View
0
Download
0
Category
Preview:
Citation preview
REVIEW Open Access
Determinants of intra-household foodallocation between adults in South Asia – asystematic reviewHelen Harris-Fry1* , Niva Shrestha1, Anthony Costello2 and Naomi M. Saville3
Abstract
Background: Nutrition interventions, often delivered at the household level, could increase their efficiency bychannelling resources towards pregnant or lactating women, instead of leaving resources to be disproportionatelyallocated to traditionally favoured men. However, understanding of how to design targeted nutrition programs islimited by a lack of understanding of the factors affecting the intra-household allocation of food.
Methods: We systematically reviewed literature on the factors affecting the allocation of food to adults in SouthAsian households (in Afghanistan, Bangladesh, Bhutan, India, Islamic Republic of Iran, Maldives, Nepal, Pakistan, SriLanka) and developed a framework of food allocation determinants. Two reviewers independently searched andfiltered results from PubMed, Web of Knowledge and Scopus databases by using pre-defined search terms andhand-searching the references from selected papers. Determinants were extracted, categorised into a framework,and narratively described. We used adapted Downs and Black and Critical Appraisal Skills Programme checklists toassess the quality of evidence.
Results: Out of 6928 retrieved studies we found 60 relevant results. Recent, high quality evidence was limited andmainly from Bangladesh, India and Nepal. There were no results from Iran, Afghanistan, Maldives, or Bhutan. At theintra-household level, food allocation was determined by relative differences in household members’ income, bargainingpower, food behaviours, social status, tastes and preferences, and interpersonal relationships. Household-leveldeterminants included wealth, food security, occupation, land ownership, household size, religion / ethnicity /caste, education, and nutrition knowledge. In general, the highest inequity occurred in households experiencingsevere or unexpected food insecurity, and also in better-off, high caste households, whereas poorer, low castebut not severely food insecure households were more equitable. Food allocation also varied regionally and seasonally.
Conclusion: Program benefits may be differentially distributed within households of different socioeconomic status,and targeting of nutrition programs might be improved by influencing determinants that are amenable to change,such as food security, women’s employment, or nutrition knowledge. Longitudinal studies in different settings couldunravel causal effects. Conclusions are not generalizable to the whole South Asian region, and research is needed inmany countries.
Keywords: Food allocation, Equity, Intra-household, South Asia, Food habits, Gender, Systematic review
* Correspondence: helen.harris-fry@lshtm.ac.uk1London School of Hygiene and Tropical Medicine, Keppel Street, LondonWC1E 7HT, UKFull list of author information is available at the end of the article
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 DOI 10.1186/s12939-017-0603-1
http://crossmark.crossref.org/dialog/?doi=10.1186/s12939-017-0603-1&domain=pdfhttp://orcid.org/0000-0003-2367-908Xmailto:helen.harris-fry@lshtm.ac.ukhttp://creativecommons.org/licenses/by/4.0/http://creativecommons.org/publicdomain/zero/1.0/
BackgroundEvery day, households must make difficult decisionsabout how limited food should be shared among theirmembers. In low-income countries, around 793 millionpeople are undernourished [1] and over 3.5 millionmothers and children under five die every year becausethey are undernourished [2]. So, these food allocationdecisions have important nutritional, and sometimeslife-critical, consequences.It is often assumed that food is allocated inequitably in
households in the United Nations-defined region ofSouth Asia (in Afghanistan, Bangladesh, Bhutan, India,Islamic Republic of Iran, Maldives, Nepal, Pakistan, SriLanka) [3]. Tangential evidence of significantly higherfemale than male infant mortality rates [4] and socialand cultural gender discrimination that pervades nume-rous cultural and religious practices [5, 6] are suggestiveof a plausible pathway by which discrimination againstwomen leads to them not receiving their ‘fair share’ offood [7]. For instance, young women often stay at homeand avoid moving around or interacting with the com-munity (a practice known as purdah), women often de-scribe their husband as their God, and it is common forwomen to serve men first and themselves last [8, 9].International reviews have not found a consistent
global trend of inequitable intra-household food allo-cation, except in South Asia [10–12]. In South Asiathe scant evidence available suggests that women arediscriminated against and receive less than their ‘fairshare’, particularly in the allocation of high status,nutrient-rich foods [9, 13, 14]. This means that nutri-tion programs providing social transfers at the hou-sehold level may fail to reach the intended targetrecipients, such as the most undernourished or preg-nant women. On the other hand, if program imple-menters know which factors affect food allocation,and can identify those amenable to change, programscould be designed to target those individuals more ef-fectively. Furthermore, behaviour change interventionswithout social transfers, or other programs or policiesnot directly related to nutrition, may be able to in-crease intra-household equity and improve nutritionaloutcomes by pushing the right levers.However, intra-household food allocation has often
been described as a ‘black box’ that is poorly under-stood [15, 16]. This may be because economic con-sumption and nutrition surveys are typically collectedat the household rather than individual level, and alsobecause evidence has been segregated by academicdiscipline. Thus, this study aims to identify the deter-minants of intra-household food allocation, focusingon allocation between adults from South Asian house-holds, using a systematic and multidisciplinary lite-rature review.
Materials and methodsWe followed guidelines on systematic search protocolsfrom Reeves et al. [17] and Petticrew and Roberts [18]and reporting guidelines (Preferred Reporting Items forSystematic Reviews and Meta-Analyses (PRISMA) [19])when relevant.
Search methodIn two phases between October 2015 and January 2017,two authors independently ran and filtered the searchesin PubMed, Web of Science Core Collection, and Scopusdatabases. The former two databases were included be-cause they contain multi-disciplinary peer-reviewed lite-rature, and the latter database (Scopus) because itcontains non peer-reviewed literature.We used a combination of medical subject heading
(MeSH) and free text search terms, using asterisks to in-dicate a wildcard operator. The search syntax used inPubMed was:(((((family OR household*)) AND (food OR
energy intake OR food habits OR diet ORnutrition*)) AND (allocat* OR distribut*OR decision* OR shared OR sharing ORshare))) AND (age factors OR "age" OR sex OR"gender")South Asia was not used as a search term because da-
tabases have different systems for cataloguing studiesgeographically, and because international or theoreticalresults were also included. Instead, studies from otherregions were excluded in the filtering process. This in-creased the sensitivity of the search at the expense ofspecificity, giving more irrelevant (but also more rele-vant) results. Other search terms relating to ‘inequity’,and ‘determinants’ were also not included to ensure ad-equate sensitivity.
Inclusion and exclusion criteriaWe included quantitative, qualitative, anthropological,and theoretical studies from peer-reviewed and non-peer-reviewed sources that referred to the determinantsof inequity in intra-household food allocation amongadults. Inequity was defined as occurring when one per-son’s food needs were met more adequately than another’sneeds [12]. ‘Food’ could refer to calories, nutrients, foodquantities, food types, or dietary diversity, and ‘needs’were defined as biological requirements. We also includedany papers that described the determinants of inequitablefood allocation without explicitly measuring food intakesor needs if the paper referred to relative food allocationswithin households, rather than effects on absolute intakes.Recognising that the capability to control food distri-
bution decisions might be as important for wellbeingoutcomes as the resulting food allocation [20, 21], weconsidered including literature on the determinants of
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 2 of 21
control over food selection, preparation, and serving.However, since this review is intended to inform nutri-tion programming, and to limit the scope, we focussedon food allocation outcomes only. Therefore, this reviewexcluded the extensive and multidisciplinary body ofwork on intra-household influences on food purchasingand preparation decisions, unless they were shown toaffect the distribution of food.We included any studies explaining inequity between
different adults (different age-sex groups or differentpregnancy or lactating status). The cut-off for adults was≥15 years because studies often include ‘women of re-productive age’ by using the age range of 15 to 49 years.We excluded papers that only referred to food allocationamong children (or between adults and children), or didnot report on food allocation directly, for example by re-ferring to anthropometry or energy expenditure. We alsoexcluded any papers that did not refer to any determi-nants other than age-sex group or pregnancy status.We included all studies relating to the UN-defined re-
gion of South Asia: Afghanistan, Bangladesh, Bhutan,India, Islamic Republic of Iran, Maldives, Nepal, Pakistan,Sri Lanka. We also included theoretical and internationalreferences that were not specific to a particular researchsetting. There was no publication date cut-off, because wepredicted that empirical and high quality evidence wouldbe limited and so did not want to exclude dated but rele-vant literature.The search results were exported into EndNote refer-
ence management software for filtering. Excluded paperswere organised into different folders labelled accordingto their reason for exclusion. Disparities in inclusion /exclusion decisions were resolved by consultation with athird author. From the systematic search results, wesearched through the reference lists to find additionalrelevant results. We added additional sources by com-munication with the authors.
Study quality assessmentTo assess the quality of quantitative results we used amodified Downs and Black checklist [22], and for qua-litative results we used the Critical Appraisal SkillsProgramme checklist [23]. Given the breadth and hete-rogeneity of results expected, a meta-analysis was notpossible and so sources of bias were not factored intothe data synthesis.
