1
Diffusion of information among small-scale farmers in Senegal: the concept
of Farmer Field Schools
Rudolf Witt, Hermann Waibel and Diemuth E. Pemsl
Research Associate, Professor, and formerly Research Associate respectively, Institute of Development and
Agricultural Economics, School of Economics and Management, University of Hannover, Germany
Abstract
Recent research on the Farmer Field School (FFS) approach in agriculture in developing
countries has raised some doubts on the economic impacts of this concept and especially the
knowledge diffusion effects from trained to non-trained farmers. Based on a study in Senegal
this paper hypothesizes that the question of the project placement strategy is vital when
analyzing knowledge diffusion effects of FFS in Africa. Results show that the share of trained
farmers in a community is a decisive factor for adoption behavior and knowledge diffusion. It
is concluded that when introducing an FFS, a critical mass of trained farmers is important to
attain effective dissemination of information and to generate positive stimuli for adoption and
learning among non-participants.
Key words: Africa, Senegal, agricultural extension, Farmer Field School, diffusion
2
1. Introduction
Knowledge is an important factor to realize productivity increases in agriculture in developing
countries. The generation and diffusion of knowledge on sustainable farming practices has
long been a problem in promoting rural development especially in Africa. . A new concept of
farmer training called the "Farmer Field School" (FFS) was developed in the 1980s by the
Food and Agriculture Organization (FAO) in Indonesia for the promotion of integrated pest
management (IPM1), and promised to be an effective tool to extend knowledge to farmers
(Pontius et al. 2002). It has been shown that FFS helps to increase farmer knowledge
(Godtland et al. 2004), and studies in several Asian countries demonstrated that FFS can be
effective in reducing the excessive use of chemical pesticides (e.g. Tripp et al. 2005; Winarto
2004; Praneetvatakul and Waibel 2005). However, the expected economic benefits are not
always unambiguously ascertainable as shown for example by a study of Feder et al. (2002)
in Indonesia. While much of the investment in FFS has taken place in Asia, more recently
FAO has introduced FFS in Africa, which some analysts have questioned from a strategic
point of view (Eicher 2003). In particular, doubts were raised regarding the expected diffusion
effects of knowledge from trained farmers to non-participants, which are essential for
achieving large-scale impact of FFS (Rola et al. 2002; Feder et al. 2004).
Empirical studies on diffusion of innovations and knowledge in agriculture show that
diffusion2 is a complex process, which depends on multidimensional, interrelated factors
(Rogers 2003; Roeling 1988; Palis et al. 2002; Fuglie and Kascak 2001). Apparently,
interpersonal networks are the predominant method by which farmers acquire knowledge
(Rola et al. 2002; Birkhaeuser et al. 1991; Tripp et al. 2005). Thus, psychosocial determinants
play an important role for the flow of information in a community. The investigation by Palis
et al. (2002) in the Philippines showed that family relations and farm neighborhood compose
homophilous3 social clusters, which offer good conditions for spontaneous diffusion of FFS
3
knowledge. This poses the question, whether knowledge transmission is determined by the
intrinsic characteristics of knowledge alone, e.g. the complexity or the abstract nature of IPM
knowledge as hypothesized by Feder et al. 2004 and Rola et al. 2002. Or whether it is rather
dependent on outside, social, conditions, i.e. the type of farmers selected for training and the
number of farmers trained. For example, the strategy of FFS placement in the context of a
national program could be an important factor for the diffusion of knowledge taught to the
participants of the FFS. When implementing a development program agricultural
administrators often try to cover large geographical areas in order to be politically visible and
to reach nation-wide impact. Thus a typical placement strategy is to introduce one FFS per
village and therefore maximize the number of “FFS villages” in a country for a given budget.
Consequently, the proportion of trained farmers in a given area is small. The alternative
strategy is to concentrate on fewer villages, which may be selected due to their history of pest
outbreaks, excessive use of pesticides or reported problems with pesticide intoxication. In this
case, the project budget would be spent to train a critical mass of farmers (in several FFS)
including follow-up training. Both approaches can be expected to have implications for the
diffusion of knowledge. In the latter strategy, due to the high visibility of FFS in a village,
trained farmers may have a stronger influence on non-participants as compared to a village
where only a few farmers attended an FFS. This influence could then result in higher adoption
through farmer-to-farmer communication. Hence the question of knowledge diffusion is
coupled with the question of project placement in the context of an overall extension strategy.
This paper investigates knowledge transfer from FFS training in Senegal, one among several
West African countries where FAO has introduced a project on FFS in 2001. It is
hypothesized that the degree and intensity of information diffusion is affected by the degree
of training exposure as measured by the proportion of farmers trained in a community. First,
the factors that can explain the likelihood and the intensity of information diffusion are
4
identified. In a second step, the effects of information diffusion on the motivation of farmers
to learn and to adopt FFS-specific knowledge are analyzed.
The paper aims to introduce a methodology that could capture the factors driving diffusion of
knowledge in agriculture in the context of an African country. Thereby it could contribute to a
better understanding of the mechanisms that determine the success or failure of the FFS
approach and provide some hints for an effective strategy of spatial placement of FFS in the
context of agricultural extension in Africa.
The paper is organized as follows: In the next section the state of diffusion theory and
research is presented. This serves as the basis for the hypotheses of the study, which are
derived in section 3. Then the methodology used for data collection and data analysis is
discussed. Section 5 presents the results of the study and finally, conclusions and implications
are discussed.
