Product variety and firm survival in the microcomputer software industry

Strategic Management JournalStrat. Mgmt. J., 25: 1005–1025 (2004)

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/smj.408

PRODUCT VARIETY AND FIRM SURVIVAL IN THEMICROCOMPUTER SOFTWARE INDUSTRY

TOM COTTRELL* and BARRIE R. NAULTHaskayne School of Business, University of Calgary, Calgary, Alberta, Canada

This article provides an analysis of product variety and scope economies in the microcomputersoftware industry by using detailed firm-level and product-level information on firms’ bundlingof functionalities over application categories and computing platforms. We find that the man-agement of product variety through the way different application categories are integrated inproducts and the platforms on which these products are offered can be as important as the sig-nificance of scope economies at the more aggregated firm level. Specifically, we find that thereis little evidence of firm benefits from economies of scope in production, but there is substantialevidence that products benefit from economies of scope in consumption. In addition, we findthat firms with products that encapsulate more application categories perform better, and thosewith products that cover more computing platforms perform worse. Finally, changes in productvariety through new product introductions improve firm performance, but extensions to exist-ing products hinder the performance of the firm and the product. We conclude that research inscope economies can benefit from a more detailed model of the evolution of product variety thatincludes data and analysis at the firm level and at the product level. Copyright 2004 JohnWiley & Sons, Ltd.

INTRODUCTION

The software industry has witnessed a wide assort-ment of product strategies over its short history,ranging from multiple stand-alone niche appli-cations to integrated application suites. Scopeeconomies play a critical role in the success ofdifferent product variety and product integrationstrategies. In this article, we study firms’ firm-level and product-level strategies in the productionand consumption of microcomputer software inthe early 1980s. We explore the impact of scope

Keywords: firm survival; diversification; product strategy∗ Correspondence to: Tom Cottrell, Haskayne School of Busi-ness, University of Calgary, 2500 University Drive NW,Calgary, Alberta, Canada T2N 1N4.E-mail: [email protected]

economies, product integration or focus, and prod-uct variety on firm and product performance. Weexamine these impacts in the context of applica-tion categories (e.g., spreadsheets, word process-ing, databases) and hardware/operating systemsplatforms (e.g., MS DOS, Apple II, CP/M).

Firms often have broad flexibility in their prod-uct-level implementation of a product variety strat-egy, and this is widely recognized in the soft-ware industry. Such flexibility includes bundlingof different application categories into products,the offering of products over different computingplatforms, introduction of new products into old ornew categories and for old and new platforms, andthe extension of existing products into new cate-gories or for new platforms. When firms have suchflexibility in their product variety strategy, detailedknowledge of product-level implementation and its

Copyright 2004 John Wiley & Sons, Ltd. Received 11 June 1997Final revision received 13 January 2004

1006 T. Cottrell and B. R. Nault

outcomes is as essential as knowledge of firm-level outcomes to evaluate the impact of scopeeconomies and product integration or focus on per-formance.

We use detailed information on microcomputersoftware product offerings in the early 1980s inorder to investigate the management of productvariety in the context of scope economies and inte-gration in software. We uncover that the manage-ment of product variety through the way differentapplication categories are integrated in productsand the platforms on which these products areoffered can be as important as the significanceof scope economies at the more aggregated firmlevel. Specifically, we find that the effects of scopeeconomies in production differed from effects ofscope economies in consumption: there is little evi-dence of firm benefits from economies of scope inproduction, but there is substantial evidence thatproducts benefit from economies of scope in con-sumption. In addition, we find that firms with prod-ucts that encapsulate more application categoriesperform better, and those with products that covermore computing platforms perform worse. Finally,changes in a firm’s product variety through newproduct introductions improve firm performance,but extensions to existing products hinder the per-formance of the firm and the product.

The remainder of this section provides a shorthistory of development in the microcomputer soft-ware industry in the 1980s. The next sectionreviews the literature and arguments underlyingdifferent relationships between product variety andscope economies. The section that follows pro-vides the definitions and hypotheses used in thestudy, and the subsequent section details the empir-ical approach, data and methodology. Our fifthsection describes our results, and the final sectiondiscusses the implications of our findings.

Development of microcomputer software in the1980s

During the early stages of the microcomputer soft-ware industry, product developers were convincedof the benefits of scope economies. But the chal-lenges of managing product variety in order toappropriate these benefits were significant, as illus-trated by the range of approaches taken during theearly life of this industry. Product design occurson two levels. First, it requires that firms deter-mine how to bundle capabilities such as addition,

multiplication, text manipulation, presentation ofgraphs and charts, search and replace commands,etc., in the development of software applications.Second, product design must determine how tomake these capabilities available. One option is asa stand-alone product in a single application cate-gory. Examples of the most common stand-aloneapplication categories circa 1980 were: spreadsheet(e.g., Visicalc), database (e.g., dBASE II), wordprocessors (e.g., WordPerfect) and programminglanguages (e.g., MS-BASIC). Another option is asproduct suites that include several application cat-egories.

In the early 1980s significant development effortwas required to extend capabilities and diversifyinto new application categories. Once the develop-ment of new capabilities was complete, they mightbe incorporated into the firm’s existing products orused to develop a stand-alone product in the newcategory. For example, when a firm offering onlya word processor decided to develop database-related capabilities (e.g., sorting and searching),it may have incorporated these functions into itsword processor, or offered a stand-alone databaseproduct. Scope economies in the production ofsoftware follows from the ability to spread fixeddevelopment costs across several product markets.(Nobeoka and Cusumano, 1997, evaluate the prod-uct development implications.)

Early in the first decade of competition, soft-ware developers anticipated significant economiesfrom combining different application categoriesinto integrated products. By 1983, microcomputersoftware developer Lotus had introduced its 1-2-3 package, combining capabilities in spreadsheet,database, programming, and graphics categoriesin a single package. Prior to 1-2-3’s introduc-tion, these capabilities were commonly providedin stand-alone products such as those noted earlier.Many developers sought to replicate Lotus’s mar-ket dominance by extending the features of alreadyestablished products. Ashton-Tate, for example,combined its successful stand-alone database prod-uct dBASE II with text and graphical capabili-ties to produce Framework. Adding capabilitiesfrom more categories is an example of increas-ing product-level scope coincident with increas-ing firm-level ‘scope’ at Ashton-Tate. In an effortto extend its market dominance Lotus introducedSymphony, which supplemented the 1-2-3 functionset with text processing and telecommunicationscapabilities. But the complexity of the product met

Copyright 2004 John Wiley & Sons, Ltd. Strat. Mgmt. J., 25: 1005–1025 (2004)

Product Variety and Firm Survival 1007

market resistance, suggesting that there are limita-tions to the extension of product breadth.

These limitations to more extensive integra-tion were anticipated by software developers whobelieved that customers would pay a premium forbroadly functional, interoperable applications in aseries of stand alone products designed to workeffectively in concert. These products could be soldeither individually or, as eventually occurred, in aproduct bundle or ‘suite.’ Hence, interoperabilityhas become an alternative to integration. A decadelater, product bundles such as Lotus’s SmartSuite,Corel Office, or Microsoft Office play an impor-tant role in competition. This strategy of interop-erability was also attempted early in the industry;for example, in Microsoft’s first applications. Inthe early 1980s, Microsoft designed its suite ofseparate spreadsheet (Multiplan), word processor(Multi-Word) and business graphics (Multi-Chart)products as the Multi-Tool set. Products sharedcommand terminology and menu structures, andMicrosoft planned a macro-language programmingtool that would improve interoperability. How-ever, this approach stalled and the term ‘multi’was retained only for Microsoft’s spreadsheet,Multiplan.

PRODUCT VARIETY AND SCOPEECONOMIES

Scope economies in production

In an overview of market structure in multi-productindustries, Bailey and Friedlaender (1982) empha-size firm operating characteristics in determiningindustry competitiveness. Their model shows thatfirm-level economies of scope are necessary formulti-product industries. In a contestable marketsmodel, multi-product firms better survive the threatof potential single-product entrants. These scopeeconomies arise as a result of excess capacity in theface of indivisibilities when diverse product linesbetter utilize sunk investments. This is consistentwith Montgomery and Wernerfelt (1988): ‘diver-sification is prompted by excess capacity in rent-yielding factors that are subject to market failure.’

Product family sharing of key components andassets is one method of obtaining scope economies.This takes the form of a product platform where thedesigns, components, and other assets are sharedby a set of products (Meyer and Utterback, 1993;

Robertson and Ulrich, 1998; Krishnan and Gupta,2001). Scope economies occur because platformsmake it cheaper to tailor products to segments, anddiversification results from building and extend-ing platform capabilities. That is, platforms reduceincremental costs of product variety yielding scopeeconomies. But as Krishnan and Gupta (2001)found analytically, use of platforms may result inthe overdesign of low end products. Not surpris-ingly, there is prior research suggesting that diver-sification correlates with poorer performance; firmsin the Montgomery and Wernerfelt (1988) modelthat elect to diversify most widely also expect thelowest average rents.

