Professional Documents
Culture Documents
Rey & Verge (2008) Resale Price Maintenance and Interlocking Relationships
Rey & Verge (2008) Resale Price Maintenance and Interlocking Relationships
Relationships∗
Patrick Rey† Thibaud Vergé‡
July 1, 2008
Abstract
∗
We benefited from helpful discussions with Rodolphe Dos Santos Ferreira, Bruno Jullien and Jean
Tirole on an earlier draft. We are also grateful to Eric Avenel, Bill Rogerson, the Editor (Pierre Régibeau)
and two anonymous referees for their remarks.
†
Toulouse School of Economics (GREMAQ, IDEI) and Institut Universitaire de France.
‡
CREST-LEI.
1
1 Introduction
The attitude of competition authorities and courts towards vertical restraints varies sig-
nificantly from one country to another or from one period to another.1 Still, a consen-
sus emerges against resale price maintenance (RPM), a restraint according to which the
manufacturer sets the final price that retailers charge to consumers. While competition
authorities are sometimes tolerant towards some variants of RPM such as price ceilings
and recommended or advertised prices, they usually treat price floors and strict RPM as
per se illegal. For example, when the European Commission adopted a more open attitude
towards non-price restrictions, it maintained RPM on a black list — with only one other
restraint. In France, price floors are per se illegal and, in Lypobar vs. La Croissanterie
(1989), the Paris Court of Appeal ruled that RPM was an abuse of franchisees’ economic
dependency. A recent exception to this consensus concerns the U.S., where the Supreme
Court overturned last year the long-established per se illegality of price floors, adopting
instead a rule of reason approach.2
The economic analysis of vertical restraints is more ambiguous: it is not clear that
RPM has a more negative impact on welfare than other vertical restraints that limit
intrabrand competition. Instead, both price (e.g., RPM) and non-price restraints (e.g.,
exclusive territories) may have positive or negative effects on welfare, depending on the
context in which they are used.3 In particular, both price and non-price vertical restraints
can deal with vertical coordination problems.4 For instance, combined with non-linear
wholesale tariffs, RPM or exclusive territories can equally limit free-riding problems cre-
ated by strong intrabrand competition.5 Quite a few papers have moreover pointed at
specific efficiency benefits of RPM.6 Vertical restraints may also affect interbrand com-
petition. Manufacturers can for example impose restraints on retailers so as to become
“less aggressive”. Through strategic complementarity, this in turn induces their rivals
to respond less aggressively (e.g., increase their wholesale prices) ultimately leading to
1
For an overview of the legal frameworks regarding vertical restraints, see OECD (1994) or the Euro-
pean Commission’s Green Paper on Vertical Restraints (1996). Comanor and Rey (1996) also compares
the evolution of the attitudes of the U.S. competition authorities and within the European Community.
2
See Leegin Creative Leather Products, Inc. v. PSKS, Inc., 127 S.Ct 2705 (2007).
3
See Motta (2004, chapter 6) or Rey and Vergé (2008) for recent surveys of that literature.
4
Rey and Tirole (1986) offers an overview of the relative merits of price and non-price restrictions in
improving vertical coordination.
5
Note however, that depending on the structure of consumer demand, such restraints may harm or
enhance economic welfare. See (among others) Spence (1975), Comanor (1985), Caillaud and Rey (1987),
or more recently Schulz (2007).
6
For example, Marvel and McCafferty (1984) stress that RPM can help manufacturers to purchase
certification from reputable dealers. Deneckere, Marvel and Peck (1996, 1997) and Wang (2004) show
that RPM can encourage retailers to hold inventories in the presence of demand uncertainty, while Chen
(1999) shows that RPM may help controlling retail price discrimination.
2
higher prices and profits.7 To achieve this, manufacturers must however give retailers
some freedom in their pricing policies. Granting exclusive territories (thus eliminating
intrabrand competition) would for example serve this purpose and have an adverse effect
on consumer surplus and economic welfare, whereas RPM would have no impact since it
also eliminates the retailers’ freedom to choose their retail prices. Overall, a comparison
of the welfare effects of exclusive territories, RPM and exclusive dealing does not clearly
justify a more lenient attitude towards non-price restrictions.8
There is however one last argument that has often been made by courts to justify
a negative attitude towards price restrictions. For example, in Business Electronics, the
Supreme Court justified the per se illegality of RPM by claiming that “there was sup-
port for the proposition that vertical price restraints reduce inter-brand price competition
because they facilitate cartelizing.” This type of argument has been informally used by
Telser (1960) and Mathewson and Winter (1998). It was however formalized only very
recently by Jullien and Rey (2007) who stress that, by making retail prices less responsive
to local shocks on retail cost or demand, RPM yields more uniform prices that facilitate
tacit collusion — by making deviations easier to detect.
This paper analyzes this “facilitating practice” argument from a different perspec-
tive. We show that, even in the absence of repeated interactions, RPM can eliminate
any scope for effective competition when manufacturers and retailers engage in “inter-
locking relationships”, that is, when manufacturers distribute their goods through the
same competing distributors. The intuition is relatively simple. In the case of a (local)
retail monopoly we know that, through “common agency”, competing manufacturers can
avoid interbrand competition, e.g., by selling at cost in exchange for a fixed fee: since
manufacturers internalize through fixed fees the impact of prices on the retailer’s profit,
eliminating the upstream margin on one brand transforms a rival manufacturer into a
residual claimant on the sales of both brands. As a result, rival manufacturers have in-
centives to maintain retail prices at the monopoly level, which can be achieved precisely
by selling at cost. Simple two-part tariffs therefore suffice to maintain monopoly prices and
profits.9 This is no longer the case when there is competition not only between brands,
but also between retailers, which tends to reduce retail margins. Manufacturers then have
conflicting incentives: they still want to keep low upstream margins in order to avoid
interbrand competition but they need now to increase their wholesale prices in order to
maintain high retail prices despite intrabrand competition. As we will see, two-part tariffs
no longer suffice to maintain industry profits, and retail prices are instead set below their
monopoly level. Manufacturers can however use RPM to eliminate intrabrand competi-
7
See for example Rey and Stiglitz (1988, 1995) and Bonanno and Vickers (1988).
8
See Caballero-Sanz and Rey (1996).
9
See Bernheim and Whinston (1985, 1998) and O’Brien and Shaffer (1997).
3
tion and restore monopoly prices and profits. In particular, selling at cost (for a fee) still
makes rivals internalize (through their own fixed fees) the full impact of their prices on
the sales of a manufacturer’s brand. At the same time, a manufacturer can now main-
tain high retail prices for its brand through RPM. Combining two-part tariffs with RPM
thus provides a mechanism through which manufacturers can give each other incentives
to maintain high retail prices and profits. Both interbrand and intrabrand competition
are then totally eliminated, even though contracts (including retail prices) are negotiated
on a purely bilateral basis. In the absence of any retail bottleneck (e.g., when there are
potential competitors for each retail location), manufacturers clearly benefit from this,
since they can appropriate most of the profits that they generate. When instead retailers
have market power, manufacturers need to leave them some rents, thus reducing their
incentives to deal with both retailers and to maintain monopoly prices. As a result, all
channels may not be active and manufacturers may moreover favor lower prices, in order
to keep a larger share of an admittedly smaller pie, whereas retailers would instead favor
higher retail prices.
Note that the mechanism identified here could not be replicated through other stan-
dard means of reducing intrabrand competition, e.g., by granting an exclusive right over
some territory. Indeed, RPM allows manufacturers to avoid interbrand competition even
when, due to retailers’ differentiation strategies, meeting consumer demand makes it un-
desirable to grant exclusive territories and exclude some of the established retailers.
This paper is closely related to that of Dobson and Waterson (2007), who study a
similar bilateral duopoly with interlocking relationships. Assuming that manufacturers
use (inefficient) linear wholesale prices, they show that the welfare effects of RPM depend
on the relative degree of upstream and downstream differentiation as well as on retailers’
and manufacturers’ bargaining powers; RPM can be socially preferable when retailers are
in a weak bargaining position, because the double-marginalization problems generated by
the restriction to linear wholesale prices is more severe in such circumstances.10 In order
to eliminate double marginalization problems and focus instead on the impact of RPM
on interbrand and intrabrand competition, we do not restrict attention to linear tariffs
but allow for bilaterally efficient (two-part) wholesale tariffs.11
Our analysis sheds an interesting light on recent legal developments. While the US
Supreme Court recently overturned the per illegality of RPM, in France, RPM, together
10
In a similar context, Allain and Chambolle (2007) moreover show that non-discriminatory price floors
can help maintain high retail prices even when manufacturers can grant secret rebates.
