of the figures shown indicate differences between the two resource measures. This is an important area
for further investigation. Large differences in poverty rates from using the income- and expenditure-based measures
suggest that income and expenditures do not reflect the same ability of families and consumers to meet their basic
needs. More likely, this could be an indication that our measures need refinement. For example, is too much income
being assigned to families based on the addition of net realized capital gains based on the computation of the current
CPS tax model? Are all expenditures being captured in our measure? Should the value of goods and services drawn
from one’s business for personal use be counted in expenditures? Should they be counted in income?
Is the time period for expenditures, quarterly, too short to adequately reflect expenditures for poverty
measurement? When quarterly reports of expenditures are used for the analysis, negative medical care expenditure
are more likely to result than when twelve months of expenditures are used. Negative medical expenditures result
when reimbursements are made for expenditures made in an earlier period. Should these reimbursements be
considered income and thereby not be counted in an expenditure-based resource measure? Setting negative medical
expenditure expenditures to zero results in poverty rates that are about three percentage points than the ones
presented in this study (results not shown). Bavier (2005) used out-of-pocket expenditures, as defined by the BLS,
26
with subtractions for necessary expenses but limited his sample to consumer units with four complete quarters of
data. By restricting his sample in this way (and not subtracting as much for work expenses), he produced poverty
rates of 16 percent for consumer units. The poverty rate for consumer units in this study using a similar measure (the
second expenditure-based resource measure), results in consumer unit poverty rates of 21.5 percent for
approximately the same time period. In additional work by Bavier (2006), consumption expenditure-based poverty
rates resulted in annual poverty rates 1.5 percentage points lower than rates based on quarterly consumption
expenditures. Evidence from the SIPP also suggests that higher poverty rates, based on income measures, result
when the reference period is shorter as well (Short, 1990).
V. Conclusions
The thresholds presented in this study, unlike the official thresholds, reflect recent spending needs in levels
and patterns. Unlike the official measure, they account for changes in living standards over time. A focus on
meeting spending needs versus consumption needs dictated the construction of the poverty measure presented here.
The thresholds produced follow the same procedure over the 2001-2003 period. These, along with the thresholds for
1993-2000 presented in Garner (2005), represent the most recent series of thresholds available that uses the same
method over time.
The results presented reveal thresholds that have been increasing at a faster rate than official poverty
thresholds. The experimental threshold has been increasing at a faster rate than the official threshold since 2001. It
was shown that changes in implicit expenditures in the experimental threshold highlight differences between the
official poverty measure and the experimental threshold, and provide insight regarding why the official and
experimental thresholds move differently. The largest share of expenditures in the CPI (about 40 percent) is
represented by goods and services not represented by the basic needs in the thresholds: food, clothing, shelter,
utilities, and medical care. This means that the official poverty threshold is influenced more by goods and services
other than food, clothing, shelter, utilities, or medical care. It was demonstrated that implicit expenditures for
shelter, utilities, and medical care in the thresholds increased much more than the CPI for these items over this
period of time.
Included in the discussion of a resource measure for poverty statistics was an emphasis on consistency with
the selected threshold. It is useful to calculate a resource measure that represents the ability of the family or
27
consumer unit to meet the needs as measured in the particular threshold selected. In this paper experimental poverty
thresholds were compared to two measures of resources that represented our best effort at conceptual consistency,
one reflecting “inflows” into the household and the other “outflows.” These experimental poverty measures showed
trends and levels of poverty that differed from the official measure for the overall population and for some specific
subgroups.
Many believe that one’s command over economic resources should be measured in terms of household
income (Atkinson et al., 2002). However, research using expenditures, including that presented in this study, can
“increase our understanding of how people actually choose, or are constrained, to allocate their economic
resources.” 28
28
The quote is from Bavier (2006) who supports research using expenditure data for poverty measurement. We
agree with his statement.
28
REFERENCES
Atkinson, Tony, Bea Cantillon, Eric Marlier, Brian Nolan, Social Indicators: The EU and Social Inclusion, Oxford,
Oxford University Press, 2002.
Atkinson, A. B. and J.E. Stiglitz, Lectures in Public Economics, McGraw-Hill, Singapore, 1980.
Banthin, Jessica, “Where Do We Stand in Measuring Medical Care Needs for Poverty Definitions?” Available at
http://www7.nationalacademies.org/cnstat/Workshop_on_Experimental_Poverty_Measures.html, August
2004.
