Content uploaded by Beatrice Rogers
Author content
All content in this area was uploaded by Beatrice Rogers on Mar 16, 2014
Content may be subject to copyright.
ORIGINAL PAPER
“He said, she said”: who should speak for households
about experiences of food insecurity in Bangladesh?
Jennifer C. Coates &Patrick Webb &Robert F. Houser &
Beatrice Lorge Rogers &Parke Wilde
Received: 26 June 2009 / Accepted: 30 January 2010 /Published online: 2 February 2010
#Springer Science+Business Media B.V. & International Society for Plant Pathology 2010
Abstract This paper examines the extent to which males and
females from the same household respond differently to
household food insecurity questions, and explores the reasons
for these differences and the impact for measurement. The
data derive from the 2001–2003 Bangladesh Food Insecurity
Measurement and Validation Study. Male and female enu-
merators administered the food insecurity questionnaire to
women and men in the same household during three survey
rounds and debriefed a subsample of men and women
regarding their response discrepancies. The rate of discor-
dance in male-female responses to individual items was
examined using contingency tables. Potential explanations for
the discordance were informed by the joint respondent
debriefing. These hypotheses were assessed through an
examination of response patterns. To assess the impact of
discordance on measurement, female and male responses to a
scale of 13 food insecurity items were compared and the
degree of differential classification was assessed. On average
the rate of discordance was 15%, but it ranged for particular
items from less than 1% to upwards of 53%. Item content
interacted with gender to produce discordance; women and
men seemed to respond differently due to separate spheres of
responsibility within the same household, power imbalances
influencing intra-household food allocation, and because men
seemed to take more psychological responsibility for ensuring
the household food supply. Nearly one-third of households
were classified in a different food security category using
female versus male responses to the items. The results suggest
that the household food insecurity construct is not as useful in
places like Bangladesh where certain food insecurity-related
manifestations are not collectively or similarly shared by
members of the same living space. Individual-level measures
of food insecurity are needed to complement household data,
along with surveys that allow for proportionate representation
of potentially vulnerable individuals with different demo-
graphic characteristics across the population.
Keywords Gender .Household food insecurit y .
Measurement .Scale
Introduction
Food security is commonly defined as a state in which “all
people, at all times, have physical and economic access to
sufficient, safe and nutritious food to meet their dietary
needs and food preferences for an active and healthy life”
(FAO 1996). Despite the primacy of the individual’s well-
being in this widely accepted World Food Summit
definition, food insecurity measurement efforts have tradi-
tionally assessed the problem at the household and national
levels. As policy is shaped by the ways in which problems
are framed and measured, excluding individual experience
from food security assessment risks misestimating the
magnitude and nature of the problem and overlooking the
intra-household implications of the policy response.
This has been the case with a common approach to food
insecurity assessment, which measures a household’s self-
reported experience of food insecurity along a scale of
severity. Household food insecurity scales are used to
calculate official estimates of food insecurity in the United
States, and are increasingly used by governments in Latin
America and in other developing countries (see Nord et al.
J. C. Coates (*):P. Webb :R. F. Houser :B. L. Rogers :
P. Wilde
Gerald J. and Dorothy R. Friedman School of Nutrition Science
and Policy, Tufts University,
150 Harrison Avenue,
Boston, MA 02111, USA
e-mail: jennifer.coates@tufts.edu
Food Sec. (2010) 2:81–95
DOI 10.1007/s12571-010-0052-9
2009; Perez-Escamilla et al. 2004; Perez-Escamilla et al.
2007). These food insecurity survey modules intend to
capture food insecurity at the household level by interview-
ing one household member, often an adult female. The
female’s responses to questions about their subjective
experience of different aspects of the phenomenon are then
extrapolated to the entire household (eg. Derrickson and
Anderson 2000; Studdert et al. 2001; Welch et al. 1998;
Perez-Escamilla et al. 2004). The assumption is that the
individual’s responses will provide a complete and accurate
picture of the household food situation.
This approach was recently upheld in a report published
by the panel commissioned to review the U.S. Department
of Agriculture’s measurement of food insecurity and hunger
(National Research Council 2005). The report concluded
that: “the concepts of food uncertainty and food insuffi-
ciency are really household-level concepts. Each implies
decisions about household resource allocation (e.g., how
much of a limited budget can be spent on food compared
with other goods and how much of the food budget is spent
for food for different household members). Worrying about
having enough money to pay for food is a response that
considers constraints on the household’s resources. Cutting
meal size and not being able to afford a balanced meal are
also adaptations made with consideration of the entire
household’s resources”(National Research Council 2005
p.32). The review panel’s interpretation of household
behavior follows a unitary model, implying that there is
one decision-maker who is acting for the benefit of all
household members, that resources are pooled, and that
worry about food is a collective experience. However, the
unitary model of the household has been widely challenged,
both in and outside the United States, in favor of a more
nuanced understanding of power imbalances and differ-
ences in domains of responsibility that influence resource
allocation within the same living space (see Alderman et al.
1995; Quisumbing et al. 2000a,b; Haddad et al. 1996;
Lundberg et al. 1997).
A key question, then, relates to the extent to which
current household-based measurement approaches that rely
on a single respondent mistakenly classify a household as
food secure when individual members are not. Rose (1999)
observed that US Food Security Survey Measure (HFSSM)
data, based on a female respondent’s self-report, are
inconclusive regarding the food security condition of the
other household members. While the measure may be able
to differentiate between adults and children due to
questions that specifically ask about their condition, the
differential situation of women and men living in the same
household is left entirely unexplained. Even in countries
where women are traditionally disempowered relative to
men, it may not be true that men are always more food
secure. A recent study of non-income poverty indicators in
Bangladesh demonstrated that impoverished men consis-
tently reported significantly higher levels of anxiety-related
symptoms—losing sleep over worry, feelings of unhappi-
ness, strain—than women (Mujeri 2004), suggesting that
men may be affected disproportionately by certain mani-
festations of food insecurity. Their wives, traditionally the
ones to respond to food and nutrition surveys, may or may
not be aware of their suffering. Though there exists a
significant literature related to intra-household food alloca-
tion, the food security measurement literature has not
engaged with these issues, opting to apply new measure-
ment methods to the household only.
In order to address these gaps, this study for the first
time compares the way in which males and females within
the same households respond to questionnaire items that
form a food insecurity measurement scale. The paper seeks
to answer the following questions: 1) to what extent do
males and females across the population and in the same
household respond discordantly to the same food insecurity
scale questions? 2) Do certain characteristics of the
questions themselves explain the observed discordance?
And, 3) what are the implications of household level male-
female response differences for household food insecurity
measurement?
