M.I.T.
UBRARIES
Digitized
by
the
Internet
Archive
in
2011
with
funding
from
Boston
Library
Consortium
IVIember
Libraries
H631
DEweyj.
working
paper
department
of
economics
COLLUSION
AND
PRICE
RIGIDITY
Susan Athey
Kyle Bagwell
Chris Sanchirico
No.
98-23
November,
1998
massachusetts
institute
of
technology
50 memorial
drive
Cambridge,
mass.
02139
WORKING
PAPER
DEPARTMENT
OF
ECONOMICS
COLLUSION
AND
PRICE
RIGIDITY
Susan Athey
Kyle Bagwell
Chris Sanchirico
No.
98-23
November,
1998
MASSACHUSEHS
INSTITUTE
OF
TECHNOLOGY
50
MEMORIAL
DRIVE
CAMBRIDGE,
MASS. 02142
MASSftCh JSETTSINSTITUTE
OF^'ECHMQLOGY
DEC
4
1998
Collusion
and
Price Rigidity
Susan
Athey,
Kyle
Bagwell,
and
Chris Sanchirico*
Last Revised:
November
1998.
Abstract
We
consideran
infinitely-repeatedBertrandgame,
inwhich
prices are perfectly ob-servedand each
firm receives a privately-observed, i.i.d. costshock
ineach
pciiod.We
focuson
symmetric
perfectpublic equilibria(SPPE),
v^'hereinany "punishments"
areborne
equallyby
allfirms.
We
identifya
tradeoff thatis associated with collusive pricingschemes
inwhich
the price tobe
charged
by
each firm
is strictly increasing in its cost level:such
"fully-sorting"schemes
offerefficiency benefits, as they ensurethatthelowest-costfirm
makes
thecurrentsale, but theyalso
imply an
informational cost (distorted pricing and/or equilibrium-path price wars), since a higher-cost firmmust
be
deterredfrom mimicking
a lower-costfirm
by
charginga
lower
price.A
rigid-pricing
scheme,
where
afum's
collusive priceisindependent of
itscurrent costposition,sac-rifices efficiency benefitsbutalsodiminishestheinformationalcost.
For
a
wide
range
of
settings,the optimal
symmetric
collusivescheme
requires (i). theabsence of equiUbrium-path
price wars,and
(ii).a
rigidprice. Iffirmsare sufficiently impatient,however,
the rigid-pricingscheme
cannot
be
enforced,and
the collusive priceof
lower-cost firmsmay
be
distorteddownward,
in ordertodiminishtheincentiveto cheat.
JEL
ClassificationNumbers: C73,
L13, L14.
Keywords:
Collusion,repeatedgames,
privateinformation,pricerigidity, auctions.*M.I.T.and
NBER
(Athey);ColumbiaUniversityEconomics Department and Graduate School ofBusiness,andNBER
(Bagwell);andColumbiaUniversityEconomics Department andLaw School(Sanchirico). >\feare grateful toGleimEllison,EricMaskin, DavidPearce, LarsStole,JeanTirole, and seminarparticipants atBarcelona(PompeuFabra), BostonCollege,Columbia,
Har-vard,M.LT,N.YU., Northwestern and Tbulouse O.D.E.L)for helpful discussions. Atheythanks the National ScienceFoundation (SBR-9631760)forgenerousfinancialsupport,andflieCowlesFoundationatYale Universityfor hospitalityandsupport.
1.
Introduction
In thestandard
model
of
collusion,symmetric
firmsinteractinan
infinitely-repeatedBertrandgame
in
which
past prices are perfectly observed.While
the standardmodel
deUvers
anumber
of
usefulinsights, a limitation
of
thismodel
is thatitpresumes
an
unchanging
market
environment.This
limitation isimportant fortwo
reasons. First,thescope
for testinga
proposed
theoryof
collusionis greater, ifthe theory offers predictions
concerning
themanner
inwhich
collusive pricesmay
vary with underlying
market
conditions.Second,
a theoryof
collusion isof
potential value formacroeconomists,
ifthe theorydetermines
theresponse
of
collusive prices todemand
and
costshocks.
These
limitationshave been
partiallyaddressed
intwo
celebratedextensionsof
thestandard col-lusionmodel.
Rotemberg
and
Saloner(1986) extend
the standardmodel
to includethe possibilityof publicly-observed
demand
shocks
that are i.i.d.over
time.When
thedemand
shock
is large,theincentive tocheat (undercutthe collusive price)isacute,
and
collusionbecomes
more
difficult toenforce.Rather
than foregocollusive activity altogether, firms thenreduce
the collusive price,therebydiminishingthe incentive tocheat
and
bringingincentivesback
inline. Inthisway,
Rotem-berg
and
Saloner (1986)offera predictionof
countercyclicalmarkups
forcollusivemarkets
thatare subject todemand
shocks.'Like
thestandardmodel,
theirmodel
does
notpredict that actual "pricewars" occur
on
the equilibriumpath;rather, thesuccess of collusion varies along theequilibrium pathwith
thedemand
shocks
thatareencountered.Following
the seminalwork
of
Stigler (1964),a
second
literature stresses that a firmmay
be
unable
to perfectlymonitor
thebehavior
of
itsrivals.Green
and
Porter (1984) explorethispossi-bilityin
an
infinitely-repeatedCoumot
model.
They
assume
thata firm
cannot observe
theoutput choicesof
rivalsbutthatallfirmsobserve a
publicsignal (themarket
price) thatisinfluencedboth
by
output choicesand
an unobserved
demand
shock.^A
colludingfirm
thatwitnessesa
low
mar-ket pricethenfacesan
inferenceproblem,
as itisunclearwhether
thelow-priceoutcome
aroseasa
consequence of
abad
demand
shock
or a secretoutputexpansion
by
arival.The
Green-Porter
(1984)
model
thus representscollusion in thecontextof a
repeatedmoral-hazard
(hidden-action)model,
and a
centralfeatureof
theiranalysisisthatwars occur along
theequilibriumpath followingbad
demand
shocks.In this paper,
we
propose a
third extensionof
the standard collusionmodel.