2. Diffusion theory
Diffusion is a dynamic process that focuses on the penetration of a social system by some
kind of new knowledge or an introduced technological innovation. Usually, it is defined as the
“path of aggregate adoption” (Fuglie and Kascak 2001) by a multiplicity of decision units.
According to general diffusion theory, the spread of an innovation usually follows a common
pattern (Rogers 2003). Most of the psycho-social factors that affect the pace and pattern of
diffusion are normally distributed, which leads to a Gaussian distribution of adoption behavior
(Rogers 2003). Diffusion of knowledge, innovations or technologies usually is preceded by
awareness. While at the beginning the flow of information is slow, the message is spreading
more rapidly, when a greater proportion of the population is aware of the new idea, i.e. is
5
exposed to it. If the cumulative number or percentage of adopters is plotted, the result is the
typical S-shaped diffusion path, or rate of adoption (Figure 1).
Insert Figure 1 about here
The shape of the diffusion curve depends on two factors: the rate of awareness and the
innovation-decision period (see Figure 1). The first factor is the speed of information
dissemination. A more rapid communication of information about a new idea leads to an
earlier creation of knowledge. This is portrayed as a left-shift of the rate of awareness. The
second is the time required for an individual to decide to adopt after becoming aware of the
innovation, i.e. the innovation-decision period. Supplying the individual with additional
information and decision support can shorten the time of decision-making, or more generally,
the time of forming an opinion concerning the innovation. This again leads to a left-shift of
the rate of adoption (Rogers 2003). The result of both effects is the acceleration of the
diffusion process.
Both processes - the spread of awareness and the adoption of a new technology - are
reciprocal and influence each other. At a certain point, which varies depending on the social
system or the nature of the innovation, these effects become self-sustained and the speed of
diffusion accelerates enormously without any external influence (Rogers 2003). The so called
critical mass is particularly relevant for interactive technologies (e.g. telephone, e-mail) where
the utility of the innovation increases for all adopters in the course of diffusion. Not only do
earlier adopters influence later adopters, but later adopters also influence earlier adopters in
this process of reciprocal interdependence.
Critical mass technologies show a very low variance of individual thresholds, i.e. a great share
of decision makers is willing to adopt when the critical mass has been reached. For an
innovation such as IPPM, where, due to the common property nature of pest control decisions,
6
synergetic effects occur with an increasing rate of adoption, the critical mass theory is likely
to be relevant. Early adopters still have to put up with negative externalities from their farm
neighbors relying on chemical pesticides as a major method of control. Furthermore, the
existence of markets for pesticide-reduced, pesticide free or organically produced agricultural
products is related to reaching a critical mass. With an increasing rate of adoption, early and
late adopters alike benefit in terms of pest control and better commercialization possibilities.
An important assumption for this study is the distinction between information, knowledge and
adoption. Farmers participating in FFS have the opportunity to increase their knowledge and
then may or may not apply this knowledge on their own fields. Untrained farmers in the same
village or the surrounding areas may interact with trained farmers and therefore have the
chance to access information on alternative ways of managing pests, which in turn they may
try to apply in their fields. Obviously, some of the technologies that participants learn in FFS
(e.g. identification of natural enemies of pests) are less transferable than other type of
information (e.g. choice of a resistant variety).
The distinction between different forms of information and knowledge is blurred, and drawing
a line between explicit and implicit knowledge, or between knowledge and information, is
often difficult. This is also true for adoption. Particularly in the case of a complex technology
such as IPPM, adoption is not a clear-cut decision but has to be defined carefully. Information
is defined here as "awareness" or "knowledge about something" and the reception of
information can be voluntary or accidental, through visual or oral information channels.
Knowledge is understood as the outcome of an active learning process, driven and influenced
by different factors, such as the ability to learn and the educational level of the learner, the
complexity of knowledge, the access to information, the information channel, and the intrinsic
as well as extrinsic motivation of the individual (e.g. through social pressure). It is defined as
explicit and implicit knowledge of something.4 The definition of adoption (of IPPM) is given
7
by the national coordinator of the IPPM programme in Senegal: "The two first and most
important issues of IPPM are reduction of pesticide use and the conscious and informed
management of farm inputs. The rest is a bundle of different production and pest control
techniques. Of course, rarely will the whole package be adopted, but without implementation
of the two most important issues you cannot talk of adoption of IPPM." In this study, adoption
is further specified as "partial adoption", which means that farmers apply IPPM techniques on
a part of their farm only (for different reasons), and "full adoption", which means that farmers
changed their production completely, according to IPPM principles.
It is assumed that the impact of information can be far-reaching, since it is a major driving
force of human behavior, and especially of learning processes (Bandura 1986; Deci and Ryan
1993; Wild et al. 1997). Bandura's social cognitive theory (1986) postulates that
environmental events, personal factors and behavior, all operate as interacting determinants of
each other. Thus, in the process of learning, people not only draw conclusions from their own
experiences, but are considerably influenced by extrinsic feedback. In this regard the theory of
cognitive dissonance (Festinger 1957) provides a helpful model of human behavior and
motivation. Festinger found that a main motivator of human behavior is an internal or
cognitive dissonance, which he defines as “the existence of non-fitting relations among
cognitions” (Festinger 1957). The term cognition means any information, opinion, or belief
about the environment, about oneself, or one’s behavior. Cognitive dissonance can be seen as
an antecedent condition, which leads to activity, oriented toward dissonance reduction just as
hunger leads to activity, oriented toward hunger reduction. For example, a farmer, who has
learned that chemical pesticides are extremely perilous for his and his family's health as well
as for the environment, experiences a state of psychological discomfort, which drives him
either to change his behavior (reducing the application of pesticides) or to change his
“knowledge” about the effects of pesticides.
