-
Joint Venture for Product Innovationand Cartel Stability under
Vertical
Di¤erentiation
Cristina Iori# - Luca Lambertinix
Department of Economics, University of BolognaStrada Maggiore
45, 40125 Bologna, Italy
fax +39-051-2092664# e-mail: [email protected]
§ e-mail: [email protected]
July 18, 2000
Abstract
We describe a vertically di¤erentiated market where …rms
choosebetween activating either independent ventures leading to
distinctproduct qualities, or a joint venture for a single quality.
Then, …rmseither repeat the one-shot Nash equilibrium forever, or
behave col-lusively, according to discount factors. We prove that
there exists aparameter region where the joint venture makes it
more di¢cult for…rms to sustain collusive behaviour, as compared to
independent ven-tures. Therefore, public policies towards R&D
behaviour should bedesigned so as not to become inconsistent with
the pro-competitiveattitude characterising the current legislation
on marketing practices.
J.E.L. classi…cation: C72, D43, L13Keywords: product quality,
R&D investment, implicit collusion,
joint venture, independent ventures
1
-
1 IntroductionOligopoly theory has produced a relevant
literature on repeated market inter-action. The relative e¢ciency
of Bertrand and Cournot competition in stabi-lizing cartels
composed by …rms whose products are imperfect substitutes hasbeen
analysed by Deneckere (1983), Rothschild (1992) and Albæk and
Lam-bertini (1998), showing that when substitutability between
products is high,collusion is better supported in price-setting
games than in quantity-settinggames, while the reverse is true in
case of low substitutability.1 Majerus(1988) has proved that this
result is not con…rmed as the number of …rmsincreases. These
contributions compare Cournot and Bertrand supergamesto conclude
that a quantity-setting cartel should almost always be preferredto
a price-setting cartel on stability grounds.2 Finally, the in‡uence
of en-dogenous product di¤erentiation on the stability of collusion
in prices hasbeen investigated by Chang (1991, 1992), Ross (1992)
and Häckner (1994,1995, 1996). The main …nding reached by these
contributions is that, undervertical di¤erentiation, collusion is
more easily sustained, the more similarthe products are, while the
opposite applies under horizontal di¤erentiation.
The consequences of collusion on the extent of optimal
di¤erentiation inthe horizontal di¤erentiation model have also
received attention. Friedmanand Thisse (1993) have considered a
repeated price game in the horizontalframework and found out that
minimum di¤erentiation obtains if …rms col-lude in the market
stage. In most of these models, although di¤erentiationcan be
endogenously determined by …rms through strategic interaction,
theissue of cartel stability is studied by making the degree of
di¤erentiationvary symmetrically around the ideal midpoint of the
interval of technolog-ically feasible or socially preferred
varieties, leading to the conclusion thatproducers may prefer to
choose the characteristics of their respective goodsdi¤erently from
what pro…t maximization would suggest, if this helps themminimize
the incentive to deviate from the implicit cartel agreement.
1The same question is addressed in Lambertini (1996), where the
evaluation of cartelstability under Bertrand and Cournot behaviour
is carried out in terms of the concav-ity/convexity of the market
demand function.
2This approach cannot grasp any strategic interaction behind the
choice of the marketvariable. Using the same demand structure as in
Deneckere (1983) and analysing asymmet-ric cartels where one …rm is
a Bertrand agent while the other is a Cournot one, Lambertini(1997)
proves that the choice of the market variable in order to stabilize
implicit collusionproduces a Prisoner’s Dilemma.
2
-
To our knowledge, little attention has been paid so far to the
interplaybetween …rms’ technological decisions and their ability to
build up and main-tain collusive agreements over time. This is a
relevant issue, in that publicauthorities prosecute collusive
market behaviour, while they seldom discour-age cooperation in
R&D activities. Indeed, there exist many examples ofpolicy
measures designed so as to stimulate the formation of research
jointventures.3 However, encouraging cooperative R&D and
discouraging marketcollusion can be mutually inconsistent moves, if
R&D cooperation tends tofacilitate collusion in the product
market.
In this respect, Martin (1995) analyses the strategic e¤ects of
a researchjoint venture (JV henceforth) designed to achieve a
process innovation foran existing product. Then, the product is
marketed by …rms engaging inrepeated Cournot behaviour over an
in…nite time horizon. Martin showsthat cooperation in process
innovation enhances implicit collusion, which canjeopardise the
welfare advantage of eliminating e¤ort duplication through theJV .
