Idenifying Merger Unilerl Effecs: HHI or Simulion? Jerome FONCEL Universiy of Lille, Frnce erome.foncel@univ-lille3.fr Mrc IVALDI Toulouse School of Economics, nd CEPR, Frnce ivldi@cic.fr Jrissy MOTIS Toulouse School of Economics, Frnce nd Universiy of Cree, Greece. rissy.mois@univ-lse.fr December 2006 Absrc We propose mehodology o compre he predicive nd screening power of nicompeiive horizonl mergers of ess. The mrke concenrion es (HHI) nd he es of unilerl effecs vi simulion. For his, we genere n economy wih muli-produc firms compeing in differenied produc mrkes nd consumers purchsing ccording o common nd idiosyncric ses for producs chrcerisics. Wih he d genered in his rue economy, n pproxime economy h fis he pre-merger equilibrium is esimed. We hen esime he unilerl merger effecs in prices of hypoheicl merger in boh, he rue nd he pproximed economies. We nex compre he pos-merger prediced unilerl effecs s well s he HHI levels nd heir corresponding dels in boh economies. We find h when srucurl economeric pproch ( subsnive es) is missed, he mrke shres pproch HHI (he dominnce es) ends o bis unilerl effecs upwrds. We confirm h he evluion of unilerl effecs by mens of simulion is very much influenced by he mrke size under scruiny. In priculr, when he shre of he ouside good is lrge decision bsed on he chnge of HHI would hve nohing o do wih decision bsed on unilerl effecs. Keywords: mergers, nirus, differenied produc mrkes, simulion. JEL codes: L0, L4, K2, D78
Accurcy of Merger Simulion. Inroducion The effecs of horizonl mergers re mong he min concerns for compeiion uhoriies. The reson is h horizonl mergers my enhnce nd/or srenghen mrke power. Mrke power in urn fcilies firms o increse prices, reduce oupu, choice or quliy nd deer innovion he expense of consumers. Since compeiion uhoriies generlly delinee welfre s consumers surplus (no including firms surplus), he previous merger effecs re perceived s welfre derimenl. Hence, subsequen concern for nirus gencies is he vilbiliy of suible n effecive nlyic nd quniive ools h fcilie hem o deec wheher nd o wh exen merger will induce mrke power. Responding o hese concerns, growing body of oligopolisic models of compeiion o nlyze merger effecs hs been proposed in he heoreicl indusril economics lierure. On is side, he empiricl lierure hs lso considerbly dvnced in providing quniive mehods h nswer o he problems cused by he nlyicl complexiy of hese heoreicl models. Incresing relince on hese dvncemen nd improvemens, compeiion gencies hve lso upded heir modes of invesigion nd implemenion. The Europen Communiy Merger Regulion, ECMR hereinfer, hs for insnce undergone ino reform h hs enlrged is scope of merger effecs invesigion. This reform kes concree form in he ccompnying Horizonl Merger Guidelines (HMG) in which wo imporn conceps hve been inroduced o ssess he scope of mrke power induced by he merger. The firs one is he Herfindhl Hirschmn Index, (HHI) nd is corresponding chnges (he dels). They re considered indicors of mrke srucure. The second one is he differeniion beween he coordined nd non-coordined (or unilerl) effecs of he mergers. Previously, invesigion ws focused on he coordined effecs of he merger nd i ended o rely on he mrke shres of he underking firms. The prohibiion crieri relied on wheher he merger creed or srenghened dominn posiion s resul of which effecive compeiion would be significnly impeded. This crierion could no cover, for insnce, cses in which he merged eniy would no be dominn bu sill hve mrke power. Afer hving recognized h focusing on coordined effecs, oher nicompeiive mergers (below he dominnce crieri) could no be covered, he new prohibiion crierion now reds h merger mus be blocked if i would would significnly impede effecive compeiion, in priculr s resul of he creion or srenghening of dominn posiion. This is referred s he SIEC (significnly impedimen in effecive compeiion) subsnive es. The mrke power subsnive es hs hen been doped o Foncel-Ivldi-Mois/ Februry 2007
Accurcy of Merger Simulion include ny form of mrke dominnce nd in priculr he unilerl effecs of he merger. Indeed, he guidelines ppel for n evluion of he merger pos-merger unilerl effecs hrough merger simulion models. To sum up, wo mesures of mrke power re ough o be used in merger cses. However, he relionship beween hese wo mesures poenil mrke power, he HHI nd he unilerl effecs, is cler only for he cse of quniy compeiion mong homogeneous producs. How much hey converge or diverge ccording o chnges in he differen fcors of more generl economic model remins o be sysemiclly invesiged. Acully, n open quesion for he ECMR is wheher he new subsnive es poins owrds modifying he nlysis of he HHI pproch owrds more effecs-driven pprisl of mergers. In sense, he quesion is wheher he wo ypes of mesures of mergers effecs re subsiues or complemens. In his sudy, we im deermining under which circumsnces he vrious srnds of he merger simulion pproch my bring vlue dded in ligh of he new subsnive es wih respec o he rdiionl HHI pproch. In priculr, his sudy ries o nswer o he following quesion: Cn simulion nlysis be used s ool for predicing he price effecs of merger? Should simulion be used primrily s screening device, nlogously o concenrion mesures? How ccurely cn n impedimen o compeiion be esimed by simulion? Cn his mehod reduce he probbiliy of incurring boh Type I nd Type II errors, nmely, blocking pro-compeiive merger nd llowing nd nicompeiive merger respecively? To ddress hese quesions, we propose generic mehodology h llows evluing he wo ess, he HHI nd merger simulion models, o ssess merger effecs. In priculr, our sudy criiclly evlues he dvnges nd he drwbcks of he simulion echniques relive o he concenrion es. Our mehodology consiss in genering by rndom drws rue mrke of differenied producs nd consumers, in which merger kes plce. In his rue economy we simule merger nd compue is effecs in erms of HHI s chnges nd unilerl effecs (price increses). Then, wihin his rue economy, n pproximed economy is esimed o recover he cul equilibrium of he mrke. Noe h his is wh is usully done in he cse by cse bsis. In his pproximed economy, he sme merger is simuled nd he chnges in HHI nd unilerl effecs re mesured. By compring he relionship beween he wo One is compulsory, he HHI (nd is del) nd he merger simulion is highly recommended. Foncel-Ivldi-Mois/ Februry 2007 2
Accurcy of Merger Simulion mesures of mergers effecs we find h he HHI es ends o be upwrds bised compred o he subsnive es of unilerl effecs. Moreover, we find h hese resuls re sensiive o he choice of he size of he mrke. The second resul is reled o he size of he mrke, or more precisely o he size of he mrke shre of he ouside good (ll he goods h do no belong o he relevn mrke). In priculr, when he size of he ouside good is lrge, predicion bsed wih he HHI es hs nohing o do wih predicion bsed on he SIEC es. Th is, in his cse he chnge in prices nd he chnges in HHI induced by he merger re no reled o ech oher i.e., decision bsed on he chnge in HHI would hve nohing in common wih decision bsed on he chnge in prices. On he oher side, when he size of he ouside good is smll, predicion bsed wih he HHI es will be in ccordnce wih predicion bsed on he SIEC es. These resuls hve imporn implicions for merger conrol. They sugges h when uniquely relying on he HHI es, uhoriies risk incurring on ype II error, i.e., on prohibiing merger lhough here is no serious compeiion concern. In Secion 2 we describe he concenrion nd he unilerl effecs ess, nd he ools required o be implemened. Th is we describe he HHI nd he simulion models. In Secion 3 we expose our mehodology of comprison. In Secion 4 we develop our mehodology by mens of simulion ools. Secion 5 exposes he resuls of our experimens nd Secion 6 concludes. 2. Tools for he Assessmen of Merger Effecs In nlyzing mergers nd heir compeiive effecs, he min quesion is wheher he merger will enhnce or srenghen mrke power, or more specificlly, wheher he merger will led o subsnil price increses. In heory, mrke power my be induced by merger hrough les wo disinc mechnisms: collusion (or coordined effecs) beween oherwise compeing firms or independen recion of he merging firms (non-coordined effecs). In he ECMR he firs mechnism is nlyzed under he collecive dominnce concep, while he second refers o he more recenly inroduced concep of unilerl effecs nd which does no require dominnce o be considered s ni-compeiive. Coordined effecs would resul from he merger if due o fewer compeiors in he mrke, hey re enbled o (implicily) coordine heir behvior o se higher h compeiive prices. Unilerl effecs refer o he biliy of he merging firm o rise prices, irrespecive of he pricing decisions or cions of heir compeiors. Th is, unilerl effecs Foncel-Ivldi-Mois/ Februry 2007 3
