Liye Ma. Carnegie Mellon University Tepper School of Business, Posner Room Forbes Avenue Pittsburgh, PA 15213

Transcription

1 Liye Ma, Posner Room Forbes Avenue Phone: Cell: (401) Office: (412) Web: EDUCATION, Pittsburgh, PA Ph.D. in Industrial Administration Machine Learning Department M.S. in Machine Learning Brown University, Providence, RI Sc.M. in Computer Science Tsinghua University, Beijing, China B.E. in Computer Science 2011(expected) 2011(expected) RESEARCH INTERESTS Dynamic Consumer Decisions, Internet/Social Media Marketing, Hi-Tech Marketing Choice Models, Structural Models, Empirical IO, Analytical Models DISSERTATION Essays on Network Mediated Marketing Interactions Committee: Kannan Srinivasan (co-chair) Kinshuk Jerath Yaroslav Kryukov Baohong Sun (co-chair) Ramayya Krishnan Alan Montgomery My dissertation focuses on network mediated marketing interactions. Technology advancement has led to ever increasing interactions of consumers and firms. Such interactions often occur in and are mediated by networks, which present an exciting opportunity for marketing research. In my final essay and job market paper, A Dynamic Competitive Analysis of Content Production and Link Formation of Internet Content Developers, I study how Internet content developers compete for viewership through producing content and linking to one another. Hundreds of revenue sharing content websites have greatly contributed to the recent proliferation of social media. Content at these websites is supplied by external developers, whom the websites attract through revenue sharing. This leads to a competition for viewership among developers at a website. A feature recently introduced at many sites, namely allowing developers to link to one another, brings intriguing interactions to this competition. I develop a dynamic oligopoly model, following the framework of Ericson and Pakes (1995), to investigate the interaction of production and linking decisions of content developers, the tradeoffs they face over time, and the resulting market structure. The model is estimated using the data obtained 1

2 from a popular Internet product review site, applying the two-step estimator recently developed by Bajari, Benkard, and Levin (2007). I find that reciprocal links are naturally encouraged by a promote-the-promoter effect, which then induces certain developers to strategically initiate non-reciprocal links, in anticipation of reciprocation. I find that the prospect of receiving incoming links can either encourage or discourage content production, depending on the situation a developer is in. Furthermore, I find that although both more content and higher network position increase viewership, only the latter leads to higher net benefit once cost is accounted for. This suggests that linking may impede competition, by giving competitive advantage to a subgroup of content developers, and policy simulation suggests that limiting links could lead to higher overall viewership at the website. In the second essay, Homophily or Influence? An Empirical Analysis of Purchase within a Social Network, I study consumer s product purchase in a social network environment. Consumers who are close to one another often behave similarly, attributable to either their intrinsic similarity, i.e. an unobserved homophily effect, or their influence on one another. Teasing out the two factors is important for target marketing. In this study, I use a dynamic model to incorporate both factors, where identification is achieved through separating static product taste from dynamic influence arising from communications, and estimate it using a dataset obtained from a large Indian telecom company, which contains information on repeated purchase and communication. I find strong homophily effect in consumer s product choice decision. In contrast, the purchase timing decision is heavily influenced by others. I show that ignoring either effect will lead to an overestimation of the other. Furthermore, I show that detailed communication data is crucial for accurately evaluating influence effects. In the first essay, A Position Paradox in Sponsored Search Auctions, I study how quality differentiated firms bid for online advertising positions when they explicitly account for consumers knowledge structure and search behavior. This study uses a game theoretic approach. Features such as click-through-rate and value-per-click, often taken as exogenous in extant literature, become endogenous in this setting. Modeling at this lower level of primitive reveals a interesting position paradox, where a low quality firm bids higher than a high quality firm, is placed higher, yet still receives fewer clicks. Analysis shows that explicitly modeling of consumer search is essential for understanding the key drivers of auction outcome, including residual demand, i.e. the amount of clicks a firm can receive at lower position, incremental value, i.e. the additional value a higher position offers, and differential cost, i.e. the different cost implication of informed consumers to both firms under pay-perclick. UNDER REVIEW / WORKING PAPERS A Dynamic Competitive Analysis of Content Production and Link Formation of Internet Content Developers (Job Market Paper) 2

