Spatial analysis of Twitter sentiment and district-level housing prices

Purpose: Studies have shown a correlation and predictive impact of sentiment on asset prices, including Twitter sentiment on markets and individual stocks. This paper aims to determine whether there exists such a correlation between Twitter sentiment and property prices. Design/methodology/approach: The authors construct district-level sentiment indices for every district of Istanbul using a dictionary-based polarity scoring method applied to a data set of 1.7 million original tweets that mention one or more of those districts. The authors apply a spatial lag model to estimate the relationship between Twitter sentiment regarding a district and housing prices or housing price appreciation in that district. Findings: The findings indicate a significant but negative correlation between Twitter sentiment and property prices and price appreciation. However, the percentage of check-in tweets is found to be positively correlated with prices and price appreciation. Research limitations/implications: The analysis is cross-sectional, and therefore, unable to answer the question of whether Twitter can Granger-cause changes in housing markets. Future research should focus on creation of a property-focused lexicon and panel analysis over a longer time horizon. Practical implications: The findings suggest a role for Twitter-derived sentiment in predictive models for local variation in property prices as it can be observed in real time. Originality/value: This is the first study to analyze the link between sentiment measures derived from Twitter, rather than surveys or news media, on property prices.