You are on page 1of 69

Antitrust Defined

January 31 2010| Filed Under Business Ethics, M&A, Microeconomics, herman Antitrust Act,
tandard !il
"n the s#irit o$ $ree trade and com#etition, e$$orts ha%e &een made since the late 1'th century to #ut
an end to (&i) &usiness(, or mono#olies, that dominate the local and $orei)n mar*et share, control
#rices and+or #re%ent com#etition, -ere .e ta*e a loo* at the history o$ antitrust la.s and e/amine
some o$ the ar)uments a)ainst them,
General Background: the Rise of Consolidation
Be$ore any le)islation deemed mono#olies ille)al in the United tates, an act .as #assed in En)land
in 1021 #re%entin) the 2ro.n $rom )rantin) mono#olies to &usinesses in return $or cash, 3he
tatute o$ Mono#olies made the #ractice ille)al, thus hel#in) ensure an e$$icient economy,
By the 1'th century, the U,, .as startin) to e/#erience .hat came to &e *no.n as &i) &usinesses,
"n the a)e o$ industriali4ation the emer)ence o$ the railroad allo.ed $or the $irst real $orm o$ mass
trans#ortation, 3his meant that #eo#le and &usinesses could coordinate themsel%es .ithin one day
o%er distances that #re%iously .ere ne%er mana)ea&le, !%er time, trans#ortation com#etition &e)an
to rise and railroads e%entually turned to consolidation in order to deal .ith decreasin) #ro$its,
2onsolidation continued in the 1'th century as &usinesses $aced increasin) com#etition, 5ith the
intention o$ controllin) #rices, com#etin) &usinesses coordinated #rices in order to minimi4e
com#etition and increase #ro$it, "n essence, it .as a $orm o$ mono#oli4ation as com#anies
consolidated their o#erations in the mar*et,
Standard Oil
"t .asn(t until the case o$ John 6, 7oc*e$eller(s tandard !il in 1882 that &i) &usiness &ecame an
issue that the #u&lic needed to con$ront, tandard !il mer)ed .ith its com#etition &y $ormin) the
tandard !il 3rust,
3his .as &asically the esta&lishment o$ .hat is *no.n no. as a &oard o$ trustees, 95ith time,
tandard !il heralded the idea o$ a holdin) com#any, .hich $unctioned as a &oard o$ trustees:,
3hese trustees &ecame le)ally res#onsi&le $or all assets and #ro#erties o$ tandard !il and the
mer)ed com#anies, toc*holders recei%ed trust certi$icates, each o$ .hich re#resented 20 shares o$
tandard !il:, and all #ro$its made $rom the nine com#onent industries o$ the trust .ere sent to the
trustees, .ho .ould then determine the di%idends, 3he trustees .ere res#onsi&le also $or a##ointin)
all directors and mana)ers o$ the com#any, there&y solidi$yin) their control o%er tandard !il and
$undamentally, the industry,
;rices across the .hole industry .ere essentially &ein) directed &y one entity, Any other com#any
not a #art o$ tandard !il(s trust .as una&le to com#ete a)ainst the mono#olistic )iant, and .as
there$ore una&le to enter the industry, tandard !il 3rust could &asically run out any ri%al &usiness
&y settin) a lo.er #rice that .ould &e im#ossi&le $or a sin)le<entity com#any to o#erate on,
By this time, the #u&lic .as .itnessin) limited com#etition and restricti%e #rice controls in t.o
industries= in the rail.ay industry, &y American 7ail.ay Union, and in the oil industry, &y the
tandard !il 3rust, 3hese t.o entities s#urred #u&lic discontent, .hich in turn led to a reaction to
Object 1
control &i) &usiness and mono#olistic #ractices,
The Sherman Antitrust Act
John herman, !hio la.yer and senator, .as com#elled to dra$t the herman Antitrust Act o$ 18'0,
3he act(s #rimary )oal .as to limit the e/#ansion o$ mono#olies, the restriction o$ $ree trade
9com#etition: and the im#osition o$ #rice $i/in) &y industry mem&ers or any com&ination o$
&usiness #ractices that led to the restriction o$ trade,
-o.e%er, the terms >restraint o$ trade>, >com&ination> and >mono#oli4e> .ere not #ro#erly de$ined,
turnin) the act into a %a)ue and .ea* le)islation, "n addition, there .as no inde#endent commission
esta&lished at the time to in%esti)ate #ossi&le antitrust cases, ma*in) the act di$$icult to en$orce,
The Clayton Antitrust Act
"n 1'11, a ne. act .as #assed &y la.yer -enry 6e ?amar 2layton o$ Ala&ama and .as called the
2layton Antitrust Act, "t aimed to rein$orce the herman Act &y )i%in) the )o%ernment more #o.er
.hen $aced .ith mono#olies, "t strictly outla.ed com#anies $rom en)a)in) in #rice<$i/in)
a)reements to lessen com#etition, 3he act also #re%ented indi%iduals $rom ser%in) as directors at
com#etin) com#anies,
AT&T and the Eolution of Antitrust
3he American 3ele#hone and 3ele)ra#h 2or#oration 9A3&3:, set u# in 188@, .as the $irst com#any
to o$$er lon)<distance #hone ser%ices across the U,, "n the meantime, Ale/ander Araham Bell,
9.ho had a #atent on the tele#hone until 18'1:, created a com#any .ith his $inanciers, called
American Bell 3ele#hone 9chan)ed $rom the #re%ious names o$ Bell 3ele#hone 2om#any and
Bational Bell 3ele#hone 2om#any:, For some time a$ter Bell(s #atent e/#ired, American Bell
mana)ed to .ard o$$ com#etition throu)h #rice $i/in) and a))ressi%e mer)ers .ith com#etitors,
By 6ecem&er 18'', A3&3 &ou)ht American Bell and )re. as it acCuired stoc* in other com#anies
#ro%idin) tele#hone ser%ices, E%entually, it &ecame the nation(s #rimary tele#hone ser%ice #ro%ider,
and it #ro#osed to esta&lish itsel$ as a mono#oly, 3he #ro#osal .as a)reed u#on &y the )o%ernment
in 1'13 throu)h the Din)s&ury 2ommitment,
!%er the years, A3&3 9a,*,a, >Ma Bell>: enEoyed almost a com#lete mono#oly o%er lon)<distance
tele#hone ser%ices, .hile also controllin) 22 local tele#hone ser%ice #ro%iders across the country,
And .hile there .as com#etition, the si4e and in$luence o$ A3&3 essentially made the com#any a
mono#oly throu)h and throu)h,
?on)<distance com#etition .as introduced &y the Federal 2ommunications 2ommission 9F22: &y
1'F@, endin) Ma Bell(s mono#oly in that area, Mean.hile, a 1'F1 antitrust la.suit &rou)ht u#on &y
the 6e#artment o$ Justice .as $inally settled in 1'82, $orcin) Ma Bell to &rea* u# her mono#oly
com#letely and di%est o$ her local tele#hone ser%ice #ro%iders, 3he local #ro%iders .ere
reor)ani4ed into se%en inde#endent, re)ional Bell o#eratin) com#anies, commonly *no.n as >&a&y
&ells>,
The !rice of Success
Ar)uments ha%e &een made that antitrust la.suits o$ten #unish success$ul &usinesses, 3hese
ar)uments say that com#anies that ha%e &een deemed mono#olies < such as Microso$t < as o##osed
to >harm$ul> mono#olies created &y the )o%ernment < such as the old A3&3 < are in reality &ein)
#unished $or their success, .hich comes not $rom .i#in) out com#etition &ut $rom the ina&ility o$
ri%al com#anies to &e e$$icient enou)h to com#ete,
3hus, the ine$$iciency o$ smaller com#etitors is actually &ein) re.arded .hen antitrust la.s are
enacted, .hile success$ul &i) &usinesses must endure restrictions on their #roducti%ity,
com#etiti%eness and e$$iciency, 3he ar)ument )oes on to state that com#etition in some industries
under scrutiny does in $act continue to e/ist= i$, $or e/am#le, a client does not .ant to &uy
Microso$t, he or she can al.ays turn to another alternati%e such as A##le,
Conclusion
Bonetheless, antitrust la.s remain ada#ta&le to time and le)al o#inion, "t there$ore di$$icult to
strictly de$ine .hat is le)al and ille)al .hen it comes to &i) &usiness, and o$ten it is sa$er $or
com#anies to see* le)al a##ro%al $rom the 6e#artment o$ Justice &e$ore concludin) a mer)er that
may cause it to &e la&eled as mono#olistic or anti<com#etiti%e,
AB3"37U3
?!U" DA;?!5
School of Law, Harvard University, and National Bureau of Economic Research
2A7? -A;"7!
Haas School of Business and Department of Economics, University of California, Bereley
Contents
1, "ntroduction 10FF
2, Mar*et #o.er 10F8
2,1, 6e$inition o$ mar*et #o.er 10F'
2,2, in)le<$irm #ricin) model accountin) $or ri%als 1080
2,3, Multi#le<$irm models 1083
2,3,1, 2ournot model .ith homo)eneous #roducts 1083
2,3,2, Bertrand model .ith di$$erentiated #roducts 108@
2,3,3, !ther )ame<theoretic models and collusion 1080
2,1, Means o$ in$errin) mar*et #o.er 108F
2,1,1, ;rice<cost mar)in 108F
2,1,2, FirmGs elasticity o$ demand 10'0
2,1,3, 2onduct 10'1
2,@, Mar*et #o.er in antitrust la. 10'@
3, 2ollusion 10'8
3,1, Economic and le)al a##roaches= an introduction 10''
3,1,1, Economic a##roach 10''
3,1,2, ?e)al a##roach 1101
3,2, !li)o#oly theory 1103
3,2,1, Elements o$ success$ul collusion 1103
3,2,2, 7e#eated oli)o#oly )ames and the $ol* theorem 1101
3,2,3, 7ole o$ communications 1100
3,3, "ndustry conditions &earin) on the li*elihood o$ collusi%e outcomes 1108
3,3,1, ?imited )ro.th $or de$ectin) $irm 1108
3,3,2, "m#er$ect detection 110'
3,3,3, 2redi&ility o$ #unishment 1110
3,3,1, Mar*et structure 1112
Hand!oo of Law and Economics, "olume #
Edited !y $% &itchell 'olinsy and Steven Shavell
H #(() Elsevier B%"% $ll ri*hts reserved
D+,- .(%.(./0S.1)23()4(5()6(#(.132
10F1 L% 7aplow and C% Shapiro
3,3,@, ;roduct di$$erentiation 111F
3,3,0, 2a#acity constraints, e/cess ca#acity, and in%estment in ca#acity 111F
3,3,F, Mar*et dynamics 111'
3,1, A)reements under antitrust la. 1121
3,1,1, !n the meanin) o$ a)reement 1121
3,1,2, A)reement, economics o$ collusion, and communications 1121
3,@, !ther hori4ontal arran)ements 112'
3,@,1, Facilitatin) #ractices 1130
3,@,2, 7ule o$ reason 1132
3,0, Antitrust en$orcement 1130
3,0,1, "m#act o$ antitrust en$orcement on oli)o#olistic &eha%ior 113F
3,0,2, 6eterminants o$ the e$$ecti%eness o$ antitrust en$orcement 113F
1, -ori4ontal mer)ers 1138
1,1, !li)o#oly theory and unilateral com#etiti%e e$$ects 113'
1,1,1, 2ournot model .ith homo)eneous #roducts 113'
1,1,2, Bertrand model .ith di$$erentiated #roducts 1113
1,1,3, Biddin) models 1118
1,2, !li)o#oly theory and coordinated e$$ects 111'
1,3, Em#irical e%idence on the e$$ects o$ hori4ontal mer)ers 11@2
1,3,1, toc* mar*et #rices 11@3
1,3,2, Accountin) measures o$ $irm #er$ormance 11@1
1,3,3, 2ase studies 11@@
1,1, Antitrust la. on hori4ontal mer)ers 11@F
1,1,1, Bac*)round and #rocedure 11@F
1,1,2, Anticom#etiti%e e$$ects 1100
1,1,3, E$$iciencies 1102
1,@, Mar*et analysis under the -ori4ontal Mer)er Auidelines 110'
1,@,1, Mar*et de$inition= )eneral a##roach and #roduct mar*et de$inition 11F0
1,@,2, Aeo)ra#hic mar*et de$inition 11F@
1,@,3, 7i%alsG su##ly res#onse 11FF
1,0, ;redictin) the e$$ects o$ mer)ers 11F8
1,0,1, 6irect e%idence $rom natural e/#eriments 11F8
1,0,2, Mer)er simulation 11F'
@, Mono#oli4ation 1180
@,1, Mono#oly #o.er= economic a##roach 1181
@,1,1, 7ationale $or mono#oly #o.er reCuirement 1181
@,1,2, A##lication to challen)ed #ractices 1183
@,2, ?e)al a##roach to mono#oli4ation 1180
@,2,1, Mono#oly #o.er 118F
@,2,2, E/clusionary #ractices 11'1
@,3, ;redatory #ricin) 11'1
@,3,1, Economic theory 11'@
@,3,2, Em#irical e%idence 11'0
Ch% .1- $ntitrust 10F@
@,3,3, ?e)al test 11'8
@,1, E/clusi%e dealin) 1203
@,1,1, Anticom#etiti%e e$$ects 1203
@,1,2, E$$iciencies 120'
@,1,3, ?e)al test 1210
0, 2onclusion 1213
Ac*no.led)ements 1211
7e$erences 1211
2ases 1221
10F0 L% 7aplow and C% Shapiro
A"stract
3his is a sur%ey o$ the economic #rinci#les that underlie antitrust la. and ho. those
#rinci#les relate to com#etition #olicy, 5e address $our core su&Eect areas= mar*et
#o.er, collusion, mer)ers &et.een com#etitors, and mono#oli4ation, "n each area, .e
select the most rele%ant #ortions o$ current economic *no.led)e and use that *no.led)e
to critically assess central $eatures o$ antitrust #olicy, !ur o&Eecti%e is to $oster the
im#ro%ement o$ le)al re)imes and also to identi$y to#ics .here $urther analytical and
em#irical e/#loration .ould &e use$ul,
#ey$ords
antitrust, com#etition #olicy, mono#oly, mar*et #o.er, mar*et de$inition, oli)o#oly,
collusion, cartels, #rice<$i/in), $acilitatin) #ractices, mer)ers, hori4ontal mer)ers,
unilateral e$$ects, coordinated e$$ects, mono#oli4ation, e/clusionary #ractices,
#redatory #ricin), e/clusi%e dealin)
8EL classification= D21, ?12, ?13, ?10, ?11, ?12
Ch% .1- $ntitrust 10FF
%& 'ntroduction
"n this cha#ter, .e sur%ey the economic #rinci#les that underlie antitrust la. and use
these #rinci#les to illuminate the central challen)es in $ormulatin) and a##lyin) com#etition
#olicy, !ur t.in )oals are to in$orm readers a&out the current state o$ *no.led)e
in economics that is most rele%ant $or understandin) antitrust la. and #olicy and to
critically a##raise #re%ailin) le)al #rinci#les in li)ht o$ current economic analysis,
ince the #assa)e o$ the herman Act in 18'0, antitrust la. has al.ays re%ol%ed
around the core economic conce#ts o$ com#etition and mar*et #o.er, For o%er a century,
it has &een ille)al in the United tates $or com#etitors to enter into #rice<$i/in)
cartels and related schemes and $or a mono#olist to use its mar*et #o.er to sti$le com#etition,
"n inter#retin) the antitrust statutes, .hich s#ea* in %ery )eneral terms, U,,
courts ha%e al.ays #aid attention to economics, Iet the role o$ economics in sha#in)
antitrust la. has e%ol%ed )reatly, es#ecially o%er the #ast $e. decades, 3he )ro.in) in$luence
o$ economics on antitrust la. can &e traced in #art to the 2hica)o chool, .hich,
startin) in the 1'@0s, launched a #o.er$ul attac* on many antitrust rules and case outcomes
that seemed to lac* solid economic under#innin)s, But the )ro.in) in$luence o$
economics on antitrust la. also has resulted $rom su&stantial theoretical and em#irical
ad%ances in industrial or)ani4ation economics o%er the #eriod since then, 5ith a la),
o$ten s#annin) a cou#le o$ decades, economic *no.led)e sha#es antitrust la., "t is our
ho#e in this essay &oth to shar#en economistsG research a)endas &y identi$yin) o#en
Cuestions and di$$iculties in a##lyin) economics to antitrust la., and also to accelerate
the dissemination o$ economic *no.led)e into antitrust #olicy,
Antitrust economics is a &road area, o%erla##in) to a )reat e/tent .ith the $ield o$ industrial
or)ani4ation, 5e do not o$$er a com#rehensi%e e/amination o$ the areas .ithin
industrial or)ani4ation economics that are rele%ant $or antitrust la., 3hat tas* is $ar
too dauntin) $or a sin)le sur%ey and is already accom#lished in the $orm o$ the three%olume
Hand!oo of ,ndustrial +r*ani9ation 91'8'a, 1'8'&, 200F:,1 "nstead, .e $ocus
our attention on $our core economic to#ics in antitrust= the conce#t o$ mar*et #o.er
9section 2:, the $orces that $acilitate or im#ede e$$orts &y com#etitors to en)a)e in collusion
9section 3:, the e$$ects o$ mer)ers &et.een com#etitors 9section 1:, and some &asic
$orms o$ sin)le<$irm conduct that can constitute ille)al mono#oli4ation, namely #redatory
#ricin) and e/clusi%e dealin) 9section @:,2 "n each case, .e attem#t to select $rom
the &road &ase o$ models and a##roaches the ones that seem most hel#$ul in $ormulatin)
a .or*a&le com#etition #olicy, Furthermore, .e use this analysis to scrutini4e the
corres#ondin) $eatures o$ antitrust la., in some cases #ro%idin) a $irmer rationali4ation
1 chmalensee and 5illi) 91'8'a, 1'8'&: and Armstron) and ;orter 9200F:,
2 ince the $ield o$ antitrust economics and la. is $ar too lar)e to co%er in one cha#ter, .e are $orced to omit
some to#ics that are %ery im#ortant in #ractice and ha%e themsel%es &een su&Eect to e/tensi%e study, includin)
Eoint %entures 9touched on &rie$ly in su&section 3,@,2:, %ertical mer)ers, &undlin) and tyin), %ertical intra&rand
restraints, the intersection o$ antitrust la. and intellectual #ro#erty la., and most $eatures o$ en$orcement
#olicy and administration, includin) international dimensions,
10F8 L% 7aplow and C% Shapiro
$or current #olicy and in others identi$yin) im#ortant di%er)ences,3 For reasons o$ concreteness
and o$ our o.n e/#ertise, .e $ocus on antitrust la. in the United tates, &ut
.e also em#hasi4e central $eatures that are #ertinent to com#etition #olicy else.here
and $reCuently relate our discussion to the #re%ailin) re)ime in the Euro#ean Union,1
(& )arket *o$er
3he conce#t o$ mar*et #o.er is $undamental to antitrust economics and to the la., E/ce#t
$or conduct su&Eect to per se treatment, antitrust %iolations ty#ically reCuire the
)o%ernment or a #ri%ate #lainti$$ to sho. that the de$endant created, enhanced, or e/tended
in time its mar*et #o.er, Althou)h the reCuisite de)ree o$ e/istin) or increased
mar*et #o.er %aries &y conte/t, the nature o$ the inCuiry is, $or the most #art, Cualitati%ely
the same,
"t is im#ortant to em#hasi4e at the outset that the mere #ossession o$ mar*et #o.er
is not a %iolation o$ antitrust la. in the United tates, 7ather, the inCuiry into mar*et
#o.er is usually a threshold CuestionJ i$ su$$icient mar*et #o.er is esta&lished, it is
then as*ed .hether the conduct in CuestionKsay, a hori4ontal mer)er or an alle)ed act
o$ mono#oli4ationKconstitutes an antitrust %iolation, "$ su$$icient mar*et #o.er is not
demonstrated, the inCuiry terminates .ith a %ictory $or the de$endant,
-ere, .e &e)in our treatment o$ antitrust la. and economics .ith a discussion o$
the &asic economic conce#t o$ mar*et #o.er and its measurement, "nitially, .e de$ine
mar*et #o.er, em#hasi4in) that, as a technical matter, mar*et #o.er is a Cuestion o$ de)ree,
3hen .e e/#lore the $actors that determine the e/tent o$ mar*et #o.er, $irst .hen
e/ercised &y a sin)le $irm and then in the case in .hich multi#le $irms interact, 5e also
consider %arious methods o$ in$errin) mar*et #o.er in #ractice and o$$er some $urther
remar*s a&out the relationshi# &et.een the conce#t o$ mar*et #o.er as understood &y
3 3here are a num&er o$ &oo*s that ha%e o%erla##in) #ur#oses, includin) Bor* 91'F8:, -ylton 92003:, ;osner
92001:, and 5hinston 92000:, the latter &ein) closest to the #resent essay in the .ei)ht )i%en to $ormal
economics,
1 As im#lied &y the discussion in the te/t, our re$erences to the la. are #rimarily meant to ma*e concrete
the a##lication o$ economic #rinci#les 9and secondarily to o$$er s#eci$ic illustrations: rather than
to #ro%ide detailed, de$initi%e treatments, !n U,, la., the interested reader should consult the e/tensi%e
treatise $ntitrust Law &y Areeda and -o%en*am#, many %olumes o$ .hich are cited throu)hout
this essay, !n the la. in the Euro#ean Union, see, $or e/am#le, Bellamy and 2hild 92001:, 6a&&ah
92001:, and 5alle de Ahelc*e and Aer%en 92001:, A .ide ran)e o$ additional in$ormation, includin)
$ormal #olicy statements and en$orcement statistics, are no. a%aila&le on the "nternet, -el#$ul lin*s
are= Antitrust 6i%ision, 6e#artment o$ Justice= htt#=++...,usdoE,)o%+atr+inde/,htmlJ Bureau o$ 2om#etition,
Federal 3rade 2ommission= htt#=++...,$tc,)o%+$tc+antitrust,htmJ Euro#ean Union, 6A 2om#etition=
htt#=++ec,euro#a,eu+comm+com#etition+inde/Len,htmlJ Antitrust ection o$ the American Bar Association=
htt#=++...,a&anet,or)+antitrust+home,html,
Ch% .1- $ntitrust 10F'
economists and as em#loyed in antitrust la.,@ Further ela&oration a##ears in sections 1
and @ on hori4ontal mer)ers and mono#oli4ation, res#ecti%ely,
#%.% Definition of maret power
Microeconomics te/t&oo*s distin)uish &et.een a #rice<ta*in) $irm and a $irm .ith some
#o.er o%er #rice, that is, .ith some mar*et #o.er, 3his distinction relates to the demand
cur%e $acin) the $irm in Cuestion, "ntroducin) our standard notation $or a sin)le $irm
sellin) a sin)le #roduct, .e .rite P $or the #rice the $irm recei%es $or its #roduct, X $or
the $irmGs out#ut, and X(P) $or the demand cur%e the $irm #ercei%es that it is $acin),
.ith X
(P ) 0,0 5hen con%enient, .e .ill use the in%erse demand cur%e, P(X),
A #rice<ta*in) $irm has no control o%er #rice= P(X) = P re)ardless o$ X, o%er some
rele%ant ran)e o$ the $irmGs out#ut, "n contrast, a $irm .ith #o.er o%er #rice can cause
#rice to rise or $all &y decreasin) or increasin) its out#ut= P
(X) < 0 in the rele%ant
ran)e, 5e say that a $irm has Mtechnical mar*et #o.erN i$ it $aces a do.n.ard slo#in)
9rather than hori4ontal: demand cur%e,
"n #ractice almost all $irms ha%e some de)ree o$ technical mar*et #o.er, Althou)h the
notion o$ a #er$ectly com#etiti%e mar*et is e/tremely use$ul as a theoretical construct,
most real<.orld mar*ets de#art at least some.hat $rom this ideal, An im#ortant reason
$or this #henomenon is that mar)inal cost is o$ten &elo. a%era)e cost, most nota&ly
$or #roducts .ith hi)h $i/ed costs and $e. or no ca#acity constraints, such as com#uter
so$t.are, &oo*s, music, and mo%ies, "n such cases, #rice must e/ceed mar)inal cost $or
$irms to remain %ia&le in the lon) run,F Althou)h in theory society could mandate that
all #rices eCual mar)inal cost and #ro%ide su&sidies .here a##ro#riate, this de)ree o$
re)ulation is )enerally re)arded to &e in$easi&le, and in most industries any attem#ts
to do so are &elie%ed to &e in$erior to reliance u#on decentrali4ed mar*et interactions,
Antitrust la. has the more modest &ut, it is ho#ed, achie%a&le o&Eecti%e o$ en$orcin)
com#etition to the e/tent $easi&le, Ai%en the near u&iCuity o$ some de)ree o$ technical
mar*et #o.er, the im#ossi&ility o$ eliminatin) it entirely, and the ine%ita&le costs o$
antitrust inter%ention, the mere $act that a $irm enEoys some technical mar*et #o.er is
not %ery in$ormati%e or use$ul in antitrust la.,
@ ;rior discussions o$ the )eneral relationshi# &et.een the economic conce#tion o$ mar*et #o.er and its
use in antitrust la. include Areeda, Da#lo., and Edlin 92001, ##, 183O1'':, Da#lo. 91'82:, and ?andes and
;osner 91'81:, For a recent o%er%ie., see American Bar Association, ection o$ Antitrust ?a. 9200@:,
0 For sim#licity, unless .e indicate other.ise, .e assume throu)hout this cha#ter that each $irm sells a
sin)le #roduct, 5hile this assum#tion is almost al.ays $alse, in many cases it amounts to loo*in) at a $irmGs
o#erations #roduct<&y<#roduct, !&%iously, a multi<#roduct $irm mi)ht ha%e mar*et #o.er .ith res#ect to one
#roduct &ut not others, 5hen interactions &et.een the di$$erent #roducts sold &y a multi<#roduct $irm are
im#ortant, nota&ly, .hen the $irm sells a line o$ #roducts that are su&stitutes or com#lements $or each other,
the analysis .ill need to &e modi$ied,
F Ed.ard 2ham&erlin 91'33: and Joan 7o&inson 91'33: are classic re$erences $or the idea that $irms in
mar*ets .ith lo. entry &arriers &ut di$$erentiated #roducts ha%e technical mar*et #o.er,
1080 L% 7aplow and C% Shapiro
Bonetheless, the technical, te/t&oo* notion o$ mar*et #o.er has the considera&le ad%anta)e
that it is amena&le to #recise measurement, .hich ma*es it #ossi&le to identi$y
#ractices that enhance a $irmGs #o.er to a su&stantial de)ree, 3he standard measure
o$ a $irmGs technical mar*et #o.er is &ased on the di$$erence &et.een the #rice the
$irm char)es and the $irmGs mar)inal cost, "n the standard theory o$ mono#oly #ricin),
a $irm sets the #rice $or its #roduct to ma/imi4e #ro$its, ;ro$its are )i%en &y
= PX(P) C(X(P)), .here C(X) is the $irmGs cost $unction, 6i$$erentiatin) .ith
res#ect to #rice, .e )et the standard e/#ression )o%ernin) #ricin) &y a sin)le<#roduct
$irm,
91:
P &C
P
= 1
|F | ,
.here &C is the $irmGs mar)inal cost, C
(X), and F dX
dP
P
X is the elasticity o$ demand
$acin) that $irm, the M$irm<s#eci$ic elasticity o$ demand,N8 3he le$t<hand side o$ this
e/#ression is the ?erner "nde/, the #ercenta)e )a# &et.een #rice and mar)inal cost,
.hich is a natural measure o$ a $irmGs technical mar*et #o.er=
m P &C
P
.
As noted earlier, some de)ree o$ technical mar*et #o.er is necessary $or $irms to co%er
their costs in the #resence o$ economies o$ scale, For e/am#le, i$ costs are )i%en &y
C(X) = F + CX, then #ro$its are )i%en &y = PX CX F and the condition that
#ro$its are non<ne)ati%e can &e .ritten as m F/PX, that is, the ?erner "nde/ must &e
at least as lar)e as the ratio o$ the $i/ed costs, F, to the $irmGs re%enues, R PX,
Be$ore #roceedin) .ith our analysis, .e note that, althou)h anticom#etiti%e harm can
come in the $orm o$ reduced #roduct Cuality, retarded inno%ation, or reduced #roduct
%ariety, our discussion .ill $ollo. much o$ the economics literature and most antitrust
analysis in $ocusin) on consumer harm that comes in the $orm o$ hi)her #rices, 3his
limitation is not as serious as may $irst a##ear &ecause hi)her #rices can ser%e as a loose
#ro/y $or other $orms o$ harm to consumers,
#%#% Sin*le3firm pricin* model accountin* for rivals
3o aid understandin), .e #resent a &asic &ut $le/i&le model sho.in) ho. underlyin)
su##ly and demand conditions determine the elasticity o$ demand $acin) a )i%en $irm,
3his model allo.s us to &e)in identi$yin) the $actors that )o%ern the de)ree o$ technical
mar*et #o.er enEoyed &y a $irm, 5e also note that this same model .ill #ro%e %ery
use$ul conce#tually .hen .e e/#lore &elo. the im#act o$ %arious #ractices on #rice,
tudyin) the e$$ects o$ %arious #ractices on #rice reCuires some theory o$ ho. $irms set
8 trictly s#ea*in), the elasticity o$ demand $acin) the $irm is endo)enous, e/ce#t in the s#ecial case o$
constant elasticity o$ demand, since it %aries .ith #rice, an endo)enous %aria&le, All the usual $ormulas re$er
to the elasticity o$ demand at the eCuili&rium 9#ro$it<ma/imi4in): #rice le%el,
Ch% .1- $ntitrust 1081
their #rices, 3he &uildin) &loc* $or these %arious theories is the &asic model o$ #ricesettin)
&y a sin)le, #ro$it<ma/imi4in) $irm, "n addition, as a matter o$ lo)ic, one must
&e)in .ith such a model &e$ore mo%in) on to theories that in%ol%e strate)ic interactions
amon) ri%al $irms,
3he standard model in%ol%es a dominant $irm $acin) a com#etiti%e $rin)e,' A #ro$itma/imi4in)
$irm sets its #rice accountin) $or the res#onses it e/#ects $rom its ri%als and
customers to the #rice it sets,10 3his is a decision<theoretic model, not a )ame<theoretic
model, so it does not ma*e endo)enous the &eha%ior o$ the other $irms in the mar*et
or o$ #otential entrants, 3his is the #rimary sense in .hich the )enerality o$ the model
is limited, 3he model also is limited &ecause it assumes that all $irms in the mar*et
#roduce the same, homo)eneous #roduct and do not en)a)e in any #rice discrimination,
althou)h the core ideas underlyin) it e/tend to models o$ di$$erentiated #roducts,
3he $irm $aces one or more ri%als that, as noted, sell the same, homo)eneous #roduct,
5hen settin) its #rice, P, the $irm reco)ni4es that ri%als .ill li*ely res#ond to hi)her
#rices &y #roducin) more out#ut, 3he com&ined out#ut o$ the $irmGs ri%als increases
.ith #rice accordin) to Y(P), .ith Y
(P ) 0, 3otal 9mar*et: demand declines .ith
#rice accordin) to Z(P), .ith Z
(P ) 0, "$ the $irm in Cuestion sets the #rice P, then
it .ill &e a&le to sell an amount )i%en &y X(P) Z(P) Y(P), 3his is the lar)est
Cuantity that the $irm can sell .ithout dri%in) #rice &elo. the le%el P that it selectedJ i$
the $irm .ants to sell more, it .ill ha%e to lo.er its #rice, 3he $irmGs so<called Mresidual
demand cur%eN is there$ore )i%en &y X(P),
"$ .e di$$erentiate the eCuation de$inin) X(P) .ith res#ect to P, and then multi#ly
&oth sides &y P/X to con%ert the le$t<hand side into elasticity $orm, .e )et
P
X
dX
dP
= P
X
dZ
dP
+ P
X
dY
dP
.
Be/t, multi#ly and di%ide the dZ/dP term on the ri)ht<hand side &y Z and the dY/dP
term &y Y, 3his )i%es
P
X
dX
dP
= P
Z
dZ
dP
Z
X
+ P
Y
dY
dP
Y
X
.
6e$ine the mar*et share o$ the $irm &ein) studied &y S = X/Z, 3he corres#ondin)
mar*et share o$ the ri%als is 1S = Y/Z, 7e#lacin) Z/X &y 1/S and Y/X &y (1S)/S
in the e/#ression a&o%e )i%es
P
X
dX
dP
= P
Z
dZ
dP
1
S
+ P
Y
dY
dP
(1 S)
S
.
' For a recent te/t&oo* treatment o$ this model, see 2arlton and ;erlo$$ 9200@, ##, 110O11':, ?andes and
;osner 91'81: #ro%ide a nice e/#osition o$ this model in the antitrust conte/t,
10 As .ith the standard theory o$ #ure mono#oly #ricin) as tau)ht in microeconomics te/t&oo*s, the results
o$ this model are unchan)ed i$ .e model the $irm as choosin) its out#ut le%el, .ith #rice adEustin) to clear
the mar*et,
1082 L% 7aplow and C% Shapiro
2all the elasticity o$ su##ly o$ the ri%als R P
Y
dY
dP , and the a&solute %alue o$ the
elasticity o$ the underlyin) mar*et demand cur%e |D| P
Z
dZ
dP , 3he a&solute %alue o$
the elasticity o$ demand $acin) the $irm, |F| P
X
dX
dP , is there$ore )i%en &y
|F 92: | =
|D| + (1 S)R
S
.
3his eCuation ca#tures the central lesson $rom this model= the a&solute %alue o$ the
elasticity o$ demand $acin) a sin)le $irm, )i%en the su##ly cur%es o$ its #rice<ta*in)
ri%als and the demand cur%e o$ the &uyers in its mar*et, is )o%erned &y three %aria&les=
91: the underlyin) elasticity o$ demand $or the #roduct, |D|, .hich is $reCuently called
the mar*et elasticity o$ demandJ 92: the elasticity o$ su##ly o$ the $irmGs ri%als, RJ and
93: the $irmGs mar*et share, S, 3he ma)nitude o$ the $irm<s#eci$ic elasticity o$ demand is
lar)er, the lar)er are the ma)nitudes o$ the mar*et elasticity o$ demand and the elasticity
o$ su##ly o$ the $irmGs ri%als and the smaller is the $irmGs mar*et share, "ntuiti%ely,
mar*et share is rele%ant $or t.o reasons= the smaller the $irmGs share, the )reater the
share o$ its ri%als and thus the )reater is the a&solute ma)nitude o$ their su##ly res#onse
to a #rice increase $or a )i%en su##ly elasticity, RJ and the smaller the $irmGs share, the
smaller is its share o$ the increase in industry #ro$its due to a )i%en sacri$ice in its o.n
sales,11
!ne #olar case in this &asic model is that o$ the traditional mono#olist, 5ith no ri%als,
S = 1, so the elasticity o$ demand $acin) the $irm is Eust the mar*et elasticity o$
demand, 5ith ri%als, ho.e%er, the ma)nitude o$ the $irm<s#eci$ic elasticity o$ demand
is lar)er than that o$ the mar*et elasticity o$ demand, 3he other #olar case is that o$ the
$irm $rom the theory o$ #er$ectly com#etiti%e mar*ets, As the $irmGs share o$ the mar*et
a##roaches 4ero, the ma)nitude o$ the $irm<s#eci$ic elasticity o$ demand &ecomes
in$inite, that is, the $irm is a #rice<ta*er,
5e can directly translate the $irm<s#eci$ic elasticity o$ demand )i%en &y e/#ression
92: into the #ro$it<ma/imi4in) #rice, As indicated in e/#ression 91:, #ro$it ma/imi4ation
in%ol%es settin) #rice so that the $irmGs )ross mar)in, m, eCuals the in%erse o$
the ma)nitude o$ the $irmGs elasticity o$ demand, "$ there are no ri%als, S = 1 and this
relationshi# sim#li$ies to the standard mono#oly $ormula, m = 1/|D|, For a $irm .ith
a tiny mar*et share, |F | is enormous, so m 0, that is, #rice nearly eCuals mar)inal
cost, For intermediate cases, as noted, in this model a lar)e mar*et elasticity o$ demand,
|D|, a hi)h elasticity o$ ri%al su##ly, R, and a small mar*et share, S, all lead to a lar)e
$irm<s#eci$ic elasticity o$ demand $acin) the #rice leader, |F |, .hich in turn im#lies a
small mar)in,
3his model #ro%ides a )uide $or studyin) the ty#es o$ conduct that may enhance
a $irmGs technical mar*et #o.er and thus allo. that $irm #ro$ita&ly to raise its #rice,
11 "t should &e noted that statements a&out the e$$ect o$ mar*et share must &e inter#reted care$ully, 3hus, an
out.ard shi$t in the su##ly cur%e o$ the ri%als, .hich lo.ers the $irmGs mar*et share at any )i%en #rice, .ill
raise the elasticity o$ demand $acin) that $irm at any )i%en #rice, -o.e%er, more &roadly, the $irmGs mar*et
share is endo)enous &ecause it de#ends on the #rice the $irm chooses,
Ch% .1- $ntitrust 1083
Aenerically, such conduct .ill &e that .hich reduces the %alue o$ the ri)ht side o$ e/#ression
92:= conduct that ma*es su&stitute #roducts less attracti%e, that causes ri%als
to reduce their su##ly, and that raises the $irmGs mar*et share 9throu)h the t.o $ormer
means or other.ise:, ?ater .e consider ho. certain ty#es o$ conduct ha%in) these e$$ects
should &e scrutini4ed under antitrust la.,
3his model is Cuite &road .hen one underta*es a##ro#riate inter#retations and e/tensions,
For e/am#le, issues relatin) to su&stitute #roducts &ear on the mar*et elasticity
o$ demand, as .ill &e noted &elo., Additionally, one can account $or entry &y re$lectin)
it in the ri%al su##ly elasticity, !ne #articular %ariant o$ the model in%ol%es in$initely
elastic ri%al su##ly, #erha#s due to entry, at some $i/ed MlimitN #rice,
#%4% &ultiple3firm models
3he model in su&section 2,2 too* the &eha%ior o$ all &ut one $irm as e/o)enous, "n this
section, .e consider )ame<theoretic models that ma*e #redictions re)ardin) the de)ree
o$ mar*et #o.er e/ercised &y interactin) $irms, First .e consider t.o standard, static,
noncoo#erati%e models= 2ournotGs model o$ oli)o#oly, $or the case .ith homo)eneous
#roducts, and BertrandGs model, $or the case .ith di$$erentiated #roducts, 3hen .e consider
&rie$ly the #ossi&ility o$ re#eated )ames and the im#act o$ collusi%e &eha%ior on
mar*et #o.er,12
#%4%.% Cournot model with homo*eneous products
3he 2ournot 91838: model o$ oli)o#oly .ith homo)eneous #roducts is similar to the
sin)le<$irm #ricin) model in that it identi$ies ho. certain o&ser%a&le characteristics o$
the mar*et determine the de)ree o$ a $irmGs mar*et #o.er, that is, the #ercenta)e mar*u#
a&o%e mar)inal cost that the $irm char)es, 3he 2ournot model )oes $urther, ho.e%er, &y
#ro%idin) #redictions a&out ho. mar*et structure a$$ects the eCuili&rium #rice, #redictions
that .ill &e im#ortant $or seein) ho. certain commercial #ractices and mer)ers
a$$ect #rice, #eci$ically, the model #redicts that $irms .ith lo.er costs .ill ha%e hi)her
mar*et shares and hi)her mar*u#s, 3he model is $reCuently em#loyed in mar*ets .ith
relati%ely homo)eneous #roducts, es#ecially i$ $irms #ic* their out#ut or ca#acity le%els,
a$ter .hich #rices are determined such that the resultin) su##ly eCuals demand,13 -o.e%er,
one should &ear in mind that the 2ournot eCuili&rium is the Bash eCuili&rium in a
one<shot )ame, As .e discuss at len)th in section 3, many di$$erent outcomes can arise
as eCuili&ria in a re#eated oli)o#oly )ame, e%en i$ the sta)e )ame #layed each #eriod
in%ol%es Cuantity<settin) P la 2ournot, "n antitrust a##lications, it is )enerally desira&le
12 3here is an enormous literature on oli)o#oly theory, .hich .e do not attem#t to co%er systematically, ee,
$or e/am#le, ha#iro 91'8':, 3irole 91'88:, and Qi%es 92001:, 5e discuss models o$ re#eated oli)o#oly at
)reater len)th in section 3 on collusion,
13 Dre#s and chein*man 91'83: use a #articular rationin) rule to sho. that ca#acity choices $ollo.ed &y
#ricin) com#etition can re#licate the 2ournot eCuili&rium,
1081 L% 7aplow and C% Shapiro
to test ro&ustness o$ results to alternati%e solution conce#ts as .ell as to test em#irically
the #redictions o$ any oli)o#oly model that is em#loyed,
"n a 2ournot eCuili&rium, a sin)le $irmGs reaction cur%e is deri%ed as a s#ecial case
o$ the &asic model o$ sin)le<$irm #ricin)= the ri%alsG out#uts are all ta*en to &e $i/ed,
so the ri%al su##ly elasticity is 4ero, As .e no. sho., the elasticity o$ demand $acin)
a sin)le $irm is eCual to the mar*et elasticity o$ demand di%ided &y that $irmGs mar*et
share, -o.e%er, the 2ournot model )oes &eyond the sin)le<$irm #ricin) model &ecause
it in%ol%es $indin) the eCuili&rium in a )ame amon) multi#le $irms,
u##ose that there are N $irms, .ith each $irm i choosin) its out#ut Xi simultaneously,
3he 2ournot eCuili&rium is a Bash eCuili&rium in these Cuantities, 3otal
out#ut is X X1 + + XN, "ndustry or mar*et 9in%erse: demand is )i%en &y
P = P(X), Ai%en the out#ut o$ the other $irms, $irm i chooses its out#ut to ma/imi4e
its o.n #ro$its, i = P(X)Xi Ci(Xi ), 3he $irst<order condition $or this $irm is
P(X) + XiP
(X) C
i(Xi ) = 0, 3his can &e .ritten as
93:
P &Ci
P
= Si
|D| ,
.here Si Xi/X is $irm iGs mar*et share, and D, as &e$ore, is the mar*et elasticity o$
demand,
3o e/#lore this result, consider the s#ecial case in .hich each $irm i has constant
mar)inal cost &Ci , Addin) u# the $irst<order conditions $or all o$ the $irms )i%es
NP(X) + XP
(X) = i &Ci , .hich tells us that total out#ut and hence the eCuili&rium
#rice de#end only u#on the sum o$ the $irmsG mar)inal costs, Moreo%er, the mar*u#
eCuation tells us that lo.er<cost $irms ha%e hi)her mar*et shares and enEoy more technical
mar*et #o.er, At the same time, the lar)er is the mar*et elasticity o$ demand $or
this homo)eneous #roduct, the smaller is the mar*et #o.er enEoyed &y each $irm and the
lo.er are the mar)ins at all $irms, -ere .e see a recurrent theme in antitrust= a lo.ercost
$irm may .ell enEoy some technical mar*et #o.er and ca#ture a lar)e share o$ the
mar*et, &ut this is not necessarily ine$$icient, "ndeed, .ith constant mar)inal costs, $ull
#roducti%e e$$iciency .ould call $or the $irm .ith the lo.est mar)inal cost to ser%e the
entire mar*et,
3he 2ournot model also #redicts that total out#ut .ill &e less than .ould &e e$$icient
&ecause none o$ the $irms #roduces u# to the #oint at .hich mar)inal cost
eCuals #riceJ they all ha%e some de)ree o$ mar*et #o.er, "n the s#ecial case .ith constant
and eCual mar)inal costs, each $irm has a mar*et share o$ 1/N, and the model
#redicts that each enEoys technical mar*et #o.er accordin) to the resultin) eCuation
(P &C)/P = 1/N|D|, "n this sim#le sense, more $irms leads to )reater com#etition
and lo.er #rices, -o.e%er, this model is clearly incom#lete $or antitrust #ur#oses=
#resuma&ly, there are $i/ed costs to &e co%ered 9.hich is .hy there is a $i/ed num&er
o$ $irms in the $irst #lace:, so addin) more $irms is not costless,11 3his ty#e o$ analy<
11 "n )eneral, there is no reason to &elie%e that the eCuili&rium num&er o$ $irms in an oli)o#oly .ith $ree
entry, that is, .here eCually e$$icient $irms enter until $urther entry .ould dri%e #ro$its &elo. 4ero, is socially
Ch% .1- $ntitrust 108@
sis .ill &e directly rele%ant .hen .e consider hori4ontal mer)ers, .hich remo%e an
inde#endent com#etitor &ut may also lead to e$$iciencies o$ %arious ty#es,
!ne o$ the attracti%e theoretical $eatures o$ the 2ournot model is that it )enerates
an ele)ant $ormula $or the industry<.ide a%era)e, out#ut<.ei)hted, #rice<cost
mar)in, that is, the e/#ression 'C& N
i=1 Si
P&Ci
P , Usin) eCuation 93:, .e )et
'C& N
i=1 Si
Si
|D| or
'C& = 91: 1
|D|
N
i=1
S2
i
= H
|D| ,
.here H S2
i is the -er$indahl<-irschman "nde/ 9--": o$ mar*et concentration
that is commonly used in antitrust analysis, es#ecially o$ hori4ontal mer)ers,
#%4%#% Bertrand model with differentiated products
3he Bertrand model .ith di$$erentiated #roducts is the other *ey static model o$
oli)o#oly used in antitrust, 3he Bertrand eCuili&rium is the Bash eCuili&rium in the
)ame in .hich the $irms simultaneously set their #rices, 5ith N $irms sellin) di$$erentiated
#roducts, .e can .rite the demand $or $irm iGs #roduct as Xi = Di(P1, . . . , PN),
As usual, the #ro$its o$ $irm i are )i%en &y i = PiXi Ci(Xi ), 3he Bertrand eCuili&rium
is de$ined &y the N eCuations i/ P i = 0, 5ritin) the elasticity o$ demand
$acin) $irm i as i X i
P i
Pi
Xi
, $irmiGs $irst<order condition is the usual mar*u# eCuation,
Pi &Ci
Pi
= 1
|i | .
