Retailer-Driven Product Bundling in a Distribution Channel

Articles in Advance, pp. 1–8ISSN 0732-2399 (print) � ISSN 1526-548X (online) http://dx.doi.org/10.1287/mksc.1120.0725

© 2012 INFORMS

Retailer-Driven Product Bundlingin a Distribution Channel

Hemant K. BhargavaGraduate School of Management, University of California, Davis, Davis, California 95616,

[email protected]

This paper studies product bundling in a distribution channel where a downstream retailer combines com-ponent goods produced by separate manufacturers acting independently. Past literature offers deep insights

about bundling by a single firm whose unit costs are not impacted by choice of selling strategy. But when theretailer bundles goods from separate manufacturers, unit costs for the bundler (retailer) are, being the prices setby the manufacturers, no longer exogenous. This alters the economic balance with respect to bundling. I showthat channel conflicts weaken the case for bundling. Although bundling is better than component selling for theintegrated firm, it is no longer so in the decentralized channel. The culprit is a combination of vertical channelconflict (incentive misalignment with respect to bundle versus component sales) and horizontal conflict (eachmanufacturer wants a higher share of profits from bundle sales), with the latter playing a dominant role. Theycause manufacturers to overprice component goods, weakening the retailer’s incentives to bundle. The compet-itive interplay between firms when one (retailer) merges the prices of several (manufacturers) leads to lowerprofits for all. Price coordination between the firms could partially restore the role of bundling and improve thefirms’ profits as well as consumer surplus.

Key words : product bundling; channel conflicts; composite goods; double marginalizationHistory : Received: November 23, 2010; accepted: May 1, 2012; Preyas Desai served as the editor-in-chief and

Ganesh Iyer served as associate editor for this article. Published online in Articles in Advance.

1. IntroductionProduct bundling involves grouping two or morecomponent goods and selling them at a discount rel-ative to the sum of the component prices. For exam-ple, Microsoft packages multiple office productivityapplications that have stand-alone value into its Officesuite, and McDonald’s restaurants sell Happy Mealbundles. Bundling raises the firm’s profit because con-sumer valuations for the bundle have less dispersion(relative to the mean) than valuations for compo-nent products (Stigler 1963, Adams and Yellen 1976).This demand-smoothing force is stronger when manygoods are being bundled and when the distribu-tions of consumer valuations for component goodsare negatively correlated, but bundling can also beprofitable under positive correlation and is mosteffective when unit marginal costs are low rela-tive to value (Schmalensee 1984, McAfee et al. 1989,Bakos and Brynjolfsson 1999). Recent extensions tobundling theory have covered facets such as cus-tomized bundling (Wu et al. 2008) and bundlingunder competition (Bakos and Brynjolfsson 2000).Venkatesh and Mahajan (2009) provide an excellentrecent survey of the literature and practice.

This paper studies product bundling in a distribu-tion channel. I examine the practice where a down-stream firm (retailer) sells a bundle of component

goods made by multiple and independent upstreamfirms (manufacturers), schematically shown in thefirst panel of Figure 1. This practice occurs widely,including in the travel, technology, media, dining, andentertainment industries. For example, theaters bun-dle multiple entertainment events from independentartists and entertainers into season passes. Travel sitessuch as Expedia bundle products (air transport, carrental, hotel, shows, etc.) from multiple providers.Firms that specialize in assembling products (e.g.,PCs) bundle components from multiple manufactur-ers, as do system integrators (e.g., defense industryfirms such as Raytheon) and information aggregators(e.g., Yahoo!). Bundling strategy can involve eitherpure bundling (selling just the bundle and not thecomponent goods) or mixed bundling.