This study was conducted as part of a multi-year
research effort by the Tufts University Friedman School
of Nutrition Science and Policy in Bangladesh, funded by
the United States Agency for International Development
(USAID) Food and Nutrition Technical Assistance
(FANTA) Project to study the adaptation and validation of
US approaches to measuring food insecurity in developing
country settings (see Coates et al. 2003; Swindale and
Bilinsky 2006). Through multiple survey rounds in 32
upazilas (subdistricts) in Bangladesh, the experience of
individual and household food insecurity in Bangladesh
was explored, and a food insecurity scale was developed
and extensively validated (see Coates et al. 2003; Coates et
al. 2006). In each survey round, women and men in the
same household were interviewed separately about the
household’s food insecurity experiences and a subset was
debriefed together to discuss the reasons for any discrep-
ancies in their responses. The setting in Bangladesh
provided the opportunity to assess issues that continue to
challenge efforts of the US and developing countries to
measure food security, while also offering important lessons
about the similarities and differences in the food insecurity
experience across dramatically different contexts. These
data represent one of the first attempts to empirically study
the gendered responses to food insecurity survey questions
and to determine the implications of relying on a single
respondent to draw household-level food security inferen-
ces. The study posited that significant response discordance
by gender on individual scale items or differential classi-
82 J.C. Coates et al.
fication of a household’s food insecurity status would suggest
the need to re-evaluate common approaches to defining and
measuring household food insecurity, both in Bangladesh and
in other countries such as the United States.
Data collection methods
Data for this paper are derived from the Bangladesh Food
Insecurity Measurement and Validation Study (FIMVS), a
three-year research initiative intended to develop and test a
process for adapting the US food insecurity scale to
developing countries. A complete account of the data
collection methods has been reported elsewhere (Dunford
2000; Coates et al. 2006). In brief, the Bangladesh FIMVS
was conducted in four phases. The first phase used
qualitative methods to explore the relevance of the US
experience of food insecurity to Bangladesh and to develop
a prototype food insecurity instrument. During the second
phase, the food insecurity instrument and a comparator
indicator questionnaire, (that collected data on demograph-
ics, anthropometry, morbidity, income sources, assets,
dietary intake and expenditure), were administered to men
and women in 600 randomly sampled households in rural
areas and small urban centers across the north, center and
southern parts of Bangladesh. The third phase re-
administered the survey to a subsample of 120 households.
The primary aim of this third phase, which commenced one
year after the first round of survey data collection, was to
investigate nuances in respondent interpretations of food
insecurity module questions by gender. Male and female
respondents in a sub-sample of 120 households were
administered the quantitative questionnaire, and were then
debriefed both individually and as a pair regarding possible
reasons for discrepancies in their answers to the food
insecurity scale questions. Enumerators conducted the
debriefing using a semi-structured checklist designed to
facilitate these interactions. The fourth phase of the study
resurveyed the original 600 households to assess changes in
household level food insecurity over the two year period.
The survey rounds were implemented approximately one
year apart over the period 2001–2003. The data for this
paper are drawn from the first round survey of men and
women in 600 households (576 male-female pairs after
accounting for missing data from 24 male respondents), and
are complemented by information obtained from the
debriefing of 120 male-female pairs about their discordant
responses following the second round survey.
The data from the surveys were entered and analyzed in
SPSS 14.0 and the debriefing checklists were translated into
English by senior staff of the survey research firm Data
Analysis and Technical Assistance, Ltd. (DATA), assisted
by a bilingual Bangladeshi masters-level student.
Table 1presents key respondent- and household-level
sample characteristics. The average household was com-
prised of around 5 people, with one income earner for every
two members. The vast majority of sample households
(93.3%) were situated in rural areas. Remaining households
were classified as urban and were located primarily on the
peri-urban edge of smaller district-level centers. Function-
ally landless households, commonly defined in Bangladesh
as having landholdings of less than 0.5 acre (0.2 hectares),
comprised nearly half the sample. Eighty-two percent of the
sample was Muslim, slightly less than the national-level
prevalence of 88% reported in the 1991 census
1
. Nearly all
of the remaining households were Hindu (17.5%). Though
the literacy rate of male respondents was 9.4 points higher
than that of female respondents (49.7 vs. 40.3%), house-
holds in which both male and female were literate (27.1%)
were more common than households in which the male was
literate and the female not (22.6%).
Analytical approach
Degree of discordance
The first question that this research sought to explore was
the degree to which men and women in the same house-
holds and across the population responded differently to
questions about the experience of food insecurity. These
questions were adapted from the US Household Food
Security Survey Measure and related to four broad categories
of the food insecurity experience: 1) insufficient quantity of
food consumed, 2) inadequate quality of food consumed, 3)
uncertainty and worry about accessing food, 4) accessing
food in social or personally undesirable ways.
Respondents had been asked to answer each food
insecurity question in terms of the frequency of its
occurrence over the prior 12 months. Though it would be
possible to define concordance, in the strictest sense, as the
extent to which males and females reported having the
same experience at the exact same frequency, in order to
minimize the potential influence of respondent recall error
on the degree of concordance, the responses to each scale
question were dichotomized so that responses of “never”,
or “rarely”were recoded as 0 (indicating that the food
insecurity manifestation was hardly ever experienced) and
responses of “sometimes”,“often”,or“mostly”were
recoded as 1 (indicating that the manifestation was
experienced relatively frequently). As collapsing responses
to ordinal data risks losing useful information, sensitivity
analyses were conducted to assess whether the choice of
cut-off affected the degree of concordance in male-female
1
(http://www.bbsgov.org/ana_vol1/religiou.htm)
Male and female food insecurity 83
responses. These analyses suggested that the cut-off choice
was not significant to the overall results.
Male and female responses to 28 food insecurity
questions were cross-tabulated. The degree of intra-
household discordance (where the female gave an affirma-
tive response and the male did not or vice versa) was
summarized for each question, along with population-level
differences in the frequency of male and female responses
to each item. Standard errors were not corrected for
complex survey design, however the sample size is
adequate as it was calculated to include the design effect.
A Pearson Chi-square test for equality of proportions was
used to determine whether the proportion of females
affirming the question differed significantly from the
proportion of males doing so
2
.
Reasons for discordance
The purpose of the second stage of the analysis was to
investigate possible reasons for discordance in male-female
responses. Prior to analyzing the survey results, researchers
drew on previously collected qualitative data related to
intra-household dynamics in Bangladesh (Coates et al.
2003), and also analyzed the explanations for discordance
given by the sub-sample of male and female respondents
when they were debriefed individually and then together.
These explanations informed a set of hypotheses about
what types of response patterns would likely be observed in
the full data set and why (see below). Actual item responses
were then analyzed descriptively to assess whether the
male-female responses did in fact follow the patterns that
were suggested by these hypotheses.
The results of the debriefing with respondents suggested
two broad types of factors that appeared to explain much of
the discordance. The first was the “question referent”–the
person, people, or thing to which the question referred.
Debriefings suggested that questions asking about the
respondent’s own experience with food insecurity manifes-
tations were likely to produce higher rates of discordance
(but elicit more accurate responses), than questions that ask
the respondent about the experience of the ‘family’,or
‘household’. This issue of the question referent appeared
likely to be salient regardless of the gender of the
respondent.
The second broad factor relates to the “question content”–
the specific food insecurity manifestation (behavior, percep-
tion, or experience) being described in the question itself.
Here, the debriefing discussions suggested that discordant
results were likely when men and women have different
experiences with, and different levels of knowledge about,
food insecurity manifestations. Where responsibilities within
the household are divided rigidly, and where women are
disadvantaged in terms of food consumption, the similarity of
their responses to questions about these experiences are likely
to be affected by their own gendered situation. The concep-
tualization of question ‘referent’and question ‘content’,their
interaction with the gender of the respondent, and the likely
effect on response patterns are described in more detail in the
next section.