This result has potential implications for the case of product
innovationas well.4
We reassess Martin’s framework, by considering a vertically
di¤erentiatedmarket where …rms are given the possibility of
choosing between activatingeither independent ventures leading to
distinct product qualities, or a jointventure for a single quality,
aimed at reducing the initial R&D expenditurevis à vis
independent ventures. Then, …rms market the product(s) over an
in-…nite horizon. In doing so, they either repeat the one-shot Nash
equilibriumforever, or behave collusively, according to their
intertemporal discounting.In such a setting, we prove that there
exists a parameter region where Mar-tin’s conclusion is reversed,
i.e., the JV makes it more di¢cult for …rms tosustain collusive
behaviour in the market supergame, as compared to inde-pendent
ventures. This holds independently of whether …rms set prices
orquantities during the supergame. Our result entails that public
policies to-wards the R&D behaviour of …rms should be tailored
case by case, so as notto become inconsistent with the
pro-competitive attitude characterising the
3See the National Cooperative Research Act in the US ; EC
Commission (1990) ; and,for Japan, Goto and Wakasugi (1988).
4Lambertini, Poddar and Sasaki (1998) adopt the same view as in
Martin (1995), al-though they consider the relationship between
standardization and the stability of implicitcartel agreeements.
See also Lambertini, Poddar and Sasaki (2000). Cabral (1996), in
asomewhat dissimilar vein, proves the possibility that competitive
pricing is needed tosustain more e¢cient R&D agreements.
3
-
current legislation on marketing practices.The remainder of the
paper is structured as follows. The basic model
of vertical di¤erentiation is described in section 2. Section 3
describes thecase of collusion along the frontier of monopoly
pro…ts. Section 4 deals withpartial collusion under either Cournot
or Bertrand behaviour. Finally, section6 provides concluding
remarks.
2 The vertical di¤erentiation modelWe adopt a well known model
of duopoly under vertical di¤erentiation (seeGabszewicz and Thisse,
1979, 1980; Motta, 1993; Aoki and Prusa, 1997;Lehmann-Grube, 1997;
Lambertini, 1999, inter alia).5 Two single-product…rms, labelled as
H and L, produce goods of (di¤erent) qualities qH andqL 2 [0;1),
with qH ¸ qL; through the same technology, C(qi) = cq2i ; withc
> 0: This can be interpreted as …xed cost due to the R&D
e¤ort needed toproduce a certain quality, while variable production
costs are assumed away.Products are o¤ered on a market where
consumers have unit demands, andbuy if and only if the net surplus
from consumption vµ(qi; pi) = µqi ¡ pi ¸ 0;where pi is the unit
price of the good of quality qi, purchased by a genericconsumer
whose marginal willingness to pay is µ 2 [0; µ]: We assume that µis
uniformly distributed with density one over such interval, so that
the totalmass of consumer is µ.
Firms interact over t 2 [0;1); as follows:² At t = 0; they
conduct R&D towards the development of product
quality, through either a joint venture (JV henceforth) or
independentventures (IV henceforth). If …rms undertake a joint
venture, thenqi = qj = q and each …rm bears half the development
cost, cq2=2.Otherwise, …rms market di¤erentiated products, each of
them bearingthe full development cost of their respective
varieties, cq2i .
6
² Over t 2 [1;1); …rms market the product(s) resulted from
previousR&D activity, either à la Cournot or à la Bertrand.
5A di¤erent model is used in Shaked and Sutton (1982, 1983),
where …xed costs areexogenous.
6The R&D e¤orts of …rms operating in vertically di¤erentited
markets are investigatedin Beath et al. (1987), Motta (1992), Dutta
et al. (1995), Rosenkranz (1995, 1997), vanDijk (1996). In
particular, Motta (1992) and Rosenkranz (1997) describe the
incentivestowards cooperative R&D.
4
-
² In the in…nitely long marketing phase, …rms may collude if
their respec-tive time discounting allows them to do so. Otherwise,
they always playà la Nash. De…ne as ±i the discount factor of …rm
i; and ±Ii (K) the crit-ical threshold for the stability of
collusion, with superscript I = B;Cstanding for Bertrand and
Cournot, and K = IV; JV; indicating theorganizational design chosen
for the R&D phase.
