Family Economics and Nutrition Review [Volume 10, Number 2]

              Feature Articles
2 Factors Contributing to Household Food Insecurity
in a Rural Upstate New York County
Christine M. Olson, Barbara S. Rauschenbach,
Edward A. Frongillo, Jr., and Anne Kendall J(t
Do Food Bars Measure Up? Nutrient Profiles of Fcwd Bar
Versus Traditional School Lunches in the CATCH StUdy y
Anne 0. Garceau, Mary Kay Ebzery, Johanna T. Dwyer,
Theresa A. Nicklas, Deanna H. Montgomery, Lynn V. Hewes,
Paul D. Mitchell, Leslie A. Lytle, and Michelle M. Zive
Research Briefs
31
34
Julia M. Dinkins
Research Summaries
46
48
Regular Items
50
52 Recent Legislation and Regulations Affecting Families
53 Research and Evaluation Activities in USDA
54 Data Sources
55 Journal Abstracts
56 Cost of Food at Home
57 Consumer Prices
UNITED STATES DEPARTMENT OF AGRICUlTURE
SERIALS DEPARTMENT
Dan Glickman, Secretary
U.S. Department of Agriculture
Mary Ann Keeffe, Acting Under Secretary
Food, Nutrition, and Consumer Services
Eileen Kennedy, Executive Director
Center for Nutrition Policy and Promotion
Carol Kramer-LeBlanc, Director
Nutrition Policy and Analysis Staff
Editorial Board
Mohamed Abdel-Ghany
University of Alabama
Rhona Applebaum
National Food Processors Association
Johanna Dwyer
New England Medical Center
Jean Mayer USDA Human Nutrition Research Center
on Aging at Tufts University
Jay Hirschman
Food and Consumer Service
U.S. Department of Agriculture
Helen Jensen
Iowa State University
Janet C. King
Western Human Nutrition Research Center
U.S. Department of Agriculture
C. J. Lee
Kentucky State University
Rebecca Mullis
Georgia State University
Suzanne Murphy
University of California-Berkeley
Donald Rose
Economic Research Service
U.S. Department of Agriculture
Ben Senauer
University of Minnesota
Laura Sims
University of Maryland
Retia Walker
University of Kentucky
Editor
Joan C. Courtless
Managing Editor
Jane W. Fleming
Family Economics and Nutrition Review is
written and published each quarter 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 busness
required by law of the Department.
This publication is not copyrighted. 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.
Family Economics and Nutrition Review is
for sale by the Superintendent of Documents.
Subscription price is $10.00 per year ($12.50
for foreign addresses). Send subscription
orders and change of address to Superinten­dent
of Documents, P.O. Box 371954,
Pittsburgh, PA 15250-7954. (See subscription
form on p. 60.) ·
Original manuscripts are accepted for
publication. (See "guidelines for authors" on
back inside cover). Suggestions or comments
concerring this publication should be
addressed to: Joan C. Courtless, Editor,
Family Economics and Nutrition Review,
Center for Nutrition Pdicy and Promotion, USDA,
112020th St., NW, Suite 200 North Lobby,
Washin~ton. DC 20036. Phone (202) 606-4816.
USDA prohibits discrimination in its programs
on the basis of race, color, national origin,
sex, religion, age, disability, political beliefs,
and marital or familial status. (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 the USDA Office of Communications
at (202) 720-2791 . To file a complaint, write
the Secretary of Agriculture, U.S. Department
of AgriciJture, Washington, DC 20250, or call
1-800-245-{;340 (voice) or (202) 720...1127
(TOO). USDA is an equal employment
opportunity employer.
Center for Nutrition Policy and Promotion
Feature Articles
2
18
Factors Contributing to Household Food Insecurity
in a Rural Upstate New York County
Christine M. Olson, Barbara S. Rauschenbach,
Edward A. Frongillo, Jr. , and Anne Kendall
Do Food Bars Measure Up? Nutrient Profiles of Food Bar
Versus Traditional School Lunches in the CATCH Study
Anne 0. Garceau, Mary Kay Ebzery, Johanna T. Dwyer,
Theresa A. Nicklas, Deanna H. Montgomery, Lynn V. Hewes,
Paul D. Mitchell, Leslie A. Lytle, and Michelle M. Zive
Research Briefs
31 Households Receiving Food or Meals as Pay
MarkLino
34 Food Preparers: Their Food Budgeting, Cost-Cutting, and
Meal Planning Practices
Julia M. Dinkins
Research Summaries
38 The State of Nutrition Education in USDA
43 Size and Growth of the Nutritionally Improved Foods Market
46 Income of the Elderly, 1994
48 The Home Computer Market
Regular Items
50 Charts From Federal Data Sources
52
53
54
55
56
57
Recent Legislation Affecting Families
Research and Evaluation Activities in USDA
Data Sources
Journal Abstracts
Cost of Food at Home
Consumer Prices
Volume 10, Number 2
1997
2
Feature Articles
Factors Contributing to
Household Food Insecurity
in a Rural Upstate New York
County
Christine M. Olson
Barbara S. Rauschenbach
Edward A. Frongillo, Jr.
Anne Kendall
Division of Nutritional Sciences
Cornell University
To identify factors contributing to household food insecurity in a rural
county in upstate New York, two personal interviews were conducted with
193 women who were between the ages of 20 and 40 years, had less than
16 years of education, and had children living at home. Data were collected
on sociodemographic characteristics, risk factors for food insecurity, food
program participation, the Radimer/Cornell hunger and food insecurity
measures, and household food supplies. Regression analyses and tree­based
partitioning were used. Variables significantly (p<0.05) contributing to
food insecurity were being a single parent, lack of savings, larger household
size, having unexpected expenses, adding $50 or more to food stamps to
purchase sufficient food, and low food expenditures. Variables contributing
to low levels of household food supplies were low educational level, low
food expenditures, not having a vegetable garden, and not receiving free
milk, eggs, and meat. These results will help improve development and
targeting of interventions to alleviate food insecurity.
[£] ood insecurity is now a
recognized public policy
concern for food-rich
countries such as the
United States, as well as a development
concern for poorer countries around
the world ( 11, 24). Furthermore, a
consensus is emerging around the
American Institute of Nutrition (AIN)
definition of food insecurity as "when­ever
the availability of nutritionally
adequate and safe foods or the ability
to acquire acceptable foods in socially
acceptable ways is limited or uncertain"
( 1 ). This definition is becoming widely
used for policy relevant nutrition research
in the United States and is consistent
with the definition of food insecurity
used in this paper.
Using both qualitative and quantitative
research methods, Radimer and colleagues
have developed a definition of food
insecurity, a conceptual framework,
Family Economics and Nutrition Review
and the Radimer/Comell hunger and
food insecurity measures (see box, p. 4)
relevant for food-rich countries ( 14, 15,
16). The validity of these measures for
identifying groups of households that
are experiencing food insecurity is now
established (8). Very recent research
indicates that the household-level food
insecurity measure has a sensitivity of
89 percent and a specificity of 63 percent
that can be improved to 71 percent with
the elimination of one item in the measure
(6). Thus, the household-level food
insecurity measure correctly identifies
89 percent of the households that are
truly insecure as insecure and correctly
identifies 71 percent of the households
that are secure as secure.
Given the previous lack of a validated
measure of household-level food
insecurity, it is not surprising that few
studies have examined the factors con­tributing
to food insecurity in populations
living in relatively food-rich countries.
Recently, Campbell (2) elaborated a
conceptualization of food insecurity and
its risk factors. She defines risk factors
for food insecurity as anything that limits
either household resources (money,
time, information, health, etc.) or the
proportion of those resources available
for food acquisition.
The conceptual framework outlines the
relationship of household resources to
food acquisition and food insecurity. The
present study draws on this conceptual
framework and examines social and
demographic characteristics that influ­ence
a household's level of resources,
as well as the household's level of
resources in relation to food insecurity.
In addition, this study examines variables
that measure food acquisition by the
household and are influenced by socio­demographic
characteristics and economic
resources and relate to household food
1997 Vol. 10 No.2
supplies and the experience of food
insecurity.
For this study, a rural population was
selected because, as Deavers and Hoppe
( 4) point out, the overall poverty rate
is higher in rural than urban areas. In
addition, the rural poor have fared
relatively badly since 1980 as the
economic performance of rural areas
has lagged behind that of the rest of
the Nation. In 1993, when the study
reported here was conducted, the non­metropolitan
poverty rate was 17.2
percent while the metropolitan poverty
rate was 14.6 percent (25 ).
Morris, Neuhauser, and Campbell ( 12)
have examined three factors that may
limit food acquisition and therefore
contribute to food insecurity in rural
areas: Limited supermarket availability,
limited food item availability, and
higher relative costs of the USDA
Thrifty Food Plan (TFP) market basket
of foods. Using a random sample of
persistently poor rural counties, the
investigators found 3.8 supermarkets
per county in rural America versus 29 in
urban America. Fresh fruits, vegetables,
and meats were very limited in the small
and medium stores that are more common
in rural areas. The average cost of the
TFP market basket was $102 in small
and medium stores and $81 in super­markets
in rural areas. The picture that
emerges from these findings is one of
limited access to supermarkets, and as
a consequence, decreased availability
of fresh foods, an increased cost of food,
and ultimately an increased risk for food
insecurity.
Additional factors may contribute to
food insecurity in rural areas. Rank and
Hirschi ( 17) have shown that qualified
families in rural areas are much less
likely to participate in food assistance
programs such as food stamps than their
urban counterparts. These researchers
found adverse attitudes toward welfare
and lack of accurate information as two
of the underlying mechanisms explaining
low food stamp participation rates in
rural areas. These studies indicate that
food acquisition may be constrained in
rural areas and that these constraints
may increase households' risk of food
insecurity.
The objectives of the research reported
here were: (1) to identify the social,
demographic, and economic charac­teristics
of households that contribute to
food insecurity; (2) to identify the food
acquisition characteristics of households
that contribute to food insecurity; and
(3) to analyze the interrelationships
between these two sets of factors, as
well as the use of coping tactics by
food insecure households. Overall, this
research aimed to increase the general
understanding of food insecurity in
order to facilitate the development of
more effective interventions to address
the problem and to improve targeting
of interventions to subgroups in the
population with the problem.
Methods
Population
The study took place in a rural county
of upstate New York that had a popula­tion
of 60,517 in 1990 ( 5 ). Nearly 77
percent of the population of this county
live in places with fewer then 2,500
people. In 1990, the unemployment rate
of the county was 5.8 percent; per capita
income was $15,503; and the percentage
of families in poverty was 12.6 percent.
This county was below both the mean
unemployment rate and the poverty
rate for similar counties in upstate
New York.
3
Selection of the Sample
A survey of women with children living
in their household was conducted in this
county between January and July 1993.
A sample of approximately 200 women
was desired because previous research
found statistically significant relationships
between risk factors and food insecurity
with a sample size of 189 ( 16 ). The
sampling frame was a 1989 health census
of the county that had a participation
rate of 86 percent. Women over the age
of 40 and those with 16 or more years
of education were excluded, resulting
in 3,433 women who were eligible for
the study. Since it was anticipated that
changes in the county's population had
taken place since the health census was
completed, a pool of 639 women was
selected from the census.
Six strata were formed based on the
demographic characteristics available
from the census most strongly associated
with low socioeconomic status: first,
whether potential subjects did or did
not have a telephone; and then, whether
they had private health insurance, Medi­caid
insurance, or no health insurance.
Each of the six strata was further stratified
into five age groups: 15-19,20-24, 25-29,
30-34, and 35-39 years.
Fifty-two percent of the women (331)
could not be located within the county
despite intensive efforts. The remaining
308 women were contacted by telephone
or, for those with no phones, at their
homes to request their participation
and set up interviews. Two hundred
women agreed to participate in the
survey. Refusal rates were 18 percent
in the strata presumed to be the lowest
income group (those having no telephone
and either Medicaid or no health insur­ance),
40 percent in the 15 intermediate
strata, and 32 percent in the 5 highest
strata (those with a telephone and
4
The Radimer/Cornell Hunger and Food Insecurity Items
1. I worry whether my food will run out before I get money to buy more.
2. The food that I bought just didn't last, and I didn't have money to get more.
3. I ran out of the foods that I needed to put together a meal and I didn't have
money to get more food.
4. We eat the same thing for several days in a row because we only have a few
different kinds of food on hand and don't have money to buy more.
5. I am often hungry, but I don'teat because I can't afford enough food.
6. I eat less than I think I should because I don't have enough money for food.
7. I can't afford to eat properly.
8. My child(ren) is/are not eating enough because I just can't afford enough food.
9. I know my child(ren) is/are hungry sometimes, but I just can't afford more
food.
10. I cannot give my child(ren) a balanced meal because I can't afford that.
Response categories for items 1 through 10 are "not true," "sometimes true," or
"often true." "Sometimes true" and "often true" are grouped together for analytical
purposes to indicate a positive response or the presence of food insecurity.
private health insurance). Because only
7 of the 200 women fell into the 15-19
age category, they were dropped from
the analysis.
Data Collection
Each respondent was interviewed twice
in her home. During the first interview,
a questionnaire was administered and an
inventory of household food supplies
was conducted by trained field workers.
The questionnaire obtained information
on sociodemographic characteristics,
methods of obtaining food, food pro­gram
participation, household expendi­tures,
and the Radimer/Cornell hunger
and food insecurity items (see box). At
the second interview, approximately 3
weeks later, the household food inven­tory
was repeated.
The survey instrument was pretested in
a sample of 20 low-income women and
afterwards a number of categories on
the food inventory instrument were
revised to better differentiate household
food supplies. The study protocol was
approved by the Cornell University
Human Subjects Committee and informed
consent was given by all respondents
prior to participation in the study. Each
respondent received $20 as compensa­tion
for participation.
Measurement of Dependent Variables
This study used two dependent measures
of food insecurity. The first was the
previously validated Radimer/Cornell
hunger and food insecurity measures
index ( 8 ). Since household-level food
insecurity was the focus of this study,
any household that had a positive response
Family Economics and Nutrition Review
to one or more of the questions was
defined as insecure. The remaining
households were defined as food secure.
