Abstracts

Can You Gig It? An Empirical Examination of the Gig-Economy and Entrepreneurial Activity
Gordon Burtch, Seth Carnahan and Brad Greenwood

We examine how the entry of gig-economy platforms influences local entrepreneurial activity. On one hand, such platforms may reduce entrepreneurial activity by offering stable employment for the un- and under-employed. On the other hand, such platforms may enable entrepreneurial activity by offering work flexibility that allows the entrepreneur to re-deploy resources strategically in order to pursue her nascent venture. To resolve this tension we exploit a set of natural experiments, the entry of the ride-sharing platform Uber X and the on-demand delivery platform Postmates into local areas. We examine the effect of each on crowdfunding campaign launches at Kickstarter, the world’s largest reward-based crowdfunding platform. Results indicate a negative and significant effect on crowdfunding campaign launches, and thus local entrepreneurial activity, after entry of Uber X or Postmates. Strikingly, the effect appears to accrue primarily to un-funded and under-funded projects, suggesting that gig-economy platforms predominantly reduce lower quality entrepreneurial activity by offering viable employment for the un- and under-employed.

Analyzing Content and Customer Engagement in Social Media with Deep Learning
Donghyuk Shin, Shu He, Gene Moo Lee and Andrew Whinston

In the present study, we investigate the effect of social media content on subsequent customer engagement (likes and reblogs) using a large-scale dataset from Tumblr. Our study focuses on company-generated posts, which consist of two main information sources: visual (images) and textual (text and tags). We employ state-of-the-art machine learning approaches including deep learning to extract data-driven features from both sources that effectively capture their semantics in a systematic and scalable manner. With such semantic representations, we develop novel complexity, similarity, and consistency measures of social media content. Our empirical results show that proper visual stimuli (e.g., beautiful images, adult-content, celebrities, etc.), complementary textual content, and consistent themes have positive effects on the engagement, and that content demanding significant concentration levels (e.g., video, images with complex semantics, text with diverse topics, complex sentences, etc.) have the opposite effects. This work contributes to the literature by exemplifying how unstructured multimedia data can be translated into insights. Our framework for semantic content analysis, particularly for visual content, illustrates how to leverage deep learning methods to better model and analyze multimedia data for effective marketing and social media strategies.

Stimulating User-Generated Content via Performance Feedback: A Randomized Mobile Field Experiment
Yili Hong, Bin Gu, Gord Burtch, Ni Huang, Chen Liang, Kanliang Wang, Dongpu Fu and Bo Yang

This study investigates whether and how a platform’s provision of performance feedback to users about the quality of their prior content contributions (user-generated content, or UGC) can help to stimulate users’ subsequent content contributions. We place a particular focus on the framing of feedback messages, considering three alternatives: pro-social (e.g., “your content benefited x other users”), pro-self (e.g., “you are in the top x% of users based on the attention your content has received”), and competitive (e.g., “you beat 1-x% other users based on the attention your content has received”). We draw on social value orientation (SVO) theory to hypothesize how these alternative framings may impact users’ likelihood of producing additional content, and we hypothesize how user gender may determine this response. To evaluate our hypotheses, we partnered with a major mobile crowdsourcing platform based in China to conduct a randomized field experiment involving the delivery of feedback messages with randomly determined framings, via mobile push notifications. By exogenously and randomly varying the message framing across users, we are able to identify a clean causal effect on each user’s subsequent contributions of UGC. We find that feedback framed either pro-socially or pro-self has a positive effect on UGC contributions, in general, whereas feedback framed competitively has no such effect. Additionally, we observe differences across genders, such that the positive effects of pro-socially and pro-self framed feedback are significantly stronger for female users. In contrast, competitively framed feedback is only effective for male users. Our findings demonstrate that platform-provided performance feedback can help to stimulate UGC and further that, when doing so, it is important for the platform to account for user characteristics when crafting these messages.

How Questions and Answers Shape Online Marketplaces: The Case of Amazon Answer
Warut Khern-Am-Nuai, Hossein Ghasemkhani and Karthik Kannan

This paper uses data from two online shopping platforms to investigate the economic implications of the question & answer system. This research problem becomes increasingly important as many small and medium players in online marketplaces start to adopt this system rather than the typical online review system. Yet, the economic implications of such a Q&A system have not been studied in the previous literature. We employ the difference-in-differences approach to empirically examine the effect of question & answer elements, which exist only on one platform, on product sales. Interestingly, we find that controlling for everything else, question elements have a negative impact on sales while answer elements, particularly the depth of the answers, have a positive impact on sales. However, as we focus on the initial sales, it turns out that the number of questions and the fraction of questions that have at least one answer positively influence the sales. We also find that there is an interaction between Q&A elements and review elements in that an increase in the number of questions seems to be positively correlated with an increase in the number of reviews in the following period. Meanwhile, an increase in the number of answers appears to reduce the average length of reviews in the subsequent period. Our findings suggest that incorporating the question & answer system could be a potential approach to drive sales. However, it is crucially important for managers to develop appropriate policies to gather necessary answers to questions asked on the platform in order to capitalize on such a system.

Demystifying Online Reviews of Physicians
Danish Saifee, Eric Zheng and Indranil Bardhan

Current trends indicate that patients who take an active role in managing their health also search and use information obtained from online reviews and ratings of physicians. However, online reviews may not accurately capture or serve as a proxy for physician quality, as measured by their patients’ outcomes. There is a dearth of research on the underlying measures of physician competence, since there are no clearly agreed upon measures that serve as proxies for the quality of care delivered provided by physicians. Our research focuses on three key questions: (1) what factors determine online ratings and reviews of physicians? (2) are online reviews and ratings related to physician quality? and (3) are online reviews and physician quality measures associated with patients’ future clinical outcomes? Drawing on a large data set of COPD patients across a ten-year period, we construct several metrics to proxy for physician competence. We operationalize perceived physician quality by calculating the sentiment score and variance, based on the online reviews of physicians collected from a major health website. We find that physicians with higher competence and poorer online reviews tend to exhibit better patient outcomes, in terms of future readmission rates and ER visits. We also find that online ratings are not significantly associated with physician competence.

Altruism or Shrewd Business? Implications of Technology Openness on Platform Innovations and Competition
He Huang, Geoffrey Parker, Yinliang Tan and Hongyan Xu

In today’s highly competitive business environment, a growing number of platforms are opening their technologies. This leads us to wonder whether sharing one’s proprietary technology is altruism or a shrewd business move. In this paper, we study the incentive of why firms share their proprietary technology with their competitors. In contrast to previous literature focusing on the network effect, our study reveals a novel explanation for why firms are willing to open their technologies. Essentially, there are two effects accompanied by the decision of technology openness. On the one hand, technology sharing commitment leads to information disclosure (i.e., information role). The competitor will perceive the other firm’s technology level whether she decides to adopt or not. On the other hand, opening one’s technology might contribute to technology access (i.e., access role). That is, the competitor has the opportunity to exert learning costs to absorb the other firm’s technology. The decision regarding technology openness critically depends on the trade-off between the information role and access role. More importantly, the interaction of these two roles is moderated by the magnitude of learning cost. We find that when the learning cost of acquiring new technology is low, given openness, the access role dominates the information role, which leads to technology closeness as the equilibrium for both firms. However, when the learning cost is substantial, the access role dissipates and technology openness can benefit the firm through the information role. This is due to the fact that openness can alleviate the unwarranted innovation competition caused by the technology uncertainty from the closeness strategy. We also discuss the impact of technology openness on the degree of innovation and find that openness does not necessarily lead to higher innovation. We further explore several extended models and show that our intuition of the base model is robust and enriched.

Online Word-of-Mouth Spillover Effects in the U.S. Automobile Industry
Yen-Yao Wang and Anjana Susarla

Online Word of Mouth (WOM) is an importance aspect of consumer-firm relationship and is a leading indicator of product performance. However, prior research focuses considerably on the static view of online WOM and its impact on online sales of non-durable or low-involvement products. The objective of this paper is to explicate the dynamics of the spillover effects between online WOM of the focal firm, online WOM of its competitors, and offline car sales (durable and high-involvement product) of the focal firm in the U.S. automobile industry. Given that the test drive is the critical aspect of the pre-purchase product evaluation for automobile purchases, I focus specifically on customers’ test drive experience as a measure of online WOM on three dimensions at the document level: volume, sentiment, and eight emotions embodied in online WOM. Using unique dataset from around 1000 different social media platforms for 32 car brands from 2009 to 2015, I employed the Bayesian approach using Markov Chain Monte Carlo (MCMC) methods for model estimation. The results suggest that there is a pressing need for extending to the dynamic view of online WOM by examining the spillover effects. The implications and future plans are discussed.

Network structure and patterns of information diversity on Twitter
Jesse Shore, Jiye Baek and Chris Dellarocas

Social media have great potential to support diverse information sharing, but there is widespread concern that platforms like Twitter do not result in communication between those who hold contradictory viewpoints. Because users can choose whom to follow, prior research suggests that social media users exist in “echo chambers” or become polarized. We seek evidence of this in a complete cross section of hyperlinks posted on Twitter, using previously validated measures of political slant to study information diversity. Contrary to prediction, we find that the average account posts links to more politically moderate news sources than the ones they receive in their own feed. However, members of a tiny network core do exhibit polarization and are responsible for the majority of tweets received overall due to their popularity and activity, which could explain the widespread perception of polarization on social media.

Impact of Easier Store Access on Customers’ Online Purchase Behavior
Anuj Kumar, Amit Mehra and Subodha Kumar

Different mechanisms through which easier access to physical stores could affect customers’ online purchase behavior are not well understood in the literature. We use customer-level data of a large apparel retailer to estimate the treatment effect of store openings on the online purchase behavior of its existing customers. We find that retailer’s store openings resulted in increase in the online purchases from its existing customers. We propose two mechanisms that could explain these results: (1) store engagement effect - higher customers’ engagement with the retailer due to higher store interactions and (2) store return effect - reduced risk of online purchase due to the low cost option of store returns. We provide direct empirical evidence of these mechanisms on our field data. We further show that these effects are caused by reduction in customers’ distances from the retailer’s store due to the store openings.

For Whom to Tweet? Evidence from a Large-Scale Social Media Platform
Zaiyan Wei and Mo Xiao

We study the effects of peer-group sizes on content generating and sharing in a large-scale and influential social media platform. User-generated content, particularly tweets in social media, disseminates information and exerts social influence. However, 50% of the users in this platform post less than 6 tweets per month and contribute to less than 15% of the total tweets in stock, while the top 10% post on average 40 tweets a month and contribute to more than half of the tweets in stock. We attribute the highly unbalanced contribution to a user's conflicting incentives of free-riding and maximizing social influence. We exploit the asymmetry of a user's peer groups (followers and followees, groups of people following and being followed by the user respectively) to disentangle these incentives, and devise empirical strategies to deal with the endogenous formation of one's networks. We find asymmetric effects, both in signs and sizes, of followers and followees on content generating and sharing. A larger group of followers leads a user to tweet more, while a larger group of followees leads a user to tweet less. As the follower effects dominate the followee effects in size, our simulations indicate that the platform could increase the number of total tweets by 25% if it randomly adds 1% new links to existing links. Targeting occasional tweeters is even more effective in promoting the activeness of this social media platform.

Complements and Substitutes in Product Recommendations: The Differential Effects on Consumers’ Willingness-to-pay
Mingyue Zhang and Jesse Bockstedt

Product recommendations have been shown to influence consumers’ preferences and purchasing behavior. However, empirical evidence has yet to be found illustrating whether and how the recommendations of other products affect a consumers’ economic behavior for the focal product. In many e-commerce websites, a product is presented with co-purchase and co-view recommendations which potentially contain complement and substitute products, respectively. Very little research has explored the differential effects of complementary and substitutable recommendations. In this study, we are interested in how the type of recommendations of other products impact the consumers’ willingness-to-pay for the focal product, and additionally how the recommendations’ price and the consumers’ decision stage moderate this effect. We conducted a 2x2x2 randomized experiment to examine how the consumers’ willingness-to-pay is affected by these factors. Experimental results provide evidence that there is no significant main effect difference between complementary and substitutable recommendations. But we observed a significant interaction effect between recommendation type and decision stage, which highlights the importance of timing in recommender systems. Other findings include that consumers are willing to pay more for a specific product when the price of a recommended product is high, as well as when they are in later decision stages. These findings have significant implications for the design and applications of recommender systems.

