Embed Size (px)
Citation preview
Interdependence in rainwater management technologies:an analysis of rainwater management adoptionin the Blue Nile Basin
Gebrehaweria Gebregziabher1• Lisa-Maria Rebelo2
•
Simon Langan1
Received: 31 May 2014 / Accepted: 20 March 2015� Springer Science+Business Media Dordrecht 2015
Abstract In the Blue Nile Basin of Ethiopian highlands, rainfall distribution is extremely
uneven both spatially and temporally. Drought frequently results in crop failure, while high
rainfall intensities result in low infiltration and high runoff causing soil erosion and land
degradation. These combined factors contribute to low agricultural productivity and high
levels of food insecurity. Poor land management practices coupled with lack of effective
rainwater management strategies aggravate the situation. Over the past two decades,
however, the Government of Ethiopia has attempted to address many of these issues
through a large-scale implementation of a number of soil and water conservation measures.
Despite the success of interventions, uptake and adoption remains low. The conceptual
framework of this study is based on the premise that farmers are more likely to adopt a
combination of rainwater management technologies as adaptation mechanism against
climate variability and agricultural production constraints. This contrasts the previous work
that typically examined a single technology without considering the interdependence be-
tween technologies. Data used in this study come from household survey in seven wa-
tersheds in the Ethiopian Blue Nile Basin. A multivariate probit model was used to account
for the potential correlation and interdependence of various components of rainwater
management technologies. Our results suggest that rainwater management technologies are
related with each other; hence, any effort to promote the adoption of rainwater manage-
ment technologies has to consider such interdependence of technologies, or failure to do so
may mask the reality that farmers face a set of choices in their adoption decisions.
& Gebrehaweria [email protected]
Lisa-Maria [email protected]
Simon [email protected]
1 International Water Management Institute (IWMI), East Africa and Nile Basin Office,Addis Ababa, Ethiopia
2 International Water Management Institute (IWMI), Southeast Asia Office, Vientiane, Lao PDR
123
Environ Dev SustainDOI 10.1007/s10668-015-9656-8
Keywords Rainwater harvesting � Technology adoption � Interdependence � Multivariate
probit � Blue Nile Basin
1 Introduction
Agriculture contributes to approximately 42 % of the gross domestic product (GDP) of the
Ethiopian economy. It generates more than 85 % of Ethiopia’s foreign exchange earnings
and employs over 80 % of the population (CSA 2004; MoFED 2010). The Government of
Ethiopia has committed to rapid agricultural growth as a means of accelerating economic
growth and reducing poverty. Despite impressive achievements over the last three decades,
Ethiopia remains one of the poorest countries in the world with over 12 million people
being food-insecure. In the Ethiopian highlands, agricultural production is low, charac-
terized by low input use and rainfed mixed crop-livestock production (Merrey and Ge-
breselassie 2011). Rainfall distribution is extremely uneven both spatially and temporally,
which has negative implications for the livelihoods of the population (FAO 2005). Drought
frequently results in crop failure, while high rainfall intensities result in low infiltration and
high runoff causing soil erosion and land degradation, which also contribute to low agri-
cultural productivity and high levels of food insecurity (Lautze et al. 2003; Deressa 2007).
In the rainfed agro-ecological landscapes, low agricultural yield (on average about 35 %
of the potential) is typically not due to the lack of water, but rather is a result of inefficient
management of water, soils and crops (Amede 2012). The gap between actual and at-
tainable yields suggests that there is a large untapped potential for yield increases
(Rockstrom et al. 2010). For example, Awulachew et al. (2012) suggest that access to
Rainwater Management (RWM) interventions can reduce poverty levels by approximately
22 %. RWM technologies can also provide a buffer against production risks associated
with increasing rainfall variability (Kato et al. 2009). The use of RWM interventions,
including soil and water conservation (SWC) techniques, is widely accepted as a key
strategy to improve agricultural productivity by alleviating the effects of drought and
worsening soil conditions (Kurukulasuriya and Rosenthal 2003). Whilst various studies
have highlighted the potential of RWM interventions to increase agricultural productivity
and improve livelihoods in Ethiopia (Pender and Gebremedhin 2007; Kassie et al. 2008;
Awulachew et al. 2010), adoption rates remain low (Santini et al. 2011). In the context of
this paper, rainwater management refers to the use of water resources to meet crop demand
and efforts to optimize productivity and environmental sustainability by managing land
and rainwater resources (van den Born 2011). Rainwater management strategies include
not only rainwater conservation structures (such as gulley rehabilitation), but also the use
of it for food production as part of rainwater management system. RWM generally
comprises small-scale systems that induce, collect, store and make use of local surface
runoff for agriculture.
The Government of Ethiopia along with development agencies has invested substantial
resources to promote RWM practices including conservation structures and crop produc-
tion systems. These have been introduced at a large scale, but with limited success (Ze-
madim et al. 2011). Since the early 1970s, for example, Food for Work (FFW) programs
have been widely implemented as a means of providing much-needed food aid to rural
communities earned by undertaking rural public works. A major component of this has
been the construction of soil and water conservation structures, with the intention of
G. Gebregziabher et al.
123
preventing and reversing erosion processes (Merrey and Gebreselassie 2011). However, the
outcome of these conservation measures was not as expected, and it has been suggested
that interventions should not only focus on the engineering and biophysical performance of
conservation measures but also on the socioeconomic and livelihood benefits (Zemadim
et al. 2011). Studies from the Ethiopian highlands show that the adoption of RWM
technologies are influenced by a variety of factors, including biophysical characteristics
such as topography, slope, soil fertility, rainfall and rainfall variability (Deressa et al.
2009). Moreover, even when technologies are appropriate for a biophysical setting, they
are not always adopted (Guerin 1999; Amsalu and de Graaff 2007) due to a variety of
factors considered by farmers when making an adoption decision (McDonald and Brown
2000; Soule et al. 2000). McDonald and Brown (2000) and Merrey and Gebreselassie
(2011) also argued that externally driven technical solutions are rarely sustained by farmers
unless consideration is given to socioeconomic, cultural and institutional, as well as bio-
physical and technical factors. Similarly, Deressa et al. (2009) demonstrated that various
factors including education, gender, age, farm and non-farm income, wealth, access to
extension and credit, information on climate, farmer-to-farmer extension and number of
relatives (as a proxy for social capital) all influence farmers’ adoption of RWM tech-
nologies. Most of these empirical studies argue that one way of improving productivity and
climate-resilient livelihoods in the Ethiopian highlands is targeting promising technologies
that are suitable to a particular biophysical and socioeconomic context. Santini et al.
(2011), for example, highlighted the need for new models of planning for RWM invest-
ments to accommodate the diversity and complexity of biophysical and socioeconomic
contexts, and by tailoring the interventions to suit the priorities and livelihood strategies of
the rural population. This helps to overcome the limited success and impact of practices
that are often adopted using ‘blanket’ approaches (ILRI and IWMI 2010), and the need to
consider a package of technologies that complement each other rather than individual
interventions.
