Degressive residential location model considering spatial correlation

To describe the spatial correlation among residential location alternatives, a paired nested Logit model was formulated based on the generalized extreme value theory, with factors including commute cost, housing cost, land use mix, and socio-economic characteristics of household defined as exogenous variables. Using traffic survey data of Beijing in 2005, the model parameters were estimated and tested, and elastic analysis was conducted to analyze the change of alternatives probability due to the variation of factors. The results show that the model is more plausible statistically than multinomial Logit model. Moreover, the spatial correlation between two alternatives decreases as their distance increases, which is more realistic than models presented in previous researches.