Org Sci Final

8/3/2019 Org Sci Final

1/18

Organization ScienceVol. 18, No. 5, SeptemberOctober 2007, pp. 763780issn 1047-7039 eissn 1526-5455 07 1805 0763

inf orms

doi 10.1287/orsc.1070.0306 2007 INFORMS

IT Assets, Organizational Capabilities, and FirmPerformance: How Resource Allocations and

Organizational Differences Explain Performance Variation

Sinan AralMassachusetts Institute of Technology Sloan School of Management, and New York University Stern School of Business,

44 West 4th Street, Room 8-81, New York, New York 10012, [email protected]

Peter WeillCenter for Information Systems Research, MIT Sloan School of Management, 3 Cambridge Center, NE20-332,

Cambridge, Massachusetts 02142, [email protected]

Despite evidence of a positive relationship between information technology (IT) investments and rm performance,results still vary across rms and performance measures. We explore two organizational explanations for this variation:differences in rms IT investment allocations and their IT capabilities. We develop a theoretical model of IT resources,dened as the combination of specic IT assets and organizational IT capabilities. We argue that investments into differentIT assets are guided by rms strategies (e.g., cost leadership or innovation) and deliver value along performance dimensionsconsistent with their strategic purpose. We hypothesize that rms derive additional value per IT dollar through a mutuallyreinforcing system of organizational IT capabilities built on complementary practices and competencies. Empirically, wetest the impact of IT assets, IT capabilities, and their combination on four dimensions of rm performance: marketvaluation, protability, cost, and innovation. Our resultsbased on data on IT investment allocations and IT capabilitiesin 147 U.S. rms from 1999 to 2002demonstrate that IT investment allocations and organizational IT capabilities drivedifferences in rm performance. Firms total IT investment is not associated with performance, but investments in specicIT assets explain performance differences along dimensions consistent with their strategic purpose. In addition, a systemof organizational IT capabilities strengthens the performance effects of IT assets and broadens their impact beyond theirintended purpose. The results help explain variance in returns to IT capital across rms and expand our understanding of alignment between IT and organizations. We illustrate our ndings with examples from a case study of 7-Eleven Japan.

Key words : business value of information technology; information technology assets; resource-based theory;

complementarities; IT infrastructure; IT capabilities; IT practices; rm performance

1. IntroductionFor more than a decade, research has attempted tountangle the relationship between information technol-ogy (IT), productivity, and organizational performance.Early results uncovered a paradox in the relationship(Loveman 1994, Strassman 1990) that has subsequentlybeen explained by both substantive and methodologi-cal considerations (Bakos 1991, Dos Santos et al. 1993,Brynjolfsson and Hitt 1996). Recently, more precisemeasurements have demonstrated a convincing (albeitvaried) positive relationship among IT investments, eco-nomic productivity, and business value across distinctmeasures (Brynjolfsson and Hitt 1996, Dewan and Min1997, Bharadwaj et al. 1999). Although this researchprovides evidence of a general relationship between ITand organizational performance, our knowledge of thespecic factors driving these general results remainsquite limited. In this paper we address two importantquestions that remain unanswered.

First, returns to IT investments exhibit substantialvariation across rms. Large sample statistical evidence

demonstrates that nearly half of the productivity benetsoriginally attributed to IT capital can be more accuratelyexplained by rm-specic factors (e.g., Brynjolfsson andHitt 1995). These results imply the existence of a set of organizational characteristics that are simultaneously andpositively correlated with both IT investment and organi-zational performance. Some rms simply derive greatervalue per IT dollar even when controlling for industry-level variation. But what types of organizational charac-teristics explain this variation? To address this question,we open the black box of the organization to examinewhat types of organizational factors and managementpractices contribute to a rms ability to generate busi-ness value from IT.

Second, the majority of rm-level analysis measuresIT in the aggregate. As a result, we know little aboutthe relative performance contributions of different typesof IT investments and whether different IT investmentsimpact different aspects of rm performance. One expla-nation for why two rms with the same amount of IT capital perform differently is that they are investing

763


2/18

Aral and Weill: IT Assets, Organizational Capabilities, and Firm Performance764 Organization Science 18(5), pp. 763780, 2007 INFORMS

in different types of technology with different goals.We therefore conceptualize IT as four distinct types of assets, implemented to achieve different managementobjectives, and test their relative performance effects.

We explore these questions using data from 147 rmsover 4 years and illustrate the results using qualitative

evidence from a case study of 7-Eleven Japan. We ndthat investments in a particular IT asset class deliverhigher performance only along dimensions consistentwith the strategic purpose of that asset. For example,investments in transactional IT applications, made toreduce costs in standard, repetitive processes, are asso-ciated with lower costs but not with more rm-levelproduct innovation. In contrast, investments in strategicIT applications are associated with more product inno-vation, but not with lower costs. These results suggestthat a monolithic view of IT may obscure the impor-tance of resource allocations within the IT function byfocusing on the performance implications of rms totalIT capital stock. We also nd evidence for comple-mentarities between IT and a system of organizationalIT capabilities (ITC). IT investments and organizationalITC covary signicantly in our sample, demonstrat-ing that rms high in IT intensity develop IT-relatedorganizational capabilities and that rms with strong ITcapabilities demand more IT. Firms with stronger orga-nizational ITC also derive greater value per IT dollar.We nd that ITC both strengthens intended performanceeffects and broadens the impact of investments in partic-ular IT assets beyond their intended performance goals.Our ndings demonstrate the importance of pursuingmore detailed and disaggregated measures of IT inten-sity, organizational IT capabilities, and rm performancein IS research.

2. Theory and Literature2.1. The Resource-Based Theory of the FirmRecent research on the relationship between IT and orga-nization describes systems of organizational practicesthat complement IT. One theoretical perspective that con-vincingly addresses the complementarity of IT and orga-nizational processes, practices, routines, and activities isthe resource-based theory of the rm (Wernerfelt 1984,Barney 1991). This theory argues that durable compet-itive advantage emerges from unique combinations of resources (Grant 1996) that are economically valuable,scarce, and difcult to imitate (Barney 1991). As theseresources are imperfectly mobile across rm boundariesand because rms pursue different strategies in deployingthese resources, they are likely to be heterogeneously dis-tributed across rms. Firm resources are insulated fromcompetitive imitation by path dependencies, embedded-ness, casual ambiguity about the source of competitiveadvantage, and time diseconomies of imitation (Barney

1991, Mata et al. 1995). These heterogeneously dis-tributed and difcult-to-imitate resources in part drivedifferences in rm performance.

From this perspective, there are compelling theoreticalreasons for investigating how rms allocate investmentsacross different types of IT assets. The resource-based

view separates stocks of undifferentiated factors of pro-duction from resources , dened as the combination of rm-specic assets (Wernerfelt 1984) and organizationalcapabilities (Richardson 1972, Nelson and Winter 1982,Dosi et al. 2000). The dynamic capabilities framework (Teece et al. 1997), which extends the resource-basedview to incorporate environmental and technologicalchange, stresses the importance of tangible and intangi-ble specic asset positions in shaping rm resources.Teece et al. (1997, pp. 522523) argue that a rmsprevious investments and repertoire of routines constrainits future behavior; and that opportunities for learn-ing will be close in to previous activities and thus will

be transaction and production specic. Taken together,these theoretical treatments of resources, assets, andcapabilities imply that rms invest in particular types of resources and learn how to use those resources over timeby developing asset-specic skills and accompanyingroutines (Cohen and Levinthal 1990). Resources are dif-cult to imitate in part because rms are unaware of theircompetitors resource allocations and how they con-tribute to performance (causal ambiguity) and becausecapability development and learning opportunities aretied to rms specic asset positions (path dependen-cies) (Dierickx and Cool 1989). If learning and behaviorinside rms are shaped by specic asset positions, thenrms that spend more heavily on particular assets shoulddisplay abnormally higher performance in measures thatreect the goals of those assets as they learn how todeploy them with complementary organizational pro-cesses. We argue that investment allocations and orga-nizational differences help shape the heterogeneous ITresources rms develop and explain variation in rm per-formance. We empirically distinguish assets, dened asinvestments in different types of IT, from capabilities,dened as practices and competencies that support theuse of IT.

2.2. The Resource-Based Theory of ITAlthough the resource-based view provides a helpfultheoretical perspective from which to evaluate the het-erogeneity of rm performance, the existing IT literaturesuffers from ambiguity in the denition and conceptual-ization of IT resources (Wade and Hulland 2004). Mostcurrent conceptualizations of IT resources equate poten-tially heterogeneous investment allocations across rmsby measuring total IT intensity. Some empirically con-found resources with capabilities by not measuring bothinvestments and organizational factors simultaneously.Others theoretically distinguish IT infrastructure from


3/18

Aral and Weill: IT Assets, Organizational Capabilities, and Firm PerformanceOrganization Science 18(5), pp. 763780, 2007 INFORMS 765

nontechnical assets (such as human capital and externalrelationships) but do not distinguish them empirically ormeasure rms specic dollar investments in differenttypes of IT assets (Ross et al. 1996, Bharadwaj 2000).Of the perspectives reviewed by Wade and Hulland(2004) that distinguish different IT resources, none iden-

ties the strategic purpose of the resource for the rmand none measures investments in different types of ITassets, implicitly discounting the possibility that invest-ment allocations help shape rms IT resources andperformance.

