De Onde Vem Os Mercados

7/30/2019 De Onde Vem Os Mercados

1/32

Where Do Markets Come From?Author(s): Harrison C. WhiteReviewed work(s):Source: American Journal of Sociology, Vol. 87, No. 3 (Nov., 1981), pp. 517-547Published by: The University of Chicago PressStable URL: http://www.jstor.org/stable/2778933 .

Accessed: 13/03/2013 13:36

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp

.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact [email protected].

.

The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to

American Journal of Sociology.

http://www.jstor.org

This content downloaded on Wed, 13 Mar 2013 13:36:02 PMAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/action/showPublisher?publisherCode=ucpresshttp://www.jstor.org/stable/2778933?origin=JSTOR-pdfhttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2778933?origin=JSTOR-pdfhttp://www.jstor.org/action/showPublisher?publisherCode=ucpress


2/32

Where Do Markets Come From?'HarrisonC. WhiteHarvardUniversity

Productionmarkets ave two sides: producers re a fully onnectedclique transacting ithbuyers s a separatebut aggregated lique.Each producers a distinctive irmwith distinctive roduct. achside continuallymonitors eactions fthe other hroughhemediumof a joint social construction,he schedule fterms ftrade. Eachproducers guided n choiceofvolumeby thetangible utcomes fother producers-not by speculation n hypothetical eactions fbuyers o itsactions. ach producerctspurely nself-interestasedon observed ctions of all others, ummarized hrough feedbackprocess.The summarys the terms-of-tradechedule,which educesto constant rice only n limiting ases. The market merges s astructurefroleswitha differentiatedichefor ach firm. xplicitformulae-both orfirms nd formarket ggregates-areobtainedby comparative-staticsethods or ne family fassumptionsboutcost structuresnd aboutbuyers' valuations fdifferentiatedrod-ucts.Not just anysetoffirms ansustain erms ftradewith nysetofbuyers. hereprove o be threemainkindsofmarkets,nd threesorts fmarket ailure, ithin parameter pacethat s specifiedndetail.Onesort fmarketPARADOX) has a MadisonAvenue lavor,anothers more onventionalGRIND), and a third CROWDED)is a newform ot ncluded n any existing heory fmarkets. urrentAmerican ndustrialmarkets re drawn on for 20 illustrations,fwhich hree representednsomedetail. nequalitynfirms'marketshares measured yGinicoefficients)s discussed.

Whydo somany four ndustrialmarkets avebuta dozenor somemberfirms, everalof whichproducesubstantial haresof the total output(Scherer 970;Porter 980)? t is not noughocite echnologicalonstraints.Why,when venthe argest ffirms ants o offer product ewto itto thepublic,does it usuallydo so by acquiring hepersonaof a firmIFinancial supportundergrants OC76-24394,SER76-17502, and SES80-08658 from heNational ScienceFoundation s gratefullycknowledged,s is technical ssistance fromHollyGrano. E. Raymond Corey,MichaelE. Porter, nd others t the Harvard BusinessSchoolgenerouslymade available teachingnotes and case studiesof ndustries r setsoffirms romndustries; heybear no responsibilityor he nterpretations propose. n thisdevelopment f earlier echnicalpapers,Ronald L. Breiger,Robert G. Eccles, Eric Leifer,JohnF. Padgett,OrlandoPatterson, nd Arthur tinchcombe ontributeddeas and otheraid, as did an anonymousreferee.Requests for reprints hould be sent to HarrisonC.White, Departmentof Sociology,WilliamJamesHall 470, Harvard University, am-bridge,Massachusetts02138.? 1981 by The University f Chicago. All rights eserved0002-9602/82/8703-0001$01.50

AJS Volume87 Number 517

http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp


3/32

American ournalf Sociologybelongingo an existingmarket? his astonishing act seems to be over-lookedbyexisting heory; ractitionersnd consultantsake t for ranted(see Salter nd Weinhold 979,pp. 7-16; Steiner 975,pp. 192,200).Whydo economistsccepta theory ffirmsnmarketswhich enies nprinciple hemostcommonly bserved ituation ffirms?Most industrialfirms ost fthetimedecideproductionolumewithin rangewhere heirunit osts re constant rdecrease s volume ncreases see jiriand Simon1977, . 7).I developed ometentative nswers o such questions n an earlier ech-nicalpaperonproduction arkets s induced ole tructuresWhite1981).HereI wish oflesh uttheargumentnd illustratet byapplication o anumber fcurrent .S. markets.WhatI haveproposed s embeddingconomists' eoclassical heory fthefirm ithin sociological iewofmarkets.Markets reself-reproducingsocial tructuresmong pecific liques ffirmsnd other ctorswho volverolesfrom bservationsfeachother's ehavior. argue hatthekey factis thatproducerswatch each otherwithin market.WithinweeksafterRogerBannister roke he four-minuteile, thersweredoing o becausethey efinedealitiesnd rewards ywatching hat ther producers" id,notbyguessingndspeculatingnwhatthecrowdswanted rthe udgessaid.Markets renotdefined y a setofbuyers,s someofourhabitsofspeechsuggest, or are theproducers bsessedwithspeculations n anamorphousemand. insist hatwhata firm oes n a market s towatchthecompetitionnterms fobservables.In myproposal,markets re social structuresnwhich roducersepro-duce their wn etofactions; he set confirmss correctach firm'sxpec-tations fwhat thopedwas an optimal olume.Thisview s a special aseof "rational xpectations" Muth 1961; fora recent urvey, ee Kantor[1979]). n thisfeedbackmodel here s also a self-selectionspectderivedfrom he"signaling heory" fAkerlof1976), Spence 1974), and othersintrigued ynotions f"imperfectnformation."A modest eneralizationfthenotion fprice s required: eneralizedoa market chedule fobserved utcomes. hese observed utcomes re asetofpairs,onepairfor achfirm: evenue eceived orvolume hipped.Look at thehypotheticalutcome ets nthepanelsoffigure . Figure1Aoutcomes annot ustain market, igure B outcomes an. Figure1C,inwhich curve s interpolatedhroughhepoints ffigure B,is onewayafirmmayvisualize hoseoutcomes, evenueW) as a functionfvolume(y). This schedulemustnotbe confounded iththe demandfunction,hypotheticalonstructfeconomists. his schedulemaybe perceivednterms fprice revenue/unit),uitt is a volume-dependentrice.As willbe seen below,thisgeneralizationf the conventional otionofprice scrucial ormyapproach.518



4/32

WhereDo MarketsCome From?wIRevenue Revenue Revenue

W(y)

Volume Volume y,VolumeA. B C

FIG. 1.-Outcomes foreach firmn a clique. A, Outcomesdo notsustaina market.B, Outcomesdo sustain a market.C, Market schedule nferred rom ; decisionsusing tsustain t.Because it frames he onlyhard,tangible vidence vailable,each firmtreatsW(y) as itsownopportunityet.Obtainingventhatmulchnforma-tionrequires lert nquiry-over uncheonswithothersn thetrade,fromtrade ssociations,rom ne's owncustomers,nd so on. Each firm nows

that tsproducts distinctive,ut it also knows hedifficultynd riskofassessing ne'sowndistinctivenesssee Corey1978;Porter 976). In par-ticular,when he totalvolume ne offersnthemarket hanges,tsattrac-tiveness obuyers hanges,nwayshardto estimate.No firm anreliablyassessrelativeualities fother irms,ndevery irmnows hat tspositioncould be affectedychoicesmadeby anyoneor moreof ts competitors.The market cheduleW(y) is a shared social constructionncorporatingall the nteractionffects.Such mutually nterlockingonfirmationsf a unique choiceby eachproducerfig. B) are notpossible or ny arbitraryollectionfproducersandanypopulationfbuyers. variety f attributes aydistinguishndi-vidualproducer irmsproduct uality, ocation, lant nvestment).n theonehand, omeofthese ttributesffectost structurese.g., figs. and 7below)and thustheproductionevel which ooksoptimum o that firm.On the otherhand,someof these attributesnfluence ttractivenessobuyers.A self-confirmingarket chedule,W(y), is inducedhereforacliqueofproductswhose ost structuresnd taste structuresan each beranked y quality.