The Impact of IT on Production Interdependence
Fengmei Gong, Barrie R. Nault and Zhuo Cheng

Information technology (IT) has improved information sharing and coordination within and between industries and firms. As a result, industries have become increasingly integrated with their suppliers’ business processes, implementing operations strategies such as just-in-time production. However, whether industries and firms within a supply chain have become more interdependent is an open question. We examine the impact of an industry’s IT investment on its production interdependence with upstream suppliers, where we measure interdependence as direct backward linkage (DBL), and examine the relationship among DBL, total factor productivity (TFP), and value-added. Using estimation models based on production theory and growth accounting theory, and employing two industry-level datasets covering 1987 to 2013 in the U.S., we estimate the impact of IT on DBL and the impact of DBL on value-added through TFP. We find that an industry’s IT investment reduces its production interdependence with suppliers and leads to greater value-added after the commercialization of the Internet. Specifically, from 1994 onwards an industry’s IT increases the efficiency with which it uses intermediate inputs from suppliers, resulting in a reduction in DBL; the reduction in DBL indicates a shift in production frontier towards more efficient production, and leads to growth in TFP and greater value-added.

How Do Complementors Respond to the Threat of Platform Owner Entry? Evidence from the Mobile App Market
Wen Wen and Feng Zhu

How do complementors respond to the threats of platform owner entry, and how do such responses differ from the responses to actual entry? Using the mobile platform Android as our research setting, we examine how app developers on Android adjust their innovation efforts and prices in response to Google’s entry threat and actual entry into the Android app markets. Based on a difference-in-differences empirical framework, we find that on average, app developers that are affected by Google’s entry threat reduces their innovation efforts under entry threats; once Google actually enters, they reduce innovation efforts further and also increase app prices. However, we find evidence that developers of top apps tend to expand their innovation efforts under entry threats.

Does Give-and-Take Really Matter? Dynamics of Social Interactions in Social Network
Sunghun Chung, Animesh Animesh, Kunsoo Han and Alain Pinsonneault

Despite the increasing attention paid to the social interaction in online social networks, it is still not clear how social media users interact with each other, consume different content, and expand their social network. This study conceptualizes two types of user engagement (internal and external) and empirically examines the dynamics between user’s engagement, friends’ engagement, and network size. Using detailed social media activity data collected from over 20,000 Facebook users for three years, we find that when people externally engage in their friends’ social space rather than one’s own space, they can make more friends and also receive friends’ engagement in one’s own social space. However, when people receive more friends’ engagement in their social space and make more friends, they are likely to reduce their engagement in social media (both externally as well as internally). Our findings can provide useful insights for the literature on social ties, user-generated content, and online peer influence.

Impact of Information Technology on Energy Productivity: An Empirical Investigation
Jiyong Park, Kunsoo Han and Byungtae Lee

Increasing energy productivity has been emphasized as an important way to reconcile the dual objectives of maintaining economic growth and reducing energy consumption. Employing a production function framework, we examine the impact of information technology (IT) on energy productivity during 1998-2014 using U.S. economy-wide data from 60 industries. Our findings suggest that IT contributes to energy productivity in the overall economy, though it is mainly driven by software. Separate analyses of economic and environmental performances reveal that this positive contribution results from both increased output and decreased energy consumption, highlighting the role of IT as a catalyst for enhancing energy productivity, beyond its role as a driver of productivity growth. Further, we find that the impact of IT on energy productivity is significant in low IT-intensity industries, but not in high IT-intensity industries, due to poor environmental performance of IT in the latter. Implications for research and practice are discussed.

IT Capability, Competitive Actions, and Firm Performance: Evidence from U.S. Firms
Inmyung Choi, David Cantor and Joey George

This research proposes to investigate the effect of information technology (IT) of a firm on its competitive actions and profitability drawing upon competitive dynamics literature. It is a timely and important issue since we have several inconsistent results about the contribution of IT on firm profitability. This research examines the impact of a firm’s IT (i.e., IT capability and IT capability gap) on its competitive actions. Then, this study seeks how competitive actions are associated with a firm’s profitability. In this study, IT capability and IT capability gap are operationalized based on a text-analysis approach. This research makes a first step towards developing a comprehensive view on how IT enhances firm profitability by supporting the exercise of competitive actions based on a large panel data set. Specifically, we support that IT capability and IT capability gap increase the competitive actions of a firm. Additionally, competitive actions enhance firm profitability. The results emphasize the need of considering competitive actions as a key factor to examine the value of IT in developing a firm’s competitiveness and complement prior studies, which focused on the relationship among IT, organizational capabilities, and firm performance.

The Potato Salad Effect: The Impact of Competition Intensity on Outcomes in Crowdfunding Platforms
Hilah Geva, Ohad Barzilay and Gal Oestreicher-Singer

Crowdfunding platforms serve as intermediaries in two-sided markets, bringing together entrepreneurs on the supply side, and potential investors on the demand side. From an economic perspective, fundraising campaigns compete with one another for investors’ money. We use a quasi-natural experiment, which occurred on Kickstarter.com, to identify how supply-side competition intensity affects campaign performance, while controlling for temporal trends and seasonal effects. We consider three metrics of performance: campaign success rate, amount of money pledged per campaign, and revenue per successful campaign. As expected, overall, an increase in competition intensity significantly decreases the revenue of successful campaigns. However, we show that a campaign’s intrinsic quality moderates the effect of competition intensity. Specifically, for low-quality campaigns, all three performance measurements are highly susceptible to the negative effect of competition intensity. Among high-quality campaigns, in contrast, none of the performance measures are affected by competition intensity. We discuss managerial implications for entrepreneurs.

Incentive Provision and Pro-Social Behaviors
Dandan Qiao, Shun-Yang Lee, Andrew Whinston and Qiang Wei

Individuals’ pro-social behaviors are driven by altruistic and selfish motivations. In this paper we explore how the introduction of external incentives would influence one’s pro-social behavior both in the short term and in the long run. Using a large data set on Amazon product reviews, we design a quasi-experimental approach where we combine a propensity score matching (PSM) and a difference-in-differences (DiD) method to empirically study the effect of incentive provision on reviewer’s behavior. We apply techniques from linguistics, language processing, and machine learning to propose several novel measures to capture reviews’ writing style and quality. We find evidences consistent with reciprocity, crowding-out, and overjustification effects. Our study contributes to the understanding of pro-social behavior and sheds light on how incentives would shift individual behavior.

Who's a Good Decision Maker? Data-Driven Expert Worker Ranking Under Unobservable Quality
Tomer Geva and Maytal Saar-Tsechansky

Evaluation of expert workers by their decision quality has substantial practical value, yet using other expert workers for decision quality evaluation tasks is costly and often infeasible. In this work, we frame the Ranking of Expert workers according to their unobserved decision Quality (REQ) -- without resorting to evaluation by other experts -- as a new Data Science problem. This problem is challenging, as the correct decisions are commonly unobservable and substantial parts of the information available to the decision maker is not available for retrospective decision evaluation. We propose a new machine learning approach to address this problem. We evaluate our method on one dataset representing real expert decisions and two public datasets, and find that our approach is successful in generating highly accurate rankings. Moreover, we observe that our approach’s superiority over the baseline is particularly prominent as evaluation settings become increasingly challenging.

Three-Part Tariff No Better than Two?
Hemant Bhargava and Manish Gangwar

Three-part tariffs (3PT) are widely used, for selling goods to compensating workers to taxing citizens. They’re particularly common in IT industries, such as telecommunications, software and web services. To a two-part tariff (which has a fixed or access fee F, and a usage fee s), a 3PT adds a free allowance Q, up to which the marginal price is zero (a 2PT forces Q = 0). Intuitively, this additional instrument should lead to higher profit, enabling the firm to segment between low types (for whom only the allowance matters) and high types (for whom the over-allowance usage fee kicks in). This paper presents a starkly counter-intuitive result: that a 3PT is no different than a 2PT in a very wide class of settings. Specifically, when the description of user heterogeneity fits a log-concave distribution (or one with non-decreasing hazard rate, or when market demand elasticity is non-decreasing in price), then the optimal 3PT has an equivalent 2PT (i.e., a 3PT with Q − 0) which produces identical outcomes for each participant in the market. This result demolishes a classic economic rationale for the free allowance. Conversely, when either (a) there is a non-economic rationale for the allowance (i.e., a user bias) or (b) the distribution does not satisfy log-concavity, then a 3PT can strictly be more profitable than a 2PT. While developing these results, the paper also makes innovations in solving the 3PT optimization problem, which is extremely challenging due to non-linearities and degeneracy.

Dynamic Pricing in E-commerce Setting
Minghong Xu and Siddhartha Bhattacharyya

We propose a Dynamic Pricing model to capture the new features of e-commerce settings. The pricing decisions are made based on market information, i.e. customer rating and sales ranking, as well as the traditional inventory consideration. We address the pricing decision questions in both single-period planning and multi-period planning. Results on implementation of our proposed model on real Amazon antivirus product data show that our pricing policy increases sales revenue, and that determining the length of planning period is critical in success of dynamic pricing policy. The MapReduce solution on Bellman Equation suggested in this paper will also help to solve the Big State Space, the dimensionality hurdle suffered by Dynamic Programming.

The Effect of Direct Marketing on Online Purchase: An Empirical Study
Xingyue Zhang and Yuliang Yao

The Internet has paved new avenues for firms to interact with consumers, either to deliver promotional marketing messages or to expand consumer purchase channels. Online marketing and online purchase, as well as interactions between marketing channels and purchase channels have gained increasing attention. Among them, the impact of online marketing on online and/or offline purchases has been intensely studied. Yet the effect of offline marketing on online sales has not been much investigated. Using direct marketing call data from one of the largest classified advertisements websites in China, we empirically analyze the effect of offline direct marketing on online purchase. We find a positive effect of direct marketing on online purchase, however this effect diminishes as the number of direct marketing calls increase. Offline direct marketing calls also demonstrate a “carry-over” effect that offline marketing calls also influence the future probability of online purchase.

Improving Weaning Decision Making with Association Rule Based Category Weighted Naïve Bayes
Yuanyuan Gao, Anqi Xu, Paul Hu and Tsang-Hsiang Cheng

Mechanical ventilation is an invasive intervention commonly used in the intensive care unit to assist patients’ respirations. Physicians’ weaning decision making that entails whether to remove a patient from the respiratory support of mechanical ventilation is essential. Effective weaning decisions improve patient care and well-being; however, ineffective decisions create severe consequences that require significant resources to overcome. Data-driven techniques can support physicians’ weaning decision making and its use needs to addresses several challenges inherent to weaning decision making, including an incomparable distribution of the target variable and feature overload. We propose a novel weighting method capable of addressing both challenges by using synthetic minority over-sampling to increase and learn the minority-class instances, and then incorporating an association rule network to extend a category weighted Naïve Bayes model for assigning differential weights to various feature categories, thus mitigating feature overload. We empirically examine the proposed method and compare its performance with those of several prevalent benchmark techniques that include one sided selection, threshold cost-sensitive technique, MetaCost, RELIEF algorithm, and forward search. Overall, our results show that the proposed method outperforms all benchmark techniques in terms of accuracy, precision, recall, and F measure.

Procurement Policies for Mobile-Promotion Platforms
Manmohan Aseri, Milind Dawande, Ganesh Janakiraman and Vijay Mookerjee

Mobile-Promotion platforms such as Cidewalk (www.cidewalk.com) and Sitescout (www.sitescout.com) enable advertisers (individual users or businesses) to directly launch their personalized mobile advertising campaigns. These platforms contract with advertisers to provide a certain number of impressions on mobile apps in their desired sets of geographic locations (usually cities or zip codes) within their desired time durations (for example, a month); the execution of each such contract is referred to as a campaign. In practice, campaigns arrive dynamically over time and the platform bids in real-time at an ad exchange to win mobile impressions arising over the desired sets of locations of these campaigns to fulfill their respective demands. Our analysis in this paper offers a rolling-horizon procedure in which the platform periodically recomputes its procurement (or bidding) policy and its policy for allocating the impressions (that have been won) to the various campaigns. For the basic problem of the rolling-horizon procedure, our main result is an effective procurement and allocation policy. Specifically, we obtain a theoretical bound on the performance of our policy and demonstrate the attractiveness of this bound for realistic values of the problem parameters, estimated using data from Cidewalk. By simulating a realistic setting of dynamically-arriving campaigns, we show that the computational time for the rolling-horizon procedure is reasonable for real-world implementation.

