Documente Academic
Documente Profesional
Documente Cultură
Jonathan Wadsworth*
Escuela Centroamericana de Ganaderia, Apartado No. 7, Atenas, Costa Rica
ABSTRACT
The development of eficient agricultural extension strategies is dependent
on the accurate prediction of farmer response to diverse extension messages
and methods of delivery. Most research on adoption behaviour has con-
centrated on farmer characteristics as the main determinants of adoption
rates. Thispaperfocuses attention on extension agency activities (EAA) as
an equally important variable.
Ex-ante studies of three farmer populations predicted the existence of
two distinct adoption scenarios: namely, managerial capacity (MC)
dependent, where farmer MC plays an important role in adoption and MC
independent, where the level of farmer MC does not appear to be involved
in the adoption process. An experiment involving 108 farmers divided
amongst four extension treatments was carried out to test the ex-ante
hypotheses. Of the nine system x technology combinations tested,Jive were
correctly predicted from the ex-ante model. This paper details reasons for
the models failure to predict accurately the four remaining cases. It is
concluded that not only farmer MC and EAA, but also variables associated
with economic climate, attributes of innovations, existence of structural
constraints and appropriateness of technologies should be incorporated into
the model in order to improve the predictability of adoption behaviour
based on ex-ante studies.
INTRODUCTION
* Present a.ddress: Kenyan Agricultural Research Institute, PO Box 7449 1, Nairobi, Kenya.
69
70 J. Waakworth
Fig. 1. Components of extension agency activity (EAA) likely to affect the flow of new
technology into farm systems.
Adoption of innovations by livestock producers 71
This paper seeks to validate and further improve the model of tech-
nology flow by critically examining the part played by extension agency
activities (EAA) on changes in farmers knowledge, evaluation and prac-
tice of innovations by means of a direct experimental approach. A field
experiment was conducted which permitted the isolation of effects due
to EAA, MC and their interactions. The four main points at which EAA
is considered able to affect the flow of new technology are illustrated in
Fig. 1. This in effect presents the mirror image of the conceptual model
tested by Wadsworth (1993) which concentrated on components of farmer
MC which affect technology flow.
The experiment took place in Costa Rica during 1987/88 and involved the
participation of 108 collaborating livestock producers representing three
different systems of production (beef, dual-purpose and intensive dairying)
who were allocated to four different extension treatments. The experiment
followed a 2-year period of surveys, farm monitoring by monthly visits
and data analysis in order to determine the appropriate extension mes-
sages and allocate MC values to each farmer. This preliminary work is
documented in greater detail by Wadsworth (1990a,b).
Treatments
?? Treatment D: The final and most intensive level of EAA was treatment
D which incorporated an element of practical skills training. At 1
month after the follow-up visit aimed at increasing levels of evaluation a
training day was held on the farm of one of the treatment D partici-
pants. All treatment D producers were invited and elaborate prepara-
tion including multiple reminders, free transport and refreshments was
undertaken to ensure high attendance rates. This was essential to
maintain the experimental design since any treatment D individual not
receiving practical training would thereafter become a member of
treatment C by default with respect to that particular innovation. The
practical skills training sessions were planned to last approximately 5 h
in total and consisted of a short introductory talk followed by practical
training in smaller groups. Not only practical skills but also a revision
of knowledge and further development of positive attitudes inevitably
formed part of each session.
The four treatments may be summarised as follows:
Treatment A = Control
Treatment B = Knowledge
Treatment C = Knowledge + Evaluation
Treatment D = Knowledge + Evaluation + Training
bulletins, training the enumerators and devising the practical skills ses-
sions with a view to orientating each extension message to farmers needs.
None of the participating farmers had been involved in organised
extension programmes during the 5 years prior to the start of the experi-
ment. However some previous sporadic contacts with extension agents
were reported by some producers (Wadsworth, 1990b). This could be
expected to influence farmers attitudes to the field staff and the extension
messages during the experiment. However the extension experiment was
preceded by a full year of data collection with monthly visits to each farm
during which time farmer confidence and rapport with the field staff
developed to excellent levels, Thus equalising starting conditions with
respect to familiarity and farmer interaction with official institutions.
and also received additional explanatory and training visits from the enu-
merators regarding each technology as did treatment B and C producers.
This activity was considered the only responsible way of repaying the
unwitting collaboration in an investigation where, if given the choice,
most producers would have elected to be subjected to treatment D.
