You are on page 1of 12

Environmental Development 40 (2021) 100672

Contents lists available at ScienceDirect

Environmental Development
journal homepage: www.elsevier.com/locate/envdev

Community-embedded experiential learning and adoption of


conservation farming practices in Eastern and Southern Africa
Paswel P. Marenya a, *, Muhammed A. Usman b, Dil B. Rahut c
a
International Maize and Wheat Improvement Center (CIMMYT), ICRAF House, UN Avenue Gigiri, P.O. Box 1041-00621, Nairobi, Kenya
b
Center for Development Research (ZEF), Bonn University, Germany
c
Asian Development Bank Institute, Tokyo, Japan

A R T I C L E I N F O A B S T R A C T

Keywords: In low-income country agriculture, there is frequently limited information about the comparative
conservation Agriculture effectiveness of alternative extension modalities for delivering information on new farming
Eastern Africa methods. This paper analyzes the impact of community-embedded farmer education programs on
Southern Africa
the adoption of conservation farming. A multinomial choice model is applied on survey data from
Experiential learning
Smallholder farmers
Kenya, Malawi, Mozambique, and Tanzania to estimate the relative risk ratios (RRR), a measure
of the probability of adopting the conservation agriculture practices studied, conditional on
attending specific extension activities in the farmers’ community. The results show that the
likelihood of adoption was generally high as measured by RRRs between 10 and 14. Suggestive of
the effectiveness of experiential learning methods, attending training sessions had consistently
higher RRRs of adopting conservation farming practices. The positive role of community-based
extension programs potentially means that the dissemination of conservation agriculture re­
quires investments by governments and development organizations in community learning sites.
Investments in programs that promote experiential, hands-on and farmer self-learning can be
consequential in enabling the diffusion of better farming methods.

1. Background and introduction

Globally, particularly in sub-Saharan Africa (SSA), the need for sustainable agricultural intensification (SAI) is critical due to the
pressing need to feed the increasing population with minimal negative impacts on the environment. The implementation of SAI in­
volves methods, technologies, and practices that improve yields while preserving (or enhancing) the environment and limiting
cultivated land expansion (The Royal Society, 2009; Godfray et al., 2010). A related concept is that of climate-smart agriculture,
whereby productivity enhancements are achieved together with improvements in the adaptive capacity to – and mitigation of – climate
change (Food and Agriculture Organisation of the United Nations [FAO], 2013). These two concepts of SAI and CSA are underpinned
by conservation farming (CF), including physical (bunds) and biological terraces (live hedges) for soil, water harvesting and storage,
conservation agriculture; such as minimising tillage, mulching, crop rotations, or inter-crops. Although these practices are promising
in terms of positive environmental impacts such as conservation of soil moisture, reducing soil erosion and nutrient recyling, their

* Corresponding author.
E-mail addresses: p.marenya@cgiar.org (P.P. Marenya), mausman25@gmail.com (M.A. Usman), drahut@adbi.org (D.B. Rahut).

https://doi.org/10.1016/j.envdev.2021.100672
Received 28 December 2020; Received in revised form 1 August 2021; Accepted 2 September 2021
Available online 9 September 2021
2211-4645/© 2021 Elsevier B.V. All rights reserved.
P.P. Marenya et al. Environmental Development 40 (2021) 100672

adoption remains low (Hobbs et al., 2008; Erenstein et al., 2012; FAO, 2012).
Many factors are reported in the literature as drivers of adoption of CF methods.1 These include socio-demographic and economic
conditions, such as labor shortages at critical times for dealing with increased weeds (Nyamangara et al., 2014), and biophysical and
institutional factors, such as the lagged yield benefits, which, in some cases, discourage farmers with low discount rates (Pannell et al.,
2014) (see also Knowler and Bradshaw, 2007; Teklewold et al., 2013; Kassie et al., 2013; Bezu et al., 2014; Shiferaw et al., 2014;
Khonje et al., 2015; Marenya et al., 2017; Tambo and Mockshell, 2018). Therefore, although factors that drive adoption are multi­
faceted, lack of farmer education and capacity building to enable them evaluate the new technologies in a low-cost manner can be an
impediment to the adoption of conservation farming practices, even under the best market and infrastructure scenarios (De Janvry,
Marcours, et al., 2017).
This paper argues that despite the presence of constraining factors, considerable adoption can still occur by providing farmers with
low-cost means of evaluating new practices and technologies. Information costs can come in many forms: farmers have to spend time
seeking out information (Mulwa et al., 2017). Alternatively, to understand how to adapt to the new technology, farmers may also need
to experiment with it for a period time and on a sufficient scale to generate useful information (Kummer et al., 2017). Farmers may
need to repeat these processes in an iterative learning process (Milestad et al., 2012). Alternatively, they may adopt a strategic
wait-and-see attitude (Sadoulet, 2017).
To support agricultural technology scaling, ways of lowering these learning costs should be identified. A network of sites that
farmers can readily visit to observe and evaluate new technologies is one way of reducing these learning costs. However, there is
limited information on the comparative effectiveness of various information delivery modalities for driving adoption of new farming
innovations. Such information, retrieved from comparative analyses of different extension methods, is necessary to guide programs for
setting up farmer learning sites, training opportunities, and other extension approaches. This paper evaluates the impact of demon­
stration sites, field days, and training sessions in the adoption of various CF practices in four countries in Eastern and Southern Africa,
namely Kenya, Malawi, Mozambique, and Tanzania. The study is conducted by using a multinomial choice model to estimate the
relative risk (odds) ratios of adopting a set of relevant CF methods. We empirically test the impact of demonstrations, field visits, and
training sessions on the probability of adoption of the selected CF methods, and related practices, among smallholder farmers in the
four countries. The data came from a total of 3400 households operating 7178 maize plots and who had been exposed to a number of
CF promotions for varying lengths of time over a period of eight years. In summary, we find that all the extension methods are useful –
farmers attending field days and training sessions showed higher and significant relative risk ratios of adoption than when they
informally visited village-based demonstration sites.
The rest of this paper is organized as follows: the next section deals with the literature review, and data and methodology are
explained in section 3. Section 4 provides results and discussion, and finally, section 5 concludes the paper by highlighting key policy
implications.

2. Literature review

Knowledge delivery to agricultural communities is often a function of agricultural extension services. Although farmers may have
the resources for supporting the costs of new technology adoption, they might still fail to adopt the practices if they cannot implement
them correctly (and profitably) due to the lack of technical information and skills. One of the most effective ways of scaling up CF
practices for smallholders is to strengthen the ability of the extension systems to provide technical backstopping to farmers. Therefore,
we propose that a high-density network of trainers, extension personnel, and demonstration sites to facilitate the interface with farmers
at the local level be prioritized.
Yet in most parts of SSA, extension services have been weak for several decades due to many years of weak budget allocations
(Marenya et al., 2017). From the late 1970s to the mid-1980s, the developing country average of extension agent-to-farmer ratio was 1
in 300, which had declined to between 1 in 1500 and 1 in 3000 by 2012 (Pye-Smith, 2012). In recent years, governments in the region
have been trying to reverse those trends (Davis et al., 2010). Broader policy attention is now geared towards demand-driven extension
models, finding optimal mixes of farmer capacity building, complemented by better access to markets and the provision of inclusive
finance. Some policy documents in the last decade highlight the renewed interest to increase investments in extension in Kenya Re­
public of Kenya (2012), Tanzania (Daniel, 2013), and Malawi (Masangano and Mthinda, 2012).
Farmer capacity-building efforts will encompass guides and manuals suitable for extensionists, development practitioners, and
farmers (Sariah, 2019). This is necessary to create a critical mass of actors familiar with these new practices. The literature has reported
the positive impacts of conscious extension services on productivity and better food security among smallholder farmers (Owens et al.,
2003; Dercon et al., 2009; Cunguara and Moder, 2011; Abdoulaye et al., 2013; Pan et al., 2018). Extension services have been shown to
improve the function of other social networks, such as farmer group organization among rural households (Ainembabazi et al., 2017).
It is estimated that more than 80 per cent of the world’s extension services are publicly-funded, especially in low-income countries
(Feder et al., 2001). This is understandable given the non-commercial and public goods nature of these services. Nevertheless, the most
effective modalities for delivering extension services for farmers are not yet clear. A variety of approaches have been implemented to
try to make extension services more effective. Yet, the top-down nature of most of these services failed to consider the heterogeneous
needs of farmers as determined by the local market and agro-ecological conditions. Although individual consultation visits may sound

