Africa

Received: 10 February 2023 | Accepted: 12 February 2024
DOI: 10.1002/jaa2.112
ARTICLE
Adoption determinants of improved cassava

varieties and intercropping among East and
Central African smallholder farmers
Paul Mwebaze1 | Sarina Macfadyen2 | Paul De Barro3 |

Anton Bua4 | Andrew Kalyebi4 | Irene Bayiyana4 | Fred Tairo5 |
John Colvin6
1
Department of Environmental Science &
Policy, the University of California Davis, Davis, Abstract
California, USA A key constraint to cassava productivity in Africa is the
2
Commonwealth Scientific and Industrial lack of adoption of improved cassava varieties tolerant to
Research Organisation, Canberra, pests and diseases. To understand the drivers of adoption
Australian Capital Territory, Australia
behavior, we examine the simultaneous adoption of
3
Commonwealth Scientific and Industrial
improved cassava varieties and intercropping by 1200
Research Organisation, Brisbane, Queensland,
Australia smallholder farmers in Malawi, Tanzania, and Uganda.
4
National Crops Resources Research Institute,
Using a linear model, we find that varietal characteristics,
Kampala, Uganda such as yield and early maturity, are critical drivers of
5
Mikocheni Agricultural Research Institute, Dar adoption. Access to extension and credit is associated
es Salaam, Tanzania with an increase in the attractiveness of yield‐improving
6
Natural Resources Institute, University of characteristics. We conclude that a more targeted
Greenwich, Kent, UK extension approach would increase technology adoption
in these countries.
Correspondence
Paul Mwebaze, Department of Environmental
KEYWORDS
Science & Policy, the University of California,
Davis, Davis, CA 95616, USA. Food security, Improved cassava varieties, Intercropping, Smallholder
Email: pmwebaze@ucdavis.edu and farmers
mpaul48@hotmail.com
JEL CLASSIFICATION
Funding information
Bill and Melinda Gates Foundation, Q1 (agriculture), Q5 (environment and ecology), O3 (innovation,
Grant/Award Number: OPP1058938 technological change)
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction
in any medium, provided the original work is properly cited.
© 2024 The Authors. Journal of the Agricultural and Applied Economics Association published by Wiley Periodicals LLC on behalf of the
Agricultural & Applied Economics Association
292 | wileyonlinelibrary.com/journal/jaa2 Jnl of Agr & App Econ Assoc. 2024;3:292–310.

27692485, 2024, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/jaa2.112 by EBMG ACCESS - COTE D'IVOIRE, Wiley Online Library on [04/07/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
MWEBAZE ET AL. | 293
1 | INTRODUCTION
Cassava is the second most important food crop in Africa after maize and provides more than half
of the dietary calories for over 200 million people in East and Central Africa (ECA) (Alene, 2013;
FAO & IFAD, 2005; Reincke et al., 2018). It is produced by smallholder farmers whose average
cultivated area is less than one hectare and often in households with female heads for home
consumption (Masamha et al., 2018).1 It is mainly grown under intercropping systems with other
crops such as maize, legumes, and bananas. Africa is the world's largest cassava‐producing region,
accounting for nearly 61% of the global cassava output (FAO, 2023).2 Moreover, cassava roots are
significantly cheaper than grains such as maize and rice. It is also relatively tolerant of poor soils
and seasonal droughts, making it a suitable food security crop, especially during conflicts
(Bennett, 2015; Parmar et al., 2017). While other crops are projected to face significant adaptation
challenges in Africa, cassava is resilient to climate change because of its drought tolerance,
resilience on marginal agricultural lands, and ability to be left in the ground without harvesting
for extended periods (Jarvis et al., 2012). Previous studies have emphasized the economic
importance of cassava to smallholder farmers (Fermont et al., 2010; Johnson et al., 2006;
Roothaert & Magado, 2011).
Cassava yields in Africa are the lowest in the world, with an average yield of 10 tonnes per
hectare (tonnes/ha) compared to 26 tonnes/ha in India (FAO, 2023). Figure 1 shows changes in
cassava yields over time in Malawi, Tanzania, and Uganda. In Tanzania and Uganda, there was a
steady decline from 1990 onwards. The potential factors explaining the trends include pests and
diseases, climate change, low‐yielding varieties, declining soil fertility, and poor agronomic practices
(Alene, 2013; Legg et al., 2006; Roothaert & Magado, 2011). By contrast, Malawian yields increased
roughly ten times from the low point in the 80 s due to the release of improved crop varieties
(IVCs) (Alene, 2013). Fermont et al. (2009) estimated that cassava yields in East Africa could reach
20 Mt/ha using existing technologies and best practices. Yield potential in a well‐controlled research
setting with irrigation and fertilization was approximately 45 Mt/Ha (Tian et al., 2009). If translated
to the farmer level, it would increase cassava production and improve food security and incomes
among cassava farmers. Dethier and Effenberger (2012) have suggested that the two main
explanations for low yields in Africa are a lack of appropriate technology and slow adoption.
However, Suri and Udry (2022) point to several other constraints, such as credit and liquidity,
insurance, information, high transaction costs, and imperfect labor and land markets.
There is extensive literature on technology adoption and the relationship between technology
adoption and socioeconomic and policy variables in developing countries (e.g., Feder et al., 1985;
Suri & Udry, 2022). Studies show that the wide adoption of new agricultural technologies
effectively increases agricultural productivity (Minten & Barrett, 2008). Adoption is a process of
willingly accepting innovation by the members of a social system (Rogers, 1962). Previous studies
have investigated factors influencing technology adoption decisions and the intensity of adoption.
Farm and farmer socioeconomic characteristics affected adoption in many of these studies. Farm
size (Kassie et al., 2013), varietal attributes, especially taste, color, and drought tolerance (Kalema
et al., 2022; Mafuru et al., 2008; Miriti et al., 2022), education (Kassie et al., 2013; Teklewold
et al., 2013), extension (Ruzzante et al., 2021; Wossen et al., 2017), household size, accessible
inputs and age of household head (Kondo et al., 2020; Simtowe et al., 2016) influenced the
adoption of technologies. Beyond these factors cited above, Michler et al. (2019) found that
profitability is more relevant than increases in yields in explaining technology adoption decisions.
1
However, the perception of cassava production as a “subsistence crop belonging to women” is changing, as a large share of the output is
intended for commercialization, and the labor requirements are partially met by male members of the households (Fermont et al., 2010; Parmar
et al., 2017).
2
Cassava is a major crop in Nigeria, Mozambique, Zambia, DR Congo, Ghana, Malawi, Uganda, and Tanzania.
294 | ADOPTION OF IMPROVED CASSAVA VARIETIES AND INTERCROPPING
FIGURE 1 Cassava yields over time in Malawi, Tanzania, and Uganda Source: FAO (2023).
One of the critical constraints to cassava productivity in ECA is the lack of adoption of improved
cassava varieties (ICV) bred to be resistant or tolerant to pests and diseases (Alene, 2013; Mbanjo
et al., 2021). Many farmers still plant local varieties with low yields and high susceptibility to certain
pests and diseases (Houngue et al., 2018). In particular, two diseases, cassava mosaic disease (CMD)
and cassava brown streak disease (CBSD), have negatively affected cassava production in Uganda,
Tanzania, and Malawi (Chikoti et al., 2019; Macfadyen et al., 2018; Patil et al., 2015; Tomlinson
et al., 2018; Vurro et al., 2010). At least three cryptic species vector the viruses that cause these
diseases in the whitefly Bemisia tabaci complex known as the sub‐Saharan African sub‐group
(Elfekih et al., 2021; Mugerwa et al., 2018). The disease can also be spread by moving infected
cassava cuttings (Boykin et al., 2018; Macfadyen et al., 2018). This paper uses the common name
“cassava whitefly” to refer to multiple species in the B. tabaci complex, including cassava within
their host‐plant range. Outbreaks of cassava whitefly and associated viruses have caused significant
crop losses in ECA countries, resulting in hunger, food insecurity, and annual crop losses of
approximately $1.25 billion (Legg et al., 2006; Macfadyen et al., 2021). For example, in 2010, CBSD
was listed among the seven most harmful plant diseases globally because of its negative impact on
food security in Africa (Pennisi, 2010). Furthermore, cassava whitefly and disease outbreaks have
contributed to high levels of technical inefficiency among cassava farmers in ECA by reducing
cassava yields (Mwebaze et al., 2022).
This paper contributes to our scientific understanding of the adoption process by jointly
analysing the decision to adopt ICVs and legume intercropping (LI) to mitigate whitefly pests and
diseases among smallholder farmers in Malawi, Tanzania, and Uganda. The results highlight the
importance of extension and credit access and their interactions with yield‐related characteristics
on the farmer's decision to adopt the ICVs and LI. The results show that access to extension is
associated with an increase in the attractiveness of yield‐improving characteristics. Similarly, we
find that access to credit was associated with an improvement in the attractiveness of the yield
attributes. These findings are robust because the conclusions are similar across both the ICV and
LI technologies in Malawi, Tanzania, and Uganda. This study will help inform the development of
appropriate technologies by identifying constraints in production, desirable characteristics, and
farmer preferences for certain improved cassava technologies. The rest of this paper is structured
as follows: The next section describes the relationship between ICVs and LI, the methods, and
the survey instrument used in the study, followed by the results, discussion, and policy
implications.
2 | I C V s A N D IN T E R C R O P P I N G S Y S T E M S
Intercropping is a powerful strategy to promote a diversified cropping system, creating