Question referent
The first part of this second stage of the analysis examined
whether the question referent influences the degree of
discordance in the responses of males and females living in
the same household. Questions in the survey were asked to
male and female respondents using three different types of
referents. Type A question referents asked the respondent
about his or her own experience. For instance, a question
with a Type A referent might ask whether the respondent
personally had to eat less food or borrow food from the
neighbors. The second type of question, (B), asked the
respondent about the experience of any individual in
“the family”. An example of a question with a Type B
referent asks whether ‘anyone in the family could not eat
balanced meals’, or whether ‘anyone in the family had to
2
The Kappa statistic, though often used to assess the significance of
inter-rater agreement, was deemed inappropriate for this purpose since
household food security questions asked primarily about individual
experiences or a collection of experiences rather than a single situation
that different people can perceive.
Table 1 Key characteristics of sample households and male and
female respondents
Characteristic NValue
a
Average household size 600 5.3 (± 2.4)
Earner dependency ratio
b
600 1.9 (± 1.4)
Average age household head 587 43.9 (±13.9)
Female household headed (%) 24 4.1
Income sources per HH 600 8.2 (± 3.3)
Religion (%)
Muslim 494 82.3
Hindu 105 17.5
Other 1 0.2
Urban 40 6.7
Rural 560 93.3
Landless (<0.5 acre) 296 49.3
Literacy
Male 297 49.7
Female 242 40.3
Tufts FSNSP/FANTA data (2001)
a
Values are percentages or means (±SD)
b
Dependency ratio is calculated as # of non-income earners/# income
earners
84 J.C. Coates et al.
purchase rice often’. Type C question referents ask about a
common household phenomenon, meaning one that is not
particular to an individual household member or members.
The household catching on fire or the family cow having a
calf are examples of phenomena that occur at the household
level. In this study, the question ‘did the household run out
of food,’is the only one classified as having a Type C
referent
3
. Based on this typology, we hypothesize that
respondents will be more likely to concur on questions that
are asked about family or household level referents (Types
B and C), as they presumably represent a more common set
of phenomena than the individual experience reported in
Type A questions.
To examine this hypothesis, male and female responses
to the food insecurity questions were cross-tabulated. Next,
the results for each question were grouped according to
referent type in a summary table. Within each referent type,
the questions were also arranged according to the severity
of the question
4
. The average degree of concordance
between males and females in the same household was
computed as an indication of the extent to which the type of
referent in the question appears to influence concordance.
Also included in the table is information on overall
concordance (“both yes”or “both no“), the percentage
discordance, and the percentage of the discordant responses
in which male said “yes”but the female said “no”and,
alternatively, in which the female said “yes”but the male
said “no”. Following this step, a subset of responses to 5
questions from Round 3 household survey data were also
cross-tabulated and analyzed. These questions had been
asked of both male and female respondents in two different
ways –at the individual (Type A) level and at the family
(Type B) level. The results of this analysis were intended to
enable a direct assessment of the effect on discordance of
asking the same questions using two different referents.
Question content
The next step in this second stage of the analysis sought to
explore whether the ‘content’of the question in relation to
the gender of the respondent influences response discor-
dance and explains discordance where it exists. Questions
in the food insecurity questionnaire can be grouped
according to three different content categories: 1) strategies
to augment the household resource base (e.g., taking food
on credit, selling or mortgaging questions), 2) strategies to
“make do”by reducing the quantity and quality of food
consumed (eg. skipping meals, eating fried wheat) 3)
perceptions of insecurity (e.g. worrying about food).
Respondent pair debriefings suggested that questions
about strategies to augment the household resource base
may elicit discordant responses because women and men
have largely separate spheres of responsibility within the
household and society. Interviews with Bangladeshi men
and women about their interpretation of the food insecurity
questions suggested that the respondents may not engage at
all in the type of resource augmentation strategy being
asked of them (for reasons of gender rather than food
insecurity) or may not be aware of the resource augmen-
tation strategies undertaken by the spouse (and the male or
female may even hide certain activities from the other). In
the sample households, men were reportedly more likely to
participate in the public realm (eg. shopping) and they were
more likely to have control over available cash resources in
order to do so. Women, on the other hand, described having
more decision-making control and responsibility over
resource augmentation strategies that are close to home –
such as borrowing food from the neighbors.
Discordance may also be evident in food insecurity
questions related to curtailing both the quantity and quality
of food consumed. Particularly in food insecure house-
holds, men and women may receive different food benefits
due to power differences influencing intra-household food
allocation. While the US HFSSM included questions about
child insecurity assuming that no child would go hungry
unless the parent had already done so, in Bangladesh the
issue is not only one of protecting children but of buffering
the income earner as well. As the primary income earner is
often male, when pressed, women may be more likely to
report cutting their own consumption in surveys more often
than men.
A third distinct question content area pertains to
perceptions of insecurity (“worrying about where food
would come from”). Though there was only one question in
this survey explicitly about perceptions of food insecurity,
debriefing data suggested that there may be a gendered
element to the way in which men and women are likely to
respond. Men and women may divide the ‘responsibility’of
worrying about food –for instance, it is possible that men’s
4
The degree of severity of a question is implied by the response
frequency of questions in which both men and women responded
positively. The more frequently the question is jointly affirmed, the
less severe it is assumed to be. For instance, the question “worried
about food”, with a combined frequency of 27%, is considered to be a
less severe (more frequent) indication of food insecurity than, say,
“skipping entire meals”(8%). Since very severe questions are likely to
exhibit very high concordance as a statistical artifact, this ordering by
severity is intended to assist the reader in discounting ‘severity’as an
actual effect influencing overall concordance.
3
Note that in this survey, the question referent does not always
correspond to the level at which the phenomenon took place. For
instance, one might consider “taking a loan”to be a household level
phenomenon, but the question referent was a Type A (“you”).
Similarly, eating high quality food is ultimately an individual level
event, but the question referent was often phrased as a Type B (“the
family”) rather than “you”or, alternatively, “the working adult”.
Male and female food insecurity 85
‘role’as income earner and food provider includes the task
of ‘worrying about where food will come from’, whereas
women’s responsibility may be restricted to worrying about
how to ensure that food is allocated according to whatever
distribution norms are followed in the household. Or, males
may tend to perceive their situation more negatively than
females. Therefore, we hypothesize that men will respond
affirmatively to questions about worrying over food more
often than their female partners when they disagree.
In order to examine the relationships between question
content and response discordance, the questions were
grouped according to the three different types of question
content defined from the qualitative interviews: ‘resource
augmentation’,‘reduction of food quantity and quality’or
‘perceptions of insecurity’. Resource augmentation ques-
tions were divided into two groups: those that debriefing
data indicated as falling primarily into the male sphere of
responsibility and those that were likely to be primarily
under the female’s sphere of responsibility. The responses
of male-female pairs were then cross-tabulated and sum-
marized in a table by type of question content to present the
percentage question concordance (“both yes”or “both no”),
the percentage discordance, and the percentage of the
discordant responses in which the male said “yes”but
the female said “no”and in which the female said “yes”but
themalesaid“no”. Question-level results were then
examined by content type to determine whether actual
response differences indeed appeared to be explained by
this categorization.
Though the emphasis of this particular set of analyses is
on the rate of discordance between responses of men and
women in the same household, it is worth noting that
discordance is not necessarily only due to the relationship
between the question and the gender of the respondent.