Unlike the channel structure identified above,extant bundling literature has studied the case wherea single firm makes the component goods, designsa bundling strategy, and sells directly to consumers(as reflected in the second panel of Figure 1). Thereare critical economic differences between the two set-tings. For the single integrated firm, unit costs of thecomponent goods are exogenous and not impactedby the choice of selling strategy. In the vertical chan-nel, however, the unit costs for the firm (retailer)making the bundling decision are, being the prices

1

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Bhargava: Retailer-Driven Product Bundling in a Distribution Channel2 Marketing Science, Articles in Advance, pp. 1–8, © 2012 INFORMS

Figure 1 Different Distribution Structures for Product Bundling

ManufacturerManufacturer1

Manufacturer2

Good1

Good2

Bundle12

Good1

Good2

Bundle12

Good1

Good2

Bundle12

Manufacturer

Good1

Good2

Bundle12

ConsumersConsumers Consumers

RetailerRetailer

Good2

Good1

Notes. The shaded boxes represent the firm making the bundling decision. The cis are unit costs of component goods.

set by the manufacturers, no longer exogenous. Thisseparation of manufacturing and retailing functionsleads to double marginalization (Spengler 1950), a mani-festation of vertical channel conflict, which raises theunit costs underlying the retail-level bundling deci-sion. The cost increase is even greater under bundlingbecause the potential to extract more surplus enticesmanufacturers into seeking a higher share of thegains from bundling. Finally, component prices arealso driven higher by horizontal channel conflict, whichoccurs because of independent price setting by manu-facturers whose goods are eventually combined into acomposite product.1 These two types of channel con-flicts raise the retailer’s unit costs and thus weakenthe case for bundling.

Still, bundling has appeal to all firms. The retailerdesires bundling because the lower dispersion in con-sumer valuations enables it to extract more surplus.Manufacturers desire bundling because it leads tohigher sales of their products. I formally investi-gate the balance between these positive and negativeforces by modeling a decentralized channel with twomanufacturers, each making a single component goodthat may then be packaged into a bundle by a retailer.I compare outcomes under the decentralized chan-nel against those for an integrated firm, which com-bines the manufacturing and retail functions, as alsoagainst a bilateral monopoly (depicted in the third panelof Figure 1), where the retailer sells a bundle of mul-tiple goods from a single manufacturer. I show thatthe basic demand-side force—which makes bundling

1 This conflict was recognized by Cournot (1929), who consideredtwo component goods made by different firms that were combinedto create utility (e.g., “zinc + copper = brass”). Although his exam-ple was about strong complements, the insight that independentmanufacturers would overprice their components carries over moregenerally.

attractive to the integrated firm, as well as in abilateral monopoly—is defeated when production ofcomponent goods is disaggregated into multiple man-ufacturers and subjected to the mix of horizontal andvertical channel conflicts. It is, however, possible forfirms to employ bundling in a profit-enhancing wayif the manufacturers could coordinate prices. Suchcoordination would keep component prices low andincrease manufacturers’ profits, the retailer’s profit,and consumer surplus.

This paper is novel in its combination of bundle-selling strategy and distribution channel structure,and it sits at the intersection of three separate lit-eratures. Past literature on bundling studies a directproducer–buyer distribution structure (e.g., a restau-rant that packages a burger, chips, and a drink;or a software firm that bundles multiple compo-nents). The literature on vertical channels and dou-ble marginalization models a retailer who essentiallypasses through the products of one or more manufac-turers to the consumer without considering the strate-gic lever of bundling these products. The literature oncomposite goods does not consider an active bundlingdecision by an intermediate firm such as a retailer.2

This paper combines these three literature streams, asa first model of bundling in a decentralized distribu-tion channel (see §2 and 3), but it is neither perfectnor complete. These limitations and the consequentresearch opportunities are discussed in §4.

2. ModelConsider a retailer who sells a bundle B of twocomponent products i = 112 made by independent

2 Bakos and Brynjolfsson (2000) study a vertical channel with bothupstream competition for content (between bundlers who competefor a single additional good) and downstream competition for con-sumers, but these two processes are modeled separately of eachother and avoid the complexity caused by multiple manufacturers.

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Bhargava: Retailer-Driven Product Bundling in a Distribution ChannelMarketing Science, Articles in Advance, pp. 1–8, © 2012 INFORMS 3

manufacturers (M1 and M2) who face unit costs c1and c2, respectively. Let vi denote consumer valua-tions for product i. A consumer v’s valuation for thebundle is the sum of her valuation for the two com-ponents, z= v1 + v2.

