Digitized
by
the
Internet
Archive
in
2011
with
funding
from
Boston
Library
Consortium
IVIember
Libraries
HB31
•M415
DEVVcY
m.C>s-C6>
Massachusetts
Institute
of
Technology
Department
of
Economics
Working
Paper
Series
COMPETITION
AND
EFFICIENCY
IN
CONGESTED
MARKETS
.Daron
Acemoglu
Asuman
E.
Ozdaglar
Working
Paper
05-06
February
1 7,2005
Room
E52-251
50
Memorial
Drive
Cambridge,
MA
021
42
This
paper can be
downloaded
withoutcharge from
the SocialScience
Research
Network
Paper
Collectionat'>3SACHUSETTS INSTITUTE1
OF TECHNOLOGY
MAR
1 2005Competition
and
Efficiency
in
Congested
Markets*
Daron
Acemoglu
Department
ofEconomics
Massachusetts
Institute
ofTechnology
Asuman
E.
Ozdaglar
Department
ofElectricalEngineering
and
Computer
Science
Massachusetts
Institute
of
Technology
February
17,2005
Abstract
We
study theefficiencyofoligopolyequilibria incongested markets.The
moti-vatingexamplesaretheallocation ofnetworkflows inacommunication networkoroftraJfi-Cina transportation network.
We
showthat increasingcompetitionamong
oligopolistscanreduce efficiency, measuredasthe differencebetweenusers'
will-ingness topayanddelaycosts.
We
characterizeatightbound
of 5/6onefficiency inpure strategyequilibria. Thisbound
istightevenwhen
thenumber
ofroutes and oligopolists is arbitrarily large.We
also study the efficiency properties of mixedstrategyequilibria.'We
thank Xin Huang,RameshJohari,EricMaskin,NicolasStierMoses,JeanTirole,JohnTsitsiklis,IvanWerning,
Muhamet
Yildizandparticipants attheINFORMS
conference, Denver 2004foruseful1
Introduction
We
analj'ze pricecompetitioninthepresenceofcongestioncosts. Considerthe following environment: oneunit oftraffic can useone of/ alternative routes.More
traffic ona particular route causes delays, exerting a negative (congestion) externahtyon
existingtraffic.^ Congestioncostsarecaptured
by
a route-specificnon-decreasingconvexlatencyfunction, li(•). Profit-maximizingoligopohstssetprices (tolls) fortravel
on
each route denotedby
pi.We
analyzesubgame
perfectNash
equilibria ofthis environment,where
foreachprice vector,p,all trafficchoosesthepaththathas
minimum
(tollplus delay)cost,
h+Pi,
and
oligopohstschooseprices tomaximize
profits.The
environmentwe
analyze is ofpractical importance for anumber
ofsettings.These
include transportationand communication
networks,where
additional use of a route (path) generates greatercongestionforallusers,and
marketsinwhichthere are "snob" effects,so thatgoodsconsumed by
fewer otherconsumers
aremore
valuable(see for example, [38]).The
key feature of these environments is the negative congestion externalitythat users exerton
others. This externalityhasbeen
well-recognized since thework by
Pigou[27] ineconomics,by
Samuelson
[33],Wardrop
[41],Beckmann
etal. [4] intransportation networks,and by
Orda
et. al. [23], Koriliset. al. [18], Kelly [17],Low
[20]incommunication
networks.More
recently,therehasbeen
agrowingliterature that focuseson
quantification ofefficiency loss(referred to asthe price ofanarchy)that resultsfrom
externaUtiesand
strategicbehaviorindifferentclasses ofproblems; selfishrouting (Koutsoupias
and
Papadimitriou [19],Roughgarden and
Tardos [31], Correa,Schulz,
and
Stier-Moses [8], Perakis [26],and Friedman
[12]), resource allocation bymarket
mechanisms
(Johariand
Tsitsiklis [16], Sanghaviand Hajek
[32]),and
networkdesign (Anshelevich et.al. [2]). Nevertheless, the game-theoretic interactions between
(multiple) serviceproviders
and
users, ortheeffectsofcompetitionamong
theproviderson
the efficiency losshas notbeen
considered. This is an important areafor analysis sinceinmost
applications, (competing) profit-maximizingentitieschargepricesfor use.Moreover,
we
willshow
that the nature of the analysis changes significantly in the presenceof pricecompetition.We
provideageneralframework
forthe analysis of pricecompetitionamong
providersin a congested (and potentiaUycapacitated) network, studyexistence ofpurestrategy
and
mixed
strategy equihbria,and
characterizeand
quantify theefficiencyproperties ofequilibria.
There
are foursetsofmajor
resultsfrom our analysis.First,
though
the equilibrium of traffic assignment without prices can be highlyinefficient (e.g., [27], [31], [8]), price-setting
by
a monopolist internalizesthe negativeexternality
and
achievesefficiency.Second,increasingcompetitioncanincreaseinefficiency. Infact,changingthe
market
structure from
monopoly
toduopoly
almost always increases inefficiency. Thisresultcontrastswith
most
existing resultsintheeconomicsliteraturewhere
greatercompetition tendstoimprovethe allocation of resources(e.g. see Tirole [36]).The
intuitionforthis resultisdrivenbythepresenceofcongestionand
isillustratedby
theexample
we
discuss below.^^
An
externalityariseswhenthe actionsofthe playerinagameaffectsthe payoff of otherplayers.Third,
we
provide a tightbound on
the extent of inefficiency in the presence ofpricing, which apphes irrespective of the
number
of routes, /.We
show
that socialsurplus (defined as the difference betweenusers'willingness to
pay and
the delay cost) in any purestrategyoligopolyequilibriumisalwaysgreater than 5/6ofthemaximum
social surplus. Simple examples reach this 5/6 bound. Interestingly, this
bound
isobtained even
when
thenumber
of routes, /,is arbitrarilylarge.Fourth, purestrategy equilibria
may
failto exist. Thisisnotsurprising inview ofthefactthat
what
we
havehereisaversion ofaBertrand-Edgeworthgame
where
pure strategy equilibriado notexist inthepresenceofconvexcostsofproductionorcapac-ity constraints (e.g.,
Edgeworth
[11], Shubik [35], Benassy [6], Vives [40]). However,in our oligopoly environment
we
show
thatwhen
latency functions are linear, apure strategyequihbrium always exists, essentiallybecausesome
amount
ofcongestion ex-ternahtiesremove
the payoff discontinuities inherentin theBertrand-Edgeworth game. Non-existencebecomes an
issuein our environmentwhen
latency functions are highly convex. Inthiscase,we
provethatmixed
strategy equilibriaalwaysexist.We
alsoshow
that
mixed
strategy equilibriacanlead to arbitrarilyinefficientworst-caserealizations; inparticular, socialsurpluscanbecome
arbitrarilysmallrelativetothemaximum
social surplus.The
followingexample
illustratessome
of theseresults.Example
1 Figure 1shows
asituation similar totheonefirstanalyzedby Pigou [27]to highlighttheinefficiency
due
tocongestionexternalities.One
unit oftrafficwilltravel fromoriginA
to destination B,using either route 1or route2.The
latency functions aregivenby
x^ 2
^1{x)
=
—,
h
[x)=
-X.Itisstraightforward to seethat theefficientallocation [i.e.,onethatminimizesthe total delay cost '^^li{xi)xi] is
xf
=
2/3and xf
=
1/3, while the (Wardrop) equilibrium allocationthatequatesdelayon
thetwo
pathsisx^^
w
.73>
xf and
x'^^«
.27<
xf
The
source of theinefficiencyisthateachunit oftrafficdoes notinternalizethe greater increaseindelayfromtravelon
route1,so thereis toomuch
useofthisrouterelativetotheefficientallocation.