Certain individual, respondent pair, or household-level
characteristics may interact with gender to magnify the
extent of discordance. Respondent recall or other measure-
ment error could also play a role in contributing to this
effect. These other potential explanatory factors did not
arise in the debriefing sessions and thus are not formally
assessed here. However, the relative contribution of
different household and individual characteristics, aside
from gender-question interactions, are important issues that
will be further examined as part of ongoing research.
Discordance in a response does not always imply
inaccuracy, nor does concordance always imply accuracy.
If males and females are questioned about their individual
experiences, then discordance could signify merely a
different experience for the man than for the woman.
Discordance only suggests inaccuracy when the question is
thought to be about a collective or shared experience in the
household. With this type of question, discordance may
either be a sign of the inaccuracy of one or both of the
responses, a sign that the experiences are not as shared as
previously thought, or that either the male or female is
being asked to respond to a poorly targeted question –a
question that is more relevant to the experience and
responsibility of one gender than the other. Concordant
responses could also be inaccurate, as in the case where
both the male and female adult respondents are convinced
that other family members were not hungry at bedtime
when in fact they were.
It is important to note that this study was designed to
quantify discordance and to explore whether potential
explanations for disagreement suggested by respondent
debriefing seemed to be upheld by the data. It was not
designed to detect conclusively whether one or both of the
responses were accurate. Short of direct observation of
individuals over the one-year recall period, this would not
have been possible. However, by assessing hypothesized
explanations for discordance through an analysis of actual
response patterns and a discussion with respondents about
their discordant answers, the researchers intended for the
results to be suggestive of which, if any, of the responses
were more or less likely to be accurate and why.
Implications of discordance for measurement
The final step in the analysis was intended to assess the
implications of discordance in male-female responses for
measurement procedures of household food insecurity that
typically rely on information provided by a single female
respondent. First, in order to test how differently men and
women in the same household would score on a set of food
insecurity questions, a food insecurity scale score was
calculated by totaling affirmative responses to a sub-set of
thirteen questions that previous analysis had demonstrated,
based on female responses only, to have high internal
reliability and validity (Coates et al. 2003). The mean
difference in the number of questions affirmed by men
versus women in the same household was computed and its
significance assessed with a paired t-test. Since the primary
interest was not only in comparing overall scores but in
assessing concordance on these 13 questions, the mean and
median of the number of questions that were responded to
concordantly was also calculated. Because food insecurity
scales, in practice, are often transformed into prevalence
estimates of food insecurity status, a separate male and a
female food insecurity categorization for the household was
created based on terciles of female food insecurity scores.
The boundaries of the terciles were the same for the men’s
and the women’s response scores. These variables denoting
status of male and female household food insecurity were
cross-tabulated to determine the extent to which households
were similarly classified using the two different sets of
responses.
86 J.C. Coates et al.
Results
Table 2shows the results of the assessment of response
differences within the household and across the population.
The first column illustrates the percentage of male-female
pairs from the same household that gave different responses
to each food insecurity question. The second and third
columns illustrate the percentage of women in the sample
that affirmed (said “yes”to) each food insecurity question
compared to the percentage of men affirming the same
question. The fourth column shows the difference between
these two percentages.
Discordance between male-female pairs ranged from just
1% for the question ‘sold or mortgaged things’to almost
53% for the question ‘could not buy snacks for the family’.
Nine of the 28 questions had a level of discordance of 20%
or greater. Six questions had a level of discordance greater
than 10% but less than 20%, and the remaining 13
Table 2 Percentage male and female respondents affirming Bangladesh food insecurity questions, sorted in order of male-female discordance
Q. ID No Question Household-level Discordance Population Level Differences
(M = yes and F = no,
F = yes and M = no)
Females
(% yes)
1
Males
(% yes)
1
Dif
(% point)
Sig.
2
1 Personally could not buy snacks for family 52.9 66.8 20.5 46.4 0.000
2 Personally took food on credit from a local shop 32.8 20.8 41.5 −20.7 0.000
3 Personally borrowed food from neighbors 31.2 31.1 13.4 17.7 0.000
4 Personally could not give kids money for snacks 27.8 25.3 13.7 11.6 0.002
5 Family could not cook rich food 27.1 65.5 52.3 13.2 0.000
6 Personally worried about where food would
come from
26.6 36.1 45.3 −9.2 0.000
7 Family did not eat big fish 26.2 47.7 37.2 10.6 0.000
8 Family not eat meat 25.0 54.3 38.0 16.3 0.000
9 Personally ate less food 20.8 45.8 37.2 8.7 0.000
10 Food stored ran out 18.6 25.9 26.4 −0.5 0.000
11 Personally skipped entire meals 18.6 22.9 13.4 9.6 0.000
12 Personally had to eat wheat 15.1 18.1 18.2 −0.2 0.000
13 Main working adult skipped entire meals 13.9 12.9 13.5 −0.7 0.000
14 Family purchased rice 12.2 56.1 53.3 2.8 0.000
15 Personally used money from another purpose
for food
10.9 2.1 3.7 −1.6
3
0.381
16 Family ate few balanced meals 9.4 11.3 7.8 3.5 0.000
17 Family sought charity meat 7.6 17.9 21 −3.1 0.000
18 Personally drank rice starch 6.8 6.9 4.3 2.6
3
0.000
19 Personally borrowed food to serve to
special guests
6.6 3.3 3.9 −0.7
3
0.139
20 Personally received or sought charity 6.3 8.2 6.8 1.4
3
0.000
21 Personally ate wild taro 5.7 4.9 1.6 3.3
3
0.015
22 Personally ate broken rice 5.2 10.4 11.9 −1.6 0.000
23 Personally ate sweet potato 4.5 3.3 4.3 −1.0
3
0.000
24 Personally ate fried wheat 4.2 3.8 4.2 −0.4
3
0.000
25 Personally drank flour and water 2.1 1.4 1 0.4
3
0.001
26 Personally did not eat for an entire day 1.6 1 0.9 0.2
4
0.000
27 Personally borrowed money from moneylenders 1.2 0.4 1.2 −0.8
4
0.000
28 Personally sold or mortgaged own things 1 0.7 1 −0.4 0.000
Tufts FSNSP/FANTA data (2001)
1. Question response frequencies were recoded to dichotomous variables. Affirmed = response of “often”“mostly”,or“sometimes”; Not affirmed =
response of “never”or “rarely”
2. Pearson Chi-square test for equality of proportions
3. 1 cell with fewer than 5 cases
4. 2 cells with fewer than 5 cases
Male and female food insecurity 87
questions had a rate of discordance under 10%. Across all
questions, the average discordance was 15%.
Looking across the sample, men reported 13 food
insecurity manifestations more often than women, while
women reported experiencing 15 of the manifestations
more commonly than men. For instance, nine percent more
men than women reported ‘worrying about where food
would come from’and 20.7% more men than women
reportedly ‘took food from the shop on credit’. In other
instances where more men reported the manifestation than
women, the difference in affirmation prevalence between
men and women was smaller, (eg. 0.2% more men than
women reported having to eat wheat rather than rice) and in
two cases (‘borrowed food to serve to special guests’and
‘used money for another purpose to buy) it was statistically
insignificant.
