뉴스 감성지수의 유용성: 소비심리지수 및 주택가격에 대한 영향-서울지역을 중심으로

The purpose of this study is to empirically evaluate whether sentiment indices extracted from domestic news texts are connected with the survey-based Housing Consumer Sentiment Index (HCSI) and whether they provide useful information for explaining and forecasting fluctuations in the Seoul Housing Price Index (HPI). For this purpose, we collected approximately 77,000 news articles related to housing prices from BigKinds between January 2000 and January 2025. These articles were classified into five categories—positive past evaluation, negative past evaluation, positive future outlook, negative future outlook, and irrelevant—and used to construct monthly Past Sentiment Index (PSI) and Future Sentiment Index (FSI). Sentiment classification was conducted by fine-tuning the KoBERT model. The analysis produced the following results. First, the PSI exhibited a clear contemporaneous relationship with the HCSI, while the FSI tended to lead changes in the HCSI. Second, the PSI, the FSI, and the HCSI were all shown to capture short-term fluctuations in Seoul housing prices in advance. These findings suggest that news text-based indices can complement the timeliness limitations of survey-based indices and can serve as useful references for assessing housing market conditions and conducting short-term forecasts.