Measuring and Explaining the Impact of Online Communities on Consumer Markets

Corporate Reputation Review Volume 13 Number 4

‘ The Equalizer’ : Measuring and Explaining the Impact of Online Communities on Consumer Markets

Mark Lee Hunter INSEAD, Fontainebleau Cedex, France

David A. Soberman Rotman School of Management at the University of Toronto , Toronto , Ontario , Canada

ABSTRACT are fragmented and diverse. The analysis Our objective is to examine how online com- shows that online communities lead to fun- munities affect the functioning of markets for damental changes in the way that informa- durable goods with particular emphasis on mar- tion moves and is used within a marketplace. kets that are both fragmented and diverse. Our As a result of these changes, online com- thesis is that online communities have the munities can transform relatively ineffi cient potential to make relatively inefficient frag- fragmented markets into markets that are mented markets more efficient. This effect will more effi cient. In particular, information manifest itself through the observed pricing for exchanged within the online community the goods in both standard commercial settings can have a direct impact on the pricing that and in internet exchange institutions. A sec- is observed in standard commercial settings ondary effect of the online communities should ( ‘ bricks and mortar’ retailers) and in internet be to amplify the impact of quality (as per- exchange institutions (eg, eBay). ceived by the user) on market transactions. We There is a signifi cant analytical literature conclude with an analysis of the foundations that examines how prices are established in of credibility for user-generated content within a market where consumers are heterogene- online communities. ous in their capability to gather pricing Corporate Reputation Review (2010) 13, 225– 247. information. In these models, competing doi: 10.1057/crr.2010.25 retailers adopt different pricing strategies. Either they set a low price to capture vol- KEYWORDS: internet marketing ; online com- ume from ‘ informed consumers’ by pricing merce ; online communities ; product innovation ; low or they set a high price to capture high user- generated content profi t per sale by only serving ‘ uninformed consumers’ (Salop and Stiglitz, 1977, 1982). INTRODUCTION Invariably, these models relate to homoge- neous goods. In contrast to this literature, Background our focus is the process by which informa- The objective of this article is to demon- tion about the quality (or value) of hetero- Corporate Reputation Review, strate how online communities can affect the geneous goods is translated into prices. Vol. 13, No. 4, pp. 225–247 © 2010 Macmillan Publishers Ltd., functioning of markets, especially those that When the quality of products cannot be 1363-3589

www.palgrave-journals.com/crr/ Impact of Online Communities on Consumer Markets

evaluated by inspection, there are many attention to strengths and weaknesses that vehicles through which consumers become manufacturers do not or cannot (due to a informed about quality including warranties, lack of resources) evoke themselves, we call standards and advertising (Spence, 1977; this an ‘ equalizer’ effect. This echoes the Leland, 1979 and Klein and Leffl er, 1981). hopes expressed by Sullivan (2008): ‘ As Moreover, in the context of online auctions, E-Commerce sites add consumer-generated there is recent evidence that auction par- review systems, marketers and consumers ticipants use the attributes of an auction hope truth trumps disingenuousness’ . environment to make inferences about qual- A second equalizing effect of the online ity (Li et al. , 2009). It is also well known community is to amplify the impact of qual- that consumers consult experts and/ or out- ity as perceived by consumers/ users on side organizations to assess the quality of market performance. Quality perceptions goods (Carlton and Perloff, 2000). clearly have an impact on the prices that Our interest is different. We wish to bet- manufacturers can obtain for new products. ter understand the process by which con- However, quality is also an important pre- sumers transmit information to each other dictor of future value for buyers of durable in a manner that was not possible prior to products that depreciate over time. Not only the penetration of the internet and the is durability ‘ a standard component’ of qual- growth of online forums. There is some ity, but the care that a manufacturer takes research in this area: a recent study demon- to build a product (and build in quality) also strates the impact of user ratings in an seems to affect the product ’ s ability to stand internet institution on sales through that the test of time. It is well known that high- institution (Chevalier and Mayzlin, 2006); quality products such as BMWs, Beneteau however, our objective is to analyze the yachts and Rolex watches are better at impact of ratings from an online agora (or retaining their value in used product markets public space) on prices observed outside the than competitive products with inferior agora. 1 A study by Hunag et al. (2009) sug- quality perceptions. gests that these ratings should have an effect We will demonstrate how a particular outside the agora: they fi nd that the presence online community has acquired signifi cant of product reviews from other consumers infl uence over the perceptions of quality in on the Web has a positive effect on con- a specifi c durable goods market. In this mar- sumer search for experience goods. In a ket, we show that this infl uence can have a larger sense, the online agora may be a crit- measurable effect on prices for used goods, ical vector of a process through which ‘ the as well as the prices that manufacturers com- accumulation of generalized reputation can mand for new goods. The existence of these begin before the fi rst direct interactions effects is shown through the analysis of data with customers or other stakeholders and from online auctions and online retailers in may in fact serve to attract such stakeholders ’ conjunction with user ratings generated ( Petkova et al. , 2008). through the online community. The implications of this dynamic for The drivers of these effects reside in qual- commercial enterprises are that for certain itative factors, which we have examined in categories of goods, freely available internet the course of three years of studying this user opinions may have as much effect on particular online community. The factors the reputation of a brand, and hence on include the credibility of user ratings pro- purchase decisions, as information provided vided within a particular online community by manufacturers or retailers to consumers. and the credibility of the community itself. To underline the power of consumers to call Another decisive factor can be found in

the nature of the information provided by communication from sellers, help from sales- online user ratings in this community, which people, the actions of the sellers themselves encompasses not only the characteristics of (such as pricing), independent evaluation or- the products under review, but the charac- ganizations, media reports (and public rela- teristics of the reviewer. In other words, tions) and word of mouth (WOM). However, the power of online reviews is based on aside from independent evaluation organiza- (a) transparency concerning the expertise, tions, media reports and WOM, most infor- interests and character of individual review- mation in this model originates from sellers. ers; and (b) the collective weight of user Seller-provided information is very important opinions. The richness of information pre- if a product category lacks independent eval- sented in a framework that allows for rapid uation organizations (eg, Consumer Reports ) and assessment of the quality of products is a does not receive extensive media coverage. critical element of user ratings. In our con- In an offl ine world, WOM is the only clusion, we provide examples of fi rms that signifi cant buyer information source that is have incorporated this power into their mar- user generated. Not surprisingly, WOM re- keting efforts, and suggest how other fi rms ceives signifi cant weight in many consumer may capitalize on this opportunity. purchase decisions. However, traditional WOM is subject to important limitations: The Impact of Online Communities on Information Flows in Consumer Markets 1. WOM is only relevant when the buyer A key proposition of this study is that online has a number of friends (acquaintances) communities may alter, and in some cases who have had experience with the prod- have clearly altered, the dynamics of infor- ucts in question. mation exchange between buyers and sellers 2. WOM is generally not quantitative. When in consumer markets for durable goods. In you ask a friend or acquaintance about a traditional markets, the fl ow of key informa- specifi c product s/ he owns, the reply is tion is straightforward. This is illustrated in typically along the lines of ‘ this product Figure 1. is great and works well’ rather than ‘ this A salient aspect of this model is that pri- product scores 5 on performance and 6 on or to the appearance of an online community, style for an overall score of 5.5’ . buyers have several sources of information 3. WOM is not statistically reliable. For upon which to base decisions. These include example, even if the information collect- (but are not limited to) advertising and ed is quantitative in nature, it is diffi cult to