Now
considera monopolist controllingboth routesand
setting prices for travel tomaximize
itsprofits.We
willshow
ingreater detailbelowthat inthis case, themonop-olist will set a price including a
markup,
x^l'^(when
li is differentiable), whichexactly internalizesthecongestionexternality. Inotherwords,thismarkup
isequivalent to the Pigovian taxthat a socialplannerwould
setin order toinducedecentralizedtraffictochoosetheefficient allocation. Consequently, inthissimple example,
monopoly
priceswin
bepf
^
=
{2/^f
+
kand
pf
^
=
(2/3^)+
^, forsome
constantk.The
resultingtraffic intheWardrop
equilibriumwill beidenticaltotheefficient allocation, i.e.,x^^
—
2/3and
x^^^=
1/3.Next
considera duopolysituation,where eachrouteiscontrolledbyadifferentprofit-maximizingprovider. Inthiscase,itcanbe
shown
that equilibrium priceswilltake theofcongestionexternalities,allmarketstructureswouldachieveefRciency,andachange frommonopoly
Ii(x)=x73
l2(x)=(2/3)x
Figure1:
A
two
linknetwork
withcongestion-dependantlatency functions.formpf-^
=
Xi{l[+
I'o) [seeEq. (20) in Section4],ormore
specifically,pf^
^
0.61and
p^^
?» 0.44.The
resulting equihbriumtrafficisxf^
w
.58<
xf and
x^^
«
.42>
xf
,
which
alsodiffersfrom
theefficientallocation.We
willshow
thatthisisgenerallythecase intheohgopoly equilibrium. Interestingly, while in theWardrop
equihbrium withoutprices, there
was
toomuch
trafficon
route 1,now
there is too little traffic because ofits greatermarkup.
It is also noteworthy that although the duopoly equilibriumis inefficient relative to the
monopoly
equihbrium, in themonopoly
equilibrium k ischosensuchthatallofthe
consumer
surplusiscapturedby
themonopolist, whileinthe oligopolyequilibriumusersmay
havepositiveconsumer
surplus.^The
intuitionfortheinefficiency ofduopolyrelativetomonopoly
can be obtainedasfollows.
There
isnow
anew
sourceof(differential)monopoly power
foreachduopolist,which
theyexploitby
distortingthe patternoftraffic:when
provider1,controllingroute 1,chargesahigherprice,itrealizesthatthis willpush
some
trafficfromroute1toroute2,raisingcongestion
on
route 2.But
thismakes
thetrafficusing route1become more
"locked-in," becausetheiroutside option, travel
on
the route2, hasbecome
worse."*As
aresult,theoptimalpricethateachduopolistchargeswillincludeanadditional
markup
overthePigovian
markup. These
areXiZj forroute 1and
X2l'i forroute2. Sincethesetwo
markups
are generally different, they will distortthe patternoftrafficaway from
theefficient allocation. Naturally, however, prices are typically lowerwith duopoly, so even
though
socialsurplus declines, userswillbe
betteroffthaninmonopoly
(i.e.,theywill
command
a positiveconsumer
surplus).Although
there isalargehterature on modelsofcongestionboth in transportationand communication
networks, very few studies have investigated the implications of havingthe"propertyrights" over routesassigned toprofit-maximizingproviders. In [3],Basar
and
Srikant analyzemonopoly
pricingunderspecificassumptions onthe utilityand
latency functions.He
and
Walrand
[15] study competitionand
cooperationamong
internet serviceprovidersunderspecific
demand
models. Issues ofefficientallocation of ^Consumersurplusisdefined asthedifferencebetweenusers'willingness topayandeffective costs,Pi
+
li{xi),andisthusdifferentfromsocialsurplus(whichisthe differencebetweenusers'willingnesstopay andlatencycost, li{xi),thusalsotakes intoaccount producersurplus/profits). See Mas-Colell,
Winston, and Green[21].
*Using economicsterminology,wecouldalsosay that thedemandforroute1becomesmore
"inelas-tic". Sincethisterm has adifferentmeaninginthecommunication networksliterature (see [34]),we
flowsor traffic acrossroutes
do
not arise inthesepapers.Our
previouswork
[1] studiesthe
monopoly problem and
contains theefficiencyofthemonopoly
result(withinelastic trafficand more
restrictiveassumptions onlatencies),butnone
of the otherresultshere.To
minimizeoverlap,we
discussthemonopoly problem
onlybrieflyin thispaper. In the rest ofthe paper,we
use the terminology of a (communication) network, though all of the analysis applies to resource allocation in transportation networks,electricity markets,
and
othereconomic applications. Section 2 describes the basic en-vironment. Section 3 briefly characterizes themonopoly
equihbriumand
establishesitsefficiency. Section 4 defines
and
characterizes the oligopoly equilibriawithcompet-ingprofit-maximizing providers. Section 5 contains the
main
resultsand
characterizes theefficiency properties ofthe ohgopoly equilibriumand
providebounds
onefficiency.Section 6containsconcluding
comments.
Regardingnotation,allvectors areviewedas
column
vectors,and
inequalitiesare to beinterpreted componentwise.We
denote byM^
thesetofnonnegative /-dimensional vectors. Letd
be aclosedsubset of[0,oo)and
let/
:Q
i—>M
beaconvexfunction.We
usedf{x) todenotethesetofsubgradientsof/
atx,and
f~{x)and
f'^{x) todenote theleftand
rightderivatives of/
at x.2
Model
We
consider a network with/parallel links. LetI =
{!,...,/} denote thesetoflinks.LetX,denote thetotalflow
on
linki,and x
=
[xi,...,xi]denotethe vectoroflinkflows.Each
link in thenetwork has aflow-dependentlatency functionk{xi),which measures the traveltime (ordelay) asa function of the total flowonlinki.We
denotethe price per unit flow(bandwidth)oflinkihyPi. Letp
=
{pi,- ,Pi]denotethe vector ofprices.We
are interestedintheproblem
ofroutingdunits of flow acrossthe/links.We
as-sume
thatthisistheaggregateflow ofmany
"small" usersand
thusadopttheWardrop's
principle(see [41]) incharacterizing the flow distributionin thenetwork;i.e.,the flows arerouted along paths with
minimum
effective cost, defined asthesum
ofthelatencyat the given flow
and
the price of thatpath (seethe definitionbelow). Wardrop'sprin-ciple is used extensively in modellingtraffic behavior in transportationnetworks ([4], [9], [25])
and
communication
networks ([31], [8]).We
alsoassume
that the usershavea reservation utility
R
and
decide not to send their fiow ifthe effectivecost exceeds the reservation utility. This implies that user preferences can be represented by the piecewisehnear aggregateutilityfunction u{-) depictedinFigure 2.^Definition 1 For a given price vector
p
>
0,^ a vectorx^^
G
Mi
is aWardrop
Thissimplifyingassumptionimpliesthatallusers arc "homogeneous"inthe sense thatthey have
thesamereservationutility,R.
We
discusspotentialissues inextendingthisworkto userswithelasticand heterogeneousrequirementsintheconcludingsection.
''Sincethe reservationutilityof usersisequal toR,wecanalso restrictattention topi<
R
forallu(x)
Figure2: Aggregateutilityfunction.
equilibrium
(WE)
if(1)
We
denotetheset ofWE
ata givenp by W{p).