The second dependent measure of food
insecurity was household food supplies
as measured by a household food
inventory. Food supplies are potentially
a physical measure of food insecurity
and have been shown to be strongly as­sociated
with food insecurity (8). Since
in this population only 9 percent of all
food expenditures are for food eaten
outside of the home, household food
supplies seem to reasonably represent
the food available for consumption.
The instrument used to measure house­hold
food supplies was based on methods
used by Sanjur et al. ( 19) and Crockett
eta!. ( 3 ). Field workers coded the pres­ence
of 51 food items in the household
into one of four categories, with zero
indicating none of the food was present
and three indicating a large amount was
present. Item-specific response categories
were determined based on the weight or
volume of each item as purchased and
judgments of differences that would be
meaningful and differentiate those with
depleted food supplies from those with
replete food supplies. These scores were
then summed over the 51 items, and the
two inventories were averaged to create
a measure of food supplies that could
range from 0 to 153. The food inventory
was normally distributed, with a range
of 19 to 115.5 and a sample mean of
71.06.
Measurement of Contributing Factors
The sociodemographic and economic
risk factors contributing to food insecurity
considered in this analysis were (table 1):
Annual income (in six categories: <$5,000,
$5,000-$10,000,$10,000-$15,000,
$15,000-$20,000, $20,000-$25,000,
>$25,000), whether the income in past
1997 Vol. 10 No.2
year was less than usual, whether income
dropped over the year, presence of
monthly variation in income, household
size (number of people eating from the
same food supply), respondent's educa­tional
level, whether the household was
headed by a single parent, respondent's
employment status and spouse's em­ployment
status, presence of savings,
and home ownership.
Food acquisition variables considered
were: Receipt of food stamps, adding
more than $50 to food stamps, 1 total
household expenditures (sum of rent/
mortgage, school and real estate taxes,
utilities, car payments, car repair costs,
car insurance and gasoline expenses,
daycare expenses, medical insurance
and other medical expenses, and food
expenses for food eaten at home and
away from home), food, housing, and
car expenditures (each expressed sepa­rately
as a dollar amount and as a percent
of total expenditures), presence of un­expected
expenses within the last year,
presence of medical expenses (other
than insurance) within the last year,
limits on store choice because of trans­portation
and/or store availability, use
of a food buying club, whether food was
obtained from vegetable gardening and
hunting or fishing, frequency of receipt
of free milk, eggs, or meat, and
frequency of shopping.
The coping strategies considered were:
Frequency of borrowing money for
food, eating with friends and relatives,
food being brought by friends and family
to the respondent's household, using a
food pantry, and whether commodity
foods were used.
1This variable measured the sufficiency of food
stamps to meet the family's food needs.
Food supplies are
potentially a physical
measure of food
insecurity and have
been shown to be
strongly associated
with food insecurity.
5
Table 1. Characteristics of food secure and food insecure households
Food secure Food insecure
Characteristic
Sociodemographic and economic factors
Income
<$5,000
$5,000- $10,000
$10,000- $15,000
$15,000- $20,000
$20,000- $25,000
>$25,000
Income last year less than usual
Income dropped in year
Income same monthly
Household size
Education
Less than high school
High school graduate
Some college or technical training
College graduate
Single-parent household
Respondent employed
Spouse employed
Have savings
Own or buying home
Food acquisition variables
Receive food stamps
Add $50 or more to food stamps
Household expenditures
Food expenditures
Housing expenditures
Car expenditures
Food/total expenditures
Housing/total expenditures
Car/total expenditures
Unexpected expenses in last year
Medical expenses in last year
Shop at store because
6
Only store in area
No transportation
(n = 90)
Percent or mean
4
7
12
9
17
51
16
33
23
4.37
12
40
25
22
8
71
82
69
76
6
2
$17,617
$4,657
$6,435
$4,779
0.28
0.38
0.25
44
82
19
3
(n = 103) p value
<0.001
10
27
14
11
12
25
26 <0.05
48 0.01
38 0.005
4.30 <0.10
<0.01
19
46
27
8
29 <0.001
59 <0.05
64 <0.001
28 <0.001
61 <0.01
33 0.001
20 <0.001
$13,613 <0.001
$3,881 <0.01
$5,438 <0.05
$3,056 <0.005
0.32 <0.05
0.41 NS
0.20 <0.01
56 <0.05
65 <0.001
20 NS
8 <0.05
table continues
Family Economics and Nutrition Review
Table 1. Characteristics of food secure and food insecure households
Food secure Food insecure
Characteristic (n= 90) (n = 103) p value
Percent or mean
Belong to buying club 15 16 NS
Vegetable garden for food 63 55 NS
Hunt or fish for food 53 53 NS
Receive free eggs, milk, or meat 27 21 NS
Frequency of major grocery shopping <0.05
Once a week or more 34 23
Once every 2 weeks 42 38
Once every 3 weeks 6 8
Once a month 17 29
Coping strategies
Frequency of eating meals at friends or family NS
Never 11 14
Hardly ever 30 33
Less than once a month 18 8
Once a month 22 22
More than once a month 18 22
Frequency of family or friends bringing food 0.05
Never 56 48
Hardly ever 28 32
Less than once a month 7 7
Once a month 6 3
More than once a month 2 10
Frequency of borrowing money for food from
family or friends 0.001
Never 87 57
Hardly ever 13 30
Less than once a month 0 7
Once a month 0 4
More than once a month 0 3
Frequency of using a food pantry 0.001
Not applicable 44 37
Never 38 24
Hardly ever 15 29
Less than once a month 2 1
Once a month 1 10
Use surplus or commodity foods 39 60 0.001
NS=p> .10.
1997 Vol.JONo. 2 7
8
Tree-based
partitioning ...
selects variables in a
sequence, choosing
at each step the
independent variable
that best distinguishes
the classes of a
categorical dependent
variable (classification
tree analysis) or the
level of a continuous
variable (regression
tree analysis).
Statistical Analysis
The first step in the statistical analysis
was to compare the food insecure and
food secure households on each of the
independent variables. Chi-square
analysis was used for categorical
variables and t-tests for continuous
variables.
The second step was to select the best
predictors from the many variables
available. The selection process was
based on Campbell's (2) conceptual
framework, other theoretical justifica­tions
outlined below, and statistical
analysis. Using logistic stepwise regres­sion
within SAS (20 ), specific variables
were selected from within each of the
major conceptual areas of Campbell's
framework as follows: (1) the social,
economic, and demographic variables
only; (2) the food acquisition variables
only; and (3) both the social, economic,
and demographic and food acquisition
variables.
A variable was selected to be added and
stay in the model if it had an F statistic
significant at the 0.05 level. If a variable
was selected in any of the three analyses,
it was included in the final models.
Linear stepwise regression was used
similarly to select a subset of the best
predictors of household food supplies.
Any variable chosen for food insecurity
or household food supplies was included
in the final models for both dependent
variables.
In the variable selection analyses, house­hold
financial resources were operation­alized
both as income and total household
expenditures, since in low-income
families expenditures may more accu­rately
characterize financial resources
than income (21 ). When the total house­hold
expenditure variable was chosen
for inclusion in the model, income was
not included. Similarly, the food expendi­ture
variable was operationalized both
as total annual food expenditures and
food expenditures as a proportion of
total expenditures. When food expendi­tures
as a proportion of all household
expenditures was included in the model,
total food expenditures was not included.
The final models presented here include
income and total annual food expendi­tures.
Income level and household size
were included in the final models even
if these variables did not survive the
selection procedure.
The third step was to create the final
logistic and linear regression models.
To address the objective of analyzing
the interrelationships among subsets
of variables, four logistic and linear
regression models were estimated for
each of the dependent variables (food
insecurity and household food supplies,
respectively) using the selected variables:
( 1) the subset of sociodemographic and
economic variables alone; (2) the subset
of food acquisition variables alone; (3)
the sociodemographic and economic
and the food acquisition variables to­gether;
and ( 4) the variables in model
number three with the addition of the
coping strategies.
To identify characteristics of households
that contribute to food insecurity, results
from the logistic regression model three
are expressed as odds ratios with associ­ated
95-percent confidence intervals
(CI). An odds ratio (OR) is a measure
of association and indicates the prob­ability
that a household with a certain
characteristic (or value on the inde­pendent
variable) will be food insecure
divided by the probability that it will
not be food insecure (7). The ratio
resulting from logistic regression analysis
compares the odds for two different
values of the independent variable and
Family Economics and Nutrition Review
can take on any value from 0 to infinity
with a value greater than I indicating
that the risk of being food insecure is
greater when the household has the
characteristic (positive association). A
value between 0 and I indicates that the
risk of being food insecure is less when
the household has the characteristic
(negative association). An odds ratio·
was considered statistically significant
if I was not in the CI.
Results from the linear regression
model three are expressed as regression
coefficients with 95-percent confidence
intervals. The coefficient resulting from
linear regression can take on a value from
negative to positive infinity. Negative
values indicate an inverse or negative
association and positive values indicate
a positive association of the variable
with household food supplies. A coeffi­cient
is significant if 0 is not in the CI.
Finally, to provide insight into possible
interactions among the most useful vari­ables
for distinguishing food secure and
insecure households and for predicting
food supplies, tree-based partitioning
analysis called S-Plus was used (23, 26).
Tree-based partitioning is particularly
useful when complicated interactions
that cannot be modeled by usual regres­sion
methods are expected. This statisti­cal
procedure selects variables in a
sequence, choosing at each step the
independent variable that best distin­guishes
the classes of a categorical
dependent variable (classification tree
analysis) or the level of a continuous
variable (regression tree analysis). An
independent variable can be included in
the tree more than once and may have
different cut-off points each time.
After the tree is constructed, it can be
pruned using various criteria to create a
simpler, more easily interpretable, and
1997 Vol. 10 No.2
more generalizable tree. We used classi­fication
tree analysis to construct a tree
for food insecurity and regression tree
analysis to create a tree for household
food supplies. We considered only the
independent variables included in the
final logistic and linear regression models
in our original tree construction. In this
paper, we show the trees down to the
level of variables found to be statistically
significant in the final models of the
logistic and linear regression analyses.
Results
Table I shows the characteristics of food
secure and insecure households for each
of the independent variables in this study.
On the sociodemographic and economic
factors, the two groups differed signifi­cantly
in the expected direction on all
independent variables. For the food ac­quisition
variables, again the two groups
differed in the expected direction on
many of the variables. For example, food
insecure households were more likely
were less than those of food secure
households.
The two groups did not differ from each
other on several strategies for acquiring
food at low cost such as belonging to a
food buying club, vegetable gardening,
hunting and fishing, and receiving eggs,
milk, and meat from friends or relatives
free or as in-kind pay for agricultural
work. Interestingly, approximately 20
percent of both groups reported they
shopped where they did because it was
the only store in the area; and while
transportation constraints on food
shopping were reported by substantially
fewer respondents, the two groups
differed significantly on this variable.
On the coping strategy variables, food
insecure households made significantly
more frequent use of all means listed
except eating meals with friends and
family. Food insecure households were
significantly more likely to have used
surplus or commodity foods than food
secure households.
to receive food stamps and to add $50
or more per month to their food stamps
to buy food for the household, but their
annual dollar expenditures for food
Table 2 presents the results of the models
for food insecurity that included various
subsets of variables. The model with
Table 2. Proportion of variance accounted for by models with various
subsets of variables
Food insecurity
area under ROC1
~odel curve
Sociodemographic and economic factors 0.77
Food acquisition variables 0.74
Sociodemographic, economic, and
food acquisition 0.81
Sociodemographic, economic, food
acquisition, and coping strategies 0.83
I Receiver operating characteristic.
Food s'fplies
R
0.26
0.31
0.41
0.43
9
only the subset of the sociodemographic
and economic variables had an area under
the receiver operating characteristic
(ROC) curve of 0. 77. The area under an
ROC curve can be interpreted like an
R2. The ROC area ranges from 0.5
(i.e., chance) to 1.0 and refers to the
probability that the logistic regression
model correctly orders pairs of food
secure and insecure households. When
the selected food acquisition variables
were considered separately, the area
under the ROC curve was 0.74 and with
both sets of variables, the value was
0.81. The addition of the selected coping
strategies resulted in an area under the
ROC curve of 0.83, not a substantial
increase. Sociodemographic and eco­nomic
factors accounted for almost
the same amount of variance in the out­come
as the food acquisition variables,
and the two sets taken together did not
account for considerably more of the
variation in food insecurity than did
either set alone.
Table 2 also presents the results from
the linear regression analysis for house­hold
food supplies. The subset of socio­demographic
and economic variables
explained 26 percent of the variance in
food supplies and the food acquisition
variables explained 31 percent of the
variance. When both subsets of variables
were included together, more of the
variance was explained, 41 percent, than
with each separately, indicating that they
both make independent contributions
to the explanation of food supplies.
The addition of coping strategies added
only 2 percentage points to the explained
variance. Based on the two sets of results,
coping strategies were not included in
the final models identifying factors
contributing to food insecurity.
10
Table 3. Odds ratios and 95-percent confidence intervals for factors
contributing to food insecurity
Confidence limits
Variable 'Odds ratio Lower Upper
Sociodemographic and
economic factors
Income1
Savings
Own/buy home
Income same in year
Education1
Single parent
Household size
Respondent employed
Food acquisition variables
Receives food stamps
Add $50 or more to food stamps
Medical expenses
Unexpected expenses
Vegetable gardening
Free milk/eggs
Food expenditures
*Statistically significant at p < 0.05.
1Treated as continuous variables in the analysis.
Table 3 presents the odds ratios with
95 percent CI for the sociodemographic
and economic factors as well as the food
acquisition variables associated with
food insecurity. These were derived
from the multivariate logistic regression
analysis of model 3 using the Radimer/
Cornell measure of food insecurity as
the dependent variable. Among the
sociodemographic and economic factors,
women with savings were much less
likely than those without to report food
insecurity (0R=0.32, CI=0.17, 0.61).