Data extraction strategyTwo authors independently extracted the author name,publication year, study location, study method, and de-terminant(s) of food allocation into Microsoft Excel da-tabases. We then mapped out each determinant onpaper, by grouping them into different themes. Theidentified themes were discussed and disparities resolved
by referring back to the text. These themes were com-piled into a conceptual framework, with discussion fromall authors.
ResultsThe result filtering process is shown in Fig. 1. Fifteen re-sults were identified from the database search, 43 resultswere added by searching references, and two paperswere added from communication with the authors, gi-ving a total of 60 results.The publication date, location, study and analysis
method, sample size and characteristics, determinants,and food allocation outcome measures for the selectedstudies are summarised in Tables 1, 2 and 3. Of the 60studies reviewed, around one quarter quantitatively esti-mated the associations between at least one determinantand food allocation, another quarter were qualitativestudies, and the remaining half made theoretical, anec-dotal, or speculative references to determinants. Nearlyevery quantitative study used a different outcome meas-ure, whereas the qualitative, theoretical and anecdotalresults tended not to define the outcome or discuss dif-ferences in nutritional requirements. Over half werefrom Bangladesh (n = 14) or India (n = 18), and aroundone third did not refer to any specific country or wereinternational reviews. Publication dates ranged between1972 and 2016, and 70% were published before 2000.
Quality assessment of selected papersThere was limited empirical evidence, a diverse range ofmethods used, heterogeneity in the outcome measure, ahigh proportion of anecdotal results, and limited metho-dological detail. This meant that it was not possible toconduct a meta-analysis of any key determinants. Qua-lity assessments for quantitative and qualitative resultsare given in Tables 4 and 5 respectively.Quality assessments showed that the results were limi-
ted by the representativeness of the samples, andmethods for quantifying and addressing non-response.Quantitative results rarely gave exact p-values (in manycases there was no statistical test) and few qualitativestudies discussed the potential influence of the inter-viewer on the respondents’ answers, the rationale fortheir sampling methods, or their analysis techniques.
Framework of determinants of intra-household foodallocationEighteen determinants emerged from the thematic ana-lysis and are illustrated in a framework in Fig. 2. Theframework also gives intuitive (rather than evidence-based) hierarchy and linkages between determinants.These linkages are not given at the intra-household level,where categorisation of a complex reality into boxes
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 3 of 21
becomes particularly difficult as determinants overlap,complement or compete with one another.The findings are narratively summarised by determi-
nant, starting at the intra-household level, then thehousehold level, and finally the distal determinants.
Intra-household level determinantsRelative economic contributions or physically strenuousworkMany theoretical or anecdotal studies suggested that se-lective investment in economically productive members(typically adult men) was a rational household survivalstrategy in the South Asian context of predominantlymanual farming labour and economic returns to health[12, 24–29]. However, these gender differences werehypothesised to differ by wealth status, with poorer house-holds being more reliant on women’s economic productiv-ity (and so more equitable) than less poor households [4].Three quantitative studies corroborated this food-
health-income link, although none directly measured theeffect of relative income on food allocation. Rathnayakeand Weerahewa [30] found that mothers’ incomes werepositively associated with their own relative calorie alloca-tions, but the authors did not adjust for total householdincome. Using body size as a proxy for economic capacity,Cantor and Associates [31] found a positive associationbetween body size and relative food allocation, although itis not known whether the authors adjusted for the higherenergy requirements associated with heavier people. Pittet al. [32] found that men’s health ‘endowments’ (pre-existing health status) determined their allocations offood, whereas women’s did not; a 10% increase in health
endowment was associated with a 6.8% increase in calorieintake for men but only one tenth of that for women [32].These differences were posited to reflect gender differ-ences in economic returns to nutritional investment.The food-health-income linkage was supported by qualita-
tive studies that reported respondents’ beliefs that incomeearners deserved to be allocated more food [33, 34], but thiswas often conflated with beliefs about physiological require-ments, particularly for manual labourers ([35] and MorrisonJ, Dulal S, Harris-Fry H, Basnet M, Sharma N, Shrestha B,Manandhar DS, Costello A, Osrin D, Saville N: Formativequalitative research to develop community-based interven-tions addressing low birth weight in the plains of Nepal.Working draft, in preparation). One study also reported be-liefs that elderly people should be allocated favoured foods toacknowledge their ‘past contributions’ to the household [36],and that men required more than women because men’swork was more physically strenuous than women’s home-based work [36]. Conversely, Hartog [28] anecdotally sug-gested that the allocation of food according to economic con-tributions was no longer justified because of the increasedmechanisation of agricultural work, and Aurino [37] sup-ported this empirically. There was no association betweengender differences in workload (with 15-year old girls workingsignificantly more than 15-year old boys) or frequency of ex-ercise and the gender differences in dietary diversity.
Cultural and religious beliefs about food properties andeating behavioursSixteen studies described cultural beliefs about foodproperties and eating behaviours as a determinant of
Results from database search(n = 6928)
Results without duplicates(n = 5985)
Duplicates excluded (n = 943)
Results after screening titles(n = 567)
Titles excluded:• Not from South Asia (n = 2524)• Not about food allocation (n = 2816)• Not about adults (n = 78)
Results after screening abstracts(n = 98)
Abstracts excluded:• Not from South Asia (n = 252)• Not about food allocation (n = 167)• Not about adults (n = 50)
Results after reading full text (n = 15)
Full texts excluded:• Not from South Asia (n = 13)• Not about food allocation (n = 47)• No mention of determinants (n = 19)• No new evidence (n = 1)• Full text not available (n = 2)• Not about adults (n=1)Results from
reference lists(n = 43)
Total studies included in review(n = 60)
Results from communication with authors
(n = 2)
Fig. 1 The exclusion of database results and inclusion of results from references and communication with authors
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 4 of 21
Table
1Quantitativestud
ies:Geo
graphicaland
metho
dologicalcharacteristicsof
selected
articles(n=16)
Autho
rYear
Stud
ymetho
dSample
size
Samplecharacteristics
Analysismetho
dDeterminant
IHFA
outcom
e
Bang
lade
sh(n=5)
Abd
ullahand
Whe
eler
[65]
1985
Long
itudinal
4x3dayWFR
53HH
RuralM
uslim
househ
olds
with
atleaston
echild
unde
r5years,fro
mon
evillage
.Men
andno
n-preg
nant,n
on-lactating
wom
en(age
notspecified
)
Analysisof
variance
Season
(March
toJulyvs
Septem
berto
Decem
ber)
RDEAR=RelativeDietaryEnergy
Adequ
acy
Ratio
(individu
alcalorie
intake
asaprop
ortion
ofbo
dyweigh
t/adultm
alecalorie
intake
asaprop
ortionof
body
weigh
t)
Bouisand
Noven
ario-Reese
[69]
1997
Long
itudinal2x
1dayWFR
590HH
Hou
seho
ldsfro
m8ruraltha
nas.Men
andwom
enaged
>18
years(average
age39
and35
yearsrespectively)
Regression
(coe
fficien
tsno
trepo
rted
)•Occup
ation(farm
eror
agriculturallabou
rer)
•Age
anded
ucationof
head
ofho
useh
old
•Land
owne
rship
FS/ES=Ratio
betw
een‘food
share’(FS),
prop
ortio
nof
totalh
ouseho
ldfood
that
ape
rson
consum
ed,and
‘ene
rgyshare’(ES),
prop
ortio
nof
househ
oldcaloriesthat
anindividu
alconsum
ed.
Kumar
and
Bhattarai[61]
1993
Long
itudinal3x
1dayWFR
300HH
Hou
seho
ldsfro
m8village
sin
4districts.
Men
andwom
enaged
>18
years
Multivariate
analysis
(morede
tailno
tgiven;
results
describ
edbu
teffect
size
notrepo
rted
)
Hou
seho
ldcaloric
adeq
uacy
Calorie‘ade
quacy’(Intakes/Requ
iremen
ts)
Pitt,Rosen
zweig
andHassan[32]
1990
Long
itudinal
1x1dayWFR
in335HH;4x
1dayWFR
in50
HH
385HH
Beng
alih
ouseho
ldsfro
m15
village
s(exclude
shilltribes).
Men
andwom
enaged
≥12
years.
Line
arregression
coefficient
Health
endo
wmen
tsCalorieintake
Tetens
etal.[72]
2003
Long
itudinal
2x1dayWFR
304HH
Tworuralvillages
inlean
andpe
akseason
s.Men
andwom
enaged
18to
<30,
30to
<60,and
≥60.
Analysisof
variance
•Season
(lean
vspe
akseason
)•Village
•Socio-econ
omicstatus
Calorieintake
India(n=8)
Aurino[37]
2016
Long
itudinal
2x1day
survey
(older
coho
rton
ly)
976HH
20clusters,w
ithover-sam
plingin
disadvantage
dareas.>90%
Hindu
,and
8%femalehe
aded
househ
olds.
