Applying Geographically Weighted Regression to Conjoint Analysis: Empirical Findings from Urban Park Amenities

The objective of this study is to develop spatially-explicit choice model and investigate its validity and applicability in CA studies. This objective is achieved by applying locally-regressed geographically weighted regression (GWR) and GIS to survey data on hypothetical dogrun facilities (off-leash dog area) in urban recreational parks in Tokyo, Japan. Our results show that spatially-explicit conditional logit model developed in this study outperforms traditional model in terms of data fit and prediction accuracy. Our results also show that marginal willingness-to-pay for various attributes of dogrun facilities has significant spatial variation. Analytical procedure developed in this study can reveal spatially-varying individual preferences on attributes of urban park amenities, and facilitates area-specific decision makings in urban park planning.