Accurcy of Merger Simulion would rrive even if he wo compeiors h merge do no cree dominn plyer in he mrke bu becuse of he removl of compeiive consrins resuling from he merger. 2 The ssessmen of he unilerl effecs sred from he inquieude h mergers mosly occur in differenied produc mrkes. The seminl work of Deneckere nd Dvidson (985) showed h in hese mrkes nd wih firms compeing in prices, merger is lwys profible (for he merging nd he non-merging firm) for insiders nd ousiders firms. The reson is h merging subsiue produc gives incenives o he firm o increse prices since in price compeiion models prices re sregic complemens, n overll increse in price is observed unless srong synergies re consequence of he merger. This resuls in higher hn prices even if he indusry members were no coordining heir cions, h is, even when collusion ws no n opion. In he conex of compeiion models in quniies, Frrell nd Shpiro (990) derive similr resul. Even hough in such models quniies re sregic subsiues, i.e., merging firms reduce oupu nd rivls increse i, horizonl mergers led o higher prices unless hey genere synergies. The reson of such similr resuls is h if he merging firm produces goods highly subsiuble, i will be rionl for i o rise prices o some degree, becuse i will recpure some of he cusomers who would hve swiched wy in fvour of wh ws previously compeing produc. In oher words, wihin he differenied produc conex, pos-merger price increse does no depend on he merged firm being he dominn plyer in he mrke, bu insed, on he subsiubiliy of he producs in quesion. Indeed he closer he subsiue goods re he greer he unilerl effecs will be. Then, he nlysis of unilerl effecs of mergers relies on he compuion of he own nd cross elsiciies of he producs in he relevn mrke becuse hey deermine firms biliy o increse prices. Even when someimes he difficuly of obining price elsiciies reurns he nlysis o he definiion of mrke concenrion nd mrke shres s indirec evidence of heir vlue, one mus keep in mind h he ler re no he issue in he nlysis of unilerl effecs. Here, he core of he nlysis i he fc h inelsic preferences of consumers llows firms o exer mrke power. The wo pproches cn predic very differen merger effecs nd herefore conclude differen decisions (llow or block) priculr merger. As mer of fc, even when elsiciies re he driver o exer mrke power, economic heory does no sugges eiher ignoring mrke shres or concenrion in merger 2 As mer of fc, in merger cse i is possible o llege boh unilerl nd coordined effecs in merger cse. They cn be regrded s cumulive effecs. Foncel-Ivldi-Mois/ Februry 2007 4
Accurcy of Merger Simulion nlysis. This migh be he reson why he HHI es remins he firs invesigion sk in merger cses. 2.. Concenrion nd HHI The curren EU HMG specify hresholds for mrke shres nd for he HHI (nd is posmerger chnges) for he ll he producs in he relevn mrke. They se h combined mrke shre of he merged firms lower hn 25% does no risk effecive compeiion wheres pos-merger mrke shre of 50% nd more my in hemselves be evidence of he exisence of dominn posiion. The HHI is clculed s he sum of squres of mrke shres i.e., HHI J 2 = s =, where J is he number of producs compeing in he relevn mrke, s, is he mrke shre of produc. Mrke shres re defined s q Q, where q is he ol quniy of produc produced in he ol relevn mrke, in urn composed of (ol producion). The HHI depends hus on he number of firms nd he dispersion of mrke shres. As i cn be noiced, he criicl poin of his pproch is he definiion of Q, i.e., he relevn mrke, which is defined in produc nd geogrphic dimensions. Is boundries re supposed o be esblished by demnd subsiubiliy. The HHI crieri se h he merger is unlikely o rise compeiive concerns if i concerns: mrkes wih pos-merger HHI below 000. mrkes wih pos-merger HHI beween 000 nd 2000 nd HHI below 250. mrkes wih pos-merger HHI bove 2000 nd HHI of less hn 50. 3 J = q = Q The ECMR reds ech of hese HHI levels, in combinion wih he relevn dels, my be used s n iniil indicor of he bsence of compeiion concerns. However, hey do no give rise o presumpion of eiher he exisence or he bsence of such concern. 4 Two disurbing piflls re worh menioning bou he merger effecs predicing power of he HHI es. Firs, he HHI is consruced under he ssumpion of homogeneous goods, however, s i is brodly recognized, horizonl mergers mosly occur in mrkes of differenied producs. In fc, he single cse in which he HHI will yield o ccure 3 There re excepions o hese rules. They re: ) If merger involves poenil enrn or recen enrn wih smll mrke shre. b) If one or boh pries re imporn innovors. c) If here re significn cross-shreholdings mong he mrke pricipns. d) If one of he merging firms is mverick wih high likelihood of disruping collusive conduc. e) If indicions of collusive behviour re presen. f) If one of he pre-merger pries hs mrke shre of 50% or more. 4 See hp://europ.eu.in/eur-lex/pri/en/o/d/2004/c_03/c_0320040205en0005008.pdf. Foncel-Ivldi-Mois/ Februry 2007 5
coss. 5 The second pifll of he mehod is he compuion of he pos-merger HHI wih which Accurcy of Merger Simulion predicions of mrke power is he one in which firms compee in quniies wih homogeneous producs, consn elsiciy of demnd nd symmeric consn mrginl he del will ferwrds be compued. I is no ccure since he pos-merger mrke shres re no reclculed ccording o new equilibrium bu only summed up in heir pre-merger vlues. This imprecision occurs since horizonl merger models predic h he pos-merger mrke shre of he merged firm will be smller hn heir pre-merger combined mrkeshres. Thus, he cul compuion of he pos-merger HHI overesimes he rue posmerger HHI vlue since i does no ke ino ccoun he inernlizion effec of he merging firms. In hese models he use of mrke shres nd he inference bou compeiion from hem is no robus o inpproprie mrke delineion nd hey (lone) migh no reflec he degree of compeiion in differenied producs mrkes. In sum, he HHI is poor indicor of mrke power in mrkes wih differenied producs nd nobly in he cse of price compeiion. Uniquely relying on he HHI es migh no void he risk of incurring on ype II error. Merger conrol my py wo coss when flling ino error Type II, direc one by forgoing welfre gin, nd more subecive one by deerring compeiive mergers h were no pu forwrd o he uhoriies becuse of perceived low probbiliy of success. In fc when looking he sisics of noified cses nd he number of prohibiions of he EC merger conrol, one observes h from 990 o 2000, he number of noified cses rose sedily nd us fer he mximum of prohibied cses occurred, in he 200, srong decrese in noified cses is observed. In models of price compeiion wih differenied producs, he relevn mesure of mrke power, he mrk-up, is deermined by he own nd cross price elsiciies of he merging producs wih respec o heir compeing producs. The subsiubiliy of producs is he key elemen in he ssessmen of he impc of mergers. In such mrkes he mos 5 To illusre his, recll h in model of quniy compeiion wih homogeneous goods, he mrkup of produc is expressed s : p c / p = s / η, where c is he consn mrginl cos of firm (or produc) nd η is he price elsiciy of ggrege demnd. Th is, he price-cos mrgin is proporionl o he mrke shre nd he inverse of he price elsiciy of demnd. In priculr, he smller he mrginl cos, he higher he mrke shre. I urns ou h if mrginl coss re ll equl nd consn he verge uni cos is: J J c= s c. Then by muliplying boh sides of he equion bove by s nd summing over ll firms yields: = J 2 / = / / 0000 HHI = p c p c s η η =. In oher words, for fixed elsiciy of demnd, he mrk-up nd herefore mrke power is proporionl o he HHI. = Foncel-Ivldi-Mois/ Februry 2007 6
Accurcy of Merger Simulion imporn drivers of compeiion re he coss srucure nd he nure of consumers demnd which cn be nlysed by mens of simulion models of unilerl effecs. 2.2. Merger Simulion Merger simulion is he forml use of gme heoreic models o mke quniive predicions of unilerl effecs. Such models ough o mch he closes possible he criicl feures of he mrke under nlysis. Indeed, his is one of he dvnges of merger simulion, h i cn mke cler on wh ssumpions he resul depends. Merger simulion of unilerl effecs proceeds in hree min seps. The firs sep is he specificion of he funcionl forms for demnd nd supply. The second sep consiss in he esimion or clibrion of he prmeers of such funcions. The hird sep uses he esimed prmeers (from he second sep) o predic he pos-merger prices nd mrke shres nd hen compred hem o he pre-merger ones. 6 From his comprison, unilerl effecs re finlly mesured. Th is, he unilerl effec of he merger is he difference beween he pre-merger nd he pos-merger price equilibrium. Noe h wihin merger simulion, he concenrion index resuled in he posmerger equilibrium cn be compued nd compred o is pre-merger vlue s well. In his sense, performing simulion does no rule ou concenrion nlysis, wheres using uniquely concenrion rules ou he unilerl effecs nlysis. Thus, he simulion pproch is more complee hn he HHI pproch becuse he former looks boh demnd nd supply wheres he former regrds only he supply side of he mrke. In wh follows we discuss hese building blocks of merger simulions. 2.2.. Demnd The depring building block for he ssessmen of unilerl effecs hrough merger simulion is produc differeniion. The reson is h in differenied producs mrkes firms cn hve mrke power if heir producs hve unique feures h mke compeing producs poor subsiues. And, his is precisely why in hese mrkes, he evluion of mrke power depends of ech produc s demnd subsibiliy. Then, he demnd side specificion plys crucil role. Indeed, models re highly sensiive o ssumpions on 6 A his ls sep efficiency gins (modeled s fixed decreses in mrginl coss) cn be ken implemened in he simulion exercise. Foncel-Ivldi-Mois/ Februry 2007 7