Actually sol%in) $or the Bertrand eCuili&rium can &e di$$icult, de#endin) on the $unctional
$orm $or the demand system and on the $irmsG cost $unctions, "n )eneral, ho.e%er,
.e *no. that a $irm $aces hi)hly elastic demand i$ its ri%als o$$er %ery close su&stitutes,
so the Bertrand theory #redicts lar)er mar*u#s .hen the #roducts o$$ered &y the %arious
$irms are more hi)hly di$$erentiated, "n #ractice, nota&ly, in the assessment o$ mer)ers,
#articular models o$ #roduct di$$erentiation are used, such as discrete choice models
.ith random utilities, includin) lo)it and nested lo)it models, or models .ith linear demand
or constant elasticities, as .e discuss $urther in section 1 on hori4ontal mer)ers,
-ere .e illustrate the o#eration o$ the Bertrand model &y e/#licitly sol%in) a sim#le,
symmetric, t.o<$irm model .ith constant mar)inal costs and linear demand, 5rite the
demand cur%es as X1 = AP1+P2 and X2 = AP2+P1, Bote that the #arameter
measures the di%ersion ratio, that is, the $raction o$ sales lost &y one $irm, .hen it
e$$icient, ee, $or e/am#le, Man*i. and 5hinston 91'80:, 3his o&ser%ation is rele%ant in assessin) certain
antitrust #olicies= i$ the eCuili&rium num&er o$ $irms is MnaturallyN too small, then e/clusionary conduct on
the #art o$ the incum&ent oli)o#olists creates an additional social ine$$iciency, -o.e%er, i$ the eCuili&rium
num&er o$ $irms is MnaturallyN e/cessi%e, di$$erent im#lications .ould $ollo.,
1080 L% 7aplow and C% Shapiro
raises its #rice, that are ca#tured &y the other $irm 9assumin) that the other $irmGs #rice
is $i/ed:, 3he di%ersion ratio, , .ill &e im#ortant .hen .e study hori4ontal mer)ers
&elo.,1@
2all the mar)inal costs #er unit &C1 and &C2, res#ecti%ely, and assume that there
are no $i/ed costs, 3hen .e ha%e 1 = (P1 &C1)(A P1 + P2), 6i$$erentiatin)
.ith res#ect to P1 and settin) this eCual to 4ero, .e )et $irm 1Gs &est<res#onse cur%e,
P1 = (A + P2 + &C1)/2, Assumin) cost symmetry as .ell, &C = &C1 = &C2,
in the symmetric Bertrand eCuili&rium .e must ha%e P1 = P2 = PB so .e )et
PB = A+&C
2 ,
5e can com#are the Bertrand eCuili&rium #rice to the #rice char)ed &y a sin)le $irm
controllin) &oth #roducts, uch a $irm .ould set P to ma/imi4e (P&C)(AP+P),
.hich )i%es the mono#oly #rice o$ PM = A+&C(1)
2(1) , 3he #ercenta)e )a# &et.een
the mono#oly #rice and the Bertrand #rice is )i%en &y PMPB
PB
=
2(1)
PB&C
PB
,10 3his
e/#ression tells us that the Bertrand eCuili&rium #rice is relati%ely close to the mono#oly
#rice .hen the t.o #roducts are rather #oor su&stitutes, that is, .hen the di%ersion ratio,
, is lo.,
3his $ormula .ill &e hi)hly rele%ant .hen studyin) the e$$ect on #rice o$ a mer)er
&et.een t.o su##liers o$ di$$erentiated #roducts, "n that conte/t, the $ormula measures
the #rice increase associated .ith the mer)er, )i%en the #rices char)ed &y other $irms
9and &e$ore accountin) $or e$$iciencies:, 3he #rice increase .ill de#end on the #remer)er
mar)in, PB&C
PB
, and on the di%ersion ratio,
#%4%4% +ther *ame3theoretic models and collusion
Both the 2ournot and Bertrand models assume that $irms en)a)e in a one<shot noncoo#erati%e
)ame, An e/tensi%e literature on re#eated )ames e/#lores the #ossi&ility that
$irms may do &etter $or themsel%es, su##ortin) .hat are more colloCuially descri&ed
as collusi%e outcomes, a##roachin) or eCualin) the industry #ro$it<ma/imi4in) #rice,
As su))ested &y ti)ler 91'01: and re$ined in su&seCuent .or*, hi)her #rices tend to
&e sustaina&le .hen cheatin) can &e ra#idly detected and e$$ecti%ely #unished, For a
)eneral discussion o$ models o$ collusion, see JacCuemin and lade 91'8': and ha#iro
91'8':,
3he #ossi&ility that $irms can su##ort alternati%e eCuili&ria $eaturin) hi)her #rices is
im#ortant to antitrust analysis, First, it su))ests that mar*et #o.er may &e hi)her than is
other.ise a##arent, econd and more im#ortant, the #ossi&ility o$ collusion a$$ects the
antitrust analysis o$ other &usiness conduct, For e/am#le, a hori4ontal mer)er may ha%e
1@ More )enerally, the di%ersion ratio $rom #roduct i to su&stitute #roduct j is de$ined as
ji
= (dXj /dPi )/(dXi/dPi ), 2on%ertin) this eCuation into elasticity $orm )i%es ji
= ji
|i
|
Xj
Xi
, .here
ji
= dXj
dPi
Pi
Xj
is the cross<elasticity $rom #roduct i to #roduct j ,
10 3he details o$ these calculations are a%aila&le at htt#=++$aculty,haas,&er*eley,edu+sha#iro+unilateral,#d$,
Ch% .1- $ntitrust 108F
only a minor im#act on #rice i$ the mer)in) $irms and their ri%als are already colludin),
&ut a $ar )reater e$$ect i$ the reduction in the num&er o$ com#etitors ma*es collusion
easier to sustain, Also, some #ractices may $acilitate collusion, in .hich case such #ractices
themsel%es should #otentially &e su&Eect to antitrust scrutiny, 3hese #ossi&ilities
are e/#lored $urther in section 3 on collusion and section 1 on hori4ontal mer)ers,
#%2% &eans of inferrin* maret power
Assessin) the e/tent o$ or increase in technical mar*et #o.er in a )i%en situation is o$ten
a di$$icult underta*in), Based u#on the $ore)oin) analysis, one can identi$y a num&er
o$ #otential strate)ies .hose use$ulness %aries )reatly &y conte/t, 3he le)al system has
tended to rely #rimarily on a su&set o$ these a##roaches, $ocusin) mostly on mar*et
de$inition, as discussed &elo., "n recent years, ho.e%er, it has increasin)ly considered
alternati%es .hen it has #ercei%ed that credi&le economic e%idence has &een o$$ered,1F
Althou)h some.hat crude, it is hel#$ul to )rou# means o$ in$errin) mar*et #o.er
into three cate)ories, First, since mar*et #o.er is technically de$ined &y the e/tent o$
the #rice<cost mar)in, one can attem#t to identi$y e%idence that &ears $airly directly on
the si4e o$ this mar)in, or to measure #ro$its 9.hich re$lect the mar)in &et.een #rice
and avera*e cost:, econd, %arious models, such as the sin)le<$irm #rice<settin) model
in su&section 2,2, indicate that the e/tent o$ mar*et #o.er .ill &e a $unction o$ the
elasticity o$ demand, a $irmGs mar*et share, and ri%alsG su##ly res#onse, Accordin)ly,
one can analy4e in$ormation indicati%e o$ the ma)nitude o$ these $actors, 3hird, one can
ma*e in$erences $rom $irm &eha%ior, nota&ly .hen o&ser%ed actions .ould &e irrational
unless a certain de)ree o$ mar*et #o.er e/isted or .as there&y con$erred,
#%2%.% 'rice3cost mar*in
#%2%.%.% Direct measurement !&ser%in) the e/tent to .hich #rice is a&o%e mar)inal
cost indicates the de)ree o$ technical mar*et #o.er, 3his direct a##roach is $easi&le i$
one can accurately measure #rice and some %ersion o$ mar)inal cost, usually a%era)e
incremental cost,18 ;rice is o$ten easy to identi$y, althou)h com#lications may arise
.hen multi#le #roducts are sold to)ether, ma*in) it di$$icult to determine the incremental
re%enue associated .ith the #roduct in Cuestion, "$ di$$erent customers are char)ed
1F For e/am#le, the u#reme 2ourt in Federal 3rade 2ommission %, "ndiana Federation o$ 6entists, 1F0 U,,
11F, 100O101 91'80: 9Cuotin) AreedaGs $ntitrust Law treatise: stated= Mince the #ur#ose o$ the inCuiries into
mar*et de$inition and mar*et #o.er is to determine .hether an arran)ement has the #otential $or )enuine
ad%erse e$$ects on com#etition, R#roo$ o$ actual detrimental e$$ects, such as a reduction o$ out#ut,G can o&%iate
the need $or an inCuiry into mar*et #o.er, .hich is &ut a Rsurro)ate $or detrimental e$$ects,G N
18 5e use the terms Mmar)inal costN and Ma%era)e incremental costN interchan)ea&ly, Both measure the e/tra
cost #er unit associated .ith increased out#ut, A%era)e incremental cost is a some.hat more accurate term,
since one is o$ten interested in increments that do not corres#ond to Mone unitN o$ out#ut, -o.e%er, i$ one
ta*es a $le/i&le a##roach to .hat constitutes a MunitN o$ #roduction, the t.o terms are e/actly the same, "n
#ractice, a%era)e incremental cost is used to determine )ross #ro$it mar)ins,
1088 L% 7aplow and C% Shapiro
di$$erent #rices, it may &e necessary to calculate the #ro$it mar)ins $or sales to di$$erent
customers 9or at di$$erent #oints o$ time:, 2om#le/ities also arise .hen some sales
im#licitly &undle other ser%ices, such as deli%ery, short<term $inancin), and customer
su##ortJ in #rinci#le, these $actors can &e accommodated &y rede$inin) the #roduct to
include these ser%ices 9and trac*in) the costs associated .ith these ser%ices:, Mar)inal
cost, &y contrast, may &e more di$$icult to measure, due &oth to di$$iculties in identi$yin)
.hich costs are %aria&le 9and o%er .hat time #eriod: and to the #resence o$ common
costs that may &e di$$icult to allocate a##ro#riately, "n #art $or this reason, the em#irical
industrial or)ani4ation literature, sur%eyed in Bresnahan 91'8':, o$ten treats mar)inal
cost as uno&ser%a&le,
"n some cases, a##ro/imate measures o$ #rice<cost mar)ins may &e su$$icient and
easy to #roduce, &ut as e%idenced &y dis#utes o%er cost in #redatory #ricin) cases and
in %arious re)ulatory conte/ts, direct measurement o$ any conce#tion o$ cost can &e
di$$icult and contentious, "n any e%ent, as .ith all measures o$ technical mar*et #o.er,
it is im#ortant to *ee# in mind the distinction &et.een the e/tent o$ mar*et #o.er and
.hether #articular conduct should )i%e rise to antitrust lia&ility, For e/am#le, as .e
ha%e already noted, es#ecially in industries in .hich mar)inal cost is &elo. a%era)e
cost and ca#acity constraints are not &indin), nontri%ial technical mar*et #o.er may &e
consistent .ith .hat are normally considered com#etiti%e industries,
#%2%.%#% 'rice comparisons Another $airly direct route to assessin) the ma)nitude o$
#rice<cost mar)ins, or at least to #ro%ide a lo.er<&ound estimate, is to com#are #rices
across mar*ets, For e/am#le, i$ a $irm sells its #roduct $or a su&stantially hi)her #rice
in one re)ion than in another 9ta*in) into account trans#ortation and other cost di$$erences:,
the #rice<cost mar)in in the hi)h<#rice re)ion should &e at least as )reat as the
9adEusted: #rice di$$erence &et.een the re)ions, 3his in$erence #resumes, o$ course, that
the #rice in the lo.<#rice re)ion is at least eCual to mar)inal cost, Bote that this method
can &e understood as a s#ecial case o$ direct measurement, "t is assumed that the lo.
#rice is a #ro/y $or 9at least an u##er &ound on: mar)inal cost, and one then is measurin)
the #rice<cost mar)in directly,
3he Staples mer)er case illustrates an a##lication o$ this method,1' 3he )o%ernment
o$$ered 9and the court .as con%inced &y: data indicatin) that #rices .ere hi)her in
re)ional mar*ets in .hich $e.er o$$ice su##ly su#erstores o#erated and that #rices $ell
.hen ne. su#erstore chains entered, 3his .as ta*en as #o.er$ul e%idence that a mer)er
o$ t.o o$ the e/istin) three su#erstores .ould lead to #rice increases,
#%2%.%4% 'rice discrimination ;rice com#arisons o$ten in%ol%e a s#ecial case o$ #rice
discrimination, .herein a )i%en $irm char)es di$$erent #rices to di$$erent consumers,
contrary to the im#licit assum#tion in the earlier analysis that each $irm sets a sin)le
1' Federal 3rade 2ommission %, ta#les, "nc,, 'F0 F, u##, 1000 96,6,2, 1''F:, For $urther discussion, see
su&section 1,0,1 in our discussion o$ hori4ontal mer)ers,
Ch% .1- $ntitrust 108'
#rice $or all o$ its customers, Accordin)ly, $or essentially the same reason as that Eust
)i%en, the a&ility o$ a $irm to en)a)e in #rice discrimination im#lies the e/istence o$
mar*et #o.er, "$ one is #re#ared to assume that the $irm is not #ricin) &elo. mar)inal
cost to any o$ its customers, and i$ one accounts $or di$$erences in the cost o$ ser%in)
di$$erent customers, the #ercenta)e di$$erence &et.een any hi)h #rice it char)es and
the lo.est #rice it char)es $or the same #roduct can ser%e as a lo.er &ound on the
#ercenta)e mar*u# associated .ith the hi)her #rice, For e/am#le, the su&stantial #rice
discrimination in sales o$ #harmaceutical dru)s on international mar*ets sho.s that
#rices in the United tates are %ery much a&o%e mar)inal cost,
3he $act that #rice discrimination technically im#lies mar*et #o.er is im#ortant &ecause
#rice discrimination is .ides#read, Familiar e/am#les include airline #ricin),
senior citi4en and student discounts, and the mundane #ractice o$ restaurants char)in)
stee# #rice increments $or alcoholic &e%era)es 9%ersus so$t drin*s: and hi)h<end entrees
that )reatly e/ceed any di$$erences in mar)inal cost, For &usiness<to<&usiness transactions,
ne)otiations that ty#ically )enerate #rice dis#ersion and #rice discrimination are
Cuite common,
!nce a)ain, ho.e%er, it is im#ortant to *ee# in mind that the e/istence o$ technical
mar*et #o.er does not im#ly antitrust lia&ility,20 As is $amiliar, #rice discrimination
)enerates )reater seller #ro$its yet may .ell &e &eni)n or e%en $a%ora&le on a%era)e $or
consumers, Moreo%er, the resultin) #ro$it mar)ins are o$ten necessary to co%er $i/ed
costs, as in models o$ mono#olistic com#etition, "$ there are no &arriers to entry so that
the resultin) mar)ins merely #ro%ide a normal rate o$ return on ca#ital, the #resence
o$ a )a# &et.een #rice and mar)inal cost is #er$ectly consistent .ith the conclusion
that the mar*et is &eha%in) is a com#etiti%e $ashion, )i%en the #resence o$ $i/ed costs
and #roduct di$$erentiation, Furthermore, in our #recedin) e/am#le o$ multinational
#harmaceutical com#anies, the mar)ins #ro%ide the re.ard $or costly and ris*y research
and de%elo#ment to create and #atent ne. dru)s, 3he e/ #ost mar*et #o.er is necessary
to #ro%ide the Cuasi<rents that induce inno%ation 9)i%en that .e rely on a #atent system
rather than a re)ime that )i%es direct re.ards to inno%ators $rom the )o%ernment $isc:,
#%2%.%2% 'ersistent profits A some.hat di$$erent a##roach to esta&lishin) antitrust
mar*et #o.er in%ol%es loo*in) at a $irmGs #ro$its, .hich amounts to com#arin) #rice
to a%era)e 9rather than mar)inal: cost, Under this a##roach, #ersistently a&o%e<normal
20 Bor is it the case that #rice discrimination in itsel$ im#lies antitrust lia&ility, des#ite the e/istence o$ the
7o&inson<;atman Act that re)ulates #articular sorts o$ #rice discrimination in certain conte/ts, As #resently
inter#reted, #rice discrimination may &e a %iolation in so<called #rimary<line cases, tantamount to #redatory
#ricin), and in secondary<line cases, such as .hen manu$acturers o$$er discounts 9that are not cost Eusti$ied:
to lar)e retailers that are not a%aila&le to smaller &uyers, Bota&ly, the Act does not co%er discriminatory #rices
to ultimate consumers 9or to intermediaries that are not in com#etition .ith each other: that are non#redatory,
Be%ertheless, it seems that de$endants in antitrust liti)ation ha%e &een reluctant to rationali4e challen)ed #ractices
that analysts ha%e su))ested .ere means o$ #rice discrimination on such )rounds, #resuma&ly $earin)
that such e/#lanations .ould &e to their detriment, !$ course, one .ay this could &e true is that the e/istence
o$ some technical mar*et #o.er .ould there&y &e conceded,
10'0 L% 7aplow and C% Shapiro
#ro$its indicate a hi)h #rice<cost mar)in and thus the e/istence o$ technical mar*et
#o.er, 3his method shares di$$iculties .ith any that rely on measures o$ cost, "n #articular,
it is o$ten %ery hard to measure the return on ca#ital earned $or a )i%en #roduct, or
in a )i%en mar*et, es#ecially $or a $irm that is en)a)ed in many lines o$ &usiness and
has su&stantial costs that are common across #roducts,21 Another #ro&lem .ith this a##roach
is that the return on ca#ital should, in #rinci#le, &e adEusted $or ris*, FreCuently,
one is loo*in) at a success$ul $irm, #erha#s one that has &een hi)hly #ro$ita&le $or many
years $ollo.in) some initial inno%ation that, e: ante, may not ha%e turned out as .ell,
"n addition, a%era)e costs o$ten di$$er $rom mar)inal costs, 5hen a%era)e costs are
hi)her, this a##roach may mas* the e/istence o$ technical mar*et #o.er, "n such circumstances,
ho.e%er, mar)inal<cost #ricin) may &e unsustaina&le in any e%entJ that is,
althou)h there may &e technical mar*et #o.er, there may not &e any .ay 9short o$ intrusi%e
re)ulation that is not contem#lated: to im#ro%e the situation, 5hen a%era)e cost
is &elo. mar)inal cost, #ro$its can e/ist des#ite the a&sence o$ any mar*u#, "n such
cases, entry mi)ht &e e/#ected, "$ #ro$its are ne%ertheless #ersistent, there may e/ist
entry &arriers, a su&Eect .e discuss &elo.,
#%2%#% ;irm<s elasticity of demand
"n the sin)le<$irm #ricin) model, the #rice<cost mar)in 9?erner "nde/: eCuals the in%erse
o$ the 9a&solute %alue o$ the: $irmGs elasticity o$ demand, as indicated &y e/#ression 91:,
Furthermore, as descri&ed in e/#ression 92:, this elasticity de#ends on the mar*et elasticity
o$ demand, the $irmGs mar*et share, and ri%alsG su##ly elasticity, "n the 2ournot,
Bertrand, and other oli)o#oly models, many o$ the same $actors &ear on the e/tent o$
the #rice<cost mar)in and thus the de)ree o$ mar*et #o.er, Accordin)ly, another route
to in$errin) mar*et #o.er is to consider the ma)nitude o$ these $actors,
#%2%#%.% Direct measurement !ne could attem#t to measure the elasticity o$ demand
$acin) the $irm in Cuestion,22 A #ossi&le a##roach .ould &e to estimate the mar*et
elasticity o$ demand and then ma*e an adEustment &ased on the $irmGs mar*et share,
Alternati%ely, one mi)ht directly o&ser%e ho. the $irmGs sales ha%e %aried .hen it has
chan)ed its #rice, As a #ractical matter, &oth o$ these methods may &e di$$icult to im#lement,
-o.e%er, they may ne%ertheless &e more relia&le than the alternati%es,
#%2%#%#% Su!stitutes, maret definition, and maret share "n antitrust analysis, &oth &y
a)encies 9nota&ly, in e/aminin) #ros#ecti%e hori4ontal mer)ers: and &y the courts, the
dominant method o$ )au)in) the e/tent o$ mar*et #o.er in%ol%es de$inin) a so<called
rele%ant mar*et and e/aminin) the share o$ a $irm or )rou# o$ $irms in that mar*et, "n
de$inin) #roduct mar*ets, the $ocus is on .hich #roducts are su$$iciently )ood demand
21 ee, $or e/am#le, Fisher and McAo.an 91'83:,
22 ee, $or e/am#le, Ba*er and Bresnahan 91'88:,
Ch% .1- $ntitrust 10'1
su&stitutes $or the #roduct in Cuestion to &e deemed in the same mar*et, ?i*e.ise, in
de$inin) the e/tent o$ the )eo)ra#hic mar*et, the Cuestion concerns the $easi&ility o$
su&stitution, $or e/am#le, &y as*in) ho. $ar #atients .ould tra%el $or hos#itali4ation,
Althou)h .e ha%e discussed the economic analysis o$ mar*et #o.er at some len)th, the
conce#t o$ mar*et de$inition has not yet a##eared directly, -ence it is use$ul to consider
the relationshi# &et.een the most common method used in antitrust la. to assess mar*et
#o.er and the im#lications o$ the $ore)oin) economic analysis,
3he connection is easiest to see &y e/aminin) e/#ression 92:, .hich relates the $irms#eci$ic
elasticity o$ demand to the mar*et elasticity o$ demand, the $irmGs mar*et share,
and ri%alsG elasticity o$ su&stitution, 2onsider the case in .hich the $irm #roduces a homo)eneous
#roduct, has a hi)h share o$ sales o$ that #roduct, and $aces a hi)hly elastic
mar*et demand cur%e due to the e/istence o$ many close su&stitutes, 3he $irm<s#eci$ic
elasticity o$ demand is hi)h and thus the e/tent o$ technical mar*et #o.er is small e%en
thou)h the $irmGs mar*et share is hi)h in the narro.ly de$ined mar*et consistin) only o$
the homo)eneous #roduct sold &y the $irm, !ne could rede$ine the Mmar*etN to include
the close su&stitutes alon) .ith the homo)eneous #roduct sold &y the $irm, 3he mar*et
elasticity o$ demand in this &roader mar*et is #resuma&ly smaller, &ut since the $irmGs
mar*et share in this mar*et is also necessarily lo.er, .e .ould a)ain conclude that
the $irm<s#eci$ic demand elasticity is lar)e and thus that the de)ree o$ technical mar*et
#o.er is lo.,
2ourtsKand thus la.yers and )o%ernment a)enciesKtraditionally eCuate hi)h mar*et
shares .ith a hi)h de)ree o$ mar*et #o.er and lo. shares .ith a lo. de)ree o$
mar*et #o.er, 3his association is hi)hly misleadin) i$ the mar*et elasticity o$ demand
is i)nored, and li*e.ise i$ ri%alsG elasticity o$ su##ly is not considered, "n #rinci#le,
as Eust e/#lained, the #aradi)m &ased on mar*et de$inition and mar*et share ta*es the
mar*et elasticity o$ demand into account, indirectly, &y de$inin) &roader mar*etsKand
thus #roducin) lo.er mar*et sharesK.hen the elasticity is hi)h, As should &e a##arent
$rom the $ore)oin) discussion, the standard antitrust a##roach is more indirect than necessary
and, due to this $act #lus its dichotomous structure 9su&stitutes are either in the
mar*et or not:, .ill tend to #roduce needlessly noisy conclusions,23 5e discuss mar*et
de$inition at )reater len)th in su&section 1,@ on hori4ontal mer)ers and su&section @,2,1
on mono#oli4ation,
FreCuently, it is use$ul to decom#ose the elasticity o$ demand $or a )i%en #roduct
into %arious cross<elasticities o$ demand .ith other #roducts, For e/am#le, i$ the #rice
o$ soda rises, consumers .ill su&stitute to other drin*s, includin), #erha#s, &eer, Euice,
mil*, and .ater, Baturally, the analysis in any )i%en case .ill de#end u#on e/actly
ho. these %arious #roducts are de$ined 9soda could &e &ro*en into re)ular soda and
diet soda, or colas and non<colas, etc,:, But the underlyin) theory o$ demand does not
%ary .ith such de$initions, 3o illustrate, su##ose that consumers allocate their total income
o$ I across N distinct #roducts, so N
i=1 PiXi = I , 3o study the elasticity o$
23 3his #oint is ela&orated in Da#lo. 91'82:,
10'2 L% 7aplow and C% Shapiro
demand $or #roduct 1, su##ose that P1 rises and the other #rices remain unchan)ed,
3hen .e )et X1 + P1
dX1
dP1
+N
i=2 Pi
dXi
dP1
= 0, 2on%ertin) this to elasticity $orm )i%es
P1
X1
dX1
dP1
= 1 +N
i=2
P1
Xi
dXi
dP1
PiXi
P1X1
, 6e$inin) the cross<elasticity &et.een #roduct i and
#roduct 1 as i1 = dXi
dP1
P1
Xi
, and the re%enues associated .ith #roduct i as Ri = PiXi ,
this can &e .ritten as
|11| = 1 + 9@:
N
i=2
i1
Ri
R1
.
"n .ords, the 9a&solute %alue o$ the: elasticity o$ demand $or #roduct 1 is eCual to
one #lus the sum o$ the cross<elasticities o$ all the other #roducts .ith #roduct 1, .ith
each cross<elasticity .ei)hted &y the associated #roductGs re%enues relati%e to those
o$ #roduct 1, "$ .e de$ine each #roductGs share o$ e/#enditures as si = Ri/I , then
e/#ression 9@: can &e .ritten as |11| = 1 + 1
s1 N
i=2 sii1, so the cross<elasticity .ith
each ri%al #roduct is .ei)hted &y its share o$ re%enues,21
3his decom#osition o$ the mar*et elasticity o$ demand is instructi%e .ith re)ard to
the standard #ractice in antitrust o$ de$inin) mar*ets &y decidin) .hether #articular
#roducts are su$$iciently )ood su&stitutesK)enerally understood as ha%in) su$$iciently
hi)h cross<elasticities o$ demandKto &e included in the mar*et, 3he e/#ression ma*es
clear that e%en a su&stitute .ith a %ery hi)h cross<elasticity may ha%e much less in$luence
than that o$ a lar)e )rou# o$ other #roducts, no one o$ .hich has a #articularly
hi)h cross<elasticity, Moreo%er, #roductsG shares o$ total re%enues are not ordinarily
considered in an e/#licit .ay, yet the $ormula indicates that a su&stitute .ith hal$ the
cross<elasticity o$ another can readily &e more im#ortant, in #articular, i$ its associated
re%enues are more than t.ice as hi)h, More &roadly, this re#resentation o$ the relationshi#
&et.een o%erall elasticity and indi%idual cross<elasticities rein$orces the #oint
that the e$$ect o$ su&stitutes is a matter o$ de)ree and thus not .ell ca#tured &y the
all<or<nothin) a##roach in%ol%ed in de$inin) antitrust mar*ets,
ome $urther comments concernin) mar*et share are in order, #articularly in li)ht
o$ the $act that a #ersistently hi)h mar*et share is %ery $reCuently #resented as com#ellin)
e%idence that a $irm has mar*et #o.er, Bo dou&t this in$erence is o$ten %alid,
s#eci$ically, i$ the mar*et demand elasticity and ri%alsG su##ly elasticities are lo. in
ma)nitude and the mar*et conditions are reasona&ly sta&le, -o.e%er, a $irm .ith only a
modest cost ad%anta)e may #ro$ita&ly maintain its hi)h share &y #ricin) lo. enou)h to
ca#ture most o$ the mar*et, 3his occurs, $or e/am#le, in the model o$ the dominant $irm
$acin) a com#etiti%e $rin)e i$ the $rin)e su##ly is %ery elastic at a #rice Eust a&o%e the
$irmGs o.n mar)inal cost, 2onsider, $or e/am#le, a truc*in) $irm that #ro%ides 100S o$
21 2ross<elasticities need not &e #ositi%e, For e/am#le, .hen the .ei)hted summation eCuals 4ero, .e ha%e
the $amiliar case o$ unit elasticityKthat is, as #rice rises, e/#enditures on the #roduct in Cuestion remain
constantKand .hen the summation is ne)ati%e, .e ha%e an elasticity less than one in a&solute %alue, o$ten
re$erred to as inelastic demand,
Ch% .1- $ntitrust 10'3
the $rei)ht trans#ortation on a #articular route &ut .ould Cuic*ly &e dis#laced &y near&y
ri%als 9.hose costs are essentially the same &ut .ho su$$er a sli)ht disad%anta)e due to
a lac* o$ $amiliarity .ith the routeGs customers: i$ it .ere to raise its #rice e%en a $e.
#ercent, Additionally, a $irm may ha%e a 100S share in a mar*et #rotected &y a #atent,
&ut i$ there are su$$iciently close su&stitutes, its mar*et #o.er is ne)li)i&le, 2on%ersely,
e%en a $irm .ith a lo. share o$ sales o$ a #articular #roduct may ha%e Cuite a &it o$
technical mar*et #o.er i$ the ma)nitude o$ the mar*et elasticity o$ demand and ri%alsG
elasticity o$ su##ly $or that #roduct are %ery lo., Aasoline re$inin) and electricity )eneration
are t.o e/am#les o$ #roducts $or .hich this latter situation can arise, "n sum, the
ri)ht side o$ e/#ression 92: indicates that mar*et share is only one $actor that determines
the elasticity o$ demand $acin) a $irm, so the ma)nitude o$ mar*et share is a rele%ant
com#onent o$ mar*et #o.er &ut not a conclusi%e indicator,
#%2%#%4% Rivals< supply response- !arriers to e:pansion, mo!ility, and entry "n e/aminin)
the ri)ht side o$ e/#ression 92: $or the $irmGs elasticity o$ demand, the #recedin)
su&section $ocused on the mar*et elasticity o$ demand and mar*et share, -o.e%er, the
elasticity o$ su##ly &y ri%als is also rele%ant, as indicated &y the Eust<mentioned contrastin)
e/am#les o$ truc*in), on one hand, and )asoline re$inin) and electricity )eneration,
on the other hand, 3he conce#t o$ ri%alsG su##ly should &e understood &roadly, to include
e/#anded out#ut $rom e/istin) #lants, shi$tin) ca#ital $rom other re)ions or $rom
the #roduction o$ other #roducts, introducin) ne. &rands or re#ositionin) e/istin) ones,
and entry &y $irms in related &usinesses or &y other $irms, "$ mar*et #o.er is si)ni$icant,
it must &e that the a))re)ate o$ these #otential su##ly res#onsesKo$ten re$erred to as
e/#ansion, mo&ility, and entryKis su$$iciently limited, at least o%er some #ertinent time
#eriod, Ail&ert 91'8': #ro%ides an e/tended discussion o$ such &arriers, Berry and 7eiss
9200F: sur%ey em#irical models o$ entry and mar*et #o.er, and utton 9200F: discusses
the relationshi# &et.een mar*et structure and mar*et #o.er,
"n some cases, the elasticity o$ ri%alsG su##ly may &e measured directly, &y measurin)
out#ut res#onses to #re%ious chan)es in #rice &y the $irm in Cuestion, or &y other $irms in
similar mar*ets, !$ten, ho.e%er, some e/tra#olation is reCuired, such as in #redictin)
.hether a hy#othetical increase in #rice to un#recedented le%els $ollo.in) a mer)er
.ould )enerate a si)ni$icant su##ly res#onse, For internal e/#ansion &y e/istin) ri%als,
the Cuestion .ould &e .hether there e/ist ca#acity constraints, stee#ly risin) mar)inal
costs, or limits on the inclination o$ consumers o$ di$$erentiated #roducts to s.itch
alle)iances, "n the case o$ ne. entry, timin), #ossi&le le)al restrictions 9intellectual
#ro#erty, 4onin), and other re)ulatory constraints:, &rand #re$erences, the im#ortance
o$ learnin) &y doin), and the a&ility to recou# $i/ed costs, amon) other $actors, .ill
determine the e/tent o$ restraint im#osed,
;articularly re)ardin) the latter, it is common to inCuire into the e/istence o$ so<called
&arriers to entry 9sometimes ta*en as a shorthand $or all $orms o$ su##ly res#onse &y
ri%als:, "n some instances, such as .hen there are le)al restrictions, the meanin) o$
this conce#t is $airly clear, -o.e%er, in many cases, it is di$$icult to ma*e sense o$ the
notion o$ entry &arriers in a %acuum, For e/am#le, there is much de&ate a&out .hether
10'1 L% 7aplow and C% Shapiro
economies o$ scale should &e %ie.ed as a &arrier to entry, "$ minimum e$$icient scale
is lar)e and incum&ent #roducers ha%e lon)<term e/clusi%e dealin) contracts .ith most
distri&utors, entry may &e rendered too costly, and e/istin) $irms mi)ht enEoy hi)h #ricecost
mar)ins 9more than necessary to co%er $i/ed costs:, "$ instead there merely e/ist
$i/ed costs and mar)inal costs are constant, in a $ree<entry eCuili&rium there .ill &e
#ositi%e #rice<cost mar)ins yet no #ro$its, 3he #ositi%e mar)ins .ill not induce $urther
entry &ecause their le%el #ost<entry .ould &e insu$$icient to reco%er $i/ed costs, As .e
ha%e o&ser%ed re#eatedly, althou)h mar*et #o.er .ould e/ist in the technical sense, the
situation should not &e %ie.ed as #ro&lematic $rom an antitrust #ers#ecti%e,
Many structural $eatures o$ mar*ets ha%e &een identi$ied as #ossi&le entry &arriers=
economies o$ scale, learnin) &y doin), re#utation, access to ca#ital, customer s.itchin)
costs, lac* o$ #roduct com#ati&ility, net.or* e$$ects, #atent #rotection, and access to
distri&ution channels, Because the im#lication o$ so<called entry &arriers de#ends on
the conte/tKand &ecause some de)ree o$ mar*et #o.er is sometimes una%oida&le yet
many are reluctant to state or im#ly its e/istence, such as &y deemin) somethin) to
&e an entry &arrier in a settin) .here antitrust inter%ention seems ina##ro#riateKthere
is no real consensus on ho. the term M&arriers to entryN should &e de$ined or a##lied
in #ractice,2@ 5e do not see clear &ene$its to $ormulatin) a canonical de$inition o$ the
conce#t, "t may &e &est sim#ly to *ee# in mind the #ur#ose o$ such inCuiries into the
e/istence o$ entry &arriers= to assess ri%alsG su##ly res#onse as an as#ect o$ an inCuiry
into the e/istence o$ mar*et #o.er, notin) that mar*et #o.er is o$ten rele%ant to antitrust
lia&ility &ut not su$$icient to esta&lish it, Beyond that, it may &e more hel#$ul to de$er
$urther analysis until considerin) s#eci$ic #ractices in s#eci$ic settin)s,
#%2%4% Conduct
"n some situations, one may &e a&le to in$er the #resence o$ mar*et #o.er $rom the
challen)ed conduct itsel$, "$ .e o&ser%e a $irm en)a)in) in a #ractice that could not &e
#ro$ita&le unless it enhanced the $irmGs mar*et #o.er to some certain de)ree, .e may
then in$er that mar*et #o.er .ould indeed increase to that de)ree, For e/am#le, i$ a $irm
#ays lar)e amounts to retailers to a)ree not to deal .ith #ros#ecti%e entrants or s#ends
lar)e sums to maintain tari$$s, .e may in$er that these #ractices create or enhance that
$irmGs mar*et #o.er,20 "$ one acce#ts the #remise that a $irmGs e/#ertise in assessin) its
o.n mar*et #o.er is li*ely to &e more relia&le than that #roduced &y a &attle o$ e/#erts
&e$ore an a)ency or in liti)ation, then the $irmGs o.n conduct may &e a sound &asis $or
in$errin) the e/istence o$ mar*et #o.er,
2@ ee 2arlton 92001:, McA$ee, Mialon, and5illiams 92001:, and chmalensee 92001: $or recent discussions
o$ ho. to a##ly the conce#t o$ entry &arriers in antitrust analysis,
20 As .e discuss in su&section @,1,2 in our analysis o$ e/clusi%e dealin) contracts .ith retailers, .e .ould
need to rule out #ro<com#etiti%e Eusti$ications, such as those &ased on $ree ridin), "n the case o$ lo&&yin)
to erect tari$$ &arriers, e%en i$ the conduct enhances mar*et #o.er, it .ould not %iolate U,, antitrust la.s
&ecause #etitionin) )o%ernment, e%en to restrict com#etition, is e/em#t acti%ity under the Noerr3'ennin*ton
doctrine,
Ch% .1- $ntitrust 10'@
3.o ca%eats should &e noted, First, the amount o$ mar*et #o.er that may &e in$erred
.ill sometimes not &e %ery )reat, A $irm .ith &illions o$ dollars in sales .ould ha##ily
s#end millions lo&&yin) $or tari$$s e%en i$ the resultin) de)ree o$ mar*et #o.er .ere
tri%ial, !n the other hand, i$ a $irm en)a)es in a #lan o$ &elo.<cost #ricin) that sacri$ices
hundreds o$ millions o$ dollars in current #ro$its, in the a&sence o$ other e/#lanations
one mi)ht .ell in$er that it antici#ates a su&stantial de)ree o$ mar*et #o.er, at least
su$$icient to recou# its in%estment,
econd, the relia&ility o$ the in$erence de#ends )reatly on the lac* o$ am&i)uity
re)ardin) the character o$ the #ractice under consideration, "$ one is certain that the
conduct .ould only &e underta*en i$ it could enhance the $irmGs mar*et #o.er 9to
some reCuisite de)ree:, then the in$erence is sound, -o.e%er, o$ten it .ill &e contested
.hether the conduct in Cuestion .as desi)ned to and .ill ha%e the e$$ect o$ increasin)
mar*et #o.er rather than constitutin) a &eni)n or e%en &ene$icial #ractice that increases
.el$are, For e/am#le, #rices &elo. cost may &e #ro$ita&le &ecause they are #redatory,
or &ecause they are introductory o$$ers that .ill enhance $uture demand $or an e/#erience
)ood, or &ecause they stimulate the demand $or other #roducts sold &y the $irm at
a healthy #rice<cost mar)in, "$ #ro<com#etiti%e e/#lanations are su$$iciently #lausi&le,
no in$erence o$ mar*et #o.er is .arranted, at least .ithout $urther in%esti)ation,
7eco)ni4in) the #ossi&ility that the conduct at issue may &e #ro<com#etiti%e is es#ecially
im#ortant )i%en the role that mar*et #o.er reCuirements o$ten #lay in antitrust,
namely, as a screenin) de%ice, 3hat is, .e may reCuire a #lainti$$ to #ro%e the e/istence
o$ mar*et #o.er &ecause .e do not .ant to su&Eect a .ide ran)e o$ &eha%ior to
the costs o$ antitrust scrutiny and the #ossi&ility o$ erroneous lia&ility, 5hen the conduct
that #ro%ides the &asis $or in$errin) mar*et #o.er is the %ery same conduct under
scrutiny, and $urthermore .hen the #ur#ose and e$$ect o$ such conduct is am&i)uous,
#ermittin) an in$erence o$ mar*et #o.er $rom the conduct some.hat undermines the
screenin) $unction o$ the mar*et #o.er threshold, 3his concern may &e es#ecially )reat
.hen Euries ser%e as the $inders o$ $act,2F
#%1% &aret power in antitrust law
As noted, in antitrust la. the notion o$ mar*et #o.er is $reCuently used as a screen=
a $irm 9or )rou# o$ $irms: must &e sho.n to ha%e some le%el o$ mar*et #o.er as a
#rereCuisite to considerin) .hether the conduct in Cuestion )i%es rise to antitrust lia&ility,
As a result, antitrust in%esti)ations and adEudications de%ote su&stantial attention
2F 3his concern may hel# to e/#lain the u#reme 2ourtGs decision in #ectrum #orts %, McTuillan, @00
U,, 11F 91''3:, .here the 2ourt held in an attem#ted mono#oli4ation case that the #lainti$$ had to meet
the mar*et #o.er reCuirement inde#endently o$ #ro%in) #redatory conduct, Althou)h the holdin) on its $ace
seems illo)ical 9i$, as the #lainti$$ ar)ued, it .ould ha%e &een irrational to ha%e en)a)ed in the conduct unless
the reCuisite contri&ution to mar*et #o.er .ere #resent:, the actual #ractice under consideration may .ell
ha%e a##eared to the 2ourt to &e non#redatory, so it .ished to hei)hten the #lainti$$Gs reCuired #roo$ &e$ore it
.ould allo. the case to &e considered &y the Eury,
10'0 L% 7aplow and C% Shapiro
to .hether or not the reCuisite mar*et #o.er e/ists, "n rhetoric and o$ten in reality,
this le)al a##roach o$ %ie.in) mar*et #o.er as somethin) either #resent or a&sentKa
dichotomous classi$icationKis at odds .ith the technical economic notion o$ mar*et
#o.er as a matter o$ de)ree, Because some de)ree o$ technical mar*et #o.er is u&iCuitous,
it is e%ident that the term Mmar*et #o.erN as used in antitrust la. has another
meanin), Be%ertheless, the la.Gs notion o$ mar*et #o.er is Cuite closely related to that
o$ economists, A le)al $indin) o$ mar*et #o.er constitutes not merely a declaration o$
the e/istence o$ technical mar*et #o.er, ho.e%er tri%ial, &ut rather a conclusion that
the de)ree o$ e/istin) or increased mar*et #o.er e/ceeds some threshold, a &enchmar*
that, as .e .ill see, %aries .ith the ty#e o$ conduct under consideration and that in most
instances is not clearly s#eci$ied,
3his $eature o$ antitrust la.Gs use o$ a mar*et #o.er reCuirement is .ell illustrated &y
the la. o$ mono#oli4ation, As .ill &e ela&orated in su&section @,2, under U,, antitrust
la. MUtVhe o$$ense o$ mono#oly , , , has t.o elements= 91: the #ossession o$ mono#oly
#o.er in the rele%ant mar*et and 92: the .ill$ul acCuisition or maintenance o$ that #o.er
as distin)uished $rom )ro.th or de%elo#ment as a conseCuence o$ a su#erior #roduct,
&usiness acumen, or historic accident,N28 3he reCuirement o$ Mmono#oly #o.erN
is conclusory in that it merely si)ni$ies that de)ree o$ mar*et #o.er deemed minimally
necessary and also su$$icient to satis$y the $irst element o$ the o$$ense o$ mono#oli4ation,
"t is understood that this le%el o$ mar*et #o.er is hi)her than that reCuired in other
areas o$ antitrust la., Bota&ly, the mar*et #o.er reCuirement is hi)hest in mono#oli4ation
cases, some.hat lo.er in attem#ted mono#oli4ation cases, and lo.er still in
hori4ontal mer)er cases, as .ill &e discussed in su&sections 1,1,2 and @,2,1, -o.e%er,
these reCuirements ty#ically are not stated Cuantitati%ely, ma*in) it di$$icult to *no.
