208
decreases (Appendix 2). This means that larger farms (with more than 20 beehives
or more than 1 acre of coffee-cultivated land) are generally less inclined than smaller
farms to innovate their processes. Finally, when access to farm input resources
increases, process innovation increases as well, when entrepreneurial competencies
are also considered (Appendix 3).
No other variable—not entrepreneurial orientation or farm characteristics—
influences process innovations as much as a farmer’s education level, a finding that
has been demonstrated frequently in many other contexts (e.g., Thangata and
Alavalapati 2003). In Uganda, however, relatively few farmers attain high levels of
education, so it is important to evaluate the findings with education level taken out
of the equation. Results suggested that farmers with smaller farm size and higher
access to resources have significantly higher levels of agricultural innovation. This
finding on farm size is somewhat surprising because it contrasts with a wide literature suggesting that larger farms engage more often in innovations (Adesina and
Baidu-Forson 1995; Weir and Knight 2004). To better understand these results, we
further analyzed the interaction effect of these farm characteristics and entrepreneurial orientation on innovation. We found:
• The higher farmers’ innovativeness is, the stronger the negative effect of farm
size on their innovations. This may suggest that smaller farmers would be the
most reactive in taking up new product, process and market innovations when
they become more innovative.
• The higher farmers’ proactiveness is, the stronger is the positive effect of farm
size on their innovations. This may suggest that larger farmers would be the most
reactive in taking up new product, process and market innovations when they
become more proactive.
These results confirm that factors such as education levels, farm size and access
to resources are key factors shaping the triggering and scaling of agricultural innovations, including those related to CSA practices.
A couple of methodological cautions are in order. First, by testing the measurement model through the CFA, we found that the measures of risk-taking as a dimension of entrepreneurial orientation did not fit the data in the Ugandan context. This
means that—in contrast to Lai et al. (2017a) in the Philippines—the farmers in this
Ugandan survey did not understand how the questionnaire items on risk-taking
together corresponded to one concept. More significantly, this signals that risktaking may not be a suitable or desirable dimension of entrepreneurial orientation in
a farm context afflicted by market, social and environmental shocks. Given the limited sample size, though, it is worth conducting further tests on the risk-taking in
other context before recommending to definitely drop this dimension in similar
study contexts.
Second, in our database, multicollinearity among variables is high (meaning that
the dimensions of entrepreneurial orientation are highly correlated with each other
and with some farm characteristics) and sample size is relatively small (n = 152).
This created statistical problems that forced us to build multiple smaller regression
models to analyze all the variables of interest. If future research allows the collecC. L. Barzola Iza et al.
decreases (Appendix 2). This means that larger farms (with more than 20 beehives
or more than 1 acre of coffee-cultivated land) are generally less inclined than smaller
farms to innovate their processes. Finally, when access to farm input resources
increases, process innovation increases as well, when entrepreneurial competencies
are also considered (Appendix 3).
No other variable—not entrepreneurial orientation or farm characteristics—
influences process innovations as much as a farmer’s education level, a finding that
has been demonstrated frequently in many other contexts (e.g., Thangata and
Alavalapati 2003). In Uganda, however, relatively few farmers attain high levels of
education, so it is important to evaluate the findings with education level taken out
of the equation. Results suggested that farmers with smaller farm size and higher
access to resources have significantly higher levels of agricultural innovation. This
finding on farm size is somewhat surprising because it contrasts with a wide literature suggesting that larger farms engage more often in innovations (Adesina and
Baidu-Forson 1995; Weir and Knight 2004). To better understand these results, we
further analyzed the interaction effect of these farm characteristics and entrepreneurial orientation on innovation. We found:
• The higher farmers’ innovativeness is, the stronger the negative effect of farm
size on their innovations. This may suggest that smaller farmers would be the
most reactive in taking up new product, process and market innovations when
they become more innovative.
• The higher farmers’ proactiveness is, the stronger is the positive effect of farm
size on their innovations. This may suggest that larger farmers would be the most
reactive in taking up new product, process and market innovations when they
become more proactive.
These results confirm that factors such as education levels, farm size and access
to resources are key factors shaping the triggering and scaling of agricultural innovations, including those related to CSA practices.
A couple of methodological cautions are in order. First, by testing the measurement model through the CFA, we found that the measures of risk-taking as a dimension of entrepreneurial orientation did not fit the data in the Ugandan context. This
means that—in contrast to Lai et al. (2017a) in the Philippines—the farmers in this
Ugandan survey did not understand how the questionnaire items on risk-taking
together corresponded to one concept. More significantly, this signals that risktaking may not be a suitable or desirable dimension of entrepreneurial orientation in
a farm context afflicted by market, social and environmental shocks. Given the limited sample size, though, it is worth conducting further tests on the risk-taking in
other context before recommending to definitely drop this dimension in similar
study contexts.
Second, in our database, multicollinearity among variables is high (meaning that
the dimensions of entrepreneurial orientation are highly correlated with each other
and with some farm characteristics) and sample size is relatively small (n = 152).
This created statistical problems that forced us to build multiple smaller regression
models to analyze all the variables of interest. If future research allows the collecC. L. Barzola Iza et al.