Citation
Independent Product Information and Marketing Strategies

Material Information

Title:
Independent Product Information and Marketing Strategies
Creator:
CHEN, YUBO ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Consumer advertising ( jstor )
Consumer choice ( jstor )
Consumer information ( jstor )
Consumer preferences ( jstor )
Consumer prices ( jstor )
Consumer tastes ( jstor )
Information attributes ( jstor )
Magazines ( jstor )
Product information ( jstor )
Recommendations ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Yubo Chen. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
8/31/2009
Resource Identifier:
439074711 ( OCLC )

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

INDEPENDENT PRODUCT INFORMATION AND MARKETING STRATEGIES By YUBO CHEN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2004

PAGE 2

Copyright 2004 By Yubo Chen

PAGE 3

To My Parents

PAGE 4

iv ACKNOWLEDGEMENTS First, I would likely to thank my advisor, Dr. Jinhong Xie. Without her continuous support, deep care and great mentorship, my graduate studies would not have been the same. I would also likely to sincerely thank my co-advisor, Dr. Steven Shugan, for his great inspiration and valuable guidance during the past several years. I would also wish to express my thanks and appreciation to my dissertation committee members, Dr. Joseph Alba, Dr. David Sappington and Dr. Barton Weitz, for their insightful advice and great help during the process of my dissertation completion and job searching. Finally, I deeply thank my parents, brother and other family members for their love and support over all these years.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGEMENTS...............................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT....................................................................................................................... ix CHAPTER 1 INTRODUCTION.....................................................................................................1 2 THIRD-PARTY PRODUCT REVIEW AND FIRM MARKETING STRATEGY.4 2.1 Introduction..........................................................................................................4 2.2 Third-Party Product Review................................................................................8 2.3 Assumptions and Model Setting........................................................................10 2.4 Description Product Review..............................................................................16 2.5 Recommendation Product Review.....................................................................21 2.6 Other Advertising Media...................................................................................27 2.7 Empirical Evidence............................................................................................31 2.8 Conclusion.........................................................................................................40 3 ONLINE CONSUMER REVIEW: A NEW MARKETING FUNCTION..............44 3.1 Introduction........................................................................................................44 3.2 Related Literature...............................................................................................48 3.3 Basic Model.......................................................................................................51 3.4 Consumer Review and Product Assortment Strategy........................................56 3.5 Information Supply Strategy..............................................................................58 3.6 Timing Decision on Offering Consumer Reviews.............................................69 3.7 Empirical Evidence............................................................................................70 3.8 Conclusion.........................................................................................................74

PAGE 6

vi APPENDIX A NOTATIONS AND PROOFS IN CHAPTER 2.....................................................79 A.1 Summary of Notations......................................................................................79 A.2 Proofs in Chapter 2...........................................................................................80 B PROOFS IN CHAPTER 3.......................................................................................88 B.1 Proof of Lemma 3.1..........................................................................................88 B.2 Proof of Lemma 3.2..........................................................................................88 B.3 Proof of Lemma 3.3..........................................................................................89 B.4 Proof of Proposition 3.4....................................................................................90 REFERENCES.................................................................................................................93 BIOGRAPHICAL SKETCH............................................................................................98

PAGE 7

vii LIST OF TABLES Table page 2-1 Examples of Third-Party Product Reviews...............................................................10 2-2 Consumer Purchase Decision....................................................................................18 2-3 Consumer Information and Preference.......................................................................22 2-4 Advertising Response to the Third-Party Product Review.........................................29 2-5 Predictions of Advertising Response..........................................................................37 2-6 Empirical Results of Advertising Responses..............................................................39 3-1 The Impact of Product Assortments...........................................................................72 3-2 The Impact of Number of Matching Consumers and Product Launch Time.............74 B-1 Novice Consumer Expected Valuations in the Presence of Consumer Reviews.......90

PAGE 8

viii LIST OF FIGURES Figure page 1-1. Independent Product Information and Marketing Strategies.......................................2 3-1. Seller Information Supply Strategy...........................................................................61

PAGE 9

ix Abstract Of Dissertation Presented To The Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy INDEPENDENT PRODUCT INFORMATION AND MARKETING STRATEGIES By Yubo Chen August 2004 Chair: Jinhong Xie Cochair: Steven Shugan Major Department: Marketing Advances in information technology make independent product information from nonmarketing sources increasingly available to consumers. This dissertation addresses how firms should respond to independent product information in designing their marketing strategies. Two complementary aspects of this problem are analyzed: (1) how a firm should strategically respond to product reviews published by third parties, and (2) when a seller will benefit from providing consumer reviews of its products to its customers. Regarding third-party product reviews, the dissertation explores when and by how much firms should change advertising intensity, advertising format, and product pricing in response to reviews by independent parties. For example, should the winner of an “Editor’s Choice” award boost advertising to spread the good news? Should firms respond differently to different types of reviews or to different advertising media? A theory is developed to address these questions, prescribing firms’ optimal responses to product reviews under different product/market/ review/media conditions. The analysis shows that it is unprofitable to use price as a strategic

PAGE 10

xvariable in response to product reviews when enough consumers value horizontal product attributes. Surprisingly, the model suggests that using advertisements containing third-party award logos to broadcast a winning review can hurt the winning product of that review. The model further suggests that it is not necessarily beneficial for winning products to boost advertising expenditures to spread the good news. Data from two industries—printers and running shoes—are used to illustrate some of these findings. Regarding online consumer review, the dissertation argues that it can serve as a free sales assistant to help consumers identify their best match products, and examines when a seller should offer consumer review information to its customers, and how this decision affects the seller’s other marketing strategies. The seller’s incentive to provide full product information is found to increase if it decides to provide consumer reviews. The seller’s consumer review supply decision is shown to depend upon the fraction of unsophisticated consumers, review informativeness, the seller’s product assortment strategy, the product cost, and product characteristics. Finally, it is shown that the seller can benefit from delaying supplying consumer review information.

PAGE 11

1 CHAPTER 1 INTRODUCTION Recent advances in information technology and the Internet make independent product information from non-marketing sources widely accessible to consumers. There are mainly two types of independent product information. The first type is the third-party product review based on independent laboratory tests or expert evaluations. Various popular consumer magazines (e.g., Consumer Reports , PC Magazine, PC World, Car and Driver ) regularly publish comprehensive reviews of products of interest to their readers. The Internet and fast-developing information technology have significantly reduced third-party reviewersÂ’ information-delivery cost and consumersÂ’ information-retrieval cost. As a result, a growing number of websites (e.g., CNET.com, ZDNET.com) are offering online third-party product reviews. Consumers can freely access third-party product reviews from these web sites. The second type of independent product information includes consumer-posted evaluative information or consumer reviews . The significantly reduced costs of collecting and distributing information create new opportunity for product evaluation sharing among consumers (Avery, Resnick and Zeckhauser 1999). An increasing number of online sellers such as Amazon.com provide platforms for consumers to publicize their personal evaluations of product performance. Independent product information has been very important for consumers in many product categories particularly durable and experience products. Walker (1995) finds more than 40% of Americans seek independent information for services such as choosing a new doctor, getting legal advice, and selecting an auto mechanic. With the explosive development of information technology in the last decade, high-tech products such as consumer electronics become widespread in consumer daily lives. Products are becoming more complicated, and more product

PAGE 12

2 categories need extensive knowledge and information for pre-purchase evaluation (Glazer 1999). Independent product information becomes increasingly important for consumer purchase decision. In a survey reported by the Los Angeles Times , 44% of online consumers said they consulted review websites before making a purchase (Piller 1999). In many product categories, buyers and sellers typically have asymmetric information. Sellers have private product information that buyers may not share. Akerlof (1970) shows that this information asymmetry may lead to market failure. To address this problem, traditional economics and marketing literature shows how firms can use different marketing strategies to communicate product information to consumers (e.g., Milgrom and Roberts 1984, Gestner 1985, Moorthy and Srinavasan 1995). In addition to sellersÂ’ marketing strategies, independent product information provides a second channel for consumers to learn product information. Given its wide accessibility and growing importance, how this second information channel for consumers affects firmsÂ’ marketing strategies has become a very important issue. Figure 1-1. Independent Product Information and Marketing Strategies Recently, there has been a growing interest among marketing scholars to study the marketing implications of independent product information (e.g., Eliashberg and Shugan 1997, Shaffer and Sellers Consumers Marketing Strategies Independent Product Information (Third-party & Consumer Reviews)

PAGE 13

3 Zettelmeyer 2002). These recent studies have advanced our understanding of the impact of independent product information on product success and firmsÂ’ profits. However, the question of how firmsÂ’ marketing strategies interact with independent product information has not been afforded detailed attention in the literature (see Figure 1-1). The main research issue of this dissertation is to investigate how firms should integrate independent product information into their marketing strategy designing. The dissertation consists of two parts on complementary aspects of this problem. Chapter 2 examines firmsÂ’ optimal pricing and advertising responses to product reviews published by third parties such as consumer magazines and online websites. Chapter 3 investigates an important marketing function of online consumer review, studies when an online seller should provide consumer reviews to its customers, and examines the interaction between an online sellerÂ’s consumer supply strategy and its other marketing strategies.

PAGE 14

4 CHAPTER 2 THIRD-PARTY PRODUCT REVIEW AND FIRM MARKETING STRATEGY 2.1 Introduction Third-party product reviews based on independent laboratory tests or expert evaluations have grown increasingly popular in recent years. Various popular consumer magazines (e.g., PC Magazine, PC World, Car and Driver , Scuba Diving Magazine , Runner's World, Entertainment Weekly , Gourmet ) regularly publish comprehensive reviews of products of interest to their readers. Moreover, the Internet and fast-developing information technology have significantly reduced reviewers’ information-delivery cost and consumers’ information-retrieval cost. As a result, a growing number of websites (e.g., CNET.com, ZDNET.com, caranddriver.com, swiminfo.com, wirelessdesign.com, enjoythemusic.com, golfdigest.com) are offering online third-party product reviews. In addition, consumers can now easily access and compare product reviews by different sources via specialized product-review sites such as ConsumerSearch.com, which collects reviews on 170 product categories from trusted publications such as Consumer Digest and PC Magazine . Market observations suggest that third-party product reviews have a significant effect on the success/failure of products. For example, USA Today reported that “[a] bad review in a computer magazine can kill a product and often does . . . [A]fter PC Magazine panned one Northgate Computer Systems Inc. computer model in early 1988, sales all but dried up . . .” Moreover, “[a]fter Clarion Software was awarded an ‘Editor's Choice’ citation for its database program, Softsel—the USA’s largest distributor of computer products—decided to carry the program, which Softsel had previously rejected, [and it] is now a best seller” (Lewyn, 1989). Third-party reviews have played a very important role in consumers’ purchasing decisions. A survey by the

PAGE 15

5 Wall Street Journal in 1994 showed that over a third of Americans sought the advice of critics when choosing a movie (Simmons, 1994). And in a survey reported by The Los Angeles Times , 44% of online consumers said they consulted review websites before making a purchase (Piller 1999). Recently, there has been growing interest among marketing scholars in studying the marketing implications of various third-party infomediaries, including third-party product reviewers (e.g., Eliashberg and Shugan 1997, Reddy, Swaminathan and Motley 1998, Chen, Iyer and Padmanabhan 2002, Shaffer and Zettelmeyer 2002, Shugan and Winner 2003). These recent studies have advanced our understanding of the impact of third-party review information on product success and firms’ profits. However, an important but under-explored area is how manufacturing firms should adapt their marketing strategy in response to such reviews. For example, should a firm receiving an unfavorable product review reduce its price or adjust its advertising in response to the negative effect of the review on the demand for its product? Should a winning product of a product review (e.g., “Editor’s Choice”) boost its advertising expenditure to spread the news of its victory or reduce its advertising and enjoy the benefit of free advertising via the product review? It is important to develop a better understanding of when and how a manufacturing firm should vary its marketing strategy to maximize its benefit (or minimize its loss) from a third-party product review. An early exploratory empirical study (Archibald, Haulman and Moody 1983) finds that, in the running shoes market, “[a]fter the ( Runner’s World review) ratings are published, firms adjust their advertising considerably … but they do not appear to adjust prices to any great degree.” Given that price is a more flexible variable than advertising, one would expect a stronger impact of product review on price than on advertising. The observation that product review did not affect firms’ prices is even more puzzling given the importance of Runner’s World in the running shoes market. Runner’s World was the dominant consumer magazine for runners and accounted for more than 70% market share ( Ayer Directory of Publications 1981). The annual running shoes

PAGE 16

6 review from Runner’s World had a significant impact on product demand. For instance, “the New Balance 320 was ranked No. 1 (by the review), and literally overnight the New Balance Company was flooded with orders” ( Runner’s World, Oct. 1980, p. 37). Intuitively, one would expect pricecutting by the losing products after the publication of such an influential product review. It is important to provide a theory that helps us to understand the considerations underlying a manufacturing firm’s decision on its choice of strategic variables. It is more important to understand how firms should adjust their advertising strategy when it is unprofitable to use price as a response variable. In this paper, we consider two popular forms of product reviews: (1) a general description format that provides objective product information while avoiding explicit comparisons of competing products, and (2) a recommendation format that selects winning products to recommend to consumers (e.g., “Editor’s Choice,” “Best Buy”). We consider two types of strategic responses to product review—pricing and advertising—but give special attention to firms’ advertising strategy. We examine a manufacturing firm’s advertising response in two different types of media: the reviewer’s publication (e.g., PC Magazine ), which publishes product reviews for printers, and the non-reviewer’s publication (e.g., PC World ). In addition, to vary the level of its advertising spending, we allow the winning product of a recommendation review to choose whether to use review-endorsed advertising (i.e., advertisements containing third-party award logos such as “Editor’s Choice by PC Magazine ”). We consider consumer heterogeneity in the importance of taste-related product attributes in consumer purchase decision as well as in consumer price sensitivity. We address five specific research questions. First, under what conditions is it optimal for a manufacturing firm to vary its advertising strategy, but not its pricing strategy, in response to third-party product reviews? Second, how should a firm adjust its advertising spending when it is unprofitable to change price? Third, does a winning product of a recommendation review always gain by using review-endorsed advertising to broadcast its superiority? Fourth, how does the

PAGE 17

7 review format (description vs. recommendation) impact a firmÂ’s strategic response? Finally, should a firmÂ’s advertising response differ across different media (the reviewerÂ’s publication vs. other media)? We find that when the number of consumers who value taste-related attributes is sufficiently large, it is best for firms to adjust advertising but not pricing strategy in response to the outcome of a third-party product review. We also discover the interaction effects between firmsÂ’ price and advertising strategies. For example, in response to a recommendation review, whether the losing product will reduce its price may depend on whether the winning product adopts review-endorsed advertising. This is because the review-endorsed advertising can significantly increase a winning productÂ’s advertising effectiveness, which may force the losing product to aggressively compete on price in order to protect its market share. Therefore, broadcasting its superiority via reviewendorsed advertising is not always beneficial for the review-winning product. Our results reveal that a third-party review has two conceptually different effects on a firmÂ’s advertising function. First, a third-party product review generates a substitutive effect because it reduces consumersÂ’ need for advertising information. Second, a third-party review also generates a complementary effect because it can increase or decrease the effectiveness of a firmÂ’s advertising. While the two effects jointly determine a firmÂ’s optimal advertising response to a product review, their strength and direction are variously subject to various product/market/review/media conditions such as the quality of the products, the penetration level of the review information, the format of the product review, and the type of media. As a result, the outcome of a product review (i.e., winning vs. losing) is neither the only nor the most important factor in determining a firmÂ’s optimal advertising strategy. For example, we find that it is not always wise for the recommended products to boost advertising expenditures to spread the good news. We also show that firmsÂ’ strategic response depends on review format. For example, description reviews have the same strategic impact on the products receiving favorable and unfavorable reviews, but recommendation reviews may have different strategic implications for

PAGE 18

8 winning and losing products. Finally, we suggest that firms adopt different advertising strategies in the reviewer’s publication and other media because the impact of product review on readers differs across different media. We conducted an exploratory empirical study based on data from two industries, printers and running shoes. The results provide preliminary support for our theoretical model. The remainder of this chapter is organized as follows. Section 2.2 reviews the relevant literature on third-party product reviews and discusses the two most popular review formats. Section 2.3 presents our model assumptions and setup. Sections 2.4 and 2.5 examine firms’ strategic responses to a description and recommendation product review, respectively. Section 2.6 considers different types of advertising media. Section 2.7 discusses empirical results, and Section 2.8 presents our conclusions. 2.2 Third-Party Product Review The emergence of third-party product reviews is a market phenomenon related to information asymmetry between sellers and buyers—sellers have product information that buyers may not share (e.g., Akerlof 1970, Nelson 1974). Marketing literature has examined how firms can communicate product information to consumers via various marketing strategies (e.g., Gerstner 1985, Wernerfelt 1994, Zhao 2000, Iyer and Soberman 2000, Villas-Boas In press). Economics literature (e.g., Faulhaber and Yao 1989, Lizzeri 1999) has shown that the problem of information asymmetry can also be resolved or mitigated by having informed third parties (infomediaries) convey product information to potential buyers. Several recent studies in the marketing literature have investigated the role of third-party product reviews. Eliashberg and Shugan (1997) show that film critics predict rather than influence movie box office revenue. Reddy et al. (1998) find that newspaper critics have a significant impact on the success of Broadway shows. Shaffer and Zettelmeyer (2002) analyze how the provision of third-party information affects the division of profits in a multi-product distribution channel. Shugan and Winner (2003) investigate the impact of firm advertising on

PAGE 19

9 third-party reviewer’s policy. While these recent studies have advanced our understanding of the impact of third-party review information on product success and firms’ profits, the interaction between third-party product review and firm marketing strategies, particularly the strategic response of manufacturing firms to third-party product reviews, remains a fascinating but underexplored area. Third-party product reviews usually provide product information (e.g., basic features/functions and prices) based on lab testing or expert evaluation using one of several different review formats. Many third-party reviewers adopt a description format to provide detailed attribute facts about a product without making overall recommendations relative to its competing products. For example, Audio , the leading U.S. audio equipment consumer magazine, provides an exhaustive list of audio components available in the U.S. in its October issue. For each component, the magazine provides manufacturer’s suggested retail price along with an extensive description of product characteristics but avoids subjective evaluation or recommendation. Other reviewers adopt a recommendation format that not only provides descriptive product attribute information but also selects winners to recommend to consumers based on overall product performance and prices. For example, PC Magazine , which regularly provides comparative product reviews on various PC-related products, such as desktop and laptop computers, printers, scanners, digital cameras, and software, bestows its "Editor's Choice" seal of approval based on overall test scores and prices. PC World applies the phrase “Best Buy,” Scuba Diving Magazine uses “Tester’s Choice,” and Runner’s World awards its “Five or Four Stars” to recommended products. Table 2-1 presents examples of consumer magazines and websites that provide comparative product reviews within various product categories and indicates which of them uses a recommendation format and which uses a description format.