A market s an "act" which an be "got together"nlybya setofpro-ducers ompatibly rrayed n thequalitieswhich onsumersee in them.Quality rrays averecently igurednmicroeconomicsnder herubricsf"hedonic rices" Terleckyj 976)and "consumers producer"Lancaster1966,1979).Nearly20years goAlonso 1964) proposed n array oroca-tions n a city,wheredistance rom hecenter rovides naturalmetric519



5/32

American ournal fSociologyfor quality ut also geometriconstraintsn areas,hisanalogue fvolumes;more ecently osen 1974) hasgeneralizedhis chematizationo mperfectcompetition.Building market s a conflict-riddennd erratic rocesswithquite arangeofoutcomes ossible n the form fmarket chedules.The variousfirms' roductswillpresumably e akin within market, ut we do not asobserversmpose ny a priori ultural r linguisticriterion f similarity.Markets re defined yself-reproducingliquesof firms,nd not the otherwayaround.Thebodyofthispaper s dividednto section n thegeneralmodel nda section n applicationsnd results.THE GENERAL MODELFirms nthemarket ifferromne another, ot only n coststructure utalso in appreciation f theirproductsby the buyers.These dispersionsoccupy entertage nmy nalysis.n this espect followhe ong-standingtradition f economic tudies of "imperfect ompetition" nitiatedbyChamberlainseeDixit andStiglitz 977; Spence 1976). But thattraditionhas firms sing conjecturesn buyertasteto decide theirmarket ffers.In contrast,myview,presentedbove, s thatfirmsecideon thebasisofobserved ositions fall other roducers.In myview,firms eek niches n a market n much the sameway asorganismseekniches n an ecology.Becauseeachfirms distinctive,heyareengaged ot npure ompetitionut nfindingndsustainingoleswithrespect oone another iven nenvironmentfdiscerninguyers. utthereis no auctioneero shapethemarket; nstead,tsstructureepends n theinterlockingf local orders.This leads to the postulatethat firmswithneighboringost schedules amountofvariable cost to producevariousvolumes)mustalso have, n theeyesofbuyers, eighboringchedules fvaluation withrespectto volume. f the postulate s not satisfied, henascentmarket ituation, setofproducerswith n attendant opulationofbuyers ttracted y them, annot ustain tself:W(y)willnot be repro-ducedthroughhe self-interestedctions ffirms,hecked y buyers.In an observedmarket, heproducing irmsre dispersedn qualityofproduct erceived y buyers s wellas in volumeproduced.By the pos-tulate of the preceding aragraph,neighboringualitiesmust lead toneighboringolumes fproduction.n themodel each firms assignedvalue on an indexofquality, enotedby n; thevalue assigned s charac-teristic:n attributewhich annotbe changed uickly s can the volumeofproduction.he volumeofproductions denoted y y.The key feature fmy approach s this:Firmscan observenlyvolumesand payments,otqualities rtheir aluations,nd they ct on thebasisof520



6/32

WhereDo MarketsCome From?these bservations,herebyeproducinghe bservations.y model, owever,canpredict ll thesevolume hoicesbydifferentirms ecause t assumesknowledge fquality, ,for achfirm,s well s valuations.Higher alueson this ndex are defined s higher uality-always as judged by buyers'evaluations.)First developthecost andvaluation chedules, nd then derive herangeofmarket chedules hatmayresuiltrom givenset offacts.Atopology fmarkets-a two-dimensionalrraywith ach point particularshape ofpayment chedule-follows s a by-product. he varieties ndimplications fmarket ailures re thendiscussedbefore hesecondpartofthispaper: applicationsndpredictionsorndustrialmarkets.The central heme roves o be a trade-offetween ispersions,n affairofvariances ather han thematter fmeansone might xpectfrom hecliche' fsupply qualingdemand.The FactsofCost and ValueacrossVolume ndQualityTheprimitivesfmydescriptionretwo chedulesof ost ndofvaluation),eachgivennterms ftwodimensionsvolumeyandqualityn). Ifthefactsabout a setoffirms nd buyers annotbe approximatednterms fnestedschedules,2hoseproducers nd customerswill not be able to sustainamarketwhich eproducestself s infigure . In order o achieve compre-hensiv~eetclear nventory,shall pecialize oparticular amiliesfpowerfunctionsodescribe ossible chedules.The schedule fthefirmharacterized y n for ariable ostsofproduc-tion s C(y; n) = qyc/nd,with andc positive (1)The firm'sontributiono buyer aluation s defined sS(y; n) = ryanb,withr, a, and b positive (2)Intuitivelyt is clear hata balancemustbe struck y themarket etweenthe trade-offakingplace between ontributionnd cost,with respect omorevolumeor lessvolume, n theone hand,and the trade-offetweendesirability nd expensewithrespect o quality,on the other. Equations2 Nestedmeanssimply hat thefunction escribing scheduledefined y one value oftheparameter e.g., n) never crosses thefunction oranothervalue of theparameter. Theconcept, hough lippery,sfamiliar nd importantnother reas, suchas Lorenzdistribu-tions; see Schwartz and Winship 1980].) This assumption, s well as the more basicassumption hat each empirical chedule an be represented ya perfectlyharpfunction(see n. 10 below), clearly s morestringenthannecessary o obtain the main findings.But to simplifyxplanation of the theory n these first apers, make not only theseassumptionsbut also thefurther ssumptionthatparticularCobb-Douglas (power-law)forms re appropriate orthe schedule.Elsewhere have explored lternative unctionalforms.

521



7/32

American ournalfSociology(1) and (2) yield the simplest amily f schedules hat allowthese fourindependentariations f schedule s measuredn total dollars nd withrespect ither ophysical olumey ofproductionr to its qualityn.Figure2 schematizeshe phenomenologynderlyinghese chedules.texplicates quation 1) as the variable cost of producing he volume ychosenby a firm eingtheproduct fthevolume-sensitiveost nd thequality-sensitivexpense.Cost must increasewithvolumeso that theexponent is positive;thisexponent ivesthe proportionateogarithmicrateof ncreasewithvolume.Quality, nlikevolume, s in theeye of thebeholder, ere the buyers, nd therefores an exogenous social fact"confrontingheproducers;o theexponent canbe either ositive r nega-tive. Whenpositive,d describeswhat I denotebelow as a PARADOXmarketnwhich producer hose roducts likedbetter indst ess costlyto make!There s in figure a parallel ationale or quation 2). By defini-tion, heexponent for esirabilitys positive; ne couldeven nsist hat tbe set equal to unity, ut I prefer ot to constrainhe scaling f qualityvaluesn fordifferentirms' roductsn thatway (see disctssion f tables1 and2 below).On thebuyer idetheresan asymmetryhichsnotcapturednfigure .Firms re the ctivedecisionmakers; ach hasan independentost chedule

i) Phenomenologyf Market ontextDispersions across Firmson Volumey) of Firms Production on Quality ndex (n)Sc Contributionf the product Desirabilityof the producth Valuation increases withvolume, s increaseswithqualityase perceivedbythe buyers judged bythe buyers

u Cost of productionncreases Expenseof building n theI Costs as volume increases quality changes(+or-)e with ncreased qualitys Increases WithVolume Changes WithQualityii) Parameters--Proportionatelog) Rates

a br | diii) The Basic Tradeoffs

--Over variation, n productvolume Contributin= a/cCost - c--Over variationin producers'quality Desirability = b/dExpense

FIG. 2522



8/32

WhereDo MarketsCome From?known o itself at least). Buyers, n theotherhand, relumped ogetheras an aggregate,n a passive role. The aggregate uyermay say "no" tothemarket ntryvolume nd pricepair) offeredya firm,ut t has onlythisbinary hoice. This binarydecisiondependson how the buyers naggregate valuatemore fonefirm'sroduct gainst essofanother's. osimplifyhemodel urther,assume hatbuyers onotseeparticular airsofproducts n interaction,ut onlythe set of products. n formal erms,thebuyersnaggregate alue their otalarray fpurchases s

V(#)= [2 S(y; n)]" . (3)The symbolV for hisvaluation s inboldface ypeto signal hat t is anaggregate uantity ver the wholemarket: adhereto thisconventionthroughout. he symbol# gives thenumber f firmsn themarket ndhencethe number fterms n thesummation. bserve hatthe contribu-tion fonefirm'sroduct olume o thetotalvaluation anreplace hecon-tribution f anyother irm.n thespecial case in which he exponent isunity, urtherolumencrementsrom given irmre neithermorehighlyvalued (as theywouldbewith > 1) nor ess highly alued thana begin-ningvolume.Moredecisive s the overallexponenty (gamma) which susually ess thanunity; hiscorrespondso a saturation ftaste,forthesumofpurchases rom ll firmsnthemarket, y the given ggregationfbuyers s theaggregate olume urchased ncreases.3The scale factors for ostscheduleseq. [1])andry or aluation ched-ules eqq. [2]and [3])are worth eeping istinct.n the section napplica-tions, hangesn thesewill be interpreteds exogenoushiftsn costanddemandmposed nthemarket. donotdefine distributionffirmsverthequality ndexn,onlytheirnumber.n thisview ofmarkets here rerelatively ewmember irmss producers,s few s a halfdozen, nd thereis no reason o suppose ny particularpacing n thequality ndex mongthosefew:These arebest treated s a set of constants o be fittedo theobserved roducersas is done nseveral mpiricalxamples ater).Certain pecial asesbear considerable eightnthepastdevelopmentftheories fmarkets, point developat length lsewhereWhite 1980).Setting he exponent to zero reduces he market o pure competition:products fdifferentirmsre ndistinguishable.erfectompetition odelsassume naddition hat there s no taste for heerdiversity n thepart ofI In this implifiedepresentationeq. [3]) of aggregate uyer valuations crossproducts,I followrecent nnovations fmicroeconomicheoristse.g., Spence 1976; Rosen 1974;and see Lancaster 1966).Note thatdemandsfordifferentroducts nteractn this repre-sentation nlybecause yis not unity or hatmarket. ach S function eports ropensitiestoward valuationbased on thatproduct's ttributes, utonly n thecontext ftheotherproducts an it be incorporated nto an actual valuation (V) with monetary imension.Observethatthecontribution,t themargin, o V from ny particular irms itsS times,Y/V(I-,Y)/-Y

