Success of Data Resource Management in Distributed Environments: An Empirical Investigation

Managing Distributed Environments

Success of DataResourceManagement inDistributedEnvironments: AnEmpiricalInvestigation1

By: Hemant JainUniversity of Wisconsin-MilwaukeeSchool of Business AdministrationMilwaukee, WI [email protected]

K. RamamurthyUniversity of Wisconsin-MilwaukeeSchool of Business AdministrationMilwaukee, WI [email protected]

Hwa-Suk RyuSamsung Data Systems Co. Ltd.Imkwang Bldg., 219-1, Migun Dong.Seodaemun-Gu, Seoul, 120-020SOUTH KOREA

Abstract

The trend toward distributed processing hassignificantly increased the awareness of dataas a key corporate resource and underscoredthe importance of its management. In spite ofthis, there is a lack of empirical investigation ofissues related to data resource management(DRM) in distributed processing environments.Being perhaps the first empirical attempt, thisexploratory study identifies four informationsystems (IS) variables related to DRM in a dis-tributed environment. It also seeks to examinethe notion of gestalt fit to describe the nature ofthe relationships among these variables. Inaddition, the study evaluates whether intemallycongruent groups outperform their opposites inrealizing DRM success. The results of clusteranalysis support the view of gestalt fit by identi-fying five clusters. The results also suggest thatorganizations represented by a well-blendedconfiguration of high intersite data dependence,high centralization of IS decisions, high con-centration IS resources at the central site, andlow DRM-related autonomy granted to localsites appear to realize a greater degree ofDRM success than the other groups. The impli-cations of the study are discussed, and furtherresearch directions are proposed.

Keywords: Data resource management, dis-tributed processing, distributed databases,gestalt fit, cluster analysis, autonomy,centralization, intersite data dependence

ISRL Categories: AI0104, AI0401, AI0402,AI0403, AI0607, AI0702, EG0101,EG0102, EG0103, EG0201, EG0202

Masoud Yasai-ArdekaniUniversity of Wisconsin-MilwaukeeSchool of Business AdministrationMilwaukee, WI [email protected]

llzak Benbaszt was the accepting senior editor for thispaper.

Introduction

Significant developments in hardware,telecommunications, and database technolo-gies, accompanied by greater computer litera-cy of end-users, allow organizations to bemore responsive by providing timely informa-tion to users in widely dispersed locations

MIS Quarter/y/March 1998 1


(Cash et al. 1992; O’Brien-1993). To effectivelyachieve this goal, organizations are movingtoward distributed processing systems thatallow local processing and storage of data andgreater end user control. Additional benefitsoffered by these systems include greater flexi-bility, enhanced reliability, faster responsetime, lower costs, and improved performance(Bray 1982; Cash et al. 1992). In order to real-ize these benefits, organizations are downsiz-ing or re-engineering fairly large and complexsystems to client/server and cooperative pro-cessing environments (Charan 1991; Schlack1991). However, in a distributed environment,there is much greater propensity for prolifera-tion of locally developed systems on a varietyof platforms. This results in redundant andfragmented data sources with their accompa-nying problems of data inconsistency, security,and integrity. Thus, there is increasing needfor companies to recognize data as an impor-tant corporate resource that they should care-fully plan and manage (Niederman et al.1991).

In many organizations, data resource manage-ment (DRM) is responsible for a number data planning and DRM policy functions, suchas developing and enforcing data naming con-ventions, data dictionary standards, and dataintegrity and security policies (Guide 1977;Kahn 1983; Tillmann 1984). The technicalDRM functions include managing operation ofthe database, data modeling and design, data-base protection, documentation, and educationand support (Auerbach 1975; Gillenson 1982;Kahn 1983; Leong-Hong and Marron 1978;Weldon 1981).

The complexity and the importance of DRMincrease significantly in a distributed process-

.ing environment. In this environment, DRMshould support local autonomy and controlwhile ensuring adherence to an organization’sdata management policies and standards.Poorly organized DRM functions in distributedenvironments can result in important corporateinformation being locked in a variety of sys-tems. This makes it difficult to compile, inte-grate, and consolidate information, and tointerpret and share data across applicationsand sites. Furthermore, conflicts may arise

among local sites and with corporate groupson issues of responsibility, accountability, andcontrol over data resources, thereby affectingorganizational performance. On the otherhand, an effectively organized DRM functionoffers a number of benefits such as reducingerrors and increasing the ability to access pre-viously unavailable information (Goodhue et al.1988).

Success of DRM requires a set of processesfor effectively formulating and implementingF)RM policies through data administration anddata base administration functions (Laudonand Laudon 1991). A hierarchical structurebased on a central DRM with global responsi-bility for DRM functions and local DRMs withresponsibility for DRM functions at local siteshas been proposed (Jain and Ryu 1988).However, adequate guidelines for organizingeffective DRM functions in a distributed envi-ronment do not exist (Goodhue et al. 1992a).