Older
coho
rtof
boys
andgirls
includ
esadolescentsaged
15years.
Line
arregression
coefficient
•Pu
berty(growth)
•Scho
olen
rolm
ent
•Timeuse
•Num
berof
meals
•Ph
ysicalactivity
Dietary
Diversity
Scoreby
gend
er
Babu
,Thirumaran
andMoh
anam
[62]
1993
Long
itudinal
6x3dayWFR
120HH
1ruralvillagein
peak
andlean
season
s.Sampleinclud
esno
n-agriculturalw
orkers(m
ainlysilk
weavers),agriculturallabou
rers,
andland
owning
subsistenceor
‘market-oriented
’cultivators.
Men
andwom
en(age
notspecified
)
Descriptive
comparison
s•Season
•Occup
ation(sub
sistence,
market-o
riented,non
-agricultural,andagricultural
labo
urerho
useholds)
RDEAR=RelativeDietaryEnergy
Adequ
acy
Ratio
(Individu
alcalorie
intake
asaprop
ortion
ofindividu
alrequ
irements/Adu
ltmaleintake
asaprop
ortionof
hisrequ
irements);RD
PAR=
RelativeDietaryProteinAdequ
acyRatio
(Individu
alproteinintake
asaprop
ortionof
requ
irements/Adu
ltmaleintake
asa
prop
ortionof
hisrequ
irements)
Barker
etal.[45]
2006
Cross-sectional
1x1daysurvey
101HH
1ruralvillage,mostly
cash
crop
farm
ers.
Selected
househ
olds
containing
aminim
umof:husband
andwife
(age
Principalcom
pone
ntanalysis
•Farm
work,ho
useh
old
chores
Oilintake
(g),andfre
quen
cyof
snacking
,fasting,
andmissing
meals
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 5 of 21
Table
1Quantitativestud
ies:Geo
graphicaland
metho
dologicalcharacteristicsof
selected
articles(n=16)(Con
tinued)
Basu
etal.[67]
1986
Cross-sectio
nal
1x1day24h
219HH
Hou
seho
ldsfro
mWestBeng
al,w
ithmen
andwom
enaged
>18
years.
Analysisof
variance
•Ru
ralvsurban
•Occup
ation(agriculturalist
vsplantatio
nworker)
•Religion
•Ethn
icity
•Microecon
omicsubg
roup
s
EI-ER(Energyintake
-Energyrequ
irements),
andage-sexgroups
ranked
inordero
fEI-ER
Behrman
and
Deo
lalikar
[63]
1990
Long
itudinal4x
1day24h
2roun
dsof
120HH
Threeruralvillages.
Samplingstratifiedto
includ
eland
less
agriculturallabou
rersandland
owning
cultivators.M
enandwom
en(age
grou
pno
tspecified
)
Line
arregression
coefficient
Food
priceelasticities
NAR=Nutrient
adeq
uacy
ratio
(Nutrient
intakes/Requ
iremen
ts)
Brahmam
,Sastry
andRao[66]
1988
Cross-sectio
nal
1x1day24h
1878
HH
10Indian
states,selectedho
useh
olds
with
atleaston
emem
berof
preschoo
lage.
Descriptive
comparison
for
adults
Hou
seho
ldcalorie
adequacy
(based
onintakesof
all
respon
dentswithinthe
household)
Calorieadeq
uacy
(‘ade
quate’=Calorie
intake
≥70%
Recommen
dedDaily
Intakes)
Chakrabarty
[73]
1996
Long
itudinal
2x2day24h
221HH
Threegrou
ps(highcaste,Sche
duled
Tribe,Sche
duledCaste)in
WestBeng
al.
Sampled
nuclearfamilies
with
both
parentsalive,no
n-working
wom
en(fo
rhigh
caste)
andworking
wom
en(fo
rSche
duledTribe).
t-test
Availabilityof
food
(lean
vspe
akseason
)Cerealintake–Recommen
dedcereal
intakesforabalanced
diet
Harriss-White
[27]
1991
Long
itudinal
4x1day24h
176HH
Sixvillage
sin
centraland
southe
rnIndia.
Men
andwom
en(age
notspecified
)t-test
•Season
•Region
•Land
holdingvs
land
less
RI=Relativecalorie
intakes(Individu
alintakes/Adu
ltmaleintakes)
Nep
al(n=1)
Gittelsohn
[9]
1991
Cross-sectio
nal
1x1day24h&
observation
115HH
Sixvillage
sin
Western
hills.M
enand
wom
enaged
18-24,25-49,and≥50
Correlatio
nFood
servinghabits,including
serving
orde
r,asking
forfood
,havingsecond
helpings,sub
stitu
tingfood
s,and
channe
lling
food
s.
FQS=Food
quantityscore(individu
alconsum
ptionas
aprop
ortio
nof
total
househ
oldconsum
ption/Individu
albo
dyweigh
tas
aprop
ortio
nof
total
househ
oldbo
dyweigh
t)
Pakistan
(n=1)
Governm
entof
Pakistan
[38]
1979
Cross-sectio
nal
1x24h
975HH
Malehe
adof
househ
old,
plus
wom
anof
childbe
aringage(preferablypreg
nant
orlactating)
andallchildrenaged
unde
r3years.
Line
arregression
(coe
fficien
tsno
trepo
rted
)
•Education
•Region
•Hou
seho
ldsize
•Income
Individu
alintake
/Hou
seho
ldintake
(calories,protein,
ironandvitamin
A)
SriLanka
(n=1)
Rathnayake
and
Weerahe
wa[30]
2002
Cross-sectio
nal
1x24h
60HH
Hou
seho
ldsfro
mlower
incomegrou
pin
urbanKand
y.Fathersandmothe
rs(age
notspecified
)
Line
arregression
coefficient
andt-test
•Mothe
r’sincome
•Mothe
r’sed
ucation
•Ethn
icity
•Family
size
RCA=Relativecalorie
allocatio
n(calorie
intake
asaprop
ortio
nof
recommen
ded
allowance
/Hou
seho
ldintake
asa
prop
ortio
nof
househ
oldallowance)
WFR
Weigh
edfood
records,24h24
-hou
rdietaryrecall,HHHou
seho
lds
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 6 of 21
Table
2Qualitativestud
ies:Geo
graphicaland
metho
dologicalcharacteristicsof
selected
articles(n=15)
Qualitativestud
ies(n=15)
Autho
rYear
Stud
ymetho
dSamplesize
Samplecharacteristics
Analysismetho
dDeterminant/them
eIHFA
outcom
e
Bang
lade
sh(n=4)
Abd
ullah[36]
1983
Unstructured
interviews
40HH
One
ruralM
uslim
village
incentral-w
estBang
lade
sh.
Mostly
malerespon
dents.
Particularlyin
poor
househ
olds,
wom
enalso
participated
.
Notes
recorded
onpape
r,andresults
analysed
bywealth
grou
p.
•Econ
omiccontrib
utions
•Normsrelatin
gto
receivinga‘fairshare
•Food
securityandscarcity
•Hou
seho
ldstructure
(allocatio
nto
wom
enin
parentalvs
maritalh
omes)
Food
allocatio
n
Mukherjee[75]
2002
Season
alcalend
ar1grou
pNot
repo
rted
Metho
dof
quantifying
discrim
inationno
tspecified
•Season
Discrim
ination
inconsum
ption
offood
items
andtype
s
Naved
[35]
2000
Focusgrou
pdiscussion
s,case
stud
ies,andothe
rmetho
ds.
Casestud
ies
of22
wom
en;
19men
Threevillage
sparticipating
inan
agriculturalp
rogram
.Maleandfemalebe
neficiaries
oftheprog
ram.
Triang
ulationof
multip
lequ
alitative
techniqu
es
•Ph
ysicallystrenu
ous
labo
urcontrib
utions
•Bargaining
power
•Individu
alstastes
and
preferen
ces(wom
eneatin
gmoreless-preferred
food
sthan
men
)•Food
availability(Seemingly
contradictoryanecdo
testhat
increasedfood
availabilitydid
notchange
food
allocatio
npatterns
much,bu
tfood
scarcity
ledto
men
beingless
likelyto
have
sufficien
tfood
than
wom
en.)
Allocatio
nof
specificfood
items
Rohn
erand
Chaki-Sircar
[71]
1988
Observatio
n?(limiteddetail)
1village
Not
repo
rted
Not
repo
rted
•Caste
-Highcastemen
andbo
yshadthebe
stqu
ality
food
,especially
eggs,m
ilkandfish.Impliedthat
this
islessthecase
with
lowercaste
households.
Food
quality
India(n=7)
Caldw
ell,Redd
yandCaldw
ell[51]
1983
In-dep
thqu
estio
nsandcase
stud
ies
50%
of4773
popu
latio
n(n=2387)
One
largevillage
andeigh
tsm
allervillage
sin
ruralarea
ofsouthe
rnKarnataka.
Individu
alrespon
dent
characteristicsno
trepo
rted
.