523



9/32

American ournal fSociologybuyers-inmy terms heexponent is set at unity. t is in thishighlyspecialized ontext hat the notionof supplyequalingdemandbecameprominent.elowwe seethat, lthough quilibrationfaggregatemountssought nd sold scertainlyecessary,t s a secondaryutcome f hemainissue,which s the terms n which ggregationan be carried utmeaning-fullywithdiverse roducts. upply nddemanddenoteone aspectofthefeedback rocesseswhich hapemarkets,ut notthebehaviorallyelevantoneor,therefore,hetheoreticallyelevant ne.MotivationsndEquilibriumchedulesEach firm,haracterizedy tsqualityn,chooses tsvolume fproduction,y(n),tomaximizets cashflow,whichs theexcessofmarket eceipts verout-of-pocketosts. (In this followneoclassic rthodoxynd disregardthebehavioristrgumentor ognitivendmotivationalimits ooptimiza-tion.) n equilibriumheframe f hoices sedmustbeexactly heobservedmarket chedule efinednfigure: theW(y)ofmarket eceipts or olumeshipped.4t is the same frame or achfirm, venthoughwe as observersknow usthow irmsiffernquality ndthusncost ndvaluationchedules.How can thisbe? It can be only feach firm, ecauseofits owncoststructure,s ledbythe amecommoncheduleW(y)to choose distinctivenichey(n)of tsown,which iche urthermoreatisfiesheaggregateuyerthat t constitutess gooda buyas anyother roducers offering.his ismydefinitionf a market,which shallnowformalizendextend.The buyers n aggregatere able to imposeonlya necessary ondition(eq. [4])for heir etbenefitobemaximized.ydefinition,heir etbenefitis their otalevaluation fpurchases,he Vofequation 3), lessthesumofthepaymentsmadeto allfi irms,hesum ofW[y(n)] vereachfirm. heI A scenariocan help as illustration. uppose some particular irm, ay Medusa (actualname of a firm n the market-for-cementllustrationater) thinks t can commandmorethanthis chedule. upposeMedusa picks particular olume,yo, t which thopes to geta revenueA, which s larger han the schedule mountW(yo)and optimizesMedusa's netreturn, iven ts hypothesized chedule yingabove the W(y) drawnthrough bservedmarket ntries. uppose yo s larger hany(m),wherey(m) is the (unique) volumewhichoptimizesMedusa's net return iventheexistingW(y) and yet keepsMedusa's offeringsin line as no more or less attractive o buyersthan the otherfirms' fferings.f volumeyo s actually sold, less thanthe volumes ndicatedfrompreviousmarketresults ortheother irms illactuallybe soldby them; o their ptimization omputations refalsified,and theymust cramble oreadjustby lowering rices nd/orvolumes n search of a self-confirmingchedule.But theneven theW(yo) evel scornedby Medusa maynotbe sus-tained n the owered chedule ventuating rom hescramble, o that Medusa's hopeforrevenueA neednot be validated, nd in anycase it willperceive changed situation nda recomputationfoptimalvolume.More likely,not all the ntendedyowillbe sold, sincethe pricewas above theprice ustainablefrom he market chedule, t y(m),and indeedat too higha pricenone at all mightbe sold, stimulatingach of the others o expandabit to fill perceivedgap. In either vent,theW(y) market chedulewill be unstable.Ithereforeuppress uch transienteadjustmentsndfocusonlyonwhat mutualconfigura-tionsofvolumes nd prices mongfirmsan sustain themselvesn equilibrium.524



10/32

WhereDo MarketsComeFrom?necessaryonditionsthat ach firm's (n)yield he ameratio fcontribu-tionto V overpaymentW[y(n)]. ncorporatehisratio nto0 (theta) seen. 3): S[y(n); n] = OW[y(n)]. (4)The firmndexed y n need notand doesnot know his ontributionunc-tion, , ofcourse, reven ts own ndex alue n; inequations2)-(4) we arestipulating ossiblefactual ituations roundthemarketn order o seewhathappensgiven hewayfirmsehave.There s nomarketmechanismwhereby uyerscan coordinatensistence n that particular alue of 0whichwouldmaximizeheir ggregate et benefit.

It is straightforwardomputationWhite1981, qq. [8]-[15]) ofind nyscheduleW(y)which an sustain tself nder hepressuresndthroughhechoices fthese ctors nthetwosidesofthe market. he resultsW(y) = (Ay(bc+ad)lb + k)b/(b+d), (5)

subject o twoauxiliaryufficiencyonditions.The constant oefficientis specified elow neq. [11].) Each producermustby tschoiceofvolumesatisfy quation 4), and thisdetermines (n) in terms f the W(y) ofequation 5) (as specified elow n the dissectionfaggregateupply nddemand).But everyproducer's wngoal is to choosethatvolumewhichmaximizes ashflow; hisfactyields hetwoauxiliaryonditions.First, he hapeofthe chedule nequation 5) guaranteesnly hat achproducer irm an find distinctive ichethatyields t an extremumfcashflow; o,toberealistic,he hapemust atisfy second-orderonditionformaximization,hich educes White1981, qq. [16]-[18]) od Ay(bc+ad)Ib> >da k (6)

for given olumey.Second, producer irm illnotremainnthemarketunless tsoptimalniche at which ashflow,W - C, ismaximized)nfactyields positive ash flow o put toward verhead ostsandprofits. hissecond nequality an be reduced o a form arallel o thefirstnein (6):Ay(bc+ad)Ib > -k . (7)

Thepresencef n arbitraryonstant,notfixednterms fother aram-eters nd rates nthesystem,s thefirstftwokeyfeaturesftheformula(5) forthemarket quilibrium chedule.We see that the two auxiliaryconditions6) and (7) can be taken as possiblyimitinghevolumes they values)andthus heunderlyingixed ttributesthen's) offirmseekingniches n thismarket.Or, in someparameter ettings,heremaybe novolumes nd thusno quality ttributes, values,and thusno firmshatcanyield market quilibrium ith omevaluesofk.Qualitatively, canbeseen osetthe ocation f hecurve escribedyW(y),positive eading525



11/32

American ournal fSociologyto higherpayoff orgiven volumeto the delighted yes of producers.Historicaldiosyncrasys onewayto think fthe formationfk-as thealmostaccidentalby-product f a numberof producers ockeying orvolume ndpayment ustainable n thepresence ftheotherproducers'offerings.t is also clearthata firman, by cuttingtsasking ricefor hesamevolume, chieve-if t has anymotive o-a loweringfthe chedule,ineffect change fvalueofk inthe viable equilibriumchedule. A firmwhich riesto raise ts chargefora givenvolumeabove the establishedterms-of-tradechedule olding t thattimemaysimply ind tselfwithnosalesat all: sharpdisciplinendeed seen. 4 above].)

The secondoutstanding spect of the formforequilibriummarketscheduleW(y) nequation5) isthemixednature f tsexponent s a (shifted)powerfunctionfvolumey. The powerfunctionorm ollows, fcourse,from uruseoffamiliesfpower unctionsodescribehefacts othofcostschedules nd of valuationcontributionseqq. [1]-[3]). The interestingquestions how thisoverall xponent,fy neglecting ,namely,exponent= -b c + bd a, (8)

represents balance between rade-offsn theunderlyingimensionsfvolume nd ofquality n the costs and valuations.The twotrade-offsfinterestre (as mentionedbove) the ratioa/c forvolumedependencies("contribution"/"cost")nd theratiob/dforqualityvariations "desir-ability"/"expense").Thisviewof ndustrialmarketsnsists n andpointsup the social com-ponentwhichsheavilynterlaced ith echnicalndengineeringacts ndwithbuyer eeds. nsteadofa uniqueequilibrationfsupply nddemand,or devious chemingsy speculatorsnpsychologynd taste,we see a his-torically haped tructurefroles mong stablesetofproducer irmsr,ifyou prefer,n adaptableecology fniches mong crowd fcompetingorganisms.n eithermetaphor he actors are making ffectiveecisionson thebasisoftangible bservationsfthe actions ftheir onfreres.hemarket s a publicfeedbackmechanismor rading ff ivergencesmongfirmsndbetween hem ndbuyers.A Topology fMarkets, ndTheirVulnerabilitiesWe canproject n a two-dimensionalcreen, igure , the mutualdistanc-ingofsorts fmarkets rom neanother,ince heprecedingection howsthat the two ratios /c and b/dsufficeo identify uantitative xes ofdifferentiation.ut to identifymarket esults nd failures crossregionson this creen ccuratelywemustbe explicit bouthowthe setof9firmsin a marketoalesce nto tablemarketggregations.here s an additional526



12/32

WhereDo MarketsCome From?feedbackoopimplicitn eqs. (4) and (5), onevulnerableo disruptionnthe terms f nequalities6) and (7).Figure3 has no lefthalf, incevariationswithvolume a and c) mustboth be positive.Figure3 has both a top half,withd > 0, and a bottomhalf,withd negative. ositive occurswhen orrelationfproductionostwith uality crossfirmss negative,ndconverselyor negative.n thebottom alf,newnotationmustbe introducedo avoidconfusion:

for d < 0: a = -d. (9)In marketsdentifiedy points nthe bottom alf ffigure , theproducerwhosegoods re ikedbetter ybuyerss aying utmorencosts fproduc-tionfor givenvolumeofgoods.As one might xpect ntuitively,hereproves obe a richerrray fmarketypes-and onessustainablecrosswiderrangeof contexts-than s trueof markets rom heupperhalfoffigure."Decreasingreturns o scale" characterizemarketsocatedto the leftofthe ine nfigure for /c = 1 (or ustc < 1 inmany exts). Increasingreturnso scale" are supposed,n economic heory Ijiri and Simon1977,

Decreasing returns Increasingreturnsto scale to scaleb

FailedPARADOX r ionCosts down asquality upd is positive Pure Competition

0 ~~~~~~~~ 1~~~~~a/cCosts up as -\ GRINDqualityupd is negative- Failure due to Failure due tofreeloading CROWDED XPLOSIVEg h

a/c=FIG. 3.-Trade-offs n costversusvaluation, crossgrowth n quality ordinate) nd involume abscissa).