Accurcy of Merger Simulion consumer preferences. 7 In hese models, produc differeniion cn be cpured by defining producs s bundle of chrcerisics. For exmple, cr my be defined by is size, horsepower, miles per gllon, quliy of is service nework, ec. hese ls re hen he chrcerisics of he produc. Two ypes of models re usully employed for he esimion of he demnd side. The one defining consumers preferences on produc spce or coninuous choice demnd nd he one defining consumers preferences in chrcerisic spce or discree choice demnd. A coninuous choice model of demnd (i.e. log-liner or AIDS) hs he inconvenien h he number of prmeers grows exponenilly wih number of producs under nlysis. The reson is h he number of prmeers o be esimed is direcly linked o he own nd cross price elsiciies. 8 For exmple, if he mrke hs J producs, demnd sysem h includes he producs hemselves requires o esime 2 J : J for ech own price elsiciy nd plus J cross price elsiciies for ech produc. This drwbck is voided when using he second cegory of demnd sysems which includes he vrious kinds of logi models. Undenibly, he min dvnge of hese models is heir prsimony in he number of prmeers o be esimed. I is subsnilly reduced from 2 J o he number of chrcerisics he produc is composed of. 9 Th is, he number of prmeers o be esimed is independen o he number of producs in he relevn mrke. Furhermore, hese prmeers deermine ll cross price elsiciies independenly of he size of he mrke. Thus, one cn ypiclly include mny more producs wihin he discree choice pproch nd sill obin firly precise prmeer esimes. Anoher dvnge of his sysem is h i llows nlyzing he likely impc of new produc before i hs been cully inroduced ino he mrke. For is dvnges, in his sudy we employ he discree choice pproch. Then, discree choice demnd is obined by defining consumers preferences s funcion of producs chrcerisics insed of producs hemselves. For furher flexibiliy consumers preferences over he vrious produc chrcerisics cn be se such h lso depend on consumers chrcerisics (for exmple, lrge fmilies migh prefer lrge crs, ec). Furhermore, he echnique enbles oo o inroduce in he nlysis no only he producs 7 In conrs, unlike he siuion wih differenied producs, merger simulion in Courno indusry wih homogenous goods requires he specificion of indusry boundries, single nd rough esime of he indusry s elsiciy of demnd supplies mrke delineion. Th is, only he ggrege demnd elsiciy of he indusry is needed. 8 For exmple if he mrke hs J producs, demnd sysem h includes he producs hemselves requires o esime J 2 cross price elsiciies becuse for ech produc, one hs is own price sensiiviy plus J- cross price elsiciies. 9 Twice he number of chrcerisics indeed, one ime for he men of such chrcerisic nd second ime for he vrince. Foncel-Ivldi-Mois/ Februry 2007 8
Accurcy of Merger Simulion observed chrcerisics bu lso he unobserved ribues o he nlys. Unobserved ribues re hose h re observed by firms h produce hem nd perceived by consumers bu no qunifible by he nlys. Quliy nd he effec of dverising re exmples of ribues h provide uiliy o buyers nd profis o firms nd cnno be qunified by he nlys. Wih his pproch simulion lso ddresses he quesion of defining he size of he relevn mrke becuse i kes ino ccoun h consumers my choose he ouside lernive. This is composed good which ghers ll he producs h re no of direc ineres for he merger under invesigion bu h could c s subsiues. Indeed, he robusness of resuls o he choice of mrke size nd he size of he ouside good is crucil. To derive demnd, consumers re ssumed o mximize heir uiliy wih respec o hese chrcerisics when mking heir choices. If he size of he mrke is M (h is, he number of consumers in he economy) 0 hen, demnd cn be direcly expressed in mrke shres. For insnce, he demnd of good is expressed s s = q M, where q is he number of imes produc ws chosen, nd herefore s is he mrke shre of firm. Noe h in his conex, esimed elsiciies will be very sensiive o he choice of he ouside good. 2.2.2. Berrnd compeiion The oher building block of merger simulion is supply. For he supply side of he mrke wo feures re required, n ssumpion bou he nure of compeiion mong firms nd n ssumpion on heir cos funcion. Concerning he nure of compeiion, he Berrnd price compeiion one is he mos convenien wihin he differenied producs frmework Indeed, some merger simulion sudies hve esed he fi of his ssumpion. For insnce, Pinkse nd Slde (2004) nd (200) performed sisicl es on observed nd esimed verge mrgins nd found h he Berrnd hypohesis could no be reeced. In ddiion, modelling firms s muliproduc producers is suggesed o ensure relism. This is imporn in he nlysis of merger effecs becuse when muliproduc firm considers incresing he price of one of is brnds, i will ke ino ccoun how much of he los demnd will go o is oher producs. In fc, no considering he muliproduc spec (nd modelling firms s single-produc eniies) would led o downwrd bis in he esime of he firm s mrk-up. The reson is h in he single produc firms cse mrk-ups re only in funcion of heir corresponding own price elsiciy. This ignores he fc h price increse 0 Recll h ech individul purchses one produc eiher from he relevn mrke se or from he ouside lernive. Foncel-Ivldi-Mois/ Februry 2007 9
Accurcy of Merger Simulion in brnd will led o n incresed demnd for oher goods produced by he sme firm. Husmn, Leonrd nd Zon (994) found h when ssuming single produc firms, he mrk-up of produc is 7% lower hn he mrk-up of he sme produc hn if ssuming muliproduc firms insed. Thus, king ino ccoun h firms produce mny producs is imporn in models of compeiive behviour wih differenied producs. Concerning he cos srucure, nlyss ofen rely on he ssumpion of consn mrginl coss. In fc, wih such cos srucure, merger would resul in incresed prices irrespecive of he presence of efficiency gins. In ddiion, efficiency gins cn be inroduced in n d hoc wy in he simulion process by, for exmple, decresing by cerin percenge he level of he consn mrginl cos curve. In wh follows we explin how we develop our mehodology of comprison of hese wo merger effecs predicive ess, he HHI nd merger simulion. 3. A Mehodology of Comprison We develop experimens in order o compre he wo differen ess of poenil merger effecs. For his, we genere n economy wih complee d bou prices, shres, nd consumer ses. Wih he genered d we esime n pproximed economy s i is usully done in merger cses. In boh, he rue nd he pproximed economy, we simule merger o obin he new equilibrium in prices nd mrke shres. In boh, he rue nd he pproximed economy, we perform he wo ess, he HHI nd he unilerl effecs es, o mesure he impc of he simuled merger. In priculr, o develop our experimens we need workbench consising on dse bou specific economy composed of consumers, firms nd producs. We genere his economy by rndom drws. Wihin his economy we deermine he cul mrke equilibrium in erm of prices nd mrke shres. We hen simule merger o predic pos-merger equilibrium, i.e., pos- merger prices nd mrke shres. However, more flexible cos srucures cn nd should be employed. Foncel-Ivldi-Mois/ Februry 2007 0
