Alternative nested logit model structures and the special case of partial degeneracy

Partially degenerate nested logit (NL) models have broad applicability in regional science. However, the literature on them is relatively sparse and confusion exists on some aspects of identification and related matters. This paper addresses a number of conceptual and econometric aspects of partially degenerate NL models. These include identification, scaling, invariance, and consistency with the utility-maximizing postulate that underlies discrete choice analysis. This is accomplished within the larger, encompassing framework of nondegenerate NL models of which the partially degenerate model is a special case.