Do household asset wealth measurements depend on who is surveyed? Asset reporting concordance within multi-adult households in rural Uganda

Background In resource-limited settings, the Filmer & Pritchett asset index is frequently used to measure household economic status. Little is known about how its validity is affected by differential reporting or recall within households. Methods As part of a whole-population survey in a rural region of southwestern Uganda, we elicited household asset information from married dyads (404 men and 404 matched women) residing within the same households. We assessed the extent to which the asset index yielded differing measures of relative household wealth, depending on whether the husband's or wife's survey data were used in its calculation. To estimate agreement, we used Cohen's κ for binary and categorical variables, and Cronbach's α for continuous variables. We also assessed the extent to which asset wealth quintiles assigned based on husbands' vs wives' reporting were concordant, and whether discordance was related to demographic characteristics. Results For most individual assets, agreement ranged from moderate to very good. Asset index scores based on husbands' vs wives' reporting were positively correlated (Pearson r = 0.85). Corresponding wealth quintiles were moderately concordant (weighted κ = 0.65); 171 households (43%) differed by one or more quintiles when the husbands' vs wives' reporting was used, and 43 (11%) differed by two or more quintiles. Concordance in asset wealth quintile could not be explained by joint educational attainment, age, or age difference. Conclusions There is significant intra-household variability in household asset reporting that can materially affect how households are classified on a widely used measure of relative household asset wealth.