Women in single-parent households
0.988 0.788 1.238
0.321 0.168 0.611 *
1.103 0.550 2.212
1.202 0.635 2.277
0.849 0.609 1.182
3.707 1.355 10.139*
1.363 1.027 1.810*
0.894 0.465 1.716
0.646 0.181 2.308
6.333 1.464 27.400*
0.771 0.345 1.723
2.317 1.269 4.231*
0.918 0.477 1.767
0.862 0.433 1.715
0.973 0.957 0.990*
were more likely to be food insecure
(0R=3.71, CI=l.36, 10.14). Women in
larger households were also more likely
to be food insecure (OR= 1.36, CI= 1.03,
1.81). Among the food acquisition variables,
those women who added $50 or more
money to food stamps were more likely
to be food insecure (OR=6.33, Cl=1.46,
27.4) as were women whose households
experienced unexpected expenses within
the last year (0R=2.32, Cl=l.27, 4.23).
Food expenditures were lower in food
insecure households (OR=0.97, Cl=0.96,
0.99).
Family Economics and Nutrition Review
Table 4. Regression coefficients and 95-percent confidence intervals
for factors contributing to household food supplies
Variable
Regression
coefficient
Confidence limits
Lower Upper
Sociodemographic and
economic factors
INTERCEPT 35.936 19.988 51.884
Income
Savings
Own/buy home
Income same in year
Education
0.364
5.610
5.693
1.285
4.134
4.980
1.421
-1.542 2.270
--0.099 11.318
--0.180 11.566
-4.170 6.741
1.368 6.899*
-2.947 12.908
--0.856 3.699
Single parent
Household size
Respondent employed --0.639 -6.113 4.836
Food acquisition variables
Receives food stamps 2.969 -7.906 13.844
Add $50 or more to food stamps
Medical expenses
-7.770 -16.570 1.031
5.235 -1.355 11.825
Unexpected expenses
Vegetable gardening
--0.293 -5.342 4.755
Free milk/eggs
Food expenditures
*Statistically significant at p < 0.05.
Table 4 presents the regression coeffi­cients
(RC) and 95 percent CI for house­hold
food supplies. Education was the
only social, demographic, or economic
factor associated with food supplies.
Women with more education had
significantly larger food inventories
(RC=4.13, Cl=l.37, 6.90). Among the
food acquisition variables, those women
who spent more on food (RC=0.24,
Cl=0.11, 0.37), had vegetable gardens
(RC=8.15, Cl=2.67, 13.64), or received
1997 Vol. 10 No. 2
8.154 2.669 13.639*
8.798 2.985 14.612*
0.244 0.114 0.374*
free milk, eggs, or meat (RC=8.80,
Cl=2.98, 14.61) had larger household
food inventories than those without these
characteristics. Several other variables
approached statistical significance
(p<0.10). Having savings and owning a
home approached significance as factors
related to greater food supplies. Women
who added $50 to food stamps had
smaller household food supplies than
those who did not do so.
The interactions between the independent
variables as well as their relative impor­tance
are indicated in the results from
the tree analysis. Figure 1 presents the
classification tree for household food
insecurity. Only a portion of the full tree
is presented. The full tree is available
from the authors. The tree had an over­all
rnisclassification rate of 16 percent.
This degree of misclassification
allowed for the production of a tree
that was understandable and acceptably
accurate.
As can be seen, if the household had
savings, it was much less likely to be
food insecure than if it didn't (31 per­cent
vs. 71 percent). Among the group
with no savings, adding $50 or more
to food stamps was the next variable
selected. Ninety-five percent of those
who added this amount of money or
more to their food stamps to feed their
family for the month were food insecure,
whereas 65 percent of those who didn't
were insecure. Among both of these
groups, annual food expenditures was
the next variable selected. Generally,
lower food expenditures were associ­ated
with greater food insecurity. To
continue on down the tree, among those
who did not add $50 or more to food
stamps and had annual food expendi­tures
of less than $3, 192, and had
unexpected expenses, 81 percent
were food insecure.
To move to the other side of the tree
and examine those who had savings
and were food insecure, annual food
expenditures was the first variable
selected. Among those with annual
expenditures less than $6,630, 36 per­cent
were food insecure whereas among
those with greater expenditures, no one
was food insecure (fig. 1). In the group
with expenditures less than $6,630, 55
percent of those with a household size
11
Figure 1. Classification tree for food insecurity {Radimer/Cornell Measure)
12
l
Not add $50
to
food stamps
N = 83
65%
I
Food
I
No savings
N = 103
71% *
N = 186
53%
I
Add $50
to
food stamps
N=20
95%
I
I
I
Savings
N =83
31% *
Food
expenditures
<$6,630
N =72
36%
Food
expenditures
~$6,630
N = 11
0%#
expenditures
<$3,192
Food
expenditures
~$3,192
Food
expenditures
<$5,360
Food
expenditures
~$5,360
Household
size
<4.5
Household
size
~4 . 5
N=22
55%
I
I
No
unexpected
expenses
N=36
58%
l
I
Nota
single Single
parent parent
N =24 N = 12
46% 83%
N =73
70%
I
I
N = 10
30%
Unexpected
expenses
N = 37
81%
No free
I
Free
N = 15
100% #
milk, eggs, milk, eggs,
meat meat
N =28 N=9
89% 56%#
N =5
80%#
Household
size
<3.5
N = 21
48%
N =50
28%
I
I
Food Food
I
Household
size
~3.5
N =29
14%
expenditures
<$2,310
N=5
80% #
expenditures
~$2,310
N = 16
38%
I
I I
Education Education
<some ~some
college college
N = 13 N =9
38% 78%#
* = Percent food insecure
# = terminal node
Family Economics and Nutrition Review
greater than 4.5 were insecure. And
following along those in this group,
among those with some college or
greater education, 78 percent were
insecure.
Regression tree analysis was used to
identify the characteristics of house­holds
with higher food supplies. The
first variable selected was annual food
expenditures. Fifty-five of 180 house­holds
spent less than $3,192 annually
on food and had a mean inventory score
of 61.1 compared with a score of74.8
for those who spent more than that
amount. Overall, the important variables
in predicting household food supplies
among those with annual food expendi­tures
ofless than $3,192 were home
ownership and income level. Owning
a home and having an income above
$20,000 were consistently associated with
larger food supplies, 69.0 versus 51.5
and 66.2 versus 47.8, respectively.
Among the food insecure with food
expenditures greater than $3,192, educa­tional
level of the respondent, whether
she added $50 or more to food stamps
to feed the family for the month, and
whether the household had a vegetable
garden were the important variables.
Those respondents with greater than
some college had a mean food inventory
score of 82.5 versus 70.3 among those
with less education. Among those with
less education, respondents who added
$50 or more to their food stamps had
a mean inventory score of 57.1 versus
73.1 for those who didn't. Among those
with more education, respondents who
did not have a vegetable garden had a
mean score of73.3 versus 88.5 for
those who did.
1997 Vol. 10 No. 2
Discussion
This paper is among the first to examine
factors contributing to food insecurity
using a validated direct measure of food
insecurity, as well as a physical measure
of food insecurity, household food
supplies. The descriptive results in
table 1 are similar to those in a recent
paper from Rose, Basi otis, and Klein
( 18) describing the correlates of food
insufficiency from USDA's 1989-91
Continuing Survey of Food Intakes by
Individuals (CSFII). They found higher
rates of food insufficiency among house­holds
with these characteristics: Low
income, renting a home, single-head
of household, low educational level,
six or more people in the household,
and minority race or ethnicity. These
authors carefully point out that their
results are descriptive and do not
control for underlying factors.
A positive aspect of the multivariate
logistic and linear regression analyses
presented in this paper is the control for
underlying factors. When this was done,
a fairly consistent picture of the factors
contributing to food insecurity e~erges
across the two dependent measures.
Measures of wealth, such as having
savings and owning a home, were related
to decreased risk of food insecurity.
Economic insecurity and limited income
earning potential, operationalized as
being in a single-parent household and
having a lower educational level, were
related to increased risk of food insecurity.
Lower levels of food expenditures and
having unexpected expenses were con­sistently
associated with increased risk
of food insecurity. This latter finding
indicates that it is not only the level of
household financial resources that is
important for food security, but it is
also their certainty.
.. .it is not only the
level of household
financial resources
that is important for
food security, but it
is also their certainty.
13
Figure 2. Regression tree for household food supplies
I
I
Not own home
N =25
Mean= 51 .5
I
I
Food
expenditures
<$3,192
N=55
Mean =61 .1
I
Food
Own home
N =30
Mean =69.0
I
Food
N = 180
Mean =70.6
I
l
Education
<some college
N =79
Mean =70.3
I
I
Food
expenditures
~$3,192
N = 125
Mean =74.8
I
I
Education
~ some college
N =46
Mean =82.5
I
I
Income Income expenditures expenditures
Add$50
to
Not add $50
to
food stamps
N=65
Mean =73.1
No vegetable
gardening
N = 18
Mean =73.3
Vegetable
gardening
N=28
Mean =88.5
<$20,000 ~$20,000 <$1,470 ~$1,470
N =20 N =5 N=6 N=24
Mean =47.8 Mean =66.2 # Mean =55.5 # Mean =72.4
food stamps
N = 14
Mean= 57.1
Senauer, Asp, and Kinsey (21, p. 218)
state, "Lack of food security and inade­quate
diets among the poor are primarily
a direct result of inadequate income to
buy sufficient food." Income, operation­alized
as a six-category variable, was
not significant in the regression analysis.
We believe this may have happened for
two reasons: first, food expenditures are
a more immediate (proximal) predictor
of food insecurity and the level of food
expenditures is determined by income.
So when the food expenditure variable
is in the model, it may mask any effect
of income on food insecurity. Second,
this finding may be a result of the way
income was measured in this study.
14
Support for this assertion is that when
total household expenditures was used
as the variable to operationalize the
concept of household financial resources,
this variable was significant (p < 0.05),
and food expenditures as a proportion of
total expenditures was not significant.
Senauer, Asp, and Kinsey (21) note that
total consumer expenditures may be a
better indicator of a household's perma­nent
income than current annual income,
especially in low-income households.
So household income, whether current
or permanent, is very likely an impor­tant
influence on household-level food
insecurity even if current income was
not significant in the multivariate
analyses shown in this paper.
Among the food acquisition variables
examined, total annual food expenditures
was strongly and consistently associated
with food insecurity and food supplies.
Food insecure households spent about
83 percent of what food secure house­holds
spent on food. Food expenditures
accounted for 32 percent of total house­hold
expenditures for food insecure
households compared with 28 percent
for food secure households. In analyzing
food expenditures for 1980 to 1988,
USDA analyst James Blaylock ( 13)
has shown that food expenditures among
the poorest one-fifth of Americans have
declined 13.1 percent while they have
grown by 2.7 percent in the wealthiest
one-fifth of the population.
Family Economics and Nutrition Review
During this period, growth in annual
income level was stagnant for the
poorest quintile, so that in 1990, this
group was spending 42 percent of their
income for food, compared with 14 per­cent
for the average household (9). The
food insecure households in this study
may well be spending as much as they
can on food, and this amount is not
sufficient to make them food secure.
Lino ( 10) recently found food stamps
to be the most common income source
among poor families with children.
In his study, 69 percent received food
stamps, and the program provided one­fifth
of these households' annual income.
Lino ( 10) states, "Probably more than
any other program, food stamps pro­vides
a safety net for poor households."
Although participation in the food
stamp program is very low in this study,
Lino's contention is supported by the
consistent association of the insuffi­ciency
of food stamps for meeting
family food needs and food insecurity.
In this sample, among households who
received food stamps, those who added
$50 or more in cash to their food stamps
to buy food for the household for the
month were more likely to be food
insecure and to have lower household
food supplies. We are inclined to evalu­ate
this finding as real not only because
of the consistency in the result across
the two methods but also because we
did the analysis with the independent
variable operationalized as "whether
food stamps lasted the whole month"
and found the same result.
An interesting finding from this study
that may only be relevant to food access
in rural areas is the positive association
of vegetable gardening with household
food supplies. Likewise, receiving milk,
1997 Vol. 10 No.2
eggs, and meat free or as in-kind payment
for agricultural labor had a positive
association with household food supplies.
This finding points to the importance of
household production in food security.
However, Shotland and Loonin (22)
note that family gardens may have only
limited potential for solving problems
of food insecurity in this population
subgroup because of the limited land
available for gardening and the high
cost of inputs such as seed, fertilizer,
and insecticides. Poor families may be
reluctant to risk their limited financial
resources on a garden.
In addition to identifying factors contrib­uting
to food insecurity, this research
aimed to understand how these factors
interrelate, thus providing insight into
the nature of food insecurity. Results
from both the staged regression analysis
and the tree analysis provide useful
insights. Clearly, the sociodemographic
and economic characteristics of house­holds
explain a substantial proportion
of the variance in food insecurity
measured both ways. These characteristics
will be helpful in identifying segments
of the population to target for inter­ventions.
But the results also show that
food acquisition factors explain addi­tional
variance, particularly in house­hold
food supplies. Two food acquisition
variables, food expenditures and having
to add $50 or more to food stamps,
appear to be particularly important
as they enter the tree analysis near the
top of the tree.
Coping strategies did not add substan­tially
to the proportion of variance
explained in either dependent variable
when the other two groups of variables
(sociodemographic factors and food
acquisition variables) were in the
model. Thus, coping tactics appear
to be coincident with food insecurity
rather than factors that contribute to or
protect against food insecurity. More
research on how coping tactics relate to
both food insecurity and its risk factors
and consequences is warranted.
The tree analysis indicates that with
information on only a very few variables,
a program manager could be confident
that a household was food insecure.