We
adoptthe followingassumptionon
thelatency functions throughoutthe paper.Assumption
1 For eachi6X, the latency functionli : [0,cx))i—> [0,oo]isproperclosed convex,nondecreasing,and
satisfies /^(O)=
0.Hence,
we
allow the latency functions to be extended real-valued (see [7]). LetCi
=
{xE
[0,oo) I li{x)<
oo} denote theeffective
domain
ofli.By
Assumption
1,Ciis aclosed interval of theform [0, b] or[0,oo). Let
bd
=
sup^g,-;^x.Without
lossof generality,we
canadd
the constraintXj£
C,inEq. (1). Usingtheoptimalityconditionsfor
problem
(1),we
seethat avectorx^^
€
M^
isaWE
ifand
onlyifX^j^jxf^^<
dand
there existssome
A
>
suchthat A( X^^^^x^^
—
d)=
and
foralli,R-k{xf^)-p,
<X
ifxf'^-O,
=
A if<
xf"^<
be>A
ifxf^
=
6c..(2)
When
thelatency functions are real-valued[i.e.,Ci=
[0,oo)],we
obtainthe following characterization of aWE,
whichisoftenusedasthe definition of aWE
intheliterature.This
lemma
states that in theWE,
the effective costs, defined as li{xf^^)+
pi, areequalized
on
alllinkswithpositiveflows(proof omitted).Lemma
1 LetAssumption
1 hold,and assume
further that Ci—
[0,oo) foralliG
I.Then
anonnegativevector x* EW{p)
ifand
onlyifli{x*)+pi
<
R,V
iwithx*>
0, I2=1
withJ2i=i^*
=
c?ifniirij{/j(xj)+Pj}
<
R.Example
2 belowshows
that condition (3) in thislemma
may
not holdwhen
the latency functions are notreal-valued.Next
we
establishthe existence ofaWE.
Proposition
1(Existence
and
Continuity)
LetAssumption
1 hold. Foranyprice vectorp
>
0, theset ofWE,
W{p),
is nonempty. Moreover, thecorrespondenceW
:R^
^
M^
isupper semicontinuous.Proof.
Given anyp
>
0,considerthe followingoptimizationproblem
maximize^>o
^^
({R
-
Pi)Xi-
j k[z)dz\ (4)subject to 2_,^i
^
'^i=l
Xi
€
Ci,V
i.Inview of
Assumption
(1) (i.e., liis nondecreasingfor all i), itcanbeshown
that the objectivefunction ofproblem
(4) isconvexoverthe constraintset, whichisnonempty
(since G Cj)
and
convex. Moreover, the firstorderoptimahtyconditions ofproblem
(4),whichare also sufRcient conditionsforoptimality,are identical to the
WE
optimality conditions [cf. Eq. (2)] .Hence
aflowvectorx^'^'^G
^{v)
if^^ndonlyif itisan optimalsolution of
problem
(4). Sincethe objectivefunction ofproblem(4) iscontinuousand
the constraintsetiscompact, this
problem
hasan optimalsolution,showingthatW{p)
isnonempty.
The
fact thatW
isan
upper semicontinuous correspondence at everyp
follows
by
using theTheorem
oftheMaximum
(seeBerge [5], chapter6) forproblem
(4).
Q.E.D.
WE
flowssatisfyintuitivemonotonicityproperties giveninthe following proposition.The
prooffollowsfromthe optimality conditions[cf. Ecj. (2)]and
isomitted (see[1]).Proposition
2(Monotonicity)
LetAssumption
1 hold. For a givenp
>
0, letP-j
=
\PiWj-(a) For
some p
<
p,letx GVV'(p)and
x G W{p). Then, X^j^jx^>
J^j^jXi.(b) For
some
pj<
Pj,letx G W{pj,p-j)and
x GVF(pj,p_j).Then
Xj>
Xjand
Xj<
Xi,foralliy^ j.
(c) For
some
I C
I, supposethatPj<
pj foralljG
X
and
Pj=
Pj forallj ^J,and
Ejgi^j(p)-For a given price vectorp, the
WE
need not be unique in general.The
followingexample
illustratessome
properties oftheWE.
Example
2 Consider atwo
hnk
network. Letthe total flowbe
d=
1and
the reserva-tionutilityheR =
1.Assume
thatthe latency functions aregivenby
^
I oo otherwise.
At
the price vector (pi,P2)=
(1,1), the set ofWE,
W{p),
is given by the set ofallvectors (xi,X2) with
<
Xj<
2/3and
X^jx,<
1.At any
price vector {pi,P2) withPi>
P2=
I,W{p)
isgivenby
all(0,X2) with<
X2<
2/3.This
example
also illustrates thatLemma
1 need not holdwhen
latency functions arenot real-valued. Consider, forinstance, the pricevector (^1,^2)=
(1~
e,1~
le) forsome
scalara>
1. Inthiscase, theuniqueWE
is(xi,X2)=
(1/3,2/3),and
clearlyeffectivecosts
on
thetwo
routes arenotequalized despite thefactthattheyboth have positive flows. This arises because the path with the lower effective cost is capacity constrained, sono
more
trafficcan use thatpath.Under
further restrictionson
the k,we
obtain the followingstandard result in the literaturewhich assumesstrictlyincreasinglatency functions.Proposition
3(Uniqueness)
LetAssumption
1 hold.Assume
further that li isstrictly increasing over Cj. For
any
price vectorp
>
0, the set ofWE,
W{p),
is a singleton. Moreover,the functionW
:W_^, >-^M^
iscontinuous.Proof.
Under
thegivenassumptions, for anyp
>
0, the objective function ofproblem
(4) is strictly convex,
and
therefore has a unique optimal solution. This shows the uniquenessoftheWE
atagivenp. Since thecorrespondenceW
isupper
semicontinuous from Proposition 1and
single-valued,itis continuous.Q.E.D.
In general,underthe
Assumption
that/j(0)=
for alliE
X,we
havethe followingresult,
which
isessentialinouranalysiswithnonuniqueWE
flows.Lemma
2 LetAssumption
1hold. For agivenp
>
0,definethesetI =
{iGl
\ 3X, XE W{p)
withx^7^x^}. (5)Then
h{xi)
-
0,Vie2,
Vxe
W{p),
Proof. Consider
some
i&T. and
xG W{p).
SinceiE
T, thereexistssome
xE
W{p)
such thatXi
^
Xi-Assume
withoutlossofgenerahtythatXt>
Xi.There
aretwo
casesto consider:
(a) IfXk
>
Xk for all A;^
i, then Xljgi^i>
'^IjeJ^i-' '^^ich implies that the
WE
optimality conditions [cf. Eq. (2)] forx hold with A
=
0.By
Eq. (2)and
Xi>
x,,we
haveli{xi)
+Pi<
R,li{xi)
+Pi>
R,which togetherimplythat /i(xi)
=
li{xi).By
Assumption
1 (i.e., liisconvexand
/,(0)
=
0),itfollows that/,(x,)=
0.(b) IfXk
<
Xkforsome
k,bytheWE
optimality conditions,we
obtainli{Xi)
+Pi<
lk{Xk)+Pk,
k{x,)
+f)i>
IkiXk)+Pk-Combining
the above with Xj>
x,and
Xfc<
x^,we
seethat /j(xj)=
li{xi),and
hi^k)=
4(^fc)-By
Assumption
1, this shows that li{xi)=
(and also thatPz
=Pk)-Next
considersome
i, j€
I.We
willshow
that pi=
pj. Since iE
T, thereexist X, X GW{p)
suchthatXi>
Xi.There
arethree cases to consider:• Xj
<
Xj.Then
a similarargument
topart (b)aboveshows
thatpi=
pj.
• Xj
>
Xj. If Xfc>
Xk for allk^
i,j, thenX^m^m ^
^'™ply™g
that theWE
optimahty conditionsholdwith A=
0. Therefore,we
havek{x,)
+Pt<
R,lj{xj)
+
Pj>
R,whichtogetherwithli{xi)
=
lj{Xj)=
implythatpi—
pj.• Xj
=
Xj. Since j ET, there existssome
xE
W{p)
suchthat Xj^
Xj. Repeating theabovetwo
stepswithXj instead ofXj yieldsthe desiredresult.Q.E.D.
