agency selling or reselling? channel structures in ... · that sales in the e-channel can also...

MANAGEMENT SCIENCEVol. 62, No. 8, August 2016, pp. 2259–2280ISSN 0025-1909 (print) � ISSN 1526-5501 (online) http://dx.doi.org/10.1287/mnsc.2015.2230

© 2016 INFORMS

Agency Selling or Reselling?Channel Structures in Electronic Retailing

Vibhanshu AbhishekHeinz College, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, [email protected]

Kinshuk JerathColumbia Business School, Columbia University, New York, New York 10027, [email protected]

Z. John ZhangThe Wharton School, University of Pennsylvania, Philadelphia, Pennsylvania 19104, [email protected]

In recent years, online retailers (also called e-tailers) have started allowing manufacturers direct access to theircustomers while charging a fee for providing this access, a format commonly referred to as agency selling.

In this paper, we use a stylized theoretical model to answer a key question that e-tailers are facing: Whenshould they use an agency selling format instead of using the more conventional reselling format? We find thatagency selling is more efficient than reselling and leads to lower retail prices; however, the e-tailers end upgiving control over retail prices to the manufacturer. Therefore, the reaction by the manufacturer, who makeselectronic channel pricing decisions based on their impact on demand in the traditional channel (brick-and-mortar retailing), is an important factor for e-tailers to consider. We find that when sales in the electronic channellead to a negative effect on demand in the traditional channel, e-tailers prefer agency selling, whereas when salesin the electronic channel lead to substantial stimulation of demand in the traditional channel, e-tailers preferreselling. This preference is mediated by competition between e-tailers—as competition between them increases,e-tailers prefer to use agency selling. We also find that when e-tailers benefit from positive externalities fromthe sales of the focal product (such as additional profits from sales of associated products), retail prices maybe lower under reselling than under agency selling, and the e-tailers prefer reselling under some conditions forwhich they would prefer agency selling without the positive externalities.

Keywords : multichannel retailing; electronic commerce; distribution channel; cross-channel spillovers; platformretailing; retail competition; game theory

History : Received February 18, 2013; accepted March 26, 2015, by Chris Forman, information systems.Published online in Articles in Advance December 7, 2015.

1. IntroductionOnline retailing has witnessed strong growth in thelast decade, and U.S. online retail sales were reportedto be $290 billion in 2013, accounting for 9% of totalU.S. retail sales (Sehgal 2014). Online retailers (alsocalled e-tailers) have primarily been resellers (i.e.,e-tailers purchase from manufacturers and resell toconsumers online). More recently, US e-tailers suchas Amazon (through Amazon Marketplace) and Sears(through Marketplace at Sears) have broken awayfrom the traditional reselling format and embracedthe agency model. Indeed, in recent years, a large frac-tion of Amazon’s sales and revenue have come fromits Marketplace (Amazon 2014). Prominent onlineretailers in other countries have also adopted agencyselling, such as Taobao in China (Bonfils 2012) andFlipkart in India (Tiwari 2014). Under this selling for-mat, e-tailers allow manufacturers (or, more generally,upstream agents) direct access to customers throughthe retailing website for a fee, and manufacturers

make decisions regarding key factors such as retailprices without making investments in retail space ore-commerce websites.

Increased transparency in factors such as retailprices has led many manufacturers and content pro-ducers to enthusiastically adopt the agency sellingformat: book publishers such as Macmillan and Ran-dom House, and magazines such as The New Yorker,have started using Kindle and iPad extensively fordistribution of their content, and app and game devel-opers use the Android-based mobile devices andiPhone to reach out to their consumers. Neverthe-less, although agency selling is becoming more pop-ular, we observe that reselling continues to prevail asthe dominant selling format in some mature digitalindustries such as music. In this research, we studydifferent factors that affect the choice of selling formatacross different industries in the online context.

A key distinction between the reselling and agencyselling formats is who sets the retail prices—in agency

2259

mailto:[email protected]



Abhishek, Jerath, and Zhang: Agency Selling or Reselling?2260 Management Science 62(8), pp. 2259–2280, © 2016 INFORMS

selling the retail prices are decided by the manufac-turer, whereas in reselling they are decided by thee-tailer, and we focus on this distinction in this paper.More specifically, in the reselling format, the e-tailerbuys goods from the manufacturer at a wholesaleprice and sets the retail price in the market (as prac-ticed by Amazon for many of its products, iTunes andSoap.com, among others). In the agency selling for-mat, the manufacturer sells directly to consumers onthe retailer’s marketplace and sets the retail price butneeds to share a certain fraction of the revenues withthe e-tailer for providing the access (as practiced byAmazon Marketplace, Kindle, iBook Store, and Alice,among others).

Online retailing is still in flux, and as it grows overtime and increasingly influences traditional retailing,it is important to understand which selling formatswill be adopted by competing e-tailers. Interestingly,online retailers have been the primary architects ofthe e-channel as opposed to manufacturers in brick-and-mortar channel. Some analysts attribute this tothe bargaining power they command in the e-channelbecause of large customer bases that span severalgeographies (Carr 2012). Even though the existenceof different selling formats in online retailing is aconspicuous phenomenon, the understanding of thisphenomenon is limited. Our aim in this paper is todevelop a deeper understanding of online retail for-mats, which can guide e-tailers in making the choicebetween using the historically dominant reselling for-mat and the agency selling format.

A key factor that plays a role in determining theselling format in online retailing is the degree of com-petition among e-tailers. Furthermore, in a multichan-nel retailing environment, sales in one channel mayhave an impact on sales in another channel. Severalempirical studies suggest that sales in the electronicchannel have a negative effect on sales in the tra-ditional brick-and-mortar channel (Brynjolfsson et al.2009 and Goolsbee 2001 show this for apparel andcomputers, respectively). Other studies suggest thatthe e-channel not only gives access to new consumersbut also has a strong stimulation effect on demand inthe traditional channel (Smith and Telang 2010 andMortimer et al. 2012 show this for media, such asmovies and music). Yet other studies suggest thateffect of e-channel sales on traditional channel can bepositive but small (Hilton and Wiley 2010 show thisfor books). Such cross-channel effects impact the man-ufacturer’s aggregate profit from the two channels.In the presence of spillovers between channels, dif-ferent selling formats in online retailing will inducedifferent strategic reactions from the manufacturer,because they will lead to different levels of sales in thee-channel, which will impact the manufacturer’s salesin the traditional channel. Moreover, these reactions

from the manufacturer will affect the e-tailer’s prof-itability as well. Therefore, when the e-tailers decidetheir selling formats, they will consider the reac-tions from the manufacturer under different kinds ofspillovers. Essentially, different selling formats leadto different degrees of channel efficiency and pric-ing power for various players, the impact of whichon players’ profits varies with the type of spillover.The equilibrium market structure is a result of thecomplex interactions among these factors. We build aparsimonious game theory model that captures thesedrivers and sheds light on the reselling versus agencyselling issue.

Our results show that if the e-channel has a nega-tive cross-effect on demand in the traditional channel,it is optimal for the e-tailers to adopt the agency sell-ing agreement. On the other hand, if the e-channelhas a strong positive cross-effect on demand in thetraditional channel, the e-tailers prefer to adopt thereselling agreement. These results are directionallyconsistent with the retailing agreements we observe indifferent industries, as we discuss later. We also findthat as the intensity of competition among e-tailersincreases, they prefer agency selling over reselling. Wealso show that agency selling is beneficial for con-sumers because prices are lower under agency sellingand consumer surplus is higher.

In our main analysis, we assume that the e-retailers’decisions are driven purely by the objective ofmaximizing profits from the sale of products fromthe manufacturer. However, in some situations, thee-tailers’ decisions might also be driven by other con-siderations such as positive externalities from salesof complementary products. An example of this isan e-tailer that sells e-books and also sells compati-ble e-readers (e.g., Amazon sells both e-books and thee-reader Kindle), and higher e-book sales also moti-vate higher e-reader sales. In other words, there is apositive externality from e-book sales, and the e-tailermay have the incentive to increase e-book sales bylowering their price at the cost of sacrificing someof the profit from e-books. On incorporating positiveexternalities from sales into the model, we find thatretail prices may be lower under reselling than underagency selling (which is contrary to the result fromthe main model). We also find that the e-tailers pre-fer reselling under some conditions where they wouldprefer agency selling without the positive externali-ties. This offers an explanation for why, around theyear 2010 when Amazon was selling the e-readerKindle at a high price, it was using the resellingarrangement and was selling e-books even below thewholesale price at which it purchased them from pub-lishers (Rich 2009).

Our research is closely related to three streams ofliterature: (i) interactions between the Internet and

Abhishek, Jerath, and Zhang: Agency Selling or Reselling?Management Science 62(8), pp. 2259–2280, © 2016 INFORMS 2261

traditional channels, (ii) retail competition within achannel, and (iii) platforms and two-sided markets.We now describe how our research relates to the lit-erature in the these areas.

A growing stream of literature looks at the inter-actions between traditional and electronic channels.Balasubramanian (1998) models a horizontally differ-entiated traditional channel and analyzes how thischannel changes in the presence of an Internet retailer.He shows that the e-tailer acts as a wedge betweenthe competing retailers and the retailers compete withthe e-tailer instead of competing with each other. Yooand Lee (2011) extend the Balasubramanian modelto account for heterogeneous customer preferencesfor the e-channel and show that introduction of thee-channel does not necessarily intensify competition,a result contrary to common intuition. Chiang et al.(2003) and Zettelmeyer (2000) suggest that a man-ufacturer can directly sell through the e-channel toimprove its bargaining position versus a traditionalretailer. These papers focus on the impact on the tra-ditional market caused by the introduction of the elec-tronic channel. Our research differs in two distinctways. First, we focus on the equilibrium selling for-mats in electronic retailing in the presence of inter-actions with the traditional channel, as opposed tofocusing on the effect of e-channel introduction ontraditional retail. Second, most research in this areaassumes that the e-channel cannibalizes sales from thetraditional channel. Noting that several studies showthat sales in the e-channel can also stimulate demandin the traditional channel, we extend the analytical lit-erature in this area by analyzing the impact on sellingformats in the e-channel under both positive and neg-ative spillovers from the e-channel into the traditionalchannel.

Our work is also related to the rich stream ofliterature on the determinants of vertical channelstructure (McGuire and Staelin 1983, Bernheim andWhinston 1985, Bonanno and Vickers 1988, Moorthy1988, Coughlan and Wernerfelt 1989). Coughlan andWernerfelt (1989) and Moorthy (1988) focus on theequilibrium channel structure when the manufactur-ers are the architects of the channel. Recent workin channels has been motivated by the observationthat retailers are increasingly gaining greater powerin the channel (Iyer and Villas-Boas 2003, Raju andZhang 2005, Geylani et al. 2006). Jerath and Zhang(2010) study the equilibrium channel structure whena monopolistic retailer is the architect of the chan-nel. Specifically, they study when a traditional retailerallows manufacturers to set up stores within a store(which can be considered as the brick-and-mortaranalogue to online marketplaces). Our research, how-ever, focuses on competing retailers and interchan-nel spillovers, which is important in a multichannel

setting. Various other papers analyze contracts inchannels assuming the channel structure to be exoge-nous (Cachon and Kok 2010, Choi 1991, Desai et al.2004, Jeuland and Shugan 1983, Iyer 1998).

A key question that we address in this paper is thefollowing: When should an e-tailer use agency sell-ing? This is related to the nascent literature on plat-form retailing (Jiang et al. 2011, Kwark et al. 2012,Hagiu and Wright 2014), because the e-tailer is essen-tially functioning as a platform on which manufac-turers interact directly with consumers and pay thee-tailer a fee. Jiang et al. (2011) consider the phe-nomenon of strategic underselling by agents sellingon platforms under the threat of being replaced bythe platform owner, whereas we study a completelydifferent problem, when will a platform structure beadopted in an electronic market? Kwark et al. (2012)study the different roles online reviews play underreselling versus agency selling in vertically differenti-ated market. Online reviews increase upstream com-petition in case of reselling and benefit the retailer, butput a downward pressure on prices under agency sell-ing and hurt the retailer. Hagiu and Wright (2014) findthat whether the marketplace or the reseller mode ispreferred depends on whether independent suppliersor the intermediary have more important informationrelevant to the optimal tailoring of marketing activi-ties for each specific product.