Who's Watching TV?
Jessica Clark, Jean-Francois Paiement and Foster Provost

Understanding the demographic makeup of TV shows' audiences is of vital concern to advertisers and other stakeholders. Such knowledge is traditionally learned using data sources such as Nielsen which measure viewership at the individual person level and report aggregate numbers. Viewership data that is now available at the individual Set-Top Box (STB) level has led to current state-of-the-art audience estimation algorithms, but there is a crucial weakness in how viewers are measured: it is impossible to tell with certainty which person is the one watching TV in multi-person households. This work introduces and formulates the problem of estimating which person is watching, which to our knowledge has not been approached in the existing literature. We develop a novel framework for estimating the likelihood that each household member in a multi-person household is watching. This method sits in the intersection of multi-instance learning and domain adaptation and leverages the single-person STBs' probabilities of watching and combines them in different ways. We also formalize two state-of-the-art industry heuristics for this problem that are currently used. A final contribution is the development of a suite of evaluation techniques for assessing the effectiveness of our model against the baseline heuristics, since the actual viewership is unknown. The solution has broad applications within the television advertising industry, as well as to any situation where multiple people share the same device or account but individual inferences are desired. Further, a major TV provider is planning on deploying this method for use in their TV ad-targeting system. No personally identifiable information (PII) was gathered or used in conducting this study. To the extent any data was analyzed, it was anonymous and/or aggregated data, consistent with the carrier's privacy policy.

The Digital Distribution Strategy in Emerging E-book Markets: An Empirical Investigation
Kyunghee Lee, Kunsoo Han, Eunkyoung Lee and Byungtae Lee

In mature e-book markets, releasing an e-book together with the corresponding print book is the norm. However, in “emerging” e-book markets, where e-books still account for a fraction of total book demand, publishers are faced with important strategic decisions, such as whether to release a digital version for a given print book, and when to release the digital version. Due to the paucity of empirical research on e-books – especially in emerging markets, little guidance is offered to publishers regarding these important decisions. To fill this gap, we investigate (1) whether e-book releases boost or hurt the demand for the corresponding print book, and (2) whether or not the timing of e-book releases matters to the relationship between e-book releases and print sales. Further, we scrutinize the moderating role of two content characteristics: content popularity (captured by print sales before e-book releases) and “e-readability,” defined as the extent to which the content of a book is suitable for e-reading (captured by the time required to finish reading a book). By employing a difference-in-differences model with entropy balancing and a large-scale book-level sales dataset from South Korea, we find that e-book releases enhance print book sales by 2.1%. We also find significant moderating effects of content characteristics. The positive effect of e-book releases on print sales is stronger for books with lower content popularity and lower e-readability. Further, we find that the influence of e-book release timing is significantly moderated by content popularity and e-readability. Delaying e-book releases strengthens the positive effect of e-book releases on print sales for books with higher content popularity and higher e-readability. This study provides theoretical explanations on how the new electronic channel can complement the existing channel in the publishing industry, as well as important managerial implications for publishers operating in emerging e-book markets.

"Reconciling the Diversity-Relevance Dilemma" — A Multi-Category Utility Model of Consumer Response to Content Recommendations
Yicheng Song, Nachiketa Sahoo and Elie Ofek

Diversity of a set of recommendations, in addition to its relevance to a consumer’s preference, affects the consumer’s satisfaction with the personalized recommender system. However, the personalized recommendation approaches that incorporate diversity risk recommending less relevant items. We address this diversity-relevance dilemma in the context of online sessions at media sites by proposing a multi-category utility model of content consumption. This model provides a framework to learn a consumer’s preferences towards different categories of content—which in turn govern the likely response to recommended content. Unlike existing approaches for personalized recommendation, ours provides a way to estimate how quickly consumers satiate while consuming in each category and how they substitute one category of content with another. The model also captures how consumers trade off relying on their own costly search effort for new content vs. relying on those proposed by a recommendation engine. Taken together, these three elements enable us to characterize how utility maximizing consumers construct diverse bundles of content over the course of each session and how likely they are to click on content recommended to them. We estimate this model using a clickstream dataset from a large international media outlet and apply it to predict the content that a consumer will select at different stages of an online session. We demonstrate that by taking into account how consumers form bundles from different categories to maximize their utility in a given session we can not only make more diverse recommendations, but also more relevant recommendations than those that do not incorporate such information.

New Entry Threats and Firm Performance: On Examining the Moderating Role of Board Capital
Yang Pan, Peng Huang and Anandasivam Gopal

A significant part of the fast-moving dynamics in the high-tech industries is due to the high rate of new entry in the form of entrepreneurial ventures. Leveraging a novel measure of new entry threats using text mining, we test the conjecture that threats from new entry lead to deterioration in operational performance. In addition, we hypothesize that the relationship is moderated by corporate board capital, a proxy of the board’s ability to bridge outside resources. Our evidences show that a higher level of new entry threats indeed leads to an incumbent’s performance deterioration. Interestingly, we find that the direction of the moderating effects of board capital depends on its nature. Board capital breadth, connecting firms with diversified external resources, mitigates the negative impact of new entry threats. Board capital depth, referred to as the embeddedness of the board in the focal firm’s industry, exacerbates the relationship between new entry threats and firm performance. We discuss the implications for research and practices.

The Role of Digital Capabilities in Converting Inventor Team Expertise to Knowledge Capital for Medical Device Innovation
Liwei Chen and Arun Rai

Medical device innovation increasingly needs inventor teams that not only have specialized expertise but also are diverse in multiple knowledge domains. Using a multi-level lens (patents by firms), this study examines how digital capabilities can empower inventor teams to solve the dilemma between broadening knowledge capital via diverse expertise and deepening knowledge capital via specialized expertise. We conceptualize multi-dimensional digital capabilities for innovation development and synthesize literatures on IT-enabled innovation and IT strategy to inform the development of our hypotheses. We constructed a multisource panel dataset by linking data from multiple sources, including the University of California (UC) Berkeley Patent Database, Computer Intelligence Technology (CI) Database, COMPUSTAT and CRSP databases. Our study enriches the literature on digital capabilities by developing and empirically validating a theoretical conceptualization of digital capabilities for innovation development in general and for medical device innovation in specific. The results shed light on how Innovation Development Digital Capabilities empower inventor teams to convert their diverse or specialized expertise into broad and deep knowledge capital and facilitate knowledge production in terms of patent innovation.

Strategic Complementarities in an Online Advertising Supply Chain
Anitesh Barua, Genaro Gutierrez and Changseung Yoo

Research in online advertising offers advice on how to increase the click-through rate (CTR) and/or conversion rate (CR) of a campaign by examining the interactions between a specific advertising channel and consumers while putting less emphasis on the interactions between the channels used in the campaign. Another stream of literature studies synergies across various forms of online advertising, e.g., display advertising and search, but do not focus on the internal supply chain structures leading to such complementarities. Aided by a proprietary dataset from a campaign executed by a publicly traded, multinational digital advertising agency, we explore the vertical (intra-channel) and horizontal (inter-channel) interactions in the online advertising supply chain, and analyzes the impact of interactions between channel structures and pricing models on the advertising agency’s decisions and performance. We show that there are quantifiable vertical interactions and horizontal synergies, and that the failure to account for such interactions may lead to overspending on some actions and underspending on others. Specifically, our results indicate that incorporating such interactions and synergies in decision making increases supply chain profit by 44% over the status quo. Moreover when we devise feasible information and profit sharing schemes, the supply chain profit more than doubles, and gets closer to the profits of a theoretically ideal, but practically infeasible, fully integrated supply chain. Our results also rationalize a positive economic role for intermediation by the digital advertising agency; by combining information from multiple channels in its decision making, and structuring contracts appropriately, the agency enables the supply chain to achieve higher levels of efficiency that would be impossible to attain if the advertising channels act individually. This goes beyond the current economic rationalization of the agency based on economies of scale and transaction costs, which led the agency to an organizational structure focused on vertical media buying. Our findings indicate that the agencies should instead be organized by campaigns in order to monetize the substantial benefits derived from cross-channel information sharing and decision making.

Does Position Matter More on Mobile? Ranking Effects across Devices
Alvin Zuyin Zheng, Ting Li and Paul Pavlou

Achieving a better rank online is often costly. Is the effect of ranking different for mobile and PC? This study empirically examines the ranking effect across different device types in an e-commerce environment. With over 4 million observations from Tweaker.net, the largest shopbot in Netherlands, we estimated the ranking effect between mobile and PC. Surprisingly, and contrary to prior findings, our results across different model specifications consistently show that ranking effect is smaller on mobile devices. This study extends the understanding about the effect of position in e-commerce context by empirically examining the ranking effect across devices. This study has important managerial implications for retailers and e-commerce platforms. As the ranking effect is smaller on mobile devices, retailers should take account of the source of traffic (mobile or PC) while bidding for a particular position. And platforms should consider the different ranking effects on different channels.

Multi-level Examination of IT-enabled Change in Healthcare Service Provision: The Interplay between Social Structure and Psychological Safety
Roopa Raman and Varun Grover

Electronic medical records and associated decision support technologies are significantly transforming the delivery of healthcare services. However, adapting to IT-enabled change is difficult, and often the service improvement promise of these IT-enabled transformations is not realized. We examine how social interaction patterns (networks) enabling the utilization of knowledge received from others can enable adaptation to IT-enabled change in healthcare service delivery. We focus on medication administration in hospitals, which is a key healthcare service spanning multiple interdependent providers in hospitals and has significant quality implications for healthcare service delivery. In a multi-level analysis of IT-enabled change in healthcare service delivery in a large urban hospital setting, we examine adaptation to IT-enabled change produced by two contrasting network structures that vary on their levels of connectivity and trust. We propose that these structures would improve service performance in two kinds of content networks, demand (where knowledge is solicited by the adapting individual) and supply (where knowledge is passively received). Our study finds that network structure and content work together to engender confidence in the ability to take risks and adapt to IT-enabled change. The results offer useful guidelines on how knowledge networks can be configured to enable better adaptation to IT-enabled change and consequently superior performance.

Selection Bias with Linear Probability Models
Suneel Babu Chatla and Galit Shmueli

Selection bias arises when the observed sample does not represent the population. It causes the estimates obtained from the sample to be biased. Self-selection bias is a specific type of selection bias in which the individuals self-select whether to be in the sample or not. It has been a primary challenge in the Information Systems (IS) field. There exist two notable approaches to correct for the selection bias: 2stage least square (2SLS) which use probit regression in the first stage and Propensity Score Matching (PSM) methods which use logistic regression (logit) to estimate propensity score. Linear probability model (LPM) - linear regression models applied to a binary outcome – is an alternative to probit or logit models. It has certain advantages over both logit and probit models, especially with large samples. Based on our extensive literature search, we find that, surprisingly, usage of LPMs is rare in the IS literature, where logit and probit models are typically used for binary outcomes. Due to the recent advances in data collection methods, large samples are now becoming popular in IS studies. Therefore, there is a need to use LPM which has clearly some advantages in the large samples. LPMs have been examined with respect to specific aspects, but a thorough evaluation of their practical pros and cons for different research goals under different scenarios is missing. We performed an extensive simulation to evaluate LPM against its alternatives under different study settings with different sample sizes. We find that, for 2SLS approach, LPM is a viable alternative to the probit model and clearly has an advantage of providing consistent estimates. Our results indicate that there is a lot of scope for the LPM and its use is not just confined to the issue of selection bias. It may be used for the other issues such as instrumental variable models which are also common in IS studies. We also plan to evaluate the LPM further as an alternative to the logistic regression in PSM studies by expanding our simulation settings and also using some real world datasets.

Salience Effect in Crowdsourcing Contests
Brian Lee, Sulin Ba, Xinxin Li and Jan Stallaert

With the emerging popularity of crowdsourcing contests, more firms have started to host crowdsourcing contests as a part of their product innovation process. In this study, we examine the role of systematic bias such as salience effect in affecting the outcome of crowdsourcing contests and how this salience effect is moderated by the number of contestants as a result of parallel path effect and competition effect. Our result suggests that the salience effect does have an impact on the performance of the contestants, even including the winners of the contests. Furthermore, the parallel path effect cannot completely eliminate the salience effect, but is able to alleviate the salience effect. In contrast, the competition effect is likely to amplify the salience effect. These results have important implications to the contest hosting firms and crowdsourcing platform designers.