Ex-post data were collected between July and September 1988 with one
technology being considered at each monthly visit. The recording sche-
dules used were identical to those originally completed in the ex-ante
investigation with the addition of two open questions put to treatment B,
C and D farmers only. These asked for participants general opinions
regarding the technical recommendations given during the programme
and the reasons, if any for not having adopted some or all components of
the innovation in question.
Ex-post indices were calculated in the same way as ex-ante indices
(Wadsworth, 1993) however, points were awarded where farmers expres-
sed a definite intention to modify farm practice at some point in the near
future and where reasons for not having adopted already were not attri-
butable to disinterest. For example, some producers had not previously
used sugar cane for cattle feeding but had planted it at the start of the
rainy season in 1988 in anticipation of feeding it in the dry season of 1988/
89. Hence, although cattle were not actually fed sugar cane during the dry
period of 1987/88, the farmers intention indicated adoption of the tech-
nology even though the actual implementation could only occur after the
forage had grown.
Statistical methods
Due to the non-orthogonality of the data resulting from the few farmers
who abandoned the programme before starting the extension experiment
and the uncontrollable migration of producers from treatment D to C the
final data were not as tidily balanced as had been planned.
The effects of covariate, MC and ET on changes in knowledge, evalua-
tion and practice of each innovation were investigated using multiple
regression and analysis of variance. A sequence of multiple regression
models were derived, each one designed to test the significance of at least
one variable (Draper and Smith, 1966). The order in which the variables are
introduced into such models is critical to the interpretation of each ana-
lysis. For example where the covariate was found to be statistically significant
this variable was obliged to be included first in all subsequent models in
order to test the effect of the other variables after having accounted for
the influence of the covariate on the dependent variable. As a general
guide variance ratios may only safely be declared to be statistically sig-
nificant when the possible confounding effect of all other variables has
already been accounted for in the regression. In this way MC could
only be declared significant when preceded by ET in the analysis of
variance and vice versa. Notwithstanding the above guide, this techni-
que is largely interactive and requires that data be analysed intelligently
with each step being derived from interpretation of the previous one.
The mathematical procedures were carried out using the statistical
programme GENSTAT (Lane et al., 1987).
In order to further explain the differences identified by the analyses of
variance, tables of treatment means were constructed and standard errors
of differences (SEDs) calculated. Where the covariate had been found to
be statistically significant all means were adjusted for covariance in order
to provide a common basis for comparison.
In some cases the SEDs are larger than their corresponding treatment
means (Tables 3,6 and 9) indicating high variability in the data. Although
ex-ante index data were normally distributed the effect of the EAA did not
follow a continuous pattern in these cases producing large and discrete
increases in ex-post treatment means. Hence changes in index values (ex-
post minus ex-ante) were affected in the same way. Presentation of the
data is further complicated by the occurrence of negative values of the
change in index scores which logically shift the arithmetic means towards
zero but do not affect the values of their corresponding SEDs. Usage of
the selected tests is not compromised by this.
As already mentioned the final distribution of participating farmers
between the ETs and the level of the MC was unbalanced; this meant that
different SEDs had to be derived for every pair of means to be compared.
In order to avoid the inclusion of large matrices of SEDs the mean SEDs
Adoption of innovations by livestock producers 77
TABLE 1
Comparison of Predicted versus Observed Adoption Behaviour Exhibited by Costa Rican
Livestock Producers
(b) Evaluation
2 (c) Practice
1
-c
A B C D
Extension treatment
Fig. 2. Change in index values of mineral supplementation technology by low (n), med-
) and high (m) MC beef farmers following exposure to different extension treatments.
Adoption of innovations by livestock producers 79
TABLE 2
Level of Probability of Variance Ratios Derived from Analysis of Variance of the Change
in Index Values of Beef Producers with Respect to Mineral Supplementation
u Ex-ante values.
TABLE 3
Mean Change in Index Values by Level of Managerial Capacity and Extension Treatment
Beef Zone Mineral Supplementation
Change in indices was derived by subtraction of ex-ante from ex-post index scores.
b Standard error of difference between any two means reported in the table are means of
individual standard errors of differences between all relevant pairs of means.
Where ex-ante indices were found to have a significant effect on ex-post values (see Table 2)
means were adjusted to account for covariance. Unadjusted means as originally recorded
have been used for cases where the covariate was not statistically significant.
80 J. Wadsworth
TABLE 4
Reasons Given for Non-adoption of Mineral Supplementation Technology by Beef
Producers
Number
Treatments B, C and D farmers were asked why they did not use at least some of the
recommended practices contained in the extension message.
b Farmers known to employ at least some of the recommended practices are referred to as
adopters although this covers a wide range of technology use.
