The DHS Wealth Index originally was constructed from existing data on household assets, services, and amenities in order to tabulate health, population, nutrition, education, and other indicators according to economic status. The Wealth Index has proved to be one of the most useful background characteristics available from the Demographic and Health Survey (DHS) data. It is now standard in DHS and UNICEF Multiple Indicator Cluster Survey (MICS) final reports and data sets. However, the Index has been criticized as being too urban in its construction and not able to distinguish the poorest of the poor from other poor households.
This paper examines the extent of these problems and suggests and evaluates several possible remedies for them. One remedy that has already been taken by the DHS is to include questions in the standard questionnaires that have been specifically designed to ascertain rural stores of wealth and to distinguish among the poor. For example, questions have been added on rural stores of wealth, such as size of landholdings and number of farm animals by type. To better distinguish among the poor, the surveys ask about possession of furniture items, such as tables, chairs, and beds, because the extremely poor may not have such items. The lack of windows and the lack of windows with glass panes may also indicate an extremely poor household.
A second remedy that could be applied is the use of urban- and rural-specific indexes, or quintile classifications of a common index. These approaches imply, respectively, the construction of separate indexes for urban and rural areas, or the calculation of wealth quintiles separately by type of area. A third approach would be to more finely divide the national index into deciles (which may distinguish better among the poor). These approaches are applied to data from the 2003 Bolivia DHS and to the 2007 Zambia DHS (in an appendix), and their advantages and drawbacks are discussed in the text.
A fourth approach is to construct totally separate indexes for each area and then scale them so that a given score on each index means the same level of wealth. The paper describes two methods of combining separate rural and urban indexes. A method based on regression coefficients is demonstrated using data from the Bolivia 2003 and Zambia 2007 DHS surveys and proves the feasibility of basing urban and rural indexes on differing sets of indicator variables and then scaling these indicators so that a composite index can be calculated. This composite index allows comparability between urban and rural areas.