시계열모형을 활용한 준공 후 미분양주택 예측

초록

This study, based on a time-series model, quantitatively predicts the future trend of houses that remain unsold after completion of construction, which is a serious risk in the housing market. A complex combination of failure to forecast demand due to the pre-sale contract system, macroeconomic and financial market changes, and supply strategy failures contributes to houses remaining unsold after construction completion. In this context, this study explains a time lag in demand and supply and a structure of accumulated inventory based on the theoretical framework of the DW four-quadrant model, empirically analyzing them based on a time series model. Considering monthly data from January 2011 to March 2025, this study sets the number of unsold houses after construction completion in the private sector as the dependent variable to predict. It then compares the predictive power of the ARIMA, ARIMAX, SARIMA, SARIMAX models This study identifies the models with the ACF and the PACF to set candidate parameters, estimates the ones with the best parameters based on the AIC through grid search, and compares the prediction performance with the MAPE and the RMSE to determine the final model. The results of the empirical analysis suggest that the ARIMAX model with the exogenous variables performed the best, and the residuals passed a white noise test. After predicting unsold houses after construction completion for the subsequent years, based on the final ARIMAX model, this study discovered a continued uptrend in the number of the unsold houses without a clear resolution in the short term, which further highlights the importance of preemptive responses and structural remedies.

키워드