527



13/32

American ournal fSociologyp. 7),to be inconsistentith ustainablemarkets. ut n myviewofmarketstructure,uchmarketsreviable as long s theexponent is negative,othatcost shigher orhigher-qualityirms.Three orts fmarket ailuresanbe distinguishedn thismodel.The only factualcontexts bsolutelynhospitableo markets, s self-supportingocial structures,re those dentified iththeupper right nfigure: increasingeturnso scale andquality orrelate egativelyd > 0inmy onvention,q. [1])with ostofproduction.n suchcontexts, heresnowhere o "shift"the trial market chedule, hat is, no value of theconstant in equation 5) forW(y)whichwillpermit atisfyingothcon-straints 6) and (7). Anypositivek yieldsa schedulewhichpermits is-tinctivenichesonly forproducerswho are masochists ttempting ominimize heir ashflows.Any negative alueof k so depresses hetermsof trade o producershatnofirmsre able to achievepositive ash flows.5Allother actual ontexts,ll other artsoffigure, should t firstightpermit t leastsomeequilibriumchedules. here s usually omevalue ofkwhich atisfiesothconstraints6) and (7).6However, heseotherwiseiableranges fmarket chedulesmayundersomecircumstancesncounterneof twoforms f nstability.irst, free-loading" anunravel market chedulewhich t first ight sviable,givenfirms ith whole ange fqualities. n formalerms,freeloading"esultsfromhewrong indofoverlap etween olume egionsllowed yoneandbythe other fthe nequalityonstraints6) and (7). Return o thepointin thePARADOX region; or nypositive value, range fyexistswhichdoesnotyieldoptimalnichesfor ndividual irms etdoesyield positivecashflows. hen t is easyto showthatall firms ith uality ndicesn intheappropriateangewouldchoose ome"edge"valueofvolume s their("corner")optimum mong possiblechoices;yet theywouldnot yieldvalueformoney obuyers: heratio /Wwouldbe lessthan (seeeq. [4]).Thus that partofthe schedule vervolume,W(y),willnot be sustained;but the ameproblemwillreemergewith somewhat ifferentet offirms,from shiftedangeon thequality ndexn) for heabbreviatedchedule,and in the endit willunravel ompletely.or anyvalue ofk, positiveornegative, he ower-leftrapezoidal egionnfigure , a region fcontextswhichyield propermarkets y existingmarket heory Spence 1976), is6 Innumerable pecific cenarioscould show how a particular et of firms nd buyersdescribableby a point n the upperrightwould blowup any trial schedule.Parsimonyrequires tating, s in the text, ust theallowedstructuresnd the staticcriteriawhichdiscriminatehem;these re given n moredetail n White 1976, 1981). But examplesofscenariosmay helpthe reader's ntuition: ee n. 4 above and thecase studiesbelow.6 In White 1981,tableA), I report, or achdetailed ubregion fthetrade-offslane,theexact rangeof k valueswhich lways yieldstable payment chedules.This rangecan belargerfnotall possible evelsof qualitynare presentn the market; o theres interactionbetween angeofk and rangeof n (worked ut in detail in White 1979],pp. 54-57).528



14/32

WhereDo MarketsComeFrom?vulnerableofreeloading.hat is, ffirmsfall possible uality evelsarein thewings eeking ossible ntrance o thatmarket, orno one ofthosecontexts an anyschedule t all (anykvalue) yield sustainablemarket.7Second, here re constraintsn viability ecauseofpossibleexplosivefeedback ffectsn the processby whichaggregatemarket ize builds.Feedback naggregationan be schematizeds

W[y(n)I]W(y) -,y(n) - W(y) (10)V(/,y(n)'s)The ratio ties marketchedule acedby ndividual irmsotheaggregatesize ofmarket.This is seen from he specificationf the constantA informula5) forW(y):

(b -)c/\1blbc-aA qr (11)This s a classicfeedbackndeterminacynwhich heproductionolumesdependontheheight fthepayment chedule,which oyieldequilibriummustdepend ust so on the actualproduction olumes hosenby all the

# irms.hearbitraryhiftonstant ,as well s theratio as shown bove,enter hedeterminationf quilibriumolumes (n)andpaymentsW[y(n)],so that ngeneral numericalolutions required. he specialcase ofk =0 yieldsguidancentheform fexplicit ormulaeor ggregateizes,whichyieldmarketoefficientsand A. TheseformulaeinWhite 981) show hattheaggregateashvolume fthemarket, amely,W, is unboundedwhenc < ay (12)

so a/c = 1/y s marked n figure by a dotted ine since t appliesonlyfork = 0) and theregion o theright s marked XPLOSIVE.8 (In theupperhalf fthefigure,ord > 0, unbounded rowth faggregate olumebegins t a/c = 1, the traditional emarcation f "increasing eturns oscale" remarkedn above.)In the nextpartofthepaper,numerical xamples eportingggregate7To put it positively, nd in moreconventional erms, market an be sustained at apoint in thisregiononlyif there re highbarriers o entry assumingthe initialset offirms eta self-sustainablechedule nmotion).Such "metastable"markets,n this ower-leftregion f thefactualtrade-offslotted n fig.3, shouldbe sought;testingpresumesnot onlymeasuresof a, b, c, and d but also following putativeexampleover timeforevidence ofbarriers. arriers an be ofmanykinds: egal,capital (sunkcostoffacilities,establishing istribution,tc.) (Scherer1970; Porter1980).8 In words:demand s suchrelative ocoststhateachproducer an keepraisinghischoiceofvolume and perhapsalso newfirms nter)and yetfind hat themarket ustainsthepricehe envisioned.My comparative taticsmodel cannot,of course,capturethis un-bounded scalation s a process, rthebounding ffects hichmust et n.Newproducts-whether olor TV in 1965 orhula hoopsin the ate 1950s-are plausible cases.

529



15/32

American ournal fSociologysizes and marketdistributions ill llustrate ow viable markets n oneregion ffigure differrom hose fanother. ne final eneral istinctionshouldbe drawnhere.The lower egionmarkedCROWDED, just short fthe EXPLOSIVE region, as "increasing eturns o scale." Such marketsare viable because buyershavenomeans to realizeor organize o exploittherealization hat they re better erved hefewer he firms here re.9Under theseconditions,he aggregate ize of the market,W, actuallydecreaseswhen newfirm nters nd finds niche n themarket chedule.APPLICATIONSI report hree lassesof applications. irst, illustraten detaileach of thethree iable market egionswith model f a current .S. industrialmar-ket.'0 Second, locate a wide varietyof empiricalmarketswithinmytopology. inally, illustrate he concentrationmplicationsf my model,via predictionfGinicoefficients.Threespecificmarkets-those or ement, ight ircraft,nd disposablediapers-providemy llustrationsfthe three iable market egions.1. Cement. mpirical ources or his nd theother wo llustrationsregiven n table3 below. Eightfirms ominate he current .S. market orcement onsidereds a national ne:#= 8. Theirnames are attached nfigure to thecostscheduleswhich assignto each. (The correspondingvaluation chedules,heS's ofmymodel, renot hown.) irmswithhighercost schedules lsohavehigher alue schedules,o thatd is negative.Mybest estimate s that the cementmarket s described y a = .8, b = 1,c = 1,d =-2 (8 2), and y= .7 (withr andq eachsetto unity).Thisputsthecementmarketnthemiddle fthetrianglenfigure justbelowthepure-competitionine segment.At the top offigure is one viableequilibriummarket chedule heavy line). It has k = 0, which n thisGRIND region s the lowestvalue ofk not vulnerable o freeloading."19When n fact buyers rehighly r totally oncentrated-e.g., n purchaseof automobilewindshieldwipers-and the rangesoftrade-offsor hisregion pply, onewould expect tto be impossible oform market forfurther iscussion ee White 1976,1978]).10Cost and valuefunctions,nd indexvaluesforn, are chosen to give thebest appearingfit for available data, within the constraints f the functional orms ssumed earlier.These are illustrationsather hanfully eveloped mplementationsrom nd testsof theW(y) model. Full implementation ould requiretests of goodnessof fitof the powerfunctionsorms o costand other chedules.Full implementations likely lso to requireintroducingtochastic fuzz" terms nto all theschedules.Grossman 1975) has sketcheda Bayesian implementationf anotherkind ofrationalexpectations heory f markets,in whichhe emphasizes tochastic ermsbut eaves producers ndifferentiated);n sec. 4Grossmanpecifiesne simplifiedstimation rameworkorCobb-Douglasformsike eqq.(1) and (2) above."Given the rangeofqualityvalues n which report orthem,noneof this set ofeightfirmswould be in the temptation egionforfreeloading. ut who knows what obscurelocal producersmight e watching rom hewings?530



16/32

WhereDo MarketsComeFrom?$ Lone Star Industry o>< Lowquality,awIdeal Basic Industry x - price firmenters:

Gifford-Hill X New equilibrium W(y)General ortland-*.-.12 - MartinMarietta-aMedusa- - Cash flowsMarquette >*UniveralAtlasCement

Cost Schedules C(y;n)06- / Cement MarketGRIND REGION)/\ Initial a=.8; b=1, c=1, d=-2, r=1; q=1, y= 7Equilibrium k=OW(y) eo=978

05 09 yFIG. 4.-Market schedulebefore nd after ntry fan additionalfirmntothecementmarket.