When women affirmed questions more often than men
the differences were often large. For example, 54.3% of
women reported not eating meat as part of a meal as
opposed to just 38% of men. Nearly 70% of women said
they could not purchase snacks for the family compared to
just 20.5% of men. And 16.6% more women than men
reported that the family could not eat meat. Overall, the
results of Table 2suggest that not only is there discordance
between male and female responses within the same
household but that the reason for the discordance is likely
to be related to the gender of the respondent rather than (or
in addition to) to some other household, individual, or
question-level characteristic. The results also suggest that
men were as likely as women to report certain manifes-
tations of food insecurity more often than their spouse. In
other words, women did not consistently report more food
insecurity than men.
Table 3presents the results of the assessment of whether
the question referent seems to explain variation in discor-
dance across questions. Questions in Table 3are grouped
by type (A, B, or C) –that is, according to whether male
and female respondents were asked to respond about their
own experience (A), about the experience or experiences of
other household members (B), or about a single household
phenomenon (C). Within each of these categories, questions
are also ordered according to their relative severity;
questions where a higher percentage of both male and
female respondents gave an affirmative answer were
considered less severe and fall before questions that were
affirmed by both respondents less often.
For Type A questions, the average discordance was
14.1%. For Type B questions, the average was 17.3% and
for type C questions, of which there was only one,
discordance was 18.6%. These initial results suggest that
Type A (individual level) questions might have the lowest,
rather than the highest, average level of discordance.
However, these results may be misleading, since the
questions were not distributed equally in terms of their
severity across Type A and Type B. As discussed in the
methods section, the severity of the question may be an
artifact that causes the greatest concordance at the extremes
(i.e. not severe at all or very severe). Because the Type A
category has more ‘very severe’questions (like ‘did not eat
for an entire day’), taking the straight average may produce
the appearance of higher concordance influenced more by
severity than by type. If instead we examine the average
discordance for all Type B questions compared to the
average for only Type A questions with a level of
discordance greater than the lowest Type B discordance
(8%), this helps to control somewhat for the statistical
artifact. According to this reanalysis, Type B discordance
averages 17.3%, while discordance in the sub-group of type
A questions is much higher, at 26.3%. Even when the Type
A question with the most extreme discordance is removed
(could not buy snacks –59%), the average discordance is
still 5% higher for Type A, individual level, questions than
for Type B, family level questions.
Table 4presents the results of the comparison of those
Round 3 questions that were asked in both the Type A and
Type B format. In every instance, the degree of discordance
was lower for questions asked at the ‘family’(Type B)
level. For instance the discordance was 51.9% when the
question “could not purchase snacks”was asked in
reference to the respondent but 20.0% when asked in
reference to the family. Aside from this one dramatic
difference, the increase in concordance with the different
question format was less pronounced but still notable and
consistent. For example, the discordance was 11.9% when
the question “ate few balanced meals”was asked about the
respondent but 9.9% when asked in reference to the family.
Tab l e 5displays the food insecurity question responses
according to the content of the question. The intention of this
step in the analysis was to investigate the hypothesis,
suggested by debriefing evidence, that the content of the
question interacts with gendered roles within the household
to effect different responses by men and women. Support for
this hypothesis is borne out by the evidence in Table 5.Here,
where there is discordance, men and women responded as
predicted along lines delineated by traditional gender roles.
Men overwhelmingly affirmed food insecurity questions
related to augmenting household resources via the public
realm when their wives did not. For example, of those pairs
that answered discordantly, 81.5% of men responded that
they took credit when their female counterpart did not,
whereas only 18.5% of women reported taking credit when
their male partners did not. In response to questions about
spending cash on snacks, engaging in sales transactions,
taking monetary loans, and seeking food from outside the
household realm, where there was discordance, men were
more likely to affirm the food insecurity-related question
88 J.C. Coates et al.
than the women. On the flip side, in response to questions
about manifestations in the female domain, women nearly
always affirmed the behavior more often then men when
they disagreed. For instance, of those that disagreed about
questions of borrowing food from neighbors, 78.3 % of
women versus 21.7% of men reported having to do so.
Table 3 Discordance in male and female responses to food insecurity questions by question type and severity; N= 576 pairs
Q. ID No Question Concordance %
of N=576
Discordance %
of N=576
Of those with discordance:
Both
“Yes ”
Both
“No”
(%) M = yes
F=no
(%) F = yes
M=no
Type A Questions
a
(Average Discordance=14.2 %) (Average discordance of Type A Questions where discordance >8%= 26.3%)
d
9 Personally ate less food 31.1 48.1 20.8 29.2 70.8
6 Personally worried about where food
would come from
27.4 46.0 26.6 67.3 32.7
1 Personally could not buy snacks for family 17.2 29.9 52.9 6.2 93.8
2 Personally took food on credit from a
local shop
14.8 52.4 32.8 81.5 18.5
12 Personally had to eat wheat 10.6 74.3 15.1 50.6 49.4
11 Personally skipped entire meals 8.9 72.6 18.6 24.3 75.7
3 Personally borrowed food from neighbors 6.6 62.2 31.2 21.7 78.3
22 Personally ate broken rice 5.7 83.3 10.9 57.1 42.9
4 Personally could not give kids money
for snacks
5.6 66.6 27.8 29.1 70.9
20 Personally received or sought charity 4.3 89.4 6.3 38.9 61.1
18 Personally drank rice starch 2.3 90.9 6.8 30.8 69.2
24 Personally ate fried wheat 1.9 93.9 4.2 54.2 45.8
23 Personally ate sweet potato 1.6 93.9 4.5 61.5 38.5
21 Personally ate wild taro 0.4 93.9 5.7 21.2 78.8
19 Personally borrowed food to serve to
special guests
0.4 93.1 6.6 55.3 44.7
28 Personally sold or mortgaged own things 0.4 98.6 1.0 66.7 33.3
25 Personally drank flour and water 0.2 97.7 2.9 41.7 58.3
26 Personally did not eat for an entire day 0.2 98.3 1.6 44.4 55.6
27 Personally borrowed money from moneylenders 0.2 98.6 1.2 85.7 14.3
15 Personally used money from another purpose
for food
0.2 94.4 5.2 64.5 35.5
Type B Questions
b
(Average Discordance=17.3%)
14 Family purchased rice 48.6 39.2 12.2 38.6 61.4
5 Family could not cook rich food 45.3 27.6 27.1 25.6 74.4
8 Family not eat meat 33.7 41.3 25.0 17.4 82.6
7 Family did not eat big fish 29.3 44.4 26.2 29.8 70.2
17 Family sought charity meat 15.6 76.7 7.6 70.5 29.6
13 Main working adult skipped entire meals 6.3 79.9 13.9 52.5 47.5
16 Family ate few balanced meals 4.9 85.8 9.4 31.5 68.5
Type C Questions
c
10 Food stored ran out 16.8 64.6 18.6 51.4 48.6
All Question Types (Average Discordance= 15.1%)
Tufts FSNSP/FANTA data (2001)
a
Type A questions asked the respondent about his or her personal experience.
b
Type B questions asked the respondent about the experience of any individual in the family
c
Type C questions asked the respondent about a common household phenomenon
d
The category “Type A questions”contains more questions that ask about severe food insecurity than Type B (or C) questions. To control
somewhat for the potential influence of question severity on the degree of concordance (likely to cause the greatest concordance at the extremes),
we also look at the average discordance of those Type A questions with discordance greater than the least discordant Type B question.