Accurcy of Merger Simulion For he generion of he economy nd he simulion exercise we define how he differenied producs nd populion of consumers wih common nd idiosyncric preferences over producs re creed. We describe firms behviour in regrd o heir rivls nd cosumers s well s consumer s behviour in regrds o produc decision mking. We ssume h muliproduc firms re proposing differenied producs composed of bundle of chrcerisics, nd individuls re selecing one mong hem. Individuls uiliy (from which demnd is derived) is modelled in wy such h heir crieri of decision rely on he chrcerisics of he produc (including he price). Individuls re ssumed o mke discree choice i.e., o selec he produc mong he whole vriey of exising differenied producs in he mrke h gives hem he grees uiliy. By summing up consumers choices demnd is hen deermined. On he supply side we ssume h rivl firms compee in prices. By fiing supply nd demnd, he equilibrium of he mrke is obined. Pre-merger prices nd quniies re herefore vilble informion in he economy. Firms nd individuls perns of behviour re prmeerized which in urn llows deermining wh hppens in he economy when merger kes plce. Then he simulion of merger is performed in he rue economy nd he pos-merger effecs compued. The pos-merger equilibrium of prices nd quniies is vilble informion oo. Besides he pre nd pos-merger equilibrium prices we re ble o compue in his economy he pre nd pos merger levels of HHI nd is corresponding del. Moreover, we esime wh we clled he rue pos-merger HHI, nd is corresponding rue del. Th is, we consider he inernlizion effec of he merging firms oin mximizion of profis fer merger o compue pos-merger mrke shres. Nex, wihin his genered economy we esime n pproximed economy. Th is, we employ sndrd economeric ools, o esime he equilibrium of he mrke nd o simule he effecs of merger from he vilble d (our genered d). By doing so, we obin esimed pre nd pos merger prices, esimed pre nd pos merger levels of HHI nd heir corresponding dels. This procedure is summrized nd depiced in Digrm. 3.. Generion of he rue economy For he merger simulion exercise we require d bou producs, firms, nd consumers. This d is genered from he underlying ssumpions bou producs chrcerisics, consumers preferences nd firms behviour nd cos srucure. Firms re modelled s muliproduc producers wih consn mrginl coss h depend on he chrcerisics of he producs nd h compee in prices o se supply. Consumers uiliy is Foncel-Ivldi-Mois/ Februry 2007
Accurcy of Merger Simulion modelled such h i depends on income, common nd own ses for he chrcerisics of he producs in he mrke. A mximizion progrm is buil up for firms profi nd consumers uiliy. Joining he soluions of hese wo mximizion progrms we obin he equilibrium of he mrke in erm of prices nd quniies (mrke shres). In wh follows we furher describe he generion process of ech elemen of our virul economy. 3... Generion of producs nd firms On he supply side of he economy we genere mrke composed of five firms indexed by f, where f=,,f nd F=5. These firms re symmeric in size nd in cos srucure. Ech f produces se of 00 differenied producs denoed Ω f, nd producs re indexed by =,,J nd so J=500. The differeniion of he producs relies on heir k chrcerisics X k s well s on is respecive price p. Th is, producs re composed of five ribues indexed by k, where k=,,k nd K=5. Two of he ribues, x nd x 4, re coninuous rndom vribles genered wih norml disribuion nd he remining hree re discree rndom vribles genered wih binomil disribuion. Th is, 2 2 h xk N( 0, σ k) for k=, 2 nd xk U( 0, σ k) X k is genered such k=3,4,5. In ddiion, ech produc belongs o ny of four he cegories g=,,g, nd G=4, where cegories cluser producs of similr chrcerisics, for exmple, lrge nd smll size would be wo disinc cegories of crs. This cegory g is genered wih uniform disribuion by giving frcion of he 2 horizonl line o ech group g, i.e., xg U( 0, σ g). Ech rndom vrible hs eiher n ssocied probbiliy funcion (in he discree cse) or probbiliy densiy funcion (in he coninuous cse). Besides hese six observed ribues, we genere sevenh one which is ssumed o be observed by consumers bu no by he nlys. I is coninuous vrible genered wih norml disribuion h we denoe 2 quliy of he produc, i.e., ξ U( 0, σξ) ξ nd h cn be inerpreed s he. As noed bove, ll he vribles following norml disribuion re ssumed o hve zero men nd heir vrince is genered by rndomizion rouine. In summry, ech produc is genered wih se of seven observed chrcerisics grouped in vecor X plus n unobserved chrcerisic, ξ. The six observed ribues of Foncel-Ivldi-Mois/ Februry 2007 2
Accurcy of Merger Simulion he produc re decomposed on X = X k + X g for k=,,5 nd g cn ke eiher vlue beween, 4. Finlly, he eighh chrcerisic of produc is is price, A funcion for he consn mrginl cos of ech produc is genered by specifying i s log-linerly dependn on he observed chrcerisics of he produc, p. X. In ddiion, he mrginl cos funcion conins noher exogenous vrible h ffecs he cos srucure bu no demnd, z, nd finlly on n unobserved erm denoed by ω. In urn, ω nd 2 2 genered wih norml disribuion such h ω N( 0, σω) nd z N( 0, σ z) z re. For simpliciy i is ssumed h mrginl coss re consn (independen of oupu) nd log liner in he vecor of coss observed chrcerisics, X, z nd unobserved chrcerisics of he produc, expressed s ω. Mrginl coss of producing he differenied produc, mc, re hen mc = exp( γ k x k + γ zz + ω ) k nd genered by giving iniil vlues o he prmeers of he exogenous vribles, γ k nd γ z. 3..2 Generion of individuls nd heir choices: demnd For he demnd side of he mrke we genere populion of M individuls who derive n uiliy from consuming produc, U i, where i=,,m nd M=500000. Individuls re ssumed o hve n income s well s common nd idiosyncric ses for he ribues of he produc. Th is, individuls uiliy funcion is expressed s i( i, i,,, ) she/he chooses ny of he producs nd ( ) i0 U y v x p ξ if U y if chooses he ouside good. Recll he ouside lernive is n ggrege h incorpores ll oher lernives o which he consumer could lloce income. If his cse, he men uiliy U i0 is normlized o zero. i The specificion we employ o genere consumer s uiliy is he rndom coefficiens one. I ensures flexible nd relible subsiuion perns becuse i llows for inercions beween consumer nd produc chrcerisics. Indeed, i kes ino ccoun consumers idiosyncric ses which in urn llows cross price elsiciies o depend on consumer rel Foncel-Ivldi-Mois/ Februry 2007 3
Accurcy of Merger Simulion preferences for he ribues of he produc nd no only on he mrke shres of he producs in quesion. Individuls uiliy funcion is specified s: ( ) β ξ ε, U = α y p + x + + i i i k k ik i U, i0 = αi yi + εi0 where y i is income of individul i, p is he price of produc, x k is chrcerisic k of produc. Then, he uiliy of ech one of he chrcerisics of he produc is decomposed ino he idiosyncric nd he common ses, h is: βik = βk + σ ɶ kβik α = α+ɶ α. i i This specificion llows he coefficiens of he chrcerisics o depend on boh, mesured (common) nd unmesured individul (idiosyncric) chrcerisics β k nd ɶ βik, respecively. This would imply for exmple h fmilies wih high mesured incomes would be less responsive o price increses. Also, since here my be sources of welh h he nlys cnno mesure n differen degrees of price sensiiviy mong individuls wih he sme welh, we llowed for inercion of price wih n unmesured vrible, uiliy is genered wih αɶ i. Thus, ( α ɶ α )( ) ( β σ ɶ β ) ξ ε, U = + y p + x + + + i i i k k k k ik i ( α ɶ α ) Ui0 = + i yi + εi0, where ε i is componen, specific o boh, individul i nd produc, h couns for he error of he rue nd he pproximed uiliy funcion of individuls. The error erm is ssumed o be disribued independenly cross produc nd consumers nd o follow n exreme vlue disribuion such h he cumulive disribuion is ( ε) = exp( exp( ε) ) F. Foncel-Ivldi-Mois/ Februry 2007 4
Accurcy of Merger Simulion The frcion h ccouns for common ses, hereinfer denoed s θ ( β, α) =, is he vecor of prmeers (common o ll consumers) o be esimed. The frcion h ccouns for idiosyncric ses vi ( βik, αi) = ɶ ɶ is vecor h we genere ccording o norml disribuion. For his, we drw vecor of individuls income following log-norml 2 disribuion such h log ( yi) N( y, σ y) 2 2 ses by βik N( 0, σ k) for k=,,k nd αi N( 0, σα). We genere he disribuion of idiosyncric ɶ for i=, M nd respecive ideniy covrince mrix, consruced such h i is independen of he level of income Then, by giving iniil prmeers o β k, σ k, nd α, nd genering ll he rndom vribles concerning consumers preferences, ( y i, α i, k y i. ɶ βik nd ε i ) s well s he vribles defining he produc, ( x k, p nd ξ ), we buil up mximizion progrm of his uiliy funcion o genere vecor of consumer choices. Given ll he exogenous vribles nd n rbirry vecor of prices, his consiss in seing consumer i condiion s o chose produc if : ( i,,, ξ, εi; θ) ( i, l, l, ξl, ε il; θ) U v p x U v p x 3..3 Generion of supply Firms re ssumed o mximize profis wih respec o he price of he goods hey produce. In doing so, hey ke ino ccoun heir compeiors sregies s well s heir poenil demnd. Then, he profi funcion of he ech firm f =,... F producing he se of differenied producs, Ω f, mrginl cos, Π = mc (previously genered) is expressed s ( p mc ) Ms ( p, x, ξ; θ) f Ω f In order o se ech of heir prices, p, firms re ssumed o perfecly know he mrginl cos of produc, mc, he size of he ol mrke M nd n pproximion of he demnd for produc, h is, nd pproximion of he mrke shre of produc, s. Th is, s in BLP we ssume h, in order o deermine prices, producers hve some informion bou he populion s preferences nd re ble o esime n expeced demnd for heir produc. For his, firms re ssumed o perfecly know he componens enering consumers uiliy funcion including heir heerogeneous chrcerisics, i.e., idiosyncric ses ν i, Foncel-Ivldi-Mois/ Februry 2007 5