We
consideran
^
As
HaltiwangerandHarrington (1991)andBagwell andStaiger (1997)explain,theassociationbetweencollusion
andcountercyclicalmarkups can beoverturned
when
theassumption ofi.i.d.demand
shocksisrelaxed.^ Whilethe Green-Porter (1984)model
isdevelopedinthe contextof
Coumot
competition, themaininsightscanbec^tured
m
arepeatedBertrandsetting,as Tirole(1988)shows.The
Green-Porter (1984)modelisfurtherextendedbyinfinitely-repeated Bertrand
game,
inwhich
each
firmis privatelyinformed of
its unit cost level ineach
period, thereisacontinuum
ofpossiblecosts,and
the cost realizationis i.i.d. across firmsand
time. Currentprice selections(butnotcostreahzations)arepubUcly observed each
period.We
thus represent collusion inthecontextof
a repeatedadverse-selection (hidden-information)model
with perfectly-observedactions (prices).^
Our
model
is well-designed to contribute to a long-standing issue in Industrial Organizationconcerning
the relationshipbetween
cdllusionand
price rigidity in the presenceof
cost shocks.Empuical
studiesby
Mills (1927),Means
(1935)and
Carlton (1986, 1989)have concluded
that prices aremore
rigid in concentrated industries. This empirical finding suggests that collusionis associated with a greater
tendency toward
price rigidity. Further support for this suggestion isoffered
by
Scherer (1980,pp. 184-93),who
summarizes
anumber
of
studiesthatfind thatcollusionisoften
implemented with
"rulesof
thumb,"
whereby
the collusive priceisfixedatsome
focallevel,such
asacertainpercentageabove
acommonly-observed
wholesaleprice."When
firmscolludeinthis
manner,
the collusiveprice isindependent
of
the privatecost positions(beyond
thewholesale
price)of
themember
firms.At
thesame
time, the Industrial Organization Uteraturehas as yet failed to provide asatisfac-tory theory that
Imks
pricerigiditywithcollusion.The
bestknown
theoryisthe"kinked
demand
curve" theory offered
by
Sweezy
(1939)and
Halland Hitch
(1939).As
Scherer (1980)and
Tu-ole(1988) discuss,
however,
this theory has important shortcomings.The
theoryemploys
an
ad
hoc
behavioralpostulate,
compresses
adynamic
story into astaticframework, and does
notdetermine
the collusiveprice.^Given
these Umitations,IndustrialOrganizationeconomists have
gravitatedto-ward
themore
informalview
that pricerigidityisappeaUng
tocollusive firms,because
arigid-price collusivescheme
preventsmistrustand
reducesthe riskof
apricewan
Carlton (1989)explains:*^ Ourmodeling frameworkisrelated torecentworkthatextendstheGreen-Porter (1984)modeltoallowfor
privately-observeddemandsignals. Compte(1998)and Kandori andMatsushima(1998),forexample,supposethatfirms
pub-liclychoose "messages" afterprivatelyobservingtheirrespective
demand
signals. Inoursetting, bycontrast,firmsare privatelyinformedas totheirrespective costsandthepublicactionisa(payoff-relevant)pricechoice.
^ Evidence fromothersourcesalso indicates that rigidpricingisaubiquitous featureofcollusiveendeavors. For example,inaBusinessVkek (1975)article,anumberofexecutivesinavarietyofindustriesdescribe informal(and
sometimesformal)agreementstomaintainaparticular price. HallandHitch (1939) providesimilaraccounts. Scherer (1980)describesacollusiveschemewhereby GeneralElectricand Westinghouse adopted apricingformulafor turbine generators.Collusionseemsespeciallyprevalentinmarketswithhomogeneousgoods(see, e.g.,
Hay
andKelley (1974)andScherer (1980,p. 203))andrelatively inelasticdemands(see, e.g.,Eckbo(1976)).
^ Inthekinked-demand-curvetheory,itisassumedthatafirm believesthat rivals will(not)match aprice (increase) decrease. This behavioralpostulateimplies akinkinthe
demand
curve,from whichitfollowsthatpriceandoutputareunresponsive tosmall maiginal costfluctuations. Maskinand lirole(1988)offer anequilibriummodelofthe
kinked-demandcurvetheory, basedonthehypothesisthatfirms
make
alternatingpricechoicesin aninfinitegame.Their theorydoesnot includethe possibilityofcostshocks,however.
^ Forasimilarview,seeScherer (1980,p. 180),
who
writes"Mostoligopolies...appearwillingtoforego themodestgains associatedwithmicro-meterlike adjustnientofprices to fleetingchangesin
demand
andcosts inordertoavoid5ie riskofmoreserious lossesfrompoorly coordinated pricingpoliciesandpricewarfare." SeealsoScherer (1980,
"Thepropertyofthekinkeddemandcurve theorythatpriceisunresponsivetosomecost f lucDiationsispreservedinmost
discussionsofoUgopolytheory whetherornotbased onthekinkeddemandcurve. Thereasoningis thatinoligopoUes
pricesfluctuate less inresponsetocostchanges(especiallysmall ones) than theywouldotherwiseinorder nottodisturb existing oligopolistic discipline.Anytimeapricechangeoccursinanoligopoly, thereisarisk thatapricewarcould break
out. Hence,firmsare reluctant tochangeprice." (Carlton,1989,pp. 914-15).
Inthepresentpaper,
we
develop a
rigorous evaluationof
thisinformal reasoning.Focusing
on
the private cost fluctuations that firms experience, v^^e explore the extent towhich
"mistrust" limitscolluding firms' ability to
respond
to their respective cost positions.The
costsof
mistrust areformalized in
terms
of
the pricewars
and
pricing distortions that are required to dissuade firmsfrom
misrepresentingtheir privateinformation.Our
formal
analysisbeginswith
the staticBertrand
game
with
private cost information.This
game,
which
hasbeen
previouslyanalyzed
by
Spulber
(1995), constitutes the stagegame
of
our
repeated-game model.
Intheunique
symmetric
Nash
equiUbrium,
the pricing strategy is strictly increasing in the firm's costlevel.An
appealingfeatureof
this pricingscheme
isthat sales in the currentperiodareallocated tothefirmwith
thelowestcost.This
isthe efficiency benefitof a
"fully-sorting" (i.e., strictly-increasing) pricingscheme.
The
pricing function is strictly increasing as aconsequence of
the winner-take-allnatureof
thegame,
togetherwith
the single-crossingproperty presentinthe firm'sexpected
profitfunction.The
single-crossingproperty arisesbecause
when
afirm
haslower
costs, itvaluesmore
theincrease inexpected
sales that alower
price implies.At
theNash
equilibrium,a
higher-costfirm
isthusunwiUing
to"mimic"
thelower
price thatitwould
charge
were
costslower, despite theincreaseinexpected
sales thatalower
pricewould
confer.We
turn nextto therepeated-game
model and
explorewhether
firmscan
supportbetter-than-Nash
profitswhen
they interact repeatedly.We
focuson
the classof symmetric
perfect publicequilibria
(SPPE).
An
SPPE
collusivescheme
atagiven
point intime
can
be
describedby
(i)a
price foreach
cost typeand
(ii)an
associated equilibrium continuation value foreach
vectorof
current prices,where
thecontinuation valueissymmetric
acrossfirms. Inan
SPPE,
therefore,col-luding firms
move
symmetricallythrough
any
cooperative orprice-war phases.By
contrast,inan
asymmetric
perfect publicequilibrium(APPE),
firmskeep
trackof
past salesand
may
coordinatecurrent price selections in orderto treat
more
favorablya firm
thathasexperienced
low
sales in therecentpast.Athey and Bagwell
(1998) study optimalAPPE.
