Adoption of Soil Organic 
Carbon-Enhancing 
Practices: A case of two 
watershed sites in Ethiopia
Bioversity International and the International Center for
Tropical Agriculture (CIAT) are part of CGIAR, a global research
partnership for a food-secure future.
Bioversity International is the operating name of the
International Plant Genetic Resources Institute (IPGRI).
www.bioversityinternational.org
www.ciat.cgiar.org
www.cgiar.org
Alliance
Africa Hub
c/o icipe (International Centre of Insect
Physiology and Ecology)
Duduville Campus Off Kasarani Road
P.O. Box 823-00621, Nairobi, Kenya
Tel. (+254) 0709134000
The Alliance of Bioversity International and the International Center for Tropical Agriculture (CIAT) delivers 
research-based solutions that address the global crises of malnutrition, climate change, biodiversity loss and 
environmental degradation.
The Alliance focuses on the nexus of agriculture, nutrition and environment. We work with local, national and 
multinational partners across Africa, Asia, Latin America and the Caribbean, and with the public and private 
sectors and civil society. With novel partnerships, the Alliance generates evidence and mainstreams innovations to 
transform food systems and landscapes so that they sustain the planet, drive prosperity and nourish people.
The Alliance is part of CGIAR, the world’s largest agricultural research and innovation partnership for a food-secure 
future dedicated to reducing poverty, enhancing food and nutrition security, and improving natural resources.
www.bioversityinternational.org            www.ciat.cgiar.org           www.cgiar.org
Stanley Karanja Ng’ang’a1
Fekadu Gelaw Mersha2
Wilson Nguru3
Evan Givertz1
Adoption of Soil Organic Carbon-
Enhancing Practices: A case of two 
watershed sites in Ethiopia
1 Alliance of Bioversity International and the International Center for Tropical Agriculture
2 Haramaya University, Ethiopia
3  University of Nairobi, Kenya
March 2020
Alliance of Bioversity International and the International Center for Tropical Agriculture (CIAT)
Sub-regional Office for Africa
Pan-Africa Bean Research Alliance National Agricultural Research Laboratories – Kawanda
P.O. Box 6247
Kampala, Uganda
Email:  s.karanja@cgiar.org
Website:  www.ciat.cgiar.org
Citation
Ng’ang’a SK; Gelaw Mersha F; Nguru W; Girvetz E. 2020. Adoption of Soil Organic Carbon-Enhancing Practices: A case of 
two watershed sites in Ethiopia. CIAT Publication No. 494. International Center for Tropical Agriculture (CIAT). Kampala, 
Uganda. 35 p. Available at: https://hdl.handle.net/10568/107916
About the authors
Stanley Karanja Ng’ang’a, Development Economist, Climate Action Research Area, Alliance of Bioversity International 
and CIAT, Africa Hub, Kampala, Uganda.
Fekadu Gelaw Mersha, Assistant Professor of Agricultural Economics, Institutional and Behavioral Economics School 
of Agricultural Economics and Agribusiness, College of Agriculture and Environmental Sciences, Haramaya University, 
Ethiopia.
Wilson Nguru, Department of Land Resource Management and Agricultural Technology, University of Nairobi, Kenya.
Evan Girvetz, Principal Scientist and Global Team Leader, Finance and Investment for Climate Action, Alliance of Bioversity 
International and CIAT, Africa Hub, Nairobi, Kenya.
The data behind this research is available for download here: https://doi.org/10.7910/DVN/QTACSN
Kindly use this citation if you plan to reuse this dataset:
Ng’ang’a SK; Gelaw F; Nguru WM; Kanyenji GM; Girvetz E. 2019. An integrated approach for understanding the factors 
that facilitate or constrain the adoption of soil carbon enhancing practices in East Africa, Kenya and Ethiopia. Harvard 
Dataverse https://doi.org/10.7910/DVN/QTACSN
See also data paper:
Nguru WM; Ng'ang'a S; Gelaw F; Kanyenji GM; Girvetz EH. 2020. Survey data on factors that constrain the adoption of soil 
carbon enhancing technologies in Ethiopia. Scientific Data 7, Article number: 93. https://doi.org/10.1038/s41597-020-0431-9
Cover photo: 
© CIAT 2020. Photo by Georgina Smith. Some Rights Reserved. This work is licensed under a 
Creative Commons Attribution NonCommercial 4.0 International License (CC-BY-NC)
https://creativecommons.org/licenses/by-nc/4.0/
Copyright © CIAT 2020. Some rights reserved.
CIAT encourages wide dissemination of its printed and electronic publications for maximum public benefit. Thus, in most cases, colleagues 
working in research and development should feel free to use CIAT materials for noncommercial purposes. However, the Center prohibits 
modification of these materials, and we expect to receive due credit. Though CIAT prepares its publications with considerable care, the 
Center does not guarantee their accuracy and completeness.
Acknowledgments
The authors would like to thank the Federal Ministry for Economic Cooperation and Development (BMZ/GTZ) for funding 
this work. We are also grateful to all the farmers from the Angacha and Bure-Yesir watersheds who provided information 
that forms the basis of this report. Our thanks also go to all the enumerators (Kabede, Bereket, Getu, Mulugeta, Ayele, 
Tesfaye, Zerihun, Tilahun, Derso, Mengista, Ayana, Ayenew, Tibebu, Yigezaw) for undertaking the household interviews, 
and agricultural extension officers (Abebe, Kidihun and Desta), who provided critical logistical and coordination support.
iii
Acronyms and abbreviations
ADLI Agricultural Development Led Industrialization
AGP Agricultural Growth Program
ATP Agricultural Transformation Programs
CIAT International Center for Tropical Agriculture
CSA Central Statistical Agency
DA development agents
DAP diammonium phosphate
DAWR Department of Agriculture and Water Resources
FAO Food and Agriculture Organization of the United Nations
FGD Focus group discussion
FTC Farmers’ Training Centers
GHK Geweke–Hajivassiliou–Keane
GO government organization
KA Kebele administrations
LR likelihood ratio
MFI micro-finance institutions
NGO nongovernment organization
NPS nitrogen, phosphorus, and sulfur fertilizer 
OLS Ordinary least squares regression
PA pastoral area
PASDEP Plan for Accelerated and Sustained Development to End Poverty
SD standard deviation
SOC soil organic carbon
TLU Tropical Livestock Unit
TV television
Contents
Acronyms and abbreviations ................................................................................................................................................iii
Abstract ....................................................................................................................................................................................... 2
1. Introduction .............................................................................................................................................................................. 4
1.1 Background of the study.............................................................................................................................................. 4
1.2 Objectives ......................................................................................................................................................................... 6
1.3 Production context and land tenure......................................................................................................................... 7
 1.3.1 The production context .................................................................................................................................. 7
 1.3.2 Land tenure ........................................................................................................................................................ 9
1.4 Adoption of carbon-enhancing technologies ....................................................................................................... 9
1.5 Conceptual framework ..............................................................................................................................................11
2. The data and analytical model ..........................................................................................................................................12
2.1 About the data ...............................................................................................................................................................12
2.2 Sampling design...........................................................................................................................................................12
2.3 Methods of collection .................................................................................................................................................13
2.4 Analytical model ...........................................................................................................................................................13
 2.4.1 Multivariate probit model ............................................................................................................................13
 2.4.2 OLS regression model ..................................................................................................................................14
2.5 Description of dependent and independent variables ....................................................................................15
3. Discussion of results .............................................................................................................................................................16
3.1. Description of socioeconomic characteristics of sample households .......................................................16
3.2. Econometric analysis .................................................................................................................................................26
 3.2.1 Determinants of level of adoption .............................................................................................................26
 3.2.2 Determinants of intensity of adoption .....................................................................................................29
4. Conclusions .............................................................................................................................................................................31
References ...................................................................................................................................................................................33
Appendix. Conversion of livestock number into TLU ......................................................................................................35
Tables
Table 1:  Distribution of sample households in the two watersheds .................................................................13
Table 2:  Demographic characteristics of sample household heads at the two sites ...................................16
Table 3:  Mean comparison of household characteristics between the two sites .........................................17
Table 4:  Descriptive statistics of wealth indicators ...............................................................................................18
Table 5:  Descriptive statistics of land and livestock holdings of sample households. ................................19
Table 6:  Description of sample households’ access to services ........................................................................20
Table 7:  Descriptive statistics of indicators of land fragmentation. ..................................................................20
Table 8:  Participation level in the different SOC-enhancing technologies between the two sites ..........21
Table 9:  Distribution of implementation of the types of SOC-enhancing practices. ....................................22
Table 10:  Distribution of households in terms of the number of plots in which  
they implemented SOC practices. ..............................................................................................................23
Table 11:  Descriptive statistics of fertilizer use and proportion of residue left in the field. ..........................23
Table 12:  Uses of crop residue for different purposes between the two sites. .................................................24
Table 13:  Extent of implementation of the three categories of SOC practices  
among sample households ..........................................................................................................................24
Table 14:  Descriptive statistics of categorical variables ........................................................................................25
Table 15:  Descriptive statistics of continuous dependent and independent variables. ...............................26
Table 16:  Estimation results of a multivariate probit model ..................................................................................27
Table 17:  OLS regression estimation results of implementation of all SOC-enhancing practices .............29
2Abstract
This study aimed at identifying the factors that determine 
the decision to adopt and the intensity of adoption of 
soil organic carbon (SOC)-enhancing practices using two 
watershed sites in Ethiopia: Yiser (Amhara region) and 
Azugashube (Southern region). The study used survey 
data collected from 379 sample households drawn from 
four Kebele/village administrations at each watershed 
site. Multivariate and ordinary least squares regressions 
were used to identify the factors that determine the 
decision to adopt the SOC-enhancing practices and the 
factors that determine the extent of adoption of these 
practices, respectively. The study classified these various 
practices into three classes: soil and water conservation, 
agronomic, and agroforestry SOC-enhancing practices. 
We find that the decision to adopt soil and water 
conservation practices is negatively related to both the 
decision to adopt agronomic and to adopt agroforestry 
SOC-enhancing practices. On the contrary, we find 
that the decision to adopt agronomic and agroforestry 
practices is complementary. The study also identified 
diverse agroecological, farming system, institutional, and 
household characteristics that determine the decision to 
adopt and the intensity of adoption of the three SOC-
enhancing practices. Among the different variables, the 
study found location as a strong determinant of the type 
and intensity of adoption of the SOC practices. 
3Working Paper
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia4
1. Introduction
1.1.  Background of the study
Natural resource degradation has remained the primary 
source of the poor performance of the agricultural 
sector in developing countries. Natural resource 
degradation coupled with the increase in adverse 
impacts of climate change has caused the productivity 
of agriculture to stagnate even in the face of increased 
development interventions to improve its performance. 
Natural resource degradation not only increases 
farmers’ discount rates (Heath and Binswanger 1996) 
by reducing land productivity, but it also undermines 
the potential impacts of increased use of productivity-
enhancing agricultural technologies. Antle and Diagana 
(2003) identify degradation of soils through the loss 
of soil organic matter as the key component of many 
unsustainable agricultural systems.
The ruling party in Ethiopia took political power in 
1991 and declared Agricultural Development Led 
Industrialization (ADLI) as its main development strategy, 
aimed at guiding all other policies and strategies. ADLI 
put the agricultural sector and rural society at the 
center of its development agenda. In addition to large 
programs designed to directly enhance the production 
and productivity of the agricultural sector and improve 
the welfare of rural society, the stated primary goal of 
most other policies and programs was also to contribute 
to the development of the agricultural sector and rural 
society. Thus, large countrywide programs such as 
Poverty Reduction Strategic Paper I, Plan for Accelerated 
and Sustained Development to End Poverty (PASDEP), 
Agricultural Transformation Programs (ATP), Agricultural 
Growth Program (AGP), Production Safety Net programs, 
Food Security programs, and many other small programs 
and projects have been implemented over the past two 
and half decades. Despite all these, ADLI has remained 
the primary development strategy for more than two 
decades, but the country has not progressed into 
industrialization. The government has claimed that the 
economy has been achieving rapid (two-digit) growth for 
successive years in the past two decades. But it is not 
clear whether this growth has been achieved as a result 
of an increase in productivity or an increase in the size 
of the production area. A Central Statistical Agency (CSA) 
report shows that there has been continuous growth 
in production area. For instance, the size of the crop 
production area between 2000 and 2010 increased by 
66.2% (CSA 2010, 2000). This indicates that only a small 
portion of the growth was achieved due to an increase 
in productivity. Thus, growth in productivity was limited. 