Figure 1 : How non-owners gather information in a traditional geographically dispersed market

assess the reliability of a product based on similar interests (Hill et al. , 2006 ). As a one or two data points. result, for almost any product, a poten- 4. Most importantly, it is diffi cult to draw tial buyer can fi nd others who have had detailed comparisons among compet- experience with the product class, and ing products through WOM. Most users often with specifi c models. only have experience with one or at most 2. Online communities can create online a few products in a given category; there- systems that allow people to provide fore comparisons need to be implied. quantitative ratings on products. Moreo- ver, online community information is For these reasons, in markets that are geo- statistically reliable because many users graphically dispersed and fragmented, the can provide rating information. impact of WOM is small in comparison to 3. Because quantitative information is being other sources of information. collected, detailed comparisons between This may not be the case in online com- products are possible and insightful. Of munities. Online communities have the course, all quantitative information collect- potential to channel and format WOM in- ed through survey technology is subject to formation from multiple sources. We use the errors, biases and limitations. Nevertheless, term ‘ online community’ to describe insti- it is clearly less subject to errors, biases and tutionalized internet-based links between limitations than WOM information col- market participants who are geographically lected from one or more friends . and socially diverse. It is only recently that these links became technically feasible due Consequently, in a market where there is an to broad penetration of high-speed internet active online community, the fl ow of informa- access (since the mid- to late 1990s). Cer- tion is richer and more diversifi ed. In Figure 2, tainly, electronic communication between we propose a framework to map this fl ow of people who are geographically diverse has information. The framework represents a for- been possible for more than a century malization of our observations gathered through (through the telegraph, the telephone, the discussions with industry participants, participa- telex, fax, and since the 1980s, electronic tion in online forums and discussions with bulletin boards). However, before the devel- owners of various products. opment of the web browser, online com- The distinction between ‘ web-active’ munication depended on the initiator of the owners, ‘ traditional’ owners and non-own- communication having the address of the ers in our diagram serves to underline the person to whom s/ he wished to communi- fact that in general, web-active individuals cate. In contrast, forums, blogs and com- have access to more information, and to munity posting boards visible through web more sources of information, than individu- browsers and search engines allow people to als who do not enter the online agora.2 This fi nd and contact each other at will, based on model implies that the online community common interests and at minimal cost in terms allows for the creation, collection and dis- of time and money. semination of information that is relevant It is useful to highlight how online com- and impactful for exchanges between buyers munity information is largely immune and sellers even when these transactions occur in from the shortcomings of traditional WOM the world of bricks and mortar. We discuss this information listed above. implication more fully below. The process by which information is collected and dif- 1. The online community by defi nition fused within online agoras is illustrated in allows people to fi nd others who have Figure 3.

Figure 2 : How non-owners gather information in a market with a Web-enabled community

Figure 3 : The creation, collection and dissemination of information in online communities

We posit that the fl ows of information in A remaining question is, how does the Figures 2 and 3 affect the functioning of a information collected through online com- market when the online community is rela- munities impact individual purchase deci- tively sophisticated. By ‘ sophisticated’ , we sions? In Figure 4, we provide an illustration refer to expertise related to product features, of the process by which this information is quality and price for value, as well as expres- likely to affect purchase decisions. sive capability. The collective knowledge of This model is based on interviews with peo- the online community is of little benefi t to ple in an actual market (the North American its members unless they are willing and able electric guitar market) about their activity in to share it. One of the most remarkable fea- the market both offl ine and online. The mod- tures of these communities is that a great el represents the procedure followed by web- deal of information is shared, even when active guitar buyers when they wish to sharing may affect the personal interests of ascertain the value of a potential purchase, or a given member. (In particular, our model are simply curious about a product they have and data suggest that by alerting other mem- never encountered before. The procedure is bers to the value of a particular product, a common knowledge among community member increases the likelihood that s/ he participants: in a recent exchange on the will pay a higher price for that item in the reverendguitars.com forum, one member ad- future.) We will consider possible reasons for vised another who was considering a brand of this apparent selfl essness in a later section. guitars to look at its user reviews on HC. The

Figure 4 : How information affects market transactions

second member replied, ‘ I have spent an in 2005), accessories such as electronic ‘ sig- inordinate time on Harmony Central’ . 3 nal processing ’ effects ( $222 m in 2005) and To sum up, the role previously played by other equipment such as strings, recording individual WOM on purchase decisions may software and microphones.4 Brick and mor- be supplemented or superseded by the col- tar dealers are the main outlets for this lective WOM of an online community equipment. Physical stores range from small when the online community offers the ad- shops run by guitar enthusiasts to the cross- vantage of access to both diversifi ed and country GC chain (Guitar Center), which quantitative information. In these situations, music trade insiders compare to Wal-Mart. buyers become active seekers of product in- In 2005, GC had 242 stores of up to 1,800 m 2 formation and a major focus of the online in size and was opening one to two new community is to gather, codify and diffuse stores per month. That year, the top 15 mu- that information. In the following section, sic retailers had sales ranging from $1.8 bn we discuss how this activity might affect be- to $ 30 m and accounted for total sales of havior in a real market. $3.1 bn, or nearly 40 percent of the total $7.8 bn business. 5 Institutional Context: The Online Electric There is also a large market for used elec- Guitar Market tric guitars. Daddy’ s Junky Music, a chain Our objective is to examine the impact of of 21 stores specializing in second-hand the online world in a market where par- instruments, sold over $ 33 m of equipment ticipants appear to be aware of and involved in 2005, and on any given day an ebay.com with an online agora. In particular, we search generated a listing of approximately require a setting (a) that is geographically 7,500 used electric guitars, about half the dispersed, where multiple manufacturers dis- total number of guitars available on eBay. tribute through local retailers to end users The world ’ s largest manufacturer of mu- and (b) where a signifi cant online commu- sical equipment, Japan’ s Yamaha Corpora- nity has developed. tion, reported sector sales of $ 2.69 bn in The setting we use to examine these issues 2006, but did not report separate fi gures for is the North American market for electric guitars. The dominant electric guitar manu- guitars. This category generated approxi- facturers in the US, Fender and Gibson, are mately $700 m in retail sales for 1.6 million both privately held and do not report sales. units sold in the US in 2005, the latest year Each of these fi rms accounts for tens of for which data were available at the time of thousands of instruments annually. this study. Electric guitar buyers also buy The diversity of models available to amplifi ers (over $400 m and 1.2 million units guitarists has exploded. As the self-defi ned

‘ international music products association’ , their use by such artists as Stevie Ray NAMM, 6 noted in its 2006 Global Report, Vaughn, Jimi Hendrix and Eric Clapton, ‘ 52 years ago, Fender offered two Strato- though lower-priced models may offer caster models [priced at $229 and $ 249]. equivalent quality. The same applies to Paul Today, including Squier [Fender’ s discount Reed Smith (PRS), for whom the endorse- brand], Fender offers upwards of 75 varia- ment of Carlos Santana was a critical factor tions on the ‘ Strat’ , ranging in price from in building the brand, and for Ibanez, whose $129 to $ 9500 ’ . association with Stevie Vai established the Another driver of product diversity is a fi rm as a leader in high-end instruments as growing number of competitors. Since the well as beginner and intermediate models. 1990s, branded and OEM manufacturers Of course, the vast majority of new from Japan, Korea, Mexico, Indonesia and entrants in guitar manufacturing, regardless of China have signifi cantly pushed down the the intrinsic value or quality of their instru- price and pushed up the quality of entry- ments, do not benefi t from major celebrity and intermediate-level (under $ 1,000) in- endorsements. Nor do many manufactur- struments. 7 The number of electric guitars ers, especially smaller ones, have the sold in the US grew by 213 percent from resources needed to actively promote their 1996 to 2005, but the average unit price fell products through paid advertising or print 46 percent, from around $630 to $ 350. 8 At media, traditionally the primary sources of present, it can be safely said that there are product information (along with retailers) more than 20,000 different models of new for guitarists. Moreover, small manufactur- and used electric guitars made by more than ers are at a disadvantage in the marketplace 1,000 manufacturers, which are currently by the fact that many of the most important bought and sold in the market. retailers do not carry their products (we Consumers are thus confronted with a discuss this more fully later). We posit that wealth of choices, and with the opportu- small manufacturers can compensate for or nity to purchase increasingly well-made ‘equalize ’ such disadvantages through a goods at steadily shrinking prices. However, strong positive relationship with online the relationship between the quality of elec- communities that vouch for the quality of tric guitars – that is, the reliability of the their products. components, the ease of use and mainte- nance, the beauty of the fi nishes and the Harmony-central.com: The Online Agora richness of the sound – and the prices that of the Guitar World many consumers pay for them is weak. In In the late 1990s, an online community of other words, the pricing of instruments is guitar players took shape. It included sites relatively ineffi cient because there is a low where players bought and sold instruments correlation between pricing and quality. A and traded performance and recording tips number of fi rms have been able to charge (such as The Gear Page), forums dedicated premium prices for instruments that do not to specifi c manufacturers (such as the Fender offer better quality of components, fi nish or Discussion Page) and instructional news- sound than other, less costly products. This groups and sites (such as Wholenote). In is largely because emotional factors have an 1998, a site called harmony-central.com, important role in determining the products which offered diverse content to musicians that customers purchase. Iconic guitars like (from song transcriptions to manufacturer the Fender Stratocaster and the Gibson Les product announcements), began to collect, Paul enjoy high prices as a function of their compile and post detailed user ratings on 21 role in the history of popular music, and categories of musical instruments, including

