KNOWLEDGE, RISK, AND BENEFIT PERCEPTIONS: USING PARTICIPATORY VIDEO
AND TAILORED MOBILE MESSAGES TO MOTIVATE FARMERS’ UPTAKE OF
DROUGHT TOLERANT (DT) MAIZE SEED IN KENYA
By
Tian Cai

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Information and Media — Doctor of Philosophy
2017

ABSTRACT
KNOWLEDGE, RISK, AND BENEFIT PERCEPTIONS: USING PARTICIPATORY VIDEO
AND TAILORED MOBILE MESSAGES TO MOTIVATE FARMERS’ UPTAKE OF
DROUGHT TOLERANT (DT) MAIZE SEED IN KENYA
By
Tian Cai
This research documents the design and testing of two strategies for using information and
communication technologies (ICTs) to increase farmers’ knowledge and uptake of DT maize: (1)
locally-made, or “participatory” video, and (2) a multichannel method that incorporates the same
video with timely mobile phone-based audio messages. We conducted a randomized field
experiment in Machakos and Makueni counties in Kenya. The experiment randomly allocated
the two strategies to farmers in the study areas.
It was found that, after implementing these interventions, farmers in the multichannel
group demonstrated significantly higher knowledge about DT maize and its accompanying
management practices. Further, they were significantly more likely to report purchasing
improved seed as well as intending to plant DT maize than the farmers in the video-only and the
control groups. Farmers in the video-only group did gain a higher level of knowledge than the
farmers in the control group; however, this difference is not statistically significant. Moreover,
farmers’ perceived risk regarding DT maize was found to be associated with their level of
knowledge on the varieties and to which kind of treatments they are exposed. Farmers in the two
treatment groups showed lower perceived risk than farmers in the control group, especially in the
multichannel group. Mitigating the perceived risk of growing the varieties played a crucial role
in motivating farmers to test DT maize. We found risk perception was more likely to decrease

willingness to test DT maize among farmers who retained lower levels of knowledge than the
farmers who had higher levels of knowledge.
The results suggest that an ICT strategy integrating multiple ICTs can effectively
communicate contextualized knowledge and timely reminders to farmers. This strategy helped
farmers gain knowledge about DT maize and induced them to test new seed varieties. This
finding contributes to existing theories on the usage of ICTs in agricultural knowledge provision.
Previous literature mainly discusses whether single ICTs can improve farmers’ knowledge and
uptake (e.g. Gandhi, et al., 2009; Cole & Fernando, 2012), while this study investigates how to
design ICT-based approaches by integrating various ICTs to make them more effective in
innovation diffusion. In this case, those designing ICT strategies should consider farmers’
cognitive capacities and the characteristics of agricultural innovations, such as complexity and
whether farmers need to practice multiple steps at various points in a growing cycle, as such
design can influence farmers’ learning and uptake of the innovations. Lastly, this study
contributes to a greater understanding of farmers’ knowledge and perceptions of DT maize and
how they are associated with farmers’ uptake decisions.

I dedicate this work to my parents and grandparents

iv

ACKNOWLEDGEMENTS

I would like to extend my deepest gratitude to many individuals who in one way or
another have contributed to making this study possible. I would like to thank my advisor, Prof.
Charles Steinfield, and committee members, Prof. John Kerr, Prof. John Besley and Prof. Susan
Wyche for their unremitting guidance and encouragements throughout my Ph.D. program.
I would also like to thank the late Prof. Mark Levy who gave me the impetus to pursue
the topic of this thesis and the hope that our work can somehow help the poor to improve their
livelihood.
This thesis would not have succeeded in having this depth of research without the
contribution of scientists in International Maize and Wheat Improvement Center and field staffs
in Farm Input Promotion Africa.
Furthermore, I would like to thank my colleagues in Department of Media and
Information and Department of Agriculture, Food and Resource Economics for their suggestions
which improved the quality of the dissertation.
I want to especially acknowledge the United States Agency for International
Development who generously provided funding for this study and supported my Ph.D. studies
through the Global Center for Food Systems Innovation at Michigan State University.
And, finally, I would like to thank farmers who participated in this study for their time
and valuable input.

v

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................................... viii
LIST OF FIGURES ....................................................................................................................... ix
CHAPTER 1 INTRODUCTION .................................................................................................... 1
CHAPTER 2 LITERATURE REVIEW ......................................................................................... 9
2.1 Drought Tolerant (DT) maize in the Eastern and Southern Africa (ESA) region ............. 9
2.2 Farmers’ Adoption of Improved Seed ............................................................................. 11
2.2.1 The Myths of Low Adoption and Dis-Adoption of Improved Varieties .................... 12
2.2.2 Step-wise adoption of improved seed ....................................................................... 15
2.3 Risky Decision-Making ................................................................................................... 16
2.3.1 Perceived Risk .......................................................................................................... 17
2.3.2 Perceived Benefit and Its Relationship with Perceived Risk ................................... 18
2.3.3 Perceived risk and risky behavior............................................................................ 20
2.4 Participatory Videos and Phone Messages in Agricultural Extension ............................ 23
CHAPTER 3 RESEARCH METHOD ......................................................................................... 29
3.1 The Intervention .............................................................................................................. 29
3.2 Study Area ....................................................................................................................... 32
3.3 Experiment Design .......................................................................................................... 35
3.4 Sampling Strategy............................................................................................................ 38
3.5 Baseline Comparisons ..................................................................................................... 39
3.6 Operational Measures ...................................................................................................... 42
3.6.1 Perceived risk and perceived benefit ....................................................................... 42
3.6.2 Knowledge Scores .................................................................................................... 45
3.6.3 Behaviors and Behavioral Intention ........................................................................ 46
3.6.4 Indirect Questioning and Visual Aids ...................................................................... 47
3.7 Questionnaire Design ...................................................................................................... 50
3.8 Estimation Strategies ....................................................................................................... 50
3.8.1 Aggregated Indices .................................................................................................. 50
3.8.2 Estimation of Treatment Impacts ............................................................................. 52
CHAPTER 4 RESULTS ............................................................................................................... 56
4.1 Study One ........................................................................................................................ 56
4.1.1 The Impacts of Treatments on Knowledge about DT Maize and Management
Practices ........................................................................................................................... 56
4.1.2 The Impacts of Treatment on Willingness to grow DT maize .................................. 60
4.1.3 The Impacts of Treatment on Probability to Purchase Improved Maize Seed ........ 63
4.1.4 Differential Impacts of the Intervention ................................................................... 66
4.2 Study Two........................................................................................................................ 72
4.2.1 Perceived Risk and Perceived Benefit by Treatment Groups .................................. 72
4.2.2 Knowledge, Farmers’ Perceptions, and Willingness to Grow DT Maize................ 78

vi

CHAPTER 5 DISCUSSION ......................................................................................................... 83
CHAPTER 6 LIMITATIONS AND FUTURE WORK ............................................................... 91
CHAPTER 7 CONCLUSION ...................................................................................................... 94
APPENDICES .............................................................................................................................. 96
Appendix 1 Summary Statistics ............................................................................................ 97
Appendix 2 Household Characteristics in Baseline by Compliance with Treatments .......... 98
Appendix 3 Household Characteristics in Baseline by NGO status1 .................................... 99
REFERENCES ........................................................................................................................... 100

vii

LIST OF TABLES

Table 1: Experimental Design ...................................................................................................... 38
Table 2: Household Characteristics in Baseline by Treatment Groups2 ....................................... 41
Table 3: Knowledge Scores 4 Regression by Treatment Groups1................................................. 59
Table 4: Marginal Impacts of Treatments on Households’ Intention to Grow DT Maize in the
Next Primary Season 3 .................................................................................................................. 62
Table 5: Marginal Impacts of Treatments on Purchasing Maize Seed by Treatment Groups ...... 65
Table 6: Differential Impacts of Treatments on knowledge about DT by NGO Status (ITT)1 .... 70
Table 7: Differential Impacts of the Treatments on Households’ Probability of Purchasing
Improved Maize Seed and Intention to Grow DT Maize in the Next Season by NGO Status2.... 71
Table 8: Farmers’ Perceptions of DT Maize by Treatment Groups1 ............................................ 76
Table 9: Regression of Farmers’ Perceived Benefit and Perceived Risk as a Function of
Knowledge Level, Treatments, and Baseline Control Variables .................................................. 78
Table 10: Differential Impacts of Risk Perception on Households’ Intention to Grow DT Maize
in the Next Season by Level of Knowledge about the Varieties .................................................. 81
Table 11: Summary Statistics ....................................................................................................... 97
Table 12: Household Characteristics in Baseline by Compliance with Treatments ..................... 98
Table 13: Household Characteristics in Baseline by NGO status1............................................... 99

viii

LIST OF FIGURES

Figure 1:Timeline of the Intervention ........................................................................................... 31
Figure 2:The Study Area ............................................................................................................... 35
Figure 3: How important is using chemical fertilizer to get good DT maize yield? .................... 49
Figure 4: Perceived Risk of Growing DT Maize by treatment groups ......................................... 74
Figure 5: Perceived Benefit of Growing DT Maize by Treatment Groups .................................. 75
Figure 6: The Interaction Effect between Perceived Risk and Knowledge of DT Maize on
Farmers’' Willing to Grow DT Maize........................................................................................... 80

ix

CHAPTER 1 INTRODUCTION
Maize contributes forty-five percent of calories to people in Eastern and Southern Africa
(ESA), and is a particularly important source of income for the poor in this region1 (Shiferaw,
2011). Farmers’ most pressing concerns in the region are harvest failures and losses attributed to
changing climate conditions, such as drought stress. These changes directly affect food
production through changes in agricultural productivity, and indirectly through the reduction of
agricultural incomes (Fisher et al., 2015). In response to changing rain patterns and climates,
scientists have been breeding improved seeds, some of which are Drought Tolerant (DT)2.
Hendrix & Glaser (2007) suggest that planting new varieties alongside improved management
practices can reduce yield loss by up to 40 percent. However, although farmers experience
prevailing low levels of yield and the productivity variation associated with changing climate
conditions in ESA, the demand for DT maize seed still falls far short (Fisher et al., 2015).
Moreover, while tests conducted in experimental stations suggest sizeable increases in
productivity, farmers’ actual profitability of the adoption of DT maize is unknown. The
successful diffusion of improved seed in Asia and other regions indicates that technology
adoption and impact at scale is due to a combination of innovative technologies and institutional

1 Maize currently covers 25 million ha in sub-Saharan African producing 38 million metric tons (Smale et
al., 2011)
2 The Drought Tolerant Maize for Africa (DTMA) project has made release 160 D maize varieties between 2007
and 2013, and disseminated to farmers in 13 African countries. In addition to drought tolerance, the varieties have
other attractive traits, such as resistance to major diseases and high protein content. Cost of DT maize seeds is
similar to other non-DT commercial varieties. The DTMA project is led by International Improved Maize and
Wheat Center.

1

and policy shifts, such as improvements to farmers’ access to information, input, output and
credit markets (Shiferaw, 2011).
Moreover, scientists have developed new technologies and practices to reduce the risks
farmers are likely to encounter in agricultural production. However, Just (2003) points out that
adopting these risk-reducing technologies like pesticides and irrigation can also be a risky
process for farmers, as they need to invest labor and capital in new technologies that they are not
familiar with. Knowledge is also limited in terms of how risk perceptions relate to farmers’
perceived benefits and decision to try a technology. Thus, it is important to know how farmers
perceive the risks, what factors can alter risk perception and how these perceptions influence
farmers’ uptake decisions. This information can enable us to understand farmers’ mental
processes in technology adoption and to develop strategies to help farmers form realistic risk
perceptions and facilitate uptake.
Scholars suggest that providing information and training can reduce perceived risk
(Wildavsky & Dake,1990; Finucane & Maybery, 1996) and facilitate uptake by helping subjects
understand potential risks and form risk perception, as well as offer related risk management
methods (Sundblad, Biel & Gärling, 2007). Slovic, Finucane, Peters & MacGregor (2004)
propose that, when individuals receive inadequate information, they can develop perceived risks
based on speculation rather than an evaluation of facts. In conventional agricultural trainings,
farmers generally learn how to practice a technology and its benefits. However, we know little
about effective ways to communicate potential risks and relevant risk management practices to
farmers. Moreover, we know little in regards to what specific risks farmers perceive about a
technology. As such, there is no consensus among scholars as to what types of information can
help individuals reduce their perceived risk.

2

In addition to posters, radio, television, and newspapers, the existing channels for
communicating benefits of DT maize to farmers include field demonstrations and field days
organized by extension officers and NGO experts (Fisher et al., 2015). However, due to a severe
shortage of extension officers and NGO experts, farmers struggle to access relevant and up-todate information (Lovo, 2013). Many scholars believe that information and communication
technologies (ICTs) can be a partial solution (e.g., Gandhi et al., 2009; Nakasone & Torero,
2014).
This research assesses how different ICT-based extension approaches can improve
farmers’ knowledge about DT maize and its complementary practices. In particular, we tested
two approaches. The first approach includes locally made, or “participatory,” videos. This
approach is utilized by agricultural extension workers and FIPS’ staff in developing countries.
These organizations shoot videos that feature farmers from the same or a similar community who
have adopted DT seed. The videos integrate both contextualized social and cultural cues, such as
how farmers traditionally choose maize seed, and the technical information about DT maize and
modern practices. The videos also use local farmers’ narrative stories to inform and educate
audiences about the potential risks of growing DT and its risk management practices. The second
approach is a multichannel method. In addition to the participatory videos, farmers received four
audio messages on their mobile phones over the course of a maize growing cycle. A voice
system complemented the information in the video by ‘pushing’ audio messages before key
decision points in the maize growing season. These audio messages communicated reminders to
reinforce the knowledge farmers learned in the video and to encourage adoption. The initial
messages reminded farmers of the timing to purchase DT maize seed. The second and third
messages included content that reminded farmers of the proper timing, practices, and quantities

3

of fertilizer micro-dosing. The last message informed farmers about post-harvesting practices for
managing DT maize, including how to differentiate between traditional varieties and other
improved varieties, and how to improve yield. However, the majority of DT maize are more
sensitive to management practices than traditional ones. Therefore, farmers are more likely to
receive poor yield if they fail to use these management practices on their improved varieties.
In this study, we partnered with Farm Input Promotions Africa (FIPS), a local NGO, to
conduct a field experiment in Machakos and Makueni counties, Kenya. FIPS has worked in the
study area for more than 10 years and operated in a variety of villages. Farmers in the FIPS’
villages receive sample DT maize seed and can purchase the seed from FIPS, as well as request
information and advice from FIPS’s village based advisors (VBA). For example, through their
VBAs, FIPS provides trainings and demonstrations, distributes free sample seed, and sells inputs.
Twenty-seven villages were randomly selected from a list provided by FIPS. The list included
villages where FIPS is currently working and villages where FIPS planned to work but had no
activity prior to this study. These villages were assigned to one of three groups. Farmers in the
first group watched a video, while farmers in the second group watched the same video and
received four correlating audio messages. The third group was designated as the control group, in
which no video was screened and no audio messages were sent. In total, 615 households were
included in the study. The video was produced in the local communities three months prior to the
first growing season of 2016 by a team of researchers from a major university in the USA, FIPS’
staff, and farmers from local communities. It was screened in the treatment villages, and the four
audio messages were then sent to farmers in the multichannel group.
This dissertation is guided by six research questions, all of which are investigated through
two studies. The first three research questions are addressed in Study One, and the remaining

4

three are addressed in Study Two. Study One investigates impacts of treatment via a comparison
of the treatment effects that occur between a video-only group, a multichannel group, and a
control group on farmers’ knowledge and uptake. First, we study whether farmers in the three
groups have different levels of knowledge about DT maize and management practices (RQ1).
We also study the treatment effects of the interventions on farmers’ uptake by comparing
farmers’ probability to purchase improved seed and their willingness to test DT maize in the next
primary maize growing season (RQ2). Lastly, we study how the services provided by FIPS
related to the treatment effects on farmers in some villages sampled in the study. Therefore, this
study aims to understand the differential treatment impacts on farmers’ knowledge and uptake
when FIPS is active in the village (RQ3).
In Study Two, we explore the pathway from farmers’ knowledge and perception of DT
maize to their uptake decisions, seeking to understand how the treatments associate with farmers’
risk and benefit perception of DT maize. As such, we compare farmers’ perceived risk and
benefits of DT maize uptake in the video-only group, the multichannel group, and the control
group (RQ4). We then explore how the perceived risk and benefit of DT maize uptake are
associated with farmers’ knowledge of DT maize (RQ5). Lastly, we ask how relationships
between knowledge and the perception influences farmers’ uptake decisions. (RQ6).
To answer each research question, we test hypotheses that guide treatment effect
comparisons between the treatment groups and the control. These hypotheses are presented at the
end of the literature review. A field experiment was conducted in Machakos and Makueni
counties, Kenya, where 581 households participated in answering these research questions. We
found that, after implementing interventions in the both communities, farmers in the
multichannel group demonstrated higher-level knowledge about both DT maize and management

5

practices. They were more likely to report purchasing improved seed and more willing to grow
DT maize than farmers in the video-only group and control groups. Moreover, households in the
FIPS-supported villages mainly drove the impact of treatment.
Additionally, farmers’ perceived risk was found to be associated with their level of
knowledge about DT maize, as well as with which kind of treatment they were exposed to.
Farmers in the two treatment groups—especially the multichannel group—showed a lower level
of perceived risk than farmers in the control group. Mitigating the perceived risks of growing the
varieties played a crucial role in motivating farmers to test DT maize. This strength of the
negative association between perceived risk and uptake is reduced by farmers’ levels of
knowledge about DT maize.
DT maize is recently introduced to the study area, Machakos and Makueni counties,
Kenya. Some DT varieties were introduced to the local market, such as drought Tego, while the
majority of DT varieties listed by the International Maize and Wheat Improvement Center
(CIMMYT) were absent. The cost of one bag of DT seed (2kg) ranged from Ksh300 to Ksh500,
similar to other improved maize varieties in the region. FIPS suggested that farmers need 8kg of
maize seed to plant one acre of farm. Research institutions such as CIMMYT and seed
companies promote that farmers need to plan different improved varieties based on the
characteristics of their differing agro-ecological conditions. Farmers commonly purchased
improved seed from local agro-vet dealers in their communities. In the interest of saving money
and in situations where farmers found certain seeds were not available in their local markets,
some planted improved seed saved from previous seasons. These recycled improved seeds often
had poor quality and a yield much lower than the fresh improved seed. Around half of the
farmers in our study reported that they had planted improved varieties before the intervention,

6

but few had planted DT maize on a large scale. We found that most farmers commonly planted
improved maize seed alongside traditional maize varieties in a same plot. Moreover, farmers
were largely unable to distinguish the important differences among the improved varieties in
order to select the ones that most suitable to their farms’ conditions. We found that farmers
commonly planted older improved varieties or any varieties that recommended by agro-vet
dealers. Some seed companies listed their varieties’ yield, maturation time, and ideal condition
on seed packages in English, but few farmers told us that they read and used this information.
This study contributes to exiting theories about the usage of ICTs in agricultural
knowledge provision. Previous literature mainly discusses whether single ICTs can improve
farmers’ knowledge and uptake, while this study investigated how to design ICT-based
approaches integrating various ICTs to make them more effective in innovation diffusion. The
results suggest the importance of designing ICT strategies based on farmers’ cognitive
capacities, characteristics of agricultural innovation (such as whether farmers need to practice
multiple steps over a long cycle), and the functions of various ICTs. In this study, we sent mobile
phone audio messages before key decision points to remind farmers about the contextualized
knowledge they learned from the video screenings about DT maize and its management practices
over the course of a growing cycle. This multichannel approach was found to be more effective
at improving farmers’ knowledge than the video-only approach that was shown to the farmers
only once before the growing season. Moreover, the knowledge gained from the ICT
intervention were shown to help farmers form perceptions of DT maize and to induce behavioral
changes. This study also offers a greater understanding of farmers’ perceptions of DT maize and
how those perceptions are associated with uptake decisions. Moreover, this study generates

7

insights into ICT design strategies for communicating knowledge to farmers about DT maize and
its correlating practices.