Dailyscrutin
yof
finding
sandon
-going
mod
ificatio
nof
questio
nsto
iden
tify
behaviou
ralp
atterns
•Beliefsabou
teq
uity
-Respon
dents
werereluctantto
talkabou
tfood
allocatio
n.This“dem
onstratesthe
existenceof
somebe
liefin
equitable
distrib
ution”.Ine
quity
was
“asmuch
amatterof
poor
commun
icationas
ofde
liberateintent”.
•Interpersonalrelationships
–Differen
tialfoo
dallocatio
nwas
inde
clinedu
eto
thestreng
then
ing
bond
betw
eenhu
sbandandwife.
Food
allocatio
n
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 7 of 21
Table
2Qualitativestud
ies:Geo
graphicaland
metho
dologicalcharacteristicsof
selected
articles(n=15)(Con
tinued)
Daivadanam
etal.[59]
2014
Interviewsandfocus
grou
pdiscussion
s17
individu
als;
3grou
ps
Ruralareas
(one
coastal
andon
eno
n-coastal)
Men
andwom
enaged
betw
een23
and75
years,
ofdifferent
religions
and
socio-econ
omicstatus.
Mostly
femalehe
adsof
househ
oldandothe
rsinvolved
indietaryde
cision
-making.
One
grou
pmostly
comprised
ofmen
.
Mod
ified
framew
ork
analysisusing
indu
ctiveand
dedu
ctivereason
ing–
didno
ttryto
fitthe
data
into
pre-iden
tified
them
es.
•Tastes
andpreferen
ces-wom
enprioritised
theirow
nfood
preferen
cestheleast
Allocatio
nof
preferredfood
s
Katona-Apte
[39]
1977
In-dep
thinterviews
62preg
nant
wom
enor
mothe
rs
Twodistrictsfro
mTamilNadu.
Allho
useh
olds
hadatotal
incomeof
<200Indian
Rupe
espe
rmon
th.
Allfemalerespon
dents,andmost
werepreg
nant,lactatin
g,or
amothe
rof
child
unde
rtw
oyears.
Analysismetho
dno
trepo
rted
.•Cu
lturalbeliefsabou
tfood
s–
pregnant
andlactatingwom
enavoidedcertainfood
s,andthis
caused
them
tohave
less
adequate
diets,particularlyif
therewas
lack
ofvarietyor
budg
etto
replaceavoided
food
swith
nutritiou
salternatives
Allocatio
nof
specificfood
type
sthat
have
different
prop
erties
Khan
etal.[33]
1987
In-dep
thinterviews
andparticipant
observations
20individu
als
One
stud
yvillage
from
western
Uttar
Pradesh
20mainfemaleinform
ants
(age
notrepo
rted
)fro
mdifferent
casteandclass
grou
ps,and
extensive
discussion
swith
othe
rvillage
rs,including
men
.
Analysismetho
dno
trepo
rted
.•Econ
omiccontrib
utions
-Respon
dentssaid
that
men
shou
ldeatmorebe
causethey
earn
andprovideforthefamily.
Thebe
liefthatmen
shou
ldbe
givenmorefood
was
rarer
(3/6respon
dents)whe
nwom
enearned
anincome.
Somewom
enateless
because
they
didno
thave
timeto
eat.
Wom
enhadless
appe
titedu
eto
fatig
ueaftercookingand
servinghe
rfamily
mem
bers.
•Religious
andcultu
ralb
eliefs-
Wom
en“enjoy
thisspiritof
sacrifice
forthefamily”.There
was
also
abe
liefthat
preg
nant
wom
enshou
ldeatgh
ee(clarifiedbu
tter)to
give
lubricationdu
ringbirth.The
cultu
raln
orm
ofthefemale
cook
eatin
glastmeant
that
wom
eneatless.
•Status
-Wom
enhadareligious
obligationto
fastforthefamily
andformen
tohave
supe
rior
status
andallocatio
nof
food
.•Hou
seho
ldincome-In
poor
househ
olds,the
eatin
gorde
rne
gativelyaffected
wom
en;in
wealth
yland
owning
families
itdidno
t.
Allocatio
nof
food
gene
rally,
andalso
ofspecificfood
items
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 8 of 21
Table
2Qualitativestud
ies:Geo
graphicaland
metho
dologicalcharacteristicsof
selected
articles(n=15)(Con
tinued)
Miller
[58]
1981
Review
ofethn
ograph
ies
58stud
ies
Review
ofmanystud
iesfro
macross
India.
Meta-analysis
•Interpersonalrelationships
-Servingfood
was
away
that
wom
enshow
love
andaffection
totheirmen
.Sim
ilarly,refusing
toeatfood
was
acommon
metho
dforaman
topu
nish
hiswife
ormothe
r.
Food
allocatio
n
Nicho
ls[74]
2016
Semi-stru
ctured
interviews,and
inform
alconversations
and
participant
observation
81individu
als
Four
village
sin
sub-Him
alayan
district.
Respon
dents:Governm
ent
workers,N
GOem
ployees,
village
men
,wom
en,and
coup
les.Co
nveniencesampling
toinclud
erespon
dentsfro
mdifferent
class,caste,age,gend
erandho
useholdcompo
sition.Plus,
national-levelNGOrepresentatives
from
Delhi.
Them
aticanalysis,
bycoding
them
esandintersectio
nsbe
tweenthem
es
•Labo
ur/ph
ysicallystrenu
ous
econ
omiccontrib
utions
-wom
enatetheleastdu
ringplantin
gand
harvestseason
swhe
nthey
were
working
thehardest(and
working
harder
than
men)d
ueto
alack
ofappetitefro
mtheexhaustio
nof
thelabo
ur
Food
allocatio
n
Palriwala[34]
1993
Participant
observation
1village
Sikardistrict,ruralagricultural
village
with
Hindu
(85%
)and
Muslim
(15%
)castes.
Individu
alparticipant
characteristicsno
trepo
rted
.
Not
repo
rted
•Culturalb
eliefs/eatin
gorde
r-
youn
gestdaug
hter
inlaw
usually
cooksandeatslast,leading
toless
diversediet
astheremay
beno
lentils
orvege
tables
left.
•Food
scarcity
–eatin
gorde
rparticularlyaffected
thedaug
hters
-in-law
durin
gfood
scarcity.
•Econ
omiccontrib
utions
affect
food
allocatio
n–incomeearners
aregivenpriorityof
delicacies
andnu
trient-richitemslikegh
ee,
•Interpersonalrelationships
–food
allocatio
naffected
bykinship
status,p
articularlyagnatio
n.
Allocatio
nof
specificfood
items
Nep
al(n=4)
Gittelsohn
,Thapaand
Land
man
[41]
1997
Keyinform
antinterviews,
participantob
servation,
unstructured
pilot
observations,focus
grou
pdiscussion
s,and
structured
pilesorts
105HH
Sixruralvillages,with
amixture
ofagriculturaland
non-agricultural
occupations.
Men
aged
18to
50years,and
wom
enaged
18to
50years,
includ
ingmen
struating,preg
nant,
lactating,
andpo
stpartum
wom
en.
Analysismetho
dof
qualitativeresults
notrepo
rted
•Culturalb
eliefs-Men
were
considered
theleastvulnerable
andthereforehadthefewest
dietaryrestrictio
ns,unlessthey
wereill.O
lder
peop
leconsidered
vulnerableandbe
lievedto
requ
irestreng
then
ingfood
s.Some
preg
nant
wom
enmen
tione
dpreferen
tially
eatin
ganim
alprod
uctsdu
eto
‘craving
’.Po
st-
partum
wom
enavoide
d‘cold’
food
sand‘indige
stible’foo
dslike
whe
atbread,
peanuts,soybeans
andcorn
porridg
e.They
preferentially
atecertain‘hot’foo
dslikefishand
Allocatio
nof
‘special’food
s
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 9 of 21
Table
2Qualitativestud
ies:Geo
graphicaland
metho
dologicalcharacteristicsof
selected
articles(n=15)(Con
tinued)
mille
troti.Lactatingwom
enavoided
fresh
greenleafyvegetables
that
wereperceivedas
‘cold’and
believed
tocausearthritis,sw
ellingandother
illnesses.
•Status
-Wom
en’sstatus
increased
byhaving
children.Before
childbe
aring,
youn
gmarriedwom
enhadlow
status
andweresubtlydiscou
rage
dfro
meatin
gspecialfoo
dslikeanim
alprod
uctsandcertainfried
food
s.Men
’shigh
erstatus
was
“recog
nised
inmanyways,includ
ingho
useh
old
food
behaviou
r”
Madjdian
andBras
[40]
2016
In-dep
thinterviews
30individu
als
TwoHim
alayan
commun
ities
from
Hum
ladistrict.
Femalerespon
dents(15Budd
hist;
15Hindu
Dalitor
Chhetri)o
freprod
uctiveage(aged15
to49
years).Selectedrespon
dentswho
werepregnant
orhadbeen
pregnant
atleasto
ncebefore.
