Você está na página 1de 11

Food Sec.

(2012) 4:581591
DOI 10.1007/s12571-012-0208-x

ORIGINAL PAPER

Vulnerability of smallholder rural households to food


insecurity in Eastern Ethiopia
Ayalneh Bogale

Received: 3 January 2012 / Accepted: 30 July 2012 / Published online: 12 August 2012
# Springer Science+Business Media B.V. & International Society for Plant Pathology 2012

Abstract Although several empirical methodologies as to score reveals the extent to which the problem has persisted.
how best assess vulnerability to food insecurity have been Although this was reduced from 43.2 to 28.7 (33.6 %)
proposed in the literature, none of these has evolved into a between 1990 and 2011, the latter figure still ranks
unanimously accepted approach. This article contributes to this Ethiopia 77th out of 81 nations and the situation remains
literature by adapting the Vulnerability as Expected Poverty alarming1 (IFPRI et al. 2011). Food insecurity and pover-
approach from poverty analysis methodology with the aim of ty have persisted even during the comparatively stable po-
scrutinizing factors determining household level vulnerability litical period following the downfall of the socialist regime
to food insecurity based on cross-section data collected from in 1991 largely due to lack of effective implementation of
277 randomly selected households in eastern Ethiopia. strategies to address the specific and lasting effects of dis-
Vulnerability to food insecurity was strongly associated with asters, which also require sustained engagement of institu-
several factors which included family size, size of cultivated tions rather than mere focus on emergency responses
landholding, soil fertility status of plots, access to irrigation, (Lautze et al. 2003).
number of extension visits, use of fertilizer and improved seed. Although the context of vulnerability to food insecurity is
The probability that any given households food consumption changing across different regions of the world, it still largely
expenditure would fall below a specified cut-off level has also remains a rural and agricultural phenomenon. Most of the
been computed and vulnerable households identified. The total poor in Ethiopia depend upon the rural sector for their live-
number of vulnerable households (111) was found to be greater lihoods (Dercon 2008). Some 84 % of the population in
than those who are currently food insecure (103). This implies Ethiopia lives in rural areas and is mainly engaged in rain-
that design and implementation of food security policies and fed subsistence agriculture. Thus, vulnerability to food in-
strategies need to focus not only on those who are observed to security in the country is also determined by external factors
be currently food insecure, but also on setting up social protec- such as rainfall patterns, land degradation, climate change,
tion mechanisms to help prevent households from falling more population density, low levels of rural investment and the
deeply into food insecurity in the future. global market. Considering these facts, increasing agricul-
tural productivity has been highlighted as a fundamental
Keywords Food insecurity . Vulnerability as expected step in reducing food insecurity because it not only can
poverty . Food consumption expenditure increase food availability and access but also rural income.
Some of the past interventions aimed at promoting agricul-
tural productivity in the country include identifying appro-
Introduction priate areas of investment, accelerating policy reforms,
developing institutions to support the private sector and
Food insecurity has characterized Ethiopia in the last few creating comprehensive development strategies to allocate
decades and the most recent Global Hunger Index (GHI) resources efficiently (World Bank 2005; Adhikari 2011).
Investment in agriculture and improving resilience among
A. Bogale (*)
1
University of KwaZulu-Natal, The categories are low (GHI4.9), moderate (5.0GHI9.9), serious
Pietermaritzburg 3209, KwaZulu-Natal, South Africa (10.0GHI19.9), alarming (20.0GHI29.9), and extremely alarm-
e-mail: bogalea@ukzn.ac.za ing (30.0)
582 A. Bogale

smallholder farmers remain key to providing sustained ac- reviews the literature on concepts and empirics of vulnerabil-
cess to food for all and reducing vulnerability to food ity. Section Data Collection and Description of the Study
insecurity due to natural disasters such as drought (FAO Area describes the data and study area. Section Empirical
2011). Such strategies at the ground level, which are pro- Methods presents the empirical methods employed. Section
moted by both the private and public sectors, include access Empirical Results and Discussion discusses the results and
to improved seeds, farm management techniques, irrigation, conclusions are provided in Section Conclusions.
fertilizer and soil and water conservation practices. These
increase productivity sustainably and reduce production
risks to farmers, especially smallholders. Furthermore im- Vulnerability: conceptual underpinning and empirics
proved access to both input and output markets is promoted.
In 2002, the Ethiopian government introduced its first There is little consensus among researchers and academics
donor-supported poverty reduction strategy, the Sustainable about the meaning of vulnerability due mainly to its diverse
Development and Poverty Reduction Program (SDPRP) use in the different disciplines of economics, psychology,
followed by the Plan for Accelerated and Sustained anthropology, environmental sciences and engineering. This
Development to End Poverty 2005/06-2009/10 (PASDEP) diversity and apparent lack of convergence over time are, in
(MOFED 2006). PASDEP intensified sectoral programs in many ways, a reflection of the divergent objectives of the
agriculture, health, education, and infrastructure in order to research and the phenomena being explained (Adger 2006).
achieve the MDGs which underscore the centrality of pov- Although human geographers and human ecologists have
erty reduction and attainment of sustained food security made significant contributions to the present understanding
(Carter and Barrett 2005). PASDEP put in place new pro- of social-ecological systems, they theorized vulnerability
grams for food security and made some improvements to the mainly in the context of environmental change and exposure
strategy of agricultural development-led industrialization. to hazards or shocks. Studies of food insecurity define
The Food Security Program (FSP) was reformulated to vulnerability through entitlement theory (Sen 1981), which
address problems of shortfalls in food production, vulnera- focuses on actual or potential resources available to house-
bility to falls in household incomes and consequent hunger, holds based on their own production, assets or reciprocal
which the country has faced repeatedly, through adaptation arrangements. However, food insecurity, which describes
of development alternatives to bring about a lasting solution. livelihood states of a household at a particular point in time,
While, on an aggregate scale, food insecurity still persists is not the same as vulnerability (Ellis 2003). Vulnerability is
at debilitating levels in Ethiopia, this does not say much a forward looking concept and a household is said to be
about the location specific extent and determinants of house- vulnerable to future loss of welfare if any risky event
hold vulnerability to food insecurity (Mwanakatwe and reduces household welfare below socially accepted norms.
Barrow 2010). Vulnerability is usually portrayed in terms It seeks to describe how prone individuals and families are
of susceptibility to harm from exposure to stress associated to being unable to cope with uncertain adverse events that
with environmental and social change and from the absence may happen to them, such as prolonged lack of rainfall, or
of capacity to adapt (Adger 2006). The term also refers to other difficulties. The degree of vulnerability depends on the
the relationship between poverty, food insecurity, risk and characteristics of the risk and the households ability to
efforts to manage risk (Alwang et al. 2001). In all formula- respond which, in turn, depends on household character-
tions, the key parameters of vulnerability are the stress to isticsnotably their asset-base (Alwang et al. 2001; Dilley
which a household is exposed, its sensitivity, and its coping and Boudreau 2001).
and adaptive capacity. After critical assessment of recent conceptual, methodo-
The present study scrutinizes factors that are associated logical and empirical developments in vulnerability assess-
with household level vulnerability to food insecurity by ments, Dilley and Boudreau (2001) concluded that, on a
adapting the Vulnerability as Expected Poverty (VEP) ap- practical level, lack of common understanding about funda-
proach (Chaudhuri 2003), which has been extensively used mental terms, data requirement and interpretation of vulner-
to analyze vulnerability to poverty (Sarris and Karfakis ability to food insecurity have hampered the development of
2006; Gaiha and Imai 2008; Gnther and Harttgen 2009; both theory and methods. However, there is increasing
Kruy et al. 2010). It makes use of household survey data recognition that there are considerable flows of households
collected from the eastern highlands of Ethiopia. The study into and out of poverty and food insecurity over time which
not only contributes to the emerging and scarce literature on instigated focus on analysis of household vulnerability as
vulnerability to expected food insecurity but also can be the basis for a social protection strategy (Hoddinott and
adapted and scaled-up to national level when such data are Quisumbing 2003). A forward looking approach, which
available. The article is structured as follows. Section identifies not only households that are presently food inse-
Vulnerability: Conceptual underpinning and empirics cure but also those that are vulnerable to shocks and other
Vulnerability of rural households to food insecurity 583