Arbitrarily,haveusedthatvalue of0 whichmerely reaks venfor hebuyer ide.12Just elow heheavy ine nfigure is a secondmarketchedule,heonewhich ouldresult f a brandnewfirm ntered t low qualityand highvolume, s shown this s a hypothetical,nnamed irm). he changes uitcommonense: eachother irmoses somevolume, s wellas bydefinitionsomemarkethare, ndloses tin sales as wellas inphysical olume-yettheaggregateize ofthemarket,ncludinghehypotheticalinth irm,sincreased. s a test funderstanding,hereader hould stablishwhathap-pensto theprices harged y the various xisting irmssee lastfigurenWhite 1981]).The companion igure , forthe same industry nd the same initialequilibriumchedule, howswhatthenewviablemarket chedule ouldbecome fgeneral ost evel q for ement roducerswentup by 20%. Inaddition, similar omparative-staticshift fschedule howswhatwouldhappen fconsumeremand roseby 10%. (In each casethenew cheduleremainst thebreak-evenvalue for he new ontextndkremainsero.)13

12 This is the0,, rt = 1 case usedfor alibration nWhite 1981), eqq. (22)-(27). A ratio-nale is thatcementbuyers re unlikely o be greatlynterestedn doingoptimallywellon their urchases,ust on getting noughof theright tuff.13 In these llustrations, irmswereselected and quality ndexvalues (the n's) were as-signedso that an orderly erms f trade schedulewould result productionvolumes ormarketshares of revenuewere the only given data, in addition to discussionsof theindustries romwhich a and c were derived). In the absence of complete data, theseunfalsifiableelectiveprocesses re unavoidable.Once the firmsnd qualitvindexvalueshave beenfixed n thebasis ofavailabledata, however, redicted ffects f"exogenous"changes embodied n r and q) can provide n opportunity orfalsificationf thetheory.

531



17/32

American ournalf Sociology$ / CementMarket 8 firms

(GRINDRegion =O,b=c=1; d=-2;-0.7).20-- k=0 t=1

InitialEquilibriumSchedule W(y)\ NewSchedule afterconsumer emandincrease f10% (in r).10New Schedule fter20 % increase inproductionostlevel q)

.05 .10VolumeyFIG. 5.-Impact on market cheduleof exogenous hiftsn cost level (q) and demandlevel (r).

In thismarket rom he GRIND region in sharpcontrast o someshownbelow)the shiftnconsumeremandmakesvery ittle ifference,ither naggregater ndivergencesmong irms. ach oftheeight irmss shown sa dot oneachmarketchedule,o that ne seesnotonly he hiftnscheduleperceived ut also thenet shift, fterfeedback,n volume chosenby agiven firm rom he shifted cheduleof perceived ffers. ote that theincreasendemand, ia thefeedback,ecreasesachfirm'solume ndprice.2. Light ircraft.he second etailedllustrations more ounterintuitiveas a market. choose locationnfigure inthe CROWDED region; heactual ndustryshalluse as an approximations themanufacturingndus-try or ight ircraftsedfor ecreationallying.here arebut three irmscurrentlyn thismarketCessna,Piper, ndBeech),theminimal umberforthetheory o makesense but also a veryconvenient umber orex-hibitingindings.Table 1 reports ourdifferenthanges f equilibriummarket chedule,and foreach of twovalues of the shift onstantk (in each case at thebreak-even forthatcontext).Because there are but threefirms, canreport ll these chedulesn one tableofthenumerical alues. Cash flow(W lesscost)andprice re ncluded s wellas theschedulesW(y).There s a changen nterpretationhich introduceustfor his xample.532



18/32

WhereDo MarketsComeFrom?In all equilibriumchedulesn table 1, thephysical olumes fproductionfor he three irms re keptfixed, nd fixed t the evels ctually bservedin the light-aircraftarket n 1969.So a changed chedule s foundbycomputinghechanged ualityndexn for ach firm eededtoyield eachthe same volume s before. ut it is just thequality ndexn that s to bethoughtf s an attributeftheproducer hichsfixed t least nthenearterm.Hencetable1 shouldbe thoughtfas the searchfor hesetofcon-

TABLE 1MARKET SCHEDULES AND QUALITY LEVELS WHICH YIELD OBSERVED PRODUCTIONVOLUMES: LIGHT-AIRCRAFT INDUSTRY, VARIOUS VALUES OFSHIFT CONSTANT kAND OF COST PLUS DEMAND LEVELS

k=.5 ~~~~~~~k=lProportion ProportionCash Cashy n W Flow* Price y n W Flow* Price

Context: q =1, r = 1, y .71 .613 .4403 .145 .440 1 .487 .3874 .388 .3871.6 .538 .6633 .364 .414 1.6 .388 .5265 .584 .3291.9 .503 .7576 .442 .399 1.9 .350 .5777 .645 .304Oo=1.63 Oo1.54

Context:q=1.2, r=1, -y=.71 .508 .3771 .315 .377 1 .372 .3069 .549 .3071.6 .446 .5800 .501 .362 1.6 .296 .4377 .708 .2741.9 .417 .6684 .566 .352 1.9 .268 .4889 .755 .257Oo=1.64 Oo=1.68

Context: q=1, r=1.2, -y=.71 .612 .4389 .146 .439 1 .489 .3895 .386 .3891.6 .537 .6614 .365 .413 1.6 .390 .5286 .582 .3301.9 .502 .7557 .443 .398 1.9 .352 .5795 .643 .305Oo= 1.96 0o=1.84

Context: q=1, r= 1, y .91 .926 .7827 (-.095) .783 1 .466 .3666 .408 .3661.6 .812 1.061 .095 .663 1.6 .371 .5043 .602 .3151.9 .759 1.163 .172 .612 1.9 .335 .5556 .662 .292Oo=1.08 Oo=1.61

Context:q=1, r=1, -y=.7 (NewFirmAdded,Third n Size)1 .552 .3726 .181 .373 1 .399 .2979 466 .2981.3 .518 .4775 .306 .367 1.3 .355 .3677 .578 .2821.6 .485 .5739 .404 .359 1.6 .318 .4273 .655 .2671.9 .453 .6618 .482 .348 1.9 .287 .4783 .712 .252Oo=1.81 oo=1.58NOTE.-Throughout; = 1 = b, c = .8, d = -2. Go s the break-even alue for , with hegivenkvalue(seeWhite 1981], qq. [20]-[25] or etails).* {W[y(n)]-{C(y(n);n] /lWy(n)I.

533



19/32

American ournal fSociologystants and parameter alues whichyield the best fit to the currentlyobservedmarket;predictionsf changes n volumewould be made onlylater.In the CROWDED region ffigure , market chedules an becomeestablished nly fthey re shiftedpward nd have a positive alueoftheshift onstantk. Regardless fother hifts n constants nd parameters,for argerk value,producers flowerqualityare sufficiento obtain theobserved olume evelsy. Thesevolume evels are associatedwithBeech(smallest, = 1), Cessna y = 1.6) andPiper y = 1.9) inascending rderofsize. At the higherk value theseproducers lso generate owerprices(and thereforeales) in all contexts, ut cash flows lways are a sharplyhigher roportionfsales.Table 1 shows also that thesemarket utcomes re quite insensitive,after he feedback ffectsaveplayed themselves ut,to changes n levelr of demand.But outcomes re extremelyensitive o changes n the"saturation" ate,theparametery. Gammacloser o 1 (less saturation)may ead,for CROWDED regionmarket ike this, omuchhigher alesvolumes;but cashflows o theproducersre sharply educed n spiteofhigher rices.'4

3. Diapers. A thirdmarketsmodeled nthedisposable-diaperndustry,a very arge nd quitenew one which lso nowhas butthree roducers,npart becauseofthehuge capitalcostsoftheequipment eededto makediapers nan efficientcale 400 per minute). his ndustrylearly elongsin the paradoxical ategory avored y Chamberlain, hich nmy frame-work s the upper-leftegionn figure : the better-likediaper actuallyhas smaller ariable ostsperunit nd thushigher-volumeroductionhanthe ess ikedbrands. igure reportsneequilibrium arketchedule orthiscase (pluscostschedules), arallel o figure for he cementndustryof eightfirms.f, analogous o figure , we drew a changed quilibriumschedule stillat k = 0 and at break-even ) after n increaseof 20%occurrednthe evelofproductionosts givenby q), this chedulewouldbe shifted ownward niformlyy just under25%0.A paralleldemandincrease r up by 20% and a correspondingew break-evenalueof00=1.497 at k = 0) leads,after eedback, o a magnifiednd quite uniformincrease fnearly 0%0nsales;the atter esultsunlike hat n figure fortheGRIND regionmarket, hichwasinsensitiveo r.As a parallel o table1 for he ircraftndustrynthe CROWDED regionoffigure , I reportn table2 also thechanges n qualitywhichmustbeattributedo thethreeproducersnorder o maintain heir olumesy attherelativeevelsobservedn themarket, orvarious hanges n assump-tions boutcost evels ndso on. n contrastothe CROWDED market f14 Yet with-ycloser to unity, n orderto match the observedvolumes, ne mustassumeextremely igh-quality roducers highvaluesfor hethree irms n thenindex). However,this same saturationvalue -ycan lead,when k is large,to lowered ales and higher rofit.534



20/32

WhereDo MarketsCome From?$ A1.50- DISPOSABLE DIAPER MARKET

(Paradox Region, a=1, b=1, c=1.2; d=15)Proctor8 Gamble bl

C(y;1.2)

