Food

The effects of product photographs and reputation systems on consumer behavior and product cost on eBay

Description
The effects of product photographs and reputation systems on consumer behavior and product cost on eBay
Categories
Published
of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  The effects of product photographs and reputation systems on consumerbehavior and product cost on eBay Brandon Van Der Heide ⇑ , Benjamin K. Johnson, Mao H. Vang School of Communication, Ohio State University, Columbus, OH 43210, United States a r t i c l e i n f o  Article history:Keywords: Computer-mediated communicationConsumer behaviorOnline commerceWarranting theory a b s t r a c t For years, computer-mediated communication (CMC) research has explored and theorized about theeffects of technology on the process of interpersonal impression formation. However, as the Internethas evolved to allow users to accomplish more and more day-to-day tasks (e. g., the buying and sellingof goods and services) little research and theory development has explored how non-interpersonalimpressions form on the internet. This work seeks to extend theoretical perspectives on online signaling(the warranting effect;Walther & Parks, 2002) to predict consumer behavior on the popular online auc-tion website, eBay.com. A content analysis of 217 completed eBay auctions revealed that auctions thatfeatured higher seller reputation scores and actual product photographs (vs. stock photographs) gener-ated more bidding interest and resulted in higher final sales prices. These findings as well as future the-oretical development in this area are discussed.Published by Elsevier Ltd. 1. Introduction Increasingly, the Internet has become a venue for informationgathering, and is frequently the context where individuals makeimportant decisions in their day-to-day lives. The magnitude of consumer decisions made through online transactions is seen inthe over $169 billion spent on online retail products in 2010according to the latest census bureau reports (U.S. Census Bureau).Purchasing decisions typically require consumers to evaluate avariety of attributes of a good in order to ensure they find the bestpossible value for their money and in a time when many of soci-ety’s face-to-face interactions are available in an online venue, vir-tual purchasing decisions are no exception. The present studyexamines the display of products that are advertised and sold on-line. Specifically, it analyzes the advertised product’s visual repre-sentation or photo, and the significance that photo type has on thefinal purchasing decision. With a theoretical grounding in signalingtheory and warranting theory, this study aims to evaluate the im-pact that a product’s main photo has on the purchasing potential of auction items on the Internet auction site, eBay.com (eBay).While much research has investigated the influence of cues ininterpersonal impression formation online such as the impact of others’ comments (Walther, Van Der Heide, Hamel, & Shulman,2009), the type of photo provided (Van Der Heide, D’Angelo, &Schumaker, 2012; Walther, Slovacek, & Tidwell, 2001), the quan-tity of information one self-discloses (Lampe, Ellison, & Steinfield,2007), and the number of ‘friends’ one accepts on a social network-ing site (Donath & boyd, 2004), less research has explored the sameimpression formation processes as they apply to inanimate objectssuch as consumer goods in online settings. Forming impressionsabout products online requires unique evaluative skills due tothe different types and amount of cues available with which tobase an evaluation (Walther & Parks, 2002).The impression formation process of online products can followsimilar patterns as the impression formation process related topeople. The impact of a seller’s feedback score (Resnick, Zeckhaus-er, Swanson, & Lockwood, 2005) and the composition of a givenphoto (Doleac & Stein, 2010) have both been shown to make an im-pact on the decision to purchase a product online. Research hasinvestigated the impact of increased uncertainty in purchasingproducts online and has created a foundation for investigatingthe specific traits that lead to online purchasing decisions (Li, Srin-ivasan, & Sun, 2009). Knowledge of these decision making pro-cesses not only provide guidance for an online seller, but alsoinform future theoretical steps toward understanding the impactof online interactions on the basic evaluative processes of anindividual.The fundamental difference between purchasing decisions on-line and offline is that when purchasing happens online, the phys-ical product is unavailable to be assessed by potential purchasers.Donath (2008)acknowledges that the face-to-face cues that aid inthe evaluation of other people—or in this case, products—is not di-rectly observable over the Web. One must therefore rely on othercues that might signal the presence of qualities they value. Previ-ous research on the influence a photo has on impression formation 0747-5632/$ - see front matter Published