The rest of the paper is organized as follows. In §2,we describe our model. In §3, we analyze the dif-ferent selling formats that are possible and derivethe equilibrium formats that emerge in the e-channelunder different conditions. In §4, we consider severalextensions of the basic model; in one of these exten-sions, we include positive externalities from sales(in addition to direct profit from sales) in the objectivefunctions of the e-tailers. In §5, we conclude with adiscussion and suggestions for future work.

2. ModelWe consider a manufacturer (M) who sells a prod-uct on the electronic channel (E) through two sym-metric “pure play” electronic retailers (e-tailer X ande-tailer Y ). Pure play implies that retailers only havean electronic/online presence. In the rest of the paper,we refer to the manufacturer as “him” and to thee-tailers as “her.”

2.1. Channel StructuresDifferent market configurations are possible in thissetting based on the contractual agreements betweenthe manufacturer and the e-tailers. An e-tailer canenter into a reselling agreement (denoted by R) withthe manufacturer, in which case the e-tailer pur-chases the product from the manufacturer at a fixedwholesale price and decides the retail price for the


consumer. This is similar to the selling format that iscommonly used in brick-and-mortar retailing. Alter-nately, the e-tailer can enter into an agency sellingagreement (denoted by A) and allow the manufac-turer to sell its products directly to consumers. In thisarrangement, the manufacturer determines the retailprice but has to pay the e-tailer a fraction �, 0 ≤ �≤ 1,of its revenues as fee for accessing her customers.Such a fee structure is used by many online market-places, e.g., Amazon Marketplace, Travelocity, Alice,iPhone App Store, iBook Store, and Kindle.1 Note thatthe key distinction between the two arrangementsthat we focus on is whether the manufacturer or theretailer has the ability to decide price.2

The e-tailers independently and simultaneouslychoose either the reselling or the agency selling for-mat. We endow the e-tailers with the power to makethis decision because online retailers have large cus-tomer bases and extensive reach, which offer themsubstantial power in determining the selling formatthat they want to use. Note, however, that in makingthese decisions the e-tailers will consider the subse-quent reactions that they expect from the manufac-turer. The e-tailers’ decisions lead to three possibleconfigurations in the market:

1. RR—Both e-tailers are resellers; we name this the“wholesale” configuration.

1 In addition to the percentage fee, marketplace providers some-times charge the sellers a small fixed fee. For instance, AmazonMarketplace charges a small fee of $39 per month to sellers whowant to sell regularly over a long time horizon on its marketplace.This fee gives the sellers access to certain infrastructure servicessuch as easy interfaces for uploading and displaying product infor-mation. Note that, because even small professional sellers havemonthly sales much higher than many thousands of dollars, thissmall fee is, in all probability, not levied as part of a two-part tariffwith the intent of coordinating the channel.2 There are other distinctions that we do not consider in this paper.For instance, who provides service and how much, and whetherthe service provided is observable, is an important issue (Jerathand Zhang 2010, Jiang et al. 2011). Another important distinction,related to logistical issues, is that under reselling the e-tailer ownsthe inventory and is responsible for order fulfillment, whereasunder agency selling the manufacturer owns the inventory and isresponsible for order fulfillment. Whereas this may not be impor-tant in the case of information goods because delivery is virtuallycostless, it may be important in the case of physical goods becausedelivery entails a cost. We do not model this distinction betweenthe two contractual agreements in our model and assume that ful-fillment entails a costless transfer to the consumer, an assumptioninvoked in most previous studies on channels. We also note thatsometimes variations of these basic selling formats are used, whichdo not have a significant impact on the basic forces at play. Forinstance, in a variation on reselling, e-tailers may decide the retailprice and negotiate with manufacturers on percentage fees that themanufacturers will get per sale (instead of a wholesale price). Sim-ilarly, in a variation of agency selling, the manufacturers, althoughsetting retail prices, do not have complete freedom and can onlychoose one of different pre-set pricing levels.

2. RA—E-tailer X chooses to sell the product asa reseller, and e-tailer Y enters into an agency sell-ing arrangement with the manufacturer; we name thisthe “hybrid” configuration. Note that the RA and ARarrangements are the same because the e-tailers areindistinguishable.

3. AA—Both e-tailers enter into agency sellingagreements with the manufacturer; we name this the“agency” configuration.

2.2. Demand Specification

2.2.1. Electronic Channel. We assume the two e-tailers are symmetric and differentiated, and they facea linear demand system given by

qX =1

1 +�−

11 −�2

pX +�

1 −�2pY 1

qY =1

1 +�−

11 −�2

pY +�

1 −�2pX1

(1)

where qi is the quantity of the product sold at e-tailer iand pi is the retail price charged by e-tailer i, i ∈

8X1Y 9.3 Linear demand models have been used inthe marketing and economics literatures previously tomodel differentiated duopolies (Shubik and Levitan1980, Singh and Vives 1984, Raju et al. 1995, Jerathand Zhang 2010).4 The demand system above givesus a model of partial competition where � ∈ 60115 isthe parameter that captures the degree of differentia-tion between the two e-tailers. If � = 0, utilities fromthe products are independent, and both retailers act asmonopolies in their respective markets. As � increase,downstream competition in the market becomes moreintense. If � → 1, the products are fully substitutableand the market becomes perfectly competitive. Inter-mediate values of � represent different degrees ofdifferentiation.

We use this particular demand specification becauseit has the following two desirable characteristics—as

3 Note that the demand system above can be rewritten in terms ofthe price differential between e-tailers X and Y as qX = 1/41 +�5−41/41 +�55pX + 4�/41−�2554pY −pX5, qY = 1/41+�5− 41/41+�55pY +

4�/41 −�2554pX − pY 5.4 This demand specification follows from the quadratic consump-tion utility of a representative consumer, which is given byU4qX1 qY 5= qX + qY − 4q2

X + q2Y + 2�qXqY 5/2. This is a special case of

the quadratic utility function in Shubik and Levitan (1980). Simi-lar quadratic utility functions have been used in Singh and Vives(1984), Raju et al. (1995), and Jerath and Zhang (2010). This utilityfunction for a representative consumer assumes that positive non-integer amounts of each product may be consumed. As pointed outin Dixit and Stiglitz (1979) and Anderson et al. (1992), this type of aformulation for a representative consumer can be consistent with aformulation in which every individual consumer in the populationconsumes zero or one unit of one product. Therefore, our formula-tion is appropriate for a digital goods setting (in which, generallyspeaking, consumers may purchase only one unit of a digital goodsuch as an e-book or a digital copy of a song).


the differentiation increases (i.e., value of � decreases),the sensitivity to price, 1/41 − �25, decreases (consis-tent with the intuition that customers are less pricesensitive for more differentiated products), and thesize of the total potential market, 2/41 + �5, increases(consistent with the intuition that more differentiatedproducts reach a wider customer base). We have alsoconsidered other popular demand systems, such as theformulation qi = 1 − pi + �4pj − pi51 i1 j ∈ 8X1Y 91� > 0.This formulation was used in Raju et al. (1995) and hasthe feature that, on changing �, the degree of differ-entiation between the products changes, whereas thetotal market size remains constant. We find that ourresults stay qualitatively the same under this alterna-tive demand formulation. (More details of the analysisare available in the appendix.)

The differentiation between the e-tailers comesfrom a combination of what is sold and where it issold, and one or both of these factors may be opera-tive. For instance, even if the manufacturer sells thesame product through the e-tailers, goods can stillbe differentiated from the perspective of a consumerbecause of various other factors at play. For exam-ple, the experience of reading an e-book on the iPadis very different from the experience of reading thesame e-book on the Kindle, and this can affect con-sumers’ preference for (and substitutability between)Apple or Amazon, even though the content of thee-book stays the same. We note that this is based onhorizontal differentiation, not vertical differentiation,because at the same price not all consumers will agreethat one is better than the other. More broadly, dif-ferentiation between e-tailers could arise because ofwebsite layout, inertia, loyalty and membership pro-grams, other products offered by the e-tailers, etc.Based on how salient these differences are from theconsumers’ perspective, the e-tailers might competevery aggressively or not so aggressively.

The total demand for the manufacturer through theelectronic channel is given by qE = qX +qY . We assumethat the manufacturer can reach the consumers in thiselectronic market only through these e-tailers, i.e., werule out the possibility that the manufacturer sets upa direct electronic channel, because direct selling is adifficult proposition for most manufacturers.

2.2.2. Traditional Channel. There is an existing(traditional) retail channel where the manufacturersells goods at a per-unit price normalized to $1 andfaces a base demand Q̄ in the absence of the elec-tronic channel. It is important to note that we do notassume that the manufacturer necessarily sells iden-tical products through the two channels; he may sellthe same product or different products through thetwo channels. For example, a publisher of books maysell e-books through the online channel and print

Figure 1 Market Structure: Interaction of the E-Channel withTraditional Retail

M

Traditionalchannel

YX

E-channel

QqX qy

+�qE

books (same and different titles) through the tradi-tional channel. Similarly, a music producer or artistmay sell individual mp3 tracks online, and sell albumCDs, concert tickets, and airplay rights offline.

The sales of the product in the traditional channelare affected by the sales in the new electronic chan-nel. We model this by assuming that sales in the tra-ditional channel are now Q̄ + �qE , where qE denotesthe total sales in the e-channel, and the parameter �captures the net overall cross-channel effect that thee-channel imposes on sales through the existing chan-nel. More specifically, � represents the net change inoffline sales for every unit sold online. In our analy-sis, we consider � ∈ 6−1117.5 This formulation mod-els, in a reduced-form way, an increase (� > 0), ordecrease (� < 0), or no effect (� = 0) in traditional salesbecause of every unit sold through the e-channel. Fig-ure 1 shows a schematic diagram explaining the cross-channel effect.

We assume a unidirectional and exogenous cross-channel effect for simplicity. In §4.3, we extend themodel to incorporate endogenous consumer choiceacross the electronic and traditional channels andshow that our results remain qualitatively the same.

When � > 0, we have a positive cross-effect, i.e.,every sale in the e-channel leads to � units ofincreased sales in the traditional channel. One factordriving this increase could be word-of-mouth, e.g.,when a consumer purchases an e-book on the Kindleand discusses it with his friends, they might buy theprint version of the book from a bookstore. Hilton and

5 For � < −1, in our model, the negative externality imposed is sostrong that the manufacturer would not want to sell through theelectronic channel at all. Since this is an uninteresting scenario, wedo not consider � < 1.


Wiley (2010) find that availability of e-books leads toa moderate increase in the sales of print books, indi-cating that � might be positive and small/moderatefor books. For media products, such as movies andmusic, we see strong evidence of positive externalities(Smith and Telang 2010). For example, positive cross-effects are present in the music industry in whichseveral independent artists and, more recently, pop-ular bands such as Coldplay, even give away theirmusic online for free to promote concert attendanceand generate demand for airplay. On the other hand,various studies also find that, in certain cases, theonline channel has a negative cross-effect on salesthrough the traditional channel, i.e., � < 0. This mightbe attributed to cannibalization, i.e., consumers whobuy online are less likely to go into stores to buy theproduct, which could happen for consumer packagedgoods sold at online stores like Soap.com and Alice(Brynjolfsson et al. 2009, Goolsbee 2001). In summary,different studies show that cross-effects can be neg-ative or positive (or absent) in different cases. Ourmodel is flexible enough to capture all of these cross-effects and understand their implications on the sell-ing formats that will result in the e-channel. Note thatmany different factors may be at play simultaneouslyin determining the aggregate effect of how demandin different channels interacts in a multichannel sce-nario.6 Our reduced-form way of modeling cross-effects is meant to capture the aggregate cross-channeleffect. Future research can look into the impact of dif-ferent factors in isolation by micro-modeling them.

