Bachelor of Science in Business Administration (BSBA)
This paper develops a textual analysis methodology to quantify sentiment on public market forums to predict outcomes in the real estate market. This paper draws inspiration from Soo (2018) which quantified sentiment through real estate news media. We believe that analyzing public forums allows us to understand public sentiment in its most unedited, casual form; whereas real estate news media is limited to perspectives and interpretations of an editor. Antweiler and Frank (2004) showed that public forums are significant when predicting stock market outcomes, lending validity to our text source. Our methodology includes identifying a relevant dictionary of positive and negative words, scraping BiggerPockets real estate forums, running a textual sentiment analysis, and finally regressing against fundamental housing market indicators in 34 large metropolitan statistical areas (MSAs) to assess sentiment’s predictability on home prices.
Our regression results suggest that sentiment significance varies more in the short-run with public forum text than it does with news media because news media is “marked-to-market daily.” Marking-to-market is the practice of valuing securities, or portfolios of securities, at their current market value, as opposed to a book value. Because news media is updated every day, and sometimes more than once a day, we find that it is capturing current market home values much more quickly than forum sentiment. Additionally, we conclude that discussion on real estate public forums can predict housing prices in the long-run, suggesting that the users are engaging in conversation that is targeting long-term investments and trying to make sense of the potential future value of a home that they are considering buying and/or selling.