risks, such as natural disasters and extreme climate conditions, household characteristics and then enabled them to profile
is imperative. This will enable the task of protecting house- the most vulnerable households. Cruces et al. (2010) pre-
holds from the adverse effects of these shocks to be performed sented cross-sectional vulnerability estimates for 18
more effectively (Holzmann and Jorgensen 2000). countries in Latin America, and also used longitudinal data
Significant attempts have been made to analyze vulnera- for a sub-set of countries for which it was available. The
bility of households in developing countries, from which validation exercise indicated that the vulnerability model
some relevant policy options have been proposed. Prowse provided good estimates of aggregate poverty level.
(2003) provides a review of the relationship between vul- The vulnerability analysis approach has also been used to
nerability and chronic poverty and argues that vulnerability classify households into those that suffer from chronic pov-
should be more widely recognized as being a cause, symp- erty and those in transient poverty. Kruy et al. (2010)
tom and constituent of poverty. He advocates placing an assessed poverty issues in Cambodia by estimating house-
increased emphasis on assets and entitlements in order to hold vulnerability to poverty and classified the poor into the
understand individual catastrophes as opposed to analysis chronic poor and the transient poor, based on poverty inci-
of the strength or severity of shocks. Hoddinott and dence and vulnerability. The main results suggested that
Quisumbing (2008) provide a comprehensive review of poverty policy needs to focus not only on poverty allevia-
literature on approaches to vulnerability including tion, but also on setting up social protection mechanisms to
Vulnerability as Expected Poverty (VEP), Vulnerability as help prevent people from falling into deeper poverty traps in
Low Expected Utility (VEU) and Vulnerability as Uninsured the future.
Exposure to Risk (VER). Other approaches to vulnerability measurement in ap-
Chaudhuri et al. (2002) and Chaudhuri (2003) provide plied work involve the use of panel data at household level.
some of the initial contributions to the recent literature on Studies using this approach include, among others, those of
Vulnerability as Expected Poverty and have broadened the Dercon (2004), Dercon et al. (2005), Skoufias and
scope of poverty analysis to take account of vulnerability. Quisumbing (2003), Dercon and Krishnan (2000) for
They developed a conceptual and methodological overview Ethiopia, Suryahadi et al. (2000) for Indonesia, Kamanou
and demonstrated this empirically using cross section data at and Morduch (2002) for Cote dIvoire and Dhamija and
the household level from the Philippines, Indonesia and Bhide (2011) for India. Although these studies are informa-
China. Their results show that the vulnerable population, tive for policy and are methodologically sound, their use is
defined by the computed expected value of a poverty score relatively limited in most developing countries where large
in the future and conditional on a series of covariates, is amounts of panel data are practically non-existant.
generally larger than the fraction observed as poor at a given
point in time. This implies that the true poverty cost of risk
is higher than that observed. Data collection and description of the study area
Other applications of the methodology, employing cross-
section household data, provide findings along similar lines. Data collection
These include those of Sarris and Karfakis (2006) for
Tanzania, Gnther and Harttgen (2009) for Madagascar, The analysis of vulnerability to food insecurity in this paper
Gaiha and Imai (2008) for Tanzania, Zhang and Wan is based on a household survey conducted from August to
(2008) for China, Azam and Imai (2009) for Bangladesh October 2010, in order to minimize the seasonal variation in
and Deressa et al. (2009) for Ethiopia. Gnther and Harttgen expenditure on household consumption following harvest
(2009) assessed empirically the vulnerability of households and the hungry period, in three districts in eastern
to idiosyncratic and covariate shocks in Madagascar and Ethiopia. The three districts, namely Haramaya, Kersa and
showed that covariate shocks have a relatively higher impact Tulo were selected purposively to capture agro-ecological,
on rural households. Deressa et al. (2009) employed the economic and social diversities within the eastern highlands
vulnerability approach to measure vulnerability of rural of Ethiopia. The household survey was carried out after a
households to climate related shocks in Ethiopia. The results two-stage random sampling procedure. In the first stage
demonstrated that vulnerability of households is highly de- five, four and three peasant associations (PAs) were selected
pendent on minimum daily requirementsthe higher the randomly from a list of PAs in the three districts. In the
daily requirements, the more vulnerable the household. second stage, sample households were randomly drawn
Capaldo et al. (2010) used a vulnerability analysis model, from a complete list of respective PA members in confor-
drawing on household survey data from Nicaragua, and mity with the proportionate to size random sampling proce-
estimated the probability that a given household would gain dure. In total, the survey covered 277 households. The
access to sufficient food in the near future. The model selection of appropriate households for the intensive house-
allowed them to analyse how vulnerability related to various hold survey was based on the intermediate results of the
584 A. Bogale

