See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281774460

Shocks, Safety-nets and Vulnerability to Poverty in Ethiopia

Article · September 2015

CITATIONS READS
5 1,201

2 authors:

Catherine Porter Ruth Hill

Lancaster University World Bank
75 PUBLICATIONS 1,424 CITATIONS 56 PUBLICATIONS 2,425 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Catherine Porter on 15 September 2015.

The user has requested enhancement of the downloaded file.

 Shocks, Safety-nets and Vulnerability to Poverty in
Ethiopia
Ruth Vargas Hill Catherine Porter
The World Bank Heriot-Watt University∗
September 15, 2015

Abstract
Whilst the measurement and determinants of poverty have been widely studied,
vulnerability, or the threat of future poverty, has been more difficult to investigate
due to data paucity. We combine nationally representative household data with
historical drought information to investigate causes of vulnerability to poverty in
Ethiopia. We model consumption as a function of household assets, aggregate
shocks (drought, food prices), and idiosyncratic shocks (job loss, death, illness).
We simulate the realization of shocks based on their (heterogeneous) cross-sectional
and historical distribution using bootstrapping and create expected future levels of
consumption. We then calculate some of the new vulnerability measures proposed
in the literature, including a vulnerability headcount as well as measures that are
more sensitive to a worsening of the worst outcomes. If vulnerable households are
defined as having a probability greater than 0.5 of falling below the poverty line,
our estimates suggest that 42 percent of the population are vulnerable to absolute
poverty (compared to 29 percent currently defined as poor). In terms of the key
shocks driving the results, urban uneducated households (who are disproportion-
ally female-headed) suffer disproportionately from food price shocks, whilst rural
drought shocks are important, but are mitigated for participants in the national
safety-net program. The results suggest that significant welfare gains can be made
from risk management and social protection.
Keywords: Vulnerability, Poverty, Ethiopia.

∗
Many thanks to Christopher Gaukler and Radhika Goyal for excellent research assistance, in partic-
ular for their dedication and patience in accessing and compiling the many datasets used in this study.
Thanks to Tassew Woldehanna, Matthew Hobson, Camilla Holmemo, participants at CSAE conference
Oxford 2015 and anonymous reviewers for comments on earlier drafts. Errors and omissions are our
own. Comments welcome to catherine.porter@hw.ac.uk

1
1 Introduction
Households in developing countries face significant uncertainty with regards to their fu-
ture consumption. There are many sources of this uncertainty, for example job insecurity,
ill health that affects the productive labor and income of a household, death of family
members, drought that reduces crop yields and incomes and, more recently, price shocks
that make purchasing basic food supplies more difficult for those with few resources.
Forward-looking quantitative research on the intersection of poverty and risk is a
relatively new field, with one of the main constraints to analysis a paucity of detailed
information on shocks that are both idiosyncratic and covariate (as pointed out by some
time ago by e.g. Ravallion (1988); Morduch (1994)). The importance of risk for decision
making (Dercon, 1998) and the consequences of realized shocks have been much more
extensively studied in recent years, and has become an ever-more important focus for
policymakers as evidenced by the World Development Reports of 2001 and 2014 (World
Bank, 2014).1
It is to this literature that this paper aims to add, by applying and extending lessons
from both theoretical and empirical work on shocks and vulnerability to a newly com-
piled dataset that is nationally representative, and used to calculate official estimates
of poverty. The methodology we use is of general interest, as it combines several new
ideas for vulnerability analysis using cross-sectional data. Whilst the availability of panel
data surveys is improving for developing countries, they are not yet widespread, often
are small in size, and not nationally representative. However, economists have recently
incorporated rainfall data as an exogenous source of shocks in various empirical models
(Dell et al., 2014), and we incorporate time-series information on rainfall-induced crop
losses to create the covariate shock component of our ex-ante distributions of consump-
tion. Our data also contain a rich set of idiosyncratic shocks. Developing an estimate
of the distribution of consumption expenditure for a household at a future point in time
requires estimating both expected household welfare in the absence of shocks, and the
distribution of shock-induced welfare losses. A particular focus of the methodology em-
ployed in this paper is in identifying an internally consistent and externally valid estimate
of the second component, the distribution of shock-induced welfare losses. When mod-
elling shock impacts, we also incorporate heterogeneity with regard to their impact on
different characteristics of households such as female-headed, urban, educated, access to
social protection. With this information we can calculate existing vulnerability measures
and analyse trends in vulnerability and differences in vulnerability between groups.
There are three reasons why we may consider Ethiopia a pertinent case study for a
focus on vulnerability. First, the Ethiopian economy is still largely comprised of rain-fed
agriculture in an agroclimatic environment that has been no stranger to extreme drought
in recent decades. Although drought has been less prevalent in the last decade than the
previous two, the underlying climate risk remains unchanged and is likely to worsen
with climate change (Robinson et al., 2013). Second, poverty in Ethiopia has fallen
considerably in the last decade. This positive change has resulted in many individuals
1
See Holzmann and Jorgensen (2000) and and Dercon (2002) for conceptual underpinnings of the
WDR 2001.

2
with per capita consumption now just above the national poverty line. These now non-
poor classified individuals are likely still exposed to considerable risk, and without safety
nets to protect them in the event that shocks do occur, could be vulnerable to falling back
into poverty. If Ethiopia continues on its successful trend of poverty reduction, programs
that are put in place to protect these gains will become increasingly important. Three,
a large safety-net program was introduced between the two surveys used in the analysis,
covering rural areas using geographical targeting. We can offer some lessons on the
program impact on vulnerability, and comment on targeting. Thus, our findings are also
relevant to other countries in Africa with large populations of subsistance farmers, and
drought-proneness.
To preview our findings, we show that shocks do have a significant impact on the
consumption of households in Ethiopia, with drought and food price increases being the
most important. We find considerable heterogeneity of impacts, with uneducated male
headed households and all female headed households in the urban areas suffering most
from food price increases. Analysing our various vulnerability measures by gender of
the household head yields surprising results: Whilst female-headed housheolds are more
vulnerable in urban areas, the opposite is true in rural areas. Relatedly, we find that
the rural safety net, the national Productive Safety Net program (PSNP) does appear to
successfully mitigate the impact of drought shocks for rural households covered by the
programme. It may then be that successful targeting is reducing vulnerability in rural
areas.
The paper is structured as follows. Section 2 briefly reviews the literature on shocks
and vulnerability measurement. Section 3 introduces the datasets. Section 4 details
our empirical methodology. In Section 5 we present results on the correlates of poverty,
and vulnerability estimates for 2005 and 2011, with a breakdown by gender and PSNP
status. Section 6 concludes with some policy recommendations.

2 Literature review and conceptual framework

We can understand vulnerability to poverty as conceptually distinct from poverty, due to
the uncertainty that is incorporated into the analysis. Many chronically poor households
are poor due to a lack of assets (including land, able bodied working age labor, good
health) and opportunities. Such households will have possibly not exited poverty in
a long time. Many households who are above the poverty line however, are at risk
of transient poverty, or of falling into chronic poverty in the future. Those below the
poverty line can of course also fall further into poverty. Alwang et al. (2001) reviewed
the concept of vulnerability across several disciplines and found agreement that is a)
forward looking with regard to potential future welfare losses; and b) concerned with the
impact of uncertainty on welfare. The authors also note that the degree of vulnerability
(however defined) will depend on the characteristics of the risk and the household ability
to respond to such risks. In terms of the latter point, an important body of research
has developed understanding of the importance of assets for protecting livelihoods under
stress, notably Moser (1998), and developed in economics through the work of Carter