However, there is no agreement in the literature on factors that encourage or discourage
adoption of specific technologies. Furthermore, relatively little empirical work has been
undertaken to understand factors that affect the adoption of RWM technologies at a wa-
tershed level as a ‘package’ or combination of technologies. For example, (Tesfay 2011);
Nata and Bheemalingeswara (2010); Deressa et al. (2009) and Amha (2006) have focused
on the adoption of individual technologies. The reality is that farmers usually adopt multiple
RWM technologies with overlapping constraints. Kato et al. (2009), for example, high-
lighted that the effectiveness of various SWC technologies in Ethiopia depend on whether
they are used independently or as a package. As the suitability and success of RWM
technologies may depend on the level of interdependence, interventions implemented in-
dependently are likely to be different from those implemented as a package. Hence, ana-
lyses that do not take into account the interdependence between the technologies may
underestimate or over-estimate the adoption and performance of RWM technologies.
The objective of this paper is to investigate and shed light on factors contributing to
farmers’ decision to adopt rainwater management technologies, and to contribute to the
existing literature on RWM with a multivariate analysis of RWM in the Blue Nile Basin. It
also contributes to informed policy making on the adoption of RWM technologies as a
strategy for sustainable agricultural production and climate-resilient livelihoods. We use
primary data to jointly analyze the factors that facilitate or impede the probability of
adopting different RWM technologies and practices, and the interdependence between
them.
Interdependence in rainwater management technologies
123
2 Study area and data
This study was undertaken in the Ethiopian part of Blue Nile Basin within the framework
of the Nile Basin Development Challenge, which aims to improve the livelihoods of the
rural people and build their resilience to climate change through a landscape approach to
rainwater management. Sample watersheds include seven landscapes (Fig. 1). The land-
scapes differ in terms of their development, agro-ecology, biophysical characteristics and
livelihood systems.
In 2012, cross-sectional data were been collected across 654 sample households, se-
lected through a multi-stage stratified random sampling procedure. Table 1 presents these
households by region, district and watershed. Meja, Dapo and Mizewa watersheds were
used as research sites of the Development Challenge Program in the Blue Nile Basin, while
the reset four watersheds, still located in the Blue Nile Basin, were selected as counter-
factuals for comparative analysis.
3 Conceptual framework and estimation methodology
The conceptual framework is based on the premise that farmers are more likely to adopt a
combination of rainwater management technologies than a single technology, in order to
deal with the multidimensional nature of agricultural water constraints that commonly
result in crop failure. These technologies can be adopted simultaneously and/or sequen-
tially as complements, substitutes or supplements.
Alternately, the choice of technology can be partly dependent on previously adopted
technologies. For example, farmers who previously adopted bund/terraces may complement it
with multipurpose trees, while bunds/terraces may enhance infiltration and ground/surface
water recharge leading to adoption of ground or surface water irrigation. These are more likely
to be sequential and suggest that the number of technologies adopted may not be independent,
but path dependent (Cowen and Gunby 1996). Some empirical studies on technology adoption
Fig. 1 Location of selected sample watersheds
G. Gebregziabher et al.
123
decisions assume that farmers consider a bundle of technologies and choose the particular
technology bundle that maximizes expected utility (Moyo and Veeman 2004; Marenya and
Barrett 2007; Nhemachena and Hassan 2007; Yu et al. 2008; Kassie et al. 2009).
In general, farmers are faced with alternative, but correlated, technologies in their
adoption decisions that indicate interdependence or path dependence. Hence, the adoption
decision is inherently multivariate. However, previous studies on rainwater management
and conservation technologies in Ethiopia (such as, Tesfay 2011; Nata and Bheema-
lingeswara 2010; Deressa et al. 2009; Amha 2006; Getnet and MacAlister 2012) assume a
single technology and apply univariate modeling without considering the correlation/in-
terdependence that exists between different technologies. When technologies are corre-
lated, univariate models are inappropriate as they exclude useful information contained in
the interdependence of technologies, because a single technology approach may under-
estimate or over-estimate the influence of factors in the adoption of technologies. In
general, univariate models ignore the potential correlation among unobserved disturbances
in the adoption equations, because farmers may consider some combination of technolo-
gies as complements, substitutes or supplements; hence, failure to capture such interde-
pendence will lead to biased and inconsistent estimates.
In this context, we employ a multivariate probit model (MVP) (Kassie et al. 2012;
Cappellari and Jenkins 2003) as shown in Eqs. (1) and (2).
Y�ht ¼ btX0ht þ eht; t ¼ 1; . . .m and ð1Þ
Y�ht ¼ 1 if Y�ht [ 0 and 0 otherwise ð2Þ
where: T = 1,…m represents the choices of rainwater management technologies. The
assumption is that hth farm household has a latent variable Yht* that captures the choices
associated with the Tth RWM technology.
The estimation is based on the observed binary discrete variables Yht* that indicate
whether or not hth farm household has adopted a particular rainwater management tech-
nology (denoted by 1 for adoption and 0 for non-adoption). The status of adoption is
assumed to be influenced by observed characteristics (Xht), including household charac-
teristics, access to services, markets, social capital (captured by household’s membership
in formal and/or informal social groups), and biophysical characteristics. The unobserved
characteristics are captured by the error term denoted by eht, while bt is a parameter to be
estimated.
Table 1 Number of sample households by watersheds
Region Woreda/District Watershed Number ofsample households
Oromia Jeldu Meja 120
Guder Boke 90
Shambu Laku 90
Diga Dapo 90
Amhara Farta Zefe 90
Fogera Mizewa 101
Gondar Zuria Gumera/Maksegnit 90
Total 671
Interdependence in rainwater management technologies
123
In line with this, we assume that RWM technologies considered in this study are
interdependent, implying that the adoption of one technology is likely to influence
(positively or negatively) the adoption of another technology; hence, the error terms
(eht, t = 1,…, m) in Eq. (1) are distributed as multivariate with a mean of 0 and a variance
of 1, where eht & MVN (0, V). The variance (V) is, therefore, normalized to unity on the
diagonal and correlations as off-diagonal elements in Eq. (3). The nonzero value of the off-
diagonal elements allows for correlation across the error terms of several latent equations,
which represent unobserved characteristics that affect the choice of alternative RWM
technologies (Kassie et al. 2012). The covariance matrix V is given in Eq. (3):
V ¼
1 q12 q13 � � � q1m
q21 1 q23 � � � q2m
q31 q33 1 � � � q3m
..
. ... ..
.1 ..
