This study uses various approaches to analyze whether bubbles existed in the housing market in the Seoul Capital Area (Seoul, Incheon, and Gyeonggi) and, if so, what their relative sizes were.
This study asserts the following, through a preliminary review prior to a rigorous econometric analysis. 1) Among the four periods of rapid housing price spikes in the housing market in the Seoul Capital Area, the housing price surge that occurred in the early 2000s can be understood as a recovery process from an undervalued phase, which was developed due to the downtrend in the housing price over the 1990s and the further big drops after the Asian financial crisis. 2) Income level, which is considered a major variable in determining housing prices, did not play a significant role in the housing market in the Seoul Capital Area. 3) The size of housing bubble in the Seoul metropolitan area, which is calculated using the present value of rent derived through a simple calculation, was the largest in the mid-2000s, and shows significant volatility at a modest level during the recent housing price surge period. This is due to the large volatility of the interest rate, which is used as the discount rate, and the movement of the jeonse price, which shows a divergence from the movement of the housing sales price.
For housing bubble estimation using an econometric methodology with more rigorous theoretical bases, this study uses a state-space model and an iterative least-squares model. The state space model is a specialized model for estimating unobservable variables by modelling the relationship between observed and unobserved variables and the dynamic motion of unobserved variables. In the context of the present value approach, which is most popular in academics, state-space model is constructed by mapping housing rent to an observable factor and housing bubble to an unobservable factor. The main result of the estimation by using a state space model is as follows. Although the bubbles in the housing market in the Seoul Capital Area have significant volatilities, they continue to expand and rise to close to 60% of the sale price during the recent surge period. However, this study also discusses that the bubble size estimated through the state-space model may be overestimated due to the inherent problems of the model.
The iterative least-square model considers the case where the estimated value of the dependent variable, which is obtained through an estimation, is used as an independent variable again, and the model needs to be recursively estimated due to the structure of the model. In this study, we consider the case where the estimate of the housing fundamental value derived through an OLS estimation is included in the model again as a variable called the degree of divergence from the fundamental housing value. On the other hand, this study solves problem that arises in the process of calculating the housing bubble ratio through differencing inversely from a arbitrary time point, which is often used in many regression analysis using the log-differenced housing sale price variable as a dependent variable. We discuss this problem and resolve it through the assumption that the positive (+) housing bubble is offset by the negative (-) housing bubble in the long run, so that the sum of the housing bubble over time approaches 0 (zero) over the entire analysis period.
The characteristics of the housing bubbles in the Seoul metropolitan area estimated through the iterative least-square method are as follows. 1) The ratio of housing bubble to housing sale price was historically the highest during the recent housing price surge period, and is 2-3 times higher than during the price surge in the mid-2000s. However, 2) the estimated relative size of the bubble ratio is approximately 10% to 20% of the sale price, revealing a large gap from the 60% level derived from the state-space model. Considering the differences in the estimation methods used and the inherent flaws of the state-space model, it is argued that the significant gap between the bubble ratios derived from the two models may not be a big problem. Rather, this study calls for attention to the fact that the bubble ratio during the recent period is very high compared to the ratio observed during the mid-2000s.
Housing bubble estimation using the iterative least-square method is applied not only at the province level but also at the county level in Gyeonggi-do. Housing bubbles are estimated in 26 cities and counties in Gyeonggi-do. Although the size of housing bubble formed during the recent surge period is larger than the bubble size in the mid-2000s, which is in line with the estimation results at the provincial level, it is also confirmed that significant differences among cities and counties are also revealed.