After describingour
resultsabout
SPPE,
we
will return tocompare
thetwo
kindsof
equilibria.We
observe
thata
collusivescheme must
satisfytwo
kindsof
incentive constraints. First, forevery firm
and
costlevel, the short-termgainfrom
cheatingwith
an
off-schedule deviation (i.e.,with a
price thatisnot assignedtoany
costtypeand
thatthus representsa
cleardeviation)must be
unattractive, inview
of
the (off-the-equilibrium-path) pricewar
thatsuch a
deviationwould
imply.As
isusual inrepeated-game
treatments ofcollusion, this constraint issure tobe
met
iffirmsaresufficiently patient.
Second,
theproposed conduct
must
alsobe such
thatno
firm
iseverattractedtoan
on-scheduledeviation,whereby
afirmof a givencosttype misrepresentsitsprivateinformationand
selectsapriceintendedfor adifferentcost type.To
characterize theoptimalSPPE,
we
buildon
thedynamic-programming
techniques putforthby
Abreu,
Pearceand
Stacchetti(1986, 1990)and Fudenberg, Levine and
Maskin
(1994).We
draw
an
analogy
between
our
repeated hidden-informationgame
and
thestaticmechanism
designliterature,in
which
theon-scheduleincentive constraintisanalogous
to thestandardincentive-compatibility constraint, the off-schedule incentive constraint serves as a counterpart to the traditional partici-pation constraint,and
the continuation values of the repeatedgame
play the roleof
"transfers."However,
unlikea
standardmechanism
designproblem
inwhich
transfersare unrestricted, theset offeasiblecontinuation valuesis limitedand endogenously
determined.For example,
apricewar
is
analogous
toatransferthatisborne
symmetricallyby
allfirms.We
break
ouranalysis of optimalSPPE
intotwo
parts.We
suppose
first thatfirms arepatient, sothattheoff-scheduleincentive constraintismet. Inthisanalysis, theon-scheduleincentivecon-straint capturesthe informational costs
of
collusion that confront privately-informedfirms.The
central
problem
isthat thescheme must
be
constructedsothata
higher-costfirmdoes
nothave an
incentive tomisrepresentits costs as lower, thereby securing for itselfa lower
priceand
a higherexpected
market
share. Inan
SPPE,
theinformationalcostsof
collusionmay
be
manifested
intwo
ways.
First,the pricesof
lower-cost firmsmay
be
distorted tosub-monopoly
levels.This
isa
poten-tiallyeffective
means
of
eliciting truthfulcostinformation,sinceunder
thesingle-crossingproperty higher-costfirmsfindlower
priceslessappealing.Second,
followingthe selectionof
lower
prices, the collusivescheme
may
sometimes
call foran
equilibrium-pathpricewar
infuture periods.The
current-periodbenefits
of
alower
pricethenmay
be
of
sufficientmagnitude
tocompensate
forthe future costof a
pricewar,onlyifthe firmtruly haslower
costs inthecurrentperiod.A
richarrayof
collusiveschemes
fitwithintheSPPE
category.At one
extreme, firmsmay
incurtheinformationalcosts
of
collusionpurelyintermsof
distorted pricing.An
example
is theNash-pricing
scheme,
inwhich
firms repeatedly play theNash
equilibriumof
the staticgame.
At
the otherextreme, firmsmay
adopta monopoly-pricing scheme,
where
each firm
observesitscostfor theperiodand
thenselectsitsmonopoly
price. Ifdemand
slopesdown
(so thatthemonopoly
price varieswith
cost), thisscheme
satisfiestheon-scheduleincentive constraintonly
ifequilibrium-pathwars sometimes
follow the selectionof
lower-costmonopoly
prices.More
generally, collusive pricingschemes
may
involveboth
distorted pricingand
equilibrium-path wars.While
allSPPE
coUusive
pricingschemes
callfor prices thatare(weakly) increasingincosts,some
schemes
allowschemes
are fully sorting.One
scheme
of
particular interest is the rigid-pricingscheme.
Under
thisscheme,
each firm
selectsthe
same
price ineach
period,regardlessof
itscurrent cost position,and
sotheinformationalcosts ofcollusion are neutralized:
a
high-costfirm hasno
incentive to misrepresentits cost, sincethe
same
price is requiredunder
any
realization.Under
a
rigid-pricingscheme,
therefore,any
deviation is oflf-schedule, representing
an
unambiguous
violationof
the agreement.The
threatof
a
pricewar
may
thusbe
limited to off-equilibrium path events.The
downside
of
therigid-pricmg scheme, however,
is that itdoes
not offer efficiency benefits:one firm
may
have lower
coststhanitsrivals,and
yetthefirms sharethemarket.We
thus seethe outlineof
abroad
tradeoff:iffirmsseek acollusive
scheme
thatemphasizes
efficiency benefits (e.g.,afiilly-sortingscheme),
then the informational costs
of
collusionmay
be
large; while if firmsseek
a
collusivescheme
that
moderates
theinformationalcostsof
collusion (e.g., the rigid-pricingscheme),
then theymay
sacrificeefficiency benefits.Our
firstmain
finding is that firms fare poorlyunder any
SPPE
collusivescheme
that insistsupon
fullsorting. Infact,considering the entire setoffully-sortingSPPE,
we
findthat in thebestsuch
equilibrium firmsemploy
theNash
pricingscheme:
itisbetter tohave
low
(Nash)
pricesand
no
actualpricewars
thanhigher(e.g.,monopoly)
pricesand
actualpricewars.Thus,
ifcolluding firms are toimprove
upon
Nash
profits,the collusive pricingscheme must
exhibitsome
rigidity.We
next considerthefullclassof
SPPE
collusionschemes
and
reportasecond
main
finding: iffirmsare sufficiently patient
and
demand
is inelastic, thenan
optimalSPPE
collusivescheme
can
be
achieved without recoursetoequilibrium-pathpricewars.We
alsoprovide
sufficientconditionsunder
which
theoptimalSPPE
does
notrelyon
equilibrium-pathpricewars,for patientfirmswhen
demand
isdownward
sloping;however,
we
identifya
limited setof
circumstanceswhere
theymight
be
used.The
finding that optimal collusion generallydoes
not require pricewars
standsinsharp contrast tothe analysis presented
by Green and
Porter (1984), suggestinga fundamental
differencebetween
repeatedgames
with
hidden
actionsand
thosewith
hidden
informationand
publicly-observedactions.
Armed
withthese findings,we
next
add
some
additional structureand
provide
a
characterizationof
theoptimalSPPE
collusivescheme.