We
nextdefinethe socialproblemand
thesocialoptimum,
whichisthe routing (flow allocation) thatwould
be chosenbyaplannerthathasfullinformationand
fullcontrol overthenetwork.Definition 2
A
flowvectorx^
isa socialoptimum
ifit is an optimalsolution of the socialproblemniaximize2;>o
2.
(-^~
ki^i) jXi (6) 2=1/
subject to
>,
^i^
^•Inviewof
Assumption
1,the socialproblem
has a continuousobjective functionand
a
compact
constraintset,guaranteeingthe existence ofasocialoptimum,
x^. Moreover, usingtheoptimality conditionsforaconvexprogram
(see [7], Section4.7),we
seethat a vectorx^ G
M^
isa socialoptimum
ifand
only if X^j^^xf
<
dand
there exists a subgradient gi.E
dli{xf) foreachi,and
a A'^>
such that A'^(X],_ixf
—
d)=
and
foreachi,
R-k[xf)-xfgi^
<X'
ifxf^O,
(7)=
A^iiO<xf<bc„
>
A^ ifif
=
6c..For future reference, fora givenvectorx
6 R^,
we
define the value of the objective functioninthe socialproblem,/
S{x)
=
Y.{R-h{x,))x,,
(8)1=1
asthe socialsurplus,i.e., thedifferencebetweenusers'willingness to pay
and
thetotal latency.3
Monopoly
Equilibrium
and
Efficiency
In this section,
we
assume
that a monopolist service providerowns
the / linksand
charges a price ofp^ perunit
bandwidth on
linki.We
consideredarelatedproblem
inAcemoglu and
Ozdaglar[1]foratomicuserswithinelastic traffic(i.e.,theutilityfunctionofeach ofafinite set ofusersisa step function),
and
withincreasing, real-valuedand
differentiable latency functions. Here
we
show
that similar results hold for themore
general latency functions
and
thedemand
model
consideredinSection2.The
monopolistsetsthe prices tomaximize
hisprofitgivenby
/
?,=i
where
x G W{p).
Thisdefinesatwo-stagedynamic
pricing-congestiongame,where
the monopolistsets prices anticipating thedemand
of users,and
given the prices (i.e., inDefinition3
A
vector {p'^'^ ,x^'^^)>
is aMonopoly
Equilibrium(ME)
if x^W{p^'^)
and
n(p*^^,x^"^)
>
n(p,x),V
p,V
XG
ly{p) .Our
definition oftheME
is strongerthanthestandardsubgame
perfectNash
equi-libriumconceptfordynamic
games.With
a sHghtabuseofterminology,letus associate asubgame
perfectNash
equilibrium with the on-the-equilibrium-path actions of the two-stagegame.Definition 4
A
vector {p*,x*)>
is asubgame
perfect equihbrium(SPE)
of the pricing-congestiongame
ifx*G W{p*) and
forallp
>
0, thereexists x6
W
{p) suchthat
U{p*,x*)
>
U{p,x).The
followingpropositionshowsthatunderAssumption
1,thetwo
solutionconceptscoincide. Sincetheproofis notrelevantfortherestof the argument,
we
provideit inAppendix
A.Proposition
4
LetAssumption
1 hold.A
vector [p^^^ ,x^'^) isanME
ifand
onlyif it isanSPE
ofthe pricing-congestiongame.Sincean
ME
{p*,x*)isan optimalsolution oftheoptimizationproblem
I
maximizep>o,x>o
^^PiXi
(9)subject to X
£ W{p),
it is slightly easier to
work
withthan anSPE.
Therefore,we
useME
asthe solution conceptinthispaper.The
precedingproblem
hasan optimalsolution,whichestablishesthe existence ofanME.
Inthenextproposition,we
show
that the flow allocation atanME
and
the socialoptimum
are thesame.Proposition
5 LetAssumption
1hold.A
vectorxisthe flow vector atanME
ifand
onlyifitisasocial
optimum.
To
provethis result,we
firstestablishthe followinglemma.
Lemma
3 LetAssumption
1hold. Let (p,x)beanME. Then
foralliwithx^>
0,we
have
p,
=
R-
li{x^).Proof.
DefineI =
{ieI
\Xi>
0}.By
the definition ofaWE,
for alliE
I,we
havej9j
+
li{xt)<
R.We
firstshow
thatli{xi)
+Pi
=
lj{xj)+
Pj,V
i,JG
X.Suppose
li{xi)+Pi
<
lj{xj)+
pjforsome
i,jG
T. Usingtheoptimality conditionsfora
WE
[cf Eq. (2)],we
seethatx
€ W{pi
+
e,p^i) forsufficientlysmalle, contradictingthe factthat(p,x)isan
ME. Assume
next thatpi+
li{xi)<
R
forsome
iE
T. Consider the price vectorp
=
p
+
eX^j^j^i'where
e^is thei*''unitvectorand
eissuchthatPi
+
e+
li(xi)<
R,y
ieJ.
Hence, x isa
WE
atpricep,and
thereforethe vector (p,x) isfeasible forproblem
(9)and
has astrictlyhigherobjectivefunction value, contradicting the fact that(p,x)isan
ME. Q.E.D.
Proof
ofProposition
5: InviewofLemma
3,we
canrewriteproblem
(9) asmaximize3;>o
2_.
{R~
hi^i)i=l I
subject to 2_]^i
^ ^
This
problem
isidentical tothe socialproblem
[cfproblem
(6)], completingtheproofQ.E.D.
In addition to thesocial surplus defined above, it is also usefulto define the
con-sumer
surplus, as thedifi^erencebetween
users'willingness topay
and
effective cost,i.e.,Y2i=i(-^
~
hi^i)—
Pi)xi (See Mas-Colell,Winston and
Green, [21]).By
Proposition5and
Lemma
3, it isclear thateventhough
theME
achievesthe socialoptimum,
allof thesurplusiscaptured bythemonopolist,and
users are just indifferentbetween
sendingtheirinformationornot(i.e., receive
no consumer
surplus).Our
major
motivation for the study of ohgopolistic settings is that they provide a better approximation toreality,where
thereis typically competitionamong
service providers.A
secondary motivationisto seewhether
an
ohgopoly equilibriumwillachieve anefhcient allocationliketheME,
while also transferringsome
or allofthe surplus to theconsumers.4
Oligopoly
Equilibrium
We
suppose that there areS
service providers, denote the set ofservice providers byiS,
and assume
thateachserviceprovider s£
S
owns
adifferentsubsetTs ofthelinks.Serviceproviderscharges a pricep^per unit
bandwidth
onlinkiE
Is- Giventhe vectorofprices oflinks
owned by
other service providers,Ps =
\pi]i^i,, theprofit of service providersisUsiPs,P-s,x)==
^PiXi,
for X e W{ps,p-s),
where
Ps=
\pi\ieXs-The
objective ofeachservice provider, hke themonopoHst
inthe previous section,is to
maximize
profits. Becausetheir profitsdepend on
the pricesset byother service providers, each service provider formsconjectures about the actions of other service providers, asweh
asthe behaviorofusers, which,we
assume, they do accordingtothe notion of(subgame
perfect)Nash
equihbrium.We
refer to thegame
among
service providers asthe price competitiongame.Definition5
A
vector{p'-'^,x'~'^)>
is a(pure strategy) OligopolyEquilibrium(OE)
if
x^^
S
W
(p?^,p°f)and
forallsG5,n,(pf^,pef,xO^)
>
n,fe,pef,x),
Vp,
>
0,V
X6 H/fe,pef).
