Community Wealth Ranking and Household Surveys: An Integrative Approach

Abstract

This article contributes to the integration of qualitative and quantitative methods of analysis. A conceptual framework systematically analysing the effects of visibility of resources and measurement errors in wealth ranking and household survey data is presented. Visibility of resources influences both the process of wealth ranking and the social value of the resources, specifically their symbolic importance. The conceptual framework is empirically examined using data from four East African countries: Kenya, Malawi, Tanzania and Uganda. The empirical results support the framework and help to more carefully interpret results and understand social values of resources in different communities.

Notes

1. For example, the visibility of the same resource, say crop output, will be different in an area with a peasant economy compared to another dominated by commercial large scale agriculture. The rural communities considered in this article do not have variations on such a scale.

2. Complicated issues like whether the weighting system reflects the value attached by the community, how individual differences among people involved in wealth ranking are reconciled, and so on, are ignored here.

3. To simplify discussion, here we have considered only one cluster. But if all households in all clusters are considered and if the weights used in wealth ranking differ by clusters – as we generally expect – the matrix of resources for each cluster should appear on the main diagonal of the matrix containing information on all households and the off-diagonal elements should be zero. Correspondingly, the dimension of the weighting vector w becomes (kxm)x1, k capturing the different weights for the same resource in the different clusters.

4. Apparently, the effects of the inclusion of irrelevant variables and the exclusion of relevant ones on using wealth ranks and household survey data are equivalent to the problems of including irrelevant and omitting relevant variables in standard econometrics.

5. LADDER didn't gather detailed information on household expenditures unlike most household surveys.

6. To ascertain that the amount of different types of income increase with total net income, first total net income is regressed on the different types of income. Except for rent income, all are positively and significantly (at 1% level) correlated to total net income.