3
and Barrett (2006) and others.2 The focus in this paper is on on vulnerability to absolute
poverty, rather than to downturns in the wellbeing indicator (Kamanou and Morduch,
2002; Dutta et al., 2011) or vulnerability as welfare loss due to increased variance (Ligon
and Schechter, 2003). We briefly review the most pertinent literature to our study below,
before discussing the Ethiopian context.
The conceptual vulnerability literature builds on the foundations of earlier work that
developed desirable properties of (unidimensional) cross-section quantitative poverty
measures, seminally axiomatized by Sen (1976), and subsequently developed by e.g.
Foster et al. (1984); Foster and Shorrocks (1991). The focus of empirical poverty mea-
surement is usually on deprivation in the space of household income or consumption,
suitably spatially deflated and adjusted for household composition.3 There is by now
somewhat of a consensus in this static poverty literature, and focus has recently turned to
the limitations of the unidimensional approach, and how to incorporate multiple dimen-
sions (such as health, education, empowerment) into a measure of poverty (Bourguignon
and Chakravarty, 2003; Alkire and Foster, 2011) as well as concepts of chronic poverty
or the dimension of duration (Bossert et al., 2012; Foster and Santos, 2013; Porter and
Quinn, 2013), as well as that of uncertainty or vulnerability to future poverty, which we
now discuss in more detail.
The empirical literature in development economics first began attempting to incorpo-
rate uncertainty into poverty in the late 1980’s. For example, Ravallion (1988) presents
an early discussion of vulnerability, conceptualized as the poverty impact of consumption
variability, and shows that if households’ consumption (in rural India) were stabilized
at their mean over time, then experienced poverty would be lower. In effect, Ravallion
estimates the Expected Value of the squared poverty gap (Foster et al., 1984). Poverty
analysts implementing vulnerability measurement as panel data became available noted
that incorporating both the mean and variance of consumption (poverty) gives a more
realistic view of the experiences of poor people in developing countries, given the high
degree of mobility between poverty and non-poverty over time (for a review of these
issues, see Baulch and Hoddinott (2000)).
The first applied vulnerability measures that calculated expected poverty and prob-
ability of future poverty were developed, see Pritchett et al. (2000); Chaudhuri (2003)
and Hoddinott and Quisumbing (2003). Empirically, poverty can be understood as the
realization of a stochastic process that generates consumption outcomes. If realized con-
sumption falls below a poverty line z, then the household is classified ex-post as a poor
household. Vulnerability measurement, being ex-ante by definition, thus needs a poverty
measure, a poverty line, and a prediction of future consumption based on assumptions
about the stochastic process generating consumption. Poverty is defined in the above
papers using the FGT family of poverty indices (Foster et al., 1984):
  α
z − cht
Pα = max 0, (1)
z
2
See e.g. Échevin (2013) for a recent economic application on vulnerability to asset poverty.
3
See Deaton and Grosh (2000) for a discussion of the issues around choosing and measuring a welfare
indicator in the field.

4
 With α = 0 being the headcount, 1 the poverty gap and 2 the squared gap. Chaudhuri
et al. (2002); Chaudhuri (2003) therefore defined vulnerability as the expectation (i.e.
the predicted value) of poverty:
  α 
z − cht
V [Pα ] = E max 0, (2)
z
Pritchett et al. (2000) and Chaudhuri (2003) further extend the definition of vulner-
ability as a high probability that the future consumption falls below the poverty line:

Vh,t = P r (ch,t+1 ) ≤ z (3)

In the binary measure, vulnerable households are defined as those with probability
of consumption falling below the poverty line being greater than 0.5. The advantage of
this measure is that is very intuitive, and policy-relevant. However, it comes with the
disadvantages that are associated with the headcount measure. that have been raised in
the literature on theoretically extending the concept of poverty measurement to a more
dynamic concept of vulnerability to poverty. It has been described as incorporating the
utility loss associated with poverty and uncertainty (Ligon and Schechter, 2003), or the
threat of future poverty (Calvo and Dercon, 2013). Calvo and Dercon (2013) take an
approach that they argue is a middle-ground between vulnerability as expected poverty
and vulnerability as utility loss of increased risk. The authors propose a set of axioms
extending those from static poverty measurement, that we summarize as focusing on the
level of welfare in a particular realized state, and the probability of that state occuring,
wherein if the probability of a lower welfare outcome increases, vulnerability should not
fall. The authors offer a measure that depends on a constant relative risk sensitive
(CRRS) measure, with risk sensitivity captured in a flexible parameter, θ:
1
1 − E[xθ ]


V (z, p, c) = (4)
θ
Where x = c/z, i.e. consumption relative to the poverty line z; p are the set of proba-
bilities for the realised states of consumption in the expectation term.
Calvo and Dercon’s “No compensation” (p730) measures incorporate the idea that
for vulnerability purposes, an improvement in the best scenario should not compensate
for bad (i.e. below the poverty line) scenarios. For example, the authors argue that
whilst an increase in the prize money of a very low-probability win in a lottery would
increase the expected value of consumption, it should not decrease any conceptually
desirable measure of vulnerability. They incorporate a censored variable e.g. c̃ into their
measures in 4. This is defined as being is equal to c when c < z, a pre-defined poverty
but is top-censored at the poverty line.4 Given their policy relevance, we present the
expected poverty measures, as well as the Dercon-Calvo no-compensation measures in
our empirical work.
4
Note here that the Expected Poverty measures outlined in equation 2 above do also incorporate this
censoring (e.g. the poverty gap is calculated, and then summed - therefore it is censored at the poverty
line in each period).

5
 We acknowledge that other perspectives have also been applied, for example, Dutta
et al. (2011) and Günther and Maier (2014) consider that a desirable property for a
vulnerability measure is to reference it to the starting point of household welfare, e.g.
that people who begin with higher levels of consumption in period one will experience
more deprivation from experiencing low levels of consumption in period two, than those
who start at lower levels. This is a normative choice, and we prefer in the context of
Ethiopia to focus on the expected FGT measures that are the vulnerability analogues of
the static poverty measures used in national policy, combined with those more sensitive
to the threat of extreme poverty.

2.1 Empirical estimation on vulnerability measurement

A practical reference on vulnerability to consumption poverty was compiled by Hoddinott
and Quisumbing (2003), and methodology in vulnerability measurement has mainly ad-
vanced since then in conceptual terms such as the research described above.5 After
defining the wellbeing metric (e.g. consumption poverty), there are two further practical
components to vulnerability analysis (Hoddinott and Quisumbing, 2003): calculating a)
the likelihood of adverse events occurring, and b) the impact of an event on wellbeing,
should the shock materialize. Measuring vulnerability is conceptually and empirically
much more complex than measuring poverty because it is about what might happen, an
expectation on the future. We therefore require information on possible shocks and the
likelihood that they might occur.
As noted above, Chaudhuri et al. (2002) define vulnerability as the expected value of
poverty, defined as the headcount index (2. We now outline in some more detail how the
empirical analysis proceeds, beginning with the empirical analogue of equation 2 below:

Vα,h,t = E [pα,h,t+1 (ch,t+1 ) | F (ch,t+1 )] (5)

Where F (ch,t+1 ) is the cumulative distribution function (CDF) of consumption in the

future time period.
It is easily seen that the expected poverty-based measures are all interlinked and the
essential underpinning of the measure is the empirical estimation of the underlying data
generating process for future consumption ch,t+1 . The CDF of ch,t+1 as noted above is
a function of household assets, the risk environment and the interactions between the
two. However for the earliest empirical work, idiosyncratic shocks are unobserved, and
they rely on the cross-sectional distribution of consumption to impute the prospective
distribution for each household. Christiaensen and Subbarao (2005) also augment the
model by including an aggregate shock (rainfall), as well as a heteroskedastic variance
of consumption (i.e. variance is explicitly modelled as depending on household and
community characteristics).
5
There has been some further progress, but not necessarily useful for our purposes, e.g. Günther
and Harttgen (2009). The authors use multilevel modelling but their statistical augmentation of the
Chaudhuri-style model are developed for the case where (complete) information on shocks is not avail-
able. The Ethiopia data are imperfect in many ways, but shocks information is quite extensive.