guitars, bass guitars, keyboards, amplifi ers, The movements of this review from the recording equipment and electronic effects, agora followed the movements shown in among others. The specifi c information Figures 3 and 4. In essence, the review was solicited for these ‘user reviews’ included (and treated like hot news by a non-offi cial user still includes) details on a broad range of network and it spread quickly beyond its quality parameters, details about the review- initial posting. Reverend owner Joe Naylor er’ s experience, such as the length of time told us, ‘ People were linking to the review the reviewer has played, in which settings from the Fender Discussion Page. People (professional or amateur performances, stu- were talking about it’ . Another HC review- dio work, etc) and other equipment the er alluded to Ray’ s infl uence, and how it fi t reviewer owns or has been able to compare with his personal search for innovative prod- fi rsthand. 9 Almost 1,000 guitar makers are ucts: ‘ Until I read some of these HC reviews, covered by user reviews. The number of including Will Ray’ s, I didn ’ t know any- reviews of specifi c models from a given thing about the guitar, where it was manu- manufacturer may vary from over 200 (for factured, etc. … . My personal goal – at this a fi rm like Fender) to 1 (for the French fi rm late date in my life – is to get off the beaten Fine Resophonic), which manufactures un- path. I haven’ t found too many new ideas der 20 instruments per year for a discreet for [creating sound] with six strings, but this celebrity clientele). The number of review- is one’ (Hunter and Soberman, 2007a). 11 ers per product may range from 1 to 453 Further anecdotal evidence for the impact (for Fender’ s mid-priced Standard Strato- of HC is plentiful. Owners of small guitar- caster, a very popular model). Reviewers manufacturing fi rms often intervene direct- assign a score of 1– 10 in various categories ly on HC to correct misleading or incorrect (features, sound quality, fi t and fi nish, cus- information about their products. Links to tomer support, overall rating). Thus the positive HC reviews are frequently posted reader of an HC review can judge both the by instrument sellers on eBay and online instrument being rated, and also the capac- ‘ fl ea market’ sites such as Craigslist, and ity of an individual reviewer to rate fairly. forum members at different sites report The impact of these reviews on guitar spontaneously that they consult HC. Online manufacturers seems to be important. 10 In a music retailers (such as the industry leader, case study of Reverend Musical Instruments, musiciansfriend.com and its competitor a single review by a ‘ musician’ s musician’ of music123.com) also allow instrument buyers a Reverend product at HC had an immedi- to post reviews, but the latter are not near- ate impact on sales of the model reviewed ly as detailed as those found on HC with and played a crucial role in establishing the regard to the instrument or the person sub- fi rm ’ s reputation. To give the fl avor of a mitting the review. Moreover, retail-spon- credible user rating, we provide a quote from sored reviews do not play a visible role on the aforementioned review. Author Will eBay, where they are rarely cited or linked. Ray of the Hellecasters band concluded: It seems that HC plays a unique role in the musical instrument market as the agora of I’ ve owned probably 500 guitars over the the web-active musical community’ s collec- years. (Wish I still had some of them). tive opinion concerning products. I also have my own signature model Fender guitar. I get a lot of instruments Poisoning and Boosting: Key Threats to thrown at me every year by companies. the Community ’s Credibility I’ m picky and hard to please. But I really An impediment to the impact of user-gener- like this guitar. It’ s a keeper. ated content on this or other online agoras

is that it can be biased or poisoned for com- than 20,000 different guitar models from petitive advantage ( Dellarocas, 2006). When almost 1,000 manufacturers are rated on it is easy for a fi rm to post negative informa- HC. A large company like Fender might tion about competitors or glowing informa- have ratings for more than 400 models. One tion about its own products, the value of the popular model can have hundreds of ratings information is reduced. In addition, there is that are posted. Thus, the task of poisoning evidence of ‘ brand communities’ that may the ratings or boosting all ratings to a degree exist within an online community such as that would affect the cumulative ratings for HC. These brand communities are cited for a given product (which are averaged accord- their potential to enhance the loyalty to a ing to the number of reviewers) is gargan- specifi c brand, but also for their tendency to tuan, if approached manually. Clearly, fi rms create ‘ oppositional loyalty’ to competing that want to poison or boost ratings would brands (Thompson and Sinha, 2008). need to develop automated systems to gen- As a result, it is important to assess the erate and post false reports. degree to which ‘ poisoning’ might be a In this regard, there are a number of safe- problem on HC. The reliability of HC rat- guards built into HC in order to prevent the ings is a function of two factors: (1) the automated poisoning or boosting of ratings. motivation of sellers to poison or boost; and First, the rating pages fi lled out by the user (2) the ability of sellers to poison or boost. are screened by automatic robots (with skill The motivation to poison or boost refl ects testing questions, mathematical problems the perspective of competitors in this market. in words and visual identifi cation tasks). Essentially, there are several very big fi rms Reviews submitted by an automated rating such as Yamaha, Fender and Gibson with submission system are rejected by these strong reputations, and hundreds of small robots. The second level of protection is a businesses that manufacture guitars. For the set of rules posted on the site that need to most part, the motivation of small fi rms (or be followed in order to validate a rating. advocates) to poison the products of big Examples of the rules are as follows: companies is low. As big companies have so many models, the vast majority of a large (a) Raters are expected to provide comments company’ s product line does not compete in at least one fi eld that go deeper than with the products of a small company. The simply saying a product is good or bad. motivation of small fi rms to poison the rat- Reviews without detailed comments are ings of other small fi rms is also low, because not used. competition among small fi rms in this market (b) Superfi cial glowing or hostile reviews may is largely atomistic (ie, they do not really not be published. compete with each other). Big fi rms do not (c) Specifi cations for products must be includ- perceive small fi rms as competitors; however, ed, so readers can see if the right product is they do have an incentive to poison other being discussed. big competitors. In sum, when a market is (d) A unique e-mail address must be submitted as fragmented as the electric guitar market, for each review. the only real motivation to ‘ poison’ exists between large fi rms. The motivation to boost Moreover, HC has staff that review ratings certainly exists for most fi rms, but primarily before they are posted. It takes between three for small fi rms since they cannot rely on ad- and fi ve days before any submitted rating vertising to the same extent as large fi rms. is added to the site, and only then are its The ability to poison or boost is another numerical ratings tabulated and factored into matter. It is important to recall that more the ratings that have already been posted

on a given model. (In one instance unre- through a network of small retailers, with lated to this study, one of the authors no retail presence in ‘ superstores’ like GC submitted a detailed but highly critical or musiciansfriend.com. In other words, review of a particular product; it was not perhaps due to Reverend’ s absence of mar- posted on HC until ten days had elapsed, keting activity, the impact of the online an unusual delay.) Finally, manufacturers agora on Reverend products relative to who are suspected of boosting or poisoning other products is high. The agora enables are banned from the community for a min- Reverend to compete on a more equal imum of one year – a sanction that can have basis, as a producer of high-quality products, grave consequences, particularly for a small with manufacturers that dispose of greater manufacturer who lacks other means of marketing resources. promotion. 12 Moreover, online community quality rat- To conclude this discussion, the validity ings appear to have a signifi cant effect on of online user ratings can indeed pose a street pricing (at brick and mortar music problem. However, if the motivation for stores). These effects are found regardless of most industry players (small and large) to brand investments (advertising and endorse- poison or boost is small, and the ability of ments) by the manufacturers. Specifi c mod- users to poison or boost is limited by a series els appear to enjoy an ‘ HC premium ’ , or of electronic and manual procedures, then conversely, to suffer from a discount, in par- scores are less likely to be distorted. It is allel with their user ratings. clear that HC is aware of this danger, and We also fi nd that user ratings do not just seeks to protect the credibility, infl uence and have an impact on selling prices but also on value of its ratings through such proce- whether or not transactions occur in the online dures. environment. Specifi cally, relatively low ratings on HC for a given product, compared Preview of Findings: Online User to competing similar products, reduce the Ratings and the Market likelihood of an eBay transaction occurring. The initial question that intrigued us was This follows from the reasoning that there whether HC reviews had an impact on the will be greater range in the expected qual- sale prices of used instruments on the eBay ity of a poorly rated product. According to internet auction site. We deduced that Akerlof (1970), the greater the variance in positive reviews would be associated with the expected quality of a product when the higher prices for specifi c used products. seller is informed about the product’ s qual- More specifi cally, our objective was to test ity and the buyer is not, the higher the like- the statistical signifi cance of some of the lihood of market failure (or no exchanges hypothesized links shown in Figures 2– 4 . occurring). If marketing can be effectively Our results show that information counteracted by internet user reviews, qual- exchanged and posted within the online ity or its lack becomes even more critical in community indeed has a signifi cant impact purchase decisions. on prices for used goods that are observed in the online environment. However, this effect appears only for Reverend (by far the DATA COLLECTION AND ANALYSIS least well known of the seven guitar brands for which data were collected). Reverend Methodology enjoys no ‘ superstar’ endorsements, under- We collected information on a representative takes little display advertising through online sample of more than 100 different instru- or print publications and is distributed ments that were on sale on eBay over an