As a …rst step, observe that the locations of indi¤erent
consumers along[0; µ] are:
µH =pH ¡ pLqH ¡ qL
; µL =pLqL
(1)
where µH is the marginal willingness to pay of the consumer who
is indi¤erentbetween qH and qL; and µL is the marginal willingness
to pay of the consumerwho is indi¤erent between qL and not buying
at all. Then, market demandsare
xH = µ ¡ µH ; xL = µH ¡ µL : (2)Notice that (2) can be inverted
to yield the relevant demand functions forthe Cournot case:
pH = qH³µ ¡ xH
´¡ qLxL ; pL = qL
³µ ¡ xH ¡ xL
´: (3)
At any t ¸ 1; …rm i obtains revenues RIi = pixi; I = B;C: The
discounted‡ow of pro…ts over the whole game is then:
¼Ii =
8>>><>>>:
±i1¡ ±i
¢RIi (qi; qj)¡ cq2i under IV±i
1¡ ±i¢RIi (q)¡
cq2
2under JV
(4)
To model collusion in marketing, we adopt the Perfect Folk
Theorem (PFThenceforth; see Friedman, 1971), where the in…nite
reversion to the one-shotNash equilibrium is used as a punishment
following any deviation from theprescribed collusive path.7 The
collusive path can instruct …rms to colludeeither fully (i.e., at
the Pareto frontier of monopoly pro…ts) or partially, at
7There exist other (less grim) penal codes (see Abreu, 1986;
1988; Abreu, Pearce andStacchetti, 1986; Fudenberg and Maskin,
1986), using symmetric optimal punishments.However, the asymmetry
of our model prevents us from adopting optimal punishments.For the
application of optimal punishments in a symmetric duopoly model
with productdi¤erentiation, see Lambertini and Sasaki (1999,
2000).
5
-
any pair of prices or quantities such that per-period individual
revenues areat least as large as the Nash equilibrium revenues.
De…ne:
[1] The instantaneous best reply of …rm i as ®¤i :
[2] The collusive action as ®coll 2³min
n®N ; ®M
o;max
n®N ; ®M
oi; ® = p; x:
[3] The collusive revenues to …rm i as RIcolli (¢) ; (¢) = f(q);
(qi; qj)g :
[4] The one-shot Nash revenues to …rm i as RINi (¢):
[5] The one-shot deviation revenues to …rm i as RIDi (¢):
The rules of the PFT establish what follows:
² At t = 0; …rms play ®coll:
² At t ¸ 1; …rms play ®coll i¤ ®i = ®coll at t¡ 1 for all i
;…rms play ®¤i otherwise.
De…nitions [3-5] and the rules of PFT yields that implicit
collusion at®coll is sustainable i¤
±i ¸ ±Ii (K) =RIDi (¢)¡RIcolli (¢)RIDi (¢)¡RINi (¢)
for all i : (5)
In the next section, we quickly deal with the case of full
collusion, where®coll = ®M :
3 Full collusionFirst, notice that when …rms operate along the
frontier of monopoly pro…ts,they are indi¤erent between settting
prices or output levels. Therefore, wecon…ne our attention to the
Bertrand case.
Suppose …rms choose independent ventures at t = 0: Then, over t
2[1;1), they should market di¤erent products. We are going to show
thatthis cannot be an equilibrium. At any t 2 [1;1), the cartel
aims at
maxpH ; pL
RM = RBH(qH ; qL) +RBL (qL; qH) : (6)
6
-
Monopoly prices are:
pMH =µqH2; pML =
µqL2; (7)
at which xMH =µ
2, while xML = 0: Therefore,
¼BH =±H
1¡ ±H¢ µ
2qH4
¡ cq2H ; ¼BL = ¡cq2L : (8)
On the basis of the above result, independent ventures imply
that, for allqL 2 (0; qH); the low-quality …rm would exit , getting
thus zero pro…ts.Alternatively, …rm L may produce qL = qH : This
immediately entails that±Bi = 1=2 for all i; as …rms o¤er
homogeneous goods.
It needs no proof to show that the same holds in the case of a
jointventure, as this would yield product homogeneity as a result
of technologicaldecisions taken at t = 0: We have thus proved the
following:
Lemma 1 Under full collusion, the low-quality product enjoys
zero demand.As a consequence, …rms will only supply homogeneous
goods, with JV Â IVdue to the cost-saving e¤ect.
Corollary 1 Under full collusion in prices, ±Bi = 1=2 for all i;
independentlyof …rms’ venture decisions.
As to the Cournot case, notice that, as long as …rms provide
di¤erentqualities, we have
xMH =µ
2; xML = 0 (9)
which again entails that the low-quality …rm survives only if qL
= qH ; ei-ther because …rms activate a JV , or because …rms develop
the same qualityindependently of each other. As a result, we can
state the following:
Lemma 2 Under full collusion in quantities, ±Ci = 9=17 for all
i; indepen-dently of …rms’ venture decisions.
