Housing Markets and Home-Sharing Services: An Analysis on Airbnb in Boston

Abstract

We assess the role of Airbnb in Boston housing market pricing mechanisms using fixed effects models with clustered standard errors. First, we find that a 10% increase in number of Airbnb listings leads to a 0.388% increase in residential property valuations as a statistically significant localized average treatment effect for Boston. Conversely, we find no statistically significant impact of number of Airbnb reviews on residential property valuations. Second, in treating the Airbnb platform as a housing market, we propose several pricing mechanisms for the base nightly rate of an Airbnb listing. We find that physical property characteristics such as accommodation capacity, number of bedrooms, and room type are significant predictors of pricing. Moreover, we find that intangible characteristics such as host verification, review count, and cancellation policies are also significant predictors of pricing. Finally, we propose sentiment of written Airbnb guest reviews as an intangible characteristic to predict nightly rates. Using unsupervised NLP algorithms, we construct four objective sentiment scores by word count weighting and by hyperbolic and exponential time decay. We find that although review sentiment alone is an insignificant predictor, it supplements review count through interaction terms that influence the Airbnb nightly rate.