PAGE 20

10 Table 2-1 Examples of Third-Party Product Reviews Third-Party Reviewer Product Category Review Format Recommendation Logo Outdoor Photographer Photographic Equipment Recommendation “Editor's Choice” PC Magazine Computer, Printer, Digital Camera, Software, Web Site, etc. Recommendation “Editor's Choice” PC World Computer, Printer, Digital Camera, Software, Web Site, etc. Recommendation “Best Buy” Rodale’s Scuba Diving Magazine Scuba Diving Equipment Recommendation “Tester's Choice” Runner's World Running shoes Recommendation “Four & Five Star Shoes” World Tennis Tennis Shoes Recommendation “Best and Good Tennis Shoes” CNET.com Computer, Printer, Digital Camera, Software, etc. Recommendation “Editor's Choice” Edutainingkids.com Toys, Kids Learning Software and Games, etc. Recommendation “Top Pick” ZDNET.com Computer, Printer, Digital Camera, Software, etc. Recommendation “Editor's Choice” Audio CD Player Description N/A Golf Magazine Golf Equipment (Club tests) Description N/A Runner’s World Running Shoes Description N/A 2.3 Assumptions and Model Setting In this section we specify assumptions and model setting. Key notations are summarized in the Appendix. 2.3.1 Third-Party Product Reviewer We make two assumptions about third-party product reviewers. First, we assume the thirdparty reviewer provides accurate product information. We do not consider cases where the thirdparty reviewer may intentionally mislead readers by providing faulty information because we are interested in the impact of product reviews published by well-known publishers such as PC Magazine and PC World, each of which boasts millions of subscribers. Their reputation among readers is critical to these well-known publishers. Further, while advertising is often an important source of third-party reviewers’ business revenue, a large reader base is crucial for attracting advertisers (Chaudhri 1998, Chen and Xie 2003). Hence, publishers have little incentive to favor

PAGE 21

11 larger advertisers at the cost of their credibility to readers. According to the New York Times (Lewis, 1989), among most review publishers “[t]he editorial and advertising staffs are usually separate, and there is no evidence that any major publication has altered or withheld an unfavorable evaluation under pressure from the advertisers.” While it is possible that a reviewer may provide faulty information with or without intention, we leave such cases to future research. Second, we assume partial penetration of review information. Specifically, we assume that in the absence of other channels of information, such as advertising, nonreaders of the reviewer’s publication will not be privy to the information contained in the review. Moreover, among readers of the publication, only percent of them read the review report. For ease of discussion, we call the penetration rate of the product review. 2.3.2 Firms We consider two competing firms, H and L . First, we allow their products to differ in two mutually independent attributes: quality (vertical) dimension and taste (horizontal) dimension (e.g., Lancaster 1966, Liu, Putler and Weinberg 2004). In the quality dimension, consumers agree on the preference order of the attributes. For instance, product reliability is an attribute in the quality dimension since all consumers agree that "the more, the better." In the taste dimension, however, different consumers may have very different preferences for the same attribute, such as design style or color (e.g., Anderson and de Palma 1992). In the quality dimension, product H has the high quality and L has the low quality. In the taste dimension, the two products offer horizontal attributes that match different consumers’ tastes. Second, we assume that the firms advertise their products in the reviewer’s publication. Later, in Section 2.6, we allow firms to advertise in both the reviewer’s publication and in other media and examine how firms’ optimal advertising strategies differ across different types of media. Third, we assume an increasing convex function of advertising cost. Following Meurer and Stahl (1994), we use a cost function, ()ln(1) j jg , where j is the reach level of firm j ’s

PAGE 22

12 advertising. Specifically, as defined by Meurer and Stahl (1994), j is the fraction of consumers who receive at least one advertisement from seller j . The parameter, j , can also be regarded as the probability that a consumer receives the advertisement from product j . The constant, , is a positive coefficient ensuring the interior choices of optimal advertising levels. This cost expression is derived by assuming that advertising technology resembles the classic statistical urn model (Butters 1977). It is in fact the mirror image of the typical concave advertising-response function in advertising literature (e.g., Little 1979). Finally, we allow the products to differ in production cost such that the high-quality product has a higher marginal cost than the low-quality product, HLcc . Without loss of generality, let the marginal cost of L be normalized to 0 and the marginal cost of H be nonnegative, 0Hcc . 2.3.3 Consumers We allow consumer heterogeneity in two dimensions. First, we allow consumers to differ in the importance of taste attributes in their purchase decision. Specifically, we assume that fraction of the consumers have a strong preference about taste-related attributes and make their purchase decisions mainly on the basis of these attributes. The remaining 1 consumers have little concern about the taste-related attribute and make their purchase decisions mainly based on the vertical product attributes. We call these two types of consumers “taste-driven” and “qualitydriven” consumers, respectively. For instance, when making a purchasing decision for SUV automobiles, quality-driven consumers are those who have strong preferences regarding quality attributes such as gas mileage, for which all consumers have the same preference order (i.e., high gas mileage is better). Taste-driven consumers are those who have strong preferences regarding taste attributes such as car design and size, for which different individuals may have different preferences. For example, many consumers strongly prefer Hummer because of its large size and special design despite its poor gas mileage. Let denote the fraction of taste-driven consumers who have matched taste with L (i.e., 1 is the fraction of taste-driven consumers who have

PAGE 23

13 matched taste with H ). Without loss of generality, the size of the whole consumers is normalized to 1. Note that the assumption about consumer heterogeneity in their types implies that if all consumers were fully informed, L would be the preferred product for fraction of consumers and H would be the preferred product for (1)(1) fraction of consumers. Second, we allow consumers to differ in their price sensitivity. Assume that is the fraction of loyal consumers who are insensitive to price and buy only their preferred product (i.e., loyal consumers who prefer H will never buy L regardless of the price of L ). 1 is then the fraction of switching consumers who are sensitive to price and would switch to a non-preferred product if the price is sufficiently low. Let vdenote consumersÂ’ reservation price for their preferred product and v denote the switchersÂ’ reservation price for their non-preferred product. By definition, the loyal consumersÂ’ reservation price for their non-preferred product is zero. This heterogeneity can be interpreted as consumers differing in the marginal rate of substitution between income and product utility (Tirole 1990). We also assume that the high-quality firmÂ’s cost is sufficiently high ( cv ) so that it is unprofitable for H to serve its least profitable segment: switchers who prefer L . This assumption ensures that the high-quality firm will not charge a price below v in the absence of product reviews, which allows us to focus on the more interesting cases and reduces the complexity of analysis. Relaxing this assumption will not alter our results qualitatively. We assume that consumers enter the market with no information about products and that advertising and product reviews are the only two sources of information available to them. We also assume that advertising can convey full information on the taste attribute (e.g., design, color), but not necessarily on quality attributes (e.g., reliability) to consumers. In the absence of product reviews, consumers who receive advertising only from H are aware of the existence of H , but only Hqof them correctly identify H to be a high-quality product. Consumers who receive advertising only from L are aware of the existence of L , and Lqof them incorrectly identify L to be a high-quality product. Consumers who receive advertising from both firms become aware of

PAGE 24

14 the existence of both products, but only qof them correctly identify H as a high-quality and L as a low-quality product. Note that q can be a function of both firmsÂ’ advertising effectiveness, (,)HLqfqq , where /0Hqq ,/0Lqq . Finally, since consumers who receive advertising from the high-quality firm alone are more likely to correctly identify H to be a high-quality product than consumers who receive advertising from both firms, we let Hqq . Similarly, since consumers who receive advertising from the low-quality firm alone are more likely to incorrectly identify L to be a high-quality product than consumers who receive advertising from both firms, we let 1Lqq . 2.3.4 Model Setting We allow firms to compete on both price and advertising. In practice, third-party reviewers often offer product reviews for products within a similar price range because buyer segments are often determined by price level. For instance, when providing reviews on PCs, CNET.com compares different products within three distinct segments: budget PC (low end), midrange PC (middle), and performance PC (high end). RunnerÂ’s World reviews running shoes with categories such as under-$50, $50 to $60, and above $60. Our data also reveal an insignificant price-quality correlation ( p =0.673) for all printer models when they were reviewed by PC Magazine . Given these market observations, we consider the situation where the third party provides information about products in the same price level. Specifically, our analysis focuses on the case where both firms charge a high price (i.e., 00 HLPPv ) in the absence of product reviews. We allow firms to choose whether to compete on price after the publication of the product review. To provide some theoretical justification for the equal-price case, we offer a detailed equilibrium analysis (see the Appendix) to show that 00 HLPPv is an equilibrium as long as advertising alone cannot sufficiently convey product quality information to consumers. An equal-price equilibrium holds under this condition because when advertising is insufficient in conveying quality information, some consumers reached by advertising may not be able to

PAGE 25

15 correctly identify the high-quality product based on firms’ advertising. When enough consumers cannot correctly identify the high-quality firm (or when enough consumers misidentify the lowquality firm), the low-quality firm has little incentive to charge a lower price. This condition is reasonable because third-party product reviews will be neither valuable to consumers nor influential to firms’ strategy if advertising can fully convey product quality information. Note that we do allow both firms to adopt different prices in the presence of product reviews. Let 0,,tdr denote the three cases we examine: in the absence of product review, in the presence of description product review, and in the presence of recommendation review, respectively. Let j t D , j tP, and j t denote firm j ’s demand, price, and profit in case t , respectively (, j HL , 0,,tdr ). Firm j ’s profit in case t is given by : (,,,)(,,,)()() j jijijjijijjj ttttttttttttPPDPPPcg, ,,,, j HLiHLji (2.1) Firm j ’s demand, (,,,) j jiji ttttt D PP, is determined not only by the firm’s advertising and pricing, but also by the availability of a product review. In Section 2.4, we first examine firms’ pricing and advertising strategies in the absence of a product review and then in the presence of a description review. In both cases we model competition with a two-stage game. Firms choose advertising reach levels in the first stage and prices in the second stage. We derive the subgame perfect Nash equilibrium (SPNE) of the twostage game. Firms’ optimal strategic responses to description product review are then derived by comparing the two cases (0t and td ). We examine the recommendation review in Section 2.5. Different from the description product reviews, recommendation product reviews explicitly identify winning products, thus offering the high-quality firm an opportunity to increase its advertising effectiveness by including third-party award logos (e.g., “Editor’s Choice by PC Magazine ”) in its ads. Dean and Biswas (2001) show that carrying third-party recommendation endorsements in the advertising can significantly increase consumers' perceived quality of the high-quality product and firm

PAGE 26

16 advertising effectiveness.1 To allow the high-quality firm to use review-endorsed advertising as a strategic variable, we consider a three-stage game. In the first stage, firm H decides whether to use the review-endorsed advertising. In the second stage, each firm chooses its advertising reach level. In the third stage, firms make pricing decisions. We derive the subgame perfect Nash equilibrium (SPNE) of the three-stage game. FirmsÂ’ optimal strategic responses to a recommendation product review are then derived by comparing the two cases (0t and tr ). 2.4 Description Product Review In this section, we first analyze the demand function in the absence of product reviews and then the demand in the presence of description reviews. Finally, we derive firmsÂ’ optimal responses to third-party product reviews by comparing firmsÂ’ competitive strategies in the two cases. 2.4.1 Analysis of Demand in the Absence of Product Review ( t =0) First, firm j Â’s demand is affected by the size of the informed consumers. In the absence of product review, consumers will make a purchase only when they are reached by a firmÂ’s advertising. There are three groups of informed consumers: (1) those reached only by L , (2) those reached only by H , and (3) those reached by both firms. Let 0 E , 0 A , and 0 B denote these three groups, respectively. The size of each group is: 000 000 000(1)Reached only by 's Advertising Size: (1)Reached only by 's Advertising Reached by both firms' AdvertisingLH HL HLEL AH B ! " # " $ (2.2) Second, firm j Â’s demand is also affected by whether informed consumers consider the advertised product(s) their preferred product(s) since consumers are willing to pay a higher price for their preferred product. For consumers in 0 E (reached only by L ), L is a preferred product for all taste-driven consumers whose tastes match with L (i.e., ) and for all quality-driven 1 While it is also possible to present the positive review information in firmsÂ’ advertising in the case of description product review, as suggested by Dean and Biswas (2001), such endorsement by limited sentence is less striking and effective.

PAGE 27

17 consumers who incorrectly identify L as a high-quality product (i.e., (1)Lq). Let L denote the fraction of consumers in 0 E who consider L to be a preferred product. It is easy to see that (1)LLq. Similarly, let H denote the fraction of consumers in0 A (reached only by H ) who consider H to be a preferred product. H is a preferred product for all taste-driven consumers whose tastes match with H (i.e., (1) ) and for all quality-driven consumers who correctly identify H as a high-quality product (i.e., (1)Hq). Hence, (1)(1)HHq. Finally, let denote the fraction of consumers in 0 B (reached by both products) who consider H to be a preferred product. It is easy to see that (1)(1) q . Since consumers in 0 B are aware of both products, the fraction of consumers who consider L to be a preferred product is 1 . We summarize consumersÂ’ preferences in (2.3): 0 0 0: consider to be preferred; Preference: : consider to be preferred; : consider to be preferred, 1 consider to be preferred. where: (1), (1L H LLHEL AH BHL q ! " # " $ )(1), (1)(1)Hqq (2.3) Finally, firm j Â’s demand is affected by consumersÂ’ price sensitivity (see Table 2-2). For consumers reached only by L (0 E ), both loyals and switchers will buy L if it is their preferred product and if LPv , regardless of their price sensitivity. If L is not their preferred product, loyals will not buy but switchers will buy if LPv . The consumers in 0 E will not buy H because they are not aware of H . The purchase behavior of consumers in 0 A about H is similar to the purchase behavior of consumers in0 E about L . For consumers reached by both firms (0 B ), as shown in Table 2-2, loyals will buy their preferred product if its price is not higher than v, and switchers will buy the product offering a higher positive surplus.

PAGE 28

18 Table 2-2 Consumer Purchase Decision FirmsÂ’ demand can be derived from Table 2-2, and the equilibrium of the two-stage game in the absence of product reviews ( t =0) can be derived by maximizing the profit function (2.1). 2.4.2 Analysis of Demand in the Presence of a Description Product Review (td ) Unlike the case without review, in this case consumers can be informed not only by firmsÂ’ advertising but also by third-party review information. Let d R denote consumers who can directly access the description product review information. As defined in Section 2.3, the size of d R is . Consumers in d R can correctly identify their preferred product regardless of whether they receive firmsÂ’ advertising. Letd E , d A , and d B denote consumers who do not have access to product review information but are reached by advertising only from L , only from H , and from both firms, respectively. The sizes of the four informed consumer groups are given in (2.4): Informed Consumer Consumer Preference Consumer Price Sensitivity Consumer Purchase Decision Loyal () L is preferred (L ) Switcher (1) Buy L if LPv Case 1 Loyal () No Purchase Case 2 Reached by L only 0E L is non-preferred (1-L ) Switcher (1) Buy L if LPv Case 3 Loyal () H is preferred (H ) Switcher (1) Buy H if HPv Case 4 Loyal () No Purchase Case 5 Reached by H only 0A H is non-preferred (1-H ) Switcher (1) Buy H if HPv Case 6 Loyal () Buy H if HPv Case 7 H is preferred ( ) Switcher (1) and and HHL LLH B uyHifPvvPvP B uyLifPvvPvP!#$ Case 8 Loyal () Buy L if LPv Case 9 Reached by both 0B L is preferred (1) Switcher (1) and and LLH HHL B uyLifPvvPvP B uyHifPvvPvP!#$ Case 10

PAGE 29

19 (1)(1)Reached only by 's advertising (1)(1)Reached only by 's advertising Size: (1)Reached only by both firms' advertising Reached by product reviewLH ddd HL ddd HL ddd dEL AH B R !"" #" "$ (2.4) Consumers in ,,dd E Aand d B have the same preferences and reservation prices as consumers in 00,, E Aand 0 B given in (2.3), respectively. Different from consumers reached only by advertising, consumers reached by product review are able to correctly identify product quality. Hence, for consumers in d R , H is a preferred product for all taste-driven consumers whose tastes match with H (i.e., (1) ) and all quality-driven consumers (i.e., (1) ). Let denote the fraction of consumers in d R who consider H to be a preferred product, (1)(1) . We summarize consumersÂ’ preferences in the presence of a description review in (2.5): : consider to be preferred : consider to be preferred Preference: : consider to be preferred, 1 consider to be preferred : consider to be preferred,1 consider to be preferL d H d d dEL AH BHL RHL red where: (1), (1)(1), (1)(1), (1)(1)LLHHqqq ! " " # " " $ (2.5) Consumers in ,,dd E Aand d B have the same purchase behavior as consumers in 00,, E Aand0 B given in Table 2-2, respectively. Consumers in d R who prefer product j have the same purchase behavior as consumers in d B who prefer product j (, j HL ). The equilibrium of the two-stage game in the presence of description product reviews ( t = d ) can be derived by maximizing the profit function given in (2.1). 2.4.3 The Optimal Response to a Description Product Review Comparing the equilibrium strategies in the two cases ( t =0 and t = d ) allows us to derive firmsÂ’ optimal strategic responses to the product review. Since market observations have suggested that third-party product reviews significantly affect a firmÂ’s advertising strategy but not its pricing strategy as discussed earlier, we devote our attention to the equilibrium where firms do not adjust their price but vary their advertising. In Proposition 2.1, we derive conditions under which such

PAGE 30

20 an equilibrium occurs and discuss how to optimally vary one’s advertising strategy when it is unprofitable to change price (see proofs of Propositions in the Appendix). PROPOSITION 2.1 (Description Review) When the size of the taste-driven segment is sufficiently large (ˆd ), both firms (I) Adjust advertising strategy but not pricing strategy in response to a description product review, i.e., ** 0 j j dPP ; (II) Reduce their advertising spending, i.e., ** 0()() j j dgg. Where * j tPand *() j tgare firm j’s optimal price and advertising expenditure in case t, and ˆd is given in (A. 18) in the Appendix. Proposition 2.1 reveals that price will not be used as a strategic variable in response to product review when there are enough consumers who value taste-related product attributes. This is because product reviews reveal information on product quality that often cannot be conveyed fully by firms’ advertising. With the help of product reviews, more consumers can correctly identify H as a high-quality product. As a result, a product review decreases the number of consumers who are willing to pay a high price for L and motivates the low-quality firm to cut price in order to protect its market. Since only the quality-driven consumers have uncertainty about their preferred products, the impact of product review on a firm’s pricing strategy depends on the relative size of the taste-driven vs. quality-driven consumers. When the segment of tastedriven consumers is very small, most consumers make purchase decisions based on product quality. Product review can significantly reduce the low-quality product’s demand and motivate the low-quality firm to reduce price. However, if a sufficient number of consumers care about horizontal attributes, it will be more profitable for both firms to use advertising rather than price as a strategic variable in response to product review. Proposition 2.1 also suggests that both highand low-quality firms should reduce their advertising expenditure in the reviewer’s publication. This is because a third-party product review is an alternative source of product information to the readers of the reviewer’s publication that reduces the value of advertisement to these consumers. This substitutive effect of product review on firm advertising function has a negative impact on marginal advertising return. As a result, a

PAGE 31

21 product review reduces firms’ advertising incentive, and all firms benefit from reducing advertising in the reviewer’s publication. 2.5 Recommendation Product Review As stated in Section 2.3, in the presence of recommendation product reviews, the high-quality firm may want to take advantage of the positive outcome of the product review by including third-party award logos (e.g., “Editor’s Choice by PC Magazine ”) in its ads. To model this, we consider a three-stage game where the high-quality firm decides whether to use review-endorsed advertising in stage 1, both firms choose their advertising reach levels in stage 2, and both firms make price decisions in stage 3. 2.5.1 Analysis of Demand in the Presence of a Recommendation Product Review (tr ) Firms’ demands under a recommendation reviews are the same as those under description reviews if the high-quality firm chooses not to use review-endorsed advertising. Therefore, we need only analyze the case where the high-quality firm adopts review-endorsed advertising. Similar to the description review, in the presence of a recommendation review, there are four groups of informed consumers: ,,,rrrr E ABR. Their definitions and sizes are given in (2.6): (1)(1)Reached only by 's advertising (1)(1)Reached only by 's advertising Size: (1)Reached only by both firms' advertising Reached by product reviewLH rrr HL rrr HL rrr rEL AH B R ! " " # " " $ (2.6) The high-quality firm’s decision to use review-endorsed advertising does not affect consumers reached only by the low-quality firm’s advertising (r E ) because these consumers do not read the high-quality firm’s ads. Such a decision also does not affect consumers reached by the product review (r R ) because these consumers are informed about quality attributes by reading product review, and not affected by advertising. Hence, consumers in r E and r R have the same purchase behavior as consumers in d E and d R , respectively. However, the high-quality firm’s decision to use review-endorsed advertising will affect consumers reached by the high-quality

PAGE 32

22 firmÂ’s advertising (r A and r B ) because review-endorsed advertising allows these consumers to learn the outcome of the product review. Hence, when review-endorsed advertising is used, all three groups, r A , r B , and r R are aware of product quality. ConsumersÂ’ preferences in the presence of a recommendation review and review-endorsed advertising are summarized in (2.7): : consider to be preferred : consider to be preferred Preference: : consider to be preferred 1 consider to be preferred consider to be preferred1 consider to be preferL r r r rEL AH BHL RHL red where: (1), (1)(1) LLq! " " # " " $ (2.7) Table 2-3 Consumer Information and Preference In the Absence of Product Review (0 t ) Reached only by L Â’s ads 000(1)LHE Reached only by H Â’s ads 000(1)HLA Reached by H/LÂ’s ads 000HLB H is preferred L is preferred H is preferred L is preferred H is preferred L is preferred N/A L H N/A 1 In the Presence of a Description Review (td ) Reached only by L Â’s ads (1)(1)LH dddE Reached only by H Â’s ads (1)(1)HL dddA Reached only by H/LÂ’s ads (1)HL dddB Reached by Review dR H is preferred L is preferred H is preferred L is preferred H is preferred L is preferred H is preferred L is preferred N/A L H N/A 1 1 In the Presence of a Recommendation Review (tr ) (With Review-endorsed Advertising)* Reached only by L Â’s ads (1)(1)LH rrrE Reached only by H Â’s ads (1)(1)HL rrrA Reached only by H/LÂ’s ads (1)HL rrrB Reached by Review rR H is preferred L is preferred H is preferred L is preferred H is preferred L is preferred H is preferred L is preferred N/A L N/A 1 1 Note: (1),(1)(1),(1)(1), (1)(1)LLHHqqq *: The case of the recommendation review is the same as that of the description product review if reviewendorsed advertising is not adopted.