8

CHAPTER 2 LITERATURE REVIEW
2.1 Drought Tolerant (DT) maize in the Eastern and Southern Africa (ESA) region
Studies have found that farmers use many different strategies to reduce the negative
impacts of climate variability and change3 on their maize production. Switching crop varieties, in
particular, is an increasingly used method in this regard (Fisher & Snapp, 2014). Drought
Tolerant crops are likely to play an increasingly important role in coping with climate change
and variation (Kassie et al., 2014). Research institutes such as CIMMYT and seed companies
have been introducing DT varieties 4 to the ESA market for about a decade5. Between 2007 and
2013, 160 DT maize varieties were released to the ESA market6. On-farm trials in ESA found
that “DT maize varieties out-yield popular commercial checks by 82-127 percent (controlled
drought), 26-47 percent (random drought), and 25-56 percent (optimal rainfall condition)7”
(unpublished data from Tesdeke Abate, DTMA project leader, March 2015, cited in Fisher et al.,
2015 pp. 284-285). La Rovere et al. (2014) found that DT maize could increase average yield
and lessen yield variability. The authors also speculated that the yield advantage between DT and
local varieties could be greater during droughts. However, the impact of these new DT maize
seeds depends on farmers’ level of adoption and their ability to profit from growing them.
Yet, scholars still know little about how farmers adopt DT maize and whether they can
profit from growing it, especially because farmers commonly apply deficient amounts of inputs,

3

Other methods include changing planting dates, switching crop species, crop diversification, and soil and water conservation.
The scientists used modern conventional methods to breed DT maize. So current DT maize seeds in the ESA market are not
genetically modified.
5 Global efforts aiming to develop DT maize germplasm include Drought-Tolerant maize for Africa (DTMA) implemented from
2006 to 2015 by CIMMYT, the International Institute for Tropical Agriculture (IITA) AND national research/extension
institutions of 13 African countries.
6 This is the total number of DT maize released to all ESA countries. Each countries have a smaller amount of varieties available
to local farmers.
7 Maize under random drought condition receive approximately less than 600mm rainfall per year and pests and diseases prone
locations under rainfed conditions. Maize under optimal rainfall condition receive more than 750mm annum and a temperature
range of 24-33 °C. Under managed drought conditions, maize grow in the off season and the irrigation interval calculated based
on the crop water balance (Setimela et al, 2017).
4

9

like fertilizer and labor, and use inadequate management practices (e.g., Tambo & Abdoulaye,
2011). Smallholders have experienced difficulties accessing new DT varieties mainly because
the seed market is not well established in ESA, and farmers lack access to the seed in areas
where no government intervention or development projects directly distribute DT maize to
farmers (Fisher et al., 2015). For example, in Malawi, farmers have mainly received their seed
from development projects. Fisher et al. (2015) found that, in six ESA countries, older farmers,
farmers with less land, and farmers with less exposure to information about DT maize were less
likely to try it. Although their study does not include data collected in Kenya, the results reflect
farmers’ perceptions and adoption statuses in the ESA area. Fisher et al., (2015) also found that,
in comparison to high-altitude farms, farmers cultivating land in low-altitude areas were more
likely to grow DT maize due to higher evapotranspiration in these areas. Scientists suggest that
modern inputs, especially fertilizer, are often a complementary input for improved varieties like
DT maize in comparison to traditional varieties, because improved maize has a steeper response
curve to fertilizer (e.g., Smale, Byerlee & Jayne, 2011). Indeed, Ertiro et al., (2017) conducted
trials on seven inbred DT lines across 11 sites in Kenya. This experiment showed that, in 2014,
the grain yield of the seven inbred lines under the low-nitrogen condition was lower than the
yield of the same varieties under well-watered and well-fertilized conditions, as well as those
under well-fertilized but drought-stressed conditions. However, only a small portion of farmers
applied sufficient amounts of chemical fertilizer and used proper approaches to apply it to their
improved maize varieties (Smale et al., 2011). It is worth noting that maize response to fertilizer
can also be influenced by factors such as the availability of soil organic content (Marenya &
Barrett, 2009; Kihara et al., 2016). Fisher et al. (2015) found labor constraints did not hinder
farmers’ adoption of DT maize, but speculated limited adoption was due to the fact that farmers

10

currently only grow DT maize on small scales. They postulated that the labor shortage may
become a salient problem if farmers decide to expand DT maize production.
Langyintuo et al. (2010) and Fisher et al. (2015) asserted that an effective strategy for
increasing adoption of improved maize seed is to enhance awareness and knowledge about DT
maize. However, most efforts mainly targeted farmers who already demonstrated a strong
demand for improved seed, rather than those with no experience of growing the seeds. Such
findings point to a need for more tailored communication strategies to increase awareness and
demand among farmers who are unfamiliar with improved seed, as different farmers have
varying information needs regarding the adoption of new technologies. Similarly, there is a need
to better inform farmers about necessary inputs and modern cultivation practices associated with
improved seed. Access to extension services is vital to bringing farmers’ information about the
existence, benefits, and usage of technologies (Kabunga, Dubois & Qaim, 2012). However,
extension services in ESA, regardless of whether they are publicly or privately funded, were
found to be less efficient in boosting adoption, especially where demand for improved seed was
low (Muyanga & Jayne, 2008). Seed companies also attempted to disseminate information about
improved seed through their market network and other information channels, including seed
packs and radio and TV programs. Some seed companies found it effective to provide small
packs of improved seeds for farmers to sample (Langyintuo et al., 2010).
2.2 Farmers’ Adoption of Improved Seed
This section presents different assessments from recent literature that analyze farmers’
awareness and knowledge about improved seed varieties and adoption, including that of DT
maize. One argument asserts that a majority of farmers are not aware of improved seed, which
hinders its adoption (e.g., Duflo et al., 2008). Another argument claims that farmers are aware of
11

improved seed, but delay adoption due to a focus on constraints such as lack of labor and capital,
which further fosters farmers’ beliefs that they will be unable to benefit from growing DT maize
(e.g., Suri, 2011). A final argument asserts the importance of knowledge-in-practice in farmers’
technology adoption. However, incomplete knowledge about “how to”—such as how to manage
maize crop to avoid potential risks—has been shown to lead to low adoption rates (e.g., Diagne
& Demont, 2007). The last two arguments, in particular, call for strategies to promote new
improved seed varieties that fit local contexts and farmers’ specific situations.
2.2.1 The Myths of Low Adoption and Dis-Adoption of Improved Varieties
Studies show that the adoption of improved maize leads to significant poverty reduction.
Khonje et al. (2015) found households that grew improved maize exhibited gains in crop
incomes, consumption expenditures, and food security. However, a substantial portion of maize
farmers still planted traditional seed. In the 2006-07 season, only 44 percent of maize land in
ESA grew improved seed8 (Abate et al., 2017). An earlier study showed that around 60 percent
of Kenyan farmers used fertilizer and hybrid seed in 2004 (Suri, 2011). Farmers who adopted
improved maize preferred old improved varieties, or jointly planted new and old hybrids in one
season (“portfolio selection”). Some of the improved seed varieties in Suri’s study were released
20 years before the study was conducted, so these older improved varieties do not have
characteristics of new and improved seed, such as drought tolerance. Farmers also grew
improved seed using farm-saved impure seed.9 Many farmers switched back and forth between
traditional and improved seed from season to season (Duflo et al., 2008).

8

This data exclude South Africa. Improved varieties include open-pollinated varieties (OPVs) and hybrids.

9

Impure improved seed could lost some improved attributes, such as high yield.

12

Extant literature cites limited access to information as a main adoption constraint, which
leads to low awareness of new seed availabiliy (Duflo et al., 2008; Fisher et al., 2015; Kabunga,
Dubois & Qaim, 2012). Heterogeneous populations do not have equal access to information
about improved seed, including its attributes and performance; this contributes to the lower
adoption rates among disadvantaged households lacking capital, labor and access to input and
output markets (Foster & Rosenzweig, 2010). Similarly, Shiferaw et al. (2015) found that factors
such as access to seed and capital, together with limited access to information, hinder DT maize
adoption. Fisher, et al. (2015) also found that some farmers failed to recognize improved seeds’
attributes, such as early maturation or drought tolerance in DT maize. They believe this explains
farmers’ reluctance to choose yield-enhancing inputs, such as DT maize and intensification
cultivation practices, as their responsive strategies to adapt to changing climates (Fisher, et al.
2015).
While some research points to a general lack of awareness among farmers in cases where
improved seeds are rejected, some research argues that farmers are very aware of improved seed
due to decades of exposure in their local markets. In such cases, farmers’ decisions to adopt or
not adopt improved seed are based on the kinds of information farmers choose to accept and how
they internalize that information. Some farmers chose not to adopt or dis-adopt improved seed as
they came to believe improved seed was not efficacious (Foster & Rosenzweig, 2010). Duflo et
al. (2011) found that farmers’ behavioral biases, such as being myopic mindedness, limited their
investment in fertilizer even when they noticed the apparent values of applying it to their fields.
Suri (2011) indicated that farmers who received low or even negative net returns from improved
seed tended to not adopt it, even when gross returns were high. When farmers had zero net
returns from growing hybridized maize, they tended to switch their adoption decisions from

13

period to period and were subject to shocks, such as the changing of input and availability of
seed. Suri (2011) ovbserved that the cost and return of hybrid seed varied by farmers, but that
farmers knew and understood the net return well before planting. Overall, this argument assumes
that farmers possess complete knowledge about improved seed, and that no cost and return
differences exist between varieties of improved seed.
Return to improved seed is sensitive in terms of how it is planted and managed. To
distinguish between farmers’ awareness of improved seed and their profound knowledge about
how to use them more efficiently, Duguid (2005) and Diagne & Demont (2007) argued that
knowledge-in-practice may better predict farmers’ adoption. Cultivating improved seed varieties
requires farmers to use a package of technologies beyond seed itself. For example, some seed
varieties are more sensitive to soil conditions than traditional varieties, so farmers need to apply
complementary inputs, such as organic or chemical fertilizer to achieve ideal yield (Byerlee &
Heisey, 1996; Foster & Rosenzweig, 2010), otherwise the return to adoption can be negative or
insignificant. Foster & Rosenzweig (1995) argue that, as long as farmers were aware of
improved varieties, they could learn the importance of complementary inputs and identify the
optimal labor and input to invest in by learning through doing or learning from their peers.
However, other studies show that this is not the case (e.g., Fisher & Snapp, 2014). Increasingly,
studies have recognized that farmers need to experiment with seed selection and the
implementation of improved cultivation practices10 to intensify production; these studies imply
that farmers need time and effort to discover how to maximize net returns from using improved
seed such as DT (Shiferaw et al., 2011; Foster and Rosenzweig, 2010). Moreover, the provision

10

The improved practices include increased plant density, line planting, place spacing, weeding and post harvesting
(Byerlee & Heisey, 1996)

14

of the information about seed selection and management practices could escalate farmers’ uptake
of knowledge-in-practice. However, knowledge about complementary inputs or improved
cultivation practices and the reasons why farmers need them are rarely communicated to farmers
together with efforts to promote improved seed.
2.2.2 Step-wise adoption of improved seed
To reduce uptake risks, farmers test and adopt new technologies in a step-wise manner,
by allocating parts of their resources, such as land, to testing new technologies at an early stage
of learning. Smale, Just & Leathers (1994) found that farmers intended to reduce the risks of
uptake by growing both new and traditional maize varieties in the same season to prevent
harvesting failure. Farmers seemed to believe that, if one variety failed, they could still harvest
enough from the other. Another type of step-wise adoption is partial adoption, which occurs if
components in a technology package can be separable. Byerlee and Polanco (1986) found that
farmers adopted individual components in a technology package sequentially. They found
farmers delayed the adoption of some package components when they perceived the components
as too risky (e.g., Abate et al., 2017). However, studies commonly treat adoption as a
dichotomous variable. That is, all respondents “either adopt completely or not at all” (Samel, Just
& Leathers, 1994, p.536). Yet, this assumption hardly reflects farmers’ adoption behaviors in the
real world (Byerlee & Polanco, 1986, p.519). As such, I employ two outcome variables in this
study to measure farmers’ decisions after receiving the intervention: (1) farmers’ probability of
paying for maize seed, and (2) the proportion of land farmers are willing to allocate to DT maize.
In the next section, I use previous literature on individuals’ perceived risk, perceived
benefit, and behavior detail to outline the conceptual framework of this study.

15

2.3 Risky Decision-Making
In this section, I review literature investigating individuals’ perceived risk, perceived
benefit, and adoption of new agricultural technologies. Slovic (2000) argued that, in general,
high levels of perceived risk were negatively associated with perceived benefit. Although a risky
object may entail the “potential for gains as well as potential for losses” (Coombs & Lehner,
1981, p. 318), individuals may not expect positive outcomes from this object if they perceive it
to be too risky (Sarin & Weber, 1993). Fischhoff, Slovic & Lichtenstein (1978) identified this
particular tradeoff as a societal dilemma because risk reductions commonly accompany a
decrease of societal benefits. It is also suggested that individuals’ perceived risk may not be
rational due to reasons such as lack of adequate information (Slovic, 2000; Sundblad, Biel &
Gärling, 2007).
However, few studies seek to specifically understand farmers’ risk and benefit
perceptions regarding improved seeds (Fisher & Snapp, 2014). Indeed, a substantial number of
studies focused on people’s perception of societal-level risk of new technologies (e.g., Slovic et
al., 1982), but little literature exists focused on people’s perception of how new technologies can
harm individuals, as well as how this risk perception is linked to technology adoption. Moreover,
few studies investigate proper approaches to help farmers form rational risk perceptions. Thus,
my study investigates how provision of contextualized knowledge and individualized reminders
intervened with farmers’ risk perception, benefit perception of DT maize, and their uptake of DT
maize. I draw findings from broader disciplines including ICT and development (ICTD), media
studies, psychology, management, and agricultural and behavioral economics to develop my
conceptual framework. In this chapter, I review the mechanism between constructs including
perceived risk, perceived benefit, and risky decision-making.

16

2.3.1 Perceived Risk
Perceived risk has been commonly cited as a crucial factor that decreases farmers’
adoption of new agricultural technologies (e.g., O’Mara, 1980; Smale, Just & Leathers, 1994;
Byerlee & Polanco, 1986). However, few studies investigate ways to reduce its impact on
technology adoption (Marra, Pannel & Ghadim, 2003). People’s perceived risk has multiple
dimensions and is regarded by scholars either as a rational judgment or as an emotional response
(e.g., “feelings”) (Loewenstein et al., 2001; Slovic et al., 2004). The basic assumption of risk
perception theories is that, at the same level of return, individuals dislike objects with more
perceived risk over fewer (von Neumann & Morgenstern, 1947). Risk perception studies do not
directly deal with actual risks, or potential risks likely to be encountered. Instead, they have
considered perceived risk as individuals’ subjective impressions of the risks of a technology
(Bauer, 1960).
Extant literature defines perceived risk in various ways. Bauer (1960) defined it as a twoaspect construct that entails individuals’ expected uncertainty and expected negative
consequences of failure. He argued that “consumer behavior involves risk in the sense that any
action of a consumer will produce consequences which he cannot anticipate with anything
approximating certainty, and some of which at least are likely to be unpleasant” (p. 24).
Cunningham (1967, p.37) advanced this definition by defining perceived risk as: “the amount
that would be lost (i.e. that which is at stake) if the consequences of an act were not favorable,
and the individual’s subjective feeling of certainty that the consequences will be unfavorable.”
Other perceived risk definitions include outcome variance of an activity (Sarin & Weber, 1993)
and subjective “evaluations of variation in the distribution of possible outcomes their likelihoods
and their subjective values” (Shapira, 1995, p.167).

17

Slovic (2000) suggested that perceived risk is a social construct. He and other scholars
have investigated perceived risk as affects, emotions, and feelings (e.g., Slovic, 2004;
Loewenstain et al., 2001). Further, scholars argued that individuals’ negative consequences of
perceived risk have diverse dimensions. Jacoby and Kaphlan (1972) proposed four dimensions
including financial, performance, physical, and psychological and social risk. Fischhoff et al.
(1978) suggested perceived risk is a qualitative construct, while Gregory and Mendelsohn (1992)
listed nine dimensions of risks and dread perceptions such as voluntariness of taking a risk,
immediacy of effect, newness, catastrophe, equity, future generations, economic benefits, etc.11
They attempted to quantify these qualitative dimensions using a psychometric paradigm. Slovic
(1987) later narrowed these dimensions down to three: dread, familiarity, and number of people
exposed.
My study focuses on farmers’ perceived risk about a new agricultural technology which
use Bauer’s (1960) two-aspect definition to investigate farmers’ perceived risk, including
uncertainty and expected negative consequences of growing DT maize.. I define perceived risk
as individuals’ subjective judgments of a risky I measure farmers’ subjective assessments on four
dimensions of the negative consequences (these four dimensions are listed in the operational
section in the method chapter).
2.3.2 Perceived Benefit and Its Relationship with Perceived Risk
Previous studies have various definition of perceived benefits. Starr (1969) defined
perceived benefit as the average contribution that a technology makes to individuals’ annual

11

They listed nine characteristics of perceived risk entailing voluntariness of risk, immediacy of effect, knowledge,
common-dread, newness, catastrophic potential, controllability, equity and risk to future generations.

18

income. Snoj et al. (2004) defined perceived benefit as bundles of attributes that together
produce a level of benefits, thereby providing utility to customers. These are tangible benefits
that individuals can directly observe or obtain by using the technology. My study aims to
understand how farmers rate specific benefits of DT maize. According to scientists, these
benefits include early maturing and good yield during drought (Fisher et al., 2015). Both benefits
were explicitly mentioned in the intervention this is tested for the study.
There are various arguments about relationship between perceived risks and perceived
benefits. Cognitive consistency theories state that people exhibit strong desires to hold consistent
judgments, so individuals tend to rate an object less beneficial if they perceive it to be risky
(Heider, 1946). Relatedly, Starr (1969) found individuals were more likely to tolerate perceived
risk when they perceived higher benefit of an object. He found that tolerance to risks was
conditioned by individuals’ judgments on the voluntariness of taking the risk and others’
perception of risks, like familiarity and catastrophic potential. Fischhoff et al. (1978) criticized
Starr’s findings, suggesting they were based on a small number of technologies. They found the
relationship between perceived risk and perceived benefit of 30 different technologies and
activities were varied and inconsistent. However, Alhakami and Slovic (1994) found an overall
inverse relationship between perceived risk and benefits. Their findings were based on the
respondents’ risk and benefit perceptions of 40 different technologies like herbicides, X-rays, airtravel and police work. They speculated that, when perceived risk was salient, it hindered the
recall of perceived benefit. Similar inverse relationships were found by later studies investigating
single technologies, such as genetically modified canola (Mauro & McLachlan, 2008) and
nanotechnology (Cobb & Macoubrie, 2004). Siegrist & Cvetkovich (2008) found that individuals

19

who perceived more benefits of nanotechnology, such as food materials and food packaging,
perceived fewer risks compared to those who perceived fewer benefits.
Moreover, individuals’ knowledge levels about a technology can alter the relationship
between perceived risk and benefit. Finucane & Maybery (1996) found that participants
perceived fewer risks of a technology when exposed to information about its benefits. In the
same study, information about risks was also found to relate to the change in benefit perception.
Wildavsky and Dake (1990) suggested that respondents who rated their knowledge of
technologies as high also tended to perceive greater benefits associated with technologies than
those who rated their knowledge as low; similarly, the perceived risks were lower when selfrated knowledge was higher. However, we know little about whether individuals’ actual
knowledge levels are associated with their perceived benefit and perceived risk of an unfamiliar
technology.
2.3.3 Perceived risk and risky behavior
Brewer et al. (2004) proposed three relationships between perceived risk and individuals’
risky behavior. The first relationship argues farmers’ perceived risk is associated with their
engagement in risky behavior. The second proposes that higher perceived risk leads to
preventative behavior. The last argument of Brewer et al. (2004) is that individuals reduce their
perceived risk of the action in which they choose to engage because those individuals tend to
justify their decisions as low risk to validate said decisions. The last argument is similar to
Fischhoff et al.’s findings (1978) that individuals’ perception of risk and benefit vary by the
degree of voluntariness of a risky behavior. Weber & Milliman (1997) suggested that reducing
individuals’ risk perception can be a driving force behind actions under risks.

