물류센터 임대시장 분석: 공간회귀모형과 혼합지리가중회귀모형의 활용

초록

The objective of this study is to investigate the spatial dimensions of warehouse rents by using spatial autoregressive regression (SAR) and mixed geographically weighted regression (MGWR) models. Two types of spatial autoregressive regression models are used to incorporate spatial autocorrelation present in the warehouse rents: spatial lag and spatial error models. Spatial lag regression model includes additional neighborhood variable to represent the influences of neighboring warehouses on the observed warehouse rent. Spatial error model incorporates spatial autocorrelation in the error term. Unlike OLS and spatial autoregressive regression models, GWR is a local model that estimates regression coefficients for each observation. Therefore, GWR allows us to assess local effects of the explanatory variables. Warehouse rent dataset for the Seoul Metropolitan Area (SMA) in South Korea is used as a case study. Monthly warehouse rent is regressed on multiple explanatory variables including transactional, physical, locational and operational characteristics of the warehouse. Model performances and prediction accuracies are compared for the OLS, spatial autoregressive, and MGWR models. Keywords: Warehouse rent; hedonic price model; spatial autoregressive regression; geographically weighted regression, logistics real estate