eight-week period starting in mid-February brand due to unique manufacturer-based 2008 and ending in April 2008. Each mod- discounts and differences in brand equity. el in the sample met the following criteria: Each manufacturer targets a specifi c retail 1. All models were manufactured in Asia: price in the market for each model. This other than the Ibanez instruments that then determines the recommended discount were produced in China, all guitars in the from list prices needed such that retailers sample were produced in Korea.13 How- post prices that are consistent with the ever, all guitars in the sample are targeted manufacturer’ s objective. Owing to (a) the to ‘ value buyers’ , who seek the intrinsic idiosyncratic nature of each manufacturer quality of a guitar, rather than simply policy with regard to its distribution net- purchasing a leading brand. Indeed, HC work; and (b) differences in brand equity, reviewers often justify purchases of off- we hypothesize that the primary factor that brand models for this reason. affects the difference between manufacturer 2. The retail prices for models selected are list prices and observed street prices is a in the same range: street prices were tak- manufacturer effect. This is driven by the en from new guitar prices at the leading reputations that competing fi rms have online retailer (see the section ‘ Data spe- developed over time within the electric cifi cs ’ below for further detail on price guitar community. As explained earlier, benchmarks). these reputations are a function of history, 3. The models selected have a broad range well-known musicians who use specifi c of features including bolt-on and set-neck brands of guitars and the performance of the models: a summary of the features of gui- instruments in terms of sound, durability and tars collected in our sample is provided in feel. Appendix A. 4. There are a minimum of three HC user H2: The HC Score will have a signifi cant effect reviews per model in the database. on the relative difference between manufac- 5. The models chosen are widely available turer list prices and street pricing observed via eBay (new and used) and at brick and for each model of guitar after accounting for mortar retailers. Thus, at some point, most brand effects. buyers can examine physical samples at traditional retailers if they wish to com- In accord with the fl ow of information pare before buying online. modeled in Figure 4, we hypothesize that ‘ HC ’ ratings will have a signifi cant effect on Through these criteria we have, in effect, street pricing of new models (at brick and ‘ leveled’ the differences between the instru- mortar and online music stores). Recall that ment models, with the signifi cant exceptions many buyers of new models are fi rst-time of advertising, artist endorsements and retail buyers of electric guitars; likewise, many distribution. experienced buyers have no experience with or physical access to a given model before Key Hypotheses purchase. For these buyers, a site like har- Our analysis will seek to test the following mony-central.com allows them to obtain hypotheses: useful information at negligible cost. In other words, the impact of ‘ online community’ H1: The relative difference between manufac- quality ratings should make it more diffi cult turer list prices and street prices for brand/ to sell poorly performing new products at models depends primarily on the specifi c high prices. Conversely, quality products

should enjoy an ‘ HC premium ’ based on buyer is not, an increase in variance of their user ratings. We evaluate this hypoth- expected quality will increase the likeli- esis by examining how HC scores affect the hood of market failure (or of no exchanges relative difference between the manufactur- occurring). er’ s list price and the observed street price after accounting for brand differences. Data Specifi cs Each model in the data set is identifi ed by H3: The HC Score for each model will have a a brand and model. We denote the brand as signifi cant effect on the percentage difference b and the model as i . For each guitar bi , we between the street price (the price for a new collected a complete brand model descrip- guitar) and the observed selling price for the tion, the ‘ overall’ rating score from HC model, in used condition, in the online auc- ( HCS bi ), the number of users who submitted tion environment. ratings, the manufacturer list price (MSL bi , which we collected from online retail sites) Consistent with the fl ow of information and the street price (SP bi ). The street price shown in Figure 2, we hypothesize that for each model is based on the retail prices ‘ HC ’ ratings will have a signifi cant effect on posted by GC. GC is America’ s largest mu- used guitar pricing (on eBay). In other sical instrument retailer with sales of more words, the impact of a weak ‘ online com- than US$ 1.78 bn through more than 214 munity’ quality rating is to make it signifi - retail stores. Owing to its size and infl uence, cantly more diffi cult to sell a used product GC is, for all intents and purposes, the price at a price that is close to the price for a new setter for instruments in the brick and mor- guitar. We also note that this effect is likely tar market, and the company guarantees that to be stronger for lesser-known brands than it will not be undersold. The prices at GC for well-known brands. Buyers of used gui- are also identical to prices on musicians- tars themselves are often members of ama- friend.com, the retailer’ s online operation. teur music communities. Here anecdotal In the case of Reverend, whose products are evidence from fellow musicians is likely to not available through GC or musicians- have a strong effect on buying decisions. friend.com, we used the lowest retail prices This would suggest reduced impact for ‘ HC ’ for new products found on eBay at the time ratings. of this study. These prices are closely con- trolled by Reverend Musical Instruments. H4: The HC Score for each model will have a Each guitar in the data set was also the signifi cant effect on the likelihood that a suc- subject of at least one internet auction on cessful transaction for a listed guitar occurs. eBay. Through the eBay tracking facility, we followed auctions for every guitar in the We hypothesize that the HC ratings will data set and collected: the Buy it Now price have an impact on selling prices and on ( BIN bi ), the initial bid (IB bi ) and the Sale 14 whether or not transactions occur. Sellers on Price (SP bi ). eBay typically set ‘ reserve’ or minimum prices for their goods. Thus low ratings on Results of the Analysis HC should reduce the likelihood of a trans- Our fi rst interest is to understand the factors action occurring. This follows from the idea that affect the relative difference between that there will be greater range in the the manufacturer’ s list price and the street expected quality of a poorly rated product price for each model. We therefore defi ne quality. In a market where the seller is a variable FN bi as the percentage decrease in informed about the product’ s quality and the manufacturer’ s selling price for product bi

that is associated with the observed street Table 1: Percentage decrease price. To evaluate H1 and H2, we estimate (versus list price) in the retail price of selected guitars the following linear regression: Parameter Estimate T-stat Signifi cance =+ + + FNbi bb01HCSbi b 2Revbi b 3Dilbi 0 54.8692 7.9579 p < 0.01 +++ bbb456Epi Prs Was 1 ( HCS ) − 3.2875 − 4.3717 p < 0.01 bi bi bi 2 ( Rev ) 4.5289 4.0636 p < 0.01 + b GGnl + e 7 bi bi 3 ( Dil ) 15.1249 14.0449 p < 0.01 (1) 4 ( Epi ) 13.3766 11.5308 p < 0.01 5 ( Prs ) 0.2776 0.2553 n.s. The variables Rev , Dil , Epi , Prs , bi bi bi bi 6 ( Was ) 24.1633 18.3759 p < 0.01 Was bi and Gnl bi are dummy variables to 7 ( Gnl ) 8.1475 6.3228 p < 0.01 allow for brand-specifi c effects for Rever- end, Dillion, Epiphone, PRS, Washburn and G &L, respectively (Ibanez is the base brand, so it does not have a dummy varia- Table 2: Percentage decrease (versus ble). The following are the results of the list price) in the retail price of selected estimation. The model is based on 109 guitars (re-estimate with the features as covariates) observations and the R -square and adjusted R -square for the estimation are 0.912 and Parameter Estimate T-stat Signifi cance 0.906, respectively (Table 1). The independent variables included in the 0 54.30621 8.203009 p < 0.01 estimation explain more than 90 percent of 1 ( HCS ) − 3.69965 − 5.12788 p < 0.01 the variance in the percentage difference 2 ( Rev ) 5.632595 1.495427 n.s. 3 ( Dil ) 14.53792 5.046497 p < 0.01 between the manufacturer list price and 4 ( Epi ) 12.55398 4.170682 p < 0.01 the street price as evidenced by the high 5 ( Prs ) − 0.62372 − 0.20851 n.s. R -square values. Moreover, all of the variables 6 ( Was ) 23.4819 7.669383 p < 0.01 except 5 are signifi cant (The insignifi cant 7 ( Gnl ) 10.12205 4.48359 p < 0.01 coeffi cient for PRS simply means that the neck type 2.510026 1.628604 n.s. average reduction in manufacturer list price pickup type 2.210105 2.513758 p < 0.01 for PRS guitars, Prs , is almost identical to wood type 0.622482 0.222734 n.s. bi 1.797518 1.398928 n.s. those observed for the base brand, Ibanez). bridge type In order, to ensure that the features of guitars (listed in Appendix A) do not affect the observed prices in the retail environ- with single coil pickups are more heavily ment, we also estimate the model in equa- discounted than those fi tted with Hum- tion (1) with neck-type, pickup type, wood bucker pickups. type, control type and bridge type as covari- Independent of whether product features ates. The results from this estimation are are included as covariates, we fi nd that the presented in Table 2. 15 effect of HCS on FN is highly signifi cant. The R -square and adjusted R -square for The negative sign implies that higher HC the re-estimated model are 0.92413 and scores lead to lower percentage reductions 0.895, respectively. The value of F -statistic from manufacturer list price. It is important for comparing nested models is 3.894, which to note that this variable captures signifi cant is signifi cant at the 5 percent level. 16 The variation in markdown that exists within the inclusion of the covariates improves the brands for different models. These results model but only pickup type is signifi cant: guitars confi rm both H1 and H2.