%ery #recisely .hat is the threshold in any o$ these areas,
"n #rinci#le, the $act that mar*et #o.er is a matter o$ de)ree should &e reco)ni4ed in
desi)nin) antitrust rules, A mono#olistic act that is unam&i)uously undesira&le mi)ht
&e condemned e%en i$ the incremental im#act on mar*et #o.er is modest, .hereas $or
conduct that is am&i)uous, .ith a hi)h ris* o$ $alse #ositi%es, it may &e a##ro#riate to
contem#late condemnation only .hen the #otential e$$ect on mar*et #o.er is su&stantial,
"$ one .ere minimi4in) a loss $unction in .hich there .as uncertainty a&out the
#ractices under scrutiny, and i$ the de)ree o$ harm conditional on the #ractices &ein)
detrimental .as risin) .ith the e/tent o$ mar*et #o.er, an o#timal rule could &e stated
as entailin) a mar*et #o.er reCuirement that .as hi)hly conte/tual,
For #ractical use &y a)encies and in adEudication, ho.e%er, a more sim#li$ied $ormulation
may economi4e on administrati%e costs, #ro%ide clearer )uidance to #arties, and
re$lect the limited e/#ertise o$ the #ertinent decision<ma*ers, Be%ertheless, some )reater
$le/i&ility may &e .arranted and is indeed increasin)ly re$lected in antitrust doctrine,
3he early emer)ence o$ a per se rule a)ainst #rice<$i/in), .hich dis#enses .ith #roo$
o$ mar*et #o.er, is one illustration, Another is the increasin) use o$ intermediate le%els
o$ scrutiny under the rule o$ reason 9see su&section 3,@,2: and the im#licit reliance
28 United tates %, Arinnell 2or#,, 381 U,, @03, @F0OF1 91'00:,
Ch% .1- $ntitrust 10'F
on di$$erent mar*et #o.er thresholds &y the antitrust a)encies in re%ie.in) hori4ontal
mer)ers in di$$erent industries, des#ite the e/istence o$ o$$icial )uidelines that #ur#ort
to &e uni$orm,
"n addition to di$$erences in the ma)nitude o$ mar*et #o.er thresholds and .hether
there is some $le/i&ility re)ardin) the reCuisite de)ree o$ mar*et #o.er, there is %ariation
across conte/ts in .hether the Cuestion #osed concerns the e/tant le%el o$ mar*et #o.er
or the amount &y .hich the actions under scrutiny .ould increase it, "n a mono#oli4ation
case, the standard Cuestion is o$ten .hether a $irmGs #ast #ractices ha%e im#ro#erly
created or maintained mono#oly #o.er, so the inCuiry is usually into .hether si)ni$icant
mar*et #o.er already e/ists, as re$lected in the #re%iously Cuoted $ormulation, By contrast,
in e/aminin) hori4ontal mer)ers, the $ocus is on .hether the #ro#osed acCuisition
.ould si)ni$icantly increase mar*et #o.er,2'
5e &elie%e that this distinction is o%erstated and #otentially misleadin) and that the
correct inCuiry should $ocus lar)ely on contri&utions to mar*et #o.er, E%en in the
ty#ical mono#oli4ation case, the rele%ant Cuestion is ho. much the #ast #ractices contri&uted
to the e/istin) situation, "$ the contri&ution is lar)e 9and i$ the #ractices are
not other.ise Eusti$ia&le:, it seems that there should &e a $indin) o$ lia&ility e%en i$ the
resultin) total de)ree o$ #o.er is not o%er.helmin), 9"n such a case, the initial le%el
o$ mar*et #o.er #resuma&ly .ill ha%e &een rather lo.,: ?i*e.ise, e%en i$ the de)ree
o$ e/istin) mar*et #o.er is )reat, in cases in .hich the #ractices in Cuestion did not
#lausi&ly contri&ute si)ni$icantly to that result, one should &e cautious in condemnin)
those #ractices, that is, they should &e condemned only i$ they are unam&i)uously undesira&le,
As an e/am#le, consider a $irm sellin) a relati%ely homo)eneous #roduct, such as in
the chemical industry, that enEoys a si)ni$icant cost ad%anta)e o%er its ri%als &ased on
#atented #rocess technolo)y, 3hat $irm mi)ht .ell enEoy a nontri%ial de)ree o$ technical
mar*et #o.er, Beither )ood sense nor e/istin) la. ordinarily condemns the disco%ery
o$ a su#erior #roduction #rocess, ?et us assume that the $irmGs technical mar*et #o.er
.as le)ally o&tained and su##ose $urther that the $irm #rices a)ainst a #er$ectly elastic
ri%al su##ly at some tri))er #rice that is &elo. the $irmGs mono#oly #rice, Antitrust
issues could arise i$ this $irm attem#ts to acCuire its ri%als or i$ the $irm en)a)es in
conduct that dri%es its ri%als out o$ &usiness, "n considerin) such cases, the de)ree o$
the $irmGs initial mar*et #o.er is o$ secondary im#ortance 9althou)h i$ it .ere near
4ero, $urther inCuiry .ould #ro&a&ly &e #ointless:, "nstead, the central Cuestion should
&e .hether and to .hat e/tent the acCuisition or e/clusionary conduct .ill au)ment
that $irmGs mar*et #o.er and thus harm consumers, For e/am#le, ho.e%er )reat is the
initial le%el o$ mar*et #o.er, the $irm .ould )ain no additional #o.er &y acCuirin)
9or destroyin): one o$ its ri%als as lon) as numerous others that remain still ha%e hi)hly
elastic su##ly at that same tri))er #rice, -o.e%er, the $irm mi)ht .ell )ain mar*et #o.er
&y acCuirin) or destroyin) a ri%al .ith uniCuely lo. costs, there&y raisin) the #rice at
2' ee su&section 1,1,2, .here .e discuss the #oint that the e/tant le%el o$ mar*et #o.er is also im#ortant,
10'8 L% 7aplow and C% Shapiro
.hich su&stantial com#etin) su##ly .ould &e tri))ered,5e return to the Cuestion o$ the
rele%ance o$ e/tant mar*et #o.er %ersus challen)ed #racticesG contri&ution to #o.er in
su&section @,1,2 .ith re)ard to mono#oli4ation and e/clusionary #ractices,
A $urther #ossi&le de%iation &et.een economic analysis and antitrust la. .ith re)ard
to mar*et #o.er concerns the &enchmar* a)ainst .hich the hei)ht o$ #rice<cost mar)ins
is assessed, 3he U,, antitrust en$orcement a)encies in the -ori4ontal Mer)er Auidelines
91''2: de$ine mar*et #o.er as Mthe a&ility #ro$ita&ly to maintain #rices a&o%e
com#etiti%e le%els $or a si)ni$icant #eriod o$ time,N and the u#reme 2ourt has similarly
stated that MAs an economic matter, mar*et #o.er e/ists .hene%er #rices can &e raised
a&o%e the le%els that .ould &e char)ed in a com#etiti%e mar*et,N30 "$ one understands
the com#etiti%e #rice to re$er to the #rice that .ould &e char)ed in a hy#othetical, te/t&oo*,
#er$ectly com#etiti%e mar*et in .hich $irms ha%e constant mar)inal costs eCual
to the mar)inal cost o$ the $irm in Cuestion at the #re%ailin) eCuili&rium, then the le)al
and economic conce#ts are essentially the same, -o.e%er, the hy#othetical com#etiti%e
scenario that underlies such statements is rather %a)ue= the counter$actual is not e/#licit,
and some s#eci$ications that may im#licitly &e contem#lated may not yield sensi&le ans.ers,
For e/am#le, .hat is meant &y the #er$ectly com#etiti%e #rice in a mar*et .ith
$i/ed costsW
2ourts ha%e stru))led .ith these issues $or many years, 3he u#reme 2ourt has stated
that MMono#oly #o.er is the #o.er to control #rices or e/clude com#etition,N31 3his
is not a meanin)$ul screen, ho.e%er, since any $irm .ith technical mar*et #o.er has
some a&ility to control #rices, 2on%ersely, in the Euro#ean Union, the Euro#ean 2ourt
o$ Justice has said that a Mdominant #ositionN corres#onds to Ma #osition o$ economic
stren)th enEoyed &y an underta*in) .hich ena&les it to hinder the maintenance o$ e$$ecti%e
com#etition on the rele%ant mar*et &y allo.in) it to &eha%e to an a##recia&le
e/tent inde#endently o$ its com#etitors and customers and ultimately o$ consumers,N32
3his test is not es#ecially use$ul either, since e%en a $irm .ith )reat mar*et #o.er does
not rationally &eha%e inde#endently o$ its com#etitors or customers, 3hat is, there is
some mono#oly #rice, PM, .hichKho.e%er hi)h it may &eKim#lies that a #rice o$,
say, 2PM .ould &e less #ro$ita&le due to $ar )reater consumer su&stitution a.ay $rom
the #roduct at that hi)her #rice,
+& Collusion
5e no. turn to collusion, includin) #rice<$i/in) cartels and other arran)ements that may
ha%e similar e$$ects, such as the allocation o$ customers or territories to di$$erent su##li<
30 Je$$erson ;arish -os#ital 6istrict Bo, 2 %, -yde, 100 U,, 2, 2F n, 10 91'81:,
31 U,, %, E,", du ;ont de Bemours & 2o, 3@1 U,, 3FF, 3'1 91'@0:,
32 2ase 322+81, Michelin %, 2ommission U1'83V E27 3101 X30,
Ch% .1- $ntitrust 10''
ers,33 For concreteness, .e .ill ordinarily $ocus on #rice<$i/in), 3here is an enormous
literature on the economics o$ collusion that .e do not attem#t to re%ie. systematically,
E/istin) sur%eys include ha#iro 91'8':, JacCuemin and lade 91'8':, Motta 92001,
ch, 1:, and o$ #articular note 5hinston 92000, ch, 2:, For in<de#th discussion o$ some
es#ecially interestin) #rice<$i/in) cases, see Borenstein 92001:, 2onnor 92001:, El4in)a
and Mills 92001:, Motta 92001, ##, 211O21':, and ;orter and Yona 92001:,
3he $ocus here, as in the rest o$ this sur%ey, is on the intersection o$ economics and
the la., 5e &e)in &y notin) the core elements $rom each $ield and #osin) Cuestions
a&out their relationshi#, Be/t, .e e/#lore the economics o$ collusion, $ocusin) on the
necessary elements $or success$ul collusion, lessons $rom )ame<theoretic models o$
oli)o#oly, and the %arious $actors that &ear on the li*elihood o$ success$ul collusion,
Finally, .e e/amine le)al #rohi&itions in li)ht o$ the &asic teachin)s o$ economics,
4%.% Economic and le*al approaches- an introduction
4%.%.% Economic approach
For as lon) as there has &een commercial com#etition, ri%als ha%e &een tem#ted to
short<circuit it &ecause sel$<interest $a%ors their o.n #ro$its at the e/#ense o$ customersG
interest in lo.er #rices and the o%erall social interest in allocati%e e$$iciency, Bo less
a cham#ion o$ the $ree<mar*et system than Adam mith 9U1FF0V 1'F0, &* 1, ch, Z:
considered collusion an e%er<#resent dan)er, M;eo#le o$ the same trade seldom meet
to)ether, e%en $or merriment and di%ersion, &ut the con%ersation ends in a cons#iracy
a)ainst the #u&lic, or in some contri%ance to raise #rices,N "$ one thin*s in terms o$ a
homo)eneous #roduct, $irms see* to esta&lish and maintain the mono#oly #rice, .hich
e/ceeds the #rice that .ould #re%ail in the a&sence o$ the a)reement,5ith di$$erentiated
#roducts or #rice discrimination, althou)h there is no sin)le mono#oly #rice, the same
idea a##lies= $irms see* to ele%ate #rices and thus raise their collecti%e #ro$its at the
e/#ense o$ consumers, "n so doin), the $irms ty#ically increase the )a# &et.een #rice9s:
and mar)inal cost9s: and thus raise dead.ei)ht loss and lo.er total .el$are, de$ined
as the sum o$ su##lier #ro$its and consumer sur#lus, 3hus, collusion is un.elcome,
.hether one is see*in) to ma/imi4e o%erall e$$iciency or consumer .el$are,
2olludin) $irms use a %ariety o$ methods to achie%e the &asic )oal o$ raisin) #rices, "n
some cases, $irms a)ree to minimum #rices, "n others, they a)ree to limit their #roduction
le%els, since out#ut restrictions translate into ele%ated #rices, Alternati%ely, $irms
can allocate customers or territories amon) themsel%es, .ith each $irm a)reein) not to
com#ete $or customers, or in territories, assi)ned to others, 3hese customer and territorial
allocation schemes e$$ecti%ely )rant each $irm a mono#oly o%er some #ortion o$
33 5e do not e/#licitly address the $ull ran)e o$ Mhori4ontal a)reements,N .hich includes )rou# &oycotts as
.ell as arran)ements amon) &uyers, nota&ly, to su##ress the #rices o$ in#uts, the latter o$ .hich are su&Eect
to similar analysis as that #resented here, althou)h they ha%e recei%ed less antitrust scrutiny,
1100 L% 7aplow and C% Shapiro
the o%erall mar*et, so they lead to hi)her #rices and reduced out#ut, e%en thou)h these
schemes do not directly s#eci$y #rice or out#ut,
Economists studyin) collusion, and more )enerally oli)o#oly, tend to inCuire into
the $actors that determine the mar*et eCuili&rium outcome in an industry, Economists
ty#ically $ocus on .hether the outcome is relati%ely com#etiti%e, .ith #rices close to
mar)inal cost, or at least some measure o$ a%era)e cost, or relati%ely collusi%e, .ith
#rices close to the le%el that a mono#olist .ould #ic* to ma/imi4e industry #ro$its, 3his
a##roach is consistent .ith economistsG traditional em#hasis on mar*et outcomes and
their im#lications $or the allocation o$ resources,
3his a##roach $ocuses on descri#tion or #rediction, not on #olicy #rescri#tions re)ardin)
ho. the )o%ernment should miti)ate the costs o$ collusion, 3here is a )eneral
consensus that clearly identi$ia&le attem#ts to en)a)e in collusi%e &eha%ior should &e
#rohi&ited, so e/#licit cartel a)reements should not &e le)ally en$orcea&le and #ri%ate
attem#ts to a)ree u#on and en$orce su#ra<com#etiti%e #rices should &e #unished, "t is
.idely reco)ni4ed, ho.e%er, that it is not al.ays #ossi&le to determine .hether collusion
is occurrin) or, e%en .hen it is, .hich s#eci$ic #ractices should &e #roscri&ed,
!ne a##roach in such settin)s .ould &e #rice re)ulation, .hich is o$ten underta*en
in the case o$ natural mono#olies &ut is )enerally thou)ht to &e in$erior to decentrali4ed
com#etition .hen such is $easi&le, "n the #ast, there ha%e &een recommendations
to deconcentrate certain industries in order to achie%e more com#etiti%e outcomes,31
uch #ro#osals ha%e not &een im#lemented, e/ce#t in some cases o$ mono#oli4ation,
and ha%e not o$ late &een acti%ely considered in the United tates, Another structural
a##roach is more #re%alent= enEoinin) hori4ontal mer)ers that ma*e collusi%e outcomes
more li*ely, a to#ic .e e/#lore in section 1, Finally, $or cases in .hich collusion can &e
identi$ied &ut the s#eci$ic #ractices ena&lin) it cannot, ;osner 91'0', 2001: interestin)ly
#ro#oses the im#osition o$ monetary #enalties on oli)o#olists i$ the mar*et eCuili&rium
outcome is collusi%e,3@ 3he idea is that, Eust as ;i)ou%ian ta/es induce $irms to re$rain
$rom ine$$icient &eha%ior, the details o$ .hich mi)ht &e di$$icult $or a re)ulator to o&ser%e
or #roscri&e, so too .ould a##ro#riate $ines or dama)e a.ards in #ri%ate liti)ation
lead $irms to a&stain $rom collusi%e &eha%ior, 3his a##roach assumes, im#ortantly, that
it is #ossi&le to measure the e/tent to .hich #rices e/ceed non<collusi%e le%els, .hich
31 ?e)islation .as introduced re#eatedly in the early 1'F0s that .ould ha%e authori4ed the dissolution o$
$irms in concentrated industries that had not en)a)ed in su&stantial #rice com#etition o%er three consecuti%e
yearsJ see, $or e/am#le, ,110F, March 12, 1'F2, 3his le)islation .as &ased on the 5hite -ouse 3as* Force
7e#ort on Antitrust ;olicy 91'08:, commonly *no.n as the Beal 7e#ort,
3@ !n its $ace, #resent #ractice a##ears to di$$er si)ni$icantly $rom ;osnerGs #ro#osal, Althou)h the United
tates and most other com#etition en$orcement re)imes do #ro%ide $or $ines or #ri%ate dama)e remedies in
cases o$ #rice<$i/in), to tri))er such #ayment o&li)ations, the )o%ernment or #ri%ate #arties ty#ically need to
sho. that in some sense there is an Ma)reement,N Furthermore, satis$action o$ this reCuirement is )enerally
understood to entail more than demonstratin) that the o&ser%ed outcome in%ol%es a Mcollusi%e #rice,N althou)h
as .e shall discuss, Eust ho. much more must &e sho.n and .hat constitutes an adeCuate demonstration is
unclear, ee su&sections 3,1,2 and 3,1,
Ch% .1- $ntitrust 1101
#oses &oth conce#tual and #ractical challen)es o$ a sort that are encountered in im#osin)
con%entional #rice<$i/in) sanctions, the ma)nitude o$ .hich de#ends on the e/tent
o$ collusi%e o%erchar)es, -o.e%er, ;osnerGs a##roach has not &een em&raced &y the
courts,
4%.%#% Le*al approach
3he le)al a##roach to collusion, at least on its $ace, di$$ers $rom the economic a##roach,
As Eust descri&ed, the economic a##roach &e)ins .ith a dia)nosis o$ the #ro&lem, then
tries to ascertain .hether and .hen collusion occurs, and $inally assesses the e$$icacy
o$ com#etin) remedies, Althou)h one .ould li*e to &elie%e that the le)al a##roach is at
some le%el )rounded in such analysis, on the sur$ace it a##ears to $ocus instead on #articular
&eha%ioral elements, As .ill &e seen in the course o$ our analysis in section 3, the
e/tent to .hich the le)al a##roach can ultimately &e rationali4ed on economic )rounds
de#ends on ho. le)al tests are inter#reted,
"n the United tates, the Euro#ean Union, and many other Eurisdictions, the structure
o$ le)al #rohi&itions re%ol%es around the distinction &et.een unilateral and )rou# &eha%ior,
Unilateral &eha%ior is circumscri&ed to a limited de)ree &y anti<mono#oli4ation
la. 9see section @: and %arious other #ro%isions &ut is not su&Eect to a re)ime o$ #rice
re)ulation or other $orms o$ internal micro<mana)ement o$ $irmsG dealin)s,30 3hus,
$irms are #ur#ortedly $ree to set #rices and other conditions o$ trade,3F 3his $reedom,
ho.e%er, is restricted to unilateral &eha%ior, "nde#endent $irms are e/#ected to com#ete,
con$errin) the &ene$its o$ com#etition on consumers and on society as a .hole,
3he central le)al Cuestion .ith .hich .e .ill &e concernedKand .ill ela&orate in
su&section 3,1Kis ho. courts or other re)ulators are to determine .hen su##osedly
com#etin) $irms are instead cons#irin), ?e)al #rohi&itions are ty#ically tri))ered &y
certain ty#es o$ conduct rather than &y outcomes themsel%es, For concreteness, .e
.ill discuss the #rohi&ition in U,, antitrust la., herman Act X1, .hich ma*es ille)al
MUeV%ery contract, com&ination . . . , or cons#iracy, in restraint o$ trade,N "n #ractice,
the standard term o$ art is Ma)reement,N e%en thou)h that term does not a##ear in the
statute,38 3hus, the le)al Cuestion is .hether $irmsG #ricin) is the result o$ an a)reement,
"$ not, there is no %iolation, "$ so, there is a %iolation, and #enalties in the United tates
include ha%in) to #ay tre&le dama)es to inEured customers, &ein) su&Eect to inEunctions
on #rohi&ited &eha%ior, and criminal #enalties, under .hich $irmsG e/ecuti%es con%icted
o$ #rice<$i/in) ser%e #rison terms and $irms #ay $ines,
30 "t is true, ho.e%er, that remedies in mono#oli4ation cases and some others can entail .hat is tantamount
to $airly detailed re)ulation,
3F 3here are im#ortant Cuali$ications, nota&ly .ith re)ard to #roscri#tions on #redatory #ricin) 9see su&section
@,3:, &ut the $ocus in this section is on #rices that are too hi)h and thus directly harm customers rather
than on #rices that are too lo. and thus directly harm com#etitors,
38 "nterestin)ly, this is the lan)ua)e o$ Article 81 o$ the com#etition la. in the Euro#ean Union, ?i*e in the
United tates, the conce#t in the Euro#ean Union em&races more than $ormal contracts yet it is uncertain Eust
ho. much more,
1102 L% 7aplow and C% Shapiro
5hat, then, is an a)reementW And ho. does this conce#t relate to the economic
analysis o$ collusi%e &eha%iorW 3hese Cuestions .ill occu#y much o$ the remainder
o$ section 3 o$ our sur%ey, 3o #ro%ide some )uidance in the interim, a $e. #reliminary
o&ser%ations are o$$ered, First, there are clear cases, At one e/treme, i$ com#etitors
meet in the #ro%er&ial smo*e<$illed room, ne)otiate a detailed cartel arran)ement,
si)n it, and im#lement itKand, im#ortantly, this all can &e #ro%ed in a le)al
#roceedin)Kan a)reement and hence a le)al %iolation .ill undou&tedly &e $ound to
e/ist, At the other e/treme, no a)reement .ould #resuma&ly e/ist and no %iolation
.ould &e $ound due to the mere $act that com#etitorsG #rices are eCualKas one e/#ects
.ith homo)eneous #roducts and #er$ect com#etition, $or e/am#leKor that they
sometimes mo%e to)etherKas tends to occur .hen there are shoc*s to in#ut #rices
9thin* o$ retail )asoline stations chan)in) sale #rices .hen #rices $rom re$ineries
chan)e:,
3he di$$icult cases $all at %arious #oints in &et.een, in terms o$ .hat actually trans#ired
and .hat can &e #ro%ed &e$ore a tri&unal, 2onsider a sim#le e/am#le, u##ose
that Eust t.o $irms, A and B, su##ly a #articular #roduct, ?et the mono#oly #rice o$ that
#roduct &e [100 and the $ully com#etiti%e #rice 9that is, the #rice at .hich the industry
mar)inal cost cur%e crosses the demand cur%e: &e only [10, u##ose $urther that the actual
industry #rice #ersists at [100, .ith sales s#lit e%enly &et.een the t.o $irms, 3his
is clearly a collusi%e outcome, &ut ha%e the $irms entered into an a)reement in restraint
o$ tradeW
As noted, i$ a .ritten a)reement, ne)otiated and si)ned in a smo*e<$illed room, is #roduced
as e%idence, a %iolation .ill &e $ound, u##ose that no such a)reement is directly
in e%idence, !ne #ossi&ility is that such an a)reement ne%ertheless e/ists, and i$ a tri&unal
can &e con%inced o$ this &y circumstantial e%idence, a case .ill ha%e &een made,
But .hat sort o$ e%idence .ould &e necessary to ma*e this in$erenceW 3he ans.er de#ends
im#ortantly on the com#etin) hy#othesesKand on .hich alternati%e e/#lanations
are li*e.ise deemed to in%ol%e a)reements and hence .ould also constitute %iolations,
5hat interactions short o$ a meetin) in a smo*e<$illed room that results in a .ritten
document .ill su$$iceW "s a $ace<to<$ace meetin) reCuired or .ould a con$erence call
or an e<mail e/chan)e &e enou)hW 5hat a&out other $orms o$ communication, such as
statements relayed throu)h third #arties or in %arious codesW !r non%er&al communication
9hand si)nals, .in*s and nods, #ostin) si)ns .ith #ro#osed #rices, and so $orth:W
Must there &e a .ritten documentW ;resuma&ly not, Must there &e a $ormal a)reement
tantamount to a le)ally en$orcea&le contractW5ell, since a contract .ould not &e le)ally
en$orcea&le in any e%ent, #resuma&ly this too is not reCuired,
"n sum, .e can &e certain that a)reements may &e deemed to e/ist .hen somethin)
.ell short o$ the $ormal meetin) and .ritten cartel document e/ists, But it is not clear
ho. much less .ill )i%e rise to lia&ility, or, #ut in the a$$irmati%e, Eust .hat is 9are:
the core underlyin) element9s: o$ an Ma)reement,N For no., .e .ill lea%e this Cuestion,
returnin) to it in su&section 3,1, a$ter .e ha%e sur%eyed some *ey as#ects o$
the #ertinent economic theory, .hich one mi)ht ho#e .ould illuminate the le)al inCuiry,
Ch% .1- $ntitrust 1103
4%#% +li*opoly theory
4%#%.% Elements of successful collusion
Economists ha%e lon) reco)ni4ed that there e/ist certain #rereCuisites to success$ul collusion,
3he classic modern re$erence is ti)ler 91'01:, Areen and ;orter 91'81: em&ed
these issues in a su#er)ame conte/t, 3he *ey elements are 91: reachin* consensus= some
understandin) must &e reached amon) the other.ise<com#etin) $irms re)ardin) .hat
conduct is #ermitted under the terms o$ the collusi%e a)reement, such as the #rices that
the $irms .ill char)eJ 92: detection= some relia&le means must e/ist &y .hich de#artures
$rom the a)reement can &e detectedJ and 93: punishment= some credi&le mechanism
must &e esta&lished &y .hich such de#artures are #unished i$ and .hen they are detected,
#eci$ically, the #ros#ect o$ detection and #unishment must &e su$$icient to deter
indi%idual $irmsG #rocli%ity to cheat on the a)reement, ty#ically &y cuttin) #rices in the
short<term, ho#in) to rea# )reater #ro$its throu)h a hi)her mar*et share at the e/#ense
o$ the other $irms, &e$ore they can res#ond, 7elated to the need to reach an a)reement
is the #ro&lem o$ 91: inclusion= a means o$ inducin) #artici#ation &y a su$$iciently lar)e
num&er o$ incum&ent su##liers so that com#etition $rom non<#artici#ants does not undermine
the #ro$ita&ility o$ the collusi%e a)reement, ?astly and relatedly, the incum&ent
$irms must &e #rotected &y 9@: entry !arriers= there must not &e so much com#etition
$rom Cuic*ly arri%in) ne. entrants so as to undermine the e$$ecti%eness o$ collusion,
ome economists consider these reCuirements to &e so dauntin) that cartels are either
una&le to $orm or Cuic* to colla#se, e%en in the a&sence o$ antitrust la.s desi)ned to sto#
collusion, For e/am#le, .hen !;E2 $irst arose, some con$idently #redicted its immediate
demise, -o.e%er, the e/#erience .ith !;E2 and em#irical e%idence on #rice<$i/in)
more &roadly does not su##ort this o#timistic %ie., For e/am#le, in the #ast decade,
the Antitrust 6i%ision o$ the U,, 6e#artment o$ Justice has &ro*en u# many lar)e, international
cartels that had o#erated $or years 9des#ite the $act that they .ere ille)al
under the antitrust la.s: and success$ully rea#ed hundreds o$ millions i$ not &illions o$
dollars in #ro$its,3' For recent e/tensi%e sur%eys o$ the e%idence o$ collusi%e acti%ity,
see 2onnor 9200F:, -arrin)ton 92000&:, and ?e%enstein and uslo. 92000:, 5hinston
92000, ##, 20O38: o$$ers a more selecti%e discussion o$ the em#irical e%idence re)ardin)
the e$$ects o$ #rice<$i/in) cons#iracies,
3' ee ?itan and ha#iro 92002: $or a discussion o$ U,, cartel en$orcement acti%ities durin) the 1''0s,
3he Antitrust 6i%ision o$ the 6e#artment o$ Justice re)ularly announces en$orcement actions a)ainst
cartels, For e/am#le, as o$ e#tem&er 2000, some [F31 million o$ $ines had resulted $rom the 6i%isionGs
in%esti)ation into #rice<$i/in) o$ 67AM 96ynamic 7andom Access Memory:, ee Mamsun) E/ecuti%e
A)rees to ;lead Auilty, er%e Jail 3ime, $or ;artici#atin) in 67AM ;rice<Fi/in) 2ons#iracy,N
htt#=++...,usdoE,)o%+atr+#u&lic+#ressLreleases+2000+218102,htm,
1101 L% 7aplow and C% Shapiro
4%#%#% Repeated oli*opoly *ames and the fol theorem
3he &asic theoretical $rame.or* used to e%aluate the #resence, a&sence, or e$$icacy
o$ collusion is that o$ dynamic or re#eated oli)o#oly, that is, situations in .hich an
identi$ia&le )rou# o$ su##liers o$$erin) su&stitute #roducts interact o%er time,10 3his
$rame.or* includes in$initely re#eated oli)o#oly )ames, so<called su#er)ames, 2artel
theory reCuires dynamic analysis &ecause the central elements o$ detection and #unishment
inherently ta*e #lace o%er time,
!ne o$ the central $indin)s in the theory o$ oli)o#oly su#er)ames is that there are
manyKindeed, in$initely manyKnoncoo#erati%e eCuili&rium outcomes, includin) outcomes
that ma/imi4e the Eoint #ro$its o$ the oli)o#olists, e%en i$ one restricts attention to
su&)ame<#er$ect Bash eCuili&ria,11 3o )i%e a $la%or $or .hy there are so many eCuili&ria
in su#er)ames, consider a )ame in .hich each o$ N $irms, sellin) the same homo)eneous
)ood and incurrin) a constant cost #er unit o$ C, sets its #rice e%ery #eriod,
u##ose that the sta)e )ame is a classic Bertrand #ricin) )ame, so the $irm .ith the
lo.est #rice in a )i%en #eriod ser%es the entire mar*et in that #eriod, and i$ multi#le
$irms char)e the same lo.est #rice out#ut is shared eCually amon) them, 9!ne can also
thin* o$ the $irms &iddin) each #eriod to ser%e the sin)le lar)e customer .ho &uys that
#eriod,: For the moment, su##ose also that, as soon as a )i%en #eriod ends, each $irm
immediately o&ser%es all the #rices char)ed &y the other $irms durin) that #eriod,
"n this sim#le re#eated Bertrand )ame, the com#etiti%e outcome, &y .hich .e mean
the Bash eCuili&rium in the one<shot 9sta)e: )ame, in%ol%es each $irm settin) its #rice
eCual to cost, C, each #eriod, 3his .ould &e the only noncoo#erati%e eCuili&rium outcome
i$ this #ricin) )ame .ere #layed only a sin)le time, or indeed any $inite num&er
o$ times 9a $amiliar conseCuence o$ &ac*.ard induction:, -o.e%er, i$ the )ame continues
inde$initely, it is easy to construct a su&)ame<#er$ect eCuili&rium in .hich the #rice
each #eriod is P > C, $or many di$$erent %alues o$ P ran)in) $rom the com#etiti%e
#rice all the .ay u# to the mono#oly #rice 9or, in $act, hi)her:, 3he tric* is to #ostulate
that, should any $irm e%er char)e &elo. P, all the other $irms .ill set a #rice o$ C in
all su&seCuent #eriods, 3his #unishment strate)y thus entails re%ersion to the one<shot
Bash eCuili&rium, Because this &eha%ior su##orts an eCuili&rium #rice o$ P e%ery #eriod,
the eCuili&rium #ro$it o$ each $irm is (P)/N each #eriod, .here (P) denotes
the #ro$its earned &y a $irm settin) #rice P and ser%in) the entire mar*et, Addin) u#
o%er all #eriods and discountin) usin) the one<#eriod discount $actor , the eCuili&rium
#ro$its o$ each $irm are (P)/N(1 ), "n contrast, in any )i%en #eriod a sin)le $irm
could de$ect &y sli)htly discountin) its #rice, there&y ca#turin) the entire mar*et that
10 3here is a %ery lar)e literature on oli)o#oly theory and su#er)ames, ince our $ocus in on antitrust, .e .ill
dra. on this literature &ut not re%ie. it in any systematic .ay, ee ha#iro 91'8': $or a sur%ey o$ oli)o#oly
theory, "%aldi et al, 92003a:, Motta 92001, ch, 1:, and 5hinston 92000, ch, 2: #ro%ide recent discussions o$ the
a##lication o$ oli)o#oly theory to collusion cases in antitrust,
11 Friedman 91'F1: sho.ed that the $ull cartel outcome can &e su##orted in a re#eated oli)o#oly i$ the #layers
are su$$iciently #atient,
Ch% .1- $ntitrust 110@
#eriod, 3his .ould, &y hy#othesis, condemn all o$ the $irmsKim#ortantly, includin)
itsel$Kto the com#etiti%e outcome $or all $uture time, 3he #ayo$$ to this de$ectin) $irm
is (P) in the current #eriod and 4ero in all su&seCuent #eriods, 6e$ectin) is there$ore
unattracti%e i$ and only i$ (P)/N(1) > (P), that is, i$ and only i$ > 11/N, "$
the #eriods are su$$iciently short, that is, i$ #rice cuts are detected ra#idly enou)h, then
is close to unity and this condition is met, e%en .hen N is lar)e,
3he $reCuency o$ sales, and hence the s#eed o$ detection, is im#licitly &uilt into this
sim#le model, As e/#ected, the $aster that ri%als can learn o$ one $irmGs de$ection and
res#ond &y reducin) their o.n #rices, the easier it is to sustain collusion, Formally,
the model assumes that #rice cuts are o&ser%ed a$ter one #eriod, &ut the len)th o$ time
com#risin) one #eriod has not &een s#eci$ied, "$ one #eriod ta*es time ! , and i$ the
interest rate #er unit o$ time is ", then the discount $actor is = #
"! , 3here$ore, a
lon)er detection la) corres#onds to a lo.er discount $actor, ma*in) it less li*ely that
collusion .ill &e sustaina&le, ceteris pari!us,
3his hi)hly sim#li$ied e/am#le illustrates that there e/ists a su&)ame<#er$ect eCuili&rium
in .hich all the $irms char)e the same #rice, P, re#eatedly, $or anyP >C, as lon)
as detection is su$$iciently ra#id, For e/am#le, i$ the len)th o$ a #eriod is one month,
and .e ima)ine that the interest rate is around 1S #er month, then the discount $actor
is rou)hly 0,'', so the mono#oly #rice can &e sustained as a #er$ect eCuili&rium in the
re#eated oli)o#oly )ame as lon) as the num&er o$ $irms is $e.er than one hundred,
5hile the s#eci$ic calculations here de#end u#on the #articular oli)o#oly )ame &ein)
#layed each #eriod, the &asic ideaKthat $or #lausi&le discount $actors there e/ist
#er$ect eCuili&ria .ith hi)h 9.ell a&o%e com#etiti%e: #rices e%en .ith many $irmsKis
&y no means s#eci$ic to this e/am#le, 2onsider, $or e/am#le, 2ournot oli)o#oly .ith
a homo)eneous #roduct, !n one hand, a $irm that de$ects cannot )ain nearly as much
as in the Bertrand )ame, 93he ratio o$ #ro$its $rom de$ectin) to the #er<#eriod #ro$its
$rom $ollo.in) the eCuili&rium strate)y de#ends u#on the sha#e o$ the demand cur%e,:
!n the other hand, re%ertin) to the static 2ournot eCuili&rium is not as se%ere a #unishment
as re%ertin) to the static Bertrand eCuili&rium, ha#iro 91'8': sho.s that, .ith
= 0.'', the mono#oly #rice can &e sustained in a re#eated 2ournot oli)o#oly .ith
constant mar)inal cost and linear demand as lon) as there are no more than $our hundred
$irms,
FurthermoreKone mi)ht say M.orse yetNKthis e/am#le is hardly anomalous, 3o the
contrary, the M$ol* theoremN $or re#eated )ames tells us that, Cuite )enerally, there e/ists
a #lethora o$ eCuili&ria in re#eated )ames, includin) eCuili&ria that corres#ond to
$ull coo#eration 9the #ro$it ma/imi4in) cartel outcome:, as lon) as the #layers are su$$iciently
#atient, that is, i$ the discount $actor is su$$iciently close to unity, Fuden&er)
and Mas*in 91'80: #ro%ide )eneral conditions under .hich any $easi&le and indi%idually
rational #ayo$$ %ector in the sta)e )ame can &e achie%ed as a #er$ect eCuili&rium in
the re#eated )ame i$ #layers are su$$iciently #atient,12 Fuden&er), ?e%ine, and Mas*in
12 A&reu, ;earce, and tacchetti 91'80, 1''0: consider #unishments stron)er than sim#ly re%ertin) to the oneshot
Bash eCuili&rium, "n the sim#le re#eated Bertrand )ame, no #unishment can &e stron)er than re%ersion
1100 L% 7aplow and C% Shapiro
91''1: e/tend this result to )ames in .hich the #layers do not o&ser%e each othersG
actions &ut only a #u&lic outcome 9such as #rice: that si)nals those actions,
3hese $ol* theorems #ose t.o related and $undamental challen)es $or the analysis o$
oli)o#oly, First, these stron) results do not com#ort .ith the o&ser%ation that ri%als o$ten
com#ete rather than coo#erate and in #articular .ith the #re%ailin) %ie. that collusion is
di$$icult i$ the num&er o$ $irms is moderately lar)eKa %ie. that Farrell 92000: calls the
Mstructural consensus,N 2learly, some im#ortant thin)s are missin) in models o$ re#eated
oli)o#oly that #redict collusi%e outcomes $or a %ery .ide ran)e o$ mar*et structures
and industry conditions, ;ut sim#ly, su#er)ame theory, at least .ith su$$iciently #atient
#layers and .ithout $urther modi$ications, is o$ limited use &ecause it #ro%es too much,
3he economics literature has res#onded to this criticism &y #ro&in) the assum#tions
underlyin) the $ol* theorem and &y e/#lorin) eCuili&rium outcomes .hen the rele%ant
discount $actor is not %ery close to unity, that is, .hen the #layers are not Msu$$iciently
#atient,N "n #articular, a lar)e literature, some o$ .hich .e e/amine in su&section 3,3,
e/#lores the conditions that ma*e it more or less di$$icult to su##ort collusi%e outcomes
in re#eated oli)o#oly,
econd, and o$ #articular rele%ance to antitrust la., there is no e/#licit role $or communications
in the &asic models o$ re#eated )ames, so these models do not hel# us
understand the im#act o$ meetin)s and other communications amon) oli)o#olists, ome
research, ho.e%er, does e/#lore as#ects o$ this limitation, as .e consider ne/t,
4%#%4% Role of communications
!ne o$ the s#eci$ic shortcomin)s o$ standard models o$ oli)o#olistic su#er)ames is that
they do not hel# us understand ho. the $irms initially determine .hich o$ the #lethora
o$ eCuili&ria to #lay, !ne inter#retation o$ these )ames is that the $irms en)a)e in e/tensi%e
communications and #erha#s ne)otiations !efore the )ame &e)ins, in order to
a)ree u#on the eCuili&rium that they .ill #lay, Under this inter#retation, the eCuili&ria
in oli)o#olistic su#er)ames re#resent sel$<en$orcin) outcomes that can arise once an
a)reement is reached, 3he alternati%e e/#lanation o$ the o&ser%ed conduct is that the
$irms someho. $ind their .ay to a relati%ely collusi%e outcome .ithout en)a)in) in any
communications other than throu)h their actions in the mar*et, such as their settin) o$
#rices,
Am&i)uity a&out the role o$ communications is inherent in the standard solution
conce#t o$ Bash eCuili&rium 9and thus #er$ect eCuili&rium, a re$inement o$ Bash eCuili&rium:,
5hen the strate)ies are hi)hly com#le/, and es#ecially .hen there are multi#le
eCuili&ria, the #er$ect eCuili&rium 9or Bash eCuili&rium: methodolo)y does not e/#lain
to the one<shot Bertrand eCuili&rium, since it in%ol%es 4ero #ro$its, .hich &y de$inition is the le%el o$ #ro$its
a $irm can o&tain &y e/itin), For the re#eated 2ournot )ame, ho.e%er, stron)er #unishments are #ossi&le
&y tar)etin) the de$ectin) $irm and #unishin) a)ain a $irm that re$uses to )o alon) #ro#erly .ith its o.n
#unishment, 5hen a%aila&le, these stron)er #unishments ma*e it #ossi&le to sustain the mono#oly #rice e%en
i$ the $irms are some.hat less #atient,
Ch% .1- $ntitrust 110F
ho. the $irms .ere a&le to coordinate to select an eCuili&rium, Iet common sense indicates
that communications can #lay a role in such coordination, and com#le/ strate)ies
su##ortin) a collusi%e outcome .ould seem to constitute e%idence in $a%or o$ the hy#othesis
that the $irms in $act met and reached some sort o$ a)reement at some #oint in
time,
3he literature on Mchea# tal*N as*s .hether communications a$$ect the eCuili&rium
outcome in a )ame<theoretic settin), Farrell and 7a&in 91''0: #ro%ide an e/cellent
o%er%ie. o$ this literature,13 "n )eneral, chea# tal*, that is, communications that do
not directly a$$ect #ayo$$s, can a$$ect eCuili&rium outcomes,11 Farrell and 7a&in )i%e
the e/am#le o$ one 2ournot duo#olist tellin) another= MIou cut your out#ut and "Gll cut
mine,N5hile this mi)ht &e a tric*Kthe s#ea*er )ains i$ the listener cuts out#ut, .hether
or not the s#ea*er does soKthis also mi)ht &e an e$$ecti%e .ay o$ initiatin) out#ut reductions
that sustain collusion in re#eated #lay, "n the end, Farrell and 7a&in conclude
that chea# tal* a&out intended #lay can ma*e a &i) di$$erence .hen the #layersG interests
are .ell ali)ned, &ut the )ains a%aila&le $rom coordination can easily &e lost due to
dis#ute and &ar)ainin) #ro&lems,
A di$$erent strand o$ the literature studies the role o$ communications to con%ey #ri%ate
in$ormation, Athey and Ba).ell 92001, 2000: #ermit the $irms to communicate
a&out #ri%ate cost in$ormation in a re#eated #ricin) )ame, 2om#te 91''8: and Dandori
and Matsushima 91''8: study communications .hen $irms o&ser%e #ri%ate &ut im#er$ect
si)nals a&out #ast #lay, Athey, Ba).ell, and anchirico 92001: study a model .ith
#ri%ate cost shoc*s and #u&licly o&ser%ed #rices,
Facin) the rather am&i)uous theoretical results in the Mchea# tal*N literature re#orted
&y Farrell and 7a&in 91''0:, economists ha%e conducted e/#eriments to learn ho.