Transaction costs reasoning asserts that theremust be some reason to own, rather than rent,the excess capacity that leads to scope economies.Teece (1982) argues that multi-product firms arisewhen a variety of products demand similar produc-tion capabilities but firms are unable to indepen-dently secure access to requisite productive assets.Teece believes that a necessary precondition formulti-product firms is not technological indivisi-bilities but rather incomplete contracting for use ofthese assets. This rationale follows from standardtransaction costs arguments that if slack produc-tion capacity could be contracted without risk ofopportunism, the firm would rent, rather than own,the productive asset. If there are scope economiesin the production of software, then this reasoningmust apply to product development.

The application of transaction cost reasoning tothe economic organization of software firms estab-lishes a rationale for scope economies in the devel-opment of software. Weak appropriability in thedesign and implementation of software algorithmssuggests that product innovations will be ownedrather than licensed (see Teece, 1982, for the argu-ment). And, while the software community clearlyshared some innovations amongst developers (e.g.,Merges, 1992), other innovations were proprietaryand aggressively protected.1 If a software producerhas a significant cost advantage in programmingboth a spreadsheet and a word processor becauseroutines in one can be inexpensively added to theother, the firm should organize to take advantageof what is learned in the first product to reduce

1 For example, Wallace and Erickson (1992) describe the uniqueway that Lotus 1-2-3 uses DOS interrupts, something discoveredonly after significant reverse engineering by Microsoft operatingsystem developers.



the cost of the second. Because this knowledge islikely to be ‘sticky’ and difficult to transfer, there isan advantage to development within the firm. Thus,economies in the way software is produced yieldsscope advantages in the product development pro-cess. And, these economies should provide a costadvantage to firms that offer a greater productvariety over more markets, resulting in improvedproduct survival and profitability.

There are several reasons to believe that sucheconomies were important in the early softwareindustry. First, some extremely successful firmsprovided significant product variety. Microsoft,for example, offered operating system, program-ming, word processing, spreadsheet and game soft-ware all by 1981. Further, the Multi-Tool ‘suite’(Microsoft Word, Multiplan, and Chart) employeda common interface so that users could moreeasily access functional variety. Although somefirms retained a single product/single applicationstrategy, most eventually turned to greater vari-ety whereby most software firms offered multipleproducts in several application categories.

Many studies of scope treat production tech-nology differences homogeneously. For example,scope economies between any two SICs depend onthe ‘distance’ between the 4-digit SIC, ignoringinherent similarities in production technologies.But firms may employ a variety of technologieswith varying degrees of transferability to other seg-ment classifications. To incorporate heterogeneoustechnologies, we proxy differences in productiontechnology as choice of computing platform. Firmswith expertise in one platform (e.g., IBMPC) willnot necessarily have the same expertise in another(e.g., Apple Macintosh). The effect of produc-tion technology choice on firm performance canbe modeled with this record of the platforms onwhich the firm chose to offer products. Firms ableto produce software for a number of platforms arediversified against the risk that any individual plat-form becomes obsolete; offering products on moreplatforms gives the firm a lower risk of exit. Sim-ilar reasoning applies to individual product lines.

However, maintaining products across a widerarray of computing platforms consumes resourcesthat might be otherwise available for new productdevelopment. The argument that platform varietydepletes firm resources would not have been con-vincing to industry participants circa 1980 (seeSherman, 1984). There are several reasons to think,

however, that this strain on resources was impor-tant. As Boehm (1981) shows, software develop-ment for an operating system in which the firm hasno experience can introduce significant uncertaintyinto the process, degrading development perfor-mance. The complexities in tailoring a product fora specific platform mean that diversity for any par-ticular product is likely to be costly to maintain.The cross-level effect depends on the firm’s abilityto deploy technical resources in one product acrossits product portfolio.

The benefits of scope economies should notbe assumed to continue indefinitely. At both thefirm and product levels, greater product varietyleads to greater complexity in managing eachproduct. Because firms in this industry have lim-ited resources, particularly small entrepreneurialfirms, product-level implementation decisions bal-ance greater variety with increasingly complexproducts. This potential for heterogeneity makesdetailed knowledge of the firm strategy essential(see Merino and Rodriguez, 1997).

Scope economies in consumption

In addition to scope economies in production,there may be consumption economies of scopein this market, where customers prefer to pur-chase a variety of products from the same vendor.Although complementary products can be manu-factured independently, goods with strong techno-logical interdependencies are commonly producedby a single firm.2 Most theoretical models do notmake the distinction between scope economiesrelated to production and those that derive fromscope effects in consumption, primarily becausein economic theory cost minimization and profitmaximization are dual. In the software industry,though, this distinction is important. Consumers ofsoftware applications appear to have a preferencefor purchasing product variety from a single ven-dor. This effect results from the way that the firm

2 There are exceptions. When design specifications are welldefined and the efficient administration of contracts can beaccomplished cheaply, concerns over transaction governancecosts can be minimized. Argyres (1996) describes the use ofinformation technology in the manufacture of the stealth bomberby Lockheed, Boeing, and McDonald Douglas. The firms col-laborate on the wing technology by writing detailed descriptionsof each component, and requiring conformance with a softwareCAD model for each change.



coordinates product design in two facets of soft-ware development: a standardized command menuand functional interoperability.

A standardized menu facilitates the use of awide variety of software by providing users witha unified command structure. With appropriatecoordination of the menu, early entrants realizeddemand-side (consumption) economies of scopeby making a series of products that lower adop-tion costs of learning the firm’s other software. Forexample, the menu system for Microsoft productscontained common commands for file manipula-tion and printing. After adopting Microsoft Mul-tiplan, a user could more easily adopt MicrosoftChart, since the command structure was similar.This advantage to incumbents is clearly seen bythe magnitude of the opportunity cost a user incursin learning to use a software application, a costusually far greater than the purchase price of thesoftware. If this common command structure is dif-ficult to provide across a suite of products, varietyis better embodied in integrated products; integra-tion should provide an advantage for the samelevel of scope. If, on the other hand, it is betterto coordinate the menu structure, firms that offer asuite of products, but coordinate interface menusacross those products, should have an advantage;firm scope over product integration.

We believe that scope economies in softwarefollow from characteristics inherent in the waysoftware products are designed, created, and used.The drive for functional integration in the early1980s under products like Lotus Symphony orAshton-Tate’s Framework suggests that develop-ers believed in both broad category scope andcategory integration. Integrated products offeredgreater benefits to consumers because ease of inter-operability was valued more than depth in anyindividual function.

Consumption side economies are closely linkedto the consumer’s willingness-to-pay for interoper-ability; appropriating these benefits requires coor-dination. Hold-up in incomplete contracting to pro-vide functional interoperability (e.g., output from adatabase into a word processor) may lead to long-term contracting hazards as the products progressthrough upgrade cycles. This also will result inconsumer preference to purchase product varietyfrom an integrated producer. If the interoperabilityinterface is standardized, specialization economiesfavor products from separate producers, as seen in

the traditional stereo components industry.3 Thisalternative strategy was employed by manufactur-ers of non-integrated products, and required thedeveloper to provide interoperability between dis-parate products, perhaps across the boundary ofthe firm. Functional interoperability allows spe-cialized applications to share common data filesor output. For example, a retailer may store cus-tomer records in a database, produce charts ofpatron buying behavior, maintain mailing lists, andintegrate these results into a word-processing doc-ument. But with the rapid technological advancesin microcomputer hardware, attempts at industrystandard interfaces came and went.

Although this level of integration was eventu-ally facilitated by the operating systems level, inthe nascent software industry interoperability wasprovided by each developer, and rarely betweencompanies. For example, spreadsheet producersprovided that files could be saved in alternative for-mats readable by other applications: the exchangefile structure for Visicalc was known as DataInterchange Format (DIF). Microsoft’s Multiplanoffered a Symbolic Link format (SYLK) for fileexchange between other spreadsheets as well as toimport data into Microsoft Chart. Walden (1986)reports that the interest in inter-product data shar-ing became so important that Lotus placed propri-etary information describing the structure of 1-2-3files into the public domain. Although the OLE(Object Linking and Embedding) level of inte-gration was possible in MS-DOS computing, itwas not feasible for a large majority of consumersbecause of the level of sophistication it required ofusers.4

The trade-off between scope economies in pro-duction and consumption is highlighted inSorenson (2000), where managers must choosebetween efficient production of a small numberof products and maintaining a range of productstailored to consumer preferences. And the latterstrategy faces diminishing returns: broad productlines are increasingly costly and product variety isless valuable as the total number of products onthe market increases.