8
3. Objectives and hypotheses of the study
In the light of this theoretical discussion, the objective of this study is to analyze how the
diffusion of information is influenced by social conditions and to estimate its effects on
learning and adoption behavior of farmers. The main hypothesis of this study is therefore as
follows: A more concentrated training effort in a village, resulting in a relatively high share
of FFS farmers (possibly above the critical mass), increases the degree and intensity of
information sharing. The more farmers are trained in IPPM through participation in farmer
field schools, the more conversations and discussions are initiated on this topic in a
community. Information is disseminated by word-of-mouth and by visual demonstration,
encouraging FFS farmers to full adoption, and motivating non-participants to learn and
possibly to adopt IPPM.
Thus, two steps can be distinguished: First, the impact of training intensity on the diffusion of
information (Hypotheses 1 and 2), and second, the impact of information on the motivation of
participants and non-participants to learn more about IPPM and to advance in the process of
adoption (Hypotheses 3 and 4).
H1:
The likelihood of information reception by non-participants increases with an
increasing share of FFS farmers.
H2:
The intensity of information sharing increases with an increasing share of FFS
farmers.
H3:
The diffusion of information affects the stage of adoption of FFS-farmers.
H4:
The diffusion of information affects the intrinsic motivation of Non-FFS farmers to
learn more of IPPM or to adopt it.
9
4. Methodology and data collection
Data collection
Data were collected in 2004 in the western part of Senegal. The approach used in this
investigation was to compare two communities with a different share of trained farmers in
IPPM. Therefore two villages had to be chosen with similar characteristics, such as ecological
conditions, infrastructure or size of village, but with different intensities of FFS training. To
choose the sample villages, seven villages of the same climatic zone, the Niayes region, were
visited, and information was gathered in group discussions with village elders and by
observation. The choice was conditional on a sufficient difference in the proportion of trained
farmers. Finally, two villages (Keur Abdou Ndoye and Gollam) at a distance of about 20 km
were selected, as they meet the established selection criteria best. In Gollam one FFS was
conducted in 2002/2003 on cabbage and onion. The 20 participants make up about 3% of
farmers in that village. In Keur Abdou Ndoye the first FFS took place in 2000/2001. From the
season of 2002/2003 on, two FFS have been conducted every year, so that at the time of the
survey, in December 2004, 105 farmers (14% of all farmers) had been trained in 5 FFS on
different vegetable crops.
Nine contiguous household clusters (carré) were randomly chosen out of the total 16 clusters
in Keur Abdou Ndoye and 20 clusters in Gollam. Within these carrés every farmer was
interviewed. The total sample size of the survey was 194 farmers in Keur Abdou Ndoye and
147 in Gollam.
Social network analysis
The way of sampling allowed for capturing the social networks in the two villages. For any
kind of relation between individuals, like kinship, friendship, advice, financial relations etc., a
specific network exists in every social system (Jansen 1999; Scott 1991). The social network
10
analysis permits to calculate certain characteristics for every farmer reflecting his or her role
in the specific relation. The relative importance of individuals within these networks is
supposed to play a role for the diffusion of information and was therefore included as an
explanatory variable.
For this study, two relationships were captured, the "respect and advice network" and the
"IPPM interaction network". The characteristics selected for the analysis are the "sociometric
status" of the "respect and advice network", the "closeness" of the "IPPM interaction
network", and the mean of the sociometric status of a Non-FFS farmer’s conversation
partners.
The centrality index of closeness reflects the reachability and power position of a person.
Farmers with high closeness are well connected within the network and can react swiftly to
changes in the network because they are closest to all other farmers. Because of this, they can
move information more quickly through a network as they will require few intermediaries to
accomplish the task. This structural advantage can be translated into power to influence
others. Closeness is defined as the inverse of the sum of the geodesic distances from one node
to all the other nodes5. The geodesic distance between two given nodes is the shortest possible
path between them. The length of a path is the number of links that comprise that path.
Closeness =
Σδ
1
g
ij
j=1
where g is the size of the network and δij is the geodesic distance from node i to node j.
Figure 2 illustrates a network of 5 farmers (or nodes). Farmer a1 has a central position within
this network and his closeness index is highest.
Insert Figure 2 about here
11
The sociometric status is the sum of its emission and reception degrees relative to the number
of all other nodes in that network.
Σx
g
Emission Degree =
ij
j=1
Σx
g
Reception Degree =
ji
j=1
1
Sociometric status = g-1
Σ (x + x )
g
ij
ji
j=1
where i is the index of the current node, xij are the link values from node i to node j, xji are the
link values from node j to node i, and g is the total number of nodes in the network.
The example in Figure 2 illustrates that farmers a1 and a4 have the same sociometric status,
since both of them have a reception and an emission degree of two, whereas the other network
members are connected to only one neighbor.
These social network characteristics have been included in the analysis in order to control for
differences in the social status of farmers. Information may be disseminated to a higher extent
from/to farmers who are more central in the village network, or who have a higher social
status. Further, the transmission of information may have different impacts on the receiver,
depending on the standing of the sender.
