媒體新聞能否預測住房市場？Can the News Media Predict the Housing Market?

本文擷取2005年至2017年間，聯合報、經濟日報中與房市相關的新聞，建構媒體新聞的情緒分數，解析新聞情緒與住房市場價格、交易量間的關係。在控制影響住房市場的總體經 濟變數後，向量自我迴歸模型(vectorautoregression model, VAR)的估計結果顯示，首先，較少證據支持房價的上漲與新聞的情緒間存在顯著關係；但交易量與新聞情緒間的關係則較為 顯著。其次，當聯合報新聞情緒越高，房市交易量可能向下修正，且在去除廣告及雜訊新聞 後，情緒對交易量的負向影響更為強烈。再者，當房市交易量越大，經濟日報新聞的樂觀情 緒越高漲，從而有更多對房市榮景的報導。
關鍵詞：房價、交易量、預售屋、中古屋、向量自我迴歸模型、新聞情緒

The study uses a vector autoregression model to investigate the influence of media sentiment on prices and trading volume in Taiwan’s housing market. It constructs sentiment indices from the housing market news of two newspapers, the United Daily News (UDN) and Economic Daily News (EDN). The empirical results based on samples from 2005 to 2017 indicate that an appreciation in lagged prices has no significant influence on media sentiment. By contrast, media sentiment and changes in trading volume are significantly interrelated. Moreover, when UDN sentiment increases, the trading volume may decrease, with this decrease becoming even greater when advertisements and noise news that are irrelevant to the housing market are excluded. Finally, the larger the trading volume, the greater the EDN sentiment, which may result in more optimistic news about the housing market.
Key words: house prices, amount, new house, existing house, vector autoregression model, media sentiment