6
 Kamanou and Morduch (2002) define vulnerability slightly differently at the popula-
tion level as the “difference between the expected value of a poverty measure in future,
and the current value” (p15). The expectation is taken in the usual way, multiplying
the probability of a state of the world with the level of consumption materializing in
that state, the shortfall from the poverty line being the poverty outcome for that state.
Rather than using the normality assumption for the error term, the potential states of
the world are generated using Monte-Carlo design (bootstrapping techniques), drawing
e.g. 1000 independent samples from the data, and predicting the change in consumption
based on the regression of household characteristics, observable shocks and estimated
residuals. Chaudhuri (2003) notes that the main drawback in this approach that the
method implicitly assumes that shocks to consumption experienced by different house-
holds are drawn from the same distribution.(p16). We outline in the methods section
below our contribution, which is to extend the method to incorporate heterogeneous
shock probability and impacts.
Ravallion (1988) and Kamanou and Morduch (2002) seek to estimate the expected
value of the squared poverty gap; in this case, poor households are implicitly assumed
to have increasing absolute risk aversion.
The methodology that we use incorporates some of the features of vulnerability mea-
sures outlined above, and we outline the details in section 4 below, once we have intro-
duced the data used.

2.2 Ethiopian Context and related Literature

Vulnerability to weather risk has historically been very high among agricultural subsis-
tance farmers, which describes the majority of Ethiopian rural households. Almost half
of rural households in Ethiopia were affected by drought in a five year period from 1999
to 2004 (Dercon et al., 2005), and drought has been found to have a significant impact
on the welfare of these households. Ten years after experiencing drought, the consump-
tion levels of those reporting a serious drought were found to be 16 percent lower than
those of the families not affected, and these households still experienced lower growth
rates (Dercon, 2004). The widespread predominance of livelihoods that are dependent
on rainfed production systems has characterized discussions of vulnerability and has
resulted in a widespread understanding of a geographic footprint of vulnerability. Un-
til recently the Government of Ethiopia framed rural policy discussions around “three
Ethiopias”: drought-prone highlands, moisture-reliable highlands and pastoral lowlands.
This classification was recently expanded to a concept of “five Ethiopias” according to
agricultural productivity and agroecological conditions (EDRI 2009). The five areas are
drought prone highlands, moisture-reliable cereals areas, moisture-reliable enset6 areas,
humid moisture-reliable lowlands and pastoral areas. The impact of shocks on consump-
tion outcomes for households has been analysed for a small sample of villages in rural
Ethiopia by amongst others Dercon and Krishnan (2000); Dercon (2004); Porter (2012),
and in urban areas using a household panel by Alem and Söderbom (2012). Dercon
6
Enset also known as “false banana” is a very reliable but low-value crop grown mainly in the
lowlands of Ethiopia.

7
(2005) and accompanying articles provide an overview of the importance of the topic,
and the need for further research into the importance of risk for vulnerability, as well
as the need for research on the impact of policies on vulnerability. As poverty becomes
more urban, sudden food price increases have also become more of a challenge for net
consumers of food (Alem and Söderbom, 2012; Headey et al., 2012).
In 2005, the Government of Ethiopia implemented of a Food Security Programme,
which included a Productive Safety Net Programme (PSNP) as well as complementary
programming to strengthen local livelihoods.7 The PNSP was put in place in an attempt
to transition food insecure households from an emergency food aid system to a more
stable and predictable safety net. As such the scheme tackles chronic poverty, but
also aims to protect beneficiaries and non-beneficiaries against drought. In recent years,
drought in Ethiopia has been increasingly well managed, in part as a result of the presence
of the PSNP.
In addition, successive years of good harvests and economic growth have not only had
an impact on poverty, but they may also have had an impact on the ability of households
to respond to shocks in the future that might hit them. Asset ownership in rural areas
was substantially higher in 2011 than 10-15 years previous . As a result, although a
significant proportion of Ethiopians report experiencing shocks to well-being in 2011,
this number has fallen since 2005, particularly in rural areas (Woldehanna and Porter,
2013). However, Devarajan et al. (2013) show that a severe drought in Ethiopia as well
as in several other countries would cause an immediate reduction in consumption at the
macro level of six percentage points.
The geographical nature of vulnerability in Ethiopia has underpinned the develop-
ment of safety net interventions. The PSNP is targeted to address vulnerability in the
most food insecure districts in Ethiopia. Interventions in less food insecure woredas have
typically not focused on providing a safety net but instead on how to improve agricul-
tural productivity and non-farm income earning enterprises. We discuss both the impact
of PSNP on vulnerability, and compare levels of vulnerability across geographic regions,
and between PSNP and non-PSNP areas in our results section below.

3 Data
Data requirements are intense for vulnerability analysis. The perfect dataset would in-
clude several years worth of observations for each household, or even better, information
on all potential outcomes and how probable these are in differing states of the world.
The base datasets for estimating the determinants of poverty are the nationally rep-
resentative Household Income Consumption Expenditure Survey (HICES) and Welfare
Monitoring Surveys (WMS) of 2011 and 2005. These contain information on just un-
der 25,000 households in each year. The key information recorded in the HICE used
7
The program was designed and financed with the support of Development Partners such as the
multilateral agencies World Bank, World Food Program, European Union and bilateral partners USAID,
the UK Department for International Development, Irish Aid the Canadian International Development
Agency and the Swedish International Development Agency.

8
to calculate vulnerability is expenditure on food and other items. The WMS records
household assets and characteristics as well as a fairly detailed module on self-reported
adverse events (referred to as shocks throughout). In both years they were administered
by Ethiopia’s Central Statistics Agency (CSA). The advantage of using the HICES-WMS
for vulnerability analysis is that they are relatively large, nationally representative, com-
parable across years and allow measures of vulnerability to be estimated at the household
level that can be related back to the official poverty measures calculated by the Govern-
ment of Ethiopia (Woldehanna and Porter, 2013). This is extremely useful in a context
where covariate shocks in particular (as our results show) can impact significantly on
poverty numbers.
The household level data allows us to look at the relative importance of geographic
and household factors in determining vulnerability, and also how vulnerability varies
across certain groups of households. Crop yields are a crucial component of rural liveli-
hoods. In Ethiopia crop yields are very dependent on rainfall and vary considerably by
year, and across regions and livelihood zones. Cross-sectional variation in rainfall and
prices could be used as a proxy for covariate shocks, but we merge the HICES-WMS with
another dataset in order to get the best proxy for the impact of aggregate shocks (likely
the greatest source of vulnerability and PSNP caseload variation). We take advantage of
the fact that in Ethiopia, much analysis has been done by policymakers on variability of
rainfall, and crop production. The Livelihoods, Early Assessment and Protection project
(LEAP) system, developed in 2006 by the Government of Ethiopia in collaboration with
the World Food Program (Hoefsloot and Calmanti, 2012), uses crop-modeling approaches
to estimate the likely rainfall-induced crop loss in woredas throughout Ethiopia based on
a water balance model and crop types.8 The LEAP database contains crop loss estimates
from 1996 to 2012, allowing us to estimate the distribution of crop-loss in each zone of
the country. We then use this distribution to predict the crop-losses that households in
this zone may experience in the future. However, recall that we only have consumption
data for 2011 and 2005 therefore the consumption model can calculate the impact of the
shock based only on information for 2005 and 2011.
In Figure 1 we examine what type of years 2005 and 2011 were relative to the distri-
bution over time.
8
See http://www.wfp.org/disaster-risk-reduction/leap, last accessed 21st October 2014. The
LEAP model also provides corresponding estimates of the number of households in need of assistance
based on a household economy model which can be seen as an alternative to our consumption based
approach.

