Bias in Estimating Equivalence Scales from Published Group Data

Abstract

This paper employs a number of different aggregation procedures to assess the bias in adult equivalence scales from using published data which are inappropriately grouped and do not contain enough information about relevant household variables. The empirical analysis is based on a non-linear extension of the Almost Ideal demand system and uses 17 years of Family Expenditure Survey (UK) data. The results suggest that (i) inappropriate grouping of data can be responsible for the relatively small and insignificant equivalence scales estimated from published grouped data and (ii) this downward aggregation bias can be reinforced by the omission of relevant household variables (e.g. ownership of durables).