(Non-adopters refers to producers who stated that they had not changed the farm prac-
tice of the technology in question over the experimental period and do not employ sig-
nificant features of the recommended practices.
Farmers described themselves as being lethargic (dejado) in not having adopted the
technology. This expression implies that they subscribe to the potential benefits but have
not yet done anything about it. (It clearly indicates that the farmer is not sufficiently con-
vinced to take action or is unable to do so for a reason he is not prepared to give.)
financial benefits of which will not be realised until at least 18 months into
the future, tends to be given low priority by farmers whose chief concerns
are more short term such as payment of labour, vaccines, essential fencing,
etc. Due to this widespread depression of the beef sector it is not unrea-
sonable to consider the cost of inputs as a structural constraint outside the
direct control of the majority of producers and therefore to acknowledge
that extension will be largely unsuccessful in affecting farm level practice
of such technologies under these circumstances.
TABLE 5
Level of Probability of Variance Ratios Derived from Analysis of Variance of Change in
Index Values of Dual-Purpose Producers with Respect to Mineral Supplementation
a Ex-ante values.
Adoption of innovations by livestock producers 83
TABLE 6
Mean Change in Index Values by Level of Managerial Capacity and Extension Treatmenta
Dual-purpose Zone, Mineral Supplementation
7i (c) Practice
6
5
4
3
2
1
-I
A B C D
Extension treatment
Fig. 3. Change in index values of mineral supplementation technology by low (Cl), med-
) and high (m) MC dual-purpose farmers following exposure to different extension
treatments.
are more short term such as payment of labour, vaccines, essential fencing,
etc. Due to this widespread depression of the beef sector it is not unrea-
sonable to consider the cost of inputs as a structural constraint outside the
direct control of the majority of producers and therefore to acknowledge
that extension will be largely unsuccessful in affecting farm level practice
of such technologies under these circumstances.
TABLE 7
Reasons Given for Non-adoption of Mineral Supplementation Technology by Dual-
Purpose Producers
Number %
the experiment only high MC producers were able to take full advantage
of the highest intensity ET. Such an observation was also noted in the case
of dry season feeding within this zone.
It is possible that, as in the case of dry season feeding, high MC pro-
ducers were more able to organise the practical application of the tech-
nology at farm level. Farmers giving a shortage of cash resources as the
main reason for non-adoption (Table 7) may also have implied that their
evaluation of the relative advantage of mineral supplementation was not
high, thus underlining the important role played by the component of
evaluation in the adoption process. The remarkable second order
improvement in the practice of high MC farmers exposed to treatment D
begs the question as to whether the use of a more intense hypothetical
treatment Ewould permit medium MC producers to equal the changes in
practice made by high MC individuals.
TABLE 8
Level of Probability of Variance Ratios Derived from Analysis of Variance of Change in
Index Values of Dairy Producers with Respect to Identification and Cow Records
TABLE 9
Mean Change in Index Values by Level of Managerial Capacity and Extension Treat-
ment. Dairy Zone Identification and Cow Records
(a) Knowledge
-21
A B C D
Extension treatment
Fig. 4. Change in index values of mineral supplementation technology by low (I-J), med-
) and high (m) MC beef farmers following exposure to different extension treatments.
Adoption of innovations by livestock producers 89
results shown in Table 9 and Fig. 4 tell the same story as found in the beef
and dual-purpose zones in that all farmers with the exception of treatment
A were able to improve record keeping technology equally well regardless
of MC. Again it would appear that the major constraint to adoption was
due to the physical unavailability of cow cards and practical know-how
for their application. However, this similarity between zones hides a major
underlying difference. Unlike the other two system types the causal path
analysis with respect to cow records in the dairy zone predicted the influ-
ence of producer MC on adoption (Wadsworth, 1993) whereas the above
results indicate that this was not in fact the case. Within the beef and dual
purpose zones the ex-ante practice of identification and cow record tech-
nologies was generally low to non-existent as shown by index values, with
only 3% and 12% of beef and dual purpose farmers, respectively, using
cow record cards. Hence it was not surprising to find that, under such
conditions of minimum technology use, no relationship existed with
respect to farmer MC.
In contrast the ex-ante situation in the dairy zone presented a much
more sophisticated set of circumstances with greater awareness and
experience in the use of individual cow records and although inputs such
as cow cards were currently not widely available the higher MC producers
had been able to acquire unused stocks from defunct technical assistance
programmes once operated by banks or Ministry of Agriculture; alter-
natively some individuals had arranged to have their own record cards
printed.