Assumption 1. Valuations for each product are dis-tributed uniformly on 60117 and are independent of theother. Equivalently, component goods i = 112 have lineardemand Di4p5= 1 − p.

Let p1, p2, and pB represent the per-unit prices setby the retailer for products i = 112 and the bundle Bin response to input costs w1 and w2 (the prices setby the manufacturer). Let Qi = Qi4p11 p21 pB5 be thesales of each product given the retail prices 4p11 p21 pB5,computed for the pure component and pure bundlestrategies as follows:

Pure Component

{

Q1 = 41 − p151

Q2 = 41 − p251(1)

Pure Bundle Q1 =Q2 =QB =DB4p5

=

{

12 42 − p25 if p ≤ 1112 42 − p52 otherwise.

(2)

The superscript C denotes firms’ equilibrium sales andprofit when the retailer pursues a pure componentselling strategy (i.e., QB = 0, or pB ≥ p1 + p2). A purebundle strategy is denoted with the superscript B andcorresponds to Q1 =Q2 = 0 (i.e., pi > pB). The retailer’sprofit çR and manufacturers’ profits �i are

çR = 4p1 −w15Q1 + 4p2 −w25Q2

+ 4pB −w1 −w25+QB1 (3a)

�1 = 4w1 − c154Q1 +QB51 (3b)

�2 = 4w2 − c254Q2 +QB50 (3c)

This two-goods setting is frequently employed toexpose insights about bundling, including in theearly work on bundling (Adams and Yellen 1976,Schmalensee 1984, Salinger 1995) and later explo-rations into the effect of correlation in product valua-tions (McAfee et al. 1989) and network effects (Prasadet al. 2010). For the classic case of a single integratedfirm (one that produces the two goods and makesthe bundling decision, hence wi = ci), McAfee et al.(1989) and others have demonstrated that bundlingis attractive and maximizes profits when costs arerelatively low; profit increases because the bundledemand curve, being flatter in the middle region,enables extraction of more of the total surplus. Foreasy reference, Lemma 1 summarizes the result forthe integrated firm, and Table 1 illustrates it for thespecial case of zero marginal costs; c1 = c2 = 0. Allproofs are in the electronic companion (available athttp://dx.doi.org/10.1287/mksc.1120.0725).

Table 1 Bundle Selling Dominates Component Selling for anIntegrated Firm Under Assumption 1 and Zero Marginal Costs

Selling strategy p1 p2 pB Q1 Q2 QB ç CS

Pure component 0.5 0.5 NA 0.5 0.5 NA 005 0025

Pure bundle NA NA√

23 NA NA 2

3 005443 002742

Lemma 1 (Integrated Firm). The optimal pricesunder the three selling strategies are

Pure Component Pure Bundle

pCi =1+ci

24i=11253 pB =

13 4c1 +c2 +

√

6+4c1 +c2525

if c1 +c2<121

23 41+c1 +c25 otherwise0

Pure bundling is better than pure components when costsare low—specifically when 4c11 c25 ∈ä1 ∪ä2, where

ä1 =

{

4c11 c252

(

c1 + c2 ≤12

)

4c1 + c253 + 46 + 4c1 + c25

253/2 − 184c1 + c25

27

≥

2∑

i=1

41 − ci52

4

}

1

ä2 =

{

4c11 c252

(

c1 + c2 ≥12

)

·

(

2(

2 − 4c1 + c25

3

)3

≥

2∑

i=1

41 − ci52

4

)}

0

The total surplus (or social welfare (SW)) undercomponent sales is

∫ 1p1xdx +

∫ 1p2ydy. Under pure

bundling, SW is 1 −∫ pB

0

∫ pB−x

0 4x + y5dydx for pB ≤ 1and

∫ 1pB−1

∫ 1pB−x

4x + y5dydx for pB ≥ 1. Consumer sur-plus (CS) is SW−ç1 −ç2 −çR.