11
Another difference concerns the equilibrium concept. To reflect different bargaining powers, Dobson
and Waterson (2007) assume that wholesale prices are determined by simultaneous pairwise bargaining.
This supposes that a manufacturer has two independent divisions, each of them negotiating with one
retailer not taking into account the impact of its own negotiation on the other division.
4
with non-linear wholesale tariffs, has instead raised concerns in markets where multiple
producers distribute their goods through the same retailers. For example, in Decem-
ber 2005, the Conseil de la Concurrence (one of the two French competition authorities)
condemned brown goods manufacturers Panasonic, Philips and Sony for “vertical collu-
sion” with their wholesalers and retailers. The Conseil de la Concurrence concluded that
there was evidence that these manufacturers were actively monitoring retailers in order to
ensure that they were actually following their recommended retail prices (this was espe-
cially the case for new lines of products) and were pushing wholesalers to refuse to supply
retailers that were cutting prices.12 In similar cases, the major perfume manufacturers
(L’Oréal, Chanel, Guerlain, Dior, ...) and retailers (Nocibé, Marionnaud, Séphora) were
fined a total of 44 million euros, and toy manufacturers (Chicco, Lego, ...) and retailers
(Carrefour, JouéClub, ...) were fined a total of 37 million euros for the same practices.13
Our analysis is also relevant for the ongoing reform of the French competition rules
banning below-cost pricing. In order to simplify billing methods and enhance trans-
parency, a law adopted in 1996 defined the relevant cost threshold as the invoice-price
paid by the retailer at the time of delivery. As a result, retailers could no longer pass on
to consumers many rebates, such as quantity and end-of-year discounts or slotting fees,
which do not usually appear on invoices. It has been argued that the regulation was legal-
izing minimum RPM: the invoice price, determined by the manufacturer’s general terms
of sales, de facto imposed a minimum retail price eliminating intrabrand competition, and
retailers then negotiated “backroom rebates” to maintain their margins. The regulation
has been heavily criticized as being responsible for the important price increases that have
taken place after 1997, especially for the major national brands present in all supermarket
chains. As we will see, our analytical framework supports this claim and has moreover
been validated by recent empirical studies of the French bottled water market.14
This paper is organized as follows. Section 2 presents our framework, where two rival
manufacturers distribute their goods through two competing retailers; this framework
allows for interlocking relationships (or “double common agency”): each manufacturer
can deal with both retailers, and conversely each retailer can carry both brands. Section
3 provides a preliminary analysis of such double common agency situations: while retail
prices are lower than the monopoly price in the absence of RPM, with RPM there exist
many equilibria, including one in which retail prices and manufacturers’ profits are at the
12
These three manufacturers were respectively fined 2.4, 16 and 16 million euros. Panasonic was later
cleared by the Court of Appeal. Other major manufacturers present on the French market were also
investigated, but the Conseil de la Concurrence did not find enough evidence to convict them. See
Conseil de la Concurrence, decision 05-D-66, December 2005.
13
See Conseil de la Concurrence, decisions 06-D-04 (March 2006, Perfumes) and 07-D-50 (December
2007, Toys).
14
See Bonnet and Dubois (2004 and 2007).
5
monopoly level. We then endogenize the market structure. Section 4 studies situations
with potential competition downstream for each retail location. Both brands are then
always present at both retail locations and the previous analysis applies; in particular,
when RPM is allowed, there always exists an equilibrium with monopoly prices and profits.
Section 5 turns to the case of retail bottlenecks, where manufacturers cannot bypass
established retailers. Manufacturers must then leave a rent to retailers to induce them
to sell their products; relatedly, they can attempt to eliminate competitors by convincing
retailers to reject their rival’s offer. As a result, it can be the case that no equilibrium
exists where both manufacturers are present in both retail outlets, even though there is
demand for each brand at each store. In addition, while there may exist a continuum
of equilibria with RPM, equilibria with higher retail prices now involve larger rents for
the retailers and lower profits for the manufacturer — implying that manufacturers favor
equilibria with rather “competitive” prices. Section 6 discusses the policy implications of
our analysis and concludes.
6
distribution unit costs are symmetric and constant, and denote them respectively by c and
P P
γ.17 The industry profit is thus equal to i=A,B j=1,2 (pij − c − γ) Dij (p) . Throughout
the paper, we assume that this industry profit is concave in p, maximal for symmetric
prices, pM = (pM , pM , pM , pM ) and denote by ΠM this maximum (from now on, we will
refer to ΠM as the monopoly profit).
To fix ideas, we assume throughout the paper that the manufacturers have all the
bargaining power. We thus consider a two-stage game where at stage 1, manufacturers
offer contracts to the retailers, and, at stage 2, retailers compete on the downstream
markets.
7
tailers then simultaneously decide whether to accept or reject the offers, and
acceptance decisions are public.
(1 − B) If all offers are accepted, the game proceeds to stage 2; otherwise, the market
breaks-down and the game ends with all firms earning zero profits.
The participation constraints are clearly binding and the program is thus equivalent to:
X ¡¡ ¢ r ¢
max Πri (w) ≡ prij (w) − c − γ Dij (w) + (prhj (w) − whj − γ)Dhj
r
(w) .
wi1 ,wi2
j=1,2
In other words, through the franchise fees each manufacturer i internalizes the impact
of its pricing decisions on (i) the entire margins (pij − c − γ) on its own product (for
i = 1, 2) and (ii) the retail margins (phj − whj − γ) on the rival’s product; it therefore
ignores the rival’s upstream margins (whj − c). As a result, (symmetric) equilibrium prices
are somewhat competitive (i.e., below the monopoly level) whenever the retail equilibrium
satisfies weak regularity conditions.
8
Assumption 1
i) For symmetric wholesale prices (wi1 = wi2 = wi for i = A, B), equilibrium re-
tail prices are symmetric: pri1 = pri2 ≡ pe (wi , wh ) for i 6= h = A, B , leading to symmetric
quantities Di1r r
= Di2 ≡De (wi , wh ); moreover:
ii) an increase in all wholesale prices increases retail prices: ∂1 pe + ∂2 pe > 0;
iii) an increase in one manufacturer’s wholesale prices decreases the demand for that
e < 0 < ∂2 D
manufacturer and increases the demand for its rival: ∂1 D e.
These conditions are for example satisfied when retail prices are strategic complements
and direct effects dominate indirect ones.21 In particular, they are satisfied in the linear
demand case analyzed in section 5.
If there were a monopoly at either level, (public) two-part tariffs would instead lead to
retail prices equal to monopoly prices. If, for example, a single manufacturer were selling
through competing retailers, it would set wholesale prices high enough to induce retail
prices at the monopoly level — and could then recover retail margins through franchise
fees. Likewise, if a single retailer were acting as a common agent for several manufacturers,
as in Bernheim and Whinston (1985), manufacturers would sell at marginal cost, thereby
inducing the retailer to adopt monopoly prices, and could recover again profits through
franchise fees.
Here, in contrast, the existence of competition at both the upstream and downstream
levels maintains retail prices below the monopoly level. This is because, as noted above,
manufacturers only take into account the retail margin on their rival’s products, and thus
fail to account that a reduction in their own prices hurt their rival’s upstream profits. If, for
example, retailers are pure Bertrand competitors (that is, assuming away any downstream
differentiation), they are both active only if wholesale prices are symmetric (wij = wi ),
in which case retail prices simply reflect wholesale prices (pij = wi ) and franchise fees
are zero, so that manufacturer i’s profit reduces to Πri (w) ≡ (wi − c − γ) D̂i (wA , wB ) ,
where D̂i (pA , pB ) represents the demand for product i = A, B when the price of product
A (respectively B) is pA (respectively pB ). The situation is then formally the same as if
the two manufacturers were directly competing against each other.