Model specification
Following the hypotheses, the analysis is conducted in two steps, first, examining the
diffusion of information, and second, its effects on the motivation of farmers.
The 21 variables used in the analysis can be summarized in five groups (see Table 1):
12
•
Social Network Characteristics (SNi)
•
Demographic Characteristics (Di)
•
Intensity of Individual Exposure to IPPM (IEi)
•
Likelihood and Intensity of Information Sharing (LIS, IISi)
•
Adoption of IPPM by FFS (AFFS,i) /Assessment of IPPM by Non-FFS Farmers (ANFFS,i)
In the first step, the diffusion of information was analyzed by use of three models. Model (1)
hypothesizes the likelihood (LIS) to depend on the intensity of FFS training as captured by the
'village' dummy variable, while for models (2) and (3) the individual exposure to IPPM (IEi)
was explicitly included. Social network characteristics (SNi) and demographic attributes (Di)
are taken into the regression model to control for other influences on the flow of information.
A binary logistic model was used to assess the likelihood of information reception while to
explain intensity of information sharing OLS regression was used.
ln(LISk) = β 0 + β SN ,i SN ik + β D ,i Dik + u k
(1)
IIS1 (= number of contacts) = β 0 + β SN ,i SN ik + β D ,i Dik + β IE ,i IEik + u k
(2)
IIS2 (= communication frequency) = β 0 + β SN ,i SN ik + β D ,i Dik + β IE ,i IEik + u k
(3)
where LISk =
p ( y k = 1)
, β 0 , β SN ,i , β D ,i , β IE ,i are the regression coefficients and u k is the
1 − p ( y k = 1)
error term.
In the second step of the analysis, the impact of information on adoption and learning
motivation of farmers was assessed. Therefore, the dependent variables of (2) and (3) were
taken as explanatory variables for the following regression estimations. The dependent
13
variables in models (4) to (6) were AFFS and ANFFS,i. AFFS, the stage of adoption by FFS
farmers is a variable with three possible values (0 = no adoption, 1 = partial adoption, 2 = full
adoption), which required a multinominal logistic model. The assessment of IPPM by NonFFS farmers, which was assumed to depend on the intensity of information about IPPM, was
captured by two variables: a personal evaluation of IPPM (on a scale from 0 to 4) in
model (5), and the wish to adopt IPPM in model (6). Again, OLS, binary and multinominal
logistic models were used.
ln(AFFS) = β 0 + β SN ,i SN ik + β D ,i Dik + β IIS ,i IIS ik + u k
(4)
ANFFS(personal evaluation) = β 0 + β SN ,i SN ik + β D ,i Dik + β IIS ,i IIS ik + u k
(5)
ln(ANFFS(wish to adopt)) = β 0 + β SN ,i SN ik + β D ,i Dik + β IIS ,i IIS ik + u k
(6)
where ln(AFFS,partial) =
p ( y k = 1)
p ( y k = 2)
and ln(AFFS,full) =
, β 0 , β SN ,i , β D ,i , β IE ,i are the
p ( y k = 0)
p ( y k = 0)
regression coefficients and u k is the error term.
5. Results
Table 1 displays all relevant variables for both villages. For most of the variables the
differences between Gollam and Keur Abdou Ndoye are highly significant.
The mean values of the individual social network characteristics display a considerable
dissimilarity of the two villages. The communication network for IPPM information is much
denser and better established in Keur Abdou Ndoye (the village with the higher FFS training
intensity) than in Gollam. In the latter it is star-like, i.e. there are only very few sources of
information on IPPM. This also explains the reverse distribution of the mean sociometric
14
status of conversation partners. All three variables are expected to have a positive effect on
the diffusion of information, as well as on the motivation of farmers to adopt IPPM.
Insert Table 1 about here
As for the demographic variables, it is noticeable that land tenure is more equally distributed
in Keur Abdou Ndoye. More than 55% of respondents were proprietors of the land they
farmed, while in Gollam only one third of farmers were landowners.
Following the hypothesis of the study the share of exposed farmers is significantly higher in
Keur Abdou Ndoye. To what extent this is attributable to the intensity of FFS training in the
village, will be investigated in the regression analysis. Table 1 shows that a higher share of
FFS farmers automatically transfers into a much higher individual exposure to IPPM by NonFFS farmers.
A positive correlation exists between FFS intensity and a more intense sharing of information.
The number of contacts as well as the communication frequency is almost twice as high for
Keur Abdou Ndoye.
The "stage of adoption" of FFS participants and the "wish to adopt" of the non-participants do
also significantly differ between the two villages, which is likewise reflected in the personal
assessment of IPPM by all farmers.
Hence, the conditions for a successful diffusion of FFS knowledge seem to be more favorable
in Keur Abdou Ndoye as compared to Gollam. It can be hypothesized that five FFS and the
continued support by the national coordinator and several dedicated farmer facilitators over
4 years have had a deeper impact on diffusion and adoption than only one FFS two years ago.
This impact is analyzed in detail in the following sections.
15
Information sharing
The models (1) to (3) concern the diffusion of information (see Table 2). Model (1) explores
the likelihood that a Non-FFS farmer receives any information about IPPM. The second and
third models deal with the intensity of information sharing.
