Estimating Wealth Effects Without Expenditure Data—Or Tears: With an Application to Educational Enrollments

tures. The econometric evidence suggests that the asset index, as a proxy of economic status for use in predicting enrollments, is at least as reliable as conventionally measured consumption expenditures, and sometimes more so. This straightforward and pragmatic method of construct-ing a proxy for household economic status produces results that are reassuringly consistent with other approaches (see Montgomery et al. 2000 and references therein) and poten-tially is broadly applicable. Demographic and Health Sur-veys (DHS) have been conducted with nearly identical sur-vey instruments in more than 40 countries. (In India this sur-vey is called the National Family Health Survey, or NFHS.) These surveys include assets and housing conditions but not consumption expenditures (except for an experimental con-sumption module included in Indonesia in 1994, a feature that we exploit later in this paper). Our proposed method for estimating household wealth in the DHS/NFHS allows estimates of the association of wealth with education across households and permits a comparison of wealth gaps across countries (and in India across states). In separate papers we use the asset index to examine wealth and gender gaps in enrollment and educational attainment in more than 35 countries (Filmer 2000; Filmer and Pritchett 1999a). We limit ourselves here to demonstrating the validity and usefulness of an asset index in analyzing education out-comes. The same method might be equally useful in examin-ing wealth differences in other socioeconomic outcomes in the DHS data, such as mortality, morbidity, utilization of health facilities, fertility, and contraceptive use (Bonilla