³ ´
21 e < 0 and ∂1 pe > −λR /λ̂R ∂2 pe ≥ 0, where λR (respectively,
For example, ∂1 pe ≥ ∂2 pe ≥ 0 implies ∂1 D
λ̂R ) denotes the impact on demand for the “product” ij of a uniform increase in retailer j’s (respectively,
e > 0.
retailer k’s) prices, implies ∂2 D
9
3.2 Resale Price Maintenance
Suppose now that manufacturers can resort to RPM. Imposing retail prices is then always
a dominant strategy for the manufacturers: whatever the strategy adopted by its rival,
a manufacturer can always replicate, with RPM, the retail prices that would emerge and
the profits it would earn without RPM.
Under RPM, the last stage of the game is straightforward. In the first stage, given the
market break-down assumption, if manufacturer h imposes retail prices (ph1 , ph2 ), manu-
facturer i will choose wholesale prices wi1 and wi2 , retail prices pi1 and pi2 , and franchises
Fi1 and Fi2 so as maximize its profit, given the retailers’ participation constraints:
P
max j=1,2 ((wij − c)Dij (p) + Fij ) ,
(wij ,pij ,Fij )j=1,2
P
s.t. h=A,B ((phj − whj − γ) Dhj (p) − Fhj ) ≥ 0, for any j = 1, 2.
As before, each manufacturer fully internalizes (through the franchise fees that it can
extract from the retailers) the entire margins on its product, but internalizes only the
retail margins on the rival’s product. But now, the manufacturer’s wholesale prices no
longer affect its profit (previously, these wholesale prices had an indirect effect through
retailers’ prices, which are now directly controlled by the manufacturer); however, as
the program (1) makes clear, these wholesale prices affect the rival’s profit and thus the
equilibrium behavior of the competitor. As a result, there usually exists a continuum of
equilibria — one equilibrium for every profile of wholesale prices w = (wA1 , wB1 , wA2 , wB2 ).
If for example manufacturer h sells at cost (wh1 = wh2 = c), program (1) becomes:
X
max ((pij − c − γ) Dij (p) + (phj − c − γ)Dhj (p))
pi1 ,pi2
j=1,2
Manufacturer i then fully internalizes the impact of its retail prices on aggregate profits,
and thus sets its prices at the monopoly level if manufacturer h does also so; there thus
exists an equilibrium in which both manufacturers set wholesale prices to c and retail
prices to the monopoly level, and share monopoly profits. RPM can thus prevent the
exercise of interbrand as well as intrabrand competition.22
If instead manufacturers adopt wholesale prices above cost, they tend to choose more
aggressive retail prices for their own brand, since they do not take into account the up-
stream margins on the rival brand. As a result, one expects an inverse relation between
22
The argument still applies when marginal costs are not constant, interpreting c as the marginal cost
for monopolistic production levels.
10
wholesale and retail prices. The next proposition confirms this intuition under the follow-
ing regularity conditions:
Assumption 2
i) For wh1 = wh2 = wh and ph1 = ph2 = ph , and i 6= h ∈ {A, B}, the revenue function
Π is single-peaked in (pi1 , pi2 ) and maximal for symmetric prices, p̂i1 = p̂i2 = p̂ (ph , wh );
ii) p̂ (., .) satisfies 0 < ∂1 p̂ < 1 and, for any w, the function p → p̂ (p, w) has a unique
fixed point.
This assumption first states that retail price responses are well defined and preserve
symmetry; in addition, for any symmetric profile of wholesale prices, there exists a unique,
stable, “retail equilibrium” (looking at a reduced strategic game where manufacturers
would simply choose retail prices, taking wholesale prices as given). We have:
There is thus a continuum of symmetric equilibria and, within this set of equilibria,
retail prices are inversely related to wholesale prices. Retail prices are at the monopoly
level when wholesale prices are equal to cost — in this equilibrium, manufacturers thus
“eliminate” any competition and achieve monopoly profits — while upstream mark-ups
sustain lower retail prices.23 In essence, with RPM, the situation is one where manu-
facturers deal with two, non-competing, common agents. Consider for example the polar
case where retailers are pure Bertrand competitors (no downstream differentiation). With
RPM the manufacturers eliminate retail competition and de facto allocate half of the de-
mand for their products to each retailer; the monopolistic equilibrium then simply mimics
the Bernheim and Whinston (1985) common agency equilibrium (without RPM) within
each half-market. The above analysis generalizes this insight to the case where retailers
are differentiated.
23
Conversely, negative upstream margins would sustain retail prices above the monopoly level. The
range of equilibrium prices depends on the domain of validity of Assumption 2. For example, for the
linear demand used in section 5, any retail price from c + γ up to the price for which quantities are 0 can
be sustained.
11
• Bilateral bargaining power
While we have assumed here that manufacturers have all the bargaining power and
make take-it or leave-it offers to retailers, the analysis is similar if retailers are the ones
that propose the contracts in stage 1 − A. Suppose for example that retailers have all
the bargaining power. With RPM, there again exists an equilibrium in which prices are
at the monopoly level — although now the retailers rather than the manufacturers get all
the profits. To achieve this, however, instead of removing the upstream margin (w∗ = c) ,
¡ ¢
the retailers remove the downstream margin w∗ = pM , so as to allow each of them to
internalize the whole margin on the manufacturers’ sales through the other retailer —
franchise fees being used to extract the manufacturers’ expected revenues (slotting fees -
i.e. negative franchise fees - are needed in this setting to transfer profits downstream).
12
with the previous situation, manufacturers are no longer indifferent as to the choice of
their wholesale prices, since they affect retail efforts. There are no longer more control
variables than targets, as a consequence, the multiplicity disappears. To provide adequate
incentives, manufacturers must make retailers residual claimants for their efforts, which
requires wholesale prices equal to marginal cost. As a result, in equilibrium the wholesale
prices are always equal to the marginal cost, and the only equilibria that are robust to
the introduction of retail efforts therefore lead to the monopoly outcome.26
4 Competitive Retailers
The previous “market break-down” assumption imposes double common agency as the
equilibrium market structure and moreover implies that manufacturers extract all prof-
its. While this assumption is clearly ad-hoc and, as such, unrealistic, it captures the
essential ingredients of potential retail competition. Indeed, if manufacturers can always
find equally efficient alternative channels for each relevant retail location then, as in the
previous section, the following two features are likely to hold:
The analysis of the precedent section is then likely to prevail: manufacturers are
deemed to “accommodate” each other and their best strategy is to maintain monopoly
prices and share the monopoly profits, which they can indeed achieve by adopting common
retailers (rather than marketing their products themselves or through different retailers)
and eliminating intrabrand competition between these common retailers through RPM.
To capture the absence of retail bottleneck in a simple way, we now interpret Dij
as the demand for brand i = A, B at retail location j = 1, 2; and assume that, for
each retail location, each manufacturer has access to at least one potential alternative,
equally efficient retailer. Manufacturers can thus either distribute their products through
the established retailers (who can carry both brands) or bypass them and use instead
alternative (exclusive) retailers. We denote 1A , 1B , 2A and 2B the alternative retailers and
assume that they face the same retail cost γ as the established retailers. In order to stick
as much as possible to the above analysis, we assume that manufacturers first try to deal
with established retailers and therefore adapt the competitive game G by modifying the
second step of the upstream competition stage as follows:
26
The complete analysis is available in an earlier version of this paper (Rey and Vergé, 2004b).
13
(1 − B) Whenever a manufacturer has an offer rejected by a retailer, it proposes a con-
tract to its relevant alternative retailer. All offers to alternative retailers are again
simultaneous and public, as well as their acceptance decisions.
The first step of the upstream competition stage thus still allows the manufacturers
to adopt a common retailer at each location, while the second step now captures the
absence of retail bottleneck: a manufacturer whose offer is rejected in step 1 − A can
still market its product through the alternative retailer in step 1 − B. This, in effect,
prevents manufacturers from trying to foreclose their rivals’ access to consumers; as we
will see, it also encourages retailers to accept any offer that gives them non-negative
profits. More generally, alternative retailers need not be exclusive and might well deal
with both manufacturers; conversely, manufacturers could also make offers to alternative
retailers at stage one as well (see the discussion below). This would not affect the essence
of the analysis but would however complicate its exposition, by increasing the number of
cases to be considered.
In the absence of RPM, a retailer that chooses to carry a single brand — brand A,
say — is likely to face tougher competition, since manufacturer B will then turn to its
alternative retailer, who will no longer internalize the impact of its price on brand A.
