Creating External Capabilities: Innovation and Vertical Disintegration in the Microcomputer Industry RichardN. Langlois1 Universityof Connecticut

In a famouspassage in Capitalism,Socialism, and Democracy, Joseph Schumpetersingled out the large capitalistfirm as the principalsource of economicprogress since the nineteenthcentury.

As soon as we go into the details and inquire into the individualitems in whichprogress was most conspicuous, the trail leads not to the doors of those firms that work under conditionsof comparativelyfree competitionbut preciselyto the doors of the large concerns--which,as in the case of agriculturalmachinery, also account for muchof the progress in the competitivesector--and a shockingsuspicion dawns upon us that big businessmay havehad more to do with creating that standardof life than with keepingit down[15, p. 82].

It is not surprisingthat businesshistorians, with their focus on large enterprises,should be sympatheticto Schumpeter'sargument. Indeed, Alfred D. Chandler,the dean of present-daybusiness historians, paints a similarpicture of the large firm as an engineof progress.In Chandler's story,however, the large enterprisecomes across less as a generatorof innovationthan as an "institutionalresponse" to innovationand growth whosesuperiority lies in its abilityto createmassive internal economies of high-volumeproduction [2, p. 12]. Yet, there is anotherimportant tradition in economicsthat seesthe sourcesof economicgrowth in a slightlydifferent light. While never denyingthe importanceto economicprogress of internaleconomies, Alfred Marshall and his followersalso highlighted the systemicinteractions among

1Theauthor would like to thank Brian Loasby and Paul Robertson foruseful ideas and conversations.

BUSINESS AND ECONOMIC HISTORY, SecondSeries, Volume Nineteen, 1990. Copyright(c) 1990 by the BusinessHistory Conference.ISSN 0849-6825.

93 a largenumber of competingand cooperating firms. 2 ForMarshall such interactioncould yield "externaleconomics" that play an importantrole in economicprogress quite in addition to that played by the economics internalto particularbnsincss organizations [10, book IV, chs.IX-XIII]. As William Lazonicksuggested in his recentkeynote address to the BusinessHistory Conference, economic progress requires the development of organizationalcapabilities.

Organizationalcapabilities represent the powerof plannedand coordinated specialized divisions of labor to achieve organizationalgoals. Through planned coordination,the speciali?edproductive activities of massesof individualscan coalesceinto a coherentcollective force. Through planned coordination,organizations can integratethe varioustypes of knowledgeneeded to developnew productsand processes. Through plannedcoordination, organizations can speedthe flow of work from purchasedinputs to soldoutputs, enabling the enterpriseto achievelower unit costs[8, p. 1].

It is dear from the dictionof this passagethat Lazonicksees organizational capabilitiesas primarilya matter of conscious,centralized administrative coordination.By contrast,Marshall envisaged the creationof organizational capabilitiesmuch more broadly. For Marshall,as Brian Loasbypoints out, capabilitiesneed not all residewithin the boundariesof the firm. Not only do firms possessorganizational assets--so also do markets. "Both are structuresfor promotingthe growth of knowledge,and both require consciousorganization" [9, p. 120]. Someimportant economic capabilities can reside within a network of competingand cooperatingfirms. And, althoughoften "conscious"and not restrictedto price signalsalone, the coordinationamong these firms is far from centrallyplanned. One areain whicha decentralized,market-like organization may have advantagesis in the generationof certain kinds of technologicaland organizationalinnovation. It is clear,as Marshallcertainly understood, that sometypes of innovationtake placemore readily within the organizational structureof a firm. I have elsewhereargued for "dynamic"transaction-cost explanationsof verticalintegration, in whichthe difficultiesof coordinating some types of innovativeactivity acrossmarket boundariescan make internalorganization a cheaperalternative [5]. And a coauthorand I have exploredone importantcase in which innovation--andrapid declinesin productprice--took place within the frameworkof internaleconomies and large-scaleproduction: the movingassembly line and the Ford Model T [7]. But we also found episodesin the historyof the automobileindustry in whichthe existenceof a varietyof competingfirms spurredinnovation and