by Elsevier Ltd.http://dx.doi.org/10.1016/j.chb.2012.11.002 ⇑ Corresponding author. Tel.: +1 614 292 4863. E-mail address: van-der-heide.1@osu.edu(B. Van Der Heide).Computers in Human Behavior 29 (2013) 570–576 Contents lists available atSciVerse ScienceDirect Computers in Human Behavior journal homepage:www.elsevier.com/locate/comphumbeh  leads us to posit that the type of photo chosen by an individual torepresent himself or herself plays a significant role in the onlinedecision making process (Van Der Heide et al., 2012; Waltheret al., 2009). Signaling theory would support the idea that thesepictures act as ‘signals’ that could contribute substantially to theevaluation of a product.Warranting theory (Walther & Parks, 2002) emphasizes theimportance of recognizing the potential discrepancies between on-line and offline information about a person or a product. Based onStone’s earlier work (1995), Walther and Parks posit that a contin-uum between the cyber-product and the physical-product exists,and displaying the product in an online setting allows for the pre-sentation to fall anywhere along that continuum. Warranting the-ory suggests that the value of information is derived from theperson’s perception about the extent to which that a piece of infor-mation is immune to manipulation by the presenter. Until now,warranting theory has primarily addressed the process by whichobservers judge a target’s personal characteristics and make infer-ences about that person’s attributes. This research seeks to extendwarranting theory by suggesting that its predictions may also holdtrue for howan observer assesses the real-world accuracyof an un-known merchant’s online product. For the present study, this leadsus to believe that for an eBay auction item, a personal photo of thephysical product will be seen as more valuable than a generic stockphoto, which holds no real warranting value concerning the cur-rent physical condition of the product itself.In this paper we will assess these theoretical perspectives fromsignaling and warranting theory and discuss their significance onexplaining online consumer behavior. We also evaluate the impactof economic returns in these purchasing decisions. A content anal-ysis examined the influence of several cues from auction itemsposted on eBay, and focuses specifically on the weight of a photoon the product’s selling potential. 1.1. Signaling theories Signaling theory describes the process by which an individual’shidden or unobservable characteristics, which an observer uses toconstruct meaning, are revealed through indirect acts performedby the individual (Donath, 2008, p. 233). This concept of signalinghas srcins in biology and economics. In offering an explanation fororganisms’ natural selection and retention of characteristics thatboth serve to attract mates but are still counterproductive to sur-vival (e.g. deer antlers, peacock plumes),Zahavi (1975)argued thatthese ‘‘handicap’’ markers demonstrate underlying qualities andadvertise that an individual passes a resource test of some relevantgenetic characteristic. The presence of a handicap in a marker re-moves the possibility of ‘‘bluff.’’ That is, Zahavi illustrates that itis very costly, if not impossible, for a male peacock, for example,to ‘‘fake’’ cues that would attract a female peacock. His tail feathersclearly signal his suitability as a mate. While the male peacock’sfertility is not directly observable, his tail feathers are (a) directlyobservable and (b) a clear indicator of balanced nutrition; his‘‘handicap’’ marker of fertility. These ‘‘unfakeable’’ characteristicsalso have human corollaries. For instance, in economics,Spence(1973)made a distinction between indices, ‘‘observable, unalter-able attributes’’ (p. 357), and signals, attributes the individualcan control and present. Signals involve costs which a person mustevaluate, and Spence asserts that people will choose signals bycomparing the costs and benefits of possible signals. The recipient,on the other hand, has an interest in interpreting the signal and thefidelity of its relationship to the underlying characteristic it is sup-posed to indicate.In a distinction similar to Spence’s indices and signals,Donath(2008)drew fromMaynard-Smith and Harper (2003)to character- ize two components of signaling behavior in social media. Anassessment signal is a sign that, by its very nature, demonstratesthat the person possesses a particular attribute which results in areliable signal. Assessment signals include handicap signals, whichdemonstrate a wealth of a given resource by expending some of that resource for signaling. The second category of signals is con-ventional signals, socially agreed-upon symbols or expressions of an attribute. These can include self-description of specific attri-butes, such as one’s age or gender. Because the presentation of these signals is not necessarily dependent upon the attribute itself,deception becomes possible. According to Donath, individuals whowish to deceive must weigh the costs and benefits of