Discovering Contextual Information from User Reviews for Recommendation Purposes
Konstantin Bauman and Alexander Tuzhilin

The paper presents a new method of systematically discovering relevant contextual information from the user-generated reviews and, therefore, converting the latent contextual information into the explicit representational view in order to provide better recommendations to the users when the reviews complement traditional ratings used in recommender systems. In particular, we classify all the user reviews into the "context rich" specific and "context poor" generic reviews and present a phrase-based and an LDA-based methods of extracting contextual information from the "specific" reviews. We also show empirically on the Yelp data that, these two methods extract almost all the relevant contextual information from the reviews across three different applications.

Case Study: TV Ads and Search Spikes for DraftKings and FanDuel
Rex Du, Kenneth Wilbur and Linli Xu

This paper reports ongoing research into how, when and why online search responds to TV ads. We quantify these effects in a case study of brands DraftKings and FanDuel. We find highly significant online search spikes in response to TV ads, positive spillovers across brands, and substantial variation in the height of those spikes across dayparts, TV networks, program genres, ad slots and creatives.

Moral Hazards and Effects of IT-enabled Monitoring Systems in Online Labor Markets
Chen Liang, Bin Gu and Yili Hong

This paper investigates how IT-enabled monitoring systems mitigate moral hazard in an online labor market and their effect on market competition. We exploit a quasi-experiment at Freelancer when it introduced an IT-enabled monitoring system in 2015. Using a large dataset including 36407 projects, we adopt a difference-in-differences (DID) approach to identify the treatment effect of the monitoring system on employer contract choice, market competition, and employer surplus. We find that the IT-enabled monitoring system lowers employers’ preference for high-reputable contractors, and thus reduces the reputation premiums. We further compare the effect of the IT-enabled monitoring system on projects vulnerable to moral hazard (hourly projects) versus that on projects resistant to moral hazard (fixed-price projects). We find that the IT-enabled monitoring system raises employer surplus in hourly projects by 0.109 and increases the number of bids by 15.7 percentage. Our result suggests that IT-enabled monitoring systems have a significant effect on alleviating moral hazards, reducing agency costs, and facilitating market competition.

Reviving Order Online: The Effect of Purchase Feature in Social Media Mobile Apps
Chenhui Guo, Bin Zhang, Xi Chen and Paulo Goes

As a new transaction channel, purchase feature in social media mobile apps enables consumers to pay merchant accounts directly. The convenience of mobile payment, the richness of product information and promotions may trigger more demand. The aim of this study is to quantify the effect of such purchase feature on businesses’ sales revenue. Using transaction data from a large hotel chain company, we rely on variations in the timing of users enabling purchase feature in the mobile app to identify the effect of adopting the mobile channel on hotel room reservation. To mitigate endogeneity caused by non-random selection of being adopters and non-adopters, we apply difference-in-differences approach with propensity score matched individuals. Further, we utilize count data model and various instrumental variables estimations to strengthen the validity of our empirical analyses. Our results provide evidence that adoption of purchase feature in mobile social media apps would significantly increase sales of the business. More importantly, a majority of the increase is attributed to more transactions from the new social media mobile channel with purchase feature, while the remaining fraction of increase is caused by channel switching.

The Effects of Optimal Matching in Social Support in an Online Weight Loss Community
Lu Yan and Yong Tan

Online healthcare communities are well-known for exchanging social support. In this study, we argue that social support may not always lead to good outcomes. Specifically, we differentiate support providers and support seekers, and examine whether the balance of needed and received social support affects individuals’ weight loss outcomes. By analyzing a group of individuals participating in an online weight loss community, we found that overprovision (receiving more support than is desired) and underprovision (receiving less support than is desired) resulted in differentiable health outcomes. In addition, by categorizing social support into different types, we found evidence suggesting that the match between the types of social support requested and provided also influenced individuals’ performance in the weight loss process. These findings have implications for maximizing the usefulness of social support for participants in the online environment as well as for clinicians who refer individuals to online weight loss communities and those who design them.

Determinants of Matching in Online Labor Markets: A Structural Two-Sided Matching Model
Jing Gong

In the past decade, IT has facilitated the shift from permanent employment to need-based outsourcing and from local labor market to global online labor markets. While prior studies have examined how global frictions affect employers’ hiring decisions on online labor markets, we have limited understanding of the inter-dependence between workers and employers and the economic impact of IT-enabled globalization on matching outcomes such as the number of matched projects, freelancer wages, and project values generated from matching. This study is an attempt to fill in the gap by examining the dual roles of IT-enabled globalization, i.e., (1) in determining the formation of matches between employers and freelancers, and (2) in affecting market outcomes. From a market perspective, we take into account two-sided decision making, competition on each side, complementarities between employer and freelancer attributes, and endogenous money transfers between employers and freelancers. In our empirical analysis, we estimate a structural two-sided matching model of the online labor market from a revealed preference perspective. The estimation is based on a dataset from a major freelancing website that connects freelancers and employers from more than 200 countries. We then conduct counterfactual analysis to quantify the economic impact of IT-enabled globalization in online labor market by comparing the current scenario with a counterfactual scenario where employers can only match with freelancers from the same country. The results from our estimation suggest that employers tend to match with freelancers from the same country, and that employers from developed countries tend to match with freelancers from developing countries. The results from the counterfactual analysis suggest that IT-enabled globalization leads to more employers and freelancers with successful matches, lower average wage among matched freelancers, and higher total project values generated on the market.

When all Products are Digital: Complexity and Intangible Value in the Ecosystem of Digitizing Firms
Pouya Rahmati, Ali Tafti, J. Christopher Westland and Cesar Hidalgo

Using a complex adaptive network setting, we examine how the integration of digital complexity into products and services affects the intangible value (Tobin’s q) of firms. We begin with the premise that if two products are related because they require similar resources and capabilities, they will tend to be produced in tandem, whereas dissimilar goods are less likely to be produced together. Accordingly, we develop the construct of digital complexity as proximity to software within a dynamic network of complementary products and services, spanning the years 1990 to 2014. We use this network to measure digital complexity at the firm level. Using a variety of panel regression methods, our analysis shows a positive relationship between firms’ digital complexity and their intangible value. Our results are also corroborated with a set of alternative firm performance measures. The identified relationship is robust for various model specifications that control for firm fixed-effects, mitigating the potential for endogeneity, and a list of possible confounding factors. Our findings suggest that digital complexity helps explain how firms develop sustainable value that is intangible, imperfectly mobile, and hence, difficult for competitors to replicate. Digital complexity, as such, is not a monetary investment or commodity, but rather is to be understood as a form of socio-technological complexity reflecting firms’ unique combination of intangible resources and capabilities that are difficult to imitate.

Managing Supply Base Complexity through All-in-One Intermediaries: IT-enabled Capabilities and Economic Outcomes
J.J. Po-An Hsieh, Arun Rai, Sean Xin Xu and Zhitao Yin

Supply base theory suggests a dilemma between transaction cost and supply risk associated with supply base complexity. Supply base complexity is particularly salient in global sourcing networks where buyers face high costs of identifying and selecting suppliers from a large pool of manufacturers and also must evaluate the risk due to low technology sophistication of some low-cost service providers and an unstable infrastructure. We identify all-in-one (AIO) online intermediaries, which leverage IT-enabled capabilities to manage supply base complexity, as a solution to break the tension between transaction cost and supply risk. Our empirical study concerns an online global chemical products sourcing AIO intermediary that provides all-in-one B2B transactions. Based on a unique and rich data set from the AIO intermediary, we find that IT-enabled capabilities for managing supply base can increase transaction profit while decreasing profit volatility, thus breaking the tension highlighted by supply base theory. Moreover, the impacts of IT-enabled capabilities on profit and profit volatility, respectively, increase and decrease as supply base complexity elevates.

Configuring Distributed Software Development Projects: Examining Multidimensional Contingencies and Equifinality
Narayan Ramasubbu and Indranil Bardhan

Using a unique dataset of more than 1800 commercial projects, we systematically examine the multidimensional contingencies resulting from work dispersion, software process diversity, and process governance choices of distributed software development teams that influence their performance. We utilize a combination of a general interaction approach and cluster analysis to examine the bundles of tradeoffs present in various project configurations. This approach helps to integrate prior conceptualizations of contingency such as fit as moderation, fit as deviation, and fit as system, and we demonstrate the different forms of equifinality regimes present among distributed project configurations. Finally, we utilize a difference-in-difference analysis using data drawn from a natural experiment that occurred at the research site to illustrate how an organizational policy that favors decentralization and provides more latitude to software teams for determining their context-specific project configurations improves their overall performance.

Ethics, Bounded Rationality and IP Sharing in IT Outsourcing
Krishnan Anand and Manu Goyal

Our dynamic model of IT outsourcing integrates (i) its major impediments-- incomplete contracts, moral hazard and adverse selection, and (ii) its major facilitators-- ethics, Intellectual Property (IP) sharing and reputations, under both perfect and, more realistically, bounded rationality. Under bounded rationality, an ethical firm-- which honors contractual obligations irrespective of legal restraints-- can obtain strictly greater profits than a profit-maximizing firm unconstrained by ethics, even when (i) the unconstrained firm has access to a superset of the ethical firm's strategies, and (ii) the ethical firm is unable to reveal its ethical commitment to its contracting partner. Further, we find a novel explanation, rooted in ethics, for IP sharing: IP sharing arises as a strategic imperative for firms, even when it lowers their profits. Our results explain why (a) a commitment to ethics can boost profits, (b) IP sharing is widespread, and (c) IT outsourcing is booming despite several formidable impediments.

Modeling User Engagement in Mobile Content Consumption with Tapstream Data and Field Experiment
Yingjie Zhang, Beibei Li, Xueming Luo and Xiaoyi Wang

Low engagement rate and high attrition rate have been formidable challenges for the long-term success of mobile apps, especially for apps whose revenues mainly come from in-app purchases. To date, little is known towards how companies can comprehensively identify user engagement stages so as to improve business revenues. This paper proposes a structural econometric framework to model consumer latent engagement stages by accounting for both the time-varying nature of engagement and consumer forward-looking consumption behavior. We instantiate our study by analyzing fine-grained mobile tapstream data with a total of 570,514 records on 4,454 consumers’ continuous content consumption behaviors in a popular mobile reading app for approximately three months in 2015. Our results enable us to tailor optimal pricing strategy to each consumer based on the model-detected engagement stages. Interestingly, we found such engagement-specific pricing strategy leads to lower average price for consumers and higher overall business revenues for the app simultaneously. We furthermore test the causal impact and generalize the effectiveness of our engagement detection method with randomized field experiments. Results from the field experiments add more causal evidence that personalized promotion strategy targeting user engagement stages could improve the overall business performance with lower costs to users. These findings from both structural models and field experiments are nontrivial and suggest potential welfare improvement in the mobile app market, with respect to the crucial role of modeling user engagement.

An Empirical Analysis of Interaction Effect between Pay-for-Performance Advertising and Price Discounting in Online Trading Platforms
Zike Cao, Junhong Chu, Kai-Lung Hui and Hong Xu

Understanding the interaction effects among multiple marketing activities is of paramount importance for firms in planning optimal "marketing mix". The development of ITs has brought forth a novel and increasingly popular advertising format: pay for performance. Based on a data set obtained from Taobao.com, the world's largest online consumer-to-consumer (C2C) trading platform, we investigate the interaction effects between performance-based advertising and price discounting, a frequently-used marketing strategy in both offline and online marketplaces, on sellers' sales performances in this study. Specifically, two forms of performance-based advertising -- sponsored search and affiliate program -- are examined. Significant negative interaction effects between performance-based advertising and price discounting are detected. The possible reason could be that the signaling strength of performance-based advertising is attenuated by price discounting, which itself acts as a negative quality signal. These results offer novel and important implications for firms' optimal planning of marketing mix in the current digital age when online advertising is becoming increasingly predominant.

Asymmetric Value Gains in Business Process Outsourcing Relationships
Sukruth Suresh and Ravichandran T.

The paper examines the drivers of differences in value gains between clients and vendor firms following the announcement of a business process outsourcing announcement. We argue that these differences are influenced by the extent of risk involved in a BPO undertaking for the client and the vendor’s ability to appropriate value out of the BPO relationship. In addition to the risk and ability of the stakeholders to extract value, we also expect the vendor and client size and the contract duration to play a moderating role in driving the difference in value gains. We test our hypothesis on a sample of 221 BPO announcements made between the years 2000-2013. We find that the potential loss of competency for the client of the task outsourced, the vendor’s operating scale and scope to be significant drivers of the difference in abnormal returns between the client and vendor firms. We also find that the client and vendor sizes play a moderating role in influencing the difference in the abnormal returns.

