Examining the determinants of the count of customer reviews in peer-to-peer home-sharing platforms using clustering and count regression techniques

Abstract

The sharing economy has experienced massive growth in the short-term shared-home rental industry. However, few studies have investigated the determinants of the number of customer reviews received by these shared-homes. To fill this gap, we were motivated to propose an analytical framework that identified these determinants, both explicit and implicit. We applied Poisson, Quasi-Poisson, and Negative Binomial regressions with a dataset consisting of Airbnb properties from ten different cities worldwide, while successful bookings were proxied by the count of customer reviews posted by guests. We performed a cluster analysis based on the properties to generate homogeneous “cluster cities” and performed the regressions separately for each cluster. Among host-generated features, superhost, host duration, bedrooms, and amenities became significant. Among user-generated features, overall review scores and negative sentiments were significant. We also found that the “superhost” badge moderated the effects of host-generated content on the count of customer reviews. Consequently, guests paid a higher “price per night” for “superhost” properties, while they overlooked crucial attributes such as “website features.” Through these novel “cluster-specific” recommendations, our study extends the existing theories and contributes to the literature of decision analytics and tourism management. Finally, we performed a sensitivity analysis to check for the timeliness and robustness of these determinants.