In addition, when dealing with its alternative retailer, manufacturer B will no longer
internalize the other retailers’ margins (since their fees have already been negotiated).
This makes manufacturer B more aggressive (through a lower wholesale price for the
alternative retailer), which further tends to result in lower retail prices and downstream
profits. As a result, refusing the offer of one manufacturer in step 1 − A is therefore likely
to make the other manufacturer’s offer less attractive and, as in the previous section the
retailers’ relevant choices are then to accept both offers or none.27 The proof of proposition
1 then carries over, ensuring that in equilibrium, retailers obtain no rent and prices are
somewhat competitive, not only when the manufacturers rely on different retailers in a
given local market, but also when they rely on common retailers.
When RPM is allowed, the preliminary analysis outlines a candidate equilibrium where
manufacturers share the monopoly profit: in this candidate equilibrium, manufacturers
adopt the established retailers as common agents, sell at cost, impose monopolistic retail
prices and extract all profits through franchise fees. By construction, no deviation is
27
Providing general conditions under which mono-branding results in lower retail prices and profits
proves cumbersome, but it holds for example in the linear model that we consider in the next section. It
holds as well if the “alternative retailer” consists of direct distribution: in that case, the wholesale price
goes down to cost and, failing to internalize the impact of its price on the other brand, a mono-brand
retailer moreover sets a lower margin than a multi-brand retailer would do. Retail prices and downstream
profits are then lower whenever retail prices are strategic complements and the retail equilibrium is stable.
14
profitable for a manufacturer if retailers keep accepting the rival’s offers.28 However, by
deviating and opting for a more aggressive behavior, a manufacturer can now discourage a
retailer from carrying the rival brand.29 In essence, such moves allow the deviating man-
ufacturer to act as a Stackelberg leader: imposing a price below the monopoly level forces
the rival to deal with the alternative retailers and therefore to set retail prices that “best
respond” to the deviating manufacturer’s prices. Such deviations are however unattrac-
tive when, as one may expect, Stackelberg profits — which involve some competition — are
lower than monopoly profits.
The following proposition confirms this intuition and shows that, under mild condi-
tions, the previous characterization of double common agency equilibrium outcomes still
applies in the absence of the "market break-down assumption. To introduce the relevant
conditions, we need to consider two hypothetical scenarios of Stackelberg competition: in
the first scenario, the leader (respectively, the follower) produces at cost c + γ the “prod-
ucts” A1 and A2 (respectively, B1 and B2); in the second scenario, the leader produces
three products, A1, A2 and B1, while the follower produces B2. The first scenario is thus
a mere extension of the standard Stackelberg price competition to a symmetric duopoly
in which each firm produces and sells two products, while the second scenario involves
asymmetric firms.
Assumption 3 In the two Stackelberg scenarios just described, the leader’s average profit
is, per product, lower than the monopoly profit.
In the first scenario, the requirement is satisfied whenever prices are strategic com-
plements: Gal-Or (1985) shows indeed that the leader’s profit is then lower than the
follower’s profit,30 and since the industry-wide profit cannot exceed the monopoly level,
the leader’s profit is thus less than half the monopoly profit. Amir and Grilo (1994) note
that the comparison between the leader’s and the follower’s profits is more ambiguous
when they are in an asymmetric position, as in the second scenario; however, there is still
some competition between the two firms, and since the follower sells one product only,
it is likely to be even more aggressive, so that the above requirement sounds again quite
reasonable. Assumption 3 is for example always satisfied in the linear case analyzed in
28
Since each manufacturer gets half the monopoly profit when its offers are accepted by the two retailers,
and retailers will not accept offers that yield negative profits.
29
Retailers will refuse the manufacturer’s offer, which involves a franchise fee equal to the monopoly
profit (per product), whenever they expect rival prices below the monopoly level.
30
When prices are strategic complements, the leader (L) is willing to increase its prices in order to
encourage the follower (F ) to (partially) follow-up and, as a result, in equilibrium L’s prices are higher
than F ’s ones; thus, F “best responds” to L’s comparatively higher prices, while L does not even best
respond to F ’s lower prices.
15
section 5 as well as when prices are strategic complements and there is strong intrabrand
or interbrand competition.31
¡ ¢
Assumption 4 The revenue function π (p) = (p − c − γ) D p, pM , pM , pM is maximal
for a price lower than pM .32
Proposition 3 When RPM is allowed, under Assumptions 3 and 4, there exists a sub-
game perfect equilibrium where manufacturers adopt common retailers (double common
agency) and set wholesale prices to marginal cost (wc = c) and retail prices to the monopoly
¡ ¢
level pc = pM , and achieve monopoly profits (that is, retail profits are zero).
16
some form of potential competition in the downstream market: (i) retailers accept any
offer as long as their expected profit is non-negative; and (ii) manufacturers cannot exclude
their competitors. This derives here from the manufacturers’ ability to use alternative
retailers when an offer has been rejected. Other situations sharing the same ingredients
(i) and (ii) would yield the same outcome:
• There could be more than one alternative retailer, and these alternative retailers
might also carry both brands; in the same vein, the manufacturers could choose
which retailer to contact first. Thus for example, the analysis would carry over
when in each location there exists a competitive supply of potential retailers, to
which the manufacturers propose contracts in turn, until an offer is accepted.
• Instead of using alternative retailers, a manufacturer could also sell directly to con-
sumers. While establishing its own retail outlet might involve some significant set-up
costs, our analysis would carry over as long as those set-up costs do not exceed the
additional profit that they would generate, and as long as the marginal cost of direct
distribution does not significantly exceed that of established retailers. This alterna-
tive might be particularly plausible in sectors where internet sales constitute a good
substitute for in-store sales.
The admittedly ad-hoc but simplifying “market break-down” of the previous section
is thus not crucial and there exists a wide range of situations for which monopoly prices
(through the adoption of common retailers and RPM) constitute a likely outcome. They
are indeed many markets with no retail bottlenecks, such as the car retailing sector for
instance.
Note finally that, while the equilibrium multiplicity issue still arises here, it is however
somewhat less acute than before: some of the previously-described equilibria involve low
industry profits and would therefore be destabilized by a manufacturer’s attempt to con-
vince established retailers to carry only its own brand — thereby placing this manufacturer
in the position of a (admittedly constrained) Stackelberg leader. In addition, the intro-
duction of (arbitrarily small) retail efforts would again single out the equilibrium where
retailers are residual claimants — and retail prices are at the monopoly level.
33
In a previous version of this paper (Rey and Vergé, 2004b), we obtained indeed a similar result using
a framework more directly inspired by Bernheim and Whinston’s original analysis of common agency.
17
5 Retail Market Power
We now turn to situations where manufacturers cannot bypass the established retailers.
The existence of retail bottlenecks raises two issues. First, a manufacturer can now try
to eliminate its rivals, by inducing retailers to carry exclusively its own brand; while
this might induce more competitive outcomes, we show that it may also prevent the
brands from being offered at both stores — despite the fact that there is demand for each
brand at each store. Second, retailers now have some market power and manufacturers
must therefore share the profits with them. As a result, while RPM may again allow
manufacturers to maintain monopoly prices, they may favor an equilibrium with lower
retail prices in order to reduce retail rents — that is, they may prefer more competitive
prices, and have a bigger share of a smaller pie.
