Robust regression discontinuity estimates of the causal effect of the TripAdvisor’s bubble rating on hotel popularity

We use detailed data on 4,599 hotels located in Rome collected from TripAdvisor, the world's largest travel platform, to examine the effects of bubble ratings (detailed to half-bubbles) on hotel popularity measured with the number of people viewing the hotel’s page. By using a regression discontinuity design, we find that bubble presentation of ratings does not create any significant jumps at cutoffs. This result is different from those obtained in previous studies of similarly designed rating systems from other industries. We provide possible explanations and implications of this result. Another finding is that web users tend to shortlist hotels with the bubble rating of at least 3. Despite that, there is no compelling evidence of review manipulation around the cutoff of 2.75 to make a transition from the 2.5-bubble rating to the 3-bubble rating. Potential uses of the number of views as a proxy of demand in hospitality research are outlined.