3 The literature is extensive, and specific works include Churchand Gandal (1992), Economides and Salop (1992), and Farrelland Saloner (1992).4 In today’s computing environment where the operating systemindustry structure is highly concentrated, OLE objects ‘drag-and-drop’ from one application to another in a seamless way.



Scope in industry evolution

In an industry evolving as rapidly as micro-computer software, some account of product andfirm change is needed. Scope advantages, andthe benefits of product variety more generally,occur as an industry evolves, and the firms andproducts must evolve with it. Traditional stud-ies of industry dynamics examine change from arelatively aggregated level, typically using snap-shots of diversification: Bailey and Friedlaen-der (1982), Evans (1987), Shapiro and Khemani(1987), Dunne, Roberts, and Samuelson (1988),Audretsch (1991), Mata and Portugal (1994),Lieberman (1989), Dunne and Hughes (1994),Baldwin and Gorecki (1994); an exception is Brud-erl, Preisendorfer, and Ziegler (1992). It is increas-ingly evident, though, that studies should accountfor the process as well as the outcome of diversi-fication strategies (see Barnett and Carroll, 1995).Firms that fail at attempts to diversify take greaterrisk, and should be treated differently than firmsthat make no effort to diversify.

Many diversification studies are cross-sectionaland do not give appropriate attention to orga-nizational change histories or market evolution.5

Because the microcomputer hardware market wasrapidly evolving, firm structures and capabilitieswere required to evolve with technological inno-vation. In most empirical work, diversification isboth a state and a process, and must be consid-ered in the context of market structure. In a rapidlydeveloping market, firms may remain undiversifiedbecause they are successful in their own marketsor unsuccessful in entering others. A longitudinalstudy records the attempted changes in the diversi-fication path in order to discriminate between thesealternatives. The degree of technological and mar-ket uncertainty in the software industry also makesit important to model attempts at change and theresulting effect on firm and product survival.

There is ample research to suggest that organiza-tional change imposes an exit risk on the firm, andthis has implications at both the firm and productlevel. At the firm level, an increase in product vari-ety depletes firm resources and may blur the firm’scompetitive profile. At a product level, increasingproduct functionality exposes the product to theuncertainty of competing in a new market. At least

5 Recent exceptions are Kesner and Sharma (1996) and Berghand Holbein (1997).

in the period of change, the product, and to someextent, the firm, face a greater risk of exit.

Difficulties in studying scope economies

It is puzzling that product diversification studiesoften fail to support scope economies, but thereare several explanations for why this might be thecase even if scope economies are important. First,any economic rationale for advantages to scopediffers across industries, and aggregation acrossSICs might blur these effects. Teece’s transactioncost argument for multi-product firms, for exam-ple, should be justified in the practice of softwaredevelopment rather than naively asserted for allindustries. Second, it may be that the studies ofscope at the aggregate business unit or corporatelevel are not accounting for product-level imple-mentation strategies. In forming hypotheses aboutscope economies in software, it is important toconsider the likely sources of these benefits inorder to determine the implications for firm andproduct strategy. This step is especially criticalin light of the array of implementation strategiesavailable to software developers, and calls for aresearch design that accounts for this heterogene-ity. Third, in a rapidly changing industry, transienteffects may overwhelm static efficiencies in scope,so that an effective research design must accom-modate industry evolution.

In many studies of diversification and scopeeconomies there is concern that managers are notacting in the interest of shareholders. Any researchon product variety must be concerned with prin-cipal/agent issues in risk diversification, becauseoptimal levels of diversification differ betweenshareholders and managers. As Amihud and Lev(1981) demonstrate, managers prefer to diversifypersonal risk at the expense of returns to sharehold-ers.6 The benefit to sale of two products is reducedvariability because sale of two products providesa smoother revenue stream than either individu-ally. This is commonly argued to be an advan-tage for capital allocation in diversified firms.Managers might be tempted to over-diversify tosmooth revenues at the expense of profits. Butin an entrepreneurial industry like microcomputersoftware, there are several reasons to believe that

6 Although some dispute this interpretation (see Lane, Cannella,and Lubatkin, 1998).



strategic decisions were made primarily by owner-managers, with the result that the firms in our studyare less likely to over-diversify under a princi-pal/agent rationale. Tarter (1996) reports revenuesand number of employees in Soft*Letter 100, thelargest microcomputer software firms measured byrevenues. These firms averaged only 90 employeesover 1985–86, the last year of our study. The bot-tom half of the list averaged only 20 employees.However, in the thousands of firms in our studyrevenues were on average far less than the top 100,and we conclude that firms in our study had evenfewer employees, and thus were closely monitoredby owners. Further, the information asymmetriesin high-technology products virtually guaranteesthat owners will be highly invested in product-linedecisions. For this reason, there is little concernthat managers will inefficiently diversify beyondprofit-maximizing levels merely to smooth the rev-enue stream and inefficiently prolong firm andproduct survival.

DEFINITIONS AND HYPOTHESES:SCOPE, INTEGRATION, AND FOCUS

Definitions

The alternative ways products incorporate capabil-ities from different application categories illustratethe complexities of the management of productvariety in the software industry. That is, firms notonly decide what capabilities to develop, but alsowhether to integrate capabilities of several appli-cation categories into a single product or maintainstand-alone products for individual application cat-egories. This is made even more complex by thedecision of whether to develop products acrosscomputing platforms, develop integrated productsfor some platforms and stand-alone products forothers, or concentrate on a subset of platforms.

We define a niche as a specific application cate-gory on a specific computing platform. A productmay cover a number of niches by including thefunctionality of several application categories, andpossibly multiple computing platforms.7 Thus, theterms ‘category,’ ‘platform,’ ‘niche,’ and ‘prod-uct’ have separate and specific meanings in ourresearch context. In this context, we use the terms

7 ‘Computing platform’ has a distinct meaning different from‘product platform’ discussed in the previous section.

‘category scope’ and ‘platform scope’ to refer tothe breadth of application categories or platformsin which the product or firm competes. We use theterm ‘category integration’ to refer to the range ofapplication categories a particular product incorpo-rates. Finally, we use the term ‘platform focus’ torefer to the range of computing platforms on whicha particular product is offered. Figure 1 shows howthese terms are used.

At the firm level Firms 1, 2, and 3 all havethe same category scope and platform scope asthey each offer software in three application cate-gories and for three computing platforms, but withdifferent category integration and platform focus.Firm 1 offers a product that incorporates database,graphics, and word processing, and has high cate-gory integration. In contrast, Firm 2 offers separateproducts for each of these categories and has lowcategory integration. Similarly, Firm 1 has a broadplatform focus and Firm 2 has a narrow platformfocus. Firm 3’s products have medium categoryintegration and an intermediate platform focus.Firm 4 has both narrow category and platformscope, and a necessarily low category integrationand narrow platform focus.

At the product level Firm 1’s Product 1 has abroad category scope and platform scope as theproduct incorporates the three application cate-gories and is offered on the three platforms. Incontrast, Firms 2 and 4’s products have narrowcategory scope and platform scope. Firm 3’s prod-ucts have intermediate category scope and platformscope.

Hypotheses

Our hypotheses concentrate on scope economies,category integration, platform focus, and prod-uct variety. As we discussed above, economiesof scope in production are supported by stan-dard economics in a multi-product firm wherebyit is possible to spread fixed costs of develop-ment across multiple products. They are also sup-ported by transaction cost theory, which argues thatbecause of incomplete contracts governing knowl-edge assets that can be used for more than onesoftware product, firms organize to own that assetand produce multiple products. In addition, thereis the more recent literature on product platforms,whereby designs, components, and other assets areshared by a variety of products.



CP/M

Product Category Integration and Firm Category Scope

Product 1

Product

Firm 1

Product 1Firm 4

Product 1Product 2Product 3

Firm 2


Firm 3

Firm

Product Platform Focus and Firm Platform Scope

Database Graphics Wordprocessor

Product 1

Product

Firm 1

Product 1Firm 4


Firm 2


Firm 3

Firm MS-DOS Apple II

Figure 1. Illustration of product integration and firm scope

Hypothesis 1: There are scope economies in theproduction of microcomputer software.

Pure scope economies in production should resultin a positive impact of broad category scope andplatform scope that a firm offers in its productportfolio. Broad category scope and platform scopeat the firm level should contribute both to firmperformance and to product performance. Becausescope economies in production occur at the firmlevel, there is not a necessary connection betweenindividual product survival, and broad categoryscope and platform scope at the product level.As we further discussed above, the principal/agent

issues in risk diversification—which would tendto increase the number of products a firm offersin absence of scope economies in production—areless critical in the microcomputer software industrybecause the firms are either owner-managed, or aresufficiently small for owners to monitor employeeproduct offering decisions closely.

On the demand side there are reasons to believethat there are also scope economies in consump-tion. During the early years of the microcomputersoftware industry, interoperability and standardswere more likely to apply to products from thesame firm. Consumers put a high premium on thesefeatures because they simplified computer work by



being able to move readily between applications,and they significantly reduced consumers’ costs oflearning and training.