Organizational capabilities moderate the relationshipbetween IT investments and different measures of rm performance (e.g., Brynjolfsson and Yang 1997,Bharadwaj 2000), but aggregate measures of IT invest-ment and ambiguous denitions of IT resources producevaried results. For example, Zhu and Kraemer (2002)identify four metrics that assess the e-commerce capa-bility of rms and demonstrate that rms with greater

e-commerce capabilities perform better on some dimen-sions of performance (e.g., supply chain optimization)but perform worse along other dimensions (e.g., thecost of goods sold). These results provide convincingevidence of complementarities between an aggregatemeasure of IT intensity and organizational capabili-ties in e-commerce. However, the aggregate measureof IT intensity also reveals some surprising results.The authors argue that inexperience and high learningcosts may explain the surprising result that the useof e-commerce, together with IT investment, is associ-ated with increased COGS for traditional manufacturingcompanies (Zhu and Kraemer 2002, p. 288). At thesame time, they acknowledge that their data did notcapture enough details of the differences in the natureof [e-commerce capabilities] and IT resources between[rms] to test whether different types of IT resources

Figure 1 Theoretical Model of IT Resources

IT assets: IT investments allocatedfor particular strategic purposes

Support entry into a new market,provide a new service, enable anew product

Strategic

Provide information for managing,accounting, reporting, planning, analysis,and data mining

Informational

Automate processes, cut costs increasevolume per unit cost

Transactional

Foundation of shared IT services. Provideflexible base for future business

Infrastructure

Strategic purposeIT asset

IT capabilities: interlocking systems ofpractices and competencies that

complement IT

Culture of IT useDigital transactionsInternet architecture

Competencies(Skills)

IT skillsIT management quality

IT resources

Practices(Routines)

are driving performance differences (Zhu and Kraemer2002, p. 288). An alternative explanation for this sur-prising result is that traditional manufacturing rms areinvesting in fundamentally different IT resources thanthe high-tech rms in their second subsample. Mod-ernization of the factory oor and trends toward exi-

ble manufacturing are requiring manufacturing rms toundertake signicant investments in new IT infrastruc-tures (Milgrom and Roberts 1990). Prior empirical work (reviewed below) suggests that investments in IT infra-structure may cause short-term disruptions that increasecosts relative to other types of IT assets, which couldexplain why traditional manufacturing rms see highercosts with more IT investment. To test this alternativeexplanation, more detailed data on how rms allocateaggregate IT investments is necessary. Our aim is tosharpen the theoretical characterization of IT resourcesby unpacking two major sources of variation in theempirical evidence on complementarities between IT

and organization: heterogeneity in IT investment alloca-tions and organizational IT capabilities.

2.3. Reconceptualizing IT Resources asCombinations of IT Assets and IT Capabilities

In our theoretical model, IT resources are combinationsof investment allocations and a mutually reinforcing sys-tem of competencies and practices that together rep-resent organizational ITC. Figure 1 depicts our modelbased on theoretical concepts drawn from reviews of the IT and organizational capabilities literatures and theresource-based theory of the rm, supported by vequalitative case studies conducted in conjunction withour quantitative analysis.

Firms make heterogeneous investment allocations inpursuit of different goals (e.g., cost leadership or inno-vation), resulting in a varying landscape of IT resources


4/18


across rms. For example, rms with cost leadershipstrategies will likely allocate investments toward trans-actional IT systems designed to cut costs, but rmspursuing innovation strategies will likely invest morein IT systems that support product and process inno-vation. Strong IT resources are scarce and difcult to

imitate because developing effective combinations of ITassets and IT capabilities takes time spent learning andoptimizing. This heterogeneity, in both investment allo-cations and capabilities, drives performance variationacross differentiated dimensions that reect rm strate-gies. Differentiated investment allocations will enablesome rms to cut costs and others to innovate; strongITC will increase the return per IT dollar invested.

The next two sections present the theoretical develop-ment of our framework for measuring IT assets and ITCas the building blocks of IT resources.

2.4. Disaggregating Total IT Capital intoIT Asset Classes

Most empirical examinations of IT business value con-sider IT as an aggregate, uniform asset (Bharadwajet al. 1999), divide IT investments into capital and laborstock (Bynjolfsson and Hitt 1996, Hitt and Brynjolffson1996, Bharadwaj 2000), or examine particular technolo-gies, such as ATMs or production control technologies(Kelley 1994, Dos Santos and Peffers 1995). Bharadwajet al. (1999, p. 1020) argue it appears that rms ben-et unequally from their different IT investments. Thusit would be interesting to examine the impact of differ-ent types of IT investments such as innovative versus

noninnovative , strategic versus nonstrategic , and inter-nally focused (e.g., process control, coordination, etc.)versus externally focused investments (customer satis-faction, relationship management, etc.)

Although IT investment allocations are likely to reectrm strategy and affect rm performance (Floyd andWooldridge 1990, Dos Santos et al. 1993), few studiesdisaggregate IT investments by asset type. To addressthis gap, we apply a framework developed by Weill(1992) and extended by Weill and Broadbent (1998) thatcategorizes rms IT investments into a portfolio of fourIT assets disaggregated by strategic purpose: infrastruc-ture, transactional, informational, and strategic assets.

This framework has been validated and empiricallytested in previous work, and we hypothesize that invest-ments in each asset class are associated with gains alongperformance dimensions consistent with their strategicpurpose.

Hypothesis 1. Investments in IT assets are associ-ated with higher rm performance only along dimen-sions consistent with the strategic purpose of the asset.

We measure IT investment allocations according tohow rms senior managers characterize spending across

the four IT asset classes:1. IT infrastructure provides the foundation of shared

IT services (both technical and humane.g., servers,networks, laptops, shared customer databases, helpdesk, application development) used by multiple ITapplications (Keen 1991, Weill and Broadbent 1998).

Infrastructure investments are typically made to pro-vide a exible base for future business initiativesand thus are made in anticipation of future businessneeds. The disruptive nature of enterprisewide infras-tructure implementations creates high up-front costs andlong benet time horizons (Duncan 1995, Weill andBroadbent 1998). However, infrastructure investmentsalso enable new applications and functionality and laythe groundwork for signicant long-term performanceimprovements (Duncan 1995, Broadbent et al. 1999).We therefore expect that infrastructure investments arepositively associated with higher short-term costs, lowershort-term protability, and higher protability and oper-

ational performance in the long run. In addition, if infrastructure investments are transparent to the market,we will likely see a positive impact on rm market value,which reects the markets assessment of rms futurevalue.

2. Transactional investments are made to automateprocesses, cut costs, or increase the volume of busi-ness a rm can conduct per unit cost (e.g., orderprocessing, point of sale processing, bank cash with-drawal, billing statement production, insurance renewal,and other repetitive transaction processing functions).We expect transactional investments are associated withimmediate cost reductions.

3. Informational investments provide information formanaging, accounting, reporting, and communicatinginternally and with customers, suppliers, and regulators.Examples include decision support, sales analysis, plan-ning, six sigma programs, and Sarbanes-Oxley reportingsystems. These investments can support the responsive-ness, control, reliability, and adaptability of rms andenable more effective decision making. Sales analysisand data mining of customer reactions to products andservices can help optimize products and pricing, thusenabling more efcient and protable operations. Weexpect informational investments to tighten reportingand control functions and to improve data collection anddecision making, thereby reducing costs and identifyingnew opportunities for revenue generation and protabil-ity improvements.

4. Strategic investments reposition rms in themarketplace by supporting entry into a new market orthe development of new products, services, or busi-ness processes. Successful strategic investments typi-cally change the nature of service delivery or organiza-tional processes in an industry, but they become non-strategic when competitors commoditize the capability.When ATMs were introduced in the banking industry,


5/18


Table 1 IT Assets and Expected Performance Benets

ExpectedIT asset Strategic purpose performance benets

Infrastructure Foundation of shared Short term: GreaterIT services. Provide costs, less protabilityexible base for (due to disruption)

future businessinitiatives

Greater market value Long term: Greater

prot, lower costs

Transactional Automate processes, Lower costscut costs, increasevolume of businessper unit cost

Informational Provide information for Lower costsmanaging, accounting, Greater protabilityreporting, decisionsupport, planning,control, analysis, anddata mining

Strategic Support entry into a More productnew market, provide innovationa new service, orenable a new product

they changed the nature of service delivery and gar-nered competitive benets for early adopters (Dos San-tos and Peffers 1995), but became nonstrategic and trans-actional as they were universally adopted. We expectstrategic investments contribute directly to product inno-vation (Samabmurthy et al. 2003). Table 1 describesexpected performance gains by asset class.

2.5. IT Capabilities as a Mutually Reinforcing

System of Practices and CompetenciesA variety of individual capabilities, practices, and pro-cesses may complement IT; however, we expect sys-tems of practices and competencies working in concertto enable greater business value generation per IT dol-lar. Milgrom and Roberts (1990) formally demonstratethat nonconvexities exist in a rms decision to adoptany or all of a set of organizational characteristics thattogether complement new technology. As the marginalbenet of adopting any one of a complementary set of activities increases with the adoption of the others, adop-tion of systems of practices (what Milgrom and Roberts1990 call groups of activities) may not be marginaldecision[s]. They argue, Exploiting such an extensivesystem of complementarities requires coordinated actionbetween traditionally separate functions (Milgromand Roberts 1990, p. 515). We use prior research andour own case studies to identify a group of interlock-ing organizational characteristics that together supportrms ability to derive value from IT. Complementaritytheory predicts both the clustering of these IT capabil-ities and their moderating effects on rm performance.To validate the systematic nature of IT complements,we identify and measure capabilities separately, test the

degree to which they covary in our sample, and sub-sequently examine their performance implications as aninterlocking system, or cluster, as depicted in Figure 1.

