Measuring household resilience in developing countries : evidence from six African countries.

Abstract

In this study, a household resilience score was developed as a measure of rural household resilience to identify households with low resilience and to measure progress towards improved household resilience. Resilience is the ability of households to cope with risk. The motivation for the study originated from the first objective of the Framework of African Food Security (FAFS) of improved household risk management, and the indicator of progress towards this objective – proposed by the FAFS - a resilience score. A review of the literature indicated that the assets owned by a household could be used as a proxy for resilience. The household component of the Demographic and Health Surveys for six African countries was used to develop and apply the resilience score. The score was estimated using an index of assets owned by the household and information regarding household access to certain services and characteristics of the dwelling. There is disagreement in the literature concerning the best method of constructing an asset index in terms of how to weight the variables included in the index. As a result, four methods of constructing an index of socio-economic status (SES) were selected for comparison in this study: two linear principal component analysis (PCA) techniques; a non-linear or categorical principal component analysis (CATPCA) method; and a simple sum of assets technique. The results from the application of each of the four indices to the country data and the resulting classification of households into quintiles of SES were compared across several assessment criteria. No single method out-performed the others across all the assessment criteria. However, the CATPCA method performed better in terms of the proportion of variance explained by the first principal component and the stability of the solution. The results showed that for all methods, SES was not evenly distributed across the sample populations for the countries analysed. This violates the assumption of uniformity implied when using quintiles as classification cut-off points. As an alternate to the quintile split cluster analysis was applied to the SES scores derived for each country. The classification of households into SES groups was repeated using k-means cluster analysis of the household SES scores estimated by the CATPCA method for each country. The results showed that a greater proportion of households fell into relatively lower levels of SES, which is in contrast to the assumption of uniformity of SES made when using the quintile cut-off approach. Cluster analysis better reflected the clustered nature of the household data analysed in this study, compared to the quintile cut-off method. In a final analysis, the index of SES along with k-means cluster analysis was applied to household data from two different time periods for five African countries to determine whether the resilience measure was able to detect changes in household SES between the two periods and, therefore, whether the tool could be used to monitor changes in household resilience over time. The results showed evidence of adjustments in SES over time: there were differences in the per cent of households allocated to the clusters of SES between the two periods. Using the CATPCA index and k-means cluster analysis, Egypt, Uganda and Mali showed an increase in the per cent of 'poor' households, while for Kenya and Tanzania there was a reduction in the per cent of households allocated to the first cluster between time periods: the decrease for Kenya from 2003 to 2008 was as much as 13 percentage points. The observed changes in SES were then compared to changes in national poverty estimates reported in the literature. The resilience score developed in the study displayed an ability to track changes in household SES over time and could be used as a measure of progress towards improved household resilience. As such, the resilience measure could be valuable to policy-makers for monitoring the impacts of policies aimed at improving household resilience. Future research is recommended before the reliability of the resilience measure developed here can be fully ascertained.