In our basic model, we also assume for simplicitythat the impact on demand in the traditional chan-nel from sales on the e-channel does not influenceprices in the traditional channel. This can be causedfor several reasons. One reason is that the baselinedemand, Q̄, is sufficiently high compared to the mag-nitude of the spillover, and therefore prices are notinfluenced much (in support of this, note that onlinesales only account for about 9% of total retail sales).Other reasons can be high menu costs, established ref-erence prices, and operational reasons. For example,the prices of print books have remained largely unaf-fected by e-book pricing. One could argue that thisis caused by a combination of historically establishedreference prices in the minds of consumers as wellas operational difficulties (e.g., prices are printed onbooks and cannot be changed with frequent changesin e-book prices). Nevertheless, in §4.2 we relax thisassumption and show that the qualitative nature of

6 The different factors at play simultaneously may include demandexpansion after a new channel is introduced, a segment of con-sumers switching between channels because of price differences,and different products being offered on different channels (e.g.,e-books through the e-channel and collector’s editions through thetraditional channel; music singles through the e-channel and albumCDs, concert tickets and airplay rights through the other channel).

Table 1 Notation

Notation Meaning

� e-channel competition� Spillover effect into traditional channelpX 1 pY Prices charged at the e-tailersqX 1 qY Demand at the e-tailers�X 1 �Y Profits of the e-tailers�ME Profit of the manufacturer from the e-channel�M Total profit of the manufacturerRR, RA, AA e-channel configurations�X 1 �Y Fees charged by the e-tailers in the agency selling formatw ∗

RR ,w ∗

RA Equilibrium wholesale prices in the RR and RA configurations�∗

AA1 �∗

RA Equilibrium fees charged by the e-tailers in the AA and RAconfigurations

the findings do not change when the prices in the tra-ditional channel are chosen endogenously.

Given the above, the profit of the manufacturerdepends on the profit from the e-channel, sales onthe traditional channel, and the spillover effects of thee-channel on the traditional channel, and is given by

�M = 1 · 4Q̄+ �qE5+�ME1

where qE is the total demand in the electronic chan-nel and �ME is the manufacturer’s profit from theelectronic channel, which depends on the e-channelstructure—RR, RA, or AA. Therefore, the manufac-turer will react to the selling format in the e-channel,and the e-tailers have to consider this reaction whilemaking their selling format decisions. A summary ofnotation used is presented in Table 1.

Finally, note that the insights from our model canbe used to inform other multichannel situations withcross-channel effects. However, using agency sellingmakes our model fairly specific to electronic channelssince agency selling is not typical in other scenarios.Therefore, any extensions to other scenarios should bedone with this caveat in mind.

3. AnalysisIn this section, we characterize the equilibriumoutcomes under the three configurations of thee-channel—RR, RA, and AA. Following this, we de-rive which of these configurations will be observedin equilibrium for different combinations of values ofmarket competition (�) and cross-channel effect (�),and the impact on different market players of the re-sulting arrangement. We also derive insights for set-ting agency fees by e-tailers.

3.1. Analysis of Different Configurations in theE-Channel

3.1.1. Electronic Channel with Two Resellers(RR). In this configuration, both e-tailers purchase theproduct at a fixed wholesale price from the manu-facturer and then set retail prices. The timing of this


Figure 2 Schematic Diagrams Showing the RR, RA, and AA Channel Structures for the E-Channel

M

w

Customers

X Y

(a) Wholesale (RR) arrangement

(R ) (R )

e-channel

pX pY

M

(b) Hybrid (RA) arrangement

w

Customers

(R ) (A)

e-channel

pX pY

X Y

�2

M

(c) Agency (AA) arrangement

Customers

(A) (A)

e-channel

pX pY

X Y

�X �Y

Notes. M represents the manufacturer and X and Y represent the e-tailers. The direction of the arrow points away from the player making the decision. Adashed line indicates that the list price at an e-tailer in case of agency selling is decided by the manufacturer.

subgame is as follows. First, the manufacturer decidesthe wholesale price it will charge to both e-tailers.7

The e-tailers then individually decide if they wishto accept this wholesale price or choose an outsideoption (which we normalize to zero). If the e-tailersaccept, they set the retail prices. This arrangement isillustrated in Figure 2(a). The profits of the manufac-turer and the two e-tailers are as follows:

�RRM = 1 · 4Q̄+ �qE5

︸︷︷︸

profits from traditional channel

+w4qX + qY 5︸︷︷︸

profits from

1

�RRi = qi4pi −w51 i ∈ 8X1Y 91

(2)

where w is the wholesale price, pi is the price chargedby e-tailer i and qi is the quantity sold by e-tailer i. Wesolve this subgame using backward induction. Theequilibrium prices and quantities under this configu-ration are as follows:

w∗

RR =1 − �

21 pRRX = pRRY =

3 − 2�− �

242 −�51

qRRX = qRRY =1 + �

242 −�541 +�50

Note that the wholesale price decreases linearly withthe spillover effect. This is because, if � is negative,then more sales in the e-channel will decrease sales inthe traditional channel, and the manufacturer keepswholesale price high to induce high retail prices andsell less in the e-channel. As � increases and becomespositive, the effects are reversed, and the manufac-turer wants to increase sales in the e-channel bycharging a low wholesale price to induce low retailprices. The equilibrium payoffs are as follows:

�RRM =

41 + �52

242 −�541 +�51 �RR

ME =1 − �2

242 −�541 +�51

7 We check that the manufacturer benefits from selling to bothe-tailers as opposed to selling to only one of them.

�RRi =

41 −�541 + �52

442 −�5241 +�5∀ i ∈X1Y 0

It is easy to see that both �RRM and �RR

i are convexand increasing in � , whereas �RR

ME is concave in � andachieves a maximum when � = 0.

3.1.2. Electronic Channel with One Reseller andOne Agency Seller (RA). In this configuration, oneof the e-tailers uses agency selling, and the othere-tailer sells as a reseller. First, the e-tailer willing toenter into an agency selling agreement announces theagency fee, �, which is the fraction of the revenueshe will keep for each item sold through the market-place. Without loss of generality, we assume that Yuses agency selling in this configuration. Next, afterobserving the agency fee, the manufacturer decideswhether he wants to sell through the marketplace oronly through the reseller (we find that M always sellsthrough the reseller). He simultaneously announcesthe wholesale price w for the reseller (X). Followingthis, X decides the retail price pX that it will charge,and the manufacturer decides the retail price pY inthe marketplace provided by Y . This arrangement isillustrated in Figure 2(b). (If the timing is changedsuch that the manufacturer simultaneously announcesthe wholesale and retail prices it will charge, theresults do not change qualitatively.) If M decides tosell through both e-tailers, his profit is

�RAM = 1 · 4Q̄+ �qE5+ qXw+ 41 −�5qY pY 0

If he decides to sell only through the reseller, weassume that the price in the marketplace is such thatthere are no sales through the marketplace.8 In this

8 We model zero sales in the marketplace j4 6= i5 by setting qj = 0, or1/41 +�5− 41/41 −�255pj + 4�/41 −�255pi = 0 ⇒ pj = 1 −�+�pi0


case, the manufacturer’s and the e-tailers’ payoffs aregiven by

�RAM = 1 · 4Q̄+ �qX5+wqX1 �RA

X = qX4pX −w51

�RAY =

{

�qY pY if M sells through Y 1

0 otherwise0

(3)

If the manufacturer chooses to sell through both thee-tailers, we observe the following prices in the elec-tronic channel:

pRAX =w4241 −�5+�25+ 41 −�5641 −�542 +�5−��7

41 −�544 −�251

pRAY =w43 −�5�+ 41 −�5641 −�542 +�5− 2�7

41 −�544 −�250

The optimal wholesale price for e-tailer X in this case,conditional on �, is given by

w∗

RA = 441−�5841−�542+�544+�4241−2�5−�41−�555

+4�48+242−�5�5−41+�548+42−�5�255�95

·42(

�4+41−�5447−�5�2

−85)

5−11

which is linearly increasing in � (as in the wholesalecase).

If the manufacturer chooses to sell only through thereseller, we observe the following wholesale and retailprices in the electronic channel:

wRA =1 − �

21 pRAX =

3 − �

40

Note that the optimal wholesale price above is thesame as in the wholesale (RR) case.

We find that, in equilibrium, the manufacturer sellsthrough both e-tailers. We compute the optimal fee�∗RA numerically over the entire parameter space,

because solving for it analytically is challenging. Weobserve that the wholesale price in this case is weaklylower than the wholesale prices offered in the whole-sale (RR) case, i.e., w∗

RA ≤w∗RR, and the retail prices at

both the e-tailers are lower than the retail prices inthe wholesale case, i.e., pRAX 1 pRAY ≤ pRRX = pRRY .

3.1.3. Electronic Channel with Two Agency Sell-ers (AA). In this configuration, both e-tailers useagency selling. First, the e-tailers announce the fees �X

and �Y . The manufacturer then decides whether hewants to sell through X or Y or both. Once he decideswhich e-tailer he wants to sell through, he decides theretail prices. This arrangement is illustrated in Fig-ure 2(c). If the manufacturer chooses both e-tailers, hisprofit can be written as follows:

�AAM = 1 · 4Q̄+ �qE5+ 41 −�X5qXpX + 41 −�Y 5qY pY 0

If the fee charged by an e-tailer is too high, themanufacturer might decide not to use the e-tailer’s

marketplace for selling his products. Under this con-dition, when he chooses to sell only through one e-tailer (say i), his profits and the e-tailer’s profits are

�AAM = 1 · 4Q̄+ �qi5+ 41 −�i5qipi1

�AAi =

{

�iqipi if M sells through i1

0 otherwise0

If the manufacturer chooses to sell through both e-tailers, we observe the following retail prices in theelectronic channel:

pAAi = 441 −�5441 −�i5

2�− 41 −�Aj 5��

+ 41 −�Ai 542 +�541 −�A

j + �555

· 4441 −�Ai 541 −�A

j 5− 42 −�Ai −�A

j 52�25−11

i1 j ∈ 8X1Y 91 i 6= j0

In case the manufacturer wishes to sell onlythrough e-tailer i, we set the quantity sold at the othere-tailer to zero, i.e., qj = 0, by using a similar treatmentas in the hybrid case. The retail price in marketplace iis then given by

pi =12

(

1 −�

1 −�i

)

0

We find that, in equilibrium, the fees are such thatthe manufacturer sells through both e-tailers. Thereis a unique and symmetric equilibrium in the feescharged by the e-tailers, given by �X = �Y = �∗

AA,where

�∗

AA = 1 −13

(

�4�− 2�5+K1/3

+�4 + 6�� − 4�3� − 3�2 + 4�2�2

K1/3

)

and

K = 3√

3441−�25�24�6+2�345−3�25�

+349−11�2+4�45�2

+4�43−2�25�3+�4551/2

+�6+9�3�−6�5�+343−2�252�2

+�49−8�25�30

The equilibrium prices and quantities as a functionof agency fee are as follows:

pAi =1 −�∗

AA − �

241 −�∗AA5

1 qAi =1 −�∗

AA + �

241 +�541 −�∗AA5

1

∀ i ∈ 8X1Y 90 (4)

Note that, after conditioning on the fees, the pricesabove do not depend directly on the level of marketcompetition (�). This is because the prices for bothe-tailers are jointly set by the manufacturer.


We can see from (4) above that � has two effects onthe retail prices—a direct effect, and an indirect effectthrough �∗

AA. The direct effect leads to the reductionof prices as � increases, because the manufacturerwants to sell more in the e-channel to advantage ofthe spillover into the traditional channel. The indi-rect effect of � manifests itself through �∗

AA, whichincreases or decreases in � under different conditions.When �∗

AA increases in � , the indirect effect leadsto a further reduction in prices (beyond the directeffect). This is because a larger fee means that themanufacturer can keep lesser revenue per unit, so hasthe incentive to reduce the price. On the other hand,when �∗

AA decreases in � , the indirect effect leads toan increase in prices. This is because a smaller feemeans that the manufacturer can keep more revenueper unit, so has the incentive to increase the price.In this case, whereas the overall effect of � on theprices is a combination of these two effects, the directeffect dominates the indirect effect and prices alwaysdecrease with � . The equilibrium payoffs for the man-ufacturer and the retailers are as follows:

�AAM = Q̄+

41 −�∗AA + �52

241 −�∗AA541 +�5

1

�AAME =

41 −�∗AA5641 −�∗

AA52 − �27

241 −�∗AA541 +�5

1

�AAi =

�∗AA641 −�∗

AA52 − �27

441 −�∗AA5

241 +�5∀ i ∈ 8X1Y 90

Here, � has both a direct and an indirect effect(through �∗

AA) on the manufacturer’s total profitsalso. The direct effect leads to an increase in prof-its as � increases. However, when �∗

AA increases in � ,the manufacturer’s profits might actually decreasebecause he keeps a smaller fraction of the e-channelrevenue. In this case, we observe that the profits arenonmonotonic in the spillover.