In order to evaluate H3, we examine the Table 5: Percentage reduction for price of impact of HC scores by model on the per- used Reverend guitar versus street price centage reduction observed from the street of new Reverend guitar price to the selling price that occurred on eBay for a used version of the same instru- Parameter Estimate T-stat Signifi cance 17 ment. First, we estimate the equation with- 18 0 2.428 1.982 p < 0.10 out the brand and feature variables. (The 1 − 0.230 − 1.814 p < 0.10 number of data points is reduced in this 0.044 0.766 n.s. analysis because 34 of the 108 guitar auctions 2 resulted in no trade.) Here we defi ne FU bi as the percentage reduction in street price that is observed for the used product that is auctioned on eBay. To evaluate these effects, In this estimation, the HC score has a we estimate the following linear regression. signifi cant effect on the percentage reduction in street price for Reverend products that FU =+bbHCS + e (2) are exchanged on eBay, yet only a small frac- bi 01 bi bi tion of the dependent variable is explained.

To further investigate the factors that affect There were 74 data points in the estimation FU , we estimate equation (2) but include and the R -square and the adjusted R -square bi the brand variables as covariates (Table 4). were 0.059 and 0.046, respectively (Table 3). There were 74 data points in the estimation, and the R -square and the adjusted Table 3: Percentage reduction for price R -square were 0.491 and 0.437, respectively. of used guitar versus street price of new This model is signifi cantly superior to the guitar simple model and suggests that the most important factor that determines the prices Parameter Estimate T-stat Signifi cance obtained for used products is the brand. It appears that buyers are willing to pay a pre- 0 0.8719 3.1722 p < 0.01 mium for brands such as Reverend, Epiphone, 1 − 0.0633 − 2.1189 p < 0.05 PRS and G& L. Moreover, the signifi cant effect of HCS bi disappears. This suggests that the effect of HCS (if signifi cant at all) is less important than the brand effect. To investi- Table 4: Percentage reduction for price gate this possibility, we conduct ‘ within brand’ of used guitar versus street price of new guitar (re-estimate with the brands estimations where we utilize HCS bi and neck- as co-variates) type (a feature that varies signifi cantly within certain brands) as explanatory variables. Parameter Estimate T-stat Signifi cance =+ + + FUbi bb01HCSbi b 2neck typebi e bi 0 0.334 0.964 n.s. 1 ( HCS ) 0.006 0.148 n.s. (3) 2 ( Rev ) − 0.180 − 3.543 p < 0.01 3 ( Dil ) 0.069 1.217 n.s. The analysis was conducted for all the 4 ( Epi ) − 0.122 − 2.185 p < 0.05 brands. For the Reverend brand, there were 5 ( Prs ) − 0.091 − 1.750 p < 0.10 29 data points (of the 31 Reverend guitars 6 ( Was ) 0.095 1.613 n.s. ( Gnl ) − 0.121 − 1.872 p < 0.10 in the sample, all but 2 resulted in a sale) 7 and the results are given in Table 5.

Table 6: The likelihood of a sale on eBay as a function of its Harmony Control Rating

Parameter Estimate SE Wald Signiﬁ cance

0 − 7.545 3.967 3.618 0.057 1 0.913 0.438 4.352 0.037

The R -square and the adjusted R -square found on HC. For this we defi ne a binary were 0.113 and 0.045, respectively. Note variable NS bi which equals 1 if a product that 1 is signifi cant at the p < 0.10 level, was sold and 0 if it was not. Thus, we esti- suggesting that Reverend models with high- mate the following model: er HC scores command higher prices in eBay auctions. We conducted similar regres- ++ exp(bb01HCS e ) sions for the other brands in the sample and P = bi bi NS ++ + (4) either the effect of the HCS score is insig- 1 exp(bb01HCSbi e bi ) nifi cant or the number of data points for the brand is insuffi cient to reach a conclusive The error term bi is assumed to be dis- estimate. tributed according to the extreme value dis- In sum, we fi nd that the effect of HCS tribution (double exponential). There were on the selling price in eBay auctions is in- 109 data points in the estimation. The fol- signifi cant compared to the effect of brand lowing are the estimation results. With a reputation. It appears that ‘ the brand’ is the logit estimation, the signifi cance of the most important determinant of the degree parameters is evaluated according to the to which used guitars retain their value as a Wald test Table 6. function of the original selling price. To be Table 4 shows that HC Score coeffi cient specifi c, when there are signifi cant numbers is signifi cant as a predictor of the probabil- of musicians who own guitars made by a ity of a sale occurring. That is, when HC specifi c manufacturer (as is the case with Score is higher for a model listed in an auc- Dilion, Epiphone, PRS, G& L, Washburn tion, the likelihood of a sale occurring is and Ibanez) or a manufacturer has a strong higher. On the other hand, the naï ve model reputation, the effect of HCS on the used (without HC Score included in the model) prices for that manufacturer’ s instruments is achieves a correct classifi cation of 67.9 negligible. However, when the stock of percent, whereas the model that includes existing owners for a given manufacturer is HC Score achieves a correct classifi cation of small and a brand is relatively unknown, like 66.1 percent (based on a cut value of 0.5). Reverend, it appears that HCS can affect Our fi ndings suggest that HC Score does the percentage reduction in street price that have an effect on the likelihood that a sell- is observed for that manufacturer’ s models er and buyer fi nd a mutually acceptable price in eBay auctions for used goods. Our anal- to exchange a used electric guitar offered ysis thus suggests limited support for H3. through eBay. On the other hand, perhaps To evaluate H4, we estimate a binary because of a lack of data, the model does logit to see if the likelihood of a product not result in an improved prediction of wheth- being exchanged after it is listed in an auc- er a sale will occur. We believe that other tion is affected by the corresponding score factors such as the Buy It Now price and/ or

the Initial Bid may have an effect on the like- internet increases, we believe that the infl u- lihood of a sale occurring; however, limitations ence of such agoras will continue to increase. of our data set prevent us from investigating We now consider the various dimensions those possibilities here. Accordingly, there is that explain how (and why) online agoras limited support for H4. We believe this is an infl uence markets. issue in need of further study. Sources of Information in Online Agoras DISCUSSION In observing HC and the Reverend forums, The impact of the online music community we note that content-creating participants con- on the marketplace for new and used instru- tributed information in at least three ways: ments can be quantifi ed. Products that the (a) They add new information, based on per- community endorses can be sold at a lesser sonal observation and objective data (such as discount, relative to their list price, than technical specifi cations) from other sources, those which the community disapproves. to begin or continue a discussion ‘ thread ’ . We note that concerning H3 – the predic- (b) They support, discount or critique existing tion that HC Score would have an effect on information within a thread. the auction price – our fi nding of limited (c) They offer independent critical opinions support (in the end, brand trumps HC Score) (positive/ negative). obscures the ‘ equalizer’ effect. At the time we collected data for this study, Reverend’ s Sources of information and products cited reputation was largely dependent on HC rat- in user forums, besides personal observation, ings, given the company’ s limited advertising include: online retailers such as Musicians- and restricted retail presence relative to the friend.com and music123.com, the mega- other brands in our sample. In addition, the store retail chain GC, enthusiast websites, participants in the company’ s web-based various guitar brand forums, print guitar forum were mainly self-declared Reverend magazines and their product reviews, friends owners and included a number of identifi - who play and own instruments, or observa- able contributors to HC reviews. Without tions gleaned from watching videos (of art- their highly positive reviews, in effect, Rev- ists and equipment demonstrations provided erend did not have a brand. Thus, while by guitar magazines or manufacturers or opinion in the online agora does not neces- record companies or individuals) on youtube. sarily replace other means of building a com and manufacturer product manuals. brand, it may nonetheless be a key vector The key point here is that while each of in creating and maintaining one. the actors in the agora generates content, a What cannot be quantifi ed, at least within major function of the user community is to the framework of this study, is the basis for collect, collate, judge and redistribute the this impact. We argue that the infl uence of information. In fact, the power of the user com- the user-generated content in this agora is munity lies in its ability to provide participants based on the credibility not only of average access to valuable information at a single location. ratings scores, but also of the information In other words, people enter the agora be- provided by individual members (given that cause it is the best place to learn about some- relatively small numbers of HC reviewers thing they care about. can have an impact on prices of products like Reverend’ s), and on the transparency The Infl uence of Expertise among Agora of a given member’ s relation to the informa- Participants tion in terms of expertise and motivation. The participation of agora members is unequal As the penetration and importance of the in terms of frequency, and also in terms of