9
 Figure 1: Meher and Belg season crop losses

Notes: The graphs show the proportion of woredas in each season, in each year, for crop losses of
between 0 and 100%, based on LEAP data. E.g. in the top left graph, we see almost 20% of woredas
experience no crop loss, and a very few woredas that experience crop loss of greater than 20% during
the Meher season. Compare this to Meher 2002 graph, directly below, which shows just under 10% that
experience no crop loss, but a much higher proportion of woreda that have crop loss greater than 20%.

In the top panel of Figure 1 the proportion of crop lost in Meher in 2010 and 2005 is
compared to 2002 which is the year in LEAP that records the largest Meher losses. In the
bottom panel of Figure 2 the proportion of crop lost in Belg in 2010 and 2005 is compared
to 2009 which is the year in LEAP that records the largest Belg losses. The values show
the proportion of crop lost, e.g. a LEAP crop value of 60 means that 60% of the crop
was lost in that year (due to rainfall deficit, pests, flood etc). We can see that in 2010,
few if any households experienced Meher crop losses of more than around 30%, whereas
in 2002, a considerable proportion did. There were some zones in 2005 that experienced
high crop losses. However, overall we can see that the years of data that are available
for the analysis do not perhaps capture the very worst that can happen, rather, we are
able to estimate the impact of moderate (yet more frequent) drought shocks. Below we
also calculate some out-of-sample estimations for severe drought shocks, though these
should be considered as more tentative.
Table 1 shows that drought and crop damage are the most prevalent shocks in 2005.
Food price shocks are clearly reported far more in 2011. The crop loss figures from LEAP
show the proportion of crops lost in each community, weighted by crop type (where zero

10
 Table 1: Frequency of shocks

Proportion of households reporting shock:

2005 2011
Food price 0.02 0.19
Drought 0.10 0.05
Crop damage 0.09 0.03
Job loss 0.01 0.00
% crop loss (from LEAP) 23.49 13.82

means no crop losses suffered). The proportion of households experiencing a shock has
fallen from 61% in 2005 to 55% of households in 2011. A general comment is that 2011
was a better year for crops than 2005 (and better than a 15 year average). Conversely,
2010/2011 was a bad year for food price inflation (as were 2008 and 2009, see figure 1).
Inflation in 2005 was not unduly high, however the crop yield was lower than average.
Compared to five years ago, the likelihood of reporting a shock has fallen across many
regions in the rural areas, with the exception of Somali and SNNP regions. However,
for urban households the number reporting shocks has increased from 41% to 50%. This
is almost entirely driven by food price shocks. Table 2 shows the mean and standard
deviations of the main variables used in the analysis.

11
 Table 2: Summary statistics

Variable Mean Std. Dev.

Consumption per adult (1996 prices) 2774.449 5871.671
Ln consumption per adult (1996 prices) 7.459 0.631
Household suffered death of member 0.045 0.207
Household suffered job loss of member 0.015 0.12
HH suffered food price shock 0.112 0.315
Rural HH 0.408 0.492
Proportion lost crops 10.088 18.4
Income from PSNP 0.05 0.218
HH owns more than 3 plots 0.27 0.444
HH can raise 200 birr in a week 0.699 0.459
Dummy: female uneducated hh head 0.198 0.399
Dummy: female uneducated hh head 0.267 0.443
Age of household head 42.662 15.596
femaleheaded hh 0.319 0.466
Highest grade completed by household head 0.835 1.127
HH sector of occupation: agriculture 0.194 0.276
HH sector of occupation: manufacturing 0.035 0.132
HH sector of occupation: construction 0.02 0.102
HH sector of occupation: mining/energy 0.005 0.049
HH sector of occupation: education, health, social services 0.059 0.174
HH sector of occupation: professional services (pub or private) 0.034 0.134
HH sector of occupation: services & trade 0.113 0.222
HH head is unemployed 0.015 0.12
Logarithm of household size 1.334 0.612
Proportion of females 16-64 0.324 0.234
Proportion of females under 15 0.18 0.191
Proportion of females over 65 0.025 0.119
Proportion of males under 15 0.181 0.194
Proportion of males over 65 0.018 0.083
HH dependency ratio 0.398 0.265
Highest grade of any household member 6.431 4.939
Highest grade of any male household member 5.686 4.97
Highest grade of any female household member 4.631 4.813
Ln Distance to market 3.092 1.342
Close to market (less than median minutes) 0.726 0.446
Distance to town 255.375 259.087
Times woreda experienced over 50% crop loss (past 15 years) 2.901 4.623
Household owns land 0.636 0.481
household has own flush toilet/pit latrine 0.337 0.473
household uses flush toilet/pit latrine 0.651 0.477
HH has corrugated iron roof 0.639 0.48
HH has electricity 0.312 0.463
Dummy: urban household that rents 0.269 0.444
12
N 40736
4 Methodology
4.1 Estimating the impact of shocks on consumption
Developing an estimate of the distribution of consumption expenditure for a household
at a future point in time, i.e. the CDF of ch,t+1 as developed in equation 5 above,
requires estimating (i) expected household welfare in the absence of shocks, and (ii)
the distribution of shock-induced welfare losses. A particular focus of the methodology
employed in this paper is in identifying an internally consistent and externally valid
estimate of the second component, the distribution of shock-induced welfare losses.
The welfare measure used is that used by the government of Ethiopia in estimating
poverty, total expenditure on food and non-food consumption per adult (Woldehanna
and Porter, 2013). A proposed empirical specification for consumption per adult is9

ln Cijt = Xijt β + Sijt δ + ijt (6)

where X a vector comprising household and woreda (community) characteristics;

and S is a vector comprising both idiosyncratic shocks (experienced by an individual
household, or an individual within a household), and covariate or aggregate shocks (which
affect a whole village, region, or possibly even the whole country). The setting in which
the household operates (including structural change, as well as the economic, legal and
political environment) determines how shocks impact on income, as well as on household
coping ability and how income/asset shocks translate into consumption changes. Xijt β
provides an estimate of the first component whilst Sijt δ is the first step in generating an
estimate of the second component.
Although equation 6 is a useful starting point, our prefered estimation is one which
allows shocks to impact different households differently. We thus estimate a heteroge-
neous model, in which we allow the impact of S to depend on the characteristics of the
household, e.g. where they live, what options they have to manage risk and shocks.
Our basic specification 6 is augmented with interaction terms for selected household
characteristics as follows:
0
ln Cijt = Xijt β + Sijt δ + Sijt ∗ Xijt γ + ijt (7)

A rich set of household and community characteristics is included in Xijt , however

with cross-sectional data there are still likely to be household characteristics that are
important for determing household welfare and are unobserved. This is a weakness of
our approach. Panel data would allow for the inclusion of fixed effects which would
ensure these characteristics are included. Christiaensen and Subbarao (2005) construct
a pseudo-panel of community-level data using the two cross-section datasets in order to
include community fixed effects. In our dataset, the communities chosen are not the same
in both rounds. Also, our drought shock variable is at community level which precludes
the inclusion of community fixed effects. A zone (equivalent of a district) fixed-effect is
included instead.
9
We use similar notation to Christiaensen and Subbarao (2005) for comparability.