But, more importantly, no perceptible change has been 
observed in the structure of the sector. Not only has the 
agricultural sector remained predominantly subsistent 
and its production system traditional but natural 
resource degradation has also increasingly become a 
bottleneck to growth of the sector.
However, conversion from natural vegetation to 
cultivated land is a well-known cause of soil organic 
carbon (SOC) loss (Post and Kwon 2000) as the SOC stock 
in the upper soil layer is highly sensitive to land-use 
change (FAO 2017). Depending on the climate, soil type, 
and historic management of the land, soils of the world’s 
Photo: CIAT/Georgina Smith
5Working Paper
agroecosystems (croplands, grazing lands, rangelands) 
are depleted of their SOC pool by 25–75%. Depletion 
of SOC adversely affects agricultural production in the 
long run by causing climate change, but it also affects 
agricultural production in the short run by reducing soil 
productivity and lowering the efficiency of added input 
(Lal 2011).
Land-use change per se may not reduce SOC stocks 
(Manna et al. 2008; Paustian et al. 2000). Instead, 
land-use change in agricultural production can even 
build the stocks if improved management practices 
are adopted that enhance soil quality, including the 
available water-holding capacity, cation exchange 
capacity, soil aggregation, and susceptibility to crusting 
and erosion (Lal 2006). Improved tillage management 
and cropping systems that reduce soil disturbance, 
increase the amount of land cover, and efficiently use 
production inputs (e.g., nutrients and water) are critical 
aspects (Follett 2001). These improved agricultural 
practices ensure sustainable production in the long run 
by enhancing SOC and, at the same time, they improve 
household welfare in the short run by improving 
agronomic productivity (Lal 2011; Pretty et al. 2006; 
Ringius 1999). 
In view of this, developing countries have made a lot 
of efforts in the past to increase farmers’ adoption of 
improved management practices. Despite these efforts 
and contrary to the suggested positive relationships 
between biological productivity and SOC stocks, 
evidence generally shows that farmers in developing 
countries continue to use extractive practices that 
deplete the SOC pool, degrade soil quality, and 
adversely affect agronomic productivity (Lal 2006). 
Some argue that the interlinked forces of population 
pressure, poverty, and environmental degradation 
pose a common challenge for the effectiveness of the 
interventions made to reverse the deterioration of 
soil quality (Shiferaw and Holden 1998). Population 
pressure, the magnitude of poverty, and the extent of 
environmental degradation may limit the feasible set 
of technologies. These factors, in general, may limit 
the technologies by limiting the investment capacity of 
households and the responsiveness of technologies, and 
by raising the discount rates of long-term investments 
in soil improvement. Hence, this may explain the limited 
adoption of SOC-enhancing technologies in these 
countries in the short run. But, these factors do not 
explain the adoption of SOC-enhancing technologies 
in the long run if they are outcomes of inappropriate 
policies as argued by Heath and Binswanger (1996). 
In the short run, evidence shows striking differences 
even among poor households in the adoption of SOC-
enhancing practices (Scherr 2000). Furthermore, 
micro-scale research evidence suggests that other 
socioeconomic and biophysical characteristics explain 
household differences in the level of adoption of SOC-
enhancing technologies. Empirical evidence suggests 
that the set of household-specific, community-specific, 
and plot-specific factors determine the adoption of these 
technologies. Household-specific characteristics such as 
age, experience, education, sex of the household head, 
family size, and farm size (Melesse 2018; Kankwamba and 
Mangisoni 2015; Laxmi and Mishra 2007); community-
specific characteristics such as imperfections in 
financial, input, and output markets, tenure security, 
and institutional and policy environments (Glover et al. 
2013; Jack 2011; Gledhill et al. 2011); and plot-specific 
characteristics such as soil types and topography 
(Kankwamba and Mangisoni 2015; Glover et al. 2013) 
determine the adoption of SOC-enhancing practices. 
Photo: CIAT/Georgina Smith
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia6
Other factors such as the existing farming system can 
also allow only a few SOC-enhancing technologies. The 
grazing pattern and the dominant cropping pattern can 
limit the feasible SOC-enhancing technologies to a few 
technologies or practices. For instance, if an open and 
free grazing system is the norm in a community, it will be 
difficult for the farmers to adopt many of the soil carbon-
enhancing practices on their croplands such as alley 
cropping, grass strips, planting of agroforestry trees on 
cropland, and planting of improved forage grasses and 
trees on the common grazing lands. Maintaining soil and 
water conservation structures will also be costly as these 
will be damaged by animals. Exclusion of forests from 
animals will also be costly. This situation also does not 
incentivize farmers to leave residues on their croplands. 
As a result, this will leave farmers with few alternative 
SOC-enhancing technologies such as fallowing, crop 
rotation, and tillage management. The current research 
aims at identifying the type of SOC-enhancing practices 
that are being implemented by farmers in the study 
areas given their circumstances and identifying the 
factors that determine potential variations among 
households in the level and extent of adoption.
The primary purpose of this study is to identify 
household, community, and plot-level factors that 
determine a household’s adoption of SOC-enhancing 
technologies in the case of two watershed areas: Yesir 
and Azugashube watersheds located in the Southern and 
Amhara regions, respectively.
We hypothesized that several household, plot-level, 
ecological, farming system, and institutional variables 
determine the level and intensity of a household’s 
decision to adopt SOC-enhancing practices. Household 
socioeconomic characteristics such as education, 
wealth (livestock, land, number of radios held), and 
family size; indicators of institutional access (markets, 
extension, credit, and farmers’ organizations); plot 
characteristics such as slope, fertility, soil type, and 
erosion problems; and farming system indicators such 
as crop diversification, land fragmentation, fertility 
management, and fertilizer application were included 
in the analyses. In addition, we analyzed the effects of 
watershed location in determining the level and intensity 
of the adoption of SOC-enhancing practices. 
1.2. Objectives
The overall objective of this research is to identify the 
factors that influence the type, level, and extent of the 
adoption of SOC-enhancing practices among farmers at 
two watershed sites.
The specific objectives are:
to measure the level and extent of adoption of the 
different SOC-enhancing practices and
to identify the household- and community-specific factors 
that explain the level and extent of adoption of  
SOC-enhancing practices. 
1
2
Photo: CIAT/Georgina Smith
7Working Paper
1.3.  Production context and land tenure
1.3.1  The production context
The agricultural production system and the activities 
(e.g., deforestation, burning, plowing, grazing) 
determine the dynamics of soil carbon (Lal and Kimble 
1997). The production system at the two study sites 
(Azugashube and Yesir watersheds) is characterized by 
a mixed crop–livestock system that is common in the 
highland parts of Ethiopia. In the typical mixed crop–
livestock production system of the country, farmers 
primarily produce crops but also rear a few cattle 
and small ruminants (sheep and goats) as a means 
to supply oxen power for plowing and as a source of 
food and to generate income. Their crop production 
is characterized by small-scale subsistence production 
systems in which farmers produce a crop primarily to 
meet the food demand of the household. The total land 
size for a typical household is not only small, usually 
less than 1 hectare, but is also highly fragmented and 
scattered over different parts of the rural kebele.1 Owing 
to the rapidly growing population, the average land 
size per household has been dwindling. Compared to 
the Yesir watershed site, the land size is even smaller 
in Azugashube watershed. In a given cropping season, 
a household produces different types of crops on 
the different plots of land. Although crop production 
usually contributes to increasing storage of carbon in 
agricultural soils (DAWR 2016), the specific production 
context may matter.
Farmers produce livestock as an integral part of 
their crop production. Their livestock production is 
characterized by a traditional production system with 
low operating costs for feed and other management 
activities. The major source of feed is communal grazing 
land. In almost all mixed crop-livestock production 
systems, grazing lands are communally owned and 
farmers use the free-grazing system. The grazing 
system is free and open to all communities. Farmers in 
the area rear livestock for many purposes: to generate 
additional income, generate food, and carry out crop 
production activities. In these areas, farming activities 
such as plowing, threshing, and land leveling (especially 
for teff plots) are all done using cattle, donkeys, and 
horse/mule draft power. In addition, cattle and equines 
are the primary means of transportation. Most of the 
crop production activities in the mixed crop–livestock 
production system are hardly carried out without 
livestock power. In sum, this situation forces every 
household to hold at least two oxen and one donkey/
horse to carry out farming activities. Livestock in the 
mixed crop–livestock farming system not only generate 
food and income but also are necessary for farming 
operations. Thus, each household in the area must 
have some livestock. But, the total number of livestock 
(cattle, small ruminants, and equines) that exist in 
each community is larger than the number that can 
be carried by the pasture lands. The grazing lands are 
not only shrinking in size but they are also becoming 
degraded due to overgrazing. The communal grazing 
lands in almost all areas cannot carry the available 
number of livestock.
The common grazing system is such that, except in 
the rainy season when crops are in the field, animals 
freely roam anywhere within the areas delineated for 
the community. The norm is that once the crops are 
harvested, all crop fields are open for common and 
free grazing. Thus, a farmer cannot let the crop residue 
(straw and stack) and biomass of other plants decay 
as it is not possible to exclude animals from grazing 
what is left on the plots. Since the communal grazing 
lands are highly degraded by overgrazing, crop fields 
are important grazing areas for some period after 
harvesting periods. Once the farmer harvests the crop 
and piles it, the crop fields will usually be open for free 
grazing. For some crops, until the harvested crop is 
threshed or taken home, the croplands may be closed 
for animal grazing. During this period, the plot owners 
allow their own animals to graze on the remnant 
biomass. After all the farmers have threshed their crop 
and taken the residues home, all crop fields in the area 
can be opened for free grazing. Thus, part of the stack 
harvested with the grain will be taken away from the 
field to use it for other purposes and the remaining 
surface biomass will be fed to livestock. As a result, no 
surface biomass returns to the soil. Even the weeds 
removed from crop fields during the rainy season are 
also taken away from the plot in order to feed the 
animals. 
The crop residues taken away from the plots have 
multiple purposes: animal feed, construction, fire 
energy, and others. Since the crop residues produced in 
a given production season are not sufficient to meet the 
various demands of farmers, the farmers use the straw 
to feed only selected animals, usually oxen and lactating 
cows. The remaining straw is used for other purposes 
such as firewood and construction. Shortage of firewood 
is a serious problem in most areas. Thus, straw and 
cattle dung are important sources of fuel. In addition, 
1 Kebele is the lowest administrative unit that is formed by bringing several neighboring villages together and can contain around 1,000 households.
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia8
the straw of some crops is used for construction. For 
instance, teff straw is an important construction material 
to make “mud-bricks.” In some cases, wheat straw is also 
used as a replacement for making mud-bricks. Similarly, 
maize stalk is used for the construction of fences 
and walls, while wheat straw is used for making local 
mattresses. In sum, farmers have to allocate the straw 
and crop residue among competitive uses. Obviously, 
the straw used for fire energy and construction never 
returns to the soil. But, even the crop residues fed to 
animals are unlikely to return back to the soil as the 
dried dung is also used for fire.
Whether the existing large number of livestock is driven 
by Hardin’s (1968) "The Tragedy of the Commons" or 
is necessitated by the farming system is an empirical 
question. In spite of the dwindling size of the common 
grazing lands as a result of conversion in land use 
and the absence of collective rules to limit livestock 
number or to improve the deteriorating grazing lands, 
the large livestock number seems to be driven more 
by "The Tragedy of the Commons" than other reasons. 
Setting aside the source of large livestock numbers, the 
existing crop and livestock production systems not only 
erode the carbon pool in grazing lands but also limit 
the feasibility of implementing many of the carbon-
enhancing practices. Given the existing livelihood and 
farming system, implementing many SOC-enhancing 
practices such as mulching, agro-forestry (grass strips 
and planting of forage trees), alley cropping, and 
manuring is technically infeasible and economically 
costly. The open grazing system is also a problem for 
sustaining physical soil conservation practices such as 
soil bunds and fanya juu terraces.
Shortage of firewood is another serious problem 
in most rural areas, and cattle dung is used as an 
important source of fire energy. Households dry 
the dung collected from livestock pens and use it 
as firewood. Even the cattle dung dropped on the 
grazing lands and crop fields is also collected by some 
poor farmers and sold to people around towns as it 
is an important source of livelihood for some poor 
households. Moreover, wet cattle dung is used for 
smoothing house floors, walls, and threshing grounds. 
Thus, most portions of the biomass fed to cattle will not 
return to the soil. In sum, almost all the surface crop 
biomass produced every year is taken away from the 
soil and never returns to it. This means that only the 
underground biomass remains in the soil. 
Furthermore, losses are exacerbated by changes in land 
use. Land conversion from natural to agricultural land 
use often results in loss of SOC (Lal and Kimble 1997). 