deceptive sig-nals, especially in the presence of constraints that restrict the abil-ity to lie about a particular conventional signal. Social networksand other contextual placement of the individual can then serveas constraints on this deceitful self-presentation.Signaling theory has also been applied to online auction envi-ronments.Li et al. (2009)investigated the use of three categoriesof possible signals used by sellers: direct quality indicators, indi-rect quality indicators, and seller credibility indicators. Their worksuggests that direct indicators (e.g. multiple pictures, product cer-tification) encouraged bidding activity, but decreased bid amountsas bidders sought to avoid overpaying in a ‘‘winner’s curse’’. Theyalso found indirect indicators (e.g. minimum bid, ‘‘Buy It Now’’ op-tion) discouraged bidding activity, but increased bid amounts, andas a third type of indicator, seller credibility indicators (e.g. feed-back scores) encouraged bid activity, decreased bid amounts, andmoderated product quality indicators. They also found that allthree forms of signaling encouraged earlier bidding. Direct qualityindicators helped bidders avoid the ‘‘lemon problem’’ of poor qual-ity goods, while indirect indicators resulted in more serious andfewer casual bids in an auction. These signal types impacted bidder judgments about ‘‘whether to participate, who bids, when to bid,and how much to bid’’ (p. 76), and, ultimately, the outcome of anauction. 1.1.1. The warranting effect  While signaling theories have proven to be productive in termsof generating findings that support the idea that signals induceparticular consumer behaviors, a signaling approach stops shortof specifying the specific relationship that assessment signals andconventional signals have on this behavior. The warranting princi-ple (Walther & Parks, 2002) offers a theoretical account of howusers of technological systems may form judgments from a varietyof cues that they encounter in online spaces. Specifically, the war-ranting principle acknowledges (as doesDonath, 2008) that mostonline cues are conventional signals and, as such, can be ‘‘faked’’.This perspective also provides a clear set of expectations abouthow observers judge whether online claims match an offlinereality.The warranting principle (Walther & Parks, 2002) proposes thatobservers tend to have a stronger belief that information is honestand truthful if the personal information is difficult or costly tofeign. In comparison to face-to-face communication, CMC offersthe presenter a greater degree of control over the informationbeing relayed. With this kind of control, the warranting value of information, which is based on the perceived accuracy of onlinepresentation in regards to offline actuality, becomes an importantcue to any perceiver. In face-to-face communication, the impres-sions’ warranting value is not generally questioned, as it is difficultto pervert, but due to the greater degree of control afforded byCMC, some interpersonal information provided may be suspect.The warranting value of information has been supported inexperimental studies of the social network site Facebook (Waltheret al., 2009). Consistent withParks and Adelman’s (1983)finding that uncertainty about one’s relational partner can be significantlyreduced when contacting members of the partner’s social network, B. Van Der Heide et al./Computers in Human Behavior 29 (2013) 570–576  571  Walther et al. found that observers of Facebook profiles judged tar-gets’ physical attractiveness on the basis of cues with high war-ranting value. Specifically, information from a third party source(in the form of a Facebook wall post) about the target’s physicalattractiveness had more warranting value than self-disclosuresfrom the target in question. As predicted by the warranting princi-ple, interpersonal information that was difficult to manipulate bythe target drove interpersonal judgments more strongly than didinformation that was more easily manipulable.Although the warranting principle has been used to predictinterpersonal judgments, it seems reasonable that the principlecould be applied to judgments of products in online commerce, aswell. There are at least two classes of cues available to consumersinonline commerce websites aboutwhich the warrantingprinciple(Walther & Parks, 2002) would make predictions. The first type of cue is product photographs, which are a frequently utilized cue ina variety of online commerce websites including the website of interest for this study, eBay. Specifically, online sellers have theopportunity to choose the pictorial presentation of the product inan online auction, and there tends to be substantial varianceregarding the type of photograph chosen by an online seller. For in-stance,thesellermaychoosetopresentaphotographthatissimplyastockphotoofanitem,orthesellermaypresentanindividualizedphotographofthephysicalitem.Theformerphotographaffordstheseller the opportunity to present an item for sale very positively;however, this type of presentation is of