Ninety-five percent of the households
that had no savings and added $50 or
more to food stamps were food insecure
(fig. 1). Furthermore, the tree analysis
offers insight into the relative impor­tance
of a variable such as single parent­hood
that was identified as significant in
the regression analysis. While being
a single parent is a risk factor for food
insecurity, it is most important for those
having no savings, who don't add $50
or more to their food stamps, have low
food expenditures, and no unexpected
expenses (fig. 1).1t does not appear to
be a risk factor for those households
with savings and higher levels of food
expenditures.
The results found here may be unique to
counties similar to the one studied here.
This county is typical of "rural-urban"
counties in New York State according
to Ebert's ( 5) classification scheme.
This type of county has a population
of less than 200,000 people, the largest
place is greater than I 0,000 people and
there is minimal commuting from the
county to a large urban center. Counties
of this type include one-third of the
population of the 44 rural counties of
New York State. The rural counties in
the United States to which these results
are applicable are likely to be in the
northern half of the country with a
predominantly white population and
some agricultural production. Further
research of this type with an urban
population is needed.
15
Conclusion
This study identified factors contributing
to food insecurity in a rural population
based on Campbell's (2) conceptual
framework of food insecurity and its
risk factors. These include lack of savings,
low educational level, low income, and
unexpected expenses-all factors that
decrease household resources in
Campbell's framework. In addition,
having to add $50 or more to food stamps
to feed the household (an indicator of
the sufficiency of food stamps) and
lower levels of food expenditures­factors
that decrease food acquisition
in Campbell's model-contributed to
food insecurity. Households with these
characteristics should be giv~n priority
in intervention programs that address
food insecurity. Furthermore, interven­tions
should be designed to address
these food acquisition characteristics
as well as other factors related to
expanding food acquisition such as
vegetable gardening.
16
References
1. Anderson, S.A. (Ed.). 1990. Core indicators of nutritional status for difficult-to­sample
populations. Journal of Nutrition 120:1559-1600.
2. Campbell, C.C. 1991. Food insecurity: A nutritional outcome or a predictor variable?
Journal of Nutrition 121:408-415.
3. Crockett, S.J., Potter, J.D., Wright, M.S., and Bacheller, A. 1992. Validation of a
self-reported shelf inventory to measure food purchase behavior. Journal of the
American Dietetic Association 92:694-697.
4. Deavers, K.L. and Hoppe, R.A. 1993. Overview of the rural poor in the 1980s.
In C.M. Duncan (Ed.), Rural Poverty in America. Auburn House, Westport, CT.
5. Eberts, P.R. 1994. Socioeconomic Trends in New York State: 1950-1990. (2d ed.).
New York State Legislative Commission on Rural Resources, Albany, NY.
6. Frongillo, E.A., Rauschenbach, B.S., Olson, C.M., Kendall, A., and Colmenares, A.
1995. Estimating the Prevalences of Hunger and Food Insecurity: Validity of
Questionnaire-Based Measures for the identification of Households. Final report to
the Institute for Research on Poverty, University of Wisconsin-Madison.
7. Kahn, H.A. and Sempos, C.T. 1989. Statistical Methods in Epidemiology.
Oxford University Press, New York, NY.
8. Kendall, A., Olson, C.M,. and Frongillo, Jr., E.A. 1995. Validation of the Radimer/
Cornell measures of hunger and food insecurity. Journal of Nutrition 125:2793-2801.
9. Kinsey, J.D. 1994. Food and families' socioeconomic status. Journal of Nutrition
124:1878S-1885S.
10. Lino, M. 1996. Income and spending of poor households with children. Family
Economics and Nutrition Review 9( 1):2-13.
11. Maxwell, S. and Frankenberger, T.R. 1993. Household Food Security: Concepts,
Indicators, Measurements. United Nations Childrens Fund-International Fund for
Agriculture Development, New York, NY.
12. Morris, P.M., Neuhauser, L., and Campbell, C.C. 1992. Food security in rural
America: A study of the availability and costs of food. Journal of Nutrition Education
24:52S-58S.
13. O'Neill, M. 1992 (March 18). As the rich get leaner, the poor get french fries.
New York Times.
Family Economics and Nutrition Review
14. Radimer, K.L. 1990. Understanding Hunger and Developing Indicators to
Assess It. Ph.D. dissertation, Cornell University, Ithaca, NY.
15. Radimer, K.L., Olson, C.M., and Campbell, C.C. 1990. Development of indicators
to assess hunger. Journal of Nutrition 120:1544-1548.
16. Radimer, K.L., Olson, C.M., Greene, J.C., Campbell, C.C., and Habicht, J-P.
1992. Understanding hunger and developing indicators to assess it in women and
children. Journal of Nutrition Education 24:36S-45S.
17. Rank, M.R. and Hirschi, T.A. 1993. The link between population density and
welfare participation. Demography 30(4):607-622.
18. Rose, D.R., Basiotis, P.P., and Klein, B.S. 1995. Improving federal efforts to
assess hunger and food insecurity. F oodReview 18(1 ): 18-23. ·
19. Sanjur, D., Haines, P., Travis, S., Brooks, M., Hammons, B., and Immink,
M.D.C. 1979. Food expenditures, consumption, and nutrient availability among
New York State EFNEP households. Human Ecology Search 9:1-58.
20. SAS Institute, Inc. 1989. SAS/STAT User's Guide, Version 6. 4th ed., Vol. 1.
SAS Institute, Cary, NC.
21. Senauer, B., Asp, E., and Kinsey, J. 1991. Food Trends and the Changing
Consumer. Eagan Press, St. Paul, MN.
22. Shotland, J. and Loonin, D. 1988. Patterns of Risk: The Nutritional Status of
the Rural Poor. Public Voice for Food and Health Policy, Washington, DC.
23. Statistical Sciences, Inc. 1993. S-Plusfor Windows User's Manual, Version 3.1.
Statistical Sciences, Inc., Seattle, W A.
24. U.S. Department of Agriculture, Food and Consumer Service and U.S. Department
of Health and Human Services, National Center for Health Statistics. 1994.
Conference on Food Security Measurement and Research. Washington, DC.
25. U.S. Department of Commerce, Bureau of the Census. 1995. Income, Poverty,
and Valuation of Noncash Benefits: 1993. Current Population Reports, Consumer
Income. Series P60-188.
26. Venables, W.N. and Ripley, B.D. 1994. Modem Applied Statistics with S-Plus.
Springer-Verlag, New York, NY.
1997 Vol. 10 No.2 17
18
Do Food Bars Measure Up?
Nutrient Profiles of Food Bar
·Versus Traditional School
Lunches in the CATCH Study
By Anne 0. Garceau
Frances Stern Nutrition Center, New England Medical Center
Mary Kay Ebzery
Frances Stern Nutrition Center, New England Medical Center
Johanna T. Dwyer
Tufts University and Frances Stem Nutrition Center, New England Medical Center
Theresa A. Nicklas
Tulane University School of Public Health and Tropical Medicine
Deanna H. Montgomery
University of Texas School of Public Health
Lynn V. Hewes
Frances Stern Nutrition Center, New England Medical Center
Paul D. Mitchell
New England Research Institutes
Leslie A. Lytle
University of Minnesota School of Public Health
Michelle M. Zive
University of California at San Diego
The nutrient content of food bars versus traditional school lunch menus is
compared using data from 16 elementary schools that participated in the
Child and Adolescent Trial for Cardiovascular Health (CATCH). Results
show that food bars offered as complete National School Lunch Program
reimbursable lunches provide less total energy (641 kcal vs. 727 kcal) and
more energy from fat (40.3 percent vs. 35.4 percent) and saturated fat
(16.3 percent vs. 14.0 percent) than traditional lunches. Sodium in the two
types of lunches did not differ, exceeding national recommendations in
each. Partial food bars offered by 24 schools in place of the traditional
vegetable/fruit meal component (as offered in 72 schools) are a higher total
fat (6.5 g vs. 3.8 g) and similar saturated fat (1.0 g vs. 1.2 g) alternative.
Food bars analyzed in this study are not necessarily lower in total fat,
saturated fat, or sodium than traditional lunches. Strategies for modifying
food bars need to be emphasized as schools work to meet the nutrition
goals of the U.S. Department of Agriculture's School Meals Initiative for
Healthy Children.
Family Economics and Nutrition Review
espite the many nutritional
benefits of the National
School Lunch Program
(NSLP) ( 5,1 0,11 ), elemen­tary
school lunches continue to be high
in fat (37 -41 percent of calories), satu­rated
fat (14-18 percent of calories), and
sodium (1,100-1,400 mg) (3,17,19,23,24).
U.S. Department of Agriculture (USDA)
regulations require that NSLP meals
meet the Dietary Guidelines for Americans
(29) by the school year 1998/99. The
Guidelines specify that dietary total fat
and saturated fat in school lunches
should not exceed 30 percent and 10
percent of total energy, respectively
(25 ). Although USDA regulations do
not state specific goals for sodium and
dietary cholesterol, it is generally
recommended that sodium in school
lunches be reduced to approximately
800 mg and cholesterol remain at or
below I 00 mg, one-third the recom­mended
daily intakes for children
(30,31).
The challenge faced by school food
services is how best to achieve these
nutritional objectives while preserving
adequate amounts of energy, protein,
vitamins, and minerals in school lunches.
The NSLP requires that lunches provide
one-third of the Recommendecl Dietary
Allowances (RDA) (1 5) for energy and
selected nutrients: Protein, vitamin A,
ascorbic acid, calcium, and iron (25 ).
Efforts are now underway nationwide,
under the leadership of USDA's Team
Nutrition, to lower the total fat, satu­rated
fat, and sodium in school meals
(26). Other school food service programs
have also been developed to assist
schools in reducing total fat and sodium
in school meals ( 1,2,8,9, 1 6,22,23,32,33 ).
Although several of these programs
proved to be effective in modifying
1997 Vol.10No.2
school meals, results reported were
limited to traditional school menus.
Many schools have expanded their menus
to include selections from "food bars"
in place of, or as part of, the more tradi­tional
lunches. Traditional lunches,
where food items are portioned and
served to the students by cafeteria staff,
are in contrast with food bars, which
offer a variety of foods from which
each student may self-serve food items.
Food bars include fruit and vegetable
salad bars, Mexican bars (taco and
fillings), potato bars (baked potato with
a variety of toppings), and deli bars (where
students can build their own sandwiches).
Whether lunches constructed from such
food bars provide less total fat, saturated
fat, and sodium than traditional lunches
has not been fully investigated due to
methodological issues associated with
their analysis.
The Child and Adolescent Trial for
Cardiovascular Health (CATCH)
evaluated both traditional school lunch
menus and food bar menus collected in
96 elementary schools ( 40 control and
56 intervention) across four States
(California, Louisiana, Minnesota,
Texas) from the fall of 1991 through
the spring of 1994. CATCH was a pro­spective,
multicenter study involving
school and family-based interventions
aimed at reducing cardiovascular risk in
a cohort of elementary school children
(12).
One component of the CATCH interven­tion
was Eat Smart, a food service
program designed to lower the total fat,
saturated fat, and sodium content of
school lunch and breakfast menus in the
intervention schools (1 7,18 ). Specific
Eat Smart objectives for school lunch
menus as offered were to reduce the
average amount of total fat to no more
than 30 percent of energy, saturated fat
to no more than 10 percent of energy,
and sodium by at least 25 percent from
baseline levels to between 600 and
1,000 milligrams (mg). Although these
nutritional objectives were applied
primarily to the traditional school lunch
menus, food bars were also included
in modification efforts using specific
Eat Smart guidelines and fat and sodium
criteria for recipes, ingredients, and
vendor products. Details of the Eat
Smart program, food service staff
training, intervention materials, and
process and outcome evaluations have
been described in detail elsewhere
(6,14,17-19,21 ).
The primary purpose of this paper was
to examine if food bar-type lunches as
offered in the CATCH study provided a
healthier alternative to traditional lunches
and whether the Eat Smart intervention
had any effects on the nutrient content
of these food bars. A secondary objective
was to determine if one type of food bar,
a vegetable and fruit salad bar, provided
a lower fat and sodium alternative to
the traditional vegetable and fruit
component of school lunch. Finally,
the method for nutrient analysis of food
bars is described and methodological
issues that arise in such analyses are
briefly discussed.
Methodology
A food bar was defined as such if it met
all of the following criteria: ( 1) the
menu was identified as a food bar by
the school cafeteria manager, (2) stu­dents
were allowed to choose among
the food items offered and serve them­selves,
and (3) foods were provided in
large serving dishes or dispensers (not
preportioned), except for more expen­sive
items such as meat, cheese, and
milk, which were usually only available
in standard, limited quantities.
19
20
Food bars were ...
either "complete" or
"partial" based on
the extent to which
they fu lfi lied the
requirements for a
NSLP reimbursable
lunch.
Food bars were also categorized as
either "complete" or "partial" based
on the extent to which they fulfilled the
requirements for a NSLP reimbursable
lunch (28). Complete food bars con­sisted
of all five meal components (that
is, a meat/meat alternate, two vegetable/
fruit, a bread/bread alternate, and milk).
Partial food bars met only the vegetable/
fruit meal component requirement of
the meal patterns.
All traditional and food bar menus and
recipes were collected over 5 consecutive
days during visits to the schools by
trained and certified CATCH staff at
3 periods: baseline (fa111991), interim
(fall 1992), and follow-up (spring 1994)
(7). Data were sent to the CATCH
Coordinating Center at New England
Research Institutes, Watertown, MA,
for coding and quality control checks.
Data entry of the school lunch menus
and recipes was performed at the
University of Minnesota's Nutrition
Coordinating Center using their Nutrition
Data System (NDS), food database
version 4A, nutrient database version 19.
It was beyond the scope of the CATCH
study to collect observational data for
type and amount of food items selected
by students from the food bars. Produc­tion
sheets were not available from all
the schools' food service departments;
therefore, data were not available on
the total amounts of each food item
taken from the food bars during the
lunch period.