communications a$$ect the strate)ies ado#ted &y #layers, 3his literature is sur%eyed &y
2ra.$ord 91''8:, As an early e/am#le, 2oo#er et al, 91'8': $ind that in the M&attle o$
the se/es )ame,N .here the #layers can )ain $rom coo#eration &ut do not a)ree a&out
.hich outcome is &est, communications )reatly increase the chance that the #layers .ill
success$ully coordinate, D\hn 92001: discusses the antitrust im#lications o$ this e/#erimental
literature, em#hasi4in) the role o$ communications in achie%in) coordination &y
reducin) uncertainty a&out .hat other $irms .ill do,
3he critical role o$ communications in sustainin) collusion is re%ealed in the $ascinatin)
study &y Aeneso%e and Mullin 92001: o$ the u)ar "nstitute, a trade association
that o#erated $rom 1'2F to 1'30, 3hey e/amine in detail ho. su)ar re$iners esta&lished
a set o$ rules to $acilitate collusion, 3he u)ar "nstitute e/#erience sho.s ho. .ee*ly
meetin)s amon) su)ar re$iners .ere used to esta&lish and inter#ret rules that en$orced
&usiness #ractices ma*in) #rice<cuttin) more trans#arent, "n contrast to the theories
13 Much earlier, chellin) 91'00: reco)ni4ed the im#ortance o$ communications and discussed the role o$
M$ocal #ointsN in coordinatin) outcomes in strate)ic situations,
11 2ommunications also may ha%e no e$$ect, 3here al.ays e/ists a M&a&&lin)N eCuili&rium in .hich #layers
i)nore the statements made &y others, As em#hasi4ed &y Farrell and 7a&in 91''0:, ho.e%er, many o$ these
eCuili&ria are im#lausi&le,
1108 L% 7aplow and C% Shapiro
descri&ed a&o%e, .hich in%ol%e no cheatin) in eCuili&rium, cheatin) did occur, &ut retaliation
.as care$ully limited, 3his e/am#le illustrates a num&er o$ $unctions ser%ed
&y re)ular communications, $unctions that could not &e ser%ed sim#ly &y initial communications,
7elated lessons can &e $ound in -arrin)ton 92000&:, .ho re#orts on some
t.enty Euro#ean 2ommission cartel cases $rom 2000 to 2001,
4%4% ,ndustry conditions !earin* on the lielihood of collusive outcomes
As noted, to )et &eyond the $ol* theorem, the theoretical literature on oli)o#oly has
e/tensi%ely e/#lored the conditions under .hich the Eoint #ro$it<ma/imi4in) outcome
can &e achie%ed as a #er$ect eCuili&rium in a re#eated oli)o#oly )ame .hen the discount
$actor is not close to unity, !ne .ay this Cuestion is #osed is to as* ho. %arious
$actors a$$ect the critical discount $actor,
, such that the $ully collusi%e outcome can
&e su##orted as a #er$ect eCuili&rium $or
,1@ 5hile a sur%ey o$ the enormous
literature on oli)o#oly theory is .ell &eyond the sco#e o$ this cha#ter, .e mention here
selected results that are es#ecially rele%ant to antitrust, 5e also e/amine a num&er o$
other $actors, not strictly #art o$ the standard oli)o#oly su#er)ame $rame.or*, that &ear
on the $easi&ility o$ collusion,
"n the $irst three su&sections, .e rela/ se%eral e/treme and unrealistic assum#tions
made in the sim#le model used a&o%e, First, the sim#le model assumed that a de$ectin)
$irm could ca#ture the entire mar*et .ith e%en a sli)ht #rice cut, econd, it assumed
that e%en a tiny #rice cut .ould surely &e o&ser%ed &y ri%als, 3hird, it assumed that e%en
the sli)htest de$ection .ould &e #unished se%erely, .ith all $irms #ricin) at mar)inal
cost in #er#etuity, leadin) to 4ero #ro$its $or all $irms, 5e then consider a %ariety o$
other $actors that ma*e it more or less di$$icult $or collusi%e outcomes to &e sustained as
#er$ect eCuili&ria in re#eated oli)o#oly,
4%4%.% Limited *rowth for defectin* firm
3here are many reasons .hy a $irm that de$ects $rom collusi%e #rices may not &e a&le to
ca#ture the entire mar*et, includin) u#.ard slo#in) mar)inal cost 9in the limit, ca#acity
constraints:, customer loyalty, customer s.itchin) costs, and #roduct di$$erentiation,
2learly, i$ the )ains $rom de$ection are limited, collusion .ill &e easier to sustain, ceteris
pari!us,
3o illustrate this idea in the sim#le model introduced a&o%e, su##ose that ca#acity
constraints only #ermit a sin)le $irm to )ro. its sales &y a $actor 1 + $ in a sin)le
#eriod,10 "n the sim#le model a&o%e, 1 + $ = N, &ut i$ a $irm can only, say, dou&le
in si4e in one #eriod, then $ = 1, For no., .e retain the assum#tion that all #rice
1@ "%aldi et al, 92003a: ta*e this a##roach in their accessi&le and in$ormati%e o%er%ie. #a#er,
10 A $ully s#eci$ied model .ould relate this )ro.th limit to underlyin) economic %aria&les, such as ca#acity
or the de)ree o$ #roduct di$$erentiation, and to ma)nitude o$ the de$ectin) $irmGs #rice cut,
Ch% .1- $ntitrust 110'
cuts are detected &y ri%als and #unished stron)ly, leadin) to 4ero $uture #ro$its $or the
de$ectin) $irm, 5ith these assum#tions, o#timal de$ection in%ol%es a tiny #rice cut and
yields #ro$its o$ (P)(1 + $)/N, As a result, char)in) the collusi%e #rice is o#timal so
lon) as >
= $
1+$ , 3he smaller is $, the smaller is
, indicatin) that collusion is
easier to sustain, For e/am#le, i$ $ = 1, then
= 0.@, com#ared .ith
= 0.' in
the case .here N = 10 and a de$ectin) $irm could ca#ture the entire mar*et, -o.e%er,
the
= 0.@ calculation ma*es the assum#tion that a tiny #rice cut &y one o$ ten $irms,
dou&lin) its mar*et share $rom 10S to 20S, .ould surely &e o&ser%a&le &y all o$ the
other $irms,
4%4%#% ,mperfect detection
As em#hasi4ed &y ti)ler 91'01:, Areen and ;orter 91'81:, and much o$ the su&seCuent
literature, .hen one $irm cuts its #rice, its ri%als may not &e a&le to o&ser%e the #rice
cut, "m#er$ect detection unCuestiona&ly ma*es it more di$$icult to sustain collusi%e outcomes,
"n $act, collusion cases in antitrust la. $reCuently re%ol%e around .hether the
$irms ha%e su$$icient a&ility accurately to o&ser%e #rice cuts so as to en$orce a collusi%e
outcome, Additionally, many o$ the $actors considered in su&seCuent su&sections ta*e
on im#ortance &ecause they a$$ect the a&ility o$ the $irms to detect and #unish those .ho
de$ect $rom a collusi%e arran)ement,
3o illustrate the $undamental im#ortance o$ detection, consider ho. the calculus o$
de$ection chan)es i$ there is an e/o)enous #ro&a&ility, %, that the #rice cut is o&ser%ed
&y ri%als,1F 7etainin) our assum#tion that a sin)le $irm can only )ro. its sales &y a
$actor 1 + $ in a sin)le #eriod, the #ayo$$ $rom cuttin) #rice $or a sin)le #eriod is
(P)
N (1 + $) + (1 %)(P)
N

1 ,18 2ollusion is sustaina&le i$ and only i$ this #ayo$$


is less than that $rom maintainin) the collusi%e #rice, (P)
N
1
1 , im#li$yin), collusion
is sustaina&le i$ and only i$ >
= $
%+$ , 3his e/#ression ca#tures a &asic tradeo$$=
collusion cannot &e sustained i$ detection is %ery unli*ely 9lo. %alues o$ %:, es#ecially
i$ a $irm can )ro. ra#idly &e$ore detection .ould ta*e #lace 9lar)e %alues o$ $:, 5ith
$ = 1 and % = 0.2@,
= 0.8, $ar hi)her than
= 0.@ .hen % = 1,
ti)ler 91'01: em#hasi4es the role o$ #rice trans#arency and secret #rice cuttin),
#ence 91'F8: ar)ues that uncertainty a&out demand conditions ma*es it more di$$icult
$or su##liers to distin)uish shi$ts in demand $rom de$ections &y their ri%als, and thus
ma*es collusi%e outcomes more di$$icult to sustain, Areen and ;orter 91'81: deri%e
tri))er strate)ies .hen #rices are only o&ser%ed .ith noise, in .hich case there is a
tradeo$$= enterin) the #unishment #hase more readily #ro%ides a stron)er deterrent &ut
can lead to #rice .ars e%en .hen no $irm has de$ected, -arrin)ton and *r4y#ac4 9200F:
1F "mmediately &elo., .e e/#lain ho. the analysis chan)es i$ the #ro&a&ility o$ detection is a $unction o$ the
#rice char)ed &y the #rice<cuttin) $irm,
18 3his e/#ression assumes, as a&o%e, that the $irm earns 4ero #ro$its in the $uture i$ its #rice cut is detected,
&ut also that i$ the #rice cut is not detected #rom#tly, it is ne%er o&ser%ed,
1110 L% 7aplow and C% Shapiro
study a model in .hich sustainin) collusion in the #resence o$ demand uncertainty
reCuires asymmetric #unishments i$ the $irms o&ser%e each otherGs out#uts &ut not their
#rices,
"n the de$ection calculations #resented so $ar, there .as no reason $or a de$ectin) $irm
to cut its #rice &y more than a small amount since dee#er #rice cuts .ere not needed
$or that $irm to sell either to the entire mar*et 9in our initial model: or to as many
customers as the $irm can ser%e )i%en its ca#acity 9in the modi$ied model:, 3he de$ection
calculations are more com#le/ i$ the $irm<s#eci$ic demand cur%e is such that a de$ectin)
$irmGs #ro$its are decreasin) in its o.n #rice at the collusi%e #rice, in .hich case the
$irmGs immediate #ro$its .ould &e hi)her .ith a discrete rather than incremental #rice
cut, 3his %ariation alone could easily &e accommodated usin) standard #ricin) theory=
the de$ectin) $irm .ould ma/imi4e its immediate #ro$its &y #ricin) at the #oint .here
mar)inal re%enue and mar)inal cost are eCual,
3he analysis &ecomes more com#le/, and more interestin), ho.e%er, i$ .e com&ine
this idea .ith im#er$ect detection, More s#eci$ically, su##ose that dee#er #rice cuts are
more li*ely to &e detected &y ri%als, 3hen o#timal de$ection in%ol%es a tradeo$$= lo.er
#rices lead to hi)her #ro$its in the immediate term &ut a hi)her #ro&a&ility o$ detection,
3he lesson is that the sustaina&ility o$ collusion may de#end in a com#le/ .ay on the
interaction &et.een the a&ility o$ a de$ectin) $irm to )ain customers in the short run
and the a&ility o$ that $irmGs ri%als to detect that the $irm has de#arted $rom collusi%e
&eha%ior,
4%4%4% Credi!ility of punishment
3he sim#le model e$$ecti%ely assumed that it .as credi&le $or the oli)o#olists, as a
)rou#, to #unish any de$ection &y re%ertin) in #er#etuity to the Mcom#etiti%eN outcome,
de$ined as the one<shot eCuili&rium, "n the case o$ a #ricin) )ame .ith homo)eneous
#roducts, this im#lied 4ero #ro$its $or all o$ the $irms, 2learly, this is an e/treme assum#tion,
one that seems to dismiss the tem#tation o$ the $irms to relent on their #unishments
and try a)ain to achie%e a more #ro$ita&le collusi%e outcome,
3o e/#lore the role o$ the #unishment strate)ies, consider $irst the im#ortance o$ the
ma)nitude o$ #unishment, u##ose that, a$ter one $irm de$ects, instead o$ #rices $ore%er
a$ter &ein) set at cost, the other $irms res#ond &y merely matchin) the initial #rice cut,
5ith this assum#tion, a $irm that cuts #rice to Pcut < P earns total #ro$its o$ (Pcut) +

(1)
(Pcut)
N , -o. does this e/#ression com#are .ith (P)
N
+
(1)
(P)
N , the #ro$its $rom
maintainin) the collusi%e #riceW 3he #ro$its $rom de$ection .ill e/ceed the #ro$its $rom
continued collusion $or %alues o$ Pcut near P since cuttin) #rice increases the $irst term
&y a $actor o$ N and has an e$$ect on the second term that %anishes as Pcut P, A$ter
all, the $ull &ene$it o$ de$ectionKca#turin) the entire mono#oly #ro$itKis o&tained .ith
a small #rice cut in the #eriod o$ de$ection, and the $uture #unishment is ne)li)i&le since
the #rice only declines %ery sli)htly $rom the initially collusi%e #rice, 3his analysis
illustrates that threats to merely match #rice cuts are insu$$icient to maintain collusion,
Ch% .1- $ntitrust 1111
econd, consider .hether the #unishment o$ cuttin) #rice to cost in all $uture #eriods
is credi&le, !ne mi)ht thin* that it is &ecause the strate)y consists o$ #layin) the
sta)e<)ame Bash eCuili&rium o%er and o%er a)ain, and &y construction no sin)le $irm
.ould $ind it o#timal to de#art $rom such conduct, )i%en the &eha%ior o$ the other $irms,
-o.e%er, there is somethin) $ishy a&out strate)ies that call $or smooth initial coordination
$ollo.ed &y a #er#etual #rice .ar i$ any one $irm de#arts $rom the a)reed<u#on
#rice, ho.e%er &rie$ the #eriod o$ time, "n #articular, .ould the $irms not &e tem#ted to
relent on the #rice .ar at some #oint and return to coo#erationW 3he tension arises &ecause
the lo)ic o$ #er$ect eCuili&rium does not consider collecti%e de#artures $rom the
#unishment re)ime, But rulin) out collecti%e de#artures is hard to de$end in a theory
that #ostulates $rom the outset that the $irms can $ind a .ay to coordinate to select an
eCuili&rium that is mutually &ene$icial,
!ne sim#le &ut ad hoc .ay o$ dealin) .ith this #oint is to limit the duration o$ #unishments
that are allo.ed a$ter one $irm de$ects &y cuttin) its #rice &elo. the initially
s#eci$ied le%el, "n the sim#le su#er)ame a&o%e, su##ose then that #unishments are limited
to & #eriods, 5ith this limitation, the #ayo$$ to a sin)le $irm $rom de$ectin) is
eCual to (P)+&+1(P)/N(1), .here the $irst term re#resents the #ro$its durin)
the #eriod .hen the $irm de$ects and ca#tures the entire mar*et and the second term
measures the #ro$its this $irm earns once the collusi%e outcome is restored a$ter & #eriods
o$ #unishment 9durin) .hich no #ro$its are earned:, For collusion to &e sustaina&le,
this e/#ression must &e no lar)er than the #ro$its $rom inde$initely char)in) the collusi%e
#rice, P, .hich as &e$ore eCual (P)/N(1 ), 2ontinuin) to use = 0.'', $or
moderate %alues o$ &, this condition im#lies that collusion is sustaina&le .ith u# to
rou)hly N & $irms,1' o, i$ .e thin* o$ one #eriod as corres#ondin) to one month
and i$ .e &elie%e that the $irms can credi&ly enter into a #rice .ar $or one year $ollo.in)
a de$ection, collusion is sustaina&le so lon) as there are no more than t.el%e
$irms,
A much dee#er, &ut $ar more com#le/, .ay o$ addressin) the credi&ility o$ #unishments
can &e $ound in Bernheim and 7ay 91'8': and Farrell and Mas*in 91'8':, 3hese
#a#ers ha%e #ro#osed re$inements o$ the #er$ect<eCuili&rium solution conce#t that rule
out continuation #lay, includin) #unishment strate)ies, that is not collecti%ely credi&le,
At the %ery least, a #er$ect eCuili&rium is reEected i$ the continuation #lay in any
su&)ame is ;areto dominated &y the continuation #lay in any other su&)ame, Usin) the
terminolo)y $rom Farrell and Mas*in, .ho study t.o<#erson )ames, an eCuili&rium that
does not contain one su&)ame that is ;areto dominated &y another su&)ame is called
M.ea*ly rene)otiation #roo$N 957;:, a re$inement o$ the su&)ame<#er$ect eCuili&rium
conce#t, "n the e/am#le a&o%e, the continuation )ame a$ter one $irm de$ected in%ol%ed
1' 6e$ection is unattracti%e so lon) as (P)+&+1(P)/N(1) < (P )/N(1), .hich is eCui%alent
to N(1 ) + &+1 < 1, For %alues o$ near 1, & 1 &(1 ), so this e/#ression is a##ro/imately
N(1)+1(& +1)(1) < 1, .hich can &e .ritten as N 1 < &, "ntuiti%ely, de$ectin) )i%es the $irm
an e/tra N 1 times its #ro$its ri)ht a.ay, .hich is &alanced a)ainst the loss o$ those #ro$its $or & #eriods,
For = 0.'' and moderate %alues o$ &, a )ood a##ro/imation $or the ma/imum num&er o$ $irms is N &,
1112 L% 7aplow and C% Shapiro
all $irms #ricin) at cost $ore%er, "n this su&)ame, all $irms earn a continuation #ayo$$
o$ 4ero, 3his su&)ame is thus ;areto dominated &y the su&)ame consistin) o$ the )ame
itsel$, in .hich each $irm earns a #ayo$$ o$ (P)/N(1), 3here$ore, the #er$ect eCuili&rium
used a&o%e to su##ort the mono#oly #rice is not .ea*ly rene)otiation #roo$,
e%en thou)h it is su&)ame #er$ect, All o$ the $irms .ould #re$er to resume coo#eratin)
rather than carry out the #unishment,
Farrell and Mas*in 91'8': sho. that the 57; condition, .hen a##lied to re#eatedduo#oly
Bertrand and 2ournot )ames, rules out hi)hly asymmetric eCuili&ria, e%en i$
the discount $actor is close to unity, -o.e%er, the 57; condition does not rule out many
eCuili&ria in%ol%in) the mono#oly #rice, includin) the symmetric eCuili&rium in .hich
the t.o $irms s#lit the mono#oly #ro$its, "$ asymmetric #unishments are s#eci$ied that
$a%or one $irm, that $irm .ill &loc* rene)otiations, .hich 57; reCuired to &e a ;areto
im#ro%ement, 3here$ore, the57; conce#t alone does not success$ully resol%e the #arado/
associated .ith the $ol* theorem as a##lied to oli)o#olies, Farrell 92000: su))ests
a $urther re$inement, .hich he calls Cuasi<symmetric 57;, moti%ated &y the notion that
all innocent $irms .ill &e treated symmetrically, 3his conce#t reCuires that all innocent
$irms #re$er to carry throu)h .ith the #unishment o$ a de$ectin) $irm rather than re%ert
&ac* to the ori)inal eCuili&rium strate)ies, Farrell sho.s that mono#oly #rices cannot
&e su##orted, re)ardless o$ the discount $actor < 1, $or moderate num&ers o$ $irms
in Bertrand or 2ournot oli)o#oly, i$ this condition must &e satis$ied, 3his a##roach is
#romisin), &ut relies on the assum#tion that the $irms .ould $ind it di$$icult to esta&lish
#unishments that treat innocent $irms asymmetrically, Further .or* is reCuired &e$ore
these ideas can &e #ut to #ractical use to hel# identi$y industry conditions under .hich
collusion is most li*ely to &e e$$ecti%e .hile accountin) $or the collecti%e credi&ility o$
res#onses to de$ections,@0
4%4%2% &aret structure
5e no. consider a series o$ $actors relatin) to mar*et structure that a$$ect the incenti%e
and a&ility o$ oli)o#olistic su##liers to sustain a collusi%e outcome in re#eated #lay,
4%4%2%.% &aret concentration 2ollusi%e outcomes are less li*ely to occur in industries
.ith more $irms &ecause )reater num&ers ma*e it more di$$icult to satis$y the $irst
$our conditions necessary $or success$ul collusion, 7eachin) consensus is harder .ith
more #arties in%ol%ed, 6etection is more di$$icult since #rice cuttin) &y one small $irm
@0 Mc2utcheon 91''F: e%en su))ests that the herman Act may help $irms collude, statin) 9#, 318:= MAo%ernment
#olicies that are desi)ned to sto# #rice<$i/in) may &ene$it $irms &y ma*in) it .orth.hile $or them to
meet to set u# collusi%e a)reements, .hile ma*in) it costly enou)h $or them to a%oid undesira&le $uture ne)otiations,N
-o.e%er, this %ie. is not su##orted &y the Su*ar ,nstitute case re#orted in Aeneso%e and Mullin
92001:, "n any e%ent, e%idence $rom the #ast decade sho.s clearly that lar)e $ines and tre&le dama)e a.ards
in #rice<$i/in) cases can and do im#ose su&stantial #enalties on $irms en)a)ed in #rice<$i/in), not the .ea*er
sanctions necessary $or Mc2utcheonGs lo)ic,
Ch% .1- $ntitrust 1113
may &e %ery di$$icult $or some or all o$ the other $irms to discern, ;unishment is less
li*ely to &e e$$ecti%e $or t.o reasons, First, a de$ectin) $irm is harder to deter &ecause it
has more to )ain $rom cheatin)= its mar*et<share )ain durin) the #eriod o$ initial de$ection
is li*ely to &e )reater, and its loss $rom #unishment smaller, the more $irms .ere
sharin) in the collusi%e #ro$its, econd, #unishment may &e more di$$icult to coordinate
&ecause o$ the $ree<rider #ro&lem, "nclusion is also harder due to $ree<rider #ro&lems,
as each indi%idual $irm may &elie%e that the others .ill coordinate, .hether or not it
#artici#ates,@1 For #recisely these reasons, one o$ the concerns underlyin) mer)er en$orcement
#olicy is that mer)ers &et.een ri%als that increase concentration can raise the
li*elihood that the remainin) $irms .ill coordinate a$ter the mer)er,
!ur sim#le su#er)ame model .ith homo)eneous )oods and re#eated #rice com#etition
illustrates ho. symmetric collusi%e outcomes are more di$$icult to sustain
.hen the num&er o$ su##liers is lar)er, 5e sho.ed a&o%e in a %ery sim#le model
that the mono#oly #rice could &e su##orted in a #er$ect eCuili&rium i$ and only i$
>
= 1 1/N, 3he lar)er the num&er o$ $irms, the lar)er is
, meanin) that
the $irms must &e more #atient to sustain the collusi%e outcome,
Asymmetries in mar*et shares tend to ma*e it more di$$icult to sustain collusion, For
illustrati%e #ur#oses, su##ose that $irm i has a mar*et share o$ si ,@2 3he condition $or
this $irm to coo#erate rather than de$ect is si(P)/(1) > (P), .hich can &e .ritten
as si > 1, 3his condition .ill &e most di$$icult to meet $or the $irm .ith the smallest
mar*et share, since this $irm has the )reatest tem#tation to )ain share &e$ore the other
$irms can res#ond and also the least #ro$its to lose $rom #unishment that renders all
$irmsG #ro$its eCual to 4ero,@3 6e$inin) smin as the mar*et share o$ the smallest $irm, .e
)et
= 1smin > 11/N, so the $irms must &e more #atient to sustain the collusi%e
outcome,@1 3he smallest $irm #lays the role o$ the ma%eric*, that is, the $irm most #rone
to de$ection $rom the collusi%e outcome,
5hile instructi%e, this sim#le model is una&le to ca#ture the other $actors noted
a&o%e, .hich tend to &e e%en more im#ortant in #ractice, !$ #articular note is the tem#tation
o$ one relati%ely small $irm to decline to #artici#ate in the collusi%e arran)ement
or secretly to cut #rices to ser%e, say, 1S rather than 2S o$ the mar*et, As lon) as #rice
cuts &y a small $irm are less li*ely to &e accurately o&ser%ed or in$erred &y the other
$irms than are #rice cuts &y lar)er $irms, the #resence o$ small $irms that are ca#a&le o$
e/#andin) si)ni$icantly is es#ecially disru#ti%e to e$$ecti%e collusion,
@1 3his #oint assumes that, due to ca#acity limits, risin) mar)inal cost, or other $actors, some de)ree o$
collusion is #ossi&le e%en i$ some $irms do not #artici#ate,
@2 3he analysis here is incom#lete &ecause it does not e/#lain the underlyin) sources o$ the di$$erences in
mar*et shares, 5e address that &elo. .hen .e consider cost asymmetries amon) the $irms,
@3 3he analysis assumes that the smallest $irm is ne%ertheless a&le to ca#ture the mar*etJ i$, ho.e%er, ca#acities
are #ro#ortional to e/istin) shares, the conclusion may not $ollo.,
@1 Bote that, .hen mar*et shares are eCual, smin = 1/N,
1111 L% 7aplow and C% Shapiro
4%4%2%#% Cost asymmetries 3he calculation Eust )i%en .ith smin is incom#lete &ecause
it does not e/#lain why the $irms ha%e di$$erent mar*et shares, A common e/#lanation
is cost asymmetries, and these also ma*e it more di$$icult $or $irms to sustain collusi%e
outcomes,@@
7eachin) consensus is clearly more di$$icult .ith cost asymmetries since there is less
li*ely to &e a $ocal #oint $or #ricin) and since the $irms may .ell disa)ree a&out the
#rice they .ould li*e to see #re%ail, Furthermore, i$ collusion is to ma/imi4e #otential
industry #ro$its, #roduction e$$iciency reCuires that the lo.<cost $irms &e allocated
a )reater share o$ sales, &ut this may reCuire contentious ne)otiations and+or side #ayments,
there&y limitin) some.hat the #otential )ains $rom collusion, 3hese #ro&lems
are e/acer&ated i$ cost in$ormation is #ri%ate, since each $irm may ha%e an incenti%e to
re#resent to the others that its costs are lo. in order to recei%e a hi)her allocation o$
out#ut or other more $a%ora&le treatment,@0
3he con%entional .isdom states that en$orcement o$ the collusi%e outcome is also
more di$$icult .ith cost asymmetries, For e/am#le, "%aldi et al, 92003a: ar)ue that cost
asymmetries hinder collusion, statin) 9#, 30:= Me%en i$ $irms a)ree on a )i%en collusi%e
#rice, lo.<cost $irms .ill a)ain &e more di$$icult to disci#line, &oth &ecause they
mi)ht )ain more $rom undercuttin) their ri%als and &ecause they ha%e less to $ear $rom
a #ossi&le retaliation $rom hi)h<cost $irms,N "%aldi et al, sho. in a sim#le duo#oly e/am#le
that a hi)her
a##lies to the lo.er<cost $irm than to the hi)her<cost $irm, i$ the
$irms di%ide the mar*et eCually, &ecause the lo.er<cost $irm earns #ositi%e #ro$its in the
#unishment #hase, Assi)nin) a lar)er share o$ the mar*et to the lo.er<cost $irm is one
.ay to o%ercome this o&stacle and restore the collusi%e outcome, -o.e%er, allocatin)
a lo.er share to the hi)her<cost $irm necessarily ma*es it more attracti%e $or that $irm
to de%iate $rom collusion, 2olludin) $irms thus $ace a tradeo$$= lo.er<cost $irms can &e
assi)ned lar)er mar*et shares, .hich reduces their incenti%e to de$ect, &ut doin) this
increases the incenti%e o$ the hi)her<cost $irms to de$ect,
3hese issues are e/#lored in )reater detail in Qasconcelos 9200@:, .ho studies re#eated
Cuantity com#etition amon) $irms .ith hetero)eneous Cuadratic cost $unctions,
.here $irms di$$er in their o.nershi# o$ an underlyin) asset that lo.ers the cost $unction,
@F -e sho.s that, in the o#timal collusi%e eCuili&rium, out#ut is shi$ted a.ay $rom
the less e$$icient $irms and to.ard the more e$$icient $irms, "n this eCuili&rium, the less
e$$icient, smaller $irms ha%e the )reatest incenti%e to de#art $rom the collusi%e outcome,
.hile the more e$$icient, lar)er $irms ha%e the )reatest incenti%e to de#art $rom the #unishments
s#eci$ied &y the eCuili&rium strate)ies, -is results are rele%ant $or the analysis
o$ hori4ontal mer)ers since he sho.s ho. a mer)er a$$ects the sco#e $or collusion &y
chan)in) not only the num&er o$ $irms &ut also the distri&ution o$ holdin)s o$ the underlyin)
asset and thus the distri&ution o$ costs amon) the $irms,
@@ ee Mason, ;hilli#s, and Bo.ell 91''2: $or e/#erimental results sho.in) that coo#eration is more li*ely
in a duo#oly i$ the $irms ha%e symmetric costs,
@0 Athey and Ba).ell 92001, 2000: study re#eated oli)o#oly .ith #ri%ate cost in$ormation,
@F ee also 7othschild 91''':,
Ch% .1- $ntitrust 111@
Firms also may di$$er in their cost o$ ca#ital and hence in the discount rate they use
to com#are current and $uture #ro$its, A $irm that is under $inancial #ressure, $or e/am#le,
may ha%e a hi)h discount rate 9lo. discount $actor: and &e es#ecially tem#ted to
de$ect, -arrin)ton 91'8': studied collusion amon) $irms .ith di$$erent discount $actors,
sho.in) a)ain ho. mar*et shares must &e allocated to su##ort a collusi%e outcome, As
another e/am#le, a $irm may &elie%e it Mdeser%esN a )reater mar*et share than it has
historically enEoyed, #erha#s &ecause it &elie%es it .ould )reatly increase its mar*et
share under com#etiti%e conditions, uch ma%eric* $irms may &e es#ecially disru#ti%e
to collusi%e #ricin),
4%4%2%4% Buyer concentration and auction marets 2ollusion is )enerally thou)ht to
&e more di$$icult to sustain in mar*ets .here the &uyin) side is hi)hly concentrated,
A#art $rom the $act that lar)er &uyers may ha%e a more credi&le threat to %ertically inte)rate
u#stream than smaller &uyers, &uyers .ho #urchase a lar)e share o$ the out#ut o$
the colludin) $irms can act strate)ically and internali4e many o$ the &ene$its o$ disru#tin)
collusion, For e/am#le, nyder 91''0: sho.s ho. a lar)e &uyer can strate)ically
accumulate a &ac*lo) o$ un$illed orders to create a &ul)e in demand that can undermine
or desta&ili4e collusion, More )enerally, a lar)e &uyer can strate)ically create %ariations
in demand o%er time, For e/am#le, &y curtailin) #urchases in one #eriod, the &uyer
may lead some or all o$ the su##liers to sus#ect that others ha%e cheated on the #ricin)
a)reement,
Additional strate)ies are a%aila&le $or a &uyer .ho is settin) u# the rules &y .hich
the su##liers .ill &id $or &usiness, Dlem#erer 92002: re#orts enormous %ariation in the
#rices recei%ed in auctions o$ third<)eneration mo&ile tele#hone licenses across di$$erent
Euro#ean countries, ar)uin) that some auction desi)ns $acilitated collusion and thus
led to $ar lo.er #rices &ein) #aid $or these licenses than .as #aid $or other, com#ara&le
licenses, Marshall and Meurer 92001: discuss some o$ the uniCue issues that arise
.hen considerin) collusion in a &iddin) conte/t, includin) a discussion o$ s#ectrum and
tim&er auctions, ar)uin) that collusion is much more di$$icult in sealed<&id, $irst<#rice
auctions than in oral ascendin)<&id auctions, a #oint #ro%en more $ormally in 7o&inson
91'8@:,
4%4%2%2% Collective maret power includin* entry !arriers "$ the $irms ha%e little collecti%e
mar*et #o.er, so they collecti%ely $ace rather elastic demand $or their #roducts,
their incenti%e to collude is corres#ondin)ly lo., 2ollecti%e mar*et #o.er may &e small
&ecause the colludin) $irms are Eust a su&set o$ the incum&ent su##liers, &ecause o$
lo. &arriers to entry into the sale o$ the #roducts the $irms o$$er, or &ecause the #roducts
they sell $ace com#etition $rom close su&stitutes sold &y other $irms, 3he smaller
is the collecti%e mar*et #o.er o$ the $irms that are alle)edly colludin), the closer .ill
&e the $irmsG #rice to the com#etiti%e #rice, and the smaller the dama)es im#osed on
consumers &y e$$ecti%e collusion,
1110 L% 7aplow and C% Shapiro
4%4%2%1% &ulti3maret contact Multi<mar*et contact re$ers to situations .here $irms
interact in more than one mar*et at the same time, Much o$ the literature su))ests that
multi<mar*et interaction tends to ma*e it easier $or the $irms to sustain collusion, 3he
standard re$erence here is Bernheim and 5hinston 91''0:, Bernheim and 5hinston
$irst #ro%e an irrele%ance result= .hen identical $irms .ith constant mar)inal cost meet
in identical mar*ets, multi<mar*et contact does not aid in sustainin) collusion, 5hile
de$ection in one mar*et can &e #unished in other mar*ets, a $irm can sim#ly de$ect
in all mar*ets simultaneously, -o.e%er, Bernheim and 5hinston )o on to sho. ho.
multi<mar*et contact can sustain collusion in many other settin)s, For e/am#le, multimar*et
contact can mute mar*et le%el asymmetries, $or e/am#le, i$ each $irm has a
maEor com#etiti%e ad%anta)e in one mar*et 9.hich could include one )eo)ra#hic area
o$ a sin)le #roduct mar*et:, u##ose, $or e/am#le, that Firm A is the leader in Mar*et A
and Firm B is the leader in Mar*et B, &ut &oth $irms com#ete in &oth mar*ets, Firm A
.ill &e es#ecially tem#ted to de$ect in Mar*et B, .here Firm A has a smaller share,
&ut may &e deterred i$ Firm B .ould res#ond in Mar*et A, Mutual $or&earance may
.ell result, Furthermore, multi<mar*et contact increases the $reCuency o$ interaction,
#ermittin) one $irm to disci#line another more ra#idly than .ould other.ise &e #ossi&le,
3here is some e%idence to su##ort the #ro#osition that multi<mar*et contact ma*es it
easier $or $irms to sustain collusi%e outcomes, "n the airline industry, E%ans and Dessides
91''1: $ind that $ares are hi)her on routes $or .hich the carriers interact on multi#le
routes, "n the mo&ile tele#hone industry, ;ar*er and 7]ller 91''F: $ind hi)her #rices in
mar*ets .here carriers ha%e multi<mar*et contact, 2ramton and ch.art4 92000: loo*
at si)nalin) to su##ort collusion in F22 s#ectrum auctions, .here multi#le auctions $or
licenses .ere conducted simultaneously,
3he %alue o$ multi<mar*et contact in sustainin) collusi%e outcomes is less clear, ho.e%er,
once one accounts $or the noisiness o$ the si)nals that the $irms recei%e re)ardin)
#ossi&le de$ections &y others, Areen and ;orter 91'81: sho. that, .ith noisy si)nals,
limited #unishments are o#timalJ in their model, the $irms re%ert to coo#eration a$ter a
limited #eriod o$ time, 5ith multi<mar*et contact, s#readin) #unishment across mar*ets
may sim#ly not &e desira&le, Eust as en)a)in) in a lon)er #rice .ar, .hile $easi&le,
may not &e o#timal in the Areen and ;orter model, A$ter all, in models .here #unishments
actually occur in eCuili&rium, stron)er #unishments are costly, 3his im#ortant
idea is a&sent $rom the many su#er)ame models in .hich #unishment ne%er ta*es #lace
in eCuili&rium, 3hus, in more realistic models in .hich de$ections and+or #unishments
actually occur, multi<mar*et contact may ha%e no e$$ect on the a&ility o$ the $irms to
collude, 3his %ie. is su##orted &y the Su*ar ,nstitute case descri&ed &y Aeneso%e and
Mullin 92001:J the u)ar "nstitute .as %ery care$ul to cali&rate #unishments to the %iolation
and certainly did not em#loy the ma/imum #ossi&le #unishment, -ad they done
so, the cartel .ould ha%e colla#sed early on, "n $act, the u)ar "nstitute steered a.ay
$rom multi<mar*et lin*a)es, care$ully limitin) #unishment to the same )eo)ra#hic re)ion
.here the %iolation occurred,
Ch% .1- $ntitrust 111F
4%4%1% 'roduct differentiation
3he traditional %ie. in antitrust circles has &een that collusion is easier to sustain amon)
$irms sellin) homo)eneous #roducts rather than hi)hly di$$erentiated #roducts, 7eachin)
consensus should &e easier .hen a)reement only reCuires that one #rice, not many,
&e esta&lished, -o.e%er, there is no com#ellin) theoretical reason to &elie%e that detection
and #unishment are more di$$icult i$ the #roducts are more di$$erentiated, !n one
hand, .ith hi)hly di$$erentiated #roducts, a sin)le $irm that cuts its #rice is li*ely to )ain
relati%ely $e. sales, since many customers .ill still #re$er the other &rands, 3here$ore,
de$ectin) is less attracti%e, !n the other hand, #unishments are .ea*er, since #rice cuts
&y the other $irms ha%e a smaller e$$ect on the #ro$its o$ the de$ectin) $irm i$ #roducts
are hi)hly di$$erentiated, 7oss 91''2: #resents t.o models o$ oli)o#olistic su#er)ames
.ith di$$erentiated #roducts that ca#ture these am&i)uities,
Another reason that collusion is more di$$icult to maintain .hen #roducts are di$$erentiated
is that dimensions o$ com#etition other than #rice and cost cuttin) can ta*e
center sta)e, 2ollusion alon) such dimensions as #roduct desi)n and mar*etin) can &e
%ery di$$icult to esta&lish and sustain, E%en i$ an initial a)reement is reached, the most
tem#tin) .ay to de$ect $rom a collusi%e a)reement may &e to im#ro%e oneGs #roduct or
to e/#and oneGs mar*etin) &ud)et rather than to cut oneGs #rice, 3he $irms may $ind it
%ery di$$icult to restrain com#etition on mar*etin) &ecause o$ the di$$iculty in dra.in)
the line &et.een #ermissi&le and im#ermissi&le mar*etin) acti%ities, ?i*e.ise, collusion
on #roduct desi)n may &e hard to sustain due to the di$$iculties o$ de$inin) .hat
ty#es o$ #roduct im#ro%ements are #ermissi&le and the $act that #roduct im#ro%ements
include an element o$ commitment, tem#tin) $irms to #reem#t their ri%als to ca#ture
more mar*et share on a sustained &asis,
4%4%/% Capacity constraints, e:cess capacity, and investment in capacity
"n many industries, certainly includin) traditional manu$acturin), ca#acity constraints
are an im#ortant as#ect o$ the com#etiti%e en%ironment, "n $act, ca#acity in%estment
decisions can &e the most im#ortant dimension alon) .hich com#etition occurs in the
lon) run, 5e no. address ca#acity decisions and their interaction .ith #ricin) and
out#ut decisions, 5e &e)in .ith a short<run analysis, .hich ta*es ca#acities as $i/ed,
and then )o on to the lon)<run analysis, .hich includes ca#acity in%estment decisions,
4%4%/%.% Collusion on prices with capacity constraints 5e ha%e already o&ser%ed that
collusion on #rices is easier to sustain i$ any sin)le $irm could only )ain limited sales
&y cuttin) its #rice, As .e noted, one reason a $irm may not &e a&le to increase its
sales much &y cuttin) #rice is that the $irm may $ace ca#acity constraints,@8 3here$ore,
@8 A de$ectin) $irm mi)ht &e a&le to rela/ this constraint &y &uildin) in%entories in antici#ation o$ cuttin) its
#rice,
1118 L% 7aplow and C% Shapiro
it .ould a##ear that collusion on #rices is %ery easy to sustain i$ the $irms all ha%e little
e/cess ca#acity,
3his ar)ument, ho.e%er, is seriously incom#lete, Most $undamentally, i$ all $irms are
#roducin) at ca#acity, it is hard to say that they are e$$ecti%ely colludin) on #rices, Full$led)ed
#rice com#etition could not cause #rices to &e lo.er than the le%el at .hich
demand and su##ly .ould &e eCuated )i%en $ull ca#acity utili4ation, E$$ecti%e collusion
on #rices must, there$ore, )o hand<in<hand .ith some de)ree o$ out#ut restriction, that
is, e/cess ca#acity, 5ith this clari$ication, one can as* ho. the #resence o$ ca#acity
constraints a$$ects the analysis already #ro%ided in .hich such constraints .ere a&sent,
;ut di$$erently, does the #resence o$ e/cess ca#acity ma*e it easier or more di$$icult to
sustain collusionW
Follo.in) the literature, .e $rame this discussion in terms o$ $irms that can #roduce at
constant mar)inal cost u# to some .ell<de$ined ca#acity le%el and not &eyond that #oint
9in the short run:, More )enerally, one could study models in .hich each $irm has a
smoothly increasin) mar)inal cost cur%e, 3he resultin) analysis .ould &e considera&ly
more com#le/ &ut lead to similar tradeo$$s and conclusions,
3he e$$ects o$ symmetric ca#acity constraints on collusion are theoretically am&i)uous,
3he )reater is the e/cess ca#acity at each $irm, the more each $irm can )ain &y
de$ectin), -o.e%er, &y the same to*en, )reater e/cess ca#acity means that the other
$irms can e/#and out#ut more to #unish the de$ectin) $irm, "n a #rice<settin) su#er)ame
.ith ca#acity constraints, Broc* and chein*man 91'8@: sho. that collusion is more
di$$icult to sustain in the #resence o$ ca#acity constraints than in their com#lete a&sence,
&ut the relationshi# &et.een
and the #er<$irm e/cess ca#acity at the mono#oly
#rice is not monotonic, ?am&son 91'8F: )enerali4es these results to o#timal cartel #unishment
strate)ies, A&reu 91'80: o&tains similar results $or re#eated Cuantity<settin)
)ames .ith ca#acity constraints,
Bot.ithstandin) these theoretical am&i)uities, in #ractice symmetric ca#acity constraints
may .ell $acilitate collusion, at least in com#arison .ith a situation in .hich
all $irms can #roduce at constant mar)inal cost, A$ter all, a $ar )reater #ercenta)e e/#ansion
o$ out#ut is li*ely to &e needed $or a lone de$ectin) $irm $ully to &ene$it $rom
#rice cuttin) than is needed $or all o$ the $irms to meet the e/#anded demand at the
lo.er com#etiti%e #rice 9es#ecially i$ N is not %ery small:, ;lus, e/#ansion $or the latter
#ur#ose can ta*e #lace o%er time,
Asymmetries per se in ca#acity constraints are li*ely to hinder collusion, More
#recisely, $or a )i%en le%el o$ total ca#acity, collusion is more di$$icult i$ ca#acity is
distri&uted une%enly across the $irms, "$ one $irm has )reater e/cess ca#acity, that $irm
has a )reater incenti%e than others to cut its #rice, and its ri%als ha%e less o$ an a&ility to
disci#line that $irm,@'
@' ee 2om#te, Jenny, and 7ey 92002:, 6a%idson and 6enec*ere 91'81, 1''0:, and ?am&son 91''1, 1''@:,
Ch% .1- $ntitrust 111'
4%4%/%#% Capacity investment decisions "n the lon)er run, the $irms can adEust their
ca#acities, !ne can thin* o$ these ca#acity choices much li*e Cuantity choices and thus
inter#ret the results $rom Cuantity<settin) su#er)ames as a##lyin) to ca#acity choices
o%er time, 3his a##roach is most reasona&le i$ ca#acities are relati%ely short<li%ed, so
there is little commitment %alue associated .ith ca#acity, or i$ the mar*et is )ro.in),
so that $irms are routinely addin) ca#acity, 5ith this inter#retation, the discount $actor
re$lects the time o%er .hich one $irm can o&ser%e other $irmsG ca#acity choices and
res#ond .ith its o.n, ince it ta*es lon)er to chan)e ca#acity than to chan)e #rice, the
discount $actor rele%ant $or ca#acity decisions is lo.er than that $or #ricin) decisions,
ma*in) collusion on ca#acities more di$$icult to sustain, ceteris pari!us, !n the other
hand, the initial ca#acity e/#ansion &y the de$ectin) $irm can itsel$ ta*e time and may
&e di$$icult to hide, ma*in) it hard $or one $irm to )ain much o$ an ed)e on its ri%als
&e$ore they are a&le to res#ond,
-o.e%er, treatin) ca#acity decisions Eust li*e out#ut decisions may $ail to re$lect accurately
some o$ their distincti%e as#ects= ca#acity in%estments tend to &e lum#y and
in%ol%e si)ni$icant sun* costs, 3he irre%ersi&le nature o$ ca#acity choices is em#hasi4ed
&y the literature on #reem#ti%e ca#acity in%estment, .hich #redicts outcomes that
are more com#etiti%e than the static 2ournot eCuili&rium,00 3he commitment as#ect o$
&uildin) ca#acity tends to ma*e collusion on ca#acity more di$$icult= a$ter one $irm adds
ca#acity, it may not &e credi&le $or the other $irms to add ca#acity as .ell, or not as
much as .ould &e needed to deter the initial e/#ansion,
2a#acity choices can interact .ith #ricin) choices o%er time in com#le/ .ays, Benoit
and Drishna 91'8F: sho. that $irms .ill choose to &uild and maintain e/cess ca#acity
to su##ort a collusi%e #ricin) outcome, 6a%idson and 6enec*ere 91''0: study a Msemicollusi%eN
eCuili&rium in .hich the $irms $irst #ic* ca#acities and then #lay a re#eated
#ricin) )ame, settin) #rices at the hi)hest sustaina&le le%el,
4%4%)% &aret dynamics
4%4%)%.% Demand *rowth, demand shocs, and !usiness cycles "n the sim#le models
o$ re#eated #ricin) com#etition, demand )ro.th ma*es collusion easier &ecause a de$ectin)
$irm sacri$ices more in $uture #ro$its in e/chan)e $or a short<term increase in
its mar*et share, ?i*e.ise, i$ demand is declinin), de$ection is more tem#tin), !ne can
easily incor#orate these ideas into the sim#le model #resented a&o%e &y addin) a mar*et
)ro.th $actor,
-o.e%er, these results rely on se%eral assum#tions that may not &e Eusti$ied in the
#resence o$ mar*et )ro.th or decline, First, they assume that de$ection today .ill $ore%er
disru#t collusion and lead to hi)hly com#etiti%e outcomes in #er#etuity,5e already
o&ser%ed that the $irms .ill &e tem#ted to rene)otiate to a%oid this un#leasant outcome,
3he incenti%e to rene)otiate is )reater i$ the mar*et is )ro.in), 5e ha%e also em#hasi4ed
that a##lyin) %ery stron) #unishments is not o#timal in the #resence o$ im#er$ect
00 ee the models o$ t.o<sta)e com#etition in ha#iro 91'8': and the citations therein,
1120 L% 7aplow and C% Shapiro
detection and that #unishments #ro#ortional to the de%iation are attracti%e, 3he lac* o$
#ro#ortionality &et.een todayGs de$ection and #er#etual #unishment is e%en )reater in
the #resence o$ )ro.in) demand, econd, the results i)nore the #ossi&ility that )ro.in)
demand .ill induce ne. $irms to enter the mar*et, 3he #ros#ect o$ $uture entry ma*es it
more tem#tin) to de$ect in the #resent and less %alua&le to maintain coo#eration amon)
the current incum&ents, 3hird, the sim#le model o$ re#eated #rice<settin) does not account
$or the $act that )ro.in) demand may tem#t the $irms to en)a)e in #reem#ti%e
ca#acity additions,
3he lo)ic o$ collusion im#lies that the tem#tation to de$ect de#ends u#on the relati%e
si4e o$ current %ersus e/#ected $uture demand, 3his has im#lications $or short<term demand
shoc*s, .hich are distinct $rom secular )ro.th or decline in demand, 7otem&er)
and aloner 91'80: study a model in .hich demand shoc*s are inde#endently and identically
distri&uted, so demand today con%eys no in$ormation a&out demand in the $uture,
A #ositi%e demand shoc* thus ma*es de$ection relati%ely more attracti%e, 3his same
lo)ic can &e a##lied to collusion o%er the &usiness cycle, -alti.an)er and -arrin)ton
91''1: sho. that collusion is more li*ely to &rea* do.n in the #ortion o$ the &usiness
cycle durin) .hich demand is declinin), Ba).ell and tai)er 91''F: )enerali4e these
results to a model in .hich demand alternates stochastically &et.een &oom and recession
#hases, ;orter 91'83: and Ellison 91''1: a##ly some o$ these ideas to the Joint
Economic 2ommittee, a railroad cartel $rom the 1880s, "n a nice em#irical a##lication,
Borenstein and he#ard 91''0: $ind that retail )asoline mar)ins are hi)her .hen $uture
demand is e/#ected to &e hi)her or $uture costs are e/#ected to &e lo.er, ?astly, .e
stress that .hen demand is un#redicta&le, as in Areen and ;orter 91'81:, collusion is
more di$$icult to sustain &ecause the $irms ha%e )reater di$$iculty distin)uishin) demand
$luctuations $rom cheatin) on the collusi%e a)reement,
4%4%)%#% Disruptive innovation 3he more li*ely it is that the mar*et .ill e/#erience
a disru#ti%e inno%ation, the harder it is to sustain collusion, 3o see this, su##ose that
each #eriod there is some #ro&a&ility, ', that a maEor ne. technolo)ical inno%ation .ill
&e introduced into the mar*et, disru#tin) the collusi%e a)reement, 9A similar analysis
a##lies to other $actors that mi)ht disru#t the a)reement,: For e/am#le, a maEor inno%ation
may disru#t the collusi%e a)reement &ecause it is introduced &y a ne. entrant or
&ecause it introduces such a shar# asymmetry amon) the e/istin) $irms that coo#eration
is no lon)er sustaina&le, u##ose that the inno%ation ends the #ro$it $lo.s $or the incum&ent
su##liers, Under these conditions, the #ayo$$ $rom de$ectin) remains at (P) &ut
the #ayo$$ $rom coo#eratin) is reduced &ecause $uture #ro$its must also &e discounted
&y the #ro&a&ility that disru#tion occurs, Formally, this is eCui%alent to chan)in) the
discount $actor $rom to (1 '), ma*in) collusion more di$$icult to sustain,
4%4%)%4% Switchin* costs, networ effects, and learnin* !y doin* 6e$ection is more
tem#tin) i$ the de$ectin) $irm can )ain a lastin) ad%anta)e o%er its ri%als, either in terms
o$ mar*et share or cost, 5ith consumer s.itchin) costs, at least some o$ the customers
)ained today $rom a #rice cut .ill remain in the $uture e%en i$ #rices $all once the de$ecCh%
.1- $ntitrust 1121
tion has &een o&ser%ed, 2a#turin) customers today has lastin) %alue $or the de$ectin)
$irm= in models o$ com#etition .ith s.itchin) costs, a $irmGs installed &ase is a %alua&le
asset, e%en i$ the $irms com#ete %i)orously to )ain ne. customers, 3he lo)ic here
is similar to that o$ cuttin) #rice .hen customersG demand is hi)h= the de$ectin) $irm
ca#tures more sales &y cuttin) its #rice today, !n the other hand, in the #resence o$
customer s.itchin) costs it can &e more di$$icult to attract customers in the $irst #lace,
2ollusion also can &e di$$icult to sustain in the #resence o$ stron) net.or* e$$ects, at
least i$ the $irms sell incom#ati&le #roducts, "n the clearest case, .here the mar*et is
&ound to ti# to.ard one #roduct standard or another, collusion &et.een incom#ati&le
#roducts is di$$icult to maintain since the $irm that is losin) the standards &attle may
&e %ery tem#ted to en)a)e in #rice<cuttin), or some other tactic, to a%oid enterin) a
do.n.ard s#iral,
A similar dynamic arises in the #resence o$ learnin) &y doin), "$ learnin) is &ased on
cumulati%e out#ut, a $irm that e/#ands its #roduction today .ill e/#erience lo.er costs
tomorro., there&y )ainin) a lastin) ad%anta)e, 6ue to the commitment and #reem#tion
as#ects o$ hi)her current #roduction, a $irm that is more a))ressi%e today ca#tures more
#ro$its in the $uture, ma*in) collusion more di$$icult to sustain in the #resence o$ stron)
learnin)<&y<doin) e$$ects,
4%2% $*reements under antitrust law
4%2%.% +n the meanin* of a*reement
As descri&ed &rie$ly in su&section 3,1, there seems to &e a contrast &et.een the economic
and le)al a##roaches to the re)ulation o$ collusi%e &eha%ior, Under the economic
a##roach, one $irst attem#ts to determine the e/istence o$ collusion and the ma)nitude
o$ its e$$ects and then considers .hich i$ any remedies are a##ro#riate, Under the le)al
a##roach ta*en &y antitrust, the $irst ste# is the determination o$ .hether there e/ists an
a)reement, and, i$ there is, certain le)al sanctions a##ly= in the United tates, these are
tre&le dama)es to inEured customers, criminal #enalties on #er#etrators includin) $ines
and im#risonment, and #ossi&ly inEunctions a)ainst #articular #ractices,
3he e/tent to .hich these a##roaches di%er)e de#ends im#ortantly on the le)al conce#t
o$ a)reement, !ne standard de$initionK$ound in dictionaries and common usa)e
in many conte/tsKis that an a)reement si)ni$ies harmony o$ o#inion or action,01 Under
that strai)ht$or.ard notion, collusion seems nearly synonymous .ith a)reement,
"ndeed, a ty#ical dictionary de$inition o$ collusion is a secret a)reement or coo#eration,
su))estin) $urther that the terms ha%e the same meanin),02
01 3he de$initions throu)hout are ta*en $rom Merriam<5e&sterGs 2olle)iate 6ictionary 910th ed, 1''3:, .ithout
Cuotation mar*s or elli#ses, ometimes other de$initions are listed as .ell,
02 "$ the le)al term Ma)reementN .as inter#reted to reCuire secrecy, then the la. .ould in essence o$$er
a com#lete de$ense .hene%er #rice<$i/ers .ere .illin) to re%eal their #lans, .hich they .ould ha%e e%ery
incenti%e to do i$ that insulated them $rom le)al lia&ility,
1122 L% 7aplow and C% Shapiro
"t seems, ho.e%er, $rom le)al materialsKcourt o#inions, a)ency #ronouncements,
and commentaryKthat the la.Gs notion o$ a)reement is di$$erent, in #articular, narro.er,
Be%ertheless, it has remained some.hat mysterious Eust .hat more is reCuired,
7eturn to the classic e/am#le o$ an undou&ted a)reement= the secret meetin) in a smo*e$illed
room at .hich com#etin) $irms su))est #rices to each other, settle on a #articular
#rice, and indicate their assent to adhere to that #rice, u##ose .e remo%e the smo*e
$rom the room, and then the room itsel$K$or e/am#le, the $irms mi)ht use a con$erence
call or e<mail 9or, as in one antitrust case, enter $ares and sym&ols on a common electronic
airline reser%ation system:, Bo., let us dis#ense .ith the secrecy= #erha#s the
$irms mi)ht s#ea* to each other throu)h seCuential #ress con$erences, At the conclusion
o$ this seCuence, .e ha%e a sort o$ &eha%ior that is o$ten o&ser%ed and is )enerally
considered to &e le)al, that is, not to constitute an a)reement, But .hyW 5hich ste# has
anythin) to do .ith .hether or not the $irms a*reed to anythin)W03 9As already mentioned,
it is the economistGs term, collusion, not the le)al term, a)reement, that o$ten
denotes secrecy,:
As one reads le)al statements on the su&Eect, it a##ears that communication is central
to the inCuiry, A)ain resortin) to standard de$initions, communication re$ers to a #rocess
&y .hich in$ormation is e/chan)ed &et.een indi%iduals throu)h a common system o$
sym&ols, si)ns, or &eha%ior, By that standard, #ress con$erences surely in%ol%e communication,
o does %irtually any other means o$ e$$ecti%e collusion,
2onsider another sim#le e/am#le, "n a some.hat remote area, there are t.o retail
)asoline stations located on o##osite corners o$ an intersection, Each #osts its #rice on
lar)e si)ns readily %isi&le $rom the roadKand, o$ course, $rom the other station, 3he
com#etiti%e #rice is [2,00 and the mono#oly #rice [3,00, !ne can easily construct a
seCuence o$ interactionsK.herein each station o.ner #osts %arious #rices, .aits to see
the otherGs res#onse, then adEusts his or her o.n #rice, and so $orth, 5e .ould #redict
that, e%en i$ neither &ene$ited $rom a $ormal course in )ame theory, they mi)ht readily
settle on a #rice near [3,00, 3he time durin) .hich a de$ector could rea# #ro$its .ithout
res#onse mi)ht &e a matter o$ minutes, not months, -ence, success$ul collusion seems
Cuite li*ely,
3he le)al Cuestion is .hether the t.o o.ners ha%e Ma)reedN to #rice at [3,00, u##ose,
as su))ested, that the le)al system )i%es content to the term a)reement &y as*in)
.hether the #arties communicated .ith each other, 5ell, they did not spea to each
otherJ they may not e%en s#ea* the same lan)ua)e, -o.e%er, in the rele%ant sense, they
did s#ea* to each other in a common lan)ua)e, that o$ #rice, 3he a&sence o$ .ords may
ha%e sli)htly len)thened the time it too* to settle 9a)reeW: on the #rice o$ [3,00, And,
should one station cut its #rice to [2,'0 9in the a&sence o$ any chan)e in mar*et conditions,
such as a dro# in the #rice o$ $uel $rom re$ineries:, the other station o.nerGs Cuic*
03 "t may matter $or other reasons .hether communications are #u&lic, For e/am#le, &uyers may %alue ha%in)
in$ormation sooner, 9-o.e%er, &uyers do not %alue means o$ communication that ma*e collusion a)ainst them
#ossi&le, e%en i$ one conseCuence is that they learn o$ adEustments to collusi%e #rices some.hat sooner,: "n
any e%ent, it is not clear ho. this consideration &ears on .hether there e/ists an a)reement,
Ch% .1- $ntitrust 1123
res#onse, cuttin) its #rice, say, to [2,80, .ill &e #retty unam&i)uousJ it .ill &e understood
as an in%itation to raise #rices, an in%itation that .ould &e acce#ted &y #ostin) a
[3,00 #rice,
E/am#les li*e these seem to su))est that there is little, i$ any, di$$erence &et.een
the le)al reCuirement o$ a)reement and the economistGs notion o$ collusion, Iet it
also seems that $e. thin* that this is actually the case, urely, it is &elie%ed, the la.