Specifically,when
firms are patientand
the distributionof
costtypesis logconcave,we
establisha
thirdmain
finding: ifdemand
isinelastic,optimalSPPE
collusionischaracterized
by
a
rigid-pricingscheme,
inwhich
firms selectthesame
price(namely,the reservation price
of
consumers)
ineach
period,whatever
theircostlevels.We
provide
as well sufficientconditions for rigid-pricingbehavior
inthe optimalSPPE when
demand
isdownward
and
collusion described above.We
then turn to thesecond
partof our
analysis,considering impatientfirms.We
begin
by
show-ingthat impatience
on
the part of firms createsan
additional disadvantage to usingprice wars: ascheme
with highprices today sustainedby
wars
in the futuremakes
a deviation especiallyprof-itabletoday, while simultaneouslyreducingthevalue
of
cooperationinthefuture.Our
second
(no-wars) findingthereforecontinuesto hold,even
when
firmsareimpatient. Next,we
observe
thatthe ofif-scheduleincentive constraintis particularlydemanding
forlower-costtypes. Intuitively,when
a firmdraws
a lower-costtype, the temptation to undercutthe assigned priceis severe, sincethe resulting market-share gainis thenespecially appealing.For
impatientfirms, acollusivescheme
thus
must
ensurethatlower-cost types receive sufficientmarket
shareand
select sufficientlylow
prices in equilibrium,sothatthegainsfrom
cheating arenot toogreat.This
logic is reminiscentof
theargument
made
by Rotemberg
and
Saloner (1986), although here it isprivatecostshocks
(asopposed
to publicdemand
shocks) that necessitate modificationof
thecollusivescheme.
We
confirm
thislogicwith
our
fourthmain
finding: iffirms arenotsuffi-cientlypatienttoenforcethe rigid-pricing
scheme,
theymay
stillsupportapartially-rigidcollusivescheme,
inwhich
the priceof
lower-cost typesisdepressedbelow
thatof
higher-costtypesinorderto mitigate the former's incentive to cheat. This finding suggeststhat
symmetric
collusionbetween
impatient firms
may
be
marked by
occasional(andperhaps
substantial)pricereductionsby
individ-ual firms.These
departuresoccur
when
a firm
receivesa
favorable costshock,and
they represent a permitted "escape clause" (i.e.,an
opportunity tocut pricesand
increasemarket
sharewithout
triggering retaliation)withinthe collusivescheme.
We
furthershow
thattheoptimalpricingproblem
forimpatientfirmscan
alsobe
studiedusingthe toolsofstatic
mechanism
design,althoughtheproblem
ismore
difficultthanthe standard casewhere
theoutside option,which
affectsthe participation constraint, isa
constant. Inthe repeatedgame
context, the valueof
the outside option (that is,openly
cheatingon
theagreement)
variesboth with
the firm's cost typeand
the collusiveagreement
itself,through
the price specified foreach
costtypeinthe present, aswellasthrough
the futurevalueof
cooperation.Finally,
we
considertheconsequences of
relaxingour
restrictiontosymmetric schemes.
Asym-metric
schemes
allowone firm
toenjoy a
more
profitable continuationvalue than another,and
inthis
way
such
schemes
facilitatetransfers^m
one firm
to another.While
we
do
notpropose a
full treatmentofAPPE
(seeAthey
and
Bagwell
(1998)),we
do
provide a simpleexample
thatcapturesthe qualitative benefits offered
by
asymmetric schemes.
The
scheme
we
study allowsonly
two
priceson
theequilibriumpath.The
low
pricemay
be
arbitrarilyclose to thehigh
one; but,over
some
efficiency benefits in the first periodof each
cycle, sincea
high-cost firm will hesitate toexpend
itsonly opportunitytochargethelow
pricewhen
itknows
thattomorrow
itsexpected
costswiU
be
lower. In contrast, low-cost firms willwish
to increase theirexpected market
share inthefirstperiod, since
on
average,winning
willbe
lessvaluablein thesecond
period.Although
APPE
can
providesuch
efficiency benefits,we
focus hereon
SPPE
for threemain
reasons. First,under
many
conditions the optimalSPPE
is stationary.While
we
do
notpropose
a
theoryof
how
firms learnand
coordinateon
aparticularequilibrium, the rigid-priceSPPE
(and
more
generally, stationarysymmetric
equilibria) are appealingly simple.By
contrast,a
necessary condition forAPPE
toimprove
upon
SPPE
is that theasymmetric
equilibrium is nonstationary.Such schemes
aremore
sophisticated,and
may
be
most
plausiblewhen
a smallnumber
of firmsinteract frequently
and
communicate
explicitly.Second,
most
of
theinformalliteratureabout
col-lusion (discussed above) highlights firms' fearof
industry-widebreakdowns
in collusion,and
itis thus interesting to analyze
SPPE, which
isolate this consideration. Finally,SPPE
are theonly
alternative iffirmscan observe
the prices offeredin themarket, but theycannot observe
(orinfer)theidentities
of
thefirmswho
offertheseprices.This
situation arisesinprocurement
auctionswith
more
thantwo
bidders,ifthewinning
bid-but notthename
of
thewinning
supplier-isannounced.
Our
findingsare related tothosedeveloped
by
McAfee
and
McMillan
(1992),intheiranalysisof
bidding
rings. Inastaticmodel,
theyshow
thata fixedprice(i.e., "identicalbidding")istheoptimalstrategy forbiddingcartelsin first-priceauctions for
a
single object,when
thecartels are"weak"
(i.e.,firmsareunableto
make
transfers).The
analysis thatwe
presentfortheinelastic-demand casemay
be
understood
intermsof
aprocurement
auction. Infact,inour
analysisof
theoptimalSPPE,
we
generalize the weak-cartelmodel,
since the staticmechanism
we
analyze is directly derivedfrom
a
repeatedgame
and
allows fora
restricted classof
transfers (corresponding tosymmetric
price wars).'Our
rigid-pricingfindingthusprovidesadditional theoreticalsupportforthecommon
practiceof
identicalbidding.*We
alsoextend
theprevious analysis inotherdirections, includingdownward-sloping
demand
and
impatientfirms.The
paper
proceedsas follows. InSections2
and
3,we
describe thestaticand
repeatedgames,
respectively.We
presentour
findings forSPPE
among
patientfirmsinSection4. Impatient firmsare considered in Section 5.
We
exploresimple
asymmetric
schemes
in Section 6.Concluding
thoughts arepresentedinSection7."^
McAfee
andMcMillan(1992)alsoconsider "strong"cartels,inwhichtransfersare unrestricted. (SeealsoCramton
andPalfrey (1990)and Kihlstrom and Vives(1992).)Incomparison, ouranalysisof
SPPE
restrictstransfers toa range ofsymmetricvalues (namely,SPPE
continuationvalues). Likewise,inAPPE,
firmshaveavailableanexpanded(butstillrestricted) classoftransfers(namely,
APPE
continuationvalues).* See,forexample,
Comanor
andSchankerman(1976),who show
thatfixed-pricerules(orcloselyrelatedsimplerotationschemes) have been usedbyasubstantialfractionofthedocumentedcasesofbidcUngringsinprocurement
2.
The
Static
Game
We
begin
by
consideringa
staticgame
of
Bertrand competitionwhen
firmspossessprivate informa-tion.This
game
illustratestheimmediate
tradeoffsthatconfront firms indetermining theirpricingpolicies. In
subsequent
sections,we
examine
adynamic game,
sothatwe may
consideraswellthelong-term
impUcations of
differentpricingpoUcies.The
staticgame
serves as the stagegame
for thedynamic
analysis.2.1
The Model
We
suppose
thattherearen
ex
ante identicalfirmsthatengage
inBertrand
competitionfor salesinahomogenous-good
market.Following Spulber
(1995),we
modify
thestandard Bertrandmodel
with
theassumption
thateach
firmisprivatelyinformed
as toitsunitcostlevel. Specifically,we
assume
thatfirmi's "type" Qiis
drawn
inan
i.i.d. fashionfrom
thesupport[^,Q\accordingtothecommonly
known
distribution function FiQ),where
thecorresponding
density j{&)=
F'{Q) iseverywhere
positive. Afterthe firms learntheirrespective cost types, they simultaneously
choose
prices.Let
Pi
G
J?+ denotethe pricechosen
by
firmi.The
associatedprice profileisthen representedby
the vector,p
=
(pi, ...,p„).On
thedemand
sideof
themarket,we
assume
a
unitmass
of
identicalconsumers,
each
ofwhom
possesses ademand
functionD
(p),where p
is the price atwhich
theconsumer
buys.We
allow that thedemand
function is either (i) inelastic, or (ii) positive, twice continuously differentiableand
decreasing (i.e.,"downward
sloping"). In theformer
case,we
suppose
thatD{p)
—
\up
tosome
reservation pricer,where
r>
6?
A
price strategy forfirm
iis a functionpi{6i)mapping from
thesetof
cost types, [9,6], totheset
of
possibleprices,R^.
The
functionpiisassumed
continuouslydifferentiable, exceptperhaps
atafinite
number
ofpoints (so as to allow for scheduleswith
jumps).'"A
price strategy profileis thus a vector
p{0)
=
(pi (9i),p_i
{0-i)),where
6
=
{$i, 0-i) isthe vectorof
cost typesand
p_i
(O-i) isthe profileof
rivalprice strategies.Each
firm
chooses
itsprice strategywith
the goalof
maximizing
itsexpected
profit,given
its cost type.To
representa
firm'sexpected
profit,we
require
two
further definitions. First,we
define it{p,6)=
[p—
6]D
(p) as the profit thata
firmreceives
when
itsetsthe pricep
and
hascosttype6
and "wins"
the entire unitmass
of consumers.
We
assume
that tt(p,6) has aunique maximizer,
p"^(9),
and
is strictly quasi-concave."Second,
we
specifya
Bertrand market-share-allocationfunction,rui (p), that indicatesfirmi'smarket
share® Inthecaseofinelasticdemand,
we
may
therefore interpret thestaticmodelasamodelofa procurementauction.
>\feassumethatr
>
flfor simplicity.^° Thisrestriction
includesawiderangeofpossible pricingstrategies,anditservestominimizetechnical
complica-tions thatarenotcentral toourdiscussion.
^^ Underinelasticdemand,p'"(6)
=
r.When
demandslopesdown,p"^{0)isincreasingin9.
when
thevectorofrealizedpricesisp.This
functionallocatesconsumers evenly
among
firmsthattie for the lowest price in the market; therefore, rrii (p)
=
1 ifp,<
mirij^ipj, rrii (p)=
ifPi
>
min.^^Pj,and
mi(p)
=
1/k
ifthere are /c—
1 other firmsthattiefirmi forthe lowestprice.We
may now
define firm z'sex
post
profit as ttj (p,6i)=
it(p^, di)mi
(p).This
is the profit thatfirmi enjoys after allprices are posted,when
itwins
a sharem^
(p)of consumers.
It is alsoconvenient
torepresentfirm
i'sinterim profit,which
istheexpected
profitforfirmiwhen
ithascosttype6i, selects the pricep^
and
anticipates thatrivalprices willbe determined
by
therival pricing strategy profile,p_t
{0-i).Firm
i's interimprofit function is thus defined as ni(pj,p_i,^j)=
E[Ki{pi,p_t
(^-t) ,Oi)],where
theexpectationistakenover
0_j. Finally,we
may
alsodefinefirm
i's
ex
ante profit.This
isthe profit thatfirm
i expects beforeany
firm (includingiitself)learns itscost type,
when
itisanticipated that firms' eventual prices aredetermined
by
thepricingstrategyprofile,
p
(0).Firm
i'sex
anteprofit isgiven
by
ni(p)
=
E[U.i{p(6) ,6i)],where
theremaining
expectationistaken
over firm
i'sown
costlevel.2.2
Equilibrium
Pricing
and
Profit
We
returnnow
to the interim profit function definedabove
and
describe the essential tradeoff that confronts a firmwhen
it sets its price. Interim profitmay
be
written as Ili{Pi,P-i,9i)=
K{Pi,
9i)Emi(pi,p-i{6
_i)), indicating thatfirm
i's price p^ affects its profitboth
by
altering its"profit-if-win" (i.e.,7r{pi,9i))
and
itsprobabilityof
winning
(i.e.,Emi{pi, p_j(0_J)). Thus,
when
afirm
of a
given type considerswhether
tolower
itspricefrom
some
statusquo
level, itmust weigh
the effect
of
an
increaseinthechance of winning
against the direct effectof
the pricereductionon
profit-if-win.An
important featureof
themodel
is that different types feel differentlyabout
this tradeoff.In particular, theinterimprofit fiinction satisfies
a
single-crossingproperty:lower
types find theexpected-market-share increasethat
accompanies a
pricereductionmore
appecding thando
highertypes. Inthefirstplace,since
lower
typeshave lower
totalcosts,theyhave
higherprofit-if-win(i.e., 7rfl(p,6)<
0). Therefore, theincreasedchance of winning
(i.e, theincrease inEmi{p, p_t(0_J)
from
lowering
priceisof
greaterbenefittoa
lower-costfirm. Secondly, the reductioninprofit-if-win from
lowering
priceislesssevereforlower
types(i.e.,tt^(p,6)>0):
revenue
changes
arethesame
acrossalltypes,but sincelower
types alsohave lower
marginalcost,their costsriseby
lessinresponse to increased
demand.
This
second
effectisstrictwhen
demand
slopesdown,
as thenthe
lower
pricedoes
raisedemand, and
itisneutral inthe caseof
inelasticdemand.
As
isstandard, the single crossing property impliesthat higher-cost firmsmust
selecthigher prices(i.e,Pi{6i) isEmploying
techniquesfrom
the auction literature(Maskin and
Riley, 1984),Spulber
(1995) analyses this stagegame
in detailand
finds in part:Proposition
1 (Spulber (1995))The
staticgame
has
a
unique
Nash
equilibrium,which
1) issymmetric: Pi
=
p^, Vi; 2)iscontinuouslydifferentiableand
strictlyincreasingover
6€
(^,6); 3)is
below
themonopoly
price: p^{6)<
p^{9),
"iO<
6 and;
4) yields positive interimprofitfor
alltypesbutthe highest6,
who
never
wins
and
whose
price p^{6)=6
would
yieldzeroprofiteven
ifitdid.