Given these limitations, one possible
approach to the nutrient analysis of food
bars is to assume that students are offered
only the minimum amounts of each meal
component necessary to satisfy the USDA­NSLP
meal pattern requirements for
a reimbursable meal. However, this
approach does not seem realistic given
that some food bar food items that are
likely to be chosen by students do not
contribute to the USDA-NSLP meal
pattern requirement, such as toppings,
condiments, and desserts.
Another approach would be to construct
a food bar meal with all the items offered
on the food bar. However, this approach
seems unrealistic given the numerous
food items on some food bars. If a student
chose a full serving of each item, the
lunch meal constructed could provide
an unusually excessive amount of calories.
Therefore, the assumption was made
that students are offered the minimum
serving of each meal component to
meet the USDA-NSLP requirements
plus toppings, condiments, and desserts
whenever available.
Food bar information was recorded on
a form by the school cafeteria manager
and reviewed with him/her by a CATCH­trained
food service data collector. Each
food bar form listed the complete food­item
description, its intended meal
component category, recipes, and vendor
product information. Most of the foods
offered on the food bars were "self-serve"
items that the children served themselves
from bulk containers and, therefore, had
no reported serving size.
For nutrient coding purposes, self-serve
food items were assigned standard portion
sizes at theCA TCH Coordinating Center.
Food items that contributed toward ful­filling
the meal pattern were assigned
portion sizes corresponding to the USDA­NSLP-
defined minimum portion sizes
for the appropriate meal components. It
was necessary to analyze the intended
USDA-NSLP lunch that coincided with
the grade of the CATCH cohort because
CATCH integrated the nutrient values
of school menu items with analyses of
Family Economics and Nutrition Review
Table 1. Nutrient analysis method of food bars: Follow-up salad bar
example
Serving
Food size1 Source2 Meal component
Lettuce, chopped 3/8 cup NSLP Salad bar vegetable
Tomatoes, diced 3/8 cup NSLP Salad bar vegetable
Carrots, shredded 3/8 cup NSLP Salad bar vegetable
Potato salad 3/8 cup NSLP Vegetable
American cheese,
shredded 2 oz NSLP Meat/meat alternate
Eggs, hard boiled 1large NSLP Meat/meat alternate
Ham, diced 2 oz NSLP Meat/meat alternate
Croutons 3/4 tbsp SNDA Topping
Olives, green 3/4 tbsp SNDA Topping
Bacon bits 3/4 tbsp SNDA Topping
Ranch dressing 1.5 tbsp FCEI Salad dressing
Italian dressing 1.8 tbsp FCEI Salad dressing
White bread 1 slice SNDA Bread/bread alternate
Pineapple chunks 3/8 cup NSLP Fruit
Milk, whole 8 fluid oz School Milk
Milk, 1% lowfat 8 fluid oz School Milk
Average nutrient content of salad bar= A + C + D + E + F + 3T
where D = (P + H) + B
2
Code3
p
p
p
B
A
A
A
T
T
T
H
H
E
c
F
F
1 All food items offered "self-serve" except milk; serving sizes for milk were reported by school cafeteria
manager.
2Sources of standard portions assigned to self-serve items:
NSLP =National School Lunch Program Meal Pattern (28)
SNDA = School Nutrition Dietary Assessment Study (4)
FCEI = Foods Commonly Eaten by Individuals (20 ).
3The alphabetic codes refer to USDA-NSLPmeal component or food category listed in previous column.
24-hour dietary recalls collected from
children at baseline and follow-up ( 13 ).
The minimum required portion size for
some food items increased from base­line
(3rd grade) to follow-up (5th grade),
so the assigned standard serving sizes
increased as well (28). For example, a
self-serve serving size of sliced turkey
presented on a food bar when the cohort
was in 3rd grade was 1.5 ounces, the
1997 Vol.l0No.2
USDA-NSLP minimum requirement for
meat/meat alternate for grades K to 3 (28).
The self-serve serving size for the same
food item presented on a food bar when
the cohort was in 5th grade was assigned
an amount of2 ounces, the USDA-NSLP
minimum requirement for meat/meat
alternate in grades 4-12 (28). Fourth
grade serving sizes were assigned for
food bars collected at interim.
For food items that did not contribute to
the meal pattern, such as salad dressings,
sour cream, mustard, and catsup, refer­ence
data on amounts children were
likely to consume (20) were used. The
assigned portion sizes for some of these
food bar items also varied by measure­ment
period, reflecting differences in
typical amounts consumed by children
at different ages. For example, the
average quantity of catsup consumed
per single eating occasion for children
age 6-8 is 1.3 tablespoons, whereas for
children age 9-14, it is 1.5 tablespoons
(20).
For other non-USDA meal pattern food
items and those for which no reference
data were available, portion sizes pub­lished
by the School Nutrition Dietary
Assessment (SNDA) Study (4) were
used as standards. Table 1 lists some of
the serving sizes assumed for self-serve
food items at follow-up (5th grade).
Daily menus were "disaggregated" into
traditional lunch menus and food bar
menus (7). Food items were then catego­rized
and assigned an alphabetic code
as follows: Meat/meat alternate (A),
nonsalad vegetable (B), fruit (C), bread/
bread alternate (E), milk (F), dessert
(G), salad dressing (H), salad vegetables
(P), and toppings (T). All menus were
cross-checked by a CATCH Coordinating
Center nutritionist for menu-coding
consistency.
Nutrient analysis of the food bar and
traditional lunch menus was completed
at the CATCH Coordinating Center.
Traditional lunches were analyzed using
the method described elsewhere by
Osganian and colleagues ( 19). Briefly,
the traditional lunches were constructed
using the minimum requirements for the
USDA-NSLP meal pattern. The sum of
one average serving of each of the meal
21
Table 2. Number of CATCH schools offering complete or partial food bars by site, treatment group, and
measurement period
Site Treatment group
California Intervention
Control
Louisiana Intervention
Control
Minnesota Intervention
Control
Texas Intervention
Control
Overall Intervention
Control
components (meat/meat alternate,
vegetable/fruit, milk) and up to two
average servings of bread/bread alternate,
plus one average serving each of condi­ments
and desserts, when offered, was
used to define a lunch meal ( 19 ).
Nutrients were calculated for each food
bar meal in a similar fashion using the
method depicted by the formula in table 1.
A school lunch menu typically included
one to two servings of bread/bread
alternates per day since the USDA-NSLP
minimum traditional meal pattern
specifies eight servings per week for
this meal component.
Toppings, which consisted of salad
garnishes and potato, taco, and sandwich
accompaniments, were analyzed
22
Measurement period
Baseline Interim Follow-up
Food bar type Food bar type Food bar type
Complete Partial Complete
0 13 0
0 IO 0
I 0 3
3 0 3
2 I 1
2 0 2
0 0 3
0 0 0
3 14 7
5 10 5
according to food bar type. Common
cultural dietary practices were taken
into account when toppings were coded
for nutrient analysis.
For potato bars and deli bars, the nutrient
contents of all nonmeat toppings or
accompaniments were averaged together
and a value for one composite topping
was used. Meat toppings counted toward
the meat/meat alternate requirement.
For taco bars, the nutrient content of all
toppings was added together, since the
underlying assumption was that all the
toppings, such as cheese, lettuce, tomato,
and sour cream, were intended to be
served with a taco. Fruit and vegetable
food bars were analyzed with up to
three toppings.
Partial Complete Partial
I4 0 I4
10 0 IO
0 3 0
0 5 0
1 2 1
2 2 0
0 2 0
0 0 0
I5 7 15
12 7 10
The effect of lunch menu type on the
average nutrient content of school
lunches as offered was investigated
using the analysis of variance (ANOV A)
technique. The primary analysis com­pared
complete food bars with traditional
lunches and was restricted to schools
offering both types of lunches. Sixteen
schools offered both complete food bars
and traditional lunches at least once
over the 3 years of study. Of these 16
schools, 8 offered both types of lunch
at baseline, while 12 and 14 did so at
interim and follow-up, respectively.
A total of 34 matched pairs of food bar
and traditional lunch data was available
for this analysis. Table 2 shows the
distribution of schools across study
sites, treatment groups, and time.
Family Economics and Nutrition Review
A secondary analysis investigated
whether partial food bars differed nutri­tionally
from the vegetable/fruit compo­nent
of traditional lunches. California
offered the vegetable/fruit component
of school lunch as partial food bars
over all 3 years, except for one school
at baseline that was omitted from the
analysis.
Ten of the 24 schools in California also
offered a pre-portioned, hot vegetable
on at least 1 of the 5 days that menus
were collected (average of 1.6 days).
This hot vegetable was included in the
analysis of partial food bars to represent
the complete vegetable/ fruit component
as offered in California. Since few
schools at the remaining sites offered
this meal component as a partial food
bar, Louisiana, Minnesota, and Texas
schools acted as a comparison group for
the vegetable/fruit component served in
the traditional manner, for a total of 287
observations.
Nutrients of interest for both sets of
analyses included energy, total fat,
saturated fat, sodium, cholesterol,
carbohydrate, protein, dietary fiber,
and vitamins and minerals (specifically
vitamin A, ascorbic acid, iron, and
calcium). The ANOV A model for the
primary analysis comparing traditional
lunches with complete food bars included
as fixed independent effects: CATCH
site (df=2: Louisiana, Minnesota, Texas);
time (df=2: baseline, interim, follow-up);
lunch type (df=1: traditional, food bar);
treatment group ( df= 1: control, interven­tion);
and all higher order interactions
among time, lunch type, and treatment
group.
The ANOV A model for the secondary
ahalysis was similar, except that the
source of the vegetable/fruit component
(partial food bar vs. traditional menu) was
1997 Vol.JONo.2
introduced as a fixed effect in place of
site. Both models incorporated school
as a random effect to account for school­to-
school variation.
When analysis of log transformed data
was necessary to satisfy modeling
assumptions, estimates of means and
standard errors were transformed back
to their original units for presentation.
All mixed models were fit with restricted
maximum likelihood estimation using
the Statistical Analysis System's Proc
Mixed (version 6.08, 1992, SAS
Institute, Inc.; Cary, NC).
Results
No differences in the mean percentage
of energy from total fat or saturated fat,
total energy, sodium, or dietary choles­terol
were evident between control and
intervention schools for either food bar
or school-matched traditional lunches at
baseline, interim, or follow-up (data not
shown). Since no intervention effects
were seen in this subsample of lunch
menus, the results reported in the fol­lowing
text and table 3 represent the
mean nutrient content of food bar and
traditional lunches as offered across all
sites for all measurement periods and
treatment groups combined. The effects
of the Eat Smart intervention on the
lunch menus in all 96 CATCH schools
are presented elsewhere ( 19 ).
Food bar lunches provided less total
food energy than school-matched tradi­tional
lunches (p<0.001); however, food
bars still contributed about one-third
of the RDA for children 7-10 years of
age ( 15 ). Food bar lunches provided a
greater percentage of energy from fat
(p<0.001) and saturated fat (p<0.001)
than traditional lunches; however, no
differences were evident in total fat and
saturated fat content.
Food bar lunches
provided less total
food energy than
school-matched
traditional lunches ....
Food bar lunches
provided a greater
percentage of energy
from fat (p<0.001)
and saturated fat
(p<0.001) than
traditional lunches ...
23
Food bars provided fewer grams of Table 3. Nutrient profiles of food bar and traditional school lunch menus1
carbohydrate (p<0.001) and protein
(p=0.015) than traditional lunches. When
Nutrient Menu type Mean (SE)2 P value3
analyzed as a percentage of the total
energy content of the lunches, food bars
provided less carbohydrate (p< 0.001) Energy (kcal) Food bar 641 (16) <0.001
than traditional lunches, whereas there Traditional 727 (16)
was no difference for protein. Total fat (g) Food bar 28.7 (1.0) 0.94
Traditional 28.7 (1.0)
No differences were noted in the mean Total fat(% of energy) Food bar 40.3 (0.9) <0.001
sodium or dietary cholesterol content Traditional 35.4 (0.9)
of meals offered when food bar and Saturated fat (g) Food bar 11.5 (0.5) 0.65
corresponding traditional lunches were Traditional 11.2 (0.5)
compared. Average sodium exceeded Saturated fat(% of energy) Food bar 16.3 (0.5) <0.001
the Eat Smart program goal of 600-
Traditional 14.0 (0.5)
1,000 mg for both food bars and tradi-tionallunches.
Both the food bar and Carbohydrate (g) Food bar 68.1 (2.3) <0.001
traditional lunches met the nationally Traditional 86.6 (2.3)
recommended dietary cholesterol goal Carbohydrate(% of energy) Food bar 42.3 (0.8) <0.001
of 100 mg or less per lunch. When Traditional 47.6 (0.8)
examined per 1,000 calories, food bars Protein (g) Food bar 29.6 (0.8) 0.015
and traditional lunches did not differ in Traditional 32.4 (0.8)
sodium provided; however, food bars Protein(% of energy) Food bar 18.5 (0.3) 0.13
provided more dietary cholesterol than Traditional 17.9 (0.3)
traditional lunches (p<0.001). Sodium (mg) Food bar 1194 (39) 0.29
Traditional 1250 (39)
Table 3 also shows mean nutrient data Sodium (mg/1000 kcal) Food bar 1900 (58) 0.054
per 1,000 calories for selected vitamins
Traditional 1731 (58)
and minerals in food bar and traditional
lunches. The amount of ascorbic acid Cholesterol (mg) Food bar 95.2 (4.7) 0.059
and calcium per 1,000 calories was Traditional 84.2 (4.7)
greater in food bars than in traditional Cholesterol (mg/1 000 kcal) Food bar 149 (5) <0.001
lunches (p<0.001 for both). No differ- Traditional 116 (5)
ences in vitamin A value, iron, or dietary Iron (mg/1000 kcal) Food bar 5.5 (0.1) 0.11
fiber content were evident. Both food Traditional 5.8 (0.1)
bar and traditional lunches as offered Calcium (mg/1000 kcal) Food bar 800 (26) <0.001
exceeded one-third the RDA ( 15) for Traditional 695 (23)
protein, vitamin A, ascorbic acid, Vitamin A value (RE/1000 kcal) Food bar 505 (28) 0.24
calcium, and iron. Traditional 465 (28)
Treatment group comparisons for single
Ascorbic acid (mg/1000 kcal) Food bar 45.9 (4.8) <0.001
meal components, such as the vegetable/
Traditional 31.4 (3.3)
fruit component, were not the focus of Dietary fiber (g/1 000 kcal) Food bar 7.5 (0.3) 0.58
this paper. Therefore, the results reported Traditional 7.4 (0.3)
in table 4 represent a comparison of the 1N =68 observations: 34 schools offered bolh menu types at baseline, interim, or follow-up. mean nutrient content of partial food bar 2 Adjusted mean (standard error).