A DRM function can be considered effectiveand successful if it meets key objectives suchas maintaining data integrity and data avail-ability and allowing standard enforcement. Theeffectiveness of the DRM function in a distrib-uted environment is likely to depend uponmany factors including the need for sharinginformation among the various sites, the role ofIS within the firm, the nature of IS leadershipstyle, the structure of the IS function, the pat-tern of IS resource deployment across sites,the demand for data security, and the organi-zation of the DRM function. However, greaterr)RM effectiveness results in conditions wherethere is congruence among key IS factors. Theiclea that congruence or fit among key organi-zational and environmental factors yieldsgreater effectiveness underscores much of thecontingency theories of organizations. Theimportance of the fit concept has been quiteextensively discussed in the organization theo-ry and strategic management literature.Several approaches for empirical testing of thefit concept have been proposed (Venkatraman11989; Venkatraman and Camillus 1984).Researchers have suggested that high perfor-rnance results froma proper alignment/congru-ence between the following: environment andorganizational structure (Burns and Stalker

2 MIS Quarterly/March 1998


1961), strategy and structure (Chandler 1962;Grinyer et al. 1980), strategy and managerialcharacteristics (Hambrick and Mason 1984),and strategy and reward systems (Norburnand Miller 1981).

In the IS literature also, researchers havebegun to examine the effect of fit among orga-nizational factors on IS performance. A taxon-omy of possible alignments between IS andorganizational structure has been presentedand the argument made that ensuring properalignment of the organization’s structuraldesign and the structure of its IS function is astrategic necessity for many organizations(Leifer 1988). However, no large scale empiri-cal study exists in the IS literature thatexplores the concept of alignment/congruenceor fit between the relevant organizational fac-tors and the performance of one or more ISsubfunctions, such as data administration.

The objective of this study is to examine howthe concept of fit can be applied to complex ISissues such as those related to designs ofeffective IS organizations. Specifically, thisstudy focuses on the nature of interrelation-ships among four IS factors and their impacton DRM success. These IS factors includeintersite data dependence, reflecting the needfor sharing of data among the various sites;centralization of IS decisions, depicting thestructure of IS decision making; concentrationof IS resources, representing the pattern of ISresource deployment across sites; and DRM-related autonomy, reflecting the organizationof the DRM function at various sites. Weexpect that congruence among these four ISfactors will enable corporations to realizegreater DRM success and propose a numberof congruent configurations. Cluster analysis isrelied upon to identify groups of firms thatshare common characteristics along the fourIS factors noted above. This method of analy-sis is most useful for evaluating fit among anumber of variables. The paper is structuredas follows: First, the conceptual framework forthe study is presented. The operationalizationof the constructs used in the study and evalua-tion of their psychometric properties follows.The data analysis approaches are explained,after which the results are presented.

Discussion of the results, their implications forIS practice and research, limitations of thestudy, and conclusions follow.

Conceptual Framework

The success of the DRM function in a distrib-uted environment can manifest in several dif-ferent ways. Success may be reflected by thedegree to which preset DRM objectives arerealized. In most organizations, DRM objec-tives relate to improvements in efficiency andeffectiveness of the DRM function. Such objec-tives include maintaining data integrity, accura-cy, security, and availability; providing timelydata; designing efficient data distributionstrategies; enhancing operational efficiency;setting and enforcing standards; facilitatingenhanced data sharing and reducing redun-dancy; developing strategic data plans; andtraining IS personnel and end users, amongothers.

In a distributed environment, success of theDRM function depends on a number of IS fac-tors. The limited amount of research in thisdomain suggests that some of the key factorsthat could influence DRM success includeintersite data dependence, centralization of ISdecisions, concentration of IS resources, andDRMorelated autonomy (Ceri and Pelagatti1984; Goodhue et al. 1992a, 1992b; Jain andRyu 1988). Some of these factors, such asintersite data dependence, may depend on thenature of the product/market domains in whichthe firm operates and the firm’s overall organi-zational structure. Thus, the IS function maynot have much influence in determining theextent of intersite data dependence. On theother hand, the IS function may have a greatdeal of influence on the degree of centraliza-tion of IS decisions, degree of concentration ofIS resources at the central site, and the levelof DRM-related autonomy granted to localsites. The following sections will provide defini-tions of each factor and discuss the impact oftheir congruence on the success of the DRMfunction.

MIS Quarterly~March 1998 3


Intersite data dependence

The degree to which local sites (divisions ordepartments) of an organization are functional-ly interdependent influences to a significantdegree their need for sharing data. This deter-mines the type of and the extent to which inter-site data/information flow takes place(Goodhue et al. 1992b). The term intersitedata dependence is defined as the need forsharing data among the various business loca-tions or sites.