relatively low warrantingvalue, as any seller with access to the Internet could conceivablypost the same stock photo. Such an image would be considered avery low-cost cue. The latter depiction should have greater war-ranting value, as posting an individualized photograph is harderto ‘‘fake’’ and would require actually having an item of the sortone was trying to sell, taking a picture of the item, and uploadingthat photo to the online auction website, a procedure that a per-ceiver would be likely to judge to be a more costly endeavor. Con-sistent withLi et al.’s (2009)findings, actual product photos,compared with stock photographs of a product, should improveconsumer confidence in an online auction and increase the likeli-hoodofanauction’ssuccess.Accordinglythishypothesisisposited: H1. Compared to auctions with stock photographs of items forsale, auctions with actual product photographs are more likely to(a) sell, (b) generate more bids, and (c) sell for higher final biddingprices. 1.1.2. Reputation systems The second type of cues about which the warranting principlemakes predictions are the system-generated (i.e., third-party) rep-utation systems present for online auctions.Walther et al. (2009)argued that one instantiation of the warranting principle suggeststhat information from a third-party, because it is difficult or costlyto fake, has a greater warranting value than first-person informa-tion. System-generated cues were found to be influential on inter-personal impressions through the number of friends that Facebooklists on a user’s profile (Kleck, Reese, Behnken, & Sundar, 2007)Tong, Van Der Heide, Langwell, & Walther, 2008; Utz, 2010). Oneparticularly salient example of this instance of the warrantingprinciple occurs in the reputation system portion of the online auc-tion website, eBay. Specifically, the warranting principle suggeststhat third-party assessments of a seller, such as the feedback scoreand percentage of positive feedback, have high-warrant. Theseareas are controlled by the auction site and affected by the feed-back from past buyer–seller transactions on eBay. System gener-ated reputation cues are extraordinarily hard to fake and should,therefore, have high warranting value. Consistent with that expec-tation, previous research has shown that reputation systems sub-stantially influence auction outcomes on eBay (Houser &Wooders, 2006; Resnick, Zeckhauser, Swanson, & Lockwood, 2006).Trust is a key element in online marketplaces due to the uncer-tainty in negotiating with unseen sellers and their wares. On siteslike eBay, reputation systems indicate the possible risks involvedfor potential buyers. AlthoughResnick and Sami (2008)point outthat reputation systems include flaws in calculating the overallscore of buyer evaluation, these systems are premised on the belief thattheycontainreliablemeasuresofaseller’sintegrity;andwouldencourage the buyer to trust the seller to a greater extent.Zachariaand Maes (2000)specifically identify two ways in which trust of anonline seller is important. The first is that buyers do not get to viewthe physical object they bid on; users need to trust that the remoteseller actually possesses the item for sale. Due to the nature of thevirtual market, a customer also takes it on faith that the seller didnot ‘‘misrepresent the condition or the quality of their products’’(p.882).Thesecondreasontrustisimportantisthatbothpartiesde-pend on the other party to honor the sales contract through to theend of the auction. As eBay closes millions of transactions monthly(Resnick & Zeckhauser, 2002), the validity of the seller’s merchan-dise, the bidder’s intent to buy, and the seller’s intent to sell mustbe evident, otherwise online auctions would not be effective.Previous research has shown that reputation systems can havean economic sway on online auction outcomes. A field study of eBay auctions conducted by Resnick et al. (2001) suggested thatitems sold by individuals with a higher reputation score were morelikely to sell items than those with a lower score. Other researchhas revealed that a seller’s reputation had a statistically significanteffect on the final sales price of an item (Houser & Wooders, 2006).Houser and Wooders conducted a content analysis on eBay itemsand found that a higher seller reputation resulted in significantlyhigher final auction prices of products. Above and beyond any of the other variables coded by their team, the reputation of the sellerwas the strongest predictor of a higher final bid of a product; sug-gesting that third party reputation systems might instill enoughtrust in a buyer to encourage higher bid offers.Consistent with the warranting principle (Walther & Parks,2002; Walther et al., 2009), a difficult-to-manipulate third-partyaccount of a seller’s reputation, as reported by eBay’s reputationsystem, should be more likely to engender trust in an online seller.This should, in turn, result in positive economic benefits for onlinesellers with positive reputations. Indeed, this proposition is consis-tent