13
 One of the concerns with omitting important household characteristics is that it
might bias the estimation of δ (and also γ in equation 7). To the extent that the propen-
sity to experience a shock is correlated with those characteristics left unobserved, then
our estimates could suffer from omitted variable bias. We rely on the following obser-
vation to address this challenge: although variation in the probability distribution of
Sijt cannot be considered exogenous to welfare across households, the timing of a shock
conditional on its distribution is exogenous (Thomas et al., 2010; Anttila-Hughes and
Hsiang, 2013). Variables that reflect the likelihood of a shock occuring, including mo-
ments of the location-specific historical distribution for covariate shocks are thus included
in Xijt . Migration in Ethiopia has been historically much lower than in other countries,
so bias from migration is less likely. To also allow for non-linearities in the relationship
between Xijt and lnCijt we test the robustness of our findings on the impact of the most
quantitatively important shock (drought) by using propensity score matching.
We estimate equation 7 pooled for 2005 and 2011, where X, the full set of household
characteristics includes education (of the household head, as well as highest education
level of the household), assets, indicators of employment, age, dependency ratio, disabil-
ity, gender, etc, the woreda10 (community) characteristics such as livelihood zone, access
to market, probability of drought occurance in the past, and urban/rural location (which
proxies for exposure to price risk).
Data was collected on a rich set of self-reported idiosyncratic shocks experienced in
the last 12 months. Here, we include job loss, illness and death of household members.
There is also a self-reported food price shock where households are asked if they have
been adversely affected by high consumer prices in the past year. Although inclusion
of data on self-reported shocks is an improvement over some of the earlier vulnerability
studies, there are concerns on the reliability of self-reported shock data. Indivdiuals for
whom the shock had a large negative impact are more likely to report its occurance than
individuals who could mitigate the effects, thus calling into question the assumption of
exogeneity. This is an assumption on which internally valid estimation of δ and γ relies.
Using objective measures of shocks helps overcome this challenge.
For aggregate shocks, which in the Ethiopian context are presumed to contribute
in larger part to vulnerability than idiosyncratic shocks, we use objective shock data
reported outside of the household survey and matched to households by location. We
use CPI data interacted with an urban dummy to calculate who experienced a price
shock. For drought data, we rely on a weighted-average of satellite rainfall data for each
woreda provided by LEAP. The weights in this average are provided by crop models and
allow an estimation of the percentage of crop-loss that has occured as a result of rainfall
falling below necessary levels at key points of the growing season.
Controlling for the probability of occurance and using objective measures of shocks
improves the internal consistency of estimates of δ and γ for drought and prices. However
in order to estimate the CDF using household survey data for two years, it is also
necessary to predict out of sample and this is discussed in the following sub-section.
10
The Woreda is the lowest administrative level in Ethiopia, equivalent to a county.

14
4.2 Estimating the distribution of shock-induced consumption
losses
The method used to generate the distribution of shock induced consumption losses varies
for idiosyncratic and covariate shocks. For the idiosyncratic shocks of death, illness and
job loss, we assume that the cross-sectional distribution of the shock for a group of
households with similar characteristics represents the probability distribution of this
shock occuring in a future period. Household characteristics such as assets, schooling
and gender of the household head, plus land, cattle, highland and pastoral in the rural
areas are used to group households and estimate the probability of the shock occuring
(given all of the idiosycratic shocks were defined as binary events).11 For drought shocks,
the 16 year historic distribution of crop losses due to inadequate rainfall for each woreda
using the LEAP dataset is used. For price shocks, the number of times large increases
in prices observed in the last 15 years is used to calculate the probability of future food
price shocks (again this was defined as a binary event).
These household-specific probability distributions are used to draw (1000) sample
realizations of each shock. The impact of the shock in each scenario is the realized shock
drawn at random, multiplied by the impact of the shocks, δ and γ as estimated from
the regression. We also incorporate a random error for each draw based on the empirical
distribution of the regression residuals.
Whilst this method is appropriate for shocks observed in the data, recall that the
drought shock included in the regression analysis is a continuous measure, reflecting the
proportion of normal crop yields lost due to insufficient rainfall, and that the two years of
household survey data capture the moderate but not severe droughts experienced during
the 16 years. Developing an accurate distribution of drought-induced consumption losses
thus requires out-of-sample predictions on the magnitude of the loss.
For out-of-sample large drought shocks, we use other empirical evidence to guide the
simulation exercise. Using the Ethiopia Rural Household Survey, Porter (2012) found
that more extreme shocks impact consumption to a far greater extent than lesser shocks.
Rainfall in the bottom quintile of the 30 year village distribution caused up to 20% drop
in household consumption, whereas for the next quintile of the rainfall distribution (i.e.
less than average rainfall, but to a lesser extent), the impact was around 2% (and non-
significant). Therefore in the simulations we also allow crop-loss to have a non-linear
impact on consumption, increasing ten-fold for shocks that are more than two standard
deviations above the long-run mean.
The baseline consumption from the “fixed effects” or household and community char-
acteristics is given by Xijt β̂. Together, with the estimates of shock-induced consumption
losses this allows us to calculate consumption for different states of the world for each
household.
11
There is a trade-off between parsimony versus comprehensiveness, see annex table for full list of
interactions.

15
4.3 Estimating vulnerability
From the results of the simulation, we can create various vulnerability measures from
the expected levels of consumption in the predicted scenarios.
We first define all those with a predicted consumption lower than the national poverty
line as poor. The national absolute poverty numbers that are used by the Government
of Ethiopia (Woldehanna and Porter, 2013) are based on the Foster-Greer-Thorbecke
(FGT) family of poverty indices (Foster et al., 1984). A household is defined as poor if
their total expenditure on all items is less than the national absolute poverty line of 3781
Birr in December 2010 prices. This is the amount of money needed to purchase food of
2200 kilocalories for every adult-equivalent in the household, and other (extremely basic,
necessary) items for everyone in the household.
We next define as “vulnerable”, anyone with a probability of being poor higher than
0.5 across the 1000 states of the world simulated (Pritchett et al., 2000; Chaudhuri,
2003). We also calculate the expected poverty gap and squared expected poverty gap
(equation 2, with α = 0, 1, 2; Chaudhuri (2003); Christiaensen and Subbarao (2005))
to show vulnerability to extreme poverty. Finally the Calvo-Dercon “no-compensation”
vulnerability measure (Calvo and Dercon, 2013), with the θ risk exposure reaction pa-
rameter set to 2, is also calculated.12

5 Results
In this section we outline what types of shocks households are most vulnerable to. We
do this by analyzing which shocks are a) the most prevalent and b) have the greatest
impact on consumption and poverty. We then explore in more depth the impact of the
food price crisis and crop loss.
As discussed in the methodology section, the crucial components of vulnerability
analysis are the likelihood of shocks occurring, and their impact on wellbeing, if they do
materialize. Table 1 showed how frequently shocks occured in the data, with the most
prevalent shocks being adverse food prices for urban areas, and drought for rural areas,
though other idiosyncratic shocks were also fairly common.
The impact of these shocks is also quite substantial, whilst a moderate drought
results in a 3% reduction in total consumption, severe drought results in 33.3% loss in
total consumption.
Table 3 shows the impact of shocks on consumption, which are extracted from the
model, and allow heterogeneous impacts. To estimate the impact of crop losses, as
outlined above we divide households into drought prone and non drought prone areas,
and then high or low land (as measured by number of plots). For those with little
land, we compare households with access to the safety net, to those without. Each
increase in crop losses of 10% is correlated with a 3% fall in consumption for drought-
prone households, without PSNP. Interestingly, land does not mitigate the impact of the
drought, though the safety-net appears to (noting that drought was not so severe in 2011
12
In robustness checks, we vary the value of θ parameter

16
 Table 3: Heterogeneous impact of shocks
(1)
Ln consumption per adult (1996 prices)
Food price shock: rural hh -0.022
(-1.19)

Food price shock: urban hh head educated 0.003

(0.14)

Food price shock: urban hh head uneducated female -0.104∗∗

(-2.76)

Food price shock: urban hh head uneducated male -0.130∗∗

(-2.92)

Drought: non drought prone community -0.001

(-0.60)

Drought: drought prone, hh many plots -0.003∗∗

(-2.82)

Drought: drought prone, hh few plots with PSNP -0.002

(-1.57)

Drought: drought prone, hh few plots no PSNP -0.003∗∗

(-2.75)
Death shock: rural hh many plots -0.008
(-0.39)

Death shock: rural hh few plots 0.003

(0.09)

