*4.2. Respondents' Agricultural Drought Resilience Profile*

As indicated, a PCA was applied to construct the outcome variable of the ADRI. Table 4 shows the communalities, component factors, and correlations of variables utilized when constructing the ADRI. All the initial commonalities were above 0.30, which was good. The component variance explained 94% of the total variance. The variables used in PCA were not inter-correlated, and Bartlett's test of sphericity and Kaiser–Meyer–Olkin (KMO) were applied. Bartlett's test of sphericity was significant (*p*-value = 0.000 with chi-square = 2224.837). As a result, the variables were suitability correlated, warranting the application of PCA, because the inter-correlation matrix did not derive from a population. The KMO was 0.64, which was above 0.5, showing that KMO was suitable for PCA. Therefore, the data set met both KMO and Bartlett's sphericity test requirements and was considered suitable for dimension reduction using PCA.


**Table 3.** Socio-economic characteristics of the respondents (*n* = 217).

Source: Authors' compilation based on survey (2020).

**Table 4.** Correlation matrix used for agricultural drought resilience index (ADRI).


Source: Authors' compilation based on survey (2020).

As a result, Equation (5) is rewritten to estimate the ADRI (Equation (8)):

$$\text{ADRI} = \text{PN} \ast 0.967 + \text{PD} \ast 0.979 + \text{Mn} \ast 0.963 + \text{Md} \ast 0.977 \tag{8}$$

Based on the findings using Equations (5) and (8), Table 5 presents the ADRI of the study area. An ADRI greater than zero represents households that were resilient to drought, while ADRI less than zero represents households that were not resilient. An estimated 81% (176) of the farming households were not resilient to agricultural drought.

**Table 5.** Agricultural drought resilience index (ADRI) of Northern Cape Province of South Africa.


Source: Authors' estimation (2020).