2Froma purelystatio point of view, it istrue that external economies areexternal to particularfirms not to the industry (appropriatelydefined) as a whole [4, p. 597]. Marshall and his followers did not reason in strictly static terms, however. See, for example, [16]. 95 even forcedsome vertical disintegration upon the large firms. Moreover, a number of other casescome to mind in which rapid progress--rapid declinesin productprice and improvementsin productquality--took place within a highlydisintegrated structure. As Nelsonand Winter haveargued, innovation in timesof uncertainty is a matterof rapidtrial-and-error learning that benefitsfrom competition among many alternatives[12]. To this extent, the ability of a large organi?ationto coordinatethe implementationof an innovation,which is dearly an advantagein somesituations, may be a disadvantagein others. Coordinationmeans getting everyone on the same wavelength. But the variationthat drivesan evolutionarylearning system depends on people beingon ddferentwavelengths--it depends, in effect,on outbreeding.This is somethingmuch more difficultto achievein a largeorgani?ation than in a disintegratedsystem. Thus, the abilityto engagein rapid trial-and-error learningcan sometimesbe an externaleconomy that cannotbe tracedback to economiesinternal to the individualfirms: it is a propertyof the system as a whole. TheMicrocomputer Industr•

One of the moststriking examples of externallycreated organizational capabilitiesis the present-daymicrocomputer industry. The historyof the industryis rife with unintendedconsequences and what we might call unplannedcoordination. It is a history in which the most successful productswere thosethat took greatestadvantage-and allowed users to take greatestadvantage--of a large networkof capabilitiesexternal to even the largestfirms. And it is a historyin which the greatestfailures occurred whenbusiness enterprises bypassed the externalnetwork and attemptedto rely significantlyon internalcapabilities. In 1975 American firms possessedthe world's most advanced capabilitiesin mainframecomputers (IBM), minicomputers(DEC), and integratedcircuits (Intel, TexasInstruments). But the microcomputerdid not emergefrom theseenterprises. Instead, the industrysprang from a welterof smallfirms who assembled machines, supplied add-on parts, wrote software,and providedknowhow and service. It is conventionalto date the beginningof the microcomputerat January1975, when that month'sissue of PopularElectronics carried a cover story on the MITS/Altair computer.Run out of an Albuquerque,New Mexicostorefront by oneEd Roberts,MITS providedvarious kinds of kits to the hobbyistmarket. The computerRoberts and his coworkersput togetherwas little more than a box with a microprocessorin it. Its only input/outputdevices were lights and toggle switches on the front panel,and its memorywas a minuscule256 bytes(not kilobytes).But the Altair was, at least potentially,a fully capablecomputer. Like a minicomputer,it

3Fora muchlonger and better-documented historyofthe microcomputer, see[6], on which this section draws, possesseda number of 'slots" that allowed for expansion--additional memory,various kinds of input/outputdevices, etc. These slotshooked into the microprocessorby a systemof wires calleda bus; the Altair bus, which came to be known as the S-100 bus becauseof its 100-line structure, wasthe early industrystandard of compatibility. The Altait's impoverishedcapabilities did not deter buyers:the machinesold beyondall expectation.But the lack of functionalitydid give rise to two phenomena:third-party suppliers of add-ohsand user groups. The latter were organi?ationsof hobbyistswho shared informationand software.The third-partysuppliers were also typically enterprising hobbyists, and the productsthese firms supplied--likememory boards--fried the gap left by MITS's tardy and low-qualityadd-ohs. In a sense,then, the Altair was quicklycaptured by the hobbyistcommunity and becamenot a self- containedproduct but a modular technologicalsystem. To accomplish anything,one needednot just the box itselfbut alsothe know-how,add-on boards,and softwareprovided by a largenetwork of externalsources. The networkcharacter of the microcomputerwas fosteredby Roberts'sdesign decisions,themselves a reflectionof hobbyistattitudes toward information sharing.More importantly,however, the capabilitiesof MITS were puny comparedto thoseof the larger community;and thoselarger capabilities were necessaryto take full advantageof a productwith suchhigh demand and so manydiverse and unforeseenuses. The earlysuccess of MITS (and later a clonecalled the IMSAI 8080) cementedthe popularityof the S-100standard, especially among hobbyists, who were still the primarybuying group. Indeed, proponents of the S-100 (and the Intel 8080 microprocessorit was built around) felt that their standardhad reached"ritical mass"and that competingchips and buses were doomed.But the predicteddominance of the S-100never materialized. In 1977, a little more than two years after the Altair's debut, three importantnew machinesentered the market,each with its own incompatible operatingsystem, and two basedaround a differentmicroprocessor. The almostsimultaneous introduction of the Apple II, the CommodorePET, and the Tandy TRS-80 Model I began a new regime of technological competitionand moved the industry away from the hobbyistinto an enormouslylarger and more diversemarket. The mostfamous and important of the threemachines was the Apple II. The early historyof Apple Computerhas become the stuff of legends. And someof the legendsare in fact true. Apple wasthe creationof Steven Jobsand StephenWozniak, two collegedropouts and tinkerers. And some of the company'searly work did take placein a garage. The Apple II reflectedin many ways a compromisebetween the divergent visions of the two founders. A quirky but determined entrepreneur,Jobs provided the drive to turn a hobbyinto a multi-million dollar corporation.He conceivedof the computeras an appliance,a well- designedmachine that was easyto comprehendbut aimedat a few specific uses.Under Jobs'sinfluence, the machinewas compact,attractive, and professionalin appearance. A gifted engineer,Wozniak was the actual designerof the Apple II. Like his fellow hobbyists,he believed in modularity,expandability, and an opensharing of informationwith buyers 97