Johnson8.60- Johnson)\ C(y;1.1)KimberlyClark

C(y,)0 1.00 175Volumey

FIG. 6.-Production volume decisions n thedisposable-diapermarket, t equilibrium.Equilibrium cheduleW(y): (k = 0; -y= 0.7; Oo 1.313).table1,herethe increase n cost evelrequires hathigher ualitybe at-tributedo each producernorder opredict he same volumesy,and theprices nd sales volumes ctually re increased s a result, s is thecashflow.Again ncontrast o table 1,but as infigure for GRIND market,theresvery ittle ensitivityo changesnthedemand evelr.LocatingU.S. IndustrialMarketsntheTrade-OffslaneIn figure (which epeats he topographyffig. )15an abbreviated ameindicates heestimatedocationnthe trade-oflslaneofeachofa scoreofcurrent omesticmarkets. ncluded are the three (cement, irplanes,diapers) ust analyzed.The primaryourceused s a business choolcasestudy serieswithacronym CCH.16 Table 3 gives ICCH identificationcodesas well s other ources.15 No dotted line at a/c = l/-y is shown because -y (crudely:a measure of aggregateelasticity fdemand) maybe differentor ach industry.My impressionsre that,ofthe20 cases in fig. 7, on carefulmeasurement nly color TV sets 1965,optical fiber,Mini-computer 980,and injectorsmight rovetohave a/c beyond thecorrespondingalue of1/-yi.e., n the EXPLOSIVE region f fig. ); and only the first womight ave valuesof-ygreater hanunity.16 The ICCH seriesused arebuta smallfraction fall Intercollegiate ase ClearingHousecase studies Graduate SchoolofBusiness,HarvardUniversity): hese reeither eliberatestudiesofa whole ndustrialmarket r studiesof ndividualfirmswhich an be combinedto yield good coverageofa givenmarketby my definition.

535



21/32

American ournal fSociologyPlacement fan industryn figure is an exercisen coding, hat s,inqualitativeudgment n cost structuresnd onvaluations.17Occasionally,as for hefarm quipmentndustry,ystematicumerical ata is given orthe quantities fconcerno that numericalstimatesf parametersan bemade.'8Other ocations re closetoguesses, uch as those hown ormain-frame omputersnd minicomputers.'9

TABLE 2MARKET SCHEDULES AND QUALITY LEVELS WHICHYIELD THE PRODUCTION VOLUMES (y) OBSERVEDIN THE DISPOSABLE-DIAPER INDUSTRY UNDER

VARIOUS CONTEXTS, WITH SHIFT CONSTANTk= -40Proportiony n W CashFlow Price

Context: =1, r= 1, ey=.730 3.33 10.38 .058 .35100 3.67 39.75 .098 .40296 4.00 128.5 .100 .43Oo=9.2

Context:q=1.2, r=1, =.730 3.61 11.07 .219 .37100 3.98 42.11 .249 .42296 4.34 136.1 .250 .45Oo=9.44

Context:q= 1, r = 1.2, y=.730 3.32 10.38 .058 .35100 3.66 39.73 .098 .40296 4.00 128.4 .100 .4300=11.05

Context: =1, r= 1, -y=.930 1.80 27.19 .094 .91100 1.98 100.2 .100 1.00296 2.16 323.6 .100 1.09O=1. 975

NOTE.-Throughout= 1=b; c = 1.2;d = 1.5.17 SiX cases were codedindependentlyy Eric Leifer. n each of these the codings greedon region. n the absence of systematicnumericaldata, however, xact agreement nlocation s virtuallympossible.18 I.e., thefarm quipment ase studyquotes theproportionateercentage ecreaseof costwith ncrease npercentage f notional)production apacitywhich s utilized.Dependingon the particular omponentreported, he exponent c lies between0.7 and 0.9. (SeeICCH-9-280-080 Rev. 2/801, able 5, p. 6.)19 he threellustrative ases presentedbove rangefrom uiteexact codingusingnumeri-cal data (disposable diapers) to more qualitative estimatesextrapolatedacross firms(cement).536



22/32

TABLE 3SOURCES AND NOTES FOR TWENTY INDUSTRIES

Name (in Fig. 7) andICCH Identification* Notes and Other ourcesDrillingmud:9-380-167................. ... Market n vibrators o cleansepollutants9-380-168................. ... from ariteVacuumtubes:

1-379-184Rev. 8/79 } Classic declining ndustryWatches:6-373-080................. Rev. 10/721-374-050................. Rev. 10/7511-374-051................. Rev. 10/759-373-090................. Rev. 9/76Baby foods1965:1-379-185. Rev. 8/79 j Unusually clearcutcase of PARADOXLog houses:1-378-195 ..|.....Firms whichrationalize ustomized1-379-(196-201) . houses3-378-193.................Circuitbreakers:9-513-152M230) .......... Rev. 10/75'9-513-151 M229) .......... Rev. 10/759-565-004................. Rev. 1/799-567-005 AM-P204) ...... ...9-578-205................. Rev. 10/78jDisposable diapers:9-380-175................. SmithBarneyHarrisUpham ResearchReports1979: variousdatestOil tankers:9-379-086 ................ ... See text.Distribution lectricalcomponents:5-379-146.... .. ...... . ...1-377-063.................5-379-146. Service irmshat tock ens f housands3-778-153.............. f..... . parts9-377-041.... .......... Rev. 1/79 o types o prs9-380-084................. ...9-377-055................. ...Strode'scables:5-377-028................. ... Market in metal-ceramiccable con-9-376-188................. Rev. 11/77k nectors, ncludes Strode division of9-377-027............. ... J EG&G, Inc.Cement1970s.... .......... ... BusinessWeek1980Sugar refining895. ......... ... Eichner 1969, chaps. 2, 3Injectorsplus electronic uelinjection EFI):9-378-219................. ... Injectorsare a component fEFI, and1-378-257................. ... a separate market;cases emphasizerolestructuren thesemarketsLight aircraft:9-369-007 BP 934R)........ ...9-369-008.... . .......... Rev. 1970 Low-cost, ecreational lanemarket9-370-036....... ......... ...

* Studies vailablefor feefromntercollegiatease ClearingHouse,Soldiers ieldRoad, Boston,Mass.02163. Each mayrunfrom ive o 25 pages of ingle-spacedica typescript. ost of these ndustrytudiesuserealnames ffirms;ome, ndmost CCH cases of ndividual irms,lterdata to conceal dentitynd/orprotect onfidentiality.)tAvailablenbusinesschoolibraries,nbound.



23/32

American ournal fSociologyTABLE 3-Continued

Name (in Fig. 7) andICCH Identification* Notes and Other ourcesMainframe omputer 980..... ... Corey ndStar1971,pp. 108-56;HarvardBusiness Review 980; Fortune 980Minicomputer980 .......... Hayes 1980Nickel..................... ... Salter and Weinhold1979,chap. 10Tractors 1970s:4-578-083................. Rev. 11/77'9-280-080................. Rev. 2/809-171-368 BC 349) ........ ...9-313-123 BP 866) ........ Rev. 7/69 ...9-313-154 PB 867R) ........ ...9-377-704........... ... I9-574-858.................Color TV:1-380-180................. ...1-380-181................. ... ...1-380-191.................Fiberoptics:1-379-136................. Rev. 1/80 Will explode nto a clusterof hugenew1-379-139................. ... markets: nmy terms,y xceedsunity1-380-1 7.

Some ICCH case study series contradict the specifications (the "givens")demandedby mymodel and so do notappear, lthough hose ndustriesmight e suitable fstudied t another ate.20 nother,n theoil tankerindustry,anbe rephrasedomeet hemodel pecifications.2'The firms hich articipatenmany fthesemarkets o so typicallysdivisions.A separatedivision,which s an independentrofit enter nd20 Onesuch s the series n turbine eneratorsfter 963. This is an industrywithbuttwomembers, pureduopolywhich s not a market n my sense.Priorto 1963,however, twas a market n my sense,one in which severalotherU.S. and foreign roducersoinedthe two neyeing ne another or he samebusiness.This earlier orm discusselsewhere(White1981) on the basis of Sultan's two-volume ccount (1974, 1975). The late nine-teenth-centuryarketn sugarrefining,blydescribedbyEichner 1969), also ended asa trust, ut withmanymembers ather hantwo. I would code it as havingbeenin theGRIND region n 1885: some productdifferentiationaccordingto impurityevel andassociated taste) but less than the variation n cost amongthe score ofmanufacturersaround the United States (b < 6). Over the next decade, greatlyexpanded factoriesyieldedstriking conomiesof scale (c < 1). The set of producerfirms elongs in theCROWDED region or1895as shown nfig. . Then thedisappearanceofproductdiffer-entiation et the stageformarket nstabilitynd theHavemeyers' ounding p of all firmsintoa trust.21 If thevolume nthismarket thedecisionvariable,y) isnumber ftanker rips uppliedbya charterer,hemarket s not productionmarket uta truck-and-barterarketwhichshouldbe referredwith ts erratic rices nd speculativefeatures) o thepuretheory fexchange Newman 1965). But one can reorient o see y as the speed at whichgiventankers resenton a journey nd thus the volumeof oil delivered ertimeperiod.Figure7 contains pointfor hismarket, orwhichthe case studymaterialyieldsquite definiteestimates c = 2, a < 1, d < 0, 6 >? b, -y 1). (The existenceof such a market as aseparate ntitys,ofcourse, n idealization, iven tsheavy dependence n terms ftradeand supplies n thespotmarket ortankers s wellas on the chartermarket.)538



24/32

WhereDo MarketsComeFrom?relativelyutonomous,s the decisionmaker yeing he other roducers.22Forexample, heceramic-metalonnector arketabeledStrode'sCable nfigure is thatparticipatedn by an independentivision f the argefirmcalledEG&G (a division cquired arlier yincorporatingn old-line irmnamedStrode).