Although Milgrom and Roberts (1990) adopt a nar-row theoretical perspective focused on complementarygroups of activities, the conceptualization of organiza-

tional complements to IT in theories drawn from evo-lutionary economics and the resource-based view of therm take a broader view. These theories address notonly the activities organizations engage in, but also theskills and competencies they develop in using assetsto accomplish organizational tasks. At least two funda-mental conceptual building blocks useful for identify-ing characteristics of rms that complement IT emergefrom these theories: competencies (or skills) and prac-tices (or routines) (Nelson and Winter 1982). Compe-tencies refer to skills embodied in individuals or groupsthat actively manage or accomplish organizational tasks(Prahalad and Hamel 1990, Dosi et al. 2000). Compe-

tencies are developed through learning and the repeatedperformance of contextual activities. As individuals andgroups interact with IT for particular purposes, theylearn, build skills, and develop competence toward effec-tive use. Practices , in contrast, refer to recurring setsof activities or routines that serve both as a means of accomplishing organizational tasks and as mechanismsfor socially storing and accessing knowledge about themost effective ways to accomplish those tasks (Cohenand Levinthal 1990). Practices and competencies sup-port and complement each other. Practices help indi-viduals and groups develop competence or skill withparticular ways of working (Dosi et al. 2000), andskills are necessary for the effective execution of orga-nizational practices toward specied goals (Nelson andWinter 1982).

To develop the construct of organizational ITC we rstidentied candidate constructs in the literature that weresupported by our case evidence. We then developed acoherent conceptualization by excluding elements unre-lated to the system of characteristics identied across allcases. We use case studies to inform our construct devel-opment and measurement and to illustrate and informthe conclusions we draw from quantitative results. Casestudies were conducted in two medium-sized manufac-turing rms, one large and one medium-sized nan-cial services rm, and 7-Eleven Japan, a large Japaneseretailer (see Weill and Aral 2005, Nagayama and Weill2004 for published case material). 1 In this paper, we useillustrative examples exclusively from 7-Eleven Japan todescribe distinctions between different IT assets and thesystematic and mutually supportive nature of individ-ual competencies and practices in a single organization.We further rened our inclusion and exclusion criteriaby testing the degree to which our candidate constructsworked together as a system, as evidenced by theircovariance across rms in our larger sample and the


6/18


Table 2 IT Capability Constructs: Theoretical Development and Examples of Supporting Qualitative Evidence

IT capability Illustrative examples of qualitative evidence fromconstruct Theoretical justication 7-Eleven Japan case data

CompetenciesHuman resource

competency Skill-biased technical

change (e.g., Autor

et al. 1998)

Immersive training of 200,000 employees in point of sale data analysis;including analysis of information on products, weather conditions, regional

demographics, and customer purchasing patterns to improve sales, customersatisfaction, and ordering.

Close contact and coaching of store franchisees by company counselorsin the use of IT to support decision making. Twice weekly visits by counselorsto stores reinforce practices and support development of skills.

IT-skill complementarity(e.g., Breshnahanet al. 2002)

Managementcompetency

Senior managementchampioning(e.g., Weill 1992)

Strong commitment of senior management to IT projects and IT-basedprocesses comes directly from the CEO, who has been committed todata-based decision making and IT-based communication sincejoining 7-Eleven Japan in 1974.

Business processes tightly integrated with and enabled by IT decisions.For a detailed set of examples see Nagayama and Weill (2004), in particularExhibit 5.

Alignment of IT andbusiness units (e.g.,Rockhart et al. 1996)

PracticesIT use intensity for

communication Systems-use theory

(e.g., Doll andTorkzadeh 1998)

Total Information System connects 70,000 computers in stores, atheadquarters, and at supplier sites to facilitate internal and externalcommunication and coordination.

Task-technology ttheory (e.g., Goodhueand Thompson 1995)

QuoteSalesperson of 7-Eleven supplier: [Their] information system is sogood that we can instantly nd out which goods of ours are selling [in theirstores] to what types of customers and how much.

Quote7-Eleven executive: Even if the point of sale data [are] used,[they] cannot be utilized for the next order unless the hypothesis ofpotential demand is shared among all store clerks as well as the storeowner. Therefore, we need to establish a system that enables store ownersand the ordering clerks to create their hypotheses and share themamong part-time workers at the store [even if they cannot communicateface to face].

Digital transactionintensity

Transaction cost theory(e.g., Williamson 1975)

Digital transactions enable order processing three times per day. Time todelivery is reduced, orders are organized for use (e.g. by temperaturefrozen,refrigerated, ambient), and costs of order processing are reduced.

Digital transactions enable tracking and analysis of point of sales data toinform daily ordering decisions. Each days data is analyzed for decisionsmade the next morning.

Customer satisfaction goal drives IT-enabled business transactions like ItemControl and Product Supply Management designed to directly addresscustomer needs and increase customer convenience.

Coordination theory(e.g., Malone et al. 1987)

Customer intimacy (e.g.,Mithas et al. 2005)

Internet architecture e-commerce capabilitytheory (e.g., Zhu andKraemer 2002)

The Internet shopping site (www.7dream.com, a strategic IT asset) isintegrated with physical stores to offer payment acceptance andpick-up and/or delivery services for products purchased on line.

Use of multipurpose, Internet-enabled store copy machines to provide newservices including preordering, printing, and purchasing of airline tickets.Also see Nagayama and Weill (2004), Exhibit 10.

results of factor analysis conducted using the broader

set of 18 factors. Using this process, we identied twocompetencies and three practices. 2 Table 2 presents asummary of the ve competencies and practices, theirtheoretical justication, and supportive examples fromcase data.

2.5.1. Competencies Skills . Competencies existacross two organizational dimensions in our data: theIT skills of employees at all levels (both technicaland business skills) and IT management competence.Shifts in labor demand over the last 25 years favoringmore skilled and educated workers have been driven in

large part by skill-biased technical change or tech-

nical progress that shifts demand toward more skilledworkers (Autor et al. 1998). Unlike shifts in labordemand during the Industrial Revolution, which favoredunskilled factory labor (Goldin and Katz 1998), todaystechnology complements greater autonomy, exibility,and skilled employees. A strong empirical relationshipbetween IT use and skill at the worker (Kreuger 1993),rm (Dunne et al. 1997), and industry (Autor et al. 1998)levels demonstrates that rms with signicant amountsof IT tend to hire more skilled workers. But few stud-ies examine the performance implications of the co-presence of IT and highly skilled labor (for an exception,


7/18


Table 3 Variable Denition and Descriptive Statistics

Variable Computation Source N Mean SD

Ln employees Natural logarithm of the total number of employees Compustat 588 14 2 30 0Sales (M$) Net sales revenue Compustat 588 3 442 7 6 936 9Advertising expenditure Advertising expenditures / Sales Compustat 588 0 03 0 08

intensity

R&D expenditure intensity R&D expenditures / Sales Compustat 588 0 02 0 08IT intensity Total IT$ / Sales MIT survey 453 0 02 0 04Infrastructure intensity IT$ spent on infrastructure / Sales MIT survey 346 0 009 0 011Transactional IT intensity IT$ spent on transactional systems / Sales MIT survey 115 0 003 0 006Informational IT intensity IT$ spent on informational systems / Sales MIT survey 119 0 004 0 009Strategic IT intensity IT$ spent on strategic systems / Sales MIT survey 118 0 002 0 002Return on assets (%) (Income before extraordinary items / Total assets) 100 Compustat 564 0 54 14 1Tobins q [Market value of common stock + Book value of Compustat 569 1 0 1 2

debt + Book value of preferred stock] / [Book value ofassets and plant, property, and equipment (PPE)+ Estimated replacement cost of PPE]

Net margin (%) (Income before extraordinary items / Total sales) 100 Compustat 564 1 1 13 3Cost of goods sold Cost of merchandise purchased + Cost of goods Compustat 569 2 395 3 5 174 3

manufactured for goods soldSales from new products Sales from new products from the previous year / Total sales MIT survey 119 0 236 0 223Sales from modied Sales from products modied or enhanced from MIT survey 119 0 333 0 278

products the previous year / Total salesITC A demeaned linear combination of capability variables. MIT survey 142 0 05 1 5

ITC = ((Capability measure 1 Mean of capabilitymeasure 1) + + (Capability measure 6 Mean ofcapability measure 6))

see Breshnahan et al. 2002). We estimate the humanresource competency (HR) of rms by assessing (a) thetechnical and business skills of IT staff, (b) the IT skillsof business users, and (c) the relative ability of rms tosatisfy their demand for highly skilled IT labor. In addi-tion, senior management championing of IT initiativesis consistently shown to improve the value created byIT investments (Weill 1992, Brynjolffson and Hitt 2000)and disconnects between business units and the IT func-tion typically hinder rms ability to generate returnsfrom IT (Rockart et al. 1996). Our measure of man-agement competency (MC) therefore assesses both thedegree of senior management commitment to IT projectsand business unit involvement in IT decisions (Weill andRoss 2004). Table 2 provides qualitative examples of HRand MC found at 7-Eleven Japan.