.
qm1 qm2 qm3 1
2666664
3777775
ð3Þ
In general, the multivariate probit model is a generalization of the probit model that is used to
estimate numerous correlated binary outcomes jointly, where the source of correlation can be
complementarity (positive correlation) and substitutability (negative correlation) between
different technologies (Belderbos et al. 2004). Four RWM technologies (bunds/terraces,
gully rehabilitation, multi-purpose trees and orchards,) were considered in the analysis.
4 Independent variables and hypotheses
The explanatory variables considered in the analysis and their expected effects on the
adoption of rainwater management technologies are discussed below.
4.1 Human capital
In this regard, we considered different household characteristics and family member
composition as a proxy of human capital endowment of the household. Education, age and
gender of family members, as well as family size, are important indicators of human
capital and its influence on the adoption of technologies. For example, households with
more educated members are more likely to access information and understand the merits
and potential drawbacks of the technologies. They are also more able to interpret new
information to make knowledge-based decisions in favor of appropriate/suitable tech-
nologies. On the other hand, households with more educated members may be less likely to
invest in labor-intensive technologies and practices, because they are more likely to earn
higher returns from their labor and capital investment in other activities (Kassie et al. 2013;
Pender and Gebremedhin 2007). Age of household head and members may capture
household’s farming experience and ability to respond to unforeseen events/shocks.
Moreover, older household heads may have an accumulation of capital and respect in their
community, implying greater social capital. On the other hand, age can be associated with
loss of energy and short-term planning horizons, and the reluctance toward new tech-
nologies due to risk aversion behavior. On these account, the effect of age on technology
adoption is ambiguous prior to empirical testing.
Gender is an important factor in terms of access to resources. The general argument is
that women have less access to resources and services, such as land, labor, credit and
G. Gebregziabher et al.
123
education, and are generally discriminated in terms of access to external inputs and in-
formation (De Groote and Coulibaly 1998; Quisumbing et al. 1995). In sub-Saharan
Africa, there are gender-specific constraints that women face, such as less education,
inadequate access to land, and production assets and livestock ownership (Ndiritu et al.
2011). These constraints will clearly have a direct effect on technology adoption (including
RWM technologies), where women are usually less likely to adopt these technologies as
they are resource-demanding and labor-intensive.
4.2 Physical capital ownership
This variable is captured by the number of livestock [Total Livestock Units (TLU)] and
farm size per adult equivalent. The assumption is that households that own more capital are
wealthier, hence, are more likely to take risks associated with the adoption of new tech-
nologies. Moreover, such households are likely to be less financial constrained and are able
to finance the purchase of inputs/technologies. Household expenditure is also considered as
a proxy for income level, and the expected effect of capital on the adoption of rainwater
management technologies is positive. However, since households with relatively large
landholdings may be able to diversify their crops and income sources, they may be less
susceptible to risks and shocks; as such, they may be less interested in investing in RWM
technologies as a coping mechanism.
4.3 Off-farm activity
Economic incentives play an important role in the adoption of RWM technologies.
Households’ access to off-farm employment and alternative sources of income are likely to
influence the adoption of technologies in different ways. For example, those who have
alternative sources of income are more likely to adopt and invest in these technologies. On
the other hand, participation in off-farm activities is likely to divert labor from on-farm
activities and working on RWM technologies, both as a private investment and as col-
lective action. The findings of Deressa et al. (2009) support this hypothesis. Off-farm
activity is captured by the participation of household members in the FFW program defined
as a dummy variable (a value of 1 for participation and zero otherwise).
4.4 Access to credit
Total credit available to the household during the preceding year was used as a proxy of
access to credit, markets and inputs. Access to markets can influence the use of various inputs
as well as access to information and support services. For example, Deressa et al. (2009)
revealed that access to credit has a significant positive impact on the likelihood of using soil
conservation techniques, changing planting dates and using irrigation in the Blue Nile Basin.
4.5 Social capital
Household’s membership in informal social networks (such as, Equib, Edir and Debo)1
was used as proxy for social capital. In Ethiopia, it is common for rural communities to
form informal groups for labor sharing, saving and risk-sharing mechanisms. This can take
1 Equib is an informal saving group, Edir is an informal group formed by members of the community,mainly for self-support, and Debo is an informal labor sharing system.
Interdependence in rainwater management technologies
123
place in the form of friendship or kinship networks, implying that households with a large
number of relatives and wider networks are likely to be more resilient to risk and have
fewer credit constraints; hence, they are more likely to adopt technologies as they are in a
better position to take risks (Fafchamps and Gubert 2007). With limited information and
imperfect markets, social networks can facilitate the exchange of information enabling
farmers to access inputs and overcome credit constraints. Social networks also reduce
transaction costs and increase farmers’ bargaining power, helping them to earn higher
returns when marketing their products, which in turn can affect technology adoption (Lee
2005; Pender and Gebremedhin 2007; Wollni et al. 2010). Moreover, farmers who have
limited contacts with extension agents can be informed about the methods and benefits of
new technologies from their networks, as they share information and learn from each other.
On the other hand, having more relatives may reduce incentives for hard work and induce
inefficiency, because farmers may apply less effort to technologies (Kassie et al. 2013).
The expected effect of the social capital coefficient is, therefore, ambiguous prior to
empirical testing.
4.6 Biophysical characteristics
The suitability of RWM technologies as coping mechanisms may depend on different
biophysical characteristics. For example, rainfall variability and desertification can be used
as proxies for the level of rainfall and moisture availability. Hawando (2000) argues that
soil moisture and temperature play major roles in controlling plant growth and develop-
ment, while large areas in Ethiopia are subjected to recurrent droughts, low and erratic
rainfall, and frequent low available soil moisture. Moreover, the same author documented
that the Ethiopian highlands are characterized by high rainfall variability (more than 30 %
coefficient of variation) and soil erosion is considerable and will continue unless reha-
bilitation measures (such as, bund/terraces and gully rehabilitation) are not implemented.
Hence, farm households may adopt certain RWM technologies to reduce their exposure to
resource depletion and climatic hazards.
To control the effect of biophysical factors on the adoption of RWM technologies, we
used various secondary data including mean Aridity Index (AI) (cgiar-csi webpage), co-
efficient of rainfall variability (CV) (NMA 2007; National Meteorology Agency (NMA),
Ethiopia, dataset and information resources (Ethiopian Metrological agency: webpage)
erosion rate/ton/ha/year (Shiferaw 2011; Haileslassie et al. 2005), average temperature
(NMA 2007; National Meteorology Agency (NMA), Ethiopia, dataset and information
resources (Ethiopian Metrological agency: webpage) and land use characteristics (MoWE
1998) Table 2.
5 Descriptive results
Analysed data indicate that 55, 173, 109, 82, 48 and 277 of the sample households, respec-
tively, have adopted a combination of multi-purpose trees and orchards, multi-purpose trees
and bunds/terraces, multi-purpose trees and gully rehabilitation, orchards and bunds/terraces,
orchards and gully rehabilitation, and bunds/terraces and gully rehabilitation (Table 3).