PAGE 33

23 Table 2-3 highlights the differences in the size of informed consumers and consumers’ preference among the three cases (0,, tdr ). Note that, the case of recommendation review ( t=r ) without using review-endorsed format is the same as the case of description review ( t=d ) as shown in Table 2-3. 2.5.2 The Optimal Response to a Recommendation Product Review In the presence of a recommendation review, the high-quality firm needs to decide whether to adopt review-endorsed advertising. Examination of firms’ equilibrium strategies and profits when the high-quality firm adopts and does not adopt review-endorsed advertising leads to the following proposition regarding the review-endorsed advertising. PROPOSITION 2.2 (Review-endorsed Advertising) (I) There exists an interaction effect between firms’ advertising and pricing strategies. Specifically, the low-quality firm is more likely to cut price if the high-quality firm adopts review-endorsed advertising than if the high-quality firm does not. (II) Adopting review-endorsed advertising can lead to a lower profit for the high-quality firm. Proposition 2.2 reveals an interesting strategic interaction between firms’ advertising and pricing strategies—the low-quality firm is more likely to engage in price-cutting if the highquality firm adopts review-endorsed advertising. This interaction is the result of the complementary effect of the recommendation review on firm advertising function. A review endorsement has a significant impact on advertising effectiveness. This complementary effect can be positive or negative depending on whether the advertiser is a highor low-quality firm. As shown in Table 2-3, for consumers who do not have direct access to the review information but are reached by H ’s advertising (r A and r B ), H (in r A ) and (in r B ) prefer H when review-endorsed advertising is not used, but (in both r A and r B ) prefer H if such advertising is used. Since H and , review-endorsed advertising has a positive complementary effect on a high-quality firm’s advertising function. Furthermore, for consumers reached by both firms’ advertising (r B ), 1 prefer L when review-endorsed advertising is not

PAGE 34

24 used, but 1 prefer L if such advertising is used. Since 11 , review-endorsed advertising has a negative complementary effect on a low-quality firm’s advertising function.2 The positive complementary effect on H ’s advertising implies that a larger number of consumers (including switchers who would buy L under a sufficiently low price) prefer H when reviewendorsed advertising is used than when it is not used. For this reason, the benefit of price-cutting for the low-quality firm is higher in the former than in the latter case. The negative complementary effect on L ’s advertising implies that a smaller number of consumers (including loyals who are willing to pay a premium price for L ) prefer L when review-endorsed advertising is used than when it is not used. For this reason, the cost of price-cutting for the low-quality firm is lower in the former than in the latter case. Therefore, price-cutting becomes a more profitable strategy for the low-quality firm when the high-quality firm adopts review-endorsed advertising. The interaction effect between firms’ advertising and pricing strategies leads to a surprising finding—the high-quality firm can be hurt by including a third-party endorsement in its advertisement. Review-endorsed advertising is a double-edged sword. On the one hand, it increases the high-quality firm’s advertising effectiveness and leads more consumers to prefer H . On the other hand, review-endorsed advertising increases the low-quality firm’s incentive to cut price. When the size of the taste-driven consumers is small, firm L may have to cut its price aggressively to compete for switchers who prefer H , which can significantly reduces H ’s profit. By comparing the equilibrium strategies in the absence of product review ( t =0) and in the presence of a recommendation product review ( t = r ), we derive the following proposition regarding firms’ optimal responses to a recommendation product review. PROPOSITION 2.3 (Recommendation Review) When the size of the taste-driven segment is sufficiently large (ˆr ), (I) Both firms adjust advertising strategy but not pricing strategy in response to a recommendation product review; (II) It is optimal for the high-quality firm to adopt review-endorsed advertising; 2 Negative complementary effect on low-quality firm indirectly comes from the increased credibility and effectiveness of high-quality firm’s advertising.

PAGE 35

25 (III) Firms adopt asymmetric advertising responses such that (a) the low-quality firm reduces its advertising expenditure, (b) the high-quality firm increases its advertising expenditure if the review penetration rate is sufficiently low, but decreases its advertising expenditure, otherwise. Mathematically, 0 000 0:(*)(*); (*)(*)*1(1*)*/; : (*)(*).LL r HHLHL r HH rLowqualityfirmgg ggif Highqualityfirm ggotherwise ! " " ! %& # " '( # " " " $ $ where (*)j tg is firm j’s optimal advertising expenditure in case t (,,0, j HLtr ), and ˆr is given in (A.28) in the Appendix. Proposition 2.3 reveals that, similar to the case of the description review, if sufficient numbers of consumers care about horizontal product attributes (ˆr ), it is optimal for firms to vary advertising but not price in response to a recommendation product review. In this case, since the low-quality firm has no incentive to cut price, the high-quality firm benefits from using review-endorsed advertising. Proposition 2.3 also shows that, in contrast to the case of a description review, where both firms reduce their advertising expenditures in response to product review (see Proposition 1), in the presence of a recommendation review, the highand low-quality firms may adopt different advertising strategies. Specifically, in response to a recommendation review, while it is best for the low-quality firm to reduce its advertising, the high-quality firm may benefit from increasing or decreasing its advertising spending depending on the review penetration rate. The asymmetric impact of the recommendation product review on highand low-quality firms can be explained by two effects: a substitutive effect and a complementary effect. On the one hand, as an alternative source of product information, a recommendation review has the negative substitutive effect on a firm’s advertising function. On the other hand, as discussed previously, a recommendation product review has a positive complementary effect on highquality firms’ advertising function and a negative complementary effect on low-quality firms’ advertising function when the high-quality firm adopts review-endorsed advertising. It is interesting to note that the magnitude of the substitutive and complementary effects depends on

PAGE 36

26 the size of the review penetration rate, . This is because the substitutive effect applies only to the consumers who directly access review information, i.e., These consumers learn about product quality from reading the product review and therefore do not need to rely on advertising information to make quality inferences. The complementary effect, however, applies to consumers who cannot directly access review information, 1Those consumers still rely on advertisements as their information source in making quality inferences. Hence, the size of affects both substitutive and complementary effects. A larger leads to a stronger substitutive effect and a weaker complementary effect. Clearly, the overall effect of product review on firmsÂ’ advertising function is determined by the combined impact of the substitutive and complementary effects. Proposition 2.3 shows that for low-quality firms, the optimal advertising response to the third-party review is to reduce advertising. This is because the low-quality firm suffers from both a negative substitutive effect and a negative complementary effect, and product review reduces the low-quality firmÂ’s incentive to invest in advertising. Hence, the optimal advertising spending is lower in the presence than in the absence of product review. For high-quality firms, the optimal advertising response depends on the penetration rate of the review information, . For high-quality firms, the substitutive and complementary effects take different directions, and is positively related to the strength of the substitutive effect but negatively related to the strength of the complementary effect. When the review penetration rate is sufficiently high, a large number of readers become aware of product quality and will not benefit from the high-quality firmÂ’s advertising. Although the product review can make a highquality firmÂ’s advertising more persuasive for consumers, this positive complementary effect may be too weak to overcome the negative substitutive effect because the former applies to a very small (1 ) and the latter a very large ( ) proportion of consumers. As a result, when a large number of consumers are aware of the review information, the high-quality firm will benefit by

PAGE 37

27 reducing advertising spending. When the review penetration rate, , is sufficiently low, the high-quality firm’s advertising can be very profitable because it will affect a large number of consumers and have high credibility. In this case, the positive complementary effect can dominate the negative substitutive effect. Therefore, when the review penetration rate is very low, the high-quality firm will benefit by increasing advertising. 2.6 Other Advertising Media In the preceding sections, we assume that the reviewer’s publication is the firms’ only advertising outlet. Now we allow the firms to advertise in two different types of media: the reviewer’s publication and other publications. We also allow the firms to adopt different advertising strategies in response to a product review in these two types of media to determine whether it affects each type of media differently. We call the non-reviewer’s publication “other media”. Let be the percentage of consumers who read the reviewer’s publication; hence, 1 read the other media. To distinguish the cases without other media (discussed in previous sections), we use an “m” subscript to denote all the variables for the case with both the reviewer’s publication and other media. For example, j tm and j tm denote firmj’s advertising reach levels in the reviewer’s publication and in other media in case t ( t =0, d, r ), respectively. Unlike the previous cases, consumers can now get product information via the firms’ ads from both the reviewer’s publication and other media. This will change the composition of different informed consumer groups. For instance, the proportion of consumers who can access review information directly was tR in previous sections, but tmR in this section, where , tdr . As a result, the size of the consumers reached only by H ’s advertising in case t , tmA, is: 00000(1)(1)(1) Size: (1)(1)(1)(1) (1)(1)(1)(1)HLHL mmmmm HLHL dmdmdmdmdm HLHL rmrmrmrmrmA A A ! " # " $ (2.8) The size of the other groups of consumers in different cases also changes accordingly. (See the Appendix for details.) Consumers’ preferences and reservation prices in each group remain the

PAGE 38

28 same. Examining the firmsÂ’ equilibrium strategies and profits leads to the following proposition regarding their advertising strategies in different media. PROPOSITION 2.4 (Other Advertising Media) (I) The existence of other media does not affect firmsÂ’ advertising responses to product review in the reviewerÂ’s publication. (II) In the presence of a description review, firms do not adjust their advertising expenditure in other media. (III) In the presence of a recommendation review, (a) firms do not adjust their advertising expenditure in other media if the high-quality firm does not adopt review-endorsed advertising, and (b) the low-quality firm reduces but the high-quality firm increases advertising expenditure in other media if the high-quality firm adopts reviewendorsed advertising. Taking Propositions 2.1, 2.3, and 2.4 together, we now are able to provide a summary of the firmsÂ’ optimal advertising strategies (see Table 2-4) in the presence of different review formats (description vs. recommendation) and in different advertising media (reviewerÂ’s publication vs. other media). The upper part of Table 2-4 shows the impact of product review on advertising function in terms of substitutive and complementary effects, and the lower part of Table 4 shows the optimal advertising strategy in response to product review. Note that both description and recommendation reviews have the same effect on the firmsÂ’ advertising strategies if the highquality firm does not adopt review-endorsed advertising after the publication of a recommendation review. To focus on the differences between the two types of reviews, in the discussion below, a recommendation review refers to the case where the review is presented in a recommendation format and review-endorsed advertising is adopted. As shown in Table 2-4, the substitutive effect depends on the type of media used. The negative substitutive effect applies only to advertising in the reviewerÂ’s publication and not to advertising in other media because readers of the former have direct access to the review information whereas readers of the latter do not. The complementary effect depends on the type of review format used and the quality of the firmÂ’s product. It applies to the recommendation format but not to the description format. It is positive for high-quality firms but negative for low-

PAGE 39

29 quality firms. Table 2-4 shows that the description format has a symmetric effect such that it affects both firms in the same direction. Conversely, the recommendation format has an asymmetric effect such that it affects highand low-quality sellers in different directions. Table 2-4 Advertising Response to the Third-Party Product Review The combined impact of substitutive and complementary effects leads to the optimal advertising response shown in the lower part of Table 2-4. For example, firms should not change their advertising spending in other media after the publication of a description product review because such reviews have neither a substitutive nor a complementary effect on their advertising in other media. Firms should decrease their advertising in other media due to the negative substitutive effect of the description format. FirmsÂ’ optimal advertising response to the recommendation format depends on the productÂ’s quality, the type of media, and the penetration rate ( of the review information. The optimal advertising response varies for firms with products of different quality because the complementary effect is positive for high-quality Recommendation Format Review (With Review-endorsed Advertising)* Description Format Review High-quality Firm Low-quality Firm High-quality Firm Low-quality Firm ReviewerÂ’s Publication Other Media ReviewerÂ’s Publication Other Media ReviewerÂ’s Publication Other Media ReviewerÂ’s Publication Other Media Substitutive Effect Negative None Negative None Negative None Negative None Complementary Effect Positive Positive Negative Negative None None None None Increase or Same (Low review penetration) Optimal Advertising Response Decrease (High review penetration) Increase Decrease Decrease Decrease Same Decrease Same *: The optimal advertising strategy in the case of recommendation review is the same as that in the case of description product review if review-endorsed advertising is not adopted. :product review can reduce consumers' need for advertising information :product review can strengthen or weaken advertising effectiveness SubstitutiveEffect ComplementaryEffect ! # $

PAGE 40

30 products but negative for low-quality products. Low-quality firms should reduce advertising in all media provided that both substitutive and complementary effects are negative. High-quality firms should increase advertising in other media due to a positive complementary effect. Their advertising strategy in the reviewer’s publication depends on the review penetration rate which decreases the positive complementary effect and increases the negative substitutive effect. Increasing advertising is optimal when fewer people are aware of the review information, but decreasing advertising is optimal when most readers of the reviewer’s publication are informed consumers. It is important to note that some third-party reviewers do not accept firms’ advertising. Rather, they charge consumers for the product review information (e.g., Consumer Reports, Zagat). We call this type of third-party information guidebooks . Although guidebooks carry no ads, our model can be applied to this type of third-party reviewer by making two modifications. First, the fraction of informed consumers who receive advertising from the reviewer’s publication is set to zero (i.e., is zero rather than positive). Second, by purchasing guidebooks, a fraction of readers of other media can directly access the review information (i.e., is defined as the percentage of readers of other media who can directly access the review information, rather than the percentage of readers of the reviewer’s publication who can directly access the review information). These minor modifications do not affect our results on firms’ pricing strategy because firms still face the same tradeoffs in deciding whether price should be a strategic variable in response to product review—the benefit (cost) of using price as a response variable is low (high) when many consumers value horizontal product attributes. Since a guidebook does not accept advertising, firms only need to consider their advertising strategy in other media. Given the modified definition of , firms’ advertising adjustments in other media in the guidebook case is the same as firms’ advertising adjustments in the reviewer’s publication shown in Table 4. In addition,

PAGE 41

31 some guidebooks such as Consumer Reports do not allow firms to carry the publications’ award logos in advertising. Hence, review-endorsed advertising may not be possible. In this case, the impact of a recommendation review will be similar to that of a description review. 2.7 Empirical Evidence To demonstrate some external validity for our theoretical model, we conduct an exploratory empirical study on the impact of third-party product reviews on firms’ marketing strategies. In selecting product categories, the following criteria must be met. (I) Third-party Product Review . (a) The category needs to have the third-party product reviews published by reputable consumer magazines; (b) the reviews must provide comprehensive information on all newly launched models in their respective markets; and (c) the reviews must use different formats in evaluating overall product performance. (II) Media and Advertising Data. (a) The category needs to have two dominating consumer magazines in the industry: the reviewer’s publication and a representative of “other media”; and (b) advertising data for all of the reviewed models must be available from both magazines before and after the review appears. (III) Pricing Data. Pricing data for all of the reviewed models must be available before and after the review appears. We are able to find two product categories that met all of these requirements: printers and running shoes. Although we were unable to obtain detailed price information for running shoes directly, Achibald et al. (1983) collected retailing price data on most of the running shoes models and reported the results of their examination of the impact of the review information on pricing in the running shoes market. In the remainder of this section, we provide detailed information on the data collected from these two industries and discuss our empirical findings.

PAGE 42

32 2.7.1 Data 2.7.1.1 Printer In the printer industry, we collected review data from PC Magazine , which is considered one of the most influential computer consumer magazines in the world (Lohr 1993). PC Magazine was launched in 1982, and in 1999 it had a circulation of over 1.23 million and the highest advertising revenue among all U.S. consumer magazines (Fost 1999). Twenty-two issues of the magazine are published each year. Between 1984 and 1992, a special issue appeared every October or November in which all new models of printers were reviewed. These special issues presented detailed information on current price, print speed, graphics output, and text output for each model. Each of the special review issues also designated some models as “Editor’s Choice” selections based on overall performance and price. We collected the printer review data from one of the special issues (Nov. 14th) published in 1989.3 This issue reviewed 106 new models, designating 20 models an ”Editor's Choice” Selections. Among the models reviewed, 27 exited the market the following year. To rule out the impact of product strategy on advertising spending, we used the remaining 79 models for our analysis. Among these models, we classified the 16 “Editor’s Choice” printers as high-quality products and the remaining 63 printers as low-quality products. To examine the firms’ advertising responses to product reviews, we collected advertising data in two magazines: the reviewer’s publication, PC Magazine, and its leading rival, PC World . These two magazines were the dominant players in their market during the period we studied. As reported in New York Times , in 1988 the subscription size was 502,700 for PC Magazine and 475,000 for PC World , and the circulation of these two magazines accounts for 75% of overall circulation size of the top 5 computer magazines (Fisher, 1988). Moreover, research showed that three-fourths of PC World ’s readers did not read PC Magazine (Lohr 1993). We counted the 3 We used the review data from this issue because we were able to find complete advertising data for 12 months before and 12 months after the publication of this issue to complete our analysis.

PAGE 43

33 number of advertising pages for all printers in both magazines one year before and one year after publication of the review issue of PC Magazine. Since sellers may advertise several models in one advertisement, when counting advertising pages we divided whole pages of this type of advertisement by the number of models and obtained the advertising level for each individual model. In addition, in each year’s review issue on printers, PC Magazine also provides the current prices for all printer models that were reviewed in previous years and are still available on the market. Therefore we were able to obtain price data for all 79 models before and after the review. 2.7.1.2 Running Shoes In the running shoes industry, we collected review data from Runner's World magazine. Runner's World had a circulation of 410,000 in 1981, and since the 1980s it has had the highest circulation among U.S. runners’ magazines. Runner's World published reviews of running shoes once a year in the October issue from 1975 to 1985 and twice a year in the April and October issue from 1986 to present. Before 1985, it employed a recommendation format with a five-star system to rate all shoes and recommended 5and 4-stars shoes to buyers. Since 1985, the magazine has adopted a description review format, providing only attribute facts on products without making recommendations. We collected recommendation review data from the October 1979 issue of Runner’s World . Of 177 shoe models reviewed, 71 models were categorized as “5star” and “4-star”. We classified these models as high-quality and the remaining 106 models as low-quality. We collected description review data on running shoes from the October 1985 issue of Runner’s World . The review presented detailed product information and pictures of 52 models. Advertising data were collected both from Runner’s World and from its leading competitor Runner. According to Ayer Directory of Publications , in 1981 the circulation for Runner’s World was about 410,000 while for Runner it was 85,000, and together these two magazines accounted for 90% of the market share in their market. For the recommendation review format, we collected advertising data by counting the number of advertising pages found in all issues of

PAGE 44

34 RunnerÂ’s World for six months before the review issue (May 1980 to October 1980) and for six months after the review issue (November 1980 to April 1981). Likewise, we gathered advertising data from all issues of Runner for four months before and four months after the review (July 1980 to February 1981).4 For the description review, we collected advertising data from all issues of both magazines for six months before and six months after the reviews appeared (May 1985 to April 1986). For the price data, we were unable to directly obtain the price data for the running shoe models after the review. However, Achibald et al. (1983) collected retailing price data on most of the running shoes models reviewed in the October 1979 issue both before and after the review issue and investigated the impact of the review information on pricing. 2.7.2 Empirical Findings 2.7.2.1 Impact of Product Review on Pricing In the running shoes market, Archibald et al. (1983) report that the manufacturers of running shoes did not adjust prices to any significant degree after the publication of a review by RunnerÂ’s World. In the printer market, our data reveal the same pattern. For example, we find that the publication of the product review by PC Magazine did not have a significant impact on the pricequality correlation ( p =0.17). In fact, the price-quality correlation is neither significant before nor after the publication of the review ( p =0.85 and p =0.95). These observations are consistent with our theoretical results. In our basic model, we assume both high-quality and low-quality firms charge a high price and prove that firms will not adjust their price policy in response to thirdparty product review if the size of taste-driven consumer is sufficiently large. This theoretical result predicts that, given the low price-quality correlation prior to the publication of the product review, the price-quality correlation will not increase after the publication of the product review when sufficient consumers value horizontal product attributes. 4 The length of this time series is constrained by the availability of data.

