Social Influence in Online Reviews : Evidence from the Steam Store

How does social influence affect consumer ratings? Using a dataset from the popular Steam gaming platform I investigate how quality judgements depend on pre-existing consumer assessments. In 2019, Steam introduced a new review system which decreased the exposure of users to previous ratings. Firstly, I find that user ratings are dependent on average ratings. A 10 percent increase in average rating increases the probability a review is positive by 5.4% before the policy change, but only by 2.8% after. The result is not due to selection, and is robust to a wide range of alternative specifications. Secondly, the effect is heavily asymmetric: individual reviewers are more negative when exposed to a lower average rating, but do not respond to a higher one. This negativity compounds and inflates the gap between lower rated and higher rated games. Overall, these social influence effects are driven by less experienced users on the platform. Finally, using estimates of owner data, I run a structural model of game choice. A 1% increase in rating is equivalent to a 2.5 dollar price reduction. This suggests social influence has large implications for buyers and sellers.