Male and female food insecurity 89
Where there was discordance among responses to
questions about resource augmentation the observed pat-
terns largely held regardless of whether the question
referred to the individual respondent or the family. For
example, the question about having to seek charity meat
was asked as a Type B (family level) referent. Nearly 80%
of men and only 20% of women affirmed this question in
households that disagreed. The question about having to
take food on credit was a Type A (individual level) referent.
In households where respondents disagreed, 81.5% of men
and versus 18.5% of women reported this behavior.
Table 5also provides support for the hypothesis that men
and women do not give the same answer to ‘food consump-
tion’questions due to receiving different food allocations
within the same household. For instance, where there was
discordance over the question of personally eating less food,
only 29.2% of men said they ate less food when the female did
not. By contrast, when there was discordance, 70.8% of
women reported eating less food when their husbands did not.
Similarly, in 75.7% of households with discordance, women
reported skipping meals where their husbands did not, while in
only 24.3% of households men reported skipping meals where
the female did not. Overall, where women and men gave
different responses to questions about involuntary reductions
in the amount food consumed women reported reducing the
quantity of food much more often than men, again, regardless
of whether the question was asked in reference to the
individual respondent (Type A) or the family (Type B).
The same pattern is largely true for questions related to
consuming food of insufficient quality. Where there was
disagreement, question responses demonstrate that women
were more likely to report eating lower quality/less
preferred food than their male counterparts. For example,
in response to a question about the family not being able to
eat meat, in households with discordance, 17.4 % of
households men said yes and women said no, vs. 82.6%
of cases where men said no and women yes. The same large
differential in favor of male consumption was observed
whether the question is about not eating big fish, not eating
balanced meals, or not eating rich, good food (bhalo
mondo).
However, where the content of the food insecurity
question was related to the consumption of a socially
unacceptable food or other food only consumed during
great scarcity, the male/female differential was less consis-
tent and pronounced. Not only was there less discordance
overall but when there was discordance, women were only
marginally more likely than men to have to eat foods like
flour and water when their counterparts did not, and men
were actually more likely than women to report eating
foods like sweet potato, fried wheat, and broken rice.
However, for these most severe questions, the sample size
is quite small so the evidence that more men than women
are likely to eat socially unacceptable foods is suggestive,
more than conclusive.
ThedatainTable5appear to uphold the third
hypothesized reason for content-related discordance. Where
male-female pairs responded differently to questions about
food worries, in significantly more households the male
reported worrying about food (67.3%), than the female did
(32.7%).
The final step in the analysis was intended to assess the
effect of using the information provided by a single female
respondent on the measurement of household food insecu-
rity using a scale or categorical prevalence indicator. The
results show that females, on average, affirmed significantly
(but only slightly) more questions than men did out of the
13 questions in the scale (3.5 +/−3.2 questions vs. 3.2 +/
−2.9 questions; paired t-test, p<.0001). Males and females
within the same household responded concordantly to an
average of just 10.6 (+/−2.1), or 81%, of the 13 questions.
Table 6presents the results of the cross-tabulation of
male-female score terciles. Cells on the left-to-right
diagonal in the table illustrate the percentage of the total
number of households in which men’sandwomen’s
responses classified households the same way (i.e. both
low food insecurity, both medium food insecurity, or
both high food insecurity). Only 68.8% of households
were classified in the same food security status tercile by
male and female responses. A total of 22.5% of house-
holds were classified differently, but within one tercile,
by male and female responses (e.g. 7.6% of households
that fell into the food secure category based on female
Table 4 Comparison of male-female response discordance from two
approaches to eliciting responses to food insecurity questions from
round 3 data
Question Of those with discordance:
Discordance
(% of total)
M=y,
F=n%
F=y,
M=n%
Personally ate few balanced
meals
11.9 38.3 61.7
Family ate few balanced
meals
9.9 38.0 62.0
Personally could not purchase
snacks
51.9 8.4 91.6
Family could not purchase
snacks
20.0 59.4 40.6
Personally food on credit
from a local shop
41.5 92.4 7.6
Family took food on credit
from a local shop
33.9 68.0 31.9
Personally borrowed food
from neighbors
34.1 4.1 95.9
Family borrowed food from
neighbors
29.4 53.4 46.6
Tufts FSNSP/FANTA data (2003)
90 J.C. Coates et al.
responses were classified as moderately food insecure
using the male responses—a one-category difference).
Finally, just under 2% of households were differently
classified by more than one category –three households
fell in the lowest food insecurity tercile according to
female responses and the highest food insecurity tercile
according to the male; while eighthouseholdsfellinto
the highest food insecurity tercile using female responses
but in the lowest food insecure tercile based on responses
from their male counterpart.
Discussion
The results of this paper have shown that men and women
in Bangladesh do respond differently to questions about
household food insecurity and, though the average degree
of discordance in this study was 15%, the mean masks
differences across the questions ranging from barely 1% to
upwards of 50%. The fact that the degree of discordance
varied so dramatically from one question to the next
suggests that there is something about the questions
Table 5 Gender specific differences related to resource augmentation questions
Q. ID No. Question Concordance %
of N=576
Discordance %
of N=576
Of those with discordance:
Both
“Yes ”
Both
“No”
(%) M = yes
F=no
(%) F = yes
M=no
Resource augmentation questions
Male sphere of responsibility
14 Family purchased rice 48.6 39.2 12.2 38.6 61.4
17 Family sought charity meat 15.6 76.7 7.6 70.5 29.6
2 Personally took food on credit from a local shop 14.8 52.4 32.8 81.5 18.5
28 Personally sold or mortgaged own things 0.4 98.6 1.0 66.7 33.3
27 Personally borrowed money from moneylenders 0.2 98.6 1.2 85.7 14.3
15 Personally used money from another purpose for food 0.2 94.4 5.2 64.5 35.5
19 Personally borrowed food to serve to special guests 0.4 93.1 6.6 55.3 44.7
Female sphere of responsibility
1 Personally could not buy snacks for family 17.2 29.9 52.9 6.2 93.8
3 Personally borrowed food from neighbors 6.6 62.2 31.2 21.7 78.3
4 Personally could not give kids money for snacks 5.6 66.6 27.8 29.1 70.9
20 Personally received or sought charity 4.3 89.4 6.3 38.9 61.1
Consumption reduction questions
5 Family could not cook rich food 45.3 27.6 27.1 25.6 74.4
8 Family not eat meat 33.7 41.3 25.0 17.4 82.6
7 Family did not eat big fish 29.3 44.4 26.2 29.8 70.2
9 Personally ate less food 31.1 48.1 20.8 29.2 70.8
12 Personally had to eat wheat 10.6 74.3 15.1 50.6 49.4
11 Personally skipped entire meals 8.9 72.6 18.6 24.3 75.7
13 Main working adult skipped entire meals 6.3 79.9 13.9 52.5 47.5
22 Personally ate broken rice 5.7 83.3 10.9 57.1 42.9
16 Family ate few balanced meals 4.9 85.8 9.4 31.5 68.5
18 Personally drank rice starch 2.3 90.9 6.8 30.8 69.2
24 Personally ate fried wheat 1.9 93.9 4.2 54.2 45.8
23 Personally ate sweet potato 1.6 93.9 4.5 61.5 38.5
21 Personally ate wild taro 0.4 93.9 5.7 21.2 78.8
25 Personally drank flour and water 0.2 97.7 2.9 41.7 58.3
26 Personally did not eat for an entire day 0.2 98.3 1.6 44.4 55.6
Perceptions of insecurity questions
6 Personally worried about where food would come from 27.4 46.0 26.6 67.3 32.7
Tufts FSNSP/FANTA data (2003)
Male and female food insecurity 91
themselves that affects how men and women will respond.