Hypothesis 2: There are scope economies in theconsumption of microcomputer software.

Scope economies in the consumption of microcom-puter software are more likely to be product-basedrather than firm-based because of superior inter-operability. These scope economies should resultin broad category scope and platform scope at theproduct level having a positive impact on productsurvival. In other words, if there are consumptionscope economies in the software industry, firmsthat offer functional variety in application cate-gories (e.g., both a word processor and a spread-sheet) in a single product should achieve better per-formance than rivals that offer single applicationcategory products. If scope economies in consump-tion occur between products, then broad categoryscope and platform scope at the firm level may alsoincrease product survival, and should increase firmsurvival.

A naive approach tests the effect of productvariety on firm survival without consideration forthe way this variety is bundled. To the extentthat scope economies in consumption are basedon compatibility across categories and platforms,these economies should also result in a positiveimpact of high category integration and of broadplatform focus on product survival. Because firmsoffered functional variety in application categorieseither at the product level (as integrated products)or at the firm level (through a common user inter-face), our approach must accommodate both levels.If a product-level approach to scope were moreeffective, products that integrate application cate-gories would outperform less varietied rival prod-ucts and contribute to firm survival. If the cost ofintegrating functional variety across several soft-ware products (as opposed to an integrated prod-ucts) is prohibitive during this time-frame, then anincreased number of products should not enhancefirm survival.

Interoperability and standards between differentapplication categories can be implemented in twoways. One is through different products in differentsets of application categories offered by a singlefirm, where the fact that the products are producedby the same firm allows for a common interfaceand data exchange. The other is through products

that cover larger numbers of application categoriesso that interoperability and a common interfaceare incorporated as part of the integrated prod-uct design. The former argues for greater productvariety through numbers of products, and the lat-ter would result in fewer products but with higherlevels of integration.

Hypothesis 3: Integration over application cat-egories yields higher firm performance.

The advantages of integration over multiple prod-ucts in delivering product variety should resultin higher category integration being positivelyassociated with firm survival and product sur-vival, while controlling for category scope. Thatis, holding category scope constant, fewer prod-ucts, but integrated over application categories,should enhance firm performance. Hypothesis 3 isstated so that the alternative to Hypothesis 3 is alsotestable—implementing product variety throughmultiple products rather than integration enhancesfirm performance.

As we discussed earlier in our overview of scopeeconomies in production, there is evidence thatmaintaining products across computing platformscan be resource-intensive due to the fundamentaldifferences in operating systems. This is consistentwith the higher costs of deriving product variantsfor multiple product platforms, rather than derivingproduct variants from a single platform.

Hypothesis 4: Focus on a smaller number ofcomputing platforms yields higher firm perfor-mance.

The advantage of focus through offering individ-ual products over a smaller number of platformsshould result in narrower platform focus havinga positive impact on firm survival and productsurvival, holding platform scope constant. Thus,a greater proportion of products that are offeredover multiple computing platforms should lowerfirm performance.

Scope and industry evolution suggest that in-creasing product variety and new product intro-ductions are important for firms to maintain theirproduct diversification advantages. It also allowsfirms to dynamically adjust strategies such as inte-gration vs. multiple products, and follow industryevolution in application categories and comput-ing platforms. In addition, accounting for industry



evolution is an important control when examiningeffects such as those in our first four hypotheses.

Hypothesis 5: Changes in product variety in-crease firm performance.

Adjustments in product variety, scope, integration,and focus strategies are adjustments in firm-levelstrategies. As measured by entry into new prod-ucts, application categories, computing platforms,etc., these entries should have a positive impact onfirm survival. To the degree that these new entriesdo not compete with existing products, they shouldalso have a positive impact on product survival.

However, product changes and introductions arerisky strategies that often negatively impact per-formance. Clark’s (1987) concept of ‘transilience’relates to the degree of transformation in technol-ogy and markets resulting from innovation, wherethe greater the technology and market transfor-mation for the firm, the greater the transilience.Product entry into new-to-the-industry applicationcategories and computing platforms are architec-tural innovations—radical technology applied tonew markets. Product entry into new-to-the-firmapplication categories and computing platforms areniche creation innovations—modifications in tech-nology applied to new applications and new cus-tomer groups. Product entry into existing appli-cation categories and computing platforms arerevolutionary innovations—changes in technologyapplied to existing markets and customers. Productrefinements such as upgrades are regular inno-vations. Architectural innovations are highest intransilience, followed by niche creation and rev-olutionary innovations, followed by regular inno-vations. Using a transilience scale developed byAbernathy, Clark, and Kantrow (1983) to mea-sure innovation, Geroski and Mazzucato (2002)found that innovation did not significantly affectfirm growth in automobile manufacturing, suggest-ing that firm performance may not be related tochanges in product variety.

EMPIRICAL APPROACH, DATA, ANDMETHODOLOGY

Empirical approach

We study the effects of scope economies, integra-tion, focus, and product variety at both the product

level and at the firm level. Use of multiple levelsof analysis is important in understanding the trade-offs at both the firm and product levels, and theeffect of firm-level decisions on individual prod-uct performance. Product and firm survival areused as our two measures of performance. Sur-vival is a particularly relevant measure of perfor-mance in high-technology industries, having beenused in studies of the computer workstation indus-try (Sorenson, 2000), among others. In a time ofrapidly expanding markets, participation in 1 yearof the industry can be seen as an option to con-tinue in later, more profitable, years. Exit from theindustry identifies firms for which the option tocontinue was worth less than the opportunity costof remaining in the industry.

Data

Two data sources are used to provide data forour analysis. Detailed information on individualsoftware product data was assembled from TheSoftware Catalog: Microcomputers published byElsevier Publishing (1980).8 The catalog lists thou-sands of software packages available by hundredsof firms in the U.S. market. The data cover theyears from 1981 through 1986 at nine catalogpublication dates. This time period is a criti-cal period in the microcomputer software indus-try since it spans the pre-IBM PC era throughthe introduction of the Apple Macintosh. Duringthis period the number of products available grewfrom 3700 to 13,000 and the number of compa-nies from 500 to 6000. In The Software Cata-log: Microcomputers each product is classified asincorporating one or more specific application cat-egories and as offered over one or more computingplatforms. We combined some of the classifica-tions from The Software Catalog: Microcomput-ers for our analysis, for example, text-processingand spell-checking software are combined withword processing. Table 1 lists the application cat-egories and platforms used in the classification.Use of various definitions of category and plat-form does not qualitatively alter the results. Forexample, using category classifications of wordprocessors, database, graphics, operating systems,spreadsheets, telecommunications, programming,games, personal, and accounting did not changethe qualitative results of our analyses.

8 The 1981/82 data were published by Imprint Software Ltd.



Table 1. Software classification definitions

Category Platform

Educational Apple IIAccounting IBM—personal

computerVertically integrated

industry softwareDigital research CP/M

familyEntertainment Radio Shack/TandyProgramming Victor/Commodore/PETOperating system software Other Intel family of

microprocessorsBusiness administration AtariMiscellaneous personal

softwareBASIC language

programsEngineering tools Hewlett-Packard family

of microsWord processors Digital Equipment

CorporationCommunications software

emulatorsNorthstar family of

microcomputersDatabase Other IBM (non IBMPC)Graphics AltosMisc. scientific MotorolaInformation retrieval Texas InstrumentsOther

To provide for a measure of market size formicrocomputer software, product market data ondollar microcomputer sales collected by the Inter-national Data Corporation (IDC) is used as a proxyof the potential market for software products.

Dependent variables

Performance is measured by product survival andfirm survival. Observations for a product (or firm)begin when the product (or firm) first appearsin The Software Catalog: Microcomputers. Theseobservations continue until the product (or firm) nolonger appears in The Software Catalog: Micro-computers. The two variables that measure sur-vival are Product Exit and Firm Exit. These arecoded as zeros during the observations when theproduct (or firm) appears, and are coded ‘1’ whenthe product (or firm) exits.

There are at least three reasons why survival isa good measure (Sorenson, 2000). First, survival isa necessary condition for positive profits. As such,survival is really a threshold measure. Second,survival (or exit) can be observed for all productsand all firms. Third, exit directly measures poorperformance.

Operationalizations of scope

Category Count and Platform Count are used tooperationalize category scope and platform scope,at both the product and firm levels. At the productlevel these variables are the number of applica-tion categories (e.g., word processor, spreadsheet)in which a given product is classified, and thenumber of computing platforms (e.g., Apple II,Apple Macintosh, IBM-PC) that a product is clas-sified under. At the firm level these counts are thenumber of application categories in which the firmhas a product classified, and the number of com-puting platforms under which the firm has at leastone product classified.