Notice that the
symmetric
equilibrium pricing strategy p^ is continuousand
strictly increasing.These
properties arise asaconsequence of
the "winner-take-all" natureof
thestaticgame.
If, forexample,
theschedulewere
tojump
up
discontinuouslyat9,thenafirm
with
type6
would do
better increasingits price intothe"gap"
(providedp^ifi)<
p^{d)); whileifthe priceschedulewere
tohave
a flat section that included type 6, thena
firm ofthis typewould
gainby
shading
its pricedownward
aslightamount
(converting "ties" into "wins").These
restrictions are standard intheliterature,but, as
we
shall see,theydisappearwhen
we move
totherepeatedgame.
3.
The
Repeated
Game
Inthis section,
we
definetherepeatedgame
thatisthe focusof our
study.We
alsopresenta "Fac-toredProgram"
and
establisha
relationshipbetween
solutions tothisprogram and
optimalSPPE.
We
usethisrelationship insubsequent
sections,where
we
characterize optimalSPPE.
3.1
The Model
Imagine
thatfirmsmeet
periodafterperiodtoplaythestagegame
describedintheprevioussection,each with
theobjectiveof
maximizing
itsexpected
discountedstream
of
profit.Assume
furtherthat,upon
entering a periodof
play,a firm
observesonly
the historyof: (i)itsown
costdraws, (ii) itsown
pricingschedules,and
(iii)the realized pricesof
itsrivals.Thus,
we
assume
thata firm does
notobserverivaltypesorrivalprice schedules.
Formally,
we
describe therepeatedgame
inthe followingterms.A
fullpathof
playisan
infi-nitesequence
{0*,p*},with
a given
pair in thesequence
representinga
vectorof
typesand
priceschedules atdatet.
The
infinitesequence
impliesa
publichistoryof
realized pricevectors {p*},and pathwise
payoffsfor firmimay
be
thusdefinedaswhere
6\ is the cost level forfirm
i atdatet,6
£
[0,1) isthediscountfactor,and
7ri(p*, 6\) isex
which
may
be
written as hi—
{Ol,pl,p*Li}J=i-(The
null history is the firm's information set atthe
beginning of
thefirstperiod.)A
(pure) strategy forfirm
iassociates apriceschedule Si{hi){9i)with
each
informationsethi.Each
strategy profile s=
(si, ...,s^)induces aprobability distributionover
play paths {0',p*}
in the usualmanner.
The
expected
discountedpayoff
from
s is thus theexpectationVi{s)
=
E[vi{{6^,p*})]takenwith
respecttothismeasure on
playpaths.We
would hke
our
dynamic
game
tohave a
recursive structure, sothatdynamic-programming
techniques (as
developed
by
Abreu, Pearce and
Stacchetti (1986, 1990)[APS])
may
be
applied.We
do
this attwo
levels. First,we
assume
thatfirm
types arei.i.d. acrosstime
(andfirms), sothat thegame
is "structurally recursive."Second,
we
followFudenberg,
Levine
and
Maskin
(1994)[FLM]
and
restrictattentiontoarecursive solutionconcept;namely,we
focusupon
thosesequential equilibria inwhich
players conditiononly
on
the historyof
realized pricesand
noton
theirown
private historyof
typesand
priceschedules.Such
strategies arecalledpublicstrategiesand
such
sequential equilibria are calledperfect public equilibria (PPE).Note
thatPPE
are tested against deviations thatdo
conditionon
private histories; that is, they are true sequential equilibria in thegame. Note
also thata
public strategymay
be
abbreviated asa
map
from
finite public histories{p*}^i
to price schedules.Using
familiararguments
(see,e.g.,FLM),
thesetwo
restrictionsimply:Lemma
2
(Recursion
of
PerfectPublic
Equilibria)The
continuationof
any
PPE
afterany
historyisitself
a
PPE
inthefullgame and
yieldspayoffsinthecontinuationequal
towhat would
have
been
obtained
had
thestrategy profilebeen used
from
thestart.Before
moving on
to aformal analysisof
the perfectpublicequilibria fortherepeatedgame,
we
highlight
some
ofthenovel
featuresof
thisgame.
First,
a
PPE
ina repeatedgame
with hidden
informationmust be
immune
totwo
kindsof
current-perioddeviations.
A
firmdeviates "off-schedule"when
itchooses a
pricenotspecified forany
costrealization: i.e.,apricenotinthe
range of
Pt.When
a firm
prices inthismanner,
ithas
unam-biguouslydeviated;consequently,ifthecollusive
scheme
prescribesa
punishment
(i.e.,a
reductionincontinuationvalues) following
such
adeviation,and
iftheprospectof
such a
punishment
detersa
firm
from
undertakingtheoff-scheduledeviation(as willbe
trueiffirms aresufficientlypatient), thenthatpunishment
willnever
actuallyoccur
(i.e., it isoff theequilibriumpath).By
contrast,a
firm
deviates "on-schedule"when
itchooses a
price that is assigned tosome
cost level,but
notits
own.
For example, a
firmmay
be
tempted
tochoose
thelower
price assignedtoa
lower
costrealization inordertoincreaseits
chances of
winning
themarket. Importantly,an on-schedule
de-viationis notperfectlyobservable,as
a
deviation, torivalplayers:a
rivalcan
notbe
surethatthe deviatingfirm
was
nottrulyof
the cost type thatit is imitating.Thus,
firmsmay
be
tempted
toconstructacollusive
scheme
thatimposes
aharshpunishment
when
low
prices are chosen,asthiswould
servetopreventon-scheduledeviationsby
higher-costtypes.But
such a
punishment
would
be
costly, sinceitwould
occur
alongtheequilibriumpathof
play,whenever
firmsactually realizedlow
costs.Secondly,
we
notethattherepeatedgame
admits aricherrangeof
equilibriumpricingschedules than didthe staticgame.
As
discussedinSection 2, theequilibriumpricing schedule forthe staticgame
can have
neitherflatsegments nor jumps. This
isno
longertrue intherepeatedgame. For
example,
we
may
specify thatundercuttinga
flatprice leads toa
harshpunishment,
and
thismay
be
sufficient to deterthe (off-schedule) deviation.More
generally, the repeatedgame
adds
athird ingredient to the tradeoffthat afirm confrontswhen
it considers reducing price.Now
lowering
pricemeans
notjust: 1)an
increasedchance
ofwinning,and
2)lower
profit-if-win, butalso 3) adifferentcontinuationvalue.
Finally,
we
observe thatthemixture
of
on-and
off-scheduleincentiveproblems
isendogenous
to the firms' choice
of
price schedules.A
completely
flat schedulemeans
a
relatively acute off-schedule incentiveproblem,
butno
on-scheduleproblem.