Indu
ctivecoding
based
onaconcep
tual
framew
ork,using
bottom
-upandtop-
downcoding
toallow
newthem
esto
emerge.
•Beliefsabou
t‘fairshare’/Religion-
Budd
histho
useh
olds
allocatedfood
accordingto
appe
tite;thiswas
not
repo
rted
inHindu
househ
olds.
•Culturalb
eliefsandfood
habits-
Certain
food
sbe
lievedto
cause
skin
allergies.Eatin
gorde
rwas
associated
with
eatin
gless.
•Food
security-Food
insecure
househ
olds
didno
tadhe
reto
food
proscriptio
nsdu
eto
alack
offood
Food
allocatio
n
Morrison
,J.etal.
Form
ativeresearch
toinform
the
developm
ent
ofinterven
tions
totacklelow
birth
weigh
tin
therural
plansof
Nep
al.
Inprep
aration.
Unp
ublishe
dob
servations
Interviewsandfocus
grou
pdiscussion
s25
wom
en,
2grou
ps.
One
districtin
Terai.
25youn
gdaug
hters-in-law
from
marginalised
grou
pslivingin
extend
edfamilies,one
focus
grou
pdiscussion
with
men
,and
onewith
FemaleCom
mun
ityHealth
Volunteerswho
were
mothe
rs-in
-law.M
ost(90%
)respon
dentswereHindu
.Respon
dent
ageno
trepo
rted
Descriptiveconten
tanalysis.D
atawere
copied
from
transcrip
tsinto
columns
of15
descrip
tiveem
erge
ntcatego
ries.
•Status
-Respon
dentsrepo
rted
that
men
atemorebe
causethey
hadhigh
erstatus
andso
deserved
to.
•Interpersonalrelationships
-Husband
smay
hide
food
for
theirpreg
nant
wives,
disrespe
ctingthemothe
r-in-law.
•Hou
seho
ldstructure-Married
wom
enwho
visitedor
lived
attheirmaternalh
omes
hadfewer
food
restrictio
ns.
•Econ
omiccontrib
utions
–Manual
labo
urerswerepe
rceivedto
deservemore
•Culturalfoo
dbe
liefs–preg
nant
wom
enateless
(fear
offull
stom
achharm
ingthebaby)
•Hou
seho
ldincome–no
effect
offood
beingbo
ught
vsgrow
non
food
decision
s.
Allocatio
nof
food
gene
rally,
andalso
of‘sp
ecial’food
s
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 10 of 21
Table
2Qualitativestud
ies:Geo
graphicaland
metho
dologicalcharacteristicsof
selected
articles(n=15)(Con
tinued)
Panter-Brickand
Egge
rman
[64]
1997
Semi-structured
interviews
120he
adsof
househ
old
Popu
latio
nof
high
andlow
casteIndo
-Nep
aleseand
Tibe
to-Burmeseethn
icgrou
psfro
mfour
Pancha
yatsin
two
districts.
Sampled
househ
olds
toen
sure
prop
ortio
nalrep
resentationof
largeandsm
allland-ho
lding
farm
ers
Age
ofrespon
dentsno
trepo
rted
.
Analysismetho
dof
qualitativeresults
notrepo
rted
.
•Food
shortage
s/Ethn
icity
-Indo
-Nep
aleseho
useh
oldused
discrim
inationagainstwom
enas
acoping
mechanism
durin
gfood
shortage
swhe
reas
Tibe
to-
Burm
eseho
useh
olds
didno
t.
Food
allocatio
n
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 11 of 21
Table 3 Studies with theoretical, hypothetical or general mention of determinants – author, year, determinant and food allocationoutcome (n=29)
Author Year Determinant IHFA outcome
Bangladesh (n = 5)
Chaudry [70] 1983 Household size Relative calorie adequacy
Chen, Huq and d’Souza [4] 1981 Relative economic contributions Allocation of food quantity andquality
Kabeer [44] 1988 Cultural beliefs / serving order Food allocation
Rizvi [50] 1981 Relative social status Food allocation
Rizvi [68] 1983 Household wealth (poverty)Household size
Food allocation
India (n = 3)
Cantor and Associates [31] 1973 Relative economic contributions (proxied by body size)Household wealth
Food allocation
Coffey, Khera andSpears [47]
2015 Relative social status Food allocation
Das Gupta [49] 1996 Relative social status (having sons)Nutrition knowledge
Relative calorie adequacy
Nepal (n = 1)
Gittelsohn, Mookherji andPelto [42]
1998 Cultural food beliefsHousehold food insecurity
Food allocation
South Asia (n = 2)
Agarwal [56] 1997 Bargaining power Food allocation
Appadurai [8] 1981 Relative cultural status / life cycle in the householdBargaining powerInterpersonal relationships
Food allocation
International (n = 16)
DeRose, Das andMillman [14]
2000 Relative social statusDecision-makingNutrition knowledge
Calorie and food allocation
Haddad and Kanbur [25] 1990 Control over incomeFood insecurity
Calorie and food allocation
Haddad et al. [11] 1996 Decision-making (identify of decision-maker)Food insecurity
Food allocation
Haddad [54] 1999 Control over incomeFood insecurity
Food allocation
Kumar [26] 1983 Decision-making Food allocation
Messer [48] 1997 Relative social status (the traditional role and perceptions of women)Beliefs about fairness
Food allocation
Pinstrup-Andersen [15] 1983 Nutritional needPreferencesDecision-makingHousehold income
Food allocation
Wheeler [12] 1991 Relative economic contributionsBeliefs about fairnessBargaining power
Allocation of nutrient-rich foods
Carloni [53] 1981 Decision-makingSocial mobility / participation in shopping
Food allocation
Hartog [28] 1972 Economic contributionsCultural beliefsSocial statusInterpersonal relationships
Food allocation
De Schutter [52] 2013 Beliefs about fairnessControl over food production or purchasingFood insecurity
Food allocation
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 12 of 21
food allocation, typically showing how these beliefscaused women to receive comparatively less.Some foods were believed to be ‘unhealthy’, ‘streng-
thening’ or ‘digestible’, and therefore deliberately avoidedor selected by vulnerable people, such as elderly people, orpregnant or lactating women [28]. Short-term ‘transitory’states, such as menstruation and illness that are perceivedto make a person ritually unclean (‘jutho’ (Nepalese)),were reported to cause inequity in food allocation [9, 38].Certain fruits and vegetables were considered unhealthy
for post-partum or lactating women [9], and other foodswere believed to cause illness [39] or skin allergies [40] inthe breastfeeding child. Elderly people were believed to re-quire soft ‘digestible’ [9] and ‘strengthening’ foods [39, 41].Pregnant women were reported to require ghee (clarifiedbutter), to facilitate an easier, lubricated birth [33], butavoid certain foods that cause illness, indigestion, fits, de-lirium, or large babies (and therefore difficult labour) [39]and eat less because of a belief that a full stomach wouldharm the baby (Morrison J, Dulal S, Harris-Fry H, Basnet
Table 3 Studies with theoretical, hypothetical or general mention of determinants – author, year, determinant and food allocationoutcome (n=29) (Continued)
Den Hartog [43] 2006 Religious beliefsBeliefs about fairness
Food allocation
Gunewardena [60] 2014 Food insecurity Food allocation
Pelto [46] 1984 Social status (in relation to modernisation and urbanisation) Food allocation
Ramachandran [55] 2007 Decision-making / control over incomeBargaining powerFood insecurityHousehold composition (nuclear vs joint households)
Food allocation
Van Esterik [29] 1985 Economic contributionsOverlap between cultural beliefs during pregnancy, social status, andpovertySocial mobilityInterpersonal relationshipsHousehold size (number of senior women)ReligionFood availability
Food allocation
No countries mentioned (n = 2)
Doss [24] 1996 Relative economic contributionsBargaining power
Food allocation
Hamburg et al. [57] 2014 Interpersonal relationships Food sharing
Table 4 Quality assessment of quantitative results (n = 16) using an adapted Downs and Black checklist
Study quality No Unable to determine Yes
n n n (%)
Is the hypothesis or aim of the study clearly described? 0 NA 16 (100)
Are the outcomes described in the Introduction or Methods? 1 NA 15 (94)
Are the characteristics of the respondents described? 6 NA 10 (63)
Are the determinants of interest described? 2 NA 14 (88)
Are the distributions of principal confounders described? 6 NA 10 (63)
Are the main findings of the study clearly described? 1 NA 15 (94)
Does the study provide estimates of random variability? 9 NA 7 (44)
Have probability values (not cutoffs) been reported? 14 NA 2 (13)
Validity, bias and confounding
Was the sample representative of the population? 1 7 8 (50)
Were the respondents representative of the population? 0 14 2 (13)
Were the statistical tests appropriate? 4 0 12 (75)
Were the main outcome measures used valid and reliable? 0 3 13 (81)
Was there adequate adjustment for confounding? 8 2 6 (38)
Were losses of respondents taken into account? 3 11 2 (13)
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 13 of 21
M, Sharma N, Shrestha B, Manandhar DS, Costello A,Osrin D, Saville N: Formative qualitative research to de-velop community-based interventions addressing lowbirth weight in the plains of Nepal. Working draft, inpreparation). Another phase condition is puberty; accord-ing to the ‘pubertal hypothesis’ households might changeintra-household food allocation in favour of adolescentsgoing through pubertal growth spurts, however this wasnot supported empirically [37].The categorisation of foods as heating, neutral or coo-
ling was also linked to differential food allocation [9].Examples of ‘heating’ foods included wheat, fish, millet,and milk, and ‘cooling’ foods included yogurt, fruits,green beans, gourds, and rice [41]. Again, these catego-ries were related to the condition of the individual; forinstance, lactating women avoided cooling foods [41].Some authors noted that these restrictions caused anoverall reduction in women’s dietary intake [41] perhapsdue to insufficient household budget for nutritious alter-natives [39]. The overlap of women being in the ‘phasecondition’ of being pregnant, having low social status,and extreme poverty, was hypothesised as a cause of ine-quitable allocation of food towards pregnant women[29], whereas another study found that food insecurehouseholds did not adhere to food proscriptions due toa lack of food [40].