2.5.2. Practices Routines . We identied three keyorganizational practices that support value creation fromIT. The rst two practices relate to two fundamentalactivities of IT-enabled organizationscommunicationand transactionand the third involves active use of the Internet, one of the most fundamental sociotechnicalinnovations in recent history.

IT Use Intensity for Communication . Devaraj andKohli (2003) make a convincing case for the measure-ment of IT use as a missing link in the relationshipbetween IT investments and rm performance. Brynjolf-sson and Yang (1997) also demonstrate that rms usingmore digital work practices obtain higher performancebenets from their IT investments. We therefore measure

the intensity of IT useboth internal and external. Inter-nal IT communication intensity describes the degree towhich internal communications and work practices areconducted electronically and measures the use of elec-tronic communication media such as email, intranets,and wireless devices for internal communications. Sup- plier facing IT communication intensity describes the

degree to which information exchanges with suppliersare conducted electronically via email, remote wirelessconnections, the Internet, and non-Internet electronicdata interchange (EDI) connections.

Digital transaction intensity (DT) measures the degreeto which both internal and external transactions are con-ducted electronically. Distinct from internal or externalcommunication intensity, transaction intensity measuresthe relative digitization of the transactions rms executewith suppliers and customers and is a linear combinationof two ratios: electronic purchase orders to total pur-chase orders and electronic sales to total sales. Processdigitization in relationships with suppliers can reduceinput costs by reducing procurement time and supplyuncertainties that necessitate stockpiles of inventories;it can do this by reducing prices through greater markettransparency and by reducing the costs of purchase orderand invoice processing. More digital transactions withsuppliers can also reduce coordination costs (Malone1987), transaction costs (Williamson 1975), and agencycosts by increasing transparency and mutual monitoring(Jensen and Meckling 1976). Today, rms with more ITcapital are smaller (Brynjolfsson et al. 1994) and lessvertically integrated (Hitt 1999), indicating that process


8/18


digitization in the supply chain is enabling increasedperformance by pushing key functions outside the rmboundary. IT also has the potential to transform rela-tionships with customers. Digitization of the customerexperience can enable greater customization and a shiftfrom build-to-stock to build-to-order processes, increas-

ing customer satisfaction (Mithas et al. 2005) and reduc-ing the cost of selling (Brynjolfsson and Hitt 2000).Finally, open Internet architectures can reduce inter-

nal and external integration costs. In contrast, proprietaryarchitectures are more complex to connect and maintain,making back-end legacy system integration less efcient.The Internet also allows rms to broaden interactionswith customers by collecting systematic data on purchas-ing decisions and the responsiveness of post-sale cus-tomer service operations. The ability to deliver onlineproduct support, technical assistance, merchandise track-ing, and customer feedback enhances the value of prod-ucts supported by Internet-based applications (Zhu andKraemer 2002). We measure the degree to which rmsemploy Internet architectures in sales force management,employee performance measurement, training, and post-sales customer support, all of which were shown tobenet from IT adoption (Brynjolfsson and Hitt 2000,Brynjolfsson 1996) and were important factors in ourcase studies (see Table 2).

Our case study evidence supports the theoretical con-ceptualization of IT capabilities as systems of interlock-ing practices and competencies and demonstrates thatinvestment allocations, driven by strategy, shape andare shaped by rms practices and skills. For example,

we found that informational investments are critical to7-Eleven Japans business strategy, which is designed tomake stores responsive to even small changes in cus-tomer demand and environmental conditions (Nagayamaand Weill 2004). 7-Elevens total information systemconnects 70,000 computers in stores, at headquarters,and at supplier sites, providing transparency across theentire value chain. Recent sales, weather conditions, andproduct range information are provided graphically toeach store to assist in ordering fresh food, which isordered and delivered three times per day. The result isthat on hot days Tokyos 7-Eleven stores have plentyof cold Bento boxes and on cold days there are lots of hot noodles for sale. The total information system alsoreduces missed opportunities from out-of-stock itemsand the need for large inventories, which in Japaneseretail are space and cost prohibitive. 7-Eleven CEOToshifumi Suzuki explains the companys information-intensive strategy: To produce the best original productswith higher quality than any competitors, we continue tocreate a hypothesis, test it, make another hypothesis, andexamine it over and over But these organizational practices alone are not enough. 7-Eleven Japan workshard to develop rmwide IT skills and/or managerial

involvement to enable and reinforce these practices. 7-Eleven Japan counselors visit each store at least twicea week to work with franchisees to improve their skillsin using data from their information systems to manageand order more effectively. Counselors train employeesto use point of sale, inventory, and weather-tracking sys-

tems to strategically stock and price goods. The point-of-sale and weather-tracking systems are examples of trans-actional and informational applications that exploit theIT infrastructure. The tight relationship between com-pany counselors and stores increases the IT skills of store operators while reinforcing critical IT practices atthe store level, demonstrating the synergy between skills,and IT practices.

7-Eleven Japans total information strategy usesinformation to make more effective business decisions.The information is extracted and summarized usingtransactional IT systems that process 35 million salestransactions and 5 million order transactions per day.

Each day, these transactions are sent to the 7-ElevenJapan information systems center, where they are inte-grated, analyzed, and shared, via informational IT, withall store owners and workers at registers in real time.In addition, as CEO Suzuki explains, the business skillsof IT employees are critical: [We] dont rely on thepoint-of-sale system. IT is merely a tool to achieve busi-ness strategy. We shouldnt use the technology unless wecan understand what the information means on paper.These examples of human resource competency, man-agement competency, and digital transactions illustratehow organizational IT capabilities support transactionaland informational IT assets at 7-Eleven Japan.

Testing Complementarity. Qualitative examples illus-trate how IT assets and organizational IT capabili-ties complement one another in a single rm, butempirical demonstrations of complementarity in largersamples require evidence of the covariance or clus-tering of complementary elements across rms andpositive effects of the copresence of complements onperformance (Milgrom and Roberts 1990, Bresnahanet al. 2002). We therefore test whether IT assets andorganizational capabilities correlate and whether theyexhibit reinforcing interaction effects on rm perfor-mance (Athey and Stern 1998). If IT investment andorganizational IT capabilities are complementary, we

expect the following:Hypothesis 2. Organizational ITC and IT investment

intensity are positively correlated.

We also expect that rms with both more IT invest-ments and stronger organizational IT capabilities per-form better. Thus:

Hypothesis 3. Variables interacting an aggregatemeasure of organizational ITC with IT investment inten-sity by asset class are positively associated with rm performance.


9/18


2.6. Dependent Variables: Distinguishing DifferentDimensions of Firm Performance

Different assessments of IT value have different theo-retical foundations, and empirical results depend heavilyon what questions are asked and how data are mod-eled (Hitt and Brynjolfsson 1996, Kohli and Devaraj

2003). The literature on IT value relates IT investmentsto a variety of performance measures, including pro-ductivity (Brynjolfsson 1996), consumer welfare (Hittand Brynjolfsson 1996, Brynjolfsson 1996), account-ing prot (Weill 1992, Bharadwaj 2000), market valu-ation (Dos Santos et al. 1993, Bynjolfsson and Yang1997, Bharadwaj et al. 1999), and operational perfor-mance (Barua et al. 1995, Zhu and Kraemer 2002).These performance dimensions are distinct (Hitt andBrynjolfsson 1996) and trade off with each other.Anderson et al. (1997) nd that productivity, protabil-ity, and customer satisfaction trade off and that rmsstrategies and industries change the nature of the trade-

offs. Quantitative empirical results concerning IT andorganizational performance vary in part because mea-sures of performance are multidimensional but measuresof IT are typically unidimensional.

To measure associations between IT investments andthe performance of rms that are potentially strate-gically differentiated, we regressed four categories of rm performanceprotability, market valuation, oper-ational performance, and innovationon total IT inten-sity, IT intensity by asset class, and the interactionbetween IT assets and organizational IT capabilities.Protability is measured by net margin and return onassets (ROA) (Bharadwaj 2000), market valuation by

Tobins q (Hitt and Brynjolfsson 1996, Bharadwaj et al.1999, Bynjolfsson and Yang 1997), operational perfor-mance by the cost of goods sold (Barua et al. 1995, Zhuand Kraemer 2002), and product innovation by revenuesfrom new and modied products.

3. Methods3.1. Data and MetricsPrevious researchers have coded types of IT investmentsaccording to the language used in media descriptions(Dos Santos et al. 1993); we asked senior IT execu-tives to subdivide their total IT budgets according todescriptions of the asset classes to better understand themanagement intention for IT investments in each rm.Descriptions of the asset classes and examples of ITassets were used to guide managers in categorizing theirIT investments. All 147 respondents were from large,publicly traded U.S. rms, and Compustat was used toobtain performance and other relevant data during 19992002. Our sample is composed of 58% manufacturingand 42% services rms, which mirrors the composi-tion of the S&P 500 and the Fortune 1000. The sam-ple includes 147 rms over 4 years for a panel of 588rm years between 1999 and 2002, accounting for $448

billion in output in 2001. The survey instrument wasdesigned and pilot tested as part of the National Sci-ence Foundation-funded MIT SeeIT Project. Using theMIT SeeIT instrument, data collection was conductedby Harte Hanks via a random sample of companies inits database, which has been used in previous research

(e.g., Brynjolfsson and Hitt 1996). Survey questions anddescriptions of the asset classes and capability metricsappear in the appendix. Table 3 provides variable de-nitions and descriptive statistics.

