계층적 베이지안 추론을 통한 아파트 단지별 실거래 기반 시세 개발

초록

The real estate price index is an indicator for monitoring the price appreciation of an asset compared to the base point. Real estate is not only a consumption good, it is also an investment asset, so its appreciation can derive decision-making on the part of market participants. Residential assets, which can be mainly represented by “apartments” in South Korea, can be explained by regional or apartment complex-level submarkets. However, the traditional way of estimating price indices by using OLS (Ordinary Least Squares) has some limitations. It cannot estimate price coefficients when there are no transactions at the specific time point, so the traditional index cannot precisely capture the ups and downs of individual submarkets. The repeat-sales model, which is widely-used for constructing housing price indices, uses transaction pairs, so the lack of samples can be a serious obstacle in accurate estimation. In order to overcome the limitations of the OLS methods, we estimate stable submarket-specific indices with hierarchical Bayesian inference by MCMC (Markov Chain Monte Carlo) sampling. This study initially focuses on estimating sub-market indices for apartment-complexes in Mangwoo-dong and Junghwa-dong in Jungrang-gu. We subsequently compare the performance of the newly constructed indices by this research and the existing indices developed by R114.

키워드