In summary, independently of the market variable chosen for the
su-pergame over t 2 [1;1), the …rms’ venture decisions at t = 0
have nobearings on the stability of collusion, as setting either
monopoly prices orquantities induces …rms to play a supergame with
homogeneous goods.
7
-
4 Partial collusionHere, we investigate the bearings of
technological choices on cartel stability,under the assumption that
…rms may activate partial collusion, i.e., they maycollude at any
®coll 2
³min
n®N ; ®M
o;max
n®N ; ®M
oi; ® = p; x:
4.1 Cournot behaviour
Consider partial collusion at xcoll 2³xM ; xN
´, for a generic quality pair
fqH ; qLg : In the limit, as qL ! qH ; we obtain the description
of the JV case.We de…ne the partially collusive output of …rm i
as:
xcolli = axNi + (1¡ a)xMi ; a 2 (0; 1) ; (10)
where xMi = xM=2 = µ=4 and8
xCNH =µ (2qH ¡ qL)4qH ¡ qL
; xCNL =µqH
4qH ¡ qL: (11)
The associated Nash equilibrium revenues are:
RCNH =µ2qH (2qH ¡ qL)2(4qH ¡ qL)2
; RCNL =µ2q2HqL
(4qH ¡ qL)2: (12)
Substituting (11) into (10) and rearranging, we have:
xcollH =µ [4qH(1 + a)¡ qL(1 + 3a)]
4 (4qH ¡ qL); xcollL =
µ [4qH ¡ qL(1¡ a)]4 (4qH ¡ qL)
(13)
which allow to calculate RCcolli :
RCcollH =µ2[4q2H(3¡ a)¡ qHqL(7¡ 3a) + q2L(1¡ a)] [4qH(1¡ a)¡
qL(1 + 3a)]
16 (4qH ¡ qL)2
RCcollL =µ2qL [2qH (2¡ a)¡ qL(1 + a)] [4qH ¡ qL(1¡ a)]
8 (4qH ¡ qL)2(14)
8We omit the explicit derivation of the Nash equilibrium
quantities, as it is well knownfrom previous literature (see Motta,
1993).
8
-
The deviation from xcolli remains to be described. The best
reply of …rmj to xcolli is given by:
9
xDCH =µ
h(4qH ¡ qL)2 ¡ aq2L
i
8qH (4qH ¡ qL); xDCH =
4µqH(3¡ a)¡ 3µqL(1¡ a)8 (4qH ¡ qL)
(15)
yielding deviation revenues:
RDCH =µ2
h(4qH ¡ qL)2 ¡ aq2L
i2
64qH (4qH ¡ qL)2; RDCL =
µ2qL
h4qH(3¡ a)¡ 3µqL(1¡ a)
i2
64 (4qH ¡ qL)2:
(16)We are now able to write the expressions for the critical
threshold of thediscount factors:
±CH =(1¡ a) (2qH ¡ qL)2 (4qH ¡ qL)2q2L [32q
2H ¡ 16qHqL + q2L(1¡ a)]
; (17)
±CL =(1¡ a) (4qH ¡ qL)2
(4qH ¡ 3qL) [4qH(5¡ a) ¡ 3qL(1¡ a)]: (18)
Notice that the above critical thresholds are independent of µ;
and can beplotted over the space fa; qLg ; after setting qH = 1.10
This is done in …gures1 and 2.
9Both xDH and xDL are admissible for all a 2 (0; 1] and qL 2 (0;
qH ]: As usual, deviation
against a collusive output never drives the cheated …rm out of
business, and never makesthe deviator a monopolist.
10Note that this normalisation involves no loss of generality,
since the same plots wouldobtain by rewriting ±Ci in terms of the
quality ratio qL=qH 2 (0; 1]:
9
-
0
0.2
0.4
0.6
0.8
1
a
0
0.2
0.4
0.6
0.8
1
qL
0
0.2
0.4 dH
0
0.2
0.4
0.6
0.8
1
a
Figure 1. Plot of ±CH over fa; qLg , with a 2 [0; 1] and qL 2
[0; 1] .