PAGE 45

35 We use price-quality correlations instead of absolute price levels to evaluate the impact of the product review on the firmsÂ’ pricing strategies also because the latter might be subject to a negative fixed time effect. A recent empirical study of the printer market (Melnikov 2000) provided evidence of such a negative fixed time effect on printer price. If product review had a significant impact on the firmsÂ’ pricing strategy, we would expect an increase in the price-quality correlation after the publication of the product review. This is because a product review increases (decreases) the number of consumers who are willing to pay a high price for the high-quality (low-quality) products, provided that the product review is informative. The fact that the pricequality correlation did not change could suggest that the product review has an insignificant impact on the firmsÂ’ pricing strategies. Our model provides a possible explanation for this interesting observation. Our theoretical results suggest that if there are enough consumers who care about horizontal product attributes, it is unprofitable for firms to use price as a response variable in the presence of a product review. This is because when there are enough consumers whose purchase decisions are driven mainly by their idiosyncratic preferences, price competition is neither necessary (i.e., a product review does not pose a crucial threat to low-quality firmsÂ’ demand) nor efficient (i.e., price-cutting offers little help in gaining market share). When buying running shoes, for some consumers quality attributes such as flexibility, weight, sole traction, and injury prevention are critical to their purchase decision. For other consumers, taste attributes such as design style and fit with foot type (pronators, supinators, or normal) may play a dominant role in their purchase decisions. When making a purchase decision about printers, many consumers focus on quality attributes such as print speed, memory buffer size, and image resolution. However, other consumers may have a strong preference toward a particular brand due to personal taste, or concerns about horizontal attributes such as color, size, design, software compatibility, and special optional functions. Horizontal attributes such as software compatibility and special functions (e.g., Postscript compatibility, Small Computer System

PAGE 46

36 Interface (SCSI), and special size paper handling) were important factors in consumers’ purchase decisions in the early stage of the printer market we examined. We found that firms tended to emphasize these horizontal attributes in their advertisements, especially for products that were not designated an “Editor’s Choice” selection. The observation that product reviews did not significantly affect firms’ pricing strategies in both running shoes and printer markets may suggest that horizontal attributes were important to many consumers in these two markets, which weakened the benefit but increased the cost of using price as a response variable in the presence of product review. 2.7.2.2 Impact of Product Review on Advertising We now examine the impact of product review on the advertising levels of firms in the two markets. Our model (see Table 2-4) suggests that advertising strategy in the presence of product review is determined by four factors: (1) review format, (2) type of media, (3) product quality, and (4) the penetration rate of the review information. In our empirical study, we are able to measure the first three factors in both industries directly but unable to measure the last factor directly. The penetration rate of review information is defined as the percentage of those subscribers of the reviewer’s publication who read the review information. Several factors may affect the review penetration rate. For example, due to limited time and search cost, a reader is more likely to learn the review information if (1) the reviewer publishes a small rather than a large number of issues per year, (2) each issue has fewer rather than more pages, and (3) the annual review is published in a fixed issue (e.g., each October) rather than in different issues every year. Comparing the two third-party product reviewers, we find major differences in these factors. PC Magazine has 22 issues each year, while Runner’s World is issued 12 times a year. Reviews of printers appeared in different issues of PC Magazine in different years, but reviews of running shoes appeared regularly in the October issue of Runner’s World. Furthermore, PC Magazine runs almost 400 pages per issue, but Runner’s World has fewer than 100 pages in each issue. These facts suggest that it is easier for readers of Runner’s World to find and remember the

PAGE 47

37 review information than it is for readers of PC Magazine . In other words, Runner’s World may have a higher review penetration rate, , than PC Magazine. Furthermore, our interviews with readers of Runner’s World suggest that most readers of this magazine are serious runners who have sophisticated knowledge of running shoes and are capable of understanding the technical details of the review report. Given the single-product category focus of the magazine (running shoes), the ease of searching the product review information (fewer issues per year, fewer pages per issue, a fixed issue for product review), and the high degree of reader sophistication, we consider that Runner’s World has a high penetration rate of product review information in developing our predictions for firms’ advertising responses. Table 2-5 Predictions of Advertising Response Based on Table 2-4, we provide 10 predictions about firms’ advertising responses to the third-party product reviews in the two industries (P1a-P6), which are summarized in Table 2-5. We use “P#” to designate the number of our prediction in Table 2-5. We use the symbols, “+”, “-”, and “0” to indicate our predicted changes in advertising level (i.e., increase, decrease, or maintain the same level) after the publication of the review. For example, Prediction 1a implies that the advertising level of high-quality printers in PC Magazine (the reviewer’s publication) is no lower after than before the recommendation review appeared in PC Magazine . Prediction 6 Recommendation Format Description Format High-quality Products Low-quality Products Magazine PC Magazine (Reviewer) PC World (Other Media) PC Magazine (Reviewer) PC World (Other Media) Printers Prediction (P#) / 0 (P1a) (P2a) (P3a) (P4a) Magazine Runner’s World (Reviewer) Runner (Other Media) Runner’s World (Reviewer) Runner (Other Media) Runner’s World (Reviewer) Runner (Other Media) Running Shoes Prediction (P#) (P1b) (P2b) (P3b) (P4b) (P5) 0 (P6)

PAGE 48

38 implies that there is no difference in the advertising level of running shoes in Runner (other media) before and after the description product review appeared in RunnerÂ’s World . These predictions directly follow Table 2-4 (Predictions 1a and 1b assume RunnerÂ’s World has a higher review penetration rate, * m and PC Magazine has a lower review penetration rate. Predictions based on recommendation review assume the use of review-endorsed advertising by the high-quality firms given that most winning products in our data did so). Table 2-6 presents the empirical results (mean advertising levels before and after the publication of product reviews). Recommendation review format As predicted by Predictions 3a and 4a, the advertising levels of low-quality printers are significantly lower in both the reviewerÂ’s publication ( PC Magazine , p <0.01) and in the non-reviewerÂ’s publication ( PC World, p <0.1) after than before the printer review appeared. The same conclusion holds for running shoes. Makers of low-quality running shoes significantly reduced advertising levels in both RunnerÂ’s World ( p <0.01) and Runner ( p <0.01), supporting Predictions 4a and 4b. Proposition 2.3 suggests that, in the presence of a recommendation review, high-quality firms will respond differently than low-quality firms. Proposition 4 suggests that they may adopt different advertising adjustments for the reviewerÂ’s publication and other media. Consistent with Predictions 1b and 2b, high-quality running shoes significantly decreased their advertising levels in the reviewerÂ’s publication ( RunnerÂ’s World , p <0.05) but significantly increased their advertising levels in the non-reviewerÂ’s publication ( Runner , p <0.01). Also, as shown in Table 25, the change of high-quality printersÂ’ advertising in the non-reviewerÂ’s publication ( PC World ) is in the direction predicted (P2a), though not to a significant degree. It is possible that the insignificant result is due to the small sample size of the high-quality printers. The change in high-quality printersÂ’ advertising in the reviewerÂ’s publication ( PC Magazine ) is also in the direction predicted (P1a) but is insignificant. One possible reason for this insignificant result is

PAGE 49

39 that the penetration rate of the review is close to the threshold penetration level. If this is the case, then the insignificant result is consistent with Propositions 2.3 and 2.4, which suggest that the high-quality firms will not change their advertising levels in the reviewerÂ’s publication if the review penetration rate is in the threshold level. Table 2-6 Empirical Results of Advertising Responses Recommendation Format Description Format High-quality Products Low-quality Products Magazine PC Magazine (Reviewer) PC World (Other Media) PC Magazine (Reviewer) PC World (Other Media) Before 1.91 (3.19) 0.69 (2.02) 0.67 (1.63) 0.25 (0.71) After 2.58 (4.84) 1.06 (2.54) 0.15 (0.55) 0.14 (0.53) Printer T-statistic 1.04ns 0.49 ns -2.85*** -1.317* Conclusion P1a: Support P3a: Support P4a: Support Magazine RunnerÂ’s World (Reviewer) Runner (Other Media) RunnerÂ’s World (Reviewer) Runner (Other Media) RunnerÂ’s World (Reviewer) Runner (Other Media) Before 0.92 (1.37) 0.22 (0.55) 0.48 (1.27) 0.12 (0.39) 0.61 (0.98) 0.72 (0.99) After 0.64 (0.87) 0.46 (0.83) 0.14 (0.43) 0.02 (0.10) 0.16 (0.42) 0.56 (1.17) Running Shoes T-statistic -1.86** 2.195*** -2.70*** -2.792*** -3.00*** -.828ns Conclusion P1b: Support P2b: Support P3b: Support P4b: Support P5: Support P6: Support Note: Advertising level is measured by the number of advertising page. ***: p < 0.01, **: p < 0.05, *: p < 0.1. ns: not significant at 0.1 level. Description review format As shown in Table 2-6, the advertising levels of firms that make running shoes are significantly lower after the publication of the description review in the reviewerÂ’s publication ( RunnerÂ’s World , p <0.01), supporting Prediction 5. Furthermore, there is

PAGE 50

40 no significant difference in advertising level before and after the publication of a description product review in the non-reviewerÂ’s publication ( Runner ), supporting Prediction 6.5 In summary, our empirical study based on data from two product categories provides some preliminary empirical support for our theoretical model. However, it is important for future research to provide more empirical evidence from a larger number of product categories. 2.8 Conclusion In this paper, we investigate a new research area: firmsÂ’ marketing strategies in response to third-party product reviews. Specifically, we study how firms should adapt their pricing and advertising strategies to third-party product reviews under different conditions. We develop a normative model to analyze firmsÂ’ strategic responses to product reviews and illustrate our findings with data from computer printers and running shoes industries. This paper contributes to the marketing literature by developing a normative theory that incorporates third-party product reviews into firmsÂ’ marketing strategies. This theory explains the strategic impact of third-party product reviews on firmsÂ’ pricing and advertising strategies, identifies key factors affecting firmsÂ’ decisions, examines the interaction between pricing and advertising response, and derives the firmsÂ’ optimal strategies in the presence of product reviews under various product/market/review/media conditions. Our findings provide the following implications for firmsÂ’ marketing strategies in the presence of third-party product reviews: First, although price is generally considered to be more flexible than advertising, varying price in response to product review can be unprofitable if there are enough consumers who value horizontal product attributes. In such markets, the optimal response to product review is to adjust advertising rather than change price. 5 Since the description review did not provide quality categorization, we also asked a 10-year amateur marathon runner and long-time subscriber to RunnerÂ’s World as the judge to pick 20 out 52 models as highquality models. The analysis based on this categorization is also consistent with our prediction.

PAGE 51

41 Second, firms’ strategic responses need to be review-format specific. For example, with review-endorsed ads, a recommendation product review generates a complementary effect, which increases winners’ but decreases losers’ advertising effectiveness. Such a complementary effect does not apply to description product reviews because description product reviews do not explicitly distinguish winners from losers. As a result, the best advertising response to a product review depends not only on the outcome (i.e., winning vs. losing) but also on the format (i.e., recommendation vs. description). Third, firms need to be aware of the potential interaction between their pricing and advertising responses. For example, it is not always optimal for a winning product of a recommendation review to use review-endorsed advertising to spread the good news about its product. This is because review-endorsed advertising not only increases the winning products’ advertising effectiveness, but also motivates price-cutting by the losing products. As a result, using review-endorsed advertising may hurt winners due to intensified price competition. Fourth, to design advertising response to third-party product review, firms need separate strategies for advertising in the reviewer’s publication versus other media. This is because a product review generates a substitutive effect on advertising in the reviewer’s publication but not (or to a lesser degree) on advertising in non-reviewers’ publication. Fifth, in responding to a recommendation product review, firms with winning products need to pay attention to the penetration rate of the review information. Increasing advertising to spread the good news is profitable for winning products only if the review information has a low penetration rate. When the review penetration rate is high, winners are better off taking the same strategy as losers—reducing advertising. Finally, in general, in the presence of a description review, both firms should reduce advertising spending in the reviewer’s publication but not vary advertising spending in other media. In the presence of a recommendation review, the low-quality firm should reduce advertising spending across all media. However, the high-quality firm should increase advertising

PAGE 52

42 spending in other media, but should always reduce spending in the reviewerÂ’s publication unless the review penetration rate is sufficiently low. While this research improves our understanding of third-party product reviewÂ’s impact on pricing and advertising strategy, many other interesting questions remain unanswered and require further investigation. One limitation of this paper is that we only examine two marketing strategies, i.e., pricing and advertising. Product review may also interact with other marketing strategies such as new product development, product line decisions, and the timing of new product launches. Second, this paper does not consider the possible interaction between manufacturing firmsÂ’ strategic response and the reviewerÂ’s policy. For example, given manufacturing firmsÂ’ response to different types of product reviews (e.g., description or recommendation review), the third-party reviewer can strategically adopt different review formats to maximize its revenue income under different market and product conditions. Our static model limits our capability to address these issues considering the dynamic nature of such strategic interactions. Future research needs to develop dynamic models to examine these issues. Third, one limitation of our empirical study is that it is based on only two product categories. More empirical evidences from more product categories are desirable to fully test our theory. Data from more product categories will not only test the generality of our theoretical findings, but also allow the examination of the impact of product characteristics on firmsÂ’ advertising response to third-party product reviews. Furthermore, it is important to develop appropriate measurements for the degree to which consumers are taste-driven vs. quality-driven for a given product. Finally, it is desirable for future empirical studies to collect data on additional control variables such as the number of new product lunched, the third-party reviewerÂ’s market share, and manufacturing firmsÂ’ financial performance. There are many other interesting questions that might be addressed in future research. For instance, what can a firm do to increase the chance of obtaining a favorable third-party review?

PAGE 53

43 How do consumers process information from multiple simultaneous product reviews? Finally, what are the strategic impacts of other types of independent product information, such as reviews posted by consumers?

PAGE 54

44 CHAPTER 3 ONLINE CONSUMER REVIEW: A NEW MARKETING FUNCTION 3.1 Introduction Internet and information technology are creating a new opportunity for consumers to share their product evaluations online (Avery, Resnick and Zeckhauser 1999). Amazon.com started to offer consumers an option to post their comments on products on its website in 1995. Currently, Amazon.com has about 10 million consumer reviews on all its product categories, and these reviews are regarded as one of the most popular and successful features of Amazon (Harmon 2004). In recent years, an increasing number of online sellers (e.g., BevMo.com, BN.com, cduniverse.com, circuitcity.com, GameStop.com, computer4sure.com , c-source.com half.com, goodguys.com, wine.com) are adopting the same strategy. These online sellers invite users of their products to post personal product evaluations on the sellersÂ’ websites or provide their customers consumer review information offered by some third-party sources such as Epinions.com and CNET.com. Online consumer reviews are common for many product categories such as apparel, books, electronics, games, videos, music, beverages, games, and wine. Recent evidence suggests that consumer reviews have become very important for consumer purchase decisions and product sales. A study by Forrester Research finds that half of those who visited the retailer sites with consumer postings reported that consumer reviews are important or extremely important in their buying decisions ( Los Angeles Times , Dec. 3, 1999). Based on the data from Amazon.com and BN.com, Chevalier and Mayzlin (2003) find that online book reviews have significant impact on book sales. Online consumer review is an emerging independent product information resource with growing popularity and importance. It has generated considerable attention in practitioners and

PAGE 55

45 popular presses. To better understand the fundamental role of this new information channel in the market place and its strategic implications to online marketers, more academic research is urgently needed. In contrast to traditional sellers, an online seller can now provide two types of product information to its customers. It can (1) offer seller-created product information to consumers via its website or other media such as advertising, and (2) offer consumer-created product information by allowing consumers to post their comments on the seller’s website. One important difference between the two types of product information is the degree of information credibility. Consumer-created information is likely to be more credible than sellercreated information because credibility of information is often positively related to the trustworthiness of the information source (Wilson and Sherrell 1993). Several recent studies have begun to examine online consumer-created information from the perspective of information credibility. For example, Dellarocas (2003) examines the relationship between online consumer feedback information and an unknown seller’s reputation. Mayzlin (In press) studies the credibility of the promotional messages in online chat rooms and the implication of such new information channels on sellers’ profitability. These studies have advanced our understanding of consumer-created information. This paper is different from these recent studies in that it focuses on an under-explored but nonetheless important function of consumer reviews. We argue that online consumer reviews can serve as free “sales assistants” to help consumers identify the products that best match their idiosyncratic usage conditions. Consumer-created review information can differ from seller-created information in the degree of relevance to consumers. Consumer-created information is likely to be more relevant to consumers than seller-created information (Bickart and Schindler 2001). Seller-created product information is more likely to be product-oriented since it often describes product attributes in terms of technical specifications and measures product performance by technical standards. It

PAGE 56

46 may not be possible for all consumers to map product attributes with their usage conditions based on the seller-created information. In contrast, the consumer-created product information is, by definition, user-oriented. It often describes product attributes in terms of usage situations and measures product performance from a user’s perspective (Bickart and Schindler 2001). More importantly, since this information is posted by users with different idiosyncratic preferences, technical backgrounds, levels of product knowledge, and usage situations, such information has the potential to be relevant to a wide variety of consumers. While sellers can also be motivated to supply highly relevant product information to consumers, sometimes it may be too costly or even impossible for a seller to acquire complete usage knowledge or to provide all possible mappings between product attributes and usage conditions, especially when consumers have multidimensional preferences and extremely idiosyncratic usage conditions. The essential contribution of this paper is to investigate the strategic implication of this information relevance advantage of consumer-created information. We propose that online consumer reviews can serve as a marketing function—they can provide relevant matching information to all kinds of consumers, including those who fail to benefit from the information provided by the seller. This marketing function is particularly important for less sophisticated consumers. Due to different levels of expertise, consumers have different information processing capabilities in diagnosing product information (Alba and Hutchinson 1987). For this reason, seller-created product information may be valuable only to more sophisticated consumers (i.e., technical experts). Consumer-created product information, however, can help less sophisticated consumers (i.e., technical novices) in finding their best-matched products. Note that experts, who are able to benefit from seller-created product information, are more likely to adopt a new product earlier than novices (Mahajan, Muller and Srivastava 1990). This suggests that the seller can benefit from offering consumer-created product information because the user-oriented information posted by the pioneer expert consumers is valuable to novice consumers. In this

PAGE 57

47 sense, the seller can create, at minimal cost, a “sales assistant” by allowing consumers to share their usage experiences via online self-posting consumer reviews. However, the marketing function of consumer reviews does not come without inherent costs. Lewis and Sappington (1994) have shown that, when providing seller-created product information, a seller may benefit from only providing partial product information but not full information (i.e., a partial information strategy can be optimal) even if information supply is costless. However, since the seller has little control over the content of the consumer-created information, by allowing consumers to post their product evaluations on the seller’s website, the seller will no longer be able to fully control the information content available to consumers and thereby fail to implement a partial information strategy. Furthermore, the seller is unable to ensure the accuracy of the consumer-created product information, and inaccurate review information may mislead consumers. This suggests that offering consumer-created product information may hurt the seller. This paper examines when an online seller should offer consumer reviews to its customers, paying special attention to the interaction between the seller’s consumer review supply decision and its other marketing strategies. We address four specific research questions. First, when should an online seller provide consumer reviews to its customers? Second, how does a seller’s product assortment strategy interact with its consumer review supply decision? Third, how does the seller’s consumer review supply strategy interact with its seller-created information supply decision (i.e., information content strategy)? Fourth, what is the optimal timing for the seller to offer consumer review information for a product? We develop a normative model to address these questions and our results reveal several new findings. First, our results show that the seller will offer consumer reviews for a product only when the reviews are sufficiently informative and the seller’s product matches the preferences of a sufficient number of consumers. Second, we discover that sellers with a wide assortment of products benefit more from offering consumer reviews than those with a narrow assortment.

PAGE 58

48 Third, we show that the sellerÂ’s information strategy on consumer-created information interacts with its information strategy on the seller-created information. Specifically, consumer review supply decision increases the sellerÂ’s incentive to provide full product information. Fourth, we find the sellerÂ’s consumer review supply decision also depends on the product cost and the consumerÂ’s expertise level. Finally, we show that if it is possible for the seller to decide the timing to offer consumer reviews at the individual product level, it may not always be optimal to offer consumer reviews at a very early stage of new product introduction, even if such reviews are available. We conducted an empirical study based on data from online sellers in several different product categories. The empirical results are consistent with the predictions of our theoretical model. The remainder of the paper is organized as follows. Section 3.2 reviews the relevant literature. Section 3.3 presents our basic model of consumer review supply decision. Section 3.4 studies how firm product assortment strategy affects its consumer review supply strategy. Section 3.5 considers a general model and examines how a sellerÂ’s decision to offer consumer reviews interacts with its other information supply strategy. Section 3.6 discusses sellerÂ’s optimal timing decision to provide consumer reviews. Section 3.7 presents the results of our empirical study, and Section 8 provides conclusions and discusses future research. 3.2 Related Literature First, this paper contributes to the emerging literature of independent product information (e.g., Eliashberg and Shugan 1997). In many product categories, buyers and sellers typically have asymmetric information. Sellers have private product information that buyers may not share, and vice versa. Akerlof (1970) shows that such information asymmetry may lead to market failure. To address this problem, traditional economics and marketing literature shows how firms can use different marketing strategies to communicate product information to consumers (e.g., Milgrom and Roberts 1984, Gestner 1985). Recently, some authors have studied product information from

PAGE 59

49 different independent sources. In general, there are two types of independent product information. The first type is third-party product review information from various third parties such as consumer magazines and websites. Third parties such as Consumer Reports conduct independent product testing and publish evaluations to consumers. Consumer-generated review is the second type of independent information. Over the past decade, the Internet and development of information technology have greatly reduced information delivery and search costs. Hence, independent product information has become widely accessible to and increasingly important for consumers. How this second information channel for consumers affects firmsÂ’ marketing strategies has become a very important research issue. Concerning independent product information from third parties, in Chapter 2, we study how third-party product reviews interact with firmsÂ’ marketing strategies. Specifically, they show when firms should choose advertising instead of pricing as the strategic responding variable, and how they should adjust their advertising formats and spending to thirdparty product reviews. Other emerging literature addresses the independent product information from online consumers (e.g., Avery et al. 1999). Current limited research on online consumer reviews focuses on the credibility function of the consumer-created information. Dellarocas (2003) shows online consumer feedback information on the seller (instead of products) can help to build reputation for unknown sellers in marketplaces such as eBay.com. Mayzlin (In press) demonstrates the credibility of firmsÂ’ promotional messages in online chartrooms and studies the implication of such new information channels on sellersÂ’ profitability. In this paper, we argue that consumer reviews can work as free sales assistants for online marketers and investigate consumer reviewsÂ’ marketing function on providing consumers matching information to map their usage conditions with product attributes. Specifically, we characterize circumstances where a seller can benefit from offering consumer review information to its customers, and show how consumer review supply decision interacts with the sellerÂ’s other marketing strategies.