Cultural beliefs about the status or value of foods werealso linked to food allocation [42]. Some people were al-located or ‘channelled’ specific foods, and these foodswere often high status and expensive. For example, ghee(a high status food) was often only given to men [9](perhaps with the exception of pregnant women, as docu-mented earlier). Food ‘channelling’ was associated withsignificantly higher food quantities, so those who weregiven special foods were also given more food in total [9].Religious beliefs may affect the distribution of food.
For example, beliefs about the meaning of food, such asthe act of eating being considered a form of worship inIslam, was suggested as a determinant of food allocation,although the direction of effects was not specified [43].Fasting caused women to be allocated comparatively lessfood, because women fasted more frequently and strictlythan men [27, 33].Eating order was another key factor in food eating and
allocation behaviours. Daughters-in-law, often young,newly married women, tended to serve themselves last(to show deference and ensure the wellbeing of malemembers [44]) and this was associated with eating less[9, 33, 40] and also lower quality diets than others [34, 45].Gittelsohn [9] found that late serving order at meal times,being the food server, and not having a second helpingwere all significantly negatively associated with the quantityof food consumed. Two authors suggested that the nega-tive effects of serving order were caused by limited foodavailability, because women tended to ensure everyoneelse had enough before serving themselves the re-mainder [9, 44].
Relative social status within the household, according tocultural normsHigh social status and perceived deservedness of house-hold members was reported by 14 studies as a determi-nant of intra-household food allocation, although nostudies quantified the effect empirically. Some men-tioned status in general terms, without ascribing high orlow status to particular household members but sugges-ting that people with lower status would receive lessfood and less preferred foods [12, 14, 28, 29, 46],whereas others reported that men had higher status andtherefore were allocated more food [9, 41, 47]. Pelto [46]noted that modernisation and increasing urbanisationwas reducing the effect of status on food allocation.Women’s identity as being lower status, frugal, modest,
and subservient was described as a determinant of foodallocation that favoured men but was often internalisedby women [33, 48]. This identity was also interlinkedwith perceptions of body image, that also led women toeat comparatively less [48].In addition to differences between genders, food allo-
cation was also determined by differences in social
Table 5 Quality assessment of qualitative results (n = 15) usingCritical Appraisal Skills Programme (CASP) checklist
Critical Appraisal Skills Programme(CASP) quality indicator
No Unable todetermine
Yes
n n n (%)
Was there a clear statement of the aimsof the research (the goal, importance,and aims)?
0 0 15 (100)
Is a qualitative methodology appropriate? 0 0 15 (100)
Was the research design justified asappropriate to address the aims of theresearch?
0 7 8 (53)
Was the recruitment strategy justified asbeing appropriate to the aims of theresearch (how and why respondentswere sampled, or discussions ofnon-response)?
0 7 8 (53)
Were the data collected in a way thataddressed the research issue (detail andjustification of methods, issues of datasaturation)?
0 9 6 (40)
Has the relationship between researcherand participants been adequatelyconsidered?
0 9 6 (40)
Have ethical issues been considered(informed consent and ethical approval)?
0 13 2 (13)
Was the data analysis sufficiently rigorous? 0 10 5 (33)
Is there a clear statement of findings? 1 0 14 (93)
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 14 of 21
rankings among women within households [47], hier-archy within the patriline [34], and variations withinindividuals over time. Women received more food astheir status increased with age [9], by having children[41], and, particularly, by having sons [49]. In contrast,Rizvi [50] anecdotally suggested that increases in femalestatus over time did not affect food allocation.
Belief that everyone should be given their ‘fair share’Overlapping these beliefs about food properties and sta-tus were locally-held beliefs about fairness, and differentdefinitions of what ‘fair’ means [48]. An Indian studyconcluded that beliefs about fairness affected food allo-cation because the respondents were reluctant to discussdisparities in food allocation [51]. Abdullah [36] also in-dicated this, by reporting a respondent’s description of amother-in-law as a “bad woman” (p112) if she did notgive her daughter-in-law special foods that others re-ceived, and Madjdian and Bras reported that Buddhisthouseholds allocated food according to appetite [40]. Pittet al. reported that the tendency for households toequalise food allocation, rather than allocate food in anincome-maximising way, was a reflection of households’aversions to inequity [32]. Pinstrup-Andersen also hypo-thesised that perceived nutritional need determined foodallocation (arguably a definition of fairness) [15].
Conversely, three others reported an ideological belief thatmen deserved to be given more than women [33, 43, 52].
Decision-making, social mobility, and control over resourcesThe ten studies on decision-making, control over resources,social mobility, or identity of the cook were all anecdotal ortheoretical. Few specified who the decision-maker was or whowould benefit from the decisions [15, 26, 53]. Haddad et al.noted that female decision-makers may be more likely thanmale decision-makers to allocate food in a way that maxi-mised the household’s nutritional outcomes [11], and otherwork suggested that women’s control over income [25, 54, 55](or production or purchasing [52]) would affect food alloca-tion in favour of women. However, decision-making in nuclearhouseholds was also a risk factor for inequitable allocation to-wards women, as female decision-makers were responsible forensuring that everyone else was sufficiently fed [55].Related to this was different household members’ social
mobility, and freedom to go food shopping or access foodoutside of the home [29, 53]. Women were described ashaving less social mobility and this was linked to lowerfood allocations [53]. However, it was also suggested thatthere may be less inequity against women than expectedbecause food allocation occurs within the household,which, compared with the public sphere, is a domain inwhich women traditionally have more control [14].
Household food security and scarcity
INTRA-HOUSEHOLD FOOD ALLOCATION
Village and regionSeasonality
Religion, ethnicity and caste
Educational levels
Land ownership
Household size / structure
HOUSEHOLD-LEVEL DETERMINANTS
Cultural and religious beliefs about food properties and eating
behaviours
Belief that everyone should be given their ‘fair share’
Relative status within the household, defined by cultural norms
Decision-making, social mobility and control over
resources
Bargaining power
Food as a means to establish and reinforce interpersonal
relationships
Relative economic contributions or
physically strenuous work
Individual tastes and
preferences
INTRA-HOUSEHOLD DETERMINANTS
DISTAL DETERMINANTS
Nutrition knowledge
Economic factors Cultural factors Social processes
Household wealth and income
Household occupation
Fig. 2 Determinants of intra-household food allocation and hypothesised hierarchical structure
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 15 of 21
Bargaining powerRelated to this, many studies suggested that ‘bargainingpower’ was an important determinant of food allocation.Some described a general linkage [55], while others de-scribed various microeconomic theories whereby indivi-duals rationally aim to maximise their utility bybargaining over household resources such as food [12,24, 56]. The strength of a person’s bargaining power isdetermined by their ‘fallback position’ (the utility thatthe individual would achieve if cooperation with otherhousehold members fails). Bargaining power may be pre-dicted by the size of dowry for women [55], social normsthat determine how goods such as food can be ex-changed or used, and /or perceived requirements [56].Naved also indicated that bargaining power might affectfood allocation, with specific reference to micronutrients[35]. Appadurai described the process of bargaining withfood, particularly in reference to daughters-in-law feelingresentment about their subservient role [8]. Daughters-in-law were reported to communicate resentment by be-ing reluctant to cook, destroying food, or making smallsigns of discontent when serving food [8].