complementary relationships between crop plants (Duchene et al., 2017). Legumes are a
fundamental functional group highly valued for their ecological services in smallholder cassava
systems. Estimates suggest that about 30%–50% of the cassava produced in Africa is intercropped
(Parmar et al., 2017; Weerarathne et al., 2017). In Malawi, Tanzania, and Uganda, LI is common
among farmers seeking to diversify their cropping systems (Koppmair et al., 2017). Short‐duration
grain legumes, such as common beans, mungbeans, cowpea, and groundnut, are the most common
crops intercropped with cassava. While intercropping can provide significant benefits, some
conditions are necessary for successful cassava‐LI systems. In general, the choice of intercrops and
the timing of planting are critical factors in avoiding competition for resources (such as light, water,
and nutrients). The biological advantage of intercropping cassava with legumes declines with the
legume's growth duration (Mutsaers et al., 1993). Parmar et al. (2017) asserted that the first 6
months are optimal for intercropping before cassava develops its canopy.
Compatible intercrops can significantly affect pest, disease, and weed suppression in smallholder
cassava production systems. Since plant pests and diseases are spread by wind, rainfall and soil,
adding an intercrop can alter population dynamics by impeding infestation, disease development,
and dispersal (Delaquis et al., 2018). Many studies show a beneficial impact of intercropping on
CMD, reporting a 10%–40% reduction in disease incidence in diversified plots (Delaquis et al., 2018;
Fondong et al., 2002; Night et al., 2011). Uzokwe et al. (2016) reported a lower incidence of CMD
and its vector whiteflies (B. tabaci) in the Lake Zone, Tanzania, when cassava was intercropped with
green gram. Besides the whitefly and associated viruses, weed competition is a significant constraint
to cassava yields (Fermont et al., 2010). Uncontrolled weed growth can result in near‐total yield loss
(Chew et al., 2012). While herbicides can be a cost‐effective method for weed control in cassava
(Chew et al., 2012), Gianessi and Williams (2011) found that only 5% of smallholder farmers in
Africa use herbicide on any crop. By contrast, Weerarathne et al. (2017) found that 30%–60% of the
weeds could be reduced by intercropping cassava with other crops, such as legumes.
ICVs can increase cassava's resistance to disease, pests, and environmental constraints, thereby
increasing yields (Acheampong et al., 2022). Crop yields have increased by up to 40% due to breeding
high‐yielding ICVs (Nassar & Ortiz, 2007). The International Centre for Tropical Agriculture (CIAT)
released ICVs in Asia that are believed to be the main drivers of Asia's cassava yield increase over the
past 40 years (Kawano, 2011). Breeding for traits such as root size, drought tolerance, adaptation to poor
soil, and disease resistance has helped improve yield (Acheampong et al., 2022; Alene et al., 2018).
However, there is still a diversity of cassava cultivars (both traditional varieties and ICVs) grown by
farmers in East Africa because not all cassava traits have been the focus of breeding efforts (e.g., complex
traits like cooking time, taste and texture of roots are very cultivar‐specific). Furthermore, some
traditional varieties have high levels of natural tolerance to certain pest species and diseases, but not all
pests and diseases (Katono et al., 2021a; 2021b). Some varieties released to help control CMD and CBSD
pandemics may also be suitable hosts for whitefly pests (Kalyebi et al., 2021; Omongo et al., 2022).
Therefore, the relative benefit of adopting ICVs with improved potential yield will depend on the specific
pest and disease pressure each farmer faces each season and, thus, their ability to realize the potential
yield gain in an uncontrolled environment.
Due to these complexities, few studies have attempted to disentangle the relative yield
impacts of these two approaches (LI and ICV) when combined or in isolation. Ogola et al.
(2013) reported mixed results showing that LI and ICVs decreased cassava's fresh root yield in
South Africa. Grain yield was lower (by 19%, 15%, and 26%, respectively, for cowpea,
groundnut, and chickpea) under intercropping compared with monocropping. The land
equivalent ratios (LER) were greater (41%, 33%, and 46%, respectively, for ICV/groundnuts,
ICV/cowpea, and ICV/chickpea intercropping systems) than mono‐cropped cassava. The study
showed that incorporating legumes into cassava‐based cropping systems could improve the
systems' overall productivity. Hidoto and Loha (2013) reported a similar finding from a study to
determine suitable legumes in cassava‐LI in Ethiopia. They showed that intercropping cassava
(traditional varieties) with haricot bean, cowpea, soybean, and mung beans reduced cassava
yield by 27%, 37%, 52%, and 50%, respectively. However, it resulted in higher land use
efficiencies (82%, 49%, 48%, and 62%, respectively) than for either sole crop. The overall LER
was higher when cassava was intercropped with legumes, suggesting higher productivity. These
studies indicate that the best combination of LI with ICV may be site‐dependent and change
depending on the needs and resources of the farmers. A better understanding of the trade‐offs
between ICV and LI seems necessary to target and increase farmer adoption.
3 | DATA AND METHODS
This study is based on a 2015/2016 cross‐sectional survey data collected from a sample of
smallholder farmers in Malawi, Tanzania, and Uganda. We used various methods to address the
study objectives, including literature reviews, workshops, household surveys, and field observations.
The socioeconomic data was analyzed using descriptive statistics such as means, standard deviation,
and frequency. In contrast, linear regression and double hurdle models (DHMs) were used to
analyze the factors influencing the adoption of ICV and LI. This section details the survey data and
econometric modeling approaches used in the study.
3.1 | Study areas

The study was conducted in Malawi, Tanzania, and Uganda and focused on districts with high
cassava whitefly populations and communities most affected by CMD and CBSD outbreaks. We
targeted areas with similar biophysical and Socioeconomic characteristics to allow the identification
of a clear causal influence and effects on households. The study involved designing and developing
survey tools, training enumerators, workshops, and pretesting survey tools and questionnaires in the
study countries. In Uganda, the study specifically targeted the major cassava‐producing districts in
the north (Apac), central (Nakasongola and Kiryandongo), and eastern regions (Serere, Tororo, and
Kamuli). In Tanzania, the survey covered the districts of Geita, Musoma, Rorya, and Ukerewe. In
Malawi, the study targeted the main growing areas in the northern belt along the lakeshore
(Karonga, Nkhata Bay, and Nkhotakota) and the central belt (Lilongwe). The distribution of sample
households across selected regions is shown in Supporting Information S1: Tables A1–A3.
3.2 | Sampling procedure and data collection

Data and information used in this study were obtained from primary and secondary sources. The
primary data were collected from a sample of 1200 cassava farmers in Malawi, Tanzania, and
Uganda. A multistage random sampling procedure was used to collect data. In the first stage of the
sampling procedure, six districts in each country were purposively selected based on their cassava
production statistics in past years. Each district was assigned an equal number of sample
households. The second stage was the random selection of 4–6 villages within the selected districts
and the random selection of 50 cassava households per village. This gave 1200 households: 450 in
Uganda, 400 in Malawi, and 350 in Tanzania. The sampling strategy used ensured an unbiased and
representative sample of the population of cassava producers in the study countries. We generated
the sampling frame from which farmers were selected randomly from a list of farmers obtained from
the district agricultural officers in each country. The distribution of sample households across
communities in each country is shown in Supporting Information S1: Appendix 1–3. It was
desirable to have a sample representative of all cassava farmers in the country for generalization. We
used sampling weights specified as the inverse of the probability of inclusion of the observations to
fit the regression models (Cameron & Trivedi, 2010).
Data were collected using a pretested structured questionnaire completed during face‐to‐face
interviews by trained enumerators and extension staff with individual farmers.3 Respondents were
interviewed by appointment, which was made through the district extension workers and contact
farmers. Pretesting was conducted using 20 farmers in each country to assess the suitability of the
questionnaire. Data collected included Socioeconomic characteristics of respondents such as age,
gender, education levels, household income, and cassava production, such as acreage under cassava,
output, and cassava variety used. The respondents were provided additional information, including
color pictures of the cassava whitefly pests and diseases with associated symptoms, to minimize
potential bias from farmers' self‐assessments. Field observations supplemented information obtained
from interviews. The final surveys were conducted during October 7–31, 2015, in Uganda and Malawi
and August 15–21, 2016, in Tanzania. The survey was completed on average in 2–3 h per respondent.
It was reviewed and approved by the CSIRO Ethics Committee before its release in 2015 and 2016.
3.3 | Descriptive statistics
Table 1 summarizes the descriptive statistics for the sample used in the final model. Overall, these
statistics suggest that the sample was representative of the general profile of farmers in the study
countries. A comparison with national statistics for Malawi (National Statistical Office [NSO]
[Malawi] and ICF, 2017), Tanzania (National Bureau of Statistics NBS, 2017), and Uganda (Uganda
Bureau of Statistics [UBOS], 2021) indicates that the sample age, education, and household size were
not significantly different to that of the general population. Concerning institutional factors, most
cassava farmers lacked access to credit, and only 40% were members of a farmer's organization. The
differences in institutional factors among the study countries may also explain adoption rates. For
example, looking at Table 1, more of the sample in Malawi has access to extension versus Uganda,
where more have access to credit; in Malawi, despite a high population density, farmers tend to be
farther from roads compared to Uganda. This suggests there may be reasons to test if these
differences impact the influence of characteristics on demand that is worth identifying. More
statistics on ICV adoption levels are provided in Supporting Information.
3.4 | Empirical strategy

A linear regression model was used to estimate the determinants of adoption and intensity of ICV
and LI in the study countries. One advantage of using the ordinary least squares (OLS) method is
that it allows researchers to check the coefficient stability of the results and test for the potential
influence of omitted variables, which are a significant limitation to the identification of causal effects
(Diegert et al., 2022; Oster, 2019). We begin with the following model specification:
Yi = βXi + εi, (1)
where: Yi is the outcome variable (adoption of ICVs and LI); Xi is a vector of explanatory variables of
interest; β is the vector of coefficients to be estimated, and εi is error term assumed to be normally
distributed with constant variance and zero mean. The main outcome variables are: (1) the adoption
3
The survey was conducted by staff from the CSIRO in collaboration with the respective countries’ national agricultural research institutions.
These included the National Root Crops and Resources Research Institute in Uganda, Mikocheni Agricultural Research Institute and
Agricultural Research Institute Ukiriguru in Tanzania, and the Department of Agricultural Research Services in Malawi.
TABLE 1 Descriptive statistics of the sample.
Malawi Tanzania Uganda

