How Social Media Sentiment Can Predict Real Estate Market Trends
Social media platforms have evolved into dynamic hubs where individuals freely exchange their views and opinions on a wide range of topics, including financial markets. Recent research has demonstrated how social media posts can predict changes in stock prices. Market participants must learn how to harness the power of sentiment analysis of social media posts or risk falling behind their competition.
One particularly interesting area of study has been the real estate market, a sector known for its lack of transparency. A notable study by Zamani and Schwartz in 2017 highlighted the potential of Tweets to predict house price changes, albeit within a limited scope. This opens an intriguing possibility: Could social media sentiment also serve as a reliable indicator in other parts of the real estate market?
Social media and real estate
A recent article in the Journal of Real Estate Research, “Social Media and Real Estate: Do Twitter Users Predict REIT Performance?”, utilized a comprehensive approach involving around four million tweets over a decade, from 2013 to 2022, focusing on U.S. Real Estate Investment Trusts (REITs). The study employed various methods to analyze social media sentiment, including dictionary-based approaches, classical machine learning with Support Vector Machines (SVMs), and more sophisticated deep learning techniques using Long Short-Term Memory (LSTM) models.
The dictionary method is the simplest and most intuitive. It involves a simple human-curated look-up table that classifies familiar words and phrases as positive, neutral, or negative. Of course, this method has little to no ability to consider context; however, it can still be a useful starting point for sentiment analysis tasks because of how intuitive and interpretable it is. The SVM method is more complicated. It starts with labeling a training dataset according to the sentiment it conveys. These texts are translated into vectors by looking at the frequency of certain words and sets of commonly occurring words, called n-grams.
The Support Vector Machine
The Support Vector Machine then attempts to separate the positive and negative sentiment examples from each other by drawing a boundary between them, specifically the widest possible boundary. An LSTM is a particular form of neural net that is helpful for context-sensitive tasks where noise must be forgotten but key features may play a role over an extended window. Like SVMs, it requires preprocessing the text strings so they can be represented numerically as well as labeling them based on the sentiments they convey. All three methods helped predict REIT performance in the study though the SVM and LSTM outperformed the older dictionary method. Furthermore, the LSTM outperformed the SVM because it could be trained on a larger dataset due to computational limitations surrounding the SVM technique.
The findings suggest that social media sentiment can indeed serve as a meaningful indicator of real estate market trends. By comparing different analytical approaches, the researchers aimed to create a standardized framework that would help investors navigate the complex real estate market. Such a framework could be beneficial not only to individual investors but also to REITs, enabling them to better understand and optimize their position within the market landscape.
This pioneering study is the first of its kind to analyze the impact of social media sentiment on real estate returns using a comprehensive national dataset. The previous attempts, while valuable pieces of research, focused on specific markets and not the US as a whole. The study highlights the potential of integrating social media-based sentiment indicators with traditional measures to provide a fuller picture of the market. As the real estate industry continues to evolve, leveraging social media sentiment could become an increasingly valuable tool for predicting market trends and making informed investment decisions.
Conclusion
Local knowledge plays a massive role in valuing and understanding real estate. This is one of the reasons humans so often outperform even the most sophisticated models when it comes to valuing property. Crowd sourcing this local knowledge by observing social media can help economists and modelers bridge this gap. Of course, modelers will not abandon tried and true leading indicators. The synergy between social media analytics and traditional forecasting could herald a new era in real estate investing.
Read more: Impact of differential measurement error in house price indices
Franklin Carroll
Franklin Carroll is Kukun's VP of Modelling and Analytics; After finishing his studies at the University of Chicago, he worked at Fannie Mae where he helped build Collateral Underwriter, roll out the Value Verify initiative, and specialized in collateral related and credit related policy.
Published June 14, 2024