1
Henceforward, we use the term conservation farming practices to allow for the broader concept of environmentally positive farming practices,
that include conservation agriculture and other practices whose objectives are to conserve soil, water and nutrients.

2
P.P. Marenya et al. Environmental Development 40 (2021) 100672

ideal, this system is infeasible due to the considerable number of farmers to visit, and only a limited number of extension personnel are
available (FAO, 2010; Swanson and Rajalahti, 2010).
Training and visit (T&V) method, promoted by the World Bank in the 1980s and 1990s, was mostly implemented through the
public extension systems. The modalities for the T&V system consisted of making the extension services more professional through
continuous training and farmer engagement by means of regular farm visits. The costs of these ambitious programs were high and
fiscally unsustainable. These were some of the first public services to face expenditure cuts during the structural adjustment programs
in the 1980s/1990s (Sulaiman and Hall, 2002; Feder et al., 2006). Decentralized extension models were similar to the T&V systems,
but implementation was devolved to local governments. Issues around incentives for effective delivery remained major concerns
(Oluwasusi and Akanni, 2014). Moreover, the programs were hardly efficient (Kondylis et al., 2017). Farmer field schools (FFS) are
meant to encourage adult learning by implementing participatory and interactive approaches to studying complex and interrelated
concepts, such as conservation farming and integrated pest management (FAO, 2016). Since its introduction in Asia in the late 1980s
(early 1990s), the approach has become popular. Reviews of the FFS approach show that it has contributed to farmer capacity,
especially decision-making capability (van de Fliert et al., 2007; David and Asamoah, 2011; Duveskog et al., 2011; Moussa et al.,
2011), and in some cases, has contributed to increased local social capital (Nakano et al., 2018). Increased social capital is an essential
enabler of technology adoption, as suggested in a recent study by Shikuku et al. (2019) on the role of social reputation to incentivize
community leaders as demonstrators of new agricultural technologies in Uganda. In a similar vein, Munshi (2004) showed that socially
similar wheat farmers operating similar soil types and climatic conditions were more likely to learn from each other than rice farmers
who operated heterogenous fields (also refer to Maartens & Barrett, 2013; and Conley & Udry, 2000).
Central to these approaches has been the search for effective learning models. Several models underpin farmer learning processes.
These include simple and complex contagion or other models, such as social imitation and conformity (Wall, 2007; BenYishay &
Mobarak, 2014; Kiptot et al., 2016). Others include the role of gathering new information from one’s own experiences, or that of
others, in a process known as Bayesian updating (Sadoulet, 2017). However, the contagion or Bayesian updating models face various
difficulties relating to the challenges of learning from others (neighbors or social contacts) whose individual circumstances may not
always be entirely clear to the observer (Tjernström, 2015). In other cases, farmers may strategically adopt a ‘wait and see behavior’,
which is cheaper, to learn from others instead of trying themselves despite receiving the information. Finally, selective use of the
information may be at play. These issues and others are described by De Janvry, Marcours, et al. (2017) and Desdoigts and Cordaro
(2017). Therefore, Desdoigts and Cordaro (2017) conclude that learning by observing others in a social network is not easy (or
straightforward).
One way of improving farmer learning is to deliver training services in local settings and use familiar methods (De Janvry, Mar­
cours, et al., 2017). Therefore, such approaches, like farmer field days run by farmers within their communities, can help in the
evaluation of new technologies alongside the existing ones (De Janvry, Sadoulet, et al., 2016; Jack, 2011). Some of the suggestions by
De Janvry, Marcours, et al. (2017) include conducting experimentation on farmer’s plots. The literature refers to these experimen­
tations as on-farm trials (e.g., demonstration of new crop varieties) when the technology is under development, and the aim is to
evaluate its prospective performance under farmers’ conditions (De Roo et al., 2019). The on-farm trials are more or less like
advertising to show the superiority of new technologies, such as a new maize hybrid.
In summary, the thrust of the literature on the role of extension is to focus on capacity building in farming communities
(Friis-Hansen and Duveskog, 2012). The challenge of delivering proven agricultural technologies to millions of smallholder farmers
across SSA remains. Public funding for agricultural extension services is likely to remain the primary funding mode since many
smallholder farmers are likely to be unable to afford privately sourced and commercially priced extension services. Our focus on
evaluating community embedded demonstrations, training, and field days are well justified as they all fall within the general idea of
allowing farmers to closely observe and try the new practices for themselves. In the next section, we provide the details on sampling
and data collection.

3. Data sources and empirical methods

3.1. Sampling approaches used to recruit respondents into the study

The data used in this paper came from an agricultural applied research project implemented by the International Maize and Wheat
Improvement Centre (CIMMYT) in collaboration with national institutions of agricultural research in the study countries (Kenya,
Malawi, Mozambique, and Tanzania). The aim of the project was to test new CF farming methods (mainly conservation agriculture) in
smallholder farmer settings. The on-farm trials and demonstrations were used as primary learning sites and supplemented by trainings
and field days in the broader communities. The goal was to educate farmers on the relevant CF methods and instigate information
diffusion and scaling of these new farming methods.
The sampling procedure was designed to yield a random sample representative of the population of smallholder farmers in project
implementation areas within a 17–25 km radius of the on-farm trial or demonstration sites. We adopted a continuous random sampling
methodology common in impact evaluation in each of the four countries. This was a two-stage approach, where the first stage involved
the selection of primary sampling units (PSUs) in each of the four countries. In each country, the PSU was a village (the smallest
administrative unit). The next step was to select households for interview of analysis. Households were randomly selected such that
from each village, 15 households were selected from a village residency list maintained by local village administrators. The actual
interviews (carried out in October/November 2018) were done in local languages by field research assistants with a minimum of a
university degree or a tertiary diploma with compensatory experience.