Professional versus Amateur Images: Investigating Differential Impact on Airbnb Property Demand
Shunyuan Zhang, Dokyun Lee, Param Vir Singh and Kannan Srinivasan

How much do property images taken by professionals impact the demand of an Airbnb property? This study answers this question. Using a unique longitudinal dataset of approximately 17000 Airbnb properties over 4 months, we find that properties with professional images are 9% more frequently booked. We find that the impact of professionally taken images is due to better reflection of quality of the property in the image (due to lower noise) and the added trust as an Airbnb professional visits the property. We find that high end properties benefit more from professionally taken images than low end properties. We further find that as more properties in a neighborhood get their properties photographed by professionals, the demand for properties with high quality pictures in a neighborhood increases but the demand for properties with low quality pictures in the neighborhood decreases. The results overall reveal that properties with professionally taken images not only bring more demand to Airbnb overall but also steal demand away from properties with pictures taken by amateurs.

Windows of Acceleration and Disconformity: Competitive Reaction to Technological Innovation
Hyunwoo Park and Rahul Basole

We study how product assortment of a service provider can be leveraged to induce supply-side competition on product innovation, which in turn may lead to a virtuous self-sustaining advantage to the service provider. We argue that a service provider's product assortment choice can provide an implicit signal to supplying manufacturers and thus influence the path of technology evolution without explicit guidelines on product specifications. We situate our investigation in the mobile handset industry where the iPhone triggered an industry-wide technology disruption. Exploiting asymmetric adoption timing of the disruptive product for different service providers, we use the difference in differences method to estimate the impact of the disruptive product on a manufacturer's product feature development and deployment decisions. We find that manufacturers first avoid competition by clinging on to old product categories, but eventually pursue head-to-head competition in the long run. Moreover, in response to the disruptive product, competing manufacturers supply their technologically superior products to the first-mover service provider over an exclusive contract period. Manufacturers particularly focus on product features in which the disruptive product excelled when improving their products. This asymmetric competition-induced product enhancement decision of manufacturers persists even after expiration of the exclusive contract. Robustness checks involving narrower time window, placebo events, and prior-event comparison corroborate our findings. Our study builds on the extant literature on strategic product design for channels and contributes to the understanding on strategic implications of disruptive technology adoption.

Altruism Pays! Towards Optimal Call-to-Action for Online Referral: A Randomized Field Experiment
Jaehwuen Jung, Ravi Bapna, Joseph Golden and Tianshu Sun

Information sharing through online WOM has become increasingly important for businesses. Despite the popularity of online referral programs, little is known about how firms can optimally design call to action to encourage referrals, as well as the motives underlying those referrals. In collaboration with a large US based online platform specialized in photo processing, we conduct a large randomized field experiment involving 100,000 customers to identify the causal effect of three types of call to action for referral (egoistic, equity and altruistic) that are widely used in practice. Our experiment shows that, surprisingly, ‘altruistic’ call to action leads to the highest likelihood of referral and best referral outcomes. Such altruistic framing is more effective for customers who had repeated purchases in the past and who reported higher Net Promoter Score. Also, we find that the effect of altruistic framing decays fast after customer’s purchase. In this way, our study provides direct managerial implications to firms on the optimal design of call to action for referral campaigns (how, to whom and when to send call to action for referral). We also show that altruism is an important driver of online referral among customers and how such motive leads to referral decision and referral outcomes. Finally, we discussed the key differences and complementarity between call for referral and call for purchase, and offer guidance on firm's integrated marketing communication strategy.

Motivating Mobile App Adoption: Evidence from a Large-scale Randomized Field Experiment
Tianshu Sun, Lanfei Shi, Siva Viswanathan and Elena Zheleva

Prior literature has established a positive association between mobile app adoption and customers’ purchase behavior. However, it is not clear whether firms can actively influence customers’ mobile app adoption and increase their purchases. Using a randomized field experiment involving 250,000 customers, we investigate i) whether and how a firm can motivate customers to adopt mobile apps and ii) the causal effect of induced mobile app adoptions on customers’ purchase behaviors. We find that i) both providing information and monetary incentive can lead to significant increase in customers’ mobile app adoption, with the relative increase 447% and 146%, respectively; ii) the effect of mobile app adoption varies greatly depending on how customers are motivated. Although providing monetary incentives may lead to larger increase in mobile app adoption, such induced adoption does not result in more purchases in the long run. In contrast, providing information leads to effective mobile adoption that sustainably increases customers’ purchases, and overall profits of the firm. We further examine customers’ multichannel browsing and purchase behavior and find additional evidence for how induced app adoptions affect customers’ purchase behaviors. In summary, by leveraging a randomized field experiment, our study provides actionable insights for firms designing interventions to motivate effective mobile adoptions.

When Streams Come True: Estimating the Impact of Free Streaming Availability on EST Sales
Uttara Ananthakrishnan, Michael Smith and Rahul Telang

The rise of online digital platforms has caused a massive increase in the number of viewers who consume entertainment via streaming. To cater to this audience, television networks have started making their content instantly available on both paid and free online streaming platforms after broadcast. However, some managers now question what impact these free streaming channels have on consumption in paid channels, such as digital Electronic Sell Through. On one hand, free streaming and EST sales could be seen by consumers as highly differentiated products in terms of the time period for which they are available, the presence or absence of advertisements, and customer platform loyalty. On the other hand, free streaming could complement EST sales or directly substitute for EST sales by the virtue of being of the same digital form. In this paper, we empirically analyze if free streaming on a major television network’s online platform cannibalizes the sales on paid channels. To do this, we use a unique dataset provided by a leading television network in the United States. We exploit the natural variation in the online streaming schedules of a prominent television show in our identification strategy. Using a difference in difference approach we find that free streaming cannibalizes EST sales by about 8.4%.

Towards Improved Education for Students of Low Socioeconomic Status: Learning Analytics of Massive Open Online Courses (MOOCs)
Sang Pil Han, Mi Hyun Lee, Sunghoon Kim and Sungho Park

Although the new era of free, online learning unfolds, the claim of ‘education for all’ appears to be overshadowed by the concern over the unequal use of Massive Open Online Courses (MOOCs hereinafter). MOOCs may not be a solution to students across all levels of educational backgrounds and economic status. Using the datasets collected from 126 countries on learners’ experiences and academic outcomes at a MOOC course and on their demographic information, we examine the effectiveness of two intervention strategies to improve educational achievements and engagements among low socioeconomic status (SES) MOOC learners: (1) monetary precommitment and (2) mobile media. In our empirical setting, we operationalize the ‘monetary precommitment’ by observing whether students enrolled into a verified track with $49 fee during the first two weeks. The verified track allows students to receive an official transcript from the hosting university if they complete the course and pay for the credit later, thereby helping students to commit throughout the course. We operationalize the ‘mobile media’ by measuring whether students have used any type of mobile devices in their learning activities. We expect that using mobile devices enables students to attend MOOCs anytime/anywhere, thus helping them enjoy great flexibility in accessing to the lecture material. In our empirical analyses, we quantify the impact of each of two interventions on outcomes and engagements among low SES learners compared to high SES learners. We employ a linear regression system with propensity score matching to control for potential self-selection bias of our two intervention variables. Results indicate that both intervention strategies – monetary precommitment through costly verified track enrollment and mobile media use – are viable solutions for stimulating educational achievements and engagements among MOOC learners with low SES in the US and around the world. Our results suggest that monetary and technological interventions studied in this research can be effective means for MOOC platforms and policy-makers to reduce educational disparities with respect to SES in MOOCs and other online education courses both in the US and around the world, ensuring educators and policy-makers to expand access to education to everyone through MOOCs.

More than Just Money: Educational Impact of Online Charitable Crowdfunding
Qiang Gao and Mingfeng Lin

Public funding for education has been dwindling in recent years in the US. An interesting emerging source of education funding is online donation-based crowdfunding, where teachers can raise funds for class-related projects. Unlike traditional donations or public sources of funding, online crowdfunding requires teachers to exert significant efforts to persuade strangers online to donate. In addition, the teachers’ identities are directly associated with the projects, therefore they are more likely to feel personally accountable for those funds. We posit that these differences can lead to better use of the funds, and therefore result in positive impacts on student performance more than just the financial aspect of funds. We empirically test this conjecture by exploiting the geographical expansion of DonorsChoose.org, the largest education-purpose donation crowdfunding site. Our results show a positive impact of these donations on classroom performance, especially when teachers are required to disclose more information about themselves and are therefore more accountable. These findings not only show the offline impacts of online crowdfunding, especially in the domain of public goods, but also have implications for the management of traditional donations for education purposes.

Choice of IT Governance Mode in Multi Business Firms: Effect of Technology Embeddedness
Pankaj Setia, Taha Havakhor and Mohammad Rahman

Information technologies have enabled growth of large firms with multiple business units spread across geographies and markets. Governance of IT infrastructure is a key strategic priority for managing digitally-enabled operations across these firms. Specifically, firms either govern the IT infrastructure centrally or decentralize the decision-making to business units. Making this choice is challenging. Offering foundation platforms, technologies enable (or limit) business units’ capability for digitizing business operations. Using an option-based perspective, we argue that firms’ assessment of new IT options requires considering various future scenarios. An effective governance model facilitates assessment of these scenarios. In general, firms consider costs of coordination and responsiveness, while choosing centralized or decentralized mode of IT infrastructure governance. Previous research examines how the characteristics of business unit or central headquarters, or the relationship between the two, influence this choice of a governance mode. However, network effects, due to similarities of technologies across business units, have been overlooked in the previous research. In this research, we argue that agglomeration effects influence the choice of a governance mode. Identifying these agglomeration effects, we assess if embeddedness of technologies–software and hardware–across business units influences firm’s choice of IT infrastructure governance mode for the unit. Specifically, we propose and test if firms centralize IT infrastructure governance for a business unit that has greater hardware and software embeddedness in the network of firm’s business units. Study also assesses the moderating effect of organizational size on these relationships. We test relationships using data from CI Technology Database and additional data from SDC Platinum Mergers and Acquisitions, Lexis-Nexis corporate affiliation, and COMPUSTAT. We test various econometric time series models, to test our hypotheses. Study discusses the academic contributions and practical implications of our findings.

Firm Size Distribution Goes Online: The Evolution of eBay Firms’ Sales Distribution
Sagit Bar-Gill, Erik Brynjolfsson and Nir Hak

The size distribution of firms is an important indicator of market concentration. Studying the distribution of sales for eBay’s commercial sellers, this paper provides the first analysis of a firm size distribution (FSD) at an online market, further examining its evolution, and the underlying growth rate patterns. The evolution of eBay’s FSD is characterized by increasing mass in its right tail, yet it remains better-fitted by a lognormal than by a power law distribution throughout our ten-year period of analysis. This is in line with a possible convergence towards a power law in the long run, while currently market concentration is lower compared to power law FSDs found in many traditional industries. Sellers’ growth rate patterns obey Gibrat’s Law only among the subset of top selling firms.

Intelligent Transportation Systems and Traffic Congestion: Evidence from US Cities
Zhi Cheng, Min-Seok Pang and Paul Pavlou

Despite substantial investments in transportation infrastructure, traffic congestion in urban areas has not abated throughout the globe. Intelligent transportation systems (ITS) have recently emerged as an alternative approach to expensive road construction. To investigate the effect of ITS on traffic congestion, we utilize a combined unique data set on traffic and ITS adoption in 99 US urban areas from 2001 to 2008. The results from fixed-effects estimations supplemented by a series robustness checks show that ITS can substantially reduce traffic congestion, saving drivers time and money and reducing carbon emissions. Our results seek to extend the IS literature on IT value creation in the public sector and the broader socioeconomic effects of IT. Our study also contributes to the transportation economics literature by showing that ITS reduce traffic congestion, informing transportation policymakers that IT could be a more cost-effective and environmentally-friendly way to tackle traffic congestion than expensive road expansion.

Information Disclosure and Crowdfunding: An Empirical Analysis of the Disclosure of Project Risk and Market Reaction
Keongtae Kim, Jooyoung Park, Pan Yang and Kunpeng Zhang

Since information asymmetry between funders and creators is a critical issue in crowdfunding, many strategies have been introduced to dampen it and make markets more efficient and sustainable. In this research, we examine one such possible mechanism, namely a platform-wide rule to require the disclosure of project risk, to assess whether and how disclosed risk information affects the behaviors of market participants. We examine this question on a popular crowdfunding site that has implemented a policy to mandate the disclosure of potential risk about projects. We find that the introduction of the new disclosure policy decreased the creation of new projects, even successful ones. Our additional analyses show that the effect is stronger for creators bearing a larger burden from the new disclosure requirement. Our online experiments imply that the decrease is partly driven by the negative perceptions of funders about the risk information disclosed, which discourages them from participating on the platform. We also conduct a text mining analysis and find that the level of project risk disclosed in the risk description is negatively associated with the success of a project campaign. Overall, our study provides implications for disclosure policies in crowd-based marketplaces.