(10)We
refertop'-'^astheOE
price.As
forthemonopoly
case,thereis acloserelationbetween
apurestrategyOE
and
a pure strategy
subgame
perfect equilibrium.Again
associatingthesubgame
perfect equilibrium withthe on-the-equilibrium-path actions,we
have the following standarddefinition.
Definition 6
A
vector {p*,x*)>
isasubgame
perfectequihbrium(SPE)
ofthe price competitiongame
ifx* E:W
{p*)and
thereexistsafunction x{p)G
W
[p)suchthatforall
seS,
^s{P*s,P*-s^X*)>Ils{Ps.pls^x{ps,P*..,))
Vp, >0.
(11)The
followingproposition generalizesProposition4and
enables us towork
withtheOE
definition,whichismore
convenientforthesubsequentanalysis.The
proofparallelsthat ofProposition4
and
isomitted.Proposition
6 LetAssumption
1hold.A
vector {p'-'^,x^^)
is anOE
ifand
onlyifit isanSPE
ofthe pricecompetitiongame.The
pricecompetitiongame
isneitherconcave norsupermodular. Therefore,classicalargumentsthat areusedto
show
the existence of apurestrategy equilibriumdo not hold(see [13], [37]). Inthenextproposition,
we show
thatforlinearlatency functions, thereexists apurestrategy
OE.
Proposition
7 LetAssumption
1 hold,and assume
furtherthatthe latency functions arelinear.Then
the pricecompetitiongame
has apurestrategyOE.
Proof.
Foralli^
I,let/j(x)=
aiX. DefinethesetIo^{ieI
\a,=
0}.Let Iq denotethe cardinality of setIq-
There
aretwo
cases to consider:• Iq
>
2:Assume
that there exist i, j € Xq such thatiE
Tsand
j€
Is' forsome
s j^ s'
E
S.Then
itcanbe
seen that a vector (p^-^,x"^-^) withpf
^
=
for alli
E
Iqand
x'-'^E
W{p'-'^)isan
OE.
Assume
next thatforalliG
2o,we
haveiE
Isfor
some
sE
S. Then,we
canassume
withoutloss ofgenerality that providersowns
asingle linki'with a^/=
and
considerthecase/q=
1.• /o
<
1: LetBsiplf)
be theset ofpf^
suchthat(pf^,x°^)Garg
max
^p^Xi.
(12)Let B{p'-'^)
=
[Bs[p2.f)]s^s- Inview of the hnearity ofthe latency functions, itfollowsthatB{p'^^)isan
upper
semicontinuousand
convex- valued correspondence. Hence,we
canuseKakutani'sfixedpointtheorem
to assertthe existence ofap'-'^suchthatB{p'-'^)
=
p'^^ (seeBerge [5]).To
complete the proof,itremainsto beshown
that there existsx°^
E
W{p'^^) suchthat Eq. (10) holds.IfIq
—
0,we
have by Proposition3thatW{p'^^) isa singleton,and
thereforeEq. (10) holdsand
(p°-^,IV
(p°^)) isanOE.
Assume
finallythat exactlyoneofthea^'s (withoutloss ofgeneralityai) isequalto 0.
We
show
thatfor all x,x E
W{p'-'^),we
have Xi=
Xi, for alli^
1. LetEC{x,p°^)
=
mmj{lj{xj)+pf^}.
Ifat leastoneofEC{x,p°^)
<
R, orEC{x,p°^)
<
R
holds, thenone can
show
that X]t=i^i—
X^i=i ^i~
'^- Substituting Xi=
d—
Y2iei i^i^i ^^problem
(4),we
see that the objective function ofproblem
(4) isstrictlyconvexinx_i
=
[x^jj^i,thusshowingthatx
=
x. IfbothEC{x,p^^)
=
R
and
EC{x,p'^^)=
R, thenXi=
Xi=
1^^{R—
pf^)
foralli^
1,establishingour claim.For
some
xE
W{p'^^),consider the vector x^-^=
(d—
X^j^j Xj,x_i) . Sincex_iisuniquelydefined
and
Xi is chosen suchthat theprovider thatowns
link1 hasno incentive to deviate, itfollows that {p^^,x'~'^)isanOE.
Q.E.D.
The
existence resultcannotbe
generalized to piecewise linear latency functions or to latency functionswhich
are linearover their effective domain, as illustrated in the followingexample.Example
3 Consideratwo
linknetwork. Letthe total flowhe d=
1.Assume
that the latency functions are givenby, ( \ n 1 f \
/O
iiO<x<S
h{x)
=
0, l2{x)=
< ^-5^>^^
for
some
e>
and
6>
1/2,withtheconventionthatwhen
e=
0,hix)=
oofor x>
6.We
firstshow
that thereexistsno
purestrategy oligopolyequihbrium forsmalle (i.e.,thereexists
no
purestrategysubgame
perfectequihbrium).The
followinglistconsidersallcandidateoligopoly priceequihbria{pi,P2)
and
profitable unilateraldeviationsforesufficientlysmall,thus estabhshingthenonexistenceofan
OE:
1. Pi
=
P2=
0:A
small increase in the price of provider 1 will generatepositiveprofits,thus provider1hasanincentive to deviate.
2. pi
=
P2>
0: Letx
bethe flow aflocationat theOE.
IfXj=
1, thenprovider 2 hasan
incentive todecreaseitsprice. IfXi<
1, thenprovider 1has anincentive to decreaseitsprice.3.
<
pi<
p2- Player 1hasanincentive toincreaseitspricesinceitsflow allocation remains the same.4.
<
p2<
Pi: Fore sufficientlysmall, theprofitfunction of player 2,givenpi, isstrictlyincreasing asafunction ofp2!showingthatprovider 2hasanincentive to increaseitsprice.
We
nextshow
that amixed
strategyOE
alwaysexists.We
defineamixed
strategyOE
asamixed
strategysubgame
perfectequilibrium of the pricecompetitiongame
(seeDasgupta
and
Maskin, [10]). LetB^
be thespace ofall (Borel) probability measureson [0,i?]". Let /<, denote the cardinality ofJj, i.e., the
number
ofhnks controlled byserviceprovider5. Letfig
G
B^'be a probabihty measure,and
denotethe vector of theseprobabilitymeasures
by
/j,and
the vector of these probabilitymeasures excluding sby
Definition 7 (/i*,x*(p))isa
mixed
strategy OligopolyEquilibrium(OE)
ifthefunction x*{p)G
W
(p) foreveryp
G [0,R]^and
/ ^s{Ps,P-s,x*{ps,p-s)) d{i2l{p,) Xi_i*_^{p_s)) J[o,rY
/ Us{ps,P^s,X*{ps,P-s))d{iJ-siPs) XIJ--,{P-s))
J\O.R]'
>
'[0,R]'
foralls
and
/igG B'\
Therefore, a
mixed
strategyOE
simplyrequires thatthereisnoprofitabledeviation toadifferentprobabilitymeasure
for eacholigopolist.Example
3(continued)
We
now
show
thatthe following strategyprofileisamixed
strategy
OE
for the abovegame when
e -^ (amixed
strategyOE
also existswhen
e
>
0,butitsstructureismore comphcated and
lessinformative);(
0<p<R{l-5),
Mp)-<
1-^
R{l-6)<p<R,
\ 1 otherwise, (0<p<R{l-6),
^2{P)=
11-^
R{l-5)<p<R,
[ 1 otherwise.Notice that pi has
an
atom
equalto 1—
5 at i?.To
verifythat this profile isamixed
strategy
OE,
letji'be
thedensity of/i,with the conventionthatjj'=
oowhen
thereisan
atom
atthat point. LetAfj=
{p |/i^{p)
>
0}.To
establish that {pi,fJ-2)isamixed
strategyequihbrium,it sufficesto
show
thattheexpectedpayoff to playeri isconstantfor allPi
G
Mi when
the other player choosesp-i according to /i_i (see [24]).These
expectedpayoffsare
n
{p, Ip-i)=
/ U,{p^,P-i,X{p^,P-i))dfl-^(p^i),
Jo
whichareconstant forallpi
G
Mi
for i=
1.2. Thisestablishesthat (/^i./io) isamixed
strategy
OE.