In this regard, owing to the rapidly growing population, 
demand has been increasing for cropping land. As a 
result, there has been a conversion of natural/pasture 
land and forest land into agricultural land. However, the 
average landholding has been declining continuously in 
the whole country while the total size of cultivated land 
has increased at the community level. These land-use 
changes can reduce the overall soil organic carbon pool.
Similar deteriorations of the natural resource base have 
occurred in forest and shrub areas. Not only has there 
been a conversion of forest land into farmland and 
pasture land, but the biomass coverage of the available 
community forest and shrubland has also been 
declining through animal grazing and deforestation. 
Most communal areas that had been covered with 
endogenous forest trees and shrubs have been 
converted into grazing land, cropland, and settlement 
places. As a result, the forest coverage around the 
crop–livestock-producing areas has been declining over 
the past decades. Although some efforts have been 
made to maintain, protect, and develop community 
forests, success has remained limited. For farmers to 
plant trees on their cropland, the costs of protecting 
the seedlings from animals during the dry season are 
high. Thus, farmers plant trees for the production of 
wood only around the residential area. But, eucalyptus 
tree is nearly the only tree species planted around 
the backyard. One good culture in both study areas is 
that, unlike most other areas of the country, farmers 
maintain a few endogenous trees such as bisana on 
their cropland. They use these trees to fix nitrogen, to 
produce wood, and to use as shade for animals and 
humans.
In general, the crop production system fails to maintain 
or enhance soil carbon in the cropland, grazing land, 
and other land uses. But, some changes have occurred 
in recent times that have the potential to reverse these 
trends. One important change is the expansion of 
electricity in the towns and some rural areas in the past 
two decades. Electrification can potentially reduce the 
use of straw and animal dung as a source of fire energy. 
For this change to impact the amount of biomass that 
returns to the soil, the amount of straw that will be free 
from this should be larger than the amount of crop 
residue required for other competing demands: animal 
feed and construction purposes. To what extent such 
changes have affected the amount of crop residue that 
is returned to the soil is not clear. Furthermore, the 
positive impacts of such changes could be offset by 
the increased exploitation of the natural resource base 
caused by the increasing pressure.
9Working Paper
1.3.2 Land tenure
The constitution of the country states that all lands 
are the property of the state and the people. With 
the exception of residential compounds and cropping 
lands, the rest of the lands are owned by the state and 
the people. Farmers have the right to use and transfer 
the residential and cropping lands that they own. But, 
they have no right to sell or to use the land as collateral 
against any obligations. In addition, their use rights are 
not even secured because the land can be expropriated 
and redistributed by the government. Such arbitrary 
expropriation and redistribution by the government 
occurred on a massive scale twice in the past. But, 
expropriations and redistributions also take place at the 
individual level under the guise of public interest, rules, 
and regulations. 
The literature shows that an insecure property rights 
regime fails to incentivize farmers to invest in long-
term land improvement practices. Furthermore, the 
restriction on the ability to use the land as collateral 
tends to limit farmers’ ability to invest in improving the 
land by limiting their access to finance. The adverse 
effects of insecure property rights on efficiency and 
long-term investments in land improvement are widely 
documented (Ambaye 2015; Grover and Temesgen 2006; 
Deininger et al. 2003). Recent evidence also suggests 
that such property rights in the long run discourage 
farmers’ mobility from agriculture to other sectors and 
hence constrain potential agricultural transformations 
(Emran and Shilpi 2015a, 2015b; Giles and Mu 2014). 
In addition, a recent study by Ali et al. (2015), using 
a choice experiment, found that the preference of 
farmers to have larger family size declines when private 
ownership of land is secure and distributions are 
made only within the household. But, apart from the 
effects of the type of land ownership on the production 
decision of households, the land property rights regime 
could have considerable effects on the evolution of the 
socioeconomic landscape of communities. 
Apart from the treatment of expropriations, the 
socioeconomic context of communities may matter in 
determining the scope of the use of rights. In addition 
to formal rules, local norms and local socioeconomic 
contexts can determine the scope of use rights: the 
exhaustiveness and freeness of the use rights and 
the ability to exclude others from using the land. As 
described above, the specific production context in the 
study areas limits the farmers’ rights to use the land 
for crops of their choice and the ability of farmers to 
exclude others from using the land for open grazing. 
This aspect of the land property rights has especially 
important implications for improved land management 
programs and SOC-enhancing technology adoption.
1.4.  Adoption of carbon-
enhancing technologies
Agricultural activities (e.g., deforestation, burning, 
plowing, intensive grazing, intensive cultivation, 
continuous cropping, and poor residue management) 
usually contribute considerably to the atmospheric 
pool by depleting the SOC stock (Post and Mann 
1990). But, the extent of this depletion depends on the 
specific production system. For instance, the amount 
of crop residue that returns to the soil depends on the 
production system. The production systems in the study 
areas, at both the individual and community levels, are 
generally exploitative in that they deplete the existing 
SOC stock. Evidence from the literature suggests diverse 
SOC-enhancing practices that can increase or maintain 
SOC without compromising the ability to meet the 
growing food demand and the ability of producers to 
improve their livelihoods (Lal 2011) and also without 
impairing the ability of future generations to meet their 
food demand (Hobbs 2006; Schahczenski and Hill 2009). 
Agricultural scientists propose diverse improved 
production and natural resource management practices 
that can increase productivity and at the same time 
maintain or increase SOC (e.g., Pretty et al. 2006; Lal 
2006). For instance, Lal (2006) estimated the yield gains 
in kilograms per hectare for every 1 Mg/ha increase 
in SOC pool in the root zone to be 20–70 for wheat, 
10–50 for rice, and 30–300 for maize. Efforts have been 
made for about half a century to improve the natural 
resource base of soil, water, and plants. The improved 
SOC-enhancing measures promoted and implemented 
in the country include construction of soil conservation 
structures such as soil/stone bunds, check dams, 
diversion canals, and interception ditches; agroforestry 
practices such as alley cropping, planting of trees, 
grass strips, and hedges; and the implementation of 
improved and traditional agronomic practices such as 
contour tillage, strip planting, manuring, mulching, crop 
rotation, and, recently, conservation tillage (zero or 
minimum tillage). In addition, efforts have been made 
to increase forest coverage and to improve the supply 
of forage through various individual-level technologies. 
We classify the practices that have been promoted 
and implemented in croplands into three categories: 
physical soil and water conservation, agroforestry, and 
improved agronomic practices. 
However, agricultural scientists claim that these 
SOC-enhancing practices also enhance agricultural 
productivity. The effect of SOC enhancement on 
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia10
productivity is more important especially in the 
tropics than in the temperate zone due to the high 
mineralization rate in the former (Woomer et al. 
1994). The positive contribution of adopting SOC-
enhancing practices to productivity is theoretically 
well-established. But farmers, in their decision to adopt 
technologies, take into account many other factors in 
addition to productivity. The potential productivity gain 
from technology is just one factor among the many 
that determine the adoption of a practice by farmers. 
Especially in mixed crop-livestock production systems, 
in addition to crop productivity, the effects of the 
practice on farming operations, livestock production, 
and demand for feed and firewood also enter into the 
household’s decision to adopt.
Unfortunately, the most feasible SOC-enhancing 
technologies/practices also involve some costs to 
households. For instance, leaving crop residue in 
the plot can increase productivity in the subsequent 
production periods but can also involve some costs 
if the crop residue has alternative uses such as for 
livestock feed, fuelwood, construction material, etc. 
The farmer will adopt the SOC-enhancing technology 
if the additional gains of the technology in terms of 
increased productivity exceed the foregone benefits 
of the specific technology. The magnitude of these 
costs and benefits depends on the specific production 
context. The farmer is expected to adopt a practice by 
equating the marginal benefits with the marginal costs. 
Since the contribution of soil-enhancing technologies 
to productivity involves a time lag (Antle and Diagana 
2003), farmers optimize by equating the discounted 
marginal benefits with the discounted marginal costs 
of the technology. The magnitudes of these discounted 
marginal costs and marginal benefits and the discount 
rate depend on the specific production context. The 
implication is that the willingness of farmers to invest 
in SOC-enhancing technologies/practices depends on 
the time preference of farmers. In subsistence farming 
communities where the sustenance of the household 
depends on current production, the time preference 
of farmers for increased production could be high. 
Furthermore, time preference tends to be higher if 
the property is insecure. The likelihood of farmers to 
realize future benefits depends on the likelihood that 
farmers secure the right to reap the fruits of the land 
until future time t. In effect, the discount rates will be 
higher when user rights are not secure. Furthermore, 
the intertemporal flow of benefits and costs matters. 
Given the production contexts discussed above, 
leaving crop residues in plots involves a lot of costs 
in livestock production, fuelwood, and supply of 
construction materials. In the absence of a well-
functioning capital market, these costs have to be 
borne fully, undiscounted. But, the productivity gains 
are not only long term but also they cannot be realized 
without incurring a lot of costs to protect the residues. 
Since farmers’ plots are scattered and unfenced, crop 
residues could be grazed by animals and even collected 
by other farmers. The same is true with the use of grass 
strips. Unless farmers take collective actions to entirely 
exclude animals throughout the year, the enforcement 
costs for an individual farmer tend to be too high to 
justify the practice. On top of all of this, the sequestered 
carbon can be released back into the atmosphere within 
a short period of time if farmers revert to conventional 
practices (Antle and Diagana 2003). This implies that 
realizing future benefits requires farmers to adopt a 
practice on a permanent basis.
These discussions are based on the assumption that 
the farmers perfectly know the future benefits. Setting 
aside the inherent uncertainties associated with future 
benefits, farmers’ expectations of these benefits 
depend on their knowledge about the effects of the 
improved practice on future productivity. In a situation 
in which farmers lack scientific knowledge about soil 
dynamics, they are likely to underestimate the future 
benefits.
Furthermore, the existing overall production and 
institutional contexts of the communities can 
encourage/discourage the adoption of one technology 
in favor of another. Some contexts provide individual 
farmers with little incentive to implement SOC-
enhancing technologies. In other cases, this can even be 
discouraged by imposing constraints on individuals. In 
other words, some contexts expand the feasible set of 
SOC-enhancing practices while others narrow it.
These discussions show that farmers’ decision to adopt 
SOC-enhancing technologies is a function of complex 
household and technical factors. While household 
and plot characteristics determine the type, level, 
and extent of adoption by farmers within a given 
homogeneous socioeconomic and agroecological 
context, differences in these contexts between 
communities also cause the farmers to behave 
differently. This means that while household and plot 
characteristics explain the differences in adoption 
among households within a community, community-
level factors explain the differences in adoption 
between communities.
One important issue that poses a serious challenge 
in our analysis is that, unlike SOC-enhancing 
practices related to agronomic and agroforestry 
practices, soil and water conservation structures 
11Working Paper
have been implemented by mobilizing all farmers 
of the community, through a campaign and without 
consultation and consent of the individual farm owners. 
The problems of soil erosion and water loss involve a 
lot of externalities. The effectiveness of any soil and 
water conservation practice implemented on a given 
plot is affected by the erosion problem in upstream 
plots, and the action can also cause erosion problems 
in downstream plots. Thus, effective conservation of 
soil and water technically requires collective actions 
along the watershed. This alone may not pose a serious 
challenge if the implementation of the practices was 
made through consultation with every individual farmer 
and with the participation of farmers that have plots in 
the area. But, in practice, the structures are constructed 
by mobilizing all farmers in the community, including 
those that have no plot of land in the conservation area. 
Thus, practices implemented in such a manner are 
hardly considered as adoption.
The preceding discussion shows that farmers’ 
knowledge of future benefits, time preference, the 
magnitude of costs, access to a well-functioning capital 
market, level of externalities, the continuity of the 
practice, and, more importantly, the enforcement 
costs all matter for farmers to adopt carbon-enhancing 
technologies. Thus, the question is: which of these 
practices were adopted by the farmers and what factors 
determined their adoption? Our interest is thus to 
measure the level and extent of adoption of the three 
classes of SOC-enhancing technologies and identify the 
factors that determine their adoption.
1.5.  Conceptual framework
Natural resources (soil, water, minerals), climate 
(temperature, rainfall, humidity), topography, and 
geographic location form the biophysical environment 
that determines the suitable plants and animals that 
can thrive in an area. Access of the community to rural 
and urban markets (product, input, factor, and financial 
markets); access to technology, extension, education, 
health, irrigation, transportation, communication, 
electricity, and other services; demographic pattern; 
culture and religion; institutional environment 
(norms, rules, codes of conduct), etc., form the overall 
socioeconomic and institutional environment. The 
interplay between the sets of socioeconomic and 
institutional contexts and biophysical factors gives the 
community a unique characteristic. This, coupled with 
external factors, determines the farming system of 
the community. It determines the types, patterns, and 
modes of crop and livestock produced, the productivity 
levels, the land-use pattern, other off-farm activities, and 
the feasible set of technologies adopted in the various 
enterprises. 
