183
were obtained by recording the costs of those inputs and activities that were done
during the plot experiment and which would also be done by typical cotton producers
in the area. These costs are similar to the costs that typical cotton producers in the area
would have, whereas we assume that the opportunity costs of unpaid household labour
are equal to the costs of hired labour for the same activity. Total revenue was calculated by multiplying the harvested quantities by the selling prices at the time of the
harvest. The production costs and output prices that were used for calculating the land
rents are presented in Tables 4 and 5, respectively. All costs and revenues were converted to US$ using an exchange rate of 2167.32 TZS US$
−1
.
2.7 Statistical Analysis
Three different outcome variables were used in the statistical analysis: seed cotton
yield (in Mg ha
−1
), total revenue (in US$ ha
−1
) and land rent (in US$ ha
−1
). For each
of these three outcome variables, the effects of the soil fertility treatments, the
effects of the pesticide treatments and the effects of selected combinations of soil
fertility and pesticide treatments were investigated. The selected combinations of
soil fertility and pesticide treatments that were included in our analysis are conventional farming without fertilisation or pesticides (CF-0 & CP-0), currently practised
conventional farming (CF-30 & CP-3), conventional farming with higher fertilisation and pesticide application rates (CF-60 & CP-6), innovative conventional farming practices (CF-30 + M & CP-3-N + CU), organic farming without fertilisation or
pesticides (OF-0 & OP-0), currently practised organic farming (OF-3 & OP-P),
organic farming with higher fertilisation rate (OF-5 & OP-P) and innovative organic
farming practices (OF-3 + L & OP-N + CU). As the conventional and organic noinput treatments (CF-0 & CP-0 and OF-0 & OP-0) were identical in the first season
but different in the second season (due to different pre-crops), the statistical analysis
of the selected combinations of treatments considered these two treatments as the
same treatment in season 1 but as two different treatments in season 2.
In order to investigate the effect of the soil fertility and pesticide treatments, we
used the ordinary least squares (OLS) method to test effects on each of the three
outcome variables using the soil fertility treatments, the pesticide treatments, the
interaction terms between the soil fertility treatments and the pesticide treatments,
and the block in which the plot was located as explanatory variables. We calculated
the least squares mean values for each soil fertility and pesticide treatment (Searle
et al. 1980) and conducted pairwise tests of equal means, where we present the
results of these tests as “compact letter display” (Piepho 2004). We investigated the
effects of the selected combinations of soil fertility and pesticide treatments in a
similar way, but we used only the selected combinations of treatments and the block
in which the plot was located as explanatory variables.
All calculations and statistical analyses were conducted with the statistical software “R” (R Core Team 2019) using the add-on packages “emmeans” (Lenth 2019),
“multcomp” (Hothorn et al. 2008) and “ggplot2” (Wickham 2016).
Yield and Profitability of Cotton Grown Under Smallholder Organic and Conventional…
were obtained by recording the costs of those inputs and activities that were done
during the plot experiment and which would also be done by typical cotton producers
in the area. These costs are similar to the costs that typical cotton producers in the area
would have, whereas we assume that the opportunity costs of unpaid household labour
are equal to the costs of hired labour for the same activity. Total revenue was calculated by multiplying the harvested quantities by the selling prices at the time of the
harvest. The production costs and output prices that were used for calculating the land
rents are presented in Tables 4 and 5, respectively. All costs and revenues were converted to US$ using an exchange rate of 2167.32 TZS US$
−1
.
2.7 Statistical Analysis
Three different outcome variables were used in the statistical analysis: seed cotton
yield (in Mg ha
−1
), total revenue (in US$ ha
−1
) and land rent (in US$ ha
−1
). For each
of these three outcome variables, the effects of the soil fertility treatments, the
effects of the pesticide treatments and the effects of selected combinations of soil
fertility and pesticide treatments were investigated. The selected combinations of
soil fertility and pesticide treatments that were included in our analysis are conventional farming without fertilisation or pesticides (CF-0 & CP-0), currently practised
conventional farming (CF-30 & CP-3), conventional farming with higher fertilisation and pesticide application rates (CF-60 & CP-6), innovative conventional farming practices (CF-30 + M & CP-3-N + CU), organic farming without fertilisation or
pesticides (OF-0 & OP-0), currently practised organic farming (OF-3 & OP-P),
organic farming with higher fertilisation rate (OF-5 & OP-P) and innovative organic
farming practices (OF-3 + L & OP-N + CU). As the conventional and organic noinput treatments (CF-0 & CP-0 and OF-0 & OP-0) were identical in the first season
but different in the second season (due to different pre-crops), the statistical analysis
of the selected combinations of treatments considered these two treatments as the
same treatment in season 1 but as two different treatments in season 2.
In order to investigate the effect of the soil fertility and pesticide treatments, we
used the ordinary least squares (OLS) method to test effects on each of the three
outcome variables using the soil fertility treatments, the pesticide treatments, the
interaction terms between the soil fertility treatments and the pesticide treatments,
and the block in which the plot was located as explanatory variables. We calculated
the least squares mean values for each soil fertility and pesticide treatment (Searle
et al. 1980) and conducted pairwise tests of equal means, where we present the
results of these tests as “compact letter display” (Piepho 2004). We investigated the
effects of the selected combinations of soil fertility and pesticide treatments in a
similar way, but we used only the selected combinations of treatments and the block
in which the plot was located as explanatory variables.
All calculations and statistical analyses were conducted with the statistical software “R” (R Core Team 2019) using the add-on packages “emmeans” (Lenth 2019),
“multcomp” (Hothorn et al. 2008) and “ggplot2” (Wickham 2016).
Yield and Profitability of Cotton Grown Under Smallholder Organic and Conventional…