PAGE 60

50 This paper also relates to the word-of-mouth (WOM) literature (e.g., Brown and Reingen 1987, Godes and Mayzlin In press). There are two major distinctions between online consumer reviews and traditional WOM in this paper. First, the influence of traditional WOM is typically limited to a local social network (Brown and Reingen 1987, Biyalogorsky, Gerstner, and Libai 2001, Shi 2003). In contrast, the impact of online consumer reviews can reach far beyond the local community since any consumer all over the world can access a consumer review via the Internet. Second, traditional WOM is not a decision variable for the seller. However, for consumer reviews, an online seller can decide whether and when to provide them to its customers on its website. The seller (e.g., Amazon.com) can provide an option on its website to allow consumers to post their reviews along the listed product. Sometimes, the seller (e.g., csource.com, half.com) can also license consumer reviews from intermediaries such as Epinions.com, and decide when to post them on its website. From a theoretical perspective, this paper is related to the agency theory literature (e.g., Sappington 1991, Stiglitz 2002). In most of the agency models, the information structure between two parties is exogenous. Moreover, the information asymmetry between the principal and agent is one-sided. The private information resides either with the principal (signaling models, e.g., Spence 1973) or with the agent (moral hazard or adverse selection models, e.g., Holmstrom 1979, Baron and Myerson 1982, Sappington 1983). Lewis and Sappington (1994) propose a model with two-sided endogenous information structure to examine a sellerÂ’s information supply decision, and show the conditions under which the seller may provide different amounts of information to consumers. In their model, the seller (principal) has private product information, but no information about consumer tastes. In contrast, the consumers (agents) have private information about their own tastes but not product attributes. The information structure is endogenous. The seller can manipulate the degree of information asymmetry by providing different amounts of product information. However, in their model, there is only one information channel between the two parties. In contrast, in our model, the seller controls two information channels and can

PAGE 61

51 provide two types of information, the seller-created information and consumer review information. We study here how these two types of information interact with each other, and what type of product information a seller should supply to its customers. Finally, this paper also relates to the marketing communications particularly informative advertising literature (e.g., Grossman and Shapiro 1984). To date, very few studies have examined a firm’s strategic decision on information content for its marketing communications. Wernerfelt (1994b) and Simester (1995) have investigated when and how firms should integrate price information in their advertising. In Chapter 2, we examine a firm’s advertising format strategy in the presence of third-party product review, and find that using review-endorsed advertising (i.e., advertisements containing third-party award logos) to broadcast its victory can hurt the winning product of a product review. In this paper, we study a firm’s information content strategy by investigating how much and what type of product information a seller should provide to its customers. 3.3 Basic Model In this section, we first specify key assumptions and setup for our basic model. We then present our basic model and discuss when the seller should offer consumer reviews. 3.3.1 Basic Model Assumptions and Setting 3.3.1.1 Model Setting In the basic model, we consider a single seller1 offering a single product assortment. We will examine the case of multiple assortments in Section 3.4. Let c denote the marginal cost of the product. 1 The seller’s monopoly position mainly results from consumers’ loyalty and limited search. Recent studies have demonstrated online consumers’ loyalty and limited search for online sellers. For example, Johnson et al. (2004) present empirical evidence that consumer online search is very limited during the shopping process. On average, consumers visit 1.2 book sites and 1.3 CD sites in each category. The monopoly model can help us understand the fundamental impact of the new information channel— online consumer review on firm marketing strategies.

PAGE 62

52 In our basic model, we allow consumer heterogeneity in two dimensions: (1) taste, (2) time to enter the market. The entire consumer size is normalized to one. First, we allow consumers to differ in their preferences toward the sellerÂ’s product. Given a product, some consumers will find that the product matches their needs and others not. Hence, there are two types of consumers in the taste dimension: matched and unmatched, i.e., ,mmTTT. Given full information on product attributes, the matched consumers find the sellerÂ’s product fits their usage conditions, and have a high valuation mv on the product. In contrast, the unmatched consumers find the sellerÂ’s product does not match their usage conditions, and have a low valuation mv on the product. Let , 1 denote the fraction of the matched and unmatched consumers. Without loss of generality, we assume mv= 0. In the absence of product information, all consumers have the same expected value, /2mvv. We further assume that the product cost is sufficiently low so that the seller can make a profit and serve some consumers in the absence of product information, /2mcvv. We will relax this assumption in section 3.5 and discuss the case where the seller can make a profit and consumers make purchases only if sufficient product information is available. Second, we allow some consumers to enter the market earlier than others. According to the diffusion literature (Rogers 1996), some consumers (i.e., innovators) are more eager to seek and try new products than others (i.e., majority consumers) due to their personality. Let denote the fraction of consumers who are innovators and enter the market earlier than majority consumers. In our basic model, we assume all consumers are novices and lack product expertise to map the product attribute information with their usage conditions or needs (e.g., Werfernelt 1994a). Due to their limited information processing capability resulting from a lack of product expertise (Alba and Hutchinson 1987), consumers are unable to identify matching or mismatching products simply based on the attribute information offered by the seller. However, the novice consumers can identify matching or mismatching product by learning from the experiences of some existing

PAGE 63

53 users. In other words, in our basic model, all consumers cannot process seller-created information but consumer review information. We will allow consumer heterogeneity in their product expertise levels in section 3.5. 3.3.1.2 Information Structure We allow a two-sided information asymmetry between the seller and consumers. The seller has private product information, but has no information on consumer characteristics. Consumers know their own tastes, but have no information on product attributes. We also allow the information structure to be endogenous. In our basic model, since all consumers are novices and cannot process seller-created information, consumer review is the only information channel for consumers to learn product information even if the seller provides product attribute information. We will allow some consumers to learn product information from the seller-created attribute information in section 3.5. In the basic model, the seller can alter the information structure by deciding whether to offer consumers the option to post their product evaluations on its website. Note that the seller is unable to ensure the accuracy of the consumer-created information. Due to the anonymity of online consumer review, the review information can also come from biased sources such as a disgruntled employee (Piller 1999). We use a measure, , to characterize the informativeness of consumer-created review information, where 01 . A higher informativeness corresponds to better information in the sense of Blackwell (1951). The consumer review information is perfectly accurate and informative when 1 , and purely uninformative when 0 . We assume the seller cannot or does not want to manipulate the review information perhaps because of reputation concerns2 or regulatory requirements. In other words, is exogenous in our model. Both parties can observe the review informativeness. 2 For instance, New York Times (Feb. 24, 2004) reported how a technological accident on Amazon.com revealed the true identity of a best-seller book authorÂ’s self-promoting reviews and jeopardized his reputation.

PAGE 64

54 Therefore is public knowledge for both sides. Without loss of generality, we assume the seller’s information supply cost is zero considering the significantly reduced costs of collecting and distributing information via the Internet (Avery et al. 1999). 3.3.1.3. Model Timing Our basic model has two periods, 1,2t. At the very beginning of the game, the seller makes a decision as to whether or not to allow consumers to post their reviews on its website. Innovators enter the market in the beginning and exit at the end of period 1 (1t). If the seller chooses to allow consumer self-posting on its website, then consumer review information will be available to potential buyers after period 1. The majority consumers enter the market at the beginning and exit at the end of period 2 (2t). In each period t, the seller can set a different pricetP. 3.3.2 Basic Model 3.3.2.1 In the Absence of Consumer Reviews In the absence of consumer reviews, all consumers have expected value /2mvv. The seller’s expected profit is ˆ /2mvc (3.1) 3.3.2.2 In the Presence of Consumer Reviews In period 1, innovators enter the market. All innovators have expected value /2mvv. Recalling is the fraction of the innovators among all consumers, the seller’s expected profit in period 1 is 1(/2)mvc. In period 2, majority consumers enter the market. With the available product information from consumer reviews, the probability that a matched consumer’s valuation is mv(or an unmatched consumer’s is 0) is an increasing function of the review informativeness, , and approaches to 1 when 1 . Following Lewis and Sappington (1994), the relevant probability function can be formalized as ()(1/2/2) q . In addition, the probability that a matched

PAGE 65

55 consumer’s valuation is 0 (or an unmatched consumer’s is mv) is a decreasing function of the review informativeness, , and approaches to 0 when 1 . Similarly, the relevant probability function can be formalized as ()(1/2/2)q . Therefore the expected values are (1/2/2)mTmVv for the matched consumers, and (1/2/2)mTmVvfor the unmatched consumers. The seller can charge a high price at 2mTPV and only serve type mT consumers or a low price at 2mTPVto gain the demand from all consumers. It is optimal for the seller to charge a high price when is sufficiently high, and a low price when is sufficiently low. Therefore, the seller’s expected profit in period 2 is 2[(1/2/2)](1),(1)(2)/(1) () [(1/2/2)](1),(1)(2)/(1)mmm mmmvcvcv vcvcv ! # $. Hence, the seller’s overall profit over two periods in the presence of consumer reviews is (/2)[(1)/2](/2)(1),(1)(2)/(1) () (/2)(1)/2,(1)(2)/(1)mmmmm mm mmvcvcvcvcv vcvvcv ! # $ (3.2) From equation (3.1) and (3.2), we have the following proposition regarding when the seller should offer consumer reviews to its customers. PROPOSITION 3.1 (Consumer Review Supply Decision) The seller’s consumer review supply decision depends on the degree of review informativeness and the number of matched consumers. Specifically, supplying consumer review information increases the seller’s profit if two conditions hold: (a) The review information is sufficiently informative; (b) The seller’s product matches sufficient consumers’ preferences. Mathematically, ˆ () if and , where (1)(2)/(1)mmvcv and 2/(1)mcv. Consumer reviews provide matching information for consumers. It can increase the matched consumer’s willingness to pay and decrease the unmatched consumer’s valuation. Due to the information asymmetry between consumers and the seller, the seller does not have information on consumer taste. Offering consumer review information can help the seller to charge a premium

PAGE 66

56 price to screen out the unmatched consumers with lower valuations, and only serve the high-value matched consumers. However, the seller can benefit from offering consumer review information only if the review is sufficiently informative. When the review is not informative enough, the increase in willingness to pay of the matched consumers is too small to provide benefit to the seller from only serving the matched consumers. The seller’s review supply decision is in fact a tradeoff between its profit gain from the matched consumers and its loss from the unmatched consumers. If the product is a mass-market product (i.e., the number of matched consumers is sufficiently large), the seller’s profit loss from the unmatched consumers is very small and dominated by the profit gain from the matched consumers. Hence, the seller will offer consumer reviews. 3.4 Consumer Review and Product Assortment Strategy In the basic model, the seller is a single assortment marketer in a product category. Now we allow the seller to carry multiple assortments and discuss how the seller’s product assortment strategy (e.g., Shugan 1989) will affect its consumer review supply decisions. Specifically, consider a seller who carries two product assortments matching two types of consumers respectively. To distinct from the basic model, we use an “ M ” subscript to denote the variables in the case of multiple assortments. 3.4.1 In the Absence of Consumer Reviews Without consumer review information, all consumers have the same expected value /2mvv for the two assortments. They randomly choose one. Seller’s expected profit is ˆ /2Mmvc . 3.4.2 In the Presence of Consumer Reviews If the seller supplies consumer review information, consumer valuation depends on the time when they enter the market. In period 1, innovators enter the market. Their expected value is /2mvv and the seller’s expected profit from period 1 is 1(/2)Mmvc.

PAGE 67

57 In period 2, majority consumers enter the market and can obtain product information from reading the reviews posted by the innovators. As explained in the basic model, consumer review information affects these consumers’ expected value. The expected value is ()(1/2/2)mTmVvfor the matched consumers and ()(1/2/2)mTmVvfor the unmatched consumers, respectively. The seller earns a highest profit by selling each assortment to only matched consumers at a price2(1/2/2)mPv. Since all consumers can find their match, the seller’s expected profit from period 2 is 2()[(1/2/2)](1)Mmvc . Hence, the seller’s overall profit is ()(/2)(1)/2Mmmvcv. The comparison of the seller’s profits in the absence and presence of consumer reviews, ˆM and ()M , reveals that supplying consumer review information is always more profitable than not supplying such information (i.e., ˆ and , ()MM). Comparing with the result of a single assortment (Proposition 3.1), we have the following proposition on how the seller’s assortment strategy affects its consumer review supply decision. PROPOSITION 3.2 (Consumer Review and Product Assortment Strategy) The seller with wider product assortments is more likely to benefit from supplying consumer review information than the seller with narrower product assortments. As we discussed in the last section, a seller’s decision to provide consumer reviews is a tradeoff between its profit gain from the matched consumers and its loss from the unmatched consumers. When the seller carries wider assortments matching different types of consumers, it is more likely that all consumers will find a perfect match from the seller’s offerings. Therefore, wider product assortments lower the seller’s profit loss from the unmatched consumers because of consumer reviews. At the same time, the seller can still enjoy a profit increase from the increased willingness to pay of the perfectly matched consumers as a result of the consumer review information.

PAGE 68

58 3.5 Information Supply Strategy In our basic model, the product cost is sufficiently low so that the seller can make a profit from consumer’s purchase even in the absence of product information (i.e., /2mcvv ). However, in reality, for many products, the product cost is sufficiently high so that the seller can profit from a consumer’s purchase only if enough product information is provided for consumers (i.e., /2mcvv ). In addition, in the basic model, all consumers are novices and have no knowledge to match their usage conditions with the product even in the presence of product attribute information. Hence, the seller-created product information has no influence on consumers. In this section, we consider a generalized model where both seller-created product attribute information and consumer review information can influence consumer purchase decision, and examine how the seller’s information supply strategies concerning these two types of information interact with each other. 3.5.1 A Generalized Model Setting In this section, we allow consumers to differ in their expertise and knowledge about the product (Alba and Hutchinson 1987). We consider two consumer segments: an expert segment and a novice segment. Let S denote the segment, where SN (Novices), E (Experts). Expert consumers are knowledgeable about the product and are able to correctly map their usage situations with the product attributes based on the attribute information offered by the seller. Novice consumers lack product knowledge to map the product attribute information with their usage conditions or needs. They are unable to identify the matching or mismatching product simply based on attribute information offered by the seller. The novice consumers can, however, identify a matching or mismatching product by learning from the experiences of some existing users. Let 01 denote the percentage of expert consumers. Let ˆSand Sdenote the seller’s profit from segment Sin the absence and presence of consumer reviews.

PAGE 69

59 Consumers can learn product information from two sources: (1) seller-created product attribute information and (2) consumer review information. Both experts and novice consumers are able to identify their matching/mismatching attributes based on the consumer review information, but only the expert consumers can identify their matching/mismatching attributes based on the seller-created information. The seller needs to make two information supply decisions. In addition to the consumer review supply strategy, the seller needs to decide its seller information content strategy on product attribute information. That is, the seller must decide how much attribute information to provide. Specifically, consider a product with two attributes, 1aand 2a. The seller has a choice of providing attribute information on both attributes or only on one of the attributes. We call the former full information content strategy and the latter partial information content strategy . Let ,FP I II denote the information content decision, where F I and P I present the case when the seller adopts full and partial information content strategy, respectively. Due to the sellerÂ’s reputation concern, we assume the seller-created product attribute information is accurate. In addition, in our basic model, there are two types of consumers in the taste dimension: matched and unmatched consumers. In reality, for any product with multiple attributes, consumers can partially match the product in the sense that they find a match on some attributes and a mismatch on others. Specifically, for a two-attribute product, consumers can be categorized into four types according to their taste-matching situations with the product: fully matched type mmT (matching on both attributes), partially matched consumers mmT and mmT (matching on attribute 1a or 2a), and fully unmatched consumers mmT (matching on neither attribute). For simplicity, we assume the sizes of four types are equal. Let 0, ,FPvvv denote consumer valuations for fully matched, partially matched and fully unmatched consumers under full information on product attributes, respectively. Without loss of generality, we assume 00 v . The consumerÂ’s expected value in the absence of product information is (+2)/4FPvvv . We assume the product

PAGE 70

60 cost is sufficiently high so that the seller can profit from the consumerÂ’s purchase only if enough product information is provided, i.e., (+2)/4FPcvvv . To distinguish consumer heterogeneity in the taste and expertise dimensions discussed above, hereafter we refer to consumers with a different taste as a different consumer type and consumers with a different expertise level as a different consumer segment . Precisely, as described earlier, there are four types of consumers with different preference-matching situations (i.e., ,,,mmmmmmmmTTTTT ) and two segments of consumers with different expertise levels (i.e., ,SEN ). The consumer preference heterogeneity and consumer expertise heterogeneity are orthogonal, i.e., for both the expert and novice consumer segments, there are four types of consumers with different preference-matching situations. Finally, since experts are more likely to search for new product information than novices (Brucks 1985, Alba and Hutchinson 1987), the former are likely to enter the market earlier than the latter. Mahajan et al. (1990) empirically demonstrate that expert consumers are more likely to read product related advertising and adopt products earlier than novice consumers. They also find that innovators are usually the expert consumers. Hence, we assume experts enter the market earlier than novice consumers. Since experts can be either innovators or majority consumers depending on their personal characteristics, we further allow some experts to be majority experts who enter the market later than innovator experts but earlier than novice consumers. Let denote the fraction of expert consumers who are innovators (i.e., there are innovator experts and 1 majority experts). Specifically, we consider three time periods, 1,2,3t . In each period, some consumers enter the market at the beginning and exit at the end of the period. The order of entry is: innovator experts (1t ), majority experts (2t ), and novice consumers (3t ). The seller makes information supply decisions at the very beginning of the three-period game. Due to concern over reputation, we assume the seller is committed to its information supply decisions once such decisions are made. However, the seller can adjust its pricing decision in each period.

PAGE 71

61 Figure 3-1. Seller Information Supply Strategy The sellerÂ’s information supply strategy is a two-stage decision, as shown in Figure 3-1. In stage 1, the seller decides whether to supply consumer review information to its customer. In stage 2, the seller decides whether to offer full or partial information to consumer via the sellercreated information. The order of the decision is determined based on the fact that the consumer review information supply decision is often less flexible than the sellerÂ’s own product information content supply decision. We solve this two-stage decision problem using backward induction. We first derive the optimal seller information content strategy (stage 2) and then the optimal consumer review supply strategy (stage 1). 3.5.2 Seller Information Content Strategy (Stage 2) In this section, we analyze the sellerÂ’s information content strategy, i.e., whether a full or partial attribute information supply strategy is optimal for seller-created information. We first derive the optimal information content strategy in both the absence and presence of consumer reviews. By comparing the sellerÂ’s optimal strategy in these two cases, we can determine the impact of a consumer review supply decision on the sellerÂ’s information content strategy. By assumption, only expert consumers can process the seller-created attribute information. The sellerÂ’s information content strategy will not affect novice consumersÂ’ valuation and the Consumer Review Supply Decision (Stage 1) Seller Information Content Decision (Stage 2) Seller Providing Consumer Reviews Not Providing Consumer Reviews Supplying Full Attribute Information Supplying Partial Attribute Information Supplying Full Attribute Information Supplying Partial Attribute Information

PAGE 72

62 seller’s demand from novice consumers. Hence, to study the seller’s information content strategy, we need only focus on its profit from the expert segment (E), i.e., the sum of its profits from innovator experts in period 1 and from majority experts in period 2. Let (,)S TVI denote the expected value of type Tconsumers in segment Sgiven attribute information I and review informativeness , where ,,,mmmmmmmmTTTTT and ,SEN . Therefore, the seller’s problem concerning information content decision can be formalized as follows: 1122 , max (,,)(,)(,,) s. t. (,) tE t IP E TIPIPIP VIc (3.3) where ,;FP I II 1,2 t . Equation (3.3) characterizes the seller’s problem for both cases: in the absence of consumer reviews (i.e., 0 ) and in the presence of consumer reviews (.e., [0,1] ). The seller maximizes its profit from the expert segment conditional on the expert consumers’ participation. In other words, the expected valuations of expert consumers have to meet IR (individual rationality) or participation constraints, (,)S TVIc. The seller either provides full attribute information on both attributes (F I I ) or provides partial information on one attribute (P I I ). In the following we first derive the seller’s optimal information content strategies in the absence and presence of consumer reviews, and then compare them to examine the impact of consumer reviews on the seller’s information content strategy. 3.5.2.1 Information Content Strategy in the Absence of Consumer Reviews In the absence of consumer reviews, novice consumers have no product information and will not make a purchase. Hence, the seller’s overall profit is the same as its profit from the expert segment, i.e., ˆˆ ()()E I I . The profit from the experts consumers, ˆ ()E I , depends on whether the seller provides full information on both attributes (i.e., F I I ) or only partial information on one of the attributes (i.e., P I I ). We examine these two cases below in turn.

PAGE 73

63 3.5.2.1.1 Providing Full Attribute Information (F I I ) If the seller provides full information on both attributes, the valuations are zero for type mmT, Pvfor type mmT and mmT, and Fv for type mmTexpert consumers. The seller can choose between selling to only one type of expert consumers (mmT) at a high price, 12 FPPv , or to three types of expert consumers (i.e., mmT, mmTand mmT) at a low price, 12 PPPv . For simplicity, we assume the perfect match provides a sufficiently high value for consumers so that, in the presence of full product information, it is more profitable for the seller charge a high price Fv to only serve the perfect matched consumers (mmT), i.e., >32FPvvc . Relaxing this assumption will not change our results qualitatively. Since of consumers are experts, the seller’s profit from the experts is ˆˆ ()()()/4EFFFIIvc (3.4) 3.5.2.1.2 Providing Partial Attribute Information (P I I ) If the seller provides information only on one attribute, for instance 1a, consumers whose tastes match the informed attribute (type mmTand mmT) are certain about their match on the informed attribute (1a) from the seller’s information. However, they remain uncertain about the uninformed attribute (2a). Without information on the uninformed attribute, it is equally likely for them to believe the uninformed attribute is a match or mismatch. In other words, they believe there is an equal probability for them to have a valuation of Pv and Fv. Hence, the expected valuation for the product is ()/2PFvv for type mmTand mmT experts. Similarly we can find the expected valuations for other types of expert consumers, and derive the seller’s profit from the experts ˆ ()EP I . Comparing the seller’s profits under full and partial information content strategy in the absence of consumer reviews, ˆ ()EF I and ˆ ()EP I , we derive the following lemma regarding the seller’s information content strategy in the absence of consumer reviews.