Assuming that only the two established retailers (1 and 2) can reach consumers, we
simply remove the part (1 − B) of our game G, i.e., once retailers have decided which
contracts to accept, the game always proceeds to stage 2 (downstream competition). In
a double common agency situation, manufacturers must now ensure that retailers get at
least as much as they could obtain by selling exclusively the rival brand; as we will see,
this implies that manufacturers must leave a rent to retailers — that is, they cannot extract
all the industry profits, even if they can make take-it-or-leave-it offers.34
The existence of these rents — and the fact that they must be evaluated for asym-
metric structures too — somewhat complicates the analysis. We could provide a partial
characterization of double common agency equilibria for general demand structures, but
it is difficult to assess the existence of these equilibria and thus to evaluate the impact of
RPM on prices and profits. In order to shed some light, we therefore restrict attention in
this section to a linear model where costs are normalized to zero, c = γ = 0, and demand
is given by:35
Dij (p) = 1 − pij + αphj + βpik + αβphk ,
18
α + β + αβ < 1 .
where p̃B1 and D̃B1 (respectively p̂A1 and D̂A1 ) denote the prices and quantities that result
from retail competition when retailer 1 carries only brand B (respectively, brand A).37
Removing one brand from one store eliminates one of the available “products”, and
thus increases the demand for the remaining products. This gives retailer 1 an incentive
to raise pB1 ,and the nature of the retail price equilibrium (strategic complementarity of
prices, stability of the equilibrium) then implies that, in the new equilibrium, all retail
prices are higher. Moreover, in the new equilibrium, retailer 1 makes more profit on prod-
uct B − 1 both because of the report from product A − 1 and of the increase in the rival’s
prices. Therefore, (p̃B1 − wB1 )D̃B1 > (prB1 − wB1 )DB1
r
, and a similar argument ensures
that (p̂A1 − wA1 )D̂A1 > (prA1 − wA1 )DA1
r
. Retailer 1 can therefore guarantee itself a pos-
itive profit, and, in a symmetric situation, the retailers’ relevant participation constraint
is thus:
19
possible existence of multiple continuation equilibria for a given set of offers complicates
the analysis. It can for instance be shown that, when the following inequalities hold:
there exist two continuation equilibria: one where both retailers carry both brands (dou-
ble common agency) and in which one retailer carries brand A while its rival carries brand
B (“mon-branding”). Such multiplicity may then be used to sustain equilibria of the con-
tracting game in which retailers obtain more than is necessary to meet their participation
constraints, by punishing deviating manufacturers through a switch to alternative, worse,
continuation equilibria. In the linear model adopted in this section, there exists a thresh-
old β (α) > 0, that guarantees that (3) never holds for any β < β (α), thereby ensuring
that the retailers’ participation constraint must be binding in any (symmetric) common
agency equilibrium.
The next proposition shows that, due to retailers’ market power, it may be the case
that no symmetric equilibrium exists where both retailers carry both brands.
Proposition 4 For any α, there exists a threshold β (α) > 0, such that, without Resale
Price Maintenance, there exists no symmetric equilibrium with double common agency for
β < β (α).
Even though there is a positive demand for each brand at each store, there often
does not exist an equilibrium where both retailers sell both products. The intuition
is the following: in equilibrium, each retailer must be indifferent between accepting or
refusing to carry each particular brand. A deviating manufacturer (manufacturer A, say)
can therefore easily break this indifference and convince one retailer to accept only its
own offer, while ensuring that the second retailer continues to carry both brands. It
can indeed slightly change its wholesale price to break the indifference between carrying
both brands and carrying brand A only (this comparison does not depend on the fixed
fee set by manufacturer A), and slightly change its fixed fee to break the indifference
between carrynig both brands and carrying brand B only. Since the deviation can be
made arbitrarily small, it does not affect the best responses to the other decisions by the
rival retailer, and this guarantees that, in any continuation equilibrium, manufacturer B
is partially excluded: a retailer then carries both brands while its rival only carries brand
A.38 Such a deviation does (almost) not affect the payments received by manufacturer A
38
If the deviation is symmetric, there exists two outcome-equilibrium continuation equilibria.
20
through the fixed fees, but it increases its sales since brand B is not longer carried by one
retailer. The deviation is therefore profitable whenever the wholesale margin is positive.
Suppose now that the wholesale margin is non-positive (w ≤ c) and consider a small
(symmetric) deviation by manufacturer A that consists of offering a wholesale price
v = w ± ε and adjusting its fixed fee to ensure that double common agency is now
the unique continuation equilibrium. This can easily be done since the wholesale price
(resp. fixed fee) can again be adjusted to break the retailers’ indifference towards prefer-
ring to carry both brands rather than brand A (resp. brand B) only. Given our linear
demand specification, it can be shown that it requires increasing the wholesale price (i.e.
v = w + ε, with ε > 0).39 Such a deviation is thus profitable (when the wholesale margin
is positive) when we have:
¡ ¢¯
r
∂ (v − c) DAj (v, w, v, w) + πr (v, w; v, w) − πr (∅, w; v, w) ¯¯
¯ > 0,
∂v ¯
v=w
where D (pij , ∅, phj , phk ) denotes the demand for brand i at retailer j when this retailer
carries only that brand. Since removing a product increases the demand for the remaining
ones, D (p, ∅, p, p) > D (p, p, p, p), and retailers thus earn again a positive rent when the
retailer margin (p − w) is positive. The next proposition shows that such equilibria do
exist and describe some of their properties:
39
This would also be the case with general demands (also it requires some additional assumptions on
the profit function π r ), whenever πr (v, w; v, w) − π r (v, ∅; v, w) is increasing in v.
21
RP M
Proposition 5 For any α, there exists a threshold β (α) > 0 such that, for any
RP M
β ≤ β (α), there exists a continuum of symmetric equilibria with RPM and dou-
RP M
ble common agency. More precisely, for any β < β (α), there exist p (α, β) ≤ pM
and p (α, β) ≥ pM (with at least one of the inequality being strict), such that for any
£ ¤
p∗ ∈ p (α, β) , p (α, β) , there exists a symmetric equilibrium with RPM where, for any
i = A, B and any j = 1, 2, pij = p∗ and:
1 − α(1 − β)w∗
p∗ = ;
2(1 − α − β)
Note that proposition 5 only provides sufficient conditions for the existence of sym-
metric equilibria with double common agency. There may exist other equilibria, including
other symmetric double common agency equilibria. Figure 1 represents the range of values
for which the results of propositions 4 and 5 apply.
Despite the presence of retail rents, the equilibrium retail price is still inversely related
to the equilibrium wholesale price. When manufacturer h offers both retailers a wholesale
price wh and imposes a retail price ph , manufacturer i’s best response, pe (ph , wh ), is given
by:
Two effects are now at work. As in the absence of retail rents, manufacturer i has an
incentive to increase the sales of its own products by being more aggressive, since it
earns the full margin on these products and only internalizes (through the franchise fees)
the retail margin on the sales of its rival’s products. Moreover, this incentive to free-
ride on the sales of the rival’s products is greater, the lower the retail margin on those
products. Therefore, in the absence of rents, an increase in the wholesale price wh makes
manufacturer i more aggressive.
However, the rent effect (the negative term on the second line of the above program)
goes in the opposite direction. In order to reduce the rent left to retailer j, a manufacturer
has an incentive to impose a low retail price on its rival, as this lowers the demand for
22
β
1 No Double Common Agency
Equilibrium with Two-Part Tariffs
0.6
0.4
0.2
retailer j. However, an increase in wh reduces, ceteris paribus, all retailers’ rents; which
in turn reduces the manufacturer incentives to behave aggressively. This effect would
thus increase manufacturer h’s reaction function to any price ph . This second effect is
however always dominated in the linear demand case; manufacturer h’s reaction function,
pe (ph , wh ), thus remains decreasing in wh , as in the absence of rents, and the equilibrium
retail p∗ is again inversely related to the equilibrium wholesale price w∗ .
This rent effect also affects the equilibrium wholesale price that is necessary to sustain
monopoly retail prices. In the absence of rents, if manufacturer h sets its wholesale prices
equal to marginal cost (wh = c), manufacturer i’s profit coincides (up to a constant, equal
to the sum of its rival’s franchise fees) with the industry profit. As a result, if manufacturer
h sets its retail prices at the monopoly level, manufacturer i’s best response is to set its
¡ ¡ ¢ ¢
own retail prices at the monopoly level too pb pM , c = pM . This is no longer the case
here, since manufacturer i has now additional incentives to lower its retail prices, in order
to reduce the rent left to retailers. Therefore, its best response to ph is lower than in the
absence of rents: pe (ph , wh ) < pb (ph , wh ) . This implies that, for wholesale prices equal to
marginal cost (w∗ = c) , the corresponding (symmetric) equilibrium retail price is lower
than the monopoly price. In the linear demand case (with c = γ = 0), we have indeed:
1
p∗ (0) = < pM .
2(1 − α − β)
23
¡ ¢
However, since p̄ ≥ pM , there still exists some wM ∈ [w, 0] such that p∗ wM = pM :
manufacturers can sustain monopoly prices, but to do so they must set wholesale prices
below their marginal cost of production.