Death shock: urban hh head educated -0.119∗∗∗

(-3.90)

Death shock: urban hh head uneducated female -0.083

(-1.88)

Death shock: urban hh head uneducated male -0.077

(-1.83)

Job loss shock: rural hh -0.067

(-0.94)

Job loss shock: urban hh head educated 0.015

(0.35)

Job loss shock: urban hh head uneducated female 0.154

(1.26)

Job loss shock: urban hh head uneducated male 0.213

(1.12)
Observations 40410
t statistics in parentheses
∗
p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

when the PSNP was in operation). Death and illness in the household have a slightly
lesser impact. A food price shock reduces consumption by 10-13% for those affected
who are in households with an uneducated household head in an urban area. Educated
households in urban areas, and rural households are not impacted by the shock. The
impacts of the shocks can then also be compared with their frequency to understand
the contribution of these shocks to poverty overall. Job loss does not appear to have
an impact on consumption if it happens, this may be due to the fact that relatively few
people are in employment, and few report such a shock, therefore this is unlikely to be
a priority area for targeting. However drought for rural areas not covered by the safety
net, and food prices for urban areas are clearly negatively impacting consumption.
In table 4 we show the results from the same model, for the “fixed” elements of
households and communities.
As a robustness check for the drought shock estimates, we use propensity score match-

17
 Table 4: Regression results: household and locality characteristics
(1)
Ln consumption per adult (1996 prices)
Age of household head -0.00222∗∗∗
(-7.38)

Female headed HH -0.0544∗∗∗

(-3.84)

Dummy: female uneducated hh head -0.207∗∗∗

(-10.20)

Dummy: male uneducated hh head -0.193∗∗∗

(-11.02)

Head primary edu -0.153∗∗∗

(-10.75)

Head nonformal edu -0.148∗∗∗

(-6.34)

Logarithm of household size -0.466∗∗∗

(-16.90)

Highest grade of any male household member 0.00509∗

(2.41)

Highest grade of any female household member 0.0112∗∗∗

(6.95)

Distance to town 0.0000556

(0.43)

Drought prone area 0.115∗∗

(3.11)

HH received income from PSNP 2010 -0.0500∗

(-2.44)

HH owns more than 3 plots 0.0532∗∗∗

(5.66)

HH has corrugated iron roof 0.0967∗∗∗

(4.18)

Goodroof*rural 0.0140
(0.54)

HH uses flush toilet/pit latrine 0.116∗∗∗

(5.87)

HH uses flush*rural -0.0585∗

(-2.17)

HH has electricity -0.121∗∗∗

(-5.98)

Electricity*rural 0.205∗∗∗
(5.35)

Household owns land 0.0428∗∗∗

(3.54)

Land*rural 0.0349
(1.29)

Rural HH 0.0329
(0.25)

Year 2011 -0.0331∗

(-2.03)
Observations 40410
t statistics in parentheses
∗
p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

18
ing methods13 to calculate the “treatment effect” of a higher than average drought. For
this we created a dummy variable for the ‘treatment’ equal to one if the woreda expe-
rienced crop loss greater than the mean for the sample (13.6% or higher), and matched
households based on their characteristics and risk exposure. The ATT estimate was -.155
(s.e. =.0001). Given the distribution of the crop losses in the sample(32% of hh have
drought shock defined in this way (for those defined as treated, the minimum shock is
13.6%, maximum shock experienced is 91%), the result is consistent with the marginal
effect estimated in the regression model.

5.1 Vulnerability estimates

Using the methodology described in section 4 above, we calculated total vulnerability
to a consumption shortfall, as shown in table 5. In 2011, vulnerability to consumption
poverty was 42%, as measured by the proportion of the population that had a higher
than 50% chance of falling below the poverty line in the next period (column 2). This
shows that vulnerability is higher than poverty in Ethiopia (with a headcount rate of
just under 30% as shown in column 1), as is to be expected.

Table 5: Vulnerability Estimates, 2011

Headcount V(prob) E(Pov Gap) E(Squared Gap) Calvo-Dercon

All 0.294 0.416 0.120 0.053 0.15

Urban Households 0.261 0.256 0.065 0.024 0.077

Rural Households 0.302 0.448 0.131 0.059 0.164
Notes: Survey-weighted proportion of population that have the characteris-
tics. Headcount is the Foster et al. (1984) headcount measure, E(pov gap) and
E(squared gap) are the expected values of the FGT poverty measures respec-
tively. The final column is Calvo and Dercon (2013) measure with parameter
θ set at one-third.

Whilst Ethiopia has experienced increased growth and substantial poverty reduction
over the past 15 years, we also noted earlier that crop conditions were relatively good
throughout Ethiopia in 2010-11. This may seem surprising given this was the onset
of the Horn of Africa drought, but the crisis only affected the pastoral regions of the
country, which are a small proportion of our sample. The relatively good crop conditions
in non-pastoral regions resulted in lower levels of poverty for many households as higher
crop incomes allowed them to consume more. However, our measures of vulnerability
captures the risk of being poor in the future, for example if crop incomes are low.
Although households were less poor in 2010-11 than in previous years, they were still
quite vulnerable to poverty in the future. Columns 3-5 outline some further measures
of vulnerability that are more sensitive to depth of future poverty episodes (rather than
simply the probability of falling below the line) as discussed in section 4. These are
the expected poverty gap, expected squared poverty gap (both as outlined in equation 2
13
teffects psmatch in stata 13

19
above). The final column shows Calvo and Dercon (2013) “no compensation” measure
(equation 4). We compare all the measures between urban and rural households. Whilst
poverty is somewhat higher in rural areas, vulnerability according to all the measures
is substantially higher in rural areas. In particular, the measures that are sensitive to
depth more than double, showing (as expected) that it is rural households that are more
vulnerable to fluctuations in rainfall, drought in particular. Below we explore whether
there are gender differences in the urban and rural vulnerability profiles.

5.2 Gender and Vulnerability

Klasen et al. (2014) examine the correlates of poverty and vulnerability in Thailand and
Vietnam with respect to gender of the household head, and find that in both countries,
female headed households are more vulnerable, though this is driven by different factors:
higher exposure to shocks in Thailand, and higher poverty in Vietnam. The authors
also call for more gender disaggregated poverty and vulnerability analysis as a guide
to policy. In Ethiopia there are still significant gaps between male and female adult
outcomes, especially with regard to education and literacy - in 2011, more than half
of Ethiopian women had no education (compared with 38% of men). We therefore
calculated our various vulnerability measures, disaggregated by gender, for both urban
and rural areas. The results show higher levels of vulnerability for rural areas overall,
consistent with higher poverty rates, however female-headed households show lower levels
of vulnerability as high predicted poverty (40%) than male-headed (45%). This result is
fairly consistent across all of our calculated vulnerability measures. In contrast, female-
headed households show higher vulnerability in the urban areas, for example with 28%
having high predicted poverty, compared to 25% of male-headed households. Referring
back to our regression results in table 3, we found that female headed households and
uneducated male headed households suffered most from food price shocks, so we can
attribute the urban vulnerability quite closely to the different levels of education of
men and women. For the rural results, we see that female headed households do have
slightly lower poverty than male headed households. We also did find that PSNP coverage
mitigated the impact of the shock for households with low land holdings (which describes
most female-headed households), and so PSNP targeting of female headed households
could explain the difference. We examine the impact of the PSNP in more detail in
section 5.4 below.