The overall community context gives a common 
opportunity and constraint to all members of the 
community. Households make production and 
consumption decisions that maximize their utility, given 
their resource constraints. Household characteristics 
such as age, education level, family size, landholding, 
wealth level, labor supply, livestock number, etc., 
determine the production and consumption pattern of 
the household. The level and extent of adoption of SOC-
enhancing technologies is thus a function of community 
characteristics and household characteristics. The 
aggregate level and extent of adoption of the SOC-
enhancing practices will in turn influence the overall 
community context. These relationships can be changed 
over time due to exogenous factors such as  
climate change.
The farming system and 
feasible technology sets
Household-specific 
characteristics
Adoption decision for soil 
organic carbon-enhancing 
technologies
COMMUNITY-LEVEL CHARACTERISTICS
Socieconomic  
and institucional
characteristics
Biophysical
characteristics
Climate
change
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia12
2.  The data and  
analytical model
2.1.  About the data
This study used both primary and secondary data. 
Primary data were collected from 160 sample households 
selected from the two watershed sites. Moreover, 
community-level data that could give an overall picture 
about the overall agroecological, socioeconomic, 
institutional, and infrastructural aspects were also 
collected from district offices of the Agricultural and 
Natural Resource Development and village-level 
development agents. In addition, results of focus group 
discussions were used to further describe and validate 
the results of the household survey. Secondary data 
were collected from diverse sources: empirical studies, 
statistical abstracts, and various reports.
2.2. Sampling design
The study used data collected from households drawn 
from two watersheds. Yiser watershed is located in Bure 
Damot Woreda, Western Gojjam Administrative Zone, in 
Amhara region, while Azugashube watershed is located in 
Woreda, Kembata and Wolaita Administrative Zone, in the 
Southern region. The central aim of the sampling design 
was to obtain households that were representative of 
the various household groups and landscapes of the 
Photo: CIAT/Georgina Smith
watershed. To obtain representative landscapes of 
the watershed, the Kebele Administrations (KAs), the 
lowest administrative structure, in the watershed were 
stratified into three zones: upper, middle, and bottom 
of the watersheds. Accordingly, on the upper side, two 
KAs from the middle and one KA from the bottom side 
of the watershed were randomly selected from the 
existing KAs in each stratum. Accordingly, households 
that had a plot of land within the watershed in the four 
selected KAs were used as the sampling frame. Using 
this sampling frame, in the second stage, the number 
of sample households in each pastoral area (PA) was 
allocated proportionally to the number of households 
in the PA. Accordingly, 161 households in the four KAs 
in Azugashube and 218 households in the four KAs in 
Yiser were drawn randomly from the PA household 
roster, which gave a total sample of 379 households. In 
order to avoid missing responses, only those households 
that had a plot of land in the watershed were included 
in the sampling. When the selected household was not 
accessible, it was replaced by the next household on 
the list. Table 1 shows the distribution of the sample 
households across the watersheds and KAs.
13Working Paper
2.3  Methods of collection
Primary data from sample households were collected 
through an interview using a structured questionnaire. 
The questionnaire focused on key household 
characteristics that were expected to influence the 
adoption of SOC-enhancing practices and included 
detailed questions about these various practices. A short 
guideline about the questions and interview procedure 
was prepared and given to the enumerators in hard copy. 
In addition, community data about the socioeconomic, 
institutional, infrastructural, and agroecological variables 
were collected at the PA level. The data were collected in 
one round. Six enumerators were recruited and trained 
at each watershed site. The enumerators were trained 
for one full day about the household and community 
questionnaire, potential difficulties they might encounter, 
and common mistakes committed by enumerators and 
the cautions they needed to take during data collection. 
The data were collected with Table using SurveyCTO 
software. The training was also given with lab exercises 
on the application of SurveyCTO. Close supervision was 
carried out using two supervisors at each watershed site. 
The collected data were checked by the supervisor on a 
daily basis and appropriate correction measures were 
taken on-site.
2.4.  Analytical model
Descriptive statistics such as percentage, mean, standard 
deviation, and, more importantly, test statistics such as 
the t-test (for continuous variables) and chi-square test 
(for categorical variables) were used to gain an overall 
picture of the socioeconomic, biophysical, institutional, 
and infrastructural contexts about the sample 
households and the watershed. Furthermore, descriptive 
analyses were made for the various SOC-enhancing 
technologies/practices and comparisons were made 
between the two watershed sites. 
With regard to analytical econometric models, the 
primary purpose was to identify the factors that 
explained households’ decision to adopt SOC-enhancing 
technologies/practices and the intensity of their adoption. 
For convenience, we divided the SOC-enhancing 
technologies into three categories: (i) physical soil and 
water conservation, (ii) agronomic practices, and  
(iii) agroforestry practices. 
2.4.1 Multivariate probit model
Since the different carbon-enhancing practices are not 
mutually exclusive, farmers can adopt one or more of the 
three classes of SOC-enhancing practices included in the 
WOREDA/DISTRICT KA N %
Azugashube
Ambercho Wasera 47 12.4
Bondenna 46 12.1
Bucha 23 6.1
Gerba Findide 45 11.9
Subtotal 161 42.5
Yiser
Gulim 51 13.5
Jib Gedel 69 18.2
Tengeha 44 11.6
Wadra 54 14.2
Subtotal 218 57.5
Grand total 379 100.0
Table 1 Distribution of sample households in the two watersheds.
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia14
questionnaire. Some studies used a multivariate probit 
model to capture the interdependence in the adoption 
decisions between SOC-enhancing practices and soil and 
water management practices (e.g., Adusumilli and Wang 
2018; Aryal et al. 2018). A multivariate probit model is an 
extension of the univariate probit model. Cappellari and 
Jenkins (2003) specified the general equation that uses a 
simulation method to maximum likelihood estimation as 
shown in Eq. 1.
Eq. 1
where εim, m = 1, ..., and M are error terms distributed 
as multivariate normal, each with a mean of zero, and 
variance-covariance matrix V, where V has values of 1  
on the leading diagonal and correlations ρjk=ρkj as  
off-diagonal elements.
The observed dichotomous outcomes are as shown in  
Eq. 2
Eq. 2
and Ω, where with Kik = 2yik − 1 for each i, k = 1, ..., 3. 
Matrix Ω has constituent elements Ωjk (Eq. 5), where
Eq. 3
Eq. 4
Eq. 5
The yim might represent outcomes for M different 
choices at the same point in time: in our case, whether 
an individual adopts each of M different SOC-enhancing 
practices.
In the trivariate probit case, the log-likelihood function for 
a sample of N independent observations is given by Eq. 3.
where wi is an optional weight for observation i = 1, ..., 
N, and Ф3(.) is the trivariate standard normal distribution 
with arguments µi  (Eq. 4).
Clearly, the log-likelihood function depends on the 
trivariate standard normal distribution function Ф3(.). For 
evaluating multivariate normal distribution functions, 
the smooth recursive conditioning simulation method 
provided by Geweke–Hajivassiliou–Keane (GHK) is most 
popular (Cappellari and Jenkins 2003).
2.4.2 OLS regression model 
The models shown in Eq. 1 to Eq. 5 are used to identify 
the factors that determine the adoption decisions 
regarding SOC-enhancing practices. But, sample 
households also differ in the intensity of their adoption. 
To identify the factors that determine the intensity of 
adoption, we estimate OLS regression using the area 
of land covered with one or more of the three SOC-
enhancing practices as a dependent variable as shown  
in Eq 6.
Eq. 6
where y is the area in hectares covered with one or 
more of the three classes of SOC-enhancing practices; 
Xk represents the explanatory variables of household, 
plot, and community characteristics; u is the error term 
assumed to be normally distributed with mean μ and 
σ; and βk' represents vectors of the parameters to be 
estimated.
15Working Paper
2.5.  Description of dependent  
and independent variables
Dependent variables
The study classified the different SOC-enhancing practices 
into three categories: soil and water conservation 
practices, agronomic practices, and agroforestry 
practices. The study used a different dependent 
variable for each model. For the multivariate model, the 
dependent variable yi takes a value of 1 if the farmer 
adopts one or more of the ith SOC-enhancing practices 
and 0 otherwise. But, for the OLS regression, the study 
measures y in terms of the total land area covered with 
one or more of the three classes of SOC-enhancing 
practices in the 2017 production season.
Independent variables
Studies suggest the contextual nature of land 
degradation, SOC stock, and intervention responses 
to improve them. Thus, spatial, temporal, economic, 
environmental, and cultural contexts matter (Warren 
2002). This study hypothesizes that diverse factors 
composing plot-level (biophysical and agroecological), 
household-level (institutional and farming system), and 
community-level (watershed) characteristics explain 
the probability of adoption and intensity of adoption of 
SOC-enhancing practices. In the absence of argument 
to the contrary, we assume that the same variables that 
determine the probability of adoption also determine the 
intensity of adoption of these practices.
VARIABLES ASSOCIATED WITH  
AGROECOLOGICAL ASPECTS
Watershed dummy: 1 if Yesir and 0 if Azugashube.
Fertile land: the ratio of fertile land in hectares to 
the total area of land operated by the farmer.
Slope: the ratio of sloping land in hectares to the 
total area of land operated by the farmer.
Clay: the ratio of land in hectares with clay soil to 
the total area of land operated by the farmer.
Erosion: the ratio of land in hectares with erosion 
problems to the total area of land operated by  
the farmer.
VARIABLES ASSOCIATED  
WITH THE FARMING SYSTEM
Crop diversification: the number of crop types 
grown by the farmer during the production 
season.
Land fragmentation: the total time it takes for 
the farmer to reach all the plots.
Fertilizer: the amount of fertilizer in kilograms 
divided by the total area.
Residue: measures the proportion of residue 
that was left in the plot expressed in terms  
of percent.
SOCIOECONOMIC CHARACTERISTICS
Family size: the number of people that live in the 
household.
Area: total area of land in hectares operated by 
the farmer during the production season.
TLU: Total Livestock Units derived using the 
conversion factor presented (see Appendix).
Education: the sum of grade levels attained by 
household members who participate in farming 
activities.
Radio: 1 if the household owns a radio and 0 
otherwise.
ACCESS TO INSTITUTIONS
Rural market: the distance in walking minutes 
from the farmer’s home.
Extension: 1 if the household has had access to 
extension and 0 otherwise.
Credit: 1 if the household has had access to 
credit and 0 otherwise.
Farm-org: 1 if the household was a member of a 
farmers’ organization and 0 otherwise.
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia16
Photo: CIAT/Georgina Smith
3.  Discussion of results
This part presents and discusses the results of the descriptive and economic analyses. The first subsection describes 
the overall socioeconomic characteristics of the sample households and the adoption of the various SOC-enhancing 
practices in detail. The next subsection presents and discusses the results of the econometric analyses.
3.1.  Description of socioeconomic characteristics of sample households
From the total sample of 379 households, 89.5% and 93.2% of the sample households are males in Yiser and Azugashube, 
respectively (Table 2). The small proportions of female sample households reflect the reality that, except for a few, most 
households are headed by a male and there is no significant difference between the two watershed sites. With regard 
to education level, 56.9% and 70.2% of the sample households are literate in Yiser and Azugashube, respectively. The 
evidence suggests that literacy is higher in Azugashube than in Yiser and the difference is significant at 1%. Similarly, 
81.2% of the household heads participate in labor activity in Yiser and almost all household heads (98.8%) participate in 
labor activity in Azugashube and the difference is significant at the 1% level.
INDICATORS DUMMY
YISER AZUGASHUBE TOTAL
Chi-2
N % N % N %
Sex
Males 195 89.5 150 93.2 345 91.0 1.57
Females 23 10.6 11 6.8 34 9.0  
Literacy
Literate 124 56.9 113 70.2 237 62.5 7.00***
Illiterate 94 43.1 48 29.8 142 37.5  
Labor 
participation
Yes 177 81.2 159 98.8 336 88.7 28.41***
No 41 18.8 2 1.2 43 11.4  
Table 2 Demographic characteristics of sample household heads at the two sites.
17Working Paper
Table 3 Mean comparison of household characteristics between the two sites.
