CHAPTER 2 LITERATURE REVIEW
4.4 An analysis of the productivity of Arado and Barka cattle in Eritrea
4.4.6 An analysis of the productivity of Arado and Barka cattle in smallholder production systems in Eritrea
The cattle productivity of the sample of 80 subsistence farmers in the Arado and Barka cattle area was examined using a set of measurements of various dimensions of cattle productivity. Because significant collinearity was detected amongst these variables, a PCA of these variables was conducted to convert the set of related productivity measurements into an index of unrelated (orthogonal) principal components that describe various
dimensions of cattle productivity. The PCA was based on the correlation matrix rather than the covariance matrix because the units of the variables included in the analysis differ (SPSS 1990). A varimax rotation (an orthogonal or unrelated criterion) was used because this type of rotation produced orthogonal principal components and simplifies the interpretation of the principal components (SPSS 1990).
The PCA produced three principal components with eigenvalues of one or greater.
Together, these three components account for 73% of the variation in the original seven livestock productivity indicators that were included in the analysis. The loadings of these principal components are presented in Table 4.13 and are used to “interpret” the dimension of cattle productivity contained in each of the three components. The results of these three components are consistent with the results observed in the ordination diagram in Figure 4.2.
PC2.1 (Principal component one of this analysis) accounts for 35% of total variance in the seven productivity measurements. It is positively related to both income from sale of live animals (ISA) per livestock unit (LU) and off- take rate (OT)/LU and is therefore interpreted as an index of ‘livestock off- take’ productivity. The mean value of PC2.1 for the Arado breed area is 0.13329, and PC2.1 for the Barka breed area is -0.13329. The mean value of PC2.1 indicates that the Arado breed area has a higher OT productivity resulting in a higher income from the sale of live animals than the Barka cattle area. Farmers with high value for PC2.1 have high livestock off-take productivity and farmers with low value for PC2.1 have low livestock off-take productivity, which seems to indicate differences in livestock off-take productivity between the Arado and Barka cattle production systems.
The negative sign for the mean value of the Barka breed area under PC2.1 was unexpected because Barka cattle are dominant in the Eritrean livestock markets and the most numerous breed of cattle in Eritrea. However, careful interpretation of the principal component reveals that the component is an index of the percentage of cattle sold, not the absolute number of cattle sold by each farmer. Results therefore reflect that cattle farming systems in Arado cattle regions, on average, had higher off-take rates than did cattle farming systems in Barka cattle regions during 2002. The high off-take rates of the Arado cattle may be linked to the fact that a shortage of grazing forage induced sales during the dry season. This is evidenced by the ordination diagram in section 4.4.5 that farmers in the
Arado cattle area had higher grazing costs, which emphasise the shortage of grazing forage, than was observed in the Barka cattle area. The shortage of grazing forage for the Arado cattle area is consistent with the results reported in Table 4.2 in section 4.3. The second principal component (PC2.2) explains 24% of the total variance in the seven productivity measurements. PC2.2 has positive loadings for milk yield (MY), live weight (LW) and calf weight at birth (wtbirth). Milk yield plus calf birth weight is interpreted as a
‘breed productivity index’. The mean value of PC2.2 for the Arado breed area is -0.69339 and PC2.2 for the Barka breed area is 0.69339. The mean values shows that compared to the Arado cattle area, subsistence farmers in the Barka cattle area, on average, achieve
Table 4.13 Principal components describing the variability in cattle productivity of a sample of 80 subsistence farmers in Eritrea, 2002
Principal components (PCs)* PC2.1 PC2.2 PC2.3
Percentage of variance explained 35 24 14
Eigenvalue 2. 5 1.7 1. 0
Variables:
Income from sale of live animal (ISA_avg) 0.951
Off-take rate (OT_avg) 0.952
Milk yield /cow (MY) 0.743
Body mass or Live weight (BM/LW) 0.653
Calf birth weight (wtbirth) 0.889
Calving rate (CR) 0.664
Mortality rate (MR) -0.816
The mean values of PCs for:
Barka breed area -0.13329 0.69339 0.28783
Arado breed area 0.13329 -0.69339 -0.28783
* = is a general method for making a factor solution easier to interpret.
higher milk productivity and farm relatively larger cattle (higher body live weight as well as a higher calf weight at birth). This result therefore reflects differences in traits of the Arado and Barka breeds of cattle, but may also reflect that there is a greater availability of forage in the Barka region compared to the Arado region.
PC2.3 accounts 14% of the total variance in the seven productivity measurements. It is positively related to calving rate and negatively related to mortality rate and is interpreted as an index of “calving rate productivity”. The mean value of PC2.3 for the Arado breed area is -0.28783 and for the Barka breed area is 0.28783. Farmers with high values for PC2.3 have high calving rate productivity and farmers with low values for PC2.3 have low calving rate productivity because the index of PC2.3 is positively related to calving rate and negatively related to mortality.
The principal compone nts elicited in this analysis closely match the results of the ordination analysis presented in the previous section. For example, “isa_avg” and
“ot_avg” are closely grouped in the ordination diagram and are also grouped together in PC2.1. Likewise, “milkyield”, “liveweight” and “wtbirth” are closely grouped in the ordination diagram and in PC2.2.
4.4.7 A regression analysis of factors affecting cattle productivity of smallholder