5.3 Worst case scenario: Vulnerability to experiencing the most

severe crop loss in the last 17 years
Our estimation results find that for these more moderate yield shocks experienced by
households at the time of survey, a 1% increase in crops lost, leads to a 0.15% fall in
consumption. Using the Ethiopia Rural Household Survey, Porter (2012) found that
more extreme shocks impact consumption to a far greater extent than lesser shocks.
Rainfall in the bottom quintile of the 30 year village distribution caused up to 20%
drop in household consumption, whereas for the next quintile of the rainfall distribution

20
 Table 6: Vulnerability by gender of household head, 2011

Headcount V(prob) E(Pov Gap) E(Squared Gap) Calvo-Dercon

Rural Households
Male-head 0.307 0.454 0.132 0.060 0.166
Female-head 0.274 0.405 0.123 0.058 0.157
Urban Households
Male-head 0.243 0.247 0.064 0.024 0.075
Female-head 0.282 0.277 0.070 0.025 0.081
Notes: Survey-weighted proportion of population that have the characteristics.
Other notes as in table 5 above.

(i.e. less than average rainfall, but to a lesser extent), the impact was around 2% (and
non-significant). Therefore in the vulnerability analysis we also allow crop-loss to have a
non-linear impact on consumption, increasing ten-fold for shocks two standard deviations
above the long-run mean. To investigate the impact on poverty of an extreme weather
shock we created scenarios whereby households are faced with the maximum crop loss
in the 17 year distribution of crop losses from the LEAP estimates for their zone. We
project this simulated consumption fall onto the poverty numbers. Experiencing the
maximum crop loss would have almost doubled the headcount poverty rate in 2011,
from 30% to 68%.
We performed several robustness checks on the modeling for the aggregate crop shock
(including a squared term on the shock and a nonlinear impact for shocks above 60%
crop losses) and conclude that results are fairly robust to our choice of specification.

5.4 Discussion of PSNP impacts and targeting

In the above analysis we also investigated whether households with access to PSNP
have been able to protect their consumption from crop losses as a result of the PSNP.
The results of an interaction (allowing a different impact of LEAP-predicted crop loss
for non-PSNP and PSNP households separately) show that the impact of crop-loss was
much larger for households that did not receive PSNP. The impact of the shock in 2011
for PSNP households is not significant and the point estimate is less than 0.02%. If the
impact of crop loss is close to zero for PSNP households, then the above simulated poverty
increase is what is being avoided when the programme effectively covers households who
experience a drought.
As an illustration, table 7 shows the vulnerability estimates for households with access
to PSNP - in this case we restrict the comparison group to other households living in
drought prone areas that also do not have high land holdings. Poverty is higher for
PSNP households, as would be expected due to the targeting criteria. The headcount
measure of vulnerability is very slightly higher for PSNP households, but interestingly
the measures that weight severe potential poverty episodes more highly (expected gap
and Calvo-Dercon) show a reversal. We interpret this as evidence that the PSNP’s
mitigating impact especially of extreme shocks does translate into an overall reduction

21
in vulnerability to extreme poverty.

Table 7: Vulnerability by PSNP status, 2011, Households in drought-prone regions with

little land

Headcount V(prob) E(Pov Gap) E(Squared Gap) Calvo-Dercon

No PSNP 0.289 0.478 0.172 0.094 0.235
Access PSNP 0.313 0.483 0.176 0.082 0.219
Notes: as in table 5

We explored the geographic footprint of vulnerability in Ethiopia by comparing the

level of vulnerability across geographic types (according to altitude and crop reliability).
We find that vulnerability does still have a strong geographic footprint in Ethiopia.
The moisture reliable lowlands are the most vulnerable places in Ethiopian in 2010-11,
followed by the Enset-growing lowlands and the drought-prone highlands. It may at
first seem surprising that the moisture reliable lowlands are the most vulnerable places
in Ethiopia, given that by their definition moisture is indeed more reliable in these areas.
However, poverty, although falling, is still very high in the moisture reliable lowlands.
More than half, 59%, of residents were poor in 2005. This fell substantially to 45% in
2010-11, though was still 13 percentage points higher than the next geographic area.
Vulnerability is high in these lowlands not because residents are more subject to climate
shocks that will drive them into poverty, but because residents are already poor. Lower
levels of education and asset ownership result in our model predicting higher vulnerability
in these areas.
We compared the proportion of vulnerable households in PSNP and non-PSNP
woredas. We did not find a big difference in the proportion of people counted as vulner-
able in PSNP and non-PSNP woredas. Also, the proportion of households measured as
vulnerable in non-PSNP woredas is quite high. For example 37% of households living
in non-PSNP woredas are vulnerable to absolute poverty. Our results do show that the
PSNP is targeted to many of the most vulnerable areas: in particular, the vulnerable
enset lowlands and to the drought prone highlands. However they also point to a need
for safety nets in the moisture reliable lowlands, and at present very few woredas in
these areas are included in the PSNP. Indeed, a safety net program that is targeted to
specific woredas will necessarily result in many vulnerable Ethiopians being left without
protection.

6 Conclusions and policy implications

We used a nationally representative dataset merged with climate and price data to
estimate vulnerability to poverty in Ethiopia. Our numbers can be compared directly to
official poverty numbers, to show how sensitive current poverty estimates are to shocks
that are prevalent in the context. We find that many Ethiopians are unable to protect
their consumption against large covariate shocks, in particular drought and food prices.

22
In particular we find considerable heterogeneity in the impact of all types of shocks,
with all rural households, and female-headed and uneducated households in urban areas
being most at risk to suffer drops in their (already low) consumption. However, much
vulnerability is not geographically determined, but instead determined by other factors
such related to gender and education. We find that female headed-housheolds are more
vulnerable in urban areas, but the situation is reversed in rural areas. With regard
to drought, we do find that the PSNP mitigates the impact quite successfully, almost
offering full protection against at least moderate levels of crop failure. The high impact
of food price shocks could be a crucial consideration for PSNP expansion in future
(which will also link in to payment types; food may be more effective than cash in a high
food-inflation environment).
We show that the results in terms of how the measures create rankings between
groups (e.g. rural female, rural male) are fairly stable across the choice of measure,
though measures that are more sensitive to depth of potential deprivation should also
be carefully checked. For example vulnerability has fallen more than proportionately for
such measures through the introduction of PSNP. The novel methodology has allowed us
to avoid the limitations of cross-section data where possible but caveats to the analysis
remain with regard to unobserved household characteristics. For this type of analysis, the
data requirements are indeed high, and future challenges include generalising estimates of
consumption variability more reliably out-of-sample in order to improve vulnerability es-
timates for forward-looking social protection programmes (Anttila-Hughes and Sharma,
2014).
In terms of policy, although safety nets can help households manage agricultural risk
such as drought, other risk management instruments can also help. Agricultural index
insurance has been piloted in a number of sites in Ethiopia, and research suggests that if
offered through the right institutional set-up, and with some subsidization of marketing
costs, it will be purchased by a number of farmers (Berhane et al., 2012; Dercon et al.,
2014). The World Food Program and Oxfam have also piloted incorporating index
insurance with the PSNP. This has resulted in high demand for insurance among PSNP
households. Further work is also ongoing on using indexes to trigger payments from a
risk pool (such as in the African Risk Capacity14 at a national level in order to support
safety net disbursements. The Global Index Insurance Facility (GIIF) launched the Index
Insurance Forum at the Agricultural Insurance Conference in Berlin in November 2014
to facilitate further policy discussion.15 This work suggests that a number of options
may be available to increase protection for rural households against bad weather.
14
A special agency of the African Union, also funded by development partners. http://www.
africanriskcapacity.org/ last accessed 26th January 2015.
15
Managed by the International Finance Corporation of the World Bank Group.

23
7 References
References
Alem, Y. and M. Söderbom (2012). Household-level consumption in urban ethiopia: The
effects of a large food price shock. World Development 40 (1), 146–162.

Alkire, S. and J. Foster (2011). Counting and multidimensional poverty measurement.

Journal of public economics 95 (7), 476–487.

Alwang, J., P. B. Siegel, and S. L. Jorgensen (2001). Vulnerability: a view from different
disciplines. Technical report, Social protection discussion paper series.