A Niche Count is used to operationalize an addi-tional dimension of scope at the product level. Thisis the number of niches, defined as the intersectionof category and platform, in which the product isclassified. A Product Count is used to operational-ize our final dimension of scope at the firm level.This is the number of products that the firm hason the market.

Operationalization of integration and focus

The operationalization of integration and focususes two measures of entropy: category and plat-form entropy. Entropy measures have long beenadvocated in studies of diversification, but rarelyapplied to product-level variety in the way nec-essary here.9 An exception is Alexander (1997),who measured the relationship between industrystructure and product variety in the music industry,measuring variety by an entropy index.

Category Entropy is used to operationalize cat-egory integration and Platform Entropy is usedto operationalize platform focus, both at the firmlevel. We use the standard method of calcu-lating entropy used elsewhere (e.g., Alexander,1997), where at the product level the calculationis − ∑

[pi ln(pi)], where pi is probability that theproduct is classified in a particular category (plat-form) and the sum is over the categories (plat-forms). The average over the firm’s products is theCategory (Platform) Entropy. Category entropymeasures the tendency of the firm to offer inte-grated functionality such that firms that tend tooffer products integrated over multiple categories

9 Jacquemin and Berry (1979), Palepu (1985), and Hoskissonet al. (1993) discuss the use of entropy measures in firm-leveldiversification.



have greater category entropy. Platform entropymeasures the extent to which firms tend to offer aproduct for several platforms such that firms thattend to offer each product across several platformshave greater platform entropy.

Using platforms to illustrate the distinction bet-ween a count and the entropy measure, supposethat two firms offer products for the same numberof platforms, but one firm offers each product foronly one platform, while the other offers eachproduct for several platforms. The firm offeringeach product for only one platform has a lowentropy measure and the other has a high entropymeasure.

Operationalizations of changes in scope

To operationalize the dynamics of scope changesat the product level three counts are used. NewCategory (or Platform) Entry is a count of thenumber of new categories (or platforms) the prod-uct was made available for during the ‘spell’—thatis, during successive publications of The SoftwareCatalog: Microcomputers. In addition, New Cat-egory and Platform Entry is a count of the newcategory/platform combinations the product wasmade available for during the spell.

New Category Entry and New Platform Entryfor products are also aggregated to the firm level.At the firm level we have three additional vari-ables. New Product Entry is a count of new prod-ucts introduced. New Product in New Category andNew Product in New Platform are counts of cat-egories and platforms in which new products areintroduced, summed over the new products.

Control variables

Control variables are used to account for industryand competitive effects. To account for differencesin overall software market potential and oppor-tunity across periods, Microcomputer Sales (in $millions) is used as a measure of market size.

To account for the competitive environment aset of different counts of products classified in thesame niche, category, and platform are used asmeasures of competition. The intuition for thesemeasures is that customers make a series of nestedbinary comparisons, where more firms or productsin the relevant market implies, all other thingsequal, greater competition for each product. Atthe product level Products in Niche is a count

of products in the same niche: for example, allword processor applications for the IBM PC, orall spreadsheets for the Apple II. If the productis classified in multiple niches, then Products inNiche is a count over the different niches. Productsin Category and Products in Platform are a countof the number of products in the same categoriesor platforms, respectively.

At the firm level Firms in Category and Firmsin Platform are a count of the number of firms inthe same categories or platforms, respectively. Inboth the product-level and firm-level measures, noadjustment is made for a competing product or firmthat is classified in more than one of the same cat-egories or platforms. Although this may cause ourmeasures to be overstated relative to measures suchas the ‘number of competing products or firms,’classification of a competing product in more thanone category or platform means competition acrossthose categories or platforms.

Table 2 provides descriptive statistics on theproduct-level and firm-level variables. To deter-mine if multicollinearity is a problem we examinedthe correlation matrix’s condition number to deter-mine if it was ‘ill-conditioned’ and it is not. Inaddition, the effect of higher correlations would beto understate the significance of our coefficients, soour results are above any variance inflation factor.

Methods

We employ event history methods to model thesurvival of software products. Petersen (1991) rec-ommends a logit specification of spell-splittingwith time-varying covariates to estimate non-pro-portional hazards, and Yamaguchi (1991) recom-mends this non-proportional hazards approach forthe continuous time models frequently employedin these kinds of studies. Following these rec-ommendations, we use a discrete-time logit forour duration equation. Details of the mathematicalform are in Yamaguchi (1991: 18–19). Survival isa censored observation (=0) and exit is an event(=1); however, for interpreting the results we havereversed the signs to conform with intuition—inthe tables, positive coefficients correlate with sur-vival (a decreased exit probability) and negativecoefficients correlate with increased exit. When-ever the discussion identifies exit risk, it shouldbe understood to mean that the probability of exitincreases with a decrease in the covariate.



Table 2. Descriptive statistics

Variables Mean S. D. Min. Max.

Product level : N = 117 ,242Product Exit 0.012 0.220 0 1Left Censored 0.270 0.896 0 1Product Age − 1 Year = 100 148.207 126.285 25 550Category Count 1.168 0.410 1 10Platform Count 1.837 1.536 1 13Niche Count 2.133 2.168 1 41New Category Entry 0.042 0.213 0 3New Platform Entry 0.139 0.688 0 10New Category and Platform Entry 0.012 0.109 0 1Products in Category 6147 5549 18 38,018Products in Platform 1229 1627 1 12,926Products in Niche 2271 1844 19 12,700Microcomputer Sales ($M) 7.738 4.615 2.390 9.704

Firm level : N = 22 ,210Firm Exit 0.027 0.163 0 1Left Censored 0.182 0.386 0 1Firm Age − 1 year = 100 166.282 131.138 25 550Product Count 3.987 5.860 1 50Category Count 3.032 1.655 1 16Platform Count 3.071 1.513 1 12Category Entropy 0.099 0.213 0 1.6094Platform Entropy 0.437 0.561 0 2.4849New Product Entry 0.265 1.309 0 20New Category Entry 0.137 0.502 0 11New Platform Entry 0.186 0.659 0 8New Product in New Category 0.090 0.583 0 14New Product in New Platform 0.113 0.888 0 20Firms in Categories 4047 1477 273 12,085Firms in Platforms 5138 2101 48 11,725

The statistical analysis below provides survivalmodels of firms, products, and the effects of firm-level covariates on product survival. We use twomethods for model selection when there is con-cern over multicollinearity. When the number ofcovariates is relatively small (e.g., less than 15),we employ the best subsets approach advocated inSen and Srivastava (1990). For a larger numberof covariates, the computational time was signif-icantly higher, and the ‘selection method’ (iter-atively applied forward and backward selection)was employed to determine covariate inclusion.After a set of coefficients was selected, these wereincluded across models to provide a consistent setin the discussion.

RESULTS

Age and survival

In our firm-level analysis Model 1 in Table 3shows firm survival with age of the firm as the

only covariate. In the absence of information aboutfirm heterogeneity, older firms are shown morelikely to survive. Later models, however, showthat in the presence of firm scope, integration andfocus, changes in scope, market size, and compet-itive environment, age of the firm increases theprobability of exit. Many studies of organizationalmortality indicate that older firms have a survivaladvantage, but it has also long been suspectedthat unobserved heterogeneity might account forthis (see Bruderl et al., 1992; Mayer, Hamerle,and Blossfeld, 1989; Mayer and Tuma, 1990). Inthe presence of covariates modeling heterogene-ity, the true effect in our data is a liability ofobsolescence, a result consistent with expectationsabout high-technology industries. Model 2 of thetable includes more detailed information on firmheterogeneity. Model 3 is a statistically signifi-cant improvement over Model 2 and, since theresults are consistent, our discussion of the firm-level analysis draws on Model 3.



Table 3. Firm survival

Variables Model 1 Model 2 Model 3

Coeff. Standard error Coeff. Standard error Coeff. Standard error

Constant 3.663 0.038∗∗ 2.270 0.100∗∗ 2.412 0.104∗∗

Left Censored 0.443 0.072∗∗ 0.426 0.073∗∗

Firm Age 0.0008 0.0002∗∗ −0.003 0.0003∗∗ −0.003 0.0003∗∗

ScopeProduct Count −0.001 0.006 −0.019 0.006∗∗

Category Count −0.143 0.029∗∗ −0.307 0.034∗∗

Platform Count 0.048 0.023∗ 0.060 0.024∗

Integration and focusCategory Entropy 0.390 0.115∗∗ 0.488 0.115∗∗

Platform Entropy −0.756 0.072∗∗ −0.789 0.073∗∗

Product varietyNew Product Entry 0.070 0.023∗∗ 0.058 0.023∗

New Category Entry −0.158 0.058∗∗ −0.136 0.058∗

New Platform Entry −0.031 0.020 −0.048 0.021∗

ControlsFirms in Categories 0.0002 0.00002∗∗ 0.0003 0.00002∗∗

Firms in Platforms 0.0002 0.00002∗∗ 0.0002 0.00002∗∗

Microcomputer Sales($M) 0.1693 0.02530∗∗ 0.1815 0.02580∗∗

Administration 0.037 0.030Education 0.242 0.057∗∗

Entertainment 0.070 0.027∗

Operating Systems 0.260 0.071∗∗

Vertical Industry Applications 0.289 0.079∗∗

Altos −0.087 0.018∗∗

Digital Research-CP/M 0.097 0.014∗∗

IBM PC 0.035 0.025North Star 0.079 0.023∗∗

Tandy/Radio Shack 0.053 0.012∗∗

Likelihood Ratio χ 2 18 1143 1308d.f. 1 13 23Concordant 31% 67% 68%Discordant 43% 30% 28%Tied 26% 4% 3%

Statistical significance: ∗∗ 1%; ∗ 5%

In our product-level analysis shown in Table 4,Model 1 reports product obsolescence with age asthe single covariate. Model 2 shows a full parame-terization of product survival, and Model 3 showsthese effects in the presence of firm-level covari-ates. Inclusion of firm-level covariates significantlyimproves model fit, even after accounting for theincrease in model degrees of freedom; firm-leveleffects are important for modeling product sur-vival. As with firms, products show a liability ofobsolescence.