Multi-purpose trees as RWM strategies are adopted for different purposes, such as for
soil and water conservation, livestock feed, to improve soil fertility or as a source of fuel
wood. About 54, 47 and 46 % of sample households have adopted multi-purpose trees in
Boke, Laku and Dapo watersheds, respectively. On the other hand, gully rehabilitation was
G. Gebregziabher et al.
123
adopted by about 18 % of the sample households, most of them in Mizewa, Maksegnit and
Zefe watersheds, all in Amhara region. About 34 % of sample households in Dapo wa-
tershed have developed orchards (Table 4).
Figure 2 and 3 show that adoption of RWM technologies is likely to be influenced by
landscape characteristics and land degradation. Figure 2, for example, shows that most
households that adopted multi-purpose trees, orchards, bunds/terraces and gully reha-
bilitation were on gentle and steep slope lands.
On the other hand, Fig. 3 shows that in comparison with less degraded land, a higher
proportion of multipurpose trees, orchard and gully rehabilitation were adopted on de-
graded lands, probably indicating that these technologies are used as ex-post land reha-
bilitation and resource conservation mechanisms and that the level of land degradation is
likely to influence the suitability and adoption of RWM technologies.
Table 4 presents definition and summary of dependent and independent variables in-
cluded in the analysis.
Table 2 Biophysical characteristics of study areas/watersheds
Study area/watershed Biophysical characteristics
Mean aridityindex (AI)
Rainfallvariability(CV)
Erosion rate/ton/ha/year
Averagetemperature(�c)
Land use
Meja 0.67 0.854 4.23 23 Agriculture
Boke 0.7 0.228 31.77 23 Agro-pastoral
Laku 1.03 0.328 16.39 22 Agro-pastoral
Dapo 1.01 0.305 13.8 24 Agro-pastoral
Zefe 0.81 0.316 27.5 22 Agriculture
Mizewa 0.64 0.348 19.17 29 Agro-pastoral
Gumera/Maksegnit 0.66 0.298 7.65 27 Agro-pastoral
Source of data http://www.cgiar-csi.org
NMA. (2007) Haileslassieet al. (2005)
NMA.(2007)
MoWE. (1998)
Soil erosion severity classes: 0–10, 10–20, 20–30, 30–45, 45–60, 60–80 and[80 classified as low, moderate,high, very high, severe, very severe and extremely severe, respectively
Aridity Index classes: \0.05, 0.05–0.2, 0.2–0.5, 0.5–0.65 and [0.65 classified as Hyper-arid, Arid, Semi-arid, Dry Sub-humid and Humid, respectively
Table 3 Number of households that adopted rainwater management technologies in watersheds
RWM technology Watersheds Total
Meja Zefe Maksegnit Boke Dapo Laku Mizewa
Bunds/terraces 13 78 78 67 76 70 86 468
Gully rehabilitation 54 74 77 36 18 36 57 352
Multi-purpose trees 22 18 16 53 52 45 15 221
Orchards 1 25 7 2 31 13 10 89
Total 90 195 178 158 177 164 168 1130
Interdependence in rainwater management technologies
123
6 Regression results
Coefficients that capture correlation between RWM technologies (Table 5) indicate a pair-
wise correlation between the error terms in the system of equations of the multivariate
probit model. With the exception of orchards versus gully rehabilitation, all RWM tech-
nologies considered in the analysis are positively correlated and support the premises that
farmers usually adopt a combination of RWM technologies. The error terms are not
Table 4 Definition and descriptive statistics of dependent and independent variables
Variable description Frequency
Yes No
Dependent variables
Bunds/terraces (1 = yes, 0 = no) 468 186
Gully rehabilitation (1 = yes, 0 = no) 352 302
Multi-purpose trees (1 = yes, 0 = no) 221 433
Orchards (1 = yes, 0 = no) 89 565
Mean SD
Independent variables
Age of household head (years) 46.996 15.342
Gender of household head (1 = male, 0 = female) 0.846 0.361
Family size in adult equivalent (number) 4.684 2.073
Household head can read and write (1 = yes, 0 = no) 0.200 0.400
Number household members with elementary (1–8) education level 1.976 1.583
Number household members with high school and above (C9) education level 0.792 1.222
Participation in off-farm activities (1 = yes, 0 = no) 0.276 0.447
Household’s livestock holding in TLU (number) 5.234 4.612
Landholding per adult equivalent (ha) 0.428 0.399
Participation in Debo (1 = yes, 0 = no) 0.890 0.313
Participation in Equib (1 = yes, 0 = no) 0.125 0.331
Participation in Edir (1 = yes, 0 = no) 0.925 0.285
Participation in women’s association (1 = yes, 0 = no) 0.201 0.401
Total credit household received (preceding year) 390.397 867.948
Household total expenditure (preceding year) 2990.173 1472.150
Aridity Index (AI) (1 = AI [ 1, 0 = AI \ 1) 0.275 0.447
Rainfall Variability (CV) (1 = CV [ 0.310, 0 = CV \ 0.310) 0.586 0.493
Average Temperature (AT) (1 = AT [ 27 �C, 0 = AT \ �C) 0.284 0.451
Land use (LU) (1 = Agro-pastoral, 0 = Agriculture) 0.703 0.457
Erosion rate 2 (1 = medium, 0 = otherwise) 0.138 0.345
Erosion rate 3 (1 = high/critical, 0 = otherwise) 3.589 4.898
Erosion rate 1 (1 = low/tolerable, 0 = otherwise) (control/omitted) 0.292 0.455
Debo is a traditional labor sharing system, Equib is traditional group saving mechanism, Edir is traditionalgroup support association
SD standard deviation
G. Gebregziabher et al.
123
Fig. 2 Adoption of RWM technologies by landscape
Fig. 3 Land degradation and adoption of RWM technologies
Table 5 Correlation coefficients for MVP Regression Equations (Robust standard error in parentheses)
RWM technology Multi-purpose trees Orchards Bunds/terraces
Orchards q21 = 0.526*** (0.093)
Bunds/terraces q31 = 0.167** (0.082) q32 = 0.243*** (0.087)
Gully rehabilitation q41 = 0.141* (0.073) q42 = 20.008 (0.072) q43 = 0.184** (0.072)
Likelihood ratio test of q21 = q31 = q41 = q32 = q42 = q43 = 0: v2 (6) = 39.199 Prob [v2 = 0.000
*, ** and *** indicates level of significance at 10, 5 and 1 %, respectively
Interdependence in rainwater management technologies
123
independent; hence, the multivariate probit approach is appropriate in this case. Similarly,
the likelihood ratio test [v2 (6) = 39.199 and probability [v2 = 0.000] indicates a sig-
nificant joint correlation between the technologies, and supports the estimation of multi-
variate probit model as opposed to univariate probit model.