with previous research (Houser & Wooders, 2006; Resnicket al., 2006). This work seeks to replicate these findings regardingthe impact of reputation systems on consumer behavior in onlineauctions and proposes the following prediction: H2. There is a positive relationship between a seller’s reputationand consumer behavior on eBay such that auctions posted bysellers whose reputation is more positive are more likely to (a) sell,(b) generate more bids, and (c) sell for higher final bidding pricesthan auctions posted by sellers whose reputation is less positive.Moreover, the present research expects that both a seller’s rep-utation and the type of product photograph presented should haverobust effects. That is, this research proposes that even when con-trolling for a number of other auction variables and each other,both photo type and seller reputation should account for a statisti-cally significant amount of the variance in consumer biddingbehavior in online auctions. Accordingly, the following hypothesisreflects that prediction: H3. After controlling for product listing price, shipping cost, andlisted item quality, seller reputation and photo type are positivelyrelated to (a) likelihood to sell, (b) number of bids generated, and(c) final sales price. 572 B. Van Der Heide et al./Computers in Human Behavior 29 (2013) 570–576   2. Method  2.1. Procedure To evaluate the influence of cues on a buyers’ purchasing deci-sion process a content analysis of completed eBay auctions wasconducted. The sample of completed auctions was collected duringthe week of October 7 through October 14 of 2010. The sample in-cluded only completed sales, in which eBay provides the productand purchasing information necessary for evaluating the hypothe-ses proposed. Completed sales retained the format and informationincluded on the webpage as if the product was still for sale, yet alsooffer additional information concerning the bidding history as wellas the final outcome of the auction. Individual coders were in-structed to select auctions that completed at various times of theday in orderto gather the mostdiverse sampleof auctionspossible.Data was collected from auctions of four products: the book‘‘Freedom’’ by Jonathan Franzen, the productivity software suiteMicrosoft Office Home and Student Edition 2007, the sports videogame Madden 2011 for the Xbox 360 home video game console,and the first season DVD set for the popular animated sitcom‘‘The Simpsons’’. The four search terms used to locate auctionsfor these items in this analysis were freedom jonathan franzen,microsoft office home and student 2007, madden 11 xbox 360,and simpsons season 1. The decision to include these items wasinfluenced by the high degree of consistency between the searchterm entered into the eBay search engine and the types of auctionsyielded. For example, there is low variability in the types of resultsproduced when one uses the search term, simpsons season 1. Gen-erally speaking this search term resulted in products that wereconsistent with the intended result from the search term. Thiswas necessary because some search terms produced a large num-ber of items that did not pertain to the particular item of study. Asanother example, the search term, ‘‘ microsoft office home and stu-dent 2007  ,’’ was also consistent in yielding auctions of installationpackets for Microsoft’s 2007 home and student office software.However, a product inquiry such as ‘‘ Barbie ’’ might contain a BarbieDVD, a collector’s edition Barbie, a Barbie dream house, various‘‘Ken’’ doll accoutrements, as wellas the entiretyof the MalibuBar-bie product line; in short, this creates excessive product variabilitywithin the search results.For each item, although every effort was made to choose prod-ucts with low variability, slight variability still existed within theitems of interest. During training, coders were instructed to onlycode items that fit the description of the product exactly. Forexample, signed copies of the Jonathan Franzen book Freedomwere excluded. Additionally, an accepted practice on eBay, whichallows the posting of auctions as ‘‘Buy It Now’’ only, effectively al-lows a seller to post an item for sale and only allows for a single bidover a set price. Once a buyer is found for that price, the item sellsto that particular buyer. As there was no bidding process and nobidding history to view, this type of auction was excluded fromthe sample. Moreover, in order to reduce the possibility of selec-tion bias, coders were instructed to code all relevant search termmatches that reflected the products described above. The finalsample included a total of 217 auctions across the four searchterms. Most of the relevant auctions conducted during the sam-pling time frame were included in the sample.  