Where i indexes the restaurants, t indexes the time when the rating is made and k indexes the
website. The dependent variable, >2 34
, is the consumer rating submitted for restaurant i at
time t on website k. GH
is a time dummy that equals to one if the time period is after the change
of rating system, and zero otherwise.
is a dummy that equals to one if the ratings are made on
treatment group which is TripAdvisor, and zero if on control group which is Yelp.
restaurant fixed effect.
5.2. DID Model: Main effect of the rating system change
We summarize difference in difference approach below:
measures the before and after rating changes for restaurant i at Yelp and is an indication of any
quality change of restaurant i. The coefficient,
on the interaction term GH
the difference due to change in rating system, after controlling for restaurant quality changes over
time and systematic website differences.
5.3. Deviation Model
we relate Deviation of ratings to the nominal sequence value of the rating at time t. Rating deviation
of restaurant i at time t is measured as the absolute difference between rating made by consumer
at time t and previous observed rating. We compute previous observed ratings as the average of
all ratings made before time t. Since the dataset consists of different restaurants, we include
restaurant fixed effects
as controls in the analysis.
measures the relationship between
deviation of current rating from previous observed ratings and sequence. A positive
deviation from previous observed ratings increase with rating sequence, while a negative