3
P.P. Marenya et al. Environmental Development 40 (2021) 100672

3.2. Data types collected

Demographic and Production Data. Household and plot-level data were elicited using a structured questionnaire. Several CF
practices were captured in the survey, including whether each of the household’s maize-legumes plots had maize-legume intercrops,
fertilizer application or used soil and water conservation measures or combinations thereof. For each maize-legume plot, we identify
whether specific CF practices or packages were implemented on the plot, including the adoption of what may be regarded as the full
SAI package (at a minimum, simultaneously applying conservation tillage, mulching and crop diversification).
Several sociodemographic characteristics, such as sex of the head of the household (or the one regarded as a household leader or
decision-maker), age and years of schooling of the head of the household were collected and included in the regression model as
control variables. Variables, such as total farm size and livestock ownership, were included as an indicator of scale of farming and
wealth. Institutional capital was identified by using questions related to access to finance or markets and whether or not the household
was a current or recent recepient of government social support payments.
Other data collected in the socio-demographics category included plot distance to the household to proxy for ease of access for the
application of bulky inputs, such as manure. Resource endowments were reflected in land size and the number of cattle owned. Market
access was measured by the distance to the nearest input trading center. Whether farmers had strong social networks was identified by
asking respondents if they were members of a social or economic interest group. They were also asked if they had non-relatives in their
networks that they could rely on in times of need and whether they had family members or relations that could be helpful when
needed.
Extension Methods Studied. Three extension methods were modelled in this paper. The first of these were individual (informal)
visits to CF demonstration sites. The demonstration sites were set up under the supervision of project agronomists in collaboration with
local representative farmers who were nominated by their communities. They were to serve as learning sites in which community
members would observe the practices all season as they went about their normal activities. Second, formal field days involved the
convening of large community meetings at some of the demonstration sites, but more commonly, at nearby agricultural research
stations or farmer training centers. Finally, several hands-on training sessions were held to demonstrate how to implement CF prac­
tices. Hands-on training sessions were important to describe the use of machinery, such as power tillers, to implement minimum tillage
and safe use of herbicide application equipment. In all cases, the project teams collaborated with local extension staff, farmer groups,
and other non-governmental organizations to convene the farmers and provide technical information. Over the 7–8 seasons during
which these activities were implemented, these three methods (farmers’ voluntary and individual assessment of demonstration plots,
attendance at formal field days, and attendance of training sessions) provided them with the opportunity to observe, gather infor­
mation, ask questions and learn about the CF methods.
Data on Conservation Farming Practices Considered. The CF methods considered were maize legume rotations or intercrops,
conservation tillage and herbicide weed control. Additionally, mulching or not harvesting crop residues as a soil and water conser­
vation measure was perceived as a CF method. Maize legume rotations and intercropping play an essential role in crop diversification
for the households, since they entail growing a cereal and a legume, thereby proving to be not only a source of calories but also of
protein. Legumes are known to fix atmospheric nitrogen into the soil, thus contributing to soil health. Legumes are also important
sources of income as they fetch twice the price of maize. In these countries, the most common legumes are common beans (Phaseolus
vulgaris). Many varieties are grown across these countries. Pigeon pea (Cajanus cajan) is also common in the drier mid-to low-altitude
maize based cropping zones of Tanzania and Kenya. Soybean (Glycine max) is increasingly being grown in areas where offtake markets
exist, or consumption preferences are developing.
Conservation tillage involved the placement of seeds in seed holes or furrows that were dug out of otherwise untilled land. In
practicing minimum tillage, the creation of furrows was the only soil disturbance done (just for sowing seed). This practice is deemed
as environmentally superior to full-scale tillage, since it conserves the soil, protects it from wind and soil erosion, and improves soil
health for long-term productivity. Since the field is only partially (minimally) tilled, weed growth is higher. The recommendation is to
use crop stover to cover the soil surface and suppress weed growth. The mulch also serves as soil protection, as it reduces wind and
rainfall erosion. Where the stover is available (not used for other purposes), mulching is an effective way to delay weed growth. Instead
of, or in combination with, mulching to control weed growth, a farmer can use herbicides, before the main crops emerge, to spray the
weeds. In many plots, there were various combinations of the four CF methods. In section 4, we provide a summary of the most
common combinations of the CF methods, as documented during data collection.
How Plot-level Information was Collected. During the (2018) in-person interviews, a plot was identified as a contiguous space of
land with a distinct and visible boundary on which maize was grown solely or in combination with other crops (mostly common beans
but other legumes or crops could be grown in rotation or as intercrop). Only plots on which maize was produced as the main or minor
crop were enumerated, and respondents asked to provide information thereon. Such information included the type of practices that
were implemented, such as maize-legume intercrops or rotations during the 2017–2018 agricultural season, the amount of labor and
fertilizer used, if minimum tillage was implemented, if weed control was done using herbicides, and so on.

3.3. Empirical model: multinomial logit model

The role of extension methods can be captured in a model explaining the adoption of CF methods as dependent on information
acquisition through attendance at farmer field days, informal visits to demonstrations and participation in training sessions. These
constitute arguments in the adoption model. Conceptually, we model technology using the random utility maximization problem. A
multinomial logit model is used to motivate the econometric model which characterizes the choice of implementing various CF

4
P.P. Marenya et al. Environmental Development 40 (2021) 100672

methods and their combinations as a multinomial discrete choice of these alternative combinations.2 As already described, these
practices were implemented in various combinations on a plot. In the estimation model, we use those combinations that appeared with
enough frequency (in at least 1% of total plots) to allow for enough degrees of freedom for econometric model convergence.
Denoting the observed combinations (including a null set where none of the practices is applied) as technology sets, let Uij represent
a farm household’s utility of choosing a technology set j (j = 0, 1, 2 … 7, and ‘j = 0’ denoting that none of the technology was adopted,
while the remaining technology sets (j = 1,…, n) contain at least one technology). A farm household chooses a technology set j if
(subject to several
⃒ controls) the utility of this j combination outweighs the utility that could be obtained from any other available
sets (Uij ≥ Uik ⃒x, ∀k ∕
= j). Further, each farmer’s expected utility under the alternative technology sets is assumed to be a function of a
vector of explanatory variablesXi (for example, access to inputs, information availability, and the like) and the random error term εij ,
which is assumed to be independent and identically distributed across the J alternatives.
Therefore, farmer i’s utility derived from choice alternative j, j = 0, 1, J is given by:
Uij = Xij βj + εij (1)

where i indexes individual farmer (decision maker) and j indexes technology set. In our model, Xij is a matrix of observable covariates
that includes household characteristics, plot and, village variables that affect the utility; and βj are unknown parameters to be esti­
mated. The probability that alternative j will be chosen can be modelled as a multinomial logit (MNL) specification, which as per
(Maddala, 1990), gives rise to a system of several probabilities :

exij βj
P(yi = j) = ∑ ′ , k = 1, . . ., J (2)
xij βk
ke

Equation (2) computes the probability of observing a particular CF practice from a set of practices y, for household i. The adoption
equation for each combination was estimated using the MNL model.3 Since all the response probabilities must sum to unity, a
convenient normalization rule is to set one of the parameter vectors, such as β0 , equal to zero. The above equation is equivalent to the
binary logit model where J = 1. The estimated parameters of a multinomial logit model are not interpretable directly. Therefore, we
recover the odds ratios (in this context the relative risk ratios4) to provide an intuitive interpretation of the impact of the promotional
activities on the probability to adopt the various CF practices.

4. Results

4.1. Descriptive statistics

Data in Table 1 shows that attendance at any of the information delivery sessions was highest in Kenya (40%), followed by Malawi
(27%), Mozambique (23%), and Tanzania (14%). In Kenya, Malawi and Mozambique, participation in training was the most popular
option for information delivery on the CF technologies.