infl uence. On the Reverend forum and oth- are asked to be explicit about their status as ers, contributors are ranked quantitatively by musicians ( professional, part-time, hobbyist, the number of their postings (at Reverend, etc), how long they have been playing and ranks range from ‘ newbie ’ to ‘senior mem- to list other equipment they own or have bers’ ). They are also ranked, implicitly, by owned. Similar features have been adopted other members of the forum, according to the in summary form in buyer review templates value of their opinions. The value of a given provided by online retailers such as musi- opinion for other members appears directly ciansfriend.com. On the Reverend forum related to a member’ s expertise concerning and The Gear Page, members often refer particular aspects of musicianship and equip- questions to another member they think is ment. This information constitutes a ‘ frame ’ an expert on the matter. In short, equipment through which participants view and consider buyers want to ascertain the expertise of who- the value of other participants’ contributions. ever is recommending products because that in- We note that contrary to the adolescent formation infl uences the weight they allocate to image of guitarists created by pop media a given opinion ( Hovland et al. , 1953 ; Lyons such as MTV, and despite the fact that forum and Henderson, 2005). participation is highest among internet users Infl uence seems related to the following aged 13– 17 years (Riegner, 2007), the most aspects of expertise: infl uential members of the online guitar community appear to be considerably older. (1) The professional experience of the member – Dick Boak, then-director of artist relations in particular, the frequency with which they for the leading acoustic guitar manufacturer use their equipment before a live audience. C.F. Martin, was quoted in 1999 as saying (2) How well the member plays. On the Reverend that ‘ 40 percent of the 2 million active gui- forum, this is shown by posting ‘ clips ’ – record- tarists in the US are over 50’ .19 It is impos- ings of the member playing the guitar. sible to say what fraction of them participates (3) How much equipment the member owns. in online communities. However, older mu- On the Reverend forum and The Gear Page, sicians are clearly present among frequent members frequently post pictures of their content contributors on HC and in the Rev- collections. erend forum. Users of brands like Reverend, (4) Whether or not the member has modifi ed whose HC reviews impact eBay and street his or her own equipment, by changing elec- prices, frequently state that they have been tronic components or its appearance. Here playing for 40 years or more. This is sig- again, photos are frequently posted. nifi cant in light of the fi nding that ‘ opinion leaders in computer-mediated environments Professional or semi-professional experience possess signifi cantly higher levels of enduring seems particularly infl uential. Most guitar involvement … exploratory behaviour and community members are amateur musicians, self-perceived knowledge than non-leaders’ and admire the performing artists. On the ( Lyons and Henderson, 2005). Older musi- Reverend forum, a ‘ newbie’ member gained cians have had more time to be involved instant infl uence owing to the obvious and explore the fi eld than younger ones, and expertise he applied to guitars and amplifi ers they also have more disposable income to and the professional clips he posted of his spend on musical equipment than do ‘ working’ (ie, for money) bands. Similar teenagers. In other words, their infl uence as phenomena appear in other online commu- consumers and as opinion leaders is dispro- nities of purchasers of durable goods. For portionate to their numbers. Recall that example, on amazon.com, a self-declared when posting HC user reviews, contributors ‘ student fi lmmaker ’ who provided an

extremely detailed working review of a behavior. One HC reviewer very accurately budget camcorder was rated ‘ helpful’ by 131 analyzed that Naylor wanted to be seen as out of 134 readers, making him the leading ‘ fast friendly and concerned … because he reviewer of the device on the site.20 Par- wants the right things said about his ticipants in online agoras are not only judg- product, which is only natural ’ . He then ing products, they are also judging each added: ‘ Joe Naylor has a right to promote a other. lower priced product that competes with Access to such expertise is clearly an at- guitars that are more than twice the price ’ . traction for new participants in online agoras. In short, Naylor’ s motivations were under- However, experts also like to encourage the stood and accepted (Hunter and Soberman, engagement of new members (experts derive 2007a). their prestige from the respect that other Transparency of motive is a recurrent members accord to them). Experts are ob- concern for HC contributors, especially served to personally counsel new members where suspicion of boosting may occur. on product choices and advise them on how Authors of glowing reviews often specify to maintain or modify their purchases to aug- that they do not work for or profi t from the ment the satisfaction of other users. Their company whose product they have lauded. motivation is implicit but clear: as the com- Thus, while participants do not insist that munity grows, so does their own prestige. contributors adopt a position of neutrality toward the information they provide, they do expect that the reasons for providing it Transparency and Trust within the Agora will be implicitly or explicitly clear. Member disclosures concerning their exper- It is worth noting that this approach cor- tise and character provide a frame that responds to recent developments in journal- enables other members to evaluate their ism, where the ‘strategic ritual’ of objectivity contributions. Self-interested motivations (Tuchman, 1972) is increasingly being sup- may also be tolerated on condition that they planted by explicitly assumed bias (eg, in the are rendered transparent . An example of this fi lms of Michael Moore), and target audi- appeared in our case study of Reverend ences are shifting from the general public Musical Instruments. Company owner Joe to specifi c communities (Beckett, 2008 and Naylor told us that in the early days of his Hunter et al. , 2008, 2009 ). Here too, the company, he frequently offered counsel in participant demands to know the frame the forums of other manufacturers, espe- within which information is delivered, the cially the Fender Discussion Page, where authority (or expertise) underlying the infor- owners of his instruments began a sub-sec- mation and the character of the individual tion called The Reverend Congregation. delivering it. In the context of our study, Explained Naylor: the outcome is trust, demonstrated by the It ’ s a PR move, obviously. If I can go fact that individuals rely and act on informa- in there and tell a guy what ’ s wrong tion from the agora in making purchase with his guitar, he’ ll like me. I still pop decisions. in from time to time to straighten them out. It ’ s my demographic – 40– 50 years old, people who started out with classic The Dynamics of Community Growth rock. And I like posting. and Infl uence We believe that in markets where online Elsewhere in the online agora, participants agora form, their scope and infl uence will ex- were neither naï ve nor offended by Naylor’ s pand over time, for the following reasons.