When the extension treatments were applied, even the minimal intensity
treatment B evoked high levels of adoption regardless of MC since the
TABLE 10
Reasons Given for Non-adoption of Identification and Cow Records Technology by
Dairy Producers
Enumeration Record.5
Number % Number %
TABLE 11
Level of Probability of Variance Ratios Derived from Analysis of Variance of Change in
Index Values of Dairy Producers with Respect to Heat Detection
TABLE 12
Mean Change in Index Values by Level of Managerial Capacity and Extension Treat-
ment. Dairy Zone, Heat Detection
n*b~c
See footnotes to Table 3.
Adoption of innovations by livestock producers 91
treatment A P < O-05). These results illustrate the way in which increasing
MC is able to compensate for limited extension input and vice versa. Fig-
ure 5(b) illustrates the different responses for each MC level and shows
how high MC producers are able to achieve a peak increase in evaluation
by exposure to written information only (treatment B), whereas medium
MC farmers require at least the addition of personal contact (treatment C)
and low MC individuals may benefit from participation with the most
intensive EAA which included skills training (treatment D).
The results pertaining to improvements in the practice of the heat
detection routine closely followed the same pattern as already described in
the case of evaluation but with the emergence of more clearly defined
trends. The significant effect of ET (P < 0.01) as demonstrated by analysis
of variance (Table 11) can be shown by t-tests to be composed of the
superiority of treatment D over all other treatments (P < 0.05) and also
the significantly greater change in practice evoked by treatments C and B
over that of treatment A (P < O-05, Table 11). This result clearly demon-
(a) Knowledge
A B C D
Extension treatment
Fig. 5. Change in index values of heat detection technology by low (a), medium (
high (m) MC dairy farmers following exposure to different extension treatments.
Adoption of innovations by livestock producers 93
TABLE 13
Reason Given for Non-adoption of Heat Detection Technology by Dairy Producers
Number % Number %
Economic climate
Attributes of innovations
Structural constraints
Possibly the most critical point which either permits or prohibits the
uptake of a new technology is that no structural constraint should make
adoption impossible. This generalisation seems too obvious to be worthy
of mention, however cases occur, all too frequently, where extension
resources are wasted in futile efforts to promote change under impossible
96 J. Wa&worth
structural conditions. The case of cow record cards in all three zones is a
case in point and has already been discussed within the individual sections.
On making previously unobtainable inputs available the whole adoption
process is transformed from enforced non-adoption (at least for the large
majority) to one of liberty to choose to adopt the innovation. With refer-
ence to cow cards this occurred even though the technology itself is
innately complex and of low initial benefit. Perhaps the initial euphoric
adoption, as a response to the sudden release of the structural constraint,
would later give way to a situation of the continuing use being dependent
on MC, whereupon extension targeting to prevent low MC producers
from abandoning the technology would be called for.
Appropriate innovations
CONCLUSIONS
reasons stated. However the four cases where causal modelling was unable
to predict the experimental result, rather than being considered to repre-
sent a failure of the technique, may be treated as an opportunity to explain
the mechanisms involved in the adoption process in more detail and
thereby improve the technology flow model.
Economic considerations
Structural constraints
become invalid. This was observed to occur with respect to the adoption
of cow record cards by dairy producers. The causal model predicted that
practice was dependent upon MC under ex-ante conditions, this was a
reasonable explanation of the prevailing use of the technology which by
no means was totally inaccessible to producers as in the other two zones.
Hence higher MC producers had been able to adopt more rapidly than
lower MC individuals. This situation changed abruptly with the applica-
tion of the extension experiment when, due to the lifting of the structural
constraint concerned with input availability, all producers were given
access to the technology and almost all adopted regardless of MC. In this
way the effect of structural change may be thought to have swamped the
predicted effect of MC.
In general terms it must be concluded that this type of modelling
approach offers significant potential benefits for increasing the efficiency
of agricultural extension activities by predicting the likely pattern of
adoption with respect to the farmer characteristic MC. Furthermore the
validity of the model will be increased if the above considerations regard-
ing economic conditions, awareness thresholds and structural constraints
are included, at least subjectively, in the interpretation of the causal
models.
ACKNOWLEDGEMENT
This work was carried out whilst the author was employed by the Over-
seas Development Administration (UK) on assignment at the Central
American School of Animal Husbandry.
REFERENCES