20

The current study adapted a risk-value framework instead of an expected utility (EU)
framework. The EU framework is widely used by previous literature to study risk and risky
behavior. Sarin & Weber (1993) suggested that a risk-value framework could more effectively
reflect individuals’ real world choices than the expected utility models because their framework
required that choices “depend on the riskiness of the gamble and its value” (Sarin and Weber,
1993, p.148).
The expected utility (EU) framework was widely used with the expectation that
individuals choose less risky objects to achieve profit maximization. The EU framework uses
individuals’ risk preferences—that is, whether an individual is risk averse or risk seeking—to
explain their choices (e.g., Binswanger, 1980). However the EU framework fails to invoke the
notion of individuals’ risk judgments in the model; such an absence makes it hard to explore
realistic human behavior in diverse application areas (e.g., King & Robison, 1981; Sarin &
Weber, 1993). In the EU framework, risk preference cannot be directly captured in a lab setting
or controlled environment by deriving individuals’ choices during a gamble or lottery
(Binswanger, 1980; Weber & Milliman, 1997). King and Robins (1981) found that participants’
perceptions of specific risks are more important in explaining adoption than their risk
preferences because individuals employ a set of mental strategies or heuristics to make decisions
in an uncertain world (Slovic, 1987).
One approach to reducing perceived risk is through assisting individuals’ cognitive
processes by providing information that aims to communicate an innovation’s values to
individuals, as well as its potential risks and risk reducing practices to help individuals better
manage perceived risk (Sundblad, Biel & Gärling, 2007). For example, Fisher, Arnold & Gibbs
(1996) showed that effective information, such as performance of the innovation in a local

21

garden, is vital to farmers’ adoption of wheat varieties. In their study, farmers mainly obtained
quality information by trying a new wheat variety over time by themselves. Sundblad, Biel &
Gärling (2007) argued that knowledge can indirectly affect individuals’ behaviors via risk
perception. They suggested that diverse components of individuals’ knowledge about climate
change, including knowledge about consequences and causes of climate change, can have
different impacts on individuals’ risk perception and behavior. Verbeke et al. (2007) suggested
that individuals were willing to spend more money to obtain information about potential risks
than on information about the benefits of a technology.
Wildavsky & Dake (1990) asserted that knowledge of a technology can help individuals
form risk perceptions of real dangers. In other words, individuals lacking knowledge of a
technology can form perceptions of said technology that are not rational and tend to be anxious
about risks that do not exist. Slovic (1982) proposed that perceived risk due to inadequate
information can be reduced by education. This is because effective information helped reduce
uncertainty of an innovation. Moreover, Slovic’s (1982) study found that individuals can more
effectively prepare to deal with potential risks by gaining more knowledge about an innovation.
Marra et al. (2003) added that, in order to reduce farmers’ perceived risk, it is important to
develop relevant risk management skills by learning from their own experience. Weber &
Milliman (1997) suggested the importance of situational factors in facilitating individuals’
decision-making under risks. These factors include times or different contexts at which an
innovation is offered.

22

2.4 Participatory Videos and Phone Messages in Agricultural Extension
Scholars and practitioners have tested participatory video in different contexts, such as
agricultural training, to achieve diverse goals like information provision and inducing behavioral
change. In contrast to non-participatory video, participatory video aims to involve farmers in
video production and distribution and to communicate contextualized information to farmers.
Early research on participatory video mainly focused on documenting locals’ participation in
video production processes and how this approach allows locals’ voices to be heard by outsiders,
like policy makers. To this effect, Kindon (2003) worked with locals to produce a video about a
tribe’s history. More recent examples include an approach developed by Digital Green, an Indian
NGO, which implements the participatory video approach in agricultural extension training.
These examples adapt some elements of participatory video steps, such as producing videos with
locals (e.g., Gandhi et al., 2009; Stassart, Mathieu & MĂŠlard, 2011; Cai et al., 2017). Their aim is
to use video to facilitate participatory learning among communities. These efforts allow
communities to view, discuss and learn from the videos as well as from one another. The videos
are lauded for a number of reasons, including their potential roles in improving farmers’
knowledge and adoption of agricultural technologies (Gandhi et al., 2009). Other studies (e.g.,
Cai et al., 2017) investigated an approach to producing participatory videos that features both
situated technical content (similar to Digital Green’s video), and social and cultural content
related to learning and the adoption of an agricultural innovation, all in the form of local farmers’
narrative stories. Further, this video initiates a participatory learning process within the same
communities where the video was produced. For example, Cai et al. (2017) found farmers’
personal stories rendered to video can motivate farmers to discuss the challenges that are
imposed by social norms, power relationships, and out-of-dated mindsets in adopting new

23

technologies. As such, farmers’ participatory learning provokes them to overcome challenges
and their own biases by encouraging them to actively test the innovation.
However, local communities commonly lack video playing devices, and training videos
are normally only shown once to a community. The complexity of some agricultural
technologies requires intensive training programs that provide relevant information at specific
times in order to improve farmers’ recall of training information and to bring farmers’ attention
to technical details at the right moment. Although this phenomenon is not documented in the
previous literature on participatory video, studies investigating other approaches used to
provision knowledge to rural communities have elicited the drawback of approaches that only
provide information one time to individuals (Casaburi & Kremer, 2014). Due to humans’
inattention and memory failure, these approaches can be ineffective at informing farmers about
complex innovations that require them to practice different steps at a variety of times. Sending
reminders to farmers can be a plausible approach to overcoming this drawback (Taubinsky,
2014).
The mobile phone is another ICT commonly used to provide agricultural extension
services to farmers. Cole & Fernando (2012) found that voice messages to be effective in helping
farmers safely handle pesticides and improve yield. Short Messages Service (SMS) is another
widely used function to deliver agricultural extension information due to its simplicity and low
cost. Whereas studies investigating the impacts of the SMS approach showed mixed results, the
impact information transmitted through SMS on knowledge learning and adoption vary
according to context, content, technology, and delivery strategies (Fafchamps and Minten, 2012).
Fafchamps & Minten (2012) investigated a program providing crop advisory tips and local
weather forecasts to farmers through SMS. They found SMS had no impact on cultivation

24

practices or harvest losses. Alternately, Casaburi & Kremer (2014) designed an intervention in
Kenya that sent out SMS information to sugar cane growers at individualized moments (harvest
cycle and age of their cane) to help them perform certain tasks. The program increased yield by
11.5 percent (Casaburi & Kremer, 2014). Similarly, a study conducted in Ecuador found that
post-training text message reminders can improve farmers’ knowledge about and adoption of
integrated pest management (IPM) practices (Larochelle, Alwang & Travis, 2016). This study
indicated that the impacts of the messages may be associated with their timing, content, and
farmers’ capacities of applying the information. The authors also attributed the positive impacts
of message reminders to farmers’ participation in formal trainings on a similar topic. Previous
studies on the impact of text reminders suggest that reminders can both increase recipients’ recall
of certain information and bring individuals’ attention to a certain topic (Casaburi & Kremer,
2014; Taubinsky, 2014; Larochelle et al., 2016). Given the general success of SMS in
agricultural extension services, I chose to implement this approach in my study. The
complementary practices—including proper planting, fertilizer application, and fertilizer microdosing—need to be implemented at key decision-making points during the maize growing cycle.
Therefore, four audio messages were sent out to selected farmers to remind them about these
practices at least two weeks before the key times.
Using the approaches of participatory video and SMS messaging synthesized in the
literature review, I test the following hypotheses. These hypotheses are organized by research
questions:
In Study One, I compare the treatment effects of farmers’ knowledge and uptake
between a video-only group, a multichannel group, and a control group.

25

RQ1: Do farmers in the three groups have different levels of knowledge about DT maize and
management practices?
H1: Farmers in the video-only group retain more knowledge about DT maize than farmers in the
control group.
H2: Farmers in the multichannel group, which include both the participatory video and mobile
phone reminders retain more knowledge about DT maize than farmers in the video-only and
control groups.
H3: Farmers in the video-only group retain more knowledge about management practices than
farmers in the control group.
H4: Farmers in the multichannel group retain more knowledge about management practices
than farmers in the video-only group and control groups.
RQ2: What are the treatment effects on farmers’ uptake as measured by their probability to
purchase improved seed and their willingness to test DT maize in the next primary maizegrowing season?
H5: Farmers in the video-only group are more willing to test DT maize than farmers in the
control group.
H6: Farmers in the multichannel group are more willing to test DT maize than farmers in the
video-only and control groups.
H7: Farmers in the video-only group are more likely to purchase DT maize than farmers in the
control group.

26

H8: Farmers in the multichannel group are more likely to purchase DT maize than farmers in
the video-only and control groups.
RQ3: What are the differential treatment impacts on farmers’ knowledge and uptake in terms of
whether or not they reside in FIPS villages?
H9: The treatment effects on farmers’ knowledge about DT maize and management practices
differ between those residing in FIPS villages and those residing in non- FIPS villages.
H10: The treatment effects on farmers’ willingness to test DT maize differ between those who
reside in FIPS villages and those who reside in non- FIPS villages.
H11: The treatment effects on farmers’ likelihood to purchase DT maize differ between those
who reside FIPS villages and those who reside in non- FIPS villages.
In Study Two, I intend to explore how the treatments associate with farmers’ risk and
benefit perceptions of DT maize, and how those perceptions associate with farmers’ uptake
decisions.
RQ4: What is the difference between farmers’ perceived risk and benefit of DT maize in the
video-only group, the multichannel group, and the control group?
H12: Risk perception differs between farmers in the treatment groups and the control.
H13: Benefit perception differs between farmers in the treatment groups and the control.
RQ5: How are the perceived risks and the perceived benefits of DT associated with farmers’
knowledge about DT maize?
H14: Association between risk perception and farmers’ knowledge differs between farmers in the
treatment groups and the control.
27

H15: Association between benefit perception and farmers’ knowledge differs between farmers in
the treatment groups and the control.
RQ6: How does the knowledge-perception relationship influence farmers’ uptake decisions?

28

CHAPTER 3 RESEARCH METHOD
In this study, we conducted a randomized field experiment in Machakos and Makueni
counties in Kenya. The experiment randomly allocated the video training and a multichannel
approach, including both audio mobile phone messages and video, to farmers in the study areas.
Through the video, farmer actors explain the benefits of DT maize, potential risks related to seed
type, how to identify and purchase seed, modern cultivation practices, and their personal stories
about uptake. Audio messages deliver advice and reminders on cultivation activities, such as
land preparation, planting, and fertilizer application. The intervention provides the study with
exogenous variation in access to knowledge about DT maize and modern practices among
similar households. The objective was to investigate whether the interventions are associated
with greater knowledge about DT maize and modern practices, differential risk and benefit
perceptions of DT maize, and uptake of DT maize. In this chapter, I explain the interventions,
study area, experimental design, sampling strategies, sample characteristics, operational
measures, questionnaire design, and identification strategies.
3.1 The Intervention
A team of researchers, FIPS staff, and local farmers produced the video in December
2015. In the video, farmer actors explained the benefits of DT maize, potential risks related to
seed type, how to identify and purchase seed, modern cultivation practices, and their personal
stories about planting. The team created four audio messages that deliver advice and reminders
related to seed purchasing and cultivation activities, such as land preparation, planting, and
fertilizer application. In February 2016, 341 households in two treatment groups were invited to
participate in the video screenings. Between March and August 2016 (long-rain reason), the
period during which they grew maize, 169 households in the multichannel group are sent four

29

different audio messages. Here is an example of the phone message received by farmers in the
multichannel group a few weeks prior to application of top dressing fertilizer to maize:
“Hello this is Faith Kilonzo. I am calling to remind you this is the time
to apply top dressing. You apply the top dressing when the maize has 8 leaves.
Apply a bottle cap full of fertilizer 5 centimeters from the base of the maize and
cover with soil. This is also the time to do the second weeding. Thank you.”
The intervention provides the study with exogenous variation in access to knowledge
about DT maize and modern practices among similar households. The objective was to
investigate whether the intervention is associated with more knowledge about and uptake of DT
maize. Figure 1 provides a calendar of intervention and data collection activities along with an
agricultural calendar. It is important to note that there are two annual growing seasons in the
study region: the “short-rains” season from September to February, and the “long-rains” season
from March to August. Although rains in the “short-rains” seasons are shorter, they are typically
more consistent than rains in “long-rains” seasons. Therefore, farmers tend to grow more maize,
the main staple crop, in the “short-rains” seasons than in the “long-rains” seasons. They are also
more likely to store the maize they harvest in the “short-rains” season as food for future
consumption.

30

Figure 1:Timeline of the Intervention

31

3.2 Study Area
The intervention took place in two counties in the southeastern region of Kenya (Figure
2). These counties are located in the dry transitional (DT)12 agro-ecological zone (AEZ) and are
considered medium drought risk zones (20–40 percent PFS) target areas13, where drought risks
are high and DT maize can have potential benefits (La Rovere et al., 2010). Maize is the critical
food and cash crop in both counties.
The majority of the agricultural activities in these two areas are rain-fed and in the smallscale semi-subsistence sector. The soil is generally of low fertility, and many soils are highly
erodible (Barber, Thomas, and Moore 1979). The maize yield is low compared to the national
average. Most farmers in this area still grow the traditional seed, and some grow both traditional
and improved seed in the same season. The Duma series and seed from Pioneer Hybrids are the
most common improved maize varieties grown in the study areas. Few farmers grew newer
improved varieties. Farming households experience food shortage regularly due to the low yield.
The average production of this zone is generally low, ranging from 300-1,200kg/ha, while the
national average is 1,600kg/ha (Muhammad et al., 2010). Farmers have no or only a small
amount of output for sale as the majority of harvest is used for domestic consumption. The
higher variation in rainfall was found to be associated with the fluctuation in maize production in

12

Other methods include changing planting dates, switching crop species, crop diversification, and soil and water
conservation.
12
The scientists used modern conventional methods to breed DT maize. So current DT maize seed in the ESA
market are not genetically modified
13
Global efforts aiming to develop drought tolerant maize germplasm include drought-tolerant maize for Africa
(DTMA) which was implemented between 2006 to 2015 by the International Maize and Wheat Improvement Center
(CIMMYT), International Institute for Tropical Agriculture (IITA) and national research/extension institutions of 13
African countries.

32

this area (Omoyo, Wakhungu & Otengi, 2015). Farmers’ adoption of improved seed, including
DT seed, is low, while the dis-adoption rates of improved seed is high (Muhammad et al., 2010).
Machakos County has a population of 442,930 with a poverty rate14 of 57 percent. The
population in Makueni County is 253,316, with a poverty rate of 63.8 percent (Government of
Kenya, 2013). These two counties encompass 368.57 thousand hectares of land and mainly
consist of hills and small plateaus rising between 700 and 1,700 meters above sea level (MASL).
The mean elevation is 1,357 MASL in Machakos and 1,047 MASL in Makueni. More than 60
percent of the region contains very erodible, relatively shallow, sticky, red, black, and brown
clay soil with variable fertility on steep slopes; 20 percent has poorly drained, shallow, stony
soils of low fertility (Bernard, Campbell, and Thom, 1989; based on findings by Jaetzold and
Schmidt, 1983). The capital towns of both counties are less than 200 km from Nairobi. The
western part of the Machakos area is considered part of the greater Nairobi area. Off-farm
employment is prevalent among men, leaving women to take responsibility for crop
management.
As previously noted, the two production seasons are between September and February,
and between March and August. Total annual precipitation ranges from 500 to 1,300 mm,
depending on altitude and other factors. The majority of farmers produce maize in both seasons,
and the primary growing season in this area is associated with the short-rains season (Hassan,
1998).
Further, farmers use traditional knowledge passed from parents to children to select the
seed. For example, farmers often pick maize cobs that have more rows of kernels to use them as

14

Population below $1.25 per day

33

seed. In our survey prior to the intervention, few farmers mentioned taste of maize as a major
concern because they were unable to tell the taste difference in varieties of maize. Some farmers
mentioned that they observed improved maize to be more vulnerable to pests than traditional
varieties.
As these study areas are close to Nairobi, the biggest city in Kenya, many male
household members work as laborers in the city. Therefore, in over 70 percent of the households
sampled in this study, women farmers made the decisions regarding which maize varieties to
plant, and managed the maize throughout the growing season. However, their decisions were
influenced by their male family members’ support, such as the provision of remittances and
information in regards to new improved seed varieties. Around four-fifths of farmers reported
fertilizer application to their maize, although we found that the majority of farmers did not apply
sufficient amounts of fertilizer. Moreover, farmers lacked knowledge about agricultural practices
that could improve the efficiency of using fertilizer, such as fertilizer micro-dosing. A small
portion of farmers in these study areas received subsidized fertilizer from the government, while
the majority of farmers reported purchasing fertilizer from their local market. The costs of a 50kg
bag of fertilizer ranges from $25 to $35 in the local market.

34

Figure 2:The Study Area
3.3 Experiment Design
In order to answer the research questions, I designed two treatment groups and one
control group in this study (Table 1). Households in the video-only treatment group watched a
video about DT maize. The video focused on introducing basic knowledge-in-practice, such as
where to purchase seed, and how to engage in modern management practices. We expected the
video to provide knowledge about DT (H1) and its associated modern practices (H3), and to
motivate farmers to test the seed (H5). Video screenings were moderated by a local FIPS. In
addition to participating in the video screening, households in the multichannel treatment group
also received four audio messages to their mobile phones. This approach was intended to
reinforce knowledge about DT maize (H2) and its management practices (H4), and to induce

35

farmers to test composting (H6) because humans have limited memory capacity and limited
attention to various characteristics of a technology. The study’s control group received no
intervention. It was expected that farmers in the multichannel and video-only groups would be
more likely to purchase DT seed than farmers in the control group (H7 and H8).
In addition to the random variation in the treatment group, the FIPS condition was also
included as a subset of the treatment and control groups in a cross-cutting experiment. Farmers
residing in FIPS and non- FIPS villages can be different in diverse ways. These differences
include that, FIPS provide services and sell inputs through its village-based advisers (VBA).
Farmers residing in FIPS villages receive sample maize seed for free (25g per sample bag) and
can purchase regular seed packages (2kg) at retail price from the VBAs. The FIPS trains their
VBAs on new varieties and their complementary technologies and practices. These advisors
work as small agro-dealers supplying inputs, such as DT maize, and provide management advice
based on farmers’ requests. Some VBAs manage a small demonstration plot in the villages and
organize “baraza”, community meetings to distribute sample seed, including DT maize.
Farmers residing in the non- FIPS villages do not receive these services. Therefore, in the
FIPS villages, the intervention is not the only information source associated with DT maize. It
was expected that households would receive services from FIPS in addition to interventions in
both the treatment and the control groups in comparison to those households in the non- FIPS
villages. Therefore, differences in knowledge improvement and increased uptake between
farmers residing in the FIPS villages and those who did not were expected (H9, H10 & H11).
This experimental design also enabled me to test how treatments associate with farmers’
risk and benefit perceptions of DT maize uptake, and how these perceptions associate with their
uptake decisions. Farmers in two treatment groups and the control are expected to have different
36

levels of perceived risks (H12) and perceived benefits (H13). Further, it is expected that the
association between farmers’ perception and knowledge will vary between the treatment groups
(H14 & H15). Lastly, I explore whether farmers’ uptake decisions relate to the association
between their knowledge about DT maize and their perceptions of its varieties (RQ6).
To obtain production information, a pre-test survey was conducted prior to the video
screenings in order to determine what maize varieties farmers grew in the previous growing
season (November 2015); other demographic data about the study area were also collected
through this survey. Then, a post-test was implemented after the first 2016 growing season
(March–August 2016) in an effort to understand farmers’ testing and perceptions of DT maize
after the intervention. Qualitative research methods like focus groups, key informant interviews,
and field observation were utilized to attain a deep understanding of how farmers’ risk and
benefit perceptions associate with their choice to grow or not grow DT maize.