and outsidesuppliers. Thus, comparedwith earlier hobbyistmachines like the Altair or the IMSAI, the Apple II wasan integratedand understandable product. Yet, thanksto its eight expansionslots--the result of Wozniak winningan argumentwith Jobs--itwas also still a system,able to draw on the large crop of externalsuppliers of softwareand add-ohsthat quickly sprangup. Indeed, Apple relied heavilyon externalsuppliers for almost everything.Like the IBM PC a few yearslater, the Apple II was almost completely"outsourced." Apple PresidentMike Scott,who wasin charge of production,did not believein automatedmanufacturing and expensive test equipment:"Our businesswas desistming,educating, and marketing. I thoughtthat Apple shoulddo the least amountof work that it could and that it shouldlet everyoneelse grow faster. Let the subcontractorshave the problems."[11, pp. 200-01.]The companyengaged in board-stuffingon the putting-outsystem before turningto a contractboard-stuffing firm in San Jose. Scott even useda contractorfor the firm's payroll. By 1981 the usesof the microcomputerwere becomingclearer than theyhad beenonly few yearsearlier, even if the full extentof the product spacelay largelyunmapped. A microcomputerwas a systemcomprising a numberof more-or-lessstandard elements: a microprocessorunit with 64K bytesof RAM memory;,a keyboard,usually built into the systemunit; one or two disk drives;a monitor;and a printer.The machineran operating- system software and applicationsprograms like word-processors, spreadsheets,and databasemanagers. The marketwas no longerprimarily hobbyistsbut was increasinglybusinesses and professionals.Total sales weregrowing rapidly. But the Apple II--with onlya 40-columndisplay and little memory--remainedat the fringesof the businessmarket. There was room for a more capablemachine targeted to businessprofessionals. It was this gap that IBM attackedwith its PersonalComputer, a machinethat wouldsoon outdistance even the fantasticallysuccessful Apple in sales.But IBM's successstemmed not from the focusingof its great internal capabilitiesbut rather from its willingnessto abandonits capabilitiesin favor of those in the external network. In July 1980William Lowe met with IBM's CorporateManagement Committee. John Opel, soonto becomeIBM's president,had charged Lowe with gettingIBM into the market for desktopcomputers. Lowe's conclusionwas a challengeto IBM's top management."The only way we can get into the personalcomputer business," he told the CMC, "s to go out and buy part of a computercompany, or buy both the CPU and softwarefrom peoplelike Apple or Atari--becausewe can't do this within the cultureof IBM" [3, p. 9]. The CMC knewthat Lowewas right, but they wereunwilling to put the IBM nameon someoneelse's computer. So they gaveLowe an unprecedentedmandate: go out and build an IBM personal computerwith completeautonomy and no interferencefrom the IBM bureaucracy.Philip Donald Estridge,who quickly succeededLowe as directorof the project,later put it thisway. "We were allowedto develop like a startupcompany. IBM acted as a venturecapitalist. It gaveus managementguidance, money, and allowedus to operateon our own"[1, October3, 1983,p. 86]. Estridgeknew that, to meet its deadlines,IBM would have to make heavyuse of outsidevendors for parts and software. The owner of an Apple II, he wasalso impressed by the importanceof expandabilityand an openarchitecture. He insistedthat his designersuse a modularbus system that wouldallow expandability and he resistedall suggestionsthat the IBM team designany of its own add-ous. Another radicaldeparture from IBM traditionwas the marketingof the PC. ShunningIBM's staff of commission-salesagents, the PC group turned to retail outlets to handle the new machine. One outlet was Sears BusinessCenters; the other was ComputerLand.Here again,the project philosophywas to do thingsin keepingwith the way theywere donein the microcomputerindustry--not the way they were done at IBM. Perhapsthe moststriking way in whichIBM relied on externalcapabilities, however, was in the actualfabrication of the PC. All partswere put up for competitive bids from outsidesuppliers. When internal IBM divisionscomplained, Fatridgetold them to their astonishmentthat they too could submitbids like anyoneelse. With a little prodding,some IBM divisionsdid win contracts.The Charlotte,North Carolinaplant won a contractfor board assemblyand the Lexington,Kentucky plant made the keyboard. But an IBM plant in Coloradocould not make quality disk drives,so Fatridge turned to Tandon as principalsupplier. Zenith made the PC's power supply,SCI Systemsstuffed the circuitboards, and Epsonmade the printer. The