reCuires more= more e%idence o$ a)reement, usually throu)h more e%idence o$ communications,
01 Iet, as should no. &e clear, it is hard to tell .hat more is &ein) sou)ht, "t
seems that some di$$erent sort o$ e%idence is reCuired, &ut e%idence o$ .hatW
ome le)al utterances distin)uish &et.een Me/#ressN a)reements and MtacitN a)reements,
3acit ordinarily means that the communication does not use .ords or s#eech
9.hich, &y contrast, is .hat is meant &y e/#ress:, By that de$inition, the #ress con$erences,
&ein) conducted usin) .ords, .ould constitute e/#ress rather than tacit a)reements,
&ut the )asoline station o.ners, usin) si)ns, .ould not &e e/#ress a)reementsK
unless, o$ course, one #ointed out that a si)n sho.in) M[3,00N is $unctionally eCui%alent
to a si)n sho.in) Mthree dollars,N the latter, containin) .ords rather than numerals,
constitutin) an e/#ress rather than tacit communication, ?i*e.ise, one could consider
si)n lan)ua)e, other hand si)nals, .in*s and nods, and so $orth, "ndeed, it is hard to
&elie%e that a sensi&le le)al re)ime .ould ma*e le)alityKand hea%y conseCuencesK
turn on su&tleties o$ modes o$ e/#ression and ta/onomic dis#utes o%er .hich constitute
Me/#ressionsN or Mcommunications,N0@
!$$icial le)al #ronouncements, althou)h sometimes seemin)ly clear, are not that
hel#$ul either, U,, u#reme 2ourt o#inions include $amous statements such as the $ollo.in)=
00 M RU2Vonscious #arallelismG has not yet read cons#iracy out o$ the herman
Act entirely,N0F But this merely indicates that #urely inde#endent actionKsuch as di$$erent
)asoline stations raisin) their #rices in #arallel .hen the #rice o$ oil risesKdoes
01 3he Mstatutory lan)ua)e Uo$ herman Act ection 1V is &road enou)h , , , to encom#ass a #urely tacit
a)reement to $i/ #rices, that is, an a)reement made .ithout any actual communication amon) the #arties to
the a)reement, , , , Be%ertheless, it is )enerally &elie%ed , , , that an e/#ress, mani$ested a)reement, and thus an
a)reement in%ol%in) actual, %er&ali4ed communication, must &e #ro%ed in order $or a #rice<$i/in) cons#iracy
to &e actiona&le under the herman Act,N "n re -i)h Fructose 2orn yru# Antitrust ?iti)ation, 2'@ F,3d 0@1,
0@1 9Fth 2ir, 2002: 9o#inion &y Jud)e ;osner:, ee also the $urther discussion o$ this case in note F2 &elo.,
0@ Bot only do le)al authorities de%ote little attention to de$inin) Ma)reement,N &ut .hen other terms li*e Me/#ressN
are em#loyed, these too are not ela&orated, Furthermore, althou)h one standard de$inition o$ e/#ress is
to re#resent in .ords, other standard meanin)s include to ma*e *no.n 9re)ardless o$ the mode:, to re%eal im#ulses
artistically, and to re#resent &y si)ns and sym&ols, .hich co%ers the $ull )amut, includin) #resuma&ly
most meanin)s that many o$ those .ho use the term Me/#ressN intend to e/clude,
00 As noted earlier, the Euro#ean Union has a similar a)reement reCuirement that li*e.ise e/tends &eyond
$ormal contracts and is im#recise, "n Dyestuffs, the Euro#ean 2ourt o$ Justice ela&orated the conce#t o$ a
concerted #ractice as Ma $orm o$ co<ordination &et.een underta*in)s .hich, .ithout ha%in) reached the state
.here an a)reement #ro#erly so called has &een concluded, *no.in)ly su&stitutes #ractical co<o#eration &et.een
them $or the ris* o$ com#etition,N "2" ?td, %, 2ommission, 2ase 18+0' U1'F2V E27 01', ^01, Althou)h
some sort o$ contact &et.een the #arties seems to &e reCuired, the 2ommission seems inclined to $ind &eha%ior
ille)al e%en .hen the contact is indirect, ee Bellamy and 2hild 92001:,
0F 3heatre Enter#rises %, ;aramount Film 6istri&utin) 2or#,, 310 U,, @3F, @11 91'@1:,
1121 L% 7aplow and C% Shapiro
not constitute an a)reement, !r consider= M3he essential com&ination or cons#iracy in
%iolation o$ the herman Act may &e $ound in a course o$ dealin) or other circumstances
as .ell as in an e/chan)e o$ .ords, , , , UA cons#iracy may &e $ound .hereV the
cons#irators had a unity o$ #ur#ose or a common desi)n and understandin), or a meetin)
o$ the minds in an unla.$ul arran)ement, , , , N08 3his 9#artially Cuestion<&e))in):
e/#ression ali)ns su&stantially .ith the idea that success$ul collusion is su$$icient, More
recently 9and more commonly Cuoted in modern cases:, the u#reme 2ourt has stated
that e%idence must &e #resented M Rthat tends to e/clude the #ossi&ilityG that the alle)ed
cons#irators acted inde#endently , , , Uthat is,V that the in$erence o$ cons#iracy is reasona&le
in li)ht o$ the com#etin) in$erenceUV o$ inde#endent action, , , , N0' -ere, the
inter#retation de#ends on the meanin) o$ Minde#endent,N "$ ta*en to mean M.ithout re)ard
to others,N then collusi%e &eha%ior is not inde#endent action and thus is su$$icient
to tri))er lia&ility, Iet another #ronouncement 9in a more recent case, &ut not one directly
addressed to the a)reement Cuestion: is that MUtVacit collusion, sometimes called
oli)o#olistic #rice coordination or conscious #arallelism UisV not in itsel$ unla.$ul,NF0
3acit collusion, ho.e%er, is unde$ined and is not )enerally understood to &e the same as
conscious #arallelism, 5hat all o$ these court decisions and most other statements ha%e
in common is that *ey terms are not de$ined, the su&Eect is not directly discussed in any
de#th 9that is, $or more than a #ara)ra#h:, and no rationale is o$$ered $or deemin) one
set o$ scenarios to &e le)al and another ille)al,
A $urther im#ortant com#lication is that it is .ell acce#ted that, .hate%er is reCuired
to esta&lish an a)reement, it is allo.a&le 9and ty#ical: $or the demonstration to &e indirect,
throu)h circumstantial e%idence, o<called Msmo*in) )unsN are not reCuired, For
e/am#le, i$ the la. demands #roo$ o$ direct %er&al communications on the s#eci$ic #rice
and #attern o$ #unishment, it mi)ht &e ar)ued that near<simultaneous #rice increases,
and then declines in res#onse to de$ections, are e%idence o$ such communications and
hence su$$icient to esta&lish a %iolation, 3he im#licit lo)ic is, M-o. else could this &eha%ior
&e e/#lainedWN 3his #er#le/in) Cuestion and some o$ the earlier discussions on
the #ossi&le meanin)s o$ a)reement reCuire $urther attention to the role o$ communications
in the economic theory o$ collusi%e &eha%ior,
4%2%#% $*reement, economics of collusion, and communications
u##ose, as seems to &e &elie%ed &y most, that the le)al reCuirement o$ an a)reement
is satis$ied only &y certain ty#es o$ communication= #erha#s %er&al statements or close
eCui%alents, su$$iciently directed at com#etitors, that relate closely to #ricin) &eha%ior,
and that may &e res#onded to reasona&ly #rom#tly, #recisely, and directly,5hat, then, is
08 American 3o&acco 2o, %, United tates, 328 U,, F81, 80'O10 91'10:,
0' Matsushita Electric "ndustrial 2o, %, Yenith 7adio 2or#,, 1F@ U,, @F1, @88 91'80:, Cuotin) Monsanto 2o,
%, #ray<7ite er%ice 2or#,, 10@ U,, F@2, F01 91'81:,
F0 Broo*e Arou# ?td, %, Bro.n & 5illiamson 3o&acco 2or#,, @0' U,, 20', 22F 91''3:,
Ch% .1- $ntitrust 112@
the relationshi# &et.een these more e/#licit sorts o$ communication and the economic
theory o$ collusionW
7e%ie.in) the economic theory o$ collusion as summari4ed in su&sections 3,2
and 3,3, communication may &e rele%ant at a num&er o$ #oints, First, the nature o$
communications may &ear on the ease o$ reachin) consensus, "n the e/am#le .ith
t.o )asoline stations, rather sim#le communications seem su$$icient, But i$ there are
more $irms, )reater hetero)eneity 9in costs, #roducts, or other $eatures:, more uncertainty
a&out &uyer &eha%ior, or other com#licatin) $actors, )reater ne)otiation may &e
reCuired, .hich in turn mi)ht &e $acilitated &y more e/#licit 9direct, head<to<head, simultaneous,
#rolon)ed: communications, 3his %ie. is not entirely o&%ious, ho.e%er,
$or i$ all #arties *ne. that they .ere limited to a $e. rounds o$ sim#le #rice su))estions,
a$ter .hich they must ha%e reached a)reement, it is #ossi&le that a)reements .ould &e
reached more Cuic*ly and .ith )reater li*elihood 9althou)h #erha#s they .ould also &e
less dura&le, due to misunderstandin)s:, 5hile this discussion is lar)ely outside sim#le
models o$ re#eated oli)o#oly, .hich ty#ically as* .hether a #rice P can &e sustained,
these Cuestions are addressed in the literature on Mchea# tal*,N cited a&o%e,
econd, in the detection o$ cheaters, e/#licit, detailed communication mi)ht also &e
hel#$ul, "$ $irm AGs cheatin) is noticed &y $irm B, $irm B could tell others, 2om#te 91''8:
and Dandori and Matsushima 91''8: address this #ossi&ility in a model .here $irms
o&ser%e and can communicate their #ri%ate in$ormation a&out #ast #lay, Alternati%ely, i$
other $irms sus#ect that $irm A is cheatin), discussions .ith $irm A 9#erha#s su##orted
&y $irm A #resentin) ori)inal in%oices or other in$ormation: mi)ht hel# clear u# the
matter, a%oidin) #rice .ars due to mista*en in$erences, 3he u)ar "nstitute o#erated
%ery much in this manner, as descri&ed &y Aeneso%e and Mullin 92001:, !li)o#oly
theory is slo.ly mo%in) more in the direction o$ modelin) these ty#es o$ issues, at least
&y e/#lorin) the role o$ communications a&out #ri%ate cost in$ormation, ee Athey and
Ba).ell 92001, 2000:,
3hird, #unishment mi)ht &etter &e coordinated .ith more e/#licit communication,
6eterminin) the ma)nitude o$ the #rice cut and its duration, #erha#s $ocusin) #unishment
.hen $irmsG #roduct lines and re)ions o$ o#erations %ary, and other as#ects
o$ strate)y mi)ht &e .or*ed out more e$$ecti%ely, As .ith reachin) consensus, ho.e%er,
)reater o##ortunity $or detailed communication may &e a dou&le<ed)ed s.ord, As
noted, the o##ortunity $or rene)otiation can undermine #unishment, "n any case, it is
)enerally assumed in $ormal models that some #articular #unishment strate)y has &een
chosen and .ill &e #ursuedJ the Cuestion e/#lored is .hether the strate)y, i$ #ursued,
.ould deter cheatin) e/ ante, or .hether the strate)y is credi&le, not .hat communications
may &e necessary to select or e$$ectuate the strate)y,
Fourth, inclusion mi)ht &e enhanced throu)h detailed ne)otiations, 3his consideration
is &ased on reasonin) similar to that o$ reachin) consensus and is li*e.ise outside
standard $ormal analysis,
"n all, there are many reasons to &elie%eKand it )enerally is &elie%edKthat )reater
o##ortunity $or $reer, more detailed, e/#licit communication tends to $acilitate collusion
9althou)h there are some counter%ailin) $actors:, "$ this is indeed the case, it $ollo.s that
1120 L% 7aplow and C% Shapiro
it is more im#ortant to #rohi&it more e/#licit $orms o$ communication, till, the Cuestion
remains, .hy not #rohi&it all communicationW 3he ans.er must &e that %arious
$orms o$ communicationKsuch as ma*in) #rice in$ormation a%aila&le to customersK
ser%e other, le)itimate #ur#oses, and that less e/#licit communicationsKsuch as sharin)
a))re)ated and la))ed sales in$ormation throu)h a trade associationKare more li*ely
to #romote socially %alua&le $unctions than to $acilitate collusion, 3his statement, too,
is not o&%ious, $or many socially %alua&le $unctions, such as the settin) o$ com#ati&ility
standards $or emer)ent technolo)ies or the sharin) o$ in$ormation a&out industry
conditions, reCuire hi)hly e/#licit communication, Furthermore, as the case o$ the t.o
)asoline stations illustrates, in some instances $acilitatin) collusion reCuires %ery little
e/#licit communication,
Additionally, directin) the le)al inCuiry at the nature o$ communicationsK.hich
themsel%es o$ten cannot &e o&ser%ed &y the tri&unal &ut must &e in$erred $rom circumstantial
e%idenceKraises .hat mi)ht &e called a #arado/ o$ #roo$, u##ose a%aila&le
e%idence indicates that, in the situation under scrutiny, collusion is es#ecially easy and
the dan)er o$ su#ra<com#etiti%e #ricin) is accordin)ly %ery hi)h, Moreo%er, e%idence
conclusi%ely demonstrates that .e ha%e e/#erienced a collusi%e outcomeKat rou)hly
the mono#oly #riceK$or years, 5ho .insW Ar)ua&ly, the de$endants, 3hey could ar)ue
that, #recisely &ecause collusion is so easy, they .ere a&le to achie%e mono#olistic
resultsKand, they )lee$ully concede, .ill &e a&le to continue to do so $or the $oreseea&le
$utureK.ithout any meetin)s in smo*ed<$illed rooms, ela&orate ne)otiations, and
so $orth, Just a $e. #u&lic #ricin) si)nals and they .ere o$$, Moreo%er, since all ha%e
ta*en courses in strate)y at &usiness school and all are ad%ised &y the leadin) consultin)
$irms and their a$$iliated )ame theory e/#erts, coordinatin) #unishment .ith only
minimal, indirect communications is a sna#, -ence, the %ery stren)th o$ the e%idence o$
the ease and success o$ collusion ma*es it im#lausi&le to in$er that the de$endant $irms
must necessarily ha%e met and had lon) discussions a&out #rice<$i/in),
7e$lectin) on this case and other #ossi&ilities, it .ould seem that the relationshi#
&et.een the ease o$ collusion and the li*elihood that there .ere su$$iciently e/#licit
communications to tri))er lia&ility under the a)reement reCuirement 9.hate%er it turns
out to &e: is not monotonic, ;ut di$$erently, .e are as*in) Eust ho. should the $actors
listed a&o%e, .hich ma*e it more or less di$$icult to sustain collusi%e outcomes, &e
incor#orated into a #rice<$i/in) case in .hich the e/istence o$ an a)reement is #ro%ed
throu)h circumstantial e%idence,
Be)innin) at one end o$ the s#ectrum, su##ose that industry conditions are such that
it is e/tremely di$$icult $or the $irms to sustain a collusi%e outcome &ecause there are
many $irms, lo. entry &arriers, #rice<cuttin) &y one $irm is %ery di$$icult $or ri%als to
o&ser%e, and demand and cost are hi)hly %aria&le, Under these industry conditions,
.e .ould not e/#ect the $irms to ha%e en)a)ed in uno&ser%ed meetin)s that satis$y
an e/#licit communication reCuirement sim#ly &ecause such meetin)s .ould li*ely &e
$utile, Moreo%er, i$ it is ne%ertheless asserted that collusion occurred, .e Eust .ill not
&elie%e that an e$$ecti%e #rice<$i/in) a)reement .as reached, 3he e%idence on #ricin)
and cost could not ha%e &een certain, and any uncertainty is naturally resol%ed a)ainst
Ch% .1- $ntitrust 112F
an in$erence o$ collusion &ecause it .ould &e nearly im#ossi&le under the o&ser%ed
industry conditions,
2onsider ne/t an industry in .hich conditions are such that collusion is some.hat
easier to sustain, #erha#s &ecause the industry is more concentrated, has moderate entry
&arriers, #ricin) is more trans#arent, and demand and cost are less %olatile, 3hese industry
conditions ma*e it more li*ely, &ut still $ar $rom ine%ita&le, that a collusi%e outcome
could arise, Under these circumstances, i$ collusion indeed seems to ha%e occurred, its
%ery di$$iculty 9&ut not such hi)h di$$iculty as to &lend to.ard im#ossi&ility: su))ests
that e/#licit communications may .ell ha%e &een em#loyed to carry it o$$,
3o.ard the o##osite end o$ the s#ectrum, consider an industry in .hich the conditions
are hi)hly conduci%e to collusion= hi)hly concentrated, no #ros#ect o$ entry, trans#arent
#ricin), and sta&le demand and cost, 3hin* o$ the t.o )as stations, 5e no. ha%e the
case .ith .hich .e &e)an, #resentin) the #arado/ o$ #roo$= the %ery ease o$ collusion
ne)ates the in$erence that there must ha%e &een ela&orate, e/#licit communications,
"n sum, as industry conditions mo%e $rom those that ma*e collusion nearly im#ossi&le
to those that ma*e it incredi&ly easy, the in$erence that there must ha%e &een hi)hly
detailed communications $irst &ecomes stron)er and then .ea*er, "t is rather hard to say
.here on this continuum the ma/imum in$erence arises, or in .hat intermediate ran)e
some )i%en #roo$ standard is satis$ied,F1
-o. does this #arado/ o$ #roo$ sCuare .ith the la. and .hat .e o&ser%e in #racticeW
U,, courts ty#ically insist on the #resentation o$ %arious so<called M#lus $actors,N Iet
these $actors are o$ten little more than indicators that collusion rather than #urely inde#endent
&eha%ior is li*ely to ha%e occurred,F2 As Eust e/#lained, such $actors indeed
F1 3he more one #ushes the lo)ic underlyin) the in$erence o$ a)reement, the more com#le/ it &ecomes, For
e/am#le, in the re)ion in .hich collusion is moderately di$$icult, a sli)ht increase in the ease o$ collusion
ma*es it more li*ely that collusion .as attem#ted, .hich raises the li*elihood o$ a )i%en ty#e o$ e/#licit communication,
&ut, conditional on collusion ha%in) &een attem#ted, reduces the li*elihood that communication
.as more e/#licit &ecause, &y hy#othesis, collusion is &ecomin) easier, 3he de#iction in the te/t, .hich assumes
a sin)le #ea*, may &e o%erly sim#listic, Moreo%er, one su##oses that di$$erent industry conditions in
di$$erent com&inations that contri&ute to the ease or di$$iculty o$ collusion may ha%e %aryin) e$$ects on the
need $or more e/#licit communication and the $orms that it .ill ta*e,
F2 ome o$ the most common $actors seem to )o little &eyond reCuirin) interde#endent rather than inde#endent
&eha%ior, For e/am#le, #rominent #lus $actors include %arious sorts o$ e%idence sho.in) that the
$irmsG actions are Ma)ainst sel$<interestN in the a&sence o$ collusion, For a sur%ey and critical commentary,
see Areeda and -o%en*am# 92002, %ol, 0, 211O2@0:, 2ourts also $reCuently rely on e%idence that #ur#orts to
directly indicate the e/istence o$ an a)reement, For e/am#le, Jud)e ;osner in "n re -i)h Fructose 2orn yru#
Antitrust ?iti)ation, 2'@ F,3d 0@1, 002 9Fth 2ir, 2002: o$$ers, amon) others, the $ollo.in) Cuotations $rom
the alle)ed cons#irators as e%idence o$ the e/istence o$ the reCuisite a)reement= M5e ha%e an understandin)
.ithin the industry not to undercut each otherGs #rices,N MU!Vur com#etitors are our $riends, !ur customers are
the enemy,N A com#etitorGs #resident is called a M$riendly com#etitorN and mention is made o$ an Munderstandin)
&et.een the com#anies that , , , causes us to , , , ma*e irrational decisions,N As the discussion in the te/t
e/#lains, ho.e%er, there can e/ist such an Munderstandin)N and $irms can %ie. com#etitors coo#erati%ely as
a result o$ education a&out collusion, )ood ad%ice, common sense and e/#erience, and o#en communications
9such as the )as stationsG #ostin) o$ #rices:, so it is di$$icult to discern in .hat sense more than the e/istence
o$ consciously interde#endent, collusi%e interaction is reCuired,
1128 L% 7aplow and C% Shapiro
$a%or the in$erence o$ an ille)al a)reement i$ &ut only i$ .e are on the Mdi$$icultN side
o$ the ma/imum, .here additional e%idence indicatin) the ease or &ene$its o$ collusion
ma*es the li*elihood o$ the reCuisite communications hi)her, !n the other side o$ the
ma/imum, they ma*e collusion more li*ely &ut e/#licit communications less li*ely,
5e are una.are o$ any cases 9nor ha%e .e e%er heard anyone su))est the e/istence
o$ any cases: in .hich #arties and courts acted as i$ they .ere in .hat .e are re$errin)
to as the #arado/ re)ion, that is, #ast the #ea*, such that e%idence that collusion is
more $easi&le ma*es the in$erence o$ detailed communications less li*ely, -o. can
one e/#lain this one<sidednessW !ne #ossi&ility is that, e%en thou)h the la. has had
this character $or o%er hal$ a century in the United tates 9and $or shorter, althou)h
si)ni$icant, #eriods in many other Eurisdictions:, no one has really understood the nature
o$ the le)al reCuirement,
Another #ossi&ility is that all cases are in $act at the di$$icult side o$ the ma/imum,
3hat is, there are no industries .here success$ul collusion is at all li*ely in the a&sence o$
hi)hly e/#licit communications, !&ser%e, ho.e%er, that i$ this .ere true, the a)reement
reCuirement .ould &e su#er$luous, 3hat is, i$ there e/ists collusion, there must ha%e
occurred the reCuisite communications to tri))er the a)reement reCuirement, 5ere this
al.ays true, nothin) .ould need to &e #ro%ed &eyond the mere e/istence o$ collusion,
93his .ould su))est that ;osnerGs a$orementioned #rescri#tion .ould &e im#lied &y
e/istin) la., and thus not constitute a si)ni$icant de#arture $rom it,:
Iet another #ossi&ility is that there are cases #ast the ma/imum, in the #arado/ ran)e,
&ut de$endants are reluctant to ad%ance the ar)ument that the #roo$ a)ainst them im#lies
the a&sence o$ any a)reement and hence %ictory, 3he reason is that, in concedin) that
collusion is easy, li*ely, and #ro&a&ly in $act has occurred and .ill continue, they $ear
that they .ill hurt their case, 6e$endants may su$$er in the determination o$ lia&ility
&ecause, as a #ractical matter, a $act $inder 9.hether a Eury, Eud)e, or e/#ert tri&unal: is
more li*ely to condemn them i$ they in $act o#erate in a situation inherently conduci%e
to collusi%e outcomes and are li*ely ta*in) ad%anta)e o$ it, 3hey .in on the $ormal la.
&ut lose &ecause they sho. themsel%es to &e )reedy and &eha%in) in an antisocial manner,
"n that e%ent, it may &e that de facto, the )reater the dan)er o$ collusion, the )reater
the li*elihood o$ lia&ility, .ithout re)ard to any in$erence that does or does not $ollo.
a&out e/#licit communications and the satis$action o$ the a)reement reCuirement, Additionally,
i$ there is a su$$icient #ros#ect that lia&ility .ill &e $ound, de$endants may &e
.orried a&out #enalties, 3he more they ar)ue that collusion is easy, the more #lausi&le
.ill &e hi)h estimates o$ o%erchar)es 9in amount and duration: and thus the )reater .ill
&e $ines and dama)e #ayments,
3his discussion may raise more Cuestions than it ans.ers, &ut .e &elie%e that it is,
ultimately, clari$yin), 3he economic analysis o$ collusion, althou)h Cuite com#le/, is
at least $airly strai)ht$or.ard in statin) the Cuestion it addresses and the moti%ation $or
the inCuiry it underta*es, U#on e/amination, the same cannot &e said a&out the la.Gs
reCuirement o$ an a)reement and the role o$ industry conditions in in$errin) that such
an a)reement e/ists, 5e ho#e to ha%e ad%anced understandin) in t.o .ays= &y &ein)
more #recise a&out .hat a)reement mi)ht mean, and, $or a )i%en de$inition, &y &ein)
Ch% .1- $ntitrust 112'
more e/#licit a&out the relationshi# &et.een the economics o$ collusion and .hether
such a)reement reCuirement is satis$ied,
3.o additional o&ser%ations a&out the inter#lay &et.een the economics o$ collusion
and antitrust la. are in order, First, under antitrust la. it is #ossi&le $or there to &e
a %iolation e%en .hen it is clear that no success$ul collusion occurred, "$ com#etitors
meet $ormally, enter into a .ritten a)reement, &ut ultimately $ail misera&ly in e/ecutin)
it, most le)al re)imes .ould $ind a %iolation, "$ $ines or dama)e a.ards .ere limited
to a multi#le o$ the o%erchar)e, this $indin) .ould &e moot, -o.e%er, other sanctions
may &e em#loyedJ nota&ly, those en)a)ed in the attem#t may &e #ut in #rison, "t is
sometimes e$$icient to #unish unsuccess$ul attem#ts 9es#ecially .hen detection is di$$icult
and limits on sanctions may ma*e it im#ossi&le to #unish %iolators su$$iciently to
achie%e e$$ecti%e deterrence:, and e/aminin) direct communication may hel# to identi$y
unsuccess$ul attem#ts, !$ course, e%idence a&out #ricin) #atterns o$ the sort that mi)ht
&e deci#hered &y economic e/#erts may also aid in the tas*, es#ecially i$ there .ere
e$$orts to #ut the a)reement into e$$ect,
econd, se#arate $rom the a)reement reCuirement, #enalties may de#end su&stantially
on the e/tent and duration o$ o%erchar)es, Underta*in) these measurements reCuires
e/#ert economic analysis, 3he )reatest di$$iculty, o$ course, is in determinin)
.hat .ould ha%e &een the #rice &ut $or the collusion, "t is necessary &oth to s#eci$y
conce#tually the nature o$ the eCuili&rium that .ould other.ise ha%e #re%ailed 9#er$ect
com#etitionW mono#olistic com#etition in #rice .ith di$$erentiated #roductsW: and
to calculate Eust .hat #rice .ould ha%e #re%ailed in that eCuili&rium,F3 3his inCuiry is
%ery closely related 9in some res#ects, identical: to that necessary to identi$y .hether
collusion e/isted in the $irst #lace,
4%1% +ther hori9ontal arran*ements
!ur analysis has $ocused almost entirely on collusion that in%ol%es arran)ements #urely
concerned .ith the $i/in) o$ #rices, im#le #rice<$i/in), in turn, is unam&i)uouslyK
M#er seNKille)al in the United tates and su&Eect to similar #rohi&itions else.here,
3here e/ists, ho.e%er, a %ariety o$ hori4ontal entitiesK#artnershi#s, trade associations,
Eoint %entures, standard<settin) &odies, to name a $e.Kand such entities en)a)e in myriad
$orms o$ conduct,
2ertain hori4ontal arran)ements can ser%e as su&stitutes $or direct #rice<$i/in), As
noted, $irms mi)ht a)ree to di%ide territories or customers so as to eliminate com#etition,
Althou)h no #articular #rice has &een set, each $irm is le$t to act as a mono#olist .ith
res#ect to its #ortion o$ the mar*et, so the result is similar to that o$ a #rice<$i/in)
cartel, 3he economic analysis is analo)ous= $irms must &e a&le to a)ree on the mar*et
F3 "$ dama)es .ere &ased not on the o%erchar)e times the Cuantity #urchased, as is ordinarily the case, &ut
instead or also on losses o$ consumer sur#lus re)ardin) units not #urchased, in$ormation on the entire rele%ant
se)ment o$ the demand cur%e .ould &e reCuired,
1130 L% 7aplow and C% Shapiro
allocation 9instead o$ the #rice:, cheatin) 9sellin) to other $irmsG allotted customers:
must &e detecta&le and su&Eect to e$$ecti%e #unishment, $irms .ith si)ni$icant ca#acity
need to &e included in the a)reement, and entry must &e limited, ?i*e.ise, le)al scrutiny
tends to &e similar= #ure hori4ontal di%isions o$ the mar*et amon) com#etitors are also
per se ille)al in the United tates,
Bot all hori4ontal arran)ements in%ol%e #ure schemes to $i/ #rices or di%ide the mar*et
9o$ten called Mna*ed restraintsN:, Be%ertheless, many hori4ontal arran)ements #ose
some ris* to com#etition, Accordin)ly, antitrust la.s need to dra. distinctions, Under
the la. in the United tates, this is done under the ru&ric o$ the Mrule o$ reasonN=
reasona&le schemes are #ermissi&leJ unreasona&le ones are #rohi&ited, !&%iously, this
conce#t needs to &e $leshed out and related to economic analysis, and that tas* .ill &e
our central $ocus in this section, But $irst .e .ill consider a #articular class o$ hori4ontal
arran)ements that is closely related to our $ore)oin) discussion o$ collusion and the
le)al #rohi&ition on #rice<$i/in),
4%1%.% ;acilitatin* practices
"n our consideration in su&section 3,3 o$ conditions &earin) on the li*elihood o$ success$ul
collusion, .e lar)ely too* such conditions to &e e/o)enous, ome $actors, ho.e%er,
are .ithin the $irmsG control, indi%idually or collecti%ely, Antitrust scrutiny has $ocused
#rimarily on the latter,F1 "n this re)ard, t.o lines o$ attac* must &e distin)uished, First,
is hori4ontal a)reement on some #ractice that $acilitates collusion itsel$ an ille)al a)reement
in restraint o$ trade and thus an inde#endent &asis $or lia&ilityW Most challen)es,
and our o.n discussion, em#hasi4e this inCuiry, econd, does the use o$ $acilitatin)
#ractices constitute e%idence o$ the e/istence o$ an underlyin) a)reement directly to
$i/ #ricesW "n some res#ects, the distinction may &e immaterial, nota&ly i$ &oth a)reements
on the $acilitatin) #ractice itsel$ and a)reements on #rice<$i/in) are ille)al and
i$ the remedy is the same, 9A remedial di$$erence is that a $acilitatin) #ractice mi)ht
inde#endently &e enEoined,: Be%ertheless, in the conte/t o$ e%aluatin) the e%idence in a
#articular case, it clari$ies thin*in) to *ee# this di$$erence in mind,
An im#ortant $acilitatin) #ractice that has lon) &een the su&Eect o$ antitrust re)ulation
concerns in$ormation e/chan)es amon) com#etitors, sometimes in the conte/t o$
trade association acti%ity and other times conducted inde#endently,F@ For e/am#le, in
F1 "n E,", du ;ont de Bemours & 2o, %, Federal 3rade 2ommission, F2' F,2d 128 92nd 2ir, 1'81:, the F32
unsuccess$ully challen)ed .hat it asserted to &e $acilitatin) #ractices that .ere unilaterally ado#ted 9althou)h
em#loyed &y all $our $irms in the industry:, claimin) authority under ection @ o$ the Federal 3rade 2ommission
Act, .hich does not reCuire the e/istence o$ an a)reement, 3o the e/tent that $acilitatin) #ractices
can only &e challen)ed .hen their e/istence is attri&uta&le to an a)reement, em#hasis is #laced on the issues
considered in su&section 3,1, nota&ly, under .hat circumstances an a)reement can &e in$erred .hen multi#le
$irms em#loy a $acilitatin) #ractice,
F@ As else.here, our discussion $ocuses on U,, antitrust la., EU la. also encom#asses $acilitatin) #ractices,
includin) e/chan)es o$ detailed in$ormation amon) com#etitors in industries #rone to collusion, ee, $or
e/am#le, Bellamy and 2hild 92001, X1<012:,
Ch% .1- $ntitrust 1131
the $merican Column = Lum!er case, a %iolation .as $ound .here $irms e/chan)ed in$ormation
on #rices in indi%idual transactions and this in$ormation .as su&Eect to audit
$or accuracy,F0 uch in$ormation )reatly eases the detection o$ cheaters, .hereas such
details do not ha%e an o&%ious and su&stantial #roducti%e use, "n Container Corporation,
$irms .ere deemed to ha%e %iolated the antitrust la.s .hen they called com#etitors
to %eri$y the accuracy o$ &uyersG assertions o$ ha%in) &een o$$ered lo.er #rices else.here,
FF
Another class o$ interest in%ol%es $irmsG contracts .ith their customers, For e/am#le,
a $irmGs use o$ a most<$a%ored<customer clauseKunder .hich it a)rees to )i%e
all customers under contract the &ene$its o$ any #rice cut e/tended to a su&seCuent
customerKmay )reatly reduce its incenti%e to de$ect $rom a collusi%e #rice since it
must sacri$ice #ro$its on its e/istin) customer &ase that .as other.ise loc*ed in $or a
#eriod o$ time at a hi)her #rice, 93his disincenti%e .ould &e immaterial $or an in$initesimal
#rice cut, &ut i$, as #re%iously discussed, )reater #rice cuts are necessary to
attract su&stantial ne. &usiness, the disincenti%e could &e su&stantial,: ome $irms em#loy
#rice<matchin) 9meetin) com#etition: clauses, under .hich they #romise to lo.er
their #rice i$ the &uyer can $ind a com#etitor that char)es less, 9ome clauses #romise to
eCual the #rice, #erha#s e%en retroacti%ely, that is, on #re%ious sales, and others #romise
to e/ceed the com#etitorGs #rice reduction,: 3his arran)ement deters other $irms $rom
lo.erin) their #rices, Moreo%er, it $acilitates detection &ecause &uyers o$$ered lo.er
#rices are more li*ely to re%eal other.ise secret #rice cuts, !&ser%e that under these
arran)ements &uyers as a .hole are disad%anta)edKi$ e$$ecti%e, the mar*et #rice is
sustained at a hi)her le%elK&ut indi%idual &uyers are su&Eect to the $ree<rider #ro&lem=
each may .ell )ain 9i$ there is any chance that some seller .ill lo.er #rice:, &ut its contri&ution
to a hi)her mar*et #rice .ill &e ne)li)i&le i$ it is a small #urchaser,F8 3hese
cross<currents are e/#lored in the economic literature on most<$a%ored<customer clauses
and meetin) com#etition clauses,F'
!ther ty#es o$ #ractices are directed at coordination #ro&lems caused &y #roduct hetero)eneity
and com#etition alon) dimensions other than #rice, Tuality or )radin) standards
may #romote uni$ormityKor at least reduce %ariety to mana)ea&le #ro#ortionsK
$acilitatin) a)reement on #rice, A)reements may limit credit 9and other: terms, lest
$irms cheat on the #rice &y o$$erin) $a%ora&le interest rates,
More &roadly, any $actor that may inhi&it collusi%e #ricin) is #otentially su&Eect
to $irmsG creati%ity in de%isin) means o$ a%oidin) its detrimental e$$ect, 3here are, o$
course, limits on .hat is $easi&le, Furthermore, to the e/tent that the use o$ $acilitatin)
#ractices itsel$ reCuires collusion, $irms must o%ercome any di$$iculties o$ coordination,
F0 American 2olumn & ?um&er 2o, %, United tates, 2@F U,, 3FF 91'21:,
FF United tates %, 2ontainer 2or#oration o$ America, 3'3 U,, 333 91'0':,
F8 3his ty#e o$ $ree<ridin) #ro&lem also arises .hen a mono#olist em#loys e/clusi%e dealin) #ro%isions .ith
its customers, as discussed in su&section @,1,1 &elo.,
F' ee 2oo#er 91'80:, Edlin 91''F:, Edlin and Emch 91''':, and alo# 91'80:, Borenstein 92001: a##lies
this idea to #rice matchin) in the airline industry and the airline tari$$ #u&lishin) case,
1132 L% 7aplow and C% Shapiro
detection, and en$orcement .ith re)ard to the $acilitatin) #ractices themsel%es, as illustrated
&y the Su*ar ,nstitute case, ome $acilitatin) #ractices may &e readily $ormulated
and o&ser%edJ others may &e com#le/ or hidden, Accordin)ly, the success$ul use o$
$acilitatin) #ractices .ill %ary )reatly,
3here remains another im#ortant consideration .ith many $acilitatin) #ractices= they
may ha%e redeemin) %irtues, ome in$ormation e/chan)e enhances #lannin), Forcin)
tradin) into $ormal mar*ets 9.hich .as #ermitted in Chica*o Board of >rade80:
#roduces &ene$its that $lo. $rom #u&lic #rices, 2ontractual arran)ements .ith &uyers
re)ardin) the sellersG and com#etitorsG #rices reduce search costs, E/chan)e o$ cost
in$ormation may enhance #roducti%e e$$iciency &y shi$tin) out#ut to more e$$icient
$irmsJ see ha#iro 91'80:, E%en .hat may seem literally to &e #rice<$i/in) .ill o$ten
&e e$$icient, such as .hen #roducti%e #artnershi#s or Eoint %entures are $ormed and
the resultant entity $i/es a sin)le #rice $or its common #roduct, ?i*e.ise, many other
arran)ements that may seem &ene$icial may also ha%e e$$ects on the $easi&ility o$ collusion,
Accordin)ly, it is necessary to $ormulate a means o$ &alancin) the costs and
&ene$its, .hich is the su&Eect o$ the ne/t su&section,
4%1%#% Rule of reason
"n the United tates, the Mrule o$ reasonN .as $ormally announced nearly a century a)o,
in the mono#oli4ation case o$ Standard +il,81 hortly therea$ter, it .as )i%en more content
in Chica*o Board of >rade in lan)ua)e that is routinely Cuoted 9or #ara#hrased: to
this day= M3he true test o$ le)ality is .hether the restraint im#osed is such as merely re)ulates
and #erha#s there&y #romotes com#etition or .hether it is such as may su##ress
or e%en destroy com#etition,N82 Althou)h more s#eci$ic than the almost<com#letely
Cuestion<&e))in) inCuiry into Mreasona&leness,N the meanin) o$ this test is hardly sel$e%ident,
Just .hat is meant &y Mcom#etitionNW "s it %alued #urely as a means or as an
end unto itsel$W "t is use$ul to &e)in .ith a $e. modern in%ocations o$ the rule o$ reasonGs
#romotin)<com#etition test, $ollo.ed &y some re$lection on the &roader Cuestion
o$ inter#retation and its relationshi# to economic analysis,83
80 Board o$ 3rade o$ 2ity o$ 2hica)o %, United tates, 210 U,, 231 91'18:,
81 tandard !il 2o, o$ Be. Jersey %, United tates, 221 U,, 1 91'11:,
82 Board o$ 3rade o$ 2ity o$ 2hica)o %, United tates, 210 U,, 231, 238 91'18:, 3he 2ourt continued= M3o
determine that Cuestion the court must ordinarily consider the $acts #eculiar to the &usiness to .hich the
restraint is a##liedJ its condition &e$ore and a$ter the restraint .as im#osedJ the nature o$ the restraint and its
e$$ect, actual or #ro&a&le, 3he history o$ the restraint, the e%il &elie%ed to e/ist, the reason $or ado#tin) the
#articular remedy, the #ur#ose or end sou)ht to &e attained, are all rele%ant $acts, 3his is not &ecause a )ood
intention .ill sa%e an other.ise o&Eectiona&le re)ulation or the re%erseJ &ut &ecause *no.led)e o$ intent may
hel# the court to inter#ret $acts and to #redict conseCuences,N
83 "n the Euro#ean Union, conduct may &e deemed e/em#t $rom the #rohi&ition in Article 8191: on anticom#etiti%e
a)reements i$ it meets certain criteria in Article 8193: that &ear resem&lance to the rule o$ reason in
the United tates, 3here are &oth &loc* 9)eneral: e/em#tions and those )ranted indi%idually, 3o enhance clarity,
the 2ommission has issued Auidelines on the A##lica&ility o$ Article 81 o$ the E2 3reaty to -ori4ontal
2oo#erati%e A)reements 9Euro#ean Union, 2001:,
Ch% .1- $ntitrust 1133
"n National Society of 'rofessional En*ineers, the ociety had an ethics rule #rohi&itin)
en)ineers $rom &ar)ainin) a&out #rice until a$ter they .ere selected $or a #roEect,81
3he #ro$$ered Eusti$ication .as that other.ise customers mi)ht &e induced to $ocus e/cessi%ely
on the #rice o$ #ro$essional ser%ices at the e/#ense o$ concerns a&out Cuality
and sa$ety, 3he u#reme 2ourt $ound a %iolation, a$ety .as not deemed unim#ortant,
&ut rather somethin) that ultimately .as $or customers to decide, 3hey could em#loy
the ocietyGs a##roach i$ they .ished, &ut the ociety could not im#ose this choice
on all customers, 2om#etition meant $ree and o#en choice, not one side o$ the mar*et
collecti%ely dictatin) terms to the other,
"n ,ndiana ;ederation of Dentists, cost<conscious insurance com#anies em#loyed an
internal #rocedure $or re%ie.in) su&missions $or reim&ursements,8@ 3he dentists o&Eected
9to non<dentists #assin) #ro$essional Eud)ment on their .or*, so they claimed:
and a)reed as a )rou# not to su##ly the necessary documentation, 3hey too lost, !nce
a)ain, it .as $or the customerKor, in essence, the customerGs a)ent, the insurance
com#anyKto ma*e .hate%er Eud)ments it .ished, Any indi%idual dentist .as $ree not
to deal .ith any insurer i$ the dentist thou)ht the insurerGs #ractices ina##ro#riate 9or
$or any other reason or $or no #articular reason:, &ut dentists could not a)ree, as a )rou#,
to im#ose their Eud)ment,
2onsider also National Colle*iate $thletic $ssociation, in%ol%in) a)reements amon)
uni%ersities re)ardin) colle)e $oot&all,80 3he u#reme 2ourt .as not &othered &y their
a)reements on rules o$ the )ame 9si4e o$ #layin) $ield, scorin), and so $orth:Krules that
.ere not challen)edK&ut did $ind their restrictions on schools sellin) tele%ision ri)hts
inde#endently o$ the AssociationGs scheme to constitute a %iolation,8F
For the most #art, cases such as these seem to see com#etition as a #rocess, 3he
%ie. seems to &e that com#etition consists o$ &uyers and sellers each decidin) $or
themsel%esKor, more #recisely, in indi%idual &uyer<seller #airsK.ith .hom they .ill
deal and on .hat terms, "nde#endent decisions are a central $eature o$ com#etition,
.hereas )rou#s 9ty#ically o$ sellers: .ho attem#t to im#ose some re)ime re)ardin) the
#ro#er terms o$ dealin) are su&%ertin) the #rocess, 3hey may or may not &e ri)ht, &ut
that is not the Cuestion, ;ut another .ay, .hat is ri)ht is essentially ta*en to &e .hate%er
is the outcome o$ the com#etiti%e #rocess, much li*e ho. one acce#ts the eCuili&rium
#rice in a com#etiti%e mar*et as Mreasona&le,N
;erha#s com#etition is %ie.ed as )ood in itsel$, !r instead the %ie. may &e that com#etition
is %alued $or its results, .hether those understood &y economists, in terms o$
81 Bational ociety o$ ;ro$essional En)ineers %, United tates, 13@ U,, 0F' 91'F8:,
8@ Federal 3rade 2ommission %, "ndiana Federation o$ 6entists, 1F0 U,, 11F 91'80:,
80 Bational 2olle)iate Athletic Association %, Board o$ 7e)ents o$ the Uni%ersity o$ !*lahoma, 108 U,, 8@
91'81:,
8F 3his case also nicely illustrates .hat is sometimes re$erred to as the Mancillary restraintsN doctrine, Bamely,
an anticom#etiti%e restraint is not deemed #ermissi&le merely &ecause it is associated .ith an other.ise
le)itimate %entureJ ho.e%er, the restraint may .ell &e allo.ed i$ it is reasona&ly necessary to accom#lish the
le)itimate o&Eecti%es o$ the %enture,
1131 L% 7aplow and C% Shapiro
allocati%e e$$iciency, or other notions concernin) $reedom o$ choice, Under this second,
instrumental %ie. o$ com#etition, the antitrust la.s are ne%ertheless inter#reted to relate
only to the #rocess= #erha#s the inte)rity o$ the com#etiti%e #rocess is much easier
to assess than the outcome o$ that #rocess, and &ene$its are assumed to $lo. as lon) as
com#etition is assured, "n cases .here it is alle)ed that the com#etiti%e #rocess is not
#ro%idin) the e/#ected &ene$its, courts in the United tates re#eatedly state that the a##ro#riate
remedy is to see* le)islati%e or re)ulatory action, E%en .hen there are mar*et
im#er$ections, it is #lausi&le to distrust the collecti%e schemes o$ sel$<interested mar*et
actors, schemes that they alle)e to &e correcti%es in the #u&lic interestKAdam mithGs
.arnin) &ein) a#ro#os,
3his #rocess %ie., ho.e%er, is #ro&lematic, Althou)h economists routinely use the
term Mcom#etition,N it does not readily &ear the .ei)ht that it must under the rule o$
reason in Eud)in) industry #ractices, "s the $ormation o$ a Eoint %enture &et.een t.o
$irms that mi)ht other.ise com#ete .ith each other, althou)h less e$$ecti%ely or .ith a
some.hat di$$erent #roduct, an enhancement to or a detraction $rom com#etitionW -o.