Rationale 1 those confl icts were resolved when the new The fi rst is that web-active individuals have arrival left the forum (or at least, ceased post- greater knowledge of the market and its of- ing). In another instance we observed, ferings (not least because many online offer- participants on the alt.fashion forum, visited ings are not visible offl ine) and have an mainly by young women comparing their advantage, in terms of opportunistic buying, cosmetic purchases or ‘ freebies’ , questioned because they know more about which prod- the motives of a new member who consist- ucts are available and at what prices. They ently attacked a particular brand. Once again, are also exposed to more products than non- the result was that the new member left the web-active buyers. This, in turn, reinforces forum, at least under that user name. their expertise, and thus increases their stat- On the one hand, members seek to expand ure within the community. In sum, web- their agoras through making them attractive active individuals become attractive role and essential resources. On the other, they models and information sources for new or seek to exclude destructive infl uences. potential participants. This dynamic is implicit in Figures 2 and 3. CONCLUSION

Rationale 2 Implications for Organizations The number of web-active individuals in the We begin this section with an anecdote: In marketplace tends to grow, not only because 1999, we called the research director of a online access continues to grow, but because major cosmetics fi rm to ask her opinion of the advantages of entering the online agora the hundreds of user comments on her fi rm ’ s are obvious to those who join it. From this products that appeared on the alt.fashion on- perspective, the role and nature of WOM is line forum. She replied, ‘ What is alt.fashion?’ changing. It serves not only to attract or A decade has elapsed and we still encounter advise information seekers concerning par- executives in different industries (from con- ticular products, but to initiate them into sumer electronics and automobiles to fresh online communities where a wider choice foods) who are unaware of the infl uential of information is available. This dynamic is opinions concerning their products that are also captured in Figures 2 and 3. emerging from online agoras. Large organizations may consider the impact of these Rationale 3 agoras to be marginal; indeed, one extensive Members protect their communities, and recent study found that ‘ less than 1 in 10 with it their infl uence. As shown above in purchases is infl uenced by online WOM’ our analysis of HC, a community can take ( Riegner, 2007). However, their market is formal steps to reduce the dangers of poison- nonetheless changing. New companies with ing or boosting, and to preserve its credibil- limited resources but good products can both ity and attractiveness for members, via both survive and prosper, if the online commu- formal and informal measures. Another way nity is willing to be their ‘ equalizer’ . This that individual members, too, seek to protect was the case for Reverend Musical Instru- their agoras is through self-policing. In ments, which was saved from failure by the one case in which we participated, a new existence of an online community that fi rst arrival to the Reverend forum half-jokingly shifted its purchases from a disintegrating accused other members of having joined a dealer network to the manufacturer’ s web- ‘ cult’ , because of their enthusiasm for the site, then actively supported the launch of products. Appalled responses from some a new product line and dealer network participants led to personal confl icts, and ( Hunter and Soberman, 2007a, b, c ).

An infl uential vector of opinion and action online agora. The medium is tailored to is taking shape, and there are signs that it individuals and not organizations. More- will have a growing impact on product over, character and expertise are critical reputations for all fi rms. The impact of on- aspects required for an individual’ s partic- line opinion on the Corporate Social ipation to have impact. Accordingly, one Responsibility reputation of fi rms is already possibility is for an organization to engage well established. Closer to our subject, a through individuals who have both the leading global toy fi rm, Lego, was able to profi le and the standing to participate. 22 reverse the steady decline of its sales partly An example was furnished by J& J in a by engaging the relevant online agora in the 2008 controversy over an advertisement development and promotion of new prod- that offended viewers, and instantly be- ucts (Schulz and Hatch, 2003) .21 (Reverend, came the object of a hostile Twitter cam- too, used HC reviews to understand the paign. Kathy Widmer engaged a leader of needs and biases of users, in creating its suc- the campaign in terms that were at once cessful new product line.) Conversely, the professional and personal: ‘ I am the Vice growth of web-active consumers in activities President of Marketing for McNeil Con- that were formerly the exclusive province sumer Healthcare. I have responsibility of fi rms (such as marketing and develop- for the Motrin Brand, and am respond- ment), coupled with their capacity to change ing to concerns about recent advertising loyalties and support new entrants, has led on our website. I am, myself, a mom of some observers to ask whether customers are 3 daughters’ . She followed with detailed ‘ partners or competitors … in the generation information about reparatory measures of economic value’ (Schultze et al. , 2007). and an acknowledgment of error on the Clearly, web-active consumers confer com- fi rm ’ s part. While aspects of her response petitive advantage on fi rms that attract their have been critiqued;23 the fundamental approval and engagement. We believe that approach is supported by our observation their importance in the overall mix of factors of the online agora. that affect the market performance of differ- 2. Further, organizations must be prepared ent brands is growing and will continue to to respond to information that falls far grow. We also believe that this importance outside the criteria to which they have will be particularly pronounced for new become accustomed through classic fi rms with limited resources, for whom rep- media relations work. While members utation-building investments in advertising of the online agora possess deep exper- and pricing would otherwise be ‘ an almost tise in given areas, they may also express impossible target’ ( Petkova et al. , 2008). opinions or biases that might be considered defamatory or outrageous in media Effective Strategies to Capitalize on the that conform to journalistic or merely Change polite conventions of discourse. Media How can organizations, large or small, par- professionals within organizations will ticipate in this movement? Further research fi nd their previous experience useful, but will be needed, beyond the best practice hardly suffi cient, in this world. They will examples cited above, to conceptualize and not start from scratch, but they will have validate such strategies. However, several plenty of things to learn, which can only factors are already clear. be learned through online participation. 3. Most important, online communities 1. First, organizations per se may face diffi - accelerate the rate at which informa- culties when they seek to participate in an tion pertaining to the performance and

quality of products becomes broadly 5 Anon. ‘ The Top 200 ’ . The Music Trades , August known. This implies that fi rms that con- 2006. This publication indicates a larger market sider the option of ‘ winging it’ for a year for instruments than does NAMM, the industry trade association. Both indicate a US market for or two with sub-standard products can no all musical instruments and gear of over $ 7 bn. longer regard it as a viable strategy. As we 6 The original acronym stood for National Associa- undertook this study, a major manufac- tion of Music Merchandisers. This US organiza- turer of musical gear, Peavey, faced just tion has become increasingly international and no such a crisis when product failures in a longer posts its original name anywhere on its website or publications. key new product line led to angry com- 7 While China accounts for the largest share of gui- plaints in numerous online forums. The tar manufacturing, the quality of Korean and fi rm engaged its director of product de- Japanese-built products is generally considered by velopment in those forums and swiftly consumers to be superior. resolved the issues. The fi rm ’ s response 8 See footnote 4. demonstrated awareness of a crucial new 9 See http://reviews.harmony-central.com/. In the spring of 2010, following this study, HC revised factor: Any individual model offered by a its user review database and rating systems; how- company can have an adverse effect on a fi rm’ s ever, reviews and ratings referred to in this article reputation due to the reach, credibility and per- are still catalogued at the cited URL under the sistence associated with online ratings. heading ‘ Harmony Central 1.0 Legacy Reviews’ . 10 We have identifi ed similar rating systems in other HC demonstrates that when a credible online markets, such as used cars and video cameras. The agora exists, a new entrant that establishes a HC database is particularly interesting from a research standpoint because it has no real competition reputation for quality can compete with fi rms in terms of authority and comprehensiveness, and it whose marketing resources are far more is largely independent of manufacturer infl uence. extensive. HC may be a unique agora in the 11 This review and Ray’ s original can be viewed at http:// sense that it offers unmatched resources to pur- reviews.harmony-central.com/reviews/Guitar/ chasers of musical equipment. However, it is product/Reverend/SlingShot+Custom/50/1. not unique in the online world. Many 12 HC states these rules on its review fi ling form at http://reviews.harmony-central.com/user_reviews/ industries offering products to a large conform/Guitar/23222. sumer base have online agoras that may infl u- 13 For this reason, country of origin cannot be used ence purchase decisions. Eventually, some of as an independent variable in our empirical these agoras will acquire the scope and credibil- analysis since it cannot be distinguished from an ity to impact their markets in a manner analo- Ibanez brand effect. gous to HC’ s effect on the market for electric 14 The column on eBay containing current bid shows the selling price once the auction is completed. guitars. Dealing with those agoras will be a If the auction does not result in a sale, then the necessity for any manufacturer whose products column is blank (indicating no sale). attract the interest of their participants. 15 The features ‘ country of origin’ and ‘ controls’ are not included as they are perfectly correlated with NOTES the brands Ibanez and Reverend, respectively. 1 This study demonstrates that user book reviews 16 The upper critical value for the F -statistic found collected on book-selling websites affect sales of in the Engineering Statistics Handbook for the com- books on each site. parison is F0.05 (4.97) = 2.485. 2 In our study, we analyze harmony-central.com, or ‘ HC ’ , 17 Auction information from eBay has been used an online agora for amateur and professional musi- in a number of studies to understand the distribu- cians. The HC website also provides a comprehensive tion of prices and potential bidders (see Adams source of user ratings for musical equipment. (2007) for example). Our approach, however, 3 http://reverendguitars.com/forum/forum_posts entails treating the eBay winning bid as a depend- .asp?TID=3069& PN=1. ent variable. 4 See NAMM, ‘ 2006 Music USA, NAMM Global 18 When brand effects are included in the model, the Report’ . adjusted R -square is lower than for the model

