A typical approach to predicting out-of-sample real estate data is the hedonic approach. Thus far, many
empirical studies have suggested that considering spatial dependence among observations or error terms
can enhance predictive accuracy [1]. Three major research fields that consider spatial dependence are
spatial econometrics [2], spatial statistics [3], and semiparametric statistics [4]. Although the modeling
techniques for considering spatial dependence in these fields are quite similar, they have not been
compared extensively, and the availability of related literature is limited [5]. Some exceptions are the
studies of [6]. In particular, from the viewpoint of spatial prediction, no significant empirical study has
attempted to compare the predictive performance of the numerical models employed in these fields.
Hence, the objective of this study is to empirically compare and discuss the predictive performance of the
models developed in these fields. For this purpose, data on apartment rent and other variables—provided