IBM PC called forth a legion of softwaredevelopers and producers of add-on peripherals. Beyond this, however, its early phenomenalsuccess also called forth competitorsproducing compatible machines.The era of the doriesfalls into two distinctperiods. The early makers of dories fed on the excess demand for PCs. With one brilliant exception(namely, Compaq), these manufacturers disappeared when IBM begancatching up with demandand loweredprices in 1983and 1984. The secondwave of doriesbegan a coupleof yearslater whenIBM abandoned its originaldesign in favorof the PC AT, whichwas built aroundthe faster Intel 80286 chip. What is especiallyinteresting is the diversityof sourcesof these compatiblemachines. Many come from Americanmanufacturers who sell under their own brand names. These would include Compaq, Zenith, Tandy,and Kaypro,the latter two havingdumped their incompatiblelines in favorof completeIBM compatibility.Another group would be foreign manufacturersselling under their ownbrand names. The largestsellers are Epsonand NEC of Japanand Hyundaiof Korea. But there is alsoa large OEM (original-equipmentmanufacturer) market, in whichfrans--typically Taiwaneseor Korean, but sometimesAmerican or European--manufacture PCs for resale under another brand name. Perhapsthe most interestingphenomenon, however, is the no-name done--thePC assembledfrom an internationalcornucopia of standardparts and resold,typically, through mail orders. Becauseof the opennessand modularityof the IBM PC and the dominanceof its bus and software standards,a hugeindustry has emergedto manufactureparts compatible with the PC. The resultingcompetition has drivendown prices in almost all areas. Most manufacturers,even the large branded ones, are really assemblers,and they draw heavilyon the wealth of availablevendors. But the parts are also availabledirectly and it is in fact quite easy to put togetherone's own PC from parts orderedfrom the back of a computer magazine. By one 1986 estimatethe stageof œmalassembly added only $10 to the cost of the œmishedmachine--two hours work for one person earningabout $5 per hour [1, July 28, 1986, p. 64]. As the œmalproduct could be assembledthis way for far less than the going price of name brands--especiallyIBM--a wealth of backroomoperations sprang up. The partslist is truly international.Most boardscome from Taiwan, stuffedwith chipsmade in the U.S. (especiallymicroprocessors and ROM BIOS) or Japan(especially memory chips). Hard-diskdrives come from the United States,but floppy drives come increasinglyfrom Japan. A powersupply might come from Taiwanor HongKong. The monitormight be Japanese,Taiwanese, or Korean. Keyboardsmight come from the U.S., Taiwan, Japan,or even Thailand. It is temptingto interpretthe successof the IBM PC as merelythe resultof the powerof IBM's name. While the namewas no doubtof some help, especiallyin forcingMS-DOS as a standardoperating system, there are enoughcounterexamples to suggestthat it was the machineitself--and IBM's approachto developingit--that must take the credit. Almost all other large firms,many with nearlyIBM's prestige,failed miserablyin the PC business. The companythat Apple and the other early computer makers feared most was not IBM but Texas Instruments,a power in integratedcircuits and systems(notably electronic calculators) [11, p. 228]. But TI floppedby enteringat the low end, seeingthe PC as akin to a calculatorrather than as a multipurposeprofessional machine. When TI did enter the businessmarket in the wake of the IBM PC, its TI Professional also failed becausethe companyrefused to make the machinefully IBM compatible.Xerox and Hewlett-Packardwere both slowout of the blocks. But the case I want to focus on is the failure of Digital Equipment Corporation.4 DEC is the second-largestcomputer maker in the world, and the largest maker of minicomputers. The companywas committedto a strategyof filling out its line of VAX minicomputersand, more broadly,to the idea of terminal-basedtime-sharing computing. In 1980,however, DEC PresidentKen Olsen becamepersuaded that the companyshould get into the personalcomputer business. The Professional300 serieswas to be the company'sprincipal entry into the fray. It would have a proprietary operatingsystem based on that of the PDP-11 minicomputer,bit-mapped graphics,and multitaskingcapabilities. The Rainbow100 was considereda lower-end alternative.It looked much like the Professional,but, rather than a proprietaryDEC microprocessor,it used two Intel chips,enabling it to run someexisting outside software. Although the Professionalwas targeted to bring in 90 per cent of the profit, most of the 300,000PCs DEC sold