3. Bundling in Decentralized ChannelThis section evaluates pure bundle versus pure com-ponent selling in a decentralized channel and com-pares these outcomes with those for an integratedfirm and a bilateral monopoly. Although there is anadditional selling strategy—mixed bundling, whichis optimal in theory—comparing the two extremecases produces sharper insights (I comment on mixedbundling in §4). Section 3.1 examines a retailer whois restricted to pure bundling (e.g., Netflix, describedbelow), and §3.3 describes a retailer who observesmanufacturer prices and then determines whether ornot to bundle. To facilitate the comparison of com-ponent selling and bundling, Table 2 provides a purecomponent-selling benchmark in the vertical channel.This solution is derived by aggregating the solution of

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.


Table 2 Pure Component-Selling Solution in Vertical Channel

Effect Manufacturers Retailer

Prices wi =1 + ci

2pi =

3 + ci4

Firm’s profit çi =18 41 − ci 5

2 çR =116 441 − c15

2 + 41 − c2525

Industry profit ç1 +ç2 +çR =316 441 − c15

2 + 41 − c2525

two separate two-stage pricing games, each of whichfeatures one manufacturer (stage 1) and one retailer(stage 2).

A retailer’s selling strategy may be restrictedto pure bundling when it can be predicted fromhistorical actions, is necessary to induce manufac-turer participation, or is the only reasonable businesspractice. An example is the movie distribution firmNetflix, which offers access at a flat price to its entirelibrary, comprising movies from several different stu-dios (the component-selling alternative would be aseparate price for access to movies from each studio).Content providers are aware of the stickiness of thisall-you-can-eat business model when they price theircontent. Other information aggregators such as newsservices also have this characteristic. The limitation topure bundling can also occur in practice when tech-nological challenges make mixed bundling infeasibleor when firms anticipate antitrust concerns (undermixed bundling) regarding the bundle discount rel-ative to component prices (Fang and Norman 2006).Pure bundling avoids these concerns because thereare no component prices with which to compare.

3.1. Retailer Committed to Pure BundlingThe sequence of events in this game is that first theretailer conveys a precommitment to bundling, thenmanufacturers set their component prices, and finally,the retailer sets the price for the bundle. In the finalstage, the retailer’s optimal pricing rule pB is obtainedby replacing the cs with ws in Lemma 1. Each firmsells QB units (see Equation (5) in the electronic com-panion), obtained by substituting pB into Equation (2).In the first stage, manufacturers set their prices whiletaking into account the retailer’s pricing rule and theirown sales volume, QB. The Nash equilibrium for man-ufacturer prices is obtained by solving

wB1 =argmax

w1

(

ç14w11w25= 4w1 −c15QB4pBB4w11w

B2 55)

1

wB2 =argmax

w1

(

ç24w11w25= 4w2 −c25QB4pBB4w

B1 1w255

)

1

(4)where each manufacturer’s best-response price (withrespect to the other manufacturer’s price) is the bet-ter of the optimal values within the two subintervalsgiven by 4w1 +w25≷1/2.

Proposition 1 (Pure Bundling Regime). The equi-librium outcome when the retailer precommits to bundlingproducts i=112 from independent firms M1 and M2 is asfollows.

Effect Manufacturer Retailer

Prices wBi =

12

+3ci−cj

4pBR=

4+c1 +c2

3

Units QB =1

1842−c1 −c25

2

soldProfit çB

1 =çB2 =

172 42−c1 −c25

3 çBR=

1108 42−c1 −c25

3.

Each manufacturer’s optimal price response is a decreasingfunction of its competitor’s price (wB

i = 4241+ci5−wj5/3).Both the retailer and manufacturers earn lower profit thanunder component selling.

Because the retailer’s bundle combines productsfrom multiple manufacturers, each manufacturer’sprice has only a partial impact on its own sales. Thisinduces overpricing by manufacturers, and then theretailer sets a higher retail price, leading to lowersales level and lower profits for all. These interfirmeffects are illustrated by the recent market evolutionof Netflix. Following consumers’ rapid acceptanceof Internet-based movie streaming (and after it hadacquired multiyear streaming rights from studios, atvery low prices, when streaming was a rarity), Netflixexperienced massive growth in movie subscribers andprofits during 2007–2010. With Netflix’s content con-tracts set for renewal around 2011, it became evi-dent that studios would demand substantially higherfees from Netflix. Indeed, during 2010–2011, Netflixcited potential price increases by movie studios as oneof the big risks facing its movie-streaming business.3

Subsequently, Netflix faced higher costs from contentproviders (and even lost some, such as Starz Play) andexperienced a dip in both subscribers and profitabilityafter raising subscription prices.