with HCS alone; moreover, all but one of the Hunter , M . L . and Soberman , D .A . ( 2007c ) ‘ Reverend coeffi cients is insignifi cant. musical instruments: Playing a different tune (C) ’ , 19 Hunter (2000). INSEAD, 2007 . 20 See http://www.amazon.com/Sony-DCR-SR47- Hunter , M . L . , Le Menestrel , M . and De Bettignies , Handycam%C2%AE-Camcorder-Silver/product- H .-C . ( 2008 ) ‘ Beyond Control: Crisis strategies and reviews/B001P3O3OK/ref=dp_top_cm_cr_acr stakeholder media in the Danone boycott of 2001 ’ , _txt?ie=UTF8& showViewpoints=1 , accessed Corporate Reputation Review , 11 (4) , 335 – 350 . Hunter , M .L . , Van Wassenhove , L .N . and Besiou , M . 2 January, 2010. ( 2009 ) ‘ Stakeholder media: The Trojan horse of 21 For an insightful interview with a Lego employee corporate responsibility ’, INSEAD Working Paper . who is a member of this community, see http:// Klein , B . and Leffl er , K .-B . ( 1981 ) ‘ The role of mar- redcouch.typepad.com/weblog/2008/06/1- ket forces in ensuring contractual performance ’ , what-years-we.html. Journal of Political Economy , 89 (4) , 615 – 641 . 22 ibid. Leland , H . E . ( 1979 ) ‘ Quacks, lemons, and licensing: 23 The incident is recounted by Dan Keeney, A theory of minimum quality standards ’ , Journal of ‘ Johnson & Johnson Redefi nes Crisis Response Political Economy , 87 , 1328 – 1346 . Again’ , at http://www.dpkpr.com/en/art/383/ , Li , S . , Srinivasan , K . and Sun , B . ( 2009) ‘ Internet accessed 28 April, 2010. auction features as quality signals ’ , Journal of Market- ing , 73 (1) , 75 – 92 . REFERENCES Lyons , B . and Henderson , K . ( 2005 ) ‘ Opinion leader- Adams , C . P . ( 2007 ) ‘ Estimating demand from eBay ship in a computer-mediated environment ’ , Journal prices ’ , International Journal of Industrial Organization , of Consumer Behaviour , 4 (5) , 319 – 329 . 25 , 1213 – 1232 . Petkova , A .P . , Rindova , V .P . and Gupta , A . ( 2008 ) Akerlof , G . A . ( 1970 ) ‘ The market for “ lemons” : ‘ How can new ventures build reputation? An ex- Quality uncertainty and the market mechanism ’ , ploratory study ’ , Corporate Reputation Review , 11 (4) , The Quarterly Journal of Economics , 84 (3) , 488 – 500 . 320 – 334 . Beckett , C . ( 2008 ) SuperMedia: Saving Journalism So It Can Save the World , Blackwell Publishing, London , Riegner , C . ( 2007 ) ‘ Word of mouth on the web: The pp. 52 – 53 . impact of Web 2.0 on consumer purchase decisions ’ , Carlton , D . W. and Perloff , J . M . ( 2000 ) Modern Journal of Advertising Research , 47 (4) , 436 – 447 . Industrial Organization , 3e edn., Addison– Wesley, Salop , S . and Stiglitz , J . E . ( 1977 ) ‘ Bargains and Reading, MA, Chapter 13, pp. 420 – 447 . ripoffs: A model of monopolistically competitive price Chevalier , J . and Mayzlin , D . ( 2006 ) ‘ The effect of dispersion ’ , Review of Economic Studies , 44 (138) , word of mouth on sales: Online book reviews ’ , 493 – 510 . Journal of Marketing Research , 43 (3) , 345 – 354 . Salop , S . and Stiglitz , J . E . ( 1982 ) ‘ The theory of sales: Dellarocas , C . ( 2006 ) ‘ Strategic manipulation of inter- A simple model of equilibrium price dispersion with net opinion forums: Implications for consumers and identical agents ’ , American Economic Review , 72 (5) , fi rms ’ , Management Science , 52 (10) , 1577 – 1593 . 1121 – 1130 . Hill , S . , Provost , F . and Volinsky , C . ( 2006 ) ‘ Net- Schulz , M . and Hatch , M .J . ( 2003 ) ‘ The cycles of work-based marketing: Identifying likely adopters corporate branding: The case of the Lego company ’ , via consumer networks ’ , Statistical Science , 21 (2) , California Management Review , 46 (1) , 6 – 26 . 256 – 276 . Schultze , U . , Prandelli , E . , Salonen , P .I . and Van Hovland , C .I . , Janis , I . L . and Kelley , H . H . ( 1953 ) Alstyne , M . ( 2007 ) ‘ Internet-enabled co-produc- Communication and Persuasion , Yale University Press, tion: Partnering or competing with customers? ’ New Haven . Communications of the Association for Information Sys- Hunag , P. , Lurie , N. H . and Mitra , S. ( 2009 ) ‘ Searching tems , 19 , 294 – 324 . for experience on the web: An empirical examination Spence , M . ( 1977 ) ‘ Consumer misperceptions, of consumer behavior for search and experience product failure and producer liability ’ , Review of goods ’, Journal of Marketing , 73 (2) , 55 – 69 . Economic Studies , 44 (3) , 561 – 572 . Hunter , M . ( 2000 ) ‘ Rock around the clock: Wood- Sullivan , E .A . ( 2008 ) ‘ Consider your source ’ , stock’s children strap on their guitars ’ , Modern Marketing News , 42 (3) , 16 – 19 . Maturity , November– December 2000 . Thompson , S .A . and Sinha , R .K . ( 2008 ) ‘ Brand Hunter , M. L. and Soberman , D. A. ( 2007a ) ‘ Reverend communities and new product adoption: The infl u- Musical Instruments: Playing a Different Tune (A ) ’, ence and limits of oppositional loyalty ’ , Journal of INSEAD published case, Fontainefl eau, France, 2007 . Marketing , 72 (6) , 65 – 80 . Hunter , M . L . and Soberman , D . A . ( 2007b ) ‘ Reverend Tuchman , G . ( 1972 ) ‘ Objectivity as strategic ritual: musical instruments: Playing a different tune (B) ’ , An examination of newsmen’s notions of objectiv- INSEAD, 2007 . ity ’ , American Journal of Sociology , 77 (4) , 660 – 679 .

model Bridge Fixed Fixed Depends on the Fixed Fixed Fixed Fixed Fixed Fixed Fixed Fixed Fixed Fixed Fixed Fixed Fixed Fixed Tremelo Tremelo Tremelo Fixed Fixed Fixed

Volume and Volume Controls Tone Proprietary Proprietary Proprietary Proprietary Standard Standard Standard Standard Standard Standard Standard Standard Standard Standard Proprietary Proprietary Proprietary Standard Standard Standard Standard Standard Standard

War m War m War m War m War m Bright War m War m War m War m Wood-Type War m War m War m War m War m Bright Bright Bright Bright Bright Bright War m

Humbucker Humbucker Single Coil Humbucker Single Coil Humbucker Single Coil Humbucker Humbucker Humbucker Humbucker Pickups Humbucker Humbucker Humbucker Humbucker Humbucker Humbucker Single Coil Single Coil Single Coil Humbucker Humbucker Humbucker Single Coil

Bolt-on Set Neck Bolt-on Bolt-on Set Neck Bolt-on Set Neck Set Neck Set Neck Set Neck Set Neck (Gibson) or Bolt-on (Fender) Set Neck Set Neck Set Neck Set Neck Set Neck Bolt-on Bolt-on Bolt-on Set Neck Set Neck Set Neck Set Neck

Korea Korea Korea Korea Korea Korea Korea Korea Korea Korea Korea Country of manufacture Korea Korea Korea Korea Korea Korea Korea Korea China China China Korea Korea

Representative selection of data on brand models within the sample

: Table A1 L Tribute L Tribute Legacy L ASAT Classic L ASAT Special & & & Reverend Charger Reverend Warhawk Reverend Flatroc Reverend Club King Reverend Dillion DR 500 Dillion DTT72 Dillion 600 Standard Epiphone Les Paul GB Standard Epiphone Les Paul PRS Santana PRS Tremonti PRS Tremonti PRS SE Custom W164 Washburn W166 Washburn W167 Washburn G G G Ibanez Artcore Ibanez AS73 Ibanez AS83 Appendix A See A1 Table model group Brand PRS SE Soap

© 2010 Macmillan Publishers Ltd. 1363-3589 Vol. 13, 4, 225–247 Corporate Reputation Review 247 Copyright of Corporate Reputation Review is the property of Palgrave Macmillan Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.