Complete sets of demand relations may be fitted using varying types of sample information and varying a priori specifications. In this paper the identification and estimation of Lluch's extended linear expenditure system (ELES) from cross‐sectional data alone is investigated. Under the most favourable conditions of data availability, all of the parameters of the ELES model are identified, and are estimable by the method of reduced form least squares. This is the case where observations on permanent income are available for the consuming units of the cross section and where, in addition, prices are recorded (even though they do not vary from one consuming unit to the next). Under the least favourable conditions only the marginal budget shares are identified. This corresponds to the case where no data on permanent income, or on savings, are available. The conventional ordinary least squares estimators of the marginal budget shares are, under these conditions, biased and inconsistent. Expressions are developed for the large‐sample biases.
|Number of pages||7|
|Journal||Australian Journal of Statistics|
|Publication status||Published - 1 Jan 1973|