With regard to other demographic and socioeconomic 
characteristics of the sample households, the 
descriptive results show significant differences between 
the two watershed sites in almost all household 
characteristics except age. As shown in Table 3, the 
mean age of households at both sites is around the age 
of 45 years and no evidence of age difference was found 
between the two sites. But, significant differences were 
found at the 1% level for all the rest of the household 
characteristics (see details in Table 3). The average family 
size of the sample households in Yiser and Azugashube 
is 8.0 and 10.6, respectively, and the difference is 
statistically significant at the 1% level. Similarly, the 
level of participation of household members in labor 
activities is 2.7 and 5.2, respectively.
The extent of education attendance and the education 
level of sample households in Azugashube is better 
in terms of both the number of educated household 
members and the aggregate grade level attained by the 
family. While on average 6.1 household members gained 
some education in Azugashube, this was only 3.8 members 
in Yiser. The average aggregate grade level achieved by 
the household members as a whole was 29.2 years in 
Azugashube while it was only 10.5 years in Yiser. Although 
the family size of sample households in Azugashube 
is higher than in Yiser by 2.6 members on average, the 
aggregate grade of the households in Azugashube is about 
three times higher than in Yiser. This means that sample 
households in Azugashube allow children to continue their 
education to higher grade levels.
HOUSEHOLD ATTRIBUTES GROUP OBS MEAN STD. ERR. STD. DEV. T-RATIO
Age of household head
Yiser 218 44.8 0.8 11.9
-1.36
Azugashube 161 46.4 0.9 11.0
Family size
Yiser 218 8.0 0.2 3.3
-7.28***
Azugashube 161 10.6 0.3 3.7
Labor participation of hh members
Yiser 218 2.7 0.1 1.8
-12.63***
Azugashube 161 5.2 0.2 2.1
Number of educated hh members
Yiser 218 3.8 0.1 1.9
-11.04***
Azugashube 161 6.1 0.2 2.1
Education level of the family  
in terms of aggregate grade level
Yiser 218 10.5 0.8 11.5
-12.89***
Azugashube 161 29.2 1.3 16.7
hh stands for households
Measuring the wealth status of households in rural 
areas is a difficult task. Although nearly all the sample 
households in Yiser watershed live in houses roofed 
with iron sheets, only about half of the sample 
households (49%) live in such types of houses in 
Azugashube (Table 4). The remaining 51% of the sample 
households live in houses roofed with grass. In the 
past, almost all rural households lived in houses made 
of grass cover. But they gradually moved to houses 
made of iron sheets. Ownership of houses made of 
iron sheets was used as an important indicator of 
wealth in the past. But it becomes a common asset in 
some areas as all farmers replace their grass-roofed 
house with iron sheet. This is especially true in Yiser 
watershed. Some also question whether the move from 
a grass-roofed to iron sheet-roofed house is a response 
to the scarcity of the grass or a deliberate action to 
improve their house. In the sense of the former, the 
transition from houses roofed with grass to iron sheet 
can indicate a deterioration of the natural resource 
base. Thus, if it was not due to scarcity, the change 
would be favorable for SOC accumulation by reducing 
the demand for grass. Sample households’ ownership 
of a TV set is almost nil at both watershed sites. On the 
other hand, sample households’ ownership of a sofa set 
is significantly (at the 1% level) higher in Azugashube 
(19%) than in Yiser, although the vast majority of the 
sample households at both sites do not have a sofa set. 
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia18
VARIABLE
YISER AZUGASHUBE TOTAL
Chi-2
N (218) % N (161) % N (379) %
Iron sheet
Yes
 138.5***
No
TV
Yes
0.7
No
Sofa set
Yes
40.9***
No
Radio
Yes
7.5***
No
Toilet
Yes
3.2*
No
Motor pump
Yes
0.2
No
Motorcycle
Yes
2.9*
No
Horse cart
Yes
1.9
No
Mobile phone
>=3
13.3***
2
1
0
Table 4 Descriptive statistics of wealth indicators.
In contrast, about half of the sample households own a 
radio, with 55% of the sample households in Azugashube 
owning a radio vis-à-vis 41% in Yiser. 
With regard to ownership of a toilet, a majority of the 
sample households in both watersheds own one. This 
was not the case some decades ago in Ethiopia. Sample 
households’ ownership of a water pump, motorcycle, and 
horse cart is considerably low at both sites. Although this 
shows the general low wealth status of the population, 
it will not be used as a useful explanatory variable as 
there is no significant difference among the sample 
households at both sites. Finally, with the exception of 
a few, a majority of the sample households at both sites 
own at least one mobile phone although a significant 
difference exists between the two sites. Except in the 
ownership of iron sheet-roofed houses, the wealth status 
of sample households in Azugashube is higher than in 
Yiser. This may thus indicate that the complete transition 
from grass-roofed houses to iron sheet-roofed houses in 
Yiser is more as a response to the scarcity of grass than 
to improvements in wealth status of the households.
19Working Paper
In addition to households’ holdings of household 
furniture, their land and livestock holdings are perhaps 
more important indicators of wealth. As described in 
Table 5, the average livestock holding in Yiser was 3.7 
measured in Total Livestock Units (TLUs) and it was 2.9 
TLUs in Azugashube, and the difference was found to 
be statistically significant at the 1% level. Similarly, the 
average total land operated by sample households in 
Yiser and Azugashube was found to be 1.3 and 0.8 ha, 
respectively. The sample households not only differ 
significantly in terms of the total size of land they 
operated, they also differ in the proportion of land 
acquired through such informal land transactions as 
sharecropping, hiring, and borrowing as 30% and 10% of 
the land in Yiser and Azugashube was acquired through 
these means, respectively.
HOUSEHOLD ATTRIBUTES GROUP OBS MEAN STD. ERR. STD. DEV. T-RATIO
Livestock holdings in TLUs
Yiser 218 3.7 0.2 2.8
3.30***
Azugashube 161 2.9 0.1 1.4
Total iron sheet of the house
Yiser 218 79.9 4.0 59.0
6.13***
Azugashube 161 44.5 4.0 50.2
Total land in hectares
Yiser 218 1.3 0.1 1.1
4.98***
Azugashube 161 0.8 0.0 0.5
Ratio of land not owned
Yiser 217 0.3 0.0 0.5
4.69***
Azugashube 161 0.1 0.0 0.2
Table 5 Descriptive statistics of land and livestock holdings of sample households
The membership of households in a farmer organization 
is higher (63%) in Azugashube than in Yiser (54%) and 
the difference is significant at the 1% level (Table 6). 
For access to credit and extension services, sample 
households in Yiser have higher access (42% and 
82%, respectively) than in Azugashube (9% and 67%, 
respectively), and, in both cases, the differences are 
significant at the 1% level. This might be because not 
only the regional state of Amhara provides extensive 
input credits to farmers, but also the Amhara Credit 
and Saving Institution is the largest MFI in the country 
in terms of outreach and number of beneficiaries. The 
government is also committed to allocating from three 
to four extension agents in each KA composed of crop 
and animal production, natural resources, and veterinary 
experts. Moreover, the government established a 
large number of Farmers’ Training Centers (FTCs) in 
rural areas. In contrast, the Southern region, where 
Azugashube watershed is located, is weaker in delivering 
these services. The proportion of sample households 
with access to multiple sources (DA, FTC, GOs, and 
NGOs) is also larger in Yiser than in Azugashube. But, in 
absolute terms, access to extension services is generally 
good as more than half of the sample households at both 
sites have access to services. Photo: CIAT/Georgina Smith
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia20
VARIABLE
YISER AZUGASHUBE TOTAL
Chi-2
N % N % N %
Membership in farmers’ 
organization
Yes 137 62.8 66 41.0 203 53.6
7.30***
No 81 37.2 95 59.0 176 46.4
Access to credit
Yes 91 41.7 14 8.7 105 27.7
50.5***
No 127 58.3 147 91.3 274 72.3
Access to extension service
Yes 179 82.1 111 68.9 290 76.5
8.9***
No 39 17.9 50 31.1 89 23.5
Table 6 Description of sample households’ access to services.
Land fragmentation
Land fragmentation has a bearing on the performance 
of agricultural and natural conservation activities, 
although fragmentation may allow households to have 
land of different quality and hence enable them to 
diversify production. But it also poses a great challenge 
for farming and other operations. This is especially true 
when the land is scattered over wide areas. When the 
livestock grazing system is open, this makes undertaking 
timely farming operations difficult and protecting 
implemented SOC practices from cattle encroachment 
is also difficult and costly. In such contexts, protecting 
physical soil and water conservation structures such 
as bunds requires fencing, and activities such as the 
planting of grasses, trees, and mulching may require 
more than fencing the plots. In this regard, the average 
number of plots of sample households in Yiser and 
Azugashube was found to be 5.6 and 3.1, respectively 
(Table 7). Not only the level of fragmentation is 
significantly higher in Yiser than in Azugashube but also 
the plots are significantly more scattered in Yiser than 
in Azugashube as it takes 147.5 and 20.8 minutes for 
a farmer to reach all plots, respectively. Similarly, the 
extent of crop diversification was found to be 4.2 and 3.0 
in Yiser and Azugashube, respectively. However, these 
contexts adversely affect the overall implementation of 
SOC-enhancing practices. This poses great challenges, 
particularly for implementing SOC-enhancing practices 
related to increasing the surface biomass of the plots. 
For these reasons, we expect that the implementation 
of agroforestry and some agronomic SOC-enhancing 
practices will be especially less attractive or costlier in 
Yiser than in Azugashube.
HOUSEHOLD ATTRIBUTES GROUP OBS MEAN STD. ERR. STD. DEV. T-RATIO
Number of plots
Yiser 218 5.6 0.17 2.55
11.48***
Azugashube 161 3.1 0.10 1.24
Total distance to plots 
in minutes
Yiser 218 147.5 12.6 185.7
8.61***
Azugashube 161 20.8 1.6 20.8
Crop diversification in 
terms of number of 
crop types
Yiser 218 4.2 0.1 1.6
8.00***
Azugashube 161 3.0 0.1 1.1
Table 7 Descriptive statistics of indicators of land fragmentation.
21Working Paper
Implementation of SOC practices
An important objective of the sustainable management 
of soil resources is to increase the SOC pool by increasing 
the passive fraction. Soil surface management, 
soil water conservation and management, and soil 
fertility regulation are all important aspects of carbon 
sequestration in soil. The survey results show that 
different types of SOC-enhancing practices were 
implemented at both sites. We divided all SOC-enhancing 
practices into three classes: physical soil and water 
conservation, agronomic practices, and agroforestry 
practices. We found high differences in the number of 
farmers who implemented the different types of SOC-
enhancing practices between the two sites. The results 
show that the number of households that implemented 
physical soil and water conservation activities is more 
or less large at both sites but the frequency of adoption 
is higher in Yiser than in Azugashube (Table 8). Although 
stone/soil bunds are implemented more or less equally 
at both sites, other types of soil and water conservation 
structures such as the construction of fanya juu, check 
dams, diversion canals, and interception ditches are 
implemented only in Yiser. One of the limitations of 
these practices is that they are largely implemented 
through heavy pressure by the development agents and 
by mobilizing all farmers. They are also implemented 
through a campaign and sometimes without the 
willingness of the plot owner. Hence, not only the quality 
of the structures is mostly poor, but the farmers also 
rarely own the structures and protect and maintain them. 
The poor quality of the structures, their construction 
through government imposition, and the high costs of 
protecting the structures against livestock encroachment 
coupled with insecure property rights on land often make 
these structures less sustainable than others.
With regard to agronomic practices, they are mostly 
implemented only in Yiser watershed. On the contrary, 
the implementation of agroforestry practices is almost 
nonexistent at both sites, except for the high adoption 
rate of grass strips observed in Azugashube. The low 
adoption rate of agroforestry practices is mainly due to 
the open grazing system discussed earlier.
Important SOC-enhancing practices such as mulching, 
minimum or zero tillage, and strip cropping are almost 
nonexistent at both survey sites. Despite conservation 
tillage being recognized as an important instrument for 
increasing SOC content of the surface layer, it has not 
been adopted at both sites. 
One interesting traditional practice that was common 
in Yiser was traditional manuring called “degele.” Degele 
is a cooperative strategy among neighboring farmers 
to fertilize their cropping land. Farmers bring all their 
animals together and make a temporary barn on the 
cropland and let the animals stay the night there. On the 
next day, they shift the barn to the next quarter until 
the crop plot of the member is fully covered with the 
dung of the animals. After the cropland is fully covered 
with dung, they shift the barn to another member’s 
cropland and do the same thing. In this way, they cover 
the croplands of group members that are located not too 
far from the village. With this strategy, farmers do not 
need to transport the manure, but they can allow all the 
animal wastes to fertilize the land. This was an excellent 
cooperative strategy to fertilize their croplands with 
manure. Sadly, the farmers reported that they have now 
abandoned this strategy for many reasons: the scarcity 
of dung, declining numbers of livestock holdings, and, 
more importantly, the security problem. Farmers now 
keep their animals at home. Some farmers still fertilize 
their cropland using manure by transporting the dry or 
wet dung: either by scattering the dry dung or by burying 
the wet dung inside a pit.
