Subjective measures of socio-economic position and the wealth index: a comparative analysis.

INTRODUCTION: The wealth index is a commonly-used measure of socio-economic position (SEP) in low- and middle-income settings, but there is concern that it is strongly influenced by community-level as well as household-level factors. Subjective SEP indicators are infrequently used in health research. METHODS: We use data from 11 280 households included in the Malawi Integrated Household Survey 2004/5. We compare the wealth index with four subjective measures of SEP: perceived food consumption adequacy, perceived overall consumption adequacy, an economic ladder question, and perceived income sufficiency. The wealth index is compared with each subjective SEP measure in terms of: (i) agreement of classification of households, (ii) targeting accuracy with respect to US$1-a-day poverty based on consumption expenditure, and (iii) the socio-economic processes (household- and community-level) giving rise to the SEP scores. RESULTS: Each subjective SEP indicator resulted in considerable differential classification of households compared with the wealth index. Three measures of subjective SEP (perceived food consumption adequacy, economic ladder question, and perceived income sufficiency) identified a higher proportion of dollar-a-day poor households as poor than the wealth index. The wealth index was strongly influenced by community infrastructure, but all subjective SEP indicators were free from strong community-level influence. CONCLUSION: The strengths and limitations of any measure of SEP depend on the context and purpose for which it is being used. In these data, the wealth index was strongly influenced by community infrastructure, whereas the subjective SEP measures were not, perhaps allowing analyses using them to disentangle household and community influences. Several subjective measures also corresponded to dollar-a-day poverty more strongly than the wealth index. Subjective measures may therefore be preferable to the wealth index in some circumstances, although they have their own set of potential biases.