A
price schedule that spans thefuU
relevantrange
ofpriceshasno
off-scheduleproblem,
but possiblyan
acuteon-scheduleproblem.
With
thechoiceof apriceschedule,therefore, firmsendogenously determine
the supportof
actual pricesthatmay
be
observed.3.2
Optimal
Symmetric
Collusion:
The Program
Ina
symmetric
perfect public equilibrium(SPPE),
followingevery publichistory,firmsadopt
sym-metric price schedules: Si{p^,...,
p^)
=
Sj{p^, ...,p^), Vz,j,T,p\
...,p^.
Symmetry
means
thatallfirms suffer future
punishments
and rewards
together.For example,
itmight be
thatfirm 1 runsthe risk
of
triggeringan
industry-widebreakdown
incollusionshould
itrepeatedlychaigeimplau-sibly
low
prices.Symmetric
equilibriathus capturethe "fearof
breakdown"
thatcollusive firmsmay
experience,asdiscussedinthe Introduction.'^Our
goal is to formulate theproblem
of
characterizing the optimalSPPE
insuch a
way
thatwe
can
employ
the toolsof
(static)mechanism
design.We
begin
by
constructinga
programming
problem
(The
"FactoredProgram")
intermsof
current-periodprice schedulesand
"continuation'^ Despite
itsclear restrictiveness,aricharrayofstrategy profilesstillfitunderthesymmetricheading.Forinstance,
anyamountofhistorydependenceis allowed. Thus,pricewars canbetriggeredbyanynumberofperiodsof"bad behaviot" Thismeansinparticular thatthelaw oflargenumbers
may
beutilized: sincetypesaredrawni.i.d. across timeandfirms,an accuratemeasure oftheaverage typeofeachfirmbecomesavailable, andthis can becomparedtotherealizedaverageannouncedtype(i.e.,the typecorrespondingto theaimouncedprice)todeterminewhether a
given firm hascheated (See,e.g.,Radno*(1981)andRubinsteinand"Ykari(1983)foradiscussionof
how
thelaw oflargenumbers
may
be usedtoimplement such "reviewstrategies.")Moreover,gradationsofpunishmentareavailable: pricewars canvaryinseverityandlengthandsomay
betailored tothe"fitthe crime."payofffunctions,"
and
claim
from
APS
logic thatany
solution to thisproblem
corresponds
to an optimalSPPE. To
this end,we
firstmake some
notational adjustments that reflectthesymmetric
setting. Let Tl{p)
denote
theex
ante profit toa
firmwhen
all firms use thesymmetric
pricing strategyp,and
letU.{p',p),representtheex
anteprofit toafirmwhen
itadoptsthepricing schedule p'whileallotherfirmsemploy
thesymmetric
pricing strategyp
.(The
notation for thecontinuationpayoff
functiondeveloped
below
is interpreted similarly.) Let V^G
$ftdenote
the setof
SPPE
continuation values
and
write^
=
infV^and
v
=
sup
Vg. Note, initially, thatwith
acontinuum
of
possiblepricing strategies thereisno
a
priori basisfrom which
to argue that eitherv
E
Vso-veVs.
Any
symmetric
public strategy profile s=
(s,...,s)can he
factored into a first-period priceschedule
p
and a
continuationpayoff
function v: 5R"—
> 5?.The
continuationpayoff
function describes therepeated-game payoff v{p) enjoyed
by
all firmsfrom
the perspectiveof
period2
onward
aftereach
first-periodprice realizationp
=
(p^, ...p„)G
SR".Each
firm'spayoff
from
scan
thenbe
writtenasll{p)+
6Ev{p).
Now
considerthefollowingproblem
inwhich
we
choose
factorizations {p,v) ratherthanchoos-ing strategiesdirectly:
The
Factored
Program:
Choose
price schedulep
and
continuationpayoff
functionv
tomaxi-mize
n(p)
+
6Ev{p)
subjectto:
Vp
e
3?^,v{p)
e
V^,and
Vp',
n(p)
+
8Ev{p)
>
n{p',p)+
6Ev{pf,p).
An
important ingredientinour
argument
iscontainedinthefollowingresult:Lemma
3
Ifthesymmetric
public strategyprofile s*=
(s*,...,s*)factorsinto(p*,v*),and
{p*,v*)solves the
Factored
Program,
thens* isan
optimal
SPPE.
The
lemma
followsfrom
the fact that (p,v) is the factorizationof
some
SPPE
ifand
only
ifitsatisfiesthe
Factored
Problem's
two
constraints.To
seethisequivalence,notefirstthat,by
Lemma
2, everySPPE
must
satisfyv{p)
G
Vg.Every
SPPE
also satisfies thesecond
constraint,because
an
SPPE
must
be
immune
todeviationsinthefirstperiod. Conversely,one
shows
thatany
(p,v)satisfyingthe constraintsisthefactorization
of
an
SPPE
by
appending
top
theSPPE
continuationstrategy profilesunderlying v{p),
and
thenarguingfrom
theone-stage-deviation-principlethattheconstructedstrategy profileisitself
an
SPPE.
We
referthe readertoAPS
forfulldetails.To
analyzetheFactoredProgram
further,we
rewriteitinequivalentinterim-profitform, parsingthe constraints into on-
and
off-schedule conditions.Using
abbreviated notation in lightof
thesymmetric
analysis:The
Interim
Program: Choose
priceschedulep
and
continuationpayoff functionv
tomaximize
Ei[Yi{p{ei),p,ei)
+
5E_iv{p{ei),p)\subjectto:
Off-Schedule Constraints: Vp'
^
p{[9,0]),(IC-ofifl)
Vp_i,
v{p',p_,)eV,
(ic-off2) yOi, u(p{ei),p,Oi)
+
8E.iv{p{ei),p)
>
n(p',p,Oi)+
8E.iv{p',p)
On-Schedule
Constraints: \/9i,(IC-onl)
Vp_i,
v{p{ei),p_i)eVs
(ic-on2) v^i, n{p{ei),p,Oi)
+
6E_iv(p{ei),p)
>
n(p{ei),p,Oi)+
6E_iv{p{9i),p).Notice
how
the on-schedule constraints are written in "direct" form: forgiven
p
and
v, the on-scheduleconstraintrequires thatafirm
with typedidoes
betterby
"announcing"
thatitstypeis 6ithan
by
announcing
some
othertype, 9i. This observation suggests thattheon-schedule
constraintcan be understood
as corresponding to a "truth-telling" constraint inan
appropriatemechanism
design formulation.
4.
Optimal
Symmetric
Collusion
Among
Patient
Firms
We
break our
analysisof
symmetric
collusion intotwo
parts. Inthe current section,we
suppose
that firms arepatient, so thatthe ofif-schedule incentive constraints aresatisfied.
The
assumption
ofpatientfirmsisthen relaxedinSection5.
Our
strategyistoshow
thatwhen
firmsarepatient,theInterimProgram
can
be
furtherrelaxedto yieldwhat
we
calltheMechanism
Design Program.