The
next propositionshows
that amixed
strategyequilibriumalwaysexists.Proposition
8 LetAssumption
1hold.Then
the pricecompetitiongame
hasamixed
strategy
OE,
(^^^,x°^(p)).The
proofofthispropositionislongand
notdirectlyrelatedtherestofthe argument, soitisgiveninAppendix
B.We
next providean
exphcit characterization ofpure strategyOE.
Though
ofalso independent interest,these results aremost
useful forus toquantify'theefficiency lossofoHgopolyinthenextsection.
The
followinglemma
shows
that an equivalent toLemma
1 (which requiredreal-valued latency functions) also holds with
more
general latency functions at the pure strategyOE.
Lemma
4 LetAssumption
1 hold. If{p'^^,x^^) isapurestrategyOE,
thenh{x?'^)+P?'^
=
min{/,(xH+p°^},
V?withxP>0,
(13)j
k{x*)+pf^
<
R,Vzwithzp^>0,
(14)^xP
<
d, (15)with
J2Li
^?^
=
d^immj{lj{xf^)
+
Pj}<
R.Proof. Let
{f^,x°^)
be anOE.
Sincex°^
£
VF(p^^), conditions(14)and
(15)follow bythe definition ofaWE.
Considercondition(13). As,sumethat thereexistsome
z,jG
J
with
xf ^
>
0,x°^
>
suchthat;,(xf^)+pp^</,(x^°^)+pf^.
Usingthe optimality conditionsfora
WE
[cf.Eq.(2)], thisimpliesthat xf-^=
hc^-Con-sider changing
pf^
to pf-^+
e forsome
e>
0.By
checkingthe optimality conditions,we
see thatwe
can chooseesufficientlysmallsuch that x'^^G H^(pf^
+
e,p?f).Hence
serviceprovider that
owns
linkicandeviate topf
-^+
eand
increaseitsprofits, contra-dictingthefactthat(p*^"^, x'~'^) isanOE.
Finally,assume
to arrive atthe contradiction thatminj{/j(x^^)-\-Pj]<
R
and
'Ylii=\^?^
<
'^- Usingthe optimality conditionsforaWE
[Eq. (2)with A=
sinceJ2i^i^?^
<
"^l'^^^^imphes
thatwe must
havexf^
=
bc^ forsome
i.With
a similarargument
toabove, adeviation topf^
+
ekeeps x'-'^ as aWE,
and
ismore
profitable,completingthe proof.Q.E.D.
We
needthe following additionalassumptionforour pricecharacterization.Assumption
2 GivenapurestrategyOE
{p^^,x'-'^),ifforsome
iE
2
withxf^
>
0,we
have/^(xf^)=
0,then1^=
{i}.Note
thatthisassumptionisautomaticallysatisfiedifalllatency functions arestrictlyincreasing orifallserviceproviders
own
onlyonelink.Lemma
5 Let {p'-^^,x'-^^) be apurestrategyOE.
LetAssumptions
1and
2 hold. LetUs
denotetheprofitofserviceproviders at {p'^^,x^^).(a) Ifn^'
>
forsome
s'G
5,thenn^
>
forallse
S.(b) If
n,
>
forsome
sG
5,thenpf^xf^
>
forallj GT,.Proof.
(a) For
some
jG
Xy,defineK
=
pf^
+
lj{xf^),whichispositivesince Ug'>
0.Assume
lis=
forsome
s. ForkE
Is,considerthe pricepfc=
A'—
e>
forsome
small e>
0. Itcan beseen that atthe price vector {pk,P-k)^ ^^^correspondingWE
linkflow
would
satisfyx^>
0. Hence,serviceprovidershasanincentive todeviate to pk atwhich hewillmake
positiveprofit, contradicting thefactthat{p^^,x^^)
isapurestrategy
OE.
(b) Since Ug
>
0,we
havep'^fx'^f>
forsome
m
G Is-By
Assumption
2,we
canassume
withoutlossofgeneralitythatImi^m^)>
(otherwise,we
are done). Let jG
Isand assume
to arrive at a contradiction thatp^^xf^
=
0.The
profitof serviceprovidersatthepurestrategyOE
canbewritten asYis
=
ns+p^fx^f,
where
lisdenotestheprofitsfrom hnksotherthanm
and
j. Letp^^
=
K
—lm{x'^)
for
some
K
. Consider changingtlieprices p'^^and
p^^
suchthat thenew
profitisn,
=
n^
+
{K -
UxZ^
-
e)){xZ^-,)
+
e{K
-
/,(e)).Note
that eunits of flow aremoved
fromlinkm
to link j suchthat the flows of other linksremainthesame
at thenew
WE.
Hence, thechangeintheprofit isn,
-
n,
=
(/„(x^,^)-
UxZ^
-
e))x^^+
e(/„(x^^-
e)-
/,(e))).Sincelm{x^^)
>
0, ecan be chosensufficientlysmallsuchthattheaboveisstrictlypositive, contradictingthefactthat
{p^^,x^^)
isanOE.
Q.E.D.
The
followingexample shows
thatAssumption
2cannot be dispensed with for part(b) ofthis
lemma.
Example
4
Consider athree linknetwork
withtwo
providers,where
provider 1owns
links1
and
3and
provider 2owns
link2. Letthe total flowbe d=
1and
the reservation.
utiUtybei?
=
L
Assume
thatthe latency functions are givenby
h{xi)=
0,l2ix2)=X2,
h{x3)=ax3,
for
some
a>
0.Any
pricevector {pi,P2,P3)=
(2/3, 1/3,b)withb>
2/3and
(xi, X2,X3)=
(2/3,1/3,0) isa purestrategyOE,
so^3X3=
contrarytopart (b) ofthelemma.
To
see
why
thisisan
equihbrium, notethat provider 2is clearlyplaying abest response. Moreover,inthis allocationHi
=
4/9.We
canrepresent anydeviation ofprovider 1by
{pi,Ps)
=
{2/3-5,2/3-ae-5),
for
two
scalarseand
6,whichwillinduce aWE
of(xi, X2,X3)=
(2/3+
5—
e,1/3—
5,e) .The
correspondingprofitofprovider 1 atthisdeviationisHi
=
4/9—
5"^<
4/9, estab-lishingthat provider1 isalsoplayinga bestresponse
and
we
have apurestrategyOE.
We
next establish that under an additional assumption, a pure strategyOE
willneverbeatapoint of non-difTerentiability of the latency functions.
Assumption
3There
existssome
sE
S
suchthat k isreal-valuedand
continuously differentiableforalliEls
Lemma
6 Let (p°^,x°^) be anOE
with min^{p^^
+
lj{xf^)}<
R
and
pf^xf
^
>
for
some
i. LetAssumptions
1,2and
3 hold.Then
where
l^{xf^)and
l~{xf^) arethe rightand
leftderivatives ofthe function li atxf^
respectively.Since theproofofthis
lemma
islong,itisgiveninAppendix
C.Note
thatAssumption
3cannotbe dispensed withinthis
lemma.
Thisisillustrated inthenextexample. 17Example
5 Consider atwo
linknetwork. Letthe total flowbe d=
1and
the reserva-tionutihtybei? =: 2.Assume
thatthelatency functions aregivenbyif
<
X<
ih{x)-l2ix)-
,2(x-i)
otherwise.Itcanbeverifiedthat the vector {Pi^,P2^)
=
(1, 1),with(xf^,x^^)
=
(1/2, 1/2) isa purestrategyOE, and
is atapoint of non-differentiabiUtyforbothlatency functions.We
next providean
exphcit characterization of theOE
prices, whichis essential inourefficiencyanalysisin Section5.