The age of the firm improves an individual prod-uct’s likelihood of survival. Firm age may sig-nal the likelihood of longer-term product upgradesand maintenance. Customers may be more willing

to sink the investment in learning new softwarewith some assurance that the investment will berecouped by future upgrades or availability on newplatforms. The effect of this covariate is oppositefrom the firm level of analysis, where firms aresubject to a liability of obsolescence.

Scope and survival

At the firm level of analysis, Model 3 in Table 3,the coefficient of product count shows a greaterprobability of exit for firms with more products,and the coefficient of category count shows agreater probability of exit for firms that have prod-ucts in a greater number of application categories.



Table 4. Product survival with firm-level covariates

Product-level variables Model 1 Model 2 Model 3

Coeff. Standard error Coeff. Standard error Coeff. Standard error

Constant 3.339 0.015∗∗ 1.9630 0.054∗∗ 2.5701 0.058∗∗

Left Censored 0.2927 0.025∗∗ 0.2995 0.029∗∗

Product Age −0.082 0.0068∗∗ −0.2100 0.0094∗∗ −0.4070 0.015∗∗

ScopeCategory Count 0.2668 0.030∗∗ 0.2218 0.032∗∗

Platform Count −0.0428 0.016 0.0548 0.018∗∗

Niche Count −0.0574 0.009∗∗ −0.0618 0.010∗∗

Product varietyNew Category Entry −0.5043 0.037∗∗ −0.4739 0.03760∗∗

New Platform Entry 0.0094 0.010 −0.0522 0.01290∗∗

New Category and Platform Entry 0.2611 0.077∗∗ 0.2323 0.07760∗∗

ControlsProducts in Category 0.0001 0.000∗∗ 0.0001 0.000∗∗

Products in Platforms 0.0001 0.000∗∗ 0.0001 0.000∗∗

Products in Niche −0.0001 0.000∗∗ −0.0001 0.000∗∗

Microcomputer Sales ($M) 0.1222 0.006∗∗ 0.1150 0.006∗∗

Firm-level variablesFirm Age 0.0021 0.0002∗∗

ScopeProduct Count −0.0136 0.001∗∗

Category Count −0.0710 0.006∗∗

Platform Count −0.0686 0.005∗∗

Integration and focusCategory Entropy 0.4810 0.082∗∗

Platform Entropy −0.3067 0.032∗∗

Product varietyNew Product Entry −0.0332 0.003∗∗

New Category Entry −0.0634 0.015∗∗

New Platform Entry 0.0887 0.009∗∗

New Product in New Category 0.0339 0.005∗∗

New Product in New Platform 0.0082 0.002∗∗

Likelihood Ratio χ 2 140 3135 5229d.f. 1 12 23Concordant 51% 62% 66%Discordant 33% 35% 31%Tied 17% 3% 2%

Statistical significance: ∗∗ 1%; ∗ 5%

Both of these results are statistically significantand indicate diseconomies of scope in the produc-tion of microcomputer software at the level of thefirm. The coefficient for platform count is posi-tive, correlating with a decreased probability ofexit for firms that have products under a greaternumber of computing platforms. Although this issuggestive of economies of scope in computingplatforms, this result is less statistically significantthan that for products and categories. In addition,the coefficient is only a fifth the size of the coef-ficient for category count, yielding a mean effect

that is less than a third of the mean effect of cat-egories.10

Examining the firm-level variables at the productlevel of analysis, Model 3 in Table 4, the coeffi-cients of product count, category count, and plat-form count are all negative and statistically signif-icant. This indicates that the exit of any individualproduct is more likely when the firm has multi-ple products, any individual product is less likely

10 Using the means for the counts from Table 2, multiplied bythe coefficients from Table 3.



to survive when the firm has products in morecategories, and when the firm has products undermore platforms. Thus, in terms of product survival,there are diseconomies of scope at the firm levelin products, categories, and platforms. With ref-erence to our first hypothesis (Hypothesis 1), wecan conclude that there are not scope economiesacross products or application categories in theproduction of microcomputer software. There maybe positive but small scope economies in pro-duction across computing platforms at the firmlevel, although individual product survival doesnot benefit from these economies. This difficulty inmanaging product market scope fits with the pop-ular sense that industry participants were far moreinterested in computing technology than consumermarkets. Understanding the emerging markets forsoftware applications, measured by the scope ofproduct markets, appears to have been a challengeto firm survival.

At the product level of analysis (Model 3,Table 4), the product-level variables for categorycount and platform count have coefficients thatare positive and significant. This indicates thatproducts which cover a greater number of applica-tion categories and products which operate undera greater number of computing platforms eachhave a greater probability of survival. This pro-vides evidence in support of our second hypothe-sis (Hypothesis 2): at least at the product level,there are scope economies in the consumptionof microcomputer software whereby products thathave functional variety that covers more applica-tion categories and products that can be used onmore computing platforms are more likely to sur-vive. The effect on survival of a product coveringa greater number of niches (categories by plat-forms) is negative—a higher niche count reducesthe probability of product survival. This reflectslimits of scope economies of consumption. How-ever, it is worth noting that the mean effect ofcategory count is almost twice that of niche count,so that the combined effects of these measures ofproduct scope increase the probability of productsurvival.11

Integration, focus, and survival

As detailed earlier, we use measures of entropyto operationalize category integration and platform

11 Coefficients are from Table 4.

focus, and these are firm-level measures: larger cat-egory entropy indicates that the firm tends to offerintegrated products, and a larger platform entropycorresponds to the firm offering each product on alarger number of computing platforms.

In the analysis of firm survival (Model 3,Table 3) the coefficient for category entropy ispositive and significant. This indicates that greatercategory entropy increases the probability of firmsurvival. That is, firms that offer integrated prod-ucts have an increased chance of survival. In theanalysis of product survival (Model 3, Table 4)the coefficient of firm-level category entropy isalso positive and significant. So, in addition tofirm survival, greater category entropy increasesthe probability of product survival. Thus, whenthe firm tends to offer integrated products, eachindividual product has an improved survival prob-ability.

Further support for the argument that consumersprefer integrated products over interoperable suitescomes from the product-level category count inModel 3, Table 4. The positive coefficient indi-cates that products with greater category scopehave a greater probability of survival, at least inthe time-frame of our study. These results sup-port our third hypothesis (Hypothesis 3)—namelythat integration over application categories yieldshigher firm performance.

In contrast to the result on category integration,the negative and significant coefficients for plat-form entropy in the firm-level analysis in Table 3(Model 3) and in the product-level analysis inTable 4 (Model 3) show that greater platformentropy correlates with increased exit both forfirms and for products. Therefore, for firms whichoffer products over a greater number of computingplatforms, the individual product is less likely tosurvive, and the firm is less likely to survive.

The results on platform entropy are consistentwith results of firm platform scope at the productlevel (Model 3, Table 4)—products from firmsthat cover a greater number of platforms are lesslikely to survive. These results support our fourthhypotheses (Hypothesis 4) that a firm’s focus ona smaller number of computing platforms yieldshigher performance.

Change in product variety and survival

We begin with the firm survival analysis of Model3 in Table 3. At the firm level there are two types



of adjustments to product variety: new productintroductions and extensions of existing productsto new application categories and/or to new com-puting platforms. The coefficient of new productentry, reflecting the number of new product intro-ductions, is positive and significant. This indi-cates that the greater the number of new prod-ucts introduced, the greater the probability of firmsurvival.

However, the coefficients of new category entryand new platform entry are both negative and sig-nificant. These variables reflect the firm’s exten-sion of one or more of its existing products to newcategories and to new platforms (new to the firm,not new to the industry), and these product exten-sions are correlated with firm exit. At the level ofanalysis of the firm our fifth hypothesis (Hypoth-esis 5) has equivocal support: changes in productvariety in the form of new products increases firmperformance, while product extensions decreasefirm performance.