Furthermore, the positive and significant correlation coefficients of the error terms
indicate that there is complementarity (positive correlation) between different RWM
technologies being used by farmers, and support the assumption of interdependence and
path dependence (sequential interdependence) between the different options. Differences
in the estimated coefficients across equations also support the appropriateness of differ-
entiating between technology options.
Parameter estimations from the multivariate probit model are presented in Table 6. The
regression result reveals that factors that determine the adoption of technologies can be
broadly classified into household and socioeconomic characteristics, access to credit, social
capital and biophysical characteristics. Social capital captured in the form of household
membership and participation in informal community networks was used as a proxy for
social capital. The mean Aridity Index (AI), coefficient of rainfall variability (CV), erosion
rate/ton/ha/year, average temperature (�c) and land use characteristics were used to capture
the effect of biophysical characteristics.
The estimation regression analysis indicates that adoption of technology has a negative
statistically significant association with the age of the household head. This is possibly due
to the fact that older farmers are less likely to adopt technologies than younger farmers, as
the latter are better able to provide the labor required implementing the technologies, and/or
older farmers may have shorter planning horizons and are more risk-averse. Kassie et al.
(2013) found similar results in the adoption of sustainable agricultural practices in Tanzania.
The results also indicate that male-headed households are more likely to adopt multi-
purpose trees compared to female-headed households, which is in agreement with the
findings of Adesina et al. (2000) and Kassie et al. (2009). Kassie et al. (2009) reported that
female-headed households are less likely to adopt sustainable agricultural technologies in
Tanzania. Male-headed households thus have a comparative advantage in the adoption of
RWM technologies in the Blue Nile Basin, because most of the agricultural work is
typically performed by men, while women are usually restricted to household and backyard
activities. Furthermore, men are more likely to have better farming experience that will
also increase their probability of technology adoption, because experienced farmers are
more likely to have better access to information and knowledge of climatic conditions and
coping mechanisms. In this context, the policy implication is that targeting women groups
to address their constraints to actively participate in rural economic activities can have a
significant impact on the adoption of these technologies. Furthermore, farmers with better
experience and access to information are most likely to take initiatives in adopting and
testing new technologies.
As expected, family size in adult equivalent has a positive significant effect on the
adoption of multi-purpose trees and orchards, possibly indicating that these are labor-
intensive—family size can determine availability of labor. Marenya and Barrett (2007)
observed a similar result in Kenya.
Participation in off-farm activities was found to have a statistically significant negative
effect on the adoption of gully rehabilitation. This is likely because participation in off-
farm activities compete for labor, which could have been used in rainwater management
activities.
Ownership of livestock has a significant positive impact on the adoption of orchards and
bunds/terraces, implying that household wealth positively affects their decision to adopt a
G. Gebregziabher et al.
123
Ta
ble
6R
esult
so
fth
em
ult
ivar
iate
pro
bit
mod
el
Ind
epen
den
tv
aria
ble
sT
ech
nolo
gie
s(d
epen
den
tv
aria
ble
s)
Mult
i-purp
ose
tree
sO
rchar
ds
Bunds/
terr
aces
Gull
yre
hab
ilit
atio
nC
oef
fici
ent
Coef
fici
ent
Coef
fici
ent
Coef
fici
ent
Hu
ma
nca
pit
al
Ag
eo
fh
ou
seh
old
hea
d(y
ears
)-
0.0
11
**
(-0
.004
)-
0.0
15
**
*(-
0.0
06
)-
0.0
14
**
*(-
0.0
05
)-
0.0
12
**
*(-
0.0
04
)
Gen
der
of
ho
use
ho
ldh
ead
(1=
mal
e)0
.358
*(-
0.1
88
)0
.23
0(-
0.2
41
)0
.282
(-0
.187
)0
.075
(-0
.166
)
Fam
ily
size
inad
ult
equiv
alen
t0.1
36***
(-0
.043
)0
.20
5*
**
(-0
.053
)0
.031
(-0
.051
)0
.034
(-0
.043
)
House
hold
hea
dis
educa
ted
(1=
yes
)0
.198
(-0
.15
6)
0.0
45
(-0
.181
)0
.174
(-0
.168
)-
0.0
01
(0.1
49)
Nu
mb
ero
fh
ou
seh
old
mem
ber
sw
ith
elem
enta
ry(1
–8
)ed
uca
tio
nle
vel
-0
.030
(-0
.04
5)
0.0
13
(-0
.058
)0
.045
(-0
.054
)0
.048
(-0
.044
)
Nu
mb
ero
fh
ou
seh
old
mem
ber
sw
ith
hig
hsc
ho
ol
and
abo
ve
(C9)
educa
tion
lev
el-
0.0
05
(-0
.05
7)
-0
.01
7(-
0.0
67
)-
0.0
87
(-0
.064
)-
0.0
14
(-0
.057
)
Ph
ysic
al
cap
ita
l
Liv
esto
ckh
old
ing
inT
LU
0.0
13
(-0
.01
5)
0.0
30
*(-
0.0
18
)0
.052
**
*(-
0.0
18
)-
0.0
01
(-0
.015
)
Lan
dhold
ing
per
adult
equiv
alen
t0.3
74**
(-0
.160
)0
.62
9*
**
(-0
.199
)-
0.3
21
*(-
0.1
64
)-
0.1
75
(-0
.174
)
Acc
ess
tom
ark
ets/
serv
ices
and
house
hold
expen
dit
ure
Tota
lcr
edit
house
hold
rece
ived
(pre
cedin
gyea
r)-
0.0
01
(-6
.38
0)
-0
.00
1(-
8.1
90
)-
2.7
60
(-7
.720
)-
9.7
10
(-6
.280
)
House
hold
’sfo
od
and
non-f
ood
expen
dit
ure
(pre
cedin
gyea
r)-
4.7
71
(-3
.96
1)
-2
.19
1(-
2.1
30
)-
1.1
71
*(-
6.9
50
)1
.411
(-3
.530
)
So
cia
lca
pit
al
Par
tici
pat
ion
inoff
-far
mac
tivit
ies
(1=
yes
)0
.146
(-0
.13
3)
-0
.16
3(-
0.1
75
)-
0.1
22
(-0
.152
)-
0.2
58
*(-
0.1
34
)
House
hold
par
tici
pat
esin
Deb
o(1
=y
es)
-0
.142
(-0
.20
3)
0.3
73
(-0
.263
)0
.532
**
*(-
0.1
89
)0
.321
(-0
.199
)
House
hold
par
tici
pat
esin
Equib
(1=
yes
)-
0.1
50
(-0
.16
5)
0.1
09
(-0
.221
)-
0.2
8(-
0.1
91
)0
.208
(-0
.170
)
House
hold
par
tici
pat
esin
Edir
(1=
yes
)0
.266
(-0
.26
7)
0.9
61
**
*(-
0.2
97
)0
.400
*(-
0.2
31
