Functional similarities in spatially correlated location choice models

Abstract

This work analyses the residential location choice problem. Location choice often is regionally clustered. The rules building up this clusters and the implementation in an appropriate model, investigates this work. Additionally to structural and spatial similarities or neighbourhoods, this work especially incorporates functional ones. The functional component is important in residential location choice models. For example, functional commuting regions and travel times are considered as functional connections. Most studies examine one spatial level only. This study suggests a spatial neighbourhood as a three hierarchical level concept – functional commuting region, accessibility including travel time, and mail delivery areas – for application in a Swiss residential choice model. We bring the application problems and theoretical concepts together and think about the appropriate representation of a residential location choice model of Switzerland. Most models use a multinomial logit (MNL) model to analyse this spatial problem, because of large set of alternatives. But the MNL model cannot count for unobserved similarities among alternatives. We give a short literature review about recent studies for residential choice models which allow spatial correlation between alternatives. This adoption is often done by GEV (generalized extreme values) models, a nested logit specification, or can be solved by implementing similarity measure into the utility function, or by a spatial probit model. These study focuses on functional closeness of a correlation coefficient not only assuming correlation in adjacent zones.