Verification of Preferential Treatment for “Official” Restaurants on a Gourmet Rating Site

Abstract

In this paper, we statistically verify whether or not the ratings on the gourmet ratings website give official preferential treatment based on actual data scraped from the site. Since “official” is done by each shop that registers on the site, there is no random assignment of official or unofficial. Therefore, we first adjust for covariates using propensity score matching to verify whether preferences exist on average. Next, we estimate the amount of movement around the rating where the official group is larger than the unofficial group by bunching, and from the fact that there is no preferential treatment around the lowest rating, we assume a monotonic preferential treatment. We then assume monotonic preferences based on the fact that there are no preferences around the lowest rating, and verify whether the result is consistent with the average amount. The result confirms that the total preferences calculated from the results of the monotone preferential assumption and the local preferential quantities are in close agreement with the total preferences calculated from the average preferential quantities.