extensive survey. One of the selection criteria was to en- replant areas where previously planted crops have been
sure representation of respondents with contrasting char- damaged by drought stress.
acteristics in terms of prevailing farming systems. The The most important perennial crop in the zone in terms of
sample size took into account the complexity of the issue, area coverage is tchat. This is a shrub, which is grown for
and the accuracy and adequacy of data necessary for the its narcotic substances, cathine and cathinone. These are
statistical analysis to be used. A structured questionnaire found in the tips of young leaves and are released when
was designed and pretested before executing the intensive they are chewed. Tchat requires little input of labour, is
household survey. The survey elicited information on fairly resistant to pests and disease and can withstand mois-
household characteristics such as demography, education, ture stress. Once the plant is established, the only manage-
inventory of assets, current production and use of non- ment practices required are weeding and harvesting. The
labor input, subjective assessment of soil fertility status plant may be harvested 3 years after planting. Owners may
and soil and water conservation measures, among other pick the crop continuously for their own consumption but it
factors. may also be harvested two to three times per year according
to the availability of irrigation water.
The study area

Agriculture is the mainstay of rural households in the east- Empirical methods


ern highlands of Ethiopia. Activities of households are cen-
tered on crop production and animal husbandry as these are This study expresses food security as the capacity of house-
important components of the integrated mixed farming sys- holds to spend a predetermined amount of money on food
tem. The scanty reconnaissance studies so far undertaken, required to achieve the daily minimum dietary requirement
localized detailed studies and general observations indicate of 2,100 kcal per adult equivalent. Conversely, a household
that the region is of high agricultural potential because it is is considered to be vulnerable to food insecurity if its
endowed with suitable and productive land resources. These expected expenditure on food will be less than the prede-
include favorable climate, fertile soils, diverse ecology, and termined amount.
length of growing season. However, in practice, high and Two approaches may be distinguished in modeling the
increasing population pressure, land and water resource determinants of household level food insecurity. The first
degradation and recurrent drought have turned the region represents food security status of households as a binary
into one of the parts of the country that is most structurally choice model where the endogenous variable is expressed as
food insecure. a dummy, with 1 representing the household being food
The region is characterized by three agro-ecological insecure and 0 otherwise. The second approach expresses
zones: the lowlands (3040 %), the midlands (3545 %) household level food insecurity based on food consumption
and the highlands (1520 %). There are two rainy seasons; expenditure as an indicator of wellbeing and defines food
belg (March to May) and meher (July to September). Belg insecurity in terms of the households per capita consump-
rains are mainly used for land preparation and planting of tion level (World Bank 2002; Demeke et al. 2003). Many
long cycle crops such as maize and sorghum and seed bed researchers have successfully employed the latter model in
preparation for meher crops. The meher rains are used for analyzing determinants of poverty and food insecurity in
planting cereal crops such as barley, teff, wheat and vegeta- developing countries including Ethiopia (Albert and
bles e.g. onion, shallot and potatoes in the mid- and the Collado 2004; Demeke et al. 2003; Datt et al. 2000;
highlands and peanuts in the lowlands. Besides, meher rains Capaldo et al. 2010).
are also responsible for the growth and development of Accordingly, in order to analyze the household level data
perennial crops such as coffee and tchat (Catha adulis). collected for this study, various empirical models were used.
Despite the privilege of being able to grow these cash crops, The study adapted the methodology proposed by Chaudhuri
the region repeatedly faces food shortages. et al. (2002) to estimate the expected mean and variance in
Cereals, specially sorghum and maize are the major an- food consumption expenditure using the cross-section data,
nual crops grown in the region. Sorghum is the most impor- and the Foster, Greer and Thorbecke (Foster et al. 1984)
tant staple food, followed by maize. The average yield of decomposable poverty index.
sorghum ranges from 1.0 to 1.4 t per ha while that of maize Gaiha and Imai (2008) extend the procedure proposed by
ranges from 1.4 to 1.8 t per ha. Teff and wheat are less Chaudhuri et al. (2002) to measure the vulnerability of
important cereals in terms of the area covered by the crops. households to poverty in rural India. Gnther and Harttgen
However, they serve the strategic purpose of covering arable (2009) empirically illustrated how the model can be
land which has not been planted, either due to lack of oxen employed to analyze households vulnerability to idiosyn-
for tilling or shortage of labour. They may also be used to cratic and covariate shocks in Madagascar using cross-
Vulnerability of rural households to food insecurity 585

