Influentials, Imitables, or Susceptibles? Virality and Word-of-Mouth Conversations in Online Social Networks

Abstract

Motivated by the rise of social media platforms that achieve a fusion of content and community, we consider the role of word-of-mouth communications (WOM) structured through a network. Using a data set from YouTube, we examine how cascades of WOM interactions enhance the popularity of videos. We first estimate the impact of channel influence and other network parameters in initiating WOM communications. The probit estimation considers the selection effect in videos that are likely to be associated with a greater propensity to trigger WOM. We find that factors related to a channel’s ability to be a connector and a translator is most likely to result in the incidence of WOM. We then examine how cascades of WOM conversations have persistent impacts on subsequent video popularity. Empirically, the main issue here is heterogeneity in the epidemic potential of a video. Since the threshold might vary across videos, we use a finite mixture model. We also conduct a simultaneous estimation using latent instrumental variables to address endogeneity from unobservables. Our research has implications for researchers and practitioners by highlighting how WOM travels through networks of influence and susceptibility in disseminating awareness, and holds insights in regard to designing social recommendation systems and identifying trending topics in social media.

Key words and phrases:

• electronic word of mouth
• eWOM
• finite mixture model
• latent instrumental variables
• opinion cascades
• peer effects
• social media
• social recommendations
• user-generated content

Acknowledgments

The authors thank Sinan Aral, Ravi Bapna, Eyal Carmi, Paul Damien, Chris Dellarocas, David Godes, David Krackhardt, Ramayya Krishnan, Yingda Lu, Gal Oestreicher-Singer, Sarah Rice, Param Vir Singh, Arun Sundararajan, Rahul Telang, and Andrew B. Whinston for their comments. They also thank seminar participants at Heinz College at Carnegie Mellon University and University of Connecticut for helpful comments. They thank participants at the Workshop in Information Systems Economics (WISE), the Winter Conference on Business Intelligence, Informs Conference on Information Systems and Technology (CIST), and the Symposium on Statistical Challenges in eCommerce Research (SCECR) for feedback on earlier versions of this paper.

Notes

1. Users are referred to as channels on YouTube. We relied on reports from the YouTube blog and TubeMogul, an analytics firm that tracks YouTube. We examine the action of YouTube users posting response videos.

2. YouTube highlights videos by featuring the most watched videos as well as promoting videos. However, such mechanisms constitute external rather than internal WOM.

3. Data API supports a number of client libraries that abstract the API into a language-specific object model such as Java, .NET, PHP, and Python (http://code.google.com/apis/youtube/overview.html).

4. Subsequent to our data collection, YouTube changed its API to restrict access to the top 200 comments so that it is no longer possible to obtain each comment posted on a video, which makes it almost impossible to study WOM interactions.

5. The k-core provides an alternative to using data-mining methods to detect influential commentators or using sentiment analysis to infer positive or negative sentiment.

6. For robustness we considered three different treatment regimes and found that a single treatment regime had the best explanatory power.

Additional information

Notes on contributors

Anjana Susarla

Anjana Susarla ([email protected]) is an associate professor at the Eli Broad College of Business, Michigan State University. Her research has appeared in Information Systems Research, Journal of Management Information Systems, Management Science, and MIS Quarterly. She received her Ph.D. in information systems from the University of Texas at Austin. Her research focuses on big data analytics, contracts and sourcing, cloud computing, social media, and network science.

Jeong-Ha Oh

Jeong-ha Oh ([email protected]) is an assistant professor of computer information systems at the Robinson College of Business, Georgia State University. She received her Ph.D. in information systems from the Foster School of Business at the University of Washington in Seattle. Her current research focuses on health-care analytics, online social networks analysis, information cascade, and content diffusion in social media. She has published in Information Systems Research.

Yong Tan

Yong Tan ([email protected]; corresponding author) is the Neal and Jan Dempsey Professor of Information Systems at the Michael G. Foster School of Business, University of Washington, and Chang Jiang Scholar Visiting Chair at the School of Economics and Management, Tsinghua University. His research interests include social media and networks, mobile and electronic commerce, big data analytics, and economics of information systems. He has published in Information Systems Research, Journal of Management Information Systems, Management Science, Management Information Systems Quarterly, and Operations Research, among others. He is a senior editor of Information Systems Research and a member of the Board of Editors of the Journal of Management Information Systems.