4.2. Demographic variables

The average family size ranged from 5.5 in Malawi to 6.7 in Mozambique. Between 74 per cent (Kenya) and 86 per cent (Tanzania)
of households were headed by a man. The average age of the household head ranged from 45.7 years in Malawi to 54 years in Kenya,
indicating that the farmers were mostly middle-aged. The average years of schooling of a household head was highest for Kenya (9.3
years), followed by Tanzania (7.0) Malawi (6.0), and Mozambique (4.9).

4.3. Plot level characteristics

The average distance to plot in minutes was highest for Tanzania (31.1), followed by Mozambique (27.4), Malawi (22.9), and then

2
The basic CF methods are crop rotation (CR), soil cover (SC), soil and water conservation (SWC), conservation tillage (CT), and herbicide usage
(HU) along with various combinations observed on farmers plots (see data description sections below).
3
A potential competing model is the multivariate model (MVP). However, our data structure did not lend themselves to MVP and there was little
convergence. For example, the model for Kenya crashed after running for 6 h; whereas the model for Tanzania was not able to converge after
running over 14 h. Given that we have seven equations and four countries, estimating an MVP will be challenging in terms of computational time.
Moreover, we have smaller observations in some technology choice sets, making the estimates of MVP unreliable. According to Cappellari and
Jenkins (2014), our models may take a couple of days, if not a few weeks, to converge. Therefore, as per Cappellari and Jenkins (2014), “… for
datasets with several thousand observations (and appropriate R), our experiments suggest that estimation may take days or even a few weeks when
the number of equations is above six or seven. Thus, although the number of equations that mvprobit can handle is unlimited in principle, there are
practical constraints that are likely to affect most users …” The violation of the independence of the irrelevant alternative (IIA) assumption needed
would suggest a multivariate model. However, our tests show that IIA has not been violated (results are presented in the supplementary online
materials).
4
The untransformed multinomial logit regression results are presented in the online supplementary materials.

5
P.P. Marenya et al. Environmental Development 40 (2021) 100672

Table 1
Selected household level descriptive statistics , by country.b
Variable Kenya (N = 851) Malawi (N = 724) Mozambique (N = 877) Tanzania (N = 948)

Mean SD Mean SD Mean SD Mean SD

Information delivery mode (% attending at least once)


Demonstration site visit 0.04 0.2 0.18 0.38 0.1 0.3 0.14 0.35
Attended field days 0.2 0.4 0.16 0.37 0.1 0.3 0.1 0.3
Training participation 0.4 0.49 0.27 0.45 0.23 0.42 0.06 0.24
Demographic variables
Family size 5.59 2.8 5.48 2.11 6.67 3.63 5.94 2.26
Male headed HH -Yes = 1 0.74 0.44 0.82 0.38 0.85 0.36 0.86 0.35
HH head age 54.01 13.15 45.67 13.81 49.1 13.58 48 10.86
HH head years of education 9.31 3.6 6.03 3.54 4.9 3.43 7.01 2.35
Plot characteristics
Distance to plot in minutes 11.82 22.93 22.79 24.69 27.4 37.79 31.11 44.17
Plot owned -Yes = 1 0.86 0.35 0.88 0.33 0.97 0.17 0.87 0.34
Soil fertility 1.8 0.59 1.63 0.73 1.65 0.65 1.73 0.56
Plot slope 1.6 0.58 1.39 0.58 1.8 0.66 1.59 0.57
Soil depth 2.08 0.57 2.21 0.82 1.88 0.71 2.23 0.67
Resource constraints
Plot size, ha 0.84 0.66 1.08 0.83 1.92 1.74 2.46 3.02
Assets values (USD)a 243.69 2443.26 286.95 1295.61 205.50 1309.75 194.23 1442.49
Livestock in TLUs 2.11 2 0.56 1.7 1.94 3.42 3.68 6.29
Credit constraint -Yes = 1 0.27 0.44 0.42 0.49 0.47 0.5 0.25 0.44
Market access
Means of transport -Foot = 1 0.45 0.5 0.47 0.5 0.35 0.48 0.2 0.4
Distance to trading center 2.71 2.74 7.05 7.52 8.3 9.91 8.17 10.45
Distance to cooperative office 4.73 5.91 6.57 8.75 14.31 11.73 10.61 11.07
Perceived road quality 2.29 0.65 2.77 0.49 2.38 0.57 2.51 0.58
Social network
Get help outside -Yes = 1 0.83 0.37 0.52 0.5 0.61 0.49 0.69 0.46
Get help from friends -Yes = 1 0.22 0.41 0.08 0.27 0.21 0.4 0.42 0.49
Group membership -Yes = 1 0.94 0.24 0.75 0.43 0.52 0.5 0.5 0.5
a
The conversion rate to US dollars is based on the 2018 average official exchange rates for the local currency units as per World Bank data found in
the World Development Indicators 2018 (https://data.worldbank.org/indicator/PA.NUS.FCRF). A reviewer pointed out to us that the asset values
have large standard deviations relative to mean values. This is likely an artefact of the subjective and imprecise estimation by respondents when asked
how much they would sell the asset presently (as an estimate of its value). These assets are in widely varying degrees of state of repair. This, combined
with lack of records on the date of purchase and the like, means that the resulting recall estimates lead to wide variations in reported values.
b
In each column, the N refers to total houshold sample size per country

Kenya, where it was only 11.8 min. More than 85 per cent of the households owned the plots they operated. The other plot-level
characteristics, such as soil fertility, plot slope, and soil depth, were also obtained and are more or less in a similar range in all
countries.

4.4. Resource constraints

We obtained three critical resource variables, namely plot size in hectares, the value of assets in US$, tropical livestock unit, and
credit constraint dummy. The average land size is highest in Tanzania (2.46 ha), followed by Mozambique (1.92 ha), Malawi (1.08 ha),
and Kenya (0.84 ha). The assets value was over US$200 for Malawi and Kenya, just about $200 for Mozambique and US$194 for
Tanzania. The tropical livestock unit (TLU) was highest for Tanzania (3.7), followed by Kenya (2.1), Mozambique (1.9), and only 0.56
for Malawi. In Mozambique and Malawi, more than 40 per cent of the farm households reported that they face credit constraints; 27 per
cent in Kenya, and 25 per cent in Tanzania reported an inability to access any type of credit (formal or informal).

4.5. Market Access

In Kenya (45%), Malawi (47%), Mozambique (35%), and Tanzania (20%), the members of the farming households travel by foot as
the primary mode of transport to do business in local markets. Distance to the trading center was highest for Mozambique (8.3 km) and
Tanzania (8.2 km). The distance to the nearest main trading center was 7.1 km for Malawi and only 2.7 km for Kenya. Similarly, the
distance to the cooperative office was 14.3 km for Mozambique, followed by Tanzania (10.6 km), Malawi (6.6 km), and Kenya (4.7
km).

4.6. Social networks

We measured the existence of social networks using the following three indicators: the ability to get help from two groups of
networks, namely non-relatives and friends and family, while the third social network variable was membership in a social or

6
P.P. Marenya et al. Environmental Development 40 (2021) 100672

economic interest group. A significant number of farm households in Kenya (83%), Tanzania (69%), Mozambique (61%), and Malawi
(52%) were able to rely on non-relatives for help. It seems that those who reported connections to social safety nets among friends and
family were less than half; 42 per cent in Tanzania, 22 per cent in Kenya, 21 per cent in Mozambique, and 8 per cent in Malawi. The
membership in the farmer’s group was about 50 per cent in Tanzania and Mozambique, 75 per cent in Malawi and 94 per cent in Kenya.
Malawi and Kenya seem to have high levels of participation in community organizations.

