Specification and estimation of the nested logit model: alternative normalisations

The nested logit model is currently the preferred extension to the simple m ultinomial logit (MNL) discrete choice model. The appeal of the nested logi t model is its ability to accommodate differential degrees of interdependen ce (i.e., similarity) between subsets of alternatives in a choice set. The received literature displays a frequent lack of attention to the very preci se form that a nested logit model must take to ensure that the resulting mo del is invariant to normalisation of scale and is consistent with utility m aximisation. Some recent papers by F.S. Koppelman, C.H. Wen [Transp. Res. B 32 (5) (1998a) 289; Transp. Res. Record 1645 (1998b) 1] and G.L. Hunt [Nes ted logit models with partial degeneracy, Department of Economics, Universi ty of Maine, December 1998 (revised)] have addressed some aspects of this i ssue, but some important points remain somewhat ambiguous. When utility function parameters have different implicit scales, imposing e quality restrictions on common attributes associated with different alterna tives (i.e., making them generic) can distort these differences in scale. M odel scale parameters are then 'forced' to take up the real differences tha t should be handled via the utility function parameters. With many variatio ns in model specification appearing in the literature. comparisons become d ifficult, if not impossible, without clear statements of the precise form o f the nested logit model. There are a number of approaches to achieving thi s, with some or all of them available as options in commercially available software packages. This article seeks to clarify the issue, and to establis h the points of similarity and dissimilarity of the different formulations that appear in the literature. (C) 2001 Elsevier Science Ltd. All rights re served.