Food as a means to establish and reinforce interpersonalrelationshipsA study on the psychology of feeding found that foodserving may be determined by the emotional state of theserver and recipient, and that feeding was done to buildrelationships [57]. Serving and eating food was reportedas a key determinant of intra-household food allocation[34] and a means to indicate rank [8], show disrespect ordispleasure ([58] and Morrison J, Dulal S, Harris-Fry H,Basnet M, Sharma N, Shrestha B, Manandhar DS, CostelloA, Osrin D, Saville N: Formative qualitative research todevelop community-based interventions addressing lowbirth weight in the plains of Nepal. Working draft, inpreparation), punish or reward people [29], or express loveand strengthen relationships [28]. Two studies suggestedthat a strong bond between husband and wife might re-duce the inequity against the wife that would otherwise beimposed by the mother-in-law ([51] and Morrison J, DulalS, Harris-Fry H, Basnet M, Sharma N, Shrestha B, Mana-ndhar DS, Costello A, Osrin D, Saville N: Formative quali-tative research to develop community-based interventionsaddressing low birth weight in the plains of Nepal. Work-ing draft, in preparation).
Individual tastes and preferencesOne study [15] reported a general hypothesis that foodpreferences determine allocations. One study found thatwomen were allocated the less-preferred foods [35], andanother found that women prioritised their own foodpreferences the least [59]. A new vegetable productionprogram, which had increased the availability of less
popular vegetables, had caused an increase in women’sconsumption of those vegetables [35].
Household-level determinantsFood insecurity and scarcityEvidence on the effect of household food security wasmixed. Seven studies provided a hypothetical link be-tween food insecurity and food allocation [11, 25, 42,52–54, 60].Most studies suggested that women were more sensi-
tive to changes in food availability, and acted as a bufferfor the household in food insecure conditions [40, 52, 55],particularly the youngest daughter-in-law [34]. Householdcalorie adequacy was a better predictor of calorie ade-quacy for women than for men [61], years of higher riceyields were linked to higher equity in calorie allocation[62], and women were more sensitive to food pricechanges (had higher food price elasticity) than men [63].Although one study found no overall effect of food inse-curity, there were differential effects between villages, andin food-scarce months landowning households favouredmen at the expense of women while labourer householdsfavoured adults at the expense of children [27]. Anotherstudy found some ethnic variability in these effects, withIndo-Nepalese household using discrimination againstwomen as a coping mechanism during food shortages andTibeto-Burmese households not changing their food dis-tribution patterns in this way [64].In contrast, some studies found opposing or no effects.
Naved found that increased food availability did notaffect food allocation patterns, but food scarcity led tomen being less likely to have sufficient food than women[35]. Abdullah also found women’s proportion of men’sintakes increased from 81 to 90% between food secureand food short seasons, in poor households [36]. A lateranalysis by the same authors found no significant effectafter adjusting for differential requirements [65], andanother also found no effect of scarcity on calorieallocation [66].
Household wealth and incomeEvidence was mixed on the effect of household wealth andincome. Three quantitative studies reported no effect [31,38, 67], while Pinstrup-Andersen’s [15] theoretical studyhypothesised that there was an effect without specifyingthe direction. Two qualitative studies reported higher in-equity in poorer households [33, 68], but a qualitative studyreported that the source of food, whether bought or grown,had little effect on food decisions (Morrison J, Dulal S,Harris-Fry H, Basnet M, Sharma N, Shrestha B, Mana-ndhar DS, Costello A, Osrin D, Saville N: Formative quali-tative research to develop community-based interventionsaddressing low birth weight in the plains of Nepal. Work-ing draft, in preparation).
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 16 of 21
Educational levelsHouseholds with more educated household heads allo-cated less meat to women than men [69], whereas an-other two studies on female education showed noassociation with food allocation [30, 37].
Nutrition knowledgeAlthough studies linked food beliefs and food allo-cation, no studies measured the association betweendietary knowledge and food allocation. Das Guptasuggested that lack of knowledge did not explain thecomparatively inadequate intakes of pregnant andbreastfeeding women in India [49], whereas others sug-gested the opposite [14, 38].
Household occupationOne study found that the greatest gender inequity oc-curred in agricultural labourer households, compared withmarket-based, subsistence farmers or non-agriculturalhouseholds [62]. Another found that farm work andhousehold chores were associated with consuming fewersnacks, and that all three (farm work, household chores,and snacks in a combined score) were significantly higherfor women than men [45]. In contrast, two studies foundno effect of being an agricultural labourer household [69]or household occupation [67] on food allocation. How-ever, if the household head was a farmer, then milk wasallocated more equitably and meat less so [69].
Land ownershipOnly Bouis and Novenario-Reese mentioned the effectof land ownership, finding that landowning householdswere less equitable with their allocation of eggs but moreequitable with fish [69].
Household size and structureThe effect of household size was also mixed. Van Esteriksuggested that the number of senior women in a house-hold would affect food allocation patterns [29], whileRizvi suggested that household size affected intakes ofindividuals, but not allocation patterns [68]. Rathnayakeand Weerahewa reported a significant positive effect ofhousehold size on mothers’ relative calorie allocations inSri Lanka [30], and a study from Pakistan also foundhigher intakes for the household head, pregnant women,and children under five years, relative to average house-hold intakes, in larger households [38]. Conversely, an-other study showed that large household size (more thaneight children) was associated with significantly highermale than female calorie adequacy [70].(Morrison J, Dulal S, Harris-Fry H, Basnet M, Sharma
N, Shrestha B, Manandhar DS, Costello A, Osrin D,Saville N: Formative qualitative research to developcommunity-based interventions addressing low birth
weight in the plains of Nepal. Working draft, in prepar-ation) reported a trend for food-related rules to be re-laxed when women returned to their parental homes,where they had fewer food restrictions. This was sup-ported by Abdullah [36], who found that women weregiven less or fewer special foods in the marital home,and so women (or adolescent, unmarried girls) receivedspecial treatment when at their parental homes. Com-paring joint and nuclear households, Ramachandran re-ported that women were allocated comparatively less innuclear households, where the women had the responsi-bility of feeding everyone; whereas, in joint householdswomen were given more because their mother-in-lawadopted the role of food planning and distribution [55].
Religion, ethnicity and casteEvidence on the effects of religion, ethnicity and castewas limited and mixed. Van Esterik [29] hypothesisedthat religion was a determinant of food allocation, par-ticularly through religious influence on food classifica-tions systems. There were no differences betweenBuddhists and Christians in food allocation [67], butBuddhist families were more equitable than Hinduhouseholds [40]. Rathnayake and Weerahewa found noeffect of ethnicity on relative calorie allocations [30],whereas Panter-Brick and Eggerman found that foodshortages caused discrimination against women duringfood shortages among Indo-Nepalese households butnot among Tibeto-Burmese households [64]. No studiesexplicitly described the effect of caste, but two studiesimplied that inequity would be greater in high castegroups [63, 71].
Distal determinantsSeasonalityTwo quantitative studies relating to seasonality foundcontradictory, non-significant and unexplained effectson intra-household food allocation. In Bangladesh,Tetens et al. found more equity in food allocation duringpeak agricultural production seasons [72], whereas inIndia Chakrabarty found less equity in cereal consump-tion in the peak season [73]. A qualitative study fromIndia found that women ate the least during plantingand harvest seasons when they were working the hardest(and working harder than men) due to a lack of appetitefrom the exhaustion [74], and a seasonal calendar fromBangladesh showed that women were allocated fewereggs and fish than men, particularly during food inse-cure months [75]. An anecdotal study suggested thatpredictable seasonal variation was unlikely to affect foodallocation patterns because households would have cop-ing strategies for predictable variations in food availabil-ity [29]. However, if this shortage overlapped withperiods of more pregnancies or more female agricultural
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 17 of 21
labour inputs, then women would have less adequate di-ets than other household members [29].
Village and regionRegional differences in intra-household food allocationwere varied. Basu et al. found no differences betweenurban and rural areas among the Lechpas ethnic groupin India [67], whereas a study from Pakistan found smallbut significant regional variation in micronutrient allo-cation [38], with higher dietary adequacy for men in cer-tain regions and lower adequacy for pregnant orlactating women in others. The authors suggested thatthis may have been a result of regional differences incultural attitudes and beliefs regarding pregnancy, butdid not describe these attitudes in detail. Harriss-Whitealso found differences in inequity between four differentvillages in India, which she hypothesised may have beendue to differences in land ownership and methods ofagricultural production [27].
DiscussionAlthough we searched for literature across South Asia,studies were mostly from Bangladesh, India, and Nepal,with few from Islamic Republic of Iran, Pakistan, and SriLanka, and none from Afghanistan, Maldives or Bhutan.We identified eighteen determinants of intra-householdfood allocation, many of which were specific to theSouth Asian context and centred on interlinked aspectsof poverty, cultural beliefs, and power. The review sug-gests that the most equitable households are low castehouseholds [63] in which women earn an income [30]and there is low educational status of the householdhead [69] (indicative of lower socioeconomic status) butthat equitable households will also be food secure [61](indicative of higher socioeconomic status).We hypothesise that these seemingly conflicting effects
could be reconciled in a non-linear relationship betweensocioeconomic status and inequity. During acute or unex-pected food shortages, households selectively invest limitedcalories in men because they have more labour opportu-nities [62, 63]. For example, recent unexpected shocks mayinclude the 2015 Nepalese earthquakes [76]. As food secur-ity increases and food shortages become more predictable,households become more equitable [65, 72, 73]. Low castehouseholds are often described as more egalitarian [71],partly because low caste men do not inherit land, [5] butalso because women have comparatively higher economiccontributions [73]. The positive effect of individuals’ rela-tive economic contributions on their shares of householdfood is also supported by evidence from the Philippines[77] and China [78]. At higher levels of socioeconomic sta-tus, inequity may increase again [27]. This is supported bythe finding that women from landowning Indian house-holds are thinner than in landless households [45].