xIVacuumDrilling Tubes 1980b/d Mud LogWatches HousesX X X /Failed/1 Baby RegioFoods1965 X


25/32

American ournal f SociologyThe ICCH reports n the baby-foodmanufacturingndustry ield aclear-cutndwell documentedxamplen thePARADOX region, o con-trastwiththe diapersmarket iscussed bove. In the late 1920s Gerbercreated nd then ed an industryorcannedbabyfoods s decisively sIBM did mainframeomputers. erber, oo, has had precursorsnd earlycompetitorsuch as Mead Johnson's abulum f 1915and Clapp's cannedbroth, s well as large competitors hichperishedwithout trace (e.g.,Libby,McNeill,BirdsEye).Motherswere onvinced nough fGerber's uality hat t could chargeprice premiums f over 10% whilespendingnearly 10% less on total

advertisingnd promotion.n 1965Gerber ad moremodern nd efficientproduction acilitieshan tsmain ompetitors-Beech-Nut,wift, einz-whichn turnweremore fficienthan he maller rand-nameompetitors,oftenregional, nd the numerous mall private-label ompetitors. eteconomiesfscale werenotsignificant,npart becausetherewere o manyfoodvarieties over 100) at a given time s well as a high rate of changeofvarieties, bout one-thirder decade.The large ariety f tems lso tended olimit oncentrationy buyers none company's roducts ince othercompanies ffered ifferentarietiesthat mightcatch taste. For complexreasons willingnesso give opendiscounts y volume, rdiscountsffnvoice, rrebates, tc.), upermarketstended o give Gerber ess shelf pace than ts market hare.And contraryto Gerber's nd the other ompanies' redictions,he birthrate ad beendeclining ince the late 1950s,and almost all mothers lreadyused thecommercial aby foods, o that n 1965the overallelasticityywas wellunder nity.The constituentstimates nderlyinghe Baby Food 1965location nfigure area = 0.6,c = 1,b = 1,andd = 0.8.23 hese differubstantiallyfrom hoseforthe diapersmarket iven above in table 2. The resultingdifferencen ocationwithinhe sameregion fthetrade-offslane eadstosubstantial ifferencesredicted or ther spectsofthe two marketsseethenext ubsection ndtheAppendix or quations).InequalitynMarket haresSinceproducers atch roducers,hey rean nterrelatedet and nthepurecase are a cliqueofmutuallyware firms.t follows hat therewillnot beverymanyfirmsna market; robably herewillbe a dozenortwo.Eachwilltendto be distinctnroleand tohave a distinct lace on thequalityindexn. It follows hat the firmwhichby the existing cheduleW(y)chooses he argest olumewill tself ave a substantial hare fthemarket,23 Furthermore,stimationof the observed terms-of-tradechedule W(y) indicates alarge,negativevalue of the shift onstantk.540



26/32

WhereDo MarketsCome From?as will the nextfew.One hazards a guessthat the smallest irm o whichthe biggestwill endan eye will be within factor f 10 in size: thus tfollows rommytheory hata markets arrayed rom leaderwith fifthormore f the totalsales downto thesmallestwith1%0 rso ofthetotal.But how an weassessthese acts ncomparison ith ther orts f nequal-ity, runderstandhemnterms f he rade-offatioswhich rame igure?I choose his opic fmarket hares orllustrating y nalysis fmarketsbecause t seems ome at onceoneofthemost mportantnd oneofthemost exed spects f hepresent ocial cience nderstandingfproductionmarkets Scherer1970; Porter 1980). Observedmarkets lmostentirelyescape he omputationalrasp f xistingmicroeconomicheoryMansfield1975; Cohen ndCyert1975),although ya variety fdevices hetheoryand whateversobserved re declared o be inharmony.aw, inparticularantitrustaw,seems o shapetheconceptsmore han doesmicroeconomictheoryWilliamson1975]; or goodsurvey ee the ast CCH study itedfor ircuit reakersn table3, and for ocumentationfantitrust olicy eeSalterand Weinhold1979], p. 289-305).The shareoftotal market eldby thetop four or 3 or 5 or 8.. .) producer irmss givenprima faciestanding s evidence ordegree fcollusionn pricefixingnd so on. Incontrast, think ssessmentfmarket haredistributionhould i) bemadeon thebasis ofan explicit heory fmarket ormationnd (ii) be assessedina more eneral omparativerameworkor he tudy f ocial nequality.This secondobjectiverequires ooking t firms s social actors; it alsocallsfor measure f nequalitywidely sed in socialscience, uch as theGini ndex.Variousdegrees f inequality mongproducers n theirmarket haresyieldquite differentorts of roles forfirms, s well as differentverallatmosphere.or example, oes the argest irmverwhelmhe otherswithitspresence, r does the traditional nglish ndustrial lavor revail,withfirmsotvarying reatlyn size"sharing p" themarketn genteel ashion?(Dispersion n profitates, tc.,also are mportant,utthey re nfluencedinpartby facts nfixednvestmentsnd the ike,whichmymodel gnores.)To achieve a simpleformula, shall further estrictmyself o marketscheduleswithk = 0.24 In this pecial ase the nequalityncashflow qualsthe nequalitynmarket olumes, owhich now turn.The Giniindex s familiar s a measure f inequality r dispersion npersonal ncomesor wealth: t is derivedfrom orenz distributions orcumulative ercentage f total incomeversuscumulative ercentage fpopulation seeSchwartz nd Winship 980). t is suitable s well ormarketvolumes see Granovetter in press]for generaltheoretical iscussion).

24 They are known to be stable in all subregions f fig. 3 except the upper portion ofCROWDED; i.e., exceptfor1/1y a/c > 1,and a/c > b/6.541



27/32

American ournal fSociologyThis index, all it G,variesbetween ero for quality) ndone (for otalinequality). s calibration: or uniformistributionver ncome he Giniis 0.33; typicalnational ncomedistributionsave G = 0.4, and nationalwealth istributions ayhaveG = 0.8.Figure shows ow hevalueofGpredictedor alesvolumes f markettends ochange ccordingo locationnthetrade-offslane. ts limitationis that thequalities ndnumber ffirmsmustbe keptthe same,for hispurpose; o the various redictionsannotbe checkedndividuallygainstvalues computed cross an assembly fobservedmarkets uch as figure7.25 The main assumptions thatall firmsn any givenmarket respaced

b/d .48 .58 .67 .821-- .5 .6 .71 .84

.53 .64 .75 .87 ( _ __ )More Equality.57 .69 .80 .90

.2 .5 .71 .87 a/c.33 .82

.67

.43

.33

.5 .27 .06

Fromeq (14): uniform istributionf firms cross all n (O


28/32

WhereDo MarketsCome From?uniformlyn thequality ndexn. For mathematicalimplicityhelowerlimit nd theupper imit ftherange or are taken obe zero nd nfinity,respectively,nd firmsre assumed o be veryclosely pacedon quality.The general ormulaeorG under hese ssumptionsincluding = 0) are:for hysical olume y),

G (13)~~1+ 2(c -) (13)for ollarvolume W), G 1 + 2(c-1 (14)

bc+ adNote thatmarkets loseto theaxiswhere erfectompetitionies (d>>b) tend ohave high nequality ither nsalesoronphysical olume.Also,inequality ecreases y either quation s one moves way from aluationchangewithvolumebeing proportionalo cost changewithvolume i.e.,as one moves wayfrom he a/c = 1 vertical),whetherntoGRIND andPARADOX regionsn the one side or ntotheCROWDED region n theother. The equations o notholdfor b/d < a/c nthe atter,where

0 is nota stable chedule.)By inspection fequation 13) one can see that equal physicalvolumesareapproachedwhen he quality rade-off/d sunity ndnegative.n thebottom ffigure (d < 0), whenthetwo trade-offatiosare equal, butoppositensign, quality n salesvolumes mong roducerss approached.But it is preciselylong this ine thatmarket tructureisintegratesustbecause t relies n divergences etween ost trade-offnd value trade-offto cue distinctiveiches or irms.CONCLUSIONSMarkets re tangible liquesofproducers bservingach other.Pressurefrom he buyer ide creates mirrorn whichproducers ee themselves,Peterson 1972]) mayyieldnostatistical industry" r"product" at all close; yetbusinessjournalsusually reportdata in terms f government efinitions. econd, thenumericallevels of Gini ndicesformarket ales came from stylizedpower-law ramework or ostand valuation; they do not make accountants'discriminations. hird, most reportsofbusinessperformance se the egal corporations unit,but typicalFortune 00 firms anhave dozensof divisions,which s mentionedbove, are commonly heeffectivectors ntheactualmarkets model.Reinforcing iaseswill tend tomakepredicted alueshigherthan observedvalues. Eq. (14) assumes an infinite angeand uniform istribution orqualityn and so shouldtend toexaggeratenequality; t also assumesa scheduledescribedby a shift ndex k of zero. Empirical stimatesusually comefrom runcated ets offirmswiththesmallestomitted nd so will tend to underestimatenequality. see no reasonwhyeither fthesereinforcingiasesshould be correlatedwith the size of G, so therankorderofpredicted ndices houldtend to be thesame as that forobserved ndices.