Before we proceed, in Table 2 we present a sum-mary of the relationships between the different equi-librium quantities under the three different e-channelstructures. To obtain a sense of consumer welfareunder the different formats, we compare the pricesand quantities sold in the three cases and observethat under all market conditions, the lowest prices are

Table 2 Comparison of Equilibrium Outcomes Under DifferentChannel Structures

Quantity Relationship

Prices pAAX = pAA

Y ≤ pRAX 1 pRA

Y ≤ pRRX = pRR

Y

Quantities qRRE ≤ qRA

E ≤ qAAE

Wholesale prices w ∗

RA ≤ w ∗

RR

Manufacturer’s profits �RAM ≤ �RR

M 1 �AAM

Manufacturer’s e-channel profits �AAE ≤ �RR

E

observed under the agency arrangement, followed bythe hybrid arrangement, and the wholesale arrange-ment has the highest prices. As a result, highest quan-tities are sold in the agency arrangement, followed byhybrid and finally by wholesale. The prices in agencyselling are lower (as compared to reselling) because,given �, a higher price would mean more payout tothe retailer (effectively, higher marginal cost) as wellas lower sales. To counteract this loss, the manufac-turer lowers the price to avoid the high payout to theretailer and achieves higher sales.

3.2. Equilibrium Selling FormatsUsing the solutions of the different subgames studiedin the previous section, we now solve the first stageof the game in which the competing e-tailers deter-mine the selling formats they will adopt. Note that weassume that selling formats are determined before thepricing decisions (in the subgames) because it is eas-ier for the manufacturer to change wholesale pricesthan it is for the e-tailers to change the selling format.In other words, the decision on the selling format is alonger-term decision than the decision on wholesaleprice. This timing is also in accordance with modelingchoices in previous papers on vertical channel struc-tures, e.g., (McGuire and Staelin 1983, Moorthy 1988,Jerath and Zhang 2010). The following matrix repre-sents the game between the e-tailers in strategic form(where �AR

X =�RAY and �AR

Y =�RAX ).

E-tailer Y

A R

E-tailer XA �AA

X 1�AAY �AR

X 1�ARY

R �RAX 1�RA

Y �RRX 1�RR

Y

Note that we have only two exogenous param-eters—� and � . We solve the preceding 2 × 2 gamefor every tuple (�1�) to derive the selling for-mats adopted by the e-tailers in equilibrium. Fromthe e-tailers’ perspectives, different formats offerthem varying levels of channel efficiency and pric-ing power. The resulting equilibrium configurationdepends on the degree of competition and the cross-channel spillover. The equilibrium selling formats forall allowable values of � and � are described in thefollowing proposition and illustrated in Figure 3.

Proposition 1. For any given �, both e-tailers preferagency selling for negative or low positive spillovers, bothe-tailers prefer reselling for high positive spillovers, andone e-tailer prefers agency selling with the other preferringreselling for medium positive spillovers.


Figure 3 Equilibrium Selling Format as a Function of Cross-ChannelEffect 4� 5 and Retail Competition 4�5

–1 10�

1

�

0

AA RR

RA

We now discuss the insights behind this pattern.First, different selling formats lead to different chan-nel efficiency. In reselling, the manufacturer sets thewholesale price, followed by each e-tailer setting herretail price, which leads to the problem of doublemarginalization. In agency selling, where a percent-age fee structure is used, if the manufacturer sets ahigher retail price, he also has to pay a larger amountas fee to the e-tailer, which leads to a downward pres-sure on prices. In other words, the problem of doublemarginalization is reduced in the case of agency sell-ing, and the channel is more efficient as comparedto the reselling agreement. If there were no spillovereffects, the e-tailers would always prefer the agencyselling format.

When faced with spillovers, the incentives of themanufacturer and the e-tailers are further misalignedbecause of another reason. This is because the man-ufacturer wants to maximize profits keeping bothchannels (electronic and traditional) in mind, whereasthe pure-play e-tailers want to maximize the prof-its they derive from the electronic channel alone. Inthe case of a negative cross-effect of e-channel saleson traditional channel sales, the manufacturer hasthe incentive to limit e-channel sales. Therefore, if areselling agreement is chosen, the manufacturer willset a high wholesale price for the e-tailer which, inturn, will lead to higher retail prices and limit the neg-ative spillover into the traditional channel. However,this implies that the double marginalization problemwill worsen in the e-channel. Therefore, the e-tailershave the incentive to choose agency selling, whichincreases channel efficiency as discussed above. Inaddition, the e-tailers can also charge lower fees foraccess to the marketplace, which gives the manufac-turer greater incentive to sell through the e-channel.

Therefore, if e-channel sales have a negative cross-effect on traditional channel sales, the e-tailers willprefer agency selling.

When e-channel sales lead to a positive cross-effect on traditional sales, the manufacturer has theincentive to increase e-channel sales through lowere-channel prices. In the agency arrangement, the man-ufacturer has control over the retail price and willcharge a retail price lower than the efficient retailprice of the e-channel. This reduces the e-tailers’ real-ized fees as well and hurts their profits. On the otherhand, in reselling, not only do the e-tailers control thefinal retail price, the manufacturer is willing to chargelower wholesale prices to the e-tailers to keep retailprices low. Therefore, if the spillover effect is positiveand large, the e-tailers prefer the reselling agreement.

Finally, when the spillover effect is positive andmedium, the channel structure is mixed, which is sur-prising because ex-ante symmetric e-tailers use asym-metric channel structures. In the mixed structure, theagency format maintains efficiency, whereas resellingformat prevents retail prices from going too low inthe e-channel.

To summarize, depending on the cross-channel ef-fect (which determines how the manufacturer willrespond in prices), the e-tailers either prefer to have amore efficient vertical channel by using agency sellingor have increased pricing power by using reselling.

The results in Proposition 1 are consistent withobserved selling formats in certain industries. Specif-ically, we find that agency selling should be preferredunder a negative or small positive cross-effect. Inthe travel industry, where we can expect a negativeeffect on traditional sales (e.g., sales through travelagents or hotels) as online sales increase, we see thatmost online travel sites such as Expedia and Traveloc-ity have agency selling agreements with airlines andhotels, which set retail prices. In the case of books,where the cross-effect is small positive (Hilton andWiley 2010), our model again predicts that we shouldobserve agency selling, which is indeed the formatused by competing e-book stores run by Amazonand Apple. On the other hand, we find that resellingshould be preferred under a positive cross-effect. Inindustries such as music, where there are strong stim-ulation effects (e.g., more song and album sales leadto more revenues to artists from other channels, suchas more airplay and concert collections), we observethat most online music stores, e.g., iTunes and Spo-tify, have a reselling agreement with artists and recordlabels.9

9 We note that these industries have complicated market structuresand our model is a simplification. Nevertheless, these examples dooffer directional support for the insights from our model.


We now discuss how the selling formats in thee-channel vary with the intensity of competition inthe market. In the presence of competing e-tailers,an advantage of agency selling is that retail pricesare jointly set for both retailers by the manufac-turer, which cushions competition at the retail level.In addition, the e-tailers are Stackelberg leaders inthe market, which gives them more control over thechannel. Thus, we expect to see that as competitionincreases, agency selling will be preferred more byboth e-tailers. In Figure 3, when � takes medium andlarge positive values, we see that as the value of thecompetition parameter � increases, the e-tailers preferto adopt the agency configuration. (For negative andsmall positive values of � , the e-tailers are already inthe agency configuration for all values of �.) We statethis result in the proposition below.

Proposition 2. As competition in the market increases(i.e., � increases), the e-tailers’ preferences shift from re-selling to agency selling. If the spillover is sufficiently pos-itive, as competition increases, the e-tailers switch from theRR arrangement to the RA and the AA arrangements.

We contrast this result with the results in McGuireand Staelin (1983) and Jerath and Zhang (2010). Inour case, as competition increases, there is a shiftaway from reselling and towards agency selling. InMcGuire and Staelin (1983), where competing manu-facturers decide the channel structure, as competitionincreases, the manufacturers choose reselling (or, intheir terminology, vertical separation). In Jerath andZhang (2010), where a monopolistic retailer decidesthe channel structure, as competition increases, thereis a shift away from the store within a store arrange-ment (which is the brick-and-mortar analogue ofagency selling). Thus, our result complements thefindings in these papers.

Interestingly, under conditions of moderate spill-over, we see a hybrid channel structure even thoughthe e-tailers are ex-ante indistinguishable. Choosingtwo different retail formats also helps reduce the com-petition between the e-tailers,10 and they make moreprofits than they could have if they had both cho-sen the same retail format. Not only does the reseller(e-tailer X) face a lower wholesale fee, but the agencyseller (e-tailer Y ) can also charge higher agency fee.(This explains why the manufacturer never prefers thehybrid arrangement, a fact we use in the next section.)Therefore, when the spillover is moderate, we see anasymmetric configuration in the e-channel.

10 Note that the hybrid channel structure is not seen when there isno market competition, i.e., � = 0, and it gains more prominenceas the market competition increases. This points to the fact that theemergence of this arrangement in equilibrium is driven by compet-itive forces.

3.3. Beneficiaries Under Agency SellingThe agency selling format confers different powersto the manufacturer and the e-tailers under differentconditions, and its adoption might lead to gains orlosses for the players in comparison to the resellingstructure. Given that the retailers decide the equi-librium selling formats, is the manufacturer alwaysworse off under the equilibrium selling formats in thee-channel?

The manufacturer is the Stackelberg leader in thereselling arrangement because it decides the whole-sale prices first, whereas it is a Stackelberg follower inthe agency arrangement because the e-tailers decidethe agency fees first. Therefore, generally speak-ing, the manufacturer prefers the reselling arrange-ment. However, in the presence of spillovers, themanufacturer wants to control the final retail price inthe e-channel, which implies that it has a preferencefor agency selling. Specifically, when the spilloveris positive, the manufacturer gains more by havingpricing power in the e-channel so that he can keepthe price low to stimulate demand in the traditionalchannel and, therefore, he prefers agency selling toreselling. For a fraction of this region, the e-tailers alsoprefer agency selling to reselling, because the gainsfrom channel efficiency are enough to compensate forthe loss of pricing power. In this situation, marked byregion I in Figure 4, the manufacturer and the e-tailersreceive the highest payoff in the agency (AA) configu-ration compared to any other configuration. Further-more, in our model, consumer welfare is greater whenhigher quantities are sold in the market, which hap-pens under the agency structure. Therefore, all marketparticipants are better off with the agency structureunder the conditions marked by region I.

In the other regions in Figure 4, the equilib-rium channel outcome is not what the manufac-turer prefers. In region II, the manufacturer prefers

Figure 4 Manufacturer’s Preference for Agency Selling

0 0.7

1

0

II

I

�

�


the agency (AA) structure, but a hybrid structure isobserved in equilibrium. (As discussed in the previ-ous section, the manufacturer never prefers the hybridchannel arrangement.) Incidentally, if conditions inthe e-book market resemble the market conditionsin this region, this might provide an explanation forpublishers’ insistence for Amazon to adopt an agencyselling format when Apple introduced agency sellingin the iBook Store.

3.4. Agency Fees Under EquilibriumAgency selling in online marketplaces is a new retail-ing innovation, which is only partially understood bythe industry. In this section, we provide some guide-lines for setting fees under agency selling. (Note that,in this section, we analyze the fee for all combina-tions of values of � and � , irrespective of whether ornot agency selling will be practiced in equilibrium.)Specifically, we address how various factors such asspillovers and competition affect the agency fee thatan e-tailer should charge.

Proposition 3. In the AA configuration, the feescharged by the e-tailers are nonmonotonic in � . As � in-creases from its lowest possible value of −1, fees increaseup to the value of � , at which 41 + �∗

AA − �25� = 41 −

�∗AA5�

∗AA� holds. Beyond this value of � , the fees decrease

in � .