3P value from analysis of variance testing lhe hypolhesis !hat lhe mean nutrient value for food bars was
equal to lhe mean nutrient value for traditional school lunches.
24 Family Economics and Nutrition Review
Table 4. Nutrient profiles of partial food bar and traditional vegetable/
fruit components of school lunch 1
Vegetable/ fruit
Nutrient component type Mean (SE)2 P value3
Energy (kcal) Partial food bar 152 (6) 0.007
Traditional 133 (3)
Total fat (g) Partial food bar 6.5 (0.4) <0.001
Traditional 3.8 (0.2)
Saturated fat (g) Partial food bar 1.0 (0.1) 0.26
Traditional 1.2 (0.1)
Carbohydrate (g) Partial food bar 23.9 (1.0) 0.060
Traditional 24.5 (0.6)
Protein (g) Partial food bar 2.1 (0.1) 0.32
Traditional 2.3(0.1)
Sodium (mg) Partial food bar 188 (18) 0.73
Traditional 196 (11)
Cholesterol (mg) Partial food bar 2.5 (0.7) 0.020
Traditional 4.4 (0.4)
Iron (mg) Partial food bar 0.75 (0.05) 0.65
Traditional 0.77 (0.03)
Calcium (mg) Partial food bar 27 (2) 0.044
Traditional 32 (1)
Vitamin A value (RE) Partial food bar 139 (17) 0.92
Traditional 141 (10)
Ascorbic acid (mg) Partial food bar 11.5 (0.6) <0.001
Traditional 14.9 (0.5)
Dietary fiber (g) Partial food bar 2.9 (0.1) 0.77
Traditional 3.0 (0.1)
1 N=287 observations: 71 schools (California) with partial food bars compared with 216 (remaining
three sites) with traditional vegetable/fruit components, over all3 years.
2 Adjusted mean (standard error).
3p value from analysis of variance testing the hypothesis that the mean nutrient value of the vegetable/
fruit component in partial food bars was equal to that in traditional school lunches.
1997 Vol.10No.2
and traditional vegetable/fruit
components for control and intervention
schools and all measurement periods
combined.
Partial food bars contributed more total
fat (p<0.001), yet similar amounts of
saturated fat, when compared with the
traditional vegetable/fruit component.
No differences in carbohydrate or
protein content were evident.
The sodium content of partial food
bars and traditional vegetable/fruit
components did not differ. Partial food
bars provided less dietary cholesterol
than the traditional vegetable/fruit
component (p=0.02).
The vitamin A value, iron, and dietary
fiber content for the two types of vegetable/
fruit component, also shown in table 4,
did not differ. Partial food bars provided
less calcium and ascorbic acid than the
traditional vegetable/ fruit component
(p=0.044 and p<0.001, respectively).
Discussion
The research methods presented in this
paper for analyzing school food bar
lunches provide an important contribu­tion
to the assessment of school-based
intervention programs aimed at modify­ing
fat, sodium, and other nutrient
intakes of children. Standardized method­ologies
for evaluating the nutrient
content of food bars have not been
available until recently. Now that such
methods are available, school food
service managers and directors can
assess meals as offered and bring them
into compliance with USDA regulations
(25).
25
The method developed for use in CATCH
and reported in this paper can serve as
a model for analyzing food bar meals.
This method for representing a complete
"lunch" was not necessarily comparable
with a lunch that a student would select
and is a limitation of this study. Studies
that characrerize and analyze lunches as
actually chosen by students are neces­sary
to lend insight into what children
are eating and how what is offered in
school lunch influences their choices.
Food bars as analyzed in this study were
not as healthful as many school food
service directors, parents, and others may
perceive them to be. They provided
significantly more total fat and satu­rated
fat as a percentage of energy than
traditional lunches; however, they also
provided significantly less total energy,
so the fat content (in grams) did not differ
between the two types of lunches as
offered.
The difference in carbohydrate content
was primarily responsible for the differ­ence
in energy levels between the lunches.
Total energy provided by food bars may
have been skewed due to assumptions
made in the coding of individual portion
sizes and their effect on the subsequent
nutrient analysis of these lunches.
The amount of total carbohydrate
observed in food bar lunches may have
been lower than in traditional lunches
because the food bar items that fulfilled
the meat/meat alternate requirement
were typically cold cuts, cheeses, cheese
sauces, and taco meat, which are low
in total carbohydrate. In contrast, tradi­tional
lunches often included entree
items that not only fulfilled the meat/
meat alternate component of the meal
but also provided a bread/bread alternate,
such as spaghetti with meat sauce,
26
breaded chicken, corn dogs, and pizza,
all of which provide considerable amounts
of energy in the form of carbohydrate.
Often a second bread/bread alternate,
for example, a roll, was served with the
traditional lunches, whereas only one
bread choice was offered with most
food bar lunches.
Also, vegetables served on many of the
food bars were those usually found in
salads, such as lettuce, tomatoes,
cucumbers, peppers, carrots, and celery,
which are lower in carbohydrate and
energy than the types of vegetables
usually served with traditional lunches,
such as peas, beans, corn, and potatoes
(most often as french fries). Although
baked potatoes served in potato bars
provided a substantial amount of
carbohydrate, these types of food bars
accounted for only 20 percent of all food
bars analyzed.
When food bars and their school-matched
traditional style lunches were examined,
we did not find effects of the Eat Smart
intervention as was found in previous
analyses of all 96 CATCH schools ( 19 ).
It is possible that this was due to the
small number of CATCH schools that
served food bar meals used in the
analysis presented here.
Sodium content did not vary by style
of lunch but was higher than the desired
range of 600-1 ,000 mg of sodium in
both lunch types. Some high sodium
foods offered on CATCH food bars
included salad dressing, cold cuts, cheese,
taco meat, salted crackers, and bacon
bits. Since a major focus of the Eat
Smart intervention was sodium reduction,
a decrease in the sodium content of the
lunch menus offered in the intervention
schools was expected. However,
increased portion sizes from 3rd to 5th
grade (CATCH analysis reflected this
trend) may have partially counteracted
Eat Smart effects on the sodium content
of the lunch menus.
One problem is that as companies lower
the fat in commercially prepared foods,
many of these food items may remain
high or may even increase in sodium
content to compensate for loss of flavor
associated with the removal of fat
( 13,19 ). Efforts are now underway by
the Federal Government to provide
schools with lower sodium commodities
to help ensure healthy school meals (27).
The cholesterol content of food bars
was higher than that of traditional
lunches when examined per 1,000
calories; however, this result seems less
important than the finding that both
types of lunches met the recommended
goal of no more than 100 mg cholesterol
per lunch ( 31 ).
Both food bars and traditional lunches
met the USDA-NSLP requirement (25)
of one-third the RDA ( 15) for selected
nutrients; however, some significant
differences were noted in the calcium
and ascorbic acid densities provided by
the two types of lunches. A major con­tributor
of calcium to the school lunch
is milk, which was offered in the same
quantity and style in both lunch types.
Cheese-also a significant source of
calcium and often served in both types
of lunches-was offered more frequently
in the food bar than in the traditional
lunch as either a topping for tacos,
potatoes, or salads, or to help fulfill the
meat/meat alternate requirement, such
as in a sandwich on a deli-type food bar.
Family Economics and Nutrition Review
It is more difficult to explain the higher
ascorbic acid content of food bars com­pared
with traditional lunches without
further analysis given that major sources
of this nutrient, such as oranges, grape­fruits,
broccoli, cauliflower, green peppers,
potatoes, and tomatoes were offered
regularly and in similar quantities in
both lunch types. The food-based analyses
that would be required to provide this
type of information were not within the
scope of this study.
Dietary fiber content of the food bar
was not greater than that of the tradi­tionallunch
as might have been expected;
rather, the dietary fiber content of the
two types of lunches was similar.
Vegetable selections offered on food
bars such as iceberg lettuce, celery,
and cucumbers are not always high
in dietary fiber when compared with
many of the vegetables served in the
traditional lunches, such as peas, beans,
and corn. The same fresh fruit and grain
products were frequently offered with
both types of lunches.
Also, it is possible that the portions of
food bar vegetables students choose to
eat are greater than the portions assigned
in this analysis due to the self-serve
nature of food bars. Therefore, the
actual food bar lunch as selected and
consumed could be higher in dietary
fiber content, as well as other nutrients,
than these results suggest.
Partial food bars are not necessarily a
healthful alternative to the traditional
vegetable/fruit component, since they
were higher in total fat and provided
similar amounts of saturated fat and
sodium. All partial food bars in this
study offered one or more types of
salad dressing, possibly reflected in
1997 Vol.J0No.2
the higher total fat content observed in
partial food bars when compared with
the traditional vegetable/fruit component.
Each partial food bar was analyzed with
one composite salad dressing serving
and serving sizes ranged from 1 to 2
tablespoons (the equivalent of 5 to 14 g
of fat for regular dressings). The tradi­tional
vegetable/fruit component some
days consisted of a vegetable with added
fat, such as butter or bacon; other days
it was without added fat. Other sources
of fat in the partial food bars included
potato salad, coleslaw, trail mix, and
in some schools, french fries or other
hot vegetables served with butter or
margarine.
Applications
Food bars can be a good nutritional
alternative to the traditional school
lunch but only with careful planning
and monitoring of the nutrient content
of each food item served and the total
amount of food energy provided.
Although food bars often contain
healthful choices such as fresh fruits
and vegetables, they also contain high
fat dishes such as pasta or vegetable
salads with mayonnaise, cheeses, cold
cuts, salad dressings, and dessert items.
Using lowfat and fat-free salad dressings,
lowfat recipes and vendor products for
the meat or meat alternate, and including
at least two bread/bread alternates in
the menu may help to decrease both the
relative and absolute amounts of fat in
food bars.
CATCH was conducted in elementary
schools where food bars are offered
less frequently than in middle and high
schools. According to the SNDA study,
16 percent of elementary schools offered
a salad bar at least once a week, increasing
to 44 percent of middle schools and
54 percent of high schools ( 3 ). A study
involving secondary schools might
yield different results, reflecting larger
food portions and/or different food
offerings.
This study examined the nutrient content
of elementary school food bars as
offered. However, the particular food
items and the nutrient composition of
the meal the students select is an impor­tant
question that needs investigation.
Until additional information is available
on the nutrient analysis of food bars,
it should not be assumed that these
alternatives to the traditional school
lunch are the healthier choice. Thus,
our data suggest that food service
interventions should target food bars
as significant contributors of total fat,
saturated fat, and sodium in elementary
school lunches.
27
28
References
1. American Cancer Society and National Cancer Institute. 1990. Changing the Course.
(Curriculum materials kit). Atlanta, GA.
2. American Heart Association. 1992. Hearty School Lunch (Resource manual and
menus). Dallas, TX.
3. Burghardt, J.A., Gordon, A.R., and Fraker, T.M. 1995. Meals offered in the National
School Lunch Program and the School Breakfast Program. The American Journal of
Clinical Nutrition 61(1 Suppl.):187S-198S.
4. Burghardt, J., Gordon, A., Chapman, N., Gleason, P., and Fraker, T. 1993. The School
Nutrition Dietary Assessment Study: School food service, meals offered and dietary
intakes. Mathematica Policy Research, Inc., Princeton, NJ.
5. Devaney, B.L., Gordon, A.R., and Burghardt, J.A. 1995. Dietary intakes of students.
The American Journal of Clinical Nutrition 61 (1 Suppl. ):205S-212S.
6. Dwyer, J.T., Hewes, L.V., Mitchell, P.D., Nicklas, T.A., Montgomery, D.H., Lytle, L.A.,
Snyder, M.P., Zive, M.M., Bachman, K.J., Rice, R., and Parcel, G.S. 1996. Improving
school meals: Effects of the CATCH Eat Smart program on the nutrient content of
school breakfasts. Preventive Medicine 24:413-422.
7. Ebzery, M.K., Montgomery, D.H., Evans, M.A., Hewes, L.V., Zive, M.M., Reed, D.,
Rice, R., Hann, B., and Dwyer, J.T. 1996. School meal data collection and documentation
methods in a multisite study. School Food Service Research Review 20:69-77.
8. Ellison, R.C., Capper, A.L., Goldberg, R.J., Witschi, J.C., and Stare, F.J. 1989. The
environmental component: Changing school food service to promote cardiovascular
health. Health Education Quarterly 16:285-297.
9. Fox, M.K. and Glantz, F.B. 1995. The School Nutrition Demonstration. Food service
reform and nutrition education in low-income schools: Real world challenges.
Abt Associates, Inc., Cambridge, MA.
10. Gordon, A.R., Devaney, B.L., and Burghardt, J.A. 1995. Dietary effects of the
National School Lunch Program and the School Breakfast Program. The American
Journal of Clinical Nutrition 61(1 Suppl.):221S-231S.
11. Hanes, S., Vermeersch, J., and Gale, S. 1984. The National Evaluation of School
Nutrition Programs: Program impact on dietary intake. The American Journal of Clinical
Nutrition 40:390-413.
12. Luepker, R.V., Perry, C.L., McKinlay, S.M., Nader, P.R., Parcel, G.S., Stone, E.J.,
Webber, L.S., Elder, J.P., Feldman, H.A., Johnson, C.C., Kelder, S.H., and Wu, M.