Estimating the Impact of User Personality Traits on Word-of-Mouth: Text-mining Microblogging Platforms
Panagiotis Adamopoulos, Anindya Ghose and Vilma Todri-Adamopoulos

Word-of-mouth (WOM) plays an increasingly important role in shaping consumers' online behaviors and preferences as users' opinions, choices, and decisions are frequently shared in social media. In this paper, we examine whether personality similarity between social media users can accentuate or attenuate the effectiveness of WOM leveraging data mining and machine-learning methods and the abundance of unstructured data in social media in combination with a novel quasi-experiment and advanced econometric techniques. Specifically, we study whether latent personality characteristics of users are associated with the effectiveness of WOM from purchases on social media platforms like Twitter and can predict their online economic behavior. Our analysis yields two main results. First, there is a positive and statistically significant effect of the level of personality similarity between two social media users on the likelihood of a subsequent purchase after exposure to WOM. In particular, exposure to WOM messages from similar users in terms of personality, rather than dissimilar users, increases the likelihood of a post-purchase by 47.58%. Second, there are statistically significant effects of specific personality characteristics on WOM effectiveness. For instance, users with low levels of extraversion are responsive to WOM, in contrast to extrovert users. In addition, WOM originating from users with high levels of emotional range affects similar users whereas for low levels of emotional range increased similarity has usually the opposite effect. By examining these effects and illustrating how companies can leverage the abundance of unstructured data and tap into users' latent personality characteristics, our paper provides insights regarding the future potential of social media advertising and advanced micro-targeting based on machine learning and natural language processing approaches.

Social Influence in Public and Private Behaviors
Shan Huang

I propose that motive and degree of peer influence are likely to differ between public and private behaviors. To compare peer influence between them quantitatively, I designed and analyzed a large-scale field experiment involving more than 37 million users on WeChat Moments ads. In the experiment, I randomized the number of social cues (i.e. peers’ endorsements of ads) and identified the effects of them on consumers’ public (i.e.liking) and private (i.e.clicking and following) responses to ads. The results show that public responses were associated with significantly more positive effects of social cues than private responses, while both of them were susceptible to a significant peer influence. Tie strength generally exerted larger effects on public responses than on private responses. Relative to homophily, influence explains more of the temporal clustering of public behaviors than private behaviors. This is among the first papers comparing the peer influence in different behaviors.

Information Technology and Income Mobility: Evidence from India
Che-Wei Liu and Sunil Mithas

This study uses a new panel data to study income mobility in India between 2005 and 2011 and the role of information technology (IT) to explain observed patterns in mobility. Income mobility is an important metric to assess the extent of equality of opportunities to move along the economic and social ladders. We discuss determinants of income mobility during the 2005-2011 period and assess the extent to which IT influenced income mobility in India at a time when it also experienced significant adoption and diffusion of computers. Our main findings show that households with higher technology-related endowments such as those experiencing a positive change in computer ownership have a higher chance of moving upward, indicating a possible technology impediment for the poverty trap, among other reasons. Our findings suggest that households with high technology literacy might have an advantage when it comes to income mobility. The change of technology ownership from 2005 to 2011 appears to suggest a link between technology literacy and income mobility. We discuss the implications of our findings for research and policies related to digitization, human capital, and income mobility.

Using IT for the Remote Treatment of Chronically Ill Patients: Differential Process and Value Implications
Balaraman Rajan, Avi Seidmann and Tolga Tezcan

We analyze the impact of telemedicine technology on patient utility and the specialists' operating decisions. Medical specialists treating chronic conditions typically face a heterogeneous set of patients. Such heterogeneity arises because of differences in medical conditions as well as the travel burden each patient faces to visit the clinic. We investigate the impact of patient heterogeneity on the strategic behavior of medical specialists in terms of their operating decisions. We prove that with the introduction of telemedicine the revenue-maximizing specialists become more productive and their service rates will move closer to the socially optimal one. Although the enhanced access to specialist care increases the overall social welfare, we explain why some patients, unexpectedly, will be even worse off with the introduction of telemedicine technology. Our analytical results lead to some important policy implications for facilitating the further deployment of telemedicine in the care of chronically ill patients.

Who Drives in My Users? Evidence for App Usage Causal Network from Graphical Model Approach
Jinyang Zheng, Zhengling Qi, Yong Tan, Baojun Ma and Yifan Dou

Identifying apps which generate usage to other app(s) is not only a crucial task for industrial practitioners, but also challenging for researchers given the larger scale of App network. With a state-of-the-art graphical model method, we overcome the limitation of traditional econometric causal inference models and examine the causal network in emerging App market of Chinese users. Our model generates a diagram displaying causality of App usages. Synergy effect of different apps developed by the same developer and that by different ones are identified. The results not only show causality consistently with intuition, but also exhibit some counterintuitive ones to generate new insights about app analytics, and suggest “mega” app might not drive in usage for other apps as we expected. Our work suggests significant role of graphical model in business analytics and big data related research.

Watch Where You Eat: On the Use of eWOM in Identifying Moral Hazard in New York City Restaurant Hygiene Inspections
Jorge Mejia, Shawn Mankad and Anand Gopal

How can recent advances in machine learning be incorporated into the efforts to reduce foodborne illness, which continues to be a major public health concern? With millions of Americans getting sick with foodborne illness every year, municipal governments are eager to monitor food establishments more closely; however, the cost of constant monitoring can be prohibitively high. Using data from New York City, we demonstrate how social media can be used to monitor hygiene in restaurants. We identify instances of moral hazard within the restaurant hygiene inspection scheme used in New York City by leveraging a dictionary specifically crafted for this purpose. To the extent that social media provide some visibility into the hygiene practices of restaurants, we show how losses from information asymmetry may be partially mitigated in this context and provide strategies for cities to better design restaurant inspection programs.

What Happens When Reviewer Start to Get Free Products? The Impact of Being Expert on Review Generation
Jingchuan Pu, Sang Pil Han and Young Kwark

Online reviews have been an influential information source to the consumers, and many websites recently begin to run expert reviewer program to motivate high-quality reviews. Existing research mainly focus on the impact of professional reviews on the product’s performance, or the difference between professional reviews and regular reviews. However, little is known about the impact of joining expert reviewer program on the participating reviewer. In this paper, we investigate how the expert reviewers’ regular review activities change after they join the expert reviewer program. We conduct this study by utilizing the big expert reviewer program from Amazon- Vine reviewer program. After analyzing the regular review activities of 4004 reviewers during 6-years long time period, we show that, after joining expert review program, reviewers’ contribution to the regular reviews and the average rating of regular reviews increase, and the rating standard deviation decreases. Meanwhile, the content quality of the reviewers’ regular reviews significantly improves, the textual content becomes more complex, and the use of emotional words decreases. In addition, we find that the expert reviewers of lower rank are more likely to be affected by joining the expert reviewer program.

How Morality Mitigates Moral Hazard in Peer-to-Peer Lending
Panagiotis Avramidis and Nikolaos Mylonopoulos

Ever since online peer-to-peer lending intermediaries centralized loan screening and approval, and blocked borrowers and lenders from direct communication, what sets them apart from traditional banking? The most significant remaining feature of p2p lending platforms is the direct matching of lenders and borrowers. Using data from Lending Club, we show that, under certain conditions, the direct matching of borrowers and lenders gives rise to moral intensity (Jones 1991). In turn, higher moral intensity leads to lower default rate among low-credit-score borrowers, effectively mitigating moral hazard. We contribute to the literature on moral hazard, showing that moral hazard is not always or exclusively the result of self-interested rational choice. We contribute to the growing literature on debt crowdfunding, showing ways in which online lending intermediaries might improve loan performance by reinforcing morality.

A Network Approach to Assessing Technology Spillovers from the ICT Industry and their Impact on R&D
Aditya Subrahmanyam, Deepa Mani and Rajib Saha

We draw on the literature on complex network analyses to theoretically and empirically characterize inter-sectoral technology spillovers. We find a significant increase over time in the centrality of ICT industries in inter-sectoral citation networks. Further, the “closeness” of an industry to the ICT industry in the network space is associated with systematic reallocation of R&D effort towards more ICT-centric R&D, greater innovative efficiency or patents per R&D dollar, and recombinant creation and reuse capabilities. Our results are evidence of the important contribution of the ICT industry beyond productivity growth to include R&D and innovation that is fundamental to firm competitiveness.

Learning in a Disruptive Enterprise Mobile Platform in the Banking Industry
Ajit Sharma, Yan Huang and Mayuram Krishnan

The increasing prevalence of enterprise mobile applications calls for research to better understand the mechanisms and factors behind success in these new platforms. In this paper, we empirically study the factors associated with business value, measured as account-opening efficiency, and employee learning dynamics of a shift from a paper based to a tablet based banking application in a large private bank in an emerging market. We find that sales officers who performed relatively well in the paper based account opening system are likely to have a head-start after switching to the tablet banking technology. However, their learning rate is lower than those that did not do as well earlier. As a result, although the high performers continue to maintain their edge in the new mobile platform, the gap between the high and low performers is reduced significantly. Our results also reveal that customer awareness of the tablet banking service and their digital literacy, mobile infrastructure and market maturity can affect sales officers’ account opening performance in the tablet based system

Medical Guideline Making under Widespread HIT Adoption when Litigation is a Concern
Yeongin Kim, Mehmet Ayvaci, Srinivasan Raghunathan and Turgay Ayer

Despite the promise of health information technology (IT), the fast-expanding health IT phenomenon creates a new risk in the U.S. healthcare system by increasing physician vulnerability to legal responsibility or Defensive Medicine. Though it is critical for policy makers to examine the undesirable behavioral shifts of physicians and to set appropriate policy instruments to deal with it, research in relevant domains is limited. Historically, tort reform has been the traditional instrument used to address the litigation concerns. The increasing reliance on evidence-based health care has shifted policymakers’ attention to CPGs as a mechanism to deal with malpractice litigation. In this study, we examine the optimal formulation of evidence-based clinical practice guidelines (CPGs) with consideration for the physicians’ increased liability stemming from the adoption of health IT. We find that when litigation is a concern, deliberately introducing vagueness in CPGs could maximize the social welfare by moderating the defensive practice. Such strategic vagueness in medical standards is desirable especially when benefits of disease detection by a medical test and costs of the test are moderate, and information sharing could not be controlled by a centralized social planner. Our results provide implications for policy makers and health IT developers as we examine several emerging issues regarding meaningful use of health IT including information sharing, medical decision support, and codification of clinical standards in health IT. Our findings also contribute to the long discussion on rigidity/flexibility of CPGs in determining legal standards.

Assessing the Effectiveness of Live Chat on Conversion Probability: Evidence from an Online Marketplace
Haoyan Sun, Jianqing Chen and Ming Fan

This paper examines how online sellers’ use of live chat influences their conversion probability. We argue that live chat manages to increase conversion probability by performing the function of “inform” and “persuade.” We further explore whether and how this effect is contingent on other factors related to consumer decision making, while accounting for the effort level of sellers. We apply a random coefficient model in a Bayesian hierarchical framework, and test the model using a panel dataset from Taboao—the giant e-commerce platform in China. Our results suggest that live chat in general has a positive impact on conversion probability, and the positive impact is stronger when product information on webpage is not comprehensive and when consumer perceive higher utility from seller products.

Peer Effects and the Social Production of Online Reviews
Zhihong Ke, De Liu and Alok Gupta

Motivated by the use of online social networks in leading online review platforms, this study examines whether there is a peer effect in the production of online reviews, namely if a user of an online review platform writes a review on a store, are her online friends more likely to write a review on the same store? We conduct two complementary studies -- an econometric analysis using a large archival dataset on Yelp reviews, and a randomized laboratory experiment. Two studies systemically show that peer effect does exists. Overall, we find a positive peer effect of friend reviews, after controlling for homophile. Interestingly, we also find reviews by future friends have a negative effect; relatedly, reviews by new offline friends in the lab also have a negative effect. These suggest that users of online review platforms may also be concerned about redundancy after seeing a “friend” review. We discuss implications of our findings for online review platform designers and marketers.