2.2. Measures Coded variables were selected to reflect the theoretical hypoth-eses presented in the previous section of this paper. The variableschosen were all suspected to either play some role in the purchas-ing probability of the product or were selected as controls to ac-count for the variety between types of products. Variablesincluded in the analysis include starting price, ending price, num-ber of bids, whether or not the product sold, type of photograph(stock photograph or individual product photograph), seller re-ported condition of the product, percentage of positive feedback,feedback score, and advertised shipping cost. The seller reportedcondition was a state chosen by the seller from a list of possibleconditions provided by eBay. The options included categories rang-ing from ‘‘brand new’’ to ‘‘acceptable’’. The percentage of positivefeedback is simply the ratio of positive feedback to total feedback.The feedback score on the other hand, is a numerical representa-tion of the positive ratings an eBay member has received in whichthe more positive ratings an individual has, the higher their feed-back score. Each buyer and seller has the opportunity to rate theother in terms of how positive an eBay transaction was. They canaward their transaction partner a positive rating (+1), a neutral rat-ing (0), or a negative rating ( À 1). These scores are then accumu-lated over time and result in a feedback score (eBay, 2010). Thenature of these variables did not allow for much variability be-tween coders as the variable items are reported as precise valueson the home page for each completed sale item. The only variablethat did require coder evaluation was the photo type variable.Photo type categories were defined as a stock photo, personalphoto: individual item, personal photo: inventory, or photo notof product for sale. (Appendix Apresents the definitions of thesecategories.)  2.3. Coding  Five trained coders were randomly assigned two products eachand instructed to code approximately 50 auctions under eachsearch term. Intercoder reliability was established by comparingrandomly double-coded auctions overlapped in product assign-ment between the coders. In total, 64 auctions were double codedand assessed for reliability. Due to the extremely objective natureof the variables coded, minimal disagreement among coders wasexpected. Consistent with that expectation, in only one auctiondid coders disagree about how a variable’s value for this one auc-tion was to be assigned. Due to the inconsistency, data from thissingle auction was discarded from the data set altogether. On allother variables and in all other auctions, coders agreed 100% of the time. No additional cases were dropped and all cases includedresponses to all variables of interest. 3. Results  3.1. Hypothesis 1 Hypothesis 1considered the bivariate effects of photo type onconsumer behavior on eBay. Specifically,Hypothesis 1a predictedthat, compared with stock photograph auctions, personal productphotographs would more likely result in a sale. A logistic regres-sion tested this hypothesis. The data suggested that the likelihoodof a sale for all auctions ( N  = 217) featuringan actual productphoto( n = 71; 83.10% sold) was not statistically significantly greater thanauctions featuring a stock photo ( n = 146; 76.71% sold), Wald = 1.157, p = 282. Thus, the data were not consistent withHypothesis 1a.Hypothesis 1b suggested that, compared with auc- tions featuring stock photographs, auctions featuring actual prod-uct photographs would receive a higher number of bids. A point-biserial correlation coefficient was calculated to assess the rela-tionship between these categorical (e.g. photo type) and continu-ous (e.g. number of bids) variables. The data suggested that therewas a significant effect in the predicted direction, r   pb (215) = 0.193, p = .004. Thus, the data were consistent with B. Van Der Heide et al./Computers in Human Behavior 29 (2013) 570–576  573  Hypothesis 1b.Hypothesis 1c predicted that, compared with auc- tions featuring stock photographs, auctions featuring actual prod-uct photographs would generate the highest final sales price.Another point-biserial correlation coefficient was calculated to testthis hypothesis, and the data were consistent with this prediction, r   pb (215) = 0.360, p < .001 (seeTable 1for descriptive statistics andTable 2for a correlation matrix of study variables).  3.2. Hypothesis 2 The second group of hypotheses examined the effects of reputa-tion systems on consumer behavior. It should be mentioned thattwo indices of seller reputation were harvested and analyzed dur-ing the data collection. The first of these indices, the percentage of positive feedback left from previous buyers of the seller’s auctions,suffered from a severe restriction of range. That is, scores for thisvariable were, with very few exceptions, uniformly high. The sec-ond reputation indicator, the feedback score, did not suffer fromthe same range restriction.Hypothesis 2a predicted that there was a positive associationbetween auctions with sellers whose reputation was more estab-lished and the likelihood of a successful auction. Logistic regressionwas used to examine the consistency of the data with this hypoth-esis. The data suggested that auctions whose seller enjoyed