The estimation of complete aggregation structures

The object of this paper is to establish several results which permit efficient estimation of behavioral parameters in models of aggregate data. The approach follows the 'minimum chi-square' estimation strategy customarily applied to contingency tables; by treating aggregates as general functionals of empirical distribution functions for each 'cell' (i.e., aggregation unit), consistent and asymptotically normal estimates are obtained for a fixed number of aggregation units as the number of observations per unit increases. The paper also discusses conditions under which the proposed estimators are asymptotically efficient, and gives examples applying the approach to continuous and discrete demand models.