Information transmission in high dimensional choice problems: The value of online ratings in the restaurant market

In choice problems with many alternatives and a priori uncertain outcomes, it has long been argued that individuals may use the observed choices of others as information to guide their own decisions. This paper analyses the role of these social interactions in the context of restaurant choice, while at the same time evaluating the value of a restaurants user rating in determining economic outcomes (as measured by restaurant checkins). The model follows a Polya urn logic, with choices being analyzed in a Dirichlet-multinomial framework. Data is provided by the online urban guide Yelp. Results show that correlation across choices of restaurant guests is present and can be modelled as a function of group-level variables related to information exchange. Ratings and other factors such as price categories do play a role, though traditional modelling approaches that ignore social interactions tend to overstate its importance both in terms of statistical and economic significance. The presented results from the restaurant market may well prove to be important in other choice contexts characterized by many alternatives and highly skewed outcome distributions.