The plot of �∗AA is shown in Figure 5. When the

e-channel has a negative cross-effect on the traditionalchannel, the manufacturer is cautious of selling toomuch in the e-channel. With a strong negative cross-effect, the e-tailers need to offer incentives to themanufacturer to sell online instead of selling throughthe traditional channel. They achieve this by settinga low agency fee so that the manufacturer makes ahigher profit per item sold through the marketplacein comparison to the loss incurred in the traditionalchannel because of this sale (i.e., because of � lowersales). As � increases and the negative cross-effect

Figure 5 (Color online) Changes in �∗

AA, the Fee Under AgencyArrangement (AA) with Changes in Spillover Effects 4� 5

–1 10

�AA

1

�

� = 0

� = 0.5

� = 0.9

*

reduces, the e-tailers need to offer smaller incentiveto the manufacturer to sell through the marketplace;therefore, the fee increases in � . As � increases fur-ther and becomes sufficiently positive, the manu-facturer now cares more about the stimulation ofdemand in the traditional channel than the revenuesfrom the e-channel. In this case, the manufacturer canreduce prices in the marketplace to stimulate demandthrough the traditional channel, which would resultin a loss of revenue for the e-tailer offering the mar-ketplace. Therefore, the e-tailer decreases the fee sothat the manufacturer gets a bigger share of the rev-enues and does not decrease retail prices in the mar-ketplace as much.

To summarize, the fee is small for a strong negativecross-effect because the e-tailer wants the manufac-turer to sell more quantity through the marketplace.It is also small for a strong positive cross-effect, how-ever this is because the e-tailer wants higher pricesto be charged in the marketplace, and for interme-diate cross-effect values the fee is larger than at theextremes.

We now study the effect of competition on the fee.It is natural to assume that the e-tailers would reducethe fee with increase in competition. However, wesee that under some conditions the fee might actuallyincrease with the degree of competition.

Proposition 4. In case of negative cross-effect (� < 0),the fee charged under agency selling always decreases in �.In case of positive cross-effect (� > 0), the fee first increasesand then decreases in �.

The above proposition shows that the relation be-tween the fees charged by the e-tailers and the com-petitive intensity in the market (�) is mediated bythe cross-channel spillover effect (�). The first part ofthe proposition states that if � increases, competitionin the market increases. The e-tailers compete for themanufacturer’s business, and they need to decreasethe fee to incentivize the manufacturer to use theirmarketplace. This is the effect for � < 0.

The second part of the proposition, which statesthat for � > 0 the fee, �∗

AA, first increases and thendecreases in �, is the more interesting case. This pat-tern is shown in Figure 6. This nonmonotonic natureof �∗

AA can be explained if we consider the impactof the positive spillover from the e-channel to thetraditional channel. It is clear from the demand sys-tem that, as � increases, the sensitivity of demand toown price (given by 1/41 − �25) increases. In addi-tion, the potential market size, given by 241/41 +�55,decreases. This implies that, all else equal, demandin the e-channel decreases as � increases. Since � > 0,the manufacturer benefits more from the positivespillover from the e-channel to the traditional channelthan from the profits in the e-channel. The decrease


Figure 6 (Color online) Changes in �∗

AA with Market Competition 4�5

1

Stimulation

Cannibalization

0

1

0

� = 0

� = 0.2

� = –0.2

� = 0.8

� = –0.8

�

�AA*

in e-channel demand hurts the manufacturer’s profit,and he reduces prices in the e-channel (as � increases)to mitigate the fall in e-channel demand. The e-tailersforesee this drop in prices and respond by increasingthe fees to get a higher share of the reduced market-place revenues. This explains why in the presence ofpositive spillovers the e-tailers may increase the feeas competition between them increases. However, as� increases further and becomes large enough, thebasic effect of competition between e-tailers domi-nates the effect described above and leads to an over-all decrease in fees in �.11

We note that the nonmonotonic nature of theagency fee, �∗

AA, with respect to the parameters �and �, is preserved in the parameter space in whichthe AA arrangement occurs in equilibrium. This canbe easily verified by jointly viewing Figures 3, 5,and 6. Overall, the model predicts that the agencyfee depends on the nature of a particular industry(specifically, competitive intensity and spillover fromone channel to the other) and should, therefore, varyacross industries.

4. ExtensionsIn this section, we extend our basic model in multi-ple ways. First, we consider a situation in which thefirm can benefit from positive externalities from salesof the product and consider the impact of this factoron pricing and channel structure. Second, we allowthe manufacturer to optimally set price in the tra-ditional channel. Third, rather than assuming a one-directional impact of online sales on offline sales, weallow both purchase decisions to influence each other.

4.1. Positive Externality from Product SalesIn the analysis in §3, we find that agency selling leadsto lower prices than in reselling because it is a moreefficient selling format compared to reselling. Thisinsight may seem contradictory to the pricing practice

11 The agency fee in the hybrid case, �RA, changes in a mannersimilar to �∗

AA.

observed in the e-book industry around the year 2010,where prices of e-books at Amazon increased afterAmazon switched from reselling to agency selling.However, the driving forces discussed in our basicmodel are not inconsistent with this phenomenon. Inthis section, we show that such a phenomenon canoccur when we consider additional firm incentives,thus enriching the basic model.

The e-books industry has gone through manychanges in the last decade, and we do not attempt toexplain all the phenomena using our model. There-fore, for clarity, we first describe the phenomena inthis industry that we address. Following this, we offeran explanation of these phenomena.

In the period around 2006–2009, Amazon was usingthe reselling format for e-books and, in many cases,was selling the e-books even below the wholesaleprice that it paid to publishers. (Specifically, Amazoncharged a retail price of $9.99 for e-books, whereasit typically paid publishers between $12 and $13 forthem (Rich 2009).) To read the e-books, consumershad to purchase the e-reader Kindle, also sold byAmazon. Notably, Amazon was making a healthyprofit on Kindle. In other words, Amazon was sell-ing e-books at a loss while making a profit throughthe paired-up product Kindle. Not surprisingly, thee-book publishers opposed Amazon’s practice of sys-tematically selling their e-books at very low prices(because it hurt their pricing ability in other chan-nels) (Meadows 2009, Rich 2009, Trachtenberg andBlackstone 2009). Then, in the early part of 2010,Apple released the iPad that, among other things,served as an e-reader. Apple also sold e-books frompublishers through an agency selling format. Becauseof competition in the e-reader market (i.e., Kindle ver-sus iPad), Amazon reduced the price of the Kindleover a period of time; in November 2008, the Kin-dle was priced at $399 per unit, and by November2012 the price had declined steadily to $79 per unit(Yarow and Angelova 2011). Concurrently, Amazonstarted moving increasingly to an agency selling setupfor e-books and also stopped selling e-books belowmarginal cost. Put differently, in contrast to its earlierstrategy of using e-books as a loss leader, Amazonwas now using the Kindle as a loss leader and mak-ing a profit on e-books (Ludwig 2011). We now offeran explanation for these events.

In the scenario described above, before the launchof the iPad, the e-tailer’s decisions are driven not onlyby the profits obtained from the focal product (in thiscase, e-books) but also by the impact of the focal prod-uct on other sources of profit—in this case, profitableKindle sales. Therefore, e-tailers may care not onlyabout the profit generated from the product sales inthe e-channel but also about the indirect incremen-tal profits induced by these sales. The externalities


might be driven by several underlying factors suchas increased sales of associated products or long-termbenefits of increased market share. Below we showthat simply adding these externalities to the basicmodel in §2 can explain the above discussed sellingformat choices and pricing decisions of e-tailers.

To incorporate positive externalities imposed byproduct sales, we modify the e-tailers’ profit functionsin the following manner:

�̃CX =�C

X + �qCX and �̃CY =�C

Y + �qCY 1

where the superscript C denotes a particular chan-nel configuration (i.e., RR, RA, AR, or AA) and � ≥ 0denotes the positive externalities from product salesfor the e-tailers. Note that the externalities might beasymmetric; however, the simpler case of symmetricexternalities is sufficient for us to develop the keyintuition. (For completeness, we present analysis forthe asymmetric case in the appendix.) The solutionprocedure for this model is the same as that in §3. Wenow discuss the insights obtained from this model.

Interestingly, if there is positive externality fromproduct sales, situations can arise in which retail priceunder reselling is lower than retail price under agencyselling. The intuition behind this is as follows. Whenan e-tailer receives a positive externality from prod-uct sales, she has a greater incentive to increase hersales, i.e., she wants to sell more with the externalitythan she would want to sell without the externality,ceteris paribus. To sell more quantity, the retail pricescharged to consumers must be lower. Consider thetwo selling formats, agency selling and reselling. Ifthe selling format is agency selling, the e-tailer candecrease prices only by decreasing the agency fee shecharges to the manufacturer, so that the manufacturerin turn reduces the retail price. On the other hand, ifthe selling format is reselling, the e-tailer can directlycontrol retail prices, which gives her greater flexibilityin lowering them to drive demand.

As an example, consider an agency selling situa-tion where the agency fee is zero and the price set bythe manufacturer under agency selling is exactly thesame as the wholesale price he would have chargedunder reselling. If the positive externality increases bya small amount, the e-tailer would want to furtherreduce retail prices to increase sales. However, shecannot do anything to reduce retail prices because theagency fee already equals zero, and since the incen-tives of the manufacturer (who sets the retail pricein this arrangement) have not changed, he does notchange the retail price. On the other hand, in the caseof reselling, the e-tailer can reduce the retail price byany desired amount to increase sales and benefit fromthe positive externality. Therefore, as shown by thisexample, with externalities because of sales, the retail

Figure 7 Price Comparison and Equilibrium Sales Formats for � = 1

1

0–1 10

�

�

–1 10�

1

0

�

AA

RR

RA

RR

(a) In the lighter region, retail pricesunder reselling are lower than retail

prices under agency selling

(b) Selling formats inequilibrium

price under reselling can be lower than the retail priceunder agency selling.

The region where the retail price under resellingis lower than that under agency selling, as a func-tion of � and �, is shown in Figure 7(a). We notethat this is a departure from the result presentedearlier in §3, and this difference arises because ofthe additional positive externality from sales that weincorporate into the basic model in this section. Inter-estingly, for a sufficiently high level of externality, in areselling arrangement, the e-tailer might even reducethe retail price below the wholesale price offered bythe manufacturer.12 This offers a possible explanationfor why Amazon was selling e-books below wholesaleprice in the reselling arrangement when Kindle priceswere high but switched to agency selling, which alsoled to an increase in e-book prices, as Kindle pricesreduced.

The equilibrium selling formats with positive exter-nalities are shown in Figure 7(b). Comparing withFigure 3, we see that the the agency selling for-mat is more prevalent (the agency (AA) and hybrid(RA) arrangements cover larger areas), whereas thewholesale (RR) arrangement is less prevalent. Thisis because, in general, as explained earlier, pricesare lower and sales are higher under agency sell-ing, which is to the benefit of the e-tailers with posi-tive externalities from sales. Interestingly, we see thatfor large negative � the RR arrangement prevails(whereas without positive externalities from sales, theAA arrangement prevailed in this region). The rea-son for this change in equilibrium channel structureis that a large negative effect on traditional sales frome-channel sales hurts the manufacturer. Under the AA

12 The level of externality that induces the e-tailers(s) to adoptthis strategy is given by �∗

RR = 441 − �541 + �55/43 − �5 and �∗

RA =

441 − �2542 + �2 − 2�∗

RA54241 − �541 + �5− 4241 + �5− �43 + �55�∗

RA −

��∗

RA255/4241 − �∗

RA54641 − �∗

RA5 − �246 − 46 − �∗

RA5�∗

RA555, where �∗

RR

and �∗

RA are the thresholds for the wholesale and hybrid cases,respectively.


arrangement, the manufacturer sets the retail pricesin the e-channel to be high to limit e-channel sales,thereby limiting the negative effect on its traditionalchannel sales. However, the e-tailers want high e-channel sales to benefit from the positive externali-ties. Therefore, the e-tailers prefer to take control ofthe retail pricing in the e-channel by adopting the RRarrangement for large negative � , and they charge alower retail price than what the manufacturer wouldcharge in the agency-selling arrangement. The follow-ing proposition summarizes the above results.

Proposition 5. In the presense of positive externali-ties because of sales through the e-channel on the e-tailer’sprofit,

(a) prices under reselling may be lower as compared toprices under agency selling;

(b) for any value of �, for large negative � , the wholesale(RR) arrangement emerges in equilibrium.