37

Table 1: Experimental Design
Treatments

Number of
villages

FIPS

Video-only
Multichannel
Control

6
6
6

Number of
households at
the baseline
104
102
100

Non- FIPS

Video-only
Multichannel
Control

2
2
5

68
67
169

Number of
households in
the post test
97
99
100
64
63
159

3.4 Sampling Strategy
Two wards, Kola and Kee, were selected together with FIPS. Kola is located in the
southern part of Machachos County and Kee is located in the northern part of Makueni County.
The Kola ward has a population of 24,264; while the Kee ward has a population of 20,926
(GOK, 2013). Summary statistics are presented in Table 1.
This study implemented a cluster-randomized field experiment where 27 villages were
randomly assigned to a control group (n =11) or one of two treatment groups (n = 8 each). The
27 villages15 were selected among villages within a local FIPS’s current or future operating area.
A two-stage sampling strategy was used to select households for the study. FIPS offered a list of
villages (n=50) to this effect. The list showed that FIPS had no activities in nine out of the 50
villages. All nine non- FIPS villages were included in the study and were randomly allocated to
two treatment groups and the control. FIPS continued to have no activities in these villages
during the study. 18 villages were randomly selected from the remaining 41 where FIPS had
activities prior to and during the study. These 18 FIPS villages were then randomly allocated to
the two treatment groups and the control. Control villages were randomly selected from the same

15

This is the total number of DT maize

38

sampling frame to preserve comparability to the treatment villages. FIPS also provided a list of
households in each village, of which 30 households in each non- FIPS village and 17 households
in each FIPS village were randomly selected for inclusion in the study. A larger number of
households in the non- FIPS villages were selected, as there more households resided in the nonFIPS villages than in the FIPS villages in the selected study areas.
3.5 Baseline Comparisons
In this section, I show that the randomization process delivered three similar groups:
households in the video-only treatment, households in the multichannel treatment, and
households in the control group. I compare the baseline characteristics of those in the two
treatment groups and those in the control with the following Ordinary Least Squares (OLS)
Regression:

(1)
Yi is a characteristic of the ith household before intervention and Âľi is a zero-mean householdspecific error term. The coefficients Îą1 and Îą2 provide estimates of the differences in Yi of the
video and multichannel groups relative to the control group. Sample means of the video, the
multichannel and the control group, as well as estimates for Equation (1), are presented in Table
2. The sample is relatively well balanced in terms of gender of household head, characteristics of
main maize manager within a household (age, gender, education), land, distances to seed and
trading center, number of members within a household, wealth (livestock index), and mobile
phone ownership. Households in the multichannel group were richer (higher livestock index)
than those in the video group. Further, more maize managers in the video-only group were

39

illiterate in comparison to both the multichannel and control groups, while the overall number of
people with no formal education in the sample was less than 7 percent.

40

Table 2: Household Characteristics in Baseline by Treatment Groups2

Age of household
Number of household
members between 15-65
Number of household
members
Household’s distance to
market selling seed
Household’s distance to
trading center
Size of land (acres)
Livestock index
Women-headed household
Own mobile phone
Household with no
education
Household with secondary
education or higher
Household in FIPS
villages
N

Control

Video

(C)
48.464
(0.91)
3.117
(0.10)
5.595
(0.16)
9.02
(0.67)
6.34
(0.43)
3.769
(0.27)
176.394
(10.02)
0.212
(0.03)
1.964
(0.01)

(V)
51.017
(1.31)
3.174
(0.13)
5.401
(0.21)
8.1
(0.77)
6.827
(0.58)
3.619
(0.30)
157.483
(12.54)
0.297
(0.04)
1.936
(0.02)

Multichann
el
(M)
50.686
(1.29)
3.036
(0.13)
5.379
(0.20)
7.56
(0.62)
7.486
(0.75)
4.195
(0.45)
198.852
(13.77)
0.201
(0.03)
1.964
(0.01)

0.055

0.116

0.036

(0.01)
0.168
(0.02)

(0.03)
0.174
(0.03)

(0.01)
0.136
(0.03)

0.386

0.64

0.515

(0.03)
274

(0.04)
172

(0.04)
169

1.

Mean Differences
(C)-(V)
-2.554
(1.55)
-0.058
(0.17)
0.194
(0.26)
0.92
(1.04)
-0.488
(0.71)
0.151
(0.41)
18.912
(16.09)
-0.085
(0.04)
0.027
(0.02)
0.062**
(0.03)
-0.007
(0.04)
0.254**

(C)-(M)
-2.223
(1.54)
0.081
(0.16)
0.216
(0.25)
1.46
(0.98)
-1.146
(0.80)
-0.425
(0.49)
-22.458
(16.73)
0.01
(0.04)
-0.001
(0.02)

(V)-(M)
0.331
(1.83)
0.139
(0.19)
0.022
(0.29)
0.54
(0.99)
-0.658
(0.94)
-0.576
(0.54)
-41.37*
(18.62)
0.095
(0.05)
-0.028
(0.02)

0.019

0.081**

(0.02)
0.032
(0.04)

(0.03)
0.038
(0.04)
0.125

*

0.129***

(0.05)
446

(0.05)
443

(0.05)
341

For the first three columns, the means and standard deviations of each variable in the control, video-only, and
multichannel are reported. In the last three columns, the differences were calculated using the following
regression: Yi = Îą1 Videoi + Îą2Multichannlei + Âľi. Regression standard errors are reported in parentheses.
2.
We also compare baseline characteristics between households residing in and out of FIPS villages. Please see
Appendix 4.
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and
control denoted by *** 99%, ** 95%, * 90%.

41

3.6 Operational Measures
3.6.1 Perceived risk and perceived benefit
In extant literature, perceived risk and perceived benefit have been measured in various
ways, producing different findings and various conclusions about the relationship between
perceived risk and perceived benefit, and the relationship between perceived risk and risky
behavior (Fausti & Gillespie, 2006; Binder et al., 2011). One main ongoing debate focuses on
whether scholars should use a single-item measurement, which captures global risk and benefit
assessment, or a multi-item measurement, which captures individuals’ perceptions of the specific
risks and benefits of an innovation. Another debate focuses on whether perceived risk and
perceived benefit should be measured separately or together16. In line with the latter strand of
research, this study uses two sets of questions to measure perceived risk and perceived benefit
separately. This measurement includes multiple items for both constructs.
Coombs & Lehner (1981) created and tested a single item measurement of risk
perception asking one question: how risky is the prospect? Several studies tested this scale and
asserted that the single-item measurement can accurately reflect individuals’ judgments of risk.
Weber et al. (1992) found this single item measurement correlated strongly “with standard
features of risky prospects that are useful in making decisions” (e.g., Weber, Anderson &
Birnbaum, 1992). Cobb and Macoubrie (2004) suggested that a single-item question was useful
for measuring risk perception in situations where the public still viewed nanotechnology as an
abstract idea. In contrast, Fishoff, Slovic & Lichtenstein (1978) proposed a multi-item
measurement using a psychometric procedure to elicit quantitative judgments of the risks and
benefits of various activities and technologies. Further, they measured individuals’ estimation of

16

Risk and benefit are the two extrema in one scale

42

likelihood that each risk and benefit would influence them or the public. Fishoff (1978) et al.
argued that this likelihood measurement presents relative magnitude of impact of perceived risk
and perceived benefit.
Binder et al. (2011) compared a single-item with a multi-item measurement of
individuals’ perceived risk and perceived benefit of two new technologies. They found the
single-item performed poorer than the multi-item measurement in capturing perceptions (p.14).
Binder et al (2011) criticized the single item measurement because it can capture biased
responses, as it is less robust in terms of controlling for random errors captured by the
measurement. Moreover, they suggested that the multi-item measurement delivered more
“descriptive” assessments of perceived risk and perceived benefit than the single-item
measurement.
Sitkin & Weingart (1995) tested a multi-item measurement to study perceived benefit and
perceived risk together as a single construct. They examined the relationship between risk
perception and risky decision-making. Sitkin & Weingart’s (1995) measurement entails four
instruments that aim to capture information about individuals’ overall evaluation of a risky
decision. The first three items answered the question: “how would you characterize the
decision?” The three seven-point Likert type items are: threat—opportunities, potential for
loss—potential for gains, and positive situation to negative situation. The fourth item measures
participants’ estimation of likelihood that a risky decision can be successful (Sitkin & Weingart
1995). Other researchers criticized similar multi-item measurements for being incapable of
adequately distinguishing between perceived risk and perceived benefit, as well as for the lack of
clarity regarding the instrument’s face validity (Binder et al., 2011).

43

I adapted Sitkin & Weingart’s (1995) multiple items to measure risks and benefits.
However, in this study, risk and benefit were measured separately due to their qualitative
differences in the context of this study. The decisions regarding whether or not to grow DT
maize can be more complex than decisions made in lab settings about the lottery or gambling, as
they includes aspects beyond mere monetary trade-off. Furthermore, rather than engage a global
evaluation, I intend to understand the specific risks and benefits farmers perceived to be
associated with DT maize. All specific perceived risks and perceived benefits were tailored to
regional conditions and to the technology. Four dimensions of the perceived risk of DT maize
include: (1) having a poor yield, (2) losing the money spent on the seeds, (3) losing spent on
growing the maize, and (4) not having food in the next season. Each dimension was comprised of
two items: (1) likelihood of the negative impact of the risks and (2) frequency of the risks. Two
dimensions of perceived benefit of DT maize include (1) fast maturation and (2) drought
tolerance. Each dimension of perceived benefit was comprised of two items: (1) likelihood of
positive impact of DT maize and (2) beliefs in the positive impact of DT maize. Participants rated
each item on a 7-point scale. These dimensions of perceived risk and perceived benefit were
derived from previous literature (e.g., Smale, Just and Leathers, 1994) and findings from
qualitative studies conducted by other project members17.

17

Three other team members who worked in the same project conducted this qualitative study summer 2015 in the
study areas.

44

Below are the questions used to measure perceived risk and perceived benefit18:
Perceived risk

Perceived benefit

Dimensions of perceived risk:

Dimensions of perceived benefit:

a. having a poor yield

a. fast maturation

b. losing the money spent on seeds

b. drought tolerant

c. losing the time spent on growing the maize
d. do not have food in the next season

Measurement items:
i. How likely is it that any of these problems
would actually happen to John and Mary if
they planted DT maize?

Measurement items:
i. To what extent would any of these benefits
positively affect you and your family?
ii. How important are these benefits to your
family?

ii. How likely is it that each of these three
concerns would negatively influence you and
your family?

3.6.2 Knowledge Scores
Knowledge was measured by giving respondents an exam that tested whether they
retained key details of DT maize and its complementary practices. We used two outcome
variables to measure farmers’ knowledge: (1) a continuous variable measuring farmers’
knowledge about DT maize with a range from 0 to 6 in increments of 0.5, and (2) a continuous
variable measuring farmers’ knowledge about the management practices with range from 0 to 11
in increments of 1. Appendix 1 details the test questions and acceptable answers. This knowledge
test had two parts: in the first part, respondents were asked to list what they knew about DT

18

A full list of question can be found in Appendix 1

45

maize; in the second part, 11 questions were used to test respondents’ knowledge about the
complementary practices featured in the video.
3.6.3 Behaviors and Behavioral Intention
The video was screened in the treatment groups several weeks before the 2016 secondary
planting season (March to August), when the majority of farmers had not yet purchased maize
seed to plant for the season. The first audio message was sent to farmers in the multichannel
groups within a week after the screenings and encouraged farmers to purchase DT maize seed
from reliable sources, such as FIPS or valid input shops.
The measurement used by Smale, Just & Leathers (1994) was adapted for this study to
implement the proportions of land allocated to new varieties. This was the first variable used in
this study to measure farmers’ uptake of DT. Farmers were first asked to report total farm size in
their households. They were then asked what maize varieties they planned to grow in the 2016
primary planting season (September 2016 to February 2017) followed by the size of the plot in
which they intended to grow each variety. The total size of the farm on which they planned to
grow DT maize in the primary season was summed. In order to measure farmers’ willingness to
plant DT maize in the primary season, I used farm size that farmers were going to allocate to
plant DT maize to divide their total farm size. This measurement was believed to be a reliable
proxy for capturing farmers’ actual uptake decisions because we asked farmers to estimate their
land allocation to DT in the post-test. This test was conducted two to three weeks prior to
purchasing the seed in the 2016 primary season when all the intervention were implemented.
This variable captured the impact of both the video and all four audio messages, as farmers were
asked to give their responses to the question after the interventions were implemented in the
communities (see Figure 1).

46

The second variable used to measure the impact of treatment on farmers’ behavior is
whether they purchased DT maize seed. Farmers were asked what maize varieties they planted in
the 2016 secondary maize season and how much they paid for the seed. Then a dichotomous
variable was created to determine farmers’ willingness to purchase the seed, in which the
variable “1” was coded if farmers paid for the seed, and the variable “0” was coded if they did
not. This variable captured an aspect of farmers’ responses following the video screening and the
first audio message, which both occurred prior to farmers’ seed purchases in the 2016 secondary
maize season (see Figure 1).
3.6.4 Indirect Questioning and Visual Aids
People tend to show their best side and hide their negative side during interactions with
others (Fisher, 1993). This tendency is called social desirability bias. It can substantially distort
the information gathered through self-reported measurement across almost all social science
disciplines. The bias can be fraught as it leads to spurious theoretical and practical conclusions
(Mensch & Denise, 1988).
As such, an indirect questioning technique is widely used by researchers to minimize the
impacts of social desirability bias (e.g., Gregson, Zhuwau, Ndlovu, 2002). It is a projective
technique through which study subjects respond to structured questions from another person’s or
group’s angles. Indirect questioning can reduce social desirability bias and offer reliable
measurement (Gregson et al., 2002). Thus, it can give respondents a context to report
information from the external world that is similar to them rather than of themselves directly. A
typical way to enact indirect questioning is to ask a respondent to guess “what would other
people would do?” instead of “what would I do?”

47

Despite the advantages of using indirect questioning, the limitations of this measurement
approach (e.g., Hofmann & Schmitt, 2008) include that it is not suitable for all kinds of behavior,
and its reliability and validity can be lower than those of more direct measures. Further, scholars
need to be cautious when they use the data collected from indirect questioning (Bishop &
Heberlein, 1979).
While cautious of these potential limitations, the current study uses this indirect
questioning technique in order to increase validity of the measurement. For instance, when
measuring farmers’ perceived risk, farmers were asked to respond to “what do farmers John and
Mary need to worry about with DT maize if they plan to grow them next season?” Here is the
question:
Farmer John and Mary, they are husband and wife. They have three kids and
live in your neighboring village. They have some land and grow maize every
year. They only grow traditional maize and Duma 43, so they don’t know
anything about DT maize. They come to you and ask your suggestions. What
things do John and Mary need to worry about if they are going to grow DT
next year?
Most of the participants in this study have low-level literacy skills. They also have little
experience answering survey questions that ask them to weigh the importance, agreement, and
likelihood of statements or situations (Bernard & Taffesse, 2014). Chachamovich, Flech &
Power (2009) suggested that illiterate respondents failed to distinguish a five-point scale
(extremely high – extremely low) and a three point scale (low, medium, high). The use of visual
scales can prevent this difficulty. In addition, (Reid, Maag & Vasa, 1994) found visual scales can
reduce social desirability bias. As such, the visual scales featured in Gore & Kahler (2015) were
48

adapted to this study in self-report questions, such as perceived risk, perceived benefit, and
expected outcome. One example is:

a

b

c

d

e

f

Very important

g
Not important at all

Figure 3: How important is using chemical fertilizer to get good DT maize yield?
All questions in the questionnaire, including the indirect questions and visual scales, were tested
before the actual survey. A pilot study was conducted with the enumerators in a Kamba
community to ensure that wording and measurements, including perceived risk, perceived
benefits, and expected outcomes, were appropriate to the local context.

49

3.7 Questionnaire Design
The pretest survey was used to gather the following information: basic household
demographics and households’ maize production information of all types of maize19, including
size, seed varieties, productivity, etc. These questionnaires collected information on up to four
maize varieties that households planted in the secondary planting season as well as what they
intended to plant in the 2016 primary season. Other modules in the questionnaire included
experience of drought, seed and other input access, information sources regarding maize
production, household GPS, and distance to the nearest market.
In the post-test survey, the following information was collected: households’ maize
production information of up to four types of maize, size of land for growing certain maize
varieties, seed varieties, costs of seed, usage of modern practices, and what maize varieties
farmers intended to plant in the next primary season. Similarly, the post-test questionnaire
includes farmers’ knowledge about DT and modern practices, risk perception, benefit perception,
experience of drought, seed, and access to the input and output market.
The schedules of both surveys as well as the interventions are presented above in Figure
1.
3.8 Estimation Strategies
3.8.1 Aggregated Indices
Several studies combined the value of perceived risk and perceived benefit as a single
variable by subtracting perceived benefit from perceived risk. According to Binder et al. (2010),
the disadvantage of this analysis method is that it potentially loses valuable information, such as

19

Including improved and traditional seed

50

the distinct influences of risk and benefit perception to an individual’s risky choices. They
invoked further studies focusing on how risk and benefit perception converge and diverge.
Relatedly, this study aggregates perceived risk and perceived benefit indices separately. Clearly,
aggregation still leads to loss of information for each of the specific dimensions, however, can
also reduce errors by cancelling out some of the random noise inherent in self-reporting
measurements. In this case, the aggregation across the dimensions of each construct can offer a
useful measurement (Bernard & Taffesse, 2014).
As previously mentioned, for each dimension of perceived risk and perceived benefit,
two items were included to measure the likelihood of a risk dimension and the severity of its
impact. To obtain the aggregated perceived risk where PR below is perceived risks, two items
were multiplied for each perception dimension. Then the four dimensions of perceived risk were
summed and divide by the number of perceived risks:

where, PR is perceived risk.
Similarly, to obtain the aggregated perceived benefits, we sum the two dimensions of
perceived benefit, the likelihood of positive impact of DT maize and beliefs in its positive
impact, and divide it by number of the perceived benefits:

51

where, PB is perceived benefit.
The same wording and scale (seven-point Likert scale) was used for the questions
measuring a given construct. To analyze the constructs together in a model, respondents’ ratings
for the questions of each construct were summed.
3.8.2 Estimation of Treatment Impacts
A key goal of this study was to analyze whether the two treatments increased households’
level of knowledge and willingness to test DT maize. As such, the following factors in treatment
groups were compared with those of the control group: level of knowledge, proportion of maize
land that households allocated to DT, the probability of paying for improved seed, and the
perception of DT maize. Additionally, the presence of FIPS effects within the treatment groups
was tested by comparing the outcomes of households in FIPS villages with those of households
in non-FIPS villages.
Throughout the analysis, we use the following operationalization for treatment variables:
Video takes a value of 1 if a household is in a video-only treatment. It takes a value of zero
otherwise.
Multichannel takes a value of 1 if a household is in a multichannel treatment. It takes a
value of zero otherwise. The remaining households (i.e., those with Video=0 and Multichannel
=0) are in control villages.
The impact of the two treatments was calculated by comparing: households’ post-test
knowledge scores, whether or not they purchased improved seed, the proportion of land they
were willing to allocate to DT, and the perceived risk and perceived benefit of DT maize uptake.
Namely, I estimated the following regression:

52

μi= β1 Videoi + β 2Multichanneli + γ Zi + Ɛi

(2)

