Sellers’ behavior and online rating bias: A sentiment analysis on airbnb reviews

We investigate the bias of online ratings stemming from the personal interaction between sellers and buyers, and quantify its effect on ratings and consumer demand. Using data and text reviews from Airbnb in Barcelona, we develop a text analytic algorithm and semantic analysis to measure host kindness. To expose the bias, we exploit the rating of the listing’s location, which should not be influenced by host behavior. We find that kindness is positively related to the location’s rating and to the listing’s demand. Moreover, host kindness mitigates the negative impact of an inconvenient position on both the location score and the listing demand. We address endogeneity concerns by exploiting the shock to tourism caused by COVID-19.