Table 8 Participation level in the different SOC-enhancing technologies between the two sites.
SOC-ENHANCING ACTIVITIES
YISER AZUGASHUBE BOTH
N (218) % N (161) % N (379) %
Physical soil and water conservation activities
Soil/stone bunds 58 26.5 41 25.5 99 35.4
Terrace 109 49.8 31 19.3 140 50.0
Fanya juu terrace 43 19.6 - - 43 15.4
Check dam 12 5.5 - - 12 4.3
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia22
SOC-ENHANCING ACTIVITIES
YISER AZUGASHUBE BOTH
N (218) % N (161) % N (379) %
Diversion canal 59 26.9 1 0.6 60 21.4
Interception ditch 14 6.4 - - 14 5.0
Agronomic SOC-enhancing practices
Contour 115 52.5 5 3.1 120 42.9
Minimum or zero tillage 3 1.4 4 2.5 7 2.5
Crop rotation 115 52.5 1 0.6 116 41.4
Strip cropping 1 0.5 - - 1 0.4
Mulching 8 3.7 - - 8 2.9
Manuring 62 28.3 - - 62 22.1
Agroforestry SOC-enhancing practices
Agroforestry 5 2.3 - - 5 1.8
Grass strip 3 1.4 71 44.1 74 26.4
Alley cropping 5 2.3 - - 5 1.8
Hedge 1 0.5 - - 1 0.4
Others 2 0.9 3 1.9 5 1.8
Integration of the various SOC practices
As described in Table 9, of the total of nine types of SOC-enhancing practices, a majority of the sample households (64%) 
implemented one or two types of SOC-enhancing practices, of which about half implemented only one type of practice. 
The remaining 34.3% and 1.3% of the sample households implemented three to five and six to nine types of SOC-
enhancing practices, respectively. The difference in the distribution between the two watershed sites was significant at 
the 1% level.
Table 9 Distribution of implementation of the types of SOC-enhancing practices.
FREQUENCY OF FARMERS WHO IMPLEMENTED
YISER AZUGASHUBE TOTAL PEARSON 
Chi-2N % N % N %
One or two types of SOC practices 98 45.0 146 90.7 244 64.4
84.71***Three to five types of SOC practices 115 52.8 15 9.3 130 34.3
More than six types of SOC practices 5 2.3 0 0.0 5 1.3
Total 218 100.0 161 100 379 100
23Working Paper
The extent of implementation of SOC practices
The survey results show that farmers implemented these SOC-enhancing activities in their different plots. Thus, farmers 
differ in the extent of their implementation, although the frequencies of the number of plots in which one or more types 
of SOC practices are implemented are higher among sample farmers in Yiser than in Azugashube (Table 10).
Table 10 Distribution of households in terms of the number of plots in which they implemented SOC practices.
NUMBER OF PLOTS
YISER AZUGASHUBE TOTAL
N (218) % N (161) % N (379) %
At least one plot 9 4.1 61 37.9 70 18.5
Two or three plots 69 31.7 89 55.3 158 41.7
Four or five plots 86 39.5 10 6.2 96 25.3
Six plots or more 54 24.8 1 0.6 55 14.5
As mentioned above, the decline in the traditional 
manuring system pushed farmers to use inorganic 
fertilizer. As shown in Table 11, on average, sample 
farmers in Yiser and Azugashube were found to use 
560 kg and 202 kg of a combination of DAP, urea, and, 
recently, NPS fertilizer, respectively. The extent of 
fertilizer use between the two sites was found to be 
significant at the 1% level. With respect to the proportion 
of crop residue left in the plot, the results show that 
about 23% of the surface biomass is left in the plot at 
both sites. The remaining 77% is removed from the crop 
field and used for many purposes discussed earlier. 
But, due to the open grazing system discussed earlier, 
except for crops whose residues are not fed to animals 
such as maize stock, most portions of the residues that 
were left on the cropland during harvesting are browsed 
by animals. Thus, the proportion of crop residue that 
ultimately goes back to the soil is much lower than 23%.
HOUSEHOLD 
ATTRIBUTES
GROUP OBS MEAN STD. ERR. STD. DEV. T-RATIO
Fertilizer
Yiser 218 559.9 29.5 435.1
9.83***
Azugashube 161 202.4 14.1 178.7
Proportion of residue 
left in the plot
Yiser 218 23.6 1.9 27.9
0.09
Azugashube 161 23.3 1.2 15.7
Table 11 Descriptive statistics of fertilizer use and proportion of residue left in the field.
A majority of the farmers in Yiser use crop residues for multiple purposes such as cooking, feed, and construction 
purposes, with a small segment of the farmers using it for fertility improvements (Table 12). In contrast, sample farmers 
in Azugashube frequently use the residues for construction and feed with about 40 using it for cooking and fertility 
management. In each case, the differences between the two sites are strongly significant.
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia24
Table 12 Uses of crop residue for different purposes between the two sites.
RESIDUE USE
YISER AZUGASHUBE
Chi-2
N % N %
Cooking 179 82 68 42 64.86***
Feed 175 80 106 66 10.07***
Construction 160 73 152 94 28.10***
Fertility 42 19 69 43 24.89***
Finally, Table 13 shows the extent of implementation of 
the three categories of SOC-enhancing practices among 
sample households at the two sites. The results show 
that the frequency of implementation (in terms of the 
number of plots) of soil and water conservation and 
agronomic practices was found to be larger in Yiser 
than in Azugashube. On the contrary, the frequency of 
implementation of agroforestry practices was higher 
in Azugashube than in Yiser. The differences between 
the two sites in all three cases were highly significant 
because the physical soil and water conservation, 
agronomic, and agroforestry SOC practices were used 
in 3.90, 3.19, and 0.78 plots in Yiser. The corresponding 
figure was 1.10, 0.07, and 1.14 plots in Azugashube. 
Similarly, the extent of implementation (in terms of 
the percentage) of land covered with the three SOC-
enhancing practices between the two sites was also 
highly significant. This is because households in Yiser 
covered 32%, 25%, and 1% of their land with physical soil 
and water conservation, agronomic, and agroforestry 
SOC practices, respectively, while the corresponding 
numbers for sample households in Azugashube were 
22%, 3%, and 19%.
PARAMETER TYPES OF SOC PRACTICES GROUP OBS MEAN STD. ERR. STD. DEV. T-RATIO
Number of 
plots under
Physical soil and water conservation SOC practices
Yiser 218 3.90 0.22 3.31
10.27***
Azugashube 161 1.10 0.09 1.16
Agronomic SOC practices
Yiser 218 3.19 0.23 3.45
11.43***
Azugashube 161 0.07 0.02 0.31
Agroforestry SOC practices
Yiser 218 0.78 0.09 1.39
-2.82***
Azugashube 161 1.14 0.08 1.04
Area covered
Physical soil and water conservation SOC practices
Yiser 218 0.32 0.03 0.38
3.04***
Azugashube 161 0.22 0.02 0.22
Agronomic SOC practices
Yiser 218 0.25 0.02 0.30
8.79***
Azugashube 161 0.03 0.01 0.14
Agroforestry SOC practices
Yiser 218 0.01 0.00 0.06
-11.10***
Azugashube 161 0.19 0.02 0.24
Table 13 Extent of implementation of the three categories of SOC practices among sample households.
25Working Paper
Description of dependent and independent variables
The following section describes the dependent and independent variables used in the econometric analyses.  
Table 14 describes the categorical dependent and independent variables. The study classified the different SOC-
enhancing practices into three categories: soil and water conservation practices, agronomic practices, and agroforestry 
practices. From the total of 379 sample households, 80.7%, 38.3%, and 49.1% have adopted at least one practice from 
the three SOC-enhancing practices, respectively. 
Ownership of a radio apparatus is taken as a measure of a household’s access to information. From the total sample, 
47.0% of the sample households own a radio. The corresponding figures for access to extension and credit are 76.5% 
and 27.7%, respectively. Finally, 53.6% of the sample households are members of farm organizations.
Table 14 Descriptive statistics of categorical variables.
DEPENDENT AND INDEPENDENT VARIABLES N %
Adoption of soil and water conservation SOC practices
No 73 19.3
Yes 306 80.7
Adoption of agronomic SOC practices
No 234 61.7
Yes 145 38.3
Adoption of agroforestry SOC practices
No 193 50.9
Yes 186 49.1
Radio ownership
No 201 53.0
Yes 178 47.0
Extension contact
No 89 23.5
Yes 290 76.5
Credit access
No 274 72.3
Yes 105 27.7
Membership in farm organization
No 176 46.4
Yes 203 53.6
From the total land farmers operated in the production 
season, sample households on average covered 57.0% 
of their land with one or more of the SOC-enhancing 
practices (Table 15). In terms of agroecological aspects, 
on average, 77%, 26%, 31%, and 25% of a farmer’s land 
is fertile, clay soil, sloping, and erosion-prone. In terms 
of crop diversification, a farmer on average has grown 
3.7 crop types. As a measure of land fragmentation, on 
average, a farmer has to walk for 93.7 minutes in order to 
reach all of his/her plots. The average fertilizer application 
in the study area was found to be 419.62 kg/ha. For 
residue management, farmers reported that on average 
they leave 23% of the residue in the field. Farmers in the 
study area have a large family size in that on average they 
have 9.09 members. The average land size in the area 
is 1.08 hectare. The livestock holding in terms of Total 
Livestock Units reported was 3.31. In terms of education 
level, the grade levels of all household members, including 
the household head, were added as most family members 
participate in farming activities. Accordingly, on average, 
the education level of the household was found to be 
equivalent to the completion of grade 18.44. Finally, on 
average, a household travels 39.6 minutes to reach the 
nearest market.
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia26
VARIABLES MEAN SD MIN MAX
Dependent variable
Proportion of land covered with SOC practices 0.57 0.28 0 1
Agroecological variables
Proportion of fertile land 0.77 0.27 0 1
Proportion of clay land 0.26 0.38 0 1
Proportion of sloping land 0.31 0.43 0 1
Proportion of land with erosion problems 0.25 0.36 0 1
Farming system
Crop diversification by number of crop types 3.66 1.52 1 11
Land fragmentation in terms of distance to all plots 93.70 154.61 1 1470
Fertilizer rate (kg/ha) 419.62 296.33 0 2000
Residue proportion left on the land in percentage 23.19 23.98 0 100
Family size 9.09 3.72 2 23
Land area in hectares 1.08 0.95 0 8
Livestock ownership (TLU) 3.31 2.35 0 14
Education at the household level, sum of grades by the family 18.44 16.74 0 82
Distance from the market in walking minutes 39.60 29.65 1 180
Table 15 Descriptive statistics of continuous dependent and independent variables.
3.2   Econometric analysis
3.2.1  Determinants of level of adoption
The study used two models to identify the factors 
that explain the level and extent of adoption of SOC-
enhancing practices. We identify the factors that affect 
the probability of adoption of the three categories of 
SOC-enhancing practices using a multivariate probit 
model. Using the same explanatory variables, we also 
identify the factors that affect the extent of adoption 
of the three practices using OLS regression. We 
divided the explanatory variables into four categories: 
agroecological characteristics of the plots (watershed 
sites and plot characteristics such as fertility, slope, 
soil type, and perceived erosion problems), farming 
system-related variables (number of crops, land 
fragmentation, fertilizer use, and residue management), 
household socioeconomic characteristics (livestock 
holdings, family size, education, and landholding), 
and institutional access (access to rural markets, 
credit, social organization, and extension). 
The decisions to adopt alternative SOC-enhancing 
practices are mutually exclusive and, instead, they 
are interdependent (Adusumilli and Wang 2018). We 
thus expect interdependence in adoption decisions 
between the pair of each class of SOC-enhancing 
practices. Following Aryal et al. (2018) and Adusumilli 
and Wang (2018), the study used a multivariate probit 
model to capture possible interdependence between 
the adoption decision combinations of the three 
classes of SOC-enhancing practices: physical soil and 
water conservation, agronomic, and agroforestry. The 
maximum-likelihood estimates of the multivariate probit 
models with associated standard errors for the joint 
decisions of soil and water conservation and agronomic, 
agronomic and agroforestry, and soil and water 
conservation and agroforestry SOC-enhancing practices 
are given in Table 16.