Before we conclude this section, we note that, in thelong run, such a positive externality may reduce inmagnitude or altogether disappear, either because ofmarket saturation or because of shrinking profit mar-gin from the complementary products. For instance,after the advent of the iPad and other e-readers inthe market, Amazon’s profit margins for the Kindlehave reduced significantly, so that the positive exter-nality from Kindle sales has reduced. In this case, ourmain model in which e-tailers maximize profit willbe more relevant in determining long-run market out-comes. Nevertheless, it is useful to understand the(short-term) impact of such externalities on marketoutcomes.

The main message of this section is the follow-ing. Whereas agency selling incentivizes a lower retailprice because of channel efficiency, this does not nec-essarily imply that retail price under reselling cannotbe lower. In fact, as the preceding analysis illustrated,if there are external influences on the incentives of theplayer who sets the retail price, then reselling couldindeed lead to a lower retail price.

4.2. Endogenous Pricing in theTraditional Channel

In the analysis in §3, we assumed that the prices in thetraditional channel are exogenously fixed. In this sec-tion, we introduce a pricing stage for the traditionalchannel, which allows prices in the traditional chan-nel to be determined endogenously and to depend onthe outcome of the e-channel.

Suppose that in the traditional channel the manufac-turer sells his product through a reselling arrangementwith a traditional retailer. He charges the traditionalretailer a wholesale price wo and the traditionalretailer sets a retail price po. We assume that thedemand function for the traditional channel is given

by qo = Q̄+ �qE − po. This is equivalent to assumingthat the spillover from the e-channel changes the basedemand in the traditional channel (the base demandbeing given by Q̄+�qE). Solving for the optimal retailand wholesale prices set by the traditional retailer andthe manufacturer, respectively, we find that the man-ufacturer’s profit is given by

�M =4Q̄+ �qE5

2

8+�ME1 (5)

as derived in the appendix. qE and �ME are the quan-tity sold and the profit obtained, respectively, fromthe outcome in the e-channel. The above expressionimplies that quantity sold in the e-channel impactsthe manufacturer’s price and profit in the tradi-tional channel, which, in turn, can affect the equilib-rium configuration in the e-channel. We solve for theequilibria in this scenario using backward induction,and Figure 8 shows the selling formats observed inthis case.

By comparing Figures 8 and 3, we see that evenwhen the prices in the traditional channel are deter-mined endogenously, we observe outcomes qualita-tively similar to those described in Proposition 1.Specifically, we observe the agency structure for neg-ative and small positive � , the hybrid structure formedium positive � , and the wholesale structure forlarge positive � . The slight difference is that a smallerpositive spillover (�) is sufficient for the retailersto switch from an agency structure to a wholesalestructure.

It may be the case that costs in the traditional chan-nel are higher, such that the retailer incurs a marginalretailing cost and the manufacturer incurs a marginalproduction cost for goods sold via the traditionalchannel. For example, in the case of a printed book,the manufacturer incurs a significant cost of printingit, and the retailer incurs a significant cost of stocking

Figure 8 Channel Structure When the Manufacturer Sells Through aRetailer and the Prices on the Traditional Channel AreEndogenously Determined 4Q̄= 45

AA RR

RA

–1 10�

1

0

�


Figure 9 (Color online) Illustrative Figures for Analysis with Endogenous Consumer Choice Across Channels

–0.6 0.60

1

0

AA RR

RA

(a) Equilibrium channel configuration (b) Variation of agency fee with � (c) Variation of agency fee with �

�

�

–0.6 0.6�

� = 0

� = 0.7 � = 1

� = 0.4

1

�AA*

10 �

1

0� = –0.7

� = –0.2

� = 0.2

� = 0

� = 0.7

�AA*

it in store. Incorporating these costs does not changethe results qualitatively. Furthermore, if the manu-facturer sells directly to consumers in the traditionalchannel (rather than through a retailer) and endoge-nously decides prices, our results are again qualita-tively unchanged.13 The main findings of our modelcontinue to hold even if the timing of the game ischanged, i.e., the prices in the traditional channel aredecided simultaneously or earlier than prices in thee-channel.

4.3. Endogenous Consumer Choice AcrossChannels

In the analysis in §3, we assumed for simplicity thatthe magnitude of the cross-channel effect (i.e., theeffect of online sales on offline sales) is exogenous.Here, we model the cross-channel effect by modify-ing the utility function of the representative consumersuch that the quantities consumed in the online andoffline channels are decided jointly. Specifically, weuse the following consumption utility function:

U4qX1 qY 1 qT 5 = qX + qY + qT −12q

2X −

12q

2Y

−12q

2T −�qXqY +�qEqT 1

where qE = qX + qY denotes the total quantityconsumed in the online channel. In this utilityspecification, � , the coefficient of the term qEqT , is ameasure of the nature and magnitude of the interac-tion between the sales in the traditional channel andthe sales in the electronic channel. If � > 0, then thereis complementarity between the two channels and if� < 0, then there is substitutability between the twochannels. Using the above, we obtain the followinglinear demand system:

qX =

(

41 +�5−1 −�2

1 −�pX +

�−�2

1 −�pY −�pT

)

ì1

13 In this case, the manufacturer’s profit is given by 4Q̄+ �qE52/4 +

�ME . On comparing this to (5), it is clear that the results in the twoscenarios will be qualitatively similar.

qY =

(

41 +�5−�−�2

1 −�pX +

1 −�2

1 −�pY −�pT

)

ì1

qT = 441 +�+ 2�5−�pA −�pB − 41 +�5pT 5ì1

where ì=1

1 +�− 2�20

Given the above demand system, the model canbe solved using the solution procedure in §3. Theanalysis is analytically intractable, so we resort to anumerical solution. The equilibrium selling format fordifferent values of � and � is shown in Figure 9(a).It is clear on comparing Figures 9(a) and 3 that theresults for the equilibrium selling formats in the cur-rent model with endogenous consumer choice acrosschannels (where � > 0 denotes complementarity and� < 0 denotes substitutability between the channels)are qualitatively the same as those in the main modelwith an exogenous cross-channel effect (where � > 0denotes complementarity and � < 0 denotes substi-tutability between the channels). In other words, theresults corresponding to Propositions 1 and 2 hold inthis case.

Next, we show the variation of the agency fee with� in Figure 9(b). This figure corresponds to Figure 5(with � on the x-axis instead of �). It is easy to seethat this figure is very similar to Figure 5—the feefirst increases in � and then decreases. This finding isconsistent with Proposition 3.

Finally, in Figure 9(c), we show the variation in theagency fee with �. This figure corresponds to Fig-ure 6. We observe that, if � ≤ 0, the fee monotonicallydecreases in �, and if � > 0, the fee first increasesand then decreases in �. In other words, the results ofProposition 4 are qualitatively replicated.

The key difference between the model used for thisanalysis and the main model is that, in the mainmodel, for simplicity, we had assumed an exogenousand unidirectional impact of online sales on offlinesales, whereas here we develop the demand system


from first principles. As shown above, the key insightsobtained from this extended model are the same asthose obtained from the main model. This lends con-fidence to our results and shows that our main modelis an appropriate and acceptable simplification.

5. Discussion and ConclusionsIn recent years, we have seen significant growth inthe use of agency selling in e-commerce across indus-tries. Given the proliferation of agency selling, thereis a need to understand the conditions under which e-tailers should adopt them, especially keeping in mindthe impact of sales in the e-channel on sales in tra-ditional retail channels through which manufacturersalso sell their products. In this paper, we identify con-ditions under which agency selling should be usedand the implications of using them on various marketparticipants, i.e., competing e-tailers, manufacturers,and consumers.

Our analysis suggests that agency selling is an effi-cient selling format, but whether an e-tailer shoulduse it or not depends on the extent of the demandspillover between the electronic channel and thetraditional channel, and on the level of competi-tion between e-tailers. Specifically, e-tailers shoulduse agency selling when spillovers from the elec-tronic channel to the traditional channel are negative,absent, or small positive. This result is anecdotallysupported by the observation that agency selling isused by travel websites in the hotel industry (suchas Expedia and Travelocity), where we can expect asignificant negative effect on traditional sales (e.g.,sales through travel agents) as online sales increase.As another example, the spillover effect for books hasbeen found to be small positive (Hilton and Wiley2010), and large e-book sellers such as Amazon andApple indeed use the agency model. On the otherhand, setting up marketplaces and using agency sell-ing implies giving control over retail pricing to themanufacturers. When sales in the e-channel have alarge stimulation effect on sales in the traditionalchannel, agency selling might be detrimental fore-tailers because manufacturers have the incentiveto reduce e-channel prices, which will reduce thee-tailers’ revenues as well. In accordance with thisprediction, we find that reselling is the dominant sell-ing format used by online stores in the music industry(such as iTunes and Spotify), where there are strongstimulation effects on offline sales (e.g., more onlinesales of music singles lead to more sales of album CDsand concert tickets). These examples lend anecdotalsupport to our results.

Furthermore, e-tailers have greater incentive to useagency selling when there is increased competitionin the market, because they can control the marketbetter by virtue of becoming Stackelberg leaders and

also partially mitigate retail competition by allow-ing the manufacturer to jointly set retail prices acrosscompeting e-tailers. Interestingly, we might notice ahybrid configuration in certain industries as the e-tailers adopt different selling formats to moderateintra-format competition.

We also provide directions for setting agency fees.We find that the fee in agency selling should dependboth on the competition in the market and on thespillover that the e-channel has on traditional retail-ing. The fee charged for the use of the market-place is small when the absolute magnitude of thespillover is large. As this magnitude of the spilloverbecomes smaller, implying that there is a reductionin the interaction between the two channels, the feeincreases. Surprisingly, our model also shows thatthe fee might not always decrease with competition.When the spillover is positive, the fee might actu-ally increase with increasing competition between thee-tailers because of the demand-stimulation effect.

Our model also provides some broader implicationsfor the future of e-commerce. Currently, e-commerceis a small fraction of overall retailing, and we canexpect both a positive or a negative spillover fromonline sales into traditional retailing. As e-commercecontinues to grow, one can expect a negative spillovereffect to dominate. In this case, our model predictsthat we will see increased adoption of agency sellingby online retailers. Furthermore, our model also pre-dicts that if e-commerce gets more competitive overtime, agency selling will increase.

We extend our basic model to assume that e-tailersmay benefit from positive externalities across productsales. In this case, e-tailers may have the incentive toreduce prices to increase sales while sacrificing directprofits from the focal product. When this holds, wefind that e-tailers may prefer the reselling arrange-ment to maintain direct control over retail pricesand reduce them appropriately to increase sales. Inthis case, contrary to the result of the main model,retail prices in reselling may be lower than underagency selling. This scenario with positive externali-ties provides an explanation for why Amazon, whichsells both e-books and the e-reader Kindle, preferredreselling with low prices to agency selling when Kin-dle prices were high.

We assume in the model that the manufacturerdirectly sells through the e-tailers and the traditionalchannel. This is indeed true for a number of manu-facturers. For instance, many e-book publishers selldirectly on Amazon in agency selling and manufac-turers such as Lee, Kate Spade, Burberry, Lacoste, andKimberly Clark sell directly on Amazon Marketplace.However, a large number of sellers on Amazon Mar-ketplace (and other e-tailing platforms) are not man-ufacturers but are third-party resellers themselves. Inthis case, the insights from our model would be valid


as long as the third-party reseller (i.e., the upstreamagent) also sells in a traditional channel (which isindeed true for many sellers), and there are demanddependencies between the channels.