sectional data. Whereas Capaldo et al. (2010) provided three-step Feasible Generalized Least Squares (FGLS)
estimates of the probability that a given household would (Christiaensen and Subbarao 2005; Chaudhuri et al. 2002).
lose or gain access to sufficient food in the near future, This involves first estimating Eq. (2) employing the
drawing on household survey data from Nicaragua. The Ordinary Least Square (OLS) method. Then Eq. (3) is
same procedure can be followed in estimating vulnerability estimated by OLS using the squared residuals from the
to food insecurity. In this case, vulnerability of households estimation of Eq. (2) as dependent variables. The predictions
is defined as the probability that a household will be food from this regression are used to re-estimate Eq. (3) by OLS
insecure in the future and can be expressed as: after having weighted each residual by Xi. The new esti-
  mates of are asymptotically efficient, and are used to
Vit Pr Ci;t1  Z 1 weigh Eq. (1) and re-estimate it by weighted least squares,
to obtain asymptotically efficient estimates of (Chaudhuri
where vulnerability of household to food insecurity at time t, et al. 2002). Then, using the estimates b b and b , we can
Vit, is the probability that the i-th households level of food compute the expected log food consumption expenditure
consumption expenditure at time t+1, Ci,t+1, will be below and the variance of log food consumption expenditure for
the minimum daily expenditure required to meet the calorie each household as follows (Gaiha and Imai 2008):
requirement of 2,100 kcal per adult equivalent, Z. This
minimum threshold can alternatively be referred to as the E ln Ci jXi  Xi b
b 4
food poverty line.
Following Chaudhuri et al. (2002), Gaiha and Imai
(2008) and Gnther and Harttgen (2009) derived empirical- V ln Ci jXi  Xib
5
ly a variant of VEP from the food consumption expenditure
function as: By assuming that food consumption expenditure is log-
normally distributed, and using the estimated parameters of
ln Ci Xi b "i 2 the model, the probability that a household will be food
insecure in the near future (say, at time t+1), is expressed
where Ci represents the food consumption expenditure per as (Gaiha and Imai 2008; Gnther and Harttgen 2009):
adult equivalent for the i-th household, Xi represents a !
bundle of observable household characteristics, is a vector ln Z  ln Cbi
b b
Vi Pln Ci < ln ZjXi f p 6
of coefficients of household characteristics to be estimated,
b2i
and i is a disturbance term with mean zero and variance of
2e;i. Food consumption expenditure Ci is assumed to be log- Where (.) is the cumulative density of the standard normal
normally distributed and as such the disturbance term, i distribution function,
bis the standard error of the regression;
will be normally distributed. Furthermore, it is assumed that Z is the prescribed threshold food expenditure to meet the
i captures the idiosyncratic shocks that contribute to the minimum energy requirement, in short, food poverty line. Vbi
difference in food consumption expenditure levels for is a set of estimates one for each household and denotes the
households that share the same characteristics. But it is probability that each household faces of falling below the
unlikely that it captures covariate shocks which can affect minimum threshold in the future. Its value ranges between
all households at a given time and unexpected very large zero and 1. When Vbi 0; household i will spend an ade-
negative shocks such as economic crises. quate amount on food for consumption in the future with
Furthermore, we assume that the variance of the unex- certainty that at least the minimum amount of calories
plained part of food consumption expenditure i depends on required will be obtained, and when Vbi 1; household i
household is observable characteristics: will consume fewer calories than the prescribed threshold.
Even though various options have been discussed in the
2e;i Xi t i 3
literature as to how to determine the cut-off point to classify
where represents a vector of parameters to be estimated, households as vulnerable and non-vulnerable, Capaldo
is the vector of residuals of this second estimation. Standard et al. (2010) justify the use of 0.5 as appropriate for studies
regression analysis based on OLS assumes homoscedastic- in undernutrition and food insecurity. Accordingly, in this
ity, and estimates of and will be unbiased but inefficient study a household is classified as vulnerable if Vbi is greater
if this assumption does not hold. To deal with this problem than or equal to 0.5, and non-vulnerable, otherwise.
and obtain consistent estimate of parameters, it is necessary Even though such a model enables researchers to esti-
to allow heteroskedasticity, that is, variances of the distur- mate vulnerability from cross-section data in the absence of
bance term to vary across households depending on Xi. time-variant information on consumption expenditure, it
Thus the estimates of and could be obtained using should be noted that the measure can correctly reflect
586 A. Bogale

vulnerability assuming that the distribution of consumption the basket was determined in such a way that the given
expenditure across households at a given time represents the bundle meets the predetermined level of minimum daily
time-series variation of consumption expenditure of the dietary requirement i.e. 2,100 kcal per day per adult equiv-
household (Hoddinott and Quisumbing 2003; Gaiha and alent. This basket was valued at local prices and the value
Imai 2008). Obviously, the existence of measurement error of the food poverty line was determined. As a result, the
is a major concern for the estimation of vulnerability which food poverty line for the study area was found to be ETB
could lead to an overestimation of variance in consumption 2018. In order to make possible a comparison of this figure
(Gnther and Harttgen 2009). Thus, in the regression of with those of other regions and the nationwide figure as well
consumption expenditure of households, it is assumed that as taking into account the effect of inflation, this poverty
the data are fairly accurate. Unfortunately, household line figure for food was deflated by the survey year con-
expenditure surveys, particularly in developing countries, sumer price index (CPI) of the Oromiya region. The food
do not meet this ideal situation for a variety of reasons. In CPI of the survey year was reported to be 187.8 % (2,0060
case the error of measurement in the dependent variable is 100) (CSA 2011). Thus the deflated food poverty line was
uncorrelated with the exogenous variables, the estimated estimated to be ETB 1075 per adult equivalent per year at
coefficients will be unbiased and the estimated variances December 2006 constant price. These results are extensively
are expected to be larger but still unbiased (for details see used in the subsequent analysis of food insecurity in the
Gujarati 2004: pp. 525526). Therefore, even though present study.
lengthy panel data are preferred for the estimation of house-
hold vulnerability, the VEP approach is an attempt at assess- Variable description
ing vulnerability of households where there is no panel data,
as is often the case in developing countries (Gnther and The summary statistics of variables hypothesized to
Harttgen 2009). influence vulnerability of households to food insecurity
Finally, a weighted average of the probabilities of being and included in the empirical model are given in Tables
food insecure provides the predicted aggregate index for the 1, 2. These include demographic and social character-
sample as a whole. Thus, the aggregate head count food istics, provision and use of productivity enhancing inputs and
insecurity index V0 of the sample was calculated as the service infrastructure, and productive asset holding of the
weighted mean of the above estimated probability that a household.
household will be food insecure, where the weights are Information on the demographic structure and character-
given by households size (hi) expressed in adult equivalent istics of a household includes gender of the household head,
terms (Demeke et al. 2003; Datt et al. 2000). family size, dependency ratio and age of the head of house-
hold. In the study area, male household heads have better
opportunity to have access to assets and therefore female
Empirical results and discussion headed households are expected to be more susceptible to
food insecurity (Bigsten et al. 2002; Bogale et al. 2005). It is
Determination of food poverty line also hypothesized that households with larger numbers of
members and a higher dependency ratio have a greater
Quite a wide range of methods of food security measure- probability of being food insecure. As age captures the
ment have been developed and used for empirical research, experience accumulated in farming, it may therefore be
among which individual dietary intake, dietary diversity and associated with better welfare (Albert and Collado 2004).
indices of household coping strategies are the most com- Exposure of the household head to education can affect a
monly used (Maxwell 1996; Hoddinott 2002). Each of these households standard of living as it helps in the development
methods has specific data requirements and analytical pro- of skills. These may result in higher marginal productivity
cedures. In this study, we adapted the household dietary of labor that eventually enables members of the household
intake in order to determine the expenditure on food re- to engage in more remunerative jobs. Hence it is expected
quired to meet the minimum daily dietary energy require- that education would be positively correlated with house-
ment, i.e. the food poverty line. Although the data holds consumption levels. Social capital is captured by the
requirement of this method is considerable and its collection number of people, including traditional self-help associations
necessitates a very comprehensive questionnaire, it remains and individuals living outside the location, that the household
a widely accepted measure of food insecurity as far as its resorts to in case of facing idiosyncratic shocks. Number of
economic dimensions are concerned. Accordingly, the fol- household members who were sick during the past 12 months
lowing procedure has been used. In the first run a basket of at the time of survey also includes information on idiosyncrat-
food items typically consumed by the poor were identified ic shock. It is assumed that poor health affects productivity
from the food consumption questionnaire. The quantity of and living standards in many ways, such as through low labor
Vulnerability of rural households to food insecurity 587