Subsidizing wholesale prices increases retail rents, however. In equilibrium, this rent
(per retailer and per brand) is equal to:40
Therefore,
1 dπ ∗R d (p∗ − w∗ ) ∗ ∗ ∗ dD
∗
= D + (p − w ) .
α dp∗ dp∗ dp∗
Given the inverse relationship between p∗ and w∗ , the mark-up (p∗ − w∗ ) increases with
p∗ and this effect dominates when p∗ is small, since then (p∗ − w∗ ) is small and D∗ is
large.
Manufacturers’ profits (per retailer) are of the form π ∗P = p∗ D∗ − π ∗R . Hence, starting
from p∗ = p, manufacturers face a trade-off between increasing industry profits (by raising
retail prices to the monopoly level) and reducing retail rents (by maintaining low retail
prices). Proposition 5 shows that in this linear model, the rent effect dominates; therefore:
Corollary 1 Among the equilibria with double common agency described in proposition 5,
manufacturers prefer the equilibrium with the lowest retail price, whereas retailers prefer
the equilibrium with the highest retail price, which exceeds the monopoly level.
6 Policy Discussion
This paper highlights how RPM can eliminate any scope for effective competition when
producers distribute their goods through the same competing retailers (“interlocking re-
lationships”). The intuition is relatively simple. As in the single common agent case,
distributing their products through the same retailers allow the manufacturers to elim-
inate, or at least soften, interbrand competition. However, when dealing with several
(common) retailers, intrabrand competition dissipates profits and prevents manufacturers
from maintaining monopoly prices. In this context, RPM can restore monopoly prices
and profits. In essence, RPM eliminates competition between retailers, while “common
agency” eliminates competition between manufacturers. Since the mechanism identified
by our analysis cannot be replicated through other vertical restraints (e.g., exclusive deal-
ing or exclusive territories), this paper offers one of the few arguments to justify the
negative attitude of the courts towards price restrictions.
40
By definition, D (p∗ , ∅, p∗ , p∗ ) = D (p∗ , p, p∗ , p∗ ) , where p = 1 + (α + β + αβ) p∗ . We thus have:
D (p∗ , ∅, p∗ , p∗ ) = 1 − p∗ + αp + βp∗ + αβp∗ = (1 + α) D (p∗ , p∗ , p∗ , p∗ ) .
24
Our analysis thus supports the concerns of the French Conseil de la Concurrence when,
as mentioned in the introduction, it condemned (in three separate cases) brown goods,
perfume and toy manufacturers for engaging, through RPM, into “vertical collusion” with
leading multi-brand retailers. It also supports the ongoing efforts to reform the French law,
adopted in 1996, that allowed manufacturers to impose de facto price floors by abusing no-
resale-below-cost regulations, and which has been blamed for the important price increases
that have taken place in the last decade, especially for national brands in supermarket
chains. Our analysis supports this claim and shows that RPM can actually eliminate
competition, not only among competing fascias, but also among competing brands. This
possibility has been validated by recent empirical studies. Using data about retail prices
of food products in French retail chains during the period 1994-1999, Biscourp, Boutin
and Vergé (2008) find that the correlation between retail prices and the concentration of
local retail markets was important before 1997 and no longer significant after that date.
This suggests that the price increases that occurred after 1997 were indeed due to the
impact of the new legislation on intrabrand competition.
Our analysis also shows that double agency equilibria may fail to exist in the pres-
ence of real retail bottlenecks, while there can be a continuum of equilibrium retail prices
with RPM (including at the monopoly level). The existence of equilibria of given types
(common multi-brand retailers versus mono-brand retailers, for example), as well as the
choice of a particular equilibrium, constitute interesting questions, which recent empiri-
cal analyses have started to address: Bonnet and Dubois (2004 and 2007), for instance,
study the French market of bottled water during the 1998-2001 period, using a structural
econometric model based on micro-level data. They build on Berto Villas-Boas (2007),
who extends the empirical approach developed by Berry, Levinsohn and Pakes (1995) to
multiple stages of competition (upstream competition among manufacturers and down-
stream competition among retailers), to compare different scenarii: linear or two-part
tariffs for wholesale prices, RPM or not for retail prices, etc. In particular, Bonnet and
Dubois (2004) carry a structural estimation of a model directly inspired by the analysis
presented in our section 4. Comparing different types of vertical contracts, they conclude
that the most likely scenario is the one with two-part tariffs, RPM and no retail margin.41
Their finding thus supports the interpretation of the 1996 law as legalizing RPM, as well
as our analysis of its impact on prices and profits. When simulating the impact of an ef-
fective ban on RPM, they find that retail prices would decrease by about 7% on average.
Bonnet and Dubois (2007) extends the analysis to allow for endogenous retail rents (as in
our section 5) and conclude again that two-part tariffs and resale price maintenance are
widely used, but also that the most likely scenario is one where retailers’ outside options
41
Berto Villas-Boas (2007) studies the distribution of yoghurts by supermarkets in California and also
finds that non-linear prices are widely used. However, she does not test for the presence of RPM.
25
are fixed (thus closer to the variant we study in section 4).
Our analysis thus suggests a cautious attitude towards price restrictions in situations
where rival manufacturers rely on the same competing retailers, even — and possibly more
so — in the absence of retail bottlenecks.
26
References
[1] Allain, Marie-Laure and Claire Chambolle (2007), “Forbidding Resale at a Loss: A
Strategic Inflationary Mechanism”, mimeo.
[2] Amir, Rabah and Isabel Grilo (1994), “Stackelberg versus Cournot/Bertrand Equi-
librium”, Universite Catholique de Louvain CORE Discussion Paper 9424.
[3] Bernheim, Douglas and Michael Whinston (1985), “Common Agency as a Device for
Facilitating Collusion”, Rand Journal of Economics, 16(2), 269-281.
[4] Bernheim, Douglas and Michael Whinston (1998), “Exclusive Dealing”, Journal of
Political Economy, 106, 64-103.
[5] Berry, Steven, James Levinsohn and Ariel Pakes (1995), “Automobile Prices in Mar-
ket Equilibrium”, Econometrica, 63(4), 841-890.
[6] Berto Villas-Boas, Sofia (2007), “Vertical Relationships Between Manufacturers and
Retailers: Inference With Limited Data”, Review of Economic Studies, 74(2), 625-
652.
[7] Biscourp, Pierre, Xavier Boutin and Thibaud Vergé (2008), “The Effects of Retail
Regulations on Prices: Evidence from the Loi Galland”, INSEE-DESE Discussion
Paper G2008/02.
[8] Bonanno, Giacomo and John Vickers (1988), “Vertical Separation”, Journal of In-
dustrial Economics, 36, 257-265.
[9] Bonnet, Céline and Pierre Dubois (2004), “Inference on Vertical Contracts between
Manufacturers and Retailers Allowing for Non Linear Pricing and Resale Price Main-
tenance”, mimeo.
[10] Bonnet, Céline and Pierre Dubois (2007), “Non Linear Contracting and Endogenous
Buyer Power between Manufacturers and Retailers: Identification and Estimation on
Differentiated Products”, mimeo.
[11] Caballero-Sanz, Francesco and Patrick Rey (1996), “The Policy Implications of the
Economic Analysis of Vertical Restraints”, Economic Papers n ◦ 119, European Com-
mission.
[12] Caillaud, Bernard and Patrick Rey (1987). “A Note on Vertical Restraints with the
Provision of Distribution Services.” Working Paper INSEE and MIT.
27
[13] Caillaud, Bernard and Patrick Rey (1995), “Strategic Aspects of Vertical Delegation”,
European Economic Review, 39, 421-431.
[14] Chen, Yongmin (1999), “Oligopoly Price Discrimination and Resale Price Mainte-
nance”, Rand Journal of Economics, 30(3), 441-455.
[15] Comanor, William (1985), “Vertical Price Fixing and Market Restrictions and the
New Antitrust Policy”, Harvard Law Review, 98, 983-1002.
[16] Comanor, William and Patrick Rey (1997), “Competition Policy towards Vertical
Restraints in the US and Europe”, Empirica, 24(1-2), 37-52.
[17] Deneckere, Raymond, Howard P. Marvel and James Peck (1996), “Demand Uncer-
tainty, Inventories, and Resale Price Maintenance”, Quarterly Journal of Economics,
111, 885-913.