In the analysis of product survival in Model3 of Table 4, the product-level variables of newcategory entry and new platform entry also havenegative coefficients, indicating that extending theproduct to a new application category or to anew computing platform is correlated with productexit. The coefficient for new category and platformentry is significant and positive, but when com-bined with the negative (exit) effects of both newcategory entry and new platform entry, the overalleffect is a lower probability of product survival.Thus, at the product level, extensions into newcategories and new platforms are correlated withproduct exit.

The firm-level variables in the product survivalanalysis indicate that individual product exit ismore likely in a period when the firm introducesa new product, and this is accelerated when thenewly introduced product is a diversification intoa new category. The firm-level effect of entering apreviously unsupported platform improves productsurvival probabilities. The coefficient signs sug-gest that new products cannibalize old ones, sothat unless the firm appeals to new set of cus-tomer needs on a different platform, new prod-uct introductions lead to product exit. More dra-matic diversification by producing a new prod-uct into a new category or onto a new plat-form correlates with improved survival—such anew product entry would not be a competingproduct.

Controls

In the firm survival analysis (Model 3, Table 3) thepositive and significant signs of the coefficients ofthe environmental variables show that the largerthe number of firms in categories and the greaternumber of firms in platforms, the less likely is firmexit. Moreover, the greater the size of the market(measured by microcomputer sales), the less likelyis firm exit. This suggests that the market sizewas growing faster than the suppliers of softwareover the duration of the study, so much so thatadditional competitors was an indication of a goodcategory or good platform to have products in.

In the product survival analysis (Model 3, Table4), more products in category and more prod-ucts in platform correlates with increased productsurvival, and more products in niche correlateswith decreased product survival. At this productlevel, products in niche more precisely measuresinter-product competition. The negative sign of theproducts in niche coefficient reflects the greaterexit probability when there are more products inthe same functional categories on the same plat-forms. This is analogous to local and non-localdensity measures (see, for example, Baum andKorn, 1996; Baum, 1995; Baum and Singh, 1994).

DISCUSSION AND IMPLICATIONS

This study contributes to our understanding ofscope, integration, focus, and product variety. Theregressions tell a consistent story about scopeeconomies: although firm-level scope measures donot provide evidence of scope economies in pro-duction, at the product level they provide sup-port for scope economies in consumption. In addi-tion, the dispersion (entropy) measures demon-strate that category integration and platform focusare important: high application category integra-tion increases survival of the firm and the firm’sproducts, and similarly for a narrow computingplatform focus. Finally, adjustments in productvariety through new product introductions andproduct extensions have distinctly different impli-cations for firm and product survival: the introduc-tion of new products increases the probability offirm survival, and extensions of existing productsto new categories and new platforms reduce thelikelihood of both firm and product survival.

The contrast between the effects of our opera-tionalizations of scope at the firm level—increased



probability of firm exit with greater product scopeand category scope—and at the product level—in-creased probability of product survival with greatercategory scope and platform scope—are stark. Webelieve this confirms the distinction between ourfirst two hypotheses. Scope economies in produc-tion, our first hypothesis, should affect firm sur-vival by spreading fixed production costs acrossproducts, and should not affect product survival,as the allocation of those cost savings to individ-ual products is arbitrary. Scope economies in con-sumption, our second hypothesis, are demand-sideeffects where the impact is on individual productsurvival. This simple result has profound impli-cations for the structure of the software industry.During the period of this study, there was a strongtendency to integrate product functionality througha single user interface. This imperative contin-ues as new products and technologies evolve inthe micro-computer market, where users clearlyprefer seamless integration of functionality. Asthe arbitrator of the operating system, Microsofthas aggressively participated in this evolution. Wehave seen this drama play out most clearly inthe anti-trust arena, where product integration andtechnology evolution met in the Microsoft anti-trust trial.

Implications of our study beyond the softwareindustry certainly depend on the nature of the prod-uct variety decisions, but our results capture anumber of difficult trade-offs that firms and man-agers face in managing their portfolio of products.For example, our results indicate that introducingnew products increases the probability of firm sur-vival. But, in spite of being less costly and possiblyquicker to implement, the extension of existingproducts to new platforms reduces both productand firm survival. These results are consistent withpreemption, whereby successful firms are forcedto ‘eat their own lunch’—that is, launching newproducts that cannibalize existing profitable (andpossibly extendable) products—in order to main-tain market leadership (Nault and Vandenbosch,1996, 2000).

Scope economies have always been importantto software, but the integration decision criti-cally determines the extent to which users accessthis scope. As Porter (1985) notes, the abilityof customers to assemble their own bundles isan important determinant of the extent of scopeeconomies realized. The strength of integratedproducts over the time-frame of the study was in

two key attributes: command interface and inter-operability. No matter how well developed theinterface to different products, during the era ofMS-DOS the extent of product interoperabilityrequired significant user expertise. It is ironic thatthe industry has turned to product suites at atime when command interfaces and interoperabil-ity are increasingly standardized. Conformance tostandardized command structures appears to befar stronger under a graphical user interface thanunder character-based operating systems. Func-tional interoperability has also been dramaticallystandardized through Object Linking and Embed-ding (OLE). Although OLE was a long-standinggoal of operating system developers, its implemen-tation in a broadly accessible, intuitively appealingway was long and labored.

Challenges in supporting a larger number ofcomputing platforms draw attention to the ques-tion of the firm’s choice of platform focus. Greaterdiversity in computing platforms appears to bene-fit firm survival. However, the standardization ofthe installed base of hardware to a few comput-ing platforms—platform focus—foreshadows theeventual importance of facility with a particularcomputing environment. Generalist approaches toplatform development, where software was avail-able for any and all platforms, were lauded earlyin the industry. Eventually, however, these werenot competitive as bandwagon effects resulted inan industry with fewer platforms. However, itremains unclear whether founding as a special-ist in a single platform would have been thebest approach. It may be that a generalist strat-egy provides greater flexibility against the even-tual narrowing on a few platforms. Even if firmswidely anticipated platform dominance by IBM, itis not obvious what the best strategy would havebeen.

Our analysis across firm and product levels pro-vides an additional perspective on firm evolution.Our results are the first to combine detailed prod-uct knowledge with firm-level characteristics, anddemonstrate the necessity of multiple levels ofanalysis in the study of scope economies. Thiscross-level analysis supports a ‘flagship’ modelof competition, also referenced in Nobeoka andCusumano (1997). Firms such as Lotus (1-2-3),Ashton-Tate (dBASE II), Borland (Turbo Pascal),and Novell (Netware) all had a central successfulproduct that was complemented by other product



lines, consistent with the studies of product plat-forms as the basis of product variety discussed inthe second section.

Our results also demonstrate that care shouldbe taken to evaluate the importance of firm-levelinitiatives in the context of individual productstrategies. The age of the firm, perhaps as asignal of longer-term maintenance and upgrades,enhances survival of individual products. This pro-vides some mechanism for the finding in manyprevious studies that older firms have a lower haz-ard rate. Perhaps surprisingly, the number of otherproducts offered by the firm decreases the proba-bility of survival of any individual product. Thisgreater probability of product exit may be due todivided interests, with a broad portfolio resultingin/from a lack of commitment to any particularproduct initiative.

This research also shows that studies of scopeeconomies should control for process of productdiversification. As we found, adjustments in prod-uct variety can have positive or negative effectson firm survival depending on whether they arenew product introductions or product extensions:introductions of a new product correlates with adecrease in exit, while entering a new category orplatform increases exit. It is likely that the newproduct effect is due to the nature of innovationin product development that is so important inthis industry vs. the extension of a product thatis already in the process of exit.

Although many characteristics of the firm playa significant role in survival of any individualproduct, the influence of firm change on individ-ual product survival is dramatic; firm expansion isnot frictionless. In our analyses, firm expansionin products and application categories producesincreased risk of product exit, but this is compen-sated by access to greater opportunities.

Scope economies and focus strategies have beenevolving throughout the history of the softwareindustry. Management of product variety has toaccount for these changes. Clearly, the relationshipbetween scope and integration/focus has changedover the past 20 years, particularly as the operat-ing system facilitates greater product market andplatform integration. In the emergent microcom-puter software industry, integrated products hadan advantage, but the results might be very differ-ent in today’s computing environment. The evolu-tion in product bundling and its implications foreconomies of scope remains an important issue.

This research has attempted to separate scopeeconomies in consumption from scope economiesin production by giving attention to the indus-try context. Even greater attention to a model ofthe customer should provide the rationale for bothwhere scope economies should be expected andwhy. Research that ignores the production and con-sumption aspects of industry context will miss thisimportant subtlety. The value we found of sep-arating firm-level from product-level effects sug-gests that a much more detailed model of scopeeconomies is warranted—one that pays attentionto these different sources of scope economies.