0
0.2
0.4
0.6
0.8
1
a
0
0.2
0.4
0.6
0.8
1
qL
0
0.2
0.4
dL
0
0.2
0.4
0.6
0.8
1
a
Figure 2. Plot of ±CL over fa; qLg , with a 2 [0; 1] and qL 2
[0; 1] .Observe …gure 1. The range of ±CH is truncated at 9/17 to
put into evi-
dence the parameter region wherein independent ventures make it
easier for
10
-
the high-quality …rm to sustain quantity collusion, as compared
to a jointventure. The equation of the border at which ±CH = 9=17
is:
ba =2q4L ¡ 15q3L + 149q2L ¡ 408qL + 2722q4L ¡ 51q3L + 221q2L ¡
408qL + 272
: (19)
All combinations of fa; qLg de…ning a point along the downward
slopingsurface in …gure 1, de…ne levels of partial collusion and
low quality such thatindependent ventures favour collusion as
compared to a joint venture. Theopposite holds for any point such
that
a 2Ã0 ;
2q4L ¡ 15q3L + 149q2L ¡ 408qL + 2722q4L ¡ 51q3L + 221q2L ¡ 408qL
+ 272
!: (20)
Consider now …gure 2. For any combination of a and qL in the
admissiblerange, ±CL · 9=17; holding as an equality at fa = 0; qL =
qHg :11
The foregoing analysis allows us to state the following:
Proposition 1 For all a 2 (ba; 1] ; implicit collusion is more
easily sustainedunder independent ventures than under a joint
venture. For all a 2 [0; ba) ;the opposite holds.
This means that, given a generic quality ratio qL=qH ;
independent ven-tures are preferable to a joint venture in terms of
cartel stability, if …rms col-lude not too far above the
disagreement point given by the one-shot Cournotequilibrium. The
shape of ba shows that, as far as cartel stability is con-cerned,
IV tends to become more and more advantageous compared to JVas
product di¤erentiation decreases. In the limit, as qL=qH ! 1; IV
ensures±Bi < 1=2 for all a 2 (0; 1] :
Alternatively, the above result can be reformulated as follows.
As a in-creases (that is, as the level of collusion weakens towards
the Cournot-Nashoutput), the range of qL=qH wherein IV ensures
±
Bi < 1=2 increases. The
intuition is that, if collusion is only slightly above the Nash
equilibrium prof-its, than deviation is scarcely pro…table and this
drastically contributes tostabilise implicit collusion.
11Notice that, in both plots, ±Ci becomes negative if a is
su¢ciently large and qL=qH issu¢ciently low, due to the fact that
deviation pro…ts become lower than collusive pro…ts.In such a case,
it can be assumed ±Ci = 0; so that any ±i ¸ 0 ensures that the
low-quality…rm does not cheat. Clearly, this has no particular
bearings on our analysis.
11
-
4.2 Bertrand behaviour
Turn now to the case where …rms are price-setters and try to
collude atpcoll 2
³pN ; pM
´, for a generic quality pair fqH ; qLg : Again, in the limit,
as
qL ! qH ; we obtain the picture of the JV case.De…ne the
partially collusive price of …rm i as:
pcolli = apNi + (1¡ a)pMi ; a 2 (0; 1) ; (21)
where pMi = µqi=2 and12
pNH =2µqH (qH ¡ qL)4qH ¡ qL
; pNL =µqL (qH ¡ qL)4qH ¡ qL
: (22)
The associated Nash equilibrium revenues are:
RBNH =µ2qH (2qH ¡ qL)2(4qH ¡ qL)2
; RBNL =µ2q2HqL
(4qH ¡ qL)2: (23)
Substituting (22) into (21) and rearranging, we have:
pcollH =µqH [4qH ¡ qL(1 + 3a)]
2 (4qH ¡ qL); pcollL =
µqL [2qH (2¡ a)¡ qL(1 + a)]2 (4qH ¡ qL)
(24)
which allow to calculate RBcolli :
RBcollH =µ2qH [4qH ¡ qL(1¡ a)] [4qH ¡ qL(1 + 3a)]
4 (4qH ¡ qL)2
RBcollL =µ2aqHqL [2qH (2¡ a)¡ qL(1 + a)]
2 (4qH ¡ qL)2(25)
Now consider the deviation from pcolli . The best reply of …rm i
againstthe collusive price pcollj is:
pBDH =µ [8q2H ¡ 2qHqL (3 + a) + q2L(1¡ a)]
4 (4qH ¡ qL)pBDL =
µqL [4qH ¡ qL(1 + 3a)]4 (4qH ¡ qL)
(26)
12Again, the explicit derivation of the Nash equilibrium prices
is omitted for the sake ofbrevity (see Choi and Shin, 1992; Motta,
1993).