PAGE 74

64 LEMMA 3.1 In the absence of consumer reviews, a partial information content strategy is optimal if the cost is sufficiently low, but otherwise a full information content strategy is optimal. Mathematically, the seller’s maximum profit from the experts is ˆˆ *()EEP I iff Pcv , where ˆ ()EP I is given in Appendix B.1. 3.5.2.2 Information Content Strategy in the Presence of Consumer Reviews In contrast to the case considered above in section 3.5.2.1, in the presence of consumer reviews, the information structure is different in periods 1 and 2 because expert consumers in period 2 can observe both the seller-created information and consumer reviews. 3.5.2.2.1 Providing Full Attribute Information (F I I ) If the seller provides full attribute information, expert consumers have full product information, and consumer reviews have no influence on the experts. The seller’s maximum profit from the expert segment, (,)EFI , is the same as in the full information case in the absence of consumer reviews. 3.5.2.2.2 Providing Partial Attribute Information (P I I ) When the seller only provides partial attribute information, in period 1, the expected valuations of four types of early experts (innovators) are the same as in the absence of consumer reviews. In period 2, when the seller only provides information on one attribute, for instance 1a, for type mmTand mmTmajority experts, due to their matched tastes with attribute 1a, the seller’s information on 1aexcludes the possibility that their valuations is zero. Without information on the second attribute or when the review information is perfectly uninformative 0 , both types of consumers assign the same probabilities, 1/2, to valuation Pv and Fv. With the informative consumer reviews, the probability that type mmT’s valuation is Fv (or type mmT’s is Pv) is an increasing function of the review informativeness, , and approaches to 1 when 1 . Similarly in the basic model, this probability function can be formalized as ()1/2/2 q . In the same

PAGE 75

65 time, the probability that type mmTÂ’s valuation is Pv (or type mmTÂ’s is Fv) is a decreasing function of , and approaching to 0 when 1 , and can be formalized as ()1/2/2 q . Therefore the expected valuations for type mmT and mmT experts are (,)mmP TVI ()Fvq ()Pvq and ()()mmFP TVvqvq , respectively. Similarly, we can find the expected valuations for mmT and mmT majority experts. The sellerÂ’s overall profits from all expert consumers, (,)EPI , can be derived based on the valuations of different consumers segments. Comparing the sellerÂ’s profits under full and partial information content strategy in the presence of consumer reviews, (,)EFI and (,)EPI , we derive the following lemma regarding the sellerÂ’s information content strategy in the presence of consumer reviews. LEMMA 3.2 In the presence of consumer reviews, a partial information content strategy is optimal if the cost is sufficiently low and the review informativeness is either small or large, and, otherwise a full information content strategy is optimal. Mathematically, the sellerÂ’s maximum profit from the experts is *(,)EEPI iff Pcv and [0,)[,1] !! !, where (,)EPI , ! and !are given in Appendix B.2. Lemma 3.1 reveals that, in the absence of consumer reviews, a partial information content strategy is optimal as long as the cost is low. Lemma 3.2 reveals that, in the presence of consumer reviews, an additional condition on review informativeness is required for the optimality of partial information. Comparing the conditions given in Lemmas 3.1 and 3.2 leads to Proposition 3.3, which states the impact of consumer reviews on the sellerÂ’s information content strategy. PROPOSITION 3.3 (The Impact of Consumer Reviews on Information Content Strategy) Offering consumer reviews increases the sellerÂ’s incentive to provide product attribute information. Specifically, when it is optimal to offer only partial attribute information in the absence of consumer reviews (i.e., with a low cost), the seller can earn a higher profit by offering full attribute information in the presence of consumer reviews if the review is sufficiently informative.

PAGE 76

66 Proposition 3.3 reveals that the seller’s consumer review supply decision will increase its incentive to provide full attribute information. This positive effect requires a low product cost and sufficient review informativeness. When the product cost is sufficiently low (Pcv ), it is optimal for the seller to provide only partial information in the absence of consumer review information (see Lemma 3.1). This is because with a low cost, it is more profitable to provide partial information and sell to different types of consumers at a low price, and less profitable to provide full information and serve only fully matched consumers at a high price (full information increasing valuation for the fully matched type but driving away other types). However, in the presence of consumer reviews, some mismatched consumers will find full attribute information from reading the consumer reviews and drop out of the market even if the seller only offers partial information. For this reason, a partial information strategy generates a smaller demand in the presence of consumer reviews than in their absence. This negative effect increases with the degree of review informativeness. As a result, the seller can achieve higher benefit from providing full information and charging a high price than offering partial information and charging a low price if the review information is sufficiently informative. 3.5.3 Seller’s Consumer Review Supply Decision (Stage 1) We now solve for the seller’s decision on whether to provide consumer reviews by comparing its overall maximum profit in the absence of consumer reviews (ˆ * ) and that in the presence of consumer reviews (* ). The seller’s overall profit is the sum of its profit from the expert consumers and novice consumers as shown in (3.5): 123 123ˆˆˆ *(,,)()In the absence of consumer reviews *(,,,)(,)In the presence of consumer reviewsEN ENIPPP IPPP!" #" $ (3.5) where ˆSand Sdenote the seller’s profit from the segment S in the absence and presence of consumer reviews, and S denotes consumer segment, , SEN (i.e., expert or novice segment). We first compare the seller’s maximum profit from the expert segment in the absence of consumer reviews (i.e. ˆ *E given in Lemma 1) with that in the presence of consumer reviews

PAGE 77

67 (i.e., *E given in Lemma 3.2). Comparing ˆ *E with *E, Lemma 3.3 follows (see the proof in Appendix B.3). LEMMA 3.3 Offering consumer reviews will either decrease or have no impact on the seller’s profit from the expert consumer segment. Mathematically, ˆ **EE. The sign is strict when ˆˆ *()EEP I . Lemma 3.3 shows that supplying consumer review information can lead to a profit loss from the expert segment (i.e., ˆ **0EEE"). This is because the seller can fully control the information content available to the expert segment in the absence of consumer reviews but not in their presence. When the seller’s optimal information content strategy in the absence of consumer reviews is to provide partial attribute information, the seller makes profit from both perfectly matched and some partially unmatched experts. However, consumer reviews will reveal the mismatch information to those partially unmatched experts, which will decrease their willingness to pay for the product and thereby drive down the seller’s profit from the expert segment. We now compare the seller’s profit from the novice segment in the absence of consumer reviews (i.e., ˆN) with its profit from the novice segment in the presence of consumer reviews (i.e.,N). Note that novice consumers will not make a purchase in the absence of consumer reviews, i.e., ˆN=0. However, some novice consumers may make a purchase in the presence of consumer reviews because the matching information provided by the review increases their willingness to pay. As a result, ˆ 0NNN". It is clear that the seller’s decision as to whether or not to provide consumer reviews is based on the tradeoff between its gain from the novice segment, 0N", and its profit loss from the expert segment, 0E". Comparison of N" and E" leads to Proposition 3.4 (see the proof in Appendix B.4).

PAGE 78

68 PROPOSITION 3.4 (Consumer Review Supply Strategy) The seller’s decision to supply consumer reviews depends on product cost and consumer expertise level. Specifically, supplying consumer review information increases the seller’s profit if (a) Consumer reviews are sufficiently informative, and (b) i) The product cost is sufficiently high, or ii) the product cost is low, but the expert segment is sufficiently small. Mathematically, ˆ (,)** I when (a) ' , and (b) i)Pcv , or ii) Pcv and * , where * and ' are defined in Appendix B.4. As stated earlier, the seller’s decision as to whether or not to provide online consumer review information depends on the tradeoff between its profit gain from novice consumers and loss from expert consumers. By offering consumer review information, the seller can bring novice consumers into the purchase process if the consumer reviews are sufficiently informative to significantly increase matched consumer’s willingness to pay. However, consumer review information can incur profit loss from expert consumers. When the product cost is sufficiently low, the seller’s optimal information content strategy in the absence of consumer reviews is to provide partial product information and serve some partially unmatched expert consumers. However, the consumer review information will reveal the mismatch information to these consumers, decrease their willingness to pay, and thereby reduce the seller’s profit from the expert segment. If there are fewer expert consumers in a market, the loss from the unmatched expert consumers is more likely to be negligible relative to the profit gain from novice consumers. The seller will offer consumer review information. When the product cost is sufficiently high, the seller must provide full product attribution in the absence and presence of consumer reviews. In this case, consumer review information supply decision will not incur profit loss from expert consumers. The seller will thus provide consumer reviews as long as the review is sufficiently informative.

PAGE 79

69 3.6 Timing Decision on Offering Consumer Reviews In our basic model, the timing for offering consumer review is not a decision variable. In practice, many online sellers do not have the flexibility to choose the timing for offering consumer reviews at the individual product level. For example, for a seller who adopts a general consumer review policy by allowing consumers to post their product valuation on its website (e.g., Amazon.com), consumers can post product reviews for any product as early as the first day of the product launch. However, it is possible for some online retailers to decide when to offer consumer reviews for a given product if they license consumer reviews from third-party sources. For example, c-source.com licenses consumer reviews from Epinions.com, and can flexibly decide when to make these third-party consumer reviews available to its consumers. To model this flexibility in timing, in this section we allow timing of offering consumer reviews to be a decision variable. Specifically, we allow the seller to decide whether to offer consumer reviews at the end of period 1 (i.e., right after the review information is available) or the end of period 2. (Note that offering reviews at the end of period 3 is the same as the case without consumer reviews.) Without loss of generality, we assume the license cost is zero. We use an “ f ” subscript to denote the variables when the review posting time is flexible. Specifically, we use “ f1 ” and “ f2 ” to denote the variables when reviews are provided in the end of period 1 and 2, separately. If consumer reviews are provided at the end of period 1, it is the same situation as discussed in the previous model. This is because consumers in period 2 and 3 can observe consumer reviews. The seller’s profit is 1*f **EN. If consumer reviews are provided at the end of period 2, only seller-created product information is available for consumers in the first two periods. The seller’s decision and profit from the expert segment is the same as in the absence of consumer reviews. In period 3, the consumer reviews are provided. The seller gets the same profit from the novice consumers as in

PAGE 80

70 the previous model. Therefore, its overall profit is 2*f ˆ **EN. Given ˆ **EE from Lemma 3.3, we find 21**ff. We have the following proposition regarding the timing decision of supplying consumer review. PROPOSITION 3.5 (Timing Decision on Consumer Review Offering) If the seller has control over the timing when consumer reviews become available at the individual product level, the seller will benefit from not offering consumer reviews too early. Mathematically, 21**ff. Proposition 3.5 reveals that, if possible, the seller should not provide consumer reviews too early even if such consumer reviews are available. Expert consumers are more likely to adopt a product earlier than novice consumers (Mahajan et al. 1990). Providing consumer reviews relatively late can decrease the seller’s profit loss from consumer reviews in the expert segment because the majority of them will have already made the purchase. At the same time, the seller can still enjoy the profit gain from purchases made by novice consumers. 3.7 Empirical Evidence To provide some external validity for our theoretical model, we conduct an exploratory empirical study to examine online sellers’ consumer review supply decisions based on data collected from different online sellers in several product categories. Our empirical study focuses on the impact of several factors identified in our theoretical model (i.e., width of product assortment, number of matched consumers, and length of time of the product on the market) on online sellers’ consumer review information supply decisions. We choose these factors because they can be reasonable measured directly or indirectly. Our results offer some preliminary evidences that are consist with our theoretical results. 3.7.1 Data First, to test whether the width of product assortment affects online sellers’ incentive to supplying consumer reviews as we predicted, we collected data in three product categories: MP3

PAGE 81

71 Players, PDAs, and video games. For these three product categories, we identified a list of 68 online sellers from the referral list of the shop agent mySimon.com in June 18, 2003. Among 68 sellers, 22 stores offer consumer reviews. We also collected data on the number of assortments a seller carried in each product category. Second, to test the impact of the number of matched consumers and the length of time of the product being on the market, we collected data from Dell.com and CNET.com. Dell.com sells computer and electronic products from different competing brands in various categories. For each product on its website, Dell.com provides product attribute information. For some selected products, it offers consumer review information from CNET.com. We collected our data in April 18, 2003. Our data include the products in six product categories: digital cameras, PDAs, digital camcorders, Web cameras, printers and scanners. For each of the six product categories, Dell.com carries multiple brands from different manufactures, and provides consumer reviews for only some selected models. There are overall 121 models in six categories. We collected data on product review and product characteristics from CNET.com. CNET.com lists almost all the available models for many product categories. For each product, it asks consumers to post detailed comments and to vote on whether or not they regard this product positively. Consumers can find the descriptive review information about the product and a summary statistic on what percentage of consumers regard the product positively. In addition, for many products, CNET.com also publishes its own product review and presents an overall rating. We can also collect data on when the product was launched to the market from CNET.com. For each of 121 models, we collected data on product launch time, product rating by CNET.com, number of available consumer postings, and the percentage of positive consumer votes. When Dell.com offers product review information for a model, it provides both CNET.com rating and consumer reviews. Therefore, to study its incentive to provide consumer reviews, we have to confine our sample in those products with both CNET review ratings and consumer

PAGE 82

72 reviews available. We found 57 out of 121 models have both CNET review ratings and consumer reviews available. We use these 57 models as the sample for our empirical analysis. 3.7.2 Empirical Findings 3.7.2.1 Impact of Product Assortment Proposition 3.2 suggests that the width of product assortment positively affects the profitability of the sellerÂ’s consumer review supply policy. A binary Logit model is used to examine the impact of product assortment on online sellersÂ’ consumer review supply decisions. The dependent variable is the binary choice for an online seller on whether it offered consumer reviews in a product category (i.e., either 1 or 0). The independent variable is the number of assortments a seller carries in a product category. In addition, we include the category dummy variables as the control variables. Table 3-1 presents our empirical results. Table 3-1. The Impact of Product Assortments Dependent Variable: Whether Consumer Reviews were Offered by a Seller (N=68) Coefficient Wald Statistic P-Value Number of Assortments in a Product Category .03* 6.11 .01 Product Category Dummy 1 .19 .05 .83 Product Category Dummy 2 .01 .00 .99 Model Correct Classification Rate 75 % 2 Log-likelihood (-2LL) 67.68 Note: A constant intercept is in the regression. *: Significant at .05 level. As shown in Table 3-1, the coefficient of the number of assortments is positive and significant (p<0.01), and both category dummy variable coefficients are insignificant. This suggests that, as predicted in Proposition 3.2, there is a significant positive relationship between the sellerÂ’s width of assortment and the likelihood for the seller to offer consumer review information. As discussed in our theoretical analysis, the seller with wider assortments is more likely to benefit from supplying consumer reviews and is thus more likely to facilitate such a new information channel.

PAGE 83

73 3.7.2.2 Impact of Number of Matching Consumers and Product Launch Time Proposition 3.1 suggests that the seller benefits from supplying consumer reviews if its product matches the preferences of a sufficient number of consumers. Proposition 3.4 and 3.5 suggest that offering consumer review too early can reduce its profit. As we described earlier, Dell.com offered consumer reviews for some selected models rather than all models it carries. The data we collected from Dell.com and CNET.com contain: 1) the launch time of 57 models, 2) the percentage of consumers who voted positively for each model, 3) the overall rating of each model by the third-party reviewer, CNET.com, 4) the number of available consumer postings for each model, and 5) which of the 57 models were offered consumer reviews. We used these data to examine empirically the impact of two factors on DellÂ’s consumer review supply decision: (1) the number of matched consumers and (2) the length of time since product launch. A binary Logit model is used to test the impact of various factors on the consumer review information supply decision. The dependent variable is whether Dell.com offered consumer reviews for a product model. The independent variables are (1) length of time since product launch into the market, and (2) percentage of consumers who vote positively for a product. We use this percentage statistic as the measure for how many consumers find a product to be a match. To control the influence of CNET third-party review on Dell.comÂ’s review supply decision, we add the CNET rating for a product as a control variable. In addition, to rule out the influence of product category and the number of available consumer postings, we also add product category dummies and the number of available consumer postings as the control variables. Table 3-2 presents the test results. As shown in Table 3-2, consistent with our theoretical predictions, we find both the coefficients of the length of time since launch and percentage of positive votes (the number of matched consumers) are significant and have the expected signs. Interestingly, we find the thirdparty product reviews from CNET is not a significant variable in explaining Dell.comÂ’s behavior on offering review information. As Chapter 2 shows, the third-party review focuses the product

PAGE 84

74 quality information. However, in our theoretical model, we argue that consumer reviews can also provide taste-matching information for consumers. The empirical results of the significant sign of the number of matched consumers and insignificant sign of the CNET rating seem consistent with our theory. In addition, the influence of the number of available consumers is not significant in the model, either. The number of postings seems not to be the determining factor in Dell.comÂ’s consumer review supply decision. Consistent with our theory, Dell.com provided consumer reviews on those models that have been on the market for a longer time period. Table 3-2. The Impact of Number of Matching Consumers and Product Launch Time Dependent Variable: The Availability of Consumer Reviews for a Product Model (N=57) Coefficient Wald Statistic P-Value Launch Time Length .02* 9.64 .00 Percentage of Positive Consumer Votes 6.82* 3.75 .05 CNET.com Rating .66 1.17 .28 Number of Available Consumer Postings .01 .53 .47 Product Category Dummy 1 -8.25 .07 .80 Product Category Dummy 2 -9.49 .09 .80 Product Category Dummy 3 -7.96 .06 .80 Product Category Dummy 4 -8.22 .07 .80 Product Category Dummy 5 -8.26 .07 .80 Model Correct Classification Rate 80.7 % 2 Log-likelihood (-2LL) 45.17 Note: A constant intercept is in the regression. *: Significant at .05 level 3.8 Conclusion Recent developments in information technology have significantly increased online sellersÂ’ information capacity. With the help of new technology, an online seller can not only present traditional seller-created information at a lower cost, but also has the new attractive option of

PAGE 85

75 supplying information to consumers by allowing them to post their product evaluations on the seller’s website or licensing consumer review information from third-party sources. In this paper, we investigate an emerging research area: online consumer reviews and their implications on firm marketing strategies. Specifically, we study the marketing function of consumer reviews, show when an online seller should offer consumer reviews to its customers, and examine how the decision to provide consumer-created information affects its other marketing strategies. An empirical study provides support for our theoretical model. 3.8.1 Online Consumer Review as a Marketing Function New technology now is making it possible for an online seller to efficiently provide two different forms of product information to its potential buyers: (1) seller-created product information supplied by the seller via its website or other media, and (2) consumer-created information self-posted by consumers on the seller’s website. We argue that, since consumercreated information is user-oriented but seller-created information is product-oriented, the former has an advantage over the latter in helping consumers to find products matching their preferences. This is particularly important for those product categories where consumers’ usage conditions are so idiosyncratic that it is impossible, or very costly, for the seller to acquire this knowledge and list all different mappings between the product attribute and usage condition spaces. As a result, consumer reviews can work as an online seller’s free “sale assistants” to help consumers to identify products that best match their needs. Consumer reviews are particularly important for unsophisticated consumers who will be less likely to buy the seller’s product if only seller-created product information is available. However, this free sale assistant does not come without cost. By allowing consumers to post their own product evaluations, the seller creates a new information channel for consumers, which eliminates the seller’s capability to control the supply of product information (e.g., providing full vs. partial information to consumers). In addition, our theory on the product matching function of the consumer reviews can also incorporate the frequently made argument that consumer reviews provide a more credible

PAGE 86

76 information source for consumers than firm-created information (e.g., Bickart and Schindler 2001). Recall that, in our basic model, consumer review information is the only information channel for consumers. This results from our assumption that all consumers are novices and cannot accurately interpret seller-created product information. This can also result from the assumption that seller-created information is totally noncredible. Therefore, the marketing function of online consumer reviews studied here can be summarized as providing credible information for all consumers to enable them to identify products that best match their idiosyncratic usage conditions. 3.8.2 The Strategic Implications of Online Consumer Review Our model provides several normative implications for online sellersÂ’ decisions related to online consumer reviews: First, the sellerÂ’s consumer review supply decision depends on the characteristics of its carried product category. Sellers carrying complicated, or high-tech products are more likely to benefit from providing consumer reviews than sellers carrying simple, or low-tech products. This is because (1) in the former markets, compared with seller-created information, consumer-created review information has the advantage of helping novice consumers to identify products that match their idiosyncratic preferences, and (2) there are fewer experts and more novice consumers in the former than the latter markets. When products are complex and the majority of consumers are unsophisticated, the benefit of providing consumer reviews becomes sufficiently high to outweigh a profit loss from the expert consumers. Second, the sellerÂ’s consumer review supply decision depends on whether its product is a mass-market or niche product. Sellers benefit more from offering consumers reviews when their products are mass-market products and match the preferences of a large consumer segment than when their products are niche products and match the preferences of a small consumer segment. With the help of consumer-created information, more consumers are able to correctly identify their matched products. As a result, the gain (cost) from consumer reviews is higher (lower) if a

PAGE 87

77 sellerÂ’s product matches majority consumersÂ’ preferences than if the sellerÂ’s product matches the minority consumersÂ’ preferences. Third, the sellerÂ’s consumer review supply decision depends on the informativeness of consumer reviews. Sellers can benefit from offering consumer reviews only if the consumer reviews are sufficiently informative. The consumer reviewÂ’s matching function largely depends on the review informativeness. When consumer review information is not sufficiently informative, the matched consumersÂ’, particularly novice consumersÂ’, uncertainty reduction and willingness-to-pay increase are too limited to make the sellerÂ’s decision profitable. Fourth, the sellerÂ’s consumer review supply decision depends on its product assortment strategy. Sellers carrying a wide product assortment benefit more from offering consumer reviews than sellers caring a narrow product assortment. This is because consumer reviews increase product valuations of matched consumers but decrease product valuations of unmatched consumers. For a seller carrying a wide assortment, the negative impact of consumer reviews is limited since most consumers are able to find their matched products from the same seller. For a seller carrying a narrow assortment, the negative impact of product review is significant because the seller may lose the unmatched consumers. Fifth, the timing of providing consumer review information can be an important strategic variable for a seller. When a seller is able to decide such timing at the individual product level (e.g., when the seller licenses consumer reviews from third-party sources), it may be unwise to supply consumer review information at a very early stage after a new productÂ’s introduction, even if such reviews are available. Offering consumer reviews reduces the sellerÂ’s control over the product information available to consumers. Hence, providing consumer reviews too earlier will hurt the seller if a partial information strategy is optimal in the absence of consumer reviews. Finally, sellersÂ’ two types of information supply strategies interact with each other. Specifically, the sellerÂ’s decision to offer consumer reviews will increase its incentive to provide full product attribute information. Without consumer reviews, the seller can benefit from

PAGE 88

78 providing partial product information and serve both perfectly and partially matched expert consumers. Consumer reviews take away the sellerÂ’s full control over product information, and reveal some mismatch information to partially matched expert consumers, which decreases their willingness to pay and incurs sellerÂ’s profit loss. Providing full product information can increase perfectly matched consumersÂ’ willingness to pay and reduce the sellerÂ’s profit loss from expert consumers. While this research improves our understanding of online consumer review and its implications for firm marketing strategies, many other interesting questions remain unanswered and require further investigation. One limitation of this paper is that we study a monopoly model and focus on the matching function of online consumer review. Future research may study some other functions of online consumer reviews and investigate its implications for firm competition. Second, in our model, the review informativeness is exogenous. Future research may study how an online seller can design some mechanisms to increase consumer review informativeness. Finally, future research may study from the perspective of consumer review infomediaries such as Epinons.com, and examine their optimal marketing strategies.