4.7. Adoption and empirical results

The Role of Demonstrations, Farmer Field Days and Training in the Adoption of Conservation Farming Practices.
The summaries presented in Table 2 show that about 15–34 per cent of the households had not adopted any CF across the four
countries. The highest adoption rates of the various practices were 35.9% for CR (Tanzania); the lowest rate was 0.9 per cent for
CT_SC_HU (Kenya), 1.3 per cent for SC_SWC (Tanzania), 2.97 per cent for CT_CS_SWC (Malawi) and 5.7 per cent for SWC_CT
(Mozambique). Table 2 shows that the adoption rates for the various combinations of CF practices was frequently higher than 10% but
mostly less than 30 per cent.
As the dependent variables were discrete and mutually exclusive, we estimated multinomial logit model for each country with the
number of choices observed varying per country, as previously described. The full regression results are presented in Tables S1–S4 in
the online Appendix. In this section, we summarize and discuss the statistically significant relative risk ratios by the farmer outreach
methods: privately attended demonstration sites (DEM), participated in farmer field days (FDY), and attended hands-on training
sessions (TRA). The results generally suggest that there is a positive and statistically significant relationship between these variables
and the adoption of technology sets among across the study countries. More specifically, the findings suggest that attending field days
and trainings were empirically associated with the CF technologies among smallholder farmers in Kenya and Malawi. In Kenya, for a
farmer who attended field days relative to those who did not participate in any of the extension methods, the RRR for adopting
conservation tillage or herbicide use (CT/HU) was 5.1 given the other controls. In other words, farmers who attended field days are five
times more likely to adopt CT/HU than to not adopt. (Fig. 1).
In Kenya, the highest impacts of extension activities on the adoption of CF methods were visits to demonstration sites and field
days’ attendance (Fig. 1).5 The field days had a positive and significant impact on the adoption of conservation tillage (CT), where the
odds of adoption were two times higher than those who never attended any outreach. The odds of adopting herbicides or weed control
(HU) were nearly 23 times higher if they had attended demonstrations. The dramatic effects of scorching and subsequent suppression
of weeds by herbicide applications are arguably easy to observe at a demonstration site. The odds of adopting soil cover (SC) were
approximately two for DEM and about 1.8 in the case of FDY. Similarly, in the case of conservation tillage & soil cover (CT&SC), the
odds of adoption ranged from 2.9 (DEM) to 3.5 (FDY) and 2.2 (TRA). Similar to HU, the odds of adopting SC/HU were about 23 times
higher if the farmer had reported attending demonstration sites.

Table 2
Technology sets considered (% of total observations), by country.
Technology sets Kenya (N = Malawi Mozambique (N = Tanzania (N =
2813) N= 1509) 1459)
(1397)

None 34.06 30.99 15.44 14.67


CR: Crop rotation [crop rotation & ML intercropping] 16.83 35.85
CT: Conservation tillage [zero tillage & minimum tillage] 12.83 5.70
HU: Herbicide use 0.71
SC: Soil cover [Mulching & crop residue] 26.77 16.54 6.65
SWC: Soil and water conservation [M13 – M15] 18.54 2.58 3.22
CT_SC: Conservation tillage + Soil cover 11.7 3.44
CT_HU: Conservation tillage + Herbicides use 4.69
SC_HU: Soil cover + Herbicides use 8.32
SC_SWC: Soil cover + Soil and water conservation 17.25 1.3
CR_SWC: Crop rotation + Soil and water conservation 10.93 8.36
CR_CT: Crop rotation + Conservation tillage 17.16
CR_SC: Crop rotation + Soil cover 23.58
SWC_CT: Soil and water conservation + Conservation tillage 5.70
CT_SC_HU: Conservation tillage + Soil cover + Herbicides use 0.92
CT_CS_SWC: Conservation tillage + Crop rotation + Soil, and water 2.79
conservation
CT_SC_SWC: Conservation tillage, Soil cover + Soil and water conservation 10.45
CR_SWC_CT: Crop rotation + Soil and water conservation + Conservation 25.65
tillage
CR_SWC_SC: Crop rotation + Soil and water conservation + Soil cover 6.37

5
In Fig. 1 and the rest of the figures (Figs. 2–4) the asterisk *, **, *** denote that the RRR is significant at 10, 5 and 1% level respectively.

7
P.P. Marenya et al. Environmental Development 40 (2021) 100672

Fig. 1. Impact of treatment types on SAI technology adoption in Kenya (RRR). For more details, see Supplementary Table S1. *, **, *** indicate RRR
was significant at 10, 5 and 1% respectively.

Fig. 2. Impact of treatment types on SAI technology adoption in Malawi (RRR). For more details, see Supplementary Table S2. *, **, *** indicate
RRR was significant at 10, 5 and 1% respectively.

In Malawi, trainings had by far the highest RRRs followed by field days. (Fig. 2). The relative risk ratios for TRA were positive and
significant for most of the CF combinations. Where significant, the odds ratio of adoption was four times if farmers had visited
demonstration sites (in the case of CT_SC_SWC). A farmer was ten times more likely to adopt conservation tillage with soil cover
(CT_SC) if they had attended trainings. Similarly, the odds ratio of adopting conservation tillage, soil cover, and soil and water
conservation (CT_SC_SWC) was nearly 14 times if a farmer had attended trainings than if they had not. We conclude that in Malawi, at
least, DEM alone, compared to the TRA and FDY, was not always the most effective way of encouraging the adoption of CF practices.
In Mozambique, the impacts of extension activities on the adoption of CF methods were more mixed. The impact of DEM was
positive and highly significant for the adoption of crop rotation (CR), crop rotation, and soil and water conservation (CR_SWC) but so
was RRR for FDY and TRA. Compared to those who had attended none of the outreach activities, those who had attended demon­
stration sites were twice and almost three times likely to adopt CR and about 2.6 times for CR_SWC. A similar magnitude in RRR was
observed for field days in the case of CR (RRR = 2). Those who had attended trainings were 2.5 times more likely to have adopted
CR_SWC (Fig. 3).
In Tanzania, field days and trainings were the most significant outreach methods. The RRR of adopting crop rotation and soil and
water conservation (CR_SWC) was about ten times higher if farmers reported attending trainings and three times higher if they had
participated at farmers’ field days (Fig. 4). The odds were three times higher in the case of crop rotation and soil cover (CR_SC) if the
farmer had attended training, and 2.5 times higher if he or she had attended field days.
Summary results in Table 3 shows that in Malawi, Mozambique, and Tanzania, TRA seems to have the highest impact on CF
adoption, while in Kenya and Mozambique, the FDY were also strongly associated with high odds of adoption. In Malawi and Tanzania,
field days had a mid-range impact, and demonstration sites had a mid-range impact in Kenya and Mozambique. In Kenya, trainings had
the lowest impact, while demonstration sites had the lowest impact in Malawi, Mozambique, and Tanzania. We can conclude that
across the four countries, the highest RRRs were in the case of trainings (having the highest odds ratios in three out of four countries)

8
P.P. Marenya et al. Environmental Development 40 (2021) 100672

Fig. 3. Impact of treatment types on SAI technology adoption in Malawi (RRR). For more details, see Supplementary Table S3. *, **, *** indicate
RRR was significant at 10, 5 and 1% respectively.