3 A Position Paradox In Sponsored Search Auctions, with Kinshuk Jerath, Young-Hoon Park and Kannan Srinivasan (revise and resubmit at Marketing Science) Homophily or Influence? An Empirical Analysis of Purchase within a Social Network, with Ramayya Krishnan and Alan Montgomery (to be submitted to Management Science) The Effect of Pre-Release Piracy on Movie Box-Office Sales, with Alan Montgomery, Param Vir Singh and Michael Smith (to be submitted to Journal of Marketing Research) Voluntary Open-Source, with Pei-yu Chen and Tridas Mukhopadhyay (to be submitted to Management Science) WORK IN PROGRESS A Dynamic Structural Analysis of Content Contribution with Recognition Effects, with Baohong Sun (model setup completed, estimation in progress) Quality Control and User Participation in Online Communities, with Onur Kesten and Tridas Mukhopadhyay (analysis completed, manuscript under preparation) TEACHING EXPERIENCE Instructor Management Information Systems (Undergraduate Core Course, instructor rating 4.38/5) Teaching Assistant Optimization for Interactive Marketing (MBA) Microeconomic Theory (PhD) Information Technology Management (MBA) TEACHING INTERESTS Marketing Management, Pricing, Marketing Strategy, Optimization for Interactive Marketing CONFERENCE PARTICIPATION INFORMS Marketing Science Conference, Cologne, Germany Presented A Position Paradox in Sponsored Search Auctions Attended Doctoral Consortium Workshop on Information in Networks, New York, NY Presented Homophily or Influence? An Empirical Analysis of Purchase within a Social Network INFORMS Marketing Science Conference, Ann Arbor, MI Statistical Challenges in e-commerce Research, Pittsburgh, PA Quantitative Marketing and Economics, New York, NY

4 HONORS AND AWARDS AD-HOC REVIEWER PROFESSIONAL EXPERIENCE William Larimer Mellon Fellowship Graduate Student Research Grant Review of Economics and Statistics Information Systems Research Wellington Management Company LLP, Boston, MA Development Team Lead (Responsible for the development work of the portfolio accounting and client administration systems of the firm. Managed a team of employees and external consultants.) Oracle Corporation, Redwood Shores, CA Senior Project Leader (Led various software development projects for the Oracle E-Business Suite, focusing on manufacturing and supply chain management applications.) COMPUTER SKILLS C, C++, Java, Perl, R, SAS, Matlab, Mathematica, Python, SQL GRADUATE COURSEWORK Course Marketing and Economics Advanced Data Analysis Advanced Choice Models Analytical and Structural Models in Marketing Bayesian Statistics and Marketing Review of Marketing Literature Behavioral Foundation of Marketing (planned) Advanced Behavioral Methods (planned) Microeconomic Theory I Microeconomic Theory II Econometrics I Econometrics II Econometrics III Contract Theory Game Theory Mathematics for Economists Statistics and Mathematics Real Analysis Intermediate Statistics Regression Analysis Statistical Machine Learning Instructor Baohong Sun Baohong Sun Kannan Srinivasan Alan Montgomery Peter Boatwright Joachim Vosgerau Joachim Vosgerau Steve Spear Paul Healy George-Levi Gayle Holger Sieg Ronald Goettler Onur Kesten Isa Hafalir Iulian Obreja Jack Schaeffer Matthew Harrison Kathryn Roeder Larry Wasserman, John Lafferty 4

5 Others Machine Learning Multimedia Database and Data Mining Graduate Algorithms Social Networks (audit) Seminar in Information Systems Eric Xing Christos Faloutsos Gary Miller David Krackhardt Pei-yu Chen REFERENCES Kannan Srinivasan (co-chair) Rohet Tolani Distinguished Professor in International Business H.J. Heinz II Professor of Management, Marketing and Information Systems Phone: (412) Baohong Sun (co-chair) Carnegie Bosch Professor of Marketing Phone: (412) Ramayya Krishnan H. John Heinz III Dean, Heinz College, W. W. Cooper and Ruth F. Cooper Professor of Information Systems Phone: (412) Alan Montgomery Associate Professor of Marketing Phone: (412)

