Traditionally, the difference between income and expenditure reported in household surveys is used for estimation of savings at the household level. However, persistent deviation in consumption–income ratios by household income brackets raises questions about both the quality of data and the estimation method employed. This paper provides statistical evidence to accentuate the endogeneity bias, which is connected with the choice of welfare proxy for ranking process, in savings estimates by income groups. Also, two different estimators for households spending behaviour are discussed and the statistical properties of their difference are derived by the delta method. – households ; measurement of welfare ; savings