As hypothesized, demographic characteristics play a role. Property status (D6) and right of
decision (D7) both have a significant positive effect. Thus, landowners are almost 5 times
more likely to receive information on IPPM than landless farmers (Exp (Bj) = 4.9). The size
of carré (D4), although statistically significant, has a rather negligible impact. The sign of the
gender coefficient (D1) is positive for all dependent variables, which means that women are
more likely to receive information than men, and that they interact more intensely with their
neighbors. Results for the effect of education (D3) are conflicting. It is positive for the
likelihood of information reception, but it is negative for the frequency of communication.
This could imply that educated farmers (i.e. younger farmers6) have a higher chance to be
aware of a new innovation like IPPM, but the intensity of interaction is higher for noneducated farmers. The effect of age (D2) is insignificant.
Quite affirming results are received for the variables on intensity of exposure (IE2-IE5),
captured by the number of FFS family members (IE2), friends and neighbors (IE3), the
observation of a change (IE4) and the time since first exposure (IE5). These four variables are
central to the investigation and were expected to have a significant positive impact on the
diffusion of information. The findings confirm this hypothesis. A Non-FFS farmer having
more relatives or friends, who underwent FFS training, shows a higher communication
frequency and the number of contacts with FFS farmers increases. Particularly the fact that a
farmer observed FFS participants applying the newly acquired knowledge on their farms (and
getting positive results) positively affects the intensity of information sharing.
16
The number of FFS kinsmen or friends is not the only important variable. In addition, the
social status of those FFS farmers (SN3) considerably increases the intensity of information
exchange. This result is in accordance with the cognitive dissonance theory that suggests that
the magnitude of dissonance can be decidedly influenced by public opinion. The influence of
another person’s or the public’s view on a farmer’s behavior is rising with the expertise and
the attractiveness or social status of the person voicing the opinion. In this model, this
“attractiveness” is specified by the mean sociometric status of conversation partners.
The "closeness" of Non-FFS farmers in the network (SN2) also seems to promote the diffusion
of information by increasing the communication frequency. This means, the more central a
farmer is placed within the communicational network, the lower are his “transaction costs” in
terms of time and effort, and the better is his access to information.
Insert Table 2 about here
An important outcome of the model is the coefficient of the dummy "village" variable (0.01
level of significance). This is in line with the hypothesis, because the share of trained farmers
is implicitly included in the "village" variable. The odds ratio shows that a farmer in Keur
Abdou Ndoye is over 5 times more likely to receive information on IPPM than a farmer in
Gollam. This suggests that a higher FFS intensity can stimulate diffusion.
Adoption and motivation
Regression results, however, do not straightforwardly support the hypotheses concerning
adoption behavior of farmers. Communication frequency and the number of contacts were
expected to have a positive impact on adoption by FFS- and the motivation to adopt by NonFFS farmers. As far as the first effect is concerned, the coefficients suggest that the intensity
of information (IIS1 and IIS2) indeed increases the chance of a partial adoption by a factor of
3.3, and the chance of full adoption by 2.2, if the number of contacts (IIS1) rises by 1. Also the
17
frequency of communication (IIS2) has a strong influence on full adoption. Only 5
conversations more per month increase the chance of full adoption by 2.1 (which is Exp(Bj)
to the power of five).
No significant effect was found related to the motivation of Non-FFS farmers as measured by
the stated "assessment of IPPM" as well as "wish to adopt". The secondary effect of the share
of trained farmers (through fostering of better information flow) seems to be non-existent.
This suggests that there are other factors that shape the motivation and the learning behavior
of Non-FFS farmers.
The observation of a change in the agricultural practice of FFS farmers emerges as the most
important factor for the motivation to adopt. The odds ratio is 240.7, which is defined
as:
p( ANFFS , 2 = 1)
p( ANFFS , 2 = 0)
. Hence, a farmer is much more willing to acquire FFS knowledge and to
adopt IPPM, if he had actually seen the impact of IPPM in practice. In addition, the
assessment of IPPM increases by about two points on the scale7, which means a considerable
gain in reputation. Thus, whether a farmer develops an interest and is eager to learn and to
apply FFS knowledge is mainly attributable to the image and the “perceived performance” of
the FFS farmers in the village. On the other hand, “perceived performance” is also dependent
on the intensity of information diffusion (Table 3). Thus, there is an indirect effect of
information sharing on the learning and adoption motivation of non-participants. In summary,
the results suggest that training of a higher share of farmers per village results in faster
diffusion of information about IPPM. The speed of diffusion affects the stage of adoption and
the performance of FFS farmers, which in turn implies a more convincing demonstration of
IPPM to Non-FFS farmers.
The goodness-of-fit statistics show high explanatory power of the models, even though only
few of the variables are significant. Thus, information sharing in connection with visible
18
application of knowledge are identified as the most important factors to explain adoption
behavior under the conditions found in the study area.
Insert Table 3 about here
It shall be added that for full adoption on the whole farm there are some village-specific
characteristics which were not controlled for by the model that could have further explained
diffusion and adoption. As shown by the odds ratio of the village effect in table 3, farmers in
Keur Abdou Ndoye are by far more likely to adopt IPPM than in Gollam. Casual observations
during the survey revealed that the FFS alumni network is vibrant in KAN. Farmers supported
by the FFS facilitators are meeting regularly once a week in different groups to discuss crop
management and other problems which may well help to facilitate their decision making. The
facilitators are supporting and supervising their “students”, always promoting the cause of
IPPM. To the contrary in Gollam, FFS farmers experience practically no support. With only
20 trained farmers there is little moral assistance and encouragement, which has led to a
dormancy state of IPPM in that village.