Variables (n = 400) (n = 350) (n = 450) Pooled (n = 1200)
Variety attributes
Early maturity 0.75 (0.43) 0.67 (0.47) 0.57 (0.50) 0.66 (0.47)
Cooking quality 0.57 (0.50) 0.35 (0.48) 0.64 (0.48) 0.52 (0.49)
Yield 0.92 (0.27) 0.69 (0.46) 0.80 (0.40) 0.80 (0.38)
Big root size 0.61 (0.49) 0.51 (0.50) 0.49 (0.50) 0.54 (0.50)
Pest and disease resistance 0.30 (0.46) 0.21 (0.41) 0.36 (0.48) 0.29 (0.45)
Yield attribute (combined) 0.98 (0.15) 0.86 (0.35) 0.91 (0.29) 0.92 (0.26)
Socioeconomic characteristics
Age (years) 47.41 (15.16) 51.09 (13.49) 46.06 (14.65) 48.18 (14.43)
Male household head (%) 65 80 76 73.67
Education of HHD (years) 5.88 (3.38) 8.72 (5.95) 8.13 (4.13) 7.58 (4.48)
Household size 6.32 (2.65) 7.52 (3.75) 8.56 (3.91) 7.47 (3.44)
Farm characteristics
Farm size (acres) 2.69 (1.97) 4.26 (3.54) 9.84 (20.10) 5.60 (8.54)
Proportion of land under improved 9.14 (16.76) 54 (35) 53.88 (38.46) 39.00 (30.10)
cassava variety (%)
Improved cassava variety (1/0) 0.51 (0.50) 0.11 (0.31) 0.70 (0.57) 0.44 (0.43)
Intercropping system (1/0) 0.36 (0.47) 0.71 (0.45) 0.30 (0.46) 0.46 (0.46)
Income from cassava (US$/acre) 70.40 149.42 84.80 101.54
Institutional characteristics
Access to credit (%) 16 22 33 23.67
Member of organization (%) 34 43 47 41.33
Access to extension services (%) 45 31 28 34.67
Extension visits 5.39 (6.03) 3.48 (4.24) 2.62 (2.03) 3.83 (4.10)
Distance to nearest roads (km) 2.55 (16.71) 2.70 (15.95) 1.29 (4.38) 2.18 (12.35)
Distance to market (km) 4.86 (25.39) 4.56 (12.19) 4.66 (5.73) 4.69 (4.44)
Land tenure
Freehold (dummy) 0.09 (0.29) 0.38 (0.49) 0.28 (0.45) 0.25 (0.41)
Leasehold (dummy) 0.01 (0.10) 0.01 (0.09) 0.04 (0.19) 0.02 (0.13)
Customary (dummy) 0.87 (0.34) 0.59 (0.49) 0.62 (0.49) 0.59 (0.43)
Others (dummy) 0.03 (0.17) 0.03 (0.16) 0.06 (0.23) 0.04 (0.20)
Livestock ownership (%) 85 89 97 90.33
Note: Figures in brackets are standard deviations. HHD means household head. Improved cassava varieties and intercropping were measured as
categorical/dummy variables using a 1/0 scale.
Source: Field surveys, 2015/2016.
decision (0/1) for the full sample, (2) the amount adopted conditional on adoption (the intensive
margin decision for the adopting subsample), and (3) the overall margin‐full sample‐continuous
adoption including the nonadopting 0 s. The main explanatory variables are technology
characteristics (yield‐related attributes), extension, and credit access. Farm and household
characteristics and district‐fixed effects were included as controls. There was concern that the
nuanced differences in yield‐related characteristics may cause imprecision in the model estimates of
what factors farmers find important. Specifically, models including root size, pest and disease
resistance, and yield as three separate variables may lead to inconsistent estimates because the two
former traits will lead to the latter. We solved this problem by combining all three into one binary
variable (=1 if any of the three is 1).4
We implemented the methods Diegert et al. (2022) and Oster (2019) developed to quantify
omitted variables bias. These papers defined a set of sensitivity parameters that index relaxations of
the assumption that the covariate of interest is not correlated with unobserved variables. We run the
following model specification:
Yi = βXi + γ0 W0 + γ1 W1 + γ2 W2 + εi, (2)
where: Yi is the outcome variable (adoption of ICVs and LI); Xi is the explanatory variable of interest;
W0 is a vector of the observed controls, and W2 is not observed. W2 is an omitted variable potentially
correlated with the variables (X , W0, W1). The parameter of interest is β long , the coefficient of the
covariate of interest in the infeasible regression that includes the unobserved variables. Since the
methods developed by Diegert et al. (2022) require a single “treatment” variable, we selected extension
as the primary variable of interest, while the other variables were used as control variables to help
identify the causal effects of extension in the model. Three scalar sensitivity parameters are of interest,
denoted as (rX,c). Given the joint distribution of the observed variables and values of the sensitivity
parameters, Diegert et al. (2022) illustrate how to calculate the upper and lower bounds of the
identified set for the coefficient of interest, denoted by βL ( r̅X , r̅ y , c̅ ). We also implemented the
sensitivity analysis in Oster (2019), which uses different sensitivity parameters. Readers should consult
Diegert et al. (2022) and Oster (2019) for more details about the sensitivity analysis.
As a robustness check, we estimated the DHM proposed by Cragg (1971) to investigate the factors
affecting the intensity of the adoption of ICV and LI. This method is widely applied in empirical
adoption studies (e.g., Awotide et al., 2014; Kondo et al., 2020; Opata et al., 2021; Singbo et al., 2021).
The model assumes that farmers make two different decisions, the decision on whether to adopt or not
the ICVs and LI (the first hurdle) and the decision on how much to adopt the ICVs and LI (the second
hurdle), and that the explanatory factors of these decisions may differ (Burke, 2009, 2019;
Wooldridge, 2010). The DHM relaxes the restrictive assumption of the Tobit model, which assumes
that the decisions on whether or not to adopt the ICVs/LI and how much to adopt are made jointly,
and therefore, the factors affecting the two decisions are assumed to be the same in the Tobit model
(Burke, 2009). We provide a detailed description of the DHM in Supporting Information.
4 | RESULTS A ND DISCUSSION
4.1 | Determinants of adoption of ICV and intercropping

This section presents the results of the OLS adoption models of ICV and LI pooled across the study
countries. The full estimation results for each study country separately are shown in the supporting
4
Correlation tests confirmed positive correlations among the three variables. We thank the reviewers for this observation which has led to more
consistent and statistically significant coefficient estimates.
information (Supporting Information S1: Tables A2–A4). Tables 2–3 present the results of the OLS
models using the pooled data for ICV and LI in the study countries, respectively. The R‐squared
values suggest the models provide a reasonable fit for explaining the relationship between adoption
and the explanatory variables. Several variables related to farm and socioeconomic characteristics
T A B L E 2 Coefficient estimates of the linear (OLS) regression model of improved cassava variety (ICV) adoption in
Malawi, Tanzania, and Uganda (pooled).
Pooled model
Variables Adoption decision (1) Intensive margin (2) Overall margin (3)
Technology characteristics
Early maturity 0.20 (0.04)*** 0.01 (0.03) 0.13 (0.03)***
Yield 0.06 (0.08)** 0.12 (0.07)** 0.04 (0.06)*
Cooking quality 0.03 (0.03) 0.03 (0.02) 0.01 (0.03)
Institutional variables
Group membership 0.07 (0.03)** 0.03 (0.02) 0.09 (0.04)**
Extension 0.04 (0.12)** 0.35 (0.18)** 0.06 (0.15)**
Credit 0.45 (0.17)*** 0.48 (0.18)** 0.05 (0.11)
Yield × extension 0.15 (0.13)** 0.40 (0.18)** 0.03 (0.15)**
Yield × credit 0.44 (0.17)** 0.45 (0.17)** 0.08 (0.10)**
Farm size 0.07 (0.02)** 0.19 (0.02)*** 0.15 (0.02)***
Freehold ownership 0.06 (0.08) 0.06 (0.07) 0.08 (0.06)
Leasehold ownership 0.08 (0.11) 0.14 (0.09) 0.11 (0.08)
Customary ownership 0.01 (0.07) 0.07 (0.06) 0.07 (0.05)
Age 0.01 (0.05) −0.05 (0.04) −0.03 (0.04)
Education 0.03 (0.03) 0.07 (0.03)** 0.03 (0.02)
Gender −0.05 (0.04) −0.01 (0.03) −0.02 (0.03)
Household size −0.03 (0.03) −0.03 (0.02) −0.03 (0.02)
Livestock 0.04 (0.06) 0.12 (0.06)** 0.09 (0.05)*
Dist. to the nearest road −0.01 (0.03) −0.01 (0.02) −0.01 (0.02)
Dist. to the nearest market −0.02 (0.02) −0.02 (0.01) −0.02 (0.01)
Country dummies
Malawi 0.68 (0.05)*** 0.45 (0.06)*** 0.51 (0.04)***
Uganda 0.69 (0.06)*** 0.46 (0.06)*** 0.52 (0.05)***
Constant −0.12 (0.22) 0.83 (0.21)*** 0.83 (0.17)***
Sensitivity analysis (Oster, 2019)

bp
δ̂ resid (%) 46.0 103.8 24.8
TABLE 2 (Continued)
Pooled model
Sensitivity analysis (Diegert et al.,