In this study nearly one-third of households were classified
differently by male and female responses. As suggested by
these results, the issue of which questions actually end up
in a food insecurity measure and who to interview does
matter for assessing food insecurity.
The comparison of questions by referent type suggested
that male and female answers were more concordant when
questions of roughly the same severity level referred to ‘the
family’rather than to the individual respondent. This makes
sense when we think that the ‘family’may represent a
collective, rather than individual, experience that can be
reported by either male or female proxy. But the fact that there
was still discordance when the same questions were asked
about the family suggests there are other factors influencing
discordance aside from the question referent type.
Though it is possible that some portion of the discordant
responses results from measurement error present in any
survey, the debriefing of respondents suggested that discor-
dance happened when either the male or female respondent
was asked about a situation outside of the sphere of
responsibility defined along gendered lines. These findings
also suggested that women and men may respond more
accurately to questions that inquire about their own personal
experience with things they are responsible for within their
socially prescribed, gendered domain. When they are asked
about things outside their personal experience (i.e. about
another family member going without food when the
respondent did not) or about things they are not responsible
for (eg. asking a female how often she had to take food on
credit), the unequal or inaccurate information on the part of
one or both of the respondents may produce discordant
results. Discordance also appeared to result from different
actual or perceived experiences related to hunger and worry
and to the quantity and quality of food consumed.
What does the discordance between male and female
responses imply for understanding and measuring house-
hold food insecurity? As described previously, the panel to
review the U.S Department of Agriculture’s measurement
of food insecurity and hunger concluded that “food
uncertainty and food insufficiency are really household-
level concepts”(2005 p.32). However, to our knowledge,
this assertion was not based on empirical evidence
comparing individual-level or individually reported experi-
ences by the members of the same household, and it
contradicts studies suggesting that the unitary household
model does not hold either in developing country or US
populations (see Alderman et al. 1995; Quisumbing et al..
2000a,b; Haddad et al. 1996; Lundberg et al. 1997). The
results from Bangladesh suggest a conceptualization that
differs from that of the US Review Panel, one more in line
with the complex household realities exhibited in the
comparison of male-female response patterns studied here
and supported by the intra-household food allocation
literature. The picture painted by these results indicates
that male and female adults do not always jointly
participate in the decision to undertake strategies to
augment the food resource base and that male and females
Table 6 Concordance of male and female response-generated food insecurity status
Male Food Insecurity Score Tercile
% (n)1
Female Food Insecurity
Score Tercile
% (n) 1
Food
Insecure --
Low
(0-1)
Food
Insecure --
Medium
(2-4)
Food
Insecure --
High
(5-13)
Total
Food Insecure -- Low (0-1) 31.4 (181) 7.6 (44) 0.5 (3) 39.6 (228)
Food Insecure--Medium (2-4) 6.1 (35) 13.0 (75) 7.6 (44) 26.7 (154)
Food Insecure--High (5-13) 1.4 (8) 8.3 (48) 24.0 (138) 33.7 (194)
Total 38.9(224) 29.0 (167) 32.1 (185) 100.0 (576)
Source: Tufts FSNSP/FANTA survey data (2001); Pearson chi-square=355.921, p=<.0001.
1 Male and Female Food Insecurity Scores calculated from sum of affirmative response to 13 food
insecurity questions assessed previously for content/construct validity and internal consistency reliability
[coded never, rarely=no (0), sometimes, often, mostly=yes (1)
92 J.C. Coates et al.
in the household also differ, relative to one another, in their
ability to command food for their own consumption. Men
and women in Bangladesh appear to experience the
psychological burden of food insecurity differently.
These results underscore an important question –is the
concept of household food insecurity useful if the situation
of individual family members is different, and if few food
insecurity experiences are collectively and similarly
shared? These data suggest that, in Bangladesh at least, it
would be wise to complement any extrapolation of
individuals into household units by developing individual
food security measures. In addition to (or rather than)
sampling households and relying on individuals of a single
sextorepresenttheexperience of all members, food
security surveys should consider sampling procedures that
allow for proportionate representation of respondents with
different demographic characteristics. In these surveys,
individuals should be asked to respond primarily in terms
of their own experience.
Survey questions must also be more gender appropriate –if
both male and females are to be respondents in the survey,
questions should be excluded that are likely to be answered
inaccurately by men or by women due to their separate
spheres of food-related responsibility. On the other hand,
certain questions were shown to be relevant to both genders,
and discordance appeared to be explained by different
experiences of food insecurity within the same household.
These types of questions may, after further validation, be
usefully retained as a relatively simple approach to assessing
the severity of intra-household differentials in food allocation.
In exploring the causes and consequences of food insecurity
in a population, individual characteristics should be tested as
predictors of individual food insecurity status, along with the
characteristics of the household to which the individuals
belong. Though there is likely to be a clustering of experience
within the same household, further research of this kind is
needed to assess the extent to which the experience of
individuals is more similar to others in the same household
than to randomly selected individuals with similar demo-
graphic characteristics in the population.
These recommendations are based on the results of a
developing country measurement and validation study.
However, they also have implications for how the U.S.
continues to conceptualize and measure household level
food insecurity. Given that the U.S. measurement effort has
not yet undertaken any individual comparative analysis like
this one, it would do well to consider these issues in any
further research and revisions it makes to the established
national level measure.
Conflict of Interest The authors declare that they have no conflict
of interest.