Accurcy of Merger Simulion income y i nd, ε i up o heir disribuion funcions. This mens h firms re ble cpure he heerogeneiy in he populion hrough he disribuion of consumer chrcerisics. In oher words, in order o se heir opiml prices, firms re fced wih he compuion of heir expeced demnd, which in he discree choice conex urns ou o be he probbiliy of produc being chosen by consumers. In our conex his probbiliy rnsles ino he expeced mrke shre. Then, knowing he densiy of consumers idiosyncric ribues, h we clled, P ( ν, y,ε), he firm producing good will compue expeced mrke shre for produc, denoed s, sɶ. This mrke shre will be funcion of he chrcerisics nd price of ll he goods compeing in he mrke. Then, firms re ssumed o know h here exis se of consumers choosing produc, A defined s {(,, ε) : (,,,, ξ, ε ; θ) (,,,, ξ, ε ; θ) } A = v y U v y x p U v y x p, l l l l l where A is he se of vlues for v h induces he choice of produc. Then, hey compue he mrke shre of heir produc s sɶ =Pr(consumer i choosing produc v, x,, p ; ξ θ ) = ɶ ( ξ θ) = s p, x, ; ( ) ( v, y, ) P dvdyd ε A ε Or = ( ) ( ) ( ) ( ) sɶ v, δ, p, x, θ, P f v, δ x, p, ξ, p, x, θ P dv i i which ccording o he specificion of he logi preferences rnsles ino (, δ,,, θ) sɶ v p x i J l= ( x p ) + ( x p v 2) exp δ,, ξ, θ µ i,, i, θ = + exp δ,, ξ, θ + µ i,, i, θ ( x p ) ( x p v 2) Given he mulidimensionliy of he inegrl we perform his compuion vi simulion. For his, we crry ou ns=00 rndom drws from he disribuion of P( ν, y, ε ) Foncel-Ivldi-Mois/ Februry 2007 6
Accurcy of Merger Simulion nd replce P by Pns. Giving iniil vlues of θ he vecor of simuled mrke shres h firms inegre in heir mximizion problem is sɶ vi p x Pns = f i vi p x ns = ns (, δ,,, θ, ) (, δ,,, θ) The demnd of produc firm f fces is hen he probbiliy of choosing good imes he number of consumers in he economy, h is, (,, ; ) Ms p x ξ θ. Then, he mximizion progrm of firms rnsles ino Π = ( p mc ) Ms ( p, x, ξ; θ) ɶ f Ω f Assuming h he Nsh equilibrium of he pricing gme exiss, nd h ll prices re sricly posiive, 2 nd expressing sɶ s s o simplify noion we hve h he J firs order condiions re derived ( p, x, ξ; θ) sl s ( p, x, ξ; θ) ( pl mcl) = 0, l=,..., J p l Ω f If is J by J mrix whose elemen (, l) is given by sl if nd l re produced by he sme firm p l = 0 oherwise in mrix noion he firs order condiions re expressed s ( ξ θ) ( ξ θ)[ ] s p, x, ; p, x, ; p mc = 0 2 Exisence of Nsh equilibrium; in muli-produc seing hs no been proven bu BLP mde numericl exercises o provide consisency of he esimes wih he exisence of equilibrium. Moreover, he properies of he esimes hey derive do no require h here be unique equilibrium ssocied wih ny given vlue of he prmeer vecor. Foncel-Ivldi-Mois/ Februry 2007 7
Accurcy of Merger Simulion or p= mc+ ( p, x, ξ ; θ) s( p, x, ξ; θ) ( ( ( ξ θ) ( ))) ξ θ ln p p, x, ; s p, x, ; = mc Wih his opimizion progrm of he firms we obin hen equilibrium prices nd quniies which in mrix noion re expressed s (,, ξ; θ) (,, ξ; θ), (,, ξ; θ),..., (,, ξ; θ) Ms p x = M s p x s2 p x sj p x Then, he pre-merger prices nd mrke shres re obined. 3..4. Merger Simulion Using he esimed mrginl coss nd he demnd s prmeers we cn now predic he posmerger equilibrium prices. The procedure consiss in chnging he ownership se of producs of firm f, i.e., is Ω f. For exmple if firm nd 2 merge hn, wh previously ws Ω nd Ω 2 fer merger becomes Ω, 2. A his sge, he ol number of producs, J, he chrcerisics of he producs, X k, X g nd ξ s well s mrginl coss mc re held consn. Then, i is ssumed h, before he merger, he merging firms se prices independenly, nd pos-merger hey inernlize he fc of oinly producing subsiue producs. This in urn, chnges heir mximizion problem ino new oin se of merging producs prices, sy nd 2. Then, for n increse in price of good, he merger firm nicipes h i cn compense he loss of consumers buying produc by gin of hose consumers who will swich o good 2. Since producs re subsiues hese gins (higher price of nd higher demnd for 2) ouweigh he loss of lower demnd for. The sme effec occurs for good 2, h is, when incresing he price of produc 2, he firm will ouweigh he loss of for 2 by n increse in demnd for good. Moreover, since in price compeiion prices re sregic complemens, whenever he price of compeior increses, he oher compeiors response will be o increse prices oo. As resul, in he pos-merger equilibrium, merging non-merging firms prices re higher. Foncel-Ivldi-Mois/ Februry 2007 8
Accurcy of Merger Simulion 3..5. D from he rue Economy Wih he previous genered d we obined pre-merger informion bou prices, quniies nd mrke shres wih which we cn compue he HHI nd is chnges. We denoe he premerger informion of he rue economy s p 0, s 0 nd HHI 0, s he vecors of premerger equilibrium prices, mrke shres nd he HHI, respecively. Afer he merger simulion we obin he equivlen informion, i.e., prices, mrke shres, HHI nd is chnges induced by he merger. We denoe his informion s p nd s, for he vecor of posmerger equilibrium prices nd mrke shres. By summing up he premerger mrke shres of he merging firms nd 2, sy, s 0 nd s 02, we compue wh we cll he ex-ne posmerger HHI nd denoe HHI 0,. This corresponds o he index h compeiion uhoriies do compue when fcing merger cse. In our rue economy we cn lso compue wh would be he rue pos-merger HHI for which ino ccoun he firms new mrke shres resuling from he mximizion progrm o se pos-merger prices. We denoe his erm by HHI. To summrize he supply side d includes: producs, ribues of producs, prices, coss of producion, mrke shres nd HHI. The consumer (demnd) side d includes: income, hypoheicl ses for producs (common nd idiosyncric) nd quniies bough. A his sge, we hve ll he necessry d from rue economy o perform srucurl economeric nlysis of merger unilerl effecs vi simulion s i is usully done by nlys. 3.2. Esimion of he Approximed of n Economy In order o esime he economy, we use he vilble d from he rue economy, which includes producs chrcerisics, prices nd mrke shres. To esime he demnd we use he discree choice mulilogi model. To pproxime he supply (o derive he pricing equions) we ssume h mrginl coss re liner funcion of he producs ribues. From hese wo sides of he economy we finlly observed prices, mrke shres nd HHI. We nex proceed o simule merger o predic he pos-merger equilibrium prices nd mrke shres. Foncel-Ivldi-Mois/ Februry 2007 9
Accurcy of Merger Simulion 3.2.. Demnd, Supply, Pre nd Pos-Merger Equilibrium To pproxime he demnd we use mulinomil logi model. We consider his specificion of demnd s he pproprie benchmrk becuse, s he HHI, i predics h he lrger he increse in mrke concenrion he lrger he price increse h cn be expeced. Moreover, if we prefer o be conservive in he predicion of price increses induced by he merger once gin he mulinomil logi model of demnd is he pproprie one. This ls hs been shown by Crooke e l. (999) who using Mone Crlo experimens for four demnd sysems (liner, log-liner, logi nd AIDS), show h he prediced price increse of merger is ceeris pribus is highes wih he log-liner demnd (i.e. consn elsiciy) follow by he AIDS. The prediced price increses re relively low when using he Logi demnd nd he lowes when using he liner demnd. Then, compring he Logi model wih he AIDS model we would expec hen n under-predicion of posmerger price increses. The mulinomil logi model demnd which is ssumes h he populion of he rue economy, he M individuls re sisiclly idenicl nd independenly disribued (iid), which mens h heir choices re governed by he sme probbiliy disribuion. The M individuls fce he sme choice se nd heir uiliy funcion is liner in chrcerisics, s: U = X β α p + ξ + ε, i i where X = x k re he observed nd ξ he unobserved (by he economericin) chrcerisics of produc nd k p is he price of he produc. Here, produc chrcerisics re reed s exogenous, lhough produc prices re deermined wihin he model. As explined before, ε i represens he disribuion of consumer preferences round he men uiliy nd i is ssumed o be ideniclly nd independenly disribued cross boh consumers nd producs (in his conex i ccouns for he individul specific deviion from he men). Noe h wih his specificion he prmeer α nd β re ssumed o be invrin cross consumers (lhough his is no necessry). Th is why his specificion is ofen resumed s U = δ + ε, i i where δ = X β α p + ξ ccouns for he men quliy of produc ; he erm ξ lone migh be ken s he men of consumers vluion of unobserved produc chrcerisics, Foncel-Ivldi-Mois/ Februry 2007 20
