Title: The Impact of an AirBnb Host's Listing Description 'Sentiment' and Length On Occupancy Rates

Abstract: There has been significant literature regarding the way product review sentiment affects brand loyalty. Intrigued by how natural language influences consumer choice, we were motivated to examine whether an AirBnb host's occupancy rate (how often their listing is booked out of the days they indicated their listing was available) can be determined by the perceived sentiment and length of their description summary. Our main goal, more generally, was to determine which features, including (but not limited to) sentiment and description length, most influence a host's occupancy rate. We define sentiment score through a natural language algorithm process, based on the AFINN dictionary. Using AirBnB data on New York City, our hypothesis is that higher sentiment scores (more positive descriptions) and longer summary length lead to higher occupancy rates. Our results show that while longer summary length may positively influence occupancy rates, more positive summary descriptions have no effect. Instead, we find that other factors such as number of reviews and number of amenities, in addition to summary length, are better indicators of occupancy rate.