3.2. Relative Role of Vertical vs.Horizontal Conflict

Proposition 1 demonstrated that the channel struc-ture negatively affects the attractiveness of bundling.Although demand smoothing creates a large poolof relatively high-value buyers in the retail market,the presence of this pool is exploited by manufac-turers who set higher wholesale prices and raise thebundler’s (i.e., retailer’s) unit costs, making bundlingless attractive. These outcomes are driven jointly bytwo types of channel conflict, vertical (between themanufacturers and retailer, which increases retailer’s

3 Based on analysis of the 2011 Securities and Exchange Commis-sion filings. Movie consumption costs presently account for about10% of Netflix’s cost to serve an average subscriber.

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.


costs as a result of double marginalization) andhorizontal (among manufacturers, which also raisesretailer’s costs because of each firm’s greed). The rel-ative influence of these two types of conflict can beisolated by evaluating an intermediate structure, abilateral monopoly in which a single manufacturermakes goods 1 and 2 that are sold separately or asa bundle by the retailer. Because there is no hori-zontal conflict in the bilateral monopoly, any differ-ence in outcome (over the integrated firm) must beon account of vertical conflict.

Bilateral monopoly applications in practice includethe Microsoft Office bundle and home-theater sys-tems, where component goods are designed by thesame firm but the bundle is sold through retailers.I illustrate the analysis for the special case of c1 =c2 =0for which bundling beats component selling for theintegrated firm but not in the decentralized channel.We see that vertical conflict alone does not negate thedemand-side motivation for bundling (see Proposi-tion 2). Although double marginalization in the verti-cal channel does lower the firms’ profits relative to anintegrated firm, bundle selling nevertheless increasesprofit for both the manufacturer and retailer com-pared with component sales. Moreover, the profitincrease (about 17%, under ci =0) is higher than thatfor the integration firm (about 9%), suggesting thatbundling actually helps coordinate the channel. Thisanalysis attests that horizontal conflict was the pri-mary culprit behind the failure of bundling in thedecentralized channel.

Proposition 2 (Bilateral Monopoly). Thecomponent-selling and bundle-selling equilibria in a bilat-eral monopoly are given in Table 3. Total manufacturerprofit (64/243) and the retailer’s profit (128/729) arehigher with bundle selling than under the component-selling regime.

3.3. Strategic Choice of Bundle orComponent Sales

Section 3.1 described the impact of the channel struc-ture on a retailer whose only product is a bundle ofcomponent goods from multiple manufacturers. Withthis single-product design, the retailer was vulnerableto overpricing by manufacturers, and the potential ofbundling to raise firms’ profits was wasted. Might

Table 3 Bundle Selling Dominates Component Selling for aBilateral Monopoly

Selling strategy wi p Qi çM çR ç=çM +çR CS

Components121

12

341

34

141

14

18

+18

18

38

116

Bundle131

13

109

3281

13281

64243

128729

320729

8503333729

this potential be restored if the retailer were to choosewhether to sell a bundle or the components afterobserving manufacturers’ prices? The expectation inthis analysis is that (i) the retailer would choosebundling only if manufacturers set prices low enough,and (ii) knowledge of this selection rule would steermanufacturers toward lower prices to realize thebundling (i.e., surplus-maximizing) outcome.

Let ä be the low-price region of 4w11w25s for whichthe retailer prefers to bundle the products. Figure 2provides a visual depiction of the sequence of deci-sions made by the manufacturers and retailer. Anequilibrium outcome in the overall game is a strategyprofile �w11w21BC1p11p21pB�, where BC is B or C rep-resenting the retailer’s choice of bundling or compo-nents, respectively. By convention, p1 and p2 are null(sufficiently high) when the retailer picks B, while pBis null when the retailer picks C. Because the retailerdeterministically sets prices and strategy after observ-ing w1 and w2, any outcome of the game is fullydescribed by the pair of manufacturer prices 4w11w25.Analysis of this game reveals that the equilibrium out-come is component selling rather than bundling.