We used conrmatory factor analysis to validate ourgrouping of the 18 indicators of ITC into the 6 variablesdescribed in 2.5. Following Straub (1989), Boudreauet al. (2001), and Zhu and Kraemer (2002), we con-sidered the reliability, content validity, and constructvalidity of our measures. We tested both the internalconsistency and the construct reliability of our IT capa-bility metrics. The average factor loading for indicatorsused to construct the six capability variables was 0.70,

and all factor loadings were positive, signicant, andabove the cutoff of 0.4 (Gefen et al. 2000). 3 The con-tent validity of the instruments was based on a reviewof the literature, our case studies, and discussions withmore than 100 IT managers in a variety of industriesat MIT Center for Information Systems Research work-shops. Following Straub (1989), we tested the conver-gent and discriminate validity of our measures to ensuretheir construct validity. According to the t-statistics of individual factor loadings, all independent indicators dis-played highly signicant contributions to the constructsthey were intended to measure, providing condencein their convergent validity. We tested the discriminatevalidity of our constructs by analyzing their internal(within measure) and external (across measure) correla-tions in a correlation matrix containing all independentindicators (Campbell and Fiske 1959, Straub 1989) andfound correlations within measures to be higher thancorrelations across measures in 16 of 18 cases, indi-cating strong discriminant validity. 4 The factor loadingsand t -statistics of convergent validity are reported inTable 4. 5

3.2. Reliability of the DataWe conducted several tests of the reliability of our data.First, as with any self-reported survey data, accuracydepends on the reliability of responses from the IS man-agers. To improve the accuracy of responses, all sur-veys were conducted in person or over the telephone,and efforts were made to ensure that respondents werein management positions responsible for IT investmentsand had detailed knowledge of their rms IT practices. 6Second, in testing for response bias, we found our sam-ple was no different than the largest 3,500 rms inthe United States in terms of total output measured bytotal sales (t -statistic = 0 8), the number of employ-ees (t -statistic = 1 1), total advertising expenditures


10/18


Table 4 Reliability and Validity of ITC Metrics

IT capability Factor Convergentmetric Indicator loading validity

Factor 1: Human Technical skill 0 733 11 97

resource capability Business skill 0 727 11 34

End user skill 0 564

Labor supply 0 620 10 34

Factor 3: Internal Email 0 742 IT use intensity Intranet 0 702 18 91

Wireless 0 509 24 15

Factor 4: Supplier Email 0 731 facing IT use Internet 0 807 12 66

intensity EDI 0 511 11 77

Factor 6: Internet Sales force mgmt 0 698 11 42

capability Performance 0 769 12 30

evaluationTraining 0 857 15 32

Online customer 0 654 support

p < 0 001

(t-statistic = 0 1), total R&D expenditures ( t -statistic =0 9), and the cost of goods sold ( t-statistic = 0 8). Also,the performance of rms in our sample (as measuredby ROA) does not differ from the population of the3,500 largest companies ( t -statistic = 0 5), indicating lit-tle chance of a systematic response bias along the per-formance dimension.

3.3. Control VariablesThree rm-level variables were used to control for

their effects on performance: R&D expenditure , adver-tising expenditure , and rm size . Many previous studiesdemonstrate that R&D expenditures are strongly corre-lated with rm performance and are particularly inu-ential in market valuation metrics such as Tobins q(Montgomery and Wernerfelt 1988, Capon et al. 1990).In addition, advertising expenditures are positivelyrelated to rm performance and are associated with mar-ket valuation and protability in particular (Montgomeryand Wernerfelt 1988, Capon et al. 1990). 7 Firm size iscontrolled for by ln(employees), and expenditure vari-ables (IT, R&D, advertising) are operationalized as ratiosof expenditures to sales to control for the relative pro-duction size of rms. We used industry dummy variablesfrom two-digit standard industry codes (SIC) and sep-arately input the two-digit SIC industry average foreach dependent variable into regressions to control forindustry-level variation. Both specications producedsimilar results for all regressions. We report results basedon the second approach to preserve degrees of freedom.

3.4. Model SpecicationWe rst tested two model specications of the relation-ship between total IT investment intensity and the six

performance variables P s s= 1 6 : a xed-effectsmodel with controls for year and rm effects,

P st = +i

iYear i + 5sITst + st (1)

and an ordinary least squares (OLS) model in each

year regressing performance (lagged by one year) on ITintensity (IT), organizational ITC, and the interaction of IT intensity and ITC (IT ITC):8

P st = s +j

sj C js t + 5sITst 1 + 6ITCs

+ 7 ITst 1ITCs + st (2)

where C j j = 1 4 represents the three rm-levelcontrol variables (ln employees, R&D intensity, andadvertising intensity) and the industry control, and rep-resents the error term.

We then examined relationships between IT invest-

ments in each of the four asset classes AC sk s =1 6 k= 1 4 in 2001 and performance in 2001and 2002 in OLS analysis as follows:

P st = s +j

sj C js t +k

ks ACskt 1 + st (3)

Finally, having analyzed the contributions of differentIT asset classes, we included an interaction term testingthe inuence of ITC on the performance contributionsof each IT asset class:

P st = s +j

sj C js t +k

ksACskt 1 + 9sITCs

+ 10s ACskt 1ITCs + st (4)where AC ITC represents the term interacting eachasset class with the aggregate measure of ITC. We wereunable to reject the hypothesis of no heteroscedasticityaccording to Breusch and Pagan (1979) tests and havereported standard errors according to the White correc-tion (White 1980). We tested for multicollinearity byexamining a correlation matrix of all independent vari-ables and discovered no variables entered simultaneouslyinto any regression with a correlation coefcient greaterthan 0.70.

4. ResultsOur results demonstrate that different IT assets are asso-ciated with different types of performance benets forrms that are generally consistent with their strategicgoals. For example, strategic IT investments are associ-ated with product innovation (and not with other mea-sures of performance), but only transactional investmentsare associated with lower costs. These results supportHypothesis 1 and demonstrate that distinct IT assetshelp explain variation in rm performance. In addi-tion, we nd evidence of complementarities between IT


11/18


Table 5 Total IT Intensity, ITC, and Firm Performance

ROA Net margin Tobins q COGS New Modiedproducts products

Specication FE OLS FE OLS FE OLS FE OLS OLS OLS

Employees 0 72 0 79 0 02 1 722 78 0 11 1 220 99 0 69 0 03 477 99 1 03 1 55

R&D 1 02 0 46 0 06 99 78 1 90 3 291 12 0 65 0 02 411 47 0 88 1 32

Advertising 1 60 2 12 0 04 67 22 0 30 5 60

1 06 1 23 0 04 316 91 1 40 2 38Total IT 41 77 43 76 33 36 9 38 1 06 2 05 1 993 8 2 133 77 44 72 72 62

36 49 47 05 34 95 42 29 1 73 1 94 2 636 9 9 427 10 37 52 52 81ITC 2 97 1 35 0 04 951 46 4 46 1 85

2 27 1 95 0 10 728 72 3 81 5 44Total IT ITC 9 50 10 28 2 47 8 299 96 43 32 121 18

41 80 36 65 1 76 7 389 96 37 86 53 94Cons. 6 731 1 54 5,284 0 65 187 3 0 68 0 32e6 465 93 24 29 29 64

(1,477) 2 74 (1,413) 1 76 69 85 0 08 (0.12e6) 502 18 3 47 4 31Industry No Yes No Yes No Yes No Yes Yes Yes

controls

R2 0 08 0 05 0 06 0 05 0 03 0 04 0 05 0 34 0 02 0 08F value 10 53 0 89 7 10 1 02 3 59 4 27 5 45 2 26 2 71 11 85

Obs. 375 84 376 84 376 85 373 85 75 74

Notes. FE = xed effects. Total IT intensity measured at t 1 in OLS regressions. Robust standard errors under the White correction arereported.

p < 0 10; p < 0 05; p < 0 001.

assets and organizational IT capabilities. IT assets andITC covary signicantly in our sample, demonstratingthat rms with strong ITC demand more IT and viceversa. ITC also displays positive interaction effects withIT assets on a variety of performance measures. Orga-

nizational ITCs strengthen the performance effects of IT investments and also broaden their impacts to newdimensions of performance. These results suggest twoimportant extensions to the resource-based theory of IT:a move away from monolithic conceptualizations of ITtoward a disaggregated view of IT assets and a viewof organizational ITC as a mutually reinforcing systemof practices and competencies that both strengthens andbroadens the performance impacts of IT.

