Family Economics and
Nutrition Review
CENTER FOR NUTRITION POLICY AND PROMOTION
Volume 16, Number 2
2004
Research Articles

3 Fruits and Vegetables Offered in School Lunch Salad Bars Versus

Traditional School Lunches

Stefanie R. Schmidt and Patricia McKinney
12 Explaining Variations in State Hunger Rates

John Tapogna, Allison Suter, Mark Nord, and Michael Leachman


23 The Pitfalls of Using a Child Support Schedule Based on Outdated Data

Yana van der Meulen Rodgers and William M. Rodgers III
Research Brief
41 The Food Environment and Food Insecurity: Perceptions of Rural,

Suburban, and Urban Food Pantry Clients in Iowa

Steven Garasky, Lois Wright Morton, and Kimberly Greder
Center Reports
49
Developing a Measure for the Dietary Guidelines Recommendations to Eat a

Variety of Foods


Andrea Carlson and WenYen Juan

57
The U.S. Food Supply Series: Selected Food and Nutrient Highlights, 1909 to 2000


Shirley Gerrior, Lisa Bente, and Hazel A.B. Hiza
66
Nutrition Insight 28: Report Card on the Quality of Americans’ Diets


P. P. Basiotis, A. Carlson, S.A. Gerrior, W.Y. Juan, and M. Lino


69 Nutrition Insight 29: Quality of Diets of Older Americans

W.Y. Juan, M. Lino, and P. P. Basiotis
Regular Items
Federal Studies
Journal Abstracts
Food Plans
Consumer Prices
Poverty Thresholds