Proposition 3 (Equilibrium Features Compo-nent Selling). The two-stage game has a unique equilib-rium solution in which the manufacturers set prices w1 =

w2 =1/2, and the retailer sells the component products atprices p1 =p2 =3/4. Each firm earns 1/8, the total profitacross all three firms is 3/8, and total consumer surplus is1/16, with a total system surplus of 7/16 (of a possible 1).

As with the precommitted retailer (Proposition 1),the decentralized structure negates the demand-sidemotivation for bundling. Even though the retailerholds the threat of switching to the component-sellingregime, manufacturers set their prices higher than thelevel needed to induce the retailer to bundle. Whenmanufacturers set their prices low enough to inducethe retailer to bundle (4w11w25∈ä2), the retailer’s bun-dle pricing rule awards them a lower profit thanunder their optimal component sales price. And ifmanufacturers price high enough to make bundlingattractive to them, then the retailer earns higher profitfrom selling the component goods separately. Thismisalignment of incentives prevents the emergence ofa bundling equilibrium even though bundling has thepotential to increase all firms’ profit.

3.4. ExtensionsThe weakness of pure bundling in the decentralizedchannel raises the following question: If the demand-side motivation for bundling were stronger, would itmake bundling more likely, or would it simply pre-cipitate the efforts of each firm to garner greater prof-its? I consider this question by examining two factorsthat strengthen this motivation: negative correlationin valuations for components and a larger number ofcomponent goods being bundled.

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.


Figure 2 Price and Bundling Strategy Game with Two Manufacturers and One Retailer

M1

M2

R

Sell components atp1, p2

w1,w2

∈Θ

w1, w

2 ∉Θ

Manufacturers pickproduct prices

Retailer picksselling strategy and prices

w1

w2 Q1, Q2

Consumers purchase,product sales realized

QBSell bundle atpB

The demand-smoothing effect becomes strongerwhen demands are anticorrelated (�<0), creating agreater pool of consumers willing to buy the bun-dle at a relatively high price. Will manufacturersexploit these higher reservation prices and set evenhigher prices for their individual components, therebydestroying the retailer’s incentives to sell a bundle?For the general case of dependent demand, there areno explicit forms for optimal bundle prices (McAfeeet al. 1989) due to a lack of closed-form terms for thesum of two dependent random variables (Makarov1981, Arbenz et al. 2011). I follow the approach ofMcCardle et al. (2007) in evaluating the extremecase of �=−1. Then, all consumers value the bun-dle at 1 (and the retailer sets this price); hencethe retailer prefers bundling when 41−w1 −w25≥441−w15/252 +441−w25/252; i.e., wi ≤

√2−1 (≈004142).

And, indeed, manufacturers would prefer to setthis price and induce a bundling equilibrium ratherthan deviate and earn the component-selling profit.Because the incremental profit from bundling ismonotonic (decreasing) in correlation � (Gürler et al.2009), we conclude the following.

Proposition 4. Let � be the correlation in consumervaluations for component goods 1 and 2. Then thereexists �∈ 4−1105 such that a bundling equilibrium emergeswhenever �<�.

Next, consider the case where the retailer bun-dles components from N >2 manufacturers. Severalauthors have examined bundling of large numbersof goods, including Bakos and Brynjolfsson (1999),who employ limit analysis, and Fang and Norman(2006), who study a finite number of goods by apply-ing peakedness of distributions (Proschan 1965). Withindependent product valuations, each consumer’svaluation of the bundle gets arbitrarily close to N/2 asN increases. In a bundling regime, the retailer wouldset bundle price at ≈N/2 and pick the bundlingstrategy when

(

N

2−

N∑

i=1

wi

)

×1≥

N∑

i=1

(

1−wi

2

)2

1

which yields wi ≤−1+√

2≈004142. And, indeed, eachmanufacturer earns a higher profit (004142×1) thanthe component-selling profit they would earn whendeviating from this price.