Table 1 Definition of variables and summary statistics of household characteristics

Variable code Viable label Mean Std. Dev. Min. Max.

LnCONEXAE Natural log of consumption expenditure per adult equivalent 7.13469 0.374758 6.33 8.1
SEXHH Dummy of sex of household sex (male01) 0.88 0.32 0 1
FAMSZAE Size of household in adult equivalent 4.65 1.47 1.6 9.82
EDUC Education level of the household head in years of schooling completed 2.78 2.41 0 12
AGE Age of the household head in years 35.5 8.18 25 68
DEP Dependency ratio 1.39 0.78 0 5
LDCULT Total cultivated land holding 0.84 0.34 0.35 2.42
TLUAE Livestock holding in tropical livestock unit 0.39 0.22 0 1.42
DIVCROP Crop diversification 4.60 1.33 1 8
FERT Use rate of fertilizer 0.27 0.19 0 0.95
SEED Use rate of improved seeds 0.15 0.08 0 0.32
IRRG Dummy for access to irrigation (Yes01) 0.19 0.39 0 1
EXVIST Number of visits by extension agent 1.87 1.20 0 4
CreditA Dummy for access to credit (Yes01) 0.27 0.44 0 1
DISPSER Average walking distance to public services in hours 1.68 0.81 0.5 4
SoilF Dummy for fertility status of plots (Fertile01) 0.31 0.46 0 1
RELY Number of person to rely on during shocks 2.92 1.55 0 9
SICK Number of sick persons in 1 year 0.83 0.87 0 4

Table 2 Estimates of expected log food consumption expenditure and variance (N0277)

Viable label Log food consumption Variance of food


expenditure consumption expenditure

Coef. Robust Std. Err. Coef. Robust Std. Err.

Dummy of sex of household sex (male01) 0.0238 0.0219 0.0077** 0.00332


Size of household in adult equivalent 0.0438*** 0.0088 0.00045 0.00114
Education level of the household head in years of schooling completed 0.0038 0.0047 0.0018** 0.00075
Age of the household head in years 0.0021 0.0013 0.00043 0.00034
Dependency ratio 0.01158 0.0094 0.00040 0.00125
Total cultivated land holding 0.0889*** 0.0314 0.00540 0.000435
Livestock holding in tropical livestock unit 0.0146 0.0398 0.00333 0.00576
Crop diversification 0.1149* 0.0069 0.00028 0.00106
Use rate of fertilizer 0.9038*** 0.0990 0.02006 0.01641
Use rate of improved seeds 1.2966*** 0.2012 0.02858 0.02903
Dummy for access to irrigation (Yes01) 0.0632*** 0.0208 0.00027 0.00318
Number of visits by extension agent 0.01357** 0.0067 0.00035 0.001558
Dummy for access to credit (Yes01) 0.0233 0.0187 0.00595 0.00496
Average walking distance to public services in hours 0.0176 0.0123 0.00050 0.00244
Dummy for fertility status of plots (Fertile01) 0.0505*** 0.0167 0.00051 0.00263
Number of person to rely on during shocks 0.0126* 0.0065 0.00051 0.00125
Number of sick persons in 1 year 0.0036 0.0083 0.00122 0.00154
Constant 6.6754*** 0.07555 0.0139 0.0199
F(17, 257) 178.55 6.78
Prob>F 0.000 0.000
R-squared 0.9004 0.1992
Root MSE 0.12561 0.0172
588 A. Bogale