Ann M. Veneman, Secretary
U.S. Department of Agriculture
Eric M. Bost, Under Secretary
Food, Nutrition, and Consumer Services
Eric J. Hentges, Executive Director
Center for Nutrition Policy and Promotion
P. Peter Basiotis, Director
Nutrition Policy and Analysis Staff
The U.S. Department of Agriculture (USDA) prohibits discrimination in all its programs and
activities on the basis of race, color, national origin, age, disability, and where applicable, sex,
marital status, familial status, parental status, religion, sexual orientation, genetic information,
political beliefs, reprisal, or because all or part of an individual’s income is derived from any
public assistance program. (Not all prohibited bases apply to all programs.) Persons with
disabilities who require alternative means for communication of program information (Braille,
large print, audiotape, etc.) should contact USDA’s TARGET Center at (202) 720-2600 (voice
and TDD).
To file a complaint of discrimination, write to USDA, Director, Office of Civil Rights, 1400
Independence Avenue, S.W., Washington, D.C. 20250-9410 or call (800) 795-3272 (voice) or
(202) 720-6382 (TDD). USDA is an equal opportunity provider and employer.
Center for Nutrition Policy and Promotion
Mission Statement
To improve the health of Americans by developing and promoting dietary
guidance that links scientific research to the nutrition needs of consumers.
Family Economics and
Nutrition Review
Research Articles
3
Fruits and Vegetables Offered in School Lunch Salad Bars Versus Traditional
School Lunches
Stefanie R. Schmidt and Patricia McKinney
12 Explaining Variations in State Hunger Rates
John Tapogna, Allison Suter, Mark Nord, and Michael Leachman
23
The Pitfalls of Using a Child Support Schedule Based on Outdated Data
Yana van der Meulen Rodgers and William M. Rodgers III
Research Brief
41 The Food Environment and Food Insecurity: Perceptions of Rural,
Suburban, and Urban Food Pantry Clients in Iowa
Steven Garasky, Lois Wright Morton, and Kimberly Greder
Center Reports
49 Developing a Measure for the Dietary Guidelines Recommendations to Eat a Variety of Foods
Andrea Carlson and WenYen Juan
57
The U.S. Food Supply Series: Selected Food and Nutrient Highlights, 1909 to 2000
Shirley Gerrior, Lisa Bente, and Hazel A.B. Hiza
66 Nutrition Insight 28: Report Card on the Quality of Americans’ Diets
P.P. Basiotis, A. Carlson, S.A. Gerrior, W.Y. Juan, and M. Lino
69 Nutrition Insight 29: Quality of Diets of Older Americans
W.Y. Juan, M. Lino, and P.P. Basiotis
Regular Items
72
Federal Studies
80
Journal Abstracts
82
Price Changes in the Thrifty Food Plan Versus the Consumer Price Index for Food:
Why the Difference?
84 Official USDA Food Plans: Cost of Food at Home at Four Levels, U.S. Average, December 2004
85 Consumer Prices
86 U.S. Poverty Thresholds and Related Statistics
87 Reviewers of Manuscripts for the 2004 Issues
Volume 16, Number 2
2004
Editor
Julia M. Dinkins
Associate Editor
David M. Herring
Associate Editor
Mark Lino
Managing Editor
Jane W. Fleming
Peer Review Coordinator
Hazel Hiza
Family Economics and Nutrition Review is
published semiannually by the Center for Nutrition
Policy and Promotion, U.S. Department of
Agriculture, Washington, DC.
The Secretary of Agriculture has determined that
publication of this periodical is necessary in the
transaction of the public business required by law
of the Department.
This publication is not copyrighted. Thus, contents
may be reprinted without permission, but credit to
Family Economics and Nutrition Review would be
appreciated. Use of commercial or trade names
does not imply approval or constitute endorsement
by USDA. Family Economics and Nutrition Review
is indexed in the following databases: AGRICOLA,
Ageline, Economic Literature Index, ERIC, Family
Studies, PAIS, and Sociological Abstracts.
Suggestions or comments concerning this
publication should be addressed to Julia M.
Dinkins, Editor, Family Economics and Nutrition
Review, Center for Nutrition Policy and Promotion,
USDA, 3101 Park Center Drive, Room 1034,
Alexandria, VA 22302-1594.
Family Economics and Nutrition Review
is available at
www.cnpp.usda.gov.
CENTER FOR NUTRITION POLICY AND PROMOTION
Front and Center
Serving the American People: From 1943 to 2005
This issue of Family Economics and Nutrition Review contains three research articles and briefs that, respectively,
examine variations in State hunger rates; focus on fruits and vegetables offered in school lunch salad bars; and
describe the perceptions of rural, suburban, and urban residents who use food pantries.
The issue also includes reports by the Center for Nutrition Policy and Promotion: one describing the nutrient content of
the U.S. food supply and the other detailing how variety—one of the 10 components of the Healthy Eating Index—was
calculated. The nutrient content of the food supply provides information on nutrient availability and is often used in setting
fortification policy. The Healthy Eating Index, representing a report card on the American diet, gives policymakers a picture
of the overall status of the American diet and where changes need to be made. In addition to these reports, the Center for
Nutrition Policy and Promotion uses a brief article to explain why cost updates of the Thrifty Food Plan, the basis for food
stamp allotments, differ from price changes as measured by the Consumer Price Index for food.
Although the name of this USDA publication has changed over the years (Wartime Family Living in 1943, Rural Family
Living in 1945, Family Economics Review in 1957, and Family Economics and Nutrition Review in 1995), its goal of
reaching American consumers with current, science-based information has remained constant. The USDA agencies or
divisions that had the privilege of producing this publication met a perennial need of linking research to the needs of
consumers. These USDA agencies or divisions were the Bureau of Human Nutrition and Home Economics, Home
Economics Research Branch, Institute of Home Economics, Consumer and Food Economics Research Division, Consumer
and Food Economics Institute, and Family Economics Research Group. The agencies’ or divisions’ contributions formed
the foundations upon which actionable consumer strategies were based. Similarly, we believe that the Center for Nutrition
Policy and Promotion, with its Family Economics and Nutrition Review, has added to that substantial tradition and has thus
improved the well-being of all Americans.
As Americans began using more electronic means of communications, the Center for Nutrition Policy and Promotion
decided to use a variety of other information-multiplying strategies that could meet the demands of consumers who are
obtaining information at the “click of the mouse.” With this final issue of Family Economics and Nutrition Review, the
Center for Nutrition Policy and Promotion concludes the chapter on this paper form of providing information to the
economic and nutrition professional communities. We invite the readers of Family Economics and Nutrition Review to use
our Web site (www.cnpp.usda.gov) to learn more about our other publications and links that provide nutrition and economic
information that can be used to help Americans develop and maintain a healthful lifestyle.
Eric J. Hentges, PhD
Executive Director
Center for Nutrition Policy and Promotion
2004 Vol. 16 No. 2
3
Fruits and Vegetables Offered
in School Lunch Salad Bars
Versus Traditional School Lunches
Most U.S. school-age children do not eat enough fruits and vegetables, both in terms of
the number of servings and variety. One proposed way to improve children’s consumption
of fruits and vegetables is to increase the number of schools that offer salad bars as part
of the National School Lunch Program. This study presented the first analysis of nationally
representative data on foods offered in school lunch salad bars. The data were collected
during the 1998-99 school year as part of USDA’s School Nutrition Dietary Assessment
Study-II. The study presented here examined whether schools with salad bars offered a
greater variety of fruits and vegetables than did schools without salad bars. The study also
examined items other than fruits and vegetables that were commonly offered in school
lunch salad bars, with a focus on dietary fat content. Results showed that salad bars were
associated with a greater variety of fruit and vegetable offerings. Schools with salad bars
were much more likely to serve lettuce, tomatoes, other raw vegetables, and fresh fruit
than were schools without salad bars. In addition, schools with salad bars were more
likely than their counterparts, to offer nutrient-dense vegetables (like carrots and broccoli).
Stefanie R. Schmidt, PhD
Institute of Education Sciences
U.S. Department of Education
Patricia McKinney, MS, RD
Food and Nutrition Service
U.S. Department of Agriculture
chool-age children in the United
States eat fewer fruits and vege-
tables than are recommended by
the Dietary Guidelines for Americans
(U.S. Department of Agriculture
[USDA] and U.S. Department of
Health and Human Services [HHS],
2000). In 1994-96, only 14 percent
of school-age children met the target
of consuming at least two servings of
fruits a day; only 17 percent met the
target of consuming at least three
servings of vegetables a day (Gleason
& Suitor, 2000). Even fewer met the
recommended standards for consuming
a variety of fruits and vegetables.
The Dietary Guidelines for Americans
recommends that all people ages 2 and
older choose a wide variety of fruits
and vegetables each day because
different fruits and vegetables are rich
in different nutrients. One target for
variety, which is used in the Federal
Healthy People 2010 objectives, is an
increase in the percentage of children
who consume one-third of their
vegetable servings from dark-green or
orange vegetables. In 1994-96, only
6 percent of 6- to 19-year-old females
and about 5 percent of 6- to 19-year-
old males met that goal (HHS, 2001).
One proposed way to improve chil-
dren’s consumption of fruits and vege-
tables is to increase the number of
schools that offer salad bars as part of
the National School Lunch Program. A
group of policy officials, the National
5-A-Day Partnership, has proposed
that all schools have salad bars as a
way to increase the number and variety
of fruits and vegetables that children
consume at school (U.S. General
Accounting Office [GAO], 2002).
Our study expanded upon a previous
USDA study (Schmidt, Hirschman, &
McKinney, 2002) on salad bars that
examined whether salad bars were
associated with a greater variety of
fruits and vegetables being offered in
school lunches. It was the first analysis
of nationally representative data on
S
Research Articles
4
Family Economics and Nutrition Review
foods offered in school lunch salad
bars.
In the interest of presenting a balanced
view of salad bars, this study also
described items other than fruits and
vegetables in salad bars to provide
a sense of how often high-fat salad
bar ingredients (including regular
salad dressing, regular cheese, and
mayonnaise-based salads) are offered.
Any policy discussion of school lunch
salad bars should consider whether
these ingredients also could contribute
to an increase in children’s total dietary
fat intake because school-age children
consume too much dietary fat. In 1994-
96, only 25 percent of school-age
children met the Dietary Guidelines
for Americans’ recommendation of
consuming no more than 30 percent
of calories from fat (Gleason & Suitor,
2000).
Previous Research
Previous research on the foods offered
in salad bars has been limited. One
study (Garceau et al., 1997) examined
directly the nutrient content of food
bars, including salad bars, in 96
elementary schools that participated
in an intervention designed to reduce
the total fat, saturated fat, and sodium
content of school lunches and break-
fasts. It found that side salad bars had
more total fat than was found in the
regular fruit and vegetable components
of traditional school lunches. It also
found that, compared with the vege-
tables and fruits served in the regular
serving line, side salad bars had similar
amounts of saturated fat, vitamin A,
iron, and dietary fiber but less calcium
and ascorbic acid. One study limita-
tion, however, was that the nutrient
analysis was based on assumptions
about foods selected from salad bars
because data on foods selected were
not available. In particular, the results
were sensitive to assumptions about
how much salad dressing children
placed on salads. The report did not
examine which foods were offered, so
it did not investigate the issue of fruit
and vegetable variety.
Methods
This analysis used data from the
School Nutrition Dietary Assessment
Study-II (SNDA-II), which was
designed to produce national cross-
sectional estimates of the nutrient
composition of USDA meals served in
elementary and secondary schools. The
data were collected in late September
1998 to May 1999. The study focused
exclusively on public schools, which
account for roughly 90 percent of all
participants in the National School
Lunch Program. The study design
included separate nationally repre-
sentative probability samples of public
elementary schools, middle schools,
and high schools participating in the
National School Lunch Program (Fox,
Crepinsek, Connor, & Battaglia, 2001).
Alaskan and Hawaiian schools were
not included in the study.
The sample of schools was developed
in several steps. First, a stratified
random sample of School Food
Authorities,1 which are typically
school districts, was selected. To the
extent possible, one elementary school,
one middle school, and one high school
were chosen from each School Food
Authority. Finally, the schools in the
sample were recruited, and 80 percent
of the schools agreed to participate in
the study.
1School Food Authorities are the governing
bodies responsible for the administration of
one or more schools and have the legal right
to operate a National School Lunch Program.
Definitions
Salad bar is a self-serve station where students can select two or more fruits
and/or vegetables.
Green salad bars are those in which lettuce is intended to serve as the base of
the salad.
Entrée salad bars are green salad bars that include a meat or meat alternate.
Side salad bars are green salad bars that do not include a meat or meat
alternate.
Theme salad bars include potato bars, taco salad bars, soup and salad bars,
salad and sandwich bars, and potato and salad bars.
“Other” self-serve bars include theme salad bars, fruit bars, and assorted raw
vegetable bars.
A serving day for a school is a day on which the school cafeteria serves
National School Lunch Program meals. The terms “serving day” and “daily
menu” are used interchangeably in this paper.
High-fat items are foods that have more than 38 percent of their calories from
fat.
Low-fat items are foods that have no more than 30 percent of calories from fat.
2004 Vol. 16 No. 2
5
The data analyzed in this study came
from a survey of school cafeteria
managers, which was collected via
mail. Among the schools that agreed
to participate in the study, the response
rate for the menu survey was 88
percent (Fox et al., 2001). A total of
435 elementary schools, 390 middle
schools, and 407 high schools com-
pleted the survey. Cafeteria managers
were asked to provide detailed infor-
mation about all foods served as part
of the National School Lunch Program
during a 5-day period, as well as to
provide a description of each item.
For the 258 schools with salad bars,
respondents were asked to list all
ingredients, including salad dressings
and toppings. SNDA-II did not collect
data on the amount and types of food
that children consumed.
The statistical techniques used in this
study were relatively straightforward.
The weighted averages and percentages
were calculated by using sampling
weights that adjusted for nonresponse.
The standard errors were adjusted to
account for the geographic clustering
of schools,2 and a 5-percent level of
significance was used for statistical
significance.
Results and Discussion
Availability of Salad Bars
Sixteen percent of public schools
(n =1,042 in fiscal year 1999) partici-
pating in the National School Lunch
Program offered salad bars daily;
21 percent offered salad bars at least
once a week (table 1). School lunch
2The SAS macro program, smsub.sas, was used
to calculate the correct standard errors. This
program is available at www.SAS.com.
salad bars were more widely available
for children in the higher grades: 41
percent of high schools, compared
with 26 percent of middle schools
and 14 percent of elementary schools
offering some type of salad bar at least
once a week. The differences among
the three grade levels were statistically
significant.
Green salad bars, including entrée
salad bars and side salad bars, were
the most common forms of salad bars
offered by National School Lunch
Program schools. Entrée salad bars
were present at least once per week
in 12 percent of all schools, and side
salad bars were offered at least once
per week in 9 percent of all schools.
Entrée salad bars can be used instead
of traditional entrées because these
types of salad bars include a meat or
meat alternate. The foods in side salad
Table 1. Percentage of public schools1 offering different types of salad bars as part of the National School Lunch Program
Variables
Elementary schools
Middle schools
High schools
All schools
Sample size (number of schools)
385
329
328
1,042
Percent
All types of salad bars
Salad bar of any type daily
10*+
20*
32
16
Any type of salad bar at least once per week
14*+
26*
41
21
Green salad bars
Entrée salad bar daily
4*+
12*
22
9
Entrée salad bar at least once per week
6*+
18*
31
12
Side salad bar daily
6
8
7
7
Side salad bar at least once per week
8
10
10
9
Other salad bars
Theme salad bar (potato bar or combination salad/sandwich,
salad/soup or salad/potato bar) daily
0.3
0
1
0.4
Theme salad bar at least once per week
2
1*
3
2
Self-serve fruit bar daily
2
1
1
2
Self-serve fruit bar at least once per week
2
1
3
2
Self-serve assorted raw vegetables daily
1
1
1
1
Self-serve assorted raw vegetables at least once per week
1
1
1
1
1 Based on 5-day menu data from SNDA-II.
* Difference, when compared with high schools, is statistically significant at the .05 level.
+ Difference, when compared with middle schools, is statistically significant at the .05 level.
6
Family Economics and Nutrition Review
bars count only as fruit or vegetable
components of a meal.3
Other types of self-serve bars were
offered less frequently in the schools
offering the National School Lunch
Program. Two percent of all the
schools offered theme salad bars at
least once a week, 2 percent offered
self-serve fruit bars, and 1 percent
offered self-serve raw vegetables at
least once a week. Theme bars count
as entrées; whereas, fruit bars and
assorted self-serve raw vegetables
count as the fruit or vegetable com-
ponent of the meal. For the remainder
of this paper, schools with salad bars
are defined as those that offer any type
of salad bar at least once per week.
The Variety of Fruits and
Vegetables Offered by
Schools With Salad Bars
and Without Salad Bars
On average, the typical high school
salad bar offered a variety of vege-
tables (6.3) and fruits (1.7) (fig. 1).
In particular, high school salad bars
included a wide variety of raw vege-
tables (3.9 on average) other than
lettuce or tomato. The results for
middle schools were similar. Elemen-
tary schools offered significantly fewer
vegetables on their salad bars than
did middle or high schools, with an
average of 4.8 vegetables and 3.1
raw vegetables other than lettuce
and tomatoes.
The remainder of the paper focuses
on findings for high school salad bars
3To count as a reimbursable traditional meal
of the National School Lunch Program, a lunch
must include a meat or meat alternate, grain or
bread, a fruit or vegetable, and milk. However,
students in high schools and some middle and
elementary schools may choose three of the five
food items under the Offer versus Serve option.
because they are the most common.4
With a few exceptions, the results for
middle schools and elementary schools
are qualitatively similar to those for
high schools.
Categories of Vegetables and
Fruits Offered by High Schools
High schools with salad bars offered a
greater variety of vegetables and fruits
than did schools without salad bars.
The analysis focused on fruits and
vegetables served both in the salad bar
and in the traditional serving line; in
schools with salad bars, the analysis
4Statistics comparing schools at all Grade levels
with and without salad bars can be misleading.
Elementary schools comprise a disproportionate
share of schools without salad bars, and high
schools comprise a disproportionate share of
schools with salad bars. Therefore, differences
in food offerings among schools at all Grade
levels with and without salad bars are partly
driven by the fact that high schools tend to offer
different types of fruits and vegetables than do
elementary schools, regardless of whether the
schools have salad bars.
focused on both serving days with and
without salad bars on the menu. The
most striking results were for lettuce,
raw tomato, and other raw vegetables,
which were offered on 91, 73, and 87
percent of serving days, respectively,
in high schools with salad bars (table
2). In schools without salad bars,
lettuce, raw tomato, and other raw
vegetables were significantly less
common, being offered on 49, 13,
and 15 percent of serving days,
respectively. (The results for lettuce
and raw tomato are shown because
traditional serving lines frequently
offer lettuce and raw tomatoes in green
salads or as sandwich toppings.5)
5High schools without salad bars offered chef’s
salads or green side salads more frequently
than did schools with salad bars. Chef’s salads,
which count as an entrée because they include
meat or meat alternates, were served on 8
percent of serving days in schools with salad
bars and 21 percent of serving days in schools
without salad bars. Green side salads were
offered in schools with salad bars on 18 percent
of serving days and 29 percent of serving days
in schools without salad bars.
Figure 1. Mean number of fruits and vegetables offered in salad bars,
by Grade level
*Difference, when compared with high schools, is statistically significant at the .05 level.
+Difference, when compared with middle schools, is statistically significant at the .05 level.