Intersite data dependence is affected by thedesign of overall organizational structure andthe interdependencies among the product/market domains in which the firm operates.Product/market interdependencies are signifi-cant in the case of larger, divisionally struc-tured firms that have pursued a strategy ofrelated diversification. Such firms typicallyoperate in a large number of product/marketdomains that share market and technologicalsimilarities. For example, consider a firm thatoperates in banking, insurance, and real estatemarkets. In this case, the various product/mar-ket domains share significant similarities interms of their client bases as well as technolo-gies used for delivery of products and ser-vices. This situation leads to a complex web ofinternal and external interdependencies thatmust be managed effectively to realize syner-gistic benefits. Effective management in suchan environment, however, requires a signifi-cant amount of data sharing across sites,resulting in higher levels of intersite data flow.

By contrast, unrelated diversified firms operatein product/market domains that share littlemarket or technological similarities. Consider,for example, a firm that operates in medicalelectronics, jet engines and paper-makingmachinery markets. Given the unrelatednature of the product/market domains, thepotential for realization of market-related ortechnologically related synergistic benefitsmay be very limited. In such an environment,each division may be responsible for its ownproduct development, manufacturing, market-ing, and distribution. The need for sharing datain such an environment is generally limited totransferring financial information between the

business segments/divisions and the corporateoffice for financial planning and control purpos-es. This results in low levels of intersite dataflow.

When the data/information flow is primarilywithin local units, the need for global enforce-ment of standards can be minimal. This resultsin a reduced need for hardware, software, ands;ystems compatibility across sites. On theother hand, when a significant amount of dataflow occurs among sites, compatibility andstandards become important and conflicts mayarise. In this environment, it is important tohave processes that promote cooperation andaid in the resolution of conflicts between orga-nizational units (Goodhue et al. 1992b;Thomas and Burns 1982). In the absence ofcoordination and control enabled by the DRMfunction, important corporate information mayeither get locked in incompatible local systemsor may get held up in turf battles of data own-ership. This could cause delays when man-agers need such information to make key deci-sions affecting organizational performance.

Centralization of IS. decisions

Organizations differ on the extent to which ISdecisions are centralized. Decisions that typi-cally fall within the domain of the IS organiza-tion include technology acquisition, IS person-nel management, selection of systems devel-opment projects,, and day to day IS operatingdecisions. Centralization of IS decisions isdefined as the degree to which the authority tomake IS decisions is concentrated at the apexof the IS organization.

Centralization of IS decisions regarding tech-nology acquisition makes it easier to imple-ment various standards, thereby ensuringhardware and software compatibility acrosssites. Furthermore, centralization of decisionson systems development projects makes iteasier to control data redundancy and toenforce uniform data-naming standards. Suchcentralized IS decision-making structures aredeemed suitable for organizations that operatein a single or a few related product/market

4 MIS Quarterly~March 1998


domains where the level of intersite datadependence is expected to be relatively high.In such an environment, the need for compati-bility of both hardware and software platformsacross sites is greatest and users at local sitesare not likely to have demands for manyunique pieces of data/information.

Decentralization of IS decisions is deemedsuitable for large, unrelated diversified organi-zations because such organizations have alow level of intersite data dependence.Decentralization of IS decisions on hardwareand software acquisition and systems develop-ment projects increases flexibility when select-ing hardware and software and systems devel-opment projects that best meet the needs ofthe local environments. However, such flexibili-ty may lead to incompatibility, making it difficultto implement various standards. It may alsocreate data redundancy. But this may not be aproblem in conditions where intersite datadependency is relatively low. Hence, designsthat either match high centralization of IS deci-sions with conditions of high intersite datadependence, or match low centralization of ISdecisions with conditions of low intersite datadependence, are likely to result in higher DRMsuccess.

Concentration of IS resources

In a distributed environment, IS resourcessuch as hardware, systems and applicationsoftware, data, and IS personnel can be con-centrated to various degrees at the centralsite. For example, companies may have largemainframes located at a central site, whilemid-range and microcomputers may be distrib-uted ’across various local sites. Similarly, theIS development staff may either be located atone central location or distributed to variouslocations. Concentration of IS resources isdefined as the degree to which various ISresources such as hardware, software, data,and development staff are concentrated at acentral site.

Concentration of hardware and softwareresources at the central site makes it easier to

ensure compatibility among hardware and soft-ware platforms. Furthermore, concentration ofother IS resources such as data at the centralsite may facilitate closer control and manage-ment of the data resources in terms of theirintegrity, security, and compatibility. This mayalso enable better enforcement of data stan-dards. Moreover, concentration of ISresources will facilitate development of a holis-tic and comprehensive strategic data plan forthe entire corporation and support queriesrequiring integrated information. Concentrationof IS resources may thus be most appropriatefor organizations with high levels of intersitedata dependence and high degrees of central-ization of IS decisions where the need forcoordination and control of IS resources maybe the greatest. However, concentration ofresources also may lead to inefficiencies anddelays because all requests for data/informa-tion will need to be channeled to the centralsite for processing.

On the other hand, distribution of IS resourcesthroughout the organization will enable storageof all the data relevant to each site at that siteand facilitate rapid responses to local queries.Distribution of IS resources may be mostappropriate for organizations with low levels ofintersite data dependence and low degrees ofcentralization of IS decisions, where needs foradaptability and local responsiveness are thegreatest. Distribution of