543



29/32

American ournal fSociologynot consumers. eterogeneousroducers ith heir ifferentiatedroductsmayfind nd maintain tablerolesor niches. elf-interestedptimizing yeachof them an sustain globalmarket cheduleW(y), but t is exposedto three orts f market ailure. asic limits or quilibriumonfigurationsturn ut to dependon just two trade-offatioswhich ummarize he factsof costs and tastes.Thus markets re shaped by trade-offsetweendis-persions, ot by averages s suggested y the cliche that supplyequalsdemand.APPENDIXChanges nCost Level andProductPopularityTwodifferentinds fsensitivityfthe market s a whole o an exogenouschange see fig.5), such as an increase n q orr,shouldbe distinguished.The first s the effectn the market cheduleW(y) offigure . This is achange n height ftheprice chedule, scale change n a graph uchasfigure. The secondkind stheresultant hange, ncluding eedback ffects,on the actual volumeofsalesby eachfirm ndby the market s a whole.The latter, verallresultants easierto compute:when he shift onstantk = 0 (White1981,eqq. [25]and [26]),all regions ield the same propor-tionality or he aggregate olume fthemarketW,namely,

W, (rclq) y/(c-ay). (A1)It befits he nature fmarket ormation,s a balancing ftrade-offsndispersionssee fig. ), thatthe market olumevarieswithvaluation calefactor to theexponent ofcostvariationwithvolume; and, similarly,dependence f W on costsscale q is through raisedto thepowera of

volume's ontributiono valuation.But note that market ales totalgoesdown s cost scale goes up. It is alsoobvious hatthe sensitivityfmarketsize to either cale factor ncreases s the demand-saturationxponentrises-as y becomes loser o c/a. Even more bvious s thefactthatthisnet feedback esultant f change n costsor valuationscales on marketsales naggregate oesnotdepend t all onthetrade-off/d nthequalitydimension! hese findingseneralizehenumerical esults resentedbovefor hree llustrative arkets.But visible changes n the market chedule W(y), the price scheduleperceived y producers,ake quite a differentorm, nd thisformtselfdiffersccordingo regionn thetrade-offslaneoffigure . RatherthanquotetheformulaefromWhite1981, qq. [28]-[30]), have llustratednthe three etailed xamples owdramatic hechangenforms.Andtherethe changed ocations feachindividual irmre indicated n thechangedschedule.544



30/32

WhereDo MarketsCome From?PriceStructureImplicit n the divergencein text above,eqq. [13]and [14]) between hetwo Gini ndices, or ales volume ndforphysical olume,s variation fprice withvolume.The last figuren White 1981) reports hepredictedvariationnpriceperunitaccording o totalvolumesfromvariouspro-ducers n that market, eparately or ubregionsf the trade-offslane.(Onlymonotonicitynd existencefextremarespecified;o thepredictedcurveholdsthroughouthe givensubregion egardless f theparticularvalueson quality ndex.The simplicityfthese shapesfollows rommyapproximationfcostand valuation chedules y families fpowerfunc-tions.)At various ocationswithin ne region fthe trade-offslane, theCROWDED region, ne canfind rice-by-volumeurves isingmonotoni-cally,orfallingmonotonically,rconcavedownward,rconcaveupward!This plasticity nderlinesmy contentionhat prices re a secondary ndexpost phenomenon.Consider numericalrice chedule rom neoftheearlier llustrations,for he ight-aircraftndustrywith = 1 = r, y= 0.7 andtake0 = 1.63,see table 1). In the schedulewith hift onstant = 0.5, the price-volumepairs correspondingo Beech, Cessna,and Piper,respectively,re (wherepn. is priceper unit)

Pn .440 .414 .399Yn 1.0 1.6 1.9n .613 .538 .503.

Thenvalueswere hosen o yield heobserved atios f alesvolumes monglight-aircraftanufacturers:igher aluesmean as before) igher uality,so thatBeech s theCadillac andPiperthe Chevy.It isdifficultoobtain eliable nformationn effectiverices nd physicalvolumes old for ll ofthe firmsn a givenmarket. t is not sufficientoobtainpostedprice chedules r pricebooks, incetheseoften epresentbasis fornegotiationmore hana reliable eport fpractices. n any eventmymodelpresupposes ggregationfall sales by the givenfirm nto oneaggregate olume, nd aggregationf the correspondingevenue crossdispersionfconcrete alespricesndifferentatches odifferentustomers.

REFERENCESAkerlof,George.1976. "The Economies of Caste and the Rat Race and OtherWoefulTales." Quarterly ournal f Economics 4:599-617.Alonso, William.1964. Location nd Land Use: Toward GeneralTheory fLand Rent.Cambridge,Mass.: HarvardUniversity ress.Business Week.1980. "Building Up to a Shortage n Cement," no. 2629 (May 5), pp.138-44.545



31/32

American ournal fSociologyCohen, Kalman, and RichardCyert. 1975. Theory f theFirm. 2d ed. Englewood Cliffs,N.J.: Prentice-Hall.Corey,E. Raymond. 1978. Procurement anagement: trategy, rganizationnd Decision-Making. Boston: CBI.Corey, E. Raymond,and S. H. Star. 1971. Organization trategy. oston: HarvardGraduate School of Business Administration.Dixit, A. K., and J. E. Stiglitz.1977. "MonopolisticCompetition nd Optimum roductDiversity."American conomicReview 7:297-308.Eichner,Alfred . 1969. The EmergencefOligopoly. altimore:JohnsHopkins UniversityPress.Fortune. 980. "IBM's Battle to Look SuperhumanAgain" (May 19), pp. 106-12.Granovetter,Mark. In press. "Toward a Sociological Theoryof Income Differences."In Sociological erspectivesnLaborMarkets,ditedby var Berg. New York: AcademicPress.Grossman, . 1975. "Rational Expectations nd the EconometricModelingofMarketsSubject to Uncertainty."Journal fEconometrics:255-72.HarvardBusiness Review. 980. Vol. 58 (March-April):82-94.Hayes, Thomas. 1980. "Digital's Masteryin Computers."New YorkTimes (April 7),pp. D1-D5.Ijiri, Yuji, andHerbertA. Simon.1977.SkewDistributionsndthe izes ofBusinessFirms.Amsterdam:North-Holland.Kantor, Brian. 1979. "Rational Expectations nd EconomicThought." Journal fEco-nomicLiterature 7:1422-41.Lancaster, Kelvin J. 1966."A New Approachto ConsumerTheory." Journal fPoliticalEconomy 4:132-57.

. 1979. Variety, quity nd Efficiency:roductVarietyn an Industrial ociety.NewYork: Columbia University ress.Mansfield,Edwin. 1975. Microeconomics: heory nd Applications.2d ed. New York:Norton.Muth, JohnF. 1961. "Rational Expectationsand the Theory of Price Movements."Econometrica 9:315-35.Newman,Peter. 1965. The Theory f Exchange. nglewood Cliffs, .J.: Prentice-Hall.Porter,Michael E. 1976. Interbrand hoice, trategy,nd BilateralMarketPower. Cam-bridge,Mass.: HarvardUniversity ress.. 1980. Competitivetrategy. ew York: AcademicPress.Rosen, Sherwin. 974."Hedonic Pricesand ImplicitMarkets: Product DifferentiationnPure Competition."Journal fPoliticalEconomy 2:34-55.--. 1978. "Substitution nd Division of Labour." Economica 5:235-70.Salter,Malcolm S., and W. A. Weinhold. 1979.Diversificationhrough cquisition.NewYork: Free Press.Scherer, . M. 1970. IndustrialMarket tructurend Economic erformance. ew York:Rand McNally.Schwartz,J. E., and ChristopherWinship.1980. "The WelfareApproachto MeasuringInequality." Pp. 1-37 in SociologicalMethodology980, ditedbyKarl Schuessler. anFrancisco: Jossey-Bass.Shiskin,Julius, nd Milo Peterson, ds. 1972. Standard ndustrialClassification anual.Washington, .C.: OfficefManagement nd theBudget,StatisticalPolicyDivision.Spence,A. Michael. 1974.Market ignaling: nformational ransfern Hiring ndRelatedScreening rocesses.Cambridge,Mass.: HarvardUniversity ress.

. 1976. "ProductSelection, ixed Costsand MonopolisticCompetition."Review fEconomic tudies43:217-35.Steiner,Peter 0. 1975. Mergers:Motives,Effects, olicies. Ann Arbor: University fMichiganPress.Sultan, Ralph G. M. 1974.Pricing n theElectricalOligopoly. ol. 1, CompetitionrCollu-sion. Boston: Harvard Graduate Schoolof Business Administration.. 1975.Pricingn the lectrical ligopoly. ol. 2,BusinessStrategy.oston: HarvardGraduateSchoolofBusiness Administration.546



32/32

WhereDo MarketsComeFrom?Terleckyj, Nestor E., ed. 1976. Household Production nd Consumption.New York:Columbia University ress (forNational Bureau ofEconomicResearch).White,H. C. 1976."Subcontracting ith n Oligopoly: pence Revisited." RIAS WorkingPaper no. 1, Harvard University, eptember.--. 1978. "Markets and HierarchiesRevisited."Mimeographed.Cambridge,Mass.:Department fSociology,Harvard University,March.1979. "On Markets."RIAS Working aper no. 16,HarvardUniversity, pril.1980. "Markets as Social Productions." Mimeographed. Cambridge,Mass.:Department f Sociology,Harvard University.--. 1981. "ProductionMarkets s Induced Role Structures." p. 1-57 inSociologicalMethodology981,editedby S. L. Leinhardt.San Francisco:Jossey-Bass.Williamson, . E. 1975. Markets nd Hierarchies:Analysis and Antitrustmplications.New York: Free Press.

547

Documents

De Onde Vem Os Mercados