Fig. 4. Impact of treatment types on SAI technology adoption in Malawi (RRR). For more details, see Supplementary Table S4. *, **, *** indicate
RRR was significant at 10, 5 and 1% respectively.

Table 3
Impact of treatment types on SAI technology adoption by countries.
Country RRR coefficient estimates

Highest Mid-range Lowest

Kenya FDY: Field Days DEM: Demonstration site visit TRA: Training
Malawi TRA: Training FDY: Field Days DEM: Demonstration site visit
Mozambique TRA: Training, FDY: Field Days DEM: Demonstration site visit DEM: Demonstration site visit
Tanzania TRA Training FDY: Field Days DEM: Demonstration site visit

and field days (highest RRR in Kenya and variously significant in multiple CF combinations in Malawi). Demonstrations had the lowest
RRRs in three out of the four countries.

5. Discussion

Enabling farmers to access low-cost opportunities to learn about new agricultural technologies is essential for agricultural

9
P.P. Marenya et al. Environmental Development 40 (2021) 100672

modernization. Establishing demonstration sites where farmers can regularly observe and organizing formal field days and on-hands
training can contribute to this process. However, there is limited information on the comparative effectiveness of various extension
modalities that can be used to facilitate both pathways of learning. Although farmers learn through the various channels, our empirical
analysis suggests that attending field days and trainings was strongly associated with the adoption of CF practices among smallholder
farmers, particularly in Kenya and Malawi. The findings support our hypothesis that providing farmers with low-cost means of
observing and evaluating new farming practices and technologies increases the adoption of CF technologies in the study countries. The
results lend weight to the need for providing platforms for personal observations and interactions with experts such as through
formalized training sessions or famer field days. For example, in a synthesis paper, De Janvry, Sadoulet and Rao (2016) list ‘Bayesian
updating’, in which farmers experiment on their own, generate new data, and update their information on what works best. Similar
conclusion was reached by Munshi (2004) on rice farmers in India, who could not learn effectively through social networks given the
heterogenous farming conditions for rice but had to learn from own practices instead. Wheat farmers, Munshi (2004) found, could
easily learn through social networks given the more homogenous growing conditions. The takeaway message being that in heterog­
enous environments, experimentation and experiential approaches to farmer learning will be critical.
It is worth pointing out that the adoption of CF practices can be influenced by various socio-economic, demographic, and insti­
tutional factors. It is, therefore, not striking, perhaps expected, to see variations across countries in the adoption of different CF
technology sets. In Malawi, for example, trainings had the highest RRRs followed by field days in the case of conservation tillage (CT),
soil, and water conservation (SWC), and herbicides usage (HU) (see Fig. 2). This is because these practices are knowledge-intensive in
the sense that they are disembodied technologies. Therefore, convincing farmers to adopt them may require more effort through
explanations and hands-on experience, which is not possible with passive visits to demonstration sites.
The passive nature of observing demonstration sites (perhaps in passing) might reduce their influence. In keeping with our results
which suggests the power of interactive engagement, demonstration sites need to be implemented as focal points for conducting formal
farmer meetings. This way, they can interact with extension experts and subject matter specialists to explain matters such as the input
rates used, types of varieties, timings, residue application, weed management and a host of details not accessible via passive obser­
vation of a demonstration plot, absent further interactions with extension specialists. The lesson for agricultural technology scaling
programs in Africa (and beyond) is that multi-pronged approaches to scaling are critical for success. Therefore, a judicious combi­
nation and sequencing of extension methods including their optimal implementation duration (Khainga et al., 2021) are matters to be
considered carefully in extension program design. These issues also offer opportunities for further research.
In rural areas of developing countries, the social network indicates the linkages within the communities in which farmers are living,
and such social networks tend to provide safety nets during difficult times and play a significant role as avenues for information
exchange (Van Rijn, Bulte, & Adekunle, 2012). Kenya has a long history with cooperative and savings organizations as far back as 1908
(Republic of Kenya, 2020), and Malawi has a strong network of farmers organizations under the umbrella of the National Smallholder
Farmers’ Association of Malawi (NASFAM (National Smallholder Farmers’ Asscoiation of Malawi), 2021) (https://www.nasfam.org/
index.php/about-us). Our results suggest that farmers belonging to groups were more likely to adopt the CF practices (see Tables S1–S4
in the Appendix). In support of our findings, Rai et al. (2018) showed that farmer-group membership facilitates the diffusion and
adoption of CF practices through interaction and the exchange of ideas.
Additionally, the adoption and performance of various CF practices depend on the biophysical, socio-technical, and institutional
structure of a country (e.g., Andrieu et al., 2020; Thierfelder et al., 2016). The physical assets and resources are important measures of
a farmer’s ability (or capacity) to access and adopt new agricultural technologies (Andersson & D’Souza, 2014). Credit constraint is
one of the main challenges for smallholder farmers across the study countries, notably Mozambique and Malawi. Prior studies in
Malawi and Zimbabwe have found that access to formal extension services is highly effective in the promotion of CF technologies
adoption among smallholders even where access to credit is low (Makate et al., 2019). Finally, regarding the other control variables,
for example, the household head’s level of education is often a significant predictor in the adoption of the CF practices (or improved
agricultural technologies more generally).

5.1. Conlusions and policy implications

As the world population increases and demand for food rises, the pressure to conserve resources will become more exigent than ever
(FAO, 2013). Moreover, climate change and soil degradation are posing big challenges to global food security. In the next decades,
agriculture will have to sustainably produce more food from less land through more efficient use of natural resources and with minimal
impact on the environment while meeting the needs of growing populations (FAO, 2013). As a result, there is a growing emphasis on
more sustainable and climate-smart agriculture practices. Promoting the adoption of CF practices by providing smallholders with
low-cost means of evaluating the various CF practices can mitigate the challenges of slow adoption and scaling of environmentally
positive agricultural practices.

5.2. Conclusions and policy implications

Using recently collected data from four countries (Kenya, Malawi, Mozambique, and Tanzania), we implement a multinomial
choice model in this paper to analyze the impacts of community-embedded demonstrations, trainings, and field days on the adoption of
conservation tillage and related practices. The results show that field days and training sessions had significant impacts on the
probability of adopting conservation farming methods.
In general, all the outreach programs appear impactful depending on the country and the conservation farming practice. In three

10
P.P. Marenya et al. Environmental Development 40 (2021) 100672

countries, training sessions caused consistently higher RRRs of adopting conservation farming practices. The relative risk ratios of
adopting the practices were between 10 (in Kenya and Tanzania) and 14 (in Malawi). Based on the number of countries where the
RRRs were significantly high, we conclude that attending training sessions were more consistently associated with higher odds of
adopting the conservation farming practices across the board. Visiting demonstration sites had substantial impacts on the odds of
adopting herbicide weed control, showing that the relative risk ratios of using herbicides for weed control were 23 times in Kenya.
Nevertheless, visiting demonstration sites had the lowest ratios in three out of the four countries. Field days had the highest RRR in
Kenya and Malawi for several conservation farming practices. We conclude that all three outreach programs positively impact the
adoption of the practices considered in this study, while trainings and field days were more consistently associated with the probability
of adopting the practices. Our results seem consistent with the notion that experiential learning approaches (field days and hands-on
training) are likely to be more effective in farmer education. Future information dissemination programs in these (and similar envi­
ronments) should combine the usage of existing social networks with complementary approaches that emphasize active farmer
engagement and self-learning. Hence, an outstanding research gap relates to examining the role of autonomous farmer experimen­
tation and finding ways to help underwrite the associated costs to speed up the learning and technology adoption process. Finally, the
matter of joint impact of different methods requires further quantitative assessments in future studies. We encourage other researchers
in this area to look into these outstanding issues.