)0
.036
(-0
.224
)
House
hold
par
tici
pat
esin
wom
en’s
asso
ciat
ions
(1=
yes
)0
.503
**
*(-
0.1
38
)0
.39
5*
*(-
0.1
76
)0
.009
(-0
.163
)0
.004
(-0
.140
)
Bio
ph
ysic
al
Ch
ara
cter
isti
cs
Ari
dit
yIn
dex
(AI)
(1=
AI[
1,
0=
AI\
1)
0.1
78
(-0
.51
9)
5.2
50
**
*(-
1.0
29
)4
.973
**
*(-
0.5
94
)2
.943
**
*(-
0.5
07
)
Rai
nfa
llV
aria
bil
ity
(CV
)(1
=C
V[
0.3
10
,0
=C
V\
0.3
10
)-
0.1
77
(-0
.49
1)
-3
.74
7*
**
(-0
.883
)-
4.8
56
**
*(-
0.5
51
)-
2.7
03
**
*(-
0.4
64
)
Av
erag
eT
emp
erat
ure
(AT
)(1
=A
T[
27
�C,
0=
AT
\�C
)-
0.8
04
(-0
.49
5)
5.5
83
**
*(-
1.0
44
)5
.538
**
*(-
0.5
76
)3
.298
**
*(-
0.4
89
)
Interdependence in rainwater management technologies
123
Ta
ble
6co
nti
nued
Ind
epen
den
tv
aria
ble
sT
ech
nolo
gie
s(d
epen
den
tv
aria
ble
s)
Mult
i-purp
ose
tree
sO
rchar
ds
Bunds/
terr
aces
Gull
yre
hab
ilit
atio
nC
oef
fici
ent
Coef
fici
ent
Coef
fici
ent
Coef
fici
ent
Lan
du
se(L
U)
(1=
Ag
ro-p
asto
ral,
0=
Agri
cult
ure
)0.4
94
(-0
.64
2)
-6
.46
6*
**
(-1
.146
)-
5.5
19
**
*(-
0.7
35
)-
4.0
85
**
*(-
0.6
25
)
Ero
sio
nra
te2
(1=
med
ium
,0
=o
ther
wis
e)-
0.0
87
4(-
0.4
65
)4
.27
7*
**
(-0
.738
)3
.691
**
*(-
0.5
33
)-
0.4
81
(-0
.443
)
Ero
sio
nra
te3
(1=
hig
h/c
riti
cal,
0=
oth
erw
ise)
0.0
07
(-0
.05
5)
-0
.51
0*
**
(-0
.106
)-
0.6
02
**
*(-
0.0
63
)-
0.2
12
**
*(-
0.0
52
)
Ero
sio
nra
te1
(1=
low
/tole
rable
,0
=o
ther
wis
e)(c
on
trol/
om
itte
d)
––
––
Co
nst
ant
-1
.445
**
(-0
.700
)1
.43
5(-
1.0
37
)5
.508
**
*(-
0.7
50
)3
.850
**
*(-
0.6
61
)
Reg
ress
ion
dia
gnost
ics
Nu
mb
ero
fo
bse
rvat
ion
s6
54
LR
test
of
q=
0:v2
(6)
45
.89
8*
**
Wal
d(v
2)
71
8.4
10
Lo
gp
seu
do
lik
elih
oo
d-
11
20
.67
9
Pro
b[
v20
.000
**
*
*,
**
and
***
indic
ates
level
sof
signifi
cance
at10,
5an
d1
%,
resp
ecti
vel
y.
Fig
ure
sw
ithin
par
enth
esis
are
robust
stan
dar
der
rors
G. Gebregziabher et al.
123
technology. On the other hand, ownership of land is positively correlated with the adoption
of multi-purpose trees and orchards, but negatively correlated with the adoption of bunds/
terraces. This is likely because households that have large farm sizes are less resource-
constrained and have better options to diversify their income, which in turn may negatively
affect willingness to invest in bunds/terraces. The negative effect on the adoption of bunds/
terraces is contrary to previous findings (Pender and Gebremedhin. 2007; Kassie et al.
2013). Since multi-purpose trees and orchards are typically private investments as opposed
to bunds/terraces and gully rehabilitation (which are usually carried out on a FFW basis),
those who own more land are more likely to defect collective action as they may not expect
to benefit from FFW payments. In this respect, the policy implication is that tenure ar-
rangement is likely to facilitate long-term investments in technologies.
Social capital captured by household membership in social networks captured by Debo,
Equib, and Edir was defined as binary (equal to 1 if the household is a member, and zero
otherwise). The result indicates that social capital positively affects a household’s decision
to adopt RWM technologies. For example, household’s membership in Debo has a sig-
nificant positive effect on the probability of adoption of bunds/terraces. Similarly,
household’s participation in Edir has a positive relationship with the adoption of orchards
and bunds/terraces. Both results suggest that informal social networks help members to use
their peer support to overcome labor and/or credit constraints. Our results are consistent
with the findings of previous studies (Kassie et al. 2009, 2013).
With respect to biophysical characteristics, the analysis shows that the adoption of orchards,
bund/terrace and gully rehabilitation is more likely in areas with a high aridity index and high
average temperatures. Unsurprisingly, egression results show that the adoption probability of
orchard, bund/terraces and gully rehabilitation is less likely in agro-pastoral areas compared to
in crop-based agricultural areas. This may be because these areas are relatively densely
populated and agriculture has been practiced for generations, resulting in land degradation. As
farmers who live the highland areas where crop-livestock farming system dominate are more
dependent on their land resources as a source of their livelihood, they may have more incentive
to invest and conserve their land, while communities in the agro-pastoral areas may not be as
vulnerable to land degradation as those who live in the densely populated crop-based agricul-
tural areas. Our regression results show that the effect of erosion on the probability of RWM
technology adoption is mixed. For example, the probability of adoption of orchard and bund/
terraces is high in areas with moderate (medium) erosion rates, while it is low in areas with a high
erosion rate. This suggests that the amount of recourse (investment) needed in highly degraded
areas is beyond the capacity of the farm households and there is a need for external support.
Finally, our regression results show that the adoption of multipurpose trees is not related
to biophysical factors, possibly indicating that the adoption is induced by prior invest-
ments, such as bund/terraces adopted sequentially, demonstrating path dependence be-
tween some of the technologies.
7 Discussion and conclusions
A significant contributor to the low agricultural productivity is not merely due to the lack
of water, but also as a result of poor water resources management. Hence, understanding
the probability of adoption of appropriate rainwater management technologies is a policy
issue. This paper uses a large household survey data from seven watersheds across the Blue
Nile Basin, and a multivariate probit regression model, to investigate factors that influence
farmers’ decisions to adopt a suite of rainwater management technologies.
