상태공간모형을 활용한 부동산실거래가격지수 추정에 관한 연구: 거래빈도가 낮은 지역을 중심으로

This paper estimates the transaction-based real estate price indices for thin markets that have few transactions by using a state space model. The thinness of the market does have a marked effect on the precision of the price index estimate. Since the volitility of the price indices for thin markets estimated by the hedonic price model or repeated- sales model tends to be high, the precision of them is question of our interest. In this paper, we suggest an alternative approach to make stable price indices even when the number of transactions are small or even does not exist. We have developed the transaction-based price indices for the apartment of Gangnam-Gu and Jongno-Gu, Seoul, by using state space models which are estimated by the Kalman filtering and EM(Expectation and Maximization) algorithm. In order to consider the thin markets, we divide housing market by its size: small-sized and medium and large sized apartment. We find that our suggested apartment price indices have lower volatility and much more accurate than the traditional repeated sales indices.