Author contribution

Paswel Marenya: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project
administration, Resources, Supervision; Validation, Visualization, Roles/Writing - original draft; Writing - review & editing. Usman
Mohammed: Software, Formal analysis, Validation, Roles/Writing - original draft. Dil Rahut: Funding acquisition, Methodology;
Project administration, Validation, Roles/Writing - original draft, Writing - review & editing.

Declaration of competing interest

We would like to inform that we do not have any conflict of interest.

Acknowledgments

This study was supported by the Australian Centre for International Agricultural Research (ACIAR) through grant no. CSE/2009/
024. Additional funding was made available through the CGIAR Research Program on Maize Agri-food Systems (CRP MAIZE). The
authors are solely responsible for the content and any remaining errors in this paper. The views expressed here are those of the authors
and do not necessarily reflect the views of the funding organizations or the authors’ affiliations. None of the funding entities directed or
participated in the execution of this research or the writing of this article.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.envdev.2021.100672.

References

Abdoulaye, T., Abass, A., Maziya-Dixon, B., Tarawali, G., Okechukwu, R., Rusike, J., Alene, A., Manyong, V., Ayedun, B., 2013. Awareness and adoption of improved
cassava varieties and processing technologies in Nigeria. J. Dev. Agric. Econ. 7 (4).
Ainembabazi, J.H., van Asten, P., Vanlauwe, B., Ouma, E., Blomme, G., Birachi, E.A., Manyong, V.M., 2017. Improving the speed of adoption of agricultural
technologies and farm performance through farmer groups: evidence from the Great Lakes region of Africa. Agric. Econ. 48 (2), 241–259. https://doi.org/
10.1111/agec.12329.
Andrieu, N., Dumas, P., Hemmerlé, E., Caforio, F., Falconnier, G.N., Blanchard, M., Vayssières, J., 2020. Ex ante mapping of favorable zones for uptake of climate-
smart agricultural practices: a case study in West Africa. Environmental Development, 100566.
Bezu, S., Kassie, G.T., Shiferaw, B., Ricker-Gilbert, J., 2014. Impact of improved maize adoption on welfare of farm households in Malawi: a panel data analysis. World
Dev. 59, 120–131.
Cunguara, B., Moder, K., 2011. Is agricultural extension helping the poor? Evidence from rural Mozambique. J. Afr. Econ. 20 (4), 562–595. https://doi.org/10.1093/
jae/ejr015.
Daniel, E., 2013. Assessment Of Agricultural Extension Services in Tanzania. A Case Study of Kyela, Songea Rural And Morogoro Rural Districts. Internship Report.
Wegeningen University, Wageningen, Netherlands.
David, S., Asamoah, C., 2011. The impact of farmer field schools on human and social capital: a case study from Ghana. J. Agric. Educ. Ext. 17 (3), 239–252. https://
doi.org/10.1080/1389224X.2011.559076.
Davis, K., Swanson, B., Amudavi, D., Mekonnen, D.A., Flohrs, A., Riese, J., Lamb, C., Zerfu, E., 2010. In-depth Assessment of the Public Agricultural Extension System
of Ethiopia and Recommendations for Improvement. Discussion Paper No. 01041. International Food Policy Research Institute (IFPRI), Washington DC.
De Janvry, A., Sadoulet, E., Rao, M., 2016. Adjusting Extension Models to the Way Farmers Learn. Learning for Adopting: Technology Adoption in Developing Country
Agriculture. FERDI.
Learning for adopting: technology adoption in developing country agriculture. In: De Janvry, A., Macours, K., Sadoulet, E. (Eds.), 2017. FERDI.
De Roo, N., Andersson, J., Krupnik, T., 2019. On-farm trials for development impact? The organisation of research and the scaling of agricultural technologies. Exp.
Agric. 55 (2), 163–184. https://doi.org/10.1017/S0014479717000382.