4Thereisin fact a similarstory to be told about the development ofthe Apple III. See [S]. 100 were Rainbows,and many of thosewere sold at fire-saleprices. All told, the companylost about$900 million on its foray into microcomputers[13, p. 238]. One mightbe temptedto see this as a matter of strategicmistakes. But the real mistakewas the company'sunwillingness to take advantageof externaleconomies. A technicalperfectionist, Olsen believedthat DEC couldbe successfulby creatinga superiorproduct. This had worked in minicomputers:put together a machine that would solve a particular problemfor a particularapplication. The PC is not, however,a machine for a particularapplication; it is a machineadaptable to many applications, includingsome its usershadn't imagined when they bought their machines. Moreover,Olsen underrated the valueof software.In minicomputersDEC couldgenerate adequate software inhouse, and users,who are highlyskilled technically,could write their ownapplications. But thiswas not the casein the wide-openmicrocomputer market. And, unlike IBM, DEC choseto ignoreexisting third-party capabilities. Except for the hard diskand the line cord, DEC designedand built every piece of the Professional. The companytooled the sheetmetal and plastics,manufactured the floppydrive, and even developedthe microprocessor.

Conclusion

Although dependenton, and in many ways driven by, economies internal to the semiconductorindustry, the rapid growth and development of the microcomputerindustry is largelya storyof externaleconomies. To put it anotherway, it is a storyof the developmentof capabilitieswithin the context of a decentralizedmarket rather than within large, vertically integratedfirms. Indeed, the microcomputerindustry represents in many waysa caseexactly opposite to the pictureof economicprogress one gets from readingSchumpeter, Chandler, or Lazonick.Rather than a few large firms supplantinga decentralizedmarket in an act of innovation,we find insteadthe largefirms lagging behind the small,especially in the beginning. And when large, verticallyintegrated firms do enter the picture,even the largestof them is forced to rely on a multitude of outsidesuppliers for parts,software, knowhow, and sales. There are, I think, a number of reasonsfor the importanceof externaleconomies in this industry.First of all, the size, diversity,rapid development,and unknowncharacter of the market for microcomputers meantthat no singleorganization could develop the necessarycapabilities with anything like the speed those capabilitiescould develop in a decentralizedmarket. Henry Ford wasforced to integratevertically because the externalmarkets could not createnew capabilitiesas fast as he could. But IBM was forcedto disintegratealmost completely to make the original PC becausethe companycould not createcapabilities nearly as fast as the market could. Second,microcomputers are not appliances--inthe way toastersare appliances--butare modular systems. This is not becauseof any technologicalnecessity but becausemodularity allows for a minuteand well- coordinateddivision of labor in the market, which in turn allows for the lol