productivity due to physical weakness and the high cost of Of the seventeen variables considered in the model, nine
medicines. Thus larger numbers of sick persons are hypothe- had a statistically significant impact on expected expendi-
sized to be positively related to food insecurity. ture on food and therefore food security status in the future.
Household asset holdings were assessed using owned and These included household size, number of individuals to
operated cultivated land and its fertility status, ownership of rely upon during shocks, number of visits by extension
livestock and access to credit. Losses of farm land to other agent, use of fertilizer and improved seeds, soil fertility
uses because of population growth and resource degrada- status, crop diversification, total land holding and access to
tion, as well as limits to the amount of new land that can be irrigation.
brought into cultivation are major constraints for food pro- Household size, expressed as adult equivalents, had a
duction in the study area. An average household owns only significant negative impact on expected food expenditure
0.84 ha of cultivated land on which an average of about 5 (P<0.01), the chance of falling into food insecurity increas-
household members has to be supported. In a farming sys- ing with the size of this parameter. This result clearly indi-
tem where agricultural productivity remains at about 1 t of cates the importance of family planning in the study area
cereal per hectare, attainment of sustainable food security is and the necessity of improving access of the poor to educa-
challenging with such very small land holdings. Moreover, tion and information on family planning methods.
nearly 79 % of the households consider that their cultivated Adoption of yield enhancing technologies including use
plots are degraded and they rely on rain-fed agriculture with of fertilizer and improved seeds, as well as support by
only 19 % having access to irrigation. extension agents, also improved expected food expenditure
Finally, we also used a set of variables that reflect provi- of households and therefore enhanced food security. These
sion and use of productivity enhancing inputs and service inputs and services contributed towards increase in agricul-
infrastructure. It is very evident that the use of fertilizer and tural income leading to higher levels of consumption.
improved seeds make direct positive contributions to agri- Moreover, higher agricultural income, besides reducing
cultural productivity and production and the probability of food insecurity, improved the welfare status of households
being food secure. In the study area, however, average which, in turn, enabled them to invest more in such items as
households use fertilizer at only 26.6 % of the recommended livestock and access to education.
rate and improved seeds are planted on only 15.1 % of the Cultivated land is the most fundamental resource of rural
cultivated plots. Moreover, members of an average house- households in Ethiopia: it is essential for the generation of
hold need to walk an average of 96 min to basic services income, accumulation of wealth and its transference be-
including health centre, veterinary services, education and tween generations and, most importantly, it is a hedge
water services. Restricted access to these services means against food insecurity. The size of land cultivated had a
less time for productive activities and results in lower stan- statistically significant (P<0.01) and positive influence on
dard of living. Table 1 provides a list of all variables, expected food expenditure per adult equivalent. The size of
including their mean values and standard deviations. landholding is a surrogate for a host of factors including
wealth, access to credit and capacity to bear risk. Larger
Determinants of vulnerability to food insecurity farms are associated with greater wealth and availability of
capital which increases the probability of investment in farm
We estimate the expected log food consumption expenditure inputs that increase food production. Crop diversification
and variance of log food consumption after accounting for was also positively correlated with expected food consump-
heteroskedasticity through the use of generalized least squares tion and was statistically significant.
method discussed earlier. Generally, the model had good Better land quality is also positively related to expected
overall fit and most variables performed as expected. food expenditure. However, Ethiopia, as one of the few
Although endogeneity may be concerns for some of the ex- African countries with a long history of settled agriculture
planatory variables, this exercise, as is the case for any econo- has exhausted its soil fertility much earlier than most.
metric analysis, serves as a vehicle to estimate relative Moreover, failure to make the crucial transition to sustain-
vulnerability and not to identify direct causes of inadequate able and intensive use of the land and failure to introduce the
expenditure on food (Capaldo et al. 2010). The F statistic of institutional adjustments required to reverse land degrada-
the model is highly significant (P<0.001 with 17,259 degrees tion has constrained the flow of investment capital, knowl-
of freedom), indicating that the expected expenditure on food edge and skilled management to the rural areas where the
by households is related to the independent variables. Based majority of the Ethiopian population lives.
on a simple multicolinearity test using Variance Inflation Access to irrigation is also among variables with a statis-
Factor (VIF) for continuous variables and Contingency tically significant and positive coefficient. Research findings
Coefficient for dummy variables, it was confirmed that there consistently indicate that irrigation development enhances
is no strong correlation among the independent variables. productivity, reduces household food insecurity and
Vulnerability of rural households to food insecurity 589

alleviates poverty in rural areas of Ethiopia (Bacha et al. It is relevant for policy makers to take account of the
2011; Tesfaye et al. 2008). In the study area, farmers use totality of vulnerable groupsabout 40 % in the sample
low-cost pumps, wells, stream diversion and water harvest- surveyedin order to effectively address the challenges of
ing technologies to supplement moisture for their crops. food insecurity. These vulnerable groups consist of house-
There are also a few modern irrigation schemes managed holds that are chronically food insecure, transiently food in-
by irrigation water users. Access to irrigation enables rural secure, and those that are currently food secure but are highly
households to produce more than one crop per year, increase vulnerable to food insecurity. Remedial action could take the
their income and consumption levels and diversify their form of some combination of prevention, protection, and
cropping systems. However, irrigation cannot be considered livelihood promotion interventions which would give them a
as a panacea for poverty alleviation and food insecurity more secure basis from which to diversify their activities and
(Bacha et al. 2011). There are farm households which may enable them to cope better with shocks and risks.
remain food insecure irrespective of their access to irrigation
because there are other correlates of food insecurity includ-
ing family size, lack of education, training and improved Conclusions
inputs. These must also be addressed in order to have a deep
and lasting impact on food insecurity. Ethiopia has enjoyed about 912 % annual expansion in the
Once food security and vulnerability thresholds have economy during the past 7 years. Nevertheless, the incidence
been established, based on current levels, these may be of food insecurity is not only higher than most developing
extended into several food insecurity and vulnerability cat- countries but still remains at unacceptable levels. Therefore,
egories as in Table 3. First, households may be classified as attaining food security will remain among the principal objec-
of low or high vulnerability according to whether the vul- tives of the development policy of the country for the foresee-
nerability score is <0.5 or 0.5, respectively (Pritchett et al. able future. Informed policy decisions to attain food security
2000). Second, the sample households may be divided into require research outputs with relevant policy implications. As
two distinct groups using the food poverty line cut-off: the a contribution to this aim, this study was carried out by
food insecure and the food secure. As discussed earlier, adapting Vulnerability as Expected Poverty (VEP) approach
those who have average food consumption expenditure be- to food insecurity. The approach was based on cross-section
low the food poverty line are generally termed as food data and so minimizes reliance on panel data which is hard to
insecure and the rest as food secure. obtain in most developing countries. The study systematically
The results reveal that 90 (32.5 %) households suffered constructed appropriate probability distribution of food con-
from chronic food insecurity as their food consumption ex- sumption expenditures of sampled households conditional on
penditure was below the cut-off and at the same time were their characteristics and subject to idiosyncratic shocks which
highly vulnerable to food insecurity. These households are were then used to estimate head count ratio. Moreover, house-
most likely to remain food insecure in the future. In contrast, hold demographic and social characteristics, provision and
13 (4.69 %) households were in transient food insecurity, use of productivity enhancing inputs and service infrastruc-
being currently food insecure but having low vulnerability to ture, and productive asset holding of the households were
food insecurity. They may escape food insecurity in the near analyzed as determinants of expected food consumption ex-
future. Moreover, 8 (2.88 %) of the sample households were penditure. We found that vulnerability to food insecurity was
food secure at present, while being at risk of being food strongly associated with several factors which included family
insecure in the future. That is, these households were living size, size of cultivated landholding, soil fertility status of plots,
under the threat of becoming food insecure in the near future. access to irrigation, number of extension visits, use of fertilizer
As Table 3 shows, only 166 (59.93 %) households enjoyed and improved seed.
relatively stable levels of food security as they were both food The approach also allowed us to derive categories of
secure and not vulnerable. households relating to their vulnerability and food security

Table 3 Classification and decomposition by vulnerability and food security status (N0277)

Food security status

Food insecure (Exp<Z) Food secure (Exp>Z) Total

Vulnerability High vulnerability Vi >0.5) CFI 90 (32.5 %) HVFS 8 (2.88 %) 98 (35.38 %)


Low vulnerability (Vi <0.5) TFI 13 (4.69 %) LVFS 166 (59.93 %) 179 (64.42 %)
Total 103 (37.18 %) 174 (62.82 %) 277 (100 %)
590 A. Bogale

status. Accordingly, sample households were grouped into Indonesia. Columbia University Department of Economics Dis-
chronic food insecure (32.5 %), transient food insecure cussion Paper No. 0102- 52, New York: Columbia University.
Christiaensen, L., & Subbarao, K. (2005). Toward an understanding of
(4.69 %), highly vulnerable-food secure (2.88 %) and low household vulnerability in rural Kenya. Journal of African Econ-
vulnerable-food secure (59.93 %), whereas the head count omies, 14(4), 520558.
index for food insecurity was 37.18 %. The results not only Cruces, G., Gasparini, L., Brgolo, M., & Ham, A. (2010). Vulnerability
depict the fact that total vulnerable households (111) are to poverty in Latin America. Evidence from cross-section and panel
data. Report prepared for the Chronic Poverty Research Centre.
greater than those considered to be food insecure (103), CEDLAS, Universidad Nacional de La Plata and CONICET.
but also allow us reiterate the importance of vulnerability CSA. (2011). Statistical abstract 2010. Addis Ababa: Central Statisti-
assessment and analysis to monitor progress of targeted cal Agency.
interventions to reduce vulnerability of households. This Datt, G., Simler K., Mukherjee S., & Dava G. (2000). Determinants of
poverty in Mozambique: 1996-97. Food Consumption and Nutri-
fact directs us to conclude that food security policy and tion Division Discussion paper No. 78, IWashington, D.C.: nter-
strategies need to focus not only on those who are currently national Food Policy Research Institute.
food insecure, but also on setting up social protection mech- Demeke, M., Guta, F., & Ferede, T. (2003). Growth, employment,
anisms to help prevent people from falling more deeply into poverty and policies in Ethiopia: an empirical investigation of
issues in employment and poverty, discussion paper 12. Geneva:
food insecurity in the future. International Labour Office.
Dercon, S. (2004). Growth and shocks: evidence from rural Ethiopia.
Acknowledgments The author thanks the Editor-in-Chief of the Journal of Development Economics, 74, 309329.
Journal and two anonymous reviewers for very helpful and profession- Dercon, S. (2008). Rural poverty: Old challenges in new contexts.
al comments and suggestions on earlier versions of this article. GPRG-WPS-072, ESRC Global Poverty Research Group, UK.
Available at http://www.gprg.org/pubs/workingpapers/pdfs/
gprgwps-072.pdf
Dercon, S., & Krishnan, P. (2000). Vulnerability, seasonality and poverty
References in Ethiopia. Journal of Development Studies, 36(6), 2553.
Dercon, S., Hoddinott, J., & Woldehanna, T. (2005). Vulnerability and
shocks in 15 Ethiopian villages, 19992004. Journal of African
Adger, W. N. (2006). Vulnerability. Global Environmental Change, 16, Economies, 14, 559585.
268281. Deressa, T., Hassan, R., & Ringler, C. (2009). Assessing household
Adhikari, B. (2011). Poverty reduction through promoting alternative vulnerability to climate change: The case of farmers in the Nile
livelihoods: implication for marginal drylands. Journal of Inter- Basin of Ethiopia. IFPRI Discussion Paper 00935. Washington,
national Development. doi:10.1002/jid.1820. D.C.
Albert, J. R. G., & Collado, P. M. (2004). Profile and determinants of Dhamija, N., & Bhide, S. (2011). Poverty in rural India. Variations in
poverty in the Philippines, 9th National Convention on Statistics factors influencing dynamics of chronic poverty. Journal of In-
(NCS) EDSA Shangri-La Hotel. ternational Development. doi:10.1002/jid.1827.
Alwang, J., Siegel P. B., & Jorgensen, S. L. (2001). Vulnerability: A Dilley, M., & Boudreau, T. E. (2001). Coming to terms with vulnerability:
view from different disciplines. Social Protection Discussion Pa- a critique of the food security definition. Food Policy, 26, 229247.
per Series #0115. Washington, DC: The World Bank. Ellis, F. (2003). Human vulnerability and food insecurity: Policy
Azam, S., & Imai, S. I. (2009). Vulnerability and poverty in Bangla- implications. Forum for food security in Southern Africa. Over-
desh. Chronic Poverty Research Centre Working Paper No. 141. seas Development Group (ODG).
UK: University of Manchester. FAO. (2011). The state of food insecurity in the world 2011: How does
Bacha, D., Namara, R., Bogale, A., & Tesfaye, A. (2011). Impact of international price volatility affect domestic economies and food
small-scale irrigation on household poverty: empirical evidence security? Rome: Food and Agriculture Organization.
from the Ambo district in Ethiopia. Irrigation and Drainage, 60, Foster, J., Greer, J., & Thorbecke, E. (1984). A class of decomposable
110. poverty measures. Econometrica, 52, 761766.
Bigsten, A., Kebede, B., Shimeles, A., & Taddesse, M. (2002). Growth Gaiha, R., & Imai, K. (2008). Measuring vulnerability and poverty:
and poverty reduction in Ethiopia: Evidence from household estimate for rural India. Research Paper No. 40, World Institute
panel surveys, Working Papers in Economics No 65 Department for Development Economics Research, United Nations Universi-
of Economics Goteborg University. ty, Helsinki.
Bogale, A., Hagedorn, K., & Korf, B. (2005). Determinants of poverty Gujarati, D. N. (2004). Basic econometrics (4th ed.). New York:
in Ethiopia. Quarterly Journal of International Agriculture, 44(2), McGraw-Hill Companies.
101120. Gnther, I., & Harttgen, K. (2009). Estimating households vulnerabil-
Capaldo, J., Karfakis, P., & Knowles, M. (2010). A model of vulnera- ity to idiosyncratic and covariate shocks: a novel method applied
bility to food insecurity. ESA Working Paper Series No. 1003. in Madagascar. World Development, 37(7), 12221234.
Rome: Food and Agriculture Organization. Hoddinott, J. (2002). Methods for rural development projects: Food
Carter, M. R., & Barrett, C. B. (2005). The economics of poverty traps security in practice. Washington, D.C.: International Food Policy
and persistent poverty: An asset-based approach, SAGA (Strate- Research Institute.
gies and Analysis of Growth & Access) Working Paper. Univer- Hoddinott, J., & Quisumbing, A. R. (2003). Methods for microecono-
sity of Wisconsin-Madison. metric risk and vulnerability assessments. Social Protection Dis-
Chaudhuri, S. (2003). Assessing vulnerability to poverty: Concepts, cussion Paper Series No. 0324. Social Protection Unit, Human
empirical methods and illustrative examples. Mimeo. New York: Development Network. Washington D.C.: World Bank
Columbia University. Hoddinott, J., & Quisumbing, A. R. (2008). Methods for econometric
Chaudhuri, S., Jalan, J., & Suryahadi, A. (2002). Assessing household risk and vulnerability, Mimeo. Washington, D.C.: International
vulnerability to poverty: A methodology and estimates for Food Policy Research Institute.
Vulnerability of rural households to food insecurity 591