PAGE 89

79 APPENDIX A NOTATIONS AND PROOFS IN CHAPTER 2 A.1 Summary of Notations j Seller, , j HL t 0,, trd (0: without review; :rrecommendation format; : ddescription format). j t Advertising reach level of seller j in the reviewerÂ’s publication in t j t Advertising reach level of seller j in other media in t j tP Price of the seller j in t j t Profit of the seller j in t j t D Demand of the seller j in t Lc Marginal cost of seller L (Lcis normalized to 0) Hc Marginal cost of seller H (Hcis normalized to c) () g ! Advertising cost for an advertising reach level Advertising cost coefficient Penetration rate of the third-party product review The percentage of the reviewerÂ’s publication subscribers The percentage of price-insensitive consumers v ConsumersÂ’ evaluation on their preferred product v Price-sensitive consumersÂ’ evaluation on their non-preferred product The percentage of the taste-driven segment The percentage of taste-driven consumers with matched taste with L t A The size of consumers who receive product information only from HÂ’s advertising in t t B The size of consumers who receive product information only from both firmsÂ’ advertising in t t E The size of consumers who receive product information only from LÂ’s advertising in t t R The size of consumers who read review information in t Hq The fraction of consumers who correctly identify H to be high quality among those receiving ads only from H in the absence of review Lq The fraction of consumers who incorrectly identify L to be high quality among those receiving ads only from L in the absence of review q The fraction of consumers who correctly identify H to be high quality among those receiving ads from both firms in the absence of review H The fraction of consumers who consider H to be preferred among those receiving ads only from H in the absence of review L The fraction of consumers who consider L to be preferred among those receiving ads only from L in the absence of review The fraction of consumers who consider H to be preferred among those receiving ads from both firms in the absence of review The fraction of consumers who consider H to be preferred among those read review

PAGE 90

80 A.2 Proofs in Chapter 2 A.2.1 Equilibrium Analysis in the Absence of Product Review In following, we derive conditions under which equal-price, { 00 HLPPv }, is an equilibrium. In the absence of review, firm H ’s potential demand comes from two segments: (1) the loyal consumers, 0 Hl, who only buy H , and (2) the switchers, 0 H s , who will buy product H if H provides a higher positive surplus than L . H ’s loyals, 0 Hl, include price insensitive consumers who receive advertising from both firms and consider H as the preferred product (case 7 in Table 2-2), and all consumers who receive advertising only from H and consider H as the preferred product (case 4 in Table 2-2). Note vc , and it is not profitable for H to cut its price to attract price sensitive consumers who do not prefer H (case 6 and case 10 in Table 2-2). Hence, the switchers, 0 H s , are the price sensitive consumers who receive advertising from both firms and consider H as the preferred product (case 8 in Table 2-2). From Table 2-2 in the paper, 0 Hl and 0 H s are given in the following equations: 000 HHlBA (A.1) 00(1)H s B (A.2) There are three segments of consumers who may buy L : (1) all informed consumers who consider L as a preferred product, 0 Ll(case 1, 9 and 10 in Table 2-2), (2) price-sensitive consumers who receive advertising only from L but consider L as a non-preferred product, 0'Ll(case 3 in Table 2-2), and (3) the switchers, 0 H s (case 8 in Table 2-2). Similar to0 H s , 0'Ll will buy from L only when L ’s price is below v. However, 0 H s may still choose H if H ’s price provides them higher surplus than L ’s, while 0'Ll will always choose L as long as L ’s price is below v. From Table 2-2 in the paper, 0 Ll and 0'Ll are given in the following equations: 000(1)LLlBE (A.3) 00'(1)(1)LLlE (A.4) When both firms charge a price of v, their profits are 000000000()()()()=()()()HHLHHHHHPPvlsvcgBAvcg (A.5) 00000000()()(1)()LHLLLLLPPvlvgBEvg %&'( (A.6) If firm L decides to cut its price below v to attract the switchers from firm H , the maximum of profit to gain is # 000000 000Max ()(')() (1)[(1)(1)]()LLLHLL LLLPvlslvg BEvg (A.7) Note that q can be a function of both firms’ advertising effectiveness, i.e., (,)HLqfqq , where /0Hqq ,/0Lqq , Hqq , and 1Lqq . From (A.6) and (A.7), when ˆHHqq or ˆLLqq , 000()LHLPPv 00Max ()LHPv, and 00**HLPPv is the unique price equilibrium, where ˆHqand ˆLqare the solutions of the following function:

PAGE 91

81 0000 000()(1) (,)(1)0 (1)()1HLLBvvEvEE fqqq BvvBB %& )*'( (A.8) Both sellers maximize their profits with respect to advertising. According to (A.5) and (A.6), the first order conditions are: # 00000/0(1)()/(1)0HHLHLHvc+ (A.9) # 00000/0(1)(1)/(1)0LLHLHLv+ (A.10) It is straightforward that both second order conditions are negative. Solutions of equations (A.9) and (A.10), 00*and *HL, are the equilibrium advertising reach levels when ˆHHqq or ˆLLqq . Therefore when ˆHHqq or ˆLLqq , i.e., high-quality firm’s advertising cannot sufficiently convey quality information to consumers, or low-quality firm’s advertising is sufficiently misleading, 0000(*,*;**)HLHLPPvis the SPNE for the model. Q. E. D. A.2.2 Proof of Proposition 2.1 Similar to the case in the absence of review, firm H ’s potential demand comes from two segments: (1) the loyals, H dl, who will only buy H , and (2) the switchers, H d s , who will buy H only if firm H ’s price is low enough. From equation (2.4) and (2.5) in the paper, H dl and H d s are given in the following equations: HH dddlBA (A.11) #(1)H dd s B (A.12) Similarly, firm L ’s potential demand comes from three segments: (1) all informed consumers who prefer L , L dl, (2) price-sensitive consumers who receive advertising only from L but do not prefer L, 'L dl, and (3) the switchers, H d s . L dl and 'L dl are given in the following equations: (1)(1)LL dddlBE (A.13) '(1)(1)LL ddlE (A.14) The profits of firm H and L when they charge prices at v are ()()()() =()()()HHLHHH dddddd HH dddPPvlsvcg BAvcg (A.15) ()() =(1)(1)()LHLLL ddddd LL dddPPvlvg BEvg %& '( (A.16) If firm L decides to cut its price below v to attract the switchers from firm H , the maximum of profit to gain is # Max ()(')() (1)(1)[(1)(1)]()LLLHLL dddddd LLL dddPvlslvg BEvg (A.17) Therefore, Max ()()LLLHL dddddPvPPv when ˆd , where # ()[(1)](1) (1) () ˆ 1 (1)[]/ddd d L ddvvBvvEvvB vvvv BqEq (A.18)

PAGE 92

82 The firm L has no incentive to undercut to attract switcher from firm H . The pricing equilibrium is **HL ddPPv . Hence, both firms do not adjust pricing when ˆd . According to (A.15) and (A.16), the first order conditions are: # /0(1)(1)()/(1)0HHLHLH dddddvc+ (A.19) # /0(1)(1)(1)/(1)0LLHLHL dddddv+ (A.20) From (A.9) (A.10) (A.19) and (A.20), we find that 0*H and *H d are the solutions of the function 2()0HHabe, where ()[1()][(1)]HLavcv $, ()()()[1()]()[(1)]HHLLbvcvcvv$, and ()[1()]Hevcv$ ()()Hvcv . Let 24 bae ". Then *()/2Hba". 0 1**HH$and 1**HH d$. Notice # */()[1()]1[(1)]*(1*)/0HHLHHvc%&$"'(. Since 11, 0**HH d. Note advertising cost function () j tgis increasing in j t. Hence 0(*)(*)HH dgg. Similarly, we can show 0**LL d and 0(*)(*)LL dgg. Proposition 2.1 holds. Q. E. D. A.2.3 Proof of Proposition 2.2 and 2.3 In the presence of the recommendation review, in the subgame where H does not use reviewendorsed advertising, the pricing and advertising equilibria are the same as those in the description review case. In the presence of the recommendation review, in the subgame where H adopts reviewendorsed advertising, similarly to previous cases, firm H ’s potential demand comes from two segments: (1) the loyals, H rl, and (2) the switchers, H r s . From equation (2.6) and (2.7) in the paper, H rl and H r s are given in the following equations: ()H rrrlBA (A.21) (1)()H rr s B (A.22) Firm L ’s possible demand came from three segments: (1) informed consumers who prefer L , L rl, (2) price-sensitive consumers who receive advertising only from L but do not prefer L , 'L rl, and (3) switchers H r s . L rl and 'L rl are given in the following equations: (1)()LL rrrlBE (A.23) '(1)(1)LL rrlE (A.24) The profits for both firms when they maintain their prices at v are ()()()() ()()()HHLHHH rrrrrr H rrrPPvlsvcg BAvcg (A.25) ()() =(1)()()LHLLL rrrrr LL rrrPPvvlg BEvg %&'( (A.26) Firm L ’s maximum profit when cutting its price is Max ()(')()LLLLHL rrrrrrPvvllsg (A.27) From (A.26) and (A.27), when ˆr , ()Max ()LHLLL rrrrrPPvPv, where

PAGE 93

83 # [(1)] (1) ()() ˆ 1 []/()r r r L rrvvEvv vvBvv EqB (A.28) Therefore both firms maintain the same price at v when ˆr if firm H adopts reviewendorsed advertising. According to (A.25) and (A.26), the first order conditions are: /0(1)()/(1)0HHH rrrvc+ (A.29) /0(1)(1)(1)/(1)0LLHLHL rrrrrv%&+'( (A.30) Both second order conditions are negative. Firms’ optimal advertising reach levels after the publication of the recommendation format review are listed in the following equations: #*1/(1)()H rvc (A.31) (1) *1/(1)(1) ()L L rv vc ,./01 (A.32) From (A.9) and (A.29), we have 0**HH r and 0(*)(*)HH rggif *, and 0**HH rand 0(*)(*)HH rgg, otherwise, where 00*1(1*)*/LHL % & ' (. From (A.32), we know*/0L r. The maximum of *L r, *L r= 0*L r. Notice 0 0**HH r since0*. From (A.10) and (A.30), given 0*,*/0L HHL r rrrLL rr, we know 000 */L L rLL 0 000 **,*//LL LHHL rr rrrLLLL rr 0 0(*[(1)]*[(1)]HLHL rv %&'(>0. Notice 0000 */LLLL=0. Hence, 0000 */0LLLL< 000 */L L rLL. Since 22 00/()0LL, we have 0***L LL r r. Therefore 0(*)(*)LL rgg. Hence Proposition 2.3 (III) holds. Given equilibrium levels * j r, from (A.18) and (A.28), we find **ˆˆˆ *j jjj rrr drrd. Hence ˆˆˆ *,**,**,*()Max ()0jjj jjj rrrrrr dddLLHLLL ddddddPPvPv". Notice /0L ddB, /0H ddB,/0L ddE, and /0H ddE. Therefore /0LL dd", and /LH dd" (1)(1)(1)()0LL dvv. From (A.19) and (A.29), we find the marginal advertising cost **'()'()HH rdgg"" noticing (1)LHL dd . Since ''()0 g ", we have **HH rd. Similarly, from (A.20) and (A.30), we have **LL rd. Therefore ˆ *,*jj r ddL d" > ˆ *,*jj rr dL d">0. Hence ˆ *,*Max ()jj r ddLL ddPv ˆ *,*()jj r ddLHL dddPPv. Therefore the threshold level for L ’s price-cut is higher if H adopts the review-endorsed ads than if H ’s does not adopt (i.e., ˆˆrd in the equilibrium). Hence Proposition 2.2 (I) holds. Note ˆˆrd . Therefore when ˆr , **HL rrPPvis the SPNE pricing equilibrium of the three-stage game. Hence Proposition 2.3 (I) holds.

PAGE 94

84 From (A.15) and (A.25), it is straightforward that *()HH rdHHL rrrPPv *()HH ddHHL dddPPv. Note **()HH rrHHHL rrrrPPv *()HH rdHHL rrrPPv. We have **()HH rrHHHL rrrrPPv *()*HH ddHHLH ddddPPv. Therefore when ˆr , H always adopts review-endorsed advertising. Proposition 2.3 (II) holds. In the following, we prove Proposition 2.2 (II) by showing a sufficient condition for the case when firm H is less profitable to adopt the review-endorsed advertising, i.e.**HH rd. It is reasonably assume that, in the absence of product review and if all consumers receive advertising from both firms, there is enough uncertainty about product quality and L ’s advertising is sufficiently persuasive such that a significant number of quality-driven consumers would misperceive L as the high-quality product. Specifically, in the absence of product review, L is sufficiently high, i.e., (1)/()Lvvv , such that it is more profitable for firm L to charge a high price vthan to cut price to v to attract switchers, i.e., Lv (1)(1)LLv % & ' (. In the subgame where firm H adopts the review-endorsed advertising, when ˆr , firm L has incentive to cut price to attract switchers from firm H . We find (')LL rrllv = # (1)()(1)(1)LL rrr B EEv (1)()L rr B vEv # (1)()L rr B Ev L rlv. Hence, whenˆr , firm L makes tradeoffs between pricing at v to capture the surplus from its loyal segment of L rl and pricing below v to sell to 'LLH rrrlls consumers. The lower boundary (#L rP) for L ’s price support happens when # (')L LLHL r rrrrPllslv, i.e., # /(')L LLLH r rrrrPlvlls. Firm H makes tradeoffs between pricing at v to capture the surplus from its loyal segment of H rl and pricing below v to sell to HH rrls consumers. The lower boundary (#H rP) for H ’s price support happens when # ()()()H HHH r rrrPclslvc, i.e., # ()/()H HHH r rrrPlvclsc. When ˆrr , # #LH rrvPvP, i.e., #$# ()HL H rrrPPPvv, where # # # # # (1) ()(1)() 1 ()[]LL rr r rrr LL rr r L rrvPEP B EB vPvP BEq (A.33) Firm H can price little bit below $ H r P to undercut firm L ’s price #L rP, and obtain the entire segment of H r s for a total profit of $()()H HH r rrPcls . Note prices (,)L rPvv attract only loyal consumers, and hence inferior to v. Firm L randomizes its prices in the strategy set # [,]L rPvv ! to achieve the equilibrium expected profit (before advertising cost) L rlv. Firm H randomizes its prices in the strategy set $ [,]H rPv to achieve the equilibrium expected profit (before advertising cost) $()()H HH r rrPcls . The equilibrium expected profits for both firms are $ ˆ ()()()H HHHH r rrrrEPclsg (A.34)

PAGE 95

85 ˆ ()LLL rrrEvlg (A.35) For ˆrr , from (A.15) and (A.34), we find when 0' , ˆ **HHH rrdE. When 0 or 'r , ˆ **HHH rrdE, where 01()/2 ba ", and '1()/2 ba " (A.36) where # 211()[]LL rraNvEqNvvcqEN%&'( #13[]()(1)LH rddqENvcBqAqN, # #2 12121212(1)()(1(1))[](1)L L r rbvEqNNNNvvcNNqENN %&'( # #1313[](1)()()()(1(1))()(1)LHHH rdrddqENNvcggNvcBqAqN, 211(1)()(1)) eNvNvvcN #13(1))(1)()()()HH drNNvcgg,24 bae ", 0() Nvvc , 1rrNBE, 2rrNBA, 3ddNBA. When ˆr , *H d ˆ ()HHHL rrrr E PPv. Therefore 0ˆr . Hence for 'r , ˆ **HHH rrdE, and it is less profitable for firm H to adopt the review-endorsed advertising. Proposition 2.2 (II) holds. Q. E. D. A.2.4 Proof of Proposition 2.4 In the absence of product review, the sizes of informed consumer groups are 00000 00000 00000(1)(1)(1)Reached only by 's Advertising Size: (1)(1)(1)Reached only by 's Advertising (1)Reached by both firms' AdvertisingLHLH mmmmm HLHL mmmmm HLHL mmmmmEL AH B ! " # " $ (A.37) Similar to the proof of the Proposition 2.1, the profits of firm H and L when they charge prices at v are 0000000 0000()()()()() =()()()()HHLHHHH mmmmmmm HHH mmmmPPvlsvcgg BAvcgg (A.38) 000000 0000()()() (1)()()LHLLLL mmmmmm LLL mmmmPPvlvgg BEvgg %&'( (A.39) The first order conditions are: # 00000/0(1)()/(1)0HHLHLH mmmmmvc+ (A.40) # 00000/0(1)(1)/(1)0LLHLHL mmmmmv+ (A.41) # 00000/0(1)(1)()/(1)0HHLHLH mmmmmvc+ (A.42) # 00000/0(1)(1)(1)/(1)0LLHLHL mmmmmv+ (A.43) It is straightforward that all second order conditions are negative. Solutions of equations (A.40) (A.41) (A.42) and (A.43), 0000*, *, *, and *HLHL mmmm, are the equilibrium advertising reach levels. In the presence of the description review, the sizes of informed consumer groups are

PAGE 96

86 (1)(1)(1)(1)Reached only by 's advertising (1)(1)(1)(1)Reached only by 's advertising Size: (1)(1)Reached only by bothLHLH dmdmdmdmdm HLHL dmdmdmdmdm HLHL dmdmdmdmdmEL AH B firms' advertising Reached by product reviewdmR! " " # " "$ (A.44) Similar to the proof of Proposition 2.1, the profits of H and L when both firms charge the same price at v are ()()()()() =()()()()HHLHHHH dmdmdmdmdmdmdm HHH dmdmdmdmPPvlsvcgg BAvcgg (A.45) ()()() =(1)(1)()()LHLLLL dmdmdmdmdmdm LLL dmdmdmdmPPvlvgg BEvgg %&'( (A.46) According to (A.45) and (A.46), the first order conditions are: # /0(1)(1)()/(1)0HHLHLH dmdmdmdmdmvc+ (A.47) # /0(1)(1)(1)/(1)0LLHLHL dmdmdmdmdmv+ (A.48) # /0(1)(1)()/(1)0HHLHLH dmdmdmdmdmvc+ (A.49) # /0(1)(1)(1)/(1)0LLHLHL dmdmdmdmdmv+ (A.50) From (A.42) (A.43) (A.49) and (A.50), it is straightforward that 0**HH mdm and 0**LL mdm. Hence, 0(*)(*)HH mdmgg and 0(*)(*)LL mdmgg. Similar to the proof of Proposition 2.1, we can show 0(*)(*)HH dmmgg and 0(*)(*)LL dmmgg. In the presence of the recommendation review, in the subgame where H does not adopt review-endorsed advertising, the pricing and advertising equilibria are the same as those in the description review case. In the presence of the recommendation review, in the subgame where H adopts reviewendorsed advertising, the sizes of informed consumer groups are (1)(1)(1)(1)Reached only by 's advertising (1)(1)(1)(1)Reached only by 's advertising Size: (1)(1)Reached only by bothLHLH rmrmrmrmrm HLHL rmrmrmrmrm HLHL rmrmrmrmrmEL AH B firms' advertising Reached by product reviewrmR! " " # " "$ (A.51) Similar to the proof of Proposition 2.2 and 2.3, the profits for both firms when they maintain their prices at v are ()()()()() ()()()()HHLHHHH rmrmrmrmrmrmrm HH rmrmrmrmPPvlsvcgg BAvcgg (A.52) ()()() =(1)()()()LHLLLL rmrmrmrmrmrm LLL rmrmrmrmPPvvlgg BEvgg %&'( (A.53) According to (A.52) and (A.53), the first order conditions are: /0(1)()/(1)0HHH rmrmrmvc+ (A.54)

PAGE 97

87 /0(1)(1)(1)/(1)0LLHLHL rmrmrmrmrmv%&+'( (A.55) /0(1)()/(1)0HHH rmrmrmvc+ (A.56) /0(1)(1)(1)/(1)0LLHLHL rmrmrmrmrmv%&+'( (A.57) All second order conditions are negative. From (A.42) and (A.56), we know #00000 */(1)()(1)HH mrmHHLHL mmmmvc <0 = 0000 */HH mmHH mmnoticing H . Since 22 00/()0HH mm, we have 0000 */LL mmLL mm0**HH rmm and 0(*)(*)HH rmmgg. From (A.43) and (A.57), 000 */LL mrmLL mm #0(1)()*(1)*(1)HLHL rmmvc >0 = 0000 */LL mmLL mm. Since 22 00/()0LL mm, we have 0**LL rmmand 0(*)(*)LL rmmgg. Similar to the proof of Proposition 2.3, we can show (i) 0(*)(*)LL rmmgg, and (ii)0(*)(*)HH rmmgg if *m, and 0**HH rmm, otherwise, where 00*1(1*)*/LHL mmm %&'( (A.58) Proposition 2.4 holds. Q. E. D.