From the estimation results, the null hypothesis that all 
the parameter estimates are simultaneously equal to 
zero is strongly rejected at the 1% level of significance. 
In addition, the null hypothesis that all rho (ρjk) 
27Working Paper
coefficients are simultaneously equal to zero (rho21 = 
rho31 = rho32 = 0) is strongly rejected at the 1% level. 
This means that the disturbance terms between the 
possible combinations of adoption decisions for the 
three SOC-enhancing practices are not independent. 
This indicates that the multivariate probit model is more 
appropriate than the standard univariate probit model. 
The coefficient ρ21, which measures the interdependence 
between the adoption decisions of soil and water 
conservation and agronomic SOC-enhancing practices, 
was found to be negative and significant at the 5% level. 
VARIABLES
PHYSICAL SOIL AND WATER  
SOC-ENHANCING PRACTICES
AGRONOMIC SOC-ENHANCING 
PRACTICES
AGROFORESTRY SOC-ENHANCING 
PRACTICES
Coef. Std. err. z Coef. Std. err. z Coef. Std. err. z
Agroecological aspects
Watershed dummy -0.43 0.18 -2.37*** -1.35 0.22 -6.19*** 0.10 0.17 0.59
Fertility 0.18 0.30 0.61 -0.12 0.32 -0.38 -0.31 0.28 -1.13
Slope 0.51 0.30 1.67* 0.37 0.30 1.24 0.49 0.28 1.76*
Clay 0.36 0.23 1.52 0.47 0.20 2.35** 0.85 0.20 4.20***
Erosion -0.08 0.30 -0.25 -0.13 0.28 -0.46 0.03 0.27 0.10
Farming system
Crop diversification 0.10 0.07 1.49 0.24 0.06 3.76*** 0.11 0.06 1.94**
Land fragmentation 0.00 0.00 -0.73 0.00 0.00 -1.57 -0.01 0.00 -4.19***
Fertilizer 0.00 0.00 1.96** 0.00 0.00 2.60*** 0.00 0.00 -2.04**
Residue -0.01 0.00 -2.40** 0.00 0.00 1.28 0.00 0.00 0.52
Socioeconomic status
Family size -0.02 0.03 -0.72 0.00 0.03 -0.03 0.04 0.02 1.46
Area 0.66 0.21 3.06*** 0.16 0.14 1.14 0.20 0.13 1.53
TLU -0.09 0.05 -1.79* -0.03 0.04 -0.61 0.01 0.04 0.30
Educ. level 0.01 0.01 1.33 0.01 0.01 1.04 0.00 0.01 0.60
Radio 0.59 0.18 3.24*** -0.13 0.17 -0.76 -0.04 0.16 -0.26
Access to institutions
Rural market 0.00 0.00 -0.71 0.01 0.00 3.08*** -0.01 0.00 -3.40***
Extension 0.36 0.21 1.76* -0.44 0.22 -2.00** -0.05 0.20 -0.25
Credit 0.10 0.22 0.47 0.12 0.19 0.66 0.13 0.19 0.71
Farm organization 0.14 0.20 0.72 -0.20 0.20 -1.00 -0.37 0.18 -2.11**
This means that it is not likely for a farmer who adopted 
soil and water conservation SOC-enhancing practices 
to also adopt agronomic practices and vice versa. 
Similarly, from the strongly significant (1% level) negative 
coefficient of ρ31, it is highly unlikely that a farmer who 
adopts soil and water conservation will also adopt 
agroforestry practices. On the contrary, from the positive 
coefficient of ρ32, it can be inferred that the likelihood 
that a farmer who adopted agronomic SOC-enhancing 
practices will also adopt agroforestry practices is 
strongly significant at the 1% level.
Table 16 Estimation results of a multivariate probit model.
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia28
VARIABLES
PHYSICAL SOIL AND WATER  
SOC-ENHANCING PRACTICES
AGRONOMIC SOC-ENHANCING 
PRACTICES
AGROFORESTRY SOC-ENHANCING 
PRACTICES
Coef. Std. err. z Coef. Std. err. z Coef. Std. err. z
rho21 -0.23 0.11 -1.99**     
rho31 -0.42 0.09 -4.48***     
rho32 0.47 0.09 5.1***     
LR test of rho21 =  
rho31 = rho32 = 0 
  31.83***       
Log-likelihood   -476.21     
Wald Chi-2 (54)   355.53***       
Consistent with Aryal et al. (2018), the study found that 
a combination of household, plot, and institutional and 
climatic factors determines the adoption of the different 
SOC practices. In the physical soil and water conservation 
enhancement practices equation, 8 variables out of 18 
were significant (3 variables at 1% level, 2 variables at 5% 
level, and 3 variables at 10% level). Land area operated, 
slope, fertilizer, extension, and radio ownership have 
a positive and significant effect on the likelihood of 
adopting soil and water conservation practices. On 
the contrary, watershed site, residue proportion, and 
livestock ownership have significant and negative effects 
on the likelihood of adopting soil and water conservation 
practices. Farmers that operate larger land area, have 
more proportion of sloping land, have a radio, use more 
fertilizer, and have extension contact are more likely 
to adopt soil conservation practices. On the contrary, 
farmers located at the Azugashube watershed site that 
left more proportion of residue in their plots and that 
own more livestock will be less likely to adopt soil and 
water conservation practices. Surprisingly, none of the 
aggregate plot characteristics, except slope, were found 
to be significant in explaining farmers’ adoption of soil 
conservation practices. These results are conditional on 
the adoption of the other two SOC-enhancing practices.
In the agronomic SOC-enhancing practices, six 
variables were significant (4 at 1% level and 2 at 5% 
level). The proportion of land with clay soil type, crop 
diversification, fertilizer application, and access to a 
rural market have positive and significant effects on the 
adoption of agronomic SOC-enhancing practices. On 
the contrary, watershed site and extension contact have 
negative effects on the adoption of these agronomic 
practices conditional on the adoption of the other two 
SOC-enhancing practices. No evidence was found on the 
effects of other variables considered in the analysis on 
the adoption of agronomic practices.
In the agroforestry SOC-enhancing practices, seven 
variables were significant (3 variables at 1% level, 3 
variables at 5% level, and 1 variable at 10% level). The 
proportion of sloping land, proportion of land with clay 
soil type, and crop diversification have positive effects 
on the adoption of agroforestry practices, whereas 
land fragmentation, amount of fertilizer, rural market, 
and membership in a farm organization have negative 
effects. Unlike the others, watershed site was to have an 
insignificant effect. Farmers that have a large proportion 
of sloping and clay land, that have grown more diverse 
crops, that have less fragmented land, that use lower 
fertilizer rates, that are near rural markets, and that 
are not members of farm organizations are more 
likely to adopt agroforestry practices. The negative 
effects of access to extension services on the adoption 
of agronomic practices and membership in farmers’ 
organizations on the adoption of agroforestry practices, 
respectively, are hard to explain and they are contrary to 
the theory. 
The dummy variable watershed site was found to be 
strongly significant in the adoption of soil and water 
conservation and agronomic practices, implying that 
community characteristics strongly explain the likelihood 
of adoption of the different types of SOC-enhancing 
practices. Many studies found similar results (Aryal et al. 
2018; Shiferaw and Holden 1998). The reason could be 
that the location captures differences in all other factors 
that are not captured by the rest of the explanatory 
variables. The strongly significant positive coefficients of 
the watershed site in the regressions of soil and water 
conservation and agronomic SOC-enhancing practices 
show that households in Yiser watershed are more likely 
to adopt these practices than those in Azugashube. 
No evidence was found on the effect of location on the 
adoption of agroforestry practices. 
29Working Paper
3.2.2  Determinants of the intensity of adoption
The above results show the likelihood of farmers to adopt one or more of the SOC-enhancing practices. But, the 
analyses were made in terms of whether the farmer implemented one or more of the practices or not regardless of 
the size of the land they were implemented on. It is therefore essential to assess the roles of these same explanatory 
variables in determining the magnitude of the adoption of the practices.
EXPLANATORY VARIABLES COEF. STD. ERR. T
Dependent variable: 
the ratio of land covered with SOC practice
Agroecological aspects
Watershed dummy 0.14 0.03 4.75***
Fertility 0.25 0.05 4.93***
Clay 0.22 0.05 4.46***
Slope 0.12 0.03 3.50***
Erosion -0.07 0.05 -1.53
Farming system
Crop diversification -0.02 0.01 -2.13***
Land fragmentation 0.00 0.00 -2.65***
Fertilizer 0.00 0.00 3.86***
Residue 0.00 0.00 -1.31
Socioeconomic status
Family size 0.01 0.00 1.86*
Area -0.01 0.02 -0.65
TLU -0.01 0.01 -0.94
Educ. level 0.00 0.00 -2.05**
Radio 0.04 0.03 1.39
Access to institutions
Rural market 0.00 0.00 3.05***
Extension 0.09 0.04 2.66***
Credit 0.05 0.03 1.49
Farm organization -0.05 0.03 -1.65*
Number of observations =  379
F (18, 361) =  107.9***
Adj. R-squared =  0.84
Table 17 OLS regression estimation results of implementation of all SOC-enhancing practices.
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia30
We measure the intensity of adoption in terms of the 
ratio of the area of land covered by one or more of the 
SOC practices to the total land operated by the farmer 
during the 2017 production season. Since all sample 
households implemented at least one of these practices, 
we have continuous values for the dependent variable. 
We thus fit the OLS regression model to identify variables 
that explain the intensity of adoption. With an adjusted 
R-squared value of 0.84, the model fits the data well. 
Moreover, the null hypothesis that the coefficients of all 
explanatory variables are simultaneously equal to zero 
is strongly rejected at the 1% level. The OLS estimation 
result is presented in Table 17.
The results show that 12 variables out of 18 significantly 
determine the intensity of implementation of SOC-
enhancing practices expressed in terms of the ratio of 
land area covered with one or more SOC-enhancing 
practices from the three classes to the total land 
operated by the sample household. Watershed site, 
fertility, clay, slope, fertilizer, family size, rural market, 
and extension were found to significantly and positively 
determine the intensity of adoption at the 1% level 
(except family size, which was significant at a marginal 
level of 10%). When controlling all other variables, the 
intensity of implementation of SOC-enhancing practices 
was significantly higher in Yiser than in Azugashube. 
With regard to plot characteristics, the intensity of 
adoption of those farmers with a higher ratio of fertile, 
sloping, and clay lands was found to be significantly 
higher. Although the significantly positive relationship 
between the intensity of implementation of SOC 
practices and the ratios of sloping lands and clay 
lands was consistent with our prior expectations, its 
relationship with the ratio of fertile lands was contrary to 
our prior expectations. This indicates that farmers prefer 
to prioritize maintaining the fertility of fertile lands 
rather than improving the fertility of less fertile lands. 
The fertilizer application rate was also found to increase 
the intensity of adoption of SOC-enhancing practices.
The results also suggest a positive effect of family size 
on the intensity of implementation of SOC-enhancing 
practices. Contrary to our expectations, the results 
showed positive and significant relationships between 
distances to rural markets and the intensity of adoption. 
Those farmers located far from rural markets were 
found to have a higher intensity of implementation than 
those located near rural markets. Access to extension 
service was also found to significantly and positively 
determine the intensity of adoption of SOC-enhancing 
practices. On the contrary, the results also suggest a 
negative relationship between membership in farmers’ 
organizations and the intensity of implementation of 
SOC-enhancing practices.
For farming system-related variables, land 
fragmentation, crop diversification, and education levels 
were found to significantly determine the intensity 
of adoption of SOC-enhancing practices negatively. 
When controlling access to extension service, the 
effects of education level, measured by the total grade 
level achieved by all those household members who 
participate in farming activities, in determining the 
intensity of implementation of SOC-enhancing practices 
were found to be negative at the 5% significance level.
Finally, the results found no evidence on the effects of 
perceived exposure to erosion problems, land area, TLU, 
residue, and credit on the intensity of implementation of 
SOC-enhancing practices. Consistent with the results of 
the multivariate probit model, no evidence was found on 
the role of perceived erosion problems in determining 
the intensity of implementation of SOC-enhancing 
practices. The important implication of these results is 
that farmers implemented SOC-enhancing practices not 
to reduce erosion problems but to meet other objectives.