6 Abstracts of Selected Working Papers A Dynamic Competitive Analysis of Content Production and Link Formation of Internet Content Developers (Job Market Paper) The emergence of hundreds of revenue sharing content websites has greatly contributed to the proliferation of Internet social media. Content at these websites is supplied by external independent developers, whom the websites attract through revenue sharing. This leads to a competition among developers, as each tries to attract viewership to her own content. A feature recently introduced at many sites, namely allowing developers to link to one another, leads to even richer interactions in this competition, and its impact on content production and overall website viewership is little understood. In this study, we develop a dynamic oligopoly model for the competition among content developers at a website. Each developer produces content and forms links to maximize her discounted viewership net of cost of actions, and their strategic interaction is characterized as a Markov-perfect equilibrium. Applying the two-step estimator of Bajari, Benkard, and Levin (2007) to the data obtained from a popular Internet product review site, we investigate the following issues: (1) why and how do developers form links? (2) Will linking encourage or discourage content production? (3) What market structure will emerge? (4) Will linking increase or decrease the overall website viewership? We find that reciprocal links are naturally encouraged by a promote-the-promoter effect. This in turn induces developers with more content to strategically initiate links to invite reciprocation. In addition, we find that link formation affects the incentive to produce content developers with more content and unfavorable network positions are encouraged to produce, while developers in the opposite states are discouraged. Furthermore, the current linking policy may impede competition by giving competitive advantage to a subgroup of content developers, and policy simulation suggests that limiting links could increase overall viewership by 17%. Our study is among the first to examine the interdependence between online link formation and contention production in a dynamic and competitive setting. A Position Paradox in Sponsored Search Auctions with Kinshuk Jerath, Young-hoon Park and Kannan Srinivasan We study the bidding strategies of vertically differentiated firms that bid for sponsored search advertisement positions for a keyword at a search engine. We explicitly model how consumers navigate and click on search results based on their knowledge and beliefs about firm qualities. Our model yields several interesting insights and a main counter-intuitive result we focus on is the "position paradox." The paradox is that a superior firm may bid lower than an inferior firm and obtain a position below it, yet still obtain more clicks than the inferior firm. Under a pay-perimpression mechanism, the inferior firm wants to be at the top where more consumers click on its link, while the superior firm is better off by placing its link at a lower position as it pays a smaller advertising fee but some consumers will still reach it in the search of a higher-quality firm. Under a pay-per-click mechanism, the inferior firm has an even stronger incentive to be at the top since now it only has to pay for the consumers who do not know the firms' reputations 6

7 and, therefore, can bid more aggressively. As the quality premium for the superior firm increases, and/or if more consumers know the identity of the superior firm, the incentive for the inferior firm to be at the top may increase. Furthermore, the position paradox occurs even when the search engine weights the firms' bids by using quality scores. We also derive implications for search engine profit and find that it prefers the pay-per-impression or the pay-per-click mechanism under different conditions, based on which mechanism leads to a more aggressive lower bid. We find strong empirical support for the position paradox based on a data obtained from a major search engine firm in Korea. Homophily or Influence? An Empirical Analysis of Purchase within a Social Network with Ramayya Krishnan and Alan Montgomery Consumers who are close to one another in a social network are known to have similar behaviors. The focus of this study is the extent to which such observed similarity is driven by homophily or social influence. Homophily refers to the similarity in product preferences between individuals who are connected. Social influence is the dependence of consumers purchase decisions on their communication with others. We construct a hierarchical Bayesian model to study both the timing and choice of consumer purchases within a social network. Our model is estimated using a unique social network dataset obtained from a large Asian telecom operator for the purchase of caller ringer-back tones. We find strong social influence effects in both the purchase-timing and product-choice decisions of consumers. In the purchase-timing decision, we find that consumers are three times more likely to be influenced by network neighbors than by other people. In the product-choice decision we find a strong homophily effect. We show that ignoring either homophily or social influence will result in overestimated effects of the other factor. Furthermore, we show that detailed communication data is crucial for measuring influence effect, and influence effect can be either over- or underestimated when such data is not available. Finally, we conduct policy simulations on a variety of target marketing schemes to show that promotions targeted using network information is superior. For example, we find a 4-21% improvement on purchase probability, and an 11-35% improvement for promoting a specific product. 7