Using web data and services : technology, theory and evidence

Abstract

Many firms and individuals have been publishing data and services on the Web. It is necessary to develop advanced technology facilitating the use of Web data and services and to understand what impacts on firms and individuals would be. This thesis, composed of three essays, aims to explore (1) what technology could be developed to facilitate using Web data and services, and (2) what theoretical mechanisms are driving the impact of using Web data and services. The first essay describes an advanced technology for using Web services and the other two essays present some theoretical mechanisms and empirical evidences about how consumers are influenced by the data published on commercial webpages. The first essay presents a classification of the data misinterpretation problems that may occur when composing Web services. After the problem scope is identified, it proposes an approach to automatic detection and reconciliation of data interpretation conflicts in Web services composition. To validate and evaluate the approach, the first essay describes a prototype and demonstrates the approach can significantly alleviate the reconciliation efforts for Web services composition. The second essay explores how herding and social media Word of Mouth (WOM) drive product sales when commercial websites disclose the sales data in real-time on the product pages and integrate with social-networking platforms (e.g., Facebook, Twitter). Using a panel data set consisting of about 500 deals from Groupon.com, the second essay shows both herding and Facebook-mediated WOM lead to additional product sales, whereas Twitter-mediated WOM has no significant impact on sales. More importantly, it documents that herding and Facebook-mediated WOM are complements in driving sales. Given the fact that many commercial websites integrate with social-networking platforms and the importance of social media endorsements, the third essay investigates if online review ratings would affect consumers' decisions of endorsing via Facebook and purchasing products. It builds a stylized Bayesian learning model and derives three hypotheses. The empirical findings largely support the hypotheses. In particular, the results show that a favorable valence of online reviews causes to increase consumers' social media endorsements and the estimated effect is greater when the variance in the review ratings is larger. Moreover, the findings reveal that consumers exhibit different behaviors when they consider endorsing versus purchasing products.