Banthin, Jessica, Thesia I. Garner, and Kathleen Short, “Medical Care Needs in Poverty Thresholds: Problems
Posed by the Uninsured.” Agency for Healthcare Research and Quality, Presented at the American
Economic Association Meetings, January 2001.
Bavier, Richard, “Do the Current Poverty Thresholds Include Any Amount for Health Care?” Poverty Measurement
Working Paper, Census Bureau, March 2001.
Bavier, Richard, “Income and Expenditure Data in Poverty Measurement,” unpublished manuscript from the Office
of Management and Budget, April 7, 2006.
Bavier, Richard, “Medical Out-of-Pocket Spending in Poverty Thresholds,” Poverty Measurement Working Paper,
Census Bureau, October 2000.
Bavier, Richard, “Misclassification in An Experimental Poverty Measure,” unpublished manuscript, Office of
Management and Budget, April 21, 2005.
Bernstein, Jared, Chauna Brocht and Maggie Spade-Aguilar, 2000. How Much Is Enough? Basic Family Budgets
for Working Families. Economic Policy Institute, Washington, D.C.
Betson, David M., “Is Everything Relative? The Role of Equivalence Scales in Poverty Measurement,” Working
Paper, Census Bureau, December 1996.
Betson, David M., “Response to Bavier’s Critique of the NRC Panel’s Recommendations”, Poverty Measurements
Working Paper, Census Bureau, November 2000.
Bureau of Labor Statistics (BLS), 2003 Consumer Expenditure Interview Survey Public Use Microdata
Documentation, Washington, D.C., available at: www.bls.gov/cex/2003/cex/cexintvw.pdf, February 24,
2004.
Bureau of Labor Statistics (BLS), Consumer Expenditure Interview Survey Internal Data for 2001Q2-2004Q1,
Washington, D.C., 2005
Bureau of Labor Statistics (BLS), data from 2001, 2002, 2003 available at http://stats.bls.gov/cex/home.htm, 2006.
Cage, Rob, Bureau of Labor Statistics, U.S. Department of Labor, personal communications regarding the CPI and
expenditure weights, 2006.
Canberra Group, Expert Group on Household Income Statistics, Final Report and Recommendations, Ottawa, 2001.
Census Bureau (Census), data for 2001, 2002, 2003 available at
http://www.census.gov/hhes/www/income/income.html, 2006.
Charles, Kerwin Kofi, Sheldon Danziger, Laurie Pounder, and Robert F. Schoeni, “Consumption, Income, and WellBeing Among the Mature Population,” National Poverty Center paper, available at
29
http://www.npc.umich.edu/news/events/consumption06_agenda/pounder-charles-schoeni-danziger.pdf,
April 2006.
Citro, Constance F., National Academy of Sciences, personal communication regarding the use of shelter
consumption as opposed to shelter expenditures in deriving the thresholds, 1999.
Citro, Constance F., and Robert T. Michael (eds.), Measuring Poverty: A New Approach, Washington, D.C.:
National Academy Press, 1995.
Cutler, David and Lawrence Katz, “Macroeconomic Performance and the Disadvantaged,” Brookings Papers on
Economic Activity, Vol. 1991, No. 2, 1991.
Fisher, Jonathan, David S. Johnson, Joseph Marchand, Timothy Smeeding, and Barbara B. Torrey, “The Retirement
Consumption Puzzle: Evidence from a Consumption Survey,” presentation at the National Poverty Center's
"Consumption, Income, and the Well-Being of Families and Children" Conference, May 2006.
Garner, Thesia I., “Developing Poverty Thresholds 1993-2003,” 2005 Proceedings of the American Statistical
Association, Social Statistics Section [CD-ROM], Alexandria, VA: American Statistical Association, 2005.
Garner, Thesia I. and Patricia Rozaklis, “Accounting for Owner Occupied Housing in Poverty: Focus on
Thresholds,” American Statistical Association (ASA) Proceedings of the Section on Government Statistics
and Section on Social Statistic, 1999.
Garner, Thesia I. and Patricia Rozaklis, “Owner-Occupied Housing: An Input for Experimental Poverty
Thresholds,” paper presented at session organized by the Society of Government Economists at the annual
meeting of the Allied Social Sciences Associations, January 6, 2001.
Garner, Thesia I., Stephanie Shipp, Geoffrey Paulin, Kathleen S. Short, and Charles Nelson, “Poverty Measurement
in the 1990s,” Monthly Labor Review, March 1998, pp. 39-61.