[18] Deneckere, Raymond, Howard P. Marvel and James Peck (1997), “Demand Uncer-
tainty and Price Maintenance: Markdowns as Destructive Competition”, American
Economic Review, 87, 619-641.
[19] Dobson, Paul and Michael Waterson (2007), “The Competition Effects of Industry-
wide Vertical Price Fixing in Bilateral Oligopoly”, International Journal of Industrial
Organization, 25(5), 935-962.
[21] Gal-Or, Esther (1985), “First Mover and Second Mover Advantages”, International
Economic Review, 26(3), 649-653.
[22] Hart, Oliver and Jean Tirole (1990), “Vertical Integration and Market Foreclosure”,
Brookings Papers on Economic Activity: Microeconomics, 205-276.
[23] Jullien, Bruno and Patrick Rey (2007), “Resale Price Maintenance and Collusion”,
Rand Journal of Economics, 38(4), 983-1001.
[24] Marvel, Howard P. and Stephen McCafferty (1984), “Resale Price Maintenance and
Quality Certification”, Rand Journal of Economics, 15(3), 346-359.
[25] Mathewson, Frank and Ralph Winter (1998), “The Law and Economics of Resale
Price Maintenance”, Review of Industrial Organization, 13, 57-84.
[26] McAfee, R. Preston and Marius Schwartz (1994), “Opportunism in Multilateral Ver-
tical Contracting: Nondiscrimination, Exclusivity and Uniformity”, American Eco-
nomic Review 84(1), 210-230.
28
[27] Motta, Massimo (2004), Competition Policy: Theory and Practice, MIT Press.
[28] O’Brien, Daniel and Greg Shaffer (1997), “Non-Linear Supply Contracts, Exclusive
Dealing, and Equilibrium Market Foreclosure”, Journal of Economics and Manage-
ment Strategy, 6, 755-785.
[30] Rey, Patrick and Joseph Stiglitz (1988), “Vertical Restraints and Producers Compe-
tition”, European Economic Review, 32, 561-568.
[31] Rey, Patrick and Joseph Stiglitz (1995), “The Role of Exclusive Territories in Pro-
ducer’s Competition”, Rand Journal of Economics, 26, 431-451.
[32] Rey, Patrick and Jean Tirole (1986), “The Logic of Vertical Restraints”, American
Economic Review, 76, 921-939.
[33] Rey, Patrick and Jean Tirole (2007), “A Primer on Foreclosure”, in Mark Armstrong
and Rob Porter Eds., Handbook of Industrial Organization, Vol 3, North-Holland.
[34] Rey, Patrick and Thibaud Vergé (2004a), “Bilateral Control with Vertical Contracts”,
Rand Journal of Economics, 35(4), 728-746.
[35] Rey, Patrick and Thibaud Vergé (2004b), “Resale Price Maintenance and Horizontal
Cartel”, CMPO Working Paper 02/047.
[36] Rey, Patrick and Thibaud Vergé (2008), “The Economics of Vertical Restraints”, in
Paolo Buccirossi Ed., Handbook of Antitrust Economics, MIT Press.
[37] Schulz, Norbert (2007), “Does the Service Argument Justify Resale Price Mainte-
nance?”, Journal of Institutional and Theoretical Economics, 163(2), 236-255.
[38] Spence, Michael (1975), “Monopoly, Quality and Regulation”, Bell Journal of Eco-
nomics, 6, 417-429.
[39] Telser, Lester (1960), “Why Should Manufacturer Want Fair Trade,” Journal of Law
and Economics, 3, 86-105.
[40] Wang, Hao (2004), “Resale price maintenance in an oligopoly with uncertain de-
mand”, International Journal of Industrial Organization, 22, 389-411.
29
A Proof of Proposition 1
We first show that equilibrium upstream margins are positive (we > c). The conclusion
then follows from the fact that manufacturers fail to account for (and thus “free-ride” on)
their rivals’ upstream margins. At a symmetric equilibrium of the form (pij = pe , wij = we ),
manufacturer i must find it optimal to choose wi1 = wi2 = we when its rival adopts
wh1 = wh2 = we ; w = we must therefore maximize:
h i
e e e
p (w, w ) − c − γ)D (w, w ) + (e
2 (e e e e e
p (w , w) − w − γ)D (w , w) .
The first-order condition yields (with D evaluated at pe and the derivatives of D e and
e + (pe − we − γ) ∂2 D
pe evaluated at (we , we )) : (∂1 pe + ∂2 pe) D + (pe − c − γ) ∂1 D e = 0,
implying:
³ ´
e e e
(∂1 pe + ∂2 pe) D + ∂1 D + ∂2 D (pe − we − γ) = − (we − c) ∂1 D. (4)
e = λM ∂1 pe + λ̂M ∂2 pe and ∂2 D
Note that ∂1 D e = λM ∂2 pe + λ̂M ∂1 pe , where λM ≡ ∂1 D + ∂3 D
represents the marginal impact on the demand for “product” i − j of a uniform increase
in the retail prices for brand i, while λ̂M ≡ ∂2 D + ∂4 D represents instead the impact of
the rival manufacturer’s retail prices. Therefore, (4) can be rewritten as:
e
(∂1 pe + ∂2 pe) [D + λ (pe − we − γ)] = − (we − c) ∂1 D, (5)
where λ ≡ λM + λ̂M represents the impact on demand of a uniform increase in all retail
prices and is thus negative. But a symmetric retail equilibrium is characterized by the
first-order condition:
D = −λR (pe − we − γ) , (6)
where λR ≡ ∂1 D+∂2 D represents the impact on the demand for “product” ij of a uniform
increase in retailer j’s prices. Combining (5) and (6) yields:
e
(∂1 pe + ∂2 pe) λ̂R (pe − w − γ) = − (we − c) ∂1 D, (7)
e + ∂2 D
Given that ∂1 D e = λ (∂1 pe + ∂2 pe) and ∂1 pe + ∂2 pe > 0, having we > c implies that
D + λ (pe − c − γ) > 0 . This in turn implies that, starting from p = pe , a uniform increase
in all prices increases the monopoly profit. By assumption the monopoly profit is single-
peaked at pM and thus, pe < pM .
30
B Proof of Proposition 2
If manufacturer h adopts wh1 = wh2 = w∗ and ph1 = ph2 = p∗ , from Assumption 2, man-
ufacturer i’s revenue function Π is single-peaked in (pi1 , pi2 ) and maximal for symmetric
prices, p̂i1 = p̂i2 = p̂ (p∗ , w∗ ); this price maximizes Π (p, p∗ , p, p∗ , w∗ , w∗ ) and thus solves:
with λM and λ̂M as defined in the previous section. To establish that p∗ decreases when
w∗ increases, note first that ∂13
2
f = −λ̂M < 0 . Therefore, a standard revealed preference
argument leads to ∂2 p̂ < 0. From Assumption 2, 0 < ∂1 p̂ < 1, implying that the fixed
point of p → p̂ (p, w∗ ) decreases when w∗ increases.
C Proof of Proposition 3
The proof is constructive and based on the following candidate equilibrium path: both
¡ ¡ ¢ ¡ ¢¢
manufacturers offer the contract C c = wc = c, pc = pM , F c = pM − c − γ D pM to
the “established” retailers 1 and 2 and all four offers are accepted at stage 1−A. Retailers
thus make zero profits and manufacturers share the monopoly profit.
31
Manufacturer h’s offers have both been accepted
We can easily rule out any such deviation since both retailers j and k would then have
πM
accepted to pay F c = 4
each to the manufacturer h. Since the industry profit cannot
exceed π M , and a retailer would never accept an offer that generates losses, manufacturer
πM
i will never be able to achieve more than π M − 2F c = 2
.
At stage 1 − B, manufacturer h thus deals with the alternative retailers (1h and 2h ) and
chooses the prices, ph1 and ph2 , that are its best replies to the prices pi1 and pi2 that have
either been accepted by the retailer(s) at stage 1 − A, or that are set by manufacturer i
(dealing with retailers 1i and 2i ) at stage 1 − B. The prices ph1 and ph2 are therefore equal
to the prices set by the follower of our first Stackelberg scenario when the leader sets prices
pi1 and pi2 .42 Given that manufacturer i’s profit can only come from the sales of products
i − 1 and i − 2 (through either the “established” or the “alternative” retailers), its profit
cannot exceed that of the leader of this first Stackelberg scenario which, by Assumption
πM
3, is lower than 2
.