Interdependence in rainwater management technologies
123
While there is heterogeneity with regard to the factors that influence the choice of
technologies, our results underscore the importance of: household’s human capital, house-
hold’s participation in off-farm activities and informal community networks, household’s
ownership of physical capital, access to market and credit services, and biophysical factors.
The evidence of interdependence between RWM technologies suggests that the adop-
tion of a specific technology is not a standalone decision, but is linked to the adoption of
other technologies, and supports the assertion that it is important to consider a package of
technologies as opposed to promoting a single technology.
The fact that male farmers are more likely to adopt RWM technologies implies that men
have a comparative advantage in improving their management of available water. In this
context, targeting women farmers to address their constraints to actively participate in rural
economic activities can have a significant impact on the adoption of technologies and the
resulting livelihood improvements.
Farmers’ participation in off-farm activities may have a positive impact in generating
additional income to the household, but these are likely to compete for labor, which could
have been used in rainwater management activities. The mixed relationship between
landholding and different RWM technologies indicates the need to clearly understand how
tenure arrangement may affect or facilitate long-term investments in technologies.
In terms of biophysical characteristics, the results show that the adoption of RWM tech-
nologies in particular, and agricultural production enhancing technologies in general, need to
be contextualized to specific biophysical settings. In general, our results suggest that it is
important to examine the socioeconomic and demographic characteristics of households, and
biophysical suitability of watersheds, instead of promoting blanket recommendations for the
adoption of RWM technologies. The regression results, together with insights gained from
qualitative analysis, suggest that households with (a) limited landholdings, (b) limited access to
markets, information and extension services, (c) bigger family size in adult equivalent,
(d) capital constraints and limited access to credit, and (e) limited livestock and asset ownership
are the most appropriate target groups for adoption and scaling-up of RWM technologies.
Constraints that limit women’s active participation in rural economic activities are important
and need to be considered. Correlation coefficients indicate that the adoption of RWM tech-
nologies is correlated, implying interdependence between different technologies. The adoption
and promotion of RWM technologies should therefore follow a holistic approach.
References
Adesina, A. A., Mbila, D., Nkamleu, G. B., & Endamana, D. (2000). Econometric analysis of the deter-minants of adoption of alley farming by farmers in the forest zone of Southwest Cameroon. Agri-culture, Ecosystems and Environment, 80, 255–265.
Amha, R. (2006). Impact assessment of rainwater harvesting ponds: The case of Alaba Woreda, Ethiopia.Faculty of Business and Economics, Addis Ababa University (Thesis).
Amede, T. (2012). Rainwater management for sustainable agricultural intensification in the Ethiopian high-lands. Presentation from the 2012 world water week in Stockholm. http://www.worldwaterweek.org/documents/WWW_PDF/2012/Tue/Rainfed-production-under-growing/Tilahun-Amede.pdf. Accessed onAugust 13, 2013.
Amsalu, A., & de Graaff, J. (2007). Determinants of adoption and continued use of stone terraces for soiland water conservation in an Ethiopian highland watershed. Ecological Economic, 61, 294–302.
Awulachew, S. B., Demissie, S. S., Hagos, F., Erkossa, T., & Peden, D. (2012). Water managementintervention analysis in the Nile Basin. In S. B. Awulachew, V. Smakhtin, D. Molden, & D. Peden(Eds.), The Nile River Basin: Water, agriculture, governance and livelihoods (pp. 292–311). Oxon,UK: Routledge.
G. Gebregziabher et al.
123
Awulachew, S. B., Rebelo, L.-M., & Molden, D. (2010). The Nile Basin: Tapping the unmet agriculturalpotential of Nile waters. Water International, 35(5), 623–654.
Belderbos, R., Carree, M., Diederen, B., Lokshin, B., & Veugelers, R. (2004). Heterogeneity in R&Dcooperation strategies. International Journal of Industrial Organization, 22, 1237–1263.
Cappellari, L., & Jenkins, S. P. (2003). Multivariate probit regression using simulated maximum likelihood.The Stata Journal, 3(3), 278–294.
Cowen, R., & Gunby, P. (1996). Sprayed to death: Path dependence, lock-in and pest control strategies. TheEconomic Journal, 5(106), 521–542.
CSA (Central Statistical Agency). (2004). The Federal Democratic Republic of Ethiopia statistical abstractfor 2003. Addis Ababa, Ethiopia: CSA.
De Groote, H., & Coulibaly, N. (1998). Gender and generation: An intra-household analysis on access toresources in southern Mali. African Crop Science Journal, 6(1), 79–95.
Deressa, T. (2007). Measuring the economic impact of climate change on Ethiopian agriculture: Ricardianapproach. World Bank Policy Research Paper No. 4342. Washington, DC: World Bank.
Deressa, T., Hassan, R. M., Ringler, C., Alemu, T., & Yesuf, M. (2009). Determinants of farmers’ choice ofadaptation methods to climate change in the Nile Basin of Ethiopia. Global Environmental Change, 19,248–255.
FAO (Food and Agriculture Organization of the United Nations). (2005). Irrigation in Africa in Figures:AQUASTAT survey—2005. FAO Water Report No. 29. Rome: FAO. ftp://fao.org/agl/aglw/docs/wr29_eng.pdf. Accessed on August 13, 2012.
Fafchamps, M., & Gubert, F. (2007). The formation of risk sharing networks. Journal of DevelopmentEconomics, 83, 326–350.
Getnet, K., & MacAlister, C. (2012). Integrated innovations and recommendation domains: Paradigm fordeveloping, scaling-out, and targeting rainwater management innovations. Ecological Economics, 76,34–41.
Guerin, T. (1999). An Australian perspective on the constraints to the transfer and adoption of innovations inland management. Environmental Conservation, 24(4), 289–304.
Haileslassie, A., Priess, J., Veldkamp, E., Teketay, D., & Lesschen, J. P. (2005). Assessment of soil nutrientdepletion and its spatial variability on smallholders’ mixed farming systems in Ethiopia using partialversus full nutrient balances. Agriculture, Ecosystems and Environment, 108, 1–16.
Hawando, T. (2000). Desertification in Ethiopian highlands, RALA Report NO. 200, Norwegian ChurchAID, Addis Ababa, Ethiopia.
http://www.cgiar-csi.orghttp://www.ethiomet.gov.et/data_access/informationILRI (International Livestock Research Institute) and IWMI (International Water Management Institute).
(2010). Rainwater management in the Ethiopian highlands: Mapping, targeting and scaling out in-terventions. Nairobi, Kenya: International Livestock Research Institute (ILRI). http://mahider.ilri.org/handle/10568/2428. Accessed on August 18, 2012.