Accurcy of Merger Simulion i.e., quliy. Then, if εi follows he exreme vlue disribuion such h exp(-exp(-ε )), in he mulinomil logi model, he discree choice mrke shre funcion, s is derived from he principle h consumer i will purchse one uni of good if n only if for ll k 0 nd k, U i > U, he mrke shre of produc is given by ik s k= ( x β α p + ξ ) exp = J + exp + ( x β α p ξ ) k k k Normlizing he men uiliy of he ouside good o zero, δ0 0, he demnd of produc cn be expressed s ln s ln s = X β α p + ξ 0 where s 0 is he shre of he ouside good. From mulinomil logi pproch, own nd cross price elsiciies re : η l ( ) s pl α p s, if = l = = pls α pl sl oherwise The ls expression implies h he rio beween he choice probbiliies of nd l is independen of he res of he elemens included in se A h conins lernives nd l. This is he so clled Independence of Irrelevn Alernives propery (IIA herefer). I implies h, condiionl on mrke shres, subsiuion perns do no depend on he men uiliy genered from he produc δ x β α p + ξ, which in urn mens h subsiuion perns do no depend on he observble chrcerisics of he produc, x nor do hey depend on he unobserved chrcerisics of he produc, ξ. For he esimion of he supply side of he mrke s previously, we ssume h symmeric firms compee in prices nd h heir mrginl cos for producing ech differenied produc, mc, is liner in produc s chrcerisics Foncel-Ivldi-Mois/ Februry 2007 2
Accurcy of Merger Simulion ln ( mc ) = w γ + ω The fis order condiion of he Berrnd oligopolisic sregic behvior rnsles ino p s c = s p, since p nd s re known nd s / δ is obined from esimions of demnd. Then for he oin esimion of supply nd demnd, we subsiue he erm s / obin pricing equion expressed s δ, by s ( s ) o p = wγ + + ω, α ( s ) where inferred. p nd s re observed, α is esimed from he demnd equion nd finlly c is The sysem of logi demnd equions nd pricing equions is esimed by mens of non-liner hree sges les squres (NL3SLS) fer genering insrumens. The se of insrumens comprises he number of producs per cegory, he number of producs per firm nd per cegory, he number of producs per firm nd per ech of he discree vribles used in he nlysis. In oher words, we proceed in sndrd wy o esime he min prmeers of he model, nmely he mrginl uiliy of income, he prmeers h provide he mrginl effecs of ech ribues on quliy nd on mrginl coss. p = wγ + + ω α ( s ) We of course ccoun of he endogeneiy problem of prices nd mrke shres nd for hese we lso genere se of insrumens. These re: he number of ol producs in ech cegory, he number of producs ech firms produces wihin ech cegory, he men of chrcerisic k cross own-firm producs nd cross rivl-firms producs. Foncel-Ivldi-Mois/ Februry 2007 22
Accurcy of Merger Simulion 3.2.2 Simulion of Merger in he rue Economy To simule he merger, we proceed s before o compue he equilibrium of he economy, The procedure provides he pos-merger equilibrium prices nd mrke shres nd hence he HHI, we denoe his informion by p, s nd HHI. From hese d we cn hen predic he unilerl effecs of he merger, h is, he chnge in prices, nd we cn compue he chnge in he concenrion index, p nd HHI. We cn now proceed o he comprison of he rue nd pproximed economy. Th is, we confron p, HHI0, nd HHI wih p nd HHI. 4. Resuls of Comprisons Three min resuls cn be drwn from he sisicl nlysis ghered in Tbles o 4 nd Grphs o 6 below. The firs expeced resul is h he levels of pos merger HHI re bised upwrds when hey re compued ex ne, h is o sy, before he new pos merger equilibrium is obined. In our experimens, he verge pre merger vlue of he HHI, he HHI 0, is round 206 nd 208 for he cses of smll nd lrge mrke shre of he ouside good respecively. Is verge ex ne pos merger, he HHI 0, is equl o 2809 nd 2808, respecively. The verge rue pos merger HHI, h is o sy, he verge vlue of HHI compued pos merger, HHI is 2698 nd 2726, respecively. The bis is no lrge bu i is significn. Th is, we found h when producs re close subsiues nd here re no ny efficiency gins (refleced in lower mrginl coss), he merging firms find profible o increse heir prices. In urn, he non merging firms lso rise heir prices oo, bu in smller exen. I follows h he shre of he merging eniy is smller when one kes ino ccoun he unilerl effecs of he merger hn one kes ino ccoun he dominnce es. Th is, he pos merger HHI compued ex pos is smller hn he pos merger HHI compued ex ne, i.e., HHI < HHI0,. The bis is lso observed for he chnges in HHI before nd fer merger. The ex ne verge chnge in HHI in our experimens under smll mrke shre for he ouside good is equl o 793 while i is 682 for he rue chnge, i.e., HHI < HHI0,. Foncel-Ivldi-Mois/ Februry 2007 23
Accurcy of Merger Simulion I mens h when one is no ble o compue he pos merger equilibrium, he compuions of pos merger HHI nd chnge in HHI re bised upwrds, which increses he risk of incurring on ype II error, h is, he risk o prohibi merger lhough here is no serious compeiion concerns. The second resul is reled o he size of he mrke, or more precisely o he size of he mrke shre of he ouside good. In he pproximed economy, he verge esimed chnge in price, p, is 0.52 (4.79) percen while he rue chnge in price p is of 2 (.8) percen when he mrke shre for he ouside good is smll (lrge). In boh cses hese figures re higher hn he rue vlues mesured in he simuled economy. However, if we would hve ccouned ( i is usully done) for efficiency gins, resuls of he merger simulion exercise would hve been differen, in priculr, we would hve obined lower increse in price pos merger p. The predicions of mergers simulion in erms of price increse depends hen on he mrke shre of he ouside good. In cse of smll mrke shre of he ouside lernive (nd hus lrge relevn mrke), he merger would end be prohibi becuse of prediced lrge unilerl effecs. In cse of lrge mrke shre of he ouside good (nd hus smll relevn mrke), merger effec would ke he opposie decision. In oher words, he evluion of unilerl effecs by mens of simulion ool is very much influenced by he mrke size. No king ino ccoun for he fc h he mrke could be much lrger migh srongly bis upwrds he mesure of unilerl effecs. However i would be useful o evlue o wh exen his resul depends on he specificion of he simulion ool. 3 As we sid before, he HHI mesure is bsed only on he inside produc mrke shres. This implies h he HHI mesure cnno discrimine beween siuions corresponding o differen levels of he ouside good s mrke shre. This hs criicl implicion: he HHI cnno be good proxy for mesuring mrke power, i.e, for mesuring he biliy of firms o rise prices bove compeiive levels. The pos merger HHI cn be lrge even hough he merging firm hs lile mrke power over consumers. This cse occurs when cusomers re no very cpive, for insnce, when he mrke shre of he ouside good is lrge. 3 For exmple how much resuls would chnge if insed of using logi demnd we use n AIDS demnd or he Disnce Meric demnd. In hese models he shre of he ouside good is no iself problem, since wh is used here re he goods hemselves nd no heir mrke shres. However hese pproches re implusible s soon s he number of goods in he economy J is lrge (by lrge we men even 3). Foncel-Ivldi-Mois/ Februry 2007 24
Accurcy of Merger Simulion This is suppored by our hird min resul which invesiges for he correlion beween he chnge in prices nd he chnge in HHI. From Tbles 3 nd 4, one observes h here is no correlion beween hese wo mesures when he mrke size of he ouside good is lrge while hey re roughly correled hiry-five percen level when his size is smll. In oher words, when here is reson o ccoun for n ouside good, i.e, when is mrke shre is no negligible, he chnge in prices nd he chnges in HHI due o merger re independen, i.e., decision bsed on he chnge in HHI would hve nohing o do wih decision bsed on he chnge in prices. From he previous resul we conclude h simulion models re useful ool o complemen he HHI nlysis bu re by no mens subsiue for i. One of he dvnges of using simulion mehods o nlyze merger effecs, is h, while forml modeling mkes ll ssumpions explici, simulion llows us o idenify hose ssumpions which re criicl. Indeed, simulion models ough o be moived nd bsed on he circumsnces of he cse, so he ssumpions in he model correspond o observed feures of he mrke, nd model prmeers re bsed on d. 5. Discussion Firs of ll, i is imporn o hve in mind h merger simulion models do no give exc numbers of price increses bu rher give n indicion of priculr compeiive effec in mrke. Sill, even if models re by consrucion sic, simulion echniques provide resonble pproximion of he compeiive shor run dynmics of he mrke vi ssessmen of price increses induced by he merger Wih respec o he esimion of he demnd side, he implicions of IIA for subsiuion perns re, however, quie resricive. In he seing of produc differenied indusry, he IIA mens h he subsiuion perns beween one produc nd ll lernive producs re proporionl o heir respecive mrke shres. For exmple ny wo goods wih he sme mrke shres re consrined o hve he sme cross price elsiciy wih ny hird good, nd re lso consrined o hve he sme own price elsiciy. This propery will hold regrdless of wheher boh nd k re highly differenied. This propery is funcion of he iid ddiive error ssumpion nd no of ny specific disribuionl ssumpion (such s logi) on he errors. Despie hese resricions imposed by he IIA ssumpion, he muli-logi model sill provides he mos nurl deful ssumpion for subsiuion perns. In fc, ssuming Foncel-Ivldi-Mois/ Februry 2007 25