Only one of manufacturer h’s offers has been accepted (say, by retailer j)
32
πM
exceed that of the leader of the second Stackelberg scenario minus 4
. Under Assumption
3π M πM πM
3, this profit is lower than 4
− 4
= 2
.
• Suppose finally that the deviation is such that the offer i − j is rejected. For
such a situation to arise at the end of stage 1 − A, the contracts must be such that
retailer j expects its retail profit (on product h − j) to cover the franchise to be paid to
manufacturer h. This means that the profit generated by product h − j has to be larger
πM
than 4
. However, if this is the case, manufacturer i would rather make an offer to retailer
j (rather than distributing the product through the alternative retailer ji ) to recover all
M
the profit generated above π4 on product h − j.
D Proof of Proposition 4
We now focus here on values of the parameters α and β for which:
In the linear demand case, it can be shown that the condition (8) does not depend on w,
and simply rewrites as β < β (α) , where β (α) > 0 is uniquely defined. We assume, in
what follows, that β ≤ β (α) , i.e. the gain from accepting both offers rather manufacturer
A’s offer only is larger when the rival retailer has accepted manufacturer B’s offer only
than when it has accepted both offers.
In the linear demand, it can be shown, that this gain is also larger than when the rival
retailer has rejected both offers, i.e., whenever β < β (α) , we have:
In lemma 1, we had already shown that, for double common agency to be an equi-
librium, it suffices to check that each retailer preferred to accept both offers rather than
one, when its rival accepts both offers. The only relevant constraint is then:
All of this implies that whenever (9) is satisfied, the unique continuation equilibrium
is one where the two retailers decide to carry both brands. If (9) were not binding, it
would then be profitable for a manufacturer to slightly increase its franchise fees, since
this would affect neither the equilibrium market structure, nor the equilibrium prices and
quantities. As a result, in any (symmetric) double common agency equilibrium, we have:
F = π r (w, w; w, w) − π r (w, ∅; w, w) .
To show that no symmetric, double common agency equilibrium exists, we now build,
for any admissible value of w, a profitable deviation for one of the manufacturers.44 We
44
A value of w is admissible, whenever the corresponding retail price and quantities are positive.
33
first concentrate on w > 0 , and consider a deviation where manufacturer A offers the
contract (v, G) such that v = w − ε and G = π r (v, ∅; v, w) − π r (∅, w; v, w) + F − η ,with
ε, η > 0. Note that, when ε and η tend to 0, (v, G) tends to (w, F ) . Therefore, for ε and
η small enough, it is still a best reply for a retailer to accept both contracts whenever the
rival retailer has rejected at least one contract. Suppose now that retailer k has accepted
both offers. Since ε and η are small, it cannot be a best response to reject both offers. As
we now show, it is a best response to accept only manufacturer A’s offer:
Moreover,
condition which is always satisfied in the linear demand case whenever v < w. Such a
deviation thus allows manufacturer A to partially exclude its rival; it also increases the
sales of its products, without (substantially) affecting the franchise fees it receives. The
deviation is thus profitable as long as the wholesale margin is strictly positive (w > 0) .
G = π r (v, w; v, w) − π r (∅, w; v, w) − η,
with ε, η > 0 close to 0. In the linear case, it is possible to check that double common
agency is the unique continuation equilibrium, and that the deviation, which yields the
following profit to manufacturer A :
Πdev r r r
A (v; w) = 2vDij (v, w, v, w) + π (v, w; v, w) − π (∅, w; v, w) − 2η,
is indeed profitable for some ε, η > 0 close to 0. This thus concludes the proof that no
“double common agency equilibrium” exists when β < β (α) .
E Proof of Proposition 5
We look for sufficient conditions on w∗ to ensure that C ∗ = (w∗ , p∗ , F ∗ ), where
1 − α(1 − β)w∗
p∗ = and F ∗ = (1 − α)(p∗ − w∗ )D (p∗ ) ,
2(1 − α − β)
is the equilibrium wholesale contract of a symmetric double common agency equilibrium.
We only sketch the proof here; a more detailed proof is available upon request. Note that
34
we have constrained the retail price p∗ to be higher than the marginal wholesale price
1
w∗ , therefore imposing w∗ ≤ wmax = 2−α−2β−αβ
. Moreover, quantities must be positive,
thereby constraining w∗ to be such that:
In what follows, we only provide sufficient conditions that guarantee that no deviation
by a manufacturer can be profitable. Depending on the contracts Ci1 and Ci2 offered
by manufacturer i, the set of contracts accepted by the retailers differs, and we should
therefore analyze the effects of a deviation on manufacturer i’s profits for any possible
market structure. There are 16 such structures, but symmetry reduces this number to
10. Moreover it cannot be profitable for a manufacturer to deviate in such a way that its
contracts would both be rejected, leaving only 7 structures to analyze.
(pij − wij )D (pi1 , p∗ , pik , p∗ ) − Fij + (p∗ − w∗ )D (p∗ , pij , p∗ , pik ) − F ∗ ≥ ...
... ≥ 0 (10)
... ≥ (p∗ − w∗ )D (p∗ , ∅, p∗ , pik ) − F ∗ (11)
... ≥ (pij − wij )D (pij , ∅, pik , p∗ ) − Fij (12)
Constraint (11) implies that maximal fixed fee that manufacturer i can obtain from
retailer j is:
Fij = (pij − wij )D (pij , p∗ , pik , p∗ ) + (p∗ − w∗ ) (D (p∗ , pij , p∗ , pik ) − D (p∗ , ∅, p∗ , pik )) .
This profit is maximized for pij = pik = p∗ , so that a deviation to “double common
agency” cannot be profitable.
(pij − wij ) D(pij , p∗ , pik , ∅) − Fij + (p∗ − w∗ )D(p∗ , pij , ∅, pik ) − F ∗ ≥ ...
35
... ≥ 0 (13)
... ≥ (p∗ − w∗ )D(p∗ , ∅, ∅, pik ) − F ∗ (14)
... ≥ (pij − wij ) D(pij , ∅, pik , ∅) − Fij (15)
and
(pik − wik ) D(pik , ∅, pij , p∗ ) − Fik ≥ ...
... ≥ 0 (16)
... ≥ (p∗ − w∗ )D(p∗ , ∅, p∗ , pij ) − F ∗ (17)
... ≥ (pik − wik ) D(pik , p∗ , pij , p∗ ) − Fik + (p∗ − w∗ )D(p∗ , pik , p∗ , pij ) − F ∗ (18)
Wholesale prices wij and wik can be set so that constraints (15) and (18) are satisfied.
If manufacturer i sets the maximal possible fixed fees, its profit is:
π S1 (pij , pik ) = pij D(pij , p∗ , pik , ∅) − max [0, (p∗ − w∗ ) (D(p∗ , ∅, ∅, pik ) − (1 − α)q ∗ )]
+pik D(pik , ∅, pij , p∗ ) − max [0, (p∗ − w∗ ) (D(p∗ , ∅, p∗ , pij ) − (1 − α)q ∗ )]
+(p∗ − w∗ ) [D(p∗ , pij , ∅, pik ) − (1 − α)q∗ ]
It is now sufficient to compare the maximal value of this profit with π ∗P (w∗ ), that is,
consider the sign of the expression: ∆1 (w∗ ) = max π S1 (pij , pik ) − π ∗P (w∗ ) . With linear
pij ,pik
demands, we can show that there exist a non-trivial interval [w1 (α, β), w1 (α, β)] within
which such a deviation is never profitable, i.e.: ∆1 (w∗ ) ≤ 0 .
Remaining Structures
Repeating the analysis for each of the five remaining structures generates additional re-
strictions, wmin ≤ wi (α, β) ≤ w∗ ≤ wi (α, β) ≤ wmax , for any i = 2, ..., 6. Let us now de-
fine: ∙ ¸
M
w(α, β) = max w , max wi (α, β) and w (α, β) = min wi (α, β).
i=1,...,6 i=1,...,6
We then verify that, for the values of the parameters α and β given by figure 1, i.e.
RP M
whenever β ≤ β (α) , we have w(α, β) = wM ≤ w(α, β) ⇔ p(α, β) ≤ pM ≤ p(α, β) .
36