Our approach has several limitations, which pro-vide interesting directions for future research. First,we assume that firms do not explicitly influence howsales in the online and offline channels influence eachother. It might be interesting to ascertain under whatconditions manufacturers would want to promotecomplementarities between the two channels andunder what conditions they would want sales in onechannel to reduce demand in the other. In a similarvein, when a manufacturer sells different products, hecan make decisions endogenously about which prod-ucts to sell through different channels. Future workcan advance our research by micro-modeling thesefactors. Second, we find that, under some conditionse-tailers choose to not offer agency selling (in spiteof their efficiency) because in the presence of cross-channel spillovers they do not want to give pricingpower to manufacturers. This indicates that e-tailerscould benefit from using agency selling if they couldimpose restrictions on the retail prices that manufac-turers charge. The “Most Favored Nation” agreementthat Apple made with publishers of e-books is anexample of this (Palazzolo 2013). Under this agree-ment, the publishers could not sell the same e-bookat another competitor at a price lower than at Apple’se-book store. Another example is that Apple does notgive complete freedom to publishers to set prices inagency selling, and they can only choose from differ-ent pricing levels. In a similar vein, whereas Amazonuses agency pricing for e-books, there is compellingevidence that it attempts to influence the prices thatpublishers can charge (Amazon Books Team 2014).Studying such contracts in detail, which are similarto vertical price restraints used when manufactur-ers control the channel (Mathewson and Winter 1984,Mathewson and Winter 1998), could be an interest-ing and important area for future research. Third, wehave not modeled how multiple manufacturers com-pete on retail marketplaces. This could be an interest-ing direction for future research. Fourth, we assume aparticular order of moves for the players in the modelthat is reasonable and justifiable. However, variationin the order of moves is possible. We consider onesuch variation in the timing of the game; specifi-cally, in the asymmetric arrangement in which onee-tailer chooses reselling and the other chooses agencyselling, we analyze the situation when the manufac-turer is the retail price leader. We find that our maininsights are robust and remain unchanged in this case,but there can also be some interesting differences inthe outcomes. For instance, the asymmetric case doesnot occur in equilibrium (more details are availablein the appendix). Future research can explore this in

more detail. Fifth, novel logistics-related agreementsare being offered by e-tailers to manufacturers (suchas “Fulfillment By Amazon,” under which the manu-facturer decides the retail price in the marketplace butthe e-tailer takes over logistics activities such as ware-housing and delivery). Incorporating logistics-sharingaspects into our framework might lead to interestingimplications. In this context, it may be important tomodel uncertainty in demand and understand howdifferent risk-sharing characteristics of the differentselling formats impact the e-tailers’ and the manufac-turers’ decisions.

E-commerce presents an exciting area of research,and we believe that this is the first paper that ad-dresses the consequences of demand spillovers be-tween the e-channel and the traditional channel onthe selling formats that emerge in electronic retailing.We hope that the ideas presented in this paper moti-vate interesting future research.

AcknowledgmentsThe authors thank the anonymous reviewers and the par-ticipants of the Workshop on Information Systems and Eco-nomics, Marketing Science Conference, and the Conferenceon Information Systems and Technologies for several help-ful suggestions. This research was partially funded by theMack Center for Technological Innovation.

Appendix

A.1. Proofs for §3

Proofs of Propositions 1 and 2. We use backward induc-tion to solve for the equilibrium channel configuration.First, conditional on the channel structure (RR, RA, AA),we derive the equilibrium outcomes like the demand, retailprices, wholesale price, and agency fees using backwardinduction for a specific subgame. These outcomes are pre-sented in §3.1. Second, the equilibrium channel arrange-ment for every pair of � and � is determined by solving forthe Nash equilibrium (NE) of the 2 × 2 game presented in§3.2. Note that, without loss of generality, we assume thatY uses agency selling in the RA configuration. Solving forthe NE, the platform arrangement (AA) is observed when�AA

X ≥ �RAX , the hybrid arrangement (RA) when �RA

X ≥ �AAX

and �RAY ≥ �RR

Y and the wholesale agreement (RR) when�RR

Y ≥�RAY . It is easy to see that these conditions are mutu-

ally exclusive and there is a unique equilibrium channelstructure for every pair of � and �. Since �∗

RA is not analyt-ically tractable, we resort to a numerical analysis to derivethe equilibrium configuration presented in Figure 2.

Proof of Proposition 3. To show the variation in theplatform fee w.r.t. � , we first compute the derivative of e-tailer X’s profit w.r.t. the platform fee (d�AA

X /d�X) and thenequate the fee charged by both e-tailers because of the sym-metric equilibrium, which gives

d�AAX

d�X

∣

∣

∣

∣

�X=�Y =�

=41−�53 +4−2+�25�2 +41−�5�42�+�5−41−�52�4�+2�5

441−�534−1+�250


Let �4�5 = 441 − �53 + 4−2 + �25�2 + 41 − �5�42�+ �5− 41 −

�52�4�+2�55/4441−�534−1+�2550 The first-order condition.implies that �4�∗

AA5 = 0 . Using implicit differentiation, weget

¡�∗AA

¡�=−

¡�

¡�

/

¡�

¡�∗AA

=241−�∗

AA54�∗AA

2�+�−�2�−�∗AA4�−�55

4242+�∗AA5−3�25�2 +41−�∗

AA52��∗

AA2+2�∗

AA2��−2��

0

Substituting the optimal fee (�∗AA) in this equation, it is easy

to show that the denominator is always greater than 0 forthe domain of parameters, which implies ¡�∗

AA/¡� ≥ 0 if� ≤ ��∗

AA41 − �∗AA5/41 + �∗

AA − �25, and ¡�∗AA/¡� < 0 other-

wise. Let g4�5 = ��∗AA41 − �∗

AA5/41 + �∗AA − �25. The afore-

mentioned conditions state that �∗AA is increasing in � when

� ≤ g4�5 and decreasing when � > g4�5. Using Banach fixedpoint theorem, we can show that for every �, there is aunique �∗ s.t. � ≤ g4�5 ∀ � ≤ �∗ and � > g4�5 otherwise.Hence, �∗

AA increases in � ≤ �∗ and decreases thereafter.

Proof of Proposition 4. Following the previous proof,we implicitly differentiate �∗

AA w.r.t. �, which gives us

¡�∗AA

¡�=−

¡�

¡�

/

¡�

¡�∗AA

=241−�∗

AA54�41−�2 +�∗AA5−�∗

AA�41−�∗AA55

4242+�∗AA5−3�25�2 +41−�∗

AA52��∗

AA2+2�∗

AA2��−2��

0

As shown in the previous proof, the denominator is alwayspositive. Hence, the sign of ¡�∗

AA/¡� depends on the sign ofthe numerator. When � ≤ 0, the numerator is always neg-ative implying that �∗

AA is monotonically decreasing in �.For � > 0, ¡�∗

AA/¡�> 0 when

�≤

√

41 −�∗AA5

2�∗AA

2 + 441 +�∗AA5�

2 − 41 −�∗AA5�

∗AA

2�1

and ¡�∗AA/¡� ≤ 0 otherwise. We can apply Banach fixed

point theorem, in the manner illustrated in the proof forProposition 3, to show that there is a unique �∗ such that�∗AA increases in �≤ �∗ and decreases otherwise.

A.2. Analysis with Asymmetric ExternalitiesIn this section, we analyze a case where e-tailers are asym-metric in terms of the externalities from product sales. Weuse the following profit functions for the e-tailers:

�̃CX =�C

X + �1qCX and �̃C

Y =�CY + �2q

CY 1

where the superscript C denotes a particular channel con-figuration (i.e., RR, RA, AR or AA) and �1 and �2 denotethe positive externalities from product sales for the e-tailers.For the sake of simplicity, we assume that �1 > �2 and �2 = 0,i.e., only e-tailer X receives the positive externality. Underthis scenario, e-tailer Y cares only about profits throughthe e-channel, whereas e-tailer X cares about both profitsand sales. Different incentives drive different selling formatchoices by the e-tailers. This asymmetric model is analyti-cally intractable, so we derive the equilibrium configurationnumerically. We present the results in Figure A.1 for �1 = 1;the insights are qualitatively the same for other values of�1 > 0. We obtain the following proposition, which com-pares the equilibrium configurations in this case to those inthe case without externalities from sales.

Figure A.1 Channel Structure When Only One Manufacturer Receivesthe Positive Externality 4�1 = 11 �2 = 05

AA

RR

RA

AA

RA

RA

1

010–1

�

�

Proposition. Compared to the case without externalitiesfrom sales, for large values of � , agency selling is more dominant,as shown in Figure A.1. However, when the value of � is largenegative, a hybrid arrangement is observed in equilibrium, wheree-tailer X is a reseller and e-tailer Y is an agency seller.

As we mentioned in Proposition 1, the primary reason forthe emergence of the wholesale (RR) arrangement for largepositive values of � was the misalignment of incentivesbetween the manufacturer and the e-tailers—the manufac-turer wanted to maximize sales on the e-channel to drivehigher sales in the traditional channel, whereas the e-tailerswanted to maximize their e-channel profits. The presence ofthe positive externality from product sales aligns the incen-tives of the manufacturer and e-tailer X, because both ofthem want to maximize sales through the e-channel when� > 0. Therefore, in this context, the value of reselling asa strategy for the e-tailer to solve the misalignment issuedecreases. On the other hand, for large negative values of � ,the manufacturer needs a higher incentive to increase saleson the e-channel. Since the agency fee under agency sell-ing is already very low for these values of � , the onlyway e-tailer X can incentivize the manufacturer is througha higher wholesale price under reselling. However, sincee-tailer Y ’s incentives have not changed, she continues touse agency selling. Consequently, we observe the hybridstructure (RA) in equilibrium.

A.3. Derivation of Equation (5) in §4.2If the traditional retailer faces wholesale price wo andcharges retail price po, her profit is �R = 4Q̄ + �qe − po5 ·

4po − wo5. The optimal retail price is po = 4Q̄ + �qe + wo5/2,and she sells a quantity qo = 4Q̄+�qe −wo5/2. The manufac-turer’s profit from both channels is, thus, �M = 4Q̄ + �qe −

wo5wo/2 +�E . He sets a profit maximizing wholesale pricegiven by wo = 4Q̄+ �qe5/2 for the traditional retailer, whichleads to an overall profit of �M = 4Q̄+ �qe5

2/8 +�E .

A.4. Manufacturer as the Price Leader in theRA Configuration

In §3.1.2, we had assumed the following timing in theRA channel structure: First, the e-tailer willing to enter


into an agency selling agreement announces the agency fee�. Subsequently, the manufacturer decides the wholesaleprice. Finally, e-tailer X and the manufacturer simultane-ously decide retail prices, and then demand is realized.

Here, we consider a situation when the timing of thegame is slightly altered. We assume that the manufactureris the price leader in the RA case. Specifically, the orderof moves is the following. First, e-tailer Y declares theagency fee. Next, the manufacturer sets both the wholesaleprice (w) to be charged to e-tailer X, and retail price (pRAY )at retailer Y simultaneously. Finally, e-tailer X sets the retailprice. Note that in the original analysis both pRAX and pRAYwere decided simultaneously. In this case, the manufacturersets the retail price at Y before e-tailer X sets its retail price,which gives it more market power. We use backward induc-tion to derive the equilibrium outcomes. Given w and pRAY ,e-tailer X sets a retail price equal to

pRAX =1 +w−�41 − pRAY 5

20

Given �, the manufacturer simultaneously decides thewholesale and retail prices, which are as follows:

w = 441 −�58441 −�5+ 242 −�541 −�5�+��2 − 6441 −�5

+ 242 −�5�+��27�95 · 4841 −�541 −�25−�2�25−11

pRAY =41 −�56441 +�541 − �5−�44 +�43 − �557

841 −�541 −�25−�2�20

Subsequently, we derive the optimal fee (�) set by e-tailer Y .The optimal fee cannot be derived analytically (similar tothe main model), and we resort to a numerical analysis.

Using the above solution for the RA configuration, andthe results from the main paper for the RR and AA arrange-ments, we solve for the equilibrium configuration. Thisequilibrium configuration is presented in Figure A.2.

From the figure, we can see that our main insightsfrom §3 continue to hold. However, the hybrid configura-tion is no longer observed in equilibrium. Since the man-ufacturer is the price leader, and hence has more marketpower, an e-tailer no longer has an incentive to adopt the

Figure A.2 Channel Structure When the Manufacturer Is the PriceLeader in the RA Configuration

AA RA

1

010–1

�

�

reselling arrangement in equilibrium when the competinge-tailer chooses the agency selling arrangement. Therefore,the asymmetric outcome does not appear in equilibrium. Inother words, the increased market power of the manufac-turer dominates the reduction in the retail-level competition(that results when one e-tailer resells).