!LL□SCHOOLS
(IGH□SCHOOLS
-IDDLE□SCHOOLS
%LEMENTARY□SCHOOLS
6EGETABLES
2AW□VEGETABLES□□EXCLUDING□LETTUCE□AND□TOMATO
&RUITS














2004 Vol. 16 No. 2
7
In addition, cooked vegetables,
legumes, and non-green vegetable
salads were significantly more
common in high schools with salad
bars than in high schools without
salad bars.
High schools with salad bars also
offered a significantly greater variety
of fruits than did high schools without
salad bars. On 74 and 70 percent of
serving days, high schools with salad
bars offered canned and fresh fruit,
respectively, compared with 53 and 50
percent of serving days, respectively,
in high schools without salad bars.
Dried fruit was also more common in
high schools with salad bars than in
high schools without salad bars: 7
percent versus 1 percent of serving
days.
Students in schools with salad bars
need to select foods from the salad bar
to take advantage of the wider variety
of fruit and vegetable offerings in their
school cafeterias, because schools with
salad bars do not serve a greater variety
of fruits and vegetables in their regular
serving lines. All of the statistically
significant differences in fruit and
vegetable category offerings among
schools with and without salad bars
are due to the greater prevalence of
fruits and vegetables in salad bars.6
Individual Nutrient-Dense
Vegetables
Certain nutrient-dense vegetables were
much more common in salad bars than
in traditional serving lines (fig. 2), and
these differences were statistically
significant. Carrots, rich in vitamin A,
were offered in either raw or cooked
form on 70 percent of serving days in
high schools with salad bars. Broccoli,
which is rich in calcium and vitamin C,
6Tables that illustrate this finding are available
upon request from the primary author.
Table 2. Percentage of daily menu items either in salad bar or regular serving line
of public schools offering the National School Lunch Program
High schools
All Grade levels
Categories of
With
Without
With
Without
fruits and vegetables served
salad bars
salad bars
salad bars
salad bars
Sample size (number of schools)
118
210
258
784
Percent
Vegetables
Lettuce
91*
49
89*
35
Tomato, raw
73*
13
64*
7
Raw vegetables,
excluding lettuce and tomato
87*
15
84*
16
Cooked vegetables
61*
45
49
44
Legumes
18*
9
13*
7
Other (non-green) salads
30*
8
19*
7
Fruits
Canned
74*
53
73*
56
Fresh
70*
50
69*
42
Dried
7*
1
12*
1
Frozen
6
4
8
7
Notes: Green salads or salad bars with multiple vegetables are categorized in multiple rows.
Based on 5-day menu data from SNDA-II.
*Difference in those schools with and without salad bars is statistically significant at the .05 level.
Students in schools with salad
bars need to select foods from
the salad bar to take advantage
of the wider variety of fruit and
vegetable offerings in their
school cafeterias, because
schools with salad bars do not
serve a greater variety of fruits
and vegetables in their regular
serving lines.
8
Family Economics and Nutrition Review
was offered in either raw or cooked
form on half of the serving days in high
schools with salad bars. In contrast,
high schools without salad bars served
carrots on 17 percent of serving days;
and broccoli, on 7 percent of serving
days. Carrots and broccoli are the only
orange and dark-green vegetables
commonly served in school lunches.
Other types of orange and dark-green
vegetables, including sweet potatoes,
pumpkin, spinach, and other greens,
were rarely offered in school lunches—
less than 1 percent of daily menus in
schools with and without salad bars.
Similar to broccoli, cauliflower, a
cruciferous vegetable rich in vitamin C,
was offered more widely in high school
lunch salad bars than in traditional
serving lines. Cruciferous vegetables
may play a role in reducing the risk
of cancer (National Research Council,
1989). Cauliflower was served on 39
percent of serving days in high schools
with salad bars, but on only 2 percent
of serving days in high schools without
salad bars. Another vitamin-C rich
vegetable, bell pepper, was offered
on 44 percent of serving days in high
schools with salad bars, but rarely
appeared (1 percent of serving days)
in the lunch menus of high schools
without salad bars.
Other Items on Salad Bars
To provide a more balanced view
of school lunch salad bars, we now
present a description of the items other
than fruits and vegetables offered in
salad bars. Public discussions of the
benefits of school lunch salad bars
typically focus on achieving the goal
of increased vegetable and fruit
consumption. But another important
dietary goal is reducing children’s fat
consumption, because only one-quarter
of children meet the recommendation
of the 2000 Dietary Guidelines for
Americans that children should
consume no more than 30 percent
of their calories from dietary fat
(Gleason & Suitor, 2000). In 1995,
USDA launched the School Meals
Initiative for Healthy Children
(Initiative), which was designed to
improve the nutritional quality of
school meals. The Initiative requires
that school menus comply with the
Dietary Guidelines for Americans’
recommendations for fat.
On those days when high schools
offered salad bars, salad dressing,
offered on 95 percent of salad bar
serving days, was the most common
non-fruit or non-vegetable offering
in high school salad bars (table 3).
Regular salad dressing was offered on
66 percent of these serving days, and
either low-fat or fat-free salad dressing
was offered on 67 percent of serving
days. On about 28 percent of serving
days, regular salad dressing was
offered but low-fat or fat-free salad
dressings were not.7
7 The figure of 28 percent is obtained by
subtracting the percentage of serving days in
which low-fat or fat-free salad dressings were
offered (67 percent) from the percentage of
serving days in which any type of salad
dressing was offered (95 percent).
Salad bars typically include one or
more high-fat items in addition to salad
dressing. The most common high-fat
item was regular cheese, which was
offered on 61 percent of high school
salad bar serving days. Regular cheese
was much more common than was
reduced-fat cheese, which was offered
on only 22 percent of salad bar serving
days. Similarly, meat and pasta salads
made with regular mayonnaise or salad
dressing were more commonly offered
than were their low-fat versions. High-
fat meat or pasta salads were offered
on 26 percent of salad bar serving
days; whereas, their low-fat meat or
pasta salads were offered on 7 percent
of salad bar serving days. Other
common high-fat items offered on
salad bar serving days were hard-
boiled eggs and bacon bits (21 and 34
percent of serving days, respectively).
Some low-fat meat or meat alternates,
grains, and toppings were commonly
offered on salad bars. The most
common low-fat item such as turkey,
water-packed tuna, chicken, or ham,
was served on 56 percent of salad bar
Figure 2. Percentage of high school daily menus that include certain
nutrient-dense vegetables
*Difference, when compared with high schools without salad bars, is statistically significant at the .05 level.











#AULIFLOWER□
COOKED□OR□RAW
"ELL□PEPPER□
RAW
"ROCCOLI□
COOKED□OR□RAW
#ARROTS□
COOKED□OR□RAW
3ALAD□BAR□AT□LEAST□ONCE□PER□WEEK
.O□SALAD□BAR□












2004 Vol. 16 No. 2
9
serving days.8 Two-percent or one-
percent cottage cheese was also
relatively common, being offered on
17 percent of salad bar serving days.
Depending on what children select
and consume, the high-fat items could
be a significant source of added fat and
calories in salad bar meals (Flowers-
Willets, McNaughton, Levine, &
Ammerman, 1985). For example,
analyses of the USDA’s 1994-96
Continuing Survey of Food Intakes by
Individuals (CSFII) have shown that
for a significant minority of children,
serving sizes of salad dressing are
8 More detailed tables on meat and meat
alternates, grains, and toppings on salad bars
are available from the first author upon request.
fairly large (Smicklas-Wright et al.,
2002). At the 75th percentile of quantity
consumed per eating occasion, 12-
to 19-year-old males and females
consumed about 4 tablespoons of
salad dressing. For blue cheese salad
dressing, that translates into 30 grams
of total fat, which is more dietary fat
(26 grams) than the average National
School Lunch Program meal in schools
without salad bars (Schmidt, Hirsch-
man, & McKinney, 2002; USDA,
2004). The typical child eating a
salad bar lunch probably consumes a
more modest serving of salad dressing,
since the median serving size of salad
dressing, reported in the CSFII, for
12- to 19-year-olds was 2 tablespoons
for females and 2-1/3 tablespoons for
males (Smicklas-Wright et al., 2002).
Table 3. Percentage of salad bar serving days in which other selected items were
offered in public schools with salad bars, as part of the National School Lunch
Program
High schools
All Grades
Any salad dressing
95
94
Regular
66
72
Low-fat or fat-free
67
60
Low-fat
49
44
Fat-free
33
26
Selected high-fat meat or meat alternates or toppings
Regular cheese
61
52
Bacon bits
34
28
Hard-boiled eggs
21
22
Meat or pasta salad with regular mayonnaise or salad
dressing (tuna salad, chicken salad, macaroni salad)
26
17
Sunflower seeds
8
10
Olives
16
10
High-fat meat (pepperoni, breaded chicken, beef, etc.)
8
5
Creamed cottage cheese
10
5
Selected reduced-fat meat or meat alternates or toppings
Reduced-fat cheese
22
13
Selected low-fat meat or meat alternates or toppings
Low-fat meats (turkey, water-packed tuna, chicken, ham, etc.)
56
43
2% or 1% cottage cheese
17
12
Meat or pasta salad with low-fat mayonnaise or salad dressing
(tuna salad, chicken salad, macaroni salad, etc.)
7
3
Note: Based on 5-day menu data from SNDA-II.
Conclusions
This analysis has focused on the foods
offered in salad bars. In schools with
salad bars, students have the oppor-
tunity to choose from a wider range of
fruits and vegetables, including lettuce,
tomato, other raw vegetables, fresh
fruit, and canned fruit. In particular,
salad bars are the best source of orange
and dark-green vegetables in school
lunches, because salad bars commonly
offer carrots and broccoli.
The School Nutrition Dietary Assess-
ment-II (SNDA), from which our data
were derived, has several limitations.
The study did not collect data on the
quantity of foods that school children
consumed. To understand whether the
more widespread adoption of salad
bars would improve dietary quality,
one would need to know what school-
children eat from salad bars. If students
select lettuce, tomato, other raw vege-
tables, fresh fruit, low-fat or fat-free
dressings, and low-fat meats, their
salad bar meal could have a greater
variety of fruits and vegetables and
be lower in dietary fat than would be
the case for a typical meal from the
National School Lunch Program. If
students choose to load their salads
with regular salad dressing, regular
cheese, bacon bits, or mayonnaise-
based salads, then their salad bar meal
could actually be higher in total fat
than found in the average meal from
the National School Lunch Program.
Future research on what students
select and consume from school lunch
offerings is needed to examine the
implications of the wider availability
of salad bars in more schools.
Another limitation is that SNDA-II
did not collect detailed ingredient
information on non-salad bar items
(in the traditional serving line) that
contained more than one ingredient.
For example, green salads were
10
Family Economics and Nutrition Review
frequently offered in the traditional
serving line, but no information
was available on whether carrots or
broccoli was offered. The SNDA-II
did collect information on the nutri-
tional composition of foods offered
in the traditional serving line.
We analyzed the nutrient composition
of green side salads and chef’s salads,
and our results suggested that vitamin
A- and vitamin C-rich vegetables
appeared relatively infrequently in
green salads served in the traditional
serving line. In particular, only 3
percent of chef’s salads and green side
salads were a good source of vitamin
C (i.e., greater than 20 percent of the
Recommended Daily Allowance); 27
percent of chef’s salads and 20 percent
of green side salads were a good source
of vitamin A. If one assumed that all
of the vitamin A-rich chef’s salads and
green side salads contained carrots,
which is the most common vitamin A-
rich vegetable in school lunch salads,
our analysis would still show that
carrots were served much more fre-
quently in schools with salad bars than
was the case in schools without salad
bars.
Another limitation is that data are not
available on fruits and vegetables that
are included as part of entrées other
than entrée salad bars and theme bars.
For example, tomato sauce topping for
pasta would not be counted as a tomato
in our analysis examining whether
tomatoes appeared more frequently in
schools with salad bars, even though
that tomato sauce would count as at
least part of a serving of vegetables
in the USDA Food Guide Pyramid.
Despite these caveats, our study
suggests two types of policies that
might increase children’s fruit and
vegetable consumption while main-
taining or reducing dietary fat con-
sumption. The first policy would be
to encourage schools with salad bars
to continue to offer a wide variety of
fruits and vegetables and low-fat meats
and to change their offerings to include
more low-fat or fat-free salad dress-
ings, reduced-fat cheese, and low-fat
versions of meat or pasta salads. In
addition, another policy might be to
improve nutrition education, as well
as the palatability and appearance of
salad bar meals so that children in
schools with salad bars choose salad
bars rather than the traditional serving
line. In schools with salad bars, chil-
dren get the benefit of increased fruit
and vegetable offerings only if they
choose the salad bar.
2004 Vol. 16 No. 2
11
References
Flowers-Willetts, L., McNaughton, J.P., Levine, J., & Ammerman, G.R. (1985).
Energy content of selected salad bar and hot serving line meals. Journal of the
American Dietetic Association, 85(12),1630-1631.
Fox, M.K., Crepinsek, M.K., Connor, P., & Battaglia, M. (2001). School Nutrition
Dietary Assessment Study-II Final Report (Report No. CN-01-SNDAIIFR). U.S.
Department of Agriculture, Food and Nutrition Service.
Garceau, A.O., Ebzery, M.K., Dwyer, J.T., Nicklas, T.A., Montgomery, D.H,
Hewes, L.V., et al. (1997). Do food bars measure up? Nutrient profiles of food
bars versus traditional school lunches in the CATCH Study. Family Economics
and Nutrition Review, 10(2),18-30.
Gleason, P., & Suitor, C. (2000). Changes in Children’s Diets: 1989-1991 to
1994-1996. (Report No. CN-01-CD1: 98, 119). U.S. Department of Agriculture,
Food and Nutrition Service.
National Research Council, Committee on Diet and Health. (1989). Diet and
Health: Implications for Reducing Chronic Disease Risk. Committee on Diet and
Health, Food and Nutrition Board, Commission on Life Sciences, National
Research Council.
Schmidt, S., Hirschman, J., & McKinney, P. (2002). School Lunch Salad Bars
(Report No. CN-02-SB). U.S. Department of Agriculture, Food and Nutrition
Service.
Smicklas-Wright, H., Mitchell, D.C., Mickle, S.J., Cook, A.J., & Goldman J.D.
(2002). Foods Commonly Eaten in the United States: Quantities Consumed Per
Eating Occasion and in a Day, 1994-1996. Pre-publication version. U.S.
Department of Agriculture.
U.S. Department of Agriculture, Agricultural Research Service, Nutrient Data
Laboratory. (2004). USDA Nutrient Database for Standard Reference, Release
17. Available: http://www.nal.usda.gov/fnic/foodcomp/search.
U.S. Department of Agriculture, & U.S. Department of Health and Human
Services. (2000). Nutrition and Your Health: Dietary Guidelines for Americans
(5th ed.). Washington, DC: U.S. Government Printing Office.
U.S. Department of Health and Human Services. (2001). Healthy People 2010:
Objectives for Improving Health, Volume II (2nd ed.). Washington, DC: U.S.
Government Printing Office.
U.S. General Accounting Office. (2002). Fruits and Vegetables, Enhanced
Federal Efforts to Increase Consumption Could Yield Health Benefits for
Americans (Report No. GAO-02-657).
12
Family Economics and Nutrition Review
Explaining Variations in
State Hunger Rates
A large and rapidly expanding body of research has examined causes of household-level
food insecurity and hunger. A definitive explanation has not emerged that links State
prevalence rates of hunger to State-level characteristics such as poverty, employment,
and per capita income. In this article, we examined the effect of State-level economic
and demographic characteristics on State prevalence rates of food insecurity and hunger.
Using food-security data from the U.S. Department of Agriculture and Census data on all
50 States and the District of Columbia, we first estimated, by using ordinary least squares
regression, the associations of food insecurity and hunger with a small number of carefully
chosen State-level factors. Based on these associations, we then examined the extent to
which these factors explained the high rate of hunger in Oregon and, as a contrast, the
lower-than-expected rate of hunger in West Virginia. Findings of our study suggest that
to reduce hunger rates, policymakers should consider ways to mitigate income shocks
associated with high mobility and unemployment and reduce the share of income spent
on rent by low-income families.
School Lunch Programs (Food
Research and Action Center, 2003b).
America’s Second Harvest, the
Nation’s largest hunger-relief organi-
zation, has also relied on the USDA’s
hunger estimates in supporting efforts
to alleviate hunger (America’s Second
Harvest, 2002).
State government agencies and the
media have used the USDA’s State-
level statistics to draw attention to
the problem of hunger. In Idaho and
Tennessee, newspaper editorial boards
have taken the opportunity to use
hunger estimates to suggest policy
(Idaho Statesman, 2002; Cooper,
2002). The State-level estimates have
received considerable attention in the
Pacific Northwest, particularly in
Oregon, where posted rates have been
at or near the top of the USDA’s hunger
rankings (Graves, 2002; Harrison,
2002; Cook, 2002). In spring 2003,
Oregon Governor Ted Kulongoski
convened a hunger summit and
discussed possible solutions with
human service providers, business
executives, and academic experts and
he U.S. Department of
Agriculture (USDA) monitors
annually the food security of
U.S. households. This monitoring
includes calculating the share of
households that are food insecure—
meaning that they had difficulty at
times during the year having enough
to eat—and the share of households
in which people were hungry at times
during the year because of their food
insecurity. The USDA reports these
statistics for the Nation and for each
State (Nord, Jemison, & Bickel, 1999;
Nord, Andrews, & Carlson, 2002).
The USDA’s Food and Nutrition
Service (FNS) uses these statistics to
assess the level of need for its food
assistance programs and to measure
their performance. Advocates for
programs that serve low-income
families have used these statistics to
call for a variety of policy initiatives.
The Food Research and Action Center
(FRAC), a prominent national organi-
zation seeking to end hunger, recently
urged Congress to authorize additional
funding for the Summer Nutrition and
John Tapogna, MPP
ECONorthwest
Allison Suter, MPP
ECONorthwest
Mark Nord, PhD
Economic Research Service
U.S. Department of Agriculture
Michael Leachman, PhD
Oregon Center for Public Policy
T
2004 Vol. 16 No. 2
13
has since made the eradication of
hunger a top priority of his adminis-
tration. Subsequently, the Governor
announced a strategic plan—
principally focused on job creation—
to reduce the State’s hunger rate.
However, with no precise information
about how job growth or unemploy-
ment relates to hunger, the Governor
was unable to predict the degree to
which his approach would affect
the State’s hunger rate, if at all
(Kulongoski, 2003).
The high hunger rates of Oregon and
its Northwest neighbors (Washington
and Idaho) have surprised policy-
makers and the Federal officials who
oversee USDA’s Current Population
Survey Food Security Supplement
(CPS-FSS) (Nord et al., 1999). A
definitive explanation linking State
prevalence rates of hunger to State-
level characteristics such as poverty,
employment, and per capita income has
not emerged. Because the underlying
reasons have—to this point—gone
unexplained, policy responses have
been hampered and some observers
have challenged methods used in
the survey and deemed the USDA’s
findings inaccurate or misleading
(Charles, 2003).
In this article, we examined the effects
of State-level economic and demo-
graphic characteristics on State prev-
alence rates of food insecurity and
hunger. Using food-security data and
Census data of all 50 States and the
District of Columbia, we first estimated
the associations of food insecurity and
hunger with a small number of care-
fully chosen State-level factors.
Based on these associations, we then
examined the extent to which these
factors explained the high rate of
hunger in Oregon and, as a contrast,
the lower-than-expected rate of hunger
in West Virginia.
Background
In 1990, Congress enacted the National
Nutrition Monitoring and Related
Research Act (U.S. Department of
Agriculture [USDA], 2002a). Under
the national plan mandated by this Act,
the USDA and the U.S. Department
of Health and Human Services (HHS)
formed the Food Security Measure-
ment Project. Several Federal agencies,
as well as academic and private
researchers, worked as a team to
develop standardized measures of
household food security that could
be used nationally as well as in State
and local surveys.
The team working on the Food
Security Measurement Project used,
as its starting point, the definitions
of food security, food insecurity, and
hunger established by the American
Institute of Nutrition (Anderson, 1990).
Whereas food security means assured
access by all people at all times to
enough food for active, healthy lives,
food insecurity means limited or
uncertain availability of nutritionally
adequate and safe foods or limited or
uncertain ability to acquire acceptable
foods in socially acceptable ways
(Anderson, 1990).1 Hunger refers to
the uneasy or painful sensation caused
by lack of food. As measured and
described by the project, hunger refers
specifically to hunger that results
from food insecurity (USDA, 2003b).
Based on these definitions and earlier
research, the members of the project
developed a series of questions about
behaviors and experiences known to
characterize households that are having
1Current methods of measuring food insecurity
may not fully take into account whether food
was acquired in socially acceptable ways. In
particular, reliance on Federal and community
food assistance programs by a household is not
directly considered in assessing the food-
security status of the household.
difficulty obtaining enough food. These
questions (i.e., the U.S. Food Security
Survey Module) are included in an
annual nationally representative survey
as a supplement to the monthly Current
Population Survey (CPS) of the U.S.
Census Bureau. Based on the number
of food-insecure conditions they report,
surveyed households are identified as
food secure, food insecure without
hunger, or food insecure with hunger.
A large and rapidly expanding body
of research has examined causes of
food insecurity and food insufficiency
(a related measure based on a single
question used in earlier surveys).
To date, however, almost all of this
research has examined these asso-
ciations at the household level. The
annual reports of food security by the
USDA reveal that households headed
by single parents, especially women,
and Black and Hispanic households
were more likely than others to be
food insecure (Nord et al., 2002).
Poor households have rates of food
insecurity far above the national
average, and food insecurity is more
prevalent in the South and West than
in the Northeast and Midwest (Nord
et al., 2002).
Using data from the Survey of Income
and Program Participation (SIPP by
the Census Bureau), Gundersen and
Gruber (2001) used a variety of
indicators to compare food-insufficient
households with food-sufficient ones.
They found that “income shocks”
were a major factor leading to food
insufficiency (especially for house-
holds that lacked savings) and that
rates of food insufficiency were lower
among homeowners, households
headed by senior citizens, and married
couples without children than among
other households. The authors also
speculated that moves by a household
might reduce the amount of resources
available to buy food, but they found
no statistically significant differences
14
Family Economics and Nutrition Review
between food-insufficient and food-
sufficient households in this regard.
Gunderson and Gruber (2001)
concluded that, compared with their
counterparts, food-insufficient
households faced more unemployment,
losses to the receipt of food stamps,
and other income shocks and were
less able to withstand these shocks by
using savings. Thus, these researchers
suggested that food insufficiency
should be addressed with policies that
mitigate income shocks commonly
experienced by low-income families.
Other studies have also examined
causes of household-level hunger.
Similar findings have emerged. Rose,
Gundersen, & Oliveira (1998) found
that high school graduates, home-
owners, and seniors were less likely
than others to be food insufficient.
Their findings showed that Whites,
compared with other racial groups, had
the lowest rates of food insufficiency.
Not surprisingly, Rose and colleagues
also concluded that the less money a
household had, the more likely it was
to be food insufficient.
In a more recent study, Nord (2003)
found hunger to be associated strongly
with low income, as expected, and also
found that, even with analytic controls
for income, hunger was associated
strongly with unemployment, part-time
employment for economic reasons
(i.e., because more work could not
be found), not working because of a
disability, recent household moves,
and low education. Hunger rates were
found to be lower for homeowners
and for households with the elderly—
especially households with retired
elderly—compared with their
respective counterparts.
All of these analyses were based on
household-level associations. To date,
little research attention has been given
to State-level food insecurity and
hunger and the extent to which these
household-level factors account for
the differences in prevalence rates
of food insecurity and hunger across
States. In an analysis of rates of
State hunger estimated by a FRAC-
sponsored survey, Ryu and Slottje
(1999) concluded that high school
graduates were less likely to be hungry
than were those who did not receive a
high school diploma. Nord et al. (1999)
reviewed USDA-measured rates and
demonstrated a strong association
between State poverty and prevalence
rates of food insecurity. However, the
authors also acknowledged that the
association was not perfect and pointed
in particular to Washington and Oregon
as exceptions to the general pattern.
They concluded: “. . . reasons for
these unexpected high rates of food
insecurity in the Pacific Northwest
are not known, and further research
is needed on this subject” (p. 8).
Data and Empirical Model
We were interested in explaining
State-level variations in two related
prevalence rates: food insecurity and
food insecurity with hunger, the more
severe condition. State-level preva-
lence rates of food insecurity and
hunger for our analysis were taken
from work by Nord et al. (2002)—the
most recent statistics on food security
that are published by the USDA. These
statistics are particularly well suited
for analysis of the associations of
State-level characteristics with State
hunger rates, because they span 1999
to 2001—a period that overlaps the
collection of data through the 2000
Decennial Census and the Census
Supplemental Survey. State-level
statistics based on these Census data
are highly precise.
The USDA’s statistics on food in-
security and hunger are based on data
collected in the CPS-FSS of April
1999, September 2000, and December
2001. The CPS-FSS is a nationally
representative survey of about 50,000
households that is conducted annually
by the U.S. Census Bureau for the
USDA. Representative of both the
U.S. civilian noninstitutionalized
population and each State, the CPS-
FSS is conducted as a supplement to
the monthly CPS, a labor force survey
conducted by the Census Bureau for
the Bureau of Labor Statistics. House-
holds are classified as food secure,
food insecure without hunger, or food
insecure with hunger,2 a classification
that is based on the number of food-
insecure conditions they report in
response to the 18 questions in the
food-security module.
For most monitoring and analytic
purposes, the CPS sample size in most
States is too small to produce annual
food insecurity or hunger rates with
sufficient reliability. Consequently, the
USDA routinely reports State-level
food insecurity and hunger rates as
3-year averages. We used the 3-year
averages for 1999 to 2001 (Nord et al.,
2002) as our main analytic variables.
Our method to assess the associations
of State-level food insecurity and
hunger rates with State economic and
demographic characteristics was a
straightforward application of ordinary
least squares (OLS) regression
analysis. We hypothesized that a
number of State-level characteristics
independently affect State-level food-
insecurity and hunger rates. The
relationship between the State hunger
rate Y and the explanatory variables X
is generally assumed to take this form:
Yi = β0 + β1X1i + β2X2i + .... + βnXni + εi.
2A complete description of the CPS
sample design is available at http://
www.bls.census.gov/cps/tp/tp63.htm.
2004 Vol. 16 No. 2
15
OLS provides estimates of the values
of the β terms, which quantify the
relationship between each of the
explanatory variables and hunger
or food insecurity. We analyzed the
associations between food insecurity
and explanatory variables in a separate
model.
We selected the explanatory variables
(X1i, X2i, etc.) based on our review
of the literature and discussions with
experts on food insecurity and hunger.
The limited degrees of freedom in this
cross-sectional analysis called for a
parsimonious model. The literature and
program experts identified associations
between five individual characteristics
(change of residence, unemployment
status, poverty status, age, and race)
and food insecurity and hunger. We
additionally included a measure of
housing cost because a number of
observers had identified a correlation
between high housing costs and food
insecurity. Housing is a major item
in the budget of most low-income
households and, if too high, can
“crowd out” resources available for
food (Gundersen & Gruber, 2001;
Rose et al., 1998; Food Research
and Action Center, 2003a).
Hypothesized Relationships
In this section, we discuss the
hypothesized relationship between
change of residence, unemployment
status, poverty status, age, and race
and rates of food insecurity and hunger.
We describe these variables as well
as report the means and standard
deviations (table 1).
• Percentage of households in
2000 that moved within the last
year. The Census Supplemental
Survey reports the share of
households in a State that indicate
whether they changed dwellings
between 1999 and 2000.
Households can move for a number
of reasons—some positive (e.g.,
house upgrade or relocation to a
new job) and some problematic
(e.g., evictions or household
dissolutions such as divorces or
separations). Household-level
research has suggested that,
overall, households that have
moved recently, compared with
households that have not moved
recently, were more likely to be
food insecure. We hypothesized
that this measure is a proxy for
income shocks, which Gundersen
and Gruber (2001) demonstrated
had a positive relationship with
hunger. The variable’s mean across
States was 16.4 percent, and the
standard deviation was 2.7
percentage points.
Table 1. Descriptive statistics for the 50 States
Standard
Variables1
Mean
deviation
Percentage
Percent 2
points
Share of population experiencing food insecurity
with hunger
3.1
0.9
Share of population experiencing food insecurity
10.2
2.2
Share of population in a different house
16.4
2.7
Peak unemployment rates during 1999-2001
5.0
1.1
Share of population living in poverty
12.1
3.3
Share of renters paying more than 50 percent of
income on gross rent
16.4
1.8
Share of population non-Hispanic White
74.9
16.1
Share of population under age 18
25.5
1.9
1Percentages for all variables are for 2000 unless noted otherwise.
2These figures report the simple average of 50 individual State observations with each State’s observation
given equal weight. That is, California’s observation is given the same weight as North Dakota’s.
Consequently, the figure does not represent a U.S. average, which would vary the States’ weighting by
their size.
• Average of 1999, 2000, and 2001
peak unemployment rates. We
constructed this variable as the
average of the peak State un-
employment rates in each of three
years: 1999, 2000, and 2001. The
3 years coincide with the period
of measurement for the dependent
variables. We selected the peak
rate in each year, rather than the
average, to capture the worst
economic conditions reported
in the States. Peak unemployment
rate is likely to be a better measure
of the share of the labor force that
experienced job loss and a related
income shock at some time during
the year. This measure is, therefore,
temporally consistent with the
measures of food insecurity and
hunger, which reflect the most
16
Family Economics and Nutrition Review
problematic food-access conditions
of the year. (Households were
classified as food insecure or
food insecure with hunger if they
experienced these conditions at any
time during the year.) Based on the
work of Gundersen and Gruber
(2001) and others (Rose et al.,
1998), we hypothesized that high
peak unemployment would be
associated with high food insecurity
and hunger rates. We used the
applicable variable from the Local
Area Unemployment Statistics
series of the Bureau of Labor
Statistics. Its mean was 5.0 percent;
the standard deviation, 1.1 percent-
age points.
• State poverty rate. Other studies
have indicated that a household’s
income level is a determinant of
food insufficiency (Gundersen &
Gruber, 2001; Rose et al., 1998;
Gundersen & Oliveira, 2001; Nord,
2003). Moreover, the most recent
USDA report showed that 12.9
percent of households with incomes
below the Federal poverty level
experienced hunger, compared
with a national average of only
3.3 percent (Nord et al., 2002).
Therefore, we anticipated that
States with higher poverty rates
would also register higher hunger
rates. State poverty rates, measured
for calendar year 1999 through the
2000 Decennial Census, averaged
12.1 percent; the standard devia-
tion, 3.3 percentage points.
• Share of renters spending more
than 50 percent of income on
gross rent. Just as limited income
can put a household at risk for
hunger, high expenses can do the
same. Past studies have reported
that renters were more likely than
homeowners to be food insecure
(Gundersen & Gruber, 2001; Rose
et al., 1998; Gundersen & Oliveira,
2001; Nord, 2003). Therefore, we
used the share of renter-households
in the State that spent more than
50 percent of their incomes on
gross rent as an explanatory
variable.3 We anticipated that
within the group of renting house-
holds, those with high rents relative
to their incomes would be particu-
larly prone to hunger. We used the
variable from the 2000 Decennial
Census. The mean for the variable
was 16.4 percent; its standard
deviation was 1.8 percentage
points.
• Population share of non-
Hispanic Whites. Previous
research has offered mixed
findings about the effect of race
and ethnicity on hunger or food
insufficiency (Gundersen & Gruber,
2001; Rose et al., 1998; Gundersen
& Oliveira, 2001; Nord, 2003). We
included the variable that measured
the share of a State’s population that
was non-Hispanic White, but we
had no a priori assumption about its
effect on hunger rates. This variable
averaged 74.9 percent; its standard
deviation was 16.1 percentage
points.
• Population share under age 18.
Researchers have indicated that
larger households, and particularly
large households with children,
have higher hunger rates (Rose
et al., 1998). We anticipated that
as a State’s share of the population
under age 18 rose, so would its
hunger rate. The mean for this
variable was 25.5 percent; its
standard deviation was 1.9
percentage points.
Finally, we explored the extent to
which the regression model could
account for the high rate of hunger
in Oregon. Based on the regression
3Gross rent consists of direct rental costs plus
essential utilities.
coefficients and the values of each
State’s independent variables, we
calculated the rates of hunger predicted
by the regression model for each State.
We also calculated the contribution of
each factor to Oregon’s higher-than-
average hunger rates. As a counter-
example, we examined the contribution
of each factor to the hunger rate in
West Virginia, which was near the
national average despite a relatively
high State poverty rate.
Results
Because of the limited number of
observations (51) and the estimation
error associated with prevalence
rates of State-level hunger, the model
predicted State hunger rates quite well.
Overall, the six independent variables
explained 64 percent (unadjusted R2)
of the variation in State hunger rates—
a high rate for this type of model—
and 74 percent (unadjusted R2) of
the variation of State rates of food
insecurity (table 2). Moreover, the
measured relationships between most
of the independent variables and State
rates of hunger and food insecurity
were statistically significant and
sufficiently strong to be of substantive
importance. Also, both in-sample and
out-of-sample predictions ranked
Oregon with the second highest
hunger rate.
Examination of the estimated relation-
ships between each of the independent
variables and State hunger and in-
security rates revealed that the
“different house,” or mobility variable,
had the most robust and consistent
relationship with State hunger and
food insecurity rates. The hunger
model suggests that each percentage-
point increase in the share of a State’s
households that reported changing
dwellings between 1999 and 2000
was associated with a 0.13-percentage-
point increase in the State’s hunger
2004 Vol. 16 No. 2
17
rate. The magnitude of the coefficient
was roughly twice as large in the
estimate of food insecurity (but the
level of food insecurity was also much
higher, so the proportional association
was similar or somewhat smaller).
We interpret the coefficient of the
“different house” variable as primarily
measuring the associations of food
insecurity and hunger with economic
shocks and family disruptions.
Effects of peak unemployment rates
also were quite strong. A 1-percentage-
point increase in peak unemployment
rates was associated with an increase
of 0.31 percentage points in a State’s
hunger rate. This relationship is
consistent with earlier research
findings that job loss and income
shocks are associated with a higher
likelihood of food insufficiency
(Gundersen & Gruber, 2001; Nord,
2003). We also found unemployment to
put upward pressure on food insecurity
rates; this association, however, was
weaker than the one for hunger and
was not statistically significant.
As expected, high poverty rates also
put upward pressure on hunger and
food insecurity rates. This association
for hunger, however, was not statis-
tically significant. The relatively high
correlation between State-level poverty
and unemployment measures accounted
for the weakness of the estimated
relationship between poverty and
hunger on the one hand and between
peak unemployment and food in-
security on the other. Because States
with high poverty rates tended also to
Table 2. Estimated relationships between selected State characteristics and
rates of hunger and food insecurity
Food insecurity
Food insecurity with hunger
(with or without hunger)
Regression
Standard
Regression
Standard
coefficient
error
coefficient
error
Share of population in a
different house
0.132
(0.034)*
0.280
(0.073)*
Peak unemployment rates
during 1999-2001
0.314
(0.100)*
0.187
(0.215)
Share of population living
in poverty
0.034
(0.031)
0.360
(0.067)*
Share of renters paying more than
50 percent of income on gross rent
0.130
(0.055)*
0.276
(0.118)*
Share of population
non-Hispanic White
0.011
(0.006)
0.014
(0.013)
Share of population under age 18
0.112
(0.047)*
0.434
(0.101)*
Constant
-0.069
(0.018)*
-0.164
(0.040)*
R2
0.638
0.736
Adjusted R2
0.588
0.700
Note: The data are based on ordinary least squares analysis.
*p < .05.
A 1-percentage-point increase in
peak unemployment rates was
associated with an increase
of 0.31 percentage points in a
State’s hunger rate.
18
Family Economics and Nutrition Review
have high peak unemployment rates,
the models had difficulty disentangling
the independent effects of poverty and
unemployment. In the case of the
hunger model, the stronger association
with the unemployment variable left
little residual association with the
poverty rate. However, when we
removed the unemployment variable
from the model (analysis not shown),
the poverty variable became statis-
tically significant. In the case of the
food-insecurity model, poverty had
the strong relationship with food
insecurity; removing it from the model
resulted in a statistically significant
association with unemployment.
The additional analyses with poverty
rates and peak unemployment rates,
omitted in turn, also confirmed that the
peak unemployment variable was more
strongly associated with hunger rates
than with food insecurity rates while
the poverty variable was more strongly
associated with food-insecurity rates
(data not shown). These findings
suggest that economic shocks at the
household level, for which peak
unemployment is a proxy at the State
level, are associated with the more
severe hunger condition. In States
with high poverty rates, by contrast,
low-income households and their
communities are more likely to have
adjusted to sustained low levels of
income. Persistently poor households
are likely to have developed ways to
avoid hunger by relying on family,
friends, and local institutions and by
altering their consumption patterns.
Community institutions in States with
consistently high poverty rates will
have had time to adjust and better
reach families in need.
High housing costs were strongly
associated with hunger and food-
insecurity rates. Our model estimated
that a 1.0-percentage-point increase in
the share of a State’s renters who paid
more than 50 percent of income for
gross rent was related to a 0.13-
percentage-point increase in the State’s
hunger rate. For example, the 8.9-
percentage-point difference between
New York (the Nation’s highest) and
South Dakota (the Nation’s lowest)
and the housing-burden measure is
expected to result in a 1.1-percentage-
point difference in hunger rates
between the two States (data not
shown).
We had no expectations about the
effects of the non-Hispanic White
variable on rates of hunger and food
insecurity. The variable showed a
positive but weak and statistically
insignificant relationship with the
dependent variables. The lack of a
conclusive relationship is consistent
with previous, generally mixed,
findings reported by researchers
(Rose et al., 1998).
As the share of a State’s population
under age 18 increased, so did both
hunger and food insecurity. A 1-
percentage-point increase in the State’s
population share under age 18 was
significantly associated with a 0.11-
percentage-point increase in hunger
and a 0.43-percentage-point increase
in food insecurity. We were concerned
that this variable could be confounding
the effects of a larger share of children
with a smaller share of elderly in the
State. However, including the elderly
population share in the model (analysis
not shown) resulted in no substantial
change in the coefficient on the share
of the State’s population under age 18.4
The measured associations of hunger
and food insecurity with the elderly
population share remained, even when
all households with elderly were
excluded from the sample used in the
analysis for calculating rates of food
insecurity and hunger. We thus
concluded that the association was
4To obtain the detailed data for each State,
please contact the first author.
spurious, resulting from other charac-
teristics of States with large elderly
population shares.
We also examined the extent to which
the regression models accounted for
hunger rates in Oregon and West
Virginia (table 3). Oregon registered
one of the highest hunger rates (5.8
percent) in the Nation; yet, it had a
poverty rate slightly below the national
average (11.6 vs. 12.1). West Virginia,
on the other hand, had a hunger rate
near the national average (3.3 percent);
yet, it had the fifth highest poverty rate
of all States (17.9 percent). We
estimated—based on the model’s
regression coefficients and the States’
values on each independent variable—
how Oregon’s and West Virginia’s
hunger rates would change if the
State’s levels were equal to the mean
for all 50 States.5
For example, Oregon’s share of renters
paying more than 50 percent of their
income in rent is 2.9 percentage points
higher than the U.S. average (19.3
vs. 16.4 percent, table 3 and table 1,
respectively). If Oregon’s rate fell to
the 50-State mean, we estimated that
the State’s hunger rate would fall
by 0.4 percentage points (table 3).
Oregon’s high levels of peak unem-
ployment rate and residential mobility,
as measured by the share of the popu-
lation in a different house, explained
even more of the gap between
Oregon’s hunger rate and those of
other States. For each of these two
variables, if Oregon’s rate fell to the
50-State mean, the model predicted
that the State’s hunger rate would
decline by 0.6 percentage points.
In West Virginia, high peak unem-
ployment pushed the hunger rate up.
Bringing peak unemployment down to
5These values are not national averages because
they are unweighted; they are means for the 50
States.
2004 Vol. 16 No. 2
19
the 50-State mean (5 percent) would
lower the hunger rate by 0.6 percentage
points. West Virginia’s high poverty
rate (17.9 percent) was estimated to
push up the hunger rate only 0.2
percentage points. As we observed,
with peak unemployment in the model,
the effect of the poverty rate was small.
Furthermore, West Virginia’s share
(17.7 percent) of renters paying more
than 50 percent of their income for
gross rent was nearer the 50-State
mean (16.4 percent) than was Oregon’s
(19 percent), putting a smaller upward
pressure on the hunger rate. The most
important difference between the two
States, however, was that the factors
pushing the hunger rate up were largely
offset by West Virginia’s much lower
rate of residential mobility, well below
the U.S. mean, and the considerably
smaller-than-average share of children
in the population. Taken together, these
factors resulted in a hunger rate in West
Virginia that was similar to the mean
for the 50 States.
Policy Implications and
Conclusions
Prior research provided considerable
insight about factors affecting
household-level hunger, food in-
security, and food insufficiency but
little information about the extent to
which these factors explained differ-
ences in State prevalence rates.
The lack of an intuitively satisfying
explanation for high estimated hunger
rates in the Pacific Northwest left
Table 3. Estimated effect of key characteristics on hunger rates in Oregon
and West Virginia
Oregon
West Virginia
Estimated
Estimated
Rate
effect1
Rate
effect1
Percent
Percentage
Percent
Percentage
point
point
Share of population
in a different house
21.1
-0.6
12.9
0.5
Peak unemployment rates
during 1999-2001
7.0
-0.6
6.9
-0.6
Share of population living in poverty
11.6
0.0
17.9
-0.2
Share of renters paying more than
50 percent of income on gross rent
19.3
-0.4
17.7
-0.2
Share of population non-Hispanic White
83.5
-0.1
94.5
-0.2
Share of population under age 18
24.7
0.1
22.2
0.4
Total
-1.6
-0.3
State hunger rate
5.8
3.3
1The effect refers to the estimated change in hunger rate if the rate equaled the mean hunger rate of the 50
States. For example, Oregon’s share of the population in a different house in 2000 was 18 percentage points
higher than the 50-State mean (21.1 vs 3.1). If Oregon’s mean were the same as that of the 50 States,
Oregon’s hunger rate would fall by 6 percentage points.
policymakers unsure about how to
address the problem of hunger and led
critics to question the validity of the
USDA survey and its measurement
techniques. The ability to associate
State hunger rates to key social and
economic conditions at the State level,
as demonstrated in this study, sheds
light on State rankings and, by doing
so, both lends credibility to the State
hunger statistics and provides policy-
makers with some guidance about
policy responses. Nevertheless, this
relatively simple cross-sectional
analysis points only to associations
between hunger and food insecurity
and the hypothesized explanatory
variables. Our work falls short of
establishing definitive causal
relationships.
The findings suggest that highly
transient populations put upward
pressure on the hunger rates in their
States. High mobility serves as a proxy
for a variety of lifetime disruptions—
divorce, separation, eviction, and other
shocks to family income—that put
people and families at risk of hunger
and food insecurity. This risk may be
exacerbated by the diminished social
cohesion that characterizes highly
mobile populations.
Paradoxically, good regional economic
conditions often lead to high levels
of mobility. States with booming
economies attract an influx of job
seekers. States with a high percentage
of seasonal jobs may experience sub-
stantial internal migration during the
year. States with strong economies may
experience rapid growth in housing
prices, resulting in both high housing
costs for residents and relatively large
portions of the population shifting into
new or less expensive areas. People
living through these types of economic
conditions may be at a higher risk of
hunger; because, they are more likely
than others to be living in new
neighborhoods, distant from family
20
Family Economics and Nutrition Review
and friends and disconnected from the
local infrastructure of social support.
Religious institutions and government
programs may not effectively reach
people who have lived in the area for
only short periods.
In trying to lower hunger rates in
highly mobile States in the West and
South, policymakers may want to focus
their efforts on vulnerable, mobile
populations—newcomers, seasonal
workers, and displaced renters, for
example. In doing so, policymakers
in these States can assume a role in
overcoming, or partially offsetting,
the lack of social cohesion in their
communities. If some Western and
Southern States lack natural support
networks (e.g., family and long-time
neighbors) found in the Northeast or
Midwest, citizens and policymakers
can attempt to substitute for the lack
of cohesion through nonprofit or
public efforts.
For example, a highly developed
network of food banks may prove
more important in Oregon than in
States in other regions with more
stable populations. Also, a state-of-
the-art information and referral system,
as envisioned by United Way’s 211
coalition, can provide much-needed
direction to those who relocate and
need to know what resources are
available to them. Policymakers can
also reform the State unemployment
insurance programs to better reach
seasonal workers, focus food stamp
outreach efforts on newcomers, and
devise effective support programs
for displaced renters.
The association between unemploy-
ment and hunger suggests that an
economic development policy could
serve a dual purpose as an anti-hunger
strategy. Many governors have indi-
cated that they want an integrated
approach to economic development—
one that stimulates job growth and
trains workers. Plans on both fronts
are necessary to help State economies
and their hungry citizens. Economic
development efforts that lower poverty
rates, reduce seasonal fluctuations in
unemployment rates, and provide jobs
in rural areas experiencing high
unemployment may be particularly
effective in fighting hunger.
Another policy direction to emerge
relates to increasing the supply of
affordable housing. Findings of this
study indicate a substantial reduction in
the hunger rate for moderate decreases
in the share of renters who pay more
than 50 percent of their income on
gross rent. States with the largest share
of such renters, such as Oregon, have
room to improve and the potential to
address concerns of both housing and
hunger advocates. Competing pro-
posals have been offered to increase
the supply of affordable housing:
construction of more affordable
housing projects and vouchers for
existing units, on the one hand, and
relaxation of land-use controls to
lower the price of land, on the other
hand. If further research demonstrates
that these approaches do, in fact,
increase the supply of low- and
moderate-cost housing, then both
may reduce the prevalence of hunger,
whatever the other strengths and
weaknesses of these approaches
might be.
In each State that has a high prevalence
of hunger, a different combination of
factors may be responsible. The results
of this study can help policymakers and
the concerned public in each of these
States understand more fully the factors
that particularly affect their State. We
hope that this improved understanding
will lead to increasingly effective
policies, programs, and community
institutions to reduce hunger and food
insecurity.
2004 Vol. 16 No. 2
21
References
America’s Second Harvest. (2002). One in Ten Americans Face Hunger.
Retrieved July 7, 2003, from http://www.secondharvest.org/
site_content.asp?s=406&p=1.
Anderson, S.A. (Ed.). (1990). Core indicators of nutritional state for difficult-
to-sample populations. Journal of Nutrition, 120(11S), 1557-1600. A report
prepared by the Life Sciences Research Office, Federation of American Societies
for Experimental Biology, for the American Institute of Nutrition.
Charles, J. (2003, January). How hungry is Oregon? Brainstorm NW: Ideas and
Entertainment for Oregon and the Pacific Northwest, pp. 23-28.
Cook, R. (November 30, 2002). ‘Immoral Irony’: Hunger rates highest in rural
West. The Associated Press. Retrieved from http://www.nexis.com/research.
Cooper, C. (2002, September 7). Tennessee 10th in nation in hunger, study claims.
Chattanooga Times Free Press. Retrieved May 7, 2003, from http://
www.nexis.com/research.
Food Research and Action Center. (2003a). Hunger in the U.S. Retrieved January
13, 2005, from http://www.frac.org/html/news/newsdigest/digest112003.htm.
Food Research and Action Center. (2003b). National Hunger Awareness Day
Press Release, June 5, 2003. Retrieved July 7, 2003, from http://www.frac.org/
html/news/060503hungerDay.htm.
Graves, B. (2002, April 25). Living in the state of hunger. The Oregonian.
Retrieved May 7, 2003, from http://www.nexis.com/research.
Gundersen, C., & Gruber, J. (2001). The dynamic determinants of food
insufficiency. In M.S. Andrews & M.A. Prell (Eds.), Second Food Security
Measurement and Research Conference, Volume II: Papers. (Food Assistance and
Nutrition Research Report No. 11-2, pp. 91-109). U.S. Department of Agriculture,
Economic Research Service.
Gundersen, C., & Oliveira, V. (2001). The Food Stamp Program and food
insufficiency. American Journal of Agricultural Economics, 83(4), 875-887.
Harrison, N. (2002, August 23). Is Utah’s hunger rate near the top? It depends.
Deseret News. Retrieved from http://www.nexis.com/research.
Idaho Statesman. (2002, December 8). Our view: The grim and sad reality of
hunger in rural Idaho. Retrieved May 7, 2003, from http://www.nexis.com/
research.
Kulongoski, T. (2003, June 5). In my opinion no one in Oregon should go hungry.
The Oregonian. Retrieved July 7, 2003, from http://www.nexis.com/research.
22
Family Economics and Nutrition Review
Nord, M. (2003, July). Keeping Warm, Keeping Cool, Keeping Food on the
Table: Seasonal Food Insecurity and Costs of Heating and Cooling. Paper
presented at the annual meeting of the National Association for Welfare Research
and Statistics, San Diego, CA.
Nord, M., Andrews, M., & Carlson, S. (2002). Household Food Security in the
United States, 2001. (Food Assistance and Nutrition Research Report No. 29).
U.S. Department of Agriculture, Economic Research Service.
Nord, M., Jemison, K., & Bickel, G. (1999). Measuring Food Security in the
United States: Prevalence of Food Insecurity and Hunger, By State, 1996-1998.
(Food Assistance and Nutrition Research Report No. 2). U.S. Department of
Agriculture, Economic Research Service.
Rose, D., Gundersen, C., & Oliveira, V. (1998). Socio-economic Determinants
of Food Insecurity in the United States: Evidence from the SIPP and CSFII
Datasets. (Technical Bulletin 1869). U.S. Department of Agriculture, Economic
Research Service.
Ryu, H.K., & Slottje, D.J. (1999). Analyzing perceived hunger across States in the
US. Empirical Economics, 24, 323-329.
U.S. Department of Agriculture, Economic Research Service. (2002a). Food
Security in the United States: History of the Food Security Measurement Project.
Retrieved May 7, 2003, from http://www.ers.usda.gov/briefing/FoodSecurity/
history/.
U.S. Department of Agriculture, Economic Research Service. (2002b). Food
Security in the United States: Measuring Household Food Security. Retrieved
May 7, 2003, from http://www.ers.usda.gov/Briefing/FoodSecurity/measurement/.
2004 Vol. 16 No. 2
23
The Pitfalls of Using a Child Support
Schedule Based on Outdated Data
A strong rationale for updating child support guidelines arises from changes over time in
the measurement of expenditures on children, as well as from changes in the empirical
relationship between expenditures on children and the income of parents. Such changes
affect the accuracy of the numerics upon which States’ child support guidelines are based.
This study evaluated an alternative child support guideline that was proposed for Virginia
and drew lessons for other States that similarly base their guidelines on older survey data.
Regression results showed that, over time, the child expenditure and household income
relationship has changed considerably. Furthermore, the largest increases in expenditures
attributable to children have occurred for lower and middle-income households.
Yana van der Meulen Rodgers, PhD
Rutgers University
William M. Rodgers III, PhD
Rutgers University
hile the Family Support Act
of 1988 requires all States to
assess their child support
guidelines at least once every 4 years,
States are not mandated to change their
guidelines following the assessment.
A number of economic changes could
warrant the updating of a State’s child
support guidelines. One such change:
Today, most obligors are fathers who
are more involved in child-rearing than
they were 20 years ago. In addition to
paying child support, many obligors
spend money on their children during
visitation hours. This increase in father
involvement and spending provides a
rationale for implementing adjustments
to child support schedules. Another
change: A worsening in labor-market
opportunities for less-skilled men has
led to sharp increases in arrearages
(Katz & Krueger, 1999; Welch, 2001).
Including a downward adjustment for
low-income obligors in child support
schedules can help to reduce arrears
caused by child support awards that
surpass the ability of low-income
obligors to pay (Holzer, Offner, &
Sorenson, 2003; Sorenson & Zibman,
2001).
Another rationale for updating child
support guidelines arises from changes
that have occurred in the measurement
W
of expenditures on children, as well as
from changes in the empirical relation-
ship between expenditures on children
and the income of parents. These
changes affect the accuracy of the
numerics upon which States’ child
support guidelines are based. To
understand better the implications
of these changes, we examined the
costs involved when States use
schedules based on statistical relation-
ships derived from outdated survey
data. We evaluated an alternative child
support guideline that was proposed
for the Commonwealth of Virginia and
then drew lessons for other States that
similarly base their guidelines on older
estimates of child-rearing expenditures.
The alternative schedule for Virginia
proposed that total child support
awards as a share of monthly income
be raised at all income levels except
for the lowest end of the income
distribution.
Virginia’s child support schedule has
not been updated since the mid-1980s.
The schedule is based on a study of
child-rearing expenditures published in
1984 that used the 1972-73 Consumer
Expenditure Survey (CES), the best
household expenditure data available
at the time. Because the Bureau of
Labor Statistics has made significant
24
Family Economics and Nutrition Review
improvements in the quality and com-
prehensiveness of its data collection
and because the data are collected
annually, Virginia’s current schedule is
no longer tied to the best quality data
from the CES. As was the case for
Lino (2001), we found that average
total expenditures on children have
risen in past decades and have changed
in composition. However, the child
expenditure and income relationship
upon which Virginia’s schedule is
based may also have changed since
the 1970s, a hypothesis that was tested
in this study. Such a change would
imply that Virginia and 10 other States
with older guidelines are no longer
generating child support orders that are
linked to accurate estimates of the child
expenditure and income relationship.
Statistical evidence in this study
provides a strong economic rationale
for developing a new child support
schedule in Virginia and in other States
with similar guideline structures.
Underlying Models and
Measurement Issues
Federal legislation requires all States
to have formal guidelines for calcu-
lating the dollar value of child support
awards. These child support guidelines
must take into account the earnings of
the nonresidential parent, they must
base support obligations on numerical
criteria, and they must include the
child’s health care costs into the
calculations. No particular method to
determine State guidelines is mandated,
so States must make decisions about
the underlying model and measurement
issues surrounding the definition of
income and child-rearing costs (Beller
& Graham, 1993; Venohr & Williams,
1999). States have chosen versions
of three underlying models: the
“Percentage of Obligor Income”
model, the “Income Shares” model,
and the “Melson Formula” model.
The Percentage of Obligor Income
model entails the most basic calcula-
tions of the three models, in which the
noncustodial parent pays a certain
share of his or her income to the cus-
todial parent. The share rises with the
number of children; for some States,
however, the share also changes as the
income level of the obligor changes.
In contrast, the Income Shares model is
more detailed. The underlying premise
of this model is that the child should
obtain the same percentage of total
income that he or she would have
obtained if the parents were together.
In calculating the child support
amount, the income of both the mother
and father is combined to proxy for the
total income of an intact family. This
income calculation is then linked to
estimates of child-rearing expenditures
by intact families with the same income
level and number of children. In the
final basic step for converting esti-
mates of child expenditures into a
schedule of child support payments for
noncustodial parents, the estimated
child support amount is divided
between the two parents according
to their respective income shares.
Finally, the Melson Formula model is
similar to the Income Shares model
except that both parents are allowed
a reserve amount to cover their own
subsistence needs and to sustain
employment.
No matter which model is chosen,
however, States must make decisions
regarding the measurement of income
and expenditures on child-rearing.
According to Beller and Graham
(1993), to measure income, most
States use either adjusted gross income
(income adjusted for prior support
orders and health insurance) or net
income (income with these same
adjustments plus deductions for taxes,
mandated retirement contributions, and
union dues). A few remaining States
use gross income. A number of States
also build into their schedules a self-
support reserve that protects the ability
of the obligor to meet his or her basic
subsistence needs and to facilitate
employment. With a self-support
reserve, if the combined gross monthly
income is less than a certain threshold,
then the guideline is not used to com-
pute the child support order. Instead,
a fixed minimum award is applied to
the noncustodial parent. At the other
end of the income distribution, very
high income levels are sometimes
treated with an income cap, declining
percentages, or noncash transfers in the
application of child support guidelines.
There is less agreement among policy-
makers and academics about the best
estimates of child-rearing costs. These
estimates come from a number of
studies that vary in the underlying
methodology as well as the survey year
used to determine the estimations. In
a survey of this literature, Beller and
Graham (1993) point to two indirect
approaches—the Engel method and
the Rothbarth method—and the direct
approach for estimating child-rearing
costs.
The Engel method is based on the
premise that families who spend the
same share of their total consumption
expenditures on food are equally well
off. When the Engel method is used
to compute child-rearing costs, two
families, one with no children and one
with one child, are assigned equal
proportions for food spending in the
total budget. Then the cost of raising
the first child is the increase in
spending required to keep the one-
child family spending the same budget
share on food. The approach is similar
for families with more children. The
most important assumption this
approach must satisfy is separability
in consumption; that is, families will
not change the way they allocate their
spending across food and other
2004 Vol. 16 No. 2
25
consumption items as they have
children.
The Rothbarth method is similar in
notion and underlying assumptions,
except that the equalizing factor across
families is the budget share devoted to
adult goods. Deaton and Muellbauer
(1986) argue that the separability
assumption causes the Engel estimator
to overestimate child-rearing costs
(families with children are over-
compensated in computations to
keep the food share equal), while the
Rothbarth estimator underestimates
child-rearing costs (families with
children are undercompensated in
computations to keep the adult-goods
share equal). Finally, the direct
approach for estimating child-rearing
costs involves directly totaling dif-
ferent categories of spending on
children. A few categories, such as
child care or children’s clothing, can
be measured by actual spending on
children, while most other categories,
such as health care or housing, are
measured by estimates of spending
attributable to children.
By 1990, over 30 States, including
Virginia, had based their guidelines
on the Income Shares model. For most
of these States, the estimates of child-
rearing expenditures were initially
calculated from Espenshade’s work
(1984), which was based on the Engel
method and data from the 1972-73
CES. Subsequently, a number of States
have updated their child support guide-
lines to reflect more recent estimates
of child-rearing costs. These recent
estimates, drawn mostly from work in
Betson (1990), use a range of methods
applied to CES data from 1980 to
1986. Some States have also drawn
from annual reports by the U.S.
Department of Agriculture, which uses
the direct approach to total categories
of spending attributable to children.
In 2003, there were still 11 States,
including Virginia, that based their
guidelines on Espenshade’s earlier
estimates (Venohr & Griffith, 2003).
The other 10 States were Alabama,
Florida, Indiana, Louisiana, Kansas,
Kentucky, Maryland, Michigan, Rhode
Island, and Washington. However,
these older guidelines may no longer
generate realistic child support orders.
In recent decades, the CES’s sample
size has grown and the level of detail
has improved, providing better expen-
diture and income data. Concepts and
definitions have changed so much
that officials of the Bureau of Labor
Statistics warn users to exercise
caution when comparing current
survey data with data from earlier
surveys, especially with data from
surveys conducted prior to 1984.
Estimating Expenditures
on Children
This section describes a schedule
of child support that was developed
for the Quadrennial Child Support
Review Panel of the Commonwealth
of Virginia.1 The schedule has been
grounded in current economic research
on child-rearing expenditures. New
estimates of child-rearing expenditures
were developed by using micro data
on husband-wife households from
the 2000 CES. The sample criteria
included having some positive amount
of household income for the past year
and reporting one to three children
under age 18 living in the home.2 These
criteria yielded 1,987 households with
one child, 2,557 households with two
children, and 990 households with
1The full report by Rodgers (2002) can be
found at www.dss.state.va.us/pub/pdf/
dcsepanel_final.pdf.
2Sample sizes for husband-wife households
with more than three children were too small
to generate reliable results.
three children. Data were used for
households with gross monthly
incomes that ranged from $1,200 to
$8,500. Computed from the 2000
decennial census micro-data file for
Virginia, this range of the income
distribution represented 76 percent
of all Virginia married-couple house-
holds with one to three children below
age 18. Of the remainder, 2 percent
were below the specified income range
and 22 percent were above the range.
Because of the CES’s focus on lower
and middle-income families, the
Bureau of Labor Statistics cautions
researchers about making statistical
inferences on the expenditures of
households with gross incomes above
$8,500.
Identifying Total
Expenditures
This study estimated a household’s
expenditures on children by using the
direct approach of totaling different
categories of actual expenditures. A
three-step procedure was used. The
first step involved identifying the
total expenditures on food, housing,
clothing, transportation, education,
miscellaneous expenditures, and
nonextraordinary health expenditures.
In Virginia, support for extraordinary
health expenditures, child care costs,
and health insurance premiums for the
child are treated as add-ons after the
initial level of support has been
calculated.
Sample means from the 2000 CES
showed that housing, variable trans-
portation, and food expenditures
comprised 70 percent of total
household expenditures. Of note,
expenditures on housing in the CES
are underestimated because the Bureau
of Labor Statistics treats mortgage
principal payments as savings rather
than as expenditures. Because a large
26
Family Economics and Nutrition Review
Table 1. Housing and transportation expenditures attributable to children based
on per capita and average use allocation methods
Housing
Transportation
Per capita
Average use
Per capita
Average use
Number of children
1
33.3
1.0
33.3
24.0
2
50.0
9.5
50.0
44.0
3
60.0
12.4
60.0
38.0
Source: JLARC (2001).
portion of an obligor’s direct expendi-
tures on children is likely to be in
housing, the CES’s treatment of mort-
gage payments generates lower expen-
ditures on children. This downward
bias can be thought of as a discount
that all homeowners receive. Obligors
with high incomes tend to own more
expensive homes, so this treatment of
the housing data generates a larger
discount for these obligors.
Determining Proportion of
Expenditures Attributable to
Children
The second step to estimating a house-
hold’s expenditures on children was to
determine in each expense category the
proportion of expenditures attributable
to children. For some categories, such
as clothing, the CES data are reported
separately for children; thus, 100
percent of these expenditures can be
attributed to children. But for other
categories, such as housing, trans-
portation, and food, assumptions must
be made regarding the proportion
attributable to children. The most
common approaches are (1) the
“representative” approach, in which
allocations are based on averages
calculated for children and adults
based on Federal studies; (2) the “per
capita” approach, in which household
expenditures are divided by the number
of family members; and (3) the
“average use” approach, in which
allocations are based on the amount of
a certain commodity that households
with different numbers of children are
observed to use on average, compared
with households without children.
As discussed in a Virginia State
government technical report on the
costs of raising children (JLARC
2001), the choice of which assumption
to use in estimating expenditures on
children could lead to large differences
for two major categories: housing and
transportation. These differences, in
turn, have an effect on estimated
income shares that are used to compute
child support guidelines, especially for
middle- and higher income households.
For those expenditure categories re-
quiring a choice in allocation method,
we compared alternative expenditure
results and explored the reasons for
choosing a particular method.
For housing, we estimated expenditures
for four subcategories of costs: shelter,
utilities, household operations and
household equipment, and furnishings.
Housing is an excellent example of the
difficulty in assigning an expenditure
amount attributable to children. If the
per capita proportions were used, then
33 percent of expenditures in a one-
child household were attributable to
that child, compared with only 1 per-
cent for the average use proportion
(table 1). The 1-percent figure was
computed by JLARC (2001), from
American Housing Survey data, as
the percentage difference between the
estimated house size (1,776 square
feet) of a two-adult household with
one child and the estimated house size
(1,758 square feet) of a two-adult
household with no children. The other
figures for average use in housing were
constructed by using the same method.
Across household sizes, the per capita
approach generated larger expenditures
on children than did the average use
approach. In effect, the per capita
approach provided an upper bound
on the share of housing expenditures
attributable to children while the
average use approach provided a lower
bound. One explanation for why the
average use figures were so small is
that they were based on observed
data on housing size that give no
indication of housing and family
planning decisions. Households may
take longer term views of family size
when they select their homes. When
children are eventually added to the
household, the total housing size may
not increase if the children are living
in extra space that had already been
intended for their use. To estimate
housing expenditures on children, our
preferred approach was to apply the
per capita proportions shown in table
1, mainly because the approach is more
equitable in its assumption that each
household member shares equally in
the use of the home.
Following the method in JLARC
(2001), we defined two types of
transportation costs: fixed vehicle
and variable costs. Fixed vehicle costs
capture spending on new and used cars
and trucks, vehicle financing, and
vehicle insurance. This expense com-
ponent captures the start-up cost of