We
thenuse standardtoolsfrom
themechanism
designliteraturetocharacterizeoptimal
SPPE,
soas to offerpredictionsregardingobserved
pricingpatterns incollusivesettings.
4.1
The
Mechanism
Design
Program
To
begin,we
definea
new
function that represents interim profit in direct form:H(9,
9;p)=
The
Mechanism
Design
Program:
Choose
priceschedulep
and
functionT
: [d,'9\—
> 3? tomaximize
E[H{9,9;p)-T{e)]
subject to : V^,
T{e)
>
(ic-onM)
v^,e,H{e,
e]p)-
T{e)
>
h{9,
e-.p)-
T{e).To
relate the Interimand
Mechanism
Design Programs,
we
firstobserve
that,while
theMech-anism Design
Program
uses different notation, the objective function is thesame.
We
thusfo-cus
on
comparing
the constraints.Observe
thatforany
(p^,v^)which
satisfies the constraintsof
the InterimProgram,
we
can
find acorresponding
T^
such
that (p^,T^)
satisfies the constraintsof
theMechanism
Design Program.
To
see this, recall thatv
=
supK,
and
define T^{9i)=
6{v
—
E-iV^{p^(9i),p^)).Then,
the restrictionv
e
Vg implies thatv{p)
<v
for all p,and
thusT\9)
>
0. Further, {p^,v^) satisfies (IC-on2) in theInterimProgram
ifand
onlyif (p^,T^)
satis-fies(IC-onM).
Now
consider a(p^,T^) which
satisfies the constraintsof
theMechanism
Design Program.
As
we
discuss below, the ofif-schedule constraintsof
the InterimProgram
can
be
satisfiedwhen
firms are sufficientlypatient.
Turning
to theon-schedule
constraints,we
can
definev^
such
thatT^{9i)
=
5{v-
E_iV^{p^(9i),p^)),
whereby (IC-onM)
and
(IC-on2) are equivalent.But
(IC-onl)
requires further thatv^{p^{9i),p_i)
^
^a,and
so itremains
toverify thatsuch a
v^
can
be
found.
There
aretwo
cases: (a) the"no
wars"
case,where
T^
=
0;and
(b)thecasewhere
T^
issometimes
positive.Consider
firstcase(a),whereby
our
construction requiresv^{p^{9),p^))
=v
forall 0. IfK
isclosed, so that
v
E
Vs, this also satisfies (IC-onl).We
proceed
by showing
thatv
can indeed
be
obtainedas
an
SPPE
ifthe solution totheMechanism
Design
Program
is (p*,T*
=
0).Thus,
suppose
forthemoment
that{p*,T*
=
0).Ifwe
relaxtheInterimProgram by
replacingthe(IC-onl) constraint,v{p{9), p_j)
G
K,
with
the constraintv{p{9),p_^
e
K
U
v,two
conclusionsfollow. First, (p*,v*(p*{9),p*))
=
v) solvestherelaxed InterimProgram. Second,
any
solution to therelaxed InterimProgram
yields profitsatleastaslargeas thesupremum
of
the profits available in the(unrelaxed) InterimProgram,
v.Formally,EH{9,
9;p*)+6v
>
v,orEH(9,
9;p*)/(l-6)>v.
But, since(p*,T*
=
0) satisfies(IC-onM), p*
must
satisfy(IC-0n2)
forany
continuationvalue
functionthat
does
notvary
with
theannounced
costtype.Thus,
the collusivescheme where
p*
isused
inevery periodisan
SPPE
so long
as theoff-scheduleconstraints are satisfied. Ifp* yieldsper-periodprofitsgreaterthanthestatic
Nash
equilibrium(aswe
will establish inSections4.3.4and
4.4.3),theoff-scheduleconstraints
can
be
satisfiedwhen
firmsare sufficiently patient(supportedby
the threatof foreverreverting to thestaticNash
equilibrium). Therefore, sinceEH{6,
6;p*)/(l—
6)G
Vg, the definitionof
v
impliesv
>
EH{d,6;p*)/{l
—
6).As
the reverse inequalitywas
established above,we
thusconclude
that thebestSPPE
is the stationary equilibriumwhere
each
firm uses p* in everyperiod.
Summarizing:
Proposition
4
(Staticizing)Suppose
thattheMechanism
Design
Program
issolvedby
(p*,T*),where
T*
=
0.Then
36*e
(0, 1)such
that, if6
>
6*,(i)p*
maximizes
EH
{9,e;p) subjecttoyd,6,H{9,
9;p)>
H{9,
9;p).(ii)In
an
optimalSPPE,
firms adopt
p*afterallequilibrium-pathhistories. (Hi)V
=
EH
(9, 9;p*)/{I-
6)e
Vs.This resultestablishesconditions
under
which
theproblem
of
findingan
optimalSPPE
can be
reduced
to asimpleand
familiar(static)mechanism
designproblem. In theremainder
of Section4
we
show
that,under
many
conditions, the solutionto theMechanism
Design
Program
(p*,T*)
in fact entails
T*
=
0.However,
we
also describe a limited setof
circumstancesunder
which
pricewars
may
be used
(case(b)from
above). Itispossibletogeneralize part(iii)of
the staticizingpropositiontothecase
of
nonstationaryequilibria,showing
thatwhen
firmsare sufficientlypatient,E[H{9,
9;p*)—
T*{9)]/{1—
6)=
v.However,
such
aresultrequires thatK
isconvex,
sothatallpositive, finite
punishments
are availablewhen
firms are sufficiently patient."Because
most
of
our
resultsconcern
thecasewhere
T*
=
0,we
willnotpursue
thatextensionformallyhere.4.2
Consequences
of Incentive
Compatibility
Now
that theproblem
of
findingan
optimalSPPE
hasbeen
reduced
toa
standardprogramming
problem,
we
proceed
toanalyzetheprogram.
Our
firststep inthisprocessisto characterizethe im-plicationsoftheon-scheduleconstraint thatisfound
intheMechanism
Design
Program
(IC-onM).
We
do
thisin thefollowinglemma
(where
we
use
thenotationHe{9,9]p)
=
^H{9,
9;p)Ig^g):Lemma
5
(ConstraintReduction)
(p,T)
satisfies(IC-onM)
ifand
onlyif(p,T)
alsosatisfies:(i). p{9) is
weakly
increasing,and
(ii).
H{9,
9;p)-
T{9)
=
H{9,9;p)
-
T{9)-
J
He{9,
9;p)d9.Thisresultisstandardinthe
mechanism
designliterature,"and
itfollowsfrom
the single-crossing^^
As
dicussed above,we
will establish that thereisalwaysvariation inthesetV^,which can be
made
arbitrarilylargebytakingScloseto1.
One way
toguaranteeconvexitywouldbetoextendthemodeltoallow public randomization.For example,foreachprice vector,
we
couldspecifyaprobability thatthe firmsrevert tothestaticNashequilibrium.^^ For astandard