6. Conclusions
The findings of the study suggest that the current strategy of implementing one Farmer Field
School with about 25 farmers per community or village and at the same time aiming for a
maximum number of locations per country or region should be reconsidered. The oftenpolitical pressure for rapid program up-scaling, in order to achieve widespread impact may be
counter-productive to and incompatible with the genuine strengths of the FFS concept. There
exists a trade-off between achieving a rather widespread placement of FFS in a country and
the impact of the training in each FFS location. Small proportions of trained farmers in a
village might be insufficient to induce change beyond the participants of the training. This
19
study submits that a critical mass of trained farmers is needed in order to attain effective
dissemination of information and positive stimuli for adoption and learning among nonparticipants. To achieve a significant demonstration effect of FFS within the community and
thus stimulate demand for more information a high-quality training is important. It is
reasonable to assume that training quality will be higher if a more concentrated and a longerterm strategy of program placement is used. The results also lend some support to the
conclusion that the strengths of FFS projects lie in their use as an intervention in special
situations, concentrating efforts and resources on selected sites rather than using it as a
substitute for a national extension strategy of introducing participatory methods. These results
are also in line with the conclusions of a study by Fleischer et al. (2002 and 2004) who found
based on comparison of cost-effectiveness of projects in Egypt, that public investments in
participatory agricultural extension can be economically justified if the targets are well
chosen.
In addition, clustering of FFS can also have other benefits like reducing the negative
externalities of pesticide spraying and improve the general state of the ecosystem and
biodiversity. Besides, an agglomeration of FFS farmers may facilitate the formation of local
markets for pesticide-reduced, pesticide free or even organic products and the
commercialization of bio-pesticides, and may lower the costs of introducing other institutional
and technical innovations.
Hence, concentrating development efforts like FFS in well-defined target areas could be an
effective tool of poverty reduction through sustainable rural development.
Further studies should assess the productivity impact of such FFS strategy and include the
program and farm level costs in the context of a cost benefit analysis.
20
Notes
1. The introduction of IPM FFS in Africa has shown that there are broader agronomic,
management and production issues that have to be addressed by the facilitators. This has
led to the use of the term IPPM (integrated production and pest management) instead of
IPM. Hence, the term IPPM is used in the succeeding text.
2. The terms diffusion, dissemination, information sharing, or transmission of information,
are used as synonyms in this paper.
3. Homophily is the degree to which a pair of individuals is similar. The similarity may be in
certain attributes, such as beliefs, education or social status. According to Rogers (2003)
communication is more effective if individuals have much in common.
4. Explicit knowledge is formal knowledge that can be made explicit by means of a verbal
statement and that can be recorded. Implicit knowledge is personal knowledge, rooted in
individual experience and involving personal belief, perspective, and values.
5. "Node" is a technical term of Social Network Analysis, which in this case simply means
"individual" or "farmer".
6. In both villages formal school education only exists for the last 10 to 15 years, which
means that there is a relatively strong negative correlation between age and education.
7. The scale was defined as: 0 = IPPM is no solution at all; 1 = IPPM is worthwhile trying;
2 = It is as good as traditional farmer practice; 3 = IPPM is much better than farmer
practice; 4 = every farmer should practice IPPM!
21
References
ABRAHAMSON, E. (1991): “Managerial fads and fashions: the diffusion and rejection of innovations”. Academy
of Management Review – Volume 16, pp.586-612
ABRAHAMSON, E. (1996): “Managerial fashion”. Academy of Management Review – Volume 21, pp.254-285
BANDURA, A. (1986): “Social Foundations of Thought and Action – A Social Cognitive Theory” – New Jersey,
Prentice-Hall Series in Social Learning Theory, 1986.
BIRKHAEUSER D., R.E. EVENSON and G. FEDER (1991): "The Economic Impact of Agricultural Extension: A
Review." Economic Development and Cultural Change - Volume 39, no. 3, pp.607-650
DECI, E. L. and R. M. RYAN (1993): "Die Selbstbestimmungstheorie der Motivation und ihre Bedeutung für die
Pädagogik." Zeitschrift für Pädagogik - Volume 39, pp. 223-238.
EICHER, C. (2003): "Flashback: Fifty years of donor aid to African agriculture." Paper presented at an
International Policy Conference “Successes in African Agriculture: Building for the Future”, sponsored
by InWent, IFPRI, NEPAD and CTA, Pretoria, South Africa, December 1-3, 2003, pp.1-53
FEDER, G., R. MURGAI and J.B. QUIZON (2002): “Sending Farmers Back to School: The Impact of Farmer Field
Schools in Indonesia.” Review of Agricultural Economics – Volume 26, no. 1, pp.1-18
FEDER, G., R. MURGAI and J.B. QUIZON (2004): “The Acquisition and Diffusion of Knowledge: The Case of
Pest Management Training in Farmer Field Schools, Indonesia.” Journal of Agricultural Economics –
Volume 55, no. 2, pp.221-243
FESTINGER, L. (1957): “A Theory of Cognitive Dissonance” – Stanford, Stanford University Press, 1957
FLEISCHER, G, H. WAIBEL and G. WALTER-ECHOLS (2002): “Transforming top-down agricultural extension to a
participatory system: A study of costs and prospective benefits in Egypt.” Public Administration and
Development - Volume 22, pp.309-322.