Garner, Thesia I. and Kathleen S. Short, “Owner-Occupied Shelter in Experimental Poverty Measures,” paper
presented at the Annual Meeting of the Southern Economic Association, Tampa, Florida, November 1,
2001.
Garner, Thesia I. and Kathleen S. Short, “Economic Well-Being Based on Income, Consumer Expenditures and
Personal Assessments of Minimum Needs” in John A. Bishop and Yoram Amiel, eds., Studies on
Economic Well-being: Essays in the Honor of John P. Formby , Vol. 12 of the Series Research on
Economic Inequality, Oxford, UK: Elsevier Science, February 2004, pp. 319-361.
Garner, Thesia I. and Kathleen S. Short, “Personal Assessments of Minimum Income and Expenses: What Do They
Tell Us about ‘Minimum Living’ Thresholds and Equivalence Scales?” in John A. Bishop and Yoram
Amiel, eds., Inequality, Welfare and Poverty: Theory and Measurement, Vol. 9 of the Series Research on
Economic Inequality, Oxford, UK: Elsevier Science, 2003, pp.191-243.
Greenlees, John S. and Charles C. Mason, “Overview of the 1998 Revision of the Consumer Price Index,” Monthly
Labor Review, December 1996, pp. 3-9.
Iceland, John. "The CNSTAT Workshop on Experimental Poverty Measures, June 2004." Focus 23(3), 2005, pp.
26-30.
International Labour Organization (ILO), Report II: Household Income and Expenditure Statistics, Seventeenth
International Conference of Labour Statisticians, Geneva, 24 November-3 December 2003.
Johnson, David, Stephanie Shipp, and Thesia I. Garner, “Developing Poverty Thresholds Using Expenditure Data,”
in Proceedings of the Government and Social Statistics Section. Alexandria, VA: American Statistical
Association, August 1997, pp. 28-37.
30
Meyer, Bruce D. and James X. Sullivan, “Measuring the Well-Being of the Poor Using Income and Consumption,”
Journal of Human Resources, 38:S, pp. 1180-1220, 2003.
Morissette, R. and Susan Poulin, Income Satisfaction Supplement: Summary of Four Survey Years, Staff Reports,
Labour and Household Surveys Analysis Division, Statistics Canada, Ottawa, Canada, 1991.
National Research Council. Workshop on Experimental Poverty Measures: Summary of a Workshop. Washington,
D.C.: National Academy Press, 2005.
Office of Management and Budget (OMB), Statistical Policy Directive No. 14, Definition of Poverty for Statistical
Purposes, May 1978, http://www.census.gov/hhes/www/povmeas/ombdir14.html.
O'Hara, Amy, “New Methods for Simulating Taxes,” Census Bureau Working Paper, available at
http://www.census.gov/hhes/income/papers.html, 2004.
Orshansky, Mollie, “Counting the Poor: Another Look at the Poverty Profile,” Reprint from the Social Security
Bulletin, U.S. Department of Health, Education, and Welfare, Social Security Administration, January
1965.
Pearce, Diana , The Real Cost of Living in 2005: The Self-Sufficiency Standard for Connecticut, prepared for the
Wider Opportunities for Women, and Office of Workforce Competitiveness, State of Connecticut, available
at http://www.sixstrategies.org/states/states.cfm, 2005a.
Pearce, Diana, The Self-Sufficiency Standard for the Washington, D.C. Metropolitan Area 2005, prepared for the
Wider Opportunities for Women, in conjunction with Wider Opportunities for Women’s Family Economic
Self-Sufficiency Project, available at http://www.sixstrategies.org/states/states.cfm, 2005b.
Renwick, T. J., 1998. Basic Needs Budgets Revisited: Does the U.S. Consumer Price Index Overestimate the
Changes in the Cost of Living for Low-Income Families? Feminist Economics 4(3), pp. 129-142.
Renwick, T. J. and Bergmann, B.R., 1993. A Budget-Based Definition of Poverty with an Application to SingleParent Families, Journal of Human Resources 28(1), pp. 1-24.
Rogers, John M. And Maureen B. Gray, “CE Data: Quintiles of Income vs. Quintiles of Outlays,” Monthly Labor
Review, December 1994, pp. 32-37.
Ruggles, Patricia, Drawing the Line: Alternative Poverty Measures and Their Implications for Public Policy,
Washington, D.C.: The Urban Institute Press, 1990.