rapid creationof new capabilities.It is misleadingto thinkof a computer as an end-product. A computeris a meansto an end--or to a variety of ends. We thus need to think of computersin hedonicor Lancasterian terms.Computer makers offer a mixtureof attributesthat consumerscan chooseamong to producetheir favoredcombinations. For most kinds of products--toastersor automobiles, say--manufacturers offer presetpackages. One can choosefrom a multiplicityof packages,but one cannotchoose the engine from one kind of car, the hood ornamentfrom another, and the front suspensionfrom a third. Not only are there transactioncosts of such pickingand choosing,there are alsoeconomies of scalein assemblingthe parts into a finished package.In microcomputers,however, both the transactioncosts of knowingthe availableparts and the scaleeconomies of assemblingthe packageare low for a wide segmentof the userpopulation. As a result,the real winnersin the evolutionaryprocess--like the AppleII and the IBM PC--wereprotean machines that couldbe tailoredto specific user demandsand could be upgradedeasily as new end-usesand technologicalpossibilities emerged. Finally, as I have alreadyargued, a decentralizedand fragmented systemcan have advantagesin innovationto the extentthat it involvesthe trying out of many alternateapproaches simultaneously, leading to rapid trial-and-errorlearning. This kind of innovationis especiallyimportant when technologyis changingrapidly and there is a high degreeof both technologicaland market uncertainty. And thiskind of innovationcertainly characterizedthe microcomputerindustry. That the microcomputerindustry partook of externaleconomies of learningand innovationis in manyways a familiarstory that neednot be retold.Popular accounts of SiliconValley soundvery much like Marshall's localizedindustry in whichthe "mysteriesof the tradebecome no mysteries; but are as it were in the air, and children learn many of them unconsciously"[10, IV.x.3, p. 225]. Compare, for example,Moritz's discussionof the effect of SiliconValley cultureon one particularchild, Wozniak. "In Sunnyvalein the mid-sixties,electronics was like hay fever: It wasin the air and the allergiccaught it. In the Wozniakhousehold the olderson had a weakimmune system" [11, p. 29]. One couldeasily multiply citations. This learningeffect went beyondthe backgroundculture, however. It includedthe proclivityof engineersto hop jobs and start spinoffs, creating a pollination effect and tendencyto biological differentiationthat Marshallwould have appreciated. Moreover, the external economiesof ideaswere not in fact restrictedto the physicalrealm of SiliconValley--or even Silicon Valley plusRoute 128. As AustinRobinson anticipatedlong ago, externaleconomies in a developedeconomy are increasinglyintangible and therefore,in his phrase,"mobile" [14, p. 142].

References

1. Business Week. 2. AJfredD. Chandler, Jr., The Visible Hand: The Menegerial Revolutionin American Business(Cambridge, MA, 1977). 102

3. JamesChposky and Ted Leonsis,Blue Magic: The People, Power and PoliticsBehind the IBM PersonalComputer (New York, 1988). 4. Frank H. Knight, "Some Fallacies in the Interpretationof Social Cost," Quarterly Journalof Economics,38 (1924). 5. Richard N. Langlois,"Economic Change and the Boundariesof the Firm," Journal of Ins#tutionaland TheoreticalEconomics, 144 (1988), 635-57. 6. , "External Economies and Economic Progress: The Case of the MicrocomputerIndustry," manuscript, March 1990. 7. and Paul L. Robertson,"Explaining Vertical Integration: Lessons from the AmericanAutomobile Industry," Journal of EconomicHistory, 49 (June 1989), 361- 375. 8. William Lazonick,"Organizational Capabilities in American Industry:The Rise and Decline of Managerial Capitalism,"Keynote address to the Business History Conference, March 23, 1990. 9. BrianJ. Loaeby,"Firms, Markets, and the Principleof Continuity,"in J. K. Whitaker, ed., CentenaryEssays on AlfredMarshall (Cambridge,ENG, 1990). 10. AJfredMarshall, Principles of Economics(London, 1920). 11. Michael Moritz, The Little Kingdom: The PrivateStory of Apple Computer (New York, 19S4). 12. Richard R. Nelson and Sidney Winter, "In Search of More Useful Theory of Innovation,"Research Policy, 5 (Winter 1977), 36-76. 13. Glenn Rifkinand George Hatrat, The UltimateEntrepreneur: The Story of Ken Olsen and DigitalEquipment Corporation (Chicago, 1988). 14. E. A. G. Robinson,The Structureof CompetitiveIndustry (Cambridge, ENG, 1935). 15. JosephA. Schumpeter,Capitalism, Socialism, and Democracy(New York, 1950). 16. Nlyn A. Young, "IncreasingReturns and EconomicProgress," Economic Journal, 38 (1928), 523-542.