4.1. IT Assets and Firm PerformanceWe nd no association between total IT intensity andrm performance in xed effects or OLS analyses (seeTable 5). These results demonstrate that aggregate ITinvestments, taken alone, provide little advantage forrms (Bakos 1991, Clemons and Row 1991) and mirrorndings that suggest that although total IT capital stock improves rm productivity, it does not contribute to prof-itability (Hitt and Brynjolfsson 1996). Although ITCenhances the performance effects of total IT investmenton all performance dimensions, the interaction effectsare not statistically signicant except in the analysis of sales from modied products. 9

Firms in our sample allocate IT investments dif-ferently. Manufacturing rms spend 6% more on IT

infrastructure than nonmanufacturing rms, and nan-cial services rms spend 8% more on strategic, 2%more on transactional, and 8% less on informational ITassets than other rms. Signicant variation also existsacross rms within industries reecting different strate-

gic choices.Table 6 reports estimates of relationships between thefour IT asset classes in 2001 and rm performance in2001 and 2002. Infrastructure investments made in 2001are negatively associated with ROA in 2001 and net mar-gin in 2001 and 2002. However, over time, the negativeassociation with protability and return diminish. Theassociation between infrastructure investments made in2001 and ROA in 2002 is positive but not signicant,and the loss of net margin is smaller and less signi-cant in 2002. Investments in infrastructure also disruptshort-term efforts at product innovation as measured byrevenues from modied products in 2001. The associ-ation between infrastructure investments in 2001 andTobins q in the same year is positive and signicant,indicating a positive relationship between infrastructureinvestments and market valuation. These ndings areconsistent with prior research that describes infrastruc-ture investments as disruptive in the short term, pro-ducing high up-front implementation and restructuringcosts; but these investments are effective in improvingbusiness performance in the long term, enabling newapplications and reducing long-term costs through inte-gration (Weill and Broadbent 1998, Duncan 1995).


12/18


Table 6 IT Investment Allocations by Asset Class and Firm Performance

ROA Net margin Tobins q COGS New Modiedproducts products

2001 2002 2001 2002 2001 2002 2001 2002 2001 2001

Employees 0 54 0 495 1 01 0 94 0 001 0 002 1,493.1 1 623 5 1 22 0 1270 80 1 12 0 97 0 65 0 054 0 021 (503.1) 446 4 1 34 1 51

R&D 1 48 1 98 1 63 0 70 0 039 0 066 377.5 116 67 1 74 40 61

1 72 1 27 2 18 0 60 0 036 0 024 (284.2) 350 45 0 979 0 808Advertising 4 18 2 19 5 17 2 69 0 021 0 022 114.8 493 61 0 104 5 36

3 84 0 95 5 31 1 32 0 061 0 041 (402.4) 406 99 1 60 1 52

IT variablesTransactional 273 49 192 08 309 77 104 4 5 36 3 02 8,047.5 160 990 169 02 836 69

186 82 212 67 201 15 163 79 14 29 9 93 (82,468) (101,138) 344 58 557 22Informational 313 76 289 74 269 47 167 4 12 00 5 08 5,798.5 17,651 277 72 1 056 56

150 98 92 63 162 60 76 3 10 34 5 44 (22,526) (28,104) 214 26 242 11Strategic 46 99 117 60 332 87 338 8 43 49 6 39 38,544 19,598 620 59 2 891 72

438 48 231 35 330 23 300 3 37 85 18 54 (78,713) (90,352) 690 90 905 39Infrastructure 224 5 74 88 377 59 179 9 16 83 4 24 30,288 43,872 341 41 48 27

112 9 145 34 159 98 95 6 9 81 4 47 (34,889) (42,782) 222 58 264 77

Industry Yes Yes Yes Yes Yes Yes Yes Yes Yes Yescontrols?

R2 0 10 0 07 0 14 0 13 0 09 0 04 0.34 0 34 0 06 0 17F value 1 99 3 60 1 69 5 18 1 14 2 83 2.30 2 56 1 45 14 83

Obs. 103 95 103 95 104 98 103 98 90 90

Notes. OLS regressions. Robust standard errors under the White correction are reported.p < 0 10; p < 0 05; p < 0 001.

Informational investments are positively correlatedwith ROA and net margin in both 2001 and 2002,demonstrating a positive association with protability,although the expected association between informationalinvestments and lower costs is not observed. Trans-

actional investments are associated with lower COGSin 2002, supporting the argument that transactionalinvestments reduce costs. Strategic investments are pos-itively associated with revenues from modied products,demonstrating strong support for the association betweenstrategic investments and innovation. As expected, R&Dinvestment is associated with gains in product inno-vation but is (unexpectedly) negatively correlated withTobins q in 2002. 10 These results demonstrate that dif-ferent IT assets produce different types of performancebenets for rms that are consistent with their strategicgoals, as outlined in Table 1. Firms with similar levelsof total IT intensity allocate investments differently inour data. By disaggregating IT investments, we observethe relative performance effects of different types of ITthat may be obscured in total IT investment data.

4.2. Testing Complementarity Between IT Assetsand Organizational IT Capabilities

Table 7 presents the spearman partial rank order corre-lations between total IT investment intensity, IT invest-ment intensity by asset class, and ITC, controlling forrm size (see Table 7, row 6). The correlations are allpositive and signicant, indicating covariance between

assets and capabilities. These results support Hypothe-sis 2a complementary relationship between ITC andIT investment intensity.

Table 8 presents the results of independent year OLSregressions, assessing the relationships between specicIT assets, ITC, and the interaction between assets andcapabilities on performance. 11

ITC both strengthens the performance effects of ITassets on their primary performance dimensions (thoseconsistent with their strategic purpose) and broadens theimpact of IT assets to other performance dimensions.The results for net margin reveal signicant positiveinteraction effects among ITC and transactional invest-ments, informational investments, and strategic invest-ments. The coefcients on strategic and informationalinvestments are also positive and signicant, as theywere in the regressions of IT investments alone. Infra-

structure investments are again negatively associatedwith protability, but the interaction effect betweeninfrastructure and ITC is positive and signicant onprot, indicating that rms with higher ITC scoresachieve gains, not losses, from infrastructure invest-ments. Firms with higher ITC have greater protabil-ity when they invest more in informational, strategic,and infrastructure assets, relative to the average rm.We again nd a positive relationship between ROA andinformational investments, which reiterates the relation-ship reported in Table 6.


13/18


Table 7 Correlations Between IT, Organizational Capabilities, and Performance

Measure 1 2 3 4 5 6 7 8 9 10 11 12

1. Total IT 1 002. Infrastructure 0 83 1 003. Transactional 0 55 0 32 1 004. Informational 0 62 0 36 0 44 1 00

5. Strategic 0 49

0 34

0 47

0 39

1 006. ITC 0 33 0 27 0 22 0 30 0 33 1 007. ROA 0 15 0 05 0 06 0 07 0 03 0 01 1 008. Net margin 0 13 0 06 0 09 0 08 0 03 0 03 0 83 1 009. Tobins q 0 01 0 06 0 03 0 05 0 04 0 05 0 44 0 40 1 00

10. COGS 0 13 0 22 0 14 0 04 0 09 0 05 0 00 0 07 0 01 1 0011. Modied products 0 10 0 20 0 01 0 03 0 06 0 10 0 23 0 18 0 04 0 16 1 0012. New products 0 15 0 04 0 06 0 03 0 01 0 10 0 11 0 07 0 03 0 08 0 41 1 00

Notes . Spearman partial rank order correlations of total IT spending, IT spending by asset class, organizational ITC in 2001, and perfor-mance measures in 2002 controlling for rm size.

p < 0 05.

The interaction effect of infrastructure investmentsand ITC on net margin presents a clear example of

the importance of organizational capabilities in explain-ing variance in the returns to IT investments acrossrms. For a rm with average organizational capabil-ities, a $1 (sales adjusted) increase in infrastructure

Table 8 IT Investment Allocations and Interactions with IT Capabilities on Firm Performance

ROA Net margin

Asset ITC Interaction Asset ITC InteractionR2 F (SE) (SE) (SE) R2 F (SE) (SE) (SE)

Transactional t 1 ITC 0 13 3 69 19 5 0 48 310 4 0 14 6 96 86 1 1 2 348 1

244 5 0 63 253 8 179 1 1 8 182 0Informational t 1 ITC 0 13 2 79 357 8 0 21 117 9 0 15 11 3 403 7 1 5 353 0

211 7 0 98 239 9 164 4 1 7 194 4Strategic t 1 ITC 0 13 4 84 31 4 0 27 244 5 0 15 7 78 297 5 0 99 428 5

216 3 0 53 217 6 285 6 1 7 205 5Infrastructure t 1 ITC 0 13 3 29 3 61 0 25 55 2 0 15 6 97 365 5 0 91 151 6

208 9 0 94 95 5 159 4 1 7 92 4

COGS Tobins q

Transactional t 1 ITC 0 40 2 65 2 2e5 1 166 9 0 93e5 0 06 5 54 0 20 0 09 18 8

(1.3e5) (987.0) (0.65e5) 11 1 0 08 8 9Informational t 1 ITC 0 39 2 01 0.52e5 1 186 9 0.19e5 0 04 2 36 5 02 0 09 4 23

(0.79e5) (995.7) (1.0e5) 10 1 0 08 12 8Strategic t 1 ITC 0 40 2 59 0 35e5 1 142 1 0 74e5 0 05 3 28 3 4 0 08 15 4

(1.1e5) (1,002.1) (0.62e5) 18 6 0 08 8 5Infrastructure t 1 ITC 0 40 3 23 1.3e5 999 0 0 78e5 0 07 2 89 4 6 0 08 8 6

(0.83e5) (967.4) (0.46e5) 5 5 0 08 3 4

Modied products New products

Transactional t 1 ITC 0 29 20 79 525 5 1 3 84 2 0 09 3 24 734 2 0 76 722 2

669 7 1 9 661 3 369 3 1 04 354 9Informational t 1 ITC 0 29 21 45 1 471 3 2 9 371 4 0 08 1 74 22 1 1 04 468 2

411 8 2 5 577 2 240 9 0 76 576 8Strategic t 1 ITC 0 29 23 90 3 401 5 2 9 371 4 0 10 4 04 936 1 0 79 818 7