Proposition 5. There exists N such that for allN >N , there is a bundling equilibrium when the retailerhas an opportunity to bundle N components from differentmanufacturers.

4. ConclusionThis paper has analyzed bundling in a decentralizedchannel in which production and price setting of com-ponent goods is managed by separate manufacturersacting independently of each other, and the sellingstrategy (bundle or components) is determined by adownstream retailer. Industry outcomes under thisstructure follow from a complex intertwining of twohorizontal and vertical channel conflicts with the eco-nomic motivation for bundling. I show that a combi-nation of horizontal and vertical channel conflicts inthe decentralized channel weakens the firms’ incen-tives for bundling. Bundle selling reduces profits com-pared to component selling in the base case of twocomponent goods with independent demand. Consis-tent with past literature, bundle selling can becomefavorable when the component demands are nega-tively correlated or when the firm can form largeN bundles, though the gains from bundling are stilllower than for the integrated firm. Similarly, bun-dle sales would be observed when the firm practicesmixed bundling (for which closed-form solutions arenot available) because the lure of component saleswould mitigate manufacturers’ inclination to over-price. However, because cross-manufacturer bundlepricing still masks the effect of each manufacturer’sprice increase on its own sales, mixed bundlingshould generate a smaller increase in profit (comparedwith component selling) in the decentralized channelthan it would for the integrated firm.

The participating firms in this bundling game face aprisoner’s dilemma in pricing: all would earn higherprofits if manufacturers set prices lower than their

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.


Figure 3 A Few Examples of Carriage Disputes in the TV Industry (January–September 2010)

component-selling optimal and the retailer pricedbelow the level that maximized its own profit. Indeed,a bundling outcome does emerge in some indus-tries (e.g., in the TV industry). However, this out-come also exhibits price and revenue-sharing tensionsamong participants who perceive an opportunity toraise prices and attain higher profits in the short term.Such conflicts are conspicuous in the TV industry inthe form of carriage fee disputes, as illustrated byrecent news headlines shown in Figure 3. The lackof a stable bundling solution has motivated firms inthis industry to integrate vertically by moving eitherup or down the value chain (e.g., Comcast’s acquisi-tion of NBC, Amazon’s effort to make movies throughAmazon Studios).

There are many additional directions that havebeen investigated in the direct-selling setting thatcould be pursued in future work on bundling ina vertical channel. These include bundling of ver-tically differentiated products (Banciu et al. 2010),bundling of complements and substitutes (Venkateshand Kamakura 2003), and bundling under competi-tion either between retailers or manufacturers mak-ing components that are substitutes for each other.It would also be useful to study formal incentive-compatible, revenue-sharing mechanisms that canimprove profits for all parties (and possibly consumersurplus) by achieving better price coordination in thedecentralized channel. Industry-specific arrangementsmay also be relevant, such as in the TV industryor advertising. Because ad revenue accrues primar-ily to the manufacturers (content owners), they havean increased incentive to maximize subscribers ratherthan margins, pushing the outcome toward lowercomponent prices and lower bundle price.

Finally, industry structure is often more complexthan assumed in the simple model with two single-product manufacturers. TV bundles feature hundredsof channels, not just two, though most of them arefrom about 6–10 programming networks and studios.Many of the big studios prebundle more than a dozenchannels in their negotiations with the retailers. Forexample, Disney executives negotiate for the inclusionof certain less popular channels in exchange for theright to carry ESPN; similarly, News Corp. charges abundle price for the collection of FOX channels. Themodel employed in this paper abstracted these stage1 bundles into a single product offered by the manu-facturer to the retailer; however, a complete analysismight provide a better understanding of this prac-tice. Another complicating factor is that some retailersare also manufacturers who provide their own con-tent (e.g., Netflix has begun producing original TVseries and movies; Comcast owns several networks,including E!, the Style Network, G4, and the GolfChannel).4 Incorporating any of these features intothe model would be challenging but would highlyenrich the analysis. The analysis framework presentedin this paper is a useful starting point, one that iscomputationally tractable and insightful. I hope thatit will spur substantial new work in this excitingarea.