Kassie, M., Jaleta, M., Shiferaw, B., Mmbando, F., & Mekuria, M. (2012). Interdependence in farmertechnology adoption decisions in smallholder systems: Joint estimation of investments in sustainableagricultural practices in rural Tanzania. In Paper prepared and selected for presentation at the In-ternational Association of Agricultural Economists (IAAE) Triennial Conference, Foz do Iguacu,Brazil, August 18–24.
Kassie, M., Jaleta, M., Shiferaw, B., Mmbando, F., & Mekuria, M. (2013). Adoption of interrelated sus-tainable agricultural practices in smallholder systems: Evidence from rural Tanzania. TechnologicalForecasting and Social Change, 80(3), 525–540.
Kassie, M., Zikhali, P., Manjur, K., & Edwards, S. (2009). Adoption of organic farming technologies:Evidence from semi-arid regions of Ethiopia. Natural Resources Forum, 33, 189–198.
Kassie, M., Pender, J., Yesuf, M., Kohlin, G., Bluffstone, R., & Mulugeta, E. (2008). Estimating returns tosoil conservation adoption in the northern Ethiopian highlands. Agricultural Economics, 38(2),213–232.
Kato, E.; Ringler, C.; Yesuf, M.; Bryan, E. (2009). Soil and water conservation technologies: A bufferagainst production risk in the face of climate change. Insights from the Nile Basin in Ethiopia. IFPRIDiscussion Paper 00871. Washington, DC: International Food Policy Research Institute (IFPRI).
Kurukulasuriya, P.; Rosenthal, S. (2003). Climate change and agriculture: A review of impacts andadaptations. Climate change series Paper No. 91. Paper prepared and published for the RuralDevelopment Group and Environment Department of the World Bank.
Lautze, S., Aklilu, Y., Raven-Roberts, A., Young, H., Kebede, G., Learning, J. (2003). Risk and vul-nerability in Ethiopia: Learning from the past, responding to the present, preparing for the future.Addis Ababa, Ethiopia: Report for the United States Agency for International Development (USAID).
Interdependence in rainwater management technologies
123
Lee, D. R. (2005). Agricultural sustainability and technology adoption: Issues and policies for developingcountries. American Journal of Agricultural Economics, 87(5), 1325–1334.
Marenya, P. P., & Barrett, C. B. (2007). Household-level determinants of adoption of improved naturalresources management practices among smallholder farmers in western Kenya. Food Policy, 32,515–536.
McDonald, K., & Brown, K. (2000). Soil and water conservation projects and rural livelihoods: Options fordesign and research to enhance adoption and adaptation. Land Degradation and Development, 11,343–361.
Merrey, D. J., Gebreselassie, T. (2011). Promoting improved rainwater and land management in the BlueNile (Abay) basin of Ethiopia. NBDC Technical Report 1. Nairobi, Kenya: International LivestockResearch Institute (ILRI).
MoFED (Ministry of Finance and Economic Development). (2010). Annual report on macroeconomicdevelopments. http://www.mofed.gov.et/English/Resources/Documents/2002MEDRAnnual.pdf. Ac-cessed on April 8, 2013.
MoWE (Ministry of Water and Energy). (1998). Abbay River Basin integrated development master planproject (1998), phase 2 report, section II (Vol. VIII). Ethiopia: Addis Ababa.
Moyo, S., & Veeman, M. (2004). Analysis of joint and endogenous technology choice for protein sup-plementation by smallholder dairy farmers in Zimbabwe. Agroforestry Systems, 60, 199–209.
NMA (National Meteorology Agency). (2007). National Adaptation Program of Action (NAPA) on climatechange of Ethiopia. Bonn, Germany: United Nations Framework Convention on Climate Change(UNFCCC). http://unfccc.int/resource/docs/napa/cpv01.pdf. Accessed on April 17, 2013.
Nata, T., & Bheemalingeswara, K. (2010). Prospects and constraints of household irrigation practices,Hayelom Watershed, Tigray, Northern Ethiopia. Momona Ethiopian Journal of Science, 2(2), 87–109.
Ndiritu, S. W., Kassie, M., Shiferaw, B., Ouma, J., & Odendo, M. (2011). Adoption of agricultural tech-nologies in Kenya: How does gender matter? Nairobi. Kenya: International Maize and Wheat Im-provement Center (CIMMYT).
Nhemachena, C., Hassan, R. (2007). Micro-level analysis of farmers’ adaptation to climate change inSouthern Africa. IFPRI Discussion Paper No. 00714. Washington, DC: International Food PolicyResearch Institute (IFPRI).
Pender, J., & Gebremedhin, B. (2007). Determinants of agricultural and land management practices andimpacts on crop production and household income in the highlands of Tigray. Ethiopia, Journal ofAfrican Economics, 17, 395–450.
Quisumbing, A. R., Brown, L., Hillary, R., Feldsten, S., Haddad, L., Pena, C. (1995). Women: The key tofood security. Food Policy Statement No. 21. Washington, DC: International Food Policy ResearchInstitute (IFPRI).
Rockstrom, J., Karlberg, L., Wani, S. P., Barron, J., Hatibu, N., Oweis, T., et al. (2010). Managing water inrainfed agriculture—The need for a paradigm shift. Agricultural Water Management, 97, 543–550.
Santini, G., Peiser, L., Faures, J. M., Neves, B., Vallee, D. (2011). Planning smart investments in agri-cultural water management through a livelihood mapping approach: The case of Ethiopia (Draft).
Shiferaw, A. (2011). Estimating soil loss rates for soil conservation planning in the Borena Woreda of SouthWollo Highlands, Ethiopia. Journal of Sustainable Development in Africa, 13(3), 87–106.
Soule, J. M., Tegene, A., & Wiebe, D. K. (2000). Land tenure and the adoption of soil conservationpractices. American Journal of Agricultural Economics, 82(4), 993–1005.
Tesfay, G. (2011). On-farm water harvesting for rainfed agriculture development and food security inTigray, Northern Ethiopia: Investigation of technical and socioeconomic issues. Drylands Coordina-tion Group Report No. 61. http://www.drylands-group.org
Van den Born, N. (2011). Conjunctive water management: How to use the full potential: A literatureresearch (B.Sc. Thesis): Irrigation and Water Engineering: Wageningen University, The Netherlands.
Wollni, M., Lee, D. R., & Thies, J. E. (2010). Conservation agriculture, organic marketing, and collectiveaction in the Honduran hillsides. Agricultural Economics, 41, 373–384.
Yu, L., Hurley, T., Kliebenstein, J., Orazen, P. (2008). Testing for complementarity and substitutabilityamong multiple technologies: The case of US hog farms. Working Paper No. 08026. Ames, IA, USA:Iowa State University, Department of Economics.
Zemadim, B., McCartney, M., Sharma, B., & Wale, A. (2011). Integrated rainwater management strategiesin the Blue Nile Basin of the Ethiopian highlands. International Journal of Water Resources andEnvironmental Engineering, 3(10), 220–232.
G. Gebregziabher et al.
123