802 9 2 5 577 2 625 9 0 92 338 7Infrastructure t 1 ITC 0 29 22 32 482 3 1 5 7 9 0 11 3 80 375 1 2 8 422 4

485 0 2 5 254 6 328 8 1 4 152 3

Notes. OLS regressions with robust standard errors. All control variables and other IT assets are included in regressions but not reported.p < 0 10; p < 0 05; p < 0 001.

investment is associated with a $366 decrease innet margin in the following year ( = 365 5). For

a rm with below-average organizational capabilities(ITC = 3; 2 standard deviations below the mean),a $1 increase in infrastructure investment is associatedwith an $820 decrease in net margin, and for a rm


14/18


with above-average organizational capabilities (ITC = 3;2 standard deviations above the mean), a $1 increasein infrastructure investment is associated with a $89increase in net margin the year after the investmentis made.12

The coefcient on transactional investments is again

negative and signicant on COGS, but the interactioneffect, although negative, is not signicant. Infrastruc-ture investments have a marginal positive associationwith COGS, but the interaction effect of infrastructureand ITC is negative and signicant on cost. The aver-age rm sees higher short-term costs with more infra-structure investment, but rms with above-average ITCsee short-term cost reductions. All else being equal,rms with strong ITC scores are associated with highermarket value when they make transactional, infrastruc-ture, and strategic investments, and strong ITC enablesgreater revenues from new products when transactional,strategic, and infrastructure investment intensities are

high. ITC seems to have little enabling effect on rev-enues from modied products, where informational ITintensities are negative and signicant, and strategic ITintensities return strong positive associations, as they hadin regressions involving the IT investments alone. Wend the inclusion of ITC explains between 2% and 12%more performance variance than our models of IT assetsalone, as seen in increased R2 values in Table 8 com-pared with Tables 5 and 6.

4.3. DiscussionIn our data, the average rm experiences performancebenets from investments in different IT assets alongdimensions consistent with the strategic purpose of the asset. However, rms with greater ITC experienceboth stronger performance effects along expected dimen-sions and a broadening of performance impacts to othermeasures.

Infrastructure investments produce high up-front im-plementation and restructuring costs but support futurebusiness value by enabling new applications and reduc-ing long-term costs through integration, creating apattern of lagged benets (Duncan 1995, Weill andBroadbent 1998, Broadbent et al. 1999). Infrastructurebenets are lagged because new applications that lever-age new infrastructure take time to deploy, and importantorganizational factors mediate their implementation anduse. For example, governance structures in most rmsseparate decision making on applications from decisionmaking on infrastructure, with the former remainingunder the authority of the business and the latter with theIT function (Weill and Ross 2004). This organizationalseparation makes building effective applications on topof new infrastructure challenging. However, rms withstrong ITC experience short-term gains, not losses frominfrastructure investments, and broaden their perfor-mance benets to include innovation, prot, and lower

costs (see Table 8). These rms develop skills and enactpractices that enable smoother infrastructure implemen-tations and more effective decision-making processesthat govern the integration of infrastructure with newapplications. Tight relationships between business unitsand the IT function (management capabilitiesMC),

strong cross-functional IT and business skills (humanresources capabilitiesHRC), and greater digitizationof important business processes such as ordering andsales (digital transaction intensityDT) support integra-tion of infrastructure with new applications and enablerms to more quickly and effectively utilize applicationsto improve a broader set of performance dimensionsbeyond market value.

Aggregate measures of IT obscure the performanceimplications of distinct IT assets. For example, total ITintensity is not correlated with product innovation in ourdata, but strategic IT investments strongly support inno-vation, and infrastructure investments are negatively cor-related. When evaluated as a monolith, IT seems to haveno innovation effect, when in fact different IT assetshave conicting innovation implications. Performanceitself is also multidimensional. Firms can pursue differ-ent strategies with distinct and at times mutually exclu-sive performance implications. Cost leadership may beorthogonal to innovation, in terms of both IT investmentsand performance.

4.4. Limitations and Future ResearchAlthough our research opens new avenues for explainingIT value creation from a resource-based perspective, it

has some limitations. First, our data set is partly cross-sectional, and although we use lagged measures of per-formance to control for reverse causality, causal claimscannot be made about disaggregated IT assets. How-ever, simultaneity is less likely to bias our results, giventhat observed performance effects match hypothesizedIT assets. A similar defense of causality is used by Bartelet al. (2004), who argue that observation of specic tech-nology impacts on expected performance measures butnot on others supports a causal argument (see page 221).Although our data set is the largest we encountered withdetailed allocations of IT investments, we did not haveenough data to test the long-term effects of IT or toexamine path dependencies over time. As infrastructureand strategic assets may impact rm performance yearsafter investments are made, and as IT generally requiresperiods of learning and adjustment to attain full value,our results may underestimate its effects. In the future,larger longitudinal data sets will be needed to explorecausal relationships between IT investment allocationsand rm performance (Bharadwaj et al. 1999, p. 1020).

Second, our assessments of organizational capabili-ties are measured with ordinal self-reported data froma single respondent. These measures are vulnerable to


15/18


respondents subjective assessments of their organiza-tions and to single respondent bias. More objective mea-sures of organizational capabilities could be collectedin future work by logging the dollars spent on tech-nical training, the education and training backgroundsof IT employees, and policies that codify the distribu-

tion of decision rights between business units and the ITfunction.Third, our division of IT investments into the four

asset classes is but one way to characterize rmsinvestment allocations. Other breakdowns might includeaggregations of investments in particular IT projectssuch as enterprise resource planning (ERP), supply chainmanagement (SCM), or customer relationship manage-ment (CRM) projects or other theoretical frameworksdistinguishing different types of IT. We hope our work can serve as a useful starting point for these endeavors.

Finally, a natural question emerges from our results:Do certain organizational IT capabilities support certain

IT assets in particular? Our work is intended to beginthis line of inquiry, yet we examine organizational ITcapabilities as a system because of their complemen-tarity as a group of practices and skills. Specic asset-capability synergies could exist, however, and we encour-age the investigation of such relationships in future work,although larger data sets will be needed to test suchnuanced propositions with sufcient statistical power.

Appendix. Denition and Construction of IT Capability Indicators and IT Investment Asset Classes

Variable Description Construction

1. Human resource

capability

Technical skills of IT staff Given a scale of 15, with 1 being inhibits signicantly,

3 being no effect, and 5 being facilitates signicantly,please rate whether the technical skills of IT staff facilitateor inhibit new technology investments at your company.

Business skills of IT staff please rate whether the business skills of IT staff facilitateor inhibit new technology investments at your company.

IT skills of end users please rate whether the IT skills of end users facilitateor inhibit new technology investments at your company.

Ability to satisfy demand forhighly skilled IT labor

please rate whether the ability to hire new IT staff facilitatesor inhibits new technology investments at your company.

2. Managementcapability

Degree of senior managementcommitment to IT projects

please rate whether the degree of senior managementsupport for IT projects facilitates or inhibits newtechnology investments at your company.

Degree of business unitinvolvement in IT decisions

please rate whether the degree of business unitinvolvement in IT projects facilitates or inhibits new

technology investments at your company.3. Internal IT use Intensity of electronic

communication media such asemail, intranets, and wirelessdevices for internalcommunications

please rate how important the following methods are forinternal communications: (a) email, (b) companyintranets, (c) wireless devices.

4. Supplier facing ITuse

Intensity of electroniccommunication media such asemail, intranets, and wirelessdevices for communicationswith suppliers and supplierfacing work practices

please rate how important the following methods are forcommunications with suppliers: (a) email, (b) companyintranets, (c) wireless devices.

4.5. ConclusionMany researchers have examined the productivity andbusiness value of rm-level IT investments. However,results have varied across performance measures, andsignicant rm level variation in the returns to ITinvestments remained unexplained. We complement and

extend recent resource-based theories of IT value byunpacking the measurement of IT into different assettypes that explain additional performance variation. Wealso nd that rms derive greater value per IT dollarby having stronger organizational IT capabilities. Theseresults suggest a move away from monolithic concep-tualizations of IT toward a disaggregated view of ITassets and a view of organizational IT capabilities as amutually reinforcing system of practices and competen-cies that both strengthens and broadens the performanceimpacts of IT.

AcknowledgmentsThe CISE/IIS/CSS Division of the U.S. National ScienceFoundation (0085725) and the MIT Center for InformationSystems Research supported this research. The authors thank Nils Fondstad, Steve Kahl, George Westerman, seminar par-ticipants at Boston University and MIT, the Associate Editor,and three anonymous reviewers for many valuable comments.


16/18


Appendix (contd.)

Variable Description Construction

5. Digital transactions Degree of digitization inpurchasing

Electronic purchase orders / Total purchase orders

Degree of digitization in salesto customers

Electronic sales/Total sales

6. Internet capability Degree to which rms useInternet architectures in sales force management

Given a scale of 15, with 1 being no use of theInternet and 5 being fully automated via the Internet,please identify to what extent your company uses Internettechnology to perform sales force management .

in employee performance measurement

please identify to what extent your company uses Internettechnology to perform employee performance measurement .

in training please identify to what extent your company uses Internettechnology to perform employee training.

in post-sales customer support

please identify to what extent your company uses Internettechnology to perform post-sales customer support.

IT investment variablesTotal IT $ What were the total expenditures on IT in millions of dollars for the entire company, including both

internal and outsourced expenditures? Please assume that IT includes all computers, software,data communications (including via phone line), and people dedicated to providing IT services.

Infrastructure Of the rmwide IT expenditure identied in Question 2, what percentage would you classifyas IT infrastructure? Please consider IT infrastructure as the base foundation of IT capability budgeted for and provided by the I/S function and shared across multiple applications or business units . This infrastructure usually includes the network, help desk,data centers, etc. but excludes applications.

Transactional Please consider transactional IT as investments in IT made to cut operating costs(e.g., reduce costs of preparing and sending invoices).

Informational to provide information. This would include information for accounting, managing quality,executive information systems, performance management, etc.

Strategic to increase or protect your sales or market share by providing improved customerservice or products (e.g., online product catalog).

Endnotes1The case studies were developed using unstructured and semi-structured interviews with upper-level IT management employ-ees, examination of archival data, historical data from thepress, and unpublished rm documents, publicly available per-formance data, and the rms websites.2Before constructing measures of the practices and capabili-ties, we asked managers in research workshops to examine ourframework to make sure our theory reected their experiences.Based on these discussions and our case data, we identied 18indicators of the competencies and practices. Denitions of thevariables and their operationalization appear in the appendix.3As our metrics are designed to measure multiple componentsor dimensions of a construct rather than multiple measuresof the same underlying construct, Cronbachs estimates areless useful in assessing internal consistency. However, esti-mates ranged from 0.44 to 0.73. The alpha for internal IT useintensity was signicantly lower than for the rest of our con-structs (0.44). This may be because of its considerably highmean (4.4) and low variance (S.D. = 0 8), indicating that emailintensity was high for most rms in our data set.4External email intensity and internal email intensity weremore highly correlated with each other than with their respec-tive constructs (internal IT use intensity and external IT useintensity), indicating that internal and external email use arehigh for most rms in concert.

5Factors 2 (MC) and 5 (DT) are excluded from Table 4because they are two item factors to which these statistics donot readily apply.6All respondents reported intimate knowledge of their rmsIT practices and close proximity to IT investment decisions.If a respondent was unfamiliar with IT assets and budgeting,the interview was terminated and a replacement respondentwas sought. Of respondents, 59% were the CIO of the rm,25% were CTOs, 13% were IT budget analysts or administra-tors of IT systems, and 3% were CFOs.7Not all rms report R&D and advertising expenditures.After lling in data from other available sources, follow-ing Bharadwaj et al. (1999) and Montgomery and Wernerfelt

(1988), we input industry average values for each rm forwhich data remained missing.8We also tested a pooled OLS model with panel-correctedstandard errors, with corrections for both autocorrelation andheteroscedasticity, where the error term was modeled as anAR1 process with diminishing uniformly over time androbust estimation of standard errors. The results were qualita-tively unchanged.9R&D intensity is positively associated with product innova-tion measured by revenue from modied products but nega-tively associated with revenue from new products. This mayindicate that rms in our sample rely more on incremental


17/18


innovation than on discontinuous changes in their product lines(e.g., Nelson and Winter 1982, Tushman and Anderson 1986).10Bharadwaj et al. (1999) also observe the surprising nega-tive association between R&D investment and Tobins q. Cit-ing other studies that found the same result, they conclude,The results are consistent with the ndings of prior empiricalefforts that attempted to estimate a relationship between [con-trols for R&D expenditures] and q (Bharadwaj et al. 1999,p. 1019). As our coefcient estimates ( 0 066) are well withinthe range of previous estimates (0.00 to 0 15), we are sat-ised that our results are consistent with prior evidence andreect the true market value of R&D capital.11To facilitate more meaningful interpretations, invest-ment intensities and capability metrics were centered andnormalized. The results reported are of regressions on depen-dent variables measured in 2002, one year lagged from obser-vations of investment and capability measures. All controlvariables and other asset classes are included in the analy-ses, although their coefcients are not reported. As innovationvariables are derived from our survey, we do not have laggedmeasures of these variables in relation to investments in par-ticular asset classes.12The derivation of these results is as follows: Y = + 1X1+

2X2 + 3 X1X2 + = + X1 2 + 1 + X2 3 X1 + .Therefore, the coefcient on X1 = 1 + X2 2.

ReferencesAnderson, E. W., C. Fornell, R. T. Rust. 1997. Customer satisfaction,

productivity, and protability: Differences between goods andservices. Marketing Sci. 16(2) 129145.

Athey, S., S. Stern. 1998. An empirical framework for testing theoriesabout complementarity in organizational design. NBER WorkingPaper 6600, Cambridge, MA.

Autor, D., F. Levy, R. Murnane. 1998. Computing inequality: Havecomputers changed the labor market? Quart. J. Econom. 113 (4)11691214.

Bakos, Y. 1991. A strategic analysis of electronic marketplaces. MISQuart. 15(3) 295310.

Barney, J. B. 1991. Firm resources and sustained competitive advan-tage. J. Management 17(1) 99120.

Bartel, A., C. Ichniowski, K. Shaw. 2004. Using insider econometricsto study productivity. Amer. Econom. Rev. 2(1) 217223.

Barua, A., C. H. Kriebel, T. Mukhopadhyay. 1995. Information tech-nology and business value: An analytical and empirical investi-gation. Inform. Systems Res. 6(1) 223.

Bharadwaj, A. S. 2000. A resource-based perspective on IT capabilityand rm performance: An empirical investigation. MIS Quart.24 (1) 169196.

Bharadwaj, A. S., S. G. Bharadwaj, B. Konsynski. 1999. Informa-tion technology effects on rm performance as measured byTobins q. Management Sci. 45(7) 10081024.

Boudreau, D., D. Gefen, D. W. Straub. 2001. Validation in MISresearch: A state-of-the-art assessment. MIS Quart. 25(1) 124.

Bresnahan, T. F., E. Brynjolfsson, L. M. Hitt. 2002. Informationtechnology, workplace organization, and the demand for skilledlabor: Firm-level evidence. Quart. J. Econom. 117 (1) 339376.

Breusch, T., A. Pagan. 1979. A simple test for heteroscedasticity andrandom coefcient variation. Econometrica 47 12871294.

Broadbent, M., P. Weill, B. S. Neo. 1999. Strategic context and pat-terns of IT infrastructure capability. J. Strat. Inform. Sys. 8(2)157187.

Brynjolfsson, E. 1996. The contribution of information technology toconsumer welfare. Inform. Systems Res. 7(3) 281300.

Brynjolfsson, E., L. M. Hitt. 1995. Information technology as a factorof production: The role of differences among rms. Econom.

Innov. New Tech. 3(4) 183199.Brynjolfsson, E., L. M. Hitt. 1996. Paradox lost? Firm-level evidence

on the returns to information systems spending. Management Sci. 42 (4) 541558.

Brynjolfsson, E., L. M. Hitt. 2000. Beyond computation: Informationtechnology, organizational transformation and business perfor-mance. J. Econom. Perspectives 14(4) 2348.

Brynjolfsson, E., S. Yang. 1997. The intangible benets and costsof investments: Evidence from nancial markets. J. I. DeGross,K. Kuldeep, eds. Proc. 18th Internat. Conf. Inform. Systems ,Association for Information Systems, Atlanta, GA.

Brynjolfsson, E., T. Malone, V. Gurbaxani, A. Kambil. 1994. Doesinformation technology lead to smaller rms? Management Sci.40(12) 16451662.

Campbell, D. T., D. W. Fiske. 1959. Convergent and discriminant val-idation by the multitrait-multimethod matrix. Psych. Bull. 56(2)81105.

Capon, N., J. U. Farley, S. Hoenig. 1990. Determinants of nan-cial performance: A meta analysis. Management Sci. 36(10)11431159.

Clemons, E. K., M. C. Row. 1991. Sustaining IT advantage: The roleof structural differences. MIS Quart. 15 (3) 275292.

Cohen, W. M., D. A. Levinthal. 1990. Absorptive capacity: A newperspective on learning and innovation. Admin. Sci. Quart. 35(1)128152.

Devaraj, S., R. Kohli. 2003. Performance impacts of information tech-nology: Is actual use the missing link? Management Sci. 49(3)273289.

Dewan, S., C. Min. 1997. The substitution of information technologyfor other factors of production: A rm level analysis. Manage-ment Sci. 43 (12) 16601675.

Dierickx, I., K. Cool. 1989. Asset stock accumulation and sus-tainability of competitive advantage. Management Sci. 35(12)15041511.

Doll, W. J., G. Torkzadeh. 1998. Developing a multidimensional mea-sure of system use in an organizational context. Inform. Man-agement 33(4) 171185.

Dos Santos, B. L., K. Peffers. 1995. Rewards to investors in innova-tive information technology applications: First movers and earlyfollowers in ATMs. Organ. Sci. 6(3) 241259.

Dos Santos, B. L., K. Peffers, D. C. Mauer. 1993. The impact of information technology investment announcements on the mar-ket value of the rm. Inform. Systems Res. 4(1) 123.

Dosi, G., R. Nelson, S. G. Winter. 2000. The Nature and Dynamic of Organizational Capabilities . Oxford University Press, Oxford,UK.

Duncan, N. B. 1995. Capturing exibility of information technologyinfrastructure: A study of resource characteristics and th

Date post:	06-Apr-2018
Category:	Documents
Upload:	smriti-pandey
View:	220 times
Download:	0 times

Org Sci Final

Documents