The levels of analysis in this study providean unprecedented look at product and firm sur-vival. Integration and focus strategies affecting theway the firm appropriates the benefit of scopeeconomies may matter more than aggregate ben-efits of scope more generally. Managing productvariety and complexity in the firm may be allthe more important, so that the existence of scopeeconomies is less critical than the ability to man-age product and firm evolution in an advantageousway. Damanpour (1996) suggests that greater com-plexity facilitates innovation in larger organiza-tions. The insight provided in our software studysuggests that product-level measures of complexityare equally important. And in this context, deter-mination of the characteristics of production andconsumption that leads to these benefits is critical.

ACKNOWLEDGEMENTS

The authors thank the Sloan Foundation Consor-tium on Competitiveness, the AEC Future Fundprogram and the Informatics Research Centre forsupport. Insightful comments from David Mowery,Trond Petersen, J. C. Herb Emery, Joanne Oxley,Brian Silverman, Jeff Church, and two anonymousreferees are gratefully acknowledged.

REFERENCES

Abernathy WJ, Clark K, Kantrow A. 1983. IndustrialRenaissance: Producing a Competitive Future forAmerica . Basic Books: New York.

Alexander PJ. 1997. Product variety and market structure.Journal of Economic Behavior and Organization32(2): 207–214.

Amihud Y, Lev B. 1997. Risk reduction as a managerialmotive for conglomerate mergers. Bell Journal ofEconomics 12: 605–617.



Argyres N. 1996. Information technology and organiza-tional boundaries: evidence from the stealth bomber.Academy of Management Annual Meetings; Cincin-nati, OH.

Audretsch DB. 1991. New-firm survival and thetechnological regime. Review of Economics andStatistics 73: 441–450.

Bailey EE, Friedlaender AF. 1982. The theory ofoligopoly with multi-product firms: market structureand multiproduct industries. Journal of EconomicLiterature 20: 1024–1048.

Baldwin JR, Gorecki PK. 1994. Concentration andmobility statistics in Canada’s manufacturing sector.Journal of Industrial Economics 42: 93–103.

Barnett WP, Carroll GR. 1995. Organizational change. InAnnual Review of Sociology , Hagan J, Cook KS (eds).Annual Reviews: Palo Alto, CA; 217–236.

Baum JAC. 1995. The changing basis of competitionin organizational populations: the Manhattan hotelindustry, 1898–1990. Social Forces 74: 177–204.

Baum JAC, Korn HJ. 1996. Competitive dynamics ofinterfirm rivalry. Academy of Management Journal 39:255–291.

Baum JAC, Singh J. 1994. Organizational niches andthe dynamics of organizational mortality. AmericanJournal of Sociology 100: 346–380.

Bergh DD, Holbein GF. 1997. Assessment and redirec-tion of longitudinal analysis: demonstration with astudy of the diversification and divestiture relationship.Strategic Management Journal 18(7): 557–571.

Boehm BW. 1981. Software Engineering Economics .Prentice-Hall: Englewood Cliffs, NJ.

Bruderl J, Preisendorfer P, Ziegler R. 1992. Survivalchances of newly founded organizations. AmericanSociological Review 57: 227–242.

Church J, Gandal N. 1992. Integration, complementaryproducts, and variety. Journal of Economics andManagement Strategy 1: 651–675.

Clark K. 1987. Investment in new technology andcompetitive advantage. In The Competitive Challenge:Strategies for Industrial Innovation and Renewal ,Teece DJ (ed). Ballinger: Cambridge, MA; 59–81.

Damanpour F. 1996. Organizational complexity and inno-vation: developing and testing multiple contingencymodels. Management Science 42: 693–716.

Dunne P, Hughes A. 1994. Age, size, growth andsurvival: UK companies in the 1980s. Journal ofIndustrial Economics 42: 115–140.

Dunne T, Roberts MJ, Samuelson L. 1988. Patterns offirm entry and exit in U.S. manufacturing industries.Rand Journal of Economics 19: 495–515.

Economides N, Salop SC. 1992. Competition andintegration among complements. Journal of IndustrialEconomics 40: 105–123.

Elsevier Publishing. 1980. The Software Catalog:Microcomputers . Elsevier: New York.

Evans DS. 1987. The relationship between firm growth,size, and age: estimates for 100 manufacturingindustries. Journal of Industrial Economics 35:567–581.

Farrell J, Saloner G. 1992. Converters, compatibility,and the control of interfaces. Journal of IndustrialEconomics 40: 9–35.

Geroski P, Mazzucato M. 2002. Learning and sources ofcorporate growth. Industrial and Corporate Change11(4): 623–644.

Hoskisson RE, Hitt MA, Johnson RA, Moesel DD. 1993.Construct validity of an objective (entropy) categoricalmeasure of diversification strategy. Strategic Manage-ment Journal 14(3): 215–235.

Jacquemin A, Berry C. 1979. Entropy measure ofdiversification and corporate growth. Journal ofIndustrial Economics 27: 359–369.

Kesner IF, Sharma A. 1996. Diversifying entry: some exante explanations for postentry survival and growth.Academy of Management Journal 39: 635–677.

Krishnan V, Gupta S. 2001. Appropriateness and impactof platform-based product development. ManagementScience 41(1): 52–68.

Lane PJ, Cannella AA Jr, Lubatkin MH. 1998. Agencyproblems as antecedents to unrelated mergersand diversification: Amihud and Lev reconsidered.Strategic Management Journal 19(6): 555–578.

Lieberman MB. 1989. The learning curve, technologybarriers to entry, and competitive survival in thechemical processing industries. Strategic ManagementJournal 10(5): 431–447.

Mata J, Portugal P. 1994. Life duration of new firms.Journal of Industrial Economics 42: 227–245.

Mayer KU, Hamerle A, Blossfeld H. 1989. Event HistoryAnalysis: Statistical Theory and Application in theSocial Sciences . Erlbaum: Hillsdale, NJ.

Mayer KU, Tuma NB. 1990. Event History Analysis inLife Course Research . University of Wisconsin Press:Madison, WI.

Merges RP. 1992. A comparative look at property rightsand the software industry. Working paper, BostonUniversity.

Merino F, Rodriguez DR. 1997. A consistent analysisof diversification decisions with non-observablefirm effects. Strategic Management Journal 18(9):733–743.

Meyer MH, Utterback JM. 1993. The product family andthe dynamics of core capability. Sloan ManagementReview 34(3): 29–49.

Montgomery CA, Wernerfelt B. 1988. Diversification,Ricardian rents, and Tobin’s Q. Rand Journal ofEconomics 19: 623–632.

Nault BR, Vandenbosch MB. 1996. Eating your ownlunch: protection through preemption. OrganizationScience 7(3): 342–358.

Nault BR, Vandenbosch MB. 2000. Disruptive technolo-gies: explaining entry in next generation informationtechnology markets. Information Systems Research11(3): 304–319.

Nobeoka K, Cusumano M. 1997. Multiproject strategyand sales growth: the benefits of rapid design transferin new product development. Strategic ManagementJournal 18(3): 169–186.

Palepu K. 1985. Diversification strategy, profit, perfor-mance, and the entropy measure. Strategic Manage-ment Journal 6(3): 239–255.



Petersen T. 1991. Time-aggregation bias in continuous-time hazard-rate models. In Sociological Method-ology , Vol. 21. Marsden PM (ed). Basil Blackwell:Oxford; 263–290.

Porter ME. 1985. Competitive Advantage: Creating andSustaining Superior Performance. Free Press: NewYork.

Robertson D, Ulrich K. 1998. Planning for productplatforms. Sloan Management Review 39(4): 19–31.

Sen A, Srivastava M. 1990. Regression Analysis: Theory,Methods, and Applications . Springer: New York.

Shapiro D, Khemani RS. 1987. The determinants ofentry and exit reconsidered. International Journal ofIndustrial Organization 5: 15–26.

Sherman CE. 1984. Up and Running: Adventures ofSoftware Entrepreneurs . Ashton-Tate: Culver City,CA.

Sorenson O. 2000. Letting the market work for you: anevolutionary perspective on product strategy. StrategicManagement Journal 21(5): 577–592.

Tarter J. 1996. Soft*Letter 100 . Soft*Letter: Watertown,MA.

Teece DJ. 1982. Towards an economic theory of themultiproduct firm. Journal of Economic Behavior andOrganization 3: 39–63.

Walden J. 1986. File Formats for Popular PC Software:A Programmer’s Reference. Wiley: New York.

Wallace J, Erickson J. 1992. Hard Drive: Bill Gates andthe Making of the Microsoft Empire. Wiley: New York.

Yamaguchi K. 1991. Event History Analysis . Sage:Newbury Park, CA.


Documents

Product variety and firm survival in the microcomputer software industry