a moreestablished reputation were not statistically significantly morelikely to sell than auctions hosted by less established sellers, Wald = 3.565, p = .059, although at a purely descriptive level, thedata did trend in the expected direction. The data were not consis-tent withHypothesis 2a.Hypothesis 2b predicted that auctions with sellers whose reputation was more established would gener-ate more bids. A bivariate correlation showed that the data wereconsistent with this prediction, r  (215) = 0.222, p = .001.Hypothe-sis 2c expected that auctions with sellers whose reputation wasmore established would end with a higher overall sales price. Con-sistent with this expectation, there was a positive association be-tween an auction’s seller reputation and the final sales price of the auction, r  (215) = 0.630, p < .001. Thus, the datawere consistentwith bothHypotheses 2b and c.  3.3. Hypothesis 3 A third group of hypotheses suggested that even after account-ing for various control variables such as the initial price posting of an auction, the seller-reported quality of an item, and shipping costboth seller reputation and product photograph type account forstatistically significantportions of the variancein consumerbehav-ior.Hypothesis 3a considered these variables as predictors of theoverall likelihood of a successful auction by entering them into alogistic regression. The overall model fit was marginal at best,Nagelekerke R 2 = .141. After controlling for the variables listedabove, neither seller reputation (Wald = 0.613, p = .434) nor anauction’s product photograph type (Wald = 2.527, p = .112) werestatistically significant predictors of whether or not an auctionwas likely to be successful. Moreover, the only control variable thatwas a statistically significant predictor of likelihood to sell was theauction starting price, Wald = 5.562, p = .018. Specifically, the rela-tionship between starting bid price and sale likelihood was suchthat items with lower starting bid prices were more likely to sellthan items with higher starting bid prices. Thus, the data werenot consistent withHypothesis 3a.Hypothesis 3b predicted that, after controlling for the initialprice posting of an auction, the seller-reported quality of an item,and product shipping cost, seller reputation and photograph typewould remain significant predictors of the number of bids an auc-tion received. A multiple regression analysis was used to assessthis hypothesis. The overall regression model accounted for a sig-nificant amount of the variance in the number of bids an auctionreceived, Adjusted R 2 = .168, D F  (5, 209) = 11.519, p < .001. Aftercontrolling for the variables in step one, the feedback score andphotograph type explained 9% of the variance in number of bidsan auction received, D R 2 = 0.090, D F  (2, 209) = 7.666, p < .001.Additionally, after controlling for the variables mentioned above,both stronger seller reputation ( b = .135, t  [209] = 2.093, p = .038)and product photograph type (actual product photographs gener-ated more bids than stock photographs; b = .260, t  [209] = 3.953,  p < .001) were significant predictors of the number of bids an auc-tion received; the data were consistent withHypothesis 3b.Hypothesis 3c expected that seller reputation and photographtype would be significant predictors of final item sales price aftercontrolling for the variables mentioned above. The model that re-gressed final sales price upon these variables predicted a statisti-cally significant and substantial amount of the final sales price,Adjusted R 2 = .351, D F  (5, 209) = 24.195, p < .001. The addition of feedback score and photograph type to the second step of the hier-archical model that regressed final sales price upon these variablesexplained an additional 12.1% of the variance in the amount of thefinal sales price, D R 2 = 0.121, D F  (2, 209) = 19.993, p < .001. Aftercontrolling for item quality, shipping cost, and an item’s initial list-ing price, both better seller reputation ( b = .170, t  [209] = 2.966,  Table 1 Means and standard deviations of study variables by photo type. Starting price Ending price Bids Seller reputation Stock photon 146 146 146 146Mean 13.81 29.75 6.64 3498.79St. Dev. 15.29 27.11 8.65 30,080.85  Actual photon 73 73 73 73Mean 20.10 54.48 10.67 6805.83St. Dev. 26.06 37.90 10.47 33,787.77 TotalsN  219 219 219 219Mean 15.82 37.75 7.95 4570.41St. Dev. 19.66 33.13 9.44 31,243.83  Table 2 Correlations among study variables. Starting price Ending price Number of bids Seller reputation Photo typeStarting priceEnding price .47 b Number of bids À .20 b .64 b Seller reputation À .10 À .24 b .34 b Photo type .13 a .33 b .19 b .05Correlations in the photo type row are point-biserial correlations due to the categorical nature of the predictor variable. a Indicates p < .05 (two-tailed). b Indicates p < .01 (two-tailed).574 B. Van Der Heide et al./Computers in Human Behavior 29 (2013) 570–576 
Search
Similar documents
View more...
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks