Electronic Commerce Research and Applications 13 (2014) 98–109

Contents lists available at ScienceDirect

Electronic Commerce Research and Applications

journal homepage: www.elsevier .com/ locate /ecra

An empirical comparison of market efficiency: Electronic marketplacesvs. traditional retail formats

1567-4223/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.elerap.2013.11.003

⇑ Corresponding author. Tel.: +1 (215) 951 1728.E-mail addresses: jiang@lasalle.edu (P. Jiang), sivakbalas@stuart.iit.edu (S.K.

Balasubramanian).1 Tel.: +1 (312) 906 6516.

Pingjun Jiang a,⇑, Siva K. Balasubramanian b,1

a Department of Marketing, School of Business Administration, La Salle University, 1900 West Olney Avenue, Philadelphia, PA 19141, USAb Stuart School of Business, Illinois Institute of Technology, Chicago, IL 60661, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 3 April 2013Received in revised form 8 November 2013Accepted 8 November 2013Available online 28 November 2013

Keywords:Electronic marketsMarket efficiencyData Envelopment AnalysisConsumer behavioral segmentationRetailing

Researchers have found that price dispersion and market inefficiency exists in electronic marketplaces.Little attention has been bestowed to explore difference in market efficiency between traditional andelectronic marketplaces. This study integrates both product and channel preference factors to analyze dif-ferences in market efficiency between electronic and traditional shopping environments. Data Envelop-ment Analysis (DEA) is applied to calculate market efficiency for single-channel and multi-channelshoppers. Results show that market efficiencies vary across consumer segments and products. In sum-mary, this paper enhances understanding of market efficiency by incorporating behavioral segmentand product characteristics into the explanatory framework.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction

The last decade has seen an exponential increase in commercialuse (buying and selling) of Internet. Between 2002 and 2012, retaile-commerce grew at 15–25% per year, following a fairly typicaladoption pattern that transformed more mainstream consumersinto online shoppers. In 2012, around 150 million Americans madeat least one online purchase, and an additional 35 million used theWeb to gather information about products (eMarketer Inc. 2012,National Retail Foundation 2012). By far, marketing activities ac-counted for most of the growth in online traffic. Overall, Internethas evolved into a well-established electronic marketplace. B2C re-tail e-commerce (both actual purchases and purchases influencedby Web shopping but actually bought from a brick-and-mortar astore) is estimated at $1.2 trillion in 2012, or over 40% of total retailsales in the United States (Forrester Research 2011).

Electronic marketplaces have the potential to fundamentallychange how people shop. They can disrupt the structure of wellestablished industries such as retail and consumer goods (Albaet al. 1997). An early, and popular notion among economists pos-ited greater transparency and efficiency in digital markets. Forexample, online shopping was expected to diminish informationasymmetries. The transparency characteristic was expected to in-

clude price, quality, and availability of most product. This grand vi-sion of web-based information and product access was labeled‘‘frictionless commerce’’ and sharply contrasted with the unpre-dictability and opaqueness that often characterize real-world mar-kets (Smith et al. 2000). A related hypothesis predicted thatelectronic markets will reduce consumers’ information searchcosts to produce efficiency gains, while the greater transparencywas expected to increase the quality of information about price,leading to price convergence (Bakos 1997) and better market effi-ciency. However, early studies failed to support this notion of per-fect, friction-free commerce. Several researchers (Bailey 1998,Brynjolfsson and Smith 2000, Erevelles et al. 2001, Lee and Gosain2002, Clemons et al. 2002) found significant price dispersion on theInternet. Baylis and Perloff (2002) cited price discrimination as alikely cause of this unexpected finding. Some pointed to seller dif-ferentiation as a driver of price dispersion (Baye et al. 2004a,b, Panet al. 2003). Others proposed that multichannel retailers havehigher prices than Internet-only retailers (Pan et al. 2002, Ancaraniand Shankar 2004). Finally, market level forces were considered bysome as the chief cause of price dispersion online (Venkatesanet al. 2007, Pan et al. 2009).

While the preceding explanations of price dispersion in theelectronic marketplace are interesting, they also raise a fundamen-tal question: how is consumer welfare impacted if electronic mar-ketplaces dominate the future of retailing? Unfortunately, thisresearch question remains unexplored. There is a lack of researchthat compares market efficiency metrics between traditional andelectronic marketplaces. In addition, no study of price dispersion

P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109 99

or market efficiency has explicitly incorporated behavioral seg-mentation based on consumers’ single or multi-channel shoppingpreference. Furthermore, despite evidence (Walter et al. 2006) thatprice variance attributable to a variable may differ across products,the price dispersion literature has not fully examined differencesacross products. Therefore, the main objectives of our study areas follows:

(1) Integrate both product and channel preference factors toexplain differences in market efficiency when electronicand traditional shopping environments coexist.

(2) Compare market efficiency across online-only shoppers, off-line-only shoppers, and multichannel shoppers; investigateif differences in market efficiency across these three groupsalso vary by product. In particular, how do product factorsinfluence market efficiency when comparing traditionaland electronic marketplaces?

(3) Explore how the electronic marketplace affects the competi-tion structure among brands within an industry, i.e., how theefficiency frontiers for brands change when a new electronicmarketplace option coexists with the traditional retailformat.

Our study contributes to the literature on market efficiency inthe electronic marketplace by adding new knowledge or extendingexisting knowledge. We introduce behavioral segmentation intomarketing efficiency calculations. In addition, we explore effi-ciency changes by consumer segments for different products alonga set of key product characteristics. We also discuss changes in effi-ciency frontiers.

Our study is structured as follows. We begin with a review ofthe literature on market efficiency in the context of price disper-sion in traditional (offline) and electronic (online) marketplaces.We then present an exploratory framework that incorporatesbehavioral segmentation. The next section describes the researchmethod, followed by our analysis. We discuss findings, their man-agerial implications, summarize our conclusions and limitations,and provide guidance for future research.

2. Market efficiency – a literature review

The market efficiency of traditional retail formats has beencalled into question in the price–quality relationship literature.Kamakura et al. (1988) provide a comprehensive overview. Theyexamine the determinants of market efficiency with an applicationof Data Envelopment Analysis (DEA) to multiple product categoriesand conclude that a significant amount of inefficiency exists. A vastliterature that explains inefficiency and price dispersion as an out-come of costly information can be traced back to Stigler’s (1961)‘‘Economics of Information’’ (EOI) theory, which posits that con-sumers will continue search until the marginal expected costequals the marginal expected return. Nagle (1984) provides anexcellent overview of economics literature on pricing, with a par-ticular focus on asymmetric information. An exchange involvesasymmetric information when one party has more informationthan the other party. Many pricing issues are associated withasymmetric information, such as the presence of ‘‘inefficientbrands’’ in a consumer market. Nagle also finds that consumerinformation acquisition is closely related to price elasticity. Thefewer the brands about which buyers are informed, the less sensi-tive they will be to the price of any one brand. The cost of con-sumer information acquisition is dependent on productattributes. As one moves from the search attributes to experienceattributes and onto credence attributes, information about abrand’s differentiating attributes becomes more costly.

A ramification from EOI is that some inefficient brands mightcontinue to exist, because information is costly and thereforeimperfect. By examining the previous price–quality literature,Kamakura et al. attribute market inefficiency and the survival ofinefficient brands to several factors. First, since the actual productspace is not continuous but contains only discrete offerings, andsince consumers cannot buy mixtures of product offerings (espe-cially for durable goods), these gaps may give each brand a monop-oly over those consumers whose equilibrium lie in the vicinity ofthe brand (Rosen 1974). Second, ignorance of available offeringsor of their characteristics may increase consumers willingness topay more than the efficient price (Maynes and Assum 1982, Rosen1974). Finally, consumers’ buying strategies may involve a trade-off between the benefits of finding an efficient brand and the costsinvolved in this search (Pratt et al. 1979, Ratchford 1980, Stigler1961). Overall, since the benefits of searching to find the most effi-cient brand may fail to exceed the costs of doing so, the optimaldecision may be to purchase an inefficient brand—one whose priceis above the minimum for its characteristics.

2.1. Key factors affecting market efficiency of electronic marketplaces

Though many factors influence market efficiency, there arethree factors that are most salient in the market efficiency litera-ture. These factors are: information acquisition, continuous prod-uct offering, and symmetric information. Bakos (1997) showedthat reduced search costs in electronic marketplaces increased all-ocational efficiency and price competition among sellers. As a re-sult, electronic marketplaces reduce buyer search costs toimprove market efficiency. In addition, Internet agents that accessand evaluate online information represent intelligent systems withthe potential to substantially increase market efficiency (Palopoliet al. 2006). Overall, in electronic marketplaces, technology haschanged how these factors drive market efficiency, as summarizedin Table 1.

2.1.1. The impact of Internet agents on improving market efficiencyAside from the positive factors that improve market efficiency,

there are complex problems facing consumers in the electronicmarketplace that have an adverse impact. First, prior to any pur-chase, consumers have to search a large number of websites tocompare brands, products and merchants. This information acqui-sition and decision making process often entails significant timeand search costs. Because of these costs, consumers may find itrather difficult to identify products that satisfy their needs. Fortu-nately, some of this difficulty is mitigated if consumers are assistedby Internet agents (Pivk and Gams 2000; Guttman and Maes 1999;Rosaci 2004, 2005). Hostler et al. (2005) assessed the impact ofInternet agent in a retail online shopping environment.

Palopoli et al. (2006) explored the role of agents in variousphases of the consumer decision-making process. Internet agentsystems, such as intelligent agents, mobile agents, or collaborativeagents, help customers to examine and evaluate product alterna-tives efficiently (Sarwar et al. 2000, Schafer et al. 2001). The contri-bution of Internet agents to market efficiency mainly stems fromassisting and simplifying the interaction between users and com-puters. For instance, a feature-based filtering system may allowconsumers to select products at a website based on featured key-words. Similarly, a collaborative filtering system may recommendproducts based on a consumer’s similarity with other unknownconsumers, as determined through a real time comparison of prod-ucts searched and other information stored in his/her profile. Aconstraint-based filtering system may facilitate the specificationof shopping constraints for the desired product (e.g., price range,delivery time) and return information on only those that meetthese constraints. Finally, price comparison sites, such as shop-

Table 1A comparison of traditional retail format with electronic marketplace.

Factors Traditional retail format Electronic marketplace

Information acquisitionVast selection Costly Consumers may have to ignore available

offerings or their characteristics, and therebypurchase inefficient brands

Costless Consumers need not ignore availableofferings or their characteristics. This helpspurchase of efficient brands

Screening Costly CostlessProduct comparisons (Maynes andAssum 1982, Rosen 1974, Alba et al.1997, Palopoli et al. 2006)

Costly Costless

Product offering (Rosen 1974, Palopoliet al. 2006)

Discrete Consumers may not be able to purchasemixtures of product offerings

Continuous Consumers may purchase mixtures ofproduct offerings

Information availability (Nagle 1984,Palopoli et al. 2006)

Asymmetric Information overload may be a problem Symmetric Information overload not likely

100 P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109

ping.com, pricegrabber.com, nextag.com, compare the price of sev-eral sellers with respect to the same product.

3. Research propositions

According to Nelson (1970), in traditional markets, informationon product quality is often more difficult to obtain than price infor-mation. In sharp contrast, the electronic marketplace (e-com-merce) allows consumers to easily access information on productquality.

Bakos (1997) developed models that considered search costsand market efficiency in a differentiated market with heteroge-neous buyer tastes and seller product offerings (Varian 1980, Salopand Stiglitz 1982). A key difference in market efficiency ariseswhen we compare buyers with access to an electronic marketplace(who have access to no-cost or low-cost information about sellerprices and product offerings), and buyers those who use traditionalretail formats (thereby facing higher search costs). Our study ex-tends Bakos (1997) by focusing on the heterogeneity in buyers’ability or preference to obtain information within online and off-line settings.

3.1. Market efficiency with consumers’ behavioral segmentation –offline-only, online-only, and multichannel shoppers

3.1.1. Modeling Consumer Buying Behavior – existing CBB modelsThere is well-established marketing literature that models how

consumers find and buy goods and services. According to Guttmanet al. (1998), such models of Consumer Buying Behavior (CBB),such as the Nicosia Model (Nicosia 1966), the Howard–Sheth Mod-el (Howard and Sheth 1969), the Engel–Blackwell Model (Engeland Blackwell 1982), the Bettmann Information Processing Model(Bettman 1979), the Andreasen Model (Andreasen 1965), The Ex-tended CBB model (E-CBB) (He et al. 2003), and the Bohm, Feltand Uellner Model (2000), share significant descriptive steps. Typ-ically, a consumer searches the marketplace for products or ser-vices that may satisfy his/her identified need. This is followed bycareful evaluations and comparisons of product alternatives basedon key purchase decision criteria. This step concludes with a con-sideration set, a set of products capable of satisfying the con-sumer’s needs. The final step is the product purchase. We adaptthese common steps into an exploratory framework to demon-strate how behavioral segmentation is used to assess and comparemarket efficiency (Fig. 1).

Modern economic theory suggests that a competitive marketwith perfect information is characterized by efficient brands.Improving consumers’ access to product and brand informationis an effective approach to eliminate market inefficiencies (Kamak-ura et al. 1988, Dardis and Gieser 1980, Maynes and Assum 1982,Sproles 1986).

Consumers may differ in how they perceive the marketplacestructure as it relates to a purchase decision process. Some con-sumers may place all the available online and offline brands intotheir consideration set, thereby invoking a decision-making pro-cess that accommodates multichannel shopping. However, othersmay restrict search and/or purchase to either the online or offlinesetting. By explicitly choosing one of these settings, such consum-ers act as if the other market/setting does not exist. Overall, all con-sumers generally fall into one of three groups: online-onlyshoppers, offline-only shoppers, and multichannel shoppers. Ourframework in Fig. 1 delineates these three behavioral segmentsin terms of both search and purchase phases.

Using the Data Envelopment Analysis (DEA) approach, ourstudy measures market efficiency for each of these three segments.Stated differently, market efficiency may vary across these seg-ments, given that they differ on how information is obtained onprice or quality, or where the purchases are made. For example,compared with traditional retail formats, the electronic market-place is generally more convenient and less expensive for consum-ers in terms of accessing and comparing information. We thereforehypothesize that the online market is more efficient than offlinemarket.

H1a. Overall, the electronic marketplace for online-only shoppersis more efficient than traditional retail formats that serve offline-only shoppers.

H1b. Overall, for multichannel shoppers, online brands are moreefficient than offline brands.

3.2. Market efficiency improvement across product categories

Various studies show that involvement has an important mod-erating effect on consumer information processing (Cacioppo et al.1986, Petty et al. 1983, Chaiken 1980) and that the same informa-tion can be processed in different ways depending on consumerinvolvement. For instance, consumers are likely to scrutinize theattributes of a high-involvement product; they are also less likelyto engage in attributional processing of a low-involvement prod-uct. When a market is characterized by a high level of price disper-sion for a high-involvement product, consumers may becomesuspicious. In other words, a wide price range for the same productis bound to raise questions and concerns that prompt attributionalprocessing, whereby consumers ascribe a cause for the price dis-persion (Winer 1985, Kelley 1973, Heider 1958). To the extent thatconsumers are highly motivated to evaluate a product’s quality,the greater information access in the pre-purchase phase is a majorcomponent of the value that consumers perceive from the elec-tronic marketplace. For example, it would be difficult for the con-sumer to examine a complex product in all its possible attribute

Fig. 1. Consumer buyer behavior model and behavioral segmentation in electronic marketplaces.

P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109 101

combinations within a brick-and-mortar retail store. In contrast, itis possible to gather information on even unusual product attributecombinations in the electronic marketplace. Consider the offlinepurchase process for a car with a particular paint job plus other op-tions where the consumer may not be sure about the product’sappearance if some of these options were changed. In an onlinecontext, however, automakers may develop websites where all col-or combinations, wheel types, door configurations, and so on, canbe selected, whereby customized cyber-representations are gener-ated immediately.

Nevertheless, for certain experiential products, consumptionbenefits are predicted more reliably from information searchablewithin traditional stores than from surrogate information search-able through e-commerce. For instance, consumers may be unwill-ing to purchase paper towels online without touching the realproduct. In this sense, in-store shopping will provide more objec-tive quality information. Online shopping will make less improve-ment in regarding to the market efficiency within this productcategory.

Next, we examine two different product categories, candy barsand personal computers. In the case of candy bars, the importanceof externalities, signaling, and psychological status associated witheither exclusivity or popularity is low. For personal computers, posi-tive network externalities (availability of software, as well as after-sale service) are very relevant. Moreover, consumers are likely torely on an attribute-by-attribute examination of quality in order tojudge the overall quality of a computer. The quality informationfor personal computers is easy to standardize and compare; thisproduct is also expensive and purchased infrequently. For these rea-sons, the pre-purchase information search is likely to be importantin this product category. Therefore, we expect that consumers willgain more benefit by purchasing computers in the electronic mar-ketplace, compared to say, the purchase of candy bars.

To consumers, the appropriateness of online marketing dependsto a large extent on the characteristics of the product or serviceconsidered for purchase. It is therefore necessary to explicitly con-sider these characteristics when evaluating the impact of the elec-tronic marketplace, relative to traditional retail formats. This canbe done by formally incorporating a product or service classifica-tion into the analysis. In this research, we classify products or ser-vices as being either high-involvement or low-involvement. Themarket efficiency of the electronic marketplace may vary by prod-uct involvement. We predict lower market efficiency for product

categories characterized by a low purchase price and low productinvolvement. For products with a high purchase price and high-involvement, consumers may be more likely to use online informa-tion resources, thereby increasing online market efficiency. Thus,for routinely purchased low-involvement product categories suchas Laundry detergents, Butter, Margarines, Cleaners, and Batteryinvestigated in our research where consumers are likely to possessconsiderable prior personal experience, the online marketplacemay not be very efficient.

H2. For high-involvement products, the online shopping formatwill result in higher market efficiency when compared to thetraditional retail format.

3.3. Portion of shipping and handling cost

The change of market efficiency is sensitive to the nature of theproducts and services being marketed. If the product value isintangible or informational, the Internet marketer can improvemarket efficiency through increased use of digital assets. For exam-ple, software can be sampled, purchased, and distributed electron-ically. For large physical products, the online delivery cost may besignificant, so the traditional brick-and-mortar retail format tendsto dominate the online channel with respect to transaction and dis-tribution functions, primarily because these functions do not offereconomies of scale to the Internet marketer (unless there is a sub-stantial delivery charge and/or several products are purchased anddelivered together). An Internet marketer is less likely to carry sucha product, given the cost disadvantage on transaction/distribution(relative to traditional retailers) that reduces market inefficiency.For expensive and infrequently purchased products, the deliverycost is a smaller proportion of the cost of the product. Althoughthis suggests that an online retailer is likely to carry such products,we also recognize that, if consumers insist on inspecting (or phys-ically experiencing) an expensive product prior to purchase, theyare more likely to prefer the traditional retail channel over onlineshopping.

H3. Product categories where shipping and handling represents asmall proportion of total price may have higher efficiency com-pared to products where shipping and handling is a large propor-tion of total price.

102 P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109

3.4. Number of salient attributes

Another key dimension to be considered is the degree to whicha given product or service is differentiable. This takes into accountwhether a seller is able to create a sustainable competitive advan-tage through product or service differentiation. If products or ser-vices are differentiated, effective segmentation can guide buyerstoward their ideal product or service. Consequently, sellers mayhave the opportunity to charge a higher price, by providing oppor-tunities to enhance fit between buyer requirements and productcharacteristics.

A brand has more opportunity to differentiate if more attributesare employed by consumers to evaluate its product category. Weuse the number of salient attributes as a proxy to measure abrand’s ability to differentiate within a category. Higher the num-ber of salient characteristics, higher the brand’s ability to moderateprice competition. We expect higher market efficiency for suchproduct categories.

H4. Product categories with more salient attributes may reflecthigher market efficiency, when compared to product categorieswith fewer salient attributes.

4. Research methodology

This study adopts the Kamakura et al. methodology to defineand measure market efficiency. Kamakura et al. employ DataEnvelopment Analysis (DEA) to assess the degree of inefficiencyof each brand within a product category. They define the measureof price efficiency in the ‘‘characteristics-space.’’ This measurecompares the price of each brand with the best possible price un-der the available technology at its particular combination of char-acteristics. This approach also identifies the set of efficient brandsused as the standard of comparison. Finally, this measure providesguidance on steps needed to render any given brand price-efficient.A brand is defined as efficient if it provides the highest value perdollar spent for that set of characteristics, where an efficient brandwould be assigned a value of 1 on a price efficiency scale. Efficiencyof any particular brand is measured as the ratio between the ‘‘effi-cient’’ price and the actual price paid.

The DEA approach described above can generate useful infor-mation to consumers, such as lists of inefficient brands, or priceranges in which specific brands may become efficient. In this study,we apply this methodology to a number of product categories inonline and offline shopping contexts to estimate gains or lossesto consumers when they purchase goods in the electronic market-place. Nevertheless, our research purpose differs from Kamakuraet al. in that their study focuses on determining if a given brandis efficient, while our study also compares market efficiency acrosstwo different retailing formats.

Since the majority of consumers are multichannel shoppers, it isuseful to develop an efficiency map for this group. Therefore, wedetermine the efficiency scores for multichannel shoppers andsubsequently analyze an integrative regression model. Fare andGrosskopf (1994), among others, have formalized input-orientedDEA models. A detailed description of a DEA model can be foundin Banker and Morey (1986) that extends the original model toempirical data involving ordinal or nominal characteristics. Ourstudy determines efficiency by mathematically solving the follow-ing problem (adapted from Banker et al. 1984).

ðMaxÞhj ;s

þr ;s�i

Zk þ nXg

r¼1

sþi þXm

i¼1

s�i

!ð1Þ

Subject to :

Prkzk �Xn

j¼1

prjhj þ sþr ¼ 0 ðr ¼ 1;2; . . . ; gÞ ð2Þ

Xn

j¼1

xijhj þ s�i ¼ xik ði ¼ 1;2; . . . ;mÞ ð3Þ

Xn

j¼1

hj ¼ 1 ð4Þ

hj; sþr ; s�i P 0 ðj ¼ 1;2; . . . ;nÞ ð5Þ

where n is the number of brands for each category, m the number ofinputs, g the number of outputs, xij the level of i-type input forbrand j, prj the level of r-type output (product characteristic) forbrand j, Zk is the efficiency ratio for the brand under consideration,n is a very small constant, sþr and s�i are slack variables for output rand input i, and hj is the weight for brand j. The slack measure, inaddition to the price reduction, focuses on additional attributes thatshould be offered by inefficient brands to make them efficient(Kamakura et al. 1988). To improve inefficient brands, the objectiveis to reduce the eventual slack in inputs (costs) without reducingthe optimal output (technical performance). If, for instance, the in-put slack is equal to unity, the observed brand is efficient. If, on theother hand, the input slack is less than unity, the brand under inves-tigation is inefficient. A computer program is utilized to solve theabove Data Envelopment Analyses (DEA) models for each productcategory.

We employ the same definition of efficiency as Hjorth-Ander-son (1984) and Kamakura et al. (1988). A given brand is inefficientif there is some other brand with a lower price and equivalent orhigher quality. The price reduction needed to make this brand effi-cient provides a measure of the loss incurred by choosing an inef-ficient brand.

For brands estimated to be lower in quality but higher in pricethan at least one other alternative, efficiencies were calculatedaccording to multichannel shoppers as per Eqs. (1)–(5) of theDEA model.

5. Data collection and data analyses

Price and quality data on several brands/models sold in tradi-tional retail formats were gathered from Consumer Reports. Onlineprice data were collected for the same brands/models from web-sites for e-tailers, and comparison shopping/manufacturers’ websites, which include: shopping.yahoo.com, www.shopping.com,www.bizrate.com, www.mysimon.com, www.nextag.com, www.amazon.com, www.dell.com, www.pricegrabber.com, www.beco-me.com, and www.pronto.com. The same set of quality indicatorsare used to compute efficiency in both offline and online settings.

To render offline and online prices comparable, online prices arecollected over the same time period as the offline data obtainedfrom Consumer Reports. Thirty-one product categories are includedin the study (Table 2). Online stores provided a comprehensiveprice range, so we employed average price as a surrogate for onlineprice. Another consideration is that online price is artificially re-duced because shipping cost is excluded, while offline price is alsoreduced because sales tax is excluded. Our analyses assume thatthese two reductions are comparable. For example, in Illinois State,sales tax paid in offline stores for electronic products is 7%, veryclose to the average shipping cost for goods delivered by onlinestores. Since prices reported by Consumer Reports did not include

Table 2DEA results for market efficiency comparisons.

Productcategories

# ofbrands

# ofattributes

Portion ofshipping in termsof product cost

# of inefficientbrands (offlineonly shoppers)

# of inefficientbrands (onlineonly shoppers)

Meanefficiency(offline onlyshoppers)

Meanefficiency(online onlyshoppers)

Mean efficiency(offline brands,multichannelshoppers)

Mean efficiency(online brands,multichannelshoppers)

1. Laptops 11 9 0.02 4 3 0.91 0.93 0.83 0.932. Printers 16 7 0.13 3 1 0.97 0.98 0.64 0.983. Receivers 15 5 0.01 5 3 0.86 0.88 0.71 0.884. Camcorders 13 4 0.01 4 3 0.94 0.97 0.74 0.975. VCR 10 3 0.03 6 5 0.88 0.92 0.82 0.886. DVD 5 3 0.06 1 1 0.97 0.99 0.70 0.997. Answer machines 8 5 0.11 2 1 0.97 0.98 0.87 0.948. Cordless phones 11 4 0.05 6 6 0.71 0.73 0.61 0.709. APS camera 7 5 0.02 2 2 0.90 0.91 0.80 0.9110. 35 mm Camera 10 5 0.03 0 0 1.00 1.00 0.94 0.9811. Digital camera 4 5 0.01 0 0 1.00 1.00 0.94 0.9912. CD players 7 4 0.05 1 2 0.93 0.94 0.78 0.9413. Alkaline battery 17 2 1.88 15 14 0.41 0.72 0.35 0.2714. Lawn tractors 9 6 0.04 2 2 0.93 0.98 0.92 0.9015. Push lawn mowers 11 6 0.24 4 4 0.87 0.89 0.85 0.7916. Self-lawn mowers 18 4 0.12 12 11 0.90 0.89 0.85 0.8817. Hedge trimmers 13 5 0.04 3 4 0.95 0.90 0.90 0.8718. Treadmills 9 5 0.06 1 1 0.96 0.96 0.94 0.9619. Strollers 21 5 0.14 9 7 0.80 0.72 0.62 0.6820. Gas ranges 12 6 0.08 7 6 0.81 0.81 0.77 0.7921. Electric ranges 15 7 0.10 7 7 0.83 0.84 0.80 0.7922. 3600 TV 10 4 0.08 7 7 0.89 0.86 0.81 0.7723. Satellite 7 2 0.03 4 4 0.77 0.81 0.51 0.8124. PDA 13 4 0.02 7 7 0.74 0.74 0.71 0.7425. Dishwashers 22 6 0.02 6 5 0.96 0.96 0.94 0.9426. Gas grills 9 5 0.02 4 5 0.75 0.71 0.74 0.6727. Vacuum cleaners 26 6 0.03 6 7 0.95 0.95 0.89 0.8928. Cleaners 32 6 0.42 15 13 0.79 0.82 0.75 0.7029. Butter 12 2 0.45 11 11 0.60 0.61 0.60 0.2930. Margarines 17 3 0.48 11 15 0.80 0.71 0.80 0.3431. Laundry detergents 22 2 0.22 19 19 0.65 0.74 0.65 0.55

P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109 103

sales tax, the online price used in our analyses did not include ship-ping costs either.

To test hypotheses H1–H4, price efficiency analysis was appliedto both offline (Consumer Reports) and online data for 31 productcategories. Most product categories include high-involvementgoods, such as electronics and computers, gardening, home appli-ances, and sporting goods. Alkaline battery, Laundry detergents,Butter, Cleaners, and Margarines were chosen as representativelow-involvement products emphasizing experience attributes. Asummary of the results and of the descriptors of each product cat-egory is presented in Table 2.

5.1. The single channel shoppers (online-only vs. offline-only shoppers)

In Table 2, the two mean efficiency columns for single channelshoppers provide approximate measures of the average price effi-ciency for all brands in each product category. Average efficiencyonline does not represent a significant improvement for all productcategories studied. Generally, the degree of improvement (11 outof 31 cases) ranges from 0.6% (Answer machines) to 9% (Laundrydetergents). The observed exception is Alkaline battery, wherethe market efficiency is 0.41 under traditional retail format but0.72 in the electronic marketplace. Interestingly, for both 35 mmcamera and digital camera, market efficiency is 1.0 on both offlineand online formats.

With respect to single-channel shoppers, another aspect tocompare is the number of inefficient brands in online and offlinesettings. For 17 out of the 31 product categories considered, thenumber of inefficient brands is not lower in online settings whencompared to offline settings. To summarize, the evidence suggeststhat the online setting does not decrease the number of inefficientbrands nor does it improve the overall efficiency. The latter finding

suggests that the inefficient brands are not closer to efficientbrands despite the greater accessibility/transparency of informa-tion in the online setting. For the remaining 14 product categories,the online setting yielded a lower number of inefficient brands. Theexceptions included CD player and Margarine product categories.

Overall, the above results suggest that the online marketplace isnot more efficient than the traditional retail setting. We conducteda Paired Sample T-test to test the significance of the efficiency dif-ference between these two marketplaces. The analysis is run on theefficiency data from Table 2, and results are shown in Table 3.1.Those differences are not significant at alpha = 0.05. Therefore,H1a is not supported.

5.2. Multichannel shoppers

The two mean efficiency columns for multichannel shoppers inTable 2 show that average efficiency of online brands is higher thanthat of offline brands for 21 out of the 31 product categories stud-ied. Ten more product categories show more efficiency improve-ment from offline to online for multichannel shoppers than forsingle-channel shoppers. The improvement ranges from 0.4%(Dishwashers) to 33.7% (Printers). Another observation involvesAlkaline battery, where the market efficiency is 0.35 under tradi-tional retail format but 0.27 in the electronic marketplace. This re-sult is directionally opposite to findings for single-channelshoppers. For multichannel shoppers, both 35 mm and digital cam-era products yield market efficiency that is higher for online thanoffline formats. It appears that, more product categories gain effi-ciency online for multichannel shoppers than for single-channelshoppers. However, the Paired Sample T-test of the efficiency dif-ference between offline and online brands for multichannel shop-

Table 3.1Variation in efficiency across products – online vs. offline (online-only shoppers vs.offline-only shoppers, Paired Sample T-test results).

N Mean Std. deviation p Value

Offline mean efficiency 31 .8515 .1312Online mean efficiency 31 .8651 .1101

.258

104 P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109

pers is not statistically significant at alpha = 0.05 (see Table 3.2). Inthis case, H1b is not supported.

5.3. Single-channel shoppers vs. multichannel shoppers

We have reported separate sets of efficiency scores for multi-channel shoppers and single-channel shoppers (Table 2). ThePaired Sample T-tests of the variation in efficiency across productsbetween these two types of shoppers for online brands is statisti-cally significant at alpha = 0.05 (see Table 3.3). The results showthat multichannel shoppers see significantly lower efficiency foronline brands when compared to online-only shoppers. Similarly,multichannel shoppers also encounter lower efficiency for offlinebrands when compared to offline-only shoppers (see Table 3.4).

5.4. Hypotheses testing with an integrative model (H1, H2, H3, andH4)

In order to simultaneously test our hypotheses about the im-pact on market efficiency of factors such as online/offline format,high/low product involvement, number of salient attributes, andshipping and handling costs as a proportion of the total productprice, we ran a multiple regression model with dependent andindependent variables as described below. The model parameterswere estimated for multichannel shoppers as they have grown intoa dominant shopping segment, and are likely to remain so in theforeseeable future.

5.5. Dependent variable OFF_ON_T

This variable represents the average efficiency scores for each ofthe 31 product categories for offline format, and an additional setof 31 average efficiency scores for the same 31 product categoriesextracted with data from online shopping sources.

5.6. Independent variables

(1) FACTOR_T (dummy variable, with a 0 value for offline for-mat, and 1 for online format).

(2) HI_LO_IN (dummy variable, with a 0 value for high-involve-ment products or less frequently purchased products, and 1for low-involvement products or more frequently purchasedproducts).

(3) SALIENT (number of attributes for a particular productcategory).

(4) SHIPPING (shipping & handling costs as a proportion of totalproduct price).

Table 3.2Variation in efficiency across products – online vs. offline for multichannel shoppers(online brands vs. offline brands, Paired Sample T-test results).

N Mean Std. deviation p Value

Offline mean efficiency 31 .7666 .1371Online mean efficiency 31 .7975 .1998

.287

This regression model is acceptable, with a R Square value of0.525, and F = 17.828, p value = .000. The parameter estimates forthis model are shown in Table 4.1. The average efficiency scores(as measured by DEA) do not vary between offline and online for-mats (Beta = .091, p value = 0.306). We note that this result is notaligned with the prediction in H1b. The average efficiency scorevaries considerably across product categories along the continuumfrom the high-involvement product category to low-involvementproduct category (Beta = �.300, p value = 0.020). Thus, H2 is sup-ported. Results for the impact of ‘‘shipping and handling as a pro-portion of total product price’’ on market efficiency follow:Beta = �.357, p value = .005. Thus, H3 is supported. With respectto the impact of number of salient attributes on efficiencyimprovement: Beta = 0.238, p value = 0.020 (as shown in Table 4.1).We therefore conclude that the number of salient attributes in aproduct category has a significant impact on efficiency. This resultstrongly supports H4.

5.6.1. Additional analyses on high-involvement productsTable 4.2 is obtained for multichannel shoppers, but the analy-

sis is limited to 26 high-involvement products (HI_LO_IN = 0).

5.7. Dependent variable OFF_ON_T

The first 26 observations are average efficiency scores of the 26high-involvement product categories for offline format, and thenext 26 observations are average efficiency scores of the same 26product categories for online format.

5.8. Independent variables

(1) FACTOR_T (dummy variable, 0 for offline format, 1 for onlineformat).

(2) SALIENT (number of attributes for a particular productcategory).

(3) SHIPPING (portion of shipping & handling cost in terms ofthe whole product price).

The R Square is 0.200, and F = 4.008, sig. = .013, which meansthis model is highly significant at alpha = 0.05. Coefficients esti-mates are shown in Table 4.2. The following results show thatthe online format has significantly improved efficiency, therebylending support to H1b when the analysis is restricted to these26 product categories.

5.8.1. Additional analyses on low-involvement productsTable 4.3 is obtained by limiting the analysis to 5 low-involve-

ment products (HI_LO_IN = 1). The R Square is 0.786, and F = 7.359,sig. = .020, which means this model is highly significant at al-pha = 0.05. Coefficients estimates are shown in Table 4.3. The re-sults show that the online format has significantly decreasedefficiency for the low-involvement products.

6. Discussions, implications, and limitations

6.1. Three patterns of frontier change (competition structurediscussion) – single channel vs. multichannel shoppers

Zettelmeyer (2000) explores how the existence and the size ofthe Internet affects firms’ optimal pricing strategies. According toZettelmeyer, firms can charge different prices in the offline and on-line marketplaces, but these prices cannot be too different fromeach other before it is profitable to arbitrage between these two.For most products currently offered, the Internet reaches only asubset of potential customers for the product. Two predictions

Table 3.3Variation in efficiency across products for online brands (online-only shoppers vs. multichannel shoppers, Paired Sample T-test results).

N Mean Std. deviation p Value

Efficiency of online brands for online-only shoppers 31 .8515 .1312Efficiency of online brands for multichannel shoppers 31 .7666 .1371

.0000

Table 3.4Variation in efficiency across products for offline brands (offline-only shoppers vs. multichannel shoppers, Paired Sample T-test results).

N Mean Std. deviation p Value

Efficiency of offline brands for offline only shoppers 31 .8651 .1101Efficiency of offline brands for multichannel shoppers 31 .7975 .1998

.0000

Table 4.1Coefficients estimates for thirty-one product categories.

Model Standardized coefficients (Beta) p Value

Constant .000HI_LO_IN �.300 .022FACTOR_T .091 .306SALIENT .238 .020SHIPPING �.357 .005

Table 4.2Coefficients estimates for twenty-six high-involvement product categories.

Model Standardized coefficients (Beta) p Value

Constant .000SALIENT .265 .049SHIPPING �.191 .154FACTOR_T .336 .012

Table 4.3Coefficients estimates for five low-involvement product categories.

Model Standardized coefficients (Beta) p Value

Constant .002SALIENT .423 .075SHIPPING �.471 .053FACTOR_T �.526 .032

P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109 105

stand out. First, some companies price lower on the Internet thanin their traditional retailing format. Second, companies facilitateconsumer search more (or at least as much) on the Internet thanthey do in their traditional retail channels. Firms such as Dell runwebsites that enable customization of computer systems that fitconsumers’ needs exactly, and to link into product reviews andproduct comparisons. Additionally, many firms charge substan-tially lower prices for identical products on the Internet (Brynjolfs-son and Smith 1999). As more consumers increase their use of theInternet and become multichannel shoppers, the opportunity forthe Web to influence their online and offline shopping behaviorgrows. Multichannel shoppers represent a very powerful audienceand tend to be channel-agnostic. Simply put, it is therefore desir-able for businesses to integrate across channels.

Although hybrid firms (firms relying on both traditional and on-line channels) have registered impressive growth, these enter-prises seem to operate in the two marketplaces by treating themas separable. While many businesses see their online and offlinebusiness as separate sales channels, our study suggests that multi-channel customers do not make such distinctions. Since multi-channel shopping is a dominant segment and also represents agrowing trend, failing to understand this finding may lead to lostopportunities to create competitive advantage from market struc-ture changes attributable to the electronic marketplace.

Multichannel shoppers and single-channel shoppers have dif-ferent perceptions on current market structure, and these differ-ences may cause firms to position their brands differently whentargeting these two segments in a competitive environment. Thecompetition structure in terms of efficiency scores among brandsare significantly different when offline and online domains are as-

sumed to represent two separate marketplaces (single-channelshoppers), when compared to another viewpoint that assumes ahybrid marketplace, or that offline and online markets are insepa-rable (multichannel shoppers).

To assess efficiency when comparing a single (offline-only oronline-only shoppers) marketplace with the integrated hybridmarketplace (multichannel shoppers), we develop four indices:OFEC (Offline Efficiency Change), OFFC (Offline Frontier Change),ONEC (Online Efficiency Change), and ONFC (Online FrontierChange), where

OFEC ðor ONECÞ ¼Xn

i¼1

ðXi � X�Þ2=n

OFFC ðor ONFCÞ ¼Xm

i¼1

ðXi � X�Þ2=m

Xi: The difference between ith brand’s single efficiency score inoffline market or online market and the score in a combinedmarketplace of offline and online.n: number of brands in the product category.m: number of efficient brands in this category.

Among the 31 product categories we investigated, four out offive experience products (Alkaline battery, Butter, Margarines,and Laundry detergents) reflect a ‘‘0’’ value both for offline-effi-ciency change and offline-frontier change (see Table 5), while allfive experience product categories reflect higher values for bothonline efficiency change and online frontier change. These resultssupport our hypothesis that electronic marketplace has little influ-ence on offline (traditional) retailing formats for experience-fea-tured goods. Electronic marketplace does improve the efficiencyfor experience goods, but it does not influence the competitionstructure for experience-featured products.

Conversely, 12 (search attributes featured) of the original 31categories reflect a ‘‘0’’ value on online efficiency change and on-line frontier change. But they have a relatively higher score on off-line efficiency and offline frontier change (see Table 5). Thesefindings provide support for H2, such that search products gainefficiency when placed online, and that traditional retailing for-

106 P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109

mats are influenced the most in terms of frontier change. Twoadditional product categories need some attention: Digital cameraand Gas grill. For both these products, there is significant change inoffline efficiency, but this does not appear to impact the efficientbrands that appear immune to marketplace changes (see Table 5).

6.2. Market competition structure

We analyzed the benchmark changes from our DEA results.Three major patterns were found:

(1) Online brands dominate the same brands offline.(2) Other online brands dominate the former efficient offline

brands.(3) No changes for the previously efficient brands.

We attribute the first pattern to ‘‘firm’s multi-channel influence’’,the second pattern to competition structure changes because of on-line retailing. Pattern (3) occurs when firms simply attempt to dupli-cate their conventional pricing and communications strategies (asapplicable to traditional retail formats) in the online retail setting,an approach that is not necessarily optimal. As shown in Table 6,most product categories advance the efficiency frontier toward theirown corresponding online format. This is consistent with Zettle-meyer’s (2000) argument that most brands are sold at lower pricesin the online setting. In our research, we found evidence of pattern(3) for all efficient brands in experience product categories like But-ter, Laundry detergents, Margerines and Battery.

These analyzes also sensitize us to multi-channel conflict andpossible arbitrage. The risk especially arises from pattern (2). Forexample, for ‘‘cleaners’’, 13 out of 17 efficient brands become inef-ficient when subjected to multichannel shoppers’ unique percep-tions of the marketplace, as a result of competition from otherbrands. Only 4 out of 17 brands change the efficiency frontier be-cause of their own online channel influence. In general, firms have

Table 5Efficiency and frontier changes (offline-only shoppers to multichannel shoppers for offline

Product categories Off_Effi_Change (OFEC) Off_Frontier_Change

Laptops (April) 0.10 0.12Printers 0.33 0.34Receivers 0.16 0.19Camcorders 0.22 0.25VCR 0.09 0.11DVD 0.27 0.28Answer machine 0.14 0.15Cordless phones 0.18 0.24APS camera 0.13 0.1535 mm Camera 0.08 0.08Digital camera 0.16 0.00CD players 0.18 0.19Alkaline battery 0.00 0.00Lawn tractors 0.01 0.01Push lawn mowers 0.05 0.07Self-lawn mowers 0.07 0.08Hedge trimmers 0.08 0.07Treadmills 0.04 0.04Strollers 0.30 0.29Gas ranges 0.05 0.05Electricity ranges 0.07 0.103600 TV 0.12 0.15Satellite 0.28 0.29PDA 0.05 0.08Dishwashers 0.05 0.05Gas grills 0.03 0.04Vacuum cleaners 0.10 0.10Cleaners 0.11 0.09Butter 0.00 0.00Margarines 0.00 0.00Laundry detergents 0.00 0.00

to be alert so that they do not lose competitive advantage becauseof the emergence of the online marketplace. Put differently, theelectronic marketplace may introduce new turbulence in someproduct categories, such as satellite, gas grill, hedge trimmers,self-mowers, APS camera, CD players and cleaners, as evident fromour research. Most offline brands for search product categories lostefficiency as a result of being dominated by their own brands in theonline setting, such as PDA and Digital Receivers.

Consider, for example, a firm that is a low-service retailer. Aslong as the reach of the Internet is limited to single-channel shop-pers, it may be in the retailer’s best interest to reorient its onlineprofile to facilitate consumers’ information search. On the otherhand, a retailer whose traditional (brick and mortar) store facili-tates consumers’ information search will benefit from continuingthe same orientation toward consumer search in the online setting.

Even if buyers have objective information on brand attributes,the value of an individual attribute or the superiority of one attri-bute combination over another may not be obvious. Hence, thereare no uniform criteria to evaluate the objective quality of informa-tion across individuals. Furthermore, no product’s quality can beappraised perfectly through inspection, even if full information isavailable. If product characteristics change frequently, consumersmay never possess complete information about the quality of anyproduct. Moreover, consumers are not perfect information proces-sors. For all these reasons, they are likely to possess imperfectinformation about brand quality and may rely on cues or heuristicsto judge quality in online settings. This occurs, either becauseinformation on an attribute is simply not accessible in a givenmedium (e.g., the scent of a detergent or softness of paper towelare difficult to discern online) or the information can only be ob-tained with considerable effort (e.g., generating a comparative list-ing of nutrition attributes for different brands of margarine is moreeffortful offline than online).

Firms can reduce customer price sensitivity for products at theirwebsites by: (1) providing useful information through multiple

brands; online-only shoppers to multichannel shoppers for online brands).

(OFFC) On_Effi_Change (ONEC) On_Frontier_Change (ONFC)

0.00 0.000.00 0.000.00 0.000.00 0.000.07 0.060.00 0.000.08 0.090.00 0.000.00 0.000.05 0.110.02 0.040.00 0.000.47 0.450.16 0.140.12 0.120.01 0.010.07 0.080.00 0.000.08 0.100.04 0.050.06 0.060.00 0.000.00 0.000.00 0.000.032 0.030.06 0.000.09 0.100.18 0.210.32 0.720.39 0.550.20 0.24

Table 6Efficiency frontier change (from separate offline market to baseline market under consumer perception).

Product category # Of efficient brands in off market # Of change because of pattern (1) # Of change because of pattern (2) # Of no change pattern (3)

3600 TV 4 3 0 1Butter 1 0 0 1Cleaners 17 4 13 0Digital camera 4 3 0 1Dishwashers 16 5 0 11Electric ranges 8 2 0 6Gas grills 5 1 1 3Hedge trimmers 10 5 2 3Laptops 7 6 0 3Laundry detergents 3 0 0 3Lawn tractors 7 5 0 2Margarines 6 0 0 6PDA 6 6 0 0Push lawn mowers 7 1 0 6Digital receivers 9 9 0 0Satellite 3 1 2 0Self mowers 6 4 1 1Strollers 12 8 0 4Answer machines 6 5 0 1Camcorders 9 9 0 035 mm Camera 10 8 0 2APS camera 5 4 1 0CD players 6 5 1 0Cordless phones 5 4 0 1Printers 13 13 0 0Battery 2 0 0 2DVD 4 4 0 0VCR 4 3 0 1Treadmills 8 8 0 0Vacuum cleaners 20 10 2 8Gas ranges 6 4 0 2

P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109 107

web pages, in particular, on non-price attributes, (2) making thesite highly interactive, whereby price becomes just one of themany attributes that interested shoppers can browse, (3) offeringa wide range of product assortments and prices, and (4) makingit easier to search on non-price attributes relative to price.

Online customers may be more predisposed to search for betterprices. But, if a firm offers rich and deep information, an interactivesearch process, and a wide product selection, customers may havethe time to also search non-price attributes. Other strategies thatencourage non-price attribute search include a strong editorialcontent, making customer testimonials available, and embeddingspecial features (e.g., a map with locations of attractions near ahotel).

Consumers may be less price-sensitive on search attributes inonline settings, especially if they perceive significant value fromproducts or services that offer integrated solutions. However,although online transactions appear transparent, informationsearch activity is constrained by the marginal cost/benefit rulewhereby search activity stops when the benefits from search equalthe costs due to search. Regardless of a product category is per-ceived, having little or no product-specific knowledge implies thatbrands are perceived as undifferentiated (because there is no infor-mation with which to differentiate them).

Consumers who are uncertain about the evolution of futureconsumption experience may be less willing or able to use objec-tive/current product quality information in online settings. Thisis especially likely for a product or service that includes experienceattributes (attributes that must be experienced to be evaluated) orcredence attributes (those that are very difficult to evaluate andforce the customer to rely on the product’s reputation to evaluatethem (Darby and Karni 1973, Nelson 1970, Zeithaml 1981). In ourstudy, Laptop users may encounter unanticipated benefits (e.g., asnew applications are identified in future) or setbacks (as compet-ing product categories develop over time). In this sense, even if

quality information is relatively current, complete and easy toobtain online, this type of information uncertainty may diminishefficiency in the electronic marketplace.

Findings from a number of studies on the price–quality rela-tionship reaffirm the significant inefficiency that characterizesboth traditional and electronic markets for consumer products.Our study highlights the problem of inefficient brands in electronicmarkets. One solution to reduce this inefficiency is to improve con-sumer information on a larger number of product attributes, whichhas the further advantage of improving choice between brandsthat are not inefficient, or even eliminating inefficient brands. Aunique contribution of this study is to empirically establish thatthe electronic marketplace does improve market efficiency forsome product categories (e.g. high-involvement products) therebyimproving consumer well being without damaging competition(dynamic efficiency).

6.2.1. Limitations of applying DEAOur DEA model assumes that competing products and services

are completely undifferentiated beyond the number of investi-gated attributes. This is a very strong assumption because it canbe argued that marketers will always be able to find a non-pricebasis for differentiation (e.g., warranties, post-sale service, image,and so on). Even minute differences in differentiation, such ashow price is bundled with other offering attributes, may allowmarketers to price their products or services at a higher level.

7. Conclusion

By distinguishing shoppers via behavioral segmentation, and byassessing market efficiency within each consumer segment’s de-fined market boundary, a unique contribution of this study is tocompare market efficiency across online-only shoppers, offline-

108 P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109

only shoppers, and multichannel shoppers. The results reaffirmearlier findings on price dispersion that persistent market ineffi-ciency exists in markets that comprise either single channel shop-pers or multichannel shoppers. In addition, we shed new light onwhether market efficiency varies across products. The electronicmarketplace increases market efficiency significantly for high-involvement products; in contrast, low-involvement products arecharacterized by lower market efficiency in the electronic market-place. Furthermore, products with larger number of attributes gainmore efficiency online, and the products with a large proportion ofshipping and handling costs (relative to total price) are likely toshow diminished efficiency in online retail settings. Finally, ourstudy analyzes efficiency scores for single-channel (offline-onlyor online-only shoppers) and multichannel (integrated hybrid mar-ketplace) settings and finds three major patterns: (1) online brandsdominate the same brands offline; (2) other online brands domi-nate the former efficient offline brands; and (3) no changes forthe previously efficient brands. Future research should relate themarket share changes of brands with changes in their efficiencyscores, thereby enabling the prediction of market share of eachbrand based on inputs derived from efficiency – frontier orientedcompetition structure.

References

Alba, J., Lynch, J., Weitz, B., Janiszewski, C., Lutz, R., Sawyer, A., Wood, S., 1997.Interactive home shopping: consumer, retailer, and manufacturer incentives toparticipate in electronic marketplaces. Journal of Marketing 61 (July), 38–53.

Ancarani, F., Shankar, V., 2004. Price levels and price dispersion within and acrossmultiple retailer types: further evidence and extension. Journal of Academy ofMarketing Science 32, 176–187.

Andreasen, A., 1965. Attitudes and Customer Behavior: A Decision Model. CaliforniaInstitute of Business and Economics Research, University of California.

Bailey, J. Intermediation and Electronic Markets: Aggregation and Pricing in InternetCommerce. Doctoral dissertation, 1998. Available at http://www.smith.umd.edu/faculty/jbailey/pub/phdthesis.pdf.

Bakos, Y.J., 1997. Reducing buyer search costs: implications for electronicmarketplaces. Management Science 43 (12), 1–22.

Banker, Rajiv D., Morey, Richard C., 1986. The use of categorical variables in dataenvelopment analysis. Management Science 32 (December), 1613–1627.

Banker, R.D., Charnes, A., Cooper, W.W., 1984. Some models for estimation oftechnical and scale inefficiencies in data envelopment analysis. ManagementScience 30 (September), 1078–1092.

Baye, M.R., Morgan, J., Scholten, P., 2004a. Price dispersion in the small and in thelarge: evidence from an Internet price comparison site. Journal of IndustrialEconomics 52, 463–496.

Baye, M., Morgan, J., Scholten, P., 2004b. Temporal price dispersion: evidence froman on-line consumer electronics market. Journal of Interactive Marketing 18,101–115.

Baylis, K., Perloff, J.M., 2002. Price dispersion on the Internet: good firms and badfirms. Review of Industry Organization 21, 305–324.

Bettman, J., 1979. An Information Processing Theory to Consumer Choice. AddisonWesley.

Bohm, A., Felt, E., and Uellner, S. An e-commerce framework. In Proc. of theInternational Conference on Advances in Infrastructure for Electronic Business,Science, and Education on the Internet (SSGRR 2001), L’Aquila, Italy, 2000. SSGRR.

Brynjolfsson, E., Smith, M., 2000. Frictionless commerce? A comparison of Internetand conventional retailers. Management Science 46, 563–585.

Cacioppo, J.T., Petty, R.E., Kao, C.F., Rodriguez, R., 1986. Central and peripheral routesto persuasion: an individual difference perspective. Journal of Personality andSocial Psychology 51 (5), 1032–1043.

Chaiken, S., 1980. Heuristic versus systematic information processing and the use ofsource versus message cues in persuasion. Journal of Personality and SocialPsychology 39 (5), 752–766.

