The more the better? Exploring the effects of reviewer social networks on online reviews

ABSTRACT

Conventional wisdom suggests that firms leverage key influencers (e.g. individuals with high centrality) in online communities to stimulate buzz. Using a large panel dataset including 1,569,264 online Yelp reviews and the ego-network of 366,715 individual reviewers over a nine-year period, this study examines the effects of number of ties and network density on the volume and valence of online reviews. In contrast with the general belief that key influencers always generate positive buzz, the findings show that they can adversely affect future review valence. Specifically, reviewers with many connections on Yelp can reduce the positivity of reviews of the same business in the next period. This finding has implications for marketing practice in online community management and social media intervention.

KEYWORDS:

• Online reviews
• eWOM
• social networks
• review volume
• review valence
• ego-network

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. We are using internet-mediated communities and online networks interchangeably in this paper.

2. For a comprehensive review of the effects of eWOM on firm performance, see Babić Rosario et al. (2016), and You et al. (2015).

3. We selected the studies in Table 1 in two steps. First, because review volume and valence are dependent variables in our study, we focus on recent studies that examine what factors influence review volume and valence or their variants. We did not include any articles that treat review volume and valence as independent variables or examine the effects of review volume and review valence on firm performance and refer the reader to published meta-analyses (e.g. Babić Rosario et al., 2016; You et al., 2015) for a comprehensive review on those studies. Second, we focus on recent papers published in the following premier marketing journals – Journal of Marketing, Journal of Marketing Research, Journal of Consumer Research, Marketing Science, Journal of Marketing Management, and Journal of Interactive Marketing. However, we hasten to admit that Table 1 is not inclusive on all the articles that study review volume and valence.

4. One possible reason for the limited number of studies examining online social networks is the difficulty of obtaining online network data. An exception is Yelp, which is is willing to release its user network data for academic research.

5. For example, Yelp displays the number of friends of a reviewer publicly on its platform.

6. The unit of analysis is at business-level. We first calculate the number of ties and network density for each reviewer, then we average the network characteristics at business-level. So, the number of ties (network density) in Figure 1 refers to the average number of ties (network density) of reviewers who have reviewed the same business in the same period. The dependent variables refer to the number of reviews and review valence of a business in the next period.

7. Previous literature has shown that eWOM valence tends to go negative as time goes by (e.g. Godes & Silva, 2012). Different from this temporal effect, our argument here is that, all else equal, the reviews of a business will be more negative in the next period if the business have been reviewed by reviewers with more online connections.

8. It is possible that reviewers with larger network may write reviews with certain valence (either positive or negative) or write more reviews motived by impression management. To examine whether such effect exists, it is necessary to examine the data at individual-reviewer-level to see whether a reviewer’s network characteristics impact the same reviewer’s future review valence and volume. However, our unit of analysis is at restaurant-level and, in each period, different set of reviewers write reviews for a business. And our focus is on the impact of past reviewers’ social network on a restaurant’s future review valence and volume (written by a different set of reviewers). Future research can examine whether such individual reviewer behaviour exists. We thank an anonymous reviewer for raising this issue.

9. See http://www.yelp-support.com/article/How-do-I-add-friends-on-Yelp?l=en_US.

10. See https://www.yelp.com/elite/.

11. See https://www.yelpblog.com/2017/01/10-yelp-elite-events-that-take-the-party-to-the-next-level.

12. Besides temporal effect, literature has also documented sequential effect – review ratings change over order (Godes & Silva, 2012). In order to examine the sequential effect of review order, the unit of analysis must be at the individual review-level. However, our unit of analysis is at restaurant-level.