a&out a #artnershi# or a hori4ontal mer)erW5hat o$ curin) a mar*et $ailureW E%en i$ the
result o$ coordinated action is unam&i)uously more e$$icient, is it more com#etiti%eW
6oes the com#etiti%e #rocess include com#etition amon) institutional $orms, includin)
%arious $orms o$ coo#eration amon) )rou#s o$ $irms that o#erate in the same industryW
More &roadly, .hen the conditions $or #er$ect, te/t&oo* com#etition $ail 9that is, #retty
much al.ays:, is there an unam&i)uous .ay to descri&e one or another arran)ement or
outcome as more com#etiti%eW88
Economists do not traditionally ans.er such Cuestions, "nstead, they underta*e #ositi%e
analysis o$ &eha%ior and outcomes under %arious mar*et arran)ements, For normati%e
#ur#oses, the ordinary metric is .el$are, or e$$iciency, or #erha#s utility to each
#arty or class o$ #arties, not the de)ree o$ com#etition accordin) to some com#etition
inde/, Iet, i$ the rule o$ reason is le)ally de$ined in terms o$ com#etition itsel$Kthat
.hich #romotes com#etition is le)al, that .hich su##resses com#etition is ille)al, end
o$ storyKthen economics cannot directly address the le)al test,
As it turns out, no matter ho. o$ten the #romote<%ersus<su##ress<com#etition test
is in%o*ed, it is not adhered to uni$ormly, and le)al authorities seem to de#art $rom
it $airly readily in many o$ the cases in .hich its a##lication seems #ro&lematic, As
noted, in National Colle*iate $thletic $ssociation, the u#reme 2ourt $inds hori4ontal
a)reement on rules o$ the )ame to &e un#ro&lematic, 3here is a sense in .hich this
$le/i&ility may ha%e &ene$ited $rom a $ortuitous #lay on .ords, in that such rules .ere
seen as creatin) com#etitionKs#orts com#etition, that is, But e%en the case that $irst
announced the no.<canonical lan)ua)e on com#etition, Chica*o Board of >rade, .as
88 As .ith Ma)reement,N little aid comes $rom standard de$initions, 2om#etition is ordinarily ta*en to mean
the act or #rocess o$ com#etin), ri%alry, or s#eci$ically the e$$ort o$ #arties to secure &usiness o$ a third
#arty, Under that ru&ric, e%en a sim#le #artnershi# o$ t.o indi%iduals .ho other.ise mi)ht #roduce 9ho.e%er
ine$$iciently: on their o.n can readily &e seen as Manticom#etiti%e,N 3his de$inition is reasona&ly clear, &ut as
.ill &e discussed it is one that antitrust tri&unals o$ten disre)ard, and .ith )ood reason,
Ch% .1- $ntitrust 113@
one that condoned restrictions on indi%idual actorsG $reedom o$ action to #roduce a
)reater )ood, tradin) in the #u&lic mar*et,8'
!ther modern cases rein$orce a more com#le/ inter#retation, Bota&ly, in Broadcast
&usic, there .ere t.o lar)e entities 9BM" and A2A;: that, &et.een them, licensed
the ri)hts to nearly all domestic co#yri)hted music to %arious users 9$or e/am#le, radio
and tele%ision stations:,'0 3he entities set a sin)le $ee $or &loc* licenses, .hich .as
challen)ed, amon) other reasons, as constitutin) ille)al #rice<$i/in), And in $act, each
o$ these t.o entities did set sin)le #rices $or &undles o$ millions o$ musical com#ositions
that other.ise mi)ht &e #riced inde#endently, Iet, economies o$ scale in contractin) and
co#yri)ht en$orcement 9that is, monitorin) the ille)al use o$ the entitiesG #ort$olios o$
music &y unlicensed #arties: induced the 2ourt to $ind no %iolation, 3he result in the
mar*et .as nothin) li*e atomistic com#etition under .hich indi%idual com#osers #aired
%oluntarily .ith indi%idual &uyers, an alternati%e the 2ourt $ound to &e cum&ersome,
"nstead, hu)e collections o$ other.ise<com#etitors used a sales a)ent to dictate #rice
and other terms o$ dealin), u##osin) one acce#ts that the #ermitted arran)ements .ere
on<&alance desira&le on e$$iciency )rounds, there remains the Cuestion o$ .hether the
arran)ements in%ol%ed more or less Mcom#etition,N
!ne de%ice em#loyed in Broadcast &usic and in some other cases is to treat the
challen)ed %enture as a sin)le entity= once %ie.ed in this manner, there is no lon)er a
hori4ontal a)reement and thus no %iolation o$ herman Act ection 1, ?oo*in) ahead
to section 1, hori4ontal mer)ers are not themsel%es %ie.ed as #rice<$i/in) cartelsKe%en
thou)h the mer)ed $irms #resuma&ly $i/ a common #riceK&ut rather as sin)le entities,
"n such cases, ho.e%er, there remains the Cuestion .hether the a)reement creatin* .hat
is su&seCuently %ie.ed as a sin)le entity constitutes a %iolation, 2arte &lanche .ould
authori4e $ormal cartels, say, incor#orated as a sin)le $irm, !$ course, Eurisdictions do
not $reely #ermit $ormal cartel arran)ements or hori4ontal mer)ers, Bor do they automatically
a##ro%e e%en loose trade associations i$, $or e/am#le, mem&er $irms en)a)e
in in$ormation e/chan)es o$ a sort that $acilitate collusion and )enerate little o$$settin)
&ene$it, 3hat is, .hen trade association acti%ity has &een challen)ed success$ully, no
sin)le<entity de$ense has &een reco)ni4ed, 3hus, the sin)le<entity characteri4ation is
more o$ a conclusion than a reason to decide one .ay or the other,
5hat, then, is the underlyin) meanin) o$ the rule o$ reasonW !n one hand, antitrust
la. does not insist on #ure atomistic com#etition, #rohi&itin) all com&inations $rom
small #artnershi#s to trade associations to Eoint %entures to mer)ers, !n the other hand,
hori4ontal arran)ements are not $reely #ermitted, "nstead, they are su&Eect to some
sort o$ &alancin) test, .hether under the rule o$ reason in the United tates or under
other ru&rics else.here, 5hen the arran)ement loo*s li*e little more than a #ure
8' !ur #oint is not to a)ree .ith the 2ourtGs analysis in Chica*o Board of >rade, .hich .as #ro&lematic in a
num&er o$ res#ects, &ut rather to indicate that the #ure, atomistic, hands<o$$ #rocess %ie. o$ com#etition .as
ne%er the com#lete story,
'0 Broadcast Music, "nc, %, 2olum&ia Broadcastin) ystem, "nc,, 111 U,, 1 91'F':,
1130 L% 7aplow and C% Shapiro
inter$erence .ith ordinary com#etition, it is li*ely to &e condemned .ith little $urther
inCuiry, "n many cases in the United tates, $or e/am#le, mar*et #o.er need not
&e demonstrated, and ad%erse e$$ects need not &e #ro%ed 9althou)h $or an a.ard o$
dama)es these considerations .ill &e im#ortant:, E/am#les include National Society of
'rofessional En*ineers, ,ndiana ;ederation of Dentists, and the tele%ision mar*etin) restrictions
in National Colle*iate $thletic $ssociationKin addition to na*ed #rice<$i/in)
and related #ractices, 5hen, ho.e%er, there a##ear to &e &ene$itsK$rom com&inin)
#roduction, conductin) research, settin) standards, or other.iseKcondemnation is not
)uaranteed, as demonstrated &y Broadcast &usic, And some hori4ontal arran)ements,
li*e #artnershi#s and mer)ers that do not #roduce su&stantial mar*et #o.er, are routinely
allo.ed,
3he #rimary area o$ am&i)uity concerns the many #ractices that $all in &et.een the
e/tremes, Economists can analy4e the arran)ementsG e$$ects and assess their e$$iciency,
But ho. do such assessments relate to the le)al testW Most modern antitrust rule<ma*ers
and adEudicators seem to #ay su&stantial attention to economic considerations, at least
in many settin)s, But under $ormulations li*e the rule o$ reason, the conce#tion o$
reasona&lenessK.hether or not concreti4ed as a determination o$ #romotion %ersus
su##ression o$ com#etitionKis not .ell s#eci$ied, 5e *no. $rom cases li*e Broadcast
&usic that #ros and cons .ill sometimes &e &alanced, &ut .hat counts as a &ene$it or
cost o$ an arran)ement, .hat metric is em#loyed $or measurement and con%ersion to
a common denominator 9i$ this is done at all:, and .hat is the ultimate decision rule
remain some.hat o$ a mystery,'1 A #urely economic criterion has not &een e/#licitly
em&racedJ nor has it &een reEected,'2
4%/% $ntitrust enforcement
5e close this section &y commentin) &rie$ly on some o$ the la. and economics issues
that arise in antitrust en$orcement,'3
'1 M2ourts sometimes descri&e their tas* under the rule o$ reason as one o$ R&alancin)G #otential harms a)ainst
li*ely )ains or de$enses, But &alancin) im#lies that one #laces some measura&le Cuantity o$ somethin) on
one side o$ a scale, a Cuantity o$ somethin) else on the other side, and determines .hich side out.ei)hs the
other, 3he set o$ rou)h Eud)ments .e ma*e in antitrust liti)ation does not e%en come close to this R&alancin)G
meta#hor, "ndeed, most courts do not e%en de$ine a unit o$ measurement in .hich the Cuantities to &e &alanced
can &e measured, Assumin) the rele%ant unit is dollars, one .ould need to #lace at least a rou)h dollar estimate
on the dan)ers to com#etition , , , and a similar estimate on li*ely cost sa%in)s, out#ut increases, or other
&ene$its, 3o the &est o$ our *no.led)e, this has ne%er &een done in any antitrust case,N -o%en*am# 9200@,
%ol, 11, #, 33':, -o%en*am#, it should &e noted, does not o$$er this de#iction as a criticism, "nstead, he sees
such &alancin) as &eyond the institutional com#etence o$ courts and &elie%es that in #ractice they em#loy a
structured seCuence o$ 9essentially dichotomous: inCuiries that usually ena&les them to resol%e cases one .ay
or the other .ithout e%er ha%in) to &alance costs and &ene$its,
'2 !ne Cuestion o$ #articular interest is, su##osin) that the criterion is economic, .hether it in%ol%es e$$iciency
as a .hole or only consumer sur#lus, 2om#are our discussion o$ this issue in the conte/t o$ hori4ontal
mer)ers, in su&section 1,1,3,
'3 For discussion o$ additional issues, see, $or e/am#le, ;osner 92001:,
Ch% .1- $ntitrust 113F
4%/%.% ,mpact of antitrust enforcement on oli*opolistic !ehavior
3he .or*horse model o$ oli)o#oly used to study collusion, namely the model o$ re#eated
#rice< or Cuantity<settin), does not e/#licitly include any antitrust en$orcement,
At $irst &lush, this seems rather #eculiar, at least $rom the #ers#ecti%e o$ la. and economics,
-o.e%er, i$ this &asic model ca#tures conduct that is &elie%ed &y the #arties
to &e &eyond the reach o$ the antitrust la.sKre#eated #rice<settin) .ithout any other
communicationsKthen the omission is Eusti$ied, 3his is another reminder that economic
theory may &e most rele%ant in determinin) the e/istence o$ #rice<$i/in) .hen it hel#s
us understand .hether additional conduct, such as communications or $acilitatin) #ractices,
si)ni$icantly increases the li*elihood that a collusi%e outcome .ill occur,
"n contrast, .here%er antitrust la. is a##lica&le, it is im#ortant to consider the in$luence
o$ e/#ected sanctions on $irmsG &eha%ior, -arrin)ton 92001a, 2001&, 200@: introduces
en$orcement #olicy into oli)o#oly su#er)ames, -e #osits that a ne.ly $ormed
cartel .ill &e more li*ely to attract the attention o$ antitrust en$orcers 9#erha#s &ased
on com#laints &y customers: i$ it ra#idly raises #rice $rom the com#etiti%e le%el to the
cartel le%el, -e sho.s ho. the #rice #ath ado#ted &y the cartel and the steady<state cartel
#rice are a$$ected &y antitrust en$orcement, -e also studies the relationshi# &et.een
dama)es rules in #rice<$i/in) cases and cartel #ricin), "n the #rocess, he identi$ies some
com#le/ and e%en #er%erse e$$ects o$ antitrust en$orcement on cartel #ricin),
4%/%#% Determinants of the effectiveness of antitrust enforcement
A num&er o$ other as#ects o$ antitrust en$orcement ha%e recently &een illuminated
&y economic analysis, !ne increasin)ly acti%e a##roach to en$orcement is the )o%ernmentGs
attem#t to strate)ically induce some colludin) $irms to turn on their #eers,
Enhanced leniency to.ard coo#eratin) $irms 9as .ell as increased international coo#eration:
led to success$ul #rosecutions in a series o$ maEor international #rice<$i/in) cases
durin) the 1''0s, -arrin)ton 92000a: discusses the im#act o$ cor#orate leniency #ro)rams
on collusion, ee also Motta and ;olo 92003: and Motta 92001, #, 1'1: on the
Euro#ean 2ommissionGs ne.ly ado#ted leniency #olicy, and ?itan and ha#iroGs 92002:
discussion o$ cartel en$orcement durin) the 1''0s,
Another im#ortant en$orcement su##lement that is #articularly im#ortant in the
United tates in%ol%es #ri%ate la.suits $or 9tre&le: dama)es, 5hen the 6e#artment o$
Justice &rin)s a #rice<$i/in) case, there ty#ically are immediate $ollo.<on #ri%ate actions
&rou)ht &y #arties claimin) to ha%e &een o%erchar)ed, FreCuently, these cases
are &rou)ht as class actions, and many ha%e resulted in lar)e #ayments, Althou)h only
direct #urchasers can claim dama)es under U,, $ederal antitrust la.s, many states allo.
indirect #urchasers to reco%er dama)es as .ell, "n all o$ these settin)s, economists
are relied u#on to estimate dama)es $or o%erchar)es, As #re%iously noted, the challen)e
they con$rontKdeterminin) .hat #rices .ould ha%e e/isted &ut $or the ille)al
collusionKis closely related to the underlyin) analysis o$ collusi%e &eha%ior,
1138 L% 7aplow and C% Shapiro
3here is also a )ro.in) economic literature on cartel detectionKaddressin) .hat
#atterns o$ #ricin), or &iddin), are indicati%e o$ collusionKthat is im#ortant $or settin)
en$orcement #riorities, determinin) lia&ility, and assessin) dama)es, ;orter 9200@:,
-arrin)ton 9200F:, and 5hinston 92000, ##, 38O@2: #ro%ide hi)hly in$ormati%e sur%eys,
BaEari and ummers 92002: discuss the detection o$ collusion amon) &idders in
an auction settin),
,& -ori.ontal mergers
3he #rimary concern a&out hori4ontal mer)ersKthat is, mer)ers &et.een direct
com#etitorsKis that they may lead to anticom#etiti%e #rice increases, either &ecause
the mer)ed entity on its o.n .ill $ind it #ro$ita&le to raise #rices $rom #re<mer)er le%els
9so<called unilateral e$$ects: or &ecause the increase in concentration enhances the
#ros#ects $or success$ul collusion 9coordinated e$$ects:,'1 Accordin)ly, .e &e)in &y o$$erin)
an economic analysis o$ these #ossi&ilities, dra.in) on our analysis in sections 2
and 3, Be/t, .e &rie$ly re%ie. em#irical e%idence on the actual e$$ects o$ hori4ontal
mer)ers,'@
Antitrust en$orcement #lays an acti%e role .ith re)ard to hori4ontal mer)ers &ecause
in the United tates nontri%ial mer)ers must &e re%ie.ed &y one o$ the t.o $ederal
authorities .ith o%erla##in) Eurisdiction, the Antitrust 6i%ision o$ the 6e#artment o$
Justice 96!J: and the Federal 3rade 2ommission 9F32:, 9imilar mer)er re%ie. ta*es
#lace in other Eurisdictions, such as the Euro#ean Union,: 3hese re%ie.s are )o%erned
&y the -ori4ontal Mer)er Auidelines, the current %ersion o$ .hich .as 9mostly: #romul)ated
in 1''2 &y the 6!J and F32,'0 5e descri&e the #ertinent #rocedures, $rom initial
$ilin) to the a)enciesG analysis to court challen)es and remedies, Althou)h our $ocus is
on the Mer)er Auidelines due to their current centrality, .e also discuss the #ertinent
antitrust statutes and the e%olution o$ hori4ontal mer)er doctrine in the courts, 5e #ay
#articular attention to the role o$ #ros#ecti%e mer)er syner)ies, usually re$erred to in
antitrust discussions as mer)er e$$iciencies, 3hese &ene$its are im#ortant in determinin)
the threshold o$ anticom#etiti%e e$$ects that must &e #resent to challen)e a mer)erKthat
is, the la. im#licitly #resumes mer)ers to &e ad%anta)eous to some de)reeKand also in
o$$erin) a #ossi&le a$$irmati%e de$ense to a mer)er that other.ise .ould &e #rohi&ited,
Furthermore, in assessin) the role o$ e$$iciencies in Eusti$yin) hori4ontal mer)ers, it is
'1 A #rice increase o$ten ser%es as a #ro/y $or other #ossi&le anticom#etiti%e e$$ects, such as a reduction in
#roduct Cuality or ser%ice or a decrease in the #ace o$ inno%ation,
'@ A su&Eect related to hori4ontal mer)ers that .e do not consider here is the tendency o$ #artial crosso.nershi#
to so$ten com#etition and thus increase #rice, ee Bresnahan and alo# 91'80:, 7eynolds and
na## 91'80:, Farrell and ha#iro 91''0a:, !GBrien and alo# 92000:, and Ailo, Moshe, and #ie)el 92000:,
'0 Federal 3rade 2ommission and U,, 6e#artment o$ Justice, -ori4ontal Mer)er Auidelines 9A#ril 2, 1''2:
9as re%ised A#ril 8, 1''F .ith res#ect to ection 1, relatin) to E$$iciencies:, ee also the en$orcement a)enciesG
detailed commentary on the )uidelines, F32 and 6!J 92000:,
Ch% .1- $ntitrust 113'
necessary to s#eci$y more #recisely the )oals o$ the antitrust la.s, in #articular, .hether
the o&Eecti%e is to ma/imi4e total economic .el$are or instead Eust consumer sur#lus,
Finally, .e consider in )reater de#th the economics underlyin) the analysis dictated
&y the Mer)er Auidelines, #articularly .ith re)ard to mar*et de$inition, relatin) the
Auidelines a##roach to the economic analysis o$ mar*et #o.er #resented in section 2,
"n this re)ard, .e also discuss the )ro.in) &ody o$ em#irical methods $or #redictin) the
e$$ects o$ #articular hori4ontal mer)ers, ;art o$ the challen)e is theoretical= )i%en that
there are a num&er o$ theories o$ oli)o#oly, .ith rather di$$erent #redictions, .hich one
should &e used in a )i%en mer)erW ;resuma&ly, the one that &est $its the $acts o$ that
mer)er, But all o$ these theories are hi)hly sim#li$ied in com#arison .ith the ine%ita&le
com#le/ity o$ real .orld com#etition, so #ic*in) the most suita&le model o$ oli)o#oly
is $ar $rom strai)ht$or.ard,
2%.% +li*opoly theory and unilateral competitive effects
3he &asic idea underlyin) theories o$ unilateral e$$ects is that the mer)ed $irm .ill ha%e
an incenti%e to raise its #rice9s:, in com#arison .ith the #re<mer)er #rice9s:, &ecause o$
the elimination o$ direct com#etition &et.een the t.o $irms that ha%e mer)ed, 3he e/amination
o$ s#eci$ic oli)o#oly models ma*es it #ossi&le to Cuanti$y the e$$ects, .hich
is im#ortant $or mer)er en$orcement, First, Cuanti$ication can hel# to identi$y the mer)ers
that are most li*ely to ha%e si)ni$icant #rice e$$ects and thus cause si)ni$icant harm
to consumers, 3hese are the mer)ers that #resuma&ly .arrant $urther scrutiny, i$ not #rohi&ition,
econd, Cuanti$ication allo.s us to estimate the mer)er e$$iciencies necessary
to o$$set the loss o$ com#etition and there&y allo. the mer)er to #ass muster accordin)
to the consumer sur#lus or total .el$are standard,
2%.%.% Cournot model with homo*eneous products
5e &e)in &y studyin) the e$$ects o$ mer)ers in the 2ournot oli)o#oly model descri&ed
in su&section 2,3,1, 3he 2ournot model seems li*e a )ood startin) #lace since it )enerates
a num&er o$ sensi&le #redictions relatin) mar*et structure to the eCuili&rium
outcome, "n #articular, .e deri%ed eCuation 93: P&Ci
P
= Si
|D| that relates a $irmGs
#rice<cost mar)in to its mar*et share and the mar*et elasticity o$ demand, "n the s#ecial
case .ith constant and eCual mar)inal costs, each $irm has a mar*et share o$
1/N, and the 2ournot model #redicts that the mar)in o$ each $irm .ill &e )i%en &y
(P &C)/P = 1/N|D|,5e also deri%ed an e/#ression $or the industry<.ide a%era)e,
out#ut<.ei)hted, #rice<cost mar)in, that is, 'C& N
i=1 Si
P&Ci
P , namely e/#ression
91:= 'C& = 1
|D| N
i=1 S2
i
= H
|D| , .here, recall, H S2
i is the -er$indahl<
-irschman "nde/ 9--": o$ mar*et concentration,
3he idea that a $irm .ith a lar)e share .ill ha%e more mar*et #o.er, and thus .ill
char)e a hi)her #rice 9&ut still less than the mono#oly #rice: has &een %ery in$luential
in hori4ontal mer)er en$orcement, o has the idea that mar)ins are hi)her in more concentrated
industries, "n $act, &ased #artially on the e/#ression $or the ;2M, the Mer)er
1110 L% 7aplow and C% Shapiro
Auidelines measure mar*et concentration usin) the --", At the same time, it is reco)ni4ed
that the mar)ins o$ all $irms in a )i%en mar*et are lo.er i$ the elasticity o$ demand
in that mar*et as a .hole is lar)e, a su&Eect to .hich .e .ill return in discussin) mar*et
de$inition in su&section 1,@ &elo.,
5hile all o$ these e/#ressions accurately characteri4e the 2ournot eCuili&rium, none
o$ them actually tells us .hat ha##ens to #rice, consumer sur#lus, #ro$its, or total .el$are
as a result o$ a mer)er &et.een t.o $irms in a 2ournot oli)o#oly, 3o ans.er those
Cuestions, .hich are central to the analysis o$ hori4ontal mer)ers, it is necessary to
com#are the 2ournot eCuili&ria &e$ore and a$ter the mer)er and, in #articular, to s#eci$y
.hat is in%ol%ed .hen t.o $ormerly inde#endent $irms &ecome one,
alant, .it4er, and 7eynolds 91'83: address this Cuestion, em#hasi4in) the #eculiar
result that mer)ers in a 2ournot oli)o#oly can &e un#ro$ita&le, 3he reduction in the
num&er o$ $irms raises #rice, "nitially, the mer)in) $irms reduce their out#ut &ecause
they internali4e more o$ the e$$ect o$ their out#ut on #rice than they did #re%iously,
"n turn, non<mer)in) $irms raise out#ut some.hat, leadin) the mer)in) $irms to cut
out#ut $urther, At the ne. eCuili&rium, #rice is hi)her, 9"ndeed, this is .hy #rice is
ordinarily hi)her in 2ournot eCuili&rium .hen there are $e.er $irms,: But the mer)ed
$irmGs com&ined share o$ total industry #ro$its is lo.erJ a$ter all, other $irmsG Cuantities
rise and the out#ut o$ the mer)in) $irms $alls, alant et al, $ocus on the symmetric
case .ith constant mar)inal costsJ in this settin), a mer)er o$ any num&er o$ $irms is
eCui%alent to all &ut one o$ the mer)in) $irms shuttin) do.n, As a result, the cost due
to the smaller #ro$it share .ill e/ceed the &ene$it $rom a hi)her industry #rice unless, in
their e/am#le, the mer)in) $irms constitute 80S or more o$ the industry_ "n this sim#le
model, a mer)er does not lead to a Mstron)erN $irm in any senseKas noted, it is as
i$ the acCuired $irm sim#ly e/its, "$ this story de#icted ho. mer)ers .or*, $e. mer)ers
9short o$ mer)ers to mono#oly: .ould &e o&ser%ed, Accordin)ly, a theory that #lausi&ly
e/#lains mer)ers that actually occur reCuires that the mer)in) $irms o.n assets that can
&e use$ully com&ined in some .ay,
;erry and ;orter 91'8@: #ursue this #oint usin) a model in .hich each $irm o.ns a
certain amount o$ ca#ital, "n their model, each $irmGs mar)inal cost increases linearly
.ith that $irmGs out#ut, and the slo#e o$ the mar)inal cost cur%e is lo.er, the lar)er is
the $irmGs ca#ital stoc*, 3hus, $irms that o.n more ca#ital are lar)er in the resultin)
2ournot eCuili&rium, ;erry and ;orter assume that .hen t.o $irms mer)e, the mer)ed
entity o.ns their com&ined ca#ital stoc* and thus has a lo.er mar)inal cost cur%e than
either o$ the constituent $irms, "n addition, since the mar)inal cost o$ each ri%al $irm
rises .ith its out#ut, the a&ility o$ ri%al $irms to e/#and in res#onse to the mer)er is
not as )reat as in the #rior e/am#le in .hich mar)inal cost is constant, As a result,
hori4ontal mer)ers are much more li*ely to &e #ro$ita&le in this model, ?e%in 91''0:
)enerali4es the alant, .it4er, and 7eynolds 91'83: model in a di$$erent direction &y
allo.in) the $irms to di$$er in their 9constant: mar)inal costs,'F McA$ee and 5illiams
'F ?e%in also allo.s the mer)ed $irm to &eha%e other than as a 2ournot oli)o#olist, $or e/am#le, as a tac*el&er)
leader,
Ch% .1- $ntitrust 1111
91''2: $urther e/#lore models .ith Cuadratic cost $unctions .here the mar)inal cost
o$ a $irm is #ro#ortional to the ratio o$ its out#ut to its ca#ital stoc*, sho.in) ho. the
ma)nitude o$ the #rice increase resultin) $rom a mer)er de#ends on the ca#ital stoc*s
o$ the mer)in) and non<mer)in) $irms,
Farrell and ha#iro 91''0&: si)ni$icantly )enerali4e these results and #ro%ide an
analysis o$ the #rice and .el$are e$$ects o$ hori4ontal mer)ers in 2ournot oli)o#oly,
3hey start .ith a 2ournot eCuili&rium amon) N $irms, .here the cost $unction o$ $irm i
is )i%en, as &e$ore, &y Ci(Xi ), A mer)er in 2ournot oli)o#oly can &e modeled as the
re#lacement o$ t.o e/istin) $irms .ith cost $unctions C1(X1) and C2(X2) &y a sin)le
mer)ed $irm .ith its o.n, ne. cost $unction, C12(X12),
Farrell and ha#iro say that a mer)er )enerates no syner)ies i$ the mer)er sim#ly
allo.s the mer)in) $irms to rationali4e out#ut &et.een their e/istin) o#erations or $acilities,
that is, i$ C12(X12) = min
X1,X2
[C1(X1)+C2(X2)] su&Eect to X1 +X2 = X12, 6e$ine
the #re<mer)er out#uts o$ the t.o mer)in) $irms as X1 and X2, and the #re<mer)er #rice
as P, ?a&el the t.o mer)in) $irms so that $irm 2Gs #re<mer)er out#ut is at least as lar)e
as $irm 1Gs #re<mer)er out#ut, X2 X1, Usin) the #re<mer)er 2ournot eCuili&rium
relationshi# 93:, P&Ci
P
= Si
|D| , .e *no. that lar)er $irms ha%e hi)her mar*u#s, so $irm
1Gs mar)inal cost in the #re<mer)er eCuili&rium, &C1 = &C1(X1), is at least as lar)e
as $irm 2Gs, &C2 = &C2(X2), 6enote the mer)ed $irmGs mar)inal cost at the com&ined
#re<mer)er out#ut &y &C12 = &C12(X1 + X2),
Usin) this $rame.or*, Farrell and ha#iro #ro%e )enerally the im#ortant result that
mer)ers )eneratin) no syner)ies raise #rice,5ithout syner)ies, the mer)ed $irmGs a&ility
to rationali4e #roduction &et.een its e/istin) o#erations 9&y eCuatin) the mar)inal cost
o$ #roduction in the t.o o#erations: is not su$$icient to o$$set the incenti%e to raise #rice
that results $rom com&inin) the o.nershi# interests o$ the t.o o#erations,
Farrell and ha#iro also as* a&out the ma)nitude o$ syner)ies necessary $or a hori4ontal
mer)er to lead to a reduction rather than an increase in #rice, 3his is an im#ortant
Cuestion in #ractice &ecause, as discussed in su&section 1,1,3, mer)ers tend
to &e Eud)ed &ased on their im#act on consumers, Farrell and ha#iro #ro%ide a
%ery )eneral necessary and su$$icient condition= a mer)er reduces #rice i$ and only i$
&C2 &C12 > P &C1, 3hat is, the mer)er .ill reduce #rice i$ and only i$ the
mar)inal cost o$ the mer)ed $irm 9at the #re<mer)er com&ined out#ut: is less than the
mar)inal cost o$ the more e$$icient $irm 9at its o.n #re<mer)er out#ut: &y an amount
that e/ceeds the di$$erence &et.een the #rice and the mar)inal cost o$ the smaller, less
e$$icient $irm #rior to the mer)er, 3his ineCuality can &e e/#ressed in #ro#ortion to the
#re<mer)er #rice as
90:
&C2 &C12
P
>
P &C1
P
= S1
|D| ,
.here .e ha%e added the #re<mer)er relationshi# &et.een $irm 1Gs mar)in and its share,
3his is a %ery demandin) condition in an industry .ith moderate to lar)e #re<mer)er
mar)ins, For e/am#le, consider a 2ournot industry in .hich the mar*et elasticity o$
1112 L% 7aplow and C% Shapiro
demand at the #re<mer)er #rice is D = 1.0, normali4e the #re<mer)er #rice at
P = 100, and su##ose that the #re<mer)er mar*et shares o$ the t.o $irms are 10S
and 30S, so S1 = 0.1 and S2 = 0.3, Usin) the #re<mer)er 2ournot eCuili&rium conditions,
the #re<mer)er mar)inal costs o$ the t.o $irms must &e '0 and F0 res#ecti%ely,
3he ineCuality a&o%e tells us that the mer)er .ill lo.er #rice i$ and only i$ the mar)inal
cost o$ the mer)ed $irm, at the com&ined out#ut o$ t.o mer)in) $irms, is less than 00,
Usin) these )eneral results, Froe& and5erden 91''8: #ro%ide calculations that relate
the reCuired ma)nitude o$ the syner)ies to the #re<mer)er shares o$ the mer)in) $irms,
"n the symmetric case, they sho. that the #ro#ortionate reduction in mar)inal cost necessary
$or #rice not to rise is eCual to S/(|D| S), .here S is the #re<mer)er mar*et
share o$ each mer)in) $irm,
Analy4in) the .el$are im#act o$ such mer)ers is more com#le/, in #art &ecause .el$are
e$$ects de#end hea%ily on the cost $unction o$ the mer)ed entity in com#arison .ith
the cost $unctions o$ the t.o constituent $irms, .hich ca#tures any syner)ies resultin)
$rom the mer)er, -o.e%er, Farrell and ha#iro 91''0&: are a&le to o&tain )eneral results
a&out the Me/ternalN e$$ect o$ the mer)er, that is, the com&ined e$$ect o$ the mer)er on
consumers and ri%als, "$ .e are #re#ared to #resume that a #ro#osed mer)er raises the
com&ined #ro$its o$ the mer)in) $irms 9$or other.ise they .ould not choose to mer)e:,
then any mer)er that )enerates #ositi%e e/ternal e$$ects must raise total .el$are, For a
ran)e o$ demand and cost conditions, Farrell and ha#iro #ro%ide an u##er &ound on
the com&ined share o$ the mer)in) $irms such that their mer)er must )enerate #ositi%e
e/ternal e$$ects, "$ the com&ined share o$ the mer)in) $irms is small, they .ill not $ind it
#ro$ita&le to restrict out#ut much, i$ at allJ .hen they do restrict out#ut the lar)er $irms
are li*ely to e/#and, and shi$tin) out#ut to.ard lar)er $irms actually &oosts .el$are,
since the lar)er $irms ha%e lo.er #re<mer)er mar)inal costs, 3his a##roach has the si)ni$icant
%irtue that it does not in%ol%e an inCuiry into the e$$iciencies )enerated &y the
mer)er, .hich can &e di$$icult to Cuantity and %eri$y, as .e discuss &elo.,
Until no., .e ha%e e/amined the e$$ects o$ mer)ers on #rice and .el$are &ut ha%e
not related this analysis to the e$$ect o$ the mer)er on industry concentration, a ty#ical
$ocus o$ hori4ontal mer)er en$orcement #olicy 9as re$lected in the Mer)er Auidelines:,
#eci$ically, concern is ty#ically thou)ht to &e )reater, the hi)her is #re<mer)er concentration
and the )reater is the mer)er<induced increase in concentration, nota&ly, as
measured &y the --", Farrell and ha#iro sho., ho.e%er, that increases in the --"
may .ell increase total .el$are, "n #articular, they sho. that, startin) $rom a 2ournot
eCuili&rium, an ar&itrary small chan)e in the out#uts o$ all o$ the $irms raises .el$are
i$ and only i$ dX
X
+ 12
dH
H > 0, .here X, as &e$ore, is industry out#ut, Baturally, an
increase in out#ut raises .el$are, since #rice is a&o%e mar)inal cost $or all o$ the $irms,
More sur#risin)ly, $or a )i%en chan)e in total out#ut, .el$are is hi)her the )reater is
the chan)e in concentration, 5hyW Each $irmGs #rice<cost mar)in is #ro#ortional to its
mar*et share, so the lar)er $irms ha%e hi)her mar)ins and thus lo.er mar)inal costs,
As a result, shi$tin) out#ut to.ard them, .hich raises concentration, raises .el$are as
.ell, 3his o&ser%ation tells us that an increase in concentration cannot ser%e as a #ro/y
Ch% .1- $ntitrust 1113
$or a decrease in total .el$are .hen studyin) hori4ontal mer)ers,'8 9"t should &e noted
in this re)ard that, ordinarily, .hen en$orcement a)encies and courts consider increases
in concentration, this is %ie.ed dia)nostically and #ros#ecti%ely under the maintained
assum#tion that the share o$ the mer)ed $irm .ill eCual the com&ined #re<mer)er shares
o$ the mer)in) $irms,:
3he a##lica&ility o$ the 2ournot model is limited to industries .here com#etition is
accurately modeled as a Cuantity<settin) )ame, or #erha#s as a ca#acity<settin) )ame
$ollo.ed &y #ricin) com#etition, .ith $airly homo)eneous #roducts, and .here the #redictions
o$ the one<shot 2ournot model 9rather than a model o$ re#eated 2ournot: $it the
industry reasona&ly .ell, 3he 2ournot model is not suita&le $or industries .ith hi)hly
di$$erentiated #roducts, es#ecially i$ ca#acity constraints are unim#ortant in the medium
to lon) run, "n those industries, a Bertrand model .ith di$$erentiated #roducts $its &etter,
5e no. study mer)ers in that model,
2%.%#% Bertrand model with differentiated products
A %ery e/tensi%e literature has de%elo#ed to e/#lore the e$$ects o$ hori4ontal mer)ers
in models o$ Bertrand com#etition .ith di$$erentiated #roducts,'' 3hese models are
e/tensi%ely used in #ractice to estimate and simulate the e$$ects o$ #ro#osed mer)ers,
#articularly in mar*ets .ith &randed #roducts, ran)in) $rom consumer )oods such as
&rea*$ast cereal to com#uter so$t.are,
6enec*ere and 6a%idson 91'8@: #ro%ide a nice entry #oint into this literature, "n contrast
to the results o$ alant, .it4er, and 7eynolds 91'83:, they $ind that mer)ers are
al.ays #ro$ita&le and .ill al.ays in%ol%e #rice increases, ;rior to the mer)er, the #rice
o$ each #roduct .as set to ma/imi4e the #ro$its earned on that #roduct, )i%en the #rices
o$ all other #roducts, Bo. consider .hat ha##ens i$ the #rice o$ one o$ the mer)in)
#roducts, say #roduct 1, is raised sli)htly, 3his .ill lo.er the #ro$its earned on #roduct
1, &ut the $irst<order e$$ect .ill &e 4ero since the #rice o$ #roduct 1 .as already
o#timi4ed, 3he hi)her #rice $or #roduct 1 .ill, ho.e%er, increase sales o$ #roduct 2,
thus raisin) the #ro$its o$ the mer)ed $irm 9a #ositi%e e/ternality that $irm 1 i)nored
#rior to the mer)er:, 3he increase in #ro$its $rom #roduct 2 .ill &e lar)er, the )reater is
the increase in sales o$ #roduct 2 that results $rom the increase in the #rice o$ #roduct 1
and the lar)er is the #rice<cost mar)in on #roduct 2, 5hat a&out chan)es in the #rices
set &y the other $irmsW "n Bertrand eCuili&rium, &est<res#onse cur%es slo#e u#.ards,
so the other $irms .ill $ind it o#timal to raise their #rices in res#onse to the hi)her
#rice $or #roduct 1 9and $or #roduct 2, the #rice o$ .hich it .ill also &e #ro$ita&le to
increase:, 3hese hi)her #rices $or other #roducts increase the demand $or #roducts 1
and 2, $urther addin) to the #ro$its o$ the mer)ed $irm, .hich #ros#ecti%ely ma*es the
'8 Farrell and ha#iro 91''0a: sho. more )enerally ho. chan)es in the o.nershi# o$ assets in 2ournot
oli)o#oly a$$ect out#ut, .el$are, and the --",
'' ee "%aldi et al, 92003&:, Motta 92001, ##, 213O20@:, and es#ecially 5erden and Froe& 9200F: $or more
e/tensi%e re%ie.s o$ this literature, Ba*er and Bresnahan 91'8@: is an im#ortant early contri&ution,
1111 L% 7aplow and C% Shapiro
mer)er e%en more attracti%e,100 Bote also that each non<mer)in) $irm .elcomes the
mer)er since it earns hi)her #ro$its &ecause, as e/#lained, the mer)ed $irm char)es
hi)her #rices $or &oth o$ its #roducts, .hich increases the demand $or the ri%al #roducts,
3hese ideas are %ery )eneral= in models .ith di$$erentiated #roducts and Bertrand
com#etition, mer)ers that in%ol%e no syner)ies are #ro$ita&le $or the mer)in) $irms,
raise the #rices char)ed &y the mer)in) $irms, and raise the #rice and #ro$its o$ the nonmer)in)
$irms as .ell, 2learly, such mer)ers lo.er consumer sur#lusJ they also tend
to lo.er .el$are, "t is #ossi&le that such mer)ers raise .el$are, ho.e%er, i$ they in%ol%e
si)ni$icant syner)ies or i$ the mer)in) $irms are ine$$icient, so shi$tin) out#ut a.ay $rom
them and to.ard the other $irms is e$$icient,
3o a##ly these ideas in #ractice, .here the em#hasis tends to &e on .hether, and ho.
much, a #ro#osed mer)er .ill raise #rice, it is hel#$ul to understand .hat economic
%aria&les tend to ma*e the #rice e$$ects o$ a mer)er &et.een t.o su##liers o$ di$$erentiated
#roducts lar)e or small, 5e return to this issue &elo., .here .e discuss the
so#histicated simulation methods no. used to estimate the #rice e$$ects o$ such mer)ers,
A )ood sense o$ the &asic $orces at .or* can &e )leaned &y com#arin) the #rices
in a Bertrand duo#oly .ith t.o di$$erentiated #roducts, each sold &y one $irm, .ith
the #rice char)ed &y a sin)le $irm sellin) &oth #roducts, Focusin) on Eust t.o #roducts
is not as restricti%e as it mi)ht a##ear= one can inter#ret the demand $unctions $or the
t.o #roducts in this model as demand in a )eneral oli)o#olistic mar*et, ta*in) as )i%en
the #rices o$ all o$ the other #roducts, "n the a&sence o$ any e$$iciencies, the lo)ic o$
6enec*ere and 6a%idson 91'8@: tells us that the mer)ed $irm .ill ha%e an incenti%e
to raise its #rice, )i%en the #rices o$ the other $irms, and that the o#timal #rice $or
the mer)ed $irm, )i%en those other #rices, is less than the ne. Bertrand eCuili&rium
#rice once one accounts $or the #rice increases &y the other $irms, 3here$ore, the #rice
increases calculated usin) a duo#oly model .ill 9some.hat: underestimate the #rice
increases in the $ull oli)o#oly model,
5e deri%ed a $ormula in su&section 2,3,2 $or the di$$erence &et.een the mono#oly
#rice and the Bertrand eCuili&rium #rice in a sim#le, symmetric Bertrand duo#oly model
.ith linear demand and constant mar)inal cost, Follo.in) ha#iro 91''0:, .e sho.ed
that the #ercenta)e )a# &et.een the mono#oly #rice and the Bertrand #rice is )i%en
&y PMPB
PB
=
2(1)
PB&C
PB
, .here dX2
dP1
/|dX1
dP1
| is the di%ersion ratio, that is, the
$raction o$ the lost unit sales o$ #roduct 1, .hen the #rice o$ #roduct 1 is raised, that
are ca#tured as unit sales o$ #roduct 2, as #re%iously de$ined in su&section 2,3,2, "$ .e
de$ine the #re<mer)er #rice<cost mar)in as m PB&C
PB
, then, in this %ery sim#le model,
the #ercenta)e #rice increase #redicted $rom the mer)er o$ the t.o $irms is
2(1)m,
100 3he lo)ic in the 2ournot case is di$$erent &ecause &est<res#onse $unctions slo#e do.n, 5hen the mer)ed
$irm o#timally reduces its out#ut, the other $irms e/#and out#ut, .hich reduces the #ro$its o$ the mer)ed $irm,
Ch% .1- $ntitrust 111@
3his #rice increase is #ro#ortional to the #re<mer)er #rice<cost mar)in, m, 3his com#orts
.ith intuition= since the #ro$its )ained on the sale o$ #roduct 2, .hen the #rice o$
#roduct 1 is raised, are #ro#ortional to the mar)in on #roduct 2, the ma)nitude o$ the
mar)in on #roduct 2 is #ro#ortional to the incenti%e to increase the #rice o$ #roduct 1
9and con%ersely:, 3here$ore, ceteris pari!us, mer)ers &et.een $irms sellin) di$$erentiated
#roducts are li*ely to raise #rice more, the )reater are the #re<mer)er mar)ins on
their #roducts,
3he #rice increase associated .ith the mer)er is also #ro#ortional to the $actor
1 ,
.hich is increasin) in the di%ersion ratio and .hich is 4ero i$ the di%ersion ratio is
4ero,101 3his, too, is intuiti%e= the )reater is the di%ersion ratio, the )reater is the share
o$ the lost sales $rom #roduct 1 that are ca#tured &y #roduct 2 and thus internali4ed
a$ter the mer)er, 3here$ore, ceteris pari!us, mer)ers &et.een $irms sellin) di$$erentiated
#roducts are li*ely to raise #rice more, the closer is the de)ree o$ su&stitution &et.een
their #roducts, as measured usin) the di%ersion ratio, Bote that a hi)h )ross mar)in is
consistent .ith a hi)h di%ersion ratioJ this #attern arises i$ the demand $or #roduct 1 is
not %ery elastic and i$ a si)ni$icant $raction o$ the 9relati%ely $e.: sales lost .hen the
#rice o$ #roduct 1 rises are di%erted to #roduct 2,
As sho.n &y ha#iro 91''0:, ho.e%er, a rather di$$erent $ormula a##lies .ith
constant<elasticity 9rather than linear: demand, "n this case, the #ercenta)e #rice increase
#redicted $rom the mer)er o$ the $irms 1 and 2 is m
1m,102 3his ratio is lar)er
than in the case o$ linear demand, and #ossi&ly much lar)er $or #lausi&le #arameter
%alues, 3o illustrate, su##ose that the #re<mer)er )ross mar)in is m = 0.3@, not an uncommon
num&er $or &randed #roducts, and that one<Cuarter o$ the sales lost .hen the
#rice o$ #roduct 1 is raised are ca#tured &y #roduct 2 9and %ice %ersa:, so = 0.2@,
5ith these #arameters, the #ost<mer)er #rice increase .ith linear demand is a&out 0S,
.hile the #ost<mer)er #rice increase $or constant<elasticity demand is nearly 22S,
u##ose that one o&ser%ed the #re<mer)er mar)in o$ 3@S and .as a&le to estimate
the di%ersion ratio o$ 2@S &et.een these t.o #roducts, Both o$ these modelsKone .ith
linear demand, one .ith constant elasticity o$ demandKcan &e #arameteri4ed to &e
consistent .ith these o&ser%ations, Iet the t.o models )i%e si)ni$icantly di$$erent #redictions
$or the #rice increase associated .ith a mer)er &ecause the t.o demand systems
di%er)e some.hat as #rices de#art $rom their #re<mer)er eCuili&rium le%els, 3his should
not &e totally sur#risin)= mer)ers are discrete e%ents, and i$ nontri%ial #rice chan)es are
#ossi&le, their ma)nitude must in $act de#end u#on demand at #rices distinctly di$$erent
$rom the #re<mer)er #rices,
All o$ this tells us that, in a mer)er in%ol%in) di$$erentiated #roducts, ma*in) relia&le
#redictions o$ unilateral #rice e$$ects &ased on a model o$ Bertrand oli)o#oly
reCuires an accurate structural model o$ the demand system, and that the sha#e o$ the
demand system at #rices some distance a.ay $rom the #re<mer)er eCuili&rium a$$ects
101 5e reCuire < 1 or else the mer)ed entity $aces #er$ectly inelastic demand at all #ositi%e #rices,
102 5e reCuire 1 m > 0 so that the elasticity o$ demand $acin) the mer)ed $irm is )reater than unity,
1110 L% 7aplow and C% Shapiro
the #ost<mer)er #rice increase,103 "$ a structural model can &e estimated that $its industry
demand, then, .ith luc*, the #ost<mer)er eCuili&rium can &e simulated usin) that
model, there&y #redictin) the ma)nitude o$ #ost<mer)er #rice increases, 3his #romisin)
a##roach has &een e/#lored analytically and a##lied in #ractice in recent years, as
.e discuss in su&section 1,0,2, 3he lo)it model $or di$$erentiated #roducts, cham#ioned
&y5erden and Froe& 9200F:, is es#ecially tracta&le and has &een used e/tensi%ely
to estimate the e$$ects o$ hori4ontal mer)ers, "n this model, each consumer #ic*s one
unit o$ a sin)le &rand $rom a set o$ choices that includes the di$$erentiated #roducts,
i = 1, . . . , N, alon) .ith the alternati%e o$ an outside )ood, .hich can sim#ly &e inter#reted
as #ic*in) none o$ these #roducts, 3he consumerGs utility $rom selectin) &rand
i = 1, . . . , N is the sum o$ a MsystematicN com#onent associated .ith that &rand, (i ,
and an uno&ser%a&le idiosyncratic com#onent, Under suita&le assum#tions a&out the
distri&ution o$ the idiosyncratic terms, the #ro&a&ility that a consumer .ill #ic* &rand i
is )i%en &y )i = #(i /N
*=0 #(* , .here the inde/ 4ero corres#onds to the outside )ood,
"$ .e de$ine + = N
*=1 )* as the #ro&a&ility that the consumer .ill #ic* one o$ the N
&rands 9rather than the outside )ood:, then $irm iGs mar*et share is Si = )i/+, so the
mar*et shares are #ro#ortional to the choice #ro&a&ilities )i ,
"n the sim#le s#eci$ication descri&ed in 5erden and Froe& 9200F:, the systematic
com#onent o$ utility $or &rand i = 1, . . . , N is )i%en &y (i = ,i-Pi , .here ,i re$lects
the underlyin) Cuality or a%era)e attracti%eness o$ &rand i, Pi is the #rice o$ &rand i,
and - is a constant that determines the de)ree o$ su&stituta&ility amon) the di$$erent
#roducts, For lar)e %alues o$ -, the com#etin) &rands are %ery close su&stitutes, and
#rice<cost mar)ins are lo., 6i$$erentiatin) the demand $or &rand i .ith res#ect to the
#rice o$ &rand i )i%es d)i/dPi = -)i (1)i ), 3rans$ormin) this e/#ression into elasticity
$orm, the o.n<#rice elasticity $or &rand i is )i%en &y -Pi (1)i ), 6i$$erentiatin)
the demand $or &rand i .ith res#ect to the #rice o$ &rand j )i%es d)i/dPj = -)i)j ,
3rans$ormin) this e/#ression into elasticity $orm, the cross<#rice elasticity o$ demand
$or &rand i .ith res#ect to the #rice o$ &rand j is -Pj)j , 3here$ore, the di%ersion ratio
$rom &rand j to &rand i, .hen the #rice o$ &rand j rises, is )i%en &y dXi/dPj
|dXj /dPj | = )i
1)j
,
3his model has the attracti%e, &ut restricti%e, #ro#erty that the di%ersion ratio $rom
&rand j to &rand i is #ro#ortional to $irm iGs mar*et share, "n this im#ortant sense, the
lo)it model is the antithesis o$ s#atial models in .hich some #roducts are %ery close
su&stitutes, others are distant su&stitutes, and #ro/imity need not &ear any #articular
relationshi# to #o#ularity, 3he lo)it model is a )ood startin) #oint in a situation .here
all o$ the &rands com#ete a)ainst one another and it is not clear .hich are McloseN to each
other, Bested lo)it models can &e used .hen additional in$ormation a&out #ro/imity is
a%aila&le,
5erden and Froe& 9200F: sho. that, in the Bertrand eCuili&rium .ith sin)le#roduct
$irms, the )a# &et.een $irm iGs #rice and $irm iGs mar)inal cost is )i%en &y
103 All o$ these ideas carry o%er to mer)ers &et.een multi<#roduct $irms, &ut the #ertinent calculations are
more com#le/,
Ch% .1- $ntitrust 111F
Pi &Ci = 1
-(1)i ) , 3his e/#ression tells us that the $irms .ith more attracti%e
#roducts, and thus lar)er mar*et shares, ha%e hi)her mar*u#s, much li*e $irms .ith
lo.er costs and thus lar)er shares ha%e hi)her mar)ins in a 2ournot eCuili&rium, "n
the Bertrand eCuili&rium that results a$ter the mer)er o$ &rands 1 and 2, the eCuili&rium
)a# &et.een #rice and mar)inal cost $or each o$ the mer)in) &rands is )i%en &y
P1 &C1 = P2 &C2 = 1
-(1)1)2) , 97emem&er, in inter#retin) this eCuation, that
the mar*et shares o$ the mer)in) &rands are not constantsJ they .ill $all as a result o$
the mer)er,: 3o illustrate, consider the situation in .hich #roduct 2, say, is inherently
more attracti%e, that is, in .hich ,2 > ,1, For sim#licity, su##ose that &oth #roducts
are #roduced at constant and eCual mar)inal cost, ;rior to the mer)er, #roduct 2 .ould
ha%e a lar)er mar*et share and a lar)er )a# &et.een #rice and mar)inal cost than .ould
#roduct 1, A$ter the mer)er, the #rices o$ &oth #roducts .ould &e hi)her than their #remer)er
le%els, and the )a# &et.een #rice and mar)inal cost $or the t.o #roducts .ill &e
eCual, .hich in this case $urther im#lies that the #rices .ill &e eCualJ #roduct 2 .ould
ha%e a lar)er mar*et share than #roduct 1, 3here$ore, the #ost<mer)er #rice increase
.ill &e lar)er $or #roduct 1 than $or #roduct 2, 3his $its .ith intuition= the incenti%e to
raise the #rice o$ #roduct 1 is )reater since a relati%ely lar)e $raction o$ its sales .ill &e
di%erted to #roduct 2, due to that #roductGs #o#ularity, Furthermore, the #re<mer)er )a#
&et.een #rice and mar)inal cost $or #roduct 2 is lar)er than that $or #roduct 1, so any
di%erted sales are actually addin) to the #ro$its o$ the mer)ed entity,
3he symmetric lo)it model .ith constant mar)inal cost can readily )enerate #redictions
a&out the #rice e$$ects o$ mer)ers, )i%en an estimate o$ the elasticity o$ demand
$or the mar*et as a .hole and an estimate o$ the #re<mer)er )a#s &et.een #rices and
cost, As an e/am#le, 5erden and Froe& 9200F: re#ort that .ith si/ 9symmetric: $irms,
a mar*et elasticity o$ demand o$ 0.@, a normali4ed #re<mer)er #rice o$ [1, and a #remer)er
)a# &et.een #rice and mar)inal cost o$ [0,10, so mar)inal cost is [0,00 9all this
corres#ondin) to - 2.':, the mer)er o$ any t.o &rands causes their #rices to increase
&y a&out 0S, As they note, the lo)it model, .ith its lac* o$ locali4ation in com#etition,
sho.s that a mer)er &et.een t.o &rands can easily raise #rice si)ni$icantly e%en i$ the
mer)in) &rands are not each otherGs ne/t closest su&stitutes in any mar*et<.ide sense,
;rices rise &ecause, .ith only si/ $irms, there are a nontri%ial $raction o$ consumers $or
.hom the mer)in) &rands are the $irst and second choices, "n this model, the mer)ed
$irm cannot identi$y and #rice discriminate a)ainst those consumers, so the mer)ed $irm
raises #rice some.hat to all consumers,101
Until no., the analysis has $ocused on the #rice e$$ects o$ mer)ers that in%ol%e no
#roduction syner)ies, 3he consideration o$ e$$iciencies is $acilitated &y a con%enient
$eature o$ models o$ Bertrand com#etition .ith di$$erentiated #roducts= the ma)nitude
o$ the e$$iciencies necessary $or a mer)er to reduce rather than raise #rice de#ends only
101 "$ #rice discrimination .ere #ossi&le, the mer)ed $irm .ould raise #rice much more to the identi$ia&le
customers .ho ran*ed #roduct 1 and #roduct 2 as their $irst and second choice, and not at all to other customers,
E$$ecti%ely, one can com#ute a ne. #ost<mer)er eCuili&rium $or each identi$ia&le customer or )rou#,
1118 L% 7aplow and C% Shapiro
u#on the sha#e o$ the demand system 9the di%ersion ratio &et.een the mer)in) #roducts:
at #rices in the immediate nei)h&orhood o$ the #re<mer)er eCuili&rium #rices,
A reduction in mar)inal cost o$ #roduct 1 at the mer)ed $irm increases the )a# &et.een
the $irmGs #rice and mar)inal cost on that #roduct, )i%in) the $irm an incenti%e to lo.er
its #rice, ;rice .ill in $act $all i$ this incenti%e is stron)er than the incenti%e to raise
the #rice o$ #roduct 1 &ased on internali4in) the di%ersion to #roduct 2, no. o.ned
&y the same $irm, ince &oth o$ these e$$ects are e%aluated at the #re<mer)er #rices, no
in$ormation is reCuired a&out the sha#e o$ the demand system at other #rices,
Based on this lo)ic, 5erden 91''0: deri%es an e/#ression $or the cost reductions
necessary to #re%ent a mer)er $rom raisin) #rice,10@ "n the symmetric case, .here the
t.o mer)in) $irms ha%e eCual mar*et shares and )ross mar)ins #rior to the mer)er, he
sho.s that a mer)er .ill reduce #rice i$ and only i$ the cost reduction satis$ies
9F:
&C1 &C12
&C1
>
m
1 m

1
,
.here m P1&C1
P1
is a)ain the #re<mer)er mar)in, 3his is a rather strin)ent condition
in mer)ers &et.een close ri%als, Usin) our #re%ious numerical e/am#le o$ m = 0.3@
and = 0.2@, the mer)er must reduce mar)inal cost &y a&out 18S to lead to a #rice
reduction rather than a #rice increase, Bote that reductions in $i/ed cost ha%e no &earin)
on 9short<run: #rice e$$ects,
2%.%4% Biddin* models
"n the Bertrand model, each $irm sets a #rice, and &uyers ma*e their #urchasin) decisions
)i%en these #rices, Bertrand models are es#ecially .ell<suited $or mar*ets .ith
di$$erentiated consumer #roducts in .hich there are a lar)e num&er o$ relati%ely #assi%e
consumers,100 "n many other settin)s, ho.e%er, there are lar)e &uyers .ho &eha%e
strate)ically, desi)nin) their #rocurement #rocedures so they can o&tain the &est #rice
$rom their su##liers, "n these settin)s, com#etition ty#ically ta*es the $orm o$ &iddin)
to .in the &usiness o$ a sin)le customer .ho has desi)ned a #rocurement #rocedure,
Many #urchasin) situations $it this #attern, includin) #rocurement auctions, 3he
#recise manner in .hich com#etition ta*es #lace de#ends u#on the auction rules esta&lished
&y the customer, Dlem#erer 92001: #ro%ides an e/cellent o%er%ie. o$ the
enormous literature on auctions, 5erden and Froe& 9200F: discuss mer)er analysis in
a situation .here a seller is auctionin) o$$ an item usin) an ascendin) oral auction and
the &idders ha%e #ri%ate %alues $or the item, 9;recisely the same ideas .ould arise in a
10@ 3his is the analo)ue in a Bertrand model o$ the necessary and su$$icient condition $or a mer)er to reduce
#rice in 2ournot oli)o#oly deri%ed &y Farrell and ha#iro 91''0&:,
100 E%en in mar*ets $or &randed consumer #roducts, lar)e &uyers such as lar)e retailers may #lay a si)ni$icant
role, 3hese &uyers may &e more acti%e and strate)ic in dealin) .ith manu$acturers, in #art &y settin) u#
&iddin) contests amon) their .ould<&e su##liers,
Ch% .1- $ntitrust 111'
situation .here a &uyer is runnin) a #rocurement auction and the &idders are su##liers
.ho di$$er in their costs o$ ser%in) the &uyer,: 3his auction $ormat is eCui%alent to a
second<#rice sealed<&id auctionJ it is a dominant strate)y $or each &idder to &id u# to its
%alue, and the #rice ultimately #aid, P, is eCual to the second<hi)hest %aluation amon)
the &idders,
"n this conte/t, consider a mer)er &et.een &idder 1 and &idder 2, and la&el the t.o
&idders so that &idder 1Gs %aluation, B1, is at least as lar)e as &idder 2Gs %aluation,
B2, ?a&el &idder 3 as the one .ith the hi)hest %aluation, B3, amon) the other &idders,
A mer)er &et.een &idder 1 and &idder 2 9.hich is eCui%alent here to collusion &et.een
these t.o &idders: .ill ha%e no e$$ect on the #rice #aid $or the item unless B2 > B3,
that is, unless the t.o mer)in) &idders ha%e the t.o hi)hest %aluations on the item, "$
they do, #rice .ill $all $rom B2 to B3, Qie.ed statistically, mer)er e$$ects de#end on
the Eoint distri&ution o$ the %aluations o$ the &idders, includin) the mer)in) &idders,
5aehrer and ;erry 92003: sho. ho. the #rice e$$ect o$ a mer)er can &e estimated $or
certain cumulati%e distri&utions o$ %aluations,
2%#% +li*opoly theory and coordinated effects
Mer)ers also can #ose a ris* to com#etition &y increasin) the li*elihood that a collusi%e
outcome .ill #re%ail, uch coordinated<e$$ects theories o$ harm $rom hori4ontal
mer)ers are $eatured in the Mer)er Auidelines, .hich state in X2,1= MA mer)er may
diminish com#etition &y ena&lin) the $irms sellin) in the rele%ant mar*et more li*ely,
more success$ully, or more com#letely to en)a)e in coordinated interaction that harms
consumers,N Mer)er en$orcement &ased on coordinated e$$ects is more im#ortant, the
more one &elie%es that increased concentration contri&utes to coordinated outcomes and
the less one &elie%es that collusi%e &eha%ior is readily deterred &y antitrust la.,
As discussed in section 3, collusion is )enerally thou)ht to &e easier to achie%e and
sustain .hen there are $e.er su##liers in the industry, 3here$ore, at the sim#lest le%el,
reducin) the num&er o$ com#etitors &y one tends to increase the li*elihood o$ collusion,
3his idea underlies .hat is re$erred to as the Mstructural #resum#tionNK that increases
in concentration lead to less com#etiti%e interactionsKthat has lon) #layed a central
role in antitrust, 3he heyday o$ the structural #resum#tion corres#onded .ith a time
.hen industrial or)ani4ation economists de%oted su&stantial e$$orts to %alidatin) em#irically
the core idea o$ the structure<conduct<#er$ormance #aradi)m= mar*ets that are
hi)hly concentrated tend to ha%e hi)her #rices and hi)her #ro$its, and thus tend to ser%e
consumers less .ell, than do mar*ets .ith more com#etiti%e structures, ceteris pari!us,
6emset4 91'F3: mounted a stron) attac* on those .ho claimed that a #ositi%e crosssectional
relationshi# &et.een concentration and #ro$its .as indicati%e o$ mar*et $ailure
or the need $or an inter%entionist antitrust #olicy, 6emset4 #ointed out that a #ositi%e
correlation .ould also arise i$ some $irms .ere more e$$icient than their ri%als, and
i$ the more e$$icient $irms had lar)e mar*et shares, Mar*et concentration .ould then
result $rom the #resence o$ lar)e, e$$icient $irms, Under this hy#othesis, small $irms in
concentrated mar*et .ould earn normal #ro$its, .ith the lar)e, e$$icient $irm earnin)
11@0 L% 7aplow and C% Shapiro
#ro$its due to 7icardian rents, "$ mar)ins are associated .ith a $irmGs mar*et share, not
o%erall mar*et concentration, this may .ell re$lect the )reater e$$iciency o$ lar)er $irms,
at least in the short run, 3he im#lications $or mer)er #olicy are #ro$ound= i$ a lar)e $irm
see*s to &uy a smaller ri%al, the resultin) increase in concentration mi)ht )o alon) .ith
lo.er #rices and consumer &ene$its i$ the lar)e and e$$icient $irm is a&le to im#ro%e the
e$$iciency .ith .hich the assets o$ the smaller acCuired $irm are used, Bor* 91'F8: is
also .ell *no.n $or attac*in) the #resum#tions under mer)er #olicies o$ the 1'00s and
1'F0s,
7e%ie.in) the enormous literature on the cross<sectional industry relationshi#s
amon) concentration, #rices, mar)ins, and #ro$its in order to distin)uish amon) these
com#etin) hy#otheses is &eyond the sco#e o$ this cha#ter, chmalensee 91'8': and
alin)er 91''0: are )ood startin) #laces $or readers interested in learnin) more, ;autler
92003: #ro%ides a more recent summary o$ the literature, .hich has made #ro)ress in
distin)uishin) e$$ects on a $irmGs #ro$its that are related to mar*et concentration, the
$irmGs mar*et share, or the $irmGs identity 9loo*in) across multi#le mar*ets:, !%erall,
economists ha%e )ro.n less con$ident o%er the #ast se%eral decades in statin) that there
is a systematic relationshi# &et.een mar*et concentration and mar*et #er$ormance, at
least o%er the ran)e o$ mar*et structures in .hich there are more than t.o or three $irms,
E%en so, the cautionary statement made &y alin)er 91''0, #, 28F: &ears re#eatin) today=
First, des#ite the .ell<*no.n #ro&lems .ith this literature, it continues to a$$ect
antitrust #olicy, 3he ina##ro#riate in$erences used to Eusti$y an acti%e antitrust #olicy
ha%e )i%en .ay to eCually incorrect in$erences that ha%e &een used to Eusti$y a
rela/ed mer)er #olicy, econd, the alternati%e to cross<industry studies is to study
s#eci$ic industries, , , , U"Vt is im#ortant to reali4e that it .as the $ailure o$ studies
o$ indi%idual industries to yield )eneral insi)hts that made cross<industry studies
#o#ular,
5hate%er one thin*s o$ this literature, one should &ear in mind that these cross<industry
studies do not directly measure the e$$ects o$ hori4ontal mer)ers, .hich .e ta*e u# in
su&section 1,3, 3he #rimary %ariation studied is across industries, not .ithin an industry
o%er time, Furthermore, throu)h the early 1'80s, hi)hly concentratin) hori4ontal
mer)ers .ould sim#ly not ha%e &een allo.ed, o, to the e/tent that one sees e$$icient
lar)er $irms in certain industries, throu)h at least the early 1'80s these $irms mostly
arose throu)h internal )ro.th, non<hori4ontal mer)ers, or hori4ontal mer)ers in%ol%in)
$irms .ith relati%ely small mar*et shares, not throu)h hi)hly concentratin) hori4ontal
mer)ers,
3he *ey Cuestion re)ardin) coordinated e$$ects in mer)er analysis is .hether a )i%en
mer)er .ill si)ni$icantly increase the li*elihood that a collusi%e outcome .ill arise,10F
10F 2oncern .ould also arise i$ the mer)er ma*es collusion more e$$ecti%e, $or e/am#le, &y raisin) the #rice
at .hich collusion can &e maintained to a le%el closer to the mono#oly #rice or &e reducin) the $reCuency
and duration o$ #rice .ars, For sim#licity, in our discussion &elo. .e use the shorthand o$ tal*in) a&out the
li*elihood that a collusi%e outcome .ill arise,
Ch% .1- $ntitrust 11@1
"n section 3, .e e/#lored in considera&le detail ho. %arious industry conditions a$$ect
the li*elihood o$ e$$ecti%e collusion, All o$ that theory and e%idence can &e &rou)ht
to &ear .hen considerin) coordinated e$$ects in hori4ontal mer)ers, 5hile .e lac*
methods, such as those .e Eust discussed re)ardin) unilateral e$$ects, to Cuanti$y these
coordinated e$$ects, .e *no. Cuite a lot a&out ho. a chan)e in mar*et structure resultin)
$rom a mer)er .ill a$$ect the li*elihood o$ e$$ecti%e collusion, "n #rinci#le, then,
one can trade o$$ the increased costs $rom #otential collusion a)ainst any e$$iciencies
associated .ith a mer)er,
ome hi)hly rele%ant and ro&ust lessons emer)ed $rom the analysis o$ collusion in
section 3, A hori4ontal mer)er &et.een t.o si)ni$icant su##liers, &y reducin) the num&er
o$ #layers &y one, can si)ni$icantly increase the li*elihood that the remainin) $irms
.ill &e a&le to reach a collusi%e a)reement, !ne #ossi&ility is that a mer)er may esta&lish
a clear mar*et leader .ho can #lay the role o$ #rice leader, ser%in) the $unction o$
esta&lishin) and adEustin) collusi%e #rices, .ith the other $irms $ollo.in), ;erha#s most
im#ortant, .hen a $irm that .ould ha%e &een reluctant to Eoin in a collusi%e scheme
9.hich, as #re%iously noted, is sometimes termed a ma%eric*: is acCuired &y another
su##lier .ho is lar)er or other.ise more inclined to #artici#ate, collusion can &e )reatly
$acilitated, Beyond these #oints, a mer)er reduces the num&er o$ &ilateral lin*s &et.een
$irms in a mar*et, .hich is some measure o$ the di$$iculty o$ reachin) an a)reement,
5ith N su##liers, the num&er o$ such lin*s is N(N 1)/2, A @<to<1 mer)er reduces
the num&er o$ lin*s $rom 10 to 0J a 1<to<3 mer)er reduces the num&er o$ lin*s $rom 0
do.n to 3,
For similar reasons, hori4ontal mer)ers also can ma*e it easier to sustain a collusi%e
outcome, A $irm .ith a lar)er mar*et share tends to ha%e less to )ain $rom cheatin) and
more to lose i$ a #rice .ar eru#ts than do smaller $irms, As a result, the mer)er o$ t.o
smaller $irms may increase the #rice at .hich collusion can &e sustained, "n )eneral,
a mer)er that si)ni$icantly increases concentration .ill tend to ma*e cheatin) on the
collusi%e #rice less attracti%e, at least $or the mer)in) #arties,
3hese o&ser%ations are surely im#ortant $or mer)er en$orcement #olicy, e%en i$ our
*no.led)e a&out the relationshi# &et.een industry conditions and the li*elihood o$ collusion
does not )i%e us a s#eci$ic Cuantitati%e #rocedure to .ei)h the increased dan)er
o$ collusion in, say, a 1<to<3 mer)er a)ainst e$$iciencies #romised &y that mer)er, -o.e%er,
a #arado/ o$ #roo$ 9di$$erent $rom the one that .e noted in su&section 3,1,2: can
#resent some #ro&lems .hen one see*s to a##ly collusion theory to hori4ontal mer)ers,
3o illustrate .ith an o%erly shar# e/am#le, su##ose that one concludes in a )i%en industry
that e$$ecti%e collusion is Cuite unli*ely i$ there are $i%e or more $irms, #ossi&le
&ut not li*ely i$ there are $our $irms, and Cuite li*ely i$ there are three or $e.er $irms,
2oncerns a&out coordinated e$$ects .ould there$ore &e minimal $or any mer)er that le$t
at least $i%e $irms in the industry, A mer)er $rom @ to 1 $irms .ould &e a cause $or concern,
as .ould &e a mer)er $rom 1 to 3 $irms, But e%en more concentratin) mer)ers,
$rom 3 to 2 $irms, and #erha#s e%en a mer)er to mono#oly, .ould cause $e.er concerns=
collusion is hy#othesi4ed to &e li*ely .ith or .ithout these mer)ers, 5hile this is surely
too stron) a conclusionKe%en .ith only t.o $irms, there #ro&a&ly is a nontri%ial chance
11@2 L% 7aplow and C% Shapiro
that collusion .ill &rea* do.nKthis lo)ic at least undermines the standard #resum#tion
that mer)ers &ecome more .orrisome as the num&er o$ $irms declines, 3his additional
#arado/ is a%oided only i$ one &elie%es that the #ro&a&ility o$ success$ul collusion is
not Eust declinin) in the num&er o$ $irms &ut also is a con%e/ $unction o$ the num&er o$
$irms,
3here is relati%ely little $ormal theory e/#lorin) the im#lications $or mer)er #olicy
o$ the relationshi# &et.een collusion and mar*et concentration, a#art $rom the #a#ers
already discussed in section 3, But se%eral o$ the them are es#ecially #ertinent $or e%aluatin)
coordinated e$$ects in hori4ontal mer)ers, Bota&ly, 2om#te, Jenny, and 7ey 92002:
and Qasconcelos 9200@: as* ho. the distri&ution o$ ca#acities a$$ects the a&ility o$ the
$irms to sustain collusion in #rice<settin) and Cuantity<settin) su#er)ames, res#ecti%ely,
6a%idson and 6enec*ere 91'81: #oint out that re%ertin) to the static Bash eCuili&rium
ty#ically is a less se%ere #unishment .hen there are $e.er $irms 9in a Cuantity<settin)
su#er)ame or a #rice<settin) su#er)ame .ith ca#acity constraints:, ma*in) $or a com#le/
relationshi# &et.een mar*et concentration and the li*elihood o$ collusion,
Do%acic et al, 92000: #ro#ose an interestin) ne. .ay to Cuanti$y the dan)ers associated
.ith coordinated e$$ects in a situation .here a num&er o$ su##liers are &iddin) $or
the customerGs #atrona)e, 3hey #ro#ose measurin) the e$$ects o$ incremental collusion,
that is, collusion that only in%ol%es t.o $irms, &e$ore and a$ter the #ro#osed mer)er,
3hey sho. ho. this calculation can &e #er$ormed in a #articular &iddin) model, 5hile
a lar)e num&er o$ calculations are necessary to im#lement their method, these calculations
are all .ell rooted in oli)o#oly theory, and in $act use the results already discussed
in the analysis o$ unilateral e$$ects,
Ba*er 92002: has em#hasi4ed the im#ortant role o$ ma%eric* $irms in disru#tin) or
#re%entin) collusion and thus the #articular dan)ers that arise .hen a mer)er eliminates
such a $irm 9an idea em&raced in the Mer)er Auidelines as .ell:, 2ollusion theory
indicates that reachin) an a)reement and sustainin) an a)reement may &e di$$icult i$
one o$ the $irms e/#ects to )ain si)ni$icant mar*et share in the a&sence o$ collusion,
3here$ore, $irms .ith strate)ies, #roducts, or costs that are distinct $rom those o$ their
ri%als, and $irms that are o#timistic and )ro.in) ra#idly, #erha#s &ecause they recently
entered the mar*et, are o&%ious candidates to &e ma%eric*s, Accordin)ly, Ba*er ad%ocates
an a##roach to mer)er en$orcement #olicy that )oes &eyond the measurement o$
increases in mar*et concentration &y em#hasi4in) the identi$ication o$ ma%eric*s, -e
ar)ues that #lacin) the $ocus on identi$yin) ma%eric* $irms .ill reduce Eudicial errors
&y allo.in) the en$orcement a)encies and the courts to identi$y more accurately those
mer)ers that are li*ely to ha%e coordinated anticom#etiti%e e$$ects $or any )i%en le%el
o$ and chan)e in mar*et concentration, -e also notes that a mer)er may actually create
a ne. ma%eric*,
2%4% Empirical evidence on the effects of hori9ontal mer*ers
Ai%en the lar)e num&er o$ mer)ers that are consummated e%ery year, includin) many
hori4ontal mer)ers, one mi)ht thin* that there .ould &e e/tensi%e, de$initi%e e%idence
Ch% .1- $ntitrust 11@3
re)ardin) the e$$ects o$ these mer)ers, Under .hat circumstances ha%e hori4ontal mer)ers
&een $ound to raise or lo.er #rices, or more )enerally to &ene$it or harm consumersW
And .hat has their im#act &een on the #ro$its o$ the mer)in) #arties and on the #ro$its
o$ their ri%alsW
adly, there is no such clear and de$initi%e &ody o$ e%idence, 3o some e/tent, this
re$lects a lac* o$ data= e%en in those cases .here one can accurately measure the #rices
char)ed &e$ore and a$ter a mer)er, it may &e hard to attri&ute #rice chan)es to the mer)er
rather than to other chan)es in industry conditions, Also, the e$$ects o$ a mer)er may
arise in non<#rice dimensions such as #roduct Cuality, customer ser%ice, or inno%ation,
Furthermore, i$ mer)er en$orcement #olicy is .or*in) .ell, the mer)ers most li*ely to
ha%e lar)e ad%erse #rice e$$ects are ne%er #ro#osed or are &loc*ed on antitrust )rounds,
5e do not mean to su))est that it is im#ossi&le to identi$y the e$$ects o$ hori4ontal
mer)ersJ &ut nor is it easy, ee F32 92002: $or some recent e%idence,
;autler 92003: o$$ers an e/tensi%e re%ie. o$ em#irical .or* on the e$$ects o$ mer)ers
and acCuisitions, 7eaders interested in e/#lorin) this literature in )reater detail
should turn to his #a#er, .hich contains a treasure tro%e o$ in$ormation on the su&Eect,
5hinston 92000, ##, 110O12F: also #ro%ides a %alua&le discussion o$ the e%idence,
5e e/amine here se%eral distinct methods $or identi$yin) and measurin) the e$$ects o$
mer)ers, "n e%aluatin) this e%idence, one should &ear in mind that o%er the #ast 2@ years,
only a&out 2S to 1S o$ the mer)ers re#orted e%ery year under the -art<cott<7odino
Act .ere considered to raise su$$icient antitrust issues to .arrant a second reCuest $rom
the F32 or the 6!J, so data on the e$$ects o$ all mer)ers may not re$lect the e$$ects o$
the maEor hori4ontal mer)ers that are most li*ely to &e scrutini4ed &y antitrust authorities,
108
2%4%.% Stoc maret prices
!ne .ay to measure the e$$ects o$ mer)ers is to study the stoc* mar*et #er$ormance o$
the mer)in) $irms, Usually, this is done usin) an e%ent study around the time o$ the announcement
o$ the mer)er, 3his a##roach has &een e/tensi%ely e/#lored in the $inance
literature, Andrade, Mitchell, and ta$$ord 92001: #ro%ide an e/cellent introduction,
3he ad%anta)e o$ this a##roach is that it relies on detailed and accurate stoc* mar*et
data, -o.e%er, &y its nature, this a##roach cannot distin)uish &et.een $a%ora&le stoc*
mar*et returns &ased on e$$iciencies %ersus mar*et #o.er,10' "n addition, this a##roach
108 ;autler 92003: #ro%ides data on F32 and 6!J second reCuests, ee F32 and 6!J 92003: and 6!J
92000: $or more detailed recent data on mer)er challen)es, ?eary 92002: also #ro%ides some data on mer)er
en$orcement acti%ities and mer)er acti%ity, Ba*er and ha#iro 9in #ress: u#date these data and comment on
the inter#retation o$ en$orcement data,
10' "n #rinci#le, a mer)er that .ould lead to syner)ies and lo.er #rices .ould de#ress the stoc* mar*et %alue
o$ ri%als, .hile an anticom#etiti%e mer)er that .ould lead to hi)her #rices throu)h unilateral or coordinated
e$$ects .ould &oost the stoc* mar*et %alue o$ ri%als, ;autler 92003: re%ie.s studies that attem#t to measure
the im#act o$ hori4ontal mer)ers on the stoc* #rice o$ ri%als, uch e$$ects are more di$$icult to measure
11@1 L% 7aplow and C% Shapiro
measures the e/#ectations o$ in%estors a&out mer)er e$$ects, not the actual e$$ects o$
mer)ers, Furthermore, this literature is not $ocused on hori4ontal mer)ers, 3hus, the
$inance literature is &est seen as addressin) a more )eneral Cuestion= do mer)ers and
acCuisitions #roduce .ealth $or shareholders or do they re$lect mana)erial hu&risW Finally,
e%ent studies do not readily disentan)le #redicted e$$ects o$ the mer)er and other
in$ormation that may &e si)naled &y the announcement,
Andrade, Mitchell, and ta$$ord 92001: re#ort a&normal ne)ati%e returns $or acCuirin)
$irms, &ased on 1801 deals $rom the 1''0s= 1,0S durin) a three<day .indo. around
the announcement and 3,'S durin) a lon)er .indo. $rom 20 days #rior to the announcement
throu)h closin) o$ the deal, -o.e%er, tar)et $irms sho.ed a 10S a&normal
#ositi%e return durin) the three<day .indo., 3he com&ined $irms )ained a&out 1,@S
o%er the short or lon)er .indo., 3hey also re#ort se%eral studies that $ound ne)ati%e
a&normal returns o%er the three to $i%e years $ollo.in) the com#letion o$ mer)ers, statin)
9#, 112:= M"n $act, some authors $ind that the lon)<term ne)ati%e dri$t in acCuirin)
$irm stoc* #rices o%er.helms the #ositi%e com&ined stoc* #rice reaction at announcement,
ma*in) the net .ealth e$$ect ne)ati%e,N -o.e%er, Andrade et al, are s*e#tical o$
these results, dis#utin) the relia&ility o$ these lon)er<term studies, in #art since it is hard
to *no. .hat the MnormalN return should &e o%er these lon)er #eriods o$ time,
"n the end, Andrade et al, state 9#, 11F:= M5e are inclined to de$end the traditional
%ie. that mer)ers im#ro%e e$$iciency and that the )ains to shareholders at mer)er announcement
accurately re$lect im#ro%ed e/#ectations o$ $uture cash $lo. #er$ormance,
But the conclusion must &e de$ended $rom se%eral recent challen)es,N !ne o$ these
challen)es arises $rom the $act that the source o$ the stoc* mar*et )ains to the com&ined
$irms $rom mer)ers has not &een identi$ied, "n the case o$ hori4ontal mer)ers, at least,
those )ains could .ell come $rom enhanced mar*et #o.er, Another challen)e arises
&ecause acCuirin) $irms do not a##ear to &ene$it $rom mer)ers, .hich at the least is an
uncom$orta&le $act $or those .ho &elie%e in a reasona&ly e$$icient stoc* mar*et, "n $act,
there is some e%idence that many mer)ers in%ol%e mana)erial hu&ris or em#ire &uildin),
Bar)eron et al, 9200F: $ind that #u&lic $irms #ay a @@S hi)her #remium to tar)ets than
do #ri%ate acCuirers, -ar$ord and ?i 9200F: $ind that in mer)ers that lea%e acCuirin)
$irm shareholders .orse o$$, &iddersG 2E!s are &etter o$$ F@S o$ the time, 3his issue
.ill &e im#ortant &elo. .hen .e consider mer)er syner)ies= i$ there truly are uniCue
syner)ies resultin) $rom the mer)er, .hy do acCuirin) $irms $ail to ca#ture any o$ these
)ains $rom tradeW
2%4%#% $ccountin* measures of firm performance
A second method $or measurin) the e$$ects o$ mer)ers is to study accountin) data $or
the $irms in%ol%ed to loo* $or chan)es in %arious measures, such as rates o$ return,
relia&ly than are e$$ects on the stoc* mar*et %alue o$ the mer)in) #arties, es#ecially i$ the ri%als are di%ersi$ied
com#anies .ith a relati%ely small share o$ their re%enues comin) $rom the sale o$ #roducts in mar*ets .here
the mer)in) $irms are si)ni$icant hori4ontal ri%als,
Ch% .1- $ntitrust 11@@
cash $lo.s, or #ro$it mar)ins, 7a%enscra$t and cherer 91'8F, 1'8':, usin) .idely cited
F32 ?ine o$ Business 6ata, reach rather ne)ati%e conclusions= many o$ the mer)ers
and acCuisitions they study .ere unsuccess$ul, leadin) to a decline in the #ost<mer)er
#ro$ita&ility o$ the acCuired line o$ &usiness, 3heir study su##orts the %ie. o$ e/cessi%e
mana)erial 4eal a&out acCuisitions, -o.e%er, they mostly e/amine con)lomerate
mer)ers, not hori4ontal mer)ers, so much o$ their e%idence is not directly rele%ant to
hori4ontal mer)er control #olicy, Also, they $ind that hori4ontal mer)ers tended to &e
more #ro$ita&le than con)lomerate mer)ers 9althou)h, a)ain, this result does not distin)uish
mar*et #o.er $rom the #ossi&ility o$ )reater syner)ies in hori4ontal mer)ers:,
-ealy, ;ale#u, and 7u&ac* 91''2: e/amine #ost<mer)er o#eratin) #er$ormance $or the
$i$ty lar)est mer)ers that too* #lace $rom 1'F' to 1'81, 3hey $ind that the mer)ed $irms
e/hi&ited im#ro%ed o#eratin) #er$ormance, as measured &y o#eratin) cash $lo.s, relati%e
to their industry #eers, 3hey attri&ute these )ains to increased o#eratin) e$$iciency,
Alon) similar lines, ?ichten&er) and ie)el 91'8F: and McAuc*in and B)uyen 91''@:
$ind #lant le%el #roducti%ity )ains associated .ith mer)ers in manu$acturin) industries,
usin) the 2ensus BureauGs ?on)itudinal Esta&lishment 6ata $or 1'F2O1'81, 3his .as
not a #eriod, ho.e%er, .hen hi)hly concentratin) hori4ontal mer)ers .ere #ermitted &y
antitrust en$orcers,
2%4%4% Case studies
A third a##roach is to study s#eci$ic mer)ers, trac*in) the $irms or industries in%ol%ed,
loo*in) at such measures as #rices, out#ut, #roduct Cuality, or 7&6 intensity, "n #rinci#le,
one can also try to measure the im#act o$ a mer)er on ri%als or customers, Da#lan
92000: #ro%ides a use$ul collection o$ case studies o$ mer)ers in a di%erse set o$ industries,
includin) hos#itals, tires, &an*s, #harmaceutical dru)s, airlines, and oil $ield
ser%ices, 3he cases studied .ere not selected s#eci$ically to shed li)ht on maEor hori4ontal
mer)ers, 3hese studies illustrate the )reat %ariety o$ $act #atterns that arise in mer)er
analysis, the im#ortant role o$ mer)ers as a means &y .hich industry #artici#ants adEust
to chan)in) mar*et conditions 9ma*in) it es#ecially hard to distin)uish the e$$ects o$
mer)ers $rom other chan)es ta*in) #lace in the industry, es#ecially once one reco)ni4es
that $irms sel$<select to #artici#ate in mer)ers:, and the ris*s as .ell as o##ortunities
associated .ith mer)ers,
For antitrust #ur#oses, it is most use$ul to study hori4ontal mer)ers that raised serious
antitrust concerns .hen #ro#osed &ut ultimately .ent $or.ard, 3his a##roach has
the %irtue o$ $ocusin) attention on the %ery small $raction o$ all mer)ers that are most
rele%ant $or assessin) mer)er control #olicy,
Airline mer)ers ha%e recei%ed a )reat deal o$ attention, in no small #art &ecause )ood
data on $ares are a%aila&le and one can use $ares on other routes as a )ood &enchmar*
.hen measurin) the e$$ects o$ mer)ers on $ares, Borenstein 91''0:, 5erden, Jos*o.,
and Johnson 91''1:, and ;eters 92003: study t.o airline mer)ers $rom the mid<1'80s
that .ere a##ro%ed &y the 6e#artment o$ 3rans#ortation o%er the o&Eections o$ the
6!J= the mer)er o$ Borth.est Airlines .ith 7e#u&lic Airlines, and the mer)er o$ 3rans
11@0 L% 7aplow and C% Shapiro
5orld Airlines 935A: .ith !4ar* Airlines, 3hese mer)ers raised si)ni$icant antitrust
issues &ecause they com&ined directly com#etin) hu&s= Borth.est and 7e#u&lic &oth
had hu&s at Minnea#olis, and 35A and !4ar* &oth had hu&s at t, ?ouis, Borenstein
91''0: $ound si)ni$icant $are increases $ollo.in) the Borth.est+7e#u&lic mer)er &ut
not $ollo.in) the 35A+!4ar* mer)er, 5erden, Jos*o., and Johnson 91''1: $ound that
the Borth.est+7e#u&lic mer)er raised $ares &y a&out @S and the 35A+!4ar* mer)er
raised $ares &y a&out 1,@S, and that &oth mer)ers led to si)ni$icant ser%ice reductions,
Dim and in)al 91''3: e/amine $ourteen airline mer)ers $rom the mid<1'80s, 3hey
com#are #rice chan)es on the routes ser%ed &y the mer)in) $irms .ith #rice chan)es
on other routes o$ the same distance and conclude that any e$$iciency )ains in mer)ers
&et.een ri%al airlines .ere more than o$$set &y enhanced mar*et #o.er, leadin) to
$ares that a%era)ed 10S hi)her a$ter si/ to nine months, Fare increases .ere es#ecially
lar)e $or mer)ers in%ol%in) airlines in &an*ru#tcy, .hich had unusually lo. 9#erha#s
unsustaina&ly lo.: #re<mer)er $ares,
3he &an*in) industry is another industry in .hich )ood #rice data are a%aila&le and
many hori4ontal mer)ers ha%e occurred, ma*in) it #ossi&le to measure the #rice e$$ects
o$ hori4ontal mer)ers, ;ra)er and -annan 91''8: study the e$$ects o$ maEor hori4ontal
mer)ers in the U,, &an*in) industry durin) the early 1''0s, 3hey loo* at chan)es in
interest rates #aid on de#osits $or se%eral ty#es o$ de#osit accounts, usin) monthly data,
3hey de$ine Msu&stantial hori4ontal mer)ersN as those that increase the --" &y more
than 200 #oints to a #ost<mer)er %alue )reater than 1800, 3hey $ind that su&stantial
hori4ontal mer)ers reduce the de#osit interest rates o$$ered &y the mer)in) &an*s,
3he #rice and Cuality e$$ects o$ hos#ital industry mer)ers ha%e &een e/amined in a
num&er o$ studies, as descri&ed in ;autler 92003:, For e/am#le, Qita and acher 92001:
$ind lar)e #rice increases, not re$lectin) increases in ser%ice Cuality, $ollo.in) a hos#ital
mer)er in anta 2ru4, 2ali$ornia,
7ecent #a#ers loo* at other industries as .ell, ;esendor$er 92003: studies the e$$ect
o$ hori4ontal mer)ers in the #a#er industry on ca#acity choices, -astin)s 92001: loo*s
at #ricin) in the retail )asoline mar*et in outhern 2ali$ornia,
!ne natural .ay to )ain in$ormation to in$orm hori4ontal mer)er #olicy .ould &e $or
the antitrust en$orcement a)encies to #er$orm retros#ecti%e studies on the deals that they
ha%e in%esti)ated closely &ut ultimately allo.ed to #roceed .ithout si)ni$icant di%estitures,
Beither the F32 nor the 6!J has o$$icially re#orted results $rom any such study,
at least in recent years,110 Barton and herman 91'81: do re#ort #rice increases $rom a
hi)hly concentratin) mer)er that .as challen)ed &y the F32 se%eral years a$ter it .as
consummated,111 "n addition, the U,D, !$$ice o$ Fair 3radin), in conEunction .ith the
6e#artment o$ 3rade and "ndustry and the U,D, 2om#etition 2ommission, s#onsored
a study o$ ten mer)ers that too* #lace durin) 1''0O2002, 3hese .ere mer)ers that the
!$$ice o$ Fair 3radin) had re%ie.ed and $ound to raise su$$icient com#etition issues
110 F32 91''': re#orts on a study desi)ned to determine the e$$icacy o$ the di%estitures it had ne)otiated,
111 3he acCuirin) com#any .as Zide/, and the #roducts in%ol%ed .ere ty#es o$ du#licatin) micro$ilm,
Ch% .1- $ntitrust 11@F
that they .ere .orthy o$ re$erral to the 2om#etition 2ommission &ut that the 2om#etition
2ommission had su&seCuently cleared, ee !$$ice o$ Fair 3radin) 9200@:, Based
on inter%ie.s .ith customers o$ the mer)in) $irms, this study did not $ind a si)ni$icant
lessenin) o$ com#etition in ei)ht o$ the ten cases studied, "n the other t.o cases, a shortterm
loss o$ com#etition .as $ound to ha%e &een corrected &y su&seCuent entry into the
mar*et,
2%2% $ntitrust law on hori9ontal mer*ers
3his su&section outlines current U,, antitrust la. on hori4ontal mer)ers, 5e start .ith
a &rie$ statutory &ac*)round and an e/#lanation o$ #rocedures, em#hasi4in) #re<mer)er
noti$ication and analysis &y en$orcement a)encies, 5e then discuss the su&stanti%e la.
re)ardin) reCuisite anticom#etiti%e e$$ects and .hether mer)er syner)ies may &e o$$ered
to de$end other.ise anticom#etiti%e mer)ers,
3hrou)hout the discussion, it is use$ul to *ee# in mind the relationshi# &et.een
mer)er la., on one hand, and the la. concernin) #rice<$i/in) and mono#oli4ation, on
the other hand, Because collusion is di$$icult to detect and #rosecute 9and, de#endin)
on the means o$ collusion, is o$ uncertain ille)ality:, as discussed in section 3, it ma*es
sense to some de)ree $or mer)er #olicy to ado#t a #ro#hylactic a##roach to.ard mer)ers
that threaten )reater coo#eration amon) $irms, ?i*e.ise, &ecause the la. on mono#oli4ation
does not re)ulate #rice<settin) once a mer)er has &een %alidated and im#oses
only modest constraints on e/clusionary #ractices, as .ill &e discussed in section @,
there is also reason to &e .ary o$ a##ro%in) a mer)er that threatens unilateral e$$ects or
e/clusionary conduct,
2%2%.% Bac*round and procedure
As else.here, our discussion .ill $ocus on antitrust la. and #rocedures in the United
tatesJ there is a )ro.in) &ut incom#lete con%er)ence in ho. hori4ontal mer)ers are
treated across Eurisdictions,112 7ele%ant U,, la. has three #rimary, o%erla##in) #ro%isions=
herman Act ection 1Gs #rohi&ition on any Mcontract, com&ination, , , , or cons#iracy
in restraint o$ tradeN 9the $ocus in section 3 on collusion:J 2layton Act ection
FGs #rohi&ition on acCuisitions o$ stoc* or assets .hose e$$ect Mmay &e su&stantially to
lessen com#etition, or to tend to create a mono#olyNJ and the Federal 3rade 2ommission
112 "n 2001, the Euro#ean Union #romul)ated ne. hori4ontal mer)er )uidelines that in many res#ects are
similar to the #ree/istin) -ori4ontal Mer)er Auidelines in the United tates, Euro#ean Union 92001&:, ;rior
to that, althou)h $e. mer)ers had &een &loc*ed, the en$orcement stance o$ the 2ommission is )enerally
re)arded to ha%e &een stricter than that in the United tates, 3he Euro#ean 2ourt o$ First "nstanceGs re%ersal
in 2002 o$ three 2ommission attem#ts to &loc* mer)ers is seen as the catalyst $or the recent re$orm, !ther
nota&le administrati%e chan)es include the a##ointment o$ a chie$ com#etition economist, ee, $or e/am#le,
6a&&ah 92001:, "t is too early to tell Eust ho. much #ractice under the ne. re)ime .ill di$$er in $act $rom that
in the United tates,
11@8 L% 7aplow and C% Shapiro
Act ection @Gs #rohi&ition on any Mun$air method o$ com#etition,N113 "n s#ite o$ di%erse
histories and statutory lan)ua)e, a lar)ely uni$ied a##roach to en$orcement o$ these #ro%isions
has emer)ed, Bota&ly, the 6!J and F32 91''2: in their most recent -ori4ontal
Mer)er Auidelines ha%e #romul)ated a sin)le #olicy statement a##lica&le re)ardless o$
the statute in%ol%ed, 2ommentators and courts ha%e lar)ely ta*en a similar a##roach,111
Most challen)es to mer)ers are &rou)ht &y one o$ the t.o $ederal a)encies, althou)h
mer)ers may also &e challen)ed &y states and #ri%ate #arties,11@ ince 1'F0, the $ederal
#rocedure has ta*en its current $orm, .hich is similar to #rocedures in many other Eurisdictions,
110 Firms intendin) to mer)e are reCuired to $ile s#eci$ied in$ormation .ith
the #ertinent a)encies, Each deal is cleared to either the F32 or the 6!J, "n mer)ers $or
.hich there is any serious #ros#ect o$ a challen)e, the #arties usually su&mit su&stantial
su##lemental material, 3hey hire a team o$ la.yers and economic e/#erts 9o$ten
associated .ith consultin) $irms: that ty#ically ha%e su&stantial e/#erience in mer)er
$ilin)sJ indeed, they may ha%e handled numerous #rior mer)ers in the same or related
industries, 3his team )athers and analy4es in$ormation and #roduces an o$ten<ela&orate
study document de$endin) the mer)er .ith re)ard to com#etiti%e e$$ects and antici#ated
e$$iciencies, 3he )oal ty#ically is to #ersuade the a)encies to a##ro%e the mer)er, and
to do so #rom#tly,
An im#ortant as#ect o$ the #rocedure concerns the e$$ects o$ a)ency delayK.hich
arises .hen the a)ency $eels that it needs additional in$ormation or must underta*e more
su&stantial inde#endent in%esti)ation and analysisKor o$ an ultimate a)ency challen)e,
E%en i$ the #arties antici#ate e%entual a##ro%al, .hether $rom the a)ency or a$ter liti)ation
in court, the #ros#ect o$ delay .ill *ill many deals and im#ose su&stantial costs
on others, Dee#in) $inancin) in line, ma*in) interim in%estment decisions in #lant and
eCui#ment, decidin) on strate)ic matters such as launchin) ne. #roducts or terminatin)
old ones, maintainin) customer loyalty in the #resence o$ uncertainty a&out #roduct
113 !&ser%e that none o$ the statutes is limited to mer)ers #er seJ other $orms o$ com&ination, nota&ly includin)
acCuisitions o$ some or all o$ another $irmGs assets, are included, 93hus, $or e/am#le, i$ the only t.o
#roducts in a mar*et are #atented, the acCuisition o$ one o$ the #atents &y the o.ner o$ the other .ould &e
analy4ed similarly to a hori4ontal mer)er,: EU re)ulations ha%e a similar reach,
111 ee, $or e/am#le, Areeda, -o%en*am#, and olo. 92000, %ol, 1, ##, 13O11:,
11@ !ne mi)ht thin* that com#etitors .ould $reCuently challen)e mer)ers, -o.e%er, under the doctrine o$
Mantitrust inEury,N this is not ordinarily #ossi&le= com#etitors tend to &e inEured &y #ro<com#etiti%e mer)ers that
lead to lo.er #rices 9deemed not the sort o$ inEury that the antitrust la.s .ere enacted to #re%ent: &ut hel#ed
&y anticom#etiti%e mer)ers 9recall $rom su&section 1,1, $or e/am#le, that unilateral e$$ects tend to &ene$it
non<mer)in) #arties:, 7elatedly, .hen a)encies are in%esti)atin) #ro#osed mer)ers, they are less li*ely to
)i%e .ei)ht to the %ie.s o$ com#etitors on o%erall e$$ects 9$or $ear o$ mani#ulation: and more commonly see*
the reactions o$ lar)e #urchasers 9$or e/am#le, health insurers, in the case o$ hos#ital mer)ers:, "n this re)ard,
2oate and Ulric* 9200@: $ind that the #ro&a&ility that the F32 ta*es action a)ainst mer)ers is hi)her, ceteris
pari!us, .hen there are customer com#laints a&out the mer)er, "n recent commentary, the a)encies a$$irm
that M2onsumers ty#ically are the &est source, and in some cases they may &e the only source, o$ critical
in$ormation , , , ,N F32 and 6!J 92000, ##, 'O10:,
110 "ndeed, there has &een some e/#licit international coo#eration, moti%ated &y the $act that many su&stantial
mer)ers are su&Eect to the com#etition re)ulation o$ multi#le national and international Eurisdictions,
Ch% .1- $ntitrust 11@'
su##ort, retainin) talented em#loyees .ho may $ear Eo& loss, and so $orth may #resent
si)ni$icant challen)es to mer)in) $irms, es#ecially those &ein) acCuired, that do not
*no. i$ or .hen their deal may &e a##ro%ed, Accordin)ly, )reat ener)y is de%oted to
o&tainin) a Cuic* and success$ul conclusion to the antitrust a)encyGs deli&erations, "$
the a)ency challen)es the deal and no #rom#t settlement is reached, the mer)in) #arties
must either a&andon their transaction or con$ront considera&le $urther delay in the
#rocess o$ liti)atin) the matter in $ederal court,
3o )i%e some sense o$ the le%el o$ mer)er re%ie. acti%ity, durin) the years 2001O200@
$or the 6!J there .ere a&out 1000 to 2100 #re<mer)er noti$ications annually, o$ .hich
F0 to 100 resulted in decisions to in%esti)ate $urther, and 2 to F led to cases &ein) $iled,
de#endin) on the year,11F As already su))ested, ho.e%er, these latter statistics can &e
misleadin) &ecause some mer)ers .ill &e dro##ed alon) the .ay either &ecause the
#arties are insu$$iciently con$ident o$ success or sim#ly &ecause they cannot tolerate the
antici#ated delays, Also, no dou&t, some #otential mer)ers are deterredJ the more #redicta&le
are the a)encies, due in #art to the Mer)er Auidelines and years o$ e/#erience
under them, the more one .ould e/#ect there to &e $e. #ro#osed mer)ers .ith a hi)h
li*elihood o$ &ein) challen)ed,
Another im#ortant outcome is settlement, most $reCuently throu)h the #arties s#innin)
o$$ #lants, other o#erations, or lines o$ &usiness in areas o$ si)ni$icant com#etiti%e
o%erla#,118 3hat is, some mer)ers may &e $ound to #ose a serious com#etiti%e threat
&ut only in certain )eo)ra#hic mar*ets or only .ith res#ect to some o$ the many #roducts
the $irms #roduce, "n such cases, a##ro#riate di%estiture o$ #ertinent assets .ill
ordinarily satis$y the en$orcement a)encies,
For challen)es that do #roceed to court, the a)encies o$ten attem#t to o&tain a #reliminary
inEunction reCuirin) the mer)in) $irms to continue to o#erate inde#endently
#endin) a $inal outcome, and in the end success$ul challen)es #roduce #ermanent inEunctions
a)ainst the mer)er 9or su&seCuent ne)otiations leadin) to asset di%estitures:,
3his a##roach contrasts dramatically .ith the course o$ #roceedin)s in earlier years,
&e$ore the #re<mer)er noti$ication re)ime .as in #lace, 3hen, mer)ers .ere #rom#tly
consummated, and $inal decrees a)ainst challen)ed mer)ers ordinarily too* many years,
sometimes more than a decade, at .hich #oint the t.o $irms .ere o$ten su$$iciently inte)rated
9#lants closed, &rands discontinued, ne. Eoint o#erations .ell under.ay: to
ma*e #ractical di%estiture di$$icult or im#ossi&le,
11F 6!J 92000:, 3he F32Gs mer)er en$orcement acti%ity is com#ara&le to that o$ the 6!J, "n the Euro#ean
Union, $rom 1''0 until May 2002 9and thus &e$ore the #romul)ation o$ the ne. 2001 re)ulations and )uidelines:,
80S o$ noti$ied mer)ers .ere a##ro%ed unconditionally, @S .ere a##ro%ed su&Eect to underta*in)s
9such as s#in<o$$s: &y the end o$ the one<month initial in%esti)ati%e #eriod, an additional 3S a$ter $urther
in%esti)ation, and 1S 918: .ere #rohi&ited, 9Another 1S .ere .ithdra.n durin) in<de#th in%esti)ationsJ %arious
others .ere $ound to &e outside the 2ommissionGs Eurisdiction,: ee5alle de Ahelc*e and Aer%en 92001,
X@,01:,
118 6urin) the #ast decade, the com&ined num&er o$ transactions that .ere restructured or a&andoned a$ter
a $ormal challen)e .as announced &ut &e$ore a case .as $iled in court usually e/ceeded the num&er o$ cases
$iled, Furthermore, these statistics include only terminations that $ollo.ed the issuance o$ a $ormal challen)e,
1100 L% 7aplow and C% Shapiro
3he remainder o$ this section $ocuses on the su&stance o$ the le)al restriction on
hori4ontal mer)ers, $irst e/aminin) the core inCuiry into anticom#etiti%e e$$ects and
then considerin) the role o$ e$$iciencies in Eusti$yin) mer)ers that .ould other.ise &e
#roscri&ed, As .ill &e seen, the a##roach to.ard &oth issues has e%ol%ed a )reat deal
o%er time, Furthermore, throu)hout this e%olution these t.o issues ha%e not &een entirely
inde#endent, "n #articular, the central Cuestion o$ the li*elihood and e/tent o$
anticom#etiti%e e$$ects reCuired to condemn hori4ontal mer)ersKthe threshold $or a
success$ul challen)eKseems to &e ans.ered in a manner that su&stantially re$lects underlyin)
%ie.s a&out the ty#ical #ro&a&ility and ma)nitude o$ mer)er syner)ies,
2%2%#% $nticompetitive effects
"n the 1'00s 9in the .a*e o$ the stren)thenin) o$ 2layton Act ection F in 1'@0:,
the U,, u#reme 2ourt, $ollo.in) the lead o$ the $ederal en$orcement authorities,
ado#ted a restricti%e %ie. to.ard hori4ontal mer)ers, 3he 2ourt condemned a num&er
o$ mer)ers .here the #artiesG com&ined mar*et shares .ere under 10S, $or e/am#le,
in Brown Shoe and "on<s,11' 3he $irst )o%ernment mer)er )uidelines, #romul)ated in
1'08, ado#ted similarly strin)ent thresholds $or challen)in) mer)ers, ?i*e.ise, they
endorsed the structural #resum#tion that concentration im#lies anticom#etiti%e e$$ects,
as articulated &y the u#reme 2ourt in 'hiladelphia Ban= Ma mer)er .hich #roduces a
$irm controllin) an undue #ercenta)e share o$ the rele%ant mar*et, and results in a si)ni$icant
increase in the concentration o$ $irms in that mar*et is so inherently li*ely to lessen
com#etition su&stantially that it must &e enEoined in the a&sence o$ e%idence clearly
sho.in) that the mer)er is not li*ely to ha%e such anticom#etiti%e e$$ects,N120 A shi$t
.as si)naled &y the 1'F1 decision in ?eneral Dynamics,121 3he s#eci$ic holdin)Kthat
the prima facie case esta&lished &y mar*et share statistics could &e re&utted &y sho.in)
that the $i)ures )a%e a misleadin) de#iction o$ com#etiti%e e$$ectsK.as not itsel$ truly
no%el 9Brown Shoe had su))ested as much:, -o.e%er, the acts o$ su&Eectin) the )o%ernmentGs
case to hei)htened scrutiny and ultimately reEectin) it .ere ta*en as a si)nal
o$ a ne. direction,
11' Bro.n hoe 2o, %, United tates, 3F0 U,, 2'1 91'02:J United tates %, QonGs Arocery 2o,, 381 U,,
2F0 91'00:, "ndeed, in "on<s, a$ter #ointin) out that the com&ined share .as F,@S o$ the ?os An)eles mar*et
$or )rocery stores, that the num&er o$ inde#endent stores had $allen $rom @30@ to 3@'0, that hal$ o$ the to#
20 chains had acCuired stores $rom smaller $irms, and similar $acts, the 2ourt #roclaimed= M3hese $acts alone
are enou)h to cause us to conclude , , , that the QonGsOho##in) Ba) mer)er did %iolate XF,N 381 U,, at 2F1,
3he dissent critici4ed the maEority $or attem#tin) to Mroll &ac* the su#ermar*et re%olutionN and asserted that
MUtVhe sole consistency that " can $ind is that under XF, the Ao%ernment al.ays .ins,N 381 U,, at 288, 301,
120 United tates %, ;hiladel#hia Bational Ban*, 3F1 U,, 321, 303 91'03:, 3he 2ourt cited #rominent economic
and le)al authorities in su##ort o$ this %ie., althou)h the mar*et share le%els in 'hiladelphia Ban
and #ro#osed &y most o$ the commentators .ere su&stantially hi)her 920S or more: than the le%els deemed
su$$icient in many o$ the other cases o$ the #eriod and in the 1'08 )uidelines,
121 United tates %, Aeneral 6ynamics 2or#,, 11@ U,, 180 91'F1:,
Ch% .1- $ntitrust 1101
ince the mid<1'F0s, there ha%e not &een $urther mer)er o#inions &y the u#reme
2ourt,122 Be%ertheless, a con$luence o$ three $actors has made clear that the la. has
mo%ed su&stantially= chan)in) %ie.s to.ard com#etition and the e$$ecti%eness o$ mar*et
$orces 9&oth &roadly and in the academy:, a chan)e in the com#osition o$ the
u#reme 2ourt and in the nature o$ its o#inions on other antitrust su&Eects, and a ne.
direction $rom the )o%ernment as em&odied in the 1'82 mer)er )uidelines, 3he current
91''2: Mer)er Auidelines are a successor to the 1'82 %ersion, .hich di$$ered in many
*ey res#ects $rom those issued in 1'08, 6etails o$ the current methodolo)y $or e%aluatin)
hori4ontal mer)ers .ill &e e/amined throu)hout the remainder o$ this section,123
;erha#s the most nota&le chan)e, ho.e%er, .as in the thresholds $or challen)e= they
.ere nota&ly hi)her, su$$iciently so that a num&er o$ the $amous cases o$ the 1'00s
9that the )o%ernment .on: .ould not ha%e &een &rou)ht had the ne. )uidelines &een in
e$$ect,
3he #resum#ti%e thresholds in the Mer)er Auidelines 9once the mar*et is de$inedJ
see su&section 1,@: are as $ollo.s, "$ the #ost<mer)er --" is &elo. 1000, the mar*et
is re)arded as unconcentrated and ordinarily no $urther analysis .ill &e underta*en, "$
the #ost<mer)er --" is &et.een 1000 and 1800, si)ni$icant concerns are raised i$ and
only i$ the mer)er increases the --" &y more than 100, in .hich case $urther analysis
is underta*en,121 And i$ the #ost<mer)er --" e/ceeds 1800, si)ni$icant concerns
are deemed to e/ist .hen the mer)er raises the --" &y more than @0,12@ 3his air o$
certainty is misleadin), "n actual a##lication o$ the Mer)er Auidelines, it has &ecome
a##arent that, in certain industries, the de facto thresholds are much hi)her, For e/am#le,
most o$ hundreds o$ hos#ital mer)ers su&seCuent to the 1'82 )uidelines 9.hich
had the same thresholds as those Eust descri&ed: ha%e )one unchallen)ed e%en thou)h
#ost<mer)er --"s and --" increases .ere )reatly a&o%e 1800 and @0, res#ecti%ely,
#resuma&ly re$lectin) a %ie. o$ ty#ical e$$icient scale in this industry,120 E%en so, the
la.yers and economists .ho s#eciali4e in mer)er #ractice are )enerally a.are o$ such
122 3his a##arent anomaly is lar)ely e/#lained &y t.o #rocedural chan)es= the ne. reCuirement o$ #remer)er
noti$ication reduced the $lo. o$ Cuestiona&le mer)ers into the courts, and the elimination o$ s#ecial
rules #ermittin) automatic a##eal in some cases directly to the u#reme 2ourt )reatly reduced the #ro#ortion
o$ antitrust cases reachin) that court,
123 For $urther ela&oration on ho. the )uidelines are im#lemented in #ractice, see the en$orcement a)enciesG
commentary, F32 and 6!J 92000:,
121 Further analysis in%ol%es consideration o$ entry, e$$iciencies, and the #ossi&ility that one o$ the $irms may
&e $ailin),
12@ 3he ne. 2001 EU )uidelines 9Euro#ean Union, 2001a: are stri*in)ly similar= 2000 re#laces 1800, and the
increases must &e 2@0 and 1@0 rather than 100 and @0, res#ecti%ely, As noted in the te/t and notes to $ollo.,
ho.e%er, actual #ractice under the U,, Mer)er Auidelines indicates the use o$ hi)her de facto thresholds
9to an e/tent that %aries &y industry:, ?i*e.ise, the additional $actors considered in the ne. EU )uidelines
are lar)ely the same as in the United tatesKnota&ly, they include e$$iciencies as a de$enseKand they e%en
a##ear in the same order,
120 ee also F32 and 6!J 92003:, .hich )i%es statistics on the #ost<mer)er --" and chan)e in --" $or
mer)er challen)es in a num&er o$ industries, 3he ta&les stron)ly su))est that the thresholds $or challen)es
%ary )reatly &y industry, 2oate and Ulric*Gs 9200@: analysis o$ mer)er en$orcement at the F32 $inds that, $or
1102 L% 7aplow and C% Shapiro
#atterns, .hich themsel%es #ro&a&ly re$lect success$ul #ersuasion in #rior mer)er $ilin)s
in #articular industries,12F
E%en thou)h the Mer)er Auidelines are $ormally Eust a #u&lic statement &y the $ederal
a)encies o$ ho. they intend to #roceed internally, they ha%e to a su&stantial de)ree
dictated #artiesG #ractices in liti)ation and courtsG analyses o$ mer)ers re)ardin) the
methodolo)y $or assessin) mar*et de$inition, anticom#etiti%e e$$ects, and other $actors
as .ell as the thresholds $or condemnation, 3he courts undou&tedly .elcome the )uidance,
Moreo%er, as noted, e%en thou)h the most on<#oint u#reme 2ourt o#inions 9$rom
the 1'00s: are much stricter, there has &een a su$$icient shi$t in understandin)s and in
&eha%ior o$ the u#reme 2ourt on other issues that lo.er courts ha%e $ollo.ed the ne.
a##roach rather than adherin) to older #recedents, 3he most direct indicator o$ this $act
is that, in this more recent era, the )o%ernment loses a )ood #ro#ortion o$ the cases it
&rin)s e%en thou)h the )o%ernmentGs threshold $or challen)in) mer)ers is much hi)her
than in the #ast 9.hen they .on nearly e%ery case:,
2%2%4% Efficiencies
Mer)er syner)ies #lay an im#ortant role in com#etition #olicy to.ard hori4ontal mer)ers,
and one that has chan)ed su&stantially o%er time, "n this su&section, .e $irst descri&e
that role and its modern de%elo#ment and then consider ho. it relates to the )oals o$
com#etition #olicy more )enerally,
Althou)h e$$iciencies are usually discussed as a #ossi&le de$ense, ad%anced &y the
mer)in) #arties to Eusti$y a mer)er that mi)ht other.ise &e condemned as anticom#etiti%e,
e$$iciencies ha%e lon) had another, more si)ni$icant in$luence on mer)er #olicy=
in settin) the threshold $or antitrust scrutiny, 2onsider tri%ial com&inations, say, .hen
t.o indi%iduals $orm a #artnershi# or t.o small stores Eoin $orces, 3hese com&inations
as .ell as some su&stantially lar)er mer)ers ha%e ne%er &een su&Eect to challen)e, But
they could ha%e &een, First, as noted in the discussion in su&section 3,@,2, such com&inations
literally in%ol%e #rice<$i/in), )oin) $or.ard, .hich is automatically condemned,
su##osedly .ithout #roo$ o$ mar*et #o.er, Iet #roducti%e com&inations, under the rule
o$ reason, are #ermissi&le, #resuma&ly &ecause they o$ten create syner)ies 9des#ite the
$act that, once $ormed into a sin)le entity, #rices .ill &e determined Eointly rather than
inde#endently:, econd, e%en .hen the num&er o$ $irms is lar)e, mer)ers in 2ournot or
)i%en --"s and other $actors, en$orcement actions are more li*ely in the oil, )rocery, and chemical industries,
"n addition, they $ind that, holdin) &oth the --" and chan)e in --" constant, the #ro&a&ility o$ en$orcement
rises as the num&er o$ leadin) ri%als $alls $rom @ to 1 and $rom 1 to 3,
12F ?eary 92002: em#hasi4es that, since the Mer)er Auidelines ha%e no. &een $ollo.ed $or years, current
mer)er #olicy e/hi&its a )ood deal o$ sta&ility, Furthermore, in order Mto #ro%ide )reater trans#arency and
$oster dee#er understandin) re)ardin) antitrust la. en$orcement,N the 6!J and F32 issued a document,
M2ommentary on the -ori4ontal Mer)er Auidelines,N in 2000, F32 and 6!J 92000:, 3he a)encies also held
a mer)er en$orcement .or*sho# in 2001, F32 and 6!J 92001:,
Ch% .1- $ntitrust 1103
Bertrand oli)o#oly tend to raise #rices 9sli)htly:, Be%ertheless, nontri%ial anticom#etiti%e
e$$ects must &e demonstrated &e$ore a mer)er .ill &e challen)ed,128
Qie.ed more &roadly, settin) the threshold o$ anticom#etiti%e e$$ects si)ni$icantly
a&o%e 4ero may &e rationali4ed &y the %ie. that mer)ers ty#ically )enerate some syner)ies,
so they should not &e #rohi&ited unless the reduction in com#etition is su$$iciently
)reat, "n the 1'00s and 1'F0s .hen U,, anti<mer)er #olicy .as strict, many $airly
lar)e mer)ers .ere ne%ertheless routinely #ermitted, 2urrently, as descri&ed in su&section
1,1,2, the thresholds are much hi)herJ moreo%er, they are raised $urther in industries
.here syner)ies are thou)ht to &e unusually lar)e relati%e to the si4e o$ the mar*et 9hos#itals,
$or e/am#le:,
3here$ore, it seems a##ro#riate to understand an e$$iciencies de$ense to a mer)er
.hose sus#ected anticom#etiti%e e$$ects e/ceed the threshold as im#licitly in%ol%in)
a claim that the mer)er syner)ies are not merely su&stantial &ut are lar)e enou)h to
nota&ly e/ceed the le%el ordinarily #resumed to e/ist, A$ter all, they must &e enou)h
to Eusti$y the mer)er in li)ht o$ .hat .ould other.ise &e su&stantial, not merely tri%ial,
anticom#etiti%e e$$ects, 3his $ramin) o$ the Cuestion may hel# e/#lain .hy courts
and en$orcement a)encies are cautious in acce#tin) e$$iciency de$enses 9in addition to
the o&%ious reason that mer)in) #arties ha%e e%ery incenti%e to assert the e/istence o$
syner)ies .hen there are $e. and the mer)er is in $act anticom#etiti%e:,
"n the 1'00s, the U,, u#reme 2ourt e/hi&ited a some.hat schi4o#hrenic &ut ultimately
hostile attitude to.ard mer)er syner)ies, Most $amously, in Brown Shoe, the
2ourt %ie.ed the e$$iciencies resultin) $rom the %ertical as#ects o$ the mer)er as #ro&lematic
&ecause they .ould )i%e the com&ined entity an ad%anta)e a)ainst com#etitors,
!n one hand, the 2ourt stated= M"t is com#etition, not com#etitors, .hich the Act #rotects,N
12' -o.e%er, the #assa)e continues= MBut .e cannot $ail to reco)ni4e 2on)ressG
desire to #romote com#etition throu)h the #rotection o$ %ia&le, small, locally o.ned
&usinesses, 2on)ress a##reciated that occasional hi)her costs and #rices mi)ht result
$rom the maintenance o$ $ra)mented industries and mar*ets, "t resol%ed these com#etin)
considerations in $a%or o$ decentrali4ation, 5e must )i%e e$$ect to that decision,N
!%er time, the $irst statementKMcom#etition, not com#etitorsNKhas continued to &e
amon) the most<Cuoted #assa)es $rom all u#reme 2ourt antitrust o#inions, &y courts
and commentators ali*e, .hereas the latter clari$ication 9contradiction: is not usually
a##ended anymore, ?o.er courts in the last cou#le o$ decades 9recall, there ha%e &een
no recent u#reme 2ourt #ronouncements on mer)ers: ha%e %aried in their a##roaches,
some e/#ressin) uncertainty a&out an e$$iciencies de$ense &ut most acce#tin) it, at least
in #rinci#le,130
128 Althou)h the lan)ua)e o$ the 2layton Act demands a su&stantial e$$ect, that o$ the herman Act does not,
9-istorically, the 2layton Act, #assed in 1'11 to remedy #ercei%ed .ea*ness in the herman Act and amended
in 1'@0 to stren)then it $urther, has &een %ie.ed as stricter than the herman Act, As noted in su&section 1,1,1,
ho.e%er, the en$orcement a)encies and, increasin)ly, the courts a##ly a more unitary a##roach,:
12' Bro.n hoe 2o, %, United tates, 3F0 U,, 2'1, 311 91'02:,
130 ee, $or e/am#le, the cases cited in Areeda, -o%en*am#, and olo. 92000, %ol, 1A, ##, 31O32 n, 1F:,
1101 L% 7aplow and C% Shapiro
3he Mer)er Auidelines are no. uneCui%ocal, "n 1''F, amendments to the 1''2 %ersion
added a ne. section on e$$iciencies, 3hey state= M"ndeed, the #rimary &ene$it o$
mer)ers to the economy is their #otential to )enerate such e$$iciencies, E$$iciencies )enerated
throu)h mer)er can enhance the mer)ed $irmGs a&ility and incenti%e to com#ete,
.hich may result in lo.er #rices, im#ro%ed Cuality, enhanced ser%ice, or ne. #roducts,N
3.o main conditions are im#osed= only mer)er<s#eci$ic e$$iciencies are reco)ni4ed, and
those e$$iciencies must &e su$$iciently %eri$ia&le, 2o)ni4a&le e$$iciencies also must &e
su$$iciently lar)e to #re%ent the mer)er $rom &ein) anticom#etiti%e,131
5hich e$$iciencies are mer)er s#eci$icW Under the a)enciesG standard a##roach,
economies o$ scale $rom, say, &uildin) lar)er #lants, are li*ely to &e acce#ted, E$$iciencies
$rom some other $unctions, li*e com&inin) #ayroll o#erations, are less li*ely to
&e credited &ecause o$ the o#tion o$ contractin) out $or such ser%ices, achie%in) &ene$its
o$ scale short o$ mer)er, And i$ t.o hos#itals demonstrate a need to share modern ima)in)
eCui#ment due to hi)h $i/ed costs, they may &e #ermitted to $orm a Eoint %enture
limited to that #ur#ose &ut not allo.ed on this )round to )o $or.ard .ith an other.ise
anticom#etiti%e mer)er o$ all o$ their o#erations,
Farrell and ha#iro 92001: e/#lore .hich e$$iciencies should Cuali$y as &ein) mer)er
s#eci$ic, reco)ni4in) that in the #resence o$ economies o$ scale $irms can )ro. internally
to reduce their a%era)e costs, 3hey distin)uish &et.een e$$iciencies &ased solely
on scale economies and e$$iciencies that re$lect the com&ination o$ s#eci$ic, non<trada&le
assets o.ned &y the mer)in) #arties, .hich they consider true mer)er<s#eci$ic e$$iciencies,
or syner)ies, 3hey ar)ue that many claimed e$$iciencies are not in $act mer)er
s#eci$ic,
3here is a dee#er #ro&lem underlyin) many dis#utes a&out .hether e$$iciencies are
mer)er s#eci$ic, !$ten, the #arties .ill claim that the mer)er is necessary .hereas en$orcers
.ill Cuestion .hy the #ur#orted &ene$it cannot &e achie%ed throu)h some more
limited $orm o$ contractual arran)ement, Under .hat circumstances there e/ist &ene$its
$rom com&inin) acti%ities under the direction o$ a sin)le $irm that cannot &e achie%ed
throu)h contractin) is, o$ course, a core issue in the theory o$ the $irm, one .hose e/#loration
.as launched &y 2oase 91'3F:, e/tended &y 5illiamson 91'F@, 1'8@:, and
e/#lored in su&seCuent .or* &y Arossman and -art 91'80:, -olmstrom and 3irole
91'8':, -art and Moore 91''0:, and -art 91''@:, amon) many others, ince the $indin)s
o$ this literature are o$ten su&tle, de#endin) on $actors that may not &e readily and relia&ly
ascertained in the course o$ an antitrust dis#ute, there .ill &e an ine%ita&le area o$
uncertainty, 3he le%el o$ theoretical so#histication re$lected in modern analysis o$ contracts
and the $irm has yet to ma*e si)ni$icant inroads in mer)er analysis, "n #ractice, the
a)encies tend to loo* at the $orms o$ colla&oration short o$ mer)er that are actually used
in the industry to determine .hether certain claimed e$$iciencies are mer)er s#eci$ic,
"n the end, en$orcers do and should ha%e a healthy s*e#ticism a&out sel$<ser%in) e$$iciency
claims made &y com#etitors see*in) to mer)e, As .e noted in su&section 1,3,
131 As noted #re%iously, the 2001 EU )uidelines are Cuite similar, includin) a reco)nition o$ an e$$iciencies
de$ense and the #articulars o$ ho. it may &e esta&lished in a )i%en case,
Ch% .1- $ntitrust 110@
acCuirin) $irms o$ten seem to o%erestimate the &ene$its o$ an acCuisition, and the o%erall
record re)ardin) mer)er e$$iciencies is mi/ed at &est, "t also tends to &e %ery di$$icult
$or the en$orcement a)encies or the courts to assess .hether claimed e$$iciencies
are indeed li*ely to arise, Dolas*y and 6ic* 92003: e/amine the treatment o$ e$$iciencies
in mer)er cases, ar)uin) that the en$orcement a)encies and the courts ha%e made
su&stantial #ro)ress o%er the #ast t.enty years incor#oratin) e$$iciencies into mer)er
analysis,132
u##ose that syner)ies can &e Cuanti$ied, or that in a #articular case the en$orcement
a)ency or the court has done its &est in determinin) the matter, 3he im#ortant remainin)
Cuestion is ho. e$$iciencies and anticom#etiti%e e$$ects are to &e &alanced, hould
the decision de#end on total economic .el$areKthe sum o$ consumer and #roducer
sur#lusKthe standard normati%e economic a##roachW hould it instead turn solely on
consumer sur#lusK.hether #rices rise or $allKthat is, .hether the e$$iciencies are su$$iciently
#assed on to consumers to at least o$$set any #rice increase that .ould other.ise
ensue due to anticom#etiti%e e$$ectsW !r should some other standard &e used, .hich
seems to &e the a##roach in earlier cases li*e Brown ShoeW
3his set o$ Cuestions &rin)s to mind the discussion in su&section 3,@,2 on the meanin)
o$ the rule o$ reason,133 7ecall that the Chica*o Board of >rade case and su&seCuent decisions
de$ine reasona&leness in terms o$ .hat #romotes %ersus su##resses com#etition,
Iet there remains am&i)uity concernin) the meanin) o$ com#etition,
Under a #rocess<oriented %ie., .hich has su##ort in many modern cases, one mi)ht
&e concerned .ith #rotectin) ri%alry per se, .hich mi)ht im#ly a strict mer)er #olicy,
3hat %ie. seems consonant .ith the 1'00s mer)er decisions and lan)ua)e such as that
Cuoted a&o%e $rom Brown Shoe 9the latter #art, re$errin) to the merits o$ #rotectin) com#etitors
and #reser%in) decentrali4ation: or that in 'hiladelphia Ban re$errin) to the
#reser%ation o$ a Mtraditionally com#etiti%e economy,N131 Under this a##roach, not only
.ould one condemn a mer)er $rom 2 $irms to 1, 3 $irms to 2, and 1 $irms to 3, &ut also,
it mi)ht seem, $rom N $irms to N 1, no matter ho. lar)e .as N, But as discussed
#re%iously, e%en .hen this %ie. o$ mer)er #olicy .as dominant, smaller mer)ers .ere
#ermitted, Althou)h e$$iciencies .ere not reco)ni4ed as a de$ense and .ere sometimes
%ie.ed as an e%il in lar)er mer)ers, the threshold $or challen)e .as hi)h enou)h to
allo. countless mer)ers to )o unchallen)ed, Furthermore, modern cases such as Broadcast
&usic that inter#ret the rule o$ reason and the &ul* o$ lo.er court cases on mer)ers,
as .ell as the Mer)er Auidelines, acce#t more e/#licitly that e$$iciency counts #osi<
132 ee also ;ito$s*y 91''': and Muris 91''': $or di$$erent %ie.s on the role o$ e$$iciencies in mer)er analysis,
133 3he rule o$ reason, note, arose in inter#retation o$ herman Act ection 1, .hich is also one o$ the statutes
a##lica&le to hori4ontal mer)ers, 3hus, as a $ormal le)al matter, one mi)ht say that the rule o$ reason is the
herman Act standard $or hori4ontal mer)ers, -o.e%er, courts ha%e )enerally considered mer)ers se#arately,
and as noted, increasin)ly .ithout re)ard to .hich statute is in%o*ed in a #articular case, Be%ertheless, )i%en
the )ro.in) con%er)ence in a##roaches under all o$ the antitrust statutes, one should e/#ect some con)ruence
&et.een courtsG analyses o$ hori4ontal mer)ers and o$ .hat are denominated as rule o$ reason cases,
131 United tates %, ;hiladel#hia Bational Ban*, 3F1 U,, 321, 3F1 91'03:,
1100 L% 7aplow and C% Shapiro
ti%ely, and in #articular that &etter ser%in) customers, nota&ly, throu)h lo.er #rices, is
desira&le,13@
3hus, at #resent the main contest seems to &e &et.een consumer .el$are and total
.el$are, that is, .hether e$$iciencies should &e credited .hen they increase #roducer
sur#lus rather than &ein) #assed on to consumers, 3he in$luence o$ mer)er syner)ies
on #rice de#ends on &oth the manner in .hich the $irmsG cost $unction is altered &y the
mer)er and on the nature o$ $irmsG interaction, At one e/treme, i$ a mer)er #roduces
no 9or ne)li)i&le: cost reductions &ut reduces the num&er o$ com#etitors in a 2ournot
or Bertrand oli)o#oly, and N is not %ery lar)e, #rices .ill rise nontri%ially and the
reduction in consumer .el$are .ill &e a##ro/imately eCual to the $all in total .el$are,
Bo tradeo$$ needs to &e considered, 3o ta*e another #ossi&ility, su##ose that t.o $irms
mer)e to mono#oly and that all the sa%in)s are in $i/ed costsJ then #rices .ill rise
9unless there .as #er$ect collusion #re%iously: &ecause $i/ed costs do not ordinarily
a$$ect #ricin) decisions, -o.e%er, i$ the increase in dead.ei)ht loss is not that )reat &ut
the sa%in)s in $i/ed costs is lar)e, then consumer .el$are may $all .hile total .el$are
rises, For a $urther contrast, consider a mer)er o$ t.o $irms in a settin) .ith many
$irms and assume that the cost sa%in) in%ol%es a reduction in mar)inal costJ then the
mer)ed $irm may #rice su$$iciently more a))ressi%ely to &rin) #rices do.n, in .hich
case &oth consumer sur#lus and total .el$are .ould rise, Qie.ed more &roadly, the
analysis in su&sections 1,1 and 1,2 a&o%e 9the latter o$ .hich dre. on section 3 on
collusion: ela&orates conditions under .hich #rices may &e e/#ected to rise or $all and
ho. those conditions de#end on the manner in .hich a mer)er may a$$ect $irmsG costs,
Accordin)ly, althou)h many mer)ers raise &oth consumer and total sur#lus or reduce
them &oth, there .ill e/ist a nota&le su&set o$ mer)ers that increase total sur#lus &ut
reduce consumer sur#lus, 3he resultin) need to choose &et.een consumer .el$are and
total .el$are as )uides to mer)er #olicy .as #resented shar#ly &y 5illiamsonGs 91'08:
classic discussion o$ the tradeo$$ &et.een mar*et #o.er and e$$iciencies and, in #articular,
his demonstration in a &asic case that e%en modest )ains in #roducti%e e$$iciency
could e/ceed the losses in allocati%e e$$iciency $rom #rice increases,130 7ecent contri&utions
to the de&ate a&out the #ro#er o&Eecti%e o$ antitrust #olicy include Farrell and
Dat4 92000:, -eyer 92000:, Dolas*y and 6ic* 92003:, and alo# 9200@:,
3he #recise lan)ua)e o$ the %arious statutesK.hich as #rior discussions indicate
are not ta*en literally or inter#reted inde#endentlyK)i%es con$lictin) )uidance in de<
13@ !ther Eurisdictions, .ith di$$erent la.s and histories, mi)ht )i%e .ei)ht to other o&Eecti%es in addition to
or instead o$ economic .el$are 9.hether consumer .el$are or total .el$are:, For e/am#le, antitrust #olicy in
the Euro#ean Union has traditionally #laced some .ei)ht on the %alue o$ inte)ration into a sin)le mar*et,
130 "n a mar*et .here #rices are initially at a com#etiti%e le%el, the &ene$its $rom #roducti%e e$$iciency are
a rectan)le 9assumin) a uni$orm do.n.ard shi$t in the mar)inal cost cur%e: .hereas dead.ei)ht loss is
the $amiliar trian)le, .hich $or small #rice increases .ill &e a much smaller Cuantity, -o.e%er, i$ #rice is
nontri%ially a&o%e mar)inal cost &e$ore a mer)er, as mi)ht &e e/#ected i$ #re<mer)er concentration is hi)h,
the incremental dead.ei)ht loss $rom #rice increases .ill &e lar)er, Also, cost sa%in)s $rom mer)ers, .hen
they e/ist, can ta*e many di$$erent $orms, so the &ene$it need not &e indicated &y a rectan)le 9that is, Cuantity
times a uni$orm chan)e in mar)inal cost:,
Ch% .1- $ntitrust 110F
terminin) the underlyin) #rinci#le,13F 2layton Act ection FGs #rohi&ition on mer)ers
that may su&stantially lessen com#etition or create mono#oly mi)ht &e inter#reted to
#rohi&it any mer)ers that reduce ri%alry or at least those leadin) to hi)h concentration,
and certainly to a sin)le $irm ser%in) the mar*et, Iet it is #ossi&le that e%en a mer)er to
mono#oly could raise &oth total .el$are and consumer .el$are, 3he herman ActGs #rohi&ition
on restraints o$ trade is inter#reted under the ru&ric o$ the rule o$ reason, .hich,
as has &een e/#lained, contains su&stantial am&i)uity e%en a$ter &ein) translated into the
#romote+su##ress com#etition test and re$ined throu)h modern a##lications, 3he F32
ActGs #rohi&ition o$ Mun$airN com#etition is %a)ue and Cuestion<&e))in) on its $ace,
3hat said, the modern trend in the United tates seems to &e to.ard a consumer .el$are
standard .hen considerin) the e$$iciencies de$ense, althou)h the le)al authorities
ha%e not ela&orately rationali4ed this %ie., s#eci$ically &y de$endin) it a)ainst the alternati%e
o$ total .el$are, 3he Mer)er AuidelinesG discussion o$ e$$iciencies at one #oint
re$ers to those that Mli*ely .ould &e su$$icient to re%erse the mer)erGs #otential to harm
consumers in the rele%ant mar*et, $or e/am#le, &y #re%entin) #rice increases in that
mar*et,N ?i*e.ise, a num&er o$ lo.er courts considerin) the issue ha%e indicated the
need $or the mer)in) $irms to demonstrate that e$$iciencies .ould &e #assed alon) to
consumers,138
From an economic #oint o$ %ie., ho.e%er, it .ould seem that in #rinci#le total .el$are
should &e the standard,13' ;roducers ha%e o.ners .ho are #eo#le, Eust li*e $inal
consumers, !ne mi)ht ne%ertheless $a%or consumers on distri&uti%e )rounds &ecause
o.ners are on a%era)e richer than consumers 9althou)h o&%iously the )rou#s o%erla#
su&stantially:, -o.e%er, it is usually more e$$icient to achie%e distri&uti%e o&Eecti%es directly,
throu)h the income ta/ and trans$er system,110 "n addition, one should add that,
13F 3he le)islati%e histories, ho.e%er, are clearer, 5hen the herman Act .as enacted in 18'0, it is unima)ina&le
that the le)islators a##reciated modern conce#ts o$ e$$iciency and dead.ei)ht loss, .hich .ere un*no.n
e%en to most economists at that time, ?e)islators did seem to care a&out hi)h #rices and also a&out the #rotection
o$ small &usinesses, As discussed &y the u#reme 2ourt in Brown Shoe, the 2layton Act amendments in
1'@0 that led to the %ersion o$ ection F that is close to its current $orm .ere moti%ated &y )eneral concerns
9lar)ely o$ a #olitical and social sort: a&out concentration o$ #o.er, most o$ .hich seem disconnected $rom
either consumer or total economic .el$are, As indicated &y the te/t in this section and in section 3, ho.e%er,
those %ie.s ha%e not si)ni$icantly in$luenced courtsG decisions in recent decades,
138 ee Areeda, -o%en*am#, and olo. 92000, %ol, 1A, #, 10 n, 1:,
13' A total .el$are standard is also consistent .ith ma/imi4in) .orld.ide .el$are, 5hen there are multi#le
Eurisdictions .ith authority o%er mer)ers .ith international s#illo%ers, di$$erent Eurisdictions mi)ht $ind it
in their national sel$<interest to ado#t di$$erent standards, For e/am#le, a country that is .ei)hted to.ard
consumers in sectors most in$luenced &y antitrust #olicy mi)ht .ant to &e stricter and use a consumer .el$are
standard, .hereas a country .ith stron)er #roducer+o.ner re#resentation amon) its citi4ens mi)ht #re$er
total .el$are or e%en a $ocus on #roducti%e e$$iciency, ee Au4man 91''8:, ome ha%e su))ested that a mi/
o$ national &ias and #rotectionismKrather than 9or in addition to: di$$erences in underlyin) a##roachesK
may e/#lain the 9$e.: instances 9the most nota&le &ein) the #ro#osed &ut ultimately a&andoned mer)er o$
A,E, and -oney.ell and the ultimately a##ro%ed mer)er o$ Boein) and Mc6onnell<6ou)las, su&Eect to some
conditions: in .hich the U,, and EU antitrust authorities ha%e ta*en di$$erent %ie.s o$ mer)ers or other
#ractices,
110 ee, $or e/am#le, Da#lo. and ha%ell 91''1: and Da#lo. 92001:,
1108 L% 7aplow and C% Shapiro
in antitrust, the #rotected class is o$ten #roducers, 3his is #atently so .hen the mer)in)
#arties sell intermediate )oods 9althou)h sa%in)s to #urchasin) $irms that o#erate in
a com#etiti%e mar*et .ill ultimately &e #assed on to su&seCuent consumers:, Furthermore,
antitrust la. treats &uyin) cartels and hori4ontal mer)ers that create mono#sony
#o.er similarly to the .ay it treats #rice<$i/in) and the creation o$ mar*et #o.er &y
sellers,111 Iet lo.er in#ut #rices can, to a de)ree de#endin) on mar*et structure and
other $actors, translate into lo.er #rices to $inal consumers, "$ only consumer .el$are
mattered, increases in &uyer #o.er throu)h hori4ontal mer)ers and other.ise mi)ht &e
#raised, not condemned,
;ra)matic considerations may, ho.e%er, #ro%ide Eusti$ication $or ca&inin) e$$iciency
de$enses in %arious res#ects, includin) #erha#s a $ocus limited to consumer .el$are,
For e/am#le, ado#tin) a consumer .el$are standard may induce $irms to underta*e
deals that o&tain #otential syner)ies .hile causin) less harm to com#etition, leadin)
to e%en hi)her total .el$are than .ould a total .el$are standard, "n any e%ent, the
#re%iously<noted di$$iculties o$ ascertainin) mer)er<s#eci$ic e$$iciencies in indi%idual
mer)ers, e: ante, counsel caution, .hich is already re$lected in the Mer)er Auidelines
and also seems e%ident in court decisions, "ndeed, this may e/#lain .hy a%era)e e$$iciencies
hea%ily in$luence the thresholds $or anticom#etiti%e e$$ects, &oth )enerally and
in s#eci$ic industries, .hile at the same time e$$iciencies are o$ten )i%en short shri$t in
e/aminin) #articular cases, a&sent an unusually stron) demonstration, But once su$$iciently
#ersuasi%e #roo$ o$ aty#ically )reat e$$iciencies is o$$ered, it is not clear that
a reCuirement o$ #ass<throu)h to consumers ma*es the a)enciesG or courtsG tas* easier
rather than harder, !n one hand, determinin) #ass<throu)h reCuires resolution o$ an
additional, challen)in) issue, !n the other hand, in some instances the &est e%idence
o$ a mer)erGs e$$ects .ill &e $rom o&ser%ed #ricin) &eha%ior that may re$lect a com&ination
o$ anticom#etiti%e e$$ects and e$$iciencies, For e/am#le, in Staples, e%idence o$
hi)her #rices in more concentrated mar*ets #resuma&ly re$lected the com&ined im#act
o$ concentration and .hate%er #ass<throu)h o$ e$$iciencies may ha%e &een occurrin),
Bote that reductions in mar)inal cost are )enerally #assed throu)h to some de)ree to
consumers, e%en &y a mono#olist, althou)h the #ass<throu)h rate can &e sensiti%e to the
sha#e o$ the demand cur%e and the nature o$ oli)o#olistic interaction,112
3here are also lon)er<run concerns relatin) to dynamic e$$iciency, ome de)ree o$
com#etition has o$ten &een thou)ht conduci%e to $irmsG runnin) a ti)ht shi#, &etter ser%in)
customers, and &ein) more inno%ati%e, 3he relationshi# &et.een com#etition and
111 For e/am#le, the Mer)er Auidelines state= M3he e/ercise o$ mar*et #o.er &y &uyers 9Rmono#sony #o.erG:
has ad%erse e$$ects com#ara&le to those associated .ith the e/ercise o$ mar*et #o.er &y sellers, "n order
to assess #otential mono#sony concerns, the A)ency .ill a##ly an analytical $rame.or* analo)ous to the
$rame.or* o$ these Auidelines,N
112 Bulo. and ;$leiderer 91'83: deri%e an e/#ression $or the #ass<throu)h rate $or a mono#olist, ha#iro
91'8': discusses #ass<throu)h rates in %arious oli)o#oly models,
Ch% .1- $ntitrust 110'
inno%ation has #ro%en di$$icult to esta&lish as a )eneral theoretical matter,113 "n any
e%ent, i$ the #rocess o$ com#etition itsel$K.hich seems to ha%e &een $a%ored &y the
dra$ters o$ antitrust le)islation and earlier courts and commentatorsKis o$ some %alue,
&ut this %alue is di$$icult to measure, then it ma*es sense to tilt the &alance a)ainst concentration,
3his mi)ht &e done &y ma*in) the threshold $or challen)e lo.erKreCuirin)
less demonstration o$ anticom#etiti%e e$$ectsKor throu)h other means, such as &ein)
less )enerous in considerin) e$$iciencies in Eusti$yin) other.ise #ro&lematic mer)ers,
Areater stin)iness mi)ht &e accom#lished &y raisin) #roo$ &urdens or &y im#osin) additional
reCuirements, such as &y reCuirin) sa%in)s to &e #assed alon) to consumers, 3he
o#timal manner o$ incor#oratin) these sometimes su&tle and ty#ically un#redicta&le
dynamic concerns into hori4ontal mer)er #olicy is not o&%ious,
2%1% &aret analysis under the Hori9ontal &er*er ?uidelines
3he -ori4ontal Mer)er Auidelines are utili4ed to e%aluate s#eci$ic #ro#osed mer)ers,
reCuirin) one to mo%e $rom a&stractions and )eneral theories to decisions a&out a concrete
case, 3o moti%ate this #ro&lem, consider the #ro#osed mer)er o$ U Air.ays and
6elta Air ?ines, "n Bo%em&er 2000, U Air.ays made an unsolicited o$$er to acCuire
6elta, .hich at the time .as in &an*ru#tcy, ;redictin) the e$$ects o$ this mer)er on
airline #assen)ers is hi)hly com#le/, First, one must identi$y the routes on .hich U
Air.ays and 6elta are si)ni$icant direct ri%als, 3hen one see*s to determine .hether
there .ill &e a unilateral or coordinated #rice increase on those routes, 6i%ersion ratios
and )ross mar)ins are certainly rele%ant to this inCuiry, &ut many other $actors enter
into the #icture as .ell, 5ill com#etition $rom other carriers, includin) lo.<cost carriers,
#re%ent the mer)ed entity $rom #ro$ita&ly raisin) $aresW Ai%en the com#le/ $are
structures that are used in the airline industry, .ill the mer)er ha%e di$$erent e$$ects on
$ares $or di$$erent classes o$ customers, such as leisure %ersus &usiness tra%elersW -o.
do $reCuent<$lier #ro)rams a$$ect the analysisW 5ill the mer)er )enerate su&stantial e$$iciencies
&ased on runnin) a lar)er net.or* o$ $li)hts at a sin)le airlineW "n e%aluatin)
these e$$iciencies, ho. does one $actor in the role o$ airline alliances, a less com#lete
$orm o$ colla&oration than a mer)erW "$ e$$iciencies lead to lo.er $ares on some routes,
&ut i$ the reduced com#etition leads to hi)her $ares on other routes, ho. does one &alance
these di%erse e$$ects on di$$erent sets o$ consumersW ince one is loo*in) ahead a
year or more, is the industry chan)in) in si)ni$icant .ays, such as throu)h the )ro.in)
role o$ re)ional Eets and lo.<cost carriers, that ma*e it a##ro#riate to discount historical
e/#erience .hen ma*in) #redictionsW ?astly, ho. stron) a com#etitor .ill 6elta &e i$
it is not acCuired &y U Air.ays, es#ecially )i%en the $reCuency o$ &an*ru#tcies in this
industryW
113 2om#etition )enerally enhances inno%ati%e incenti%es, as $irms see* to )ain )round on their ri%als &y
introducin) ne. and im#ro%ed #roducts, But a com#etiti%e mar*et structure can cause #ro&lems $or inno%ation
i$ $irms ha%e di$$iculty a##ro#riatin) the returns $rom their inno%ati%e e$$orts, $or e/am#le, due to ra#id
imitation &y ri%als,
11F0 L% 7aplow and C% Shapiro
5e .ill consider only some o$ the core, recurrin) issues,111 3he $irst ste# in the
analytical $rame.or* #ro%ided &y the Mer)er Auidelines is de$inin) the rele%ant mar*et,
ty#ically a rele%ant #roduct mar*et alon) .ith a rele%ant )eo)ra#hic mar*et, 3hen, as
#re%iously descri&ed in su&section 1,1,2, the )o%ernment determines .hether there is a
li*ely com#etiti%e ris* &ased on the #ost<mer)er --" and the increase in --" due to
the mer)er, "$ there is, it considers entry and other $orms o$ su##ly res#onse 9as .ell
as e$$iciencies, already discussed in su&section 1,1,3, and the #ossi&ility that one o$
the $irms is $ailin):, "n this su&section, .e descri&e this a##roach and relate it to the
#recedin) economic analysis and $urther .or*,

You might also like