Holzmann, R., & Jorgensen, S. (2000). Social risk management: A new case of Filtino and Godino irrigation schemes in Ethiopia. Irriga-
conceptual framework for social protection and beyond. Social tion and Drainage Systems, 22(2), 145158.
Protection Discussion Paper No. 0006. Washington, D.C.: World World Bank. (2002). A source book for poverty reduction strategies,
Bank core techniques and cross cutting issues (Vol. 1). Washington,
IFPRI, Concern Woldwide and Welthungerhilfe (2011). The global D.C.: World Bank Press.
hunger index 2011. The Challenge of hunger: Taming price spikes World Bank. (2005). Agricultural growth for the poor: An agenda for
and excessive food price volatility. Washington, D.C. development. Washington, D.C.: The World Bank.
Kamanou, G., & Morduch, J. (2002). Measuring vulnerability to Zhang, Y., & Wan, G. (2008). Can we predict vulnerability to poverty?
poverty, UNU-WIDER Discussion Paper No. 2002/58. Helsinki: UNU WIDER Research Paper No. 2008/82. Helsinki, Finland.
United Nations University
Kruy, N., Kim, D., & Kakinaka, M. (2010). Poverty and vulnerability:
an examination of chronic and transient poverty in Cambodia.
International Area Review, 13(4), 323.
Lautze, S., Aklilu, Y., Raven-Roberts, A., Young, H., Kebede, G., &
Leaning, J. (2003). Risk and vulnerability in Ethiopia: Learning
from the past, responding to the present, preparing for the future. Ayalneh Bogale received his
A report for the U.S. Agency for International Development. PhD in Agricultural and Re-
Washington, D.C. source Economics from Hum-
Maxwell, D. (1996). Measuring food insecurity: the frequency and boldt University, Berlin and he
severity of coping strategies. Food Policy, 21, 291303. holds BSc and MSc degrees in
MOFED (2006). Building on progress. A Plan for Accelerated and Agricultural Economics from
Sustained Development to End Poverty (PASDEP) (2005/06- Haramaya University, the then
2009/10), Ministry of Finance and Economic Development, Alemaya University of Agricul-
Addis Ababa, Ethiopia. ture, Ethiopia. He has served at
Mwanakatwe, P., & Barrow, L. (2010). Ethiopias economic growth Haramaya University in various
performance: Current situation and challenges. The African De- academic ranks including Lec-
velopment Bank Group, Economic Brief, 1(5). turer, Assistant Professor and
Pritchett, L., Suryahadi, A., & Sumarto, S. (2000). Quantifying vulnera- Associate Professor. He has a
bility to poverty: A proposed measure, applied to Indonesia. World wide range of experience in han-
Bank Policy Research Working Paper No. 2437. Washington, D.C. dling lectures for both under-
Prowse, M. (2003). Towards a clearer understanding of vulnerability graduate and postgraduate students and has supervised MSc and PhD
in relation to chronic poverty. Chronic Poverty Research Centre students registered both at Haramaya University and Humboldt Uni-
Working Paper No 24, Manchester. versity. Research and fellowship awards include the Alexander von
Sarris, A., & Karfakis, P. (2006). Household vulnerability in rural Humboldt Research Fellowship (20072009), the Agriculture for
Tanzania. Paper presented at the CSAE Conference held on 19 Peace Research Fellowship from the International Foundation for
21 March 2006 at Oxford, United Kingdom. Science and the Institute for Advanced Studies of the United Nations
Sen, A. K. (1981). Poverty and famines: An essay on entitlement and University (2004) as well as the DAAD and Nuffic Scholarships. His
deprivation. Oxford: Clarendon. papers may be found in several academic journals, including the
Skoufias, E., & Quisumbing, A. R. (2003). Consumption insurance Journal of Development Studies, Land Degradation and Development,
and vulnerability to poverty: A synthesis of the evidence from Human Ecology, Ecological Economics, Irrigation and Drainage Sys-
Bangladesh, Ethiopia, Mali, Mexico and Russia. Paper presented tems, Quarterly Journal of International Agriculture, Journal of Rural
at the conference Staying Poor: Chronic Poverty and Develop- Development in the Tropics and Sub-tropics and Journal of Rural
ment Policy at the University of Manchester, 7 to 9 April 2003. Development. He has made numerous paper contributions to national
United Kingdom. and international conferences in order to share his findings from
Suryahadi, A., Sumarto, S., & Pritchett, L. (2000). Quantifying vulnera- collaborative research projects. His research interests are economics
bility to poverty: A proposed measure applied to Indonesia, Report of resource degradation, poverty and food security, property rights and
for the Social Monitoring and Early Response Unit, UNDP. adoption of technologies. He is currently Associate Professor in Food
Tesfaye, A., Bogale, A., Regassa, E. N., & Bacha, D. (2008). The Security and Director of the African Centre for Food Security (ACFS)
impact of small-scale irrigation on household food security: the at the University of KwaZulu Natal, South Africa.

Você também pode gostar