Short, Kathleen, “Estimating Resources for Poverty Measurement: 1993-2003,” 2005 Proceedings of the American
Statistical Association, Social Statistics Section [CD-ROM], Alexandria, VA: American Statistical
Association, 2005.
Short, Kathleen, Experimental Poverty Measures: 1999, U.S. Census Bureau, Current Population Reports,
Consumer Income, P60-216, U.S. Government Printing Office, Washington, DC, 2001.
Short, Kathleen, “The Relationship Between Monthly and Annual Income: SIPP,” paper presented at the Annual
Meeting of the American Statistical Association, 1990.
Short, Kathleen, “Work-Related Expenditures in a Measure of Poverty,” available at
http://www7.nationalacadmeies.org/cnstat/Workshop_on_Experimental_Poverty_Measures.html, August
2004.
Short, Kathleen, and Thesia I. Garner, “A Decade of Experimental Poverty Thresholds: 1990 to 2000,” The ICFAI
Journal of Public Finance, II (2), May 2004, pp. 65-89.
31
Short, Kathleen and Thesia I. Garner, “Experimental Poverty Measures Under Alternate Treatments of Medical Outof-Pocket Expenditures,” Monthly Labor Review, August 2002, pp. 3-13.
Short, Kathleen, Thesia Garner, David Johnson, and Patricia Doyle, Experimental Poverty Measures: 1990 to 1997,
U.S. Census Bureau, Current Population Reports, Consumer Income, P60-205, U.S. Government Printing
Office, Washington, DC, 1999.
Short, Kathleen and Patricia Ruggles, “Experimental Measures of Poverty and Net Worth: 1996,” Journal of Income
Distribution, forthcoming.
Slesnick, Daniel T., “Gaining Ground: Poverty in the Postwar United States,” Journal of Political Economy,
1901(1), pp.1-38, 1993.
U.S. Department of Labor (USDL), Bureau of Labor Statistics, BLS Handbook of Methods, Bulletin 2490, April
1997.
U.S. Department of Agriculture (USDA), Official USDA Food Plans: Cost of Food at Home at Four Levels, U.S.
Average, June 2003, available at www.cnpp.usda.gov, issued July 2003.
Wolff, Edward, “Wealth, Holdings, and Poor Status in the U.S.,” Review of Income and Wealth, 36(2), pp. 143-165,
1990.
(Many of the unpublished working papers listed here are available at:
http://www.census.gov/hhes/www/povmeas/papers.html
32
Table 1: Official and Experimental Poverty Thresholds for the Reference Family: 2001-2003
Year
Official
Experimental
$17,960
$21,177
2001
18,244
22,036
2002
18,660
23,170
2003
Source: Census Bureau and authors’ own calculations using U.S. Consumer Expenditure Interview Survey
data from 2000Q2 – 2003Q1.
Table 2: Average Annual CPS Household Income, CE Total Expenditures, and Thresholds
Year
2001
2002
2003
Published Before Tax
Money Income
$58,208
57,852
59,067
Published CE Total
Expenditures
$39,518
40,677
40,817
Reference Family Experimental
Threshold
$21,177
22,036
23,170
Source: Census Bureau and Bureau of Labor Statistics web sites, and authors' own calculations using U.S.
Consumer Expenditure Interview Survey data from 2000Q2 – 2003Q1.
Table 3: Shares of Aggregated Annual Expenditures Implicit in the
Experimental Threshold for the Reference Family
Year
2001
2002
2003
Food
0.280
0.276
0.267
Clothing
0.062
0.063
0.061
Shelter
0.301
0.297
0.305
Utilities
0.134
0.136
0.137
Medical
0.067
0.072
0.075
Other
0.157
0.156
0.155
Source: Authors' own calculations using U.S. Consumer Expenditure Interview
Survey data from 2000Q2 – 2003Q1.
33
Table 4a: Shares of Aggregated Annual Expenditures Based on the 1998
CPI Market Basket Structure, U.S. City Average, CPI-U Population
Year
Food1
Clothing2
Shelter3
Utilities4
Medical
Other
0.143
0.044
0.324
0.049
0.056
0.384
2001
0.145
0.047
0.305
0.043
0.055
0.403
2002
0.145
0.047
0.305
0.043
0.055
0.403
2003
Source: Basic calculations produced by Cage (2006), Bureau of Labor Statistics;
upper level aggregations produced by authors.
1
Does not include alcoholic beverages.
2
Officially referred to as "Apparel".
3
Owners’ equivalent rent for owners shelter costs.
4
Does not include telephone services.