Accurcy of Merger Simulion subsiubiliy is funcion of he shre of he good is generlly seen s n ppeling deful ssumpion, even if he IIA propery does no define wh i mens for ll goods in he choice se o be eqully good subsiues for ech oher. The mulinomil-logi model employed o esime demnd in he pproximed economy migh no be he perfec one. The model consrins ll price elsiciies o depend only in he men uiliy of good, δ. Since here is one o one relionship beween δ nd he shres s, his implies h ny wo goods wih he sme shres re consrined o hve he sme cross price elsiciy wih ny hird good, nd re lso consrined o hve he sme own price elsiciy. imposing he IIA propery ssures h ll esimed elsiciies mke sense, i.e., h goods known o be subsiues cully hve posiive esimed cross elsiciies of demnd. Moreover, if one wns o be conservive in he predicion of pos merger price increses, oping for he logi demnd sysem is he righ decision. Using Mone Crlo experimens for four demnd sysems (liner, log-liner, logi nd AIDS), Crooke e l. (999) show h he prediced price increse of merger is ceeris pribus highes wih he logliner demnd (i.e. consn elsiciy) follow by he AIDs. The prediced price increse re relively low when using he Logi demnd nd he lowes when using he liner demnd. 6. Exensions Merger simulion models should nd cn be enriched nd enlrged by relxing criicl ssumpions. Nowdys i is possible o inroduce buyer power, poenil vericl srucure, oher dimensions of compeiion such s promoion, more flexible mrginl coss funcions, firms symmeries nd dynmics llowing for enry (pkes). Efficiencies In his sic nlysis efficiencies re usully modelled s fixed decreses in he consn mrginl coss. Efficiencies which ffec fixed coss do no mer since hese do no ffec he pricing decision. If mrginl coss where such h he merged firm hs incresing reurns, hen he possible scle effec should lso be compued oo. The inroducion of efficiencies in he simulion nlysis is limied o decrese in mrginl cos uniquely of he merging firm. If he merger generes more producs or processes which will benefi compeiors oo, we would like o lso qunify he effec. This ls is reled o he nex poin. Dynmics Unil now, he performed nlysis hs been sic becuse he chrcerisics of he goods nd heir cos funcions re held fixed. If dynmics re o be inroduced one hs o consider he impc of he merger on enry nd invesmen decision. Foncel-Ivldi-Mois/ Februry 2007 26
Accurcy of Merger Simulion Enry The evluion of he likelihood of enry involves however lo of difficulies such he possible brriers o enry, he lengh of ime i would ke o enrns o se hemselves in he mrke. Brriers o enry cn be of differen nure like echnologicl (lerning know-how, pens proecing exising firms), dminisrive (when governmen licenses re required o opere), finncil nd so on. The repuion or loyly buil by exising firms my lso represen brrier o enry since new enrns would hve o considerble inves in dverising o win he confidence of consumers. Reled o his ls issue re he swiching cos of consumers, h is, when cosly for consumers o subsiue from one produc o noher (becuse of disnce, nework effec) his could lso be brrier o enry. Buyer Power The scope of buyer power my pre-emp pos-merger price increses since lrge nd srong buyer my hre he merging firm of swiching o noher supplier. Th is, i is beer for he merging firm o fce lrge number of dispersed buyers h one or few buyers wih srong brgining power. 7. Conclusion We develop experimens in order o compre he wo differen ess of poenil merger effecs. For his, we genere n economy wih complee d bou prices, shres, nd consumer ses. Wih he genered d we esime n pproximed economy s i is usully done in merger cses. In boh, he rue nd he pproximed economy, we simule merger o obin he new equilibrium in prices nd mrke shres. In boh, he rue nd he pproximed economy, we perform he wo ess, he HHI nd he unilerl effecs es, o mesure he impc of he simuled merger. The firs expeced resul is h he levels of pos merger HHI re bised upwrd when hey re compued ex ne, h is o sy, before he new pos merger equilibrium is obined. This implies h h when one is no ble o compue he pos merger equilibrium, he compuions of pos merger HHI nd dels re bised upwrd, which increse he risk of ype I error, h is o sy, he risk o prohibi merger lhough here is no serious compeiion concerns. Forgeing o ccoun for he fc h he mrke could be much lrger migh srongly bis he mesure of unilerl effecs. The second resul is reled o he size of he mrke, or more precisely on he size of he mrke shre of he ouside good. This implies h forgeing o ccoun for he fc h he mrke could be much lrger migh srongly bis Foncel-Ivldi-Mois/ Februry 2007 27
Accurcy of Merger Simulion he mesure of unilerl effecs. Reled o his is he ls finding h revels h when he mrke shre of he ouside good is lrge, he chnge in prices nd he chnges in HHI induced by he merger re independen, i.e., decision bsed on he chnge in HHI would hve nohing o do wih decision bsed on he chnge in prices. The dvnges of he merger simulion pproch compred o he mrke concenrion one is h he former cn indeed offer n esime of he price effec induced by he merger. References Ackerberg, D., Benkrd, L., Berry, S., nd Pkes, A., (2005), Economeric Tools for Anlyzing Mrke Oucomes, vilble Anderson, S., De Plm, A., nd Thisse J., (992), Discree Choice Theory of Produc Differeniion, The MIT Press. Bker, J., nd Bresnhn, T., (985), The Gins from Merger or Collusion in Produc- Differenied Indusries, Journl of Indusril Economics, Vol. 33 Issue 4, pp. 427-444. Bker, J., nd Rubinfeld, D., (999), Empiricl Mehods Used in Anirus Liigion: A Review nd Criique, Journl of Americn Ls. & Economics Review, Vol. No. /2. Berry, S., nd Pkes, A., (993), Some Applicions nd Limiions of Recen Advnces in Empiricl Indusril Orgnizion: Merger Anlysis, AEA Ppers nd Proceedings, pp. 247-252. Berry, S., (994), Esiming Discree Choice Models of Produc Differeniion, Rnd Journl of Economics, 25, pp. 242-262. Berry, T., Levinsohn, J. nd Pkes A., (995), Auomobile Prices in Mrke Equilibrium, Economeric 63, pp. 84-90. Crooke, P., Froeb, L., nd Tschnz, S., (999), Simuling Merger Effecs Among Cpciy-Consrined Firms. Crooke, P., Froeb, L., Tschnz, S., nd Werden, G., (999c), The Effecs of Assumed Demnd on Simuled Pos-Merger Equilibri, Review of Inernionl Indusril Orgnizion 5, pp. 205-27. Deneckere, R., nd Dvidson, C., (985), Incenives o Form Coliions wih Berrnd Compeiion, RAND Journl of Economics, Vol. 6, pp. 473-86. Duso, T., D. Neven, J.H Roller (2003), The Poliicl Economy of Europen Merger Conrol: Evidence using Sock Mrke D, CEPR Discussion Ppers 3880. Epsein, J., nd Rubinfeld, D., (2004), Effecs of Mergers Involving Differenied Producs, Epsein, R. nd Rubinfeld, D., (Mrch 2004), Merger Simulion wih Brnd-Level Mrgin D: Exending PCAIDS wih Ness, Advnces in Economic Anlysis nd Policy, Vol. 4, No.. Epsein, R., nd Rubinfeld, D., (2002), Merger Simulion: A Simplified Approch wih new Applicions, Anirus Lw Journl, 69, pp. 883-99. Foncel-Ivldi-Mois/ Februry 2007 28
Accurcy of Merger Simulion Frrell, J., nd Shpiro, C., (990), "Horizonl Mergers: An Equilibrium Anlysis, Americn Economic Review, Vol. 80, pp. 07-26. Foncel, J., Ivldi, M., nd Rbss, V., (2004), Economic Anlysis of Mergers under he New Merger Europen Guidelines: Lessons from he Lgrdere-VUP Cse, unpublished pper vilble Froeb, L., nd Tschnz, S., (2000), How Much Informion is Required o Accurely Predic Merger Effecs, Working Pper. Froeb, L., nd Werden, G., (994), The Effec of Mergers in Differenied Produc Indusries: Logi Demnd nd Merger Policy, 0 Journl of Lw, Economics nd Orgnizion. Froeb, L., nd Werden, G., (996), Simulion s n Alernive o Srucurl Merger Policy in Differenied Producs Indusries, The Economics 0f he Anirus Process 77, Mlcolm B. Coe & Andrew N. Klei eds. Froeb, L., nd Werden, G., (998), The enry-inducing effecs of horizonl mergers: nd explnory nlysis, Journl of Indusril Economics, Vol.46. Froeb, L., Scheffmn, D., nd Werden, G., (999), Effecs of Assumed Demnd Form on Simuled Pos merger Equilibri, Review of Indusril Orgnizion, 5, pp. 205-27. Froeb, L., Scheffmn, D., nd Werden, G., (2004) A Duber Discipline for Merger Simulion, vilble hp://fc.gov/be/duberdiscipline.pdf. Froeb, L., Tschnz, S., nd Werden, G., (200), Pss-Through Res nd he Price Effecs of Mergers, Working Pper. Husmn, J., Leonrd, G., (997), Economic Anlysis of Differenied Producs Mergers Using Rel World D, George Mson Lw Revie,. Vol. 53, pp. 32-346. Husmn, J., Leonrd, G., (2002), The compeiive effecs of new produc inroducion: A cse Sudy, The Journl of Indusril Economics, Vol., No.3, pp. 237-263. Husmn, J., Leonrd, G., nd Zon, D., (994), Compeiive Anlysis wih Differenied Producs, Annles d économie e de Sissique, No. 34. Hosken, D., O Brien, D., Scheffmn, D., nd Vi, M., (2002), Demnd Sysem Esimion nd is Applicion o Horizonl Merger Anlysis, Federl Trde, Commission Bureu of Economics, Working Pper 246, vilble hp://fc.gov/be/workppers/wp246.pdf. Ivldi, M., Jullien, B., Rey, P., Sebrigh, P., nd Tirole, J., The Economics of Unilerl Effecs, vilble hp://europ.eu.in/comm/compeiion/mergers/review/he_economics_of_unilerl_eff ecs_en)pdf. Ivldi, M., nd McCullogh, (2005), Welfre Trdeoffs in U.S. Ril Mergers, IDEI Working Ppers 344, Insiu d'économie Indusrielle (IDEI), Toulouse, vilble IDEI Websie. Ivldi, M., nd Verboven, F., (2005), Qunifying he Effecs from Horizonl Mergers in Europen Compeiion Policy, Inernionl Journl of Indusril Orgnizion. Mriuzzo, T., Wlch, P., nd Wheln, C., (2004), EU Merger Conrol in Differenied Produc Indusries, CESIFO Working Pper. McFdden, D., nd Trin, K., (2000), Mixed MNL Models for Discree Response, Journl of Applied Economerics, Vol. 5, Is. 5, 447-440. Nevo, A., (2000), A Prciioner s Guide o Esimion of Rndom Coefficien Logi Demnd Models, Journl of Economics nd Mngemen sregy, 9, pp. 53-48. Nevo, A., (2000b), Mergers in Differenied Producs: he Cse of he Redy-o-E Cerel indusry, Rnd Journl of Economics, 3, pp. 395-42. Nevo, A., (200), Mesuring Mrke Power in he Redy o E Cerel Indusry, Economeric, 69, pp. 307-42. Foncel-Ivldi-Mois/ Februry 2007 29