Anttila-Hughes, J. and M. Sharma (2014). Linking risk models to microeconomic indi-

cators. Microeconomics and Catastrophe Risk Modelling Paper 2, World Bank.

Anttila-Hughes, J. K. and S. M. Hsiang (2013). Destruction, disinvestment, and death:

Economic and human losses following environmental disaster. Available at SSRN
2220501 .

Baulch, B. and J. Hoddinott (2000). Economic mobility and poverty dynamics in devel-
oping countries. Journal of Development Studies 36 (6), 1.

Berhane, G., D. Clarke, S. Dercon, R. V. Hill, and A. S. Taffesse (2012). Financial

innovations for social and climate resilience: Ethiopia case study. Research report.

Bossert, W., S. R. Chakravarty, and C. dAmbrosio (2012). Poverty and time. The
Journal of economic inequality 10 (2), 145–162.

Bourguignon, F. and S. R. Chakravarty (2003). The measurement of multidimensional

poverty. The Journal of Economic Inequality 1 (1), 25–49.

Calvo, C. and S. Dercon (2013). Vulnerability to individual and aggregate poverty. Social
Choice and Welfare 41 (4), 721–740.

Carter, M. R. and C. B. Barrett (2006). The economics of poverty traps and persistent
poverty: An asset-based approach. The Journal of Development Studies 42 (2), 178–
199.

Chaudhuri, S. (2003). Assessing vulnerability to poverty: concepts,

empirical methods and illustrative examples. Department of Eco-
nomics Columbia University available at: http://info. worldbank.
org/etools/docs/library/97185/Keny 0304/Ke 0304/vulnerabilityassessment. pdf .

Chaudhuri, S., J. Jalan, and A. Suryahadi (2002). Assessing household vulnerability

to poverty from cross-sectional data: A methodology and estimates from indonesia.
Working paper, Department of Economics, Columbia University.

24
Christiaensen, L. J. and K. Subbarao (2005). Towards an understanding of household
vulnerability in rural kenya. Journal of African Economies 14 (4), 520–558.

Deaton, A. and M. Grosh (2000). Consumption. In M. Grosh and P. Glewwe (Eds.),

Designing Household Survey Questionnaires for Developing Countries: Lessons from
Fifteen Years of LSMS Experience. Washington, DC: Oxford University Press.

Dell, M., B. F. Jones, and B. A. Olken (2014). What do we learn from the weather? the
new climate-economy literature. Journal of Economic Literature 52 (3), 740–98.

Dercon, S. (1998). Wealth, risk and activity choice: cattle in western tanzania. Journal
of Development Economics 55 (1), 1–42.

Dercon, S. (2002). Income risk, coping strategies, and safety nets. The World Bank
Research Observer 17 (2), 141–166.

Dercon, S. (2004). Growth and shocks: evidence from rural ethiopia. Journal of Devel-
opment Economics 74 (2), 309–329.

Dercon, S. (2005). Risk, poverty and vulnerability in africa. Journal of African

Economies 14 (4), 483–488.

Dercon, S., R. V. Hill, D. Clarke, I. Outes-Leon, and A. Seyoum Taffesse (2014). Offering
rainfall insurance to informal insurance groups: evidence from a field experiment in
ethiopia. Journal of Development Economics 106, 132–143.

Dercon, S., J. Hoddinott, and T. Woldehanna (2005). Shocks and consumption in 15

ethiopian villages, 1999-2004. Journal of African economies.

Dercon, S. and P. Krishnan (2000). Vulnerability, seasonality and poverty in ethiopia.

Journal of Development Studies 36 (6), 25.

Devarajan, S., D. S. Go, M. Maliszewska, I. Osorio-Rodarte, and H. Timmer (2013).

Stress-testing africa’s recent growth and poverty performance. Technical report, The
World Bank.

Dutta, I., J. Foster, and A. Mishra (2011). On measuring vulnerability to poverty. Social
Choice and Welfare 37 (4), 743–761.

Échevin, D. (2013). Measuring vulnerability to asset-poverty in sub-saharan africa.

World Development 46, 211–222.

Foster, J., J. Greer, and E. Thorbecke (1984). A class of decomposable poverty measures.
Econometrica: Journal of the Econometric Society, 761–766.

Foster, J. E. and M. E. Santos (2013). Measuring chronic poverty. In G. Betti and

A. Lemmi (Eds.), Poverty and Social Exclusion: New Methods of Analysis, pp. 143.
Routledge.

25
Foster, J. E. and A. F. Shorrocks (1991). Subgroup consistent poverty indices. Econo-
metrica 59 (3), 687–709.

Günther, I. and K. Harttgen (2009). Estimating households vulnerability to idiosyn-

cratic and covariate shocks: a novel method applied in madagascar. World Develop-
ment 37 (7), 1222–1234.

Günther, I. and J. K. Maier (2014). Poverty, vulnerability, and reference-dependent

utility. Review of Income and Wealth 60 (1), 155–181.

Headey, D., F. B. Nisrane, I. Worku, M. Dereje, and A. S. Taffesse (2012). Urban wage
behavior and food price inflation: The case of ethiopia. Ethiopia Strategy Support
Programme II Working Paper 41.

Hoddinott, J. and A. R. Quisumbing (2003). Methods for microeconometric risk and

vulnerability assessments. Technical report, The World Bank.

Hoefsloot, P. and S. Calmanti (2012). Leap version 2.61 for ethiopia user manual. Tech-
nical report, The World Bank.

Holzmann, R. and S. Jorgensen (2000). Social risk management: A new conceptual

framework for social protection and beyond. Technical report, The World Bank.

Kamanou, G. and J. Morduch (2002). Measuring vulnerability to poverty, Vol-

ume 58. WIDER Discussion Papers–World Institute for Development Economics
(UNU-WIDER).

Klasen, S., T. Lechtenfeld, and F. Povel (2014). A feminization of vulnerability? female

headship, poverty, and vulnerability in thailand and vietnam. World Development.

Ligon, E. and L. Schechter (2003). Measuring vulnerability*. The Economic Jour-

nal 113 (486), C95–C102.

Morduch, J. (1994). Poverty and vulnerability. The American Economic Review , 221–
225.

Moser, C. O. (1998). The asset vulnerability framework: reassessing urban poverty

reduction strategies. World development 26 (1), 1–19.

Porter, C. (2012). Shocks, consumption and income diversification in rural ethiopia.

Journal of Development Studies 48 (9), 1209–1222.

Porter, C. and N. N. Quinn (2013). Measuring intertemporal poverty. In G. Betti and

A. Lemmi (Eds.), Poverty and Social Exclusion: New Methods of Analysis, pp. 166.
Routledge.

Pritchett, L., A. Suryahadi, and S. Sumarto (2000). Quantifying vulnerability to poverty:

A proposed measure, applied to indonesia. Technical Report 2437, World Bank.

26
 Ravallion, M. (1988). Expected poverty under risk-induced welfare variability. The
Economic Journal , 1171–1182.

Robinson, S., K. Strzepek, and R. Cervigni (2013). The cost of adapting to climate
change in ethiopia: Sector-wise and macro-economic estimates. Ethiopia Strategy
Support Program II (ESSP) working Paper 53.

Sen, A. (1976). Poverty: An ordinal approach to measurement. Econometrica 44 (2),

219–231.

Thomas, T., L. Christiaensen, Q.-T. Do, and L. D. Trung (2010). Natural disasters
and household welfare: evidence from vietnam. World Bank Policy Research Working
Paper Series.

Woldehanna, T. and C. Porter (2013). Poverty and growth in ethiopia (1995/96 -

2010/11): The poverty reduction legacy of meles zenawi and future challenges. Tech-
nical report, Government of Ethiopia.

World Bank (2014). World development report 2014: Risk and opportunity. Technical
report, World Bank.

27

View publication stats