FLEISCHER, G, H. WAIBEL and G. WALTER-ECHOLS (2004): "Egypt: How much does it cost to introduce
participatory extension approaches in public extension services?" - The World Bank, Agriculture and
Rural Development Discussion Paper 10, - Volume 3, pp.40-48
FUGLIE, K.O. and C.A. KASCAK (2001): “Adoption and Diffusion of Natural-Resource-Conserving Agricultural
Technology.” Review of Agricultural Economics – Volume 23, no. 2, pp.386-403
GODTLAND, E., E. SADOULET, A. DE JANVRY, R. MURGAI and O. ORTIZ (2004): “The impact of Farmer-FieldSchools on knowledge and productivity: a study of potato farmers in the Peruvian Andes.” Economic
Development and Cultural Change – Volume 53, no.1, pp.63-92.
JANSEN, D. (1999): „Einführung in die Netzwerkanalyse – Grundlagen, Methoden, Anwendungen“ - Opladen
Leske+Budrich, 1999
NEWELL, S., J.A. SWAN and R.D. GALLIERS (2000): “A knowledge-focused perspective on the diffusion and
adoption of complex information technologies: the BPR example”. Information Systems 10, pp.239-259
PALIS, F.G., S. MORIN and M. HOSSAIN (2002): "Social Capital and Diffusion of Integrated Pest Management
Technology: A Case Study in Central Luzon, Philippines.“ Paper presented at the Social Research
Conference, CIAT, Cali, Columbia, September 11-14, 2002. pp.1-14
PONTIUS, J., R. DILTS and A. BARTLETT (eds.) (2002): “Ten Years of IPM Training in Asia - From Farmer Field
School to Community IPM” – Bangkok, FAO, pp.1-110
PRANEETVATAKUL, S. and H. WAIBEL (2005): “Impact Assessment of Farmer Field School using A Multi Period
Panel Data Model” working paper. Submitted for presentation at the International Agriculture and
Economics Conference in Brisbane, 2006, pp.1-14
QUADDUS, M. and J. XU (2005): “Adoption and Diffusion of Knowledge Management Systems: Field Studies of
Factors and Variables”. Knowledge-Based Systems 18, pp.107-115
ROGERS, E. M. (2003): “Diffusion of Innovations.” 5th Edition. - New York, The Free Press, 2003
ROELING, N. (1988): "Extension Science – Information Systems in Agricultural Development“ – Cambridge,
Cambridge University Press, 1988
22
ROLA, A.C., S.B. JAMIAS, and J.B. QUIZON (2002): “Do Farmer Field School Graduates Retain and Share What
They Learn?: An Inverstigation in Iloilo, Philippines.” Journal of International Agricultural and
Extension Education – Volume 9 (Spring 2002), pp.65-76
SCOTT, J. (1991): “Social Network Analysis – A Handbook” – London, SAGE Publications, 1991
TRIPP, R., M. WIJERATNE and V.H. PIYADASA (2005): "What should we expect from farmer field schools? A Sri
Lanka case study" World Development - Volume 33, no. 10 , pp.1705-1720
WILD, K.-P., A. KRAPP, D. LEWALTER and I. SCHREYER (1997): "Der Einfluß berufsbezogener Interessen und
kognitiver Kompetenzen auf den Lernerfolg in der beruflichen Erstausbildung. Eine zweijährige
Längsschnittstudie" Gelbe Reihe: Arbeiten zur Empirischen Pädagogik und Pädagogischen Psychologie,
Volume. 39, pp.1-35
WINARTO, Y.T. (2004): "Seeds of knowledge - The beginning of integrated pest management in Java" New
Haven, CT: Yale Southeast Asia Studies, 2004
23
farmers [%]
rate of awareness
innovationdecision
period
non-exposed
farmers
exposed
farmers
rate of adoption
(= rate of knowledge)
adopters
time
Figure 1: The rate of adoption, the rate of awareness and the innovation-decision period
Source: adapted from Rogers (2003)
24
Figure 2: Schematic illustration of a social network.