12
-
The corresponding output levels for the cheating …rm are:
xBDH =µ [8q2H ¡ 2qHqL (3 + a) + q2L(1¡ a)]
4 (4q2H ¡ 5qHqL + q2L)xBDL =
µqH [4qH ¡ qL(1 + 3a)]4 (4q2H ¡ 5qHqL + q2L)
(27)
Notice that deviation outputs (27) are admissible for all values
of fa; qH ; qLgsuch that xBDi · µ; which entails the following
restrictions, for all positiveµ :
xBDH · µ for allqLqH
2"0;7¡ a¡
pa2 ¡ 22a+ 253 + a
#; (28)
xBDL · µ for allqLqH
2"0;19¡ 3a¡
p9a2 ¡ 114a+ 1698
#: (29)
The admissible range for the quality ratio in (29) is larger
than in (28), i.e.,
19¡ 3a¡p9a2 ¡ 114a+ 1698
¸ 7¡ a¡pa2 ¡ 22a+ 253 + a
8 a 2 [0; 1] :(30)
The above inequality entails that, as intuition would suggest,
it is easier forthe high-quality than for the low-quality …rm to
become a monopolist.
If (29) and (29) are met, then deviation revenues are:
RBDH =µ2[8q2H ¡ 2qHqL (3 + a) + q2L(1¡ a)]2
16 (qH ¡ qL) (4qH ¡ qL)2
RBDL =µ2qHqL [4qH ¡ qL (1 + 3a)]216 (qH ¡ qL) (4qH ¡ qL)2
(31)
Otherwise, the deviator becomes a monopolist. For the moment, we
writethe critical threshold of the discount factors by using
(31):
±BH =(1¡ a)qL (4qH ¡ qL)2
(2qH ¡ qL) [16q2H ¡ 2qHqL(7 + a) + q2L(1¡ a)]; (32)
±BL =(1¡ a) (4qH ¡ qL)23qL [8qH ¡ qL(5 + 3a)]
: (33)
Again, the above thresholds are independent of µ; and can be
plotted overthe space fa; qLg ; after setting qH = 1. This is done
in …gures 3 and 4, where
13
-
the range of both plots is bounded above at 1/2, corresponding
to the criticallevel of discounting associated with a joint
venture.13
0
0.2
0.4
0.6
0.8
1
a
0
0.2
0.4
0.6
0.8
1
ql
0
0.1
0.2
0.3
0.4
0.5
deltaH
0
0.2
0.4
0.6
0.8
1
a
Figure 3. Plot of ±BH over fa; qLg , with a 2 [0; 1] and qL 2
[0; 1] .13As in the Cournot case, whenever ±Bi < 0 because
deviation is unpro…table, the
relevant threshold becomes ±Bi = 0:
14
-
0
0.2
0.4
0.6
0.8
1
a
0
0.2
0.4
0.6
0.8
1
ql
0
0.1
0.2
0.3
0.4
0.5
deltaL
0
0.2
0.4
0.6
0.8
1
a
Figure 4. Plot of ±BL over fa; qLg , with a 2 [0; 1] and qL 2
[0; 1] .
Consider …rst ±BL (…gure 4). We have:
±BL =1
2(34)
ifqLqH=4
h5¡ 2a§ 3
p2a2 ¡ 1
i
17 + 7a: (35)
The above solutions coincide at a = 1=p2 ' 0:707; where qL '
0:653:
Then, observe the behaviour of ±BH (…gure 3). The border along
which±BH = 1=2 is everywhere to the north-west of the border
(35).
Moreover, the curve xBDH = µ is also to the north-west of the
border(35).14
14Indeed, the equation
4£5 ¡ 2a + 3
p2a2 ¡ 1
¤
17 + 7a=
7 ¡ a ¡p
a2 ¡ 22a + 253 + a
has no real root for a 2 [0; 1] ; with the r.h.s. being always
larger than the l.h.s. over theunit interval.
15
-
The cases where deviation gives rise to a monopoly remain to be
inves-tigated. This would entail recalculating ±Bi anew, taking
into account theadditional information conveyed by the complements
to (28) and (29). Yet,to the aims of the present paper, the
following argument will su¢ce.