2022)
bp
r̄ˆX (%) 26.9 16.2 10.3
R2 0.50 0.51 0.52
Obs. 1200 500 1200
Note: Dependent variable is the adoption of ICVs. We use three estimations for the dependent variable—(1) the adoption decision (0/1) for the
full sample, (2) the amount adopted conditional on adoption (the intensive margin decision for the adopting subsample), and (3) the overall
margin‐full sample‐continuous adoption including the nonadopting 0 s. Country‐fixed effects are included in the pooled models, with Tanzania
used as the reference group.
Standard errors in parenthesis. Significance levels:
*p < 0.1; **p < 0.05; ***p < 0.01.
Source: Farmer surveys (2015/2016).
have significant effects. We avoid a detailed discussion of all explanatory factors to keep the focus on
the impact of the variety characteristics, extension, and credit access on the adoption of ICVs/LI.
Results suggest some heterogeneity regarding the factors influencing the adoption decision, intensive
margins, and overall adoption margins for ICV and LI in the study countries (Supporting
Information S1: Tables A2–A4). Overall, the results indicate that varietal characteristics, extension,
and credit access were positively related to the adoption of ICV and LI across the study countries.
Early maturity and yield performance coefficients were positive and significantly related to adoption,
intensive margins, and overall adoption margins across the study countries. Furthermore, we
included interactions to test whether institutional factors increase the attractiveness of the yield‐
improving characteristics. We found that these differences impact the influence of attributes on
demand, which is worth highlighting. Results show that access to extension is associated with an
increase in the attractiveness of yield‐improving characteristics (Tables 2–3). Similarly, access to
credit is correlated to an improvement in the attractiveness of yield attributes. These results provide
valuable policy insights to increase technology adoption using a more targeted extension approach.
They show that farmers' assessment of the ICV characteristics greatly influences their adoption
behavior. Therefore, the omission of farmers' evaluation of technology‐specific attributes may bias
the results of the factors conditioning adoption choices.
These results are consistent with the literature on consumer demand, showing that demand is
affected by product attributes (Mafuru et al., 2008; Miriti et al., 2022; Opata et al., 2021). Farmers are
consumers of agricultural research products, and their preferences for the characteristics of new
agricultural technologies affect their adoption decisions (Kalema et al., 2022). The findings are
similar to those obtained by Adesina and Zinnah (1993) for improved rice varieties in Sierra Leone
and Adesina and Baidu‐Forson (1995) for improved sorghum and rice varieties in Burkina Faso and
Guinea. A more recent study by Kondo et al. (2020) showed that the high‐yielding attribute of ICV
increased the probability of adoption by 11% in Ghana. In Uganda, Iragaba et al. (2021) found the
softness of boiled roots and ground storability as key attributes influencing varietal preference
besides yield, disease resistance, early maturity, and drought resistance. In Tanzania, a study by
Mafuru et al. (2008) found that the color and taste of sorghum were the most important
determinants used by consumers to evaluate the quality of sorghum. Our results suggest that
farmers' assessments of variety‐specific attributes matter and should be considered in assessing the
drivers of adoption decisions of ICVs and LI.
Furthermore, we found a positive relationship between farm size and the intensive and overall
adoption margins for both the ICV and LI. This result indicates that farmers with larger farm sizes
were more likely to adopt the ICV and LI. Farmers holding a large piece of land are likely to have
more opportunities to learn about new technologies, have more incentives to embrace them and can
bear risks associated with new technologies. Adoption also varied by location (See district dummy
variables in Supporting Information S1: Tables A2–A4). These results will likely reflect unobservable
spatial differences among the study districts or interactions between location and the other
T A B L E 3 Coefficient estimates of the linear (OLS) regression model of legume intercropping (LI) adoption in Malawi,
Tanzania, and Uganda (pooled).
Pooled model
Early maturity 0.08 (0.05)** 0.06 (0.04)* 0.07 (0.03)**
Yield 0.14 (0.09)* 0.15 (0.10) 0.06 (0.02)**
Cooking quality 0.07 (0.04)* 0.06 (0.03)* 0.03 (0.06)
Group membership 0.03 (0.04) 0.01 (0.04) 0.01 (0.03)
Extension 0.17 (0.20)** 0.40 (0.14)** 0.13 (0.17)**
Credit 0.22 (0.14)* 0.01 (0.04) 0.21 (0.11)*
Yield × extension 0.21 (0.19) 0.34 (0.14)** 0.01 (0.07)*
Yield × eredit 0.33 (0.14)** 0.17 (0.08)**
Farm size −0.010 (0.03) 0.17 (0.03)*** 0.17 (0.02)***
Freehold ownership 0.12 (0.11) 0.10 (0.15) 0.06 (0.06)
Customary ownership 0.04 (0.10) 0.14 (0.14) 0.08 (0.06)
Age −0.13(0.06)** −0.05 (0.05) −0.06 (0.04)*
Education 0.02 (0.04) 0.03 (0.04) 0.01 (0.02)
Gender −0.01 (0.04) −0.04 (0.04) −0.01 (0.03)
Household size 0.05 (0.04) 0.03 (0.04) 0.02 (0.03)
Livestock 0.02 (0.07) 0.05 (0.07) 0.03 (0.04)
Dist. to the nearest road −0.02 (0.03) −0.02 (0.03) −0.01 (0.01)
Dist. to the nearest market −0.02 (0.03) −0.02 (0.03) −0.01 (0.02)
Country dummies
Malawi 0.01 (0.10) 0.35 (0.06)*** 0.43 (0.04)***
Uganda 0.02 (0.09) 0.36 (0.07)*** 0.19 (0.06)***
Constant 0.69 (0.29)** 1.22 (0.29)*** 0.73 (0.15)***
Sensitivity analysis (Oster, 2019)

bp
δ̂ resid (%) 7.5 39.1 25
TABLE 3 (Continued)
Pooled model
Sensitivity analysis (Diegert et al., 2022)

bp
r̄ˆX (%) 6.59 17.7 16.7
2
R 0.45 0.51 0.53
Obs. 1200 500 1200
Note: Dependent variable is the adoption of LI. We use three estimations for the dependent variable—(1) the adoption decision (0/1) for the full
sample, (2) the amount adopted conditional on adoption (the intensive margin decision for the adopting subsample), and (3) the overall
margin‐full sample‐continuous adoption including the nonadopting 0 s. Country‐fixed effects are included in the pooled models, with Tanzania
used as the reference.
*p < 0.1; **p < 0.05; ***p < 0.01.
identified factors. For instance, the location of farms influences how readily farmers can access
information on technologies. Location‐specific effects may also affect the acquisition of inputs and
credit to expand the area under cultivation (Owusu & Donkor, 2012).
We examined the sensitivity of regression coefficient estimates to the presence of omitted
variables. Tables 2 and 3 report estimates of the sensitivity parameters. Based on Oster (2019), the
absolute value of the sensitivity parameter delta |δ| > 1 implies limited scope for unobservable
bp
variables to explain the results. Based on the δ̂ resid estimates, unobservable factors are unlikely to
bp
drive our results. Tables 2 and 3 also report estimates of the breakdown point (r̄ˆX ) developed by
Diegert et al. (2022). This statistic represents the largest amount of selection on unobservable
factors, as a percentage of selection on observable factors allowed until one can no longer conclude
that the coefficient of interest (β) is nonzero. Larger values of r̄ bpX indicate the robustness of the
results. The breakdown point estimates in columns (1)–(3) are relatively small, between 10.3% and
26.9%. This implies that we only need the selection of unobservable factors to be at least
10.3%–26.9% as large as a selection of observable factors to overturn the conclusion that the
coefficient of extension (β) is zero. Thus, without imposing any restrictions on the impact of
unobservable factors on outcomes, we find that the analysis is sensitive to selection on unobservable
factors. Supporting Information S1: Figure A5 illustrates one example using the adoption decision
estimation for ICV. The solid lines show the estimated identified set for the coefficient of extension
(β) as a function of r̅X , allowing for endogenous controls and no restrictions on the outcome
bp
equation. The horizontal intercept is the estimated breakdown point r̄ˆX = 26.9, as reported in
column (1) of Table 2. Overall, while some cases could be considered robust, we find that the results
are generally susceptible to omitted variables.
4.2 | Robustness results

As a robustness check, we estimated the DHM to investigate the factors affecting the adoption and
intensity of the adoption of ICV and LI. Results of the DHM for the factors affecting the likelihood
of adoption and use intensity of the ICV using the pooled data across the study countries are
summarized in Table 4.5 The full estimation results for each study country are shown in Supporting
5
The results of the robustness check for LI were very similar to the ICV results and are not reported for brevity purposes.
TABLE 4 Robustness results using the DHM for improved cassava varieties (ICV) in Malawi, Tanzania, and Uganda
(pooled).
Pooled model
Variables First part (adoption) Second part (intensity) Marginal effects
Early maturity 0.07 (0.13)* 0.13 (0.06)** 0.02 (0.04)**
Yield 0.17 (0.33)* 0.03 (0.13)* 0.07 (0.08)**
Cooking quality 0.15 (0.13)* 0.08 (0.07)* 0.01 (0.04)
Group membership 0.03 (0.14)* 0.1 (0.09)* 0.04 (0.05)
Extension 0.19 (0.70)* 0.08 (0.19)* 0.12 (0.05)**
Credit 0.10 (0.73)* 0.10 (0.16)* 0.08 (0.05)**
Yield x Extension 0.03 (0.71)* 0.06 (0.21)**
Yield x Credit 0.07 (0.74)* 0.10 (0.15)**
Farm size 0.71 (0.13)*** 0.20 (0.08)** 0.03 (0.05)***
Freehold ownership 0.01 (0.16)* 0.15 (0.09) 0.06 (0.06)
Customary ownership 0.42 (0.29)* 0.01 (0.13) 0.08 (0.09)
Socioeconomic factors
Age 0.24 (0.19)* 0.11 (0.10)* 0.09 (0.06)*
Education 0.23 (0.12)** 0.14 (0.06)** 0.11 (0.04)***
Gender 0.02 (0.14) 0.05 (0.08) 0.03 (0.04)
Household size 0.01 (0.15) 0.09 (0.08)* 0.04 (0.05)
Livestock 0.49 (0.26)** 0.02 (0.23) 0.12 (0.07)*
Dist. to the nearest road −0.03 (0.09)* −0.01 (0.04)* −0.01 (0.03)*
Dist. to the nearest market −0.05 (0.09)* −0.02 (0.04)* −0.03 (0.02)*
Country dummies
Malawi 0.09 (0.16)** 0.06 (0.09)** 0.05 (0.06)
Uganda 0.10 (0.10)** 0.07 (0.10)**
Constant 2.68 (0.80)*** −0.67 (0.54)*
Log pseudolikelihood −425.17
Wald χ 2
36.69
Sigma (σ) 0.45 (0.04)***
Obs. 1200
Note: The dependent variable is the observed intensity of adoption of ICVs given as the proportion of land acres allocated to ICVs. Country‐
fixed effects are included in the pooled models, with Tanzania used as the reference group.
*p < 0.1; **p < 0.05; ***p < 0.01.
Information S1: Table A4. The first‐order coefficients, average marginal effects, and measures of fit
of the two‐part models are shown.6 The likelihood ratios of the estimated models are significant at
the 1% probability level, indicating the joint significance of the explanatory variables included in the
models. Several explanatory variables were again highly statistically significant (p < 0.05) as observed
in the OLS models. We find similarities and differences in the factors that explain the likelihood of
adoption and use intensity of ICVs across Malawi, Tanzania, and Uganda. This justifies using the
DHM in examining the adoption of ICVs/LI in this study. Several Socioeconomic characteristics
had the expected signs and were significant in explaining adoption decisions. As mentioned, we
refrain from a detailed discussion of all influencing factors to focus only on the key variables of
interest—variety characteristics, extension, and credit access—and we treat the rest as mere controls.
Varietal traits affect both the adoption and intensity decisions for the ICV, although at different
significance levels. This suggests that the characteristics are critical drivers of the likelihood of
adoption and use intensity of ICVs/LI, and this finding holds across the three study countries. Yield
performance was positively correlated with the probability of adoption and use intensity of ICVs at
the 5% level in the combined model. The high yield attribute is associated with a 7% increase in the
probability of adopting ICVs, followed by the early maturity period, which is associated with a 2%
increase in the likelihood of adoption (Table 4). Furthermore, we found that access to extension is
related to the rise in the attractiveness of yield‐improving characteristics (Table 4). Likewise, access
to credit was associated with an improvement in the attractiveness of the yield attributes.
Membership in farmer's organizations was also significantly (p < 0.05) related to adopting the ICV.
Farm size positively influenced the adoption and use intensity of the ICV at the 5% level across the
study countries, suggesting scale effects. An additional acre of land is associated with a 3% increase
in the probability of adopting ICVs. This finding implies a scale technology—only farmers with
certain land sizes would likely embrace it. The literature shows that ownership of larger farm sizes
increases the likelihood of adoption and use intensity of ICVs (Opata et al., 2021). Distance to the
nearest market was negatively correlated with the probability of adoption and use intensity of the
ICVs/LI at the 5% level across most of the study countries. This suggests that the closer a household
is to a market, the higher the likelihood of adopting ICVs. This finding agrees with Kondo et al.
(2020) for ICV in Ghana, who reported that differences in households' access to financial and
commodity markets significantly influenced technology adoption decisions.
There were also limitations to our study. The adoption of ICV can be affected by other factors,
such as profitability and risk associated with technology adoption (Michler et al., 2019), which were
not considered here. Future studies should examine the profitability, risk, environmental, and
welfare implications of ICV and LI adoption (e.g., see Suri & Udry, 2022 and Michler et al., 2019 for
recent analyses of technology adoption in Africa). Our results were generally sensitive to omitted
variables. It would be interesting to explore further the implications of omitted variables on
technology adoption in future empirical work.
5 | CONCLUSIONS AND POLICY IMPLICATIONS
This study investigated the factors that affect the adoption of ICV and LI among smallholder
farmers and provided guidance on the best ways to increase adoption. Our results highlight the
importance of technological characteristics and institutional factors in farmers' decisions to adopt
ICVs and LI. Technology characteristics, such as yield and early maturity, are critical drivers of
adoption behavior. Access to extension is associated with an increase in the attractiveness of yield‐
improving characteristics. Similarly, access to credit was associated with an improvement in the
6
The average marginal effects (AME) measure the change in the probability of adoption or the intensity of adoption of the ICV/LI with a
marginal change in the variable of interest for continuous variables. For an indicator variable, the AME measures the change in the probability
of adoption or intensity of adoption for a discrete change (from 0 to 1) in the indicator variable (Wooldridge, 2010).
attractiveness of the yield attributes. Therefore, varietal attributes must be integrated into ICV
adoption dissemination campaigns throughout the study countries. These findings are robust
because the conclusions are similar across technologies in Malawi, Tanzania, and Uganda.
These results have important policy implications for future interventions. Firstly, the analysis
indicates the need for improved design of extension policies for effective technology promotion to
mitigate the threat of cassava whitefly pests and diseases (also see Paul et al., 2022). These measures
should include targeting technologies and practices that work together to increase productivity and
disseminate information about cassava pest and disease management. Farmers should be given more
frequent education and training on technology adoption, especially concerning ICVs and pest
management (see Chipeta et al., 2016). Governments should expand the current efforts by local and
international nongovernmental organizations (such as the World Bank, the Bill & Melinda Gates
Foundation, FAO, UK DFID, Feed the Future Initiative by USAID, NEPAD Pan African Cassava
Initiative, etc.) to develop sustainable, market‐oriented ICV seed systems for smallholder farmers in
Malawi, Tanzania and Uganda. Ongoing field trials by the respective countries' national agricultural
research institutions to develop ICVs resistant to the cassava whitefly and diseases and high yielding
should be expanded in the study countries.
The results affect ICV research and development in ECA. They show that the yield performance
of ICVs needs to be evaluated by farmers under their field conditions. Varieties that farmers judge as
having superior yield performance over local varieties under their farm conditions tend to be
adopted. Farmers are also important sources of technology information and agents of technology
transfer. When farmers evaluate the characteristics of new ICVs and find them to match their
preferences, they often give the ICVs to other farmers for assessment, thereby setting into motion a
dynamic technology adoption process (Grisley, 1994). The low adoption levels of ICVs in ECA
suggest a lack of markets where farmers can sell their surpluses. Increased yield gains will likely not
be attractive for cassava farmers without additional economic profits, which require markets for
surpluses (see Michler et al., 2019). This implies that focusing policy solely on ICV yield gains may
be insufficient to motivate adoption. Incorporating traits other than yields and improving
households' ability to generate higher returns (through complementary investments that will
enhance cassava value chains and market access) should accompany ICV yield‐increasing programs
in Malawi, Tanzania, and Uganda. The adoption of ICV is likely to increase with improved market
access. Our results imply that the government should support local institutions and service
providers in the presence of market failure since they effectively support farmers by providing
credit, inputs, information, and market outlets.
Finally, governments and NGOs should encourage and support local, national, regional, and
international exchange of experience, skills, and technologies. We recommend field trials to replicate
good examples of optimized cassava production in other countries (e.g., Ivory Coast with no viruses,
no bacteria, clean cuttings, high density, no weeds, optimal management, and 60 t/ha yields). There
are lessons for the cassava sector in ECA based on policy responses undertaken by the Government
of Thailand (e.g., Tijaja, 2009). The Thailand cassava sector grew partly due to government support
for domestic research and development. Thailand actively supported research institutions to
promote ICVs, farmer training, and participation in policy dialogs, and they implemented policies
to encourage trade and agribusiness (Tijaja, 2009). The organization's focus on increasing yield,
market access, and attracting investors can be directly linked to growth in the cassava sector. Similar
sector‐wide initiatives could deliver benefits for East African small‐holder farmers.
ACKNOWLEDGMENTS
The wider CSIRO team, especially Andy Hulthen and Cate Paull, were very helpful in collating the
data and generating the maps. We thank the field team, especially Donald Kachigamba, Christopher
Omongo, Sophia Hamba, and George Sonda, for helping to administer the surveys in Malawi,
Tanzania, and Uganda. This study was funded by the Bill and Melinda Gates Foundation through
the Natural Resources Institute, University of Greenwich (Grant number#OPP1058938).
DATA AVAILABILITY STATE MENT

The data that support the findings of this study are available on request from the corresponding
author. The data are not publicly available due to privacy or ethical restrictions.
REFERENCES
Acheampong, Patricia Pinamang, Monica Addison, and Camillus Abawiera Wongnaa. 2022. “Assessment of Impact of
Adoption of Improved Cassava Varieties on Yields in Ghana: An Endogenous Switching Approach.” Cogent Economics
& Finance 10(1): 2008587. https://doi.org/10.1080/23322039.2021.2008587
Adesina, Akinwumi A., and Jojo Baidu‐Forson. 1995. “Farmers’ Perceptions and Adoption of New Agricultural Technology:
Evidence From Analysis in Burkina Faso and Guinea, West Africa.” Agricultural Economics 13(1): 1–9. https://doi.org/
10.1111/j.1574-0862.1995.tb00366.x
Adesina, Akinwumi A., and Moses M. Zinnah. 1993. “Technology Characteristics, Farmers’ Perceptions and Adoption
Decisions: A Tobit Model Application in Sierra Leone.” Agricultural Economics 9(4): 297–311. https://doi.org/10.1111/j.
1574-0862.1993.tb00276.x
Alene, Arega D., Tahirou Abdoulaye, Joseph Rusike, Ricardo Labarta, Bernardo Creamer, Martha del Río, Hernan Ceballos,
and Luis Augusto Becerra. 2018. “Identifying Crop Research Priorities Based on Potential Economic and Poverty
Reduction Impacts: The Case of Cassava in Africa, Asia, and Latin America.” PLoS One 13(8): e0201803. https://doi.org/
10.1371/journal.pone.0201803
Alene, Arega. D., R. Khataza, C. Chibwana, P. Ntawuruhunga, and C. Moyo. 2013. “Economic Impacts of Cassava Research
and Extension in Malawi and Zambia.” Journal of Development and Agricultural Economics 5(11): 457–69. https://doi.
org/10.5897/JDAE2013.0496
Awotide, B. A., T. Abdoulaye, A. Alene, and M. V. Manyong. 2014. “Assessing the Extent and Determinants of Adoption of
Improved Cassava Varieties in South‐Western Nigeria.” Journal of Development and Agricultural Economics 6(9): 376–
85. https://doi.org/10.5897/JDAE2014.0559
Bennett, Ben. 2015. “Guest Editorial: Smallholder Cassava Production and the Cassava Processing Sector in Africa.” Food
Chain 5(1): 1–3. https://doi.org/10.3362/2046-1887.2015.001
Boykin, Laura M., Tonny Kinene, James M. Wainaina, Anders Savill, Susan Seal, Habibu Mugerwa, Sarina Macfadyen, et al.
2018. “Review and Guide to a Future Naming System of AfricanBemisia Tabacispecies.” Systematic Entomology 43(3):
427–33. https://doi.org/10.1111/syen.12294
Burke, William J. 2009. “Fitting and Interpreting Cragg's Tobit Alternative Using Stata.” The Stata Journal: Promoting
Communications on Statistics and Stata 9(4): 584–92. https://doi.org/10.1177/1536867X0900900405
Burke, William J. 2019. “Evidence Against Imposing Restrictions on Hurdle Models As a Test for Simultaneous Versus
Sequential Decision Making.” American Journal of Agricultural Economics 101(5): 1473–81. https://doi.org/10.1093/
ajae/aaz026
Cameron, Adrian Colin, and P. K. Trivedi. 2010. Microeconometrics using Stata. College Station, Tex: Stata Press
Chew, A., E. Slakie, A. Cullen, and C. L. Anderson, 2012. Potential Environmental and Socioeconomic Effects of Herbicide
Resistant Cassava. Evans Policy Analysis & Research Group, Brief #200.
Chikoti, Patrick Chiza, Rabson Mpundu Mulenga, Mathias Tembo, and Peter Sseruwagi. 2019. “Cassava Mosaic Disease: A
Review Of a Threat to Cassava Production in Zambia.” Journal of Plant Pathology 101(3): 467–77. https://doi.org/10.
1007/s42161-019-00255-0
Chipeta, Michael M., Paul Shanahan, Rob Melis, Julia Sibiya, and Ibrahim R. M. Benesi. 2016. “Farmers’ Knowledge of
Cassava Brown Streak Disease and Its Management in Malawi.” International Journal of Pest Management 62(3): 175–
84. https://doi.org/10.1080/09670874.2016.1167268
Cragg, John G. 1971. “Some Statistical Models for Limited Dependent Variables With Application to the Demand for
Durable Goods.” Econometrica 39(5): 829. https://doi.org/10.2307/1909582
Delaquis, E., S. De Haan, and K. A. G. Wyckhuys. 2018. “On‐Farm Diversity Offsets Environmental Pressures in Tropical
Agro‐Ecosystems: A Synthetic Review for Cassava‐Based Systems.” Agriculture, Ecosystems & Environment 251: 226–35.
https://doi.org/10.1016/j.agee.2017.09.037
Dethier, Jean‐Jacques, and Alexandra Effenberger. 2012. “Agriculture and Development: A Brief Review of the Literature.”
Economic Systems 36(2): 175–205. https://doi.org/10.1016/j.ecosys.2011.09.003
Diegert, Paul, Matthew A. Masten, and Alexandre Poirier. 2022. Assessing Omitted Variable Bias when the Controls are
Endogenous. https://doi.org/10.48550/ARXIV.2206.02303
Duchene, Olivier, Jean‐François Vian, and Florian Celette. 2017. “Intercropping With Legume for Agroecological Cropping
Systems: Complementarity and Facilitation Processes and the Importance of Soil Microorganisms. A Review.”
Agriculture, Ecosystems & Environment 240: 148–61. https://doi.org/10.1016/j.agee.2017.02.019
Elfekih, S., W. T. Tay, A. Polaszek, K. H. J. Gordon, D. Kunz, S. Macfadyen, T. K. Walsh, et al. 2021. “On Species
Delimitation, Hybridization and Population Structure of Cassava Whitefly in Africa.” Scientific Reports 11(1): 7923.
https://doi.org/10.1038/s41598-021-87107-z
FAO & IFAD. 2005. “A Review of Cassava in Africa with Country Case Studies on Nigeria, Ghana, the United Republic of
Tanzania, Uganda and Benin.” In Proceedings of the Validation Forum on the Global Cassava Development Strategy, Vol.
2. ftp://ftp.fao.org/docrep/fao/009/a0154e/A0154E00.pdf
FAO. 2023. FAOSTAT. http://www.fao.org/faostat/en/#home.
Feder, Gershon, Richard E. Just, and David Zilberman. 1985. “Adoption of Agricultural Innovations in Developing
Countries: A Survey.” Economic Development and Cultural Change 33(2): 255–98. https://doi.org/10.1086/451461
Fermont, A. M., A. Babirye, H. M. Obiero, S. Abele, and K. E. Giller. 2010. “False Beliefs on the Socio‐Economic Drivers of
Cassava Cropping.” Agronomy for Sustainable Development 30(2): 433–44. https://doi.org/10.1051/agro/2009044
Fermont, A. M., P. J. A. van Asten, P. Tittonell, M. T. van Wijk, and K. E. Giller. 2009. “Closing the Cassava Yield Gap: An
Analysis From Smallholder Farms in East Africa.” Field Crops Research 112(1): 24–36. https://doi.org/10.1016/j.fcr.2009.
01.009
Fondong, V. N., J. M. Thresh, and S. Zok. 2002. “Spatial and Temporal Spread of Cassava Mosaic Virus Disease in Cassava
Grown Alone and When Intercropped With Maize and/or Cowpea.” Journal of Phytopathology 150(7): 365–74. https://
doi.org/10.1046/j.1439-0434.2002.00775.x
Gianessi, Leonard, and Ashley Williams. 2011. “Overlooking the Obvious: The Opportunity for Herbicides in Africa.”
Outlooks on Pest Management 22(5): 211–5. https://doi.org/10.1564/22oct04
Grisley, W. 1994. “Farmer‐to‐Farmer Transfer of New Crop Varieties: An Empirical Analysis on Small Farms in Uganda.”
Agricultural Economics 11(1): 43–9. https://doi.org/10.1016/0169-5150(94)90015-9
Hidoto, Legese, and Gobeze Loha. 2013. “Identification of Suitable Legumes In Cassava (Manihot esculenta Crantz)‐Legumes
Intercropping.” African Journal of Agricultural Research 8(21): 2559–62. https://doi.org/10.5897/AJAR12.1976
Houngue, Jerome Anani, Justin S. Pita, Gilles Habib Todjro Cacaï, Martine Zandjanakou‐Tachin, Emmanuel A. E. Abidjo,
and Corneille Ahanhanzo. 2018. “Survey of Farmers’ Knowledge of Cassava Mosaic Disease and Their Preferences for
Cassava Cultivars in Three Agro‐Ecological Zones in Benin.” Journal of Ethnobiology and Ethnomedicine 14(1): 29.
https://doi.org/10.1186/s13002-018-0228-5
Iragaba, Paula, Sophia Hamba, Ephraim Nuwamanya, Michael Kanaabi, Ritah Ann Nanyonjo, Doreen Mpamire,
Nicholas Muhumuza, Elizabeth Khakasa, Hale Ann Tufan, and Robert Sezi Kawuki. 2021. “Identification of Cassava
Quality Attributes Preferred by Ugandan Users Along the Food Chain.” International Journal of Food Science &
Technology 56(3): 1184–92. https://doi.org/10.1111/ijfs.14878
Jarvis, Andy, Julian Ramirez‐Villegas, Beatriz Vanessa Herrera Campo, and Carlos Navarro‐Racines. 2012. “Is Cassava the
Answer to African Climate Change Adaptation.” Tropical Plant Biology 5(1): 9–29. https://doi.org/10.1007/s12042-012-
9096-7
Johnson, Michael E., William A. Masters, and Paul V. Preckel. 2006. “Diffusion and Spillover of New Technology: A
Heterogeneous‐Agent Model for Cassava in West Africa.” Agricultural Economics 35(2): 119–29. https://doi.org/10.
1111/j.1574-0862.2006.00146.x
Kalema, Elizabeth P., Essegbemon Akpo, Geoffrey Muricho, Justin Ringo, Chris O. Ojiewo, and Rajeev K. Varshney. 2022.
“Mapping Out Market Drivers of Improved Variety Seed Use: The Case of Sorghum in Tanzania.” Heliyon 8(1): e08715.
https://doi.org/10.1016/j.heliyon.2022.e08715
Kalyebi, Andrew, Sarina Macfadyen, Andrew Hulthen, Patrick Ocitti, Frances Jacomb, Wee Tek Tay, John Colvin, and
Paul De Barro. 2021. “Within‐Season Changes in Land‐Use Impact Pest Abundance in Smallholder African Cassava
Production Systems.” Insects 12(3): 269. https://doi.org/10.3390/insects12030269
Kassie, Menale, Moti Jaleta, Bekele Shiferaw, Frank Mmbando, and Mulugetta Mekuria. 2013. “Adoption of Interrelated
Sustainable Agricultural Practices in Smallholder Systems: Evidence from Rural Tanzania.” Technological Forecasting
and Social Change 80(3): 525–40. https://doi.org/10.1016/j.techfore.2012.08.007
Katono, Kasifa, Sarina Macfadyen, Christopher Abu Omongo, John Colvin, Jeninah Karungi, and Michael Hilary Otim.
2021a. “Mortality Factors Acting on Field Populations of Bemisia tabaci (Hemiptera: Aleyrodidae) SSA1 on Cassava in
Uganda.” European Journal of Entomology 118: 148–58. https://doi.org/10.14411/eje.2021.016
Katono, Kasifa, Sarina Macfadyen, Christopher Abu Omongo, Thomas Lapaka Odong, John Colvin, Jeninah Karungi, and
Michael Hilary Otim. 2021b. “Influence of Cassava Morphological Traits and Environmental Conditions on Field
Populations of Bemisia tabaci.” Insects 12(7): 604. https://doi.org/10.3390/insects12070604
Kawano, Kazuo. 2011. “The Triumphant Cassava Chronicled by Foresight, Political Will and Accountability.” Journal of Root
Crops 37(2): 101. http://journal.isrc.in/index.php/jrc/article/view/33
Kondo, Kodjo, Oscar Cacho, Euan Fleming, Renato A. Villano, and Bright O. Asante. 2020. “Dissemination Strategies and
the Adoption of Improved Agricultural Technologies: The Case of Improved Cassava Varieties in Ghana.” Technology
in Society 63: 101408. https://doi.org/10.1016/j.techsoc.2020.101408
Koppmair, Stefan, Menale Kassie, and Matin Qaim. 2017. “The Influence of Farm Input Subsidies on the Adoption of Natural
Resource Management Technologies.” Australian Journal of Agricultural and Resource Economics 61(4): 539–56. https://
doi.org/10.1111/1467-8489.12220
Legg, J. P., B. Owor, P. Sseruwagi, and J. Ndunguru. 2006. “Cassava Mosaic Virus Disease in East and Central Africa:
Epidemiology and Management of a Regional Pandemic.” Advances in Virus Research 67: 355–418. https://doi.org/10.
1016/S0065-3527(06)67010-3
Macfadyen, S., C. Paull, L. M. Boykin, P. De Barro, M. N. Maruthi, M. Otim, A. Kalyebi, et al. 2018. “Cassava Whitefly,
Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae) in East African Farming Landscapes: A Review of the Factors
Determining Abundance.” Bulletin of Entomological Research 108(5): 565–82. https://doi.org/10.1017/
S0007485318000032
Macfadyen, Sarina, Wee Tek Tay, Andrew D. Hulthen, Cate Paull, Andrew Kalyebi, Frances Jacomb, Hazel Parry, et al. 2021.
“Landscape Factors and How They Influence Whitefly Pests in Cassava Fields Across East Africa.” Landscape Ecology
36(1): 45–67. https://doi.org/10.1007/s10980-020-01099-1
J. M. Mafuru, D. W. Norman and J. S. Fox (Eds.). 2008. Consumer Perception of Sorghum Variety Attributes in the Lake
Zone Tanzania. In African Association of Agricultural Economists (AAAE) Conference Proceedings. Second International
Conference, August 20‐22, 2007, Accra, Ghana. 171–6. AAAE.
Masamha, Blessing, Vusilizwe Thebe, and Veronica N. E. Uzokwe. 2018. “Mapping Cassava Food Value Chains in Tanzania's
Smallholder Farming Sector: The Implications of Intra‐Household Gender Dynamics.” Journal of Rural Studies 58: 82–
92. https://doi.org/10.1016/j.jrurstud.2017.12.011
Mbanjo, Edwige Gaby Nkouaya, Ismail Yusuf Rabbi, Morag Elizabeth Ferguson, Siraj Ismail Kayondo, Ng Hwa Eng,
Leena Tripathi, Peter Kulakow, and Chiedozie Egesi. 2021. “Technological Innovations for Improving Cassava
Production in Sub‐Saharan Africa.” Frontiers in Genetics 11: 623736. https://doi.org/10.3389/fgene.2020.623736
Michler, Jeffrey D., Emilia Tjernström, Simone Verkaart, and Kai Mausch. 2019. “Money Matters: The Role of Yields and
Profits in Agricultural Technology Adoption.” American Journal of Agricultural Economics 101(3): 710–31. https://doi.
org/10.1093/ajae/aay050
Minten, Bart, and Christopher B. Barrett. 2008. “Agricultural Technology, Productivity, and Poverty in Madagascar.” World
Development 36(5): 797–822. https://doi.org/10.1016/j.worlddev.2007.05.004
Miriti, Philip, Mekdim D. Regassa, Chris O. Ojiewo, and Mequanint B. Melesse. 2022. “Farmers’ Preferences and Willingness
to Pay for Traits of Sorghum Varieties: Informing Product Development and Breeding Programs in Tanzania.” Journal
of Crop Improvement. 37(2):253–72. https://doi.org/10.1080/15427528.2022.2079038
Mugerwa, Habibu, Susan Seal, Hua‐Ling Wang, Mitulkumar V. Patel, Richard Kabaalu, Christopher A. Omongo,
Titus Alicai, et al. 2018. “African Ancestry of New World, Bemisia tabaci‐whitefly Species.” Scientific Reports 8(1): 2734.
https://doi.org/10.1038/s41598-018-20956-3
Mutsaers, H. J. W., H. C. Ezumah, and D. S. O. Osiru. 1993. “Cassava‐Based Intercropping: A Review.” Field Crops Research
34(3–4): 431–57. https://doi.org/10.1016/0378-4290(93)90125-7
Mwebaze, Paul, Macfadyen Sarina, De Barro Paul, Bua Anton, Kalyebi Andrew, Tairo Fred, Kachigamba Donald,
Omongo Christopher, and Colvin John. 2022. “Impacts of Cassava Whitefly Pests on the Productivity of East and
Central African Smallholder Farmers.” Journal of Development and Agricultural Economics 14(3): 60–78. https://doi.
org/10.5897/JDAE2022.1330
Nassar, N. M. A., and R. Ortiz. 2007. “Cassava Improvement: Challenges and Impacts.” The Journal of Agricultural Science
145(02): 163. https://doi.org/10.1017/S0021859606006575
National Bureau of Statistics (NBS). 2017. 2016/17 Annual Agriculture Sample Survey. Initial Report. The United Republic of
Tanzania. https://nbs.go.tz/nbs/takwimu/Agriculture/2016_17_AASS_%20report.pdf
National Statistical Office (NSO) (Malawi) and ICF. 2017. Malawi Demographic and Health Survey 2015‐16. Zomba, Malawi,
and Rockville, Maryland, USA. NSO and ICF.
Night, G., P. Asiimwe, G. Gashaka, D. Nkezabahizi, J. P. Legg, G. Okao‐Okuja, R. Obonyo, et al. 2011. “Occurrence and
Distribution of Cassava Pests and Diseases in Rwanda.” Agriculture, Ecosystems & Environment 140(3–4): 492–7.
https://doi.org/10.1016/j.agee.2011.01.014
Ogola, J. B. O., C. Mathews, and S. M. Magongwa 2013, July. The Productivity of Cassava‐Legume Intercropping System in a
Dry Environment in Nelspruit, South Africa. In African Crop Science Conference Proceedings (Vol. 11, 61–65).
Omongo, Christopher Abu, Samuel Morris Opio, Irene Bayiyana, Michael Hillary Otim, Tom Omara, Sam Wamani,
Patrick Ocitti, et al. 2022. “African Cassava Whitefly and Viral Disease Management Through Timed Application of
Imidacloprid.” Crop Protection 158: 106015. https://doi.org/10.1016/j.cropro.2022.106015
Opata, Patience Ifeyinwa, Oguejiofor Joseph Okorie, Juliana Chinasa Iwuchukwu, Chukwuma Otum Ume, and
Oyakhilomen Oyinbo. 2021. “Adoption of Improved Cassava Varieties in Nigeria: Insights From DNA Fingerprinting
Versus Self‐Reporting Varietal Identification Approaches.” Outlook on Agriculture 50(3): 269–76. https://doi.org/10.
1177/0030727021990032
Oster, Emily. 2019. “Unobservable Selection and Coefficient Stability: Theory and Evidence.” Journal of Business & Economic
Statistics 37(2): 187–204. https://doi.org/10.1080/07350015.2016.1227711
Owusu, Victor, and Emmanuel Donkor. 2012. “Adoption of Improved Cassava Varieties in Ghana.” Agricultural Journal 7(2):
146–51. https://doi.org/10.3923/aj.2012.146.151
Parmar, Aditya, Barbara Sturm, and Oliver Hensel. 2017. “Crops That Feed the World: Production and Improvement of
Cassava for Food, Feed, and Industrial Uses.” Food Security 9(5): 907–27. https://doi.org/10.1007/s12571-017-0717-8
Patil, Basavaprabhu L., James P. Legg, Edward Kanju, and Claude M. Fauquet. 2015. “Cassava Brown Streak Disease: A
Threat to Food Security in Africa.” Journal of General Virology 96(5): 956–68. https://doi.org/10.1099/jgv.0.000014
Pennisi, Elizabeth. 2010. “Armed and Dangerous.” Science 327(5967): 804–5. https://doi.org/10.1126/science.327.5967.804
Reincke, Katrin, Elisa Vilvert, Anja Fasse, Frieder Graef, Stefan Sieber, and Marcos A. Lana. 2018. “Key Factors Influencing
Food Security of Smallholder Farmers in Tanzania and the Role of Cassava as a Strategic Crop.” Food Security 10(4):
911–24. https://doi.org/10.1007/s12571-018-0814-3
Rogers, Everett M. 1962. Diffusion of Innovations (1st ed.). New York: Free Press of Glencoe.
Roothaert, Ralph L., and Ronald Magado. 2011. “Revival of Cassava Production in Nakasongola District, Uganda.”
International Journal of Agricultural Sustainability 9(1): 76–81. https://doi.org/10.3763/ijas.2010.0547
Ruzzante, Sacha, Ricardo Labarta, and Amy Bilton. 2021. “Adoption of Agricultural Technology in the Developing World: A
Meta‐Analysis of the Empirical Literature.” World Development 146: 105599. https://doi.org/10.1016/j.worlddev.2021.
105599
Simtowe, Franklin, Solomon Asfaw, and Tsedeke Abate. 2016. “Determinants of Agricultural Technology Adoption Under
Partial Population Awareness: The Case of Pigeonpea in Malawi.” Agricultural and Food Economics 4(1): 7. https://doi.
org/10.1186/s40100-016-0051-z
Singbo, Alphonse, Felix Badolo, Jourdain Lokossou, and Hippolyte Affognon. 2021. “Market Participation and Technology
Adoption: An Application of a Triple‐Hurdle Model Approach to Improved Sorghum Varieties in Mali.” Scientific
African 13: e00859. https://doi.org/10.1016/j.sciaf.2021.e00859
Suri, Tavneet, and Christopher Udry. 2022. “Agricultural Technology in Africa.” Journal of Economic Perspectives 36(1): 33–
56. https://doi.org/10.1257/jep.36.1.33
Teklewold, Hailemariam, Menale Kassie, Bekele Shiferaw, and Gunnar Köhlin. 2013. “Cropping System Diversification,
Conservation Tillage and Modern Seed Adoption in Ethiopia: Impacts On Household Income, Agrochemical Use and
Demand for Labor.” Ecological Economics 93: 85–93. https://doi.org/10.1016/j.ecolecon.2013.05.002
Tian, Yishui, Lixin Zhao, Haibo Meng, Liying Sun, and Jinyue Yan. 2009. “Estimation of Un‐Used Land Potential for Biofuels
Development in (The) People's Republic of China.” Applied Energy 86: S77–85. https://doi.org/10.1016/j.apenergy.2009.
06.007
Tijaja, Julia. 2009. The Evolution and Organization of Cassava Value Chains in Global Trade Landscape: Lessons for Africa
from Thailand. https://www.yumpu.com/en/document/view/6644840/evolution-and-organisation-of-cassava-value-
chains-kenya
Tomlinson, Katie R., Andy M. Bailey, Titus Alicai, Sue Seal, and Gary D. Foster. 2018. “Cassava Brown Streak Disease:
Historical Timeline, Current Knowledge and Future Prospects.” Molecular Plant Pathology 19(5): 1282–94. https://doi.
org/10.1111/mpp.12613
Uganda Bureau of Statistics (UBOS). 2021. 2021 Statistical Abstract. Kampala, Uganda: Uganda Bureau of Statistics.
Uzokwe, Veronica N. E., Deusdedit P. Mlay, Habai R. Masunga, Edward Kanju, Inakwu O. A. Odeh, and Joseph Onyeka.
2016. “Combating Viral Mosaic Disease of Cassava in the Lake Zone of Tanzania By Intercropping with Legumes.” Crop
Protection 84: 69–80. https://doi.org/10.1016/j.cropro.2016.02.013
Vurro, Maurizio, Barbara Bonciani, and Giovanni Vannacci. 2010. “Emerging Infectious Diseases of Crop Plants in
Developing Countries: Impact on Agriculture and Socio‐Economic Consequences.” Food Security 2(2): 113–32. https://
doi.org/10.1007/s12571-010-0062-7
Weerarathne, L. V. Y., B. Marambe, and Bhagirath S. Chauhan. 2017. “Does Intercropping Play a Role in Alleviating Weeds
in Cassava as a Non‐Chemical Tool of Weed Management? A Review.” Crop Protection 95: 81–8. https://doi.org/10.
1016/j.cropro.2016.08.028
Wooldridge, Jeffrey M. 2010. Econometric Analysis of Cross Section and Panel Data (2nd ed.). Cambridge, Mass: MIT Press.
Wossen, Tesfamicheal, Tahirou Abdoulaye, Arega Alene, Mekbib G. Haile, Shiferaw Feleke, Adetunji Olanrewaju, and
Victor Manyong. 2017. “Impacts of Extension Access and Cooperative Membership on Technology Adoption and
Household Welfare.” Journal of Rural Studies 54: 223–33. https://doi.org/10.1016/j.jrurstud.2017.06.022
S U P P OR T I N G I N F OR M A TI O N
Additional supporting information can be found online in the Supporting Information section at
the end of this article.
How to cite this article: Mwebaze, Paul, Sarina Macfadyen, Paul De Barro, Anton Bua,
Andrew Kalyebi, Irene Bayiyana, Fred Tairo, and John Colvin. 2024. “Adoption Determinants
of Improved Cassava Varieties and Intercropping among East and Central African
Smallholder Farmers.” Journal of the Agricultural and Applied Economics Association. 3:
292–310. https://doi.org/10.1002/jaa2.112

Africa

Uploaded by

Copyright:

Available Formats

Africa

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Africa

Uploaded by

Copyright:

Available Formats

Received: 10 February 2023 | Accepted: 12 February 2024

Adoption determinants of improved cassava

Paul Mwebaze1 | Sarina Macfadyen2 | Paul De Barro3 |

292 | wileyonlinelibrary.com/journal/jaa2 Jnl of Agr & App Econ Assoc. 2024;3:292–310.

Intercropping is a powerful strategy to promote a diversiﬁed cropping system, creating

3 | DATA AND METHODS

3.1 | Study areas

3.2 | Sampling procedure and data collection

3.3 | Descriptive statistics

3.4 | Empirical strategy

Yi = βXi + εi, (1)

TABLE 1 Descriptive statistics of the sample.

Malawi Tanzania Uganda

Yield 0.92 (0.27) 0.69 (0.46) 0.80 (0.40) 0.80 (0.38)

Male household head (%) 65 80 76 73.67

Income from cassava (US$/acre) 70.40 149.42 84.80 101.54

Access to credit (%) 16 22 33 23.67

Member of organization (%) 34 43 47 41.33

Access to extension services (%) 45 31 28 34.67

Livestock ownership (%) 85 89 97 90.33

Yi = βXi + γ0 W0 + γ1 W1 + γ2 W2 + εi, (2)

4.1 | Determinants of adoption of ICV and intercropping

Early maturity 0.20 (0.04)*** 0.01 (0.03) 0.13 (0.03)***

Yield 0.06 (0.08)** 0.12 (0.07)** 0.04 (0.06)*

Cooking quality 0.03 (0.03) 0.03 (0.02) 0.01 (0.03)

Group membership 0.07 (0.03)** 0.03 (0.02) 0.09 (0.04)**

Extension 0.04 (0.12)** 0.35 (0.18)** 0.06 (0.15)**

Credit 0.45 (0.17)*** 0.48 (0.18)** 0.05 (0.11)

Yield × extension 0.15 (0.13)** 0.40 (0.18)** 0.03 (0.15)**

Yield × credit 0.44 (0.17)** 0.45 (0.17)** 0.08 (0.10)**

Farm size 0.07 (0.02)** 0.19 (0.02)*** 0.15 (0.02)***

Freehold ownership 0.06 (0.08) 0.06 (0.07) 0.08 (0.06)

Leasehold ownership 0.08 (0.11) 0.14 (0.09) 0.11 (0.08)

Customary ownership 0.01 (0.07) 0.07 (0.06) 0.07 (0.05)

Age 0.01 (0.05) −0.05 (0.04) −0.03 (0.04)

Education 0.03 (0.03) 0.07 (0.03)** 0.03 (0.02)

Gender −0.05 (0.04) −0.01 (0.03) −0.02 (0.03)

Household size −0.03 (0.03) −0.03 (0.02) −0.03 (0.02)

Livestock 0.04 (0.06) 0.12 (0.06)** 0.09 (0.05)*

Malawi 0.68 (0.05)*** 0.45 (0.06)*** 0.51 (0.04)***

Uganda 0.69 (0.06)*** 0.46 (0.06)*** 0.52 (0.05)***

Constant −0.12 (0.22) 0.83 (0.21)*** 0.83 (0.17)***

Sensitivity analysis (Oster, 2019)

Sensitivity analysis (Diegert et al.,

R2 0.50 0.51 0.52

Obs. 1200 500 1200

Early maturity 0.08 (0.05)** 0.06 (0.04)* 0.07 (0.03)**

Yield 0.14 (0.09)* 0.15 (0.10) 0.06 (0.02)**

Cooking quality 0.07 (0.04)* 0.06 (0.03)* 0.03 (0.06)

Group membership 0.03 (0.04) 0.01 (0.04) 0.01 (0.03)

Extension 0.17 (0.20)** 0.40 (0.14)** 0.13 (0.17)**

Credit 0.22 (0.14)* 0.01 (0.04) 0.21 (0.11)*

Yield × extension 0.21 (0.19) 0.34 (0.14)** 0.01 (0.07)*

Yield × eredit 0.33 (0.14)** 0.17 (0.08)**

Farm size −0.010 (0.03) 0.17 (0.03)*** 0.17 (0.02)***

Freehold ownership 0.12 (0.11) 0.10 (0.15) 0.06 (0.06)

Leasehold ownership 0.01 (0.16) 0.07 (0.15) 0.09 (0.08)

Customary ownership 0.04 (0.10) 0.14 (0.14) 0.08 (0.06)

Age −0.13(0.06)** −0.05 (0.05) −0.06 (0.04)*

Education 0.02 (0.04) 0.03 (0.04) 0.01 (0.02)

Early maturity 0.20 (0.04)* 0.01 (0.03) 0.13 (0.03)*

Yield 0.06 (0.08) 0.12 (0.07) 0.04 (0.06)*

Group membership 0.07 (0.03) 0.03 (0.02) 0.09 (0.04)

Extension 0.04 (0.12) 0.35 (0.18) 0.06 (0.15)**

Credit 0.45 (0.17)* 0.48 (0.18) 0.05 (0.11)

Yield × extension 0.15 (0.13) 0.40 (0.18) 0.03 (0.15)**

Yield × credit 0.44 (0.17) 0.45 (0.17) 0.08 (0.10)**

Farm size 0.07 (0.02) 0.19 (0.02)* 0.15 (0.02)***

Malawi 0.68 (0.05)* 0.45 (0.06)* 0.51 (0.04)***

Uganda 0.69 (0.06)* 0.46 (0.06)* 0.52 (0.05)***

Constant −0.12 (0.22) 0.83 (0.21)* 0.83 (0.17)*

Extension 0.17 (0.20) 0.40 (0.14) 0.13 (0.17)**

Yield × eredit 0.33 (0.14) 0.17 (0.08)

Farm size −0.010 (0.03) 0.17 (0.03)* 0.17 (0.02)*

Malawi 0.01 (0.10) 0.35 (0.06)* 0.43 (0.04)*

Uganda 0.02 (0.09) 0.36 (0.07)* 0.19 (0.06)*

Constant 0.69 (0.29) 1.22 (0.29)* 0.73 (0.15)***

Early maturity 0.07 (0.13)* 0.13 (0.06) 0.02 (0.04)

Farm size 0.71 (0.13)* 0.20 (0.08) 0.03 (0.05)***

Education 0.23 (0.12) 0.14 (0.06) 0.11 (0.04)***

Malawi 0.09 (0.16) 0.06 (0.09) 0.05 (0.06)

Uganda 0.10 (0.10) 0.07 (0.10)