180
varieties adoption. To establish the reliability of the estimates from the logit model,
a variance inflation factor (VIF) test for multicollinearity and Hosmer–Lemeshow
(HL) test for goodness of fit were conducted. The VIF test ruled out serious multicollinearity and the HL test showed that the logit model was properly specified.
Additionally, the log likelihood ratio obtained was −608.5264, which was statistically significant at 1%, while the pseudo-R
2
value of the model was 0.1421. This
indicated overall significance of the logistic model and a good fit for the data. As
shown in Table 15.2, the decision to adopt stress tolerant varieties was positively
influenced by household size, gender of the household head, access to agricultural
information from NGOs, perception of future changes in climate, number of years’
residence in the village and the asset index. This indicated that for every unit
increase in any of the variables, the probability of adopting improved varieties
increases by the corresponding marginal effects.
These results are in harmony with other past studies on theoretical and empirical
literature about agricultural technology adoption. For instance, farmers who had
access to NGO information were 10.33% more likely to adopt stress-tolerant varieties than their counterparts. This is partly because access to information reduces
uncertainty about new technologies as farmers become aware of the new technology
and how to use it effectively. These findings are in agreement with (BonabanaWabbi 2002) which reported that farmers who had access to agricultural information had a higher probability of adopting integrated pest management technologies
in Uganda. However, against our expectation, farmers who had access to demo plots
information were not more likely to adopt improved varieties. We hypothesise the
reason for this finding is based on the context of the study site. Communities in
northern Uganda suffered conflict and were displaced in camps and have only resettled back in their farms within the last decade. Approaches relying on trust and
social networks are, therefore, more likely to influence learning and the adoption of
stress tolerant varieties. In this case, we see that learning through NGOs—most of
which have been in the community for long periods and have built good relationships with the farmers—is likely to be more effective compared to demonstration
plots, which are often set up for short periods. In addition, CSA technologies are
context-specific and so might be the approaches used to promote CSA. In Nwoya
District, for example, households as well as villages tend to be geographically quite
far from each other. In such cases, farmers (in a previous and related study) indicate
that distance to the plot was the main reason why they were not actively participating in the demonstration plots (Shikuku et al. 2015). Such farmers often demanded
the reimbursement of transport and refreshments costs during training, without
which they were unwilling to actively learn.
Household size had a positive effect on the adoption of stress tolerant varieties. This is plausible because a greater number of household members means
there are more people available to provide the intensive labour that comes with
the adoption of new technologies. This is in agreement with Adepoju and
Obayelu (2013) who reported that household size was an important factor in
determining the type of livelihood strategies adopted. The significance of the
C. M. Mwungu et al.
varieties adoption. To establish the reliability of the estimates from the logit model,
a variance inflation factor (VIF) test for multicollinearity and Hosmer–Lemeshow
(HL) test for goodness of fit were conducted. The VIF test ruled out serious multicollinearity and the HL test showed that the logit model was properly specified.
Additionally, the log likelihood ratio obtained was −608.5264, which was statistically significant at 1%, while the pseudo-R
2
value of the model was 0.1421. This
indicated overall significance of the logistic model and a good fit for the data. As
shown in Table 15.2, the decision to adopt stress tolerant varieties was positively
influenced by household size, gender of the household head, access to agricultural
information from NGOs, perception of future changes in climate, number of years’
residence in the village and the asset index. This indicated that for every unit
increase in any of the variables, the probability of adopting improved varieties
increases by the corresponding marginal effects.
These results are in harmony with other past studies on theoretical and empirical
literature about agricultural technology adoption. For instance, farmers who had
access to NGO information were 10.33% more likely to adopt stress-tolerant varieties than their counterparts. This is partly because access to information reduces
uncertainty about new technologies as farmers become aware of the new technology
and how to use it effectively. These findings are in agreement with (BonabanaWabbi 2002) which reported that farmers who had access to agricultural information had a higher probability of adopting integrated pest management technologies
in Uganda. However, against our expectation, farmers who had access to demo plots
information were not more likely to adopt improved varieties. We hypothesise the
reason for this finding is based on the context of the study site. Communities in
northern Uganda suffered conflict and were displaced in camps and have only resettled back in their farms within the last decade. Approaches relying on trust and
social networks are, therefore, more likely to influence learning and the adoption of
stress tolerant varieties. In this case, we see that learning through NGOs—most of
which have been in the community for long periods and have built good relationships with the farmers—is likely to be more effective compared to demonstration
plots, which are often set up for short periods. In addition, CSA technologies are
context-specific and so might be the approaches used to promote CSA. In Nwoya
District, for example, households as well as villages tend to be geographically quite
far from each other. In such cases, farmers (in a previous and related study) indicate
that distance to the plot was the main reason why they were not actively participating in the demonstration plots (Shikuku et al. 2015). Such farmers often demanded
the reimbursement of transport and refreshments costs during training, without
which they were unwilling to actively learn.
Household size had a positive effect on the adoption of stress tolerant varieties. This is plausible because a greater number of household members means
there are more people available to provide the intensive labour that comes with
the adoption of new technologies. This is in agreement with Adepoju and
Obayelu (2013) who reported that household size was an important factor in
determining the type of livelihood strategies adopted. The significance of the
C. M. Mwungu et al.