Source: own illustration
25
Table 1: Descriptive statistics of variables used in the regression analysis
Variable
SN1
Adoption /
Intensity of
Assessment of information
IPPM
sharing
Individual
exposure to IPPM
Demographic
characteristics
Category
Social Network
characteristics
Gollam
1
Keur Abdou Ndoye
Significance1
FFS
Non-FFS
Total
FFS
Non-FFS
Total
Description
n = 14
n = 133
n = 147
n = 58
n = 127
n = 185
Mean Sociometric Status
(Respect and Advice Network)
0.0268
0.0362
0.0353
0.0132
0.0095
0.0106
0.0027
0.0025
0.0025
0.0615
0.0457
0.0507
-
0.0265
-
-
0.0173
-
42.86
66.17
63.95
44.83
47.24
46.49
***
ns
D1
Mean Closeness
(IPPM Interaction Network)
Mean Sociometric Status
of Conversation Partners
Gender [% women]
D2
Age [years]
35.36
35.49
35.48
32.24
31.54
31.76
**
SN2
SN3
***
***
D3
Educated Farmers [%]
28.57
29.32
29.25
27.59
33.07
31.35
ns
D4
Size of Carré [Number of Inhabitants]
13.86
14.74
14.66
19.62
17.73
18.32
***
D5
Size of Farm [ha]
2.63
2.30
2.33
2.06
1.58
1.73
***
D6
Landowners [%]
57.14
30.83
33.33
44.83
60.63
55.68
***
D7
Right of Decision [% of landless farmers]
92.86
87.97
88.44
86.21
86.61
86.49
ns
IE1
Exposed Farmers [%]
-
61.65
-
-
90.55
-
***
IE2
FFS Family Members [number]
0.00
0.74
0.67
1.16
1.68
1.52
***
IE3
FFS Friends and Neighbors [number]
0.00
2.44
2.20
3.14
4.60
4.14
***
IE4
Non-FFS Farmers having observed
a change in the practice of FFS Farmers [%]
-
59.4
-
-
85.83
-
IE5
Time since first IPPM exposure [months]
2.50
14.80
13.63
11.86
25.98
21.56
IIS1
Number of Contacts [number]
1.71
0.88
0.96
2.33
1.50
1.76
IIS2
Communication Frequency [talks per month]
4.93
2.51
2.74
6.22
4.60
5.11
0.54
-
-
0.64
-
-
3.07
1.50
1.65
3.60
2.83
3.08
-
90.25
-
-
100.00
-
AFFS
ANFFS1
ANFFS2
Stage of Adoption
[mean value of a scale 0 to 2]
Assessment of IPPM
[mean value of a scale 0 to 4]
Wish to Adopt [% of exposed farmers]
*, **, *** : significant difference between Gollam and Keur Abdou Ndoye at 0.1, 0.05 and 0.01 level of significance, respectively, based on t-test
ns
***
***
***
***
***
**
26
Table 2: Results of the first step of analysis: the likelihood and intensity of information diffusion
Intensity of information sharing
Model (1)
Exposure to information (IE1)
Model (2)
Number of
contacts
(IIS1)
Model (3)
Communication
frequency
(IIS2)
binary logistic
OLS
OLS
non-standardized
coefficient Bj
non-standardized
coefficient Bj
regression coefficient Bj
odds ratio
Exp (Bj)
Variable
Intercept
-5.628***
-0.701*
-0.836
Village
SN1
SN2
SN3
D1
1.692***
-2.346
0.306**
0.019
-0.404
9.444***
0.041
0.276
-0.178
4.873**
19.041*
0.820
0.879**
2.409
D2
D3
D4
D5
D6
D7
IE2
IE3
IE4
IE5
0.009
0.726**
0.062**
0.058
1.589***
0.830*
1.009
2.068
1.064
1.060
4.899
2.293
0.010**
0.097
-0.001
-0.036
-0.048
-0.014
0.164***
0.035**
0.688***
0.001
-0.044*
-0.887***
0.024
0.149
1.560**
-0.168
-0.289
0.346***
1.859*
0.048*
R2
0.370
0.508
0.311
N
197
197
197
5.428
0.096
* 0.1 level of significance, ** 0.05 level of significance, *** 0.01 level of significance
27
Table 3: Results of the second step of analysis: the stage of adoption and the motivation of NonFFS farmers to adopt
Motivation of Non-FFS farmers
Model (4)
Stage of adoption of FFS farmers (AFFS)
Model (5)
Assessment of
IPPM (ANFFS,1)
Model (6)
Wish to to adopt
(ANFFS,2)
multinominal logistic
OLS
binary logistic
non-standardized
coefficient Bj
Bj (Exp(Bj))
partial adoption
Bj (Exp(Bj))
full adoption on the
whole farm
Bj (Exp(Bj))
-3.091**
-8.798***
0.020
-4.916**
Village
0.762 (2.143)
3.149*** (23.316)
0.091
0.925 (2.522)
SN1
-1.343 (0.261)
0.599 (1.821)
-0.028
-10.008 (0.000)
SN2
0.528 (1.697)
1.168 (3.216)
0.060
-1,149 (0.316)
0.512
-3.208 (0.040)
Variable
Intercept
SN3
D1
0.191 (1.210)
0,174 (1.190)
0.009
0.185 (1.202)
D2
-0.006 (0.993)
-0.015 (0.985)
-0.004
0.001 (1.001)
D3
0.154 (1.166)
0.031 (1.031)
-0.047
-0.191 (0.826)
D4
0.006 (1.006)
-0.007 (0.993)
0.001
0.0191 (1.019)
D5
-0.045 (0.956)
0.241 (1.273)
-0.012
-0.161 (0.850)
D6
-0.507 (0.602)
-0.004 (0.996)
0.068
0.910 (2.485)
D7
-0.268 (0.765)
0.064 (1.065)
0.069
-0.124 (0.883)
IE2
-0.365 (0.694)
0.020 (1.020)
0.022
0.159 (1.172)
IE3
-0.192* (0.824)
-0.028 (0.972)
0.007
-0.003 (0.996)
2.161***
5.483*** (240.747)
IE4
IE5
-0.074*** (0.928)
-0.120*** (0.887)
0.007**
0.058** (1.059)
IIS1
1.207*** (3.345)
0.787*** (2.196)
-0.037
-0.329 (0.719)
IIS2
0.037 (1.037)
0.149*** (1.160)
0.003
-0.006 (0.993)
R2
0.640
0.839
0.849
N
72
197
197
Note: Odds ratio Exp(Bj) is given in parentheses
* 0.1 level of significance, ** 0.05 level of significance, *** 0.01 level of significance