11
P.P. Marenya et al. Environmental Development 40 (2021) 100672

Dercon, S., Gilligan, D.O., Hoddinott, J., Woldehanna, T., 2009. The impact of agricultural extension and roads on poverty and consumption growth in fifteen
Ethiopian villages. Am. J. Agric. Econ. 91 (4), 1007–1021.
Desdoigts, A., Cordaro, F., 2017. Learning versus status quo bias and the role of social capital in technology adoption: the case of cocoa farmers in Côte d’Ivoire. In: De
Janvry, A., Macours, K., Sadoulet, E. (Eds.), Learning for Adopting: Technology Adoption in Developing Country Agriculture. FERDI.
Duveskog, D., Friis-Hansen, E., Taylor, E.W., 2011. Farmer field schools in rural Kenya: a transformative learning experience. J. Dev. Stud. 47 (10), 1529–1544.
https://doi.org/10.1080/00220388.2011.561328.
Erenstein, O., Sayre, K., Wall, P., Hellin, J., Dixon, J., 2012. Conservation agriculture in maize-and wheat-based systems in the (sub) tropics: lessons from adaptation
initiatives in South Asia, Mexico, and Southern Africa. J. Sustain. Agric. 36 (2), 180–206.
FAO, 2012. Voluntary Guidelines on the Responsible Governance of Tenure of Land, Fisheries and Forests in the Context of National Food Security. FAO.
FAO, 2013. Climate-smart Agriculture: Sourcebook. Food and Agriculture Organization of the United Nations.
FAO, 2010. Mobilizing the Potential of Rural and Agricultural Extension. Danish Institute for International Studies.
Feder, G., Ganguly, S., Anderson, J.R., 2006. The Rise and Fall of Training and Visit Extension: an Asian Mini-Drama with an African Epilogue. The World Bank.
Feder, G., Willett, A., Zijp, W., 2001. Agricultural extension. Generic challenges and the ingredients for solutions. Knowledge Generation and Technical Change:
Institutional Innovation in Agriculture. Kluwer.
Friis-Hansen, E., Duveskog, D., 2012. The empowerment route to well-being: an analysis of farmer field schools in East Africa. World Dev. 40 (2), 414–427. https://
doi.org/10.1016/j.worlddev.2011.05.005.
Godfray, H.C.J., Beddington, J.R., Crute, I.R., Haddad, L., Lawrence, D., Muir, J.F., Pretty, J., Robinson, S., Thomas, S.M., Toulmin, C., 2010. Food security: the
challenge of feeding 9 billion people. Science‚ 327 (5967), 812–818.
Hobbs, P.R., Sayre, K., Gupta, R., 2008. The role of conservation agriculture in sustainable agriculture. Phil. Trans. Biol. Sci. 363 (1491), 543–555.
Jack, K., 2011. Constraints on the Adoption of Agricultural Technologies in Developing Countries. White Paper, Agricultural Technology Adoption Initiative. J-PAL
(MIT) and CEGA, UC Berkeley).
Kassie, M., Jaleta, M., Shiferaw, B., Mmbando, F., Mekuria, M., 2013. Adoption of interrelated sustainable agricultural practices in smallholder systems: evidence from
rural Tanzania. Technol. Forecast. Soc. Change 80 (3), 525–540. https://doi.org/10.1016/j.techfore.2012.08.007.
Khainga, D.N., Marenya, P.P., da Luz Quinhentos, M., 2021. How much is enough? How multi-season exposure to demonstrations affects the use of conservation
farming practices in Mozambique. Environ. Dev. Sustain. 1–23.
Khonje, M., Manda, J., Alene, A.D., Kassie, M., 2015. Analysis of adoption and impacts of improved maize varieties in Eastern Zambia. World Dev. 66, 695–706.
https://doi.org/10.1016/j.worlddev.2014.09.008.
Knowler, D., Bradshaw, B., 2007. Farmers’ adoption of conservation agriculture: a review and synthesis of recent research. Food Pol. 32 (1), 25–48. https://doi.org/
10.1016/j.foodpol.2006.01.003.
Kiptot, E., Karuhanga, M., Franzel, S., Nzigamasabo, P.B., 2016. Volunteer farmer-trainer motivations in East Africa: practical implications for enhancing farmer-to-
farmer extension. Int. J. Agric. Sustain. 14 (3), 339–356.
Kondylis, F., Mueller, V., Zhu, J., 2017. Seeing is believing? Evidence from an extension network experiment. J. Dev. Econ. 125, 1–20. https://doi.org/10.1016/j.
jdeveco.2016.10.004.
Kummer, S., Leitgeb, F., Vogl, C.R., 2017. Farmers’ own research: organic farmers’ experiments in Austria and implications for agricultural innovation systems.
Sustain. Agric. Res. 6, 526-2017-2663).
Maddala, G.S., 1990. Limited-dependent and Qualitative Variables in Econometrics. Cambridge University Press.
Makate, C., Makate, M., Mutenje, M., Mango, N., Siziba, S., 2019. Synergistic impacts of agricultural credit and extension on adoption of climate-smart agricultural
technologies in Southern Africa. Environmental Development 32, 100458.
Marenya, P.P., Kassie, M., Jaleta, M., Erenstein, O., 2017. Predicting minimum tillage adoption among smallholder farmers using micro-level and policy variables.
Agricultural and Food Economics 5 (1), 12.
Masangano, C., Mthinda, C., 2012. Pluralistic Extension System in Malawi. Discussion Paper No. 01171. International Food Policy Research Institute (IFPRI),
Washington DC.
Milestad, R., Dedieu, B., Darnhofer, I., Bellon, S., 2012. Farms and farmers facing change: the adaptive approach. In: Darnhofer, I., Gibbon, D., Dedieu, B. (Eds.),
Farming Systems Research into the 21st Century: the New Dynamic. Springer.
Moussa, B., Otoo, M., Fulton, J., Lowenberg-DeBoer, J., 2011. Effectiveness of alternative extension methods through radio broadcasting in West Africa. J. Agric.
Educ. Ext. 17 (4), 355–369. https://doi.org/10.1080/1389224X.2011.576826.
Mulwa, C., Marenya, P., Kassie, M., 2017. Response to climate risks among smallholder farmers in Malawi: a multivariate probit assessment of the role of information,
household demographics, and farm characteristics. Climate Risk Management 16, 208–221.
Nakano, Y., Tsusaka, T.W., Aida, T., Pede, V.O., 2018. Is farmer-to-farmer extension effective? The impact of training on technology adoption and rice farming
productivity in Tanzania. World Dev. 105, 336–351. https://doi.org/10.1016/j.worlddev.2017.12.013.
NASFAM (National Smallholder Farmers’ Asscoiation of Malawi), 2021. https://www.nasfam.org/index.php/about-us.
Nyamangara, J., Nyengerai, K., Masvaya, E.N., Tirivavi, R., Mashingaidze, N., Mupangwa, W., Dimes, J., Hove, L., Twomlow, S., 2014. Effect of conservation
agriculture on maize yield in the semi-arid areas of Zimbabwe. Exp. Agric. 50 (2), 159–177.
Oluwasusi, J.O., Akanni, Y.O., 2014. Effectiveness of extension services among food crop farmers in Ekiti state, Nigeria. J. Agric. Food Inf. 15 (4), 324–341. https://
doi.org/10.1080/10496505.2014.952175.
Owens, T., Hoddinott, J., Kinsey, B., 2003. The impact of agricultural extension on farm production in resettlement areas of Zimbabwe. Econ. Dev. Cult. Change 51
(2), 337–357.
Pan, Y., Smith, S.C., Sulaiman, M., 2018. Agricultural extension and technology adoption for food security: evidence from Uganda. Am. J. Agric. Econ. 100 (4),
1012–1031. https://doi.org/10.1093/ajae/aay012.
Pannell, D.J., Llewellyn, R.S., Corbeels, M., 2014. The farm-level economics of conservation agriculture for resource-poor farmers. Agric. Ecosyst. Environ. 187,
52–64.
Pye-Smith, C., 2012. Agricultural Extension, a Time for Change: Linking Knowledge to Policy and Action for Food and Livelihoods. Technical Centre for Agricultural and
Rural Cooperation ACP-EU (CTA), Wageningen, the Netherlands.
Rai, R.K., Bhatta, L.D., Acharya, U., Bhatta, A.P., 2018. Assessing climate-resilient agriculture for smallholders. Environmental Development 27, 26–33.
Republic of Kenya, 2012. National Agricultural Sector Extension Policy (NASEP). Government of Kenya, Agricultural Sector Coordination Unit (ASCU).
Republic of Kenya, 2020. History and Organization of Cooperative Development and Marketing Sub Sector in Kenya. http://www.industrialization.go.ke/index.php/
downloads/123-history-and-organization-of-cooperative-development-and-marketing-sub-sector-in-kenya.
Shiferaw, B., Kassie, M., Jaleta, M., Yirga, C., 2014. Adoption of improved wheat varieties and impacts on household food security in Ethiopia. Food Pol. 44, 272–284.
Sulaiman, V., Hall, A., 2002. Beyond technology dissemination – can Indian agricultural extension re-invent itself?.
Swanson, B.E., Rajalahti, R., 2010. Strengthening Agricultural Extension and Advisory Systems: Procedures for Assessing, Transforming, and Evaluating Extension
Systems. In: Agriculture and Rural Development Discussion Paper No. 45. World Bank, Washington.
Tambo, J.A., Mockshell, J., 2018. Differential impacts of conservation agriculture technology options on household income in Sub-Saharan Africa. Ecol. Econ. 151,
95–105. https://doi.org/10.1016/j.ecolecon.2018.05.005.
Teklewold, H., Kassie, M., Shiferaw, B., Köhlin, G., 2013. Cropping system diversification, conservation tillage and modern seed adoption in Ethiopia: impacts on
household income, agrochemical use and demand for labor. Ecol. Econ. 93, 85–93.
Thierfelder, C., Matemba-Mutasa, R., Bunderson, W.T., Mutenje, M., Nyagumbo, I., Mupangwa, W., 2016. Evaluating manual conservation agriculture systems in
southern Africa. Agric. Ecosyst. Environ. 222, 112–124.
van de Fliert, E., Dung, N.T., Henriksen, O., Dalsgaard, J.P.T., 2007. From collectives to collective decision-making and action: farmer field schools in Vietnam.
J. Agric. Educ. Ext. 13 (3), 245–256. https://doi.org/10.1080/13892240701427706.

12

You might also like