Electronic CompanionAn electronic companion to this paper is available as part ofthe online version at http://dx.doi.org/10.1287/mksc.1120.0725.

4 Wikipedia contributors, “Comcast.” Wikipedia, The Free Ency-clopedia. Accessed June 29, 2012, http://en.wikipedia.org/wiki/Comcast.

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.


ReferencesAdams WJ, Yellen JL (1976) Commodity bundling and the burden

of monopoly. Quart. J. Econom. 90(3):475–498.Arbenz P, Embrechts P, Puccetti G (2011) The AEP algorithm for the

fast computation of the distribution of the sum of dependentrandom variables. Bernoulli 17(2):562–591.

Bakos Y, Brynjolfsson E (1999) Bundling information goods:Pricing, profits, and efficiency. Management Sci. 45(12):1613–1630.

Bakos Y, Brynjolfsson E (2000) Bundling and competition on theInternet. Marketing Sci. 19(1):63–82.

Banciu M, Gal-Or E, Mirchandani P (2010) Bundling strategieswhen products are vertically differentiated and capacities arelimited. Management Sci. 56(12):2207–2223.

Cournot AA (1929) Researches into the Mathematical Principles of theTheory of Wealth [Translated by Bacon NT, Essay on Cournotand Mathematical Economics and a Bibliography of Mathemat-ical Economics by Fisher I] (Macmillan, New York).

Fang H, Norman P (2006) To bundle or not to bundle. RAND J.Econom. 37(4):946–963.

Gürler Ü, Öztop S, Sen A (2009) Optimal bundle formation andpricing of two products with limited stock. Internat. J. Produc-tion Econom. 118(2):442–462.

Makarov G (1981) Estimates for the distribution function of a sumof two random variables when the marginal distributions arefixed. Theory Probab. Appl. 26(4):803–806.

McAfee RP, McMillan J, Whinston MD (1989) Multiproductmonopoly, commodity bundling, and correlation of values.Quart. J. Econom. 104(2):371–383.

McCardle KF, Rajaram K, Tang CS (2007) Bundling retail products:Models and analysis. Eur. J. Oper. Res. 177(2):1197–1217.

Prasad A, Venkatesh R, Mahajan V (2010) Optimal bundling of tech-nological products with network externality. Management Sci.56(12):2224–2236.

Proschan F (1965) Peakedness of distributions of convex combina-tions. Ann. Math. Statist. 36(6):1703–1706.

Salinger MA (1995) A graphical analysis of bundling. J. Bus.68(1):85–98.

Schmalensee R (1984) Gaussian demand and commodity bundling.J. Bus. 57(1):S211–S230.

Spengler JJ (1950): Vertical integration and antitrust policy. J. Polit-ical Econom. 58(4):347–352.

Stigler GJ (1963) United States v. Loew’s Inc.: A note on block-booking. Supreme Court Rev. 1963:152–157.

Venkatesh R, Kamakura W (2003) Optimal bundling and pricingunder a monopoly: Contrasting complements and substitutesfrom independently valued products. J. Bus. 76(2):211–232.

Venkatesh R, Mahajan V (2009) The design and pricing of bundles:A review of normative guidelines and practical approaches.Rao V, ed. Handbook of Pricing Research in Marketing (EdwardElgar Publishing, Northampton, MA), 232–257.

Wu S, Hitt LM, Chen P, Anandalingam G (2008) Customized bun-dle pricing for information goods: A non-linear mixed integerprogramming approach. Management Sci. 54(3):608–622.

Copyright:

INFORMS

holdsco

pyrig

htto

this

Articlesin

Adv

ance

version,

which

ismad

eav

ailableto

subs

cribers.

The

filemay

notbe

posted

onan

yothe

rweb

site,includ

ing

the

author’s

site.Pleas

ese

ndan

yqu

estio

nsrega

rding

this

policyto

perm

ission

s@inform

s.org.

Documents

Retailer-Driven Product Bundling in a Distribution Channel