Accurcy of Merger Simulion Pkes, A., (994), Dynmic Srucurl Models, Problems nd Prospecs, in J.J-Lffon nd C. Sims, Advnces in Economerics, Cmbridge Universiy Press: New York. Pinkse, J., nd Slde, M., (2004), Mergers, Brnd Compeiion, nd he Price of Pin, Europen Economic Review, 48, pp. 67-43. Rubinfeld, D. (2000), Mrke Definiion wih Differenied Producs: The Pos/Nbisco Cerel Merger, 68 Anirus Lw Journl, 63, pp. 73-76. Scheffmn, D., (2003) Sources of Informion nd Evidence in Merger Invesigions: A FTC Economis s View vilble hp://www.fc.gov/speeches/oher/sourcesofinfobrussels03.pdf. Scheffmn, D., nd Spiller, P., (987), Geogrphic Mrke Definiion Under he US Deprmen of Jusice Merger Guidelines, 30 Journl of Lw nd Economics. Shpiro, C., (996), Mergers wih Differenied Producs, Anirus, 0:2, pp. 23-30 (Spring). Suon, J., (99), Sunk Cos nd Mrke Srucure: Price Compeiion, Adverising nd he Evoluion of Concenrion, Cmbridge MA: MIT Press. Trin, K, (2003), Discree Choice Mehods wih Simulion, Cmbridge Universiy Press. Werden, G., (997), Simuling he effecs of differenied produc mergers: A Prcicl Alernive o Srucurl Merger Policy,, George Mson Lw Review, 5, pp. 563-86. Werden, G., (997), An Economic Perspecive on he Anlysis of Merger Efficiencies, Anirus. Foncel-Ivldi-Mois/ Februry 2007 30
Accurcy of Merger Simulion TABLE STATISTICS OF THE EXPERIMENTS (SMALL s 0 ) Men S.D. Min Mx Firm 's mrke shre 9.87.69 5.8 24.43 Firm 2's mrke shre 20.0.93 4.42 25.32 Merging eniy's mrke shre 39.87 2.23 34.5 46.4 Ouside opion's mrke shre 0.03 0.00 0.02 0.04 Aggrege elsiciy 0.00 0.00 0.00 0.00 Esimed verge chnge in prices 0.52.83 6.42 7.06 True verge chnge in prices.99 0.74 0.4 4.86 Bis on chnge in prices 494.74 236.4 20.0 779.08 Averge pre-merger HHI 206.20.33 2000.79 2067.37 Averge ex-ne pos-merger HHI 2809.56 89.00 2620.73 32.53 Averge rue pos-merger HHI 2698.77 65.67 2579.44 2923.8 Bis on HHI levels 4.4 3.58-5.85 4.53 Averge ex-ne chnge in HHI 793.36 89.49 584.50 074.0 Averge rue chnge in HHI 682.57 66.56 549.30 895.62 Bis on chnge in HHI 6.95 4.74-9.73 58.7 Foncel-Ivldi-Mois/ Februry 2007 3
Accurcy of Merger Simulion TABLE 2 STATISTICS OF THE EXPERIMENTS (LARGE s 0 ) Men S.D. Min Mx Firm 's mrke shre 9.85.84 4.9 26.06 Firm 2's mrke shre 9.93.85 4.60 24.29 Merging eniy's mrke shre 39.79 2.29 33.4 46.7 Ouside opion's mrke shre 68.29.20 64.65 72.09 Aggrege elsiciy.54 0.05.38.67 Esimed verge chnge in prices 2.79 0.48.76 4.0 True verge chnge in prices 0.80 0.5 0.0 2.72 Bis on chnge in prices 84.35 2248.0 7.24 26286.68 Averge pre-merger HHI 208.59 3.27 2000.75 209.06 Averge ex-ne pos-merger HHI 2808.56 89.54 2608.67 336.29 Averge rue pos-merger HHI 2726.68 79.42 2574.98 2990.68 Bis on HHI levels 3.06 3.77-4.72 6.04 Averge ex-ne chnge in HHI 789.97 9.35 557. 076.28 Averge rue chnge in HHI 708.09 80.86 548.49 96.68 Bis on chnge in HHI 2.53 5.6-7.3 74.28 Foncel-Ivldi-Mois/ Februry 2007 32
Accurcy of Merger Simulion TABLE 3 CORRELATION STUDY ON THE EXPERIMENTS (SMALL s 0 ) HHI 0,.00 HHI 0, HHI p p HHI 0.99 (0.0).00 p 0.7 (0.0) 0.5 (0.03).00 p 0.35 (0.00) 0.34 (0.00) 0.54 (0.00).00 HHI is he verge ex-ne chnge in HHI,,0 esimed nd he rue verge chnge in prices, respecively. HHI is he verge ex-ne pos-merger HHI, p nd p re he TABLE 4 CORRELATION STUDY ON THE EXPERIMENTS (LARGE s 0 ) HHI 0,.00 HHI 0, HHI p p HHI 0.98 (0.00).00 p 0.0 (0.4) 0.09 (0.8).00 p 0.06 (0.39) 0.05 (0.44) 0.4 (0.04).00 HHI is he verge ex-ne chnge in HHI,,0 esimed nd he rue verge chnge in prices, respecively. HHI is he verge ex-ne pos-merger HHI, p nd p re he Foncel-Ivldi-Mois/ Februry 2007 33
Accurcy of Merger Simulion GRAPH 2 HISTOGRAM OF LEVELS OF HHI (SMALL s 0 ) 30 25 True pos-merger HHI Ex-ne pos-merger HHI 20 Frequency 5 0 5 0 2400 < - 2450 2450 < - 2500 2500 < - 2550 2550 < - 2600 2600 < - 2650 2650 < - 2700 2700 < - 2750 2750 < - 2800 2800 < - 2850 2850 < - 2900 2900 < - 2950 2950 < - 3000 3000 < - 3050 3050 < - 300 Inervls of level of HHI GRAPH 2 HISTOGRAM OF LEVELS OF HHI (smll s 0 ) 30 25 True pos-merger HHI Ex-ne pos-merger HHI 20 Frequency 5 0 5 0 2400 < - 2450 2450 < - 2500 2500 < - 2550 2550 < - 2600 2600 < - 2650 2650 < - 2700 2700 < - 2750 2750 < - 2800 2800 < - 2850 2850 < - 2900 2900 < - 2950 2950 < - 3000 3000 < - 3050 3050 < - 300 Inervls of level of HHI Foncel-Ivldi-Mois/ Februry 2007 34
Accurcy of Merger Simulion GRAPH 3 HISTOGRAM OF LEVELS OF CHANGES IN HHI (SMALL s 0 ) 30 25 True chnge in HHI Ex ne chnge in HHI 20 Frequency 5 0 5 0 490 < - 540 540 < - 590 590 < - 640 640 < - 690 690 < - 740 740 < - 790 790 < - 840 840 < - 890 890 < - 940 940 < - 990 990 < - 040 Inervls of chnge in HHI GRAPH 4 HISTOGRAM OF LEVELS OF CHANGES IN HHI (LARGE s 0 ) 30 25 True chnge in HHI Ex ne chnge in HHI 20 Frequency 5 0 5 0 490 < - 540 540 < - 590 590 < - 640 640 < - 690 690 < - 740 740 < - 790 790 < - 840 840 < - 890 890 < - 940 940 < - 990 990 < - 040 Inervls of chnge in HHI Foncel-Ivldi-Mois/ Februry 2007 35
Accurcy of Merger Simulion GRAPH 5 HISTOGRAM OF CHANGES IN PRICES (SMALL s 0 ) 60 50 True Chnge in Prices Esimed Chnge in Prices 40 Frequency 30 20 0 0 0 - < - 2 2 < - 3 3 < - 4 4 < - 5 5 < - 6 6 < - 7 7 < - 8 8 < - 9 9 < - 0 0 < - < - 2 2 < - 3 3 < - 4 4 < - 5 5 < - 6 6 < - 7 7 < - 8 8 < - 9 9 < - 20 Inervls of chnge in prices GRAPH 6 HISTOGRAM OF CHANGES IN PRICES (LARGE s 0 ) 35 30 True Chnge in Prices Esimed Chnge in Prices 25 Frequency 20 5 0 5 0 0-0.4 0.4 < - 0.82 0.82 < -.23.23 < -.64.64 < - 2.05 2.05 < - 2.46 2.46 < - 2.87 2.87 < - 3.28 3.28 < - 3.69 3.69 < - 4.0 Inervls of chnge in prices Foncel-Ivldi-Mois/ Februry 2007 36
Accurcy of Merger Simulion Digrm : Presenion of he esing procedure True economy Genered d on individuls nd producs HHI0 p0 Equilibrium in prices nd mrke shres HHI0 p0 Esimion of he pproximed economy HHI0, Merger Simulion of he merger HHI p Pos merger equilibrium in prices nd mrke shres HHI p Pos merger equilibrium in prices nd mrke shres HHI 0, p0 HHI p p HHI p p Tess nd Comprisons Foncel-Ivldi-Mois/ Februry 2007 37
Digrm 2: Noions nd Componens of Experimens TRUE ECONOMY FIRMS CONSUMERS. Generion of : - F firms - J producs (including ouside good) - The k ribues of he producs, X k - Mrginl coss of producion, mc 0 - M consumers 2 Equilibrium is deermined. Produce good wih ribues Xk 2. mrginl cos mc in funcion on Xk 3. Compue expeced demnd sɶ o: 0 4. Mximize profis π nd f 0 5. Se up he price of he good 0 p 0 - Uiliy of consumer i of buying produc U i 0 -Income of consumer i y i -Common Preferences β -Idiosyncric Preferences σ, v, ε p, s 0 0. Observe he se of producs =0,, J 2. Observe p nd Xk 0 3. Choose he produc h mximizes U i 4. Aggrege over ll consumers o deermine mrke shres for ech produc, 0 3 Merger Occurs: Pos-Merger Equilibrium is deermined. F- firms in he economy 2. Merging nd non-merging firms relloce producion by mximizing π f 3. Merging firm hs new 4. Ses new prices p mc 4 True Merger Effecs: Avilble informion: Pre merger HHI 0 Ex ne Pos merger 0 HHI, (s usully compued ) Ex pos Pos merger HHI HHI (righly compued). Fce new se of choices 2. Observe new prices p s p, s 3. Choose he good h mximizes U i 4. Aggregion over ll individuls deermines demnd s ESTIMATED ECONOMY FIRMS CONSUMERS 5 Esimion of Demnd nd Supply Using d from he rue economy 6 Pre Merger Equilibrium is esimed. Compeiion in prices sregy 2. Prmeers of cos funcion γ k 3. Equilibrium prices p 0 p, s 0 0. Discree choice Demnd 2. Prmeers of Demnd ˆβ 3. Equilibrium quniies nd mrke shres s 0 7 Merger Simuled: Pos-Merger Equilibrium esimed. F- firms in he economy 2. Merging nd non-merging firms relloce producion 3. merging firm hs new 4. se new prices p mc 5. Fce new se of choices 6. Observe new prices p p, s 7. Choose he good h mximizes 8. Aggregion over ll individuls deermine demnd: 8 Esimed Merger Effecs s U HHI Esimed Esimed Increse in price p
Mergers s Aucions Increse in prices p nd dels HHI 0 (rue) HHI (s usully compued) - 39 -,