A.5. Analysis with Alternative Demand FunctionIn this appendix, to address the concern that our results arenot driven by the specific demand function, we choose inEquation (1) in this paper, we consider a demand systemfrequently used in the literature (e.g., Raju et al. 1995), i.e.,

qX = 1 − pX + �4pY − pX5 and

qY = 1 − pY + �4pX − pY 51 where �> 00

Note that as � increases, the level of competition in the mar-ket goes up and the consumers become more sensitive to theprice differential between the two e-tailers. In this manner,� is analogous to � in the original model, although they areon different scales—� ∈ 40117, whereas � ∈ 401�5.14 Further-more, the market size is constant in this formulation unlikein Equation (1), where the market size changes with �. Inthe subsequent discussion, we show that our major insightscontinue to hold even with this general demand system andare not driven by our specific demand systems specified inEquation (1).

We continue to use the same timing of the game as pre-sented in §3. The solution of the channel game leads to theequilibrium channel configuration depicted in Figure A.3(a).This figure corresponds to Figure 3 in this paper. We canclearly see that Propositions 1 and 2 continue to hold qual-itatively (with slight differences in the exact shapes of theregions with different equilibria).

Next, we analyze the agency fee charged in equilibrium.We show the variation of the agency fee with � in the Fig-ure A.3(b), which corresponds to Figure 5 in this paper. Itis easy to see that this figure is very similar to Figure 5. Thefee first increases in � and then decreases. This finding isconsistent with Proposition 3 in this paper.

Figure A.3(c), demonstrates the variation in the agencyfee with � in the AA arrangement. The figure correspondsto Figure 6 in this paper and shows that the fee monotoni-cally decreases in � when � ≤ 0, and first increases and thendecreases in � when � > 0. Therefore, Proposition 4 in thispaper continues to hold. This figure is very similar to Fig-ure 6 in this paper. However, we do not see the agency feeconverging to 0, as � is small. For large values of �, the feeconverges to 0. (Note that large values of � are similar tovalues of � close to 1.)

The above analysis shows that our results are not drivenby our choice of the functional form. Furthermore, this anal-ysis also alleviates the concern that some of our resultsmight be driven by a decrease in market size (in the mainmodel) as � increases. All our main results continue to hold,

14 Loosely speaking, large values of � have a competitive effectsimilar to values of � close to 1. To see this, note that the orig-inal demand function used in this paper can be written as qX =

41/41 + �5541 − pX + 4�/41 − �554pY − pX55, and compare with qX =

1 − pX + �4pY − pX5.


Figure A.3 (Color online) Illustrative Figures for Analysis with the Alternative Demand System

(b) Variation of agency fee with � (c) Variation of agency fee with �(a) Equilibrium channel configuration

–1 10

1

0

AA

RA

RR

�

�

� = 0

� = 1

� = 5

–1 10�

1

�AA*

300

�

� = 0

� = 0.2

� = –0.2

� = –0.8

� = 0.8

1

0

�AA*

even when the market size is held constant indicating thatthe main forces driving our results are indeed the cross-effect and e-tailer competition.

ReferencesAnderson SP, de Palma A, Thisse J-F (1992) Discrete Choice Theory

of Product Differentiation (MIT Press, Cambridge, MA).Amazon (2014) Amazon enjoys record-setting year for marketplace

sellers. Amazon PR (January 9), http://phx.corporate-ir.net/phoenix.zhtml?c=176060&p=irol-newsArticle&ID=1889317.

Amazon Books Team (2014) Update re: Amazon/hachette businessinterruption. Kindle Forum (July 29), http://www.amazon.com/forum/kindle/ref=cm_cd_tfp_ef_tft_tp?_encoding=UTF8%20&cdForum=Fx1D7SY3BVSESG&cdThread=Tx3J0JKSSUIRCMT.

Balasubramanian S (1998) Mail versus mall: A strategic analysis ofcompetition between direct marketers and conventional retail-ers. Marketing Sci. 17(3):181–195.

Bernheim BD, Whinston MD (1985) Common marketing agencyas a device for facilitating collusion. RAND J. Econom. 16(2):269–281.

Bonanno G, Vickers J (1988) Vertical separation. J. Indust. Econom.36(3):257–265.

Bonfils M (2012) Taobao SEO: A guide to one of the world’s largestconsumer marketplaces. Search Engine Watch (August 24),http://searchenginewatch.com/article/2200749/Taobao-SEO-A-Guide-to-One-of-the-Worlds-Largest-Consumer-Marketplaces.

Brynjolfsson E, Hu Y, Rahman MS (2009) Battle of the retailchannels: How product selection and geography drive cross-channel competition. Management Sci. 55(11):1755–1765.

Cachon GP, Kok AG (2010) Competing manufacturers in a retailsupply chain: On contractual form and coordination. Manage-ment Sci. 56(3):571–589.

Carr D (2012) Book publishing’s real nemesis. New York Times(April 15), http://www.nytimes.com/2012/04/16/business/media/amazon-low-prices-disguise-a-high-cost.html.

Chiang WK, Chhajed D, Hess JD (2003) Direct marketing, indi-rect profits: A strategic analysis of dual-channel supply-chaindesign. Management Sci. 49(1):1–20.

Choi SC (1991) Price competition in a channel structure with a com-mon retailer. Marketing Sci. 10(4):271–296.

Coughlan AT, Wernerfelt B (1989) On credible delegation byoligopolists: A discussion of distribution channel management.Management Sci. 35(2):226–239.

Desai P, Koenigsberg O, Purohit D (2004) Strategic decentralizationand channel coordination. Quant. Marketing Econom. 2(1):5–22.

Dixit AK, Stiglitz J (1979) Monopolistic competition and optimumproduct diversity: Reply. Amer. Econom. Rev. 69(5):961–963.

Geylani T, Dukes AJ, Srinivasan K (2006) Strategic manufacturerresponse to a dominant retailer. Marketing Sci. 26(2):164–178.

Goolsbee A (2001) Competition in the computer industry: Onlineversus retail. J. Indust. Econom. 49(4):487–499.

Hagiu A, Wright J (2014) Marketplace or reseller? Management Sci.61(1):184–203.

Hilton J, Wiley D (2010) The short-term influence of free digitalversions of books on print sales. J. Electronic Publishing 13(1),http://dx.doi.org/10.3998/3336451.0013.101.

Iyer G (1998) Coordinating channels under price and non-pricecompetition. Marketing Sci. 17(4):338–355.

Iyer G, Villas-Boas M (2003) A bargaining theory of distributionchannels. J. Marketing Res. 40(1):80–100.

Jerath K, Zhang ZJ (2010) Store within a store. J. Marketing Res.47(4):748–763.

Jeuland AP, Shugan SM (1983) Managing channel profits. MarketingSci. 2(3):239–272.

Jiang B, Jerath K, Srinivasan K (2011) Firm strategies in the “mid-tail” of platform-based retailing. Marketing Sci. 30(5):757–775.

Kwark Y, Chen J, Raghunathan S (2012) Platform or wholesale:Different effects on retailers of online product reviews. ThirtyThird Internat. Conf. Inform. Systems, Orlando.

Ludwig S (2011) Loss leader alert: $79 Amazon Kindle costs $84to build. Venturebeat (November 10), http://venturebeat.com/2011/11/10/amazon-79-kindle-costs-84-to-build/.

Mathewson F, Winter R (1984) An economic theory of verticalrestraints. RAND J. Econom. 15(1):27–38.

Mathewson F, Winter R (1998) The law and economics of resaleprice maintenance. Rev. Indust. Organ. 13(1–2):57–84.

McGuire TW, Staelin R (1983) An industry equilibrium analysis ofdownstream vertical integration. Marketing Sci. 2(2):161–191.

Meadows C (2009) Amazon losing money on $9.99 e-books.TeleRead (May 13), http://www.teleread.com/ebooks/amazon-losing-money-on-999-e-books/.

Moorthy KS (1988) Strategic decentralization in channels. MarketingSci. 7(4):335–355.

Mortimer JH, Nosko C, Sorensen A (2012) Supply responses to digi-tal distribution: Recorded music and live performances. Inform.Econom. Policy 24(1):3–14.

Palazzolo J (2013) Apple ruling heaps doubt on “MFN” clauses.Wall Street Journal (July 14), http://online.wsj.com/article/SB10001424127887323664204578605880157245830.html.

Raju J, Zhang ZJ (2005) Channel coordination in the presence of adominant retailer. Marketing Sci. 24(2):254–262.

Raju JS, Sethuraman R, Dhar SK (1995) The introduction and per-formance of store brands. Management Sci. 41(6):957–978.

Rich M (2009) Steal this book (for $9.99). New York Times (May 16),http://www.nytimes.com/2009/05/17/weekinreview/17rich.html?pagewanted=all&_r=3&.

Sehgal V (2014) Forrester Research online retail forecast, 2013 to2018. Forrester Res. (March 21), http://www.forrester.com/Forrester+Research+Online+Retail+Forecast+2013+To+2018+US/fulltext/-/E-RES115941.

http://phx.corporate-ir.net/phoenix.zhtml?c=176060&p=irol-newsArticle&ID=1889317

http://phx.corporate-ir.net/phoenix.zhtml?c=176060&p=irol-newsArticle&ID=1889317

http://www.amazon.com/forum/kindle/ref=cm_cd_tfp_ef_tft_tp?_encoding=UTF8%20&cdForum=Fx1D7SY3BVSESG&cdThread=Tx3J0JKSSUIRCMT



http://searchenginewatch.com/article/2200749/Taobao-SEO-A-Guide-to-One-of-the-Worlds-Largest-Consumer-Marketplaces

http://searchenginewatch.com/article/2200749/Taobao-SEO-A-Guide-to-One-of-the-Worlds-Largest-Consumer-Marketplaces

http://www.nytimes.com/2012/04/16/business/media/amazon-low-prices-disguise-a-high-cost.html

http://www.nytimes.com/2012/04/16/business/media/amazon-low-prices-disguise-a-high-cost.html

http://dx.doi.org/10.3998/3336451.0013.101

http://venturebeat.com/2011/11/10/amazon-79-kindle-costs-84-to-build/

http://venturebeat.com/2011/11/10/amazon-79-kindle-costs-84-to-build/

http://www.teleread.com/ebooks/amazon-losing-money-on-999-e-books/

http://www.teleread.com/ebooks/amazon-losing-money-on-999-e-books/

http://online.wsj.com/article/SB10001424127887323664204578605880157245830.html

http://online.wsj.com/article/SB10001424127887323664204578605880157245830.html

http://www.nytimes.com/2009/05/17/weekinreview/17rich.html?pagewanted=all&_r=3&

http://www.nytimes.com/2009/05/17/weekinreview/17rich.html?pagewanted=all&_r=3&

http://www.forrester.com/Forrester+Research+Online+Retail+Forecast+2013+To+2018+US/fulltext/-/E-RES115941




Shubik M, Levitan R (1980) Market Structure and Behavior (HarvardUniversity Press, Cambridge, MA).

Singh N, Vives X (1984) Price and quantity competition in a differ-entiated duopoly. RAND J. Econom. 15(4):546–554.

Smith MD, Telang R (2010) Piracy or promotion? The impact ofbroadband Internet penetration on DVD sales. Inform. Econom.Policy 22(4):289–298.

Tiwari A (2014) Flipkart to grow marketplace to compete withAmazon. DNAIndia (April 17), http://www.dnaindia.com/money/report-flipkart-to-grow-marketplace-to-compete-with-amazon-1979084.

Trachtenberg J, Blackstone B (2009) Book price war escalates.Wall Street Journal (October 17), http://online.wsj.com/news/articles/SB10001424052748704322004574477050954174722.

Yarow J, Angelova K (2011) Chart of the day: When will Kin-dles be free? Business Insider (October 17), http://www.businessinsider.com/chart-of-the-day-the-shrinking-price-of-the-kindle-2011-9?IR=T&.

Yoo WS, Lee E (2011) Internet channel entry: A strategic analysis ofmixed channel structures. Marketing Sci. 30(1):29–41.

Zettelmeyer F (2000) Expanding to the Internet: Pricing and com-munications strategies when firms compete on multiple chan-nels. J. Marketing Res. 37(3):292–308.

http://www.dnaindia.com/money/report-flipkart-to-grow-marketplace-to-compete-with-amazon-1979084



http://online.wsj.com/news/articles/SB10001424052748704322004574477050954174722

http://online.wsj.com/news/articles/SB10001424052748704322004574477050954174722

http://www.businessinsider.com/chart-of-the-day-the-shrinking-price-of-the-kindle-2011-9?IR=T&



agency selling or reselling? channel structures in ... · that sales in the e-channel can also...

Documents