References
Alderman H, Chiappori P, Haddad L, Hoddinott J, Kanbur R (1995)
Unitary versus collective models of the household: is it time to
shift the burden of proof? World Bank Res Obs 10:1–19
Coates J, Webb P, Houser R (2003) Measuring food insecurity: going
beyond indicators of income and anthropometry. Washington,
Food and Nutrition Technical Assistance Project, Academy for
Educational Development
Coates J, Wilde PE, Webb P, Rogers BL, Houser RF (2006)
Comparison of a qualitative and a quantitative approach to
developing a household food insecurity scale for Bangladesh. J
Nutr 136:1420S–1430S
Derrickson J, Anderson J (2000) Face validity of the core food security
module with Asians and pacific islanders. J Nut Ed 32:21–30
Dunford B (2000) Baseline survey report (October–November 2000):
food security enhancement initiative: Fy 2000–Fy 2004. World
Vision Bangladesh, Dhaka
FAO (1996) Rome Declaration on World Food Security, World Food
Summit. Food and Agriculture Organisation, Rome
Haddad L, Peña C, Nishida C, Quisumbing A, Slack A (1996) Food
security and nutrition implications of intrahousehold bias: a
review of the literature. International Food Policy Research
Institute, Washington
Lundberg S, Pollak R, Wales T (1997) Do husbands and wives pool
their resources? Evidence from the United Kingdom child
benefit. J Human Resour 32(3):463–480
Mujeri MK (2004) The use of qualitative and quantitative indicators
for local-level poverty assessment: the experience of a pilot
survey in Bangladesh. MIMAP-Bangladesh Project, Dhaka
National Research Council (2005) Measuring Food Insecurity and
Hunger: Phase I Report. Panel to Review U.S. Department of
Agriculture’s Measurement of Food Insecurity and Hunger, Com-
mittee on National Statistics, Division of Behavioral and Social
Sciences and Education. The National Academies Press, Washington
Nord M, Andrews M, Carlson S (2009) Household Food Security in
the United States, 2008. Economic Research Report No. (ERR-
83), US Department of Agriculture, Washington
Perez-Escamilla R, Segall-Correa AM, Kurdian Maranha L, Sampaio
MdFA, Marin-Leon L, Panigassi G (2004) An adapted version of
the U.S. Department of Agriculture Food Insecurity Module is a
valid tool for assessing household food insecurity in Campinas,
Brazil. J Nutr 134:1923–1928
Pérez-Escamilla R, Melgar-Quiñonez H, Nord M, Alvarez Uribe MC,
Segall-Correa AM (2007) Escala Latinoamericana y Caribeña de
Seguridad Alimentaria (ELCSA) [Latinamerican and Caribbean
Food Security Scale]. Perspectivas en Nutrición Humana
(Colombia) (supplement):117–134
Quisumbing AR, Maluccio JA (2000) Intrahousehold allocation and
gender relations: new empirical evidence from four developing
countries. International Food Policy Research Institute, Washington
Quisumbing AR, Brière Bdl (2000) Women’s assets and intrahousehold
allocation in rural Bangladesh: testing measures of bargaining
power. International Food Policy Research Institute, Washington
Rose D (1999) Economic determinants and dietary consequences of
food insecurity in the United States. J Nutr 129:517S–520S
Studdert LJ, Frongillo EA Jr, Valois P (2001) Household food
insecurity was prevalent in Java during Indonesia’s economic
crisis. J Nutr 131:2685–2691
Swindale A, Bilinsky P (2006) Development of a universally
applicable household food insecurity measurement tool: process,
current status, and outstanding issues. J Nutr 136:1449S–1452S
Welch K, Mock N, Netrebenko O (1998) Measuring hunger in the
Russian Federation using the radimer hunger scale. Bull World
Health Organ 76:143–148
Male and female food insecurity 93
Jennifer Coates, PhD. is Assistant
Professor of Food Policy and Ap-
plied Nutrition at the Tufts Fried-
man School of Nutrition Science
and Policy, where she specializes
in food security and nutrition as-
sessment, international food and
nutrition policy spanning both ‘de-
velopment’and ‘humanitarian
emergency’contexts. Dr. Coates
also engages in the evaluation of
large-scale, multi-sectoral pro-
grams and teaches two advanced
graduate-level seminars on these
subjects. While at the Friedman School, Dr. Coates has carried out
evaluations for UN organizations like UNICEF and the World Food
Programme, provided technical assistance to Non-Governmental
Organizations implementing food security, nutrition, and health
activities, and has served as an ongoing consultant to the USAID-
funded Food and Nutrition Technical Assistance (FANTA) Project.
Prior to joining the Friedman School of Nutrition Science and Policy,
Dr. Coates worked with John Snow, Inc. for nearly seven years
managing large, USAID-funded integrated maternal/child health,
nutrition, reproductive health projects.
Patrick Webb is Dean for Aca-
demic Affairs and Alexander
MacFarlane Professor of Public
Policy at the Friedman School of
Nutrition Science and Policy, and
also Professor at the Fletcher
School of Law and Diplomacy.
He worked for six years in the
United Nations World Food
Programme (WFP), most recent-
ly as Chief of Nutrition. At WFP
he had global responsibilities for
emergency nutrition interven-
tions (such as the Asian tsunami
response), oversaw maternal and
child programs in 30 developing countries, and monitored nutrition
trends through national surveys, (including in North Korea). He
served on the Hunger Task Force of the Millennium Project reporting
to UN Secretary General Kofi Annan. Earlier, Dr. Webb spent 9 years
with the International Food Policy Research Institute, stationed mostly
in Ethiopia, Niger and The Gambia, working with national officials on
food and agriculture policy. He has researched many aspects of
malnutrition, humanitarian practice, and food security. His co-
authored book on Famine in Africa (published by Johns Hopkins
University Press in 1999), sold out of its first edition and went into a
second run. Other publications include 20 book chapters and 50 peer-
reviewed journal articles. Patrick holds honorary professor status at
the University of Hohenheim (Stuttgart, Germany) as designated by
the Minister for Education of the State of Baden-Wuerrtemburg.
Robert Houser is an Assistant
Professor and Statistical Pro-
grammer/Analyst at the Fried-
man School of Nutrition
Science and Policy. Dr. Houser
holds a Ph.D. in Psychology
from Tufts University and has
over 20 years experience ana-
lyzing nutrition and public
health data. His varied recent
research contributions include
the study of: Food and nutrition
services in assisted living facil-
ities; Food security, hunger and
food stamp program participa-
tion; Child overweight/obesity and the epidemiological transition in
developing countries; HIV, dietary intake and weight status; Defining
and understanding healthy lifestyle choices for Adolescents; Positive
deviance, Antenatal Nutrition and Birth Weight; Nutrition education
and access to health information; Nutritional impact of school feeding
programs; marketing fresh produce in the workplace; and motivations
for a vegetarian diet.
Beatrice Rogers is a professor
of Economics and Food Policy
at the Friedman School of
Nutrition Science and Policy.
She served as Academic Dean
of the School for 13 years, and
is currently Director of the
school’s Food Policy and Ap-
plied Nutrition Program. Dr.
Rogers’research focuses on
economic determinants of
household food consumption,
including price policy, food
aid, food price subsidies and
income transfers. She has con-
94 J.C. Coates et al.
ducted research on the determinants of intra-household allocation of
resources, focusing on the role of female household headship
among other factors. More recently she has been working on
applying the statistical technique of Small Area Estimation to the
assessment of the distribution of malnutrition prevalence at
geographically disaggregated levels. She is currently conducting
research in three countries on the effectiveness of alternative exit
strategies - techniques to ensure the sustainability of food aid
program impacts after the programs themselves have shut down.
Most of her research has been in less developed countries,
including Pakistan, India, Thailand, the Philippines, Cameroon,
Dominican Republic, Jamaica, Republic of Mali, Morocco,
Mozambique, Honduras, Bangladesh, Bolivia, Peru, and Brazil,
but she has also conducted research in the US on food stamps
and their relationship to household food security. Dr Rogers holds
a PhD from the Heller School of Social Welfare Policy at
Brandeis.
Parke Wilde is a food economist
in the Food Policy and Applied
Nutrition (FPAN) program at the
Friedman School of Nutrition Sci-
ence and Policy at Tufts Universi-
ty. With a Ph.D. in agricultural
economics at Cornell, he worked
five years for the U.S. Department
of Agriculture’s Economic Re-
search Service. In 2003, he joined
the faculty of the Friedman School,
where he teaches graduate-level
courses in statistics and U.S. food
policy. His research addresses food
security and hunger measurement,
food assistance programs, and the economics of the food environment
facing low-income Americans.
Male and female food insecurity 95