Inequity amongst higher socioeconomic groups may becharacterised by preferential “channelling” of high sta-tus, often micronutrient-rich luxury foods such as meator dairy [9], rather than by unequal allocation of staplesthat may occur in the poorest groups [62]. If true, thischannelling could result in social rather than nutri-tional inequity, or perhaps inequity of micronutrientsrather than calorie allocation. This may explain the sur-prising negative gradient in the prevalence of anaemiain women with increasing wealth in Nepal (anaemia:32% in the lowest wealth quintile, and 49% in the mid-dle quintile) [79], despite the relatively high cost ofmicronutrients required to reduce anaemia. This in-equity may decline in households where women arehighly educated and ‘modern’, as they may have themost knowledge of dietary requirements (to counteractdiscriminatory cultural food practices) and these house-holds may be less socially conservative and restrictivetowards women. Although no studies have providedevidence to support this directly, the prevalence of an-aemia in Nepal does fall to 36% in the highest wealthquintile [79].Any socioeconomic level, dynamic intra-household
factors (such as bargaining, preferences and interper-sonal relationships) will introduce variance within thesetrends, while cultural norms and food practices (thathave wide local variation but are slow to change [40])may introduce variance and also determine the strengthof association between socioeconomic status on foodallocation.
Limitations of the study, and future workThis study benefitted from a systematic literature searchbut the results and conclusions are limited by the lack ofrecent evidence. Given the rapid changes in labour mi-gration, engagement in non-farm work, and mechanisa-tion of farm work [80], the full framework should betested with recent data. Another limitation is that moststudies came from Bangladesh, India and Nepal, so theresults may be less or not valid for other South Asiancountries that did not have any or many studies (particu-larly Afghanistan, Bhutan, Maldives, Iran, and Pakistan).Even within the results from India, Bangladesh andNepal, there is wide heterogeneity in culture, beliefs, andinstitutions, making generalisations difficult. Many re-sults arose from snowballing, indicating poor indexingof multi-disciplinary evidence, so the included studiesmay be subject to citation bias.Most papers considered only a few determinants in
their analyses and did not control for possible con-founders, so the pathways in the framework cannot bedisentangled. For example, we cannot tell whether pa-pers that tested the effects of seasonality or wealth onfood allocation would have found the same results after
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 18 of 21
controlling for household food security. Similarly, theoverlap of cultural norms, social status, income earning,bargaining power, and participation in decision-making,means that the effects of these intra-household determi-nants cannot be distinguished. We expect that theseintra-household determinants co-exist, sometimes re-inforcing and other times opposing each other.To truly understand these intra-household processes,
longitudinal mixed methods are required to examineeach possible determinant at different stages in thehousehold life cycle (adolescents, newlyweds, seniormembers, elderly), with different phase conditions (du-ring puberty, illness, menstruation, pregnancy, post-partum, and breastfeeding), at different points in thefood acquisition–preparation–distribution pathway, indifferent socioeconomic groups, in different regions andin different seasons. Given the breadth of this evidencegap, research to inform nutrition program design couldparticularly focus on the determinants of food allocationthat may be amenable to change, such as food security,women’s employment, or nutrition knowledge. Moststudies focused on energy or food quantity, so more re-search is needed on the allocation of micronutrient-richfoods [27]. Other next steps could link the evidence onfactors affecting the determinants in this framework,such as predictors of food choice, bargaining power, andfood security.
Implications of the findings: what does this mean fornutrition interventions?The findings indicate that women are disadvantaged inthe allocation of food, and that women’s nutrition out-comes could be improved through changes in intra-household food allocation patterns. In particular, pregnantwomen tend to receive lower relative allocations, and thishas important nutritional implications because nutritionduring pregnancy is associated with maternal health out-comes, intra-uterine growth retardation, child health out-comes, and is a key point in the intergenerational cycle ofundernutrition [81]. As such, interventions may need toparticularly prioritise pregnant women.By predicting the distribution of transfers at all levels
of socioeconomic status, social transfer programs couldbe designed to ensure that the intended recipients re-ceive the transfers. For example, programs deliveringtransfers in emergency contexts may need to ensure thattransfers are large enough [82], so that transfers willreach less economically productive household members.Poorer households might require additional resourcessuch as food or cash transfers to improve the nutritionalstatus of women. Low caste households may be morelikely to share these transfers equitably, and so addi-tional intervention components (such as behaviour
change communication) to ensure the transfers reachtarget recipients may only be required if the program-mer intends to disproportionately target women (orpregnant women).In contrast, transfers to high caste or better-off groups
may need to include a behaviour change component toensure that transfers reach women. Alternatively, pro-grams delivering transfers to high caste groups couldprovide lower status, less desirable goods, such as flourrather than rice [83], to ensure that transfers will bepreferentially distributed to lower status householdmembers. This leads to a wider discussion on prog-rammatic objectives, and whether programs intend tochallenge patriarchal norms to empower women and im-prove nutrition, or to work within patriarchy to achieveoptimal nutritional outcomes. Furthermore, high castebetter-off households that can already afford moremicronutrient-rich foods may not need more resourcesto improve maternal nutrition – behaviour change inter-ventions may be sufficient.Alternative, or complementary, approaches to improve
intra-household equity in food allocation may include:interventions that provide income-generating activitiesfor women, women’s groups to increase social mobilityand bargaining power or otherwise empower women, oragricultural programs to improve food security.
ConclusionsThere are many possible household-level and intra-household determinants of intra-household food alloca-tion, but evidence is out-dated and not comprehensive.Local context and variation in social hierarchies makesgeneralised conclusions difficult. Programs deliveringsocial transfers may find differential intra-household dis-tribution of transfers in different socioeconomic groups.Programs affecting determinants that are amenable tochange, such as household food security, bargainingpower, and gender-specific labour opportunities, maycause changes in intra-household food allocation patterns.
AcknowledgementsNot applicable.
FundingThis research was funded by Child Health Research Appeal Trust(CHRAT) and UKaid from Department for International Development SouthAsia Research Hub (PO 5675). The donors had no role in the design, analysis,interpretation, or preparation of the manuscript.
Availability of data and materialsNot applicable.
Authors’ contributionsHHF designed the study protocol, with inputs from NMS and AC. HHF andNS conducted the literature search, and NMS resolved any conflicts instudy selection. HHF drafted the manuscript. All authors read andapproved the final manuscript.
Harris-Fry et al. International Journal for Equity in Health (2017) 16:107 Page 19 of 21
Competing interestsThe authors declare that they have no competing interests.
Consent for publicationNot applicable.
Ethics approval and consent to participateNot applicable.
Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims in publishedmaps and institutional affiliations.
Author details1London School of Hygiene and Tropical Medicine, Keppel Street, LondonWC1E 7HT, UK. 2Maternal Child and Adolescent Health, World HealthOrganization, Geneva, Switzerland. 3Institute for Global Health, UniversityCollege London, 30 Guilford Street, London WC1N 1EH, UK.
Received: 13 March 2017 Accepted: 12 June 2017
References1. FAO/ IFAD/ WFP. The State of Food Insecurity in the World 2015. Meeting
the 2015 international hunger targets: taking stock of uneven progress.Rome: FAO; 2015.
2. Black RE, Allen LH, Bhutta ZA, Caulfield LE, De Onis M, Ezzati M, Mathers C,Rivera J. Maternal and child undernutrition: global and regional exposuresand health consequences. Lancet. 2008;371:243–60.
3. Sen AK. Family and food: sex bias in poverty. In: Bardhan P, Srinivasan T, editors.Rural poverty in South Asia. New York: Columbia University Press; 1988. p. 453–72.
4. Chen LC, Huq E, d’Souza S. Sex bias in the family allocation of food andhealth care in rural Bangladesh. Popul Dev Rev. 1981;7:55–70.
5. Cameron MM. On the edge of the auspicious: Gender and caste in Nepal.Urbana and Chicago: University of Illinois Press; 1998.
6. Kondos V. On the Ethos of Hindu Women: Issues, Taboos, and Forms ofExpression. Kathmandu: Mandala Publications; 2004.
7. Sen A, Sengupta S. Malnutrition of rural children and the sex bias. EconPolit Wkly. 1983;18:855–64.
8. Appadurai A. Gastropolitics in Hindu South Asia. Am Ethnol. 1981;8:494–511.9. Gittelsohn J. Opening the box: intrahousehold food allocation in rural
Nepal. Soc Sci Med. 1991;33:1141–54.10. Berti PR. Intrahousehold distribution of food: a review of the literature and
discussion of the implications for food
Recommended