The
proofisgiveninAppendix
D.Proposition
9 Let {p^^,x^^)
be anOE
such thatpf^xf^
>
forsome
i GI. LetAssumptions
1, 2,and
3 hold.a)
Assume
thatmin^{pf^
+
lj{x'^^)}<
R. Then,for allse5
and
i GX,,we
have(
xf%{x°^),
ifl'j{x°^)=
forsome
j^
Is,??""={
xf^^K^P^)+
v-^^'^'^^r , otherwise.(^6)
b)
Assume
thatmin^{p'j^+
Ijix'j^)}—
R- Then, forallsG<Sand
i £Xg,we
havepf^>^,n-(:^?^)-
(17)Moreover,ifthereexists
some
iGT
suchthatX
=
{i} forsome
sG
5,thenp?^
<
^rc(^r)
+
^
', (18)
Ifthe latency functions U are all real-valued
and
continuouslydifferentiable, then analysis ofKarush-Kuhn-Tucker
conditionsforohgopolj'problem
[Eq. (78)inAppendix
D] immediatelyyieldsthe followingresult:
Corollary
1 Let (p°^,x°^)
beanOE
suchthatp°^x°^
>
forsome
iG
I. Let As-sumptions1and
2hold.Assume
alsothatZjisreal-valuedand
continuouslydifferentiableforalli. Then, foralls
G
5
and
iEiTs,^^
haver
xf
^/^(xp^), if/;(xf^)=
forsome
j^
J,Pi
^
\ r..ir.mm
JP
_
l.f^OE\^OEv(^OE\
, £,<^i?
-
/,(xf ^) , x?^l'AxY^)+
^'^^
'\ }, otherwise.
(19)
Thiscorollaryalsoimpliesthatinthetwolinkcasewithreal-valued
and
continuously differentiable latency functionsand
withminimum
effective cost lessthan R, theOE
pricesarepP
=
xf^(/;(x?^)+
/^(xn)
(20)asclaimedinthe Introduction.
5
Efficiency
of
Oligopoly
Equilibria
In thissection,
we
study theefficiencyproperties ofohgopolypricing.We
take asourmeasure
ofefficiencythe ratio ofthesocialsurplus oftheequilibriumflow allocation to the social surplus ofthe socialoptimum,
§(x*)/§(x'^),where
x* referstothemonopoly
orthe oligopolyequilibrium [cf. Eq. (8)]. In Section3,
we
showed
that the flow alloca-tion at amonopoly
equilibrium isasocialoptimum.
Hence,in congestiongames
withmonopoly
pricing, there isnoefficiencyloss.The
followingexample
showsthatthis isnotnecessarilythe caseinohgopolypricing.
Example
6 Consideratwo
linknetwork. Letthe total flowhe
d=
1and
the reservation utihtybei?=
1.The
latency functions are givenby
3
li{x)^0,
kix)=
~x.The
uniquesocialoptimum
forthisexample
isx'^=
(1,0).The
uniqueME
{p^^,x^^)
is
x^^
=
(1,0)and
p^'^=
(1,1).As
expected, the flow allocations at the socialoptimum
and
theME
are thesame.Next
consider aduopoly
where
each ofthese linksis
owned
by adifferentprovider. UsingCorollary1and
Lemma
4, it followsthat the flow allocation attheOE,
x'^^, satisfies/i(xf^)
+
x?^[/;(x?^)+
Ux'i^)]=
kix^
+
x^^[l[{x?^)+
l',ix^%
Solvingthistogetherwithxf^
+
x^^
=
1showsthattheflow allocation attheunique oligopoly equilibrium is x'-'^=
(2/3,1/3).The
socialsurplus at the socialoptimum,
the
monopoly
equilibrium,and
the oligopoly equilibrium are givenby
1, 1,and
5/6, respectively.Beforeproviding a
more
thoroughanalysis oftheefficiencyproperties oftheOE,
the nextpropositionprovesthat,asclaimedintheIntroductionand
suggestedbyExample
6, a change in the
market
structurefrommonopoly
toduopoly
in a twohnk
network typicallyreducesefficiency.Proposition 10
Consider atwo
link networkwhere
each link isowned
by adiffer-ent provider. Let
Assumption
1 hold. Let {p'^^,x^^)be
a pure strategyOE
suchthat
pP^xf^
>
forsome
iG
J
and
mm,
{pf^
-f-lj{xf^)}<
R. If/'i(xf^)/xf^ j^l'2{x^^)/x^^, then S(xO-^)/S(x^)
<
1 .Proof.
Combining
theOE
priceswiththeWE
conditions,we
have /i(xf^)+
xf^{i[{xf^)
+
ux^))
=
Hxr)
+
xriux'",")
+
ux^^)),
where
we
use thefactthatmin,{pf^
+
lj{x?^)}<
R- Moreover,we
canuse optimality conditions (7) toprove that a vector (xf ,xf)>
is asocialoptimum
ifand
onlyif/l(xf)
+
xf/;(xf)=/2(xf)+xf/^(xf),
(seealsotheproofof
Lemma
4). Since l[{x'^^)/x'^^^
{^{x^^)jx^^
, theresult follows.Q.E.D.
We
next quantify theefficiency ofohgopoly equilibriaby providing a tightbound
onthe efficiency loss incongestiongames
witholigopoly pricing.As we
haveshown
inSection4,such
games do
notalwayshave apurestrategyOE.
In the following,we
firstprovide
bounds
on congestiongames
thathave purestrategyequilibria.We
nextstudyefficiencyproperties of
mixed
strategyequilibria.5.1
Pure
Strategy
Equilibria
We
consider pricecompetitiongames
thathave purestrategy equilibria(thissetincludes, but is larger than,games
with linear latency functions, see Section 4).We
consider latency functionsthatsatisfyAssumptions
1, 2,and
3. Let £/ denote thesetoflatency functions {Zi}ieisuchthat the associated congestiongame
hasapurestrategyOE
and
theindividualZ^'ssatisfy
Assumptions
1,2,and
3.^ Givena parallel linknetwork with/links
and
latency functions {/,}iei£
C.I-, let0£/({/i})denotethesetofflow allocations atan
OE.
We
definetheefficiencymetricatsome
x'^^G
OE{{li\) aswhere
x^
isasocialoptimum
given the latency functions{li}ieiand
R
isthereservationutihty. In other words, our efficiency metric is the ratio ofthe social surplus in an equilibrium relative tothe surplus in the social
optimum.
Following theliteratureon the "price ofanarchy",inparticular [19],we
are interestedintheworstperformanceinanoligopoly equilibrium, so
we
lookfora lowerbound on
inf inf ri{{l,},x'^^).
We
first provetwo
lemmas, which reducethe set oflatency functions that needtobe consideredin
bounding
theefficiencymetric.The
nextlemma
allows us to use the oligopoly pricecharacterization giveninProposition9Lemma
7 Let [p'^^,x'^^) beapurestrategyOE
such thatpf^x^^
=
for alliG
I.Then
x'^^isasocialoptimum.
Proof.