Table 4b: Shares of Aggregated Annual Expenditures Based on Experimental
Threshold Structure, Reference Family, All U.S.
Clothing2
Shelter3
Utilities4
Medical
Year
Food1
0.143
0.047
0.292
0.073
0.056
2001
0.145
0.052
0.277
0.067
0.055
2002
0.145
0.052
0.277
0.067
0.055
2003
Source: Basic calculations produced by Cage (2006), Bureau of Labor Statistics;
upper level aggregations produced by authors.
1
Does not include alcoholic beverages.
2
Officially referred to as "Apparel".
3
Owners’ out-of-pocket expenses for owners shelter costs.
4
Includes telephone services.
Other
0.389
0.403
0.403
34
Table 5: Annual CPI-U, U.S. City Average for All Items and Selected Categories
Year
2001
2002
2003
All Items
1.77
1.80
1.84
Food
1.73
1.76
1.80
Clothing
1.27
1.24
1.21
Shelter
2.01
2.08
2.13
Utilities
1.50
1.44
1.55
Medical
Care
2.73
2.86
2.97
2001 to 2003
3.9
4.0
-5.0
6.2
2.9
percent change
Source: Bureau of Labor Statistics web site and authors’ own calculations.
8.9
Table 6: Implicit Average Annual Expenditures Based on Experimental Threshold
Structure, Reference Family, All U.S. (in threshold year dollars)
Year
2001
2002
2003
Threshold
21,177
22,036
23,170
Food
5,928
6,080
6,176
Clothing
1,315
1,392
1,413
Shelter
6,371
6,535
7,074
Utilities
2,834
3,005
3,169
Medical
Care
1,413
1,590
1,742
Other
3,317
3,433
3,596
9.4
4.2
7.5
11.0
11.8
23.3
8.4
2001 to 2003
percent change
Source: Authors' own calculations using U.S. Consumer Expenditure Interview Survey data from
2000Q2 - 2003Q1.
35
Table 7a. Annual Average Family or Consumer Unit Subtractions Used to Calculate Poverty Rates
Payroll Taxes
Work Expenses
Child Care Expenses
Year
CPS
CE
CPS
CE
CPS
CE
$2,856
$2,483
$993
$1,093
$163
$140
2001
2,990
2,608
1,038
1,146
155
152
2002
3,039
2,725
1,028
1,128
153
151
2003
Source: Authors' own calculations using U.S. Consumer Expenditure Interview Survey data from 2001Q2 - 2004Q1.
Table 7b. Annual Average Consumer Unit Addition and Additional Subtractions Used to Calculate Poverty Rates
Additions
Subtractions
Vehicle
Change in Financial
Property Capital
Child
Finance Other Finance
Year
Alimony Charges
Accounts1
Improvements2
Support
Charges3
$682
$765
$195
$27
$364
$223
2001
835
893
187
22
396
274
2002
874
1,048
165
43
369
183
2003
Source: Authors' own calculations using U.S. Consumer Expenditure Interview Survey data from 2001Q2 - 2004Q1.
1
Positive changes to savings accounts, checking accounts, and U.S. savings bonds.
2
Expenditures for capital improvements to owned home, vacation home, and other property
3
Finance charges excluding those for mortgages and vehicles
Interest on Home
Equity Loans and
Lines of Credit
$164
157
121
36
Table 7c. Reference Family Thresholds and Annual Average Household Income and
Consumer Unit Expenditures
Income
Published Household BTM Income
Income-NAS
Expenditures
Published CU Total Expenditures
Expenditures (CE outlays + additions)
Expenditures-NAS
Expenditures-NAS-Other Legal Obligations
Thresholds
Experimental Threshold
Official Threshold
2001
2002
2003
$58,208
43,282
$57,852
42,513
$59,067
42,637
39,518
39,215
35,499
34,526
40,677
40,623
36,716
35,680
40,817
41,130
37,126
36,246
21,177
17,960
22,036
18,244
23,170
18,660
Source: Published data from Census Bureau and Bureau of Labor Statistics web sites for years
2001-2003; other results based on authors' own calculations using U.S. Consumer Expenditure
Interview Survey data from 2000Q2 - 2004Q1.