31Working Paper
Conclusions
 This study aimed to identify the determinants 
of the level and intensity of adoption of SOC-
enhancing practices in the case of two watershed 
sites in Ethiopia: Yiser and Azugashube. The descriptive 
results reveal that the two watershed sites are significantly 
different in almost every aspect: socioeconomic, 
biophysical, farming system, and institutional. Moreover, 
the two sites were found to be significantly different in 
the level and intensity of adoption of SOC-enhancing 
practices. Significant differences were also found between 
the two sites in the adoption of the three classes of SOC-
enhancing practices. This indicates that, compared to the 
Azugashube watershed site, households in Yiser are more 
likely to adopt soil and water conservation and agronomic 
SOC-enhancing practices.
We assume that the decisions to adopt the three classes 
of SOC-enhancing practices are not mutually exclusive. In 
order to capture the potential interdependences between 
the adoption decisions of the three classes of practices, 
the study specified a multivariate probit model. The 
multivariate estimation reveals that interdependence 
exists between the decisions to adopt the three SOC-
enhancing practices. The results revealed that the decision 
to adopt soil and water conservation is negatively related 
to the decision to adopt both agronomic and agroforestry 
SOC-enhancing practices. But, the interdependence 
between the decisions to adopt agronomic and 
agroforestry practices is positive and significant at the 
1% level. That means that although the adoption of 
agroforestry SOC-enhancing practices is complementary 
with the adoption of agronomic practices, the adoption 
of soil and water conservation practices is competitive 
with the adoption of both agronomic and agroforestry 
practices.
The OLS regression estimated to identify the 
factors that determine the intensity of adoption 
found 12 variables out of 18 to significantly 
determine the intensity of implementation of SOC-
enhancing practices expressed in terms of the ratio 
of land area covered with one or more among the 14 
practices classified under the three classes to the total 
land operated by the sample household. Compared 
to the results of the probability of adoption, many 
explanatory variables were found to determine the 
intensity of adoption of SOC-enhancing practices in 
general. Watershed site, fertility, clay, slope, fertilizer, 
family size, rural market, and extension were found to 
significantly and positively determine the intensity of 
adoption at the 1% level (except family size, which was 
significant at a marginal level of 10%). When controlling 
all other variables, the intensity of implementation of 
SOC-enhancing practices was significantly higher in Yiser 
than in Azugashube. The implication is that encouraging 
Photo: CIAT/Georgina Smith
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia32
farmers to intensify their adoption requires more efforts 
than encouraging them to adopt one or more of the SOC-
enhancing practices.
Finally, in an attempt to check the motive for the adoption 
of SOC practices, we included the ratio of land exposed to 
soil erosion problems. In all the regressions, this variable 
was found to be consistently insignificant in determining 
both the probability to adopt and the intensity of adoption 
of SOC-enhancing practices. This implies that the adoption 
of these practices is not driven by the perceived threat 
of soil loss through erosion but by other factors. This 
result is not surprising in the case of soil and water 
conservation practices as they are implemented through 
a community-level campaign and by involving all farmers 
in the community and without consultation with individual 
owners of plots. But, the results are surprising when it 
comes to the adoption of agronomic and agroforestry 
practices that are implemented individually. 
33Working Paper
References
Adusumilli N; Want H. 2018. Analysis of soil management and water conservation practices adoption among crop and 
pasture farmers in humid-south of the United States. International Soil and Water Conservation Research 6:79–86.
Ali DA; Deininger K; Kemper N. 2015. Pronatal property rights over land and fertility outcomes: Evidence from a natural 
experiment in Ethiopia. World Bank, Policy Research Working Paper 7419.
Ambaye DW. 2015. Land rights and expropriation in Ethiopia. Doctoral Thesis accepted by the Royal Institute of 
Technology (KTH), Stockholm, Sweden.
Antle JM; Diagana B. 2003. Creating incentives for the adoption of sustainable agricultural practices in developing 
countries: The role of soil carbon sequestration. American Journal of Agricultural Economics 85(5):1178–1184. 
Aryal JP; Rahut DB; Maharjan S; Erenstein O. 2018. Factors affecting the adoption of multiple climate-smart agricultural 
practices in the Indo-Gangetic Plains of India. Natural Resources Forum 42:141-158.
Cappellari L; Jenkins SP. 2003. Multivariate probit regression using simulated maximum likelihood. The Stata Journal 
3(3):278–294.
CSA (Central Statistical Agency). 2000. Report on Area and Production of Crops (Private Peasant Holdings, Meher 
Season): Agricultural Sample Survey 1999/2000. Ethiopian Central Statistical Agency, Volume I, Statistical Bulletin 227.
CSA (Central Statistical Agency), 2010. Report on Area and Production of Crops (Private Peasant Holdings, Meher 
Season): Agricultural Sample Survey 2009/2010. Ethiopian Central Statistical Agency, Volume I, Statistical Bulletin 446.
DAWR (Department of Agriculture and Water Resources). 2016. Boosting farm productivity: Improved soils and reduced 
greenhouse gas emissions. Department of Agriculture and Water Resources, Canberra, March. CC BY 4.0.
Deininger K; Jin S; Adenw B; Gebre-Selassie S; Demeke M. 2003. Market and non-market transfers of land in Ethiopia: 
Implications for efficiency, equity, and non-farm development. World Bank. Washington, DC.
Emran MS; Shilpi F. 2015a. Land market restrictions, women’s labor force participation, and wages in a rural economy. 
World Bank, Policy Research Working Paper 7524.
Emran MS; Shilpi F. 2015b. Do land market restrictions hinder structural change in a rural economy? Evidence from Sri 
Lanka. World Bank, Policy Research Working Paper 7525.
FAO (Food and Agriculture Organization of the United Nations). 2017. Global symposium on soil organic carbon: 
Unlocking the potential of mitigating and adapting to a changing climate. 21-23 March 2017. FAO, Rome, Italy.
Follett RF. 2001. Soil management concepts and carbon sequestration in cropland soils. Soil and Tillage Research  
61:77-92.
Giles J; Mu R. 2014. Village political economy, land tenure insecurity, and the rural to urban migration decision: Evidence 
from China. World Bank, Policy Research Working Paper 7080.
Gledhill R; Herweijer C; Hamza-Goodacre D; Grant J; Webb C; Steege J. 2011. Agricultural carbon markets: Opportunities 
and challenges for sub-Saharan Africa. Sustainability and Climate Change, PwC LLP, UK.
Glover EK; Ahmed HB; Glover MK. 2013. Analysis of socio-economic conditions influencing adoption of agroforestry 
practices. International Journal of Agriculture and Forestry 3(4):178-184. 
Grover DK; Temesgen A. 2006. Enhancing land-use-efficiency through appropriate land policies in Ethiopia. Poster 
paper prepared for presentation at the International Association of Agricultural Economists Conference, Gold Coast, 
Australia, 12-18 August 2006.
Hardin G. 1968. The tragedy of the commons. Science, New Series 162(3859):1243-1248.
Heath J; Binswanger H. 1996. Natural resource degradation effects of poverty and population growth are largely policy-
induced: The case of Colombia. Environment and Development Economics 1:65–84.
Adoption of Soil Organic Carbon-Enhancing Practices: A case of two watershed sites in Ethiopia34
Hobbs PR. 2006. Conservation agriculture: What is it and why is it important for future sustainable food production? 
Paper Presented at International Workshop on Increasing Wheat Yield Potential, CIMMYT, Obregón, Mexico, 20-24 
March 2006.
Jack BK. 2011. Constraints on the adoption of agricultural technologies in developing countries. White paper, Agricultural 
Technology Adoption Initiative, J-PAL (MIT) and CEGA (UC Berkeley).
Kankwamba H; Mangisoni JH. 2015. Are sustainable agricultural practices improving output and incomes of smallholder 
farmers in Malawi? Contributed paper presented at the 29th International Conference of Agricultural Economists, 8-14 
August 2015, Milan, Italy.
Lal R. 2006. Enhancing crop yields in developing countries through the restoration of the soil organic carbon pool in 
agricultural lands. Land Degradation & Development 17:197–209.
Lal R. 2011. Sequestering carbon in soils of agro-ecosystems. Food Policy 36:S33-S39.
Lal R; Kimble JM. 1997. Conservation tillage for carbon sequestration. Nutrient Cycling in Agroecosystems 49:243–253.
Laxmi V; Mishra V. 2007. Factors affecting the adoption of resource conservation technology: Case of zero-tillage in rice-
wheat farming systems. Indian Journal of Agricultural Economics 62(1):126-138.
Manna MC; Wani SP; Rego TJ; Sahrawat KL; Bhattacharyya T; Ramesh V; Bandyopadhyay KK; Rupa TR;Singh P; Pathak 
P; Padmaja KV. 2008. Influence of different land-use management on soil biological properties and other C fractions 
under semi-arid benchmark soils of India. Global Theme on Agro-ecosystems Report no. 41. Patancheru 502 324, 
Andhra Pradesh, India: International Crops Research Institute for the Semi-Arid Tropics (ICRISAT). 
Melesse B. 2018. A review on factors affecting adoption of agricultural new technologies in Ethiopia. Journal of 
Agricultural Sciences and Food Research 9:226.
Paustian K; Six J; Elliott ET; Hunt HW. 2000. Management options for reducing CO2 emissions from agricultural soils. 
Biogeochemistry 48:147–163.
Post WM; Kwon KC. 2000. Soil carbon sequestration and land-use change: Processes and potential. Global Change 
Biology 6:317–328.
Post WM; Mann LK. 1990. Changes in soil organic carbon and nitrogen as a result of cultivation. In: Bouwman AE. (ed.), 
Soils and the greenhouse effect, Wiley & Sons, Chichester, U.K.
Pretty JN; Noble AD; Bossio D; Dixon J; Hine RE; Penning de Vries FWT; Morison JIL. 2006. Resource-conserving 
agriculture increases yields in developing countries. Environmental Science Technology 40(4):1114–1119.
Ringius L. 1999. Soil carbon sequestration and the CDM opportunities and challenges for Africa. Center for International 
Climate and Environmental Research – Oslo (CICERO) Report.
Schahczenski J; Hill H. 2009. Agriculture, Climate Change and Carbon Sequestration. NCAT, National Sustainable 
Agriculture Information Service, 1-800-346-9140, www.attra.ncat.org
Scherr SJ. 2000. A downward spiral? Research evidence on the relationship between poverty and natural resource 
degradation. Food Policy 25:479–498.
Shiferaw B; Holden ST. 1998. Resource degradation and adoption of land conservation technologies in the Ethiopian 
Highlands: A case study in Andit Tid, North Shewa. Agricultural Economics 18:233-247.
Warren A. 2002. Land degradation is contextual. Land Degradation and Development 13:449–459. DOI: 10.1002/ldr.532
Woomer PL; Martin A; Albrecht A; Resck DVS; Scharpenseal HW. 1994. The importance and management of soil organic 
matter in the tropics. In: Woomer PL; Swift MJ. (eds.). The biological management of tropical soil fertility, Wiley & Sons, 
Chichester, UK: 47–80.
35Working Paper
Appendix. Conversion of livestock number into TLU
LIVESTOCK TYPES TLU EQUIVALENT
Cattle: mature female, bulls, and oxen 1.00
Calves 0.75
Goats and sheep 0.40
Heifers 1-2 years 0.10
Donkeys 0.50
Camels 1.00
Horses and mules 0.80


Bioversity International and the International Center for
Tropical Agriculture (CIAT) are part of CGIAR, a global research
partnership for a food-secure future.
Bioversity International is the operating name of the
International Plant Genetic Resources Institute (IPGRI).
www.bioversityinternational.org
www.ciat.cgiar.org
www.cgiar.org
Alliance
Africa Hub
c/o icipe (International Centre of Insect
Physiology and Ecology)
Duduville Campus Off Kasarani Road
P.O. Box 823-00621, Nairobi, Kenya
Tel. (+254) 0709134000
Bioversity International and the International Center for
Tropical Agriculture (CIAT) are part of CGIAR, a global research
partnership for a food-secure future.
Bioversity International is the operating name of the
International Plant Genetic Resources Institute (IPGRI).
www.bioversityinternational.org
www.ciat.cgiar.org
www.cgiar.org
Alliance
Africa Hub
c/o icipe (International Centre of Insect
Physiology and Ecology)
Duduville Campus Off Kasarani Road
P.O. Box 823-00621, Nairobi, Kenya
Tel. (+254) 0709134000
Bioversity International and the International Center for 
Tropical Agriculture (CIAT) are part of CGIAR, a global research 
partnership for a food-secure future.
Bioversity International is the operating name of the 
International Plant Genetic Resources Institute (IPGRI).
Sub-regional Office for Africa
Pan-Africa Bea  Rese rch Alliance
National Agricultural Research 
Laboratories
. . x 6 47
Kampala, Uganda
www.bioversityinternational.org
www.ciat.cgiar.org
www.cgiar.org