We
firstshow
that k{xf^)=
for alli G I.Assume
that lj{xf^)>
forsome
jG
I. Thisimplies thatxf^ >
and
thereforepf^
=
0. Since lj{xf^)>
0, itfollowsby
Lemma
2that for allx GW{p),
we
haveXj=
x^^. Consider increasing p*?^ tosome
smalle>
0.By
theupper semicontinuityofW{p),
it followsthatfor allX G W{e,p9.f),
we
have\xj—Xj^\
<
5forsome
5>
0. Moreover,byLemma
2,we
have,forall
X
GW{e,p'ilf),Xi>
.xf foralli7^ j. Hence,theprofitofthe provider thatowns
^Moreexplicitly,Assumption2implies thatifany
OE
{p°^ ,x^^)associatedwith{i,}j(=ihasxf^ >
andli{xf^)
=
0,thenJ,=
{i}.link jisstrictlyhigherat price vector [e^p'^f) thanatp*^^, contradictingthe factthat (p^^,
^0£)
is^^OE.
Clearly
x^^
>
forsome
jand
hencemmi^j{pf^
+
li{xf^)}=
pf^
+
lj{xf^)=
0,which
imphes by
Lemma
4thatJ2i&x^?^
=
d- Using/^(xP'^)=
0,and
G
dli{xf^) foralli,
we
haveR
-
li{xf^)-
xf^gi^=R,
y
lel,
for
some
gi^G
dli{xf^). Hence,x^^
satisfiesthe sufficientoptimality conditions forasocial
optimum
[cf.Eq. (7) withX^
=
R],and
the resultfollows.Q.E.D.
The
nextlemma
allows us toassume
without loss ofgenerality thatR
Yli=i^f
~
X]i=ih{xf)xf
>
and
X]j^ixf^
=
dinthesubsequentanatysis.Lemma
8 Let {k}iexG
Cj.Assume
thateither (i)Yli=i ^ii^i)^i
—
•^Xli=i^2^^^^some
socialoptimum
x^,or (ii)
ELi
^?'^<
dforsome
x^^
G 0^{{h}).
Then
everyx*^^G
OE{{li}) isasocialoptimum,
implyingthat rj{{li},x'^^)—
1.Proof.
Assume
that Ei=i^i(^f)^f—
^Yli=i^i-
Since x^ is a socialoptimum
and
everyx'-^^
G
OE{{li})isafeasiblesolution tothesocialproblem
[problem(6)],we
have=
^(/? -
k{xf))xf> J^iR -
k{xf''))xf^.V x^^
G ag({/0).
2=1 1=1
By
the definition ofaWE,
we
havexf-^>
and
/?-
kixf'^)>
pf ^
>
(wherepf-^isthe price of
hnk
iattheOE)
foralli. Thiscombined
withtheprecedingrelationshows
that
x^^
isasocialoptimum.
Assume
nextthatEi=i
^?^ <
dforsome
x'-^^G
0S({/,}). Let p'^^be theassociatedOE
price.Assume
thatp^^x^^
>
forsome
j&
I
(otherwisewe
aredone by
Lemma
7). Since
Ei=i
^?^
<
d,we
haveby
Lemma
4that m.mj^j{pj-\-lj{xf^}=
R. Moreover,by
Lemma
5, itfollowsthatp^xf^
>
foralliG
X. Hence,forallse
S, {{pf^)ieis,x'-'^) isan optimalsolution oftheproblem
maximize((p,)_gj^_j.)
Y.V^X^
subject to P^
+
l^{x,)=
R,Vtels,
pf^
+
/,(x,)=
i?.-ii^ls
Tx?''<d.
Substitutingfor {pi)iei, in the above,
we
obtainmaximize2:>o
2.
[R
—
U{Xz))xisubject to XiETi,
^
i ^Is,where
T^=
{xj|
pf^
+
k{xi)=
R}
iseither asingleton or a closed interval. Sincethisisaconvex problem,using the optimality conditions,
we
obtainR-k{xf^)-xf^gi^
=
0,VzgX„Vsg5,
where
gi^ £ dk{xf^).By
Eq. (7), itfollowsthatx^^
isasocialoptimum.
Q.E.D.
Hence, infinding a lower
bound on
theefficiency metric,we
canrestrict ourselves, without loss of generahty, to latency functions {/j} G Cj such that X]i=i'i(^f)^f<
RY!i=\ ^f
forsome
socialoptimum
x^
,and
Yli=i^?^
^
d.for allx^^ G
US{{li}).By
the following
lemma, we
canalsoassume
that X]i=i^f
=
d-Lemma
9 For asetoflatency functions{/j},gj,letAssumption
1hold. Let (p*^^,x^^)
beanOE
and x^
bea socialoptimum.
Then
iex iei
Proof.
Assume
to arrive at acontradiction thatYli^jxf^
>
X^jgjxf. This implies that x'^^>
Xj forsome
j.We
alsohave lj{x^^)>
lj{x^). (Otherwise,we
would
have lj{xj)=
l'j{Xj)=
0,whichyields a contradictionby
the optimality conditions (7)and
thefactthat
^^^jxf
<
d). Usingtheoptimality conditions (2)and
(7),we
obtainR
-
/,(xf ^)- p°^
>R-
/,(xf)-
xf
5;^,
for
some
gi^ G dlj{x^).Combining
the preceding with lj{x^^)>
lj{xj)and
pj'^>
x^^Z~(x^^)
(cf.Proposition9),we
see thatxf%ixf^)<x^g„
contradicting
x^^
>
xJand
completingthe proof.Q.E.D.
5.1.1
Two
Links
We
firstconsider aparallel linknetwork with twolinksowned
by twoserviceproviders.The
next theorem providesatight lowerbound on
r2{{h},x'^^) [cf. Eq. (21)]. In the following,we assume
without loss ofgenerality that d=
I. Also recall that latency functions in£2satisfyAssumptions
1, 2,and
3.Theorem
1 Consider a two link network where each link isowned
by a difi'erentprovider.
Then
r2{{k},x^^)
>
^,V
{/,},.i,2e
£2,x^^
e ag({/J),
(22)and
thebound
istight,i.e.,there exists {li}i=i^2 S^2 and
x*^^G
OE{{li})that attains the lowerbound
inEq. (22).Proof.
The
prooffollowsanumber
ofsteps:Step1:
We
are interestedinfinding a lowerbound
for theproblem
inf inf r2({/J,x«^). (23)
Given
{/j}G
C2, letx'-'^6
OE{{li})and
letx^
be a socialoptimum.
By
Lemmas
8and
9,we
canassume
that5Zi=i^?^
~
Xli=i^f
—
1- Thisimplies that there existssome
isuchthatxf^
<
xf. Since theproblem
is symmetric,we
canrestrict ourselves to{li}G £2
suchthatxf^
<
xf.We
claiminf inf
r2{{k},x^^)>r^f,
(24)where
^2,1-
mmimize,s,
(,5),>o—
—
T5-5—
j^-^
{E) yf,yOE^o subject to if<
yfilfy, i=
1, 2, (25) il+
yUil)'=
if+
ynify,
(26) ^f+
yf(^f)'<^,
(27)E^f
=
l' (28) z=l Zi<
;f , (29) /;=
0, if /f=
0. (30) li<y?^l[. ^=
l,2, (31)E?/?^ =
l' (32) i=l+
{Oligopoly EquilibriumConstraints};, f=
1,2.Problem
(E)can be viewedasafinitedimensionalproblem
thatcapturestheequihb-rium
and
the socialoptimum
characteristics oftheinfinitedimensionalproblem
giveninEq. (23). Thisimpliesthat instead ofoptimizingoverthe entire functionl^,
we
optimizeover the possible values ofli{-)
and
dk{-) atthe equilibriumand
the socialoptimum,
whichwe
denote byk,l'^,if,(if)'[i.e., (if)' isavariable that representsallpossiblevalues ofgi^G
dli(yf)].The
constraints oftheproblem
guaranteethat these valuessatisfy the necessaryoptimahtyconditions for asocialoptimum
and
anOE.
In particular, condi-tions (25)and
(31) capturetheconvexityassumption onli{-) byrelatingthe valuesli,l[and
/f,(/f)' [notethat theassumption1^(0)