본문 바로가기
  • Home

Incorporating Subjective Priors into Mass Appraisal Modeling

Lee, Changro 1 Park, Key Ho 1

1서울대학교

Accredited

ABSTRACT

The government-announced land price has a variety of usage from providing land information to market to serving as assessment value for property. In this study, we investigate an alternative model to enhance the accuracy of this government-announced land price. The current model for the price estimation is based on a standard comparison table of price, which is a special case of hedonic pricing models. This comparison table is constructed from the ordinary least squares result, which is one of frequentist approaches. The frequentist approach, however, can only produce an output as good as the quality of input data and if the data is not enough in amount or its quality is poor, its output necessarily deteriorates. We overcome this constraint of data by incorporating substantive knowledge of real estate experts (subjective priors) into mass appraisal modeling. The improved accuracy of our model is verified by the cross-validation simulation method. Attractiveness of a Bayesian model lies in the fact that it provides a logically transparent method for combining prior information with data, rather than rejecting the knowledge in experts’ heads just because it is subjective or qualitative.

Citation status

* References for papers published after 2023 are currently being built.

This paper was written with support from the National Research Foundation of Korea.