First, observe that, in general:
@±Ii@RIDi
=RIcolli ¡RINi(RIDi ¡RINi )2
> 0 : (36)
At the boundary where xBDH = µ; critical discount factors are
given by (32)and (33). When
qLqH
2Ã7¡ a¡
pa2 ¡ 22a+ 253 + a
;19¡ 3a¡
p9a2 ¡ 114a+ 1698
!; (37)
the critical discount factor for …rm L is still given by (33),
while that associ-ated to …rm H is:
b±B
H =RM ¡RBcollHRM ¡RBNH
> ±BH =RBDH ¡RBcollHRBDH ¡RBNH
>1
2: (38)
Finally, whenqLqH>19¡ 3a¡
p9a2 ¡ 114a+ 1698
; (39)
we haveb±B
L =RM ¡RBcollLRM ¡RBNL
> ±BL =RBDL ¡RBcollLRBDL ¡RBNL
>1
2; (40)
along with (38).The above discussion su¢ces to establish the
following result:
Proposition 2 Implicit collusion in prices is more easily
sustained underindependent ventures than under a joint venture, for
all
qLqH
20@4
h5¡ 2a¡ 3
p2a2 ¡ 1
i
17 + 7a;4
h5¡ 2a+ 3
p2a2 ¡ 1
i
17 + 7a
1A
Outside the above range, the opposite holds.
As the intensity of collusion decreases towards the
Bertrand-Nash equi-librium pro…ts, i.e., as a grows larger, the
range of product di¤erentiationwherein collusion is easier under IV
than under JV increases. The intuitiveexplanation behind this
conclusion is the same as in the Cournot case.
16
-
5 Concluding remarksWe have reassessed an issue previously
raised by Martin (1995), under a newperspective, where …rms’
initial R&D e¤orts are aimed at product ratherthan process
innovation. We have analysed the relationship between
theorganizational design of R&D for product innovation and the
stability ofimplicit collusion either in quantities or in prices,
keeping unaltered the rulesgoverning the market supergame, i.e.,
using the Perfect Folk Theorem.
The main conclusion emerging from this setting is that a JV may
or maynot facilitate collusion in the market supergame, depending
upon (i) thedegree of di¤erentiation produced by …rms activating
independent ventures;and (ii) the intensity of price or quantity
collusion.
Independently of the market variable being set by …rms, we have
foundthat, the lower is the level of collusion, the lower is the
pro…tability of de-viation for any given degree of product
di¤erentiation resulting from inde-pendent ventures. This
drastically contributes to stabilise implicit collusion,in that a
reduction of deviation pro…ts goes along with a reduction in
thecritical threshold of the discount factor.
Therefore, public policies towards R&D behaviour should be
designed soas not to become inconsistent with the pro-competitive
attitude characteris-ing the current legislation on marketing
practices.
17
-
References[1] Abreu, D. (1986), “Extremal Equilibria of
Oligopolistic Supergames”,
Journal of Economic Theory, 39, 191-225.
[2] Abreu, D. (1988), “On the Theory of In…nitely Repeated Games
withDiscounting”, Econometrica, 56, 383-96.
[3] Abreu, D.J., D. Pearce and E. Stacchetti (1986), “Optimal
Cartel Equi-libria with Imperfect Monitoring”, Journal of Economic
Theory, 39,251-69.
[4] Albæk, S. and L. Lambertini (1998), “Collusion in
Di¤erentiatedDuopolies Revisited”, Economics Letters, 59,
305-8.
[5] Aoki, R. and T. Prusa (1997), “Sequential versus
Simultaneous Choicewith Endogenous Quality”, International Journal
of Industrial Organi-zation, 15, 103-21.
[6] Beath, J., Y. Katsoulacos and D. Ulph (1987), “Sequential
ProductInnovation and Industry Evolution”, Economic Journal, 97,
32-43.
[7] Cabral, L.M.B. (1996), “R&D Alliances as Non-Cooperative
Su-pergames”, CEPR Discussion Paper No. 1439.
[8] Chang M. (1991), “The E¤ects of Product Di¤erentiation on
CollusivePricing”, International Journal of Industrial
Organization, 9, 453-69.
[9] Chang, M.H. (1992), “Intertemporal Product Choice and Its
E¤ects onCollusive Firm Behavior”, International Economic Review,
33, 773-93.
[10] Choi, C.J. and H.S. Shin (1992), “A Comment on a Model of
VerticalProduct Di¤erentiation”, Journal of Industrial Economics,
40, 229-31.
[11] Deneckere, R. (1983), “Duopoly Supergames with Product
Di¤erentia-tion”, Economics Letters, 11, 37-42.
[12] Dutta, P.K., S. Lach and A. Rustichini (1995), “Better Late
than Early:Vertical Di¤erentiation in the Adoption of a New
Technology”, Journalof Economics and Management Strategy, 4,
563-89.
18
-
[13] EC Commission (1990), Competition Law in the European
Communi-ties, Volume I, Rules Applicable to Undertakings,
Brussels-Luxembourg,EC Commission.
[14] Friedman, J.W. (1971), “A Noncooperative Equilibrium for
Su-pergames”, Review of Economic Studies, 38, 1-12.