Clemons, E., Hann, I., Hitt, L., 2002. Price dispersion and differentiation in onlinetravel: an empirical investigation. Management Science 48, 534–549.

Darby, M.R., Karni, E., 1973. Free competition and the optimal amount of fraud.Journal of law and economics 16 (1), 67–88.

Dardis, R., Gieser, N., 1980. Price and quality of durable goods: are they more closelyrelated in the seventies than in the sixties? Journal of Consumer Policy (Trans.)4 (3), 238–248.

eMarketer, Inc. Comparative estimates: BC2 ecommerce sales growth for China,India, Brazil. Last accessed on June 8, 2012a.

Engel, J., Blackwell, R., 1982. Consumer Behavior, 4th edition. CBS CollegePublishing.

Erevelles, S., Rolland, E., and Srinivasan, S. Are prices really lower on the Internet?An analysis of the vitamin industry. Working paper, Department of Marketing,Riverside, University of California, 2001.

Fare, R., Grosskopf, S., 1994. Estimation of returns to scale using data envelopmentanalysis: a comment. European Journal of Operational Research 79, 379–382.

Forrester Research. Forrester research web influenced retail sales forecast, 2010 to2015 (US) (January 31, 2011).

Guttman, R. H., and Maes, P. Agent-mediated integrative negotiation for retailelectronic commerce. In Proceedings of the Workshop on Agent-MediatedElectronic Trading (AMET’98), Lecture Notes in Computer Science, Minneapolis,MN, USA, 1999, Springer-Verlag, 70–90.

Guttman, R.H., Moukas, A.G., Maes, P., 1998. Agent-mediated electronic commerce:a survey. Knowledge Engineering Review 13 (2), 147–159.

He, M., Jennings, N.R., Leung, H., 2003. On agent-mediated electronic commerce.IEEE Transactions on Knowledge and Data Engineering 15 (4), 985–1003.

Heider, F., 1958. The Psychology of Interpersonal Relations. Wiley, New York.Hjorth-Anderson, C., 1984. The concept of quality and the efficiency of markets for

consumer products. Journal of Consumer Research 11 (September), 708–718.Hostler, E.R., Yoon, V.Y., Guimaraes, T., 2005. Assessing the impact of Internet agent

on end users’ performance. Decision Support Systems 41, 313–323.Howard, J., Sheth, J., 1969. The Theory of Buyer Behaviour. John Wiley and Sons.Kamakura, Wagner A., Ratchford, Brian T., Agrawal, Jagdish, 1988. Measuring

market efficiency and welfare loss. Journal of Consumer Research 15(December), 289–302.

Kelley, H.H., 1973. The processes of causal attribution. American Psychologist 28(2), 107–128.

Lee, Z., Gosain, S., 2002. A longitudinal price comparison for music CD’s inelectronics and brick-and-mortar markets: pricing strategies in emergentelectronic commerce. Journal of Business Strategies 19 (1), 55–71.

Maynes, E. Scott, Assum, Terie, 1982. Informationally imperfect consumer markets:empirical findings and policy implications. The Journal of Consumer Affairs(Summer), 62–87.

Nagle, 1984. Economic foundations of pricing. Journal of Business 57 (1), S3.National Retail Foundation. NRF forecasts retail industry sales growth of 3.4 percent

(January 16, 2012).Nelson, Phillip, 1970. Information and consumer behavior. Journal of Political

Economy 78, 311–329.Nicosia, F., 1966. Consumer Decision Processes: Marketing and Advertising

Implications. Prentice Hall.Palopoli, L., Rosaci, D., Ursino, D., 2006. Agents’ roles in B2C e-commerce. AI

Communications 19 (2), 1–34.Pan, X., Ratchford, B., Shankar, V., 2002. Can price dispersion in online markets be

explained by differences in e-tailer service quality. Journal of Academy ofMarketing Science 30, 433–445.

Pan, X., Ratchford, B., Shankar, V., 2003. The evolution of price dispersion in Internetretail markets. Advances in Applied Microeconomics: Organizing the NewIndustrial Economy 12, 85–105.

Pan, X., Ratchford, B., & Shankar, V. Drivers of price dispersion among e-tailersduring the boom, shakeout, restructuring, and mature periods of e-commerce.Working paper, 2009. Available at http://econweb.tamu.edu/workshops/PERC%20Applied%20Microeconomics/Venkatesh%20Shankar.pdf. Retrieved onApril 19, 2009.

Petty, R.E., Cacioppo, J.T., Schumann, D., 1983. Central and peripheral routes toadvertising effectiveness: the moderating role of involvement. Journal ofConsumer Research 10 (2), 135–146.

Pivk, A., and Gams, M. E-commerce intelligent agents. In Proceedings of the ThirdInternational Conference on Telecommunications and Electronic Commerce (ICTEC2000), Dallas, Texas, USA, 2000, 418–429.

Pratt, John W., Wise, David A, Zeckhauser, Richard, 1979. Price differences in almostcompetitive markets. The Quarterly Journal of Economics (May), 189–211.

Ratchford, Brian T., 1980. The value of information for selected appliances. Journalof Marketing Research 17 (February), 14–25.

Rosaci, D., 2004. A Model Of Agent Ontologies For B 2 C e-Commerce. ICEIS 2004:Software Agents and Internet Computing 14 (17).

Rosaci, D. An ontology-based two-level clustering for supporting e-commerceagents’ activities. In Proceedings of the Sixth International Conference on E-Commerce and Web Technologies (EC-Web 2005), Springer, 2005.

Rosen, Sherwin, 1974. Hedonic prices and implicit markets. Journal of Politicaleconomy 82, 34–55.

Salop, S., Stiglitz, J.E., 1982. The theory of sales: a simple model of equilibrium pricedispersion with identical agents. The American Economic Review 72 (5), 1121–1130.

Sarwar, B. M., Karypis, G., Konstan, J. A., and Riedl, J. Analysis of recommendationalgorithms for e-commerce. In ACM Conference on Electronic Commerce 2000(EC-00), Minneapolis, USA, 2000, ACM Press, 158–167.

Schafer, J., Konstan, J., Riedl, J., 2001. E-commerce recommendation applications.Data Mining and Knowledge Discovery 5 (1/2), 115–153.

Smith, M.D., Bailey, J., Brynjolfsson, E., 2000. Understanding digital markets: reviewand assessment. In: Brynjolfsson, E., Kahin, B. (Eds.), Understanding the DigitalEconomy. MIT Press, Cambridge, MA, pp. 99–136.

Sproles, George B., 1986. The concept of quality and the efficiency of markets: issuesand comments. Journal of Consumer Research 13 (June), 146–148.

Stigler, George, 1961. The economics of information. Journal of Political Economy69, 213–225.

Varian, H., 1980. A model of sales. American Economic Review 70 (September),651–659.

Venkatesan, R., Mehta, K., Bapna, R., 2007. Do market characteristics impact therelationship between retailer characteristics and online prices? Journal ofRetailing 83, 309–324.

P. Jiang, S.K. Balasubramanian / Electronic Commerce Research and Applications 13 (2014) 98–109 109

Walter, Zhiping, Gupta, A., Su, B.C., 2006. The sources of on-line price dispersionacross product types: an integrative view of on-line search costs and pricepremiums. International Journal of Electronic Commerce 11 (1), 37–62.

Winer, B., 1985. An attributional theory of achievement motivation and emotion.Psychological Review 92 (4), 548–573.

Zeithaml, V.A., 1981. How consumer evaluation processes differ between goods andservices. Marketing of services 9 (1), 25–32.

Zettelmeyer, F., 2000. Expanding to the Internet: pricing and communicationstrategies when firms compete on multiple channels. Journal of MarketingResearch 37 (3), 292–308.