Where ÎĽi were outcomes including: (1) level of knowledge about DT maize and its
complementary practices, which household i retained, (2) whether the household purchased
improved maize, (3) farmers’ willingness to plant DT maize and the proportion of land farmers
were willing to allocate to the varieties, and (4) farmers’ risk and benefit perceptions. Zi is a
vector of household-level controls, including gender of household head, number of people in
household, livestock index (number of livestock multiplied with each type’s market price), main
income sources (agriculture or non-agriculture), total land size, households’ distance to the
nearest market selling seed, the main maize growers’ education level, whether households
received information about DT maize from other information sources such as radio, and whether
households lived in the FIPS villages. In this specification, the reference group was the control
group. Additionally, the error term iƐi was uncorrelated with other explanatory variables as
households were randomly assigned to treatments. Equation 2 is based on the original treatment
assignment (Videoi and Multichanneli)—regardless of whether households actually watched the
video or received audio messages—and provided an Intention-to-Treat (ITT) estimate of the
intervention.
Around half of the households in the video-only treatment villages attended the video
screenings. In the multichannel villages, one third of sampled households watched the video and
received the audio messages. The baseline characteristics of those who actually received the
treatments in the treatment groups and those who did not receive treatment in the control group
(compliers) were compared with those who did not receive treatments in the treatment groups
and those who received treatment in the control group (non-compliers) using formula (1), see
Appendix 3. Two significant differences between the compliers and non-compliers were found in

53

accordance with the following characteristics: (a) compliers were younger than non-compliers
and (b) more compliers were in the non- FIPS villages than in the FIPS villages. There are
several reasons that can explain the relatively large number of non-compliers (compliance rate).
One reason is that the project only had resources to screen the video once in each village.
Therefore, households did not have opportunities to watch the video if they missed the scheduled
screenings in their village. In the multichannel group, phone messages were only sent to those
who attended the video screenings, as we were not able to reach some households’ phones due to
technical issues, such as phones being turned off or having no reception20.
The Local Average Treatment Effects (LATE) were estimated using the treatment
assignment, 𝑣𝑖𝑑𝑒𝑜𝑖 and 𝑚𝑢𝑙𝑡𝑖𝑐ℎ𝑎𝑛𝑛𝑒𝑙𝑖 as instruments for compliance (Gerber & Green, 2012).
In particular, we estimated the following system of equations:
𝛶𝑖 = 𝜑0 + 𝜑1 𝑣𝑖𝑑𝑒𝑜𝑖 + 𝜑2 𝑚𝑢𝑙𝑡𝑖𝑐ℎ𝑎𝑛𝑛𝑒𝑙𝑖 + 𝜽𝑿𝒊 + 𝜀𝑖

(3)

𝑊𝑎𝑡𝑐ℎ𝑖 = 𝛿0 + 𝛿1 𝑣𝑖𝑑𝑒𝑜𝑖 + 𝛿2 𝑚𝑢𝑙𝑡𝑖𝑐ℎ𝑎𝑛𝑛𝑒𝑙𝑖 + 𝝆𝑿𝒊 + 𝜔𝑖

(3a)

𝑅𝑒𝑐𝑒𝑖𝑣𝑒𝑊𝑎𝑡𝑐ℎ𝑖 = 𝜆0 + 𝜆1 𝑚𝑢𝑙𝑡𝑖𝑐ℎ𝑎𝑛𝑛𝑒𝑙𝑖 + 𝜆2 𝑣𝑖𝑑𝑒𝑜𝑖 + 𝝈𝑿𝒊 + 𝜂𝑖

(3b)

Using these equations, we predicted the likelihood of households attending the video screening
(Watchi ) and the likelihood of households attending the video screening and receiving the audio
messages (ReceiveWatchi). In other words, we were able to obtain predicted values of the
likelihood that households were compliers in the video treatment and in the multichannel
treatment respectively, and 𝛶𝑖 were the outcome variables that included knowledge about DT
maize, knowledge about maize management practices, and DT uptake. φ was the LATE on those

20

The system tried to send the messages to households’ phone up to six times within three days

54

who were compliers in the two assigned treatment groups. In Appendix 3, I present the
comparison of baseline characteristics between households who received the treatments and
those who did not.
The following variation of Equation (4) was estimated to calculate the ITT effect on
households with different NGO statuses:
Yi = β1 Videoij + β 2Multichannelij + β3Videoij Wj+ β 4Multichannelij Wj+ γ Zij + ¾ij

(4)

where Wj is NGO status (i.e., lived in an FIPS village or lived in a non-FIPS village) or main
income source (i.e., agricultural activities and non-agricultural activities).

55

CHAPTER 4 RESULTS
In the next chapter, I present the results in two sections based on the research questions.
For Study One, the present results answer questions about the interventions’ treatment impacts
on farmers’ knowledge and uptake of DT maize. For Study Two, the treatments are investigated
in terms of how they associate with farmers’ perceived risks and benefits, and how these
perceptions relate to farmers’ uptake of DT maize.
4.1 Study One
In this study, I present the impact of treatments on: farmers’ levels of knowledge
regarding DT maize and its complementary management practices, farmers’ probability of
purchasing improved maize seed, and farmers’ willingness to plant these seed varieties in the
next primary growing season. I also present results of the heterogeneous treatment effects
between farmers who received FIPS services and farmers who did not.
4.1.1 The Impacts of Treatments on Knowledge about DT Maize and Management Practices
Results of the treatment effects on respondents’ knowledge scores are reported in Table
3. Impact of the treatments on farmers’ knowledge about DT maize are estimated in columns 14, and impact of the treatments on farmers’ management practices are estimated in columns 5-8.
Measurement of management practices includes proper planting and fertilizer application. The
video featured characteristics of DT maize and its management practices, and the phone audio
messages reinforced them. Therefore, both knowledge measurements can indicate the impact of
reminders on households’ knowledge about the practices. Columns (1), (2), (5) and (6) show
intent-to-treat (ITT) effects of treatments using simple OLS regressions.

56

As shown in column (1), although households retained a low level of knowledge about
DT maize, households in the multichannel group, on average, showed a 0.19 higher knowledge
score on DT maize than households in the control group. This difference accounts for three
percentage points of the total knowledge score and are quantitatively small, but statistically
significant. Households in the video group gained nearly the same knowledge score as the
control group. This difference in knowledge score is not significantly higher in the multichannel
group than it is in the video-group.
The proportion of farmers who watched video in the video treatment group is 47.82
percent, and in the multichannel treatment, 37.04 percent of the farmers watched the video and
received the audio messages. To solve this compliance problem, the random treatment
assignments were used as instrumental variables for treatments received. The results in column
(3) suggest sizeable impacts of the multichannel treatment on households’ level of knowledge
about DT maize and its complementary practices in cases where the households received the
treatments. Column (3) shows households who received both video and the four audio messages
earned knowledge scores on DT that were 0.52 higher than households in the control group and
0.47 higher than households in the video-only group. The multichannel treatment increased
households’ knowledge of DT maize by 8.6 percentage points of the total knowledge score, and
the video-only treatment increased it by 7.8 percentage points. The difference between the
multichannel group and the control group is statistically significant.
Similar results were found for the households’ knowledge score on DT maize
management practices. Column (5) in Table 3 shows that the multichannel treatment increased
farmers’ level of knowledge by 0.4 over farmers in the control and the video-only groups. These
differences are around 3.6 percentage points of the total knowledge score of DT maize. Column

57

(7) presents the LATE estimation showing farmers in multichannel treatment groups retaining
eight percentage points more knowledge about DT maize management practices than farmers in
the control group, and six percentage points more than farmers in the video-only group. All of
these reported differences are statistically significant.
Overall, these results indicate that multichannel treatment can significantly increase
households’ knowledge about DT maize compared to the video-only treatment and the control.
Further, the reminders about DT maize management practices sent out to the multichannel
treatment group significantly improved households’ knowledge about the complementary
practices of growing DT maize, which were demonstrated in the video and reinforced in the
audio message reminders.

58

Table 3: Knowledge Scores 4 Regression by Treatment Groups1
Knowledge about the management
practices
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
0.02
0.02
0.01 -0.05
(0.09) (0.09)
(0.15) (0.16)
*
0.19
0.14
0.40** 0.31*
(0.10) (0.09)
(0.15) (0.16)
0.05
0.02
0.20
0.11
(0.21) (0.16)
(0.23)
(0.24)
**
*
***
0.52
0.40
0.88
0.77***
(0.25) (0.24)
(0.29)
(0.30)
Knowledge about DT 4

Video (1)
Multichannel (2)
Video Complier
Multichannel
Complier
Additional baseline
controls 2
Observations
Mean Knowledge
scores in the Control
households in the
video only group
watched the video 3
households in the
multichannel group
watched the video
and received the
messages 3
F-test values for
Video only =
Multichannel

N

Y

N

Y

N

Y

N

Y

582

582

582

582

582

582

582

582

0.55

0.55

0.55

0.55

7.19

7.19

7.19

7.19

0.48

0.48

0.48

0.48

(0.02)

(0.02)

0.37

0.37

(0.04)

(0.04)

5.95**

5.45**

(0.02) (0.02)
0.37

0.37

(0.04) (0.04)

1.89

1.32

2.30

2.12

5.22** 4.54**

1.

Column 1, 2, 5 & 6 report the Intend-to-Treat (ITT) estimates of the intervention. Column 3, 4, 7 & 8 report the
Average Treatment Effect or Local Average Treatment Effects (LATE) estimates using the treatment assignments
as instruments for the compliers in each treatment group.
2.

Additional baseline controls in columns 2, 4, 6 & 8 include gender of household head, number of people in
household, livestock index, main income sources (agriculture or non-agriculture), total land size, household
distance to the nearest market selling seed, main maize growers’ education level, residence in FIPS villages or
not, and total amount of fertilizer applied in the 2016 long rain season.
3.

Predicted value of the likelihood for households in two treatment groups to participate in the intervention. For
the multichannel group, households both watched the video and received the phone messages. The results are the
first stage of the Instrumental Variable Regression (following equation (3))
4.

The range of knowledge about DT is 0-6, measured in increments of 0.5 as a unit. The knowledge measurement
of DTmanagement practices is 0-11 with 1 as a unit.

59

Standard errors in parentheses. Significance levels of the differences between the two treatment groups and
control group denoted by *** 99%, ** 95%, * 90%.

4.1.2 The Impacts of Treatment on Willingness to grow DT maize
Uptake of DT maize is modeled as a function of the treatment groups and households’
socio-economic characteristics. The same variables from the earlier knowledge regressions are
included, plus a variable that measures households’ intended uptake of DT maize a season after
the intervention. Due to the project schedule, the post-test data collection was carried out in
September 2016. As such, it was not possible to capture the households’ actual uptake of DT in
the primary maize growing season, which occurs from September to February. Therefore, I
captured the households’ post-intervention intention to grow DT maize in the next primary
season, and used it as a proxy to measure households’ uptake of DT maize.
Table 4 shows that the multichannel treatment had a positive and significant impact on
households’ intention to plant DT maize in the next primary planting season. The multichannel
treatment increased the probability of uptake by 14 percent. The video-only treatment, however,
had a positive but insignificant impact on this intention. It increased farmers’ willingness to plant
DT by three percent. This evidence is consistent with a behavioral model of inattention; in other
words, the audio message probably had a reminder effect on households’ uptake of DT maize.
The LATE estimation reported in Table 4 suggests that the multichannel treatment increased the
probability of DT uptake by 38 percent, and the video-only treatment increased it by 7 percent.
Two outcome variables were used to measure farmers’ willingness to grow DT maize in
the 2016 primary season: (1) a dichotomous variable measuring whether farmers will grow the
60

varieties in the next primary season or not, and (2) a continuous variable measuring the
proportion of land farmers were willing to allocate to DT maize that ranged from 0 to 1. Only the
regression analysis using the dichotomous variable as the outcome variable in Table 4 is
presented, as the analysis using the continuous outcome variable showed similar results.

61

Table 4: Marginal Impacts of Treatments on Households’ Intention to Grow DT Maize in the Next Primary
Season 3

Video (1)
Multichannel (2)

(1) 2

(2)

0.03

-0.02

(0.06)

(0.07)

0.14

**

0.12*

(0.08)

(0.09)

Video Complier
Multichannel Complier

(3) 2

(4)

0.07

-0.02

(0.14)

(0.09)

0.38**

0.28**

(0.18)

(0.13)

Additional baseline controls 1

N

Y

N

Y

Observations
Mean adoption rate of control
group

582

582

582

582

0.37

0.37

0.37

0.37

0.48

0.48

(0.02)

(0.02)

0.37

0.37

(0.04)

(0.04)

2.60*

6.04**

Households in the video only
group watched the video
Households in the multichannel
group watched the video and
received the messages
F-test values for
Video-only = Multichannel

3.23*

2.12

1.

Additional baseline controls in columns 2 & 4 include gender of household head, number of people in household,
livestock index, main income sources (agriculture or non-agriculture), total land size, household distance to the
nearest market selling seed, main maize growers’ education level, residence in FIPS villages or not, and total
amount of fertilizer applied in the 2016 long rain season.
2.
Columns 1 & 2 report the Intent-to-Treat (ITT) estimates of the intervention. Columns 3 & 4 report the Average
Treatment Effect or Local Average Treatment Effects (LATE) estimates using the treatment assignments as
instruments for the compliers in the treatment groups.
3.
These analyses are based on a linear specification; note that other non-linear models—such as random effects
Probits—yielded similar coefficient magnitudes (not reported).
4.
The indicator variable for households in two treatment groups participated in the intervention. For the
multichannel group, households both watched the video and received the phone messages. The results are the first
stage of the Instrumental Variable Regression (following equation (3)).
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and control
group are denoted by *** 99%, ** 95%, * 90%.

62

4.1.3 The Impacts of Treatment on Probability to Purchase Improved Maize Seed
I found the multichannel treatment tended to increase farmers’ probability of purchasing
fresh improved maize seed more than the video only treatment in cases where farmers decided to
grow DT maize and other improved maize varieties. Farmers were informed about the
importance of planting fresh improved maize seed in the video and the first audio message. This
information was sent out to farmers a few weeks before it was time to purchase the maize seed.
A little more than half of the participants who reported growing maize in the first maize season
of 2016purchased improved seed (n = 280). On average, they spent 230Ksh on one KG of the
maize seed, which is similar to the average price of improved maize seed (median = 225 & SD =
45, range 140-500Ksh). Farmers’ probability of purchasing the maize seed is modeled as a
function of the treatment groups and households’ socio-economic characteristics. In total, around
half of the farmers purchased maize seed after the intervention in the first planting season of
2016. Table 5 shows that all farmers in the multichannel group, regardless of whether or not they
received the treatments, were six percent more likely to pay for improved maize seed than
farmers in the control group. The difference was 15 percent among farmers who actually
received the treatment, which is significant at 0.05 level. Moreover, around one fifth of the
farmers reported that they purchased DT maize seed in the first planting season of 2016. Among
them, farmers in the multichannel group were nine percent more likely to pay for DT maize seed
than farmers in the control. This difference is significant at the 0.1 level among farmers who are
compliers.
In sum, households in the multichannel treatment group were more likely to purchase
improved maize seed and more willing to grow DT maize in the next primary maize planting
season than the households in the control and video-only treatment groups. The analysis shows

63

that although the multichannel treatment did not increase households’ knowledge about DT
maize practices, it reinforce households’ knowledge about DT maize, and correspondingly
reinforced farmers’ likelihood to purchase improved maize seed and their willingness to try DT
maize.

64

Table 5: Marginal Impacts of Treatments on Purchasing Maize Seed by Treatment Groups
Purchased Improved Maize
Purchased DT Maize
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
-0.03 -0.01
0.04
0.03
Video (1)
(0.05) (0.05)
(0.04) (0.04)
0.09* 0.09*
0.06* 0.04
Multichannel (2)
(0.05) (0.05)
(0.04) (0.04)
-0.07
-0.03
0.09
0.08
Video Complier
(0.08) (0.08)
(0.06)
(0.06)
**
**
*
0.20
0.19
0.15
0.13
Multichannel
Complier
(0.10) (0.10)
(0.15)
(0.08)
Additional baseline
N
Y
N
Y
N
Y
Y
Y
controls 2
Observations
582
582
582
582
582
582
582
582
Likelihood of
purchasing seed in
0.49
0.49
0.49
0.49
0.17
0.17
0.17
0.17
the control group
Households in the
0.48
0.48
0.48
video-only group
(0.02) (0.02)
(0.02)
watched the video 3
Households in the
0.37
0.37
0.37
multichannel group
watched the video
(0.04) (0.04)
(0.04)
and received the
messages 3
F-test values for
Video only =
4.72** 3.71* 8.03*** 6.17** 0.25
0.56
0.56
0.38
Multichannel
1.

Columns 1, 2, 5 & 6 report the Intend-to-Treat (ITT) estimate of the intervention. Columns 3, 4, 7 & 8 report
the Average Treatment Effect or Local Average Treatment Effects (LATE) estimate using the treatment
assignments as instruments for the compliers in each treatment group.
2.
Additional baseline controls in columns 2, 4, 6 & 8 include gender of household head, number of people in
household, livestock index, main income sources (agriculture or non-agriculture), total land size, household
distance to the nearest market selling seed, main maize growers’ education level, residence in FIPS villages or
not, and total amount of fertilizer applied in the 2016 long rain season. In Columns 6 & 8, households in FIPS
villages who have higher asset indexes are more likely to purchase DT maize seed. In Columns 2 & 4, maleheaded households and households located further away from the seed market were more likely to purchase
improved maize seed.
3.
Indicator variable for households in two treatment groups participated in the intervention. For the multichannel
group, households both watched the video and received the phone messages. The results are the first stage of the
Instrumental Variable Regression (following equation (3))
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and
control group are denoted by *** 99%, ** 95%, * 90%.

65

4.1.4 Differential Impacts of the Intervention
I found that the treatment effects were relatively heterogeneous between households
residing in the FIPS villages and households who did not. Table 6 shows that the impact of both
treatments on households’ knowledge about DT maize was mostly driven by those residing in the
FIPS villages. Households residing in the FIPS villages who were assigned to one of the two
treatments gained higher knowledge scores on DT maize than the households who received the
treatments but were located in the non-FIPS villages. Households assigned to the multichannel
treatment group who lived in the FIPS villages scored 0.26 points higher on index measuring
knowledge about DT maize than households in the control group that without the FIPS’ services.
This difference accounts for four percentage points of the total knowledge score on DT maize; it
is quantitatively small, but statistically significant. Households assigned to the video treatment
who lived in the FIPS villages scored 0.15 points higher on the index measuring knowledge
about DT maize than households in the control group. This difference is not statistically
significant. However, it is worth noting that, in the control group, there were no significant
differences in knowledge about DT maize between the farmers living in the FIPS villages and
those in the non-FIPS villages. This suggests that the FIPS’ efforts became important only when
paired with the treatments.
Table 6 shows that the treatment effects on farmers’ knowledge about management
practices were significantly higher among farmers in the multichannel group who resided in the
non-FIPS villages. These farmers scored 0.53 points higher on the index measuring knowledge
about DT maize management practices than farmers in the control outside the FIPS-supported
area. This difference constitutes nine percentage points of the total knowledge score on
management practices. I did not observe similar significant differences in knowledge about

66

practices between the farmers in the video-only treatment group who resided in the non-FIPS
villages, and those in the control group who did not reside in the FIPS-supported area. Further,
farmers in the FIPS villages retained less knowledge than farmers in non-FIPS villages in both
the video-only group and the multichannel group; however, this difference is not significant.
This might be because the FIPS already provided farmers with training regarding management
practices prior to the video screenings. These farmers may have thus learned less about the
management practices from the treatments than farmers in the same treatment groups who did
not have experience working with the FIPS.
I found that farmers living in the FIPS villages who received the multichannel treatment
were more likely to purchase improved seed in the first planting season of 2016 after watching
the video and receiving the audio message reminders compared to farmers who in the video-only
group and control groups, regardless of whether they lived in the FIPS villages or not. Moreover,
farmers’ likelihood of purchasing improved seed was also higher in the video-only treatment
group if they lived in the FIPS villages. In the control group, farmers living in the FIPS villages
were less likely to purchase improved seed than farmers in the non-FIPS villages. This may be
because the FIPS offers their farmers sample maize seed, including DT maize, for free. The two
treatments, especially the multichannel one, motivated farmers living in the FIPS village to
invest more than farmers in the non- FIPS villages in improved maize seed, instead of depending
on the free sample seed, which is normally small in quantity (25g per bag). The multichannel
treatment increased the probability of farmers’ purchasing improved seed by 18 percent when a
household was located in the FIPS villages compared to farmers who lived in the non- FIPS
villages but were assigned to the control. Compared to households in the FIPS villages who were

67

assigned to the control group, the video-only treatment increased percentage of farmers who are
willing to purchase improved seed by 12 percent if a household resided in the FIPS villages.
Households in the video treatment and multichannel treatment groups had higher
willingness to plant DT maize in the next primary season if they lived in the FIPS villages as
opposed to households who lived in villages with no FIPS services (see Table 7). In the control
group, the differences in willingness to plant DT maize were similar between households
residing in the FIPS villages and those who did not. Compared to the control group, the
multichannel treatment increased the probability of uptake by 20 percent when a household lived
in the FIPS villages, and the video-only treatment increased willingness to plant DT by 13
percent if a household resided in the FIPS villages. Both impacts are statistically significant.
VBAs in the villages mainly serve as seed suppliers, so the households in the FIPS villages have
easier access to DT maize than the households in the non- FIPS villages. I speculate this makes
the uptake more likely. In Appendix 3, I present the comparison of baseline characteristics
between households living in FIPS villages and those living in non- FIPS villages.
In Study One, I found that farmers in the multichannel group retained more knowledge
about DT maize and its management practices, and that they were more likely to purchase
improved seed and more willing to grow DT maize in the next primary season than farmers in
the video-only and control groups. Therefore, H2, H4, H6 and H8 are supported in this study.
Further, there is not a significant difference in knowledge and uptake between farmers in the
video-only group and control groups; therefore, this study fails to support H1, H3, H5 and H7.
Moreover, the services provided in FIPS villages enhanced the multichannel treatment effect on
farmers’ knowledge about DT maize and uptake, which suggests that factors such as easy access
to DT maize complement the information provision via trainings and reminders in the

68

multichannel group. However, farmers’ knowledge scores of DT maize management practices
were higher in the non- FIPS villages than farmers in the FIPS villages in both treatment groups,
though this difference is not statistically significant. Thus, these findings partially support H9,
H10 and H11.

69

Table 6: Differential Impacts of Treatments on knowledge about DT by NGO Status (ITT)1
Knowledge about DT
Knowledge about DT
management
maize
practices
(3)
(4)
(1)
(2)
0.15
0.18
-0.17
-0.13
Video (β1)
(0.23)
(0.23)
(0.22)
(0.19)
0.53**
0.46**
0.09
0.06
Multichannel (β2)
(0.23)
(0.23)
(0.17)
(0.12)
-0.40
-0.45
0.32
0.27
Video treatment x FIPS (β3)
(0.31)
(0.31)
(0.23)
(0.20)
-0.40
-0.32
0.16
0.16
Multichannel treatment x FIPS (β4)
(0.31)
(0.32)
(0.24)
(0.21)
-0.25
-0.27
0.15
0.16
Farmers assigned to video treatment residing
2
in FIPS villages
(0.22)
(0.22)
(0.12)
(0.10)
0.14
0.14
0.26***
0.25***
Farmers assigned to multichannel treatment
2
residing in FIPS villages
(0.22)
(0.22)
(0.08)
(0.09)
N
Y
Additional baseline controls
N
Y
582
582
Observations
582
582
Knowledge score in control group (FIPS)
0.55
0.55
7.49
7.49
7.00
7.00
Knowledge score in control group (non-FIPS)
0.54
0.54
1.

Additional baseline controls in columns 2 & 4 include gender of household head, number of people in
household, livestock index, main income sources (agriculture or non-agriculture), total land size, household
distance to the nearest market selling seed, main maize growers’ education level, residence in the FIPS villages or
not, and total amount of fertilizer applied in the 2016 long rain season.
2.

The coefficients are a sum of the main effects and the interaction effects.

Standard errors in parentheses. Significance levels of the differences between the two treatment groups and
control group are denoted by *** 99%, ** 95%, * 90%.

70

Table 7: Differential Impacts of the Treatments on Households’ Probability of Purchasing Improved Maize
Seed and Intention to Grow DT Maize in the Next Season by NGO Status2
Purchased
Intention To Grow
Improved Maize
DT Maize2
(1)
(2)
(3) 2
(4)
Video ( β1)

-0.18**

-0.15

-0.12*

-0.11*

(0.07)

(0.07)

(0.05)

(0.06)

0.03

0.03

0.05

0.02

(0.07)

(0.07)

(0.06)

(0.05)

0.29**

0.27

0.24***

0.23**

(0.10)

(0.10)

(0.08)

(0.09)

0.14

0.14

0.15

0.17*

(0.10)

(0.10)

(0.11)

(0.11)

0.12*

0.12*

0.13**

0.12*

(0.07)

(0.07)

(0.06)

(0.07)

0.18**

0.17

0.20**

0.19**

(0.07)

(0.07)

(0.09)

(0.09)

N

Y

N

Y

Observations

582

582

582

582

Control Mean (FIPS villages)

0.41

0.41

0.38

0.38

Control Mean (non-FIPS villages)

0.54

0.54

0.37

0.37

Multichannel (β2)
Video treatment x FIPS (β3)
Multichannel treatment x FIPS (β4)
Farmers assigned to video treatment residing in
FIPS villages 3
Farmers assigned to multichannel treatment
residing in FIPS villages 3
Additional baseline controls

1

1.

Additional baseline controls in columns 2 & 4 include gender of household head, number of people in
household, livestock index, main income sources (agriculture or non-agriculture), total land size, household
distance to the nearest market selling seed, main maize growers’ education level, residence in FIPS villages or
not, and total amount of fertilizer applied in the 2016 long rain season. In column 2, gender of household head
and house’s distance to the nearest seed market are significant baseline predictors to whether farmers purchased
improved seed after the intervention. In column 4, the number of livestock is the only significant predictor to
farmers’ willingness to plant DT in the next season.
2
. These analyses are based on a linear specification; note that other non-linear models—such as random effects
Probits—yield similar coefficient magnitudes (not reported).
3.
The coefficients are a sum of the main effects and the interaction effects.
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and
control group are denoted by *** 99%, ** 95%, * 90%.

71

4.2 Study Two
I found that farmers’ uptake of DT maize was mainly attributed to the perceived risk of
growing these varieties, which were mitigated by the knowledge farmers gained from attending
the video screening and receiving the audio messages. Farmers in the multichannel treatment had
the lowest perceived risk among the three treatment groups. They were more likely to test DT
maize if they had lower perceived risk. Farmers’ knowledge about DT maize moderated the
relationship between perceived risk and willingness to test DT maize. In fact, for farmers with
little to no knowledge about the varieties, perceived risk was more likely to reduce farmers’
willingness to grow the seed. Whereas, for farmers who had higher knowledge levels, perceived
risk was less likely to associate with farmers’ willingness to grow DT maize. Further, I found
that farmers who perceived lower risk were more likely to purchase DT maize seed and other
improved maize seed. However, perceived benefit was not significantly associated with
probability to purchase the seed, nor with farmers’ willingness to grow the seed in the next
primary season. As DT maize was a new technology for the farmers in this study, helping them
increase their knowledge about the varieties and reduce their perceived risk of the seed by
helping them tackle potential risks tended to motivate farmers to test DT maize. Therefore, future
efforts to communicate information on DT maize should be geared toward knowledge
improvement and risk mitigation.
4.2.1 Perceived Risk and Perceived Benefit by Treatment Groups
As shown in Figure 3, farmers in the multichannel group perceived lower risk in all four
dimensions. The same figure displays farmers’ perception of the four risk perception items that
constituted the perceived risk index in this study. Indeed, farmers were less likely to perceive
that growing DT maize would reduce their yield, waste the money they spent on the inputs(such
72

as DT maize), and time they used to grow the seed. However, only the perceived risk of food
security was significantly lower in the multichannel group than in the control group. Further,
when integrating the four risk dimensions together into the perceived risk index, Table 8 shows
farmers in the multichannel group had lower perceived risk of DT than farmers in the control
group, while farmers in the video-only group perceived similar amounts of risks as farmers in the
control. These differences are not statistically significant when using the Intent-to-Treat
estimation, which includes all sampled farmers in the analysis. Moreover, around half of the
households in the video-only treatment attended the video screenings, and one third of sampled
households in the multichannel treatment group watched the video and received the audio
messages. Therefore, the Local Average Treatment Effect (LATE) was measured using the
treatment assignments as the instruments for compliance in order to estimate the treatment
effects on households who received the treatments. Treated farmers in the multichannel group
had 4.37 points lower perceived risk than farmers’ in the control group and 3.24 points lower
than treated farmers in the video-only group. Both differences are statistically significant.

73

Control

Video only

Video + phone

Total

35

30

25

20

a. having a poor yield

b. losing the money
c. losing the time you d. do not have food in
you spent on the seeds and your family spent
the next season
on growing the maize

1.The

range of perceived risk is between 1 and 49
figure displays the average values of the four risk perception items we used in the current
study to constitute the perceived risk index by treatment assignments.
2. This

Figure 4: Perceived Risk of Growing DT Maize by treatment groups
The treatments presented to farmers regarding the two main characteristics of DT maize
are early maturation and drought resistance. The scientists who developed DT maize state that
these two characteristics can benefit farmers’ production by increasing farmers’ maize
production and by helping farmers to reduce harvest failure due to drought (Abate et al., 2017).
Figure 4 shows that farmers in the multichannel group had a lower level of perceived benefit in
regards to the two characteristics than farmers in the control. Further, as Table 8 shows, when
integrating the two perceived benefit dimensions together into an index, farmers in the
multichannel group reported less perceived benefit of DT maize than farmers in the video-only
and control groups,. These differences are not statistically significant.

74

25

20
Control
Video only
15

Video + phone
Total

10

a. Fast Maturation

b. Drought Tolerant

1.The

range of perceived benefit is between 1 and 49
figure displays the average values of the two benefit perception items used to
constitue the perceived benefit index by treatment assignments.
2. This

Figure 5: Perceived Benefit of Growing DT Maize by Treatment Groups

75

Table 8: Farmers’ Perceptions of DT Maize by Treatment Groups1
Perceived Risk
(1)
(2)
(3)
(4)
0.14 -0.23
Video (1)
(1.03) (1.06)
-1.46 -1.63
Multichannel (2)
(1.03) (1.06)
-0.92 -1.69
Video Complier
(1.61) (1.63)
-4.37**
Multichannel
4.49**
Complier
(2.00) (2.04)
Additional baseline
N
Y
N
Y
controls 2
Observations
582
582
582
582
Farmers’ perception in
27.81 27.81 27.81 27.81
the control group 3
Households in the
0.48
0.48
video-only group
(0.02) (0.02)
watched the video 4
Households in the
0.37
0.37
multichannel group
watched the video and
(0.04) (0.04)
received the messages

(5)
0.23
(1.54)
-1.65
(1.54)

Perceived Benefit
(6)
(7)
(8)
0.58
(1.56)
-1.23
(1.56)
1.36
2.23
(2.40)
(2.39)
-1.28

-0.96

(2.97)

(2.98)

N

Y

N

Y

582

582

582

582

30.29

30.29

30.29

30.29

0.48

0.48

(0.02)

(0.02)

0.37

0.37

(0.04)

(0.04)

0.86

1.26

4

F-test values for
Video only =
Multichannel

1.96

1.46

3.24*

2.09

1.21

1.13

1.

Columns 1, 2, 5 & 6 report the Intend-to-Treat (ITT) estimate of the intervention. Columns 3, 4, 7 & 8 report
the Local Average Treatment Effects (LATE) estimate using the treatment assignments as instruments for the
compliers in each treatment group.
2.
Additional baseline controls in columns 2, 4, 6 & 8 include gender of household head, number of people in
household, livestock index, main income sources (agriculture or non-agriculture), total land size, household
distance to the nearest market selling seed, main maize growers’ education level, residence in the FIPS villages or
not, and total amount of fertilizer applied in the 2016 long rain season.
3.
The range of the perceived risk is 1-49, and the range of perceived benefit is 1-.
4.
49 Indicator variable for households in two treatment groups participated in the intervention. For the
multichannel group, households both watched the video and received the phone messages. The results are the first
stage of the Instrumental Variable Regression (following equation (3)).
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and
control group are denoted by *** 99%, ** 95%, * 90%.

76

These results demonstrate that the multichannel treatment was associated with farmers’
lower risk perception of DT maize, but farmers in the treatment group reported lower perceived
benefits than farmers assigned to the video-only group and control groups.

77

4.2.2 Knowledge, Farmers’ Perceptions, and Willingness to Grow DT Maize
Farmers’ knowledge about DT maize was found to be associated with their risk and
benefit perceptions. Further, knowledge moderated the relationship between farmers’ risk
perception and reported willingness to plant DT maize the next season. Table 9 shows that, after
controlling for the treatments and demographic variables, farmers who gained more knowledge
about DT maize perceived significantly less risk of DT maize.
Table 9: Regression of Farmers’ Perceived Benefit and Perceived Risk as a Function of Knowledge Level,
Treatments, and Baseline Control Variables
Perceived Risk
Perceived Benefit
(3)
(4)
(1)
(2)
Video-only
Multichannel
Knowledge about DT

Observations
Control Mean 2

-0.24

0.15

0.64

(1.03)

(1.05)

(1.51)

(1.52)

-0.97

-1.24

-2.83

*

-2.19

(1.04)

(1.06)

(1.52)

(1.54)

***

3.11***

-1.90

Knowledge about the management practices
Additional baseline controls

0.19

***

***

-1.92

2.88

(0.72)

(0.74)

(1.06)

(1.07)

-0.32

-0.33

*

1.59

***

1.61***

(0.28)

(0.28)

(0.41)

(0.41)

N

Y

N

Y

582

582

582

582

30.73

30.73

15.17

15.17

1

1.

Additional baseline controls in columns 2 & 4 include treatment groups, gender of household head, number of people in
household, livestock index, main income sources (agriculture or non-agriculture), total land size, household distance to the nearest
market selling seed, main maize growers’ education level, residence in the FIPS villages or not, and total amount of fertilizer
applied in the 2016 long rain season. In column 4, the coefficient for distance from the households to the nearest seed market is
positive and significant.
2.
The range of the perceived risk and perceived benefit is 1-49, and the range of the knowledge of DT maize is 0-6 measured in
increments of 0.5 as a unit.
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and control group are
denoted by *** 99%, ** 95%, * 90%.

78

Farmers who perceived fewer risks of DT maize reported that a greater willingness to
grow DT maize in the next primary season. Figure 5 and Table 10 show a significant interaction
effect between farmers’ perceived risk and level of knowledge about DT maize to farmers’
willingness to plant DT maize. For farmers with little to no knowledge about DT maize,
perceived risk was negatively associated with their willingness to plant DT maize in the next
season. Whereas farmers who retained more knowledge about DT showed risk perception was
less likely to be associated with their willingness to grow DT maize. For example, for farmers
who received the treatments (compliers) and scored zero or 0.5 points on the DT maize
knowledge test following the intervention, there was a significant negative relationship between
perceived risks and willingness to test DT in the next primary season. However, among farmers
with higher knowledge about DT maize, perceived risk was not significantly associated with
their uptake decisions. These conclusions are valid even when the baseline control is included in
the regression.

79

Figure 6: The Interaction Effect between Perceived Risk and Knowledge of DT Maize on
Farmers’' Willing to Grow DT Maize

80

Table 10: Differential Impacts of Risk Perception on Households’ Intention to Grow DT Maize in the Next
Season by Level of Knowledge about the Varieties
Compliers
Total Sample
(3)
(4)
(1)
(2)
Perceived Risk (β1)

-0.003*

-0.003*

-0.006***

-0.006***

(0.002)

(0.001)

(0.002)

(0.002)

0.10**

0.11***

0.09**

0.07

(0.04)

(0.04)

(0.05)

(0.05)

0.002*

0.002*

0.003*

0.004**

(0.001)

(0.001)

(0.002)

(0.002)

-0.03

-0.03*

-0.006***

-0.006***

(0.002)

(0.001)

(0.002)

(0.002)

β1+β3 ( Knowledge of DT = 0.5)

-0.001

-0.002

-0.004**

-0.004**

(0.001)

(0.002)

(0.002)

β1+β3 ( Knowledge of DT = 1)

(0.001)
0.0002
(0.001)

-0.0006

-0.002

-0.002

(0.001)

(0.002)

(0.002)

0.001

0.0004

0.0001

0.0001

(0.002)

(0.001)

(0.002)

(0.002)

0.002

0.001

0.002

0.002

(0.004)

(0.002)

(0.003)

(0.003)

N

Y

N

Y

Observations

582

582

347

347

Risk perception in the control group

0.16

0.16

0.14

0.14

Knowledge of DT (β2)
Perceived Risk x Knowledge of DT (β3)
β1+β3 ( Knowledge of DT = 0)

β1+β3 ( Knowledge of DT = 1.5)
β1+β3 ( Knowledge of DT = 2)
Additional baseline controls 1

1.

Additional baseline controls in columns 2 & 4 include treatment groups, gender of household head, number of
people in household, livestock index, main income sources (agriculture or non-agriculture), total land size,
household distance to the nearest market selling seed, main maize growers’ education level, residence in the FIPS
villages or not, and total amount of fertilizer applied in the 2016 long rain season. The coefficient for size of land is
negative and significant.
2.
The range of the perceived risk is 1-49, and the range of knowledge of DT maize is measured in increments of
0.5 as a unit.
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and control
group are denoted by *** 99%, ** 95%, * 90%.

Lastly, farmers who perceived larger benefits in regards to DT maize retained more
knowledge about DT maize and its management practices. This may be because farmers with

81

knowledge about DT maize management practices had higher levels of confidence that their
family could benefit from producing DT maize. However, perceived benefit was not associated
with willingness to plant DT maize in the next season (see Appendix 13).
In sum, I found that mitigating the perceived risk of growing DT maize played a crucial
role in motivating farmers to test the varieties. Farmers in the two treatment groups, especially
the multichannel group, had lower perceived risk than farmers in the control group. Farmers’
perceived risk was associated with their level of knowledge about DT maize and to which kind
of treatment they were exposed. These findings support H12 and H14. The strength of the
association between perceived risk and uptake was lessened by farmers’ increasing of knowledge
about DT maize over time. This finding answers RQ6. However, no relationship was found
between perceived benefits, treatment assignment, and knowledge level. Therefore, H13 and
H15 are not supported by this study.

82

CHAPTER 5 DISCUSSION
Study One examined the impacts of participatory video and audio message reminders in
both enhancing farmers’ knowledge about DT maize regarding awareness of DT seed varieties as
and agricultural practices, and keeping farmers’ attention on DT maize after attending the video
screening and discussion. The findings thus provide insights that can inform the design of ICT
strategies aimed at communicating knowledge about DT maize and its complementary
agricultural practices to farmers. To my knowledge, few studies have explored strategies to
reinforce participatory video trainings by combining video with other approaches. This study
found audio messages reinforced video training content and reminded farmers about agricultural
management practices based on the maize growing stages. The multichannel treatment that
included both video and audio reminders was more effective than the video-only treatment in
communicating relatively complex agricultural technologies with multiple steps. I found farmers
in the video-only and multichannel groups retained more knowledge about DT maize and its
management practices than farmers in the control group. However, only the difference between
the multichannel group and the control group is significant. This result suggests that a
multichannel approach including both video and audio messages is a more effective design than
the video only treatment. The video provided contextualized knowledge about DT maize and its
associated agricultural management practices. The participatory content provides important
context complemented by the short audio messages at appropriate times in the growing season.
For example, two audio messages reminded farmers to engage in fertilizer micro-dosing, a topic
the farmers learned about and viewed during the video screening, more than a month after the
video screenings. Hence, one explanation for the heightened effectiveness of the multichannel

83

treatment is that the audio messages reminded farmers of the video content at key times over the
course of a relatively long growing season.
This study contributes to extant literature on participatory video by adapting this
approach to integrate contextualized content into learning materials such as training videos. The
video and all four audio messages were developed by working closely with local farmers and
inviting them to share their personal stories about: how DT seed can benefit local farmers like
them, the risks they encountered, and how they can mitigate the risks using modern practices of
maize management. In focus group discussions following the post test, farmers explained that
this situated content, such as a scene showing actors selecting various seeds in a local seed shop,
aided their learning, particularly in regards to the difference between improved seed varieties and
modern management practices like fertilizer micro-dosing. In the pre-test, we found that the
majority of farmers only planted or heard about older improved varieties, such as Duma43. Few
knew about newer improved varieties like DT maize or the difference between Duma43 and DT
maize. Therefore, farmers believed that they could just grow the older improved varieties in lieu
of newer ones because they assumed the performance would be the same. Farmers told us that
they had heard something about improved varieties through sources such as agriculture programs
on the radio, roadside advertising, demonstration plots, relatives living in cities, and directly
from seed company agents. However, they further explained that the video screening and
discussion was the first time they learned about the many improved seed varieties and how they
might choose the appropriate varieties for their farm.
In this study, the video screenings were moderated by two FIPS experts from the local
communities who were knowledgeable on DT maize . We observed that, during the screenings,
these moderators were essential to engaging farmers in watching and discussing the videos, as

84

well as to addressing farmers’ questions immediately after screening. Much as was found in
Wyche et al. (2016), we doubt that simply showing a short video alone would be sufficient for
teaching farmers about new agricultural technologies. Wyche et al. (2016) found farmers needed
moderators to explain the video content during and after viewings. Farmers also benefited from
participating in a group, as they could help each other learn and resolve questions.
However, the study extends the Wyche et al. (2016) research of questioning the
effectiveness of video-only approaches to educating rural farmers by showing a need to further
complement video-based informational strategies, even those deeply rooted in the local context,
with additional communication strategies; in this case, with mobile phone reminders to
supplement the video-only training. Farmers commonly need to understand multiple steps and
various components in order to adopt new agricultural technologies. Therefore, they need
consistent information and coaching to learn and to reinforce the new knowledge. For example,
Digital Green, an Indian NGO discussed in the literature review, has developed the participatory
video training approach, which has been adopted by various organizations in India to provide
information to farmers. The success of this approach depends on a substantial number of skilled
frontline extension workers who consistently work with communities to help them learn from the
training videos and resolve issues that impede adoption. However, in the majority of Sub
Saharan African countries (including our study area), training videos are normally only shown to
a community once, due to the lack of video playing devices and proficient personnel to handle
these devices and to moderate screening sessions. Further, due to a deficiency of skilled
extension workers and other training resources in the majority of Sub Saharan African countries,
extension workers are not able to provide frequent follow-up to enhance farmers' knowledge
about a technology after the video screening. The multichannel strategy we tested in this study

85

creates a synergy between various ICT extension approaches to provide farmers contextualized
information using the video approach. Without engaging with extension workers, the audio
messages remind farmers about the content of the video before key decision points in the maize
growing season. Therefore, the audio reminders in this study enhanced the video approach based
on the local context to compensate for the deficiencies of knowledge reinforcement after the
video screening.
The importance of the reminders has become more widely-recognized by scholars and
practitioners because of its impact on the reinforcement of knowledge and its capacity for
nudging behavior in areas such as agricultural production. This finding corroborates Larochelle,
Alwang & Travis (2016) conclusion in Northern Ecuador that reminders can reduce inattention,
especially when farmers need to make complex decisions. Further, the first reminder sent out
after the video screening in this study encouraged more farmers to purchase improved maize
seed compared to farmers who did not receive the message. This result suggests that tailored
reminders sent out at key decision-making points to encourage specific behavior in farmers’ can
be effective, a suggestion similarly made by Cole & Fernando (2012).
Larochelle et al. (2016) argued that reminders had greater impact on simple tasks or
farmers’ non-purchase-reliant behaviors. On the contrary, this study found that farmers in the
multichannel group were significantly more likely to purchase improved seed comparing with
the video-only and the control group, while there is no difference in fertilizer application to
improved maize varieties between the treatment groups and the control. In this study, farmers
generally purchased a small amount of improved seed. On average, farmers purchased 4kg of
improved seed, and one kg of the seed costs around 200KSh (slightly less than 2 USD). Farmers
tended to purchase a small quantity of seed for testing rather than purchasing larger scale maize
86

production, and it is likely that the low cost did not substantially hinder farmers’ decisions to
purchase maize seed. The labor of planting was also less likely to impede farmers’ seed
purchasing decisions than the labor of fertilizer application. However, fertilizer application can
involve a relatively larger investment in capital and labor. One bag of fertilizer costs $25 to $35
per bag (50kg) in the local market, depending on the fertilizer variety. For the farmers, the
prohibitively-high investment for fertilizers reduced the impact of treatment on the farmers’
decisions to purchase fertilizer.
Study Two shows that, in order to increase farmers’ willingness to plant DT maize, it is
important to understand farmers’ perceptions of DT maize and the relationship between these
perceptions and other impacts of treatments like improving knowledge and adoption. Study Two
tests a pathway from the treatments to farmers’ knowledge about DT maize and management
practices to farmers’ risk and benefit perception of the varieties and to their uptake decisions. If
farmers have inadequate or inaccurate knowledge about DT maize, their perception of risk can
be higher than those who have sufficient knowledge. Moreover, the association between
perceived risk and farmers’ risk-taking behavior can differ between farmers with varied levels of
knowledge. I do not intend to reduce the actual risks of growing DT because it is beyond the
scope of this study. Instead, this study tests strategies aimed at helping farmers form perceptions
of DT maize based on facts and knowledge and to improve farmers’ knowledge about
management practices in order to reduce the potential risks of growing DT maize.
Study Two was designed based on a premise by Weber & Milliman (1997) that
situational factors like proper timing can help individuals form proper risk perceptions. The
informed risk perception in turn facilitates decision-making. Sundblad, Biel & Gärling (2007)
suggested perceived risk can be properly mitigated by facilitating individuals’ cognitive
87

processes. Fisher et al. (1996) found that providing information to farmers communicating
potential risks of an innovation can mitigate their perceived risk.
I found that, if farmers had less or no knowledge of DT maize, their higher risk
perception was associated with a lower willingness to test DT maize; this result is consistent with
previous literature, which found a negative relationship between individuals’ perceived risk and
risk-taking behavior (Brewer et al., 2004). However, if farmers had higher knowledge about DT
maize, perceived risk had no impact on their willingness to plant DT maize in the next primary
season. This may be because farmers’ willingness to take risks can be conditioned by their level
of knowledge about an innovation and risk reducing practices. For example, a farmer might
report that the risk of growing DT maize is high because the maize may not perform well if he or
she chose the wrong variety of DT maize, or if no complementary practices were used to manage
the maize. However, they were also aware that the risks of growing DT maize could be reduced
if they chose corrected DT maize and used timely practices to manage the crop. In contrast,
another farmer might say his perceived risk of growing DT maize was high because he does not
know about DT maize and its management practices. Therefore, although the two farmers might
have similar levels of perceived risk, the first farmer is more likely to plant DT maize than the
second. As such, risk perception can predict future behavior differently. Thus, two people with
the same responses to a risk question—“Is growing DT maize risky?”—can have very different
degrees of interest in DT maize. Level of knowledge about the innovation and about practices
that can reduce risks moderate the relationship between perceived risks and uptake (Brewer et
al., 2004). These findings support Sundblad, Biel & Gärling’s (2007) study, which argued that
knowledge can indirectly effect individuals’ behavior via risk perception.

88

Another explanation for the finding that level of knowledge moderates the relationship
between perceived risks and uptake is due to the indirect questioning measures used to capture
farmers’ risk and benefit perception of DT maize in this study. Surveyed households were asked
about their perceptions of DT maize through invitations to evaluate what other farmers like them
might perceive the risks and benefits of DT maize to be in order to reduce potential social
desirable biases. The purpose of the indirect questioning measure was to avoid the social
desirability biases among farmers. This measure, however, can lead to third person bias (Gunther
& Mundy, 1993) or the phenomenon that people who are more knowledgeable tend to see
themselves as more resistant to risks when compared to others. That is to say, farmers with
higher levels of knowledge about DT maize choose to plant it even when they perceive other
farmers like them might experience higher risks of growing DT maize. Although this study does
not include data about farmers’ perception of risks on their own production and how this risk
perception relates to their uptake of DT maize, we speculate that farmers’ knowledge of DT
maize was a primary factor leading to their willingness to test DT maize.
Additionally, the current study found that the treatments did not increase farmers’
perceptions of the benefits of DT maize. The perceived benefit was not shown to be associated
with farmers’ willingness to grow DT maize, which differs from the relationship between
perceived risks and the risky behavior. One explanation for this is that perceived risks are more
likely to affect farmers’ uptake decisions than perceived benefits, even when farmers are
convinced of the potential benefits. Prospect theory, as described by Kahneman & Tversky
(1979), can explain this finding, as it suggests people’s behavior is conditioned by their
evaluation of losses and gains. However, people tend to be loss averse, thus, their perceived risk
has a larger influence on their behavior than perceived benefit. Another explanation is that

89

farmers can form their benefit perception by means other than the information provided by the
intervention; examples of this include their personal observations of a new technology’s benefits
and the profitability of the seed. Therefore, the farmers’ evaluation of the benefits can be differ
from the assessments made of the same object by scientists (Lee, 1981). This suggests that other
efforts are needed to help farmers understand whether and how DT maize can benefit their
households’ production, such as farmers’ own experimentation. Moreover, farmers can perceive
a new technology differently from the scientists and technicians who developed the technology.
Therefore, more studies are needed to help scholars understand farmers’ perceptions of DT
maize, how they align with or diverge from the creators’ initial intentions, and what content
should be included in interventions aiming to increase adoption.

90

CHAPTER 6 LIMITATIONS AND FUTURE WORK
A number of limitations should be kept in mind when interpreting the results of this study
and planning for future studies. The methodology, study design, and serval contextualized
factors led to these limitations. Due to resource constraints, this study was not able to measure
farmers’ actual uptake of DT maize and its management practices in their primary growing
season. Future studies should plan for a longer-term assessment of this multichannel approach to
understand its impact on farmers’ knowledge and uptake over time. Further, the current study
design does not include an audio messages only treatment, which only sends out farmers the
audio messages without videos. Further research should incorporate this treatment to understand
the marginal impact of participatory videos on farmers’ knowledge and uptake of agricultural
innovation in a multichannel approach. Future studies should also focus on investigating what
types of reminders are more effective, when to send them to farmers, and how to design different
multichannel approaches that match the characteristics of various agricultural innovations as well
as farmers’ needs and cognitive capacities. In other words, future studies can focus on content
design and creation.
Although the study area had more than a 90 percent mobile phone subscription rate, we
experienced several challenges in delivering the audio mobile phone reminders to farmers. The
audio messages were delivered by making automated phone calls with audio recordings in the
local language. This approach differs from those taken in other studies (e.g., Larochelle et al.,
2017) where text messages were sent to farmers. Our approach intentionally diverged from this
text-based approach because previous research among rural farmers in Kenya found low mobile
phone literacy among this population and limited abilities to read text messages (Wyche &
Steinfield, 2016). Despite our use of audio recordings, however, we still encountered difficulties

91

delivering messages. The main challenges were network failures and non-functioning phone
numbers. Additionally, in the focus group discussions, some farmers told us they did not pick up
the calls because they were busy and did not expect to receive the messages. Thus, we suggest
that future attempts to use this approach should consider implementing the sending of hints prior
to sending out the audio messages to inform farmers of the scheduled calls. It is also possible to
convert from a “push” strategy to a “pull” strategy, where farmers can trigger a system to send
the messages based on their demand. Lastly, although we found some farmers were able to recall
the content in the audio messages, it was impossible for them to revisit these audio messages
when they wished. Therefore, useful revisions to this design for future studies include providing
a means for farmers to re-listen to the messages and providing the information in a reviewable
form, such as in a handout.
We also found that the lack of supply of DT maize in local markets undermined treatment
impacts. For example, we found that, in a village where a local agricultural input shop supplied a
few DT maize varieties, more farmers were willing to try DT seed after the intervention than
those in villages where DT varieties were hard to find. Farmers living in the former village told
us that they asked the owners to supply the DT seed they learned about from the video. This
owner was originally from the village, worked full time in the county agriculture office, and
operated this local input shop as a part-time job. We learned that this owner had access to supply
channels through which he obtained the seed varieties requested by farmers, and therefore was
able to supply the seed directly to his community. Farmers in other villages who did not report
requesting DT maize seed from an input shop may have not been able to do this due to the lack
of an input supplier in the village, or because they did not have a trusted relationship with the
seed suppliers located there.

92

Finally, in this study, we mainly used a randomized controlled trial (RCT) approach and
focus group discussions as methodologies to evaluate the impact of the video and multichannel
strategies. The former methodology is widely used in development studies to measure the effects
of specific developmental interventions. Although we were able to use this methodology to
capture the overall impact of the interventions, it did not allow us to fully capture the
heterogeneity of the treatment impacts on various households in the study (Barrett & Carter,
2010). Moreover, it did not fully reveal the mechanics of how the two communication strategies,
particularly the multichannel approach, changed farmers’ knowledge and willingness to test DT
maize. Therefore, we encourage future studies to implement a more mixed-method approach by
incorporating participatory research methods to complement the RCT and develop a more indepth understanding of the treatment impact.

93

CHAPTER 7 CONCLUSION
Overall, this study designs and tests two communication strategies created to increase
farmers’ knowledge and uptake of DT maize, locally-made “participatory” video and a
multichannel method that incorporates the participatory video with a series of mobile phonebased audio messages. Study One found that farmers in the multichannel group had higher levels
of knowledge about DT maize and its accompanying management practices. They were shown to
be more likely to purchase improved seeds after the treatment, and were more willing to plant
DT maize in the next primary season. The effects of both treatments were largely driven by
farmers living in villages supported by FIPS that provided trained village-based advisors to
perform informational assistance and access to inputs. The second study found that farmers in
the multichannel group had lower levels of perceived risk and lower levels of perceived benefit
in regards to planting DT maize than farmers in the video-only and control groups. Perceived
risk was more likely to reduce farmers’ willingness to test DT if they retained little knowledge
about DT maize. On the other hand, among farmers who had higher levels of knowledge about
DT maize, perceived risk was less likely to be associated with willingness to grow DT maize. As
such, it appears that the multichannel treatment reduced farmers’ perceived risk in regards to
growing DT maize, which in turn encouraged them to plant DT maize.
These findings suggest that providing timely reminders and contextualized video content
can increase farmers’ knowledge about an agricultural innovation and reduce the negative impact
of perceived risk on uptake. Moreover, mitigating farmers’ perceived risk by increasing their
knowledge about an innovation can effectively increase their willingness to try the innovation.
Furthermore, this study investigated how to design ICT-based approaches to make them
more effective in innovation diffusion, which contributes to current literature on ICT usage in

94

agricultural knowledge provision. Such literature has mainly focused on testing whether ICTs
can be effective in improving farmers’ knowledge and uptake. This study demonstrates that the
design of ICT approaches depends on farmers’ cognitive capacities, characteristics of the
agricultural innovation (such as whether farmers need to practice multiple steps over a long
cycle), and the functions of various ICTs. Thus, this research can motivate implementation and
examination of knowledge learning and risk perception theories in ICTD research and practices.
This study will hopefully encourage ICTD researchers and practitioners to use these theories to
guide their own design, testing, and evaluation of ICT strategies in agricultural training and rural
development.

95

APPENDICES

96

Appendix 1 Summary Statistics
Table 11: Summary Statistics
Variable
Knowledge Scores
% of correct answers on DT maize
% of correct answers on the
complementary
practices
% of correct answers of all knowledge
questions
% of maize land allocated to plant DT
Maize grower's characteristics
Education Level
No formal education
Did not complete primary
Completed primary
Did not complete secondary
Completed secondary or higher
Age

Mean

SD

Min

Max

N

10.1

0.1

0

91.67

582

66.45

0.11

18.18

100

582

46.53

0.08

11.76

79.41

582

6.67
21.63
34.31
17.24
20.17
49.9

15.97

18

95

582

Household characteristics
Number of household members (15-65 yrs)
Kilometers to main seed source
Gender of household head

3.11
8.60
23.58

1.7
9.28
42.49

0
0
0

12
50
1

582
582
582

Assets and income
Size of farmland (acre)
Livestock index
Agriculture as primary source of income
Own mobile phone

3.27
177.78
55.94
92.94

2.67
170.04
49.69
25.63

0
0
0
0

17
782
1
1

582
582
582
582

97

Appendix 2 Household Characteristics in Baseline by Compliance with Treatments
Table 12: Household Characteristics in Baseline by Compliance with Treatments

Age of household
Number of household members
between 15-65
Number of household members
Household’s distance to market
selling seed
Household’s distance to trading
center
Size of land (acres)
Livestock index
Women headed household
Own mobile phone
Household with no education
Household with secondary
education or higher
Household in FIPS villages
N

non- Complier
51.737
(1.31)
3.07
(0.13)
5.301
(0.20)
7.27
(0.66)
7.269
(0.66)
4.189
(0.43)
181.43
(12.56)
0.247
(0.03)
0.941
(0.02)
0.081
(0.02)
0.129
(0.03)
0.602
(0.04)
186

1.

Complier
49.041
(0.75)
3.139
(0.08)
5.59
(0.13)
6.589
(0.37)
6.57
(0.38)
3.706
(0.21)
176.056
(8.54)
0.23
(0.02)
0.965
(0.01)
0.058
(0.01)
0.175
(0.02)
0.466
(0.03)
395

Mean Difference
2.696*
(1.42)
-0.069
(0.15)
-0.289
(0.24)
0.682
(0.71)
0.699
(0.71)
0.483
(0.43)
5.374
(15.13)
0.017
(0.04)
-0.024
(0.02)
0.022
(0.02)
-0.046
(0.03)
0.136*
(0.04)
581

For the first three columns, the means and standard deviations of each variable among farmers receiving the
treatments in the treatment villages who did not receive the treatment in the control (complier), those who did not
receive treatments in the treatment villages, and those who received treatment in the control villages (nonComplier) are reported. In the last three columns, the differences were calculated using the following regression:
Yi = Îą1 Complieri + Âľi. Regression standard errors are reported in parentheses.
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and
control group are denoted by *** 99%, ** 95%, * 90%.

98

Appendix 3 Household Characteristics in Baseline by NGO status1
Table 13: Household Characteristics in Baseline by NGO status1

Age of household
Number of household
members between 15-65
Number of household
members
Household’s distance to
market selling seed
Household’s distance to
trading center
Size of land (acres)
Livestock index
Women headed household
Own mobile phone
Household head with no
education
Household head with
secondary education or
higher
N

Non-FIPS
50.989
(0.97)
3.218
(0.10)
5.439
(0.15)
9.275
(0.60)
10.16
(0.61)
3.919
(0.31)
168.267
(9.64)
0.214
(0.02)
1.965
(0.01)
0.074
(0.02)
0.172

FIPS
48.858
(0.90)
3.02
(0.10)
5.554
(0.16)
4.431
(0.23)
7.027
(0.45)
3.804
(0.26)
186.932
(10.27)
0.257
(0.03)
1.949
(0.01)
0.057
(0.01)
0.149

Mean Differences
2.131*
(1.32)
0.197
(0.14)
-0.115
(0.22)
4.844***
(0.63)
3.134***
(0.76)
0.115
(0.40)
-18.666
(14.10)
-0.043
(0.04)
0.016
(0.02)
0.016
(0.02)
0.023

(0.02)
234

(0.02)
347

(0.03)
581

1.

For the first three columns, the means and standard deviations of each variable in the FIPS and non- FIPS
villages are reported. In the last three columns, the differences were calculated using the following regression: Yi
= Îą1 FIPS i + Âľi. Regression standard errors are reported in parentheses.
Standard errors in parentheses. Significance levels of the differences between the two treatment groups and
control group are denoted by *** 99%, ** 95%, * 90%.

99

REFERENCES

100

REFERENCES

Abate, T., Fisher, M., Abdoulaye, T., Kassie, G. T., Lunduka, R., Marenya, P., & Asnake, W.
(2017). Characteristics of maize cultivars in Africa: How modern are they and how many do
smallholder farmers grow? Agriculture & Food Security, 6(1), 30.
http://doi.org/10.1186/s40066-017-0108-6.
Alhakami, A. S., & Slovic, P. (1994). A psychological study of the inverse relationship between
perceived risk and perceived benefit. Risk Analysis, 14(6), 1085–1096.
Barber, R.G. Moore, T.R. Thomas, D. B. (1979). The erodibility of two soils from Kenya.
Journal of Soil Science, 30, 579–591.
Barrett, C. B., & Carter, M. R. (2010). The power and pitfalls of experiments in development
economics: Some non-random reflections. Applied Economic Perspectives and Policy,
32(4), 515–548. http://doi.org/10.1093/aepp/ppq023.
Bauer, R. A. (1960). Consumer behavior as risk taking. In Proceedings of the 43rd Conference of
the American Marketing Association.
Bernard, T., & Taffesse, A. S. (2014). Aspirations: An approach to measurement with validation
using ethiopian data. Journal of African Economies, 23(0), 189–224. doi:10.1093/jae/ejt030
Binder, A. R., Cacciatore, M. a, Scheufele, D. a, Shaw, B. R., & Corley, E. a. (2012). Measuring
risk/benefit perceptions of emerging technologies and their potential impact on
communication of public opinion toward science. Public Understanding of Science (Bristol,
England), 21, 830–47. doi:10.1177/0963662510390159.
Binswanger, H. P. (1980). Attitudes toward risk : Experimental measurement in rural india.
Agricultural & Applied Economics Association, 62(179), 395–407.
Bishop, R. G., & Heberlein, T. A. (1979). Measuring Values of Extramarket Goods: Are Indirect
Measures Biased? American Journal of Agricultural Economics, 61(5), 926–930.
http://doi.org/10.2307/3180348.
Brewer, N. T., Weinstein, N. D., Cuite, C. L., & Herrington, J. (2004). Risk Perceptions and
Their Relation to Risk Behavior. Annals of Behavior Medicine, 27(2), 125–130.
http://doi.org/10.1207/s15324796abm2702_7
Byerlee, D., & de Polanco, E. H. (1986). Farmers’ stepwise adoption of technological packages:
evidence from the mexican altiplano. American Journal of Agricultural Economics, 68(3),
519–527.

101

Byerlee, D., & Heisey, P. W. (1996). Past and potential impacts of maize research in subSaharan Africa: a critical assessment. Food Policy, 21(3), 255–277. doi:10.1016/03069192(95)00076-3.
Cai, T., Chiwasa, H., Steinfield, C. & Wyche, S. (2015). Participatory videos for smallholder
farmer training in Malawi: an analysis of knowledge gain and uptake. In Seventh
International Conference on Information and Communication Technologies and
Development. Singapore.
Cai. T., Steinfield, C., Chiwasa, H. & Ganunga, T. (2017) Using participatory video as an inquiry
method to understand Malawian farmers’ slow adoption of composting.
Casaburi, L., & Kremer, M. (2014). Harnessing ICT to Increase Agricultural Production:
Evidence From Kenya.
Chachamovich, E., Fleck, M. P., & Power, M. (2009). Literacy affected ability to adequately
discriminate among categories in multipoint Likert Scales. Journal of Clinical
Epidemiology, 62(1), 37–46. http://doi.org/10.1016/j.jclinepi.2008.03.002.
Cobb, M., & Macoubrie, J. (2004). Public perceptions about nanotechnology: risk, benefits and
trust. Journal of Nanopartical Research, (6), 395–405.
Cole, S. A., & Fernando, A. N. (2012). The Value of Advice: Evidence from Mobile PhoneBased Agricultural Extension. SSRN Electronic Journal.
http://doi.org/10.2139/ssrn.2179008.
Coombs, C. H., & Lehner, P. E. (1981). Evaluation of two alternative models for a theory of risk:
I. Are moments of distributions useful in assessing risk? Journal of Experimental
Psychology: Human Perception and Performance, 7(5), 1110–1123.
http://doi.org/10.1037/0096-1523.7.5.1110
Cunningham, S. M. (1967). The major dimensions of perceived risk. In Risk Taking and
Information Handling in Consumer Behavior. Boston, MA: Harvard University Press.
Diagne, A., & Demont, M. (2007). Taking a new look at empirical models of adoption: average
treatment effect estimation of adoption rates and their determinants. The Africa Rice Center.
(AfricaRice). Agricultural Economics, 37, 201–210. doi:10.1111/j.1574-0862.2007.00266.x
Duflo, E., Kremer, M., & Robinson, J.(2008). How high are rates of return to fertilizer? Evidence
from field experiments in Kenya. American Economic Review, 98, 482–488.
doi:10.1257/aer.98.2.482.
Duguid, P. (2005). “The Art of Knowing”: Social and Tacit Dimensions of Knowledge and the
Limits of the Community of Practice. The Information Society, 21, 109–118.
doi:10.1080/01972240590925311.

102

Ertiro, B. T., Beyene, Y., Das, B., Mugo, S., Olsen, M., Oikeh, S., … Prasanna, B. M. (2017).
Combining ability and testcross performance of drought-tolerant maize inbred lines under
stress and non-stress environments in Kenya. Plant Breeding, 136(2), 197–205.
http://doi.org/10.1111/pbr.12464.
Fafchamps, M., & Minten, B. (2012). Impact of SMS-based agricultural information on Indian
farmers. The World Bank Economic Review, 26(3), 383–414.
Fausti, S., & Gillespie, J. (2006). Measuring risk attitude of agricultural producers using a mail
survey: how consistent are the methods? Agricultural and Resource Economics, 50(2), 171–
188.
Finucane, M. L., & Maybery, M. T. (1996). Risk perceptions in Australia. Psychological
Reports, 79, 1331–1338.
Fischhoff, B., Slovic, P., & Lichtenstein, S. (1978). Fault trees: Sensitivity of estimated failure
probabilities to problem representation. Journal of Experimental Psychology: Human
Perception and Performance, 4, 330–344.
Fisher, R. (1993). Social desirability bias and the validity of indirect questioning. Journal of
Consumer Research, 20(2).
Fisher, M., Abate, T., Lunduka, R. W., Asnake, W., Alemayehu, Y., & Madulu, R. B. (2015).
Drought tolerant maize for farmer adaptation to drought in sub-Saharan Africa:
Determinants of adoption in eastern and southern Africa. Climatic Change, 133, 283–299.
doi:10.1007/s10584-015-1459-2.
Fischer, A. J., Arnold, A. J., & Gibbs, M. (1996). Information and the Speed of Innovation
Adoption. American Journal of Agricultural Economics, 78(4), 1073–1081.
http://doi.org/10.2307/1243863.
Fisher, M., & Snapp, S. (2014). Smallholder farmers’ perceptions of drought risk and adoption of
modern maize in Southern Malawi. Experimental Agriculture, 50, 533–548.
doi:10.1017/S0014479714000027.
Foster, A. D., & Rosenzweig, M. R. (2010). Microeconomics of Technology Adoption. Annual
Review of Economics, 2, 395–424. doi:10.1146/annurev.economics.102308.124433.
Foster, A. D., Rosenzweig, M. R., Journal, T., & Dec, N. (1995). Learning by doing and learning
from others : Human capital and technical change in agriculture. The Journal of Political
Economy, 103, 1176–1209.
Gandhi, R., Veeraraghavan, R., Toyama, K., & Ramprasad, V. (2009). Digital Green :
Participatory Video and Mediated Instruction for Agricultural. In Information Technologies
and International Development (Vol. 5, pp. 1–15).

103

Gerber, A., & Green, D. (2012). Field Experiments: Design, Analysis, and Interpretation. W. W.
Norton & Company.Gore, M. L., & Kahler, J. S. (2015). Using visual scales in researching
global human dimensions of wildlife. Human Dimensions of Wildlife, 20(2).
Government of Kenya. (2013). Makueni County. Data Sheets Kenya.
Gregson, S., Zhuwau, T., & Ndlovu, J. N. C. A. (2002). Methods to reduce social desirability
bias in sex surveys in low-development settings: Experience in Zimbabwe. Sexually
Transmitted Diseases, 29(10), 568–575.
Gunther, A., & Mundy, P. (1983). Biased optimism and the third-person effect. Journalism
Quarterly, 70(1), 58–67.
Hassan, R. M. (1998). Maize Technology Development and Transfer: A GIS Application for
Research Planning in Kenya.
Heider, F. (1946). Attitudes and cognitive organization. The Journal of Psychology, 21, 107–
112.
Hendrix, C. S., & Glaser, S. M. (2007). Trends and triggers: Climate, climate change and civil
conflict in Sub-Saharan Africa. Political Geography, 26(6), 695–715.
http://doi.org/10.1016/j.polgeo.2007.06.006
Hofmann, W., & Schmitt, M. (2008). Advances and challenges in the indirect measurement of
individual differences at age 10 of the implicit association test. European Journal of
Psychological Assessment, 24(4), 207–209. http://doi.org/10.1027/1015-5759.24.4.207.
Jacoby, J., & Kaplan, L. (1972). The components of perceived risk. In 3rd Annual Conference,
Association for Consumer Research (pp. 382–93). Chicago, IL.
Just, R. E. (2003). Risk research in agricultural economics: Opportunities and challenges for the
next twenty-five years. Agricultural Systems, 75, 123–159. doi:10.1016/S0308521X(02)00063-X.
Kabunga, N. S., Dubois, T., & Qaim, M. (2012). Heterogeneous information exposure and
technology adoption: The case of tissue culture bananas in Kenya. Agricultural Economics,
43, 473–486. doi:10.1111/j.1574-0862.2012.00597.x.
Kahneman, D., & Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk.
Econometrica, 47(2), 263–292. http://doi.org/10.2307/1914185
Kassie, M., Jaleta, M., & Mattei, A. (2014). Evaluating the impact of improved maize varieties
on food security in rural Tanzania: Evidence from a continuous treatment approach. Food
Security, 6, 217–230. doi:10.1007/s12571-014-0332-x.
Kihara, J., Nziguheba, G., Zingore, S., Coulibaly, A., Esilaba, A., Kabambe, V., … Huising, J.
(2016). Understanding variability in crop response to fertilizer and amendments in sub104

Saharan Africa. Agriculture, Ecosystems and Environment, 229, 1–12.
http://doi.org/10.1016/j.agee.2016.05.012.
Khonje, M., Manda, J., Alene, A. D., & Kassie, M. (2015). Analysis of adoption and impacts of
improved maize varieties in Eastern Zambia. World Development, 66, 695 706.
doi:10.1016/j.worlddev.2014.09.008.
La Rovere, R., Abdoulaye, T., Kostandini, G., Guo, Z., Wilfred, M., MacRobert, J., & Dixon, J.
(2014). Economic, production, and poverty impacts of investing in maize tolerant to
drought in Africa: An ex-ante assessment. The Journal of Developing Areas, 48(1), 199–
225.
La Rovere, R., G. Kostandini, T. Abdoulaye, J. Dixon, W. Mwangi, Z. Guo, A. M. B. (2010).
Potential impact of investments in drought tolerant maize in Africa (pp. 1–136).
Langyintuo, A. S., Mwangi, W., Diallo, A. O., MacRobert, J., Dixon, J., & Bänziger, M. (2010).
Challenges of the maize seed industry in eastern and southern Africa: A compelling case for
private-public intervention to promote growth. Food Policy, 35(4), 323–331.
doi:10.1016/j.foodpol.2010.01.005.
Larochelle, C., Alwang, J., & Travis, E. (2016). Did you really get the message ? Using text
reminders to stimulate adoption of agricultural technologies.
Loewenstein, G. F., Weber, E. U., Hsee, C. K., & Welch, N. (2001). Risk as feelings.
Psychological Bulletin, 127(2), 267–286. doi:10.1037/0033-2909.127.2.267.
Lovo, S. (2013). Tenure insecurity and investment in soil conservation. Evidence from Malawi.
Clinate Change and Environment, 1–25.
Marenya, P. P., Barrett, C. B., & Barrett, B. (2016). Agricultural & Applied Economics
Association State-Conditional Fertilizer Yield Response on Western Kenyan Farms Stateconditional on Western Fertilizer Kenyan and Yield Farms Response, 91(4), 991–1006.
http://doi.org/10.1111/j.l467-8276.2009.01313.x.
Marra, M., Pannell, D. J., & Abadi Ghadim, A. (2003). The economics of risk, uncertainty and
learning in the adoption of new agricultural technologies: Where are we on the learning
curve? Agricultural Systems, 75, 215–234. doi:10.1016/S0308-521X(02)00066-5.
Mauro, I. J., & McLachlan, S. M. (2008). Farmer knowledge and risk analysis: post-release
evaluation of herbicide-tolerant canola in Western Canada. Risk Analysis, 28(2), 463–476.
Mensch, B., & Kandel, D. (1988). Underreporting of substance use in a national longitudinal
youth cohort: individual and interviewer effects. The Public Opinion Quarterly, 52(1), 100–
124.

105

Muhammad, L., Mwabu, D., Mulwa, R., Mwangi, W., Langyintuo, A., & Rovere, R. La. (2010).
Characterization of Maize Producing Households in Machakos and Makueni Districts in
Kenya.
Muyanga, M., & Jayne, T. S. (2008). Private agricultural extension system in Kenya: Practice
and policy lessons. The Journal of Agricultural Education and Extension, 14(May), 111–
124. doi:10.1080/13892240802019063.
Nakasone, E., & Torero, M. (2014). Soap Operas for Female Micro Entrepreneur Training (pp.
1–49).
O’Mara, G. (1983). The microeconomics of technique adoption by smallholding Mexican
farmers. In The book of CHAC : programming studies for Mexican agriculture (pp. 250–
289). Baltimore and London: The Johns Hopkins Univ. Pr.
Omoyo, M. N., Wakhungu, J., & Oteng’i, S. (2015). Effects of climate variability on maize yield
in the arid and semi-arid lands of lower eastern Kenya. Agriculture Food Security, 41, 1–13.
Reid, R., Maag, J., & Vasa, S. (1994). Attention deficit hyperactivity disorder as a disability
category: a critique. Exceptional Children, 60(3), 198–214.
Sarin, R., & Weber, M. (1993). Risk-value models. European Journal of Operational Research,
70, 135–149.
Setimela, P. S., Magorokosho, C., Lunduka, R., Gasura, E., Makumbi, D., Tarekegne, A., …
Mwangi, W. (2017). On-farm yield gains with stress-tolerant maize in eastern and Southern
Africa. Agronomic Application of Genetic Resources, 109(2), 406–417.
http://doi.org/10.2134/agronj2015.0540
Shapira, Z. (1995). Risk Taking: A Managerial Perspective. Russell Sage Foundation.
Shiferaw, B., Kebede, T., Kassie, M., & Fisher, M. (2015). Market imperfections, access to
information and technology adoption in Uganda: Challenges of overcoming multiple
constraints. Agricultural Economics, 46, 475–488. doi:10.1111/agec.12175.
Shiferaw, B., Prasanna, B. M., Hellin, J., & Bannziger, M. (2011). Crops that feed the world 6.
Past successes and future challenges to the role played by maize in global food security.
Food Security, 3, 307–327. doi:10.1007/s12571-011-0140-5.
Siegrist, M., & Cvetkovich, G. (2000). Perception of hazards: the role of social trust and
knowledge. Risk Analysis, 20(5), 713–719.
Sitkin, S. B., & Weingart, L. R. (1995). Determinants of risky decision-making behavior: A test
of the mediating role of risk perceptions and propensity. The Academy of Management
Journal, 38(6), 1573–1592.
Slovic, S. (1987). Perception of risk. Science, 236, 280-285.
106

Slovic, P. (2000). Informing and Educating the Public about Risk. In R. E. Lofstedt (Ed.), The
perception of Risk (pp. 182–198). London: Earthscan Publications Ltd.
Slovic, P., Finucane, M. L., Peters, E., & MacGregor, D. G. (2004). Risk as analysis and risk as
feelings. Risk Analysis, 24(2). Retrieved from papers2://publication/uuid/F4B3EA59-C2F74F62-9223-701C27F152C2.
Slovic, P., Fischhoff, B., & Lichtenstein, S. (1982). Why study risk perception? Risk Analysis.
doi:10.1111/j.1539-6924.1982.tb01369.x.
Smale, M, Byerlee, D. and Jayne, T. (2011). Maize Revolutions in Sub-Saharan Africa. The
World Bank: Washington DC. doi:10.1007/978-94-007-5760-8_8.
Smale, M., Just, R. E., & Leathers, H. D. (1994). Land allocation in HYV adoption models: an
investigation of alternative explanations. American Journal of Agricultural Economics,
76(3), 535–546. http://doi.org/10.2307/1243664.
Snoj, B., Korda, A. P., & Mumel, D. (2004). The relationships among perceived quality,
perceived risk and perceived product value. Journal of Product & Brand Management, 13,
156–167. doi:10.1108/10610420410538050.
Starr, C. (1969). Social Benefit versus Technological Risk. Science, 165(1232).
Sundblad, E. L., Biel, A., & Gärling, T. (2007). Cognitive and affective risk judgements related
to climate change. Journal of Environmental Psychology, 27(2), 97–106.
http://doi.org/10.1016/j.jenvp.2007.01.003.
Suri, T. (2011). Selection and Comparative Advantage in Technology Adoption. Econometrica,
79(1), 159–209. doi:10.3982/ECTA7749.
Tambo, J. A., & Abdoulaye, T. (2013). Smallholder farmers’ perceptions of and adaptations to
climate change in the Nigerian savanna. Regional Environmental Change, 13, 375–388.
doi:10.1007/s10113-012-0351-0.
Verbeke, W., Frewer, L. J., Scholderer, J., & De Brabander, H. F. (2007). Why consumers
behave as they do with respect to food safety and risk information. Analytica Chimica Acta,
586(1-2), 2–7.
Von Neumann, J., & Morgenstern, O. (1947). Theory of Games and Economic Behavior (Third
Edit.). Princeton, NY: Princeton University Press.
Weber, E. U., Anderson, C. J., & Birnbaum, M. H. (1992). A theory of perceived risk and
attractiveness. Organizational Behavior and Human Decision Processes, 52(3), 492–523.
http://doi.org/10.1016/0749-5978(92)90030-B.

107

Weber, E. U., & Milliman, R. A. (1997). Perceived risk attitudes: relating risk perception to risky
choice. Management Science, 43, 123–144.
Wildavsky, a., & Dake, K. (1990). Theories of risk perception: Who fears what and why?
Daedalus, 119(4), 41–60. doi:10.2307/20025337.
Wyche, S., Steinfield, C., Cai, T., Simiyu, N. and Othieno, M.E., 2016. Reflecting on video:
exploring the efficacy of video for teaching device literacy in rural Kenya. In Proceedings
of the Eighth International Conference on Information and Communication Technologies
and Development (p. 8). ACM. DOI:10.1145/2909609.2909667.

108