[15] Friedman, J.W. and J.-F. Thisse (1993), “Partial Collusion
Fosters Min-imum Product Di¤erentiation”, RAND Journal of
Economics, 24, 631-45.
[16] Fudenberg, D. and E. Maskin (1986), “The Folk Theorem in
RepeatedGames with Discounting or with Incomplete Information”,
Economet-rica, 54, 533-54.
[17] Gabszewicz, J.J. and J.-F. Thisse (1979), “Price
Competition, Qualityand Income Disparities”, Journal of Economic
Theory, 20, 340-59.
[18] Gabszewicz, J.J. and J.-F. Thisse (1980), “Entry (and Exit)
in a Di¤er-entiated Industry”, Journal of Economic Theory, 22,
327-38.
[19] Goto, A. and R. Wakasugi (1988), “Technology Policy”, in
Komiya, R.,M. Okuno and K. Suzumura (eds.), Industrial Policy of
Japan, NewYork, Academic Press.
[20] Häckner, J. (1994), “Collusive Pricing in Markets for
Vertically Di¤eren-tiated Products”, International Journal of
Industrial Organization, 12,155-77.
[21] Häckner, J. (1995), “Endogenous Product Design in an
In…nitely Re-peated Game”, International Journal of Industrial
Organization, 13,277-99.
[22] Häckner, J. (1996), “Optimal Symmetric Punishments in a
BertrandDi¤erentiated Products Duopoly”, International Journal of
IndustrialOrganization, 14, 611-30.
[23] Lambertini, L. (1997), “Prisoners’ Dilemma in Duopoly
(Super)Games”,Journal of Economic Theory, 77, 181-91.
19
-
[24] Lambertini, L. (1999), “Endogenous Timing and the Choice of
Quality ina Vertically Di¤erentiated Duopoly”, Research in
Economics (RicercheEconomiche), 53, 101-09.
[25] Lambertini, L., S. Poddar and D. Sasaki (1998),
“Standardization andthe Stability of Collusion”, Economics Letters,
58, 303-10.
[26] Lambertini, L., S. Poddar and D. Sasaki (2000), “Research
Joint Ven-tures, Product Di¤erentiation, and Price Collusion”,
International Jour-nal of Industrial Organization, forthcoming.
[27] Lambertini, L. and D. Sasaki (1999), “Optimal Punishments
in Lin-ear Duopoly Supergames with Product Di¤erentiation”, Journal
of Eco-nomics, 69, 173-88.
[28] Lambertini, L. and D. Sasaki (2000), “Non-Negative Quantity
Con-straints and the Duration of Punishment”, Japanese Economic
Review,forthcoming.
[29] Lehmann-Grube, U. (1997), “Strategic Choice of Quality when
Qual-ity is Costly: The Persistence of the High-Quality Advantage”,
RANDJournal of Economics, 28, 372-84.
[30] Majerus, D.W. (1988), “Price vs Quantity Competition in
OligopolySupergames”, Economics Letters, 27, 293-7.
[31] Martin, S. (1995), “R&D Joint Ventures and Tacit
Product Market Col-lusion”, European Journal of Political Economy,
11, 733-41.
[32] Motta, M. (1992), “Cooperative R&D and Vertical Product
Di¤erenti-ation”, International Journal of Industrial Organization,
10, 643-61.
[33] Motta, M. (1993), “Endogenous Quality Choice: Price vs
Quantity Com-petition”, Journal of Industrial Economics, 41,
113-32.
[34] Rosenkranz, S. (1995), “Innovation and Cooperation under
VerticalProduct Di¤erentiation”, International Journal of
Industrial Organiza-tion, 13, 1-22.
[35] Rosenkranz, S. (1997), “Quality Improvements and the
Incentive toLeapfrog”, International Journal of Industrial
Organization, 15, 243-61.
20
-
[36] Ross, T.W. (1992), “Cartel Stability and Product
Di¤erentiation”, In-ternational Journal of Industrial Organization,
10, 1-13.
[37] Rothschild, R. (1992), “On the Sustainability of Collusion
in Di¤erenti-ated Duopolies”, Economics Letters, 40, 33-7.
[38] Shaked, A. and J. Sutton (1982), “Relaxing Price
Competition throughProduct Di¤erentiation”, Review of Economic
Studies, 49, 3-13.
[39] Shaked, A. and J. Sutton (1983), “Natural Oligopolies”,
Econometrica,51, 1469-83.
[40] van Dijk, T. (1996), “Patent Height and Competition in
Product Im-provements”, Journal of Industrial Economics, 44,
151-67.
21