37
Table 8: Population Poverty Rates Using Income- or Expenditure-Based Resources: 2001-2003
Official Threshold
Experimental Threshold
Income-Based Resources
Subtractions
Experimental Threshold
Expenditure-Based Resources
Subtractions
NAS + Other Legal
Year
None
NAS1
None
NAS1
Obligations2
11.7
15.5
17.4
22.7
23.9
2001
12.1
16.4
17.8
23.3
24.5
2002
12.5
17.7
19.7
25.6
26.7
2003
Source: Authors’ own calculations using 1994 - 2004 CPS ASEC and U.S. Consumer Expenditure Interview Survey 2001Q2 - 2004Q1 CE data.
1
Subtractions for payroll taxes, work related expenses, and child care expenses associated with working.
2
Child support payments, alimony, vehicle finance charges, interest on home equity loans and lines of
credit.
38
Table 9: Population Poverty Rates Using
Income- and Expenditure-Based Resources
2001
2002
2003
Children
NAS Income
Official
Experimental
Expenditure Outlays
No Subtractions
16.3
19.3
16.7
20.1
17.6
21.7
19.9
20.3
23.0
NAS Subtractions1
26.2
26.8
30.2
27.8
28.3
31.5
10.1
13.4
10.6
14.1
10.8
15.2
14.4
15.1
16.3
19.9
20.7
22.5
21.0
21.8
23.5
10.1
18.4
10.5
20.4
10.3
22.1
27.0
26.3
30.4
29.3
29.1
32.3
30.0
29.8
32.9
9.9
13.5
10.2
14.3
10.5
15.4
14.9
15.1
17.3
20.0
20.2
22.8
21.0
21.2
23.8
22.7
28.0
23.9
29.4
24.3
31.0
31.3
34.0
35.1
2
NAS + Other
Non-elderly Adults
NAS Income
Official
Experimental
Expenditure Outlays
No Subtractions
NAS Subtractions
1
NAS + Other2
Elderly
NAS Income
Official
Experimental
Expenditure Outlays
No Subtractions
NAS Subtractions
1
2
NAS + Other
White
NAS Income
Official
Experimental
Expenditure Outlays
No Subtractions
NAS Subtractions
1
NAS + Other2
Blacks
NAS Income
Official
Experimental
Expenditure Outlays
No Subtractions
1
38.1
41.6
43.2
NAS + Other2
Other races
NAS Income
Official
Experimental
Expenditure Outlays
No Subtractions
40.4
43.7
45.1
12.8
16.9
12.5
16.8
13.6
18.9
20.9
19.2
19.6
NAS Subtractions1
26.1
25.9
26.6
NAS Subtractions
2
26.7
26.5
27.3
NAS + Other
Source: Authors’ own calculations using 1994 - 2004 CPS ASEC and U.S. Consumer Expenditure Interview Survey 2001Q2 - 2004Q1 CE
data.
1
Subtractions for payroll taxes, work related expenses, and child care expenses associated with working.
2
Subtractions for child support payments, alimony, vehicle finance charges, interest on home equity loans and lines of credit.
39
Chart 1. Official and Experiment Poverty Thresholds for the Reference Family: 2001-2003
$25,000
$20,000
$15,000
$10,000
$5,000
$0
2001
2002
Official
2003
Experimental
40
Chart 2. Average Annual NAS Subtractions to Calculate Poverty Rates in Both CPS and CE
$0
-$500
-$1,000
-$1,500
-$2,000
-$2,500
-$3,000
-$3,500
CPS-Payroll Tax
CE-Payroll Tax
CPS-Work Expenses
2001
CE-Work Expenses
2002
CPS-Child Care
CE-Child Care
2003
41
Chart 3. Annualized Annual Mean Additions and Subtractions for Expenditure Resource
Poverty Measure
$1,200
$1,000
$800
$600
$400
$200
$0
-$200
-$400
-$600
Child Support
Alimony
Vehicle Finance
Charges
CC Finance
Charges
2001
2002
Home Equity
Change in
Borrowing Interest Financial Accounts
Property Capital
Improvements
2003
42
Chart 4. Average Annual Income and Expenditures
$60,000
$55,000
$50,000
Published Household BTM Income
Income-NAS
$45,000
Expenditures
$40,000
Published CU Total Expenditures
Expenditures-NAS
$35,000
Expenditures-NAS-Other Legal
Obligations
Experimental Threshold
$30,000
Official Threshold
$25,000
$20,000
$15,000
2001
2002
2003
43
Chart 5. Poverty Rates Using Official and Experimental Measures
30
25
20
15
10
5
0
Official
Experi w/ NAS Income
Experi w/ Expenditures
2001
2002
Experi w/Exp-NAS
Experi w/Exp-NAS-other
2003
44