The Invisible Barrier: The Effect of Promoting Agencies on Sales in Electronic Markets for Music
Marios Kokkodis, Theodoros Lappas and Konstantinos Pelechrinis

In this study we focus on understanding how the interplay of different music label characteristics affects the placement of a song in the charts. We collect and analyze a unique panel dataset from a major marketplace of electronic music. First, we deploy an Accelerated Failure Time model to estimate the probability of a song to appear in the charts. By doing so, we create a labeled dataset with songs that stochastically have very low propensity of entering the charts, as well as with songs that have already been listed in the charts. We then use this dataset to build a multidimensional "tree based causal inference" framework to study how the interaction of different music label characteristics affect the placement of the song in the charts. We find that certain combinations of the music label's characteristic create a 9-fold increase of a song's likelihood to enter the charts.

Online Retailer with Its Own Brand Product and a Competing Supplier in the Presence of Uncertain Consumers
Kyung Sung Jung and Young Kwark

Abstract We study the effect of consumer reviews on the dominant online retailer that launches its own brand product and the supplier that sells the product through the retailer. The online retailer acts as a supplier that sells its own brand product while acting as a platform owner or a reseller to let other suppliers sell their products. The retailer and the supplier sell two substitutable products and the products differ in both their qualities and fits to consumers' needs. Along with the products, the retailer often presents the support provided for the product on purchase. Consumers are uncertain in both the product and the support. A dominant online retailer has better reputation only in the product support whereas the two products are comparably competitive. Consumers read the user-generated content (e.g., consumer reviews) to reduce the uncertainty in product attributes and product support. We find that the increased competition because of the consumer reviews between the retailer and the supplier can help the retailer better off under reselling mode whereas competition effect only dominates under the agency mode. Regardless of the modes, the retailer can be better off when consumers rely more on the consumer reviews for the support for the competing supplier's product because of the reputation effect. We provide the results depending on whether consumers weigh more on the quality or fit. We show that when the reputable retailer enters the product market and competes with the supplier on its website, whether the retailer or the supplier can be better off because of the reviews varies depending on the different modes for different types of the products and the reputation of the retailer in the presence of the uncertain consumers.

An Interaction Analysis of Social Media and Traditional Platforms in the Consumer Purchasing Funnel
Ran Zhang, Daniel Zantedeschi and Shivendu Shivendu

This paper studies complementarity and substitutability of online platforms for targeted advertising along the consumer purchasing funnel. We use cutting-edge machine learning techniques to form case-control designs analyzed by means of post-regularized choice models. This allows us to gauge the interaction effects between activities on different platforms and within different parts of the purchasing funnel avoiding potential selection biases. Our empirical analysis delivers many interesting insights: (1) clustering based on the similarity of consumers, compared with no clustering, mitigates heterogeneity and offers more accurate associational measures related to platform effects, (2) targeting across platforms is positively associated for the lower funnel, but negatively associated with the odds ratio of purchase for the upper funnel, suggesting that too much targeting may be detrimental in converting early-stage consumers, (3) social media is generally positively associated with the odds ratio of purchase for consumers in the early stages, but has no measurable impact relative to the traditional platform when consumers move down to the lower funnel. We leverage upon these findings to discuss actionable managerial prescriptions.

The Impact of Website Blocking on Digital Piracy Activity
Filipa Reis, Miguel Godinho de Matos and Pedro Ferreira

In this work we study the effectiveness of website blocking, in particular, DNS blocking on deterring piracy activity and promoting the use of legal alternatives. In collaboration with a large telecommunications provider, we measure the impact of a website blocking protocol on household's Internet and paid Video on Demand use. We find that after the start of the protocol, daily download traffic households who had previously used BitTorrent regularly decreased by 35% and upload traffic decreased by 40%. We find no statistical significant impact on these households' VoD expenditure except for a slight increase of 1.5 minutes/week of VoD view time. We complement this analysis with a district level analysis exploring some of the mechanisms affecting website blocking effectiveness. We find that as side effect of the blocking protocol, part of the pirate population became more sophisticated in their piracy behaviour by searching for and adopting strategies to bypass blocks. Our results shows that districts where this behaviour was more prevalent presented significantly higher torrent activity. Our study contributes to the literature on piracy control strategies by being the first to conduct a household level analysis of the effectiveness of batch DNS blocking of websites providing copyright infringing content. We also provide evidence on the heterogenous effect of this measure and the associated learning and increased sophistication of pirates in what regards block circumvention.

Market Segmentation and Software Security: Pricing Patching Rights
Duy Dao, Terrence August and Kihoon Kim

The patching approach to security in the software industry has been less effective than desired. One critical issue with the status quo is that the endowment of "patching rights" (the ability for a consumer to choose whether security updates are applied) lacks the incentive structure to induce better security-related decisions. In this paper, we establish how producers can differentiate their products based on the provision of patching rights and how the optimal pricing of these rights can segment the market in a manner that leads to both greater security and greater profitability. We characterize the price for these rights, the discount provided to those who relinquish rights and have their systems automatically updated, and the consumption and protection strategies taken by users in equilibrium as they strategically interact due to the security externality associated with product vulnerabilities. We quantify the effectiveness of priced patching rights, its impact on welfare, and the ability of taxes to achieve an analogous effect in the open-source domain. In this domain, we demonstrate why large populations of unpatched users remain even when automatic updating is available, and then characterize how taxes on patching rights should optimally be structured.

Scarcity Strategy in Crowdfunding: An Empirical Exploration of Reward Limit on Kickstarter
Lusi Yang, Zhiyi Wang and Jungpil Hahn

Scarcity-based marketing strategies (e.g., limited edition products) have widely been embraced by firms to increase sales. Recently, a similar practice has been increasingly adopted in reward-based crowdfunding platforms in the form of reward limits, whereby project creators are able to restrict the quantity of contributors in each reward tier. Despite an increasing volume of research devoted to understanding project design strategy and fundraising success, the role of such scarcity strategy has been neglected. The current study strives to fill this void by uncovering the effect of reward limits in eventual and concurrent funding performance. Specifically, we performed campaign and campaign-day level analysis with data from Kickstarter. At the campaign level, we find that setting reward limits at the beginning of a campaign is beneficial for final funding outcomes across four different measures of crowdfunding performance. Further, the number of limited reward tiers was shown to have an inverted U-shaped relationship with fundraising performance. Potential endogeneity issues were addressed with propensity score matching and the Heckman selection model. At the campaign-day level, we examined the dynamics of reward limit during the course of fundraising using a two-way fixed effects panel model. Incorporating new limited reward tiers is helpful for attracting new backers, but having reward tiers being “sold out” will demotivate subsequent backers to contribute to the project and thus lead to lower funding speed in subsequent days. Our findings highlight the importance of strategically using the scarcity strategy in crowdfunding. Our research extends the crowdfunding literature by showing the dynamic influence of project attributes within the campaign duration.

Consumer Digital Shopping Journey: The Interplay of Social Information from Friends and Crowds
Baojiang Yang, Miguel Godinho de Matos and Pedro Ferreira

As information technology continues to advance rapidly, the way consumers generate and obtain information in digital shopping environment has been funda- mentally shifted. With abundant individual level data becoming broadly available, companies are moving rapidly from ”last conversion” or ”search only” models to- ward greater sophisticated models to better understand consumer’s footprint through- out the decision journey. For consumers, the reduced cost of information display and peer communication offers multiple channels of social informations that serves as guidance for their digital shopping experience. This study aims at investigating the how different sources of peer information, particularly from crowds and from friends, can impact consumers searching and decision-making process. By estimat- ing a sequential search based dynamic structural model and validating it in a digital Video-on-Demand platform, we show that positive signals from the crowds and the friends both have significant and positive predictive relationship with consumer pref- erences over a movie. Friends’ word-of-mouth has a potential to reduce search cost, serving as a recommendation signal. Our study provides new insights for practi- tioners into understanding of consumer decision making journey in a social digital shopping environment.

Price Competition in the Software Market: On-premises vs. Software as a Service
Kunhao Jia, Juan Feng and Xiuwu Liao

This research studies the price competition between a SaaS vendor, who adopts subscription-based pricing scheme, and an on-premises vendor, who uses perpetual license scheme. The SaaS vendor decides its subscription price for two periods, and the on-premises software vendor decides its initial-version software price for the first period, as well as the upgrade price for the second period. We find that firms’ pricing strategy largely depends on the differences in the customizability, as well as the implementation costs between the two vendors: (1) it is possible that the SaaS vendor becomes a monopolist in the market (when the customizability of the SaaS software, as well as the implementation cost of the on-premises vendor are sufficiently high); but it is never possible for the on-premises vendor to drive the SaaS vendor out of the market; (2) the impact of the customizability of the SaaS software is non-monotonic. More specifically, when the customizability of the SaaS software is low or moderate, an increase in the customizability can hurt the SaaS vendor’s profit if the implementation cost of the on-premises software is low, though it always harms the on-premises software vendor; when the customizability of the SaaS software is high, however, an increase in the customizability benefits both software vendors.

User-Generated Charitable Content in Social Media: Evidence from a Field Experiment
Xue Tan, Yingda Lu and Yong Tan

Leveraging an annual charitable movement that takes place in Twitter, we investigated the effect of content adoption (retweets) and social ties (followers) on users’ repeat engagement in generating charitable content. We conducted an empirical analysis (study I) to understand the effect of content adoption, and find that the retweets received by users’ prior engagement one year ago has no significant impact over their repeat engagement. We designed a field experiment (study II) to examine the role of social ties by adding compassionate followers to users, right before a charitable campaign. We find that the increase in followers does not have an overall impact. However, active users who generate more than eight tweets in a ten-day period respond positively to the increase in followers. In addition, users who included pictures in their prior charitable content respond positively to the increase in followers, despite the fact that users are generally less likely to engage repeatedly after including pictures in their charitable content.

Effects of Personalized versus Aggregate Ratings on Consumer Preference Responses
Gediminas Adomavicius, Jesse Bockstedt, Shawn Curley and Jingjing Zhang

Prior research has shown that online recommendations have significant influence on consumers’ preference ratings and their economic behavior. However, research has not examined the anchoring effects of aggregate user ratings, which are also commonly displayed in online retail settings. This research compares and contrasts the anchoring biases introduced by aggregate ratings on consumers’ preference ratings to those produced by personalized recommendations. Through multiple laboratory experiments, we show that the user preferences can be affected (i.e., distorted) by the displayed average online user ratings in a similar manner as has been shown with personalized recommendations. We further compare the magnitude of anchoring biases by personalized recommendations and aggregate ratings. Our results show that when shown separately, aggregate ratings and personalized recommendations create similar effects on user preferences. When shown together, there is no cumulative increase in the effect, and personalized recommendations tend to dominate the effect on user preferences.

Integrating "Buy-It-Now" into Sequential Auctions: Results from Lab and Field Experiments
Yixin Lu, Alok Gupta, Wolfgang Ketter and Eric van Heck

While the standard auction formats have been studied extensively in the literature, various modifications made in the real-world operating environment often post challenges in characterizing bidding behaviors and predicting market outcomes. This paper studies a hybrid sales mechanism which combines features of posted prices and sequential Dutch auctions. In particular, we adopt a multi-method approach to examine the performance of such hybrid mechanism in terms of operational efficiency, allocative efficiency, revenue generation and price stability. Using a controlled lab experiment and a large-scale quasi-natural field experiment, we find that the hybrid mechanism can substantially speed up the auction process at negligible cost of allocative efficiency. Further, by allowing multiple bidders to purchase at the same price, it also increases the stability of winning prices. Our results provide useful implications for auction design in complex markets.

Mobile Temptation and Self-Control through Precommitment Apps
Hyunji So, Jinpyo Hong, Sang Pil Han and Wonseok Oh

Given the inherent vulnerability of humans to temptation, many people resort to precommitment as an early preemptive guarantee against relenting to enticing opportunities to stray from sensible behavior. This study investigates the effectiveness of diverse precommitment mechanisms as self-control measures against mobile temptation, which refers to an uncontrollable desire and craving to consume mobile applications. These precommitment systems are made available by app-blocking options that are downloadable on smartphones at user discretion to restrict access to apps, including games and SNSs. On the basis of Thaler and Shefrin’s (1981) self-control framework, we identify and evaluate rule-based (spatial and temporal) and incentive-driven (social- and reinforcement-based) precommitment schemes. The former leverages some form of coercion that obliges adherence to rules, whereas the latter is designed to reinforce volition for self-control through incentives rather than restrictions. Unlike previous self-control studies that rely heavily on self-reported survey data, we use log data that contain the individual smartphone/app usage and blockage patterns of numerous mobile users, along with registration information that captures their demographic profiles and self-diagnosed levels of smartphone/app addiction. Mixed results are found with respect to the effectiveness of rule-driven precommitment schemes: Rigid temporal precommitment effectively facilitates sustained self-control and motivates users to increase block time, but contrary to expectations, the less stringent flexible spatial precommitment outperforms rigid spatial precommitment. The findings also suggest that both social- and reinforcement-based methods successfully advance sustainable command over oneself and therefore aid users in increasing voluntarily implemented block time. The collective monitoring and encouragement enabled by social-based precommitment appear to positively influence users’ willpower in maintaining self-discipline. Finally, reinforcement-based precommitment, which allows users to monitor the progression of their performance through push notification, helps individuals consistently exert pressure to exercise restraint.