PAGE 98

88 APPENDIX B PROOFS IN CHAPTER 3 B.1 Proof of Lemma 3.1 In the absence of consumer reviews, when the seller provides full information, its profit from the experts isˆ ()()/4EFFIvc. When the seller provides partial information on one attribute, for instance1a, the expected valuation for the product is ()/2PFvv for type mmTand mmT experts. For those consumers who have mismatched tastes in the attribute 1a(type mmT and mmT), they have equal probabilities to find match and mismatch in 2a and have a valuation of Pv and 0 on the product. Hence the expected valuation for them is /2Pv. This valuation is lower than (2)/4FPvvv and c, and fail to meet the participation constraint. Recalling that is the percentage of first period experts (innovators) among all experts and is the fraction of the experts among all consumers, the seller’s profit is12112ˆ ()()(1)()/2/2EP PFIPcPcvvc. We find ˆˆ ()()EPEF I I and ˆˆ *()EEF I iff Pvc. Q. E. D. B.2 Proof of Lemma 3.2 In the presence of consumer reviews, when the seller provides full information, from equation (3.4) in the paper, its profit from the experts is (,)()/4EFFIvc (B.1) When the seller only provides information on one attribute, for instance 1a, the seller’s profit from the experts in period 1 is 1()()/2/2PFPIvvc . In period 2, we know the expected valuations for type mmT and mmT experts are ()()()/2()/2FPPFFPvqvqvvvv and ()()()/2()/2FPPFFPvqvqvvvv, separately. Due to their mismatched tastes with attribute 1a, type mmT and mmTexperts find their valuations are Pv and 0 with the probability of ()q , and 0 and Pvwith a probability of ()q . Therefore the expected valuations for type mmT and mmT experts are ()(1)/2PPvqv and ()(1)/2PPvqv, separately. For a sufficiently small , it is profitable for the seller to charge a low price at ()/2()/2PFFPvvvv to gain the demand from both type mmT and mmTexperts. For a sufficiently large , it is profitable for the

PAGE 99

89 seller to charge a premium price at ()/2()/2PFFPvvvvto serve only type mmT experts. Therefore seller’s profit in period 2 is 2(1)1 ()(), (,1] 422 (,) (1)1 ()(), [0,] 222PFFP P PFFPvvvvcif I vvvvcif ! ,./ " "01 # ," ./ " 01 $ % % (B.2) where1/32()/3()PFPvcvv%. The seller’s profit from the experts when only providing partial attribute information in the presence of consume reviews is (,) (1)1 ()()(), [0,) 22222 (1)1 ()()(), [,1] 22422EP FP PFFP FP PFFPI vv cvvvvif vv cvvvvif ! ,./ " "01 # ," ./ " 01 $ % % (B.3) Comparing the seller’s profits from providing the full and partial attribution information in equation (B.1) and (B.3), we find, (,)EPI < (,)EFI when Pcv and [,] ! !, where ()/(1)()PFPvcvv !and ()/()PFFPvvvv ! 2()/(1)()PFPvcvv. Hence, (,)(,)EFEPII and ˆˆ ()()EFEP I I if Pcv and [,] ! !. Q. E. D. B.3 Proof of Lemma 3.3 1) When Pcv , the seller will provide full attribute information both in the absence and presence of consumer reviews. Its profits from the expert segment are the same, i.e., ˆ **EE 2) When Pcv , in the presence of consumer reviews, from equation (B.1), when it provides full product information, the seller’s expected profit from the expert segment is ˆ (,)()/4()()/4EFFPPIvcIvc (B.4) From equation (B.3), seller’s profit from the experts when it provides partial information in the presence of consume reviews is 2 2 (,) (1)1 ()()(), [0,) 22222 (1)1 ()()(), [,1] 22422 (1) ˆ ()(), [0,) 22 (1)1 ˆ ()()( 422EP FP FFP FP FFP PFP PPFI vv cvvvvif vv cvvvvif Ivvif Ivvv ! ,./ " "01 # ," ./ " 01 $ % % % ), [,1]FPvif! " " # ," ./ " 01 $ % (B.5)

PAGE 100

90 Note in the absence of consumer reviews, the seller’s maximum profit from the expert segment is ˆˆˆ ()*()*()EP I II . From equation (B.4) and (B.5), it is straightforward that both (,)EFI and (,)EPI are less than ˆ ()P I . Therefore, ˆ **EE when Pcv . Therefore ˆ (,)**EI. Q. E. D. B.4 Proof of Proposition 3.4 When the seller decides to provide consumer reviews, in period 3, as shown in Table B-1, for type mmT novices who are the perfect match consumers for the seller’s product, given the consumer review information, the probability for them to find the match in each attribute is ()1/2/2 q , which is increasing with the review informativeness. In contrast, the probability for them to find the mismatch in each attribute is ()1/2/2 q , which is decreasing with the review informativeness. Therefore the probabilities for type mmT novices to have valuations of zero, Pv and Fvare ()() qq , ()() qq , and ()() qq , respectively. As a result, the expected valuation of type mmT novices is (2)/4PFvv /2Fv2(2)/4PFvv. Similarly, as shown in Table B-1, we can find the expected valuations for other three types of novices. The expected valuations for four types of novice consumers are 2 2 2 2(,) (2)/4/2(2)/4 (,) (2)/4(2)/4 (,) (2)/4(2)/4 (,) (2)/4/2(2)/4mm mm mm mmN PFFPF T N PFPF T N PFPF T N PFFPF TVI vvvvv VI vvvv VI vvvv VI vvvvv %& % & )* ) * )* ) * )* ) * )* ) * ) * )* ' ( '( (B.6) Table B-1. Novice Consumer Expected Valuations in the Presence of Consumer Reviews Prob. of Attribute Match for Type T Novices Given Consumer Review Informativeness Consumer Type ( T ) Mismatch in 1a and 2a ( 0 v ) Mismatch in 1abut not 2a (Pvv ) Mismatch in 2abut not 1a (Pvv ) Match in both 1a and 2a (Fvv ) Consumer Expected Valuation (,)N TVI mmT ()() qq ()() qq * ()() qq ()() qq (2)/4PFvv /2Fv2(2)/4PFvv mmT ()() qq ()() qq ()()qq ()() qq (2)/4PFvv2(2)/4PFvv mmT ()() qq ()()qq ()() qq ()() qq (2)/4PFvv2(2)/4PFvv mmT ()()qq ()() qq ()() qq ()() qq (2)/4PFvv /2Fv2(2)/4PFvv Note *: ()1/2/2q , and ()1/2/2 q .

PAGE 101

91 From equation (B.6), we find ()mmN TV is less than vand therefore c, and fails the participation constraint. The participation constraint for type mmT novice consumers is: ()mmN TVc (B.7) From (B.7), ' , where 2'(2)(42)/(2)FFPFPFPFvvvvcvvvv,./ 01. The participation constraint for mmT, mmTand mmTtypes of novice consumers to buy is: () () ()mm mm mmN T N T N TVc Vc Vc ! " # " $ (B.8) From (B.8), & , where (42)/(2)PFPFcvvvv & > . The seller’s profit from the novice consumers is 3 3 2 2ˆ (,), [',)[,1] ()* ˆ (,), [,) 12(2) ˆ , [',)[,1] 4424 3(1)2(2) ˆ , [,) 444mm mmNN T N NN T PFFPF PFPFPVif PVif vvvvv cif vvvv cif ! ! " # " $ ! ,! " ./ "01 # ," ./ " 01 $ where 28(2)(4) ˆ 0 4(2)FFPFPF PFvvvvcvv vv and 28(2)(4) 1 4(2)FFPFPF PFvvvvcvv vv . Hence the seller’s profit from the novice consumers is 3 2()*(,) 12(2) 4424mmNNN T PFFPFPV vvvvv c ,./ 01 (B.9) When Pvc , the seller will provide full attribute information in the absence of consumer reviews. The seller’s decision to provide consumer reviews does not affect its profit from the expert segment. Hence, when ' , the seller will provide consumer reviews and gain extra profit from the novice segment. When Pvc , the seller always provides partial information in the absence of consumer reviews, i.e., ˆˆ *()P I . When the seller provides consumer reviews, 1) For [',], ! its maximum overall profit is 123 2 (,)*(,,)(,) (1)12(2) ˆ *() 224424mmmmEPENN TT PFFPF FPIIPPVPV vvvvv vvc ,./ 01 (B.10) ˆ (,)** I if * , where

PAGE 102

92 # 1 2 322(2) *(1)()/ 424PFFPFvvvvv vvc ,,./ ./ 01 01 (B.11) 2) For [,], ! !its maximum profit is 123 2 (,)*(,,)(,) 12(2) ˆ *() 44424mmmmEFENN TT PFFPF PIIPPVPV vvvvv vcc ,./ 01 (B.12) ˆ (,)** I if * , where 1 22(2) *()/ 424PFFPF Pvvvvv vcc ,,./ ./ 01 01 (B.13) 3) For[,1] !, its maximum profit is 123 2 (,)*(,,,)(,) (1)112(2) ˆ *()() 4224424mmmmmmEPEENN TTT PFFPF PFFPIIPVPVPV vvvvv vvvvc ,,./ ./ 01 01 (B.14) ˆ (,)** I if * , where ## 1 22(2) *(1)()()/ 424PFFPF PFFPvvvvv vvvvc ,,./ ./ 01 01 (B.15) Therefore, ˆ (,)** I when PcV , or PcV , * and . Q. E. D.

PAGE 103

93 REFERENCES Akerlof, George (1970), "The Market for `Lemons': Quality and the Market Mechanism," Quarterly Journal of Economics , 84 (August), 488-500. Alba, Joseph and Wesley Hutchinson (1987), “Dimensions of Consumer Expertise,” Journal of Consumer Research , 13 (March), 411-454. Anderson, Simon and Andre de Palma (1992), "The Logit as a Model of Product Differentiation," Oxford Economic Paper , 44, 51-67. Archibald, Robert B., Clyde A. Haulman, and Carlisle E. Moody Jr. (1983), "Quality, Price, Advertising, and Published Quality Ratings," Journal of Consumer Research , 9 (March), 347-56. Avery, Christopher, Paul Resnich, and Richard Zeckhauser (1999), “The Market for Evaluations,” American Economic Review , 89 (3), 564-584. Ayer Directory of Publications , (1981), Ayer Press, Philadelphia. Baron, David and Roger Myerson (1982), “Regulating a Monopolist with Unknown Costs”, Econometrica , 50(4), 911-930. Bettman, James (1979), An Information Processing Theory of Consumer Choice , Reading, MA: Addision-Wesley Publishing Company. Bickart, Barbara and Robert M. Schindler (2001), “Internet Forums as Influential Sources of Consumer Information,” Journal of Interactive Marketing , 15 (3), 31-40. Biyalogorsky, Eyal, Eitan Gerstner, and Barak Libai (2001), "Customer Referral Management: Optimal Reward Programs," Marketing Science, 20 (1), 82-95. Blackwell, David (1951), “Comparision of Experiments,” in J. Neyman, ed. Proceedings of the Second Berkeley Symposium on Mathematical Statistics and Probability , Berkeley: University of California Press, 93-102. Brown, Jacqueline J. and Peter H. Reingen (1987), “Social Ties and Word-of-Mouth Referral Behavior,” Journal of Consumer Research , 14 (December), 350-362. Brucks, Merrie (1985), "The Effects of Product Class Knowledge on Information Search Behavior," Journal of Consumer Research , 12 (June), 1-16. Butters, Gerard (1977), "Equilibrium Distributions of Sales and Advertising Prices," The Review of Economic Studies , 44(3), 465-91.

PAGE 104

94 Chaudhri, Vivek (1998), “Pricing and Efficiency of a Circulation Industry: The Case of Newspapers,” Information Economics and Policy , 10, 59-76. Chen, Yuxin and Jinhong Xie (2003), “Cross-market Network Effect with Asymmetric Customer Loyalty: Implications on Competitive Advantage,” Working Paper, University of Florida. Chen, Yuxin, Ganesh Iyer and V. Padmanabhan (2002), "Referral Infomediaries", Marketing Science, 21(4), 412-434. Chevalier, Judith and Dina Mayzlin (2003), "The Effect of Word of Mouth Online: Online Book Reviews", working paper, Yale School of Management. Dean, Dwane H. and Abhijit Biswas (2001), "Third-party Organization Endorsement of Products: An Advertising Cue Affecting Consumer Prepurchase Evaluation of Goods and Services," Journal of Advertising , 30 (4), 41-57. Dellarocas, Chrysanthos (2003), “The Digitization of Word of Mouth: Promise and Challenges of Online Feedback Mechanisms,” Management Science , 49 (10), 1401-1424. Eliashberg, Jehoshua and Steven Shugan (1997), “Film Critics: Influencers or Predictors?” Journal of Marketing , 61 (April), 68-78. Fader, Peter and Leigh McAlister (1990), “An Elimination by Aspects Models of Consumer Response to Promotion Calibrated on UPC Scanner Data,” Journal of Marketing Research , 27 (August), 322-332. Faulhaber, Gerald and Dennis Yao (1989), "'Fly-by-night' Firms and the Market for Product Reviews," Journal of Industrial Economics , 38 (September), 65-77. Fisher, Lawrence M. (1988), “Two Giants in PC Publishing,” New York Times , Apr. 21st, D1. Fost, Dan (1999), "Computer Magazines Slug It Out in Furious Contest for Revenue," The San Francisco Chronicle, Sept.9th, B1. Gensch, Dennis (1987), “A Two-stage Disaggregate Attribute Choice Model,” Marketing Science , 6 (3), 223-239. Gerstner, Eitan (1985), "Do Higher Prices Signal Higher Qualities?" Journal of Marketing Research , 22 (May), 209-215. Gertner, Robert and Robert Stillman (2001), “Vertical Integration and Internet Strategies in the Apparel Industry,” Journal of Industrial Economics , 49 (4), 417-440. Glazer, Rashi (1999), “Winning in Smart Markets,” Sloan Management Review , 40 (4), 59-69. Godes, David and Dina Mayzlin (In press), “Using Online Conversations to Study Word of Mouth Communication,” Marketing Science . Grossman, Gene M and Carl Shapiro (1984), “Informative Advertising with Differentiated Products,” Review of Economic Studies , 51 (1), 63-81.

PAGE 105

95 Harmon, Amy (2004), “Amazon Glitch Unmasks War of Reviewers,” New York Times (Feb. 14), A1. Holmstrom, Bengt (1979), “Moral Hazard and Observability,” Bell Journal of Economic s, 10(1), 74-91. Iyer, Ganesh and David Soberman (2000), “Markets for Product Modification Information,” Marketing Science , 19 (3), 203-225. Johnson, Eric, Wendy Moe, Peter Fader, Steven Bellman, and Jerry Lohse (2004), "On the Depth and Dynamics of World Wide Web Shopping Behavior," Management Science , 50 (3), 299-308. Kalra, Ajay, Mengze Shi and Kannan Srinivasan (2003) "Salesforce Compensation Scheme and Consumer Inference," Management Science , 49 (5), 655-672. Lancaster, Kelvin (1966), "A New Approach to Consumer Theory," Journal of Political Economy , 74, 132-157. Lewis, Peter H. (1989), “How Useful are Product Reviews?” New York Times, Sept. 10th, 12. Lewis, Tracy and David Sappington (1994), “Supplying Information to Facilitate Price Discrimination,” International Economic Review , 35(2), 309-327. Lewyn, Mark (1989),"Interfacing with Success: Ad Revenues Soar at High-tech Publications; Computer Mags 'Rule the Industry'," USA Today , February 14, 1B. Little, John (1979), “Aggregate Advertising Models: The State of the Art,” Operations Research , 27 (4), 629-667. Liu, Yong, Daniel S. Putler and Charles B. Weinberg (2004), “Is Having More Channels Really Better? A Model of Competition Among Commercial Television Broadcasters,” Marketing Science , 23 (1), 120-133. Lizzeri, Alessandro (1999), "Information Revelation and Certification Intermediaries," Rand Journal of Economics , 30(2), 214-231. Lohr, Steve (1993), “PC Magazine Is Mirroring an Industry's Success,” New York Times , Nov. 22nd, 1D. Mahajan, Vijay, Eitan Muller and Rajendra Srivastava (1990), “Determination of Adopter Categories by Using Innovation Diffusion Models,” Journal of Marketing Research , 27 (February), 37-50. Mayzlin, Dina (In press), “Promotional Chat On the Internet,” Marketing Science . Melnikov, Oleg (2000), “Demand for Differentiated Durable Products: The Case of the U.S. Computer Printer Market,” Working Paper, Yale University. Meurer, Michael and Dale Stahl II (1994), "Informative Advertising and Product Match," International Journal of Industrial Organization , 12 (1), 1-19.

PAGE 106

96 Milgrom, Paul and John Roberts (1986), "Price and Advertising Signals of Product Quality," Journal of Political Economy , 94 (4), 796-821. Moorthy, Sridhar and Kahan Srinivasan (1995), "Signaling Quality With a Money-Back Guarantee: The Role of Transaction Costs," Marketing Science , 14 (4), 442-466. Nelson, Phillip (1974), "Advertising as Information," Journal of Political Economy , 82 (4), 72954. Piller, Charles (1999), “Everyone Is A Critic in Cyberspace,” Los Angeles Times, ( Dec. 3), A1. Reddy, Srinivas, Vanitha Swaminathan and Carol Motley (1998), "Exploring the Determinants of Broadway Show Success," Journal of Marketing Research , 35(August), 370-83. Sappington, David (1983), “Limited Liability Contracts Between Principal and Agent,” Journal of Economic Theory , 29 (1), 1-21. Sappington, David (1991), “Incentives in Principal-Agent Relationships,” Journal of Economic Perspective s, 5 (2), 45-66. Shaffer, Greg and Florian Zettelmeyer, (2002), “When Good News About Your Rival Is Good For You: The Effect Of Third-Party Information On The Division Of Channel Profits," Marketing Science , 21(3), 273-293. Shi, Mengze (2003), “Social Network-Based Discriminatory Pricing Strategy,” Marketing Letters , 14 (4), 239-256 Shugan, Steven (1989), “Product Assortment in a Triopoly,” Management Science , 35 (3), 304320. Shugan, Steven M. and Larry Winner (2003), “Product Reviews and Marketing Strategy,” Working Paper, Marketing Department, University of Florida. Simester, Duncan (1995), “Signaling Price Image Using Advertised Prices,” Marketing Science , 14(2), 166-188. Simmons, Jacqueline (1994), " A 'Thumbs Up' Pulls in the Audience," Wall Street Journal, March, 25, B1. Spence, Michael (1973), "Job Market Signaling," Quarterly Journal of Economics , 87 (3), 35574. Stiglitz, Joseph (2002), “Information and the Change in the Paradigm in Economics,” American Economic Review , 92 (3), 460-501 Tirole, Jean (1990), The Theory of Industrial Organization , Cambridge, MA: MIT Press. Tversky, Amos (1972), “Elimination by Aspects: A Theory of Choice,” Psychological Review , 79 (July), 281-299.

PAGE 107

97 Villas-Boas, J. Miguel (In press), “Communication Strategies and Product Line Design,” Marketing Science. Walker, C. (1995), “Word of Mouth,” American Demographics , October, 38-44. Wernerfelt, Birger (1994a), “On the Function of Sales Assistance,” Marketing Science , 13 (1), 68 –82. Wernerfelt, Birger (1994b), “Selling Formats for Search Good,” Marketing Science , 13 (3), 298309. Wilson, E.J. and D.L. Sherrell (1993), “Sources Effects in Communication and Persuasion Research: A Meta-Analysis of Effect Size,” Journal of the Academy of Marketing Science , 21 (Spring), 101–112. Zhao, Hao (2000), “Raising Awareness and Signaling Quality to Uninformed Consumers: A Price Advertising Model,” Marketing Science , 19 (4), 390-396.

PAGE 108

98 BIOGRAPHICAL SKETCH Yubo Chen was born in April 1977. Before coming to the U. S. to pursue a Ph. D in marketing, he received his B. Eng. in industrial management engineering with highest honors from Southeast University, China, in 1997. He also holds a M. Eng. in systems engineering from Southeast University.