The Paradoxes of Word of Mouth in Electronic Commerce

Abstract

Challenging conventional wisdom, we unravel three paradoxes of word of mouth (WOM) in e-commerce. Specifically, the WOM valence paradox contends that higher WOM valence of a product results in a larger subsequent decrease in the WOM valence of the product, the WOM volume paradox propounds that higher WOM volume of a product results in a smaller subsequent increase in the WOM volume of the product, and the WOM spillover paradox proposes that an improvement in the WOM of a product also improves the WOM of connected products in a product network. These paradoxes caution online retailers that superior WOM may at times backfire and not boost further sales. Drawing theoretical support from expectation-confirmation theory and network theory, we collect data from China’s largest business-to-consumer platform, Tmall.com, and use linear panel data models to examine WOM evolution in a product network, controlling for relevant factors at the individual product, product network, and time unit levels. Importantly, we base our identification strategies on the use of instrumental variables and the difference-in-differences estimation approach. Numerous statistical checks confirm the robustness and consistency of our findings. We contribute to a much richer theoretical understanding of WOM, by extending the applicability of expectation-confirmation theory and network theory to novel predictions and contexts, adding a dynamic perspective, unveiling three important WOM paradoxes, and offering practical insights.

Key words and phrases:

• econometric analysis
• electronic commerce
• e-tail
• eWoM
• product network
• product review
• word of mouth

Acknowledgments

We thank the Editor and the anonymous reviewers for their valuable comments and suggestions. This research is partially supported by the National Natural Science Foundation of China, Project Grant 71502079, and the Singapore Ministry of Education, Project Grant R-253-000-105-112. Both authors contributed equally to this research.

Notes

1 On most e-commerce sites, each product is linked to relevant products (accompanied by the respective WOM) to assist consumers in their purchase decisions [63]. Thus, the recommendations create a visible directed product network (or WOM network) where products (or WOM) (i.e., network nodes) are explicitly connected by hyperlinks (i.e., network ties). An example is the copurchase network on Amazon.com, where recommended products are listed under the title “Customers who bought this item also bought.” Accordingly, we define the product that explicitly recommends additional products as the recommender, and products in the recommended set as the recommended products.

2 It is noteworthy that robustness checks using data from notebook computers, mobile phones, and hair-care products from different stores also reaffirm our findings.

3 Please refer to the robustness checks for more details.

4 Consistent with prior literature [20, 25, 45], WOM valence refers to the average rating of consumer reviews.

5 Consistent with prior literature [20, 25, 45], WOM volume refers to the cumulative number of consumer reviews.

6 We sincerely thank our anonymous reviewer for offering the insight that higher WOM valence of a product may imply the higher quality of customer service associated with the product, and thus may offset the impact that higher WOM valence results in a larger subsequent decrease in the WOM valence. However, as subsequently discussed in the data description, each Tmall retailer is in charge of the sales and customer service for all the products in the retailer’s store. Thus, the quality of customer service should remain relatively stable across different products in the same store. Therefore, if customer service is included as an explanatory variable to our main econometric model (i.e., fixed effects model), the estimate will be statistically omitted.

7 In our empirical analysis, we have ruled out several alternative explanations (e.g., the ceiling effect of WOM valence, the confounding effects of valence variance) to validate this hypothesis. Please refer to the robustness checks for more details.

8 In our empirical analysis, we have ruled out several alternative explanations (e.g., the ceiling effect of WOM volume, the product age) to validate this hypothesis. Please refer to the robustness checks for more details.

9 Although we focus on one major category (i.e., digital cameras) as the focal products in our empirical analysis, we also report robustness checks on the sensitivity of this operationalization by including all the related product categories (e.g., battery, lens) in our sample and find consistent results. Moreover, we also collect data on notebook computers, mobile phones, and hair-care products, and obtain similar findings. The results are reported as robustness checks.

10 For robustness checks, we also report on the sensitivity of this operationalization by organizing our data at the product-week level and find consistent results.

11 This is aggregated from the transaction records to indicate the total quantity of product i sold in day t.

12 This is shown on Tmall product web pages to indicate the sales quantity of product i during the past month prior to day t.

13 This is shown on Tmall product web pages to indicate the available quantity of product i for sale in day t.

14 This is shown on Tmall product web pages to indicate the cumulative number of product i’s web page bookmarked by consumers in day t.

15 This measures the number of products pointing to product i in the network in day t.

16 This measures the average sales quantity of products pointing to product i in the network in day t.

17 This measures the average list price of products pointing to product i in the network in day t.

18 The estimation results based on different time lag levels are consistent with those from our one-day lag models.

19 We sincerely thank our anonymous reviewers for pointing out these limitations.

Additional information

Notes on contributors

Zhijie Lin

Zhijie Lin is an assistant professor in the Department of Marketing and Electronic Business, School of Business at the Nanjing University. He received his Ph.D. in information systems from the National University of Singapore. His research interests focus on economics of information systems, electronic commerce, and social media. He has published his work in journals such as MIS Quarterly, Information Systems Research, and Decision Support Systems, and in the proceedings of conferences such as the International Conference on Information Systems.

Cheng-Suang Heng

Cheng-Suang Heng (corresponding author; [email protected]) is an associate professor in the Department of Information Systems, School of Computing at the National University of Singapore. He received his Ph.D. in organization, technology, and entrepreneurship from Stanford University. His research interests focus on organization strategies, with an emphasis on electronic commerce and social media. He has published papers in journals such as MIS Quarterly, Information Systems Research, Journal of the AIS, and in the proceedings of conferences such as the International Conference on Information Systems.