Recruitment Practices for Technology Leaders
Prasanna Tambe and Xuan Ye

Using a unique database reflecting the details of technical job interviews, we study how high-tech employers source and evaluate candidates in tight labor markets. We show that for jobs involving new technologies--for which a supply of labor may not yet be readily available--employers rely more heavily on recruiters and referrals so they can access a pool of employed workers who may have already acquired these skills on-the-job at competitor firms. Second, by text-mining interview questions, we show that employers trying to fill such positions are more likely to ask interview questions that indicate how quickly prospective IT hires can learn on the job, rather than asking about prior experience. Employers ask such questions 39% more frequently for these jobs than for jobs involving mature technologies. These effects are robust to including employer, job title, and labor market fixed effects. Our findings are consistent with the argument that employers design technical HR strategies to complement their technological choices.

Antecedents and Consequences of Electronic Medical Record System Abandonment
Kartik Krishna Ganju, Hilal Atasoy, Paul Pavlou and Pei-Yu Chen

The adoption of Electronic Medical Record (EMR) systems has been found to have a number of organization wide changes and is often met with resistance from users. Additionally, systems often do not meet requirements over time. Due to this conflict, EMR systems are occasionally switched or abandoned. In this paper, we examine the phenomenon of switching and abandoning of EMR systems and identify the impact policy and hospital characteristics have on the decisions hospitals make about their information systems. We first examine the role of the HITECH Act in facilitating s witching and abandoning of EMR systems. We also examine the choices that hospitals make and if they choose to adopt the market leader and the impact of these switches both within and outside the purview of the HITEH Act and the impact of switching of an EMR system.

Catching Them Red-handed: Optimizing the Nursing Homes’ Rating System
Xu Han, Niam Yaraghi and Ram Gopal

The Centers for Medicare & Medicaid Services (CMS) launched its nursing home rating system in 2008, which has been widely used among patients, doctors and insurance companies since its inception. The system rates nursing homes based on a combination of CMS’s onsite inspections and nursing homes’ self-reported measures. Research has shown that the rating system is subject to inflation in the self-reporting procedure which leads to biased overall ratings. Given the limited resources of CMS, it is important to optimize its inspection process and design effective audit process. In this paper, we systematically investigate the nursing home rating system inspection and audit problem. We formulate the inspection problem by using an innovative graph-based method, and test the model by using CMS’s data. The results of this paper show that the measures that CMS are currently inspecting are optimal provided that an effective audit mechanism is in place. In the study of audit problem, we take nursing homes’ reactions to different audit policies into consideration and conduct a simulation to study the optimal audit parameter settings. We then prove our results mathematically. Our result suggests that CMS should use a moderate audit policy, in order to carefully balance the tradeoff between audit net budget and audit efficiency.

The Causal Impact of Fit Valence and Fit Reference on Product Returns
Yang Wang, Vandana Ramachandran and Olivia Sheng

We investigate the causal impact of two types of online product fit information – fit valence and fit reference – on product return rate by leveraging a change in the product review system that took place at an online retailer of outdoor goods. This natural experimental setting allows us to estimate the treatment effect of making new fit-related information available to online consumers. Specifically, we find that the mere presentation of either fit valence (e.g. “true to size”) or fit-reference information (e.g. body size) does not help online shoppers reduce their purchase error. Rather, it is the combination of the two types of fit information that drives the drop in product return rate. Through the lens of semantic holism, we illustrate how customers interpret product fit opinions by using the fit-reference information provided by the same reviewer. The combination of the two types of product fit information assists a subsequent customer to infer the right size that matches her fit preference, thus leading to a decrease in product return rate. We also find such product fit information should be posted as early as possible in order to achieve better effectiveness. Our findings offer useful implications to online retailers grappling with high product return rates for merchandise where fit matters.

Understanding User Contribution in a Social Crowdsourcing Mobile App
Tae Hun Kim, Ashton Shortridge, Vallabh Sambamurthy and Anjana Susarla

We investigate a specific mobile app, used to share traffic information in real time, to explain user contribution. A theoretical framework for spatial decision support systems hypothesizes traffic (virtual and real), demographic, and economic factors as socioeconomic drivers of user contribution and its sustained effectiveness. Based on a spatial data approach, our panel analyses provide empirical evidence from a 63-day dataset, hourly collected from Waze users in New York City. Our aggregate estimations suggest that traffic and demographic factors are overall associated with initial and secondary contributions as well as sustained contribution. On the other hand, economic factors have negative or insignificant association with initial and secondary contributions; they are not associated with sustained contributions. In addition, the large-scale data provide visualized results to support our estimations for user contribution with socioeconomic factors. Our findings enhance insights about user contribution to mobile communities and suggest practical implications for mobile app businesses.

Business Models in the Sharing Economy: Manufacturing durable goods in the presence of Peer-to-Peer rental markets
Vibhanshu Abhishek, Jose Guajardo and Zhe Zhang

Business models focusing on providing access to assets rather than on transferring ownership of goods have become one of the most fundamental recent trends across various sectors and are being referred to as the sharing economy. The proliferation of sharing economy companies in many indus- tries represents a challenge for traditional firms, such manufacturers of durable goods. We analyze the impact of the emergence of a peer-to-peer market for a monopolistic firm that manufactures a durable good using an analytical model. Our analysis shows that the heterogeneity in the usage rate among consumer is an important driving factor, and both the firm and consumers benefit from peer-to-peer when there is sufficient heterogeneity in the usage rate. However, when the usage rates are similar sides can be worse off due to the sharing economy. We also observe that several low-usage customers substitute from buying to renting from the sharing economy, and the sharing economy results in shifting the surplus from high-usage, high-value consumers to low-value consumers.

Understanding, Replicating, and Leveraging Dynamics of Bidder Behavior in Continuous Combinatorial Auctions
Ali Mahdavi Adeli, Gediminas Adomavicius and Alok Gupta

Combinatorial auctions represent sophisticated market mechanisms that are becoming increasingly important in various business applications due to their ability to improve economic efficiency and auction revenue, especially in settings where participants tend to exhibit more complex user preferences and valuations. While recent studies on such auctions have found heterogeneity in bidder behavior and its varying effect on auction outcomes, the area of bidder behavior and its impact on economic outcomes in combinatorial auctions is still largely underexplored. One of the main reasons is that it is nearly impossible to control for the type of bidder behavior in real world or experimental auction setups. We propose an agent-based modeling approach to simulate human bidder behavior in continuous combinatorial auctions and demonstrate the validity of our developed agents by replicating human bidder behaviors observed in experimental combinatorial auctions. We leverage our agents to simulate a wide variety of competition types, including experimentally unobserved ones that could not otherwise be studied. The capabilities of the proposed approach enable more comprehensive studies (via richer controlled experiments) of bidding behavior in the complex and highly dynamic decision environment of continuous combinatorial auctions.

Predicting Human Capital Flow With LinkedIn Profiles
Yuanyang Liu, Gautam Pant and Olivia Sheng

Firms compete within industries and across industries for human capital which is a key component of the knowledge economy. Hence predicting the flow of human capital between firms can provide essential intelligence for all of firms' stakeholders such as managers, recruiters, government, investors, and analysts. The increasing web presence of employees through their LinkedIn profiles provides a rich source of data for studying interactions between firms and human capital over time. In particular, we model and study such interactions as a network of human capital flows between firms with the goal of predicting the inter-firm human capital movement. Through network analysis we recognize an industry-based localization in human capital movement in addition to the role of network features in determining the human capital flow. We find that models that use network features in addition to firm-level features such as revenue, industry, and the nature of human capital can provide greater predictive accuracy.

Impact of Unlimited Mobile Data Plan on Media Consumption: A Natural Experiment
Karthik Babu Nattamai Kannan, Eric Overby and Sri Narasimhan

In this article, we examine the impact of adoption of unlimited mobile data plan on data and TV consumption. We work with our partner company study a quasi-natural experiment that resulted when this company allowed its subscribers with both TV and wireless connection to switch to unlimited mobile data plan. We find that adoption of this plan increases monthly data usage by as much as 3.5GB per month. Surprisingly, this increase in data usage is moderated by the type of data plan previously held by the customer, with those who had plans with lower data caps almost doubling their consumption while others only marginally increasing their consumption. Also, TV consumption through streaming in mobile apps and browser-based access increased by as much as 6 minutes per user per month. Interestingly, music and sports related content exhibited the largest increase while local channels and family based entertainment had the smallest increase. We arrived at these results after controlling for individual heterogeneity, a time trend, and other unobserved time variant trends.

Analysis of the Efficiency of Bus Transportation Using Large Scale GPS and Smart Card Data
Yun Wang, Faiz Currim and Sudha Ram

Efficiency evaluation is crucial to the management and development of smart urban transportation, as it allows transportation planners to better understand the impact of their decisions and design targeted interventions to improve productivity. In this work, we introduce a three-layer framework to support smart urban mobility with an emphasis on bus transportation. In Layer-1, we apply novel Big Data techniques to expeditiously calculate bus travel times and passenger demands using universal data streams. Layer-2 contains two analytic components: network analysis of passenger transit patterns and causal relationship analysis for bus delays. The third layer provides interactive visualization tools for decision support. Our system is developed in cooperation with the city of Fortaleza in Brazil. The use of generally available urban transportation data makes our methodology adaptable and customizable for other cities.

The Impact of Broadband Speed and Market Competition on Economic Growth of Local Towns and Communities
Aindrila Chakraborty, Sudip Bhattacharjee and James Marsden

Fiber-to-the-home (FTTH) is an exciting development that promises to bring fast internet to all communities, thus improving the backbone of all internet related activities for businesses and consumers. Recent technological advances in fixed line (wired) and mobile broadband (wireless) sectors can be beneficial to the economic development of local towns and communities in education, healthcare, public safety, job creation and other measures. While broadband expansion decisions are made locally, on a town basis, existing studies on the impact of broadband have generally focused on macro factors and country level analysis. Our study estimates, on a local community basis (town/municipality), the impact of certain aspects of broadband on different economic measures of towns. Combining data from a wide variety of sources, we find that market competition and speed drive local communities’ economic growth, albeit under certain conditions. We also observe that these drivers have a significant effect on different factors of local economies during the post-recession period (after 2009). To our knowledge, our data driven analysis is the first one that enables towns and communities to estimate whether they benefit economically, given their local conditions, before they commit to investing in FTTH and Gigabit Internet infrastructure development.

Peer Influence and the Choice of IT Careers
Nishtha Langer and Tarun Jain

The productivity of the Information Technology (IT) and IT enabled Services (ITeS) industry depends critically on the supply of high quality human capital. While existing research has examined the role of education and training on the human capital in this industry, little research informs the role of peer influences on the decision to pursue IT/ITeS careers. Managerial employees are especially critical to IT firms because they contribute beyond technical functions to management, sales, and marketing, and leadership roles. Therefore, focusing on managerial employees, we examine the influence of peers on the choice to pursue information technology careers in India. Specifically, we analyze data on student networks at a leading business school where students are exogenously assigned to peer groups, and link these to students’ choice of post-program careers in the IT industry. Although before the program, students have experience in both IT and non-IT fields, they may switch roles and/or industries after the program. For instance, some may pursue IT roles in non-IT sectors such as retail, whereas others may pursue non-IT roles such as strategy and sales in IT companies. We posit that such career choices may be informed and driven not only by own motivation and ability, but also by the influence of peers. Our findings reveal that being part of a group that includes peers who have worked in IT increases the likelihood of accepting an offer in the IT industry. However, counter-intuitively, we find that if a student has had no IT experience, having IT peers decreases the likelihood of accepting a job in the IT industry. In other words, IT peers discourage non-IT peers from being part of the IT industry.