1996. Outcomes of a field trial to improve children's dietary patterns and physical
activity: The Child and Adolescent Trial for Cardiovascular Health (CATCH).
Journal of the American Medical Association 275:768-776.
Family Economics and Nutrition Review
13. Lytle, L.A., Stone, EJ., Nichaman, M.Z., Perry, C.L., Montgomery, D.H., Nicklas, T.A.,
Zive, M.M., Mitchell, P., Dwyer, J.T., Ebzery, M.K., Evans, M.A., and Galati, T.P.
1996. Changes in nutrient intakes of elementary school children following a school-based
intervention: Results from the CATCH study. Preventive Medicine 25:465-477.
14. Montgomery, D.H., Zive, M.M., Raizman, D.J., Nicklas, T.A., Evans, M.A., Snyder,
M.P., Baker, N., Hann, B., and Bachman, K. 1993. Description and evaluation of a food
service intervention (Eat Smart) training [abstract]. Journal of the American Dietetic
Association 93(9 Suppl.):74A.
15. National Academy of Sciences, National Research Council, Food and Nutrition
Board. 1989. Recommended Dietary Allowances (lOth ed.). National Academy Press,
Washington, DC.
16. Nicklas, T.A., Forcier, J.E., Farris, R.P., Hunter, S.M., Webber, L.S., and Berenson,
G.S. 1989. Heart Smart School Lunch Program: A vehicle for cardiovascular health
promotion. American Journal of Health Promotion 4:91-100.
17. Nicklas, T.A., Reed, D.B., Rupp, J., Snyder, M.P., Clesi, A.L., Glovsky, E.,
Bigelow, C., and Obarzanek, E. 1992. Reducing total fat, saturated fatty acids, and
sodium: The CATCH Eat Smart School Nutrition Program. School Food Service
Research Review 16:114-121.
18. Nicklas, T.A., Stone, E., Montgomery, D., Snyder, P., Zive, M., Ebzery, M.K.,
Evans, M.A., Clesi, A., Hann, B., and Dwyer, J. 1994. Meeting the dietary goals for
school meals by the year 2000: The CATCH Eat Smart School Nutrition Program
Journal of Health Education 25:299-307.
19. Osganian, S.K., Ebzery, M.K., Montgomery, D.H., Nicklas, T.A., Evans, M.A.,
Mitchell, P.D., Lytle, L.A., Snyder, M.P., Stone, E.J., Zive, M.M., Bachman, K.J., and
Parcel, G.S. 1996. Changes in the nutrient content of school lunches: Results from the
CATCH Eat Smart food service intervention. Preventive Medicine 25:400-412.
20. Pao, E.M., Fleming, K.H., Guenther, P.M., and Mickle, S.J. 1982. Foods Commonly
Eaten by Individuals: Amount Per Day and Per Eating Occasion. Home Economics
Research Report No. 44. U.S. Department of Agriculture, Human Nutrition Information
Service.
21. Raizman, D.J., Montgomery, D.H., Osganian, S.K., Ebzery, M.K., Evans, M.A.,
Nicklas, T.A., Zive, M.M., Hann, B.J., Snyder, M.P., and Clesi, A.L. 1994. CATCH:
Food service program process evaluation in a multicenter trial. Health Education
Quarterly (Suppl. 2):S51-S71.
22. Simons-Morton, B.G., Parcel, G.S., Baranowski, T., Forthofer, R., and O'Hara, N.M.
1991. Promoting physical activity and a healthful diet among children: Results of a
school-based intervention study. American Journal of Public Health 81 :986-991.
23. Snyder, M.P., Story, M., and Lytle, L. 1992. Reducing fat and sodium in school
lunch programs: The Lunchpower! Intervention study. Journal of the American Dietetic
Association 92:1087-1091.
1997 Vol.JO No.2 29
30
24. St. Pierre, R., Fox, M.K., Puma, M., Glantz, F., and Moss, M. 1992. Child Nutrition
Program Operations Study: Second Year Report. Abt Associates, Inc., Cambridge, MA.
25. U.S. Department of Agriculture, Food and Consumer Service. 1995. National
School Lunch Program and School Breakfast Program: School meals initiative for
healthy children, 7 CFR Parts 210 and 220, Final Regulation. Federal Register
60(113):31188-31222.
26. U.S. Department of Agriculture, Food and Consumer Service. 1995. Team Nutrition.
(Program description materials).
27. U.S. Department of Agriculture, Food and Consumer Service. 1995. Healthy Kids
HOTUNE. (October 16, 1995).
28. U.S. Department of Agriculture, Food and Nutrition Service. 1988. Menu Planning
Guide for School Food Service (revised). USDA FNS Publication No. PA 1260.
29. U.S. Department of Agriculture and U.S. Department of Health and Human Services.
1995. Nutrition and Your Health: Dietary Guidelines for Americans (4th ed.). Home and
Garden Bulletin No. 232.
30. U.S. Department of Health and Human Services, Public Health Service. 1991.
National Cholesterol Education Program Report of the Expert Panel on Blood Cholesterol
Levels in Children and Adolescents. Nlli Publication No. 91-2732.
31. U.S. Department of Health and Human Services, Public Health Service. 1993.
National High Blood Pressure Education Program. Working Group Report on Primary
Prevention of Hypertension. (Nlli-NHLBI Publication No. 93-2669).
32. Whitaker, R.C., Wright, J.A., Finch, A.J., and Psaty, B.M. 1993. An environmental
intervention to reduce dietary fat in school lunches. Pediatrics 91: 1107-1111.
33. Witschi, J.C., Ellison, R.C., Doane, D.D., Vorkink, G.L., Slack, W.V., and Stare, F.J.
1985. Dietary sodium reduction among students: Feasibility and acceptance. Journal of the
American Dietetic Association 85:816-821.
Family Economics and Nutrition Review
1997 Vol. 10 No.2
Research Briefs
Households Receiving
Food or Meals as Pay
Mark Uno
Economist
Center for Nutrition Policy and Promotion
In today's economy, the noncash benefits
of a job are often as important as wages
or salary. Health care insurance, child
care support, and transportation subsidies
are examples of major noncash benefits.
These benefits may be substantial,
especially as they are not subject to
taxation.1 A minor but interesting benefit
is food or meals. To supplement current
findings on noncash benefits of employ­ment,
this short paper was undertaken to
examine households reporting the receipt
of food or meals as pay.
Data
Data used are from the Diary portion
of the 1993-94 Consumer Expenditure
Survey (CE), conducted by the Bureau
of the Census for the Bureau of Labor
Statistics. The CE is an ongoing survey
that collects data on food and other
selected expenditures, income, and
major sociodemographic characteristics
of consumer units (for this study, the
term "cons timer unit" will be used
interchangeably with household). A
national sample of consumer units, rep­resenting
the civilian noninstitutional­ized
population, is selected and asked
to keep expenditure diaries, which cover
each of two consecutive 1-week periods.
The diaries are placed throughout the
year with oversampling done during the
1For an overview of these benefits, see Employee­Benefits
in a Changing Economy: A BLS CluJrtbook,
Bulletin 2394, September 1992.
fourth quarter of the year. Each week
is deemed an independent sample by
BLS. The 1993-94 Diary survey contains
information from approximately 22,000
diaries.
Although the terms "households" and
"consumer units" will be used inter­changeably,
there is a slight difference.
A consumer unit is defined by blood
relationship, marriage, or adoption;
financial interdependence; and joint
expenditures. In contrast, a household
is defined by residence in a housing unit.
It is possible for two or more consumer
units to reside in a household. For
example, two individuals may live
together but be financially independent
and make separate expenditures; this
would be categorized as two single
consumer units residing in the same
household. A small percentage (7 per­cent)
of consumer units receiving meals
as pay resided in a multiunit household.
Some of these people may be employed
by the other consumer unit in the house­hold
and may receive meals as pay from
this other unit. Child care providers
who live with the family for whom they
provide child care would be included in
this category.
The value of meals received as pay is
expressed as the average weekly dollar
value during the past year. It was derived
from the CE questions: "During the
past 12 months, have any members of
your consumer unit received any free
31
meals at work as part of their pay?",
"About what was the weekly dollar
value of such meals?", and "How many
weeks did members of your consumer
unit receive such meals during the past
12 months?" Value was readily assigned
for meals with a price attached to them;
value for meals provided without a price
attached to them needed to be estimated
by the recipient. The value of these meals
may be underestimated or overestimated
-it is impossible to assess the extent of
this with the data.
Households that were complete income
reporters were selected for analysis.
Complete income reporters are house­holds
that provide values for major
sources of income, such as wages and
salary, interest/dividends, and Social
Security; however, complete income
reporters do not necessarily provide a
full accounting of their income. The
unweighted sample of complete income
reporters consisted of 17,404 households;
of these, 675 reported receiving meals
as pay. Data were weighted to represent
the population.
In order to place households that received
meals as pay in perspective, those house­holds
that did not were also analyzed.
Tests of statistical significance (chi­square
and t-tests) were performed be­tween
the two groups using unweighted
data and reported at the 0.01level. The
0.01level of statistical significance was
selected rather than the more traditional
0.05 level to compensate for any possible
clustering effect present in the data. All
percentages and means reported, however,
are based on weighted data.
Results
Four percent of all households (2.96 mil­lion
consumer units that were complete
income reporters) reported having mem­bers
who received meals as pay. There
32
Characteristics of households receiving and not receiving meals as pay,
1993-94
Receiving Not receiving
Characteristics meals as pay meals as pay
Mean
Age of head 1 * 36 48
Household size 2.5 2.5
Before-tax income $32,300 $34,300
Percent
Family type
Husband-wife only 15 21
Husband-wife with children 31 28
Single-parent with children 8 6
Single 31 29
Other2 15 16
Race
White 88 87
Non-White 12 13
Education of head*
No high school diploma 13 21
High school diploma 29 30
Some college 31 25
College degree 27 24
Occupation of head*
Service 22 8
ManageriaVprofessional 31 22
Othe~ 47 70
Food stamp receipt
Yes 7 7
No 93 93
Region
Urban4 86 85
Rural 14 15
*Differences between the two groups were statistically significant at p ~ 0.01 based on unweighted data.
1 The household head is defined as the person who owns or rents the home; in cases where there is joint
ownership or renting status, the head is arbitrarily decided so is actually a co-head.
2Includes husband-wife or single-parent households residing with others.
30ther includes administrative support, technical, and sales; operators, assemblers, and laborers;
precision production, craft, and repairs; farming, forestry, and fishing; Armed Forces; self-employed;
and not working. ·
4Urban areas are defined as Metropolitan Statistical Areas (MSA' s) and places outside an MSA of 2,500
or more people; rural areas are places of fewer than 2,500 people outside an MSA.
Family Economics and Nutrition Review
were significant differences between
, households receiving meals as pay and
those that did not, by household head's
age, education, and occupation? It
should be noted that the head of house­hold
was not necessarily the person in
the household receiving meals as pay.
The CE question of receiving meals as
pay is a household level question that
could be answered affirmatively if any
one in the household was receiving
meals as pay.
Among households receiving meals as
pay, the household heads were younger
than those in households not receiving
meals as pay (36 vs. 48 years) (see table).
Heads in households receiving meals as
pay also had a higher level of education
than heads in households not receiving
meals as pay. Fifty-eight percent of
heads in households receiving meals as
pay had at least some college education,
compared with 49 percent of heads in
households not receiving meals as pay.
Heads in households receiving meals
as pay were more likely to be employed
in service occupations and managerial/
professional occupations (22 and 31
percent) than heads in households not
receiving meals as pay (8 and 22 percent).
Restaurant employees may account
for the greater percentage in service
occupations. People with expense ac­counts
may explain the higher percent­age
being employed in managerial/
professional occupations. Receiving
meals as pay in managerial/professional
occupations is also related to the higher
educational level of these household
heads, compared with their counterparts
in households not receiving meals as
pay.
2-rhe household head is defined as the person
who owns or rents the home; in cases where there
is joint ownership or renting status, the head is
arbitrarily decided so is actually a co-head.
1997 Vol. 10 No.2
Average weekly dollar value of meals received as pay and average
weekly food expenses of househQids, by receipt as pay, 1993-94
Receiving meals as pay
$32
$73
$29
(away from
home)
$44
(atllome)
Meals as pay Food expenses
Because heads in households receiving
meals as pay were in both high-paying
(managerial/professional) and low­paying
(service) occupations, household
average before-tax income (which does
not include the value of meals received
as pay) was not significantly different
in these households from those not
receiving meals as pay. There were no
significant differences between house­holds
receiving and not receiving meals
as pay in terms of family type, race,
size, region of residence, and food
stamp receipt.
For households receiving meals as pay,
the average weekly dollar value of these
meals was $32 (see figure). The distri­bution
of this weekly dollar value was
skewed; for 62 percent of households,
the average weekly dollar value of meals
received as pay was $20 or under and
for 7 percent of households, the value
was over $90.
The average weekly food expenditures
of households receiving meals as pay
was $73, compared with $79 for house­holds
not receiving meals as pay (a non­significant
difference). There was no
significant difference between the two
Not receiving meals as pay
$79
$27
(away from
home)
$52
(1111\ome)
Food expenses
groups in terms of average expenses
for food at home and food away from
home. This is surprising as one would
expect meals received as pay to lower
the food expenditures of households.
When the value of meals received as
pay was added to the food expenditures
of households, households receiving
such meals had significantly higher
average food expenses than households
that did not receive such meals ($1 05
vs. $79).
Meals received as pay represent a sub­stantial
noncash benefit for households
that receive them, the value being
approximately $1,660 on an annual
average basis. When the value of these
meals is added to before-tax income, the
income of households receiving meals
as pay is within a few hundred dollars
of the income of households not receiving
such meals. It may be that this in-kind
benefit is considered part of total
compensation, so that cash income is
adjusted for the value of in-kind meals
and food. Although the impact of these
meals on job selection is unknown, as
with other noncash benefits, they may
be influential.
33
34
Food Preparers: Their Food
Budgeting, Cost-Cutting, and
Meal Planning Practices
Julia M. Dinkins
Analyst
Center for Nutrition Policy and Promotion
Decisions about household budgets
can influence the economic well-being
as well as health status of families. A
decision to constrain the food budget
without adequate consideration of its
impact on dietary status may be counter­productive
by leading to higher expendi­tures
in other areas, such as short- and
long-term nutrition-related medical
costs. A recent report indicates that 4
of the 10 leading causes of deaths in
the United States are diet related-heart
disease, cancer, strokes, and diabetes.
Other diet-related conditions are over­weight,
hypertension, and osteoporosis.
Health costs associated with these seven
diet-related conditions are approximately
$250 billion each year ( 1 ). Thus, house­hold
practices related to nutrition status
may be of concern for those interested
in identifying ways to reduce public
and private health costs for American
consumers.
Limited household income can constrain
the food budget and influence food
choices and meal planning decisions.
A U.S. Department of Agriculture report
shows that households with limited
financial resources spend less per person
for most food categories and consume
less than does the general population
(2). Although low-income households
spend less on food-in absolute terms­than
other households, they spend a
higher percentage of their total income
for food. Food preparers in these house­holds
have many difficult choices to
make in order to provide nutritious
meals for family members.
This study focuses on one household
practice that may influence the nutri­tional
status of American consumers. It
examines primary food preparers who
agreed or disagreed with the following
item: I run my household on a strict
food budget. Specifically, it examines
the relationship of tli.at reported budgetary
decision with: (I) other measures food
preparers frequently use to try to cut
food costs, (2) their nutritional concerns
for the family, and (3) meal planning
considerations and practices. Results
suggest inconsistencies or limitations
in applying commonly recommended
cost-saving consumer techniques.
Data Source and Sample
For this study, consumer data from
Market Research Corporation of America
(MRCA) Information Services are used.
This data set is used to link individuals'
nutritional attitudes, food consumption
patterns, and nutrient intake. MRCA
conducts a continuous sampling program
using a multistage stratified random
design to identify participants for its
National Consumer Panel. About 5,000
Family Economics and Nutrition Review
households are selected based on demo­graphic
criteria1 matched to the U.S.
Census. This study uses the Household
Information Form that reports in<lividual,
household, and geographic characteristics
and the Psychographic Questionnaire
that reports food selection and prepara­tion
practices of the primary food
preparer. The sample consists of 5,551
primary food preparers who participated
in the 1993-94 and 1994-95 panels.
Percentages are weighted to represent the
population of interest Unweighted data are
used for Pearson's chi-square significance
testing of independence between household
budget practices and cost-cutting and meal
planning practices.
Results
Sociodemographic Characteristics
More than half of all food preparers
reported using strict food budgets (see
table). Compared with food preparers
who did not adhere to strict food budgets,
those who did were significantly more
likely to have less education and house­hold
income and more people in their
household. Two-thirds of those who
used strict food budgets had less than
a college education, three-fourths had
household income less than $40,000,
and almost three-fourths had three or
more people in the household. Gender
and race were not significantly different
between those who used and did not use
strict food budgets.
Cost-Cutting Practices
Compared with food preparers who did
not follow a strict food budget, those
who did were significantly less likely
to use four different means to cut food
expenditures (fig. 1): Make a complete
list before shopping (20 vs. 32 percent),
1The demographic factors are census regions,
metro-area size, household size, homemaker's
age, and household income.
1997 Vol.JONo.2
Food preparers' characteristics, by type of food budget, 1993-95
Type of food budget
Characteristics Overall Nonstrict Strict
Sample size 5,551 2,661 2,890
Percent
Gender
Male 46 46 46
Female 54 54 54
Race
White 88 88 88
Non-White 12 12 12
Education*
Less than high school 36 32 39
High school 26 25 27
College
Household income1*
38 43 34
<$20,000 27 22 31
$20,000- $39,000 41 36 44
$40,000 and over 32 42 25
Household size*
One 10 11 9
Two 23 30 19
Three and over 67 59 72
*Difference between food preparers adhering to a strict food budget and other food pre parers was
statistically significant at p !> 0.01 based on unweighteddata.
I Income is rounded to the nearest thousand in the data set.
stock up when their brands were on
sale (5 vs. 9 percent), comparison shop
(14 vs. 25 percent), and redeem coupons
(10 vs. 14 percent).
Why would those on reportedly strict
food budgets not take advantage of
different means of cutting food costs?
A complete shopping list includes
categories such as produce, canned goods,
dairy products, meats, and household
supplies. Having a strict food budget
may compel consumers to limit their
purchases to required items. However,
it is quite possible that these items are
so similar on each shopping occasion
that little variation is permitted-thus,
there is no need for a list. Stocking up on
sale items and comparison shopping
among food stores produce immediate
costs for the consumer (e.g., out-of­pocket
expenditures for sale items and
for public or private transportation).
Thus, the household on a reportedly
strict food budget may be more con­cerned
about current cost than about
long-term savings. Coupons, which are
generally found in newspapers, magazines,
and retail flyers, may not be available to
some families.
35
Nutrition Concerns and
Meal Planning Practices
Food preparers who reported using a
strict household food budget were
significantly more likely than others to
be concerned whether the meals they
served were nutritious, to believe they
made every effort to ensure that family
members ate nutritious foods, and to
believe they prepared each meal to be
nutritionally balanced (fig. 2). There
was no significant difference between
those with and without a strict food
budget regarding their belief that giving
the family a wide variety of foods would
result in proper nutrition. Also, groups
did not differ significantly on the prac­tice
of serving their family nutritionally
incomplete meals "once in a while."
Most households that adhere to strict
food budgets do so in order to ensure
that their money income goes further
in covering all their expenses. To deter­mine
whether income level affected
differences between those on and not
on strict food budgets, further analysis
was undertaken on the three significant
variables: Worrying if meals were nutri­tious,
making every possible effort to
see that the family eats nourishing food,
and preparing each meal to be nutrition­ally
balanced. Findings showed that
only one variable (worrying if the meals
were nutritious) at one income level
($40,000 and more) failed to indicate
significant differences between those
on strict food budgets and others (fig. 3).
Possibly, at this income level, a "strict
food budget" takes on a less literal
meaning, and worrying about nutritious
meals concerns being well informed
and pleasing family members and is not
budget-related.
36
Figure 1. Food preparers' cost reduction practices, by type of
food budget, 1993-95
"I make a complete list
before going shopping."1
"When I find a sale on the
brand of food items I like,
I stock up."1
""I save a lot of money by
shopping around for food
bargains."1
"When I get coupons,
I almost always redeem
them."1
Pe3rcent 2- 20 L_j
2:~5 - 14u-
14.
10 o-
• Nonstrtct food budget 0 Strict food budget
1 Difference between food preparers adhering to a strict food budget and other food preparers was statistically
significant at p ~ 0.01 based on unweighted data.
Figure 2. Food preparers' nutrition concerns and meal planning
practices, by type of food budget, 1993-95
'I frequently worry about
whether the meals I serve
are really nourishing."1
"I feel that if I give my family
a large variety of foods, they
will get the proper nutrition."
"I make every possible effort
to see that my family eats
really nourishing foods."1
"I prepare each meal to be
nutritionally balanced. •1
'Once in a while, I serve my
family meals that are not
nutritionally complete."
Percent
84
81 L-------------------~
90
94L-----------------------~
61
74L-----------------~
91
90
~------------------------~
• Nonstrtct food budget D Strtct food budget
1 Difference between food preparers adhering to a strict food budget and other food preparers was statistically
significant at p s 0.01 based on unweighted data.
Family Economics and Nutrition Review
Conclusions and Implications
Food preparers may use a variety of
means to cut food costs yet, at the same
time, remain concerned about meeting
their family's nutritional needs. Findings
in this study show that those who are
following a strict budget are not willing
to sacrifice nutrition. Therefore, they
should be receptive to nutrition promo­tion
strategies that focus on economical
ways of providing their households with
a healthful diet.
Findings also indicate that several of the
most commonly recommended planning
and budgeting tools for food shoppers
are not widely used by those who report
being on a strict food budget. It may
be that families on a strict food budget
cannot spare the money to "stock up"
during sales or the time and/or transpor­tation
to comparison shop among
neighborhood food stores for bargains.
However, the strict food budgeters may
employ other food shopping strategies not
identified in this study.
Additional behavioral research is needed
to determine which factors influence
consumers' use of various cost-cutting
methods. Formative research, such as
focus group interviews, may be helpful
in identifying food shopping practices
and attitudes toward common budgeting
recommendations among low-income
food preparers. This information could
then be used by programs for low-income
consumers on household budgeting and/
or dietary improvement, such as the
USDA Expanded Food and Nutrition
Education Program.
1997 Vol.JO No.2
Figure 3. Food preparers' nutrition concerns and meal planning
practices, by type of budget and income, 1993-95
Income 1 Percent "I frequently worry about whether the
meals I serve are really nourishing."
<$2o,ooif 34
48
$20,00~ 39
$39,00 47
$40,000 and over ~~~
"I make every possible effort to see that
my family eats really nourishing foods."
<$2o,ooif
89
93
$20,00~ 89
$39,00 95
$40,000t 90
and ove 93
"I prepare each meal to be
nutritionally balanced.·
<$2o,ooif
61
73
$20,00~ 57
$39,00 72
$40,000t 63
and ove 78
• Nonstrict food budget 0 Strict food budget
11ncome is rounded to the nearest thousand in the data set.
2Statistically significant at p 5 O.Q1 based on unweighted data.
References
1. Frazao, E. 1995. The American Diet: Health and Economic Consequences. U.S.
Department of Agriculture, Economic Research Service. Agriculture Information
Bulletin No. 711.
2. Lutz, S.M., Smallwood, D.M., and Blaylock, J.R. 1995. Limited financial resources
constrain food choices. F oodReview 18( 1 ): 13-17.
37
38
Research Summaries
The State of
Nutrition Education
in USDA
On November 15, 1996, an intra­departmental
working group of repre­sentatives
from USDA agencies with a
mission area related to nutrition educa­tion
submitted its report to Secretary
Glickman. The State of Nutrition
Education in USDA-A Report to the
Secretary was the culmination of a
yearlong effort facilitated by the Center
for Nutrition Policy and Promotion
(CNPP). The working group assessed
the successes as well as weaknesses
of USDA's nutrition education efforts,
reviewed characteristics of effective
contemporary delivery methods, and
analyzed trends in programs, legislative
history, and funding. The recommenda­tions
provide a useful prelude to strategic
planning for the 21st century.
A Renewed Vision for
Nutrition Education
USDA has long been committed to
improving the nutritional health of
Americans through a program of
research and education t~ maintain a
food supply of high nutritional quality
and to encourage consumption of a
healthful diet (fig. 1). In the summer
of 1995, USDA reconfirmed nutrition
as the link between agriculture and
health and espoused a renewed vision
for nutrition education that would
ensure the Department's leadership
role in Federal nutrition education
into the 21st century. This new vision
challenged USDA to revitalize its
nutrition education plan and to establish
priorities for future nutrition education
efforts.
With the 10-year tenure for USDA's
1986 Comprehensive Plan for a National
Food and Human Nutrition Research
and Education Program about to expire,
the impetus for an assessment of the
current state of nutrition education in
USDA was created. There were other
influential facts that were critical to
USDA's decision to take action.
Scientific evidence increasingly suggests
that poor diet plays an important role in
the onset of chronic diseases and other
health conditions; in many cases, medical
costs and lost productivity might be
avoided by an improved diet. Research
indicates that nutrition education can
help improve diets when behavioral
change is set as a goal and when educa­tional
strategies are designed to include
behavioral change. Motivating people to
change their behavior is seen as the key
because nutrition education that merely
emphasizes dissemination of informa­tion
and teaching of skills is seldom
sufficient to change dietary behavior.
Also, substantial changes have occurred
in communications, technology, and
consumer demographics, lifestyles, and
health status. Finally, advances in nutrition
science and food technology require
updating both delivery and content of
nutrition education to reach the public
effectively.
Background Information
The Working Group developed and
adopted a framework for describing
USDA's contributions to the nutrition
education process (fig. 2). Additional
information was compiled on the history
of nutrition education activities at USDA,
key legislation ·authorizing USDA to
conduct nutrition education, data on
USDA's annual nutrition education
expenditures, and information on some
current nutrition education program
activities at the Federal, State, and
Family Economics and Nutrition Review
Figure 1. Major milestones in nutrition education at USDA
1860
1870
1880
1890
1900
1910
1920
1930
1940
1950
1960
1970
1980
1990
'62 USDA formed
'90 W.O. Atwater-Human nutrition research
'02 Atwater-Variety, Balance, and Moderation
'14 Cooperative Extension Service
'16 Caroline Hunt-First food guide
'33 Food Plans at 4 Cost Levels
'41 National Nutrition Conference for Defense
'46 School Lunch Program began
'56 Basic Four Food Guide
'64 Food Stamp Program began
'69 White House Conference on Food, Nutrition and Health
'70 Expanded Food and Nutrition Education Program began
'71 Food and Nutrition Information Center formed at the National Agricultural Library
'75 Special Supplemental Nutrition Program for Women, Infants, and Children began
'77 Food and Agriculture Act of 1977, Nutrition Education and Training Program began; USDA named
"lead agency" for nutrition research, extension, and teaching
'80 Dietary Guidelines for Americans first issued
'82 Joint Subcommittee on Human Nutrition Research defines "nutrition education research"
'86 USDA Comprehensive Plan for Human Nutrition Research and Education
'90
'90
National Nutrition Monitoring and Related Research Act
Nutrition Labeling and Education AcVNational Exchange for Food Labeling Education
'92 Food Guide Pyramid
'94 Nutrition and Food Safety Education Task Force
'95 Dietary Guidelines for Americans, 4th edition
1997 Vol.JO No.2 39
Figure 2. Interrelationships in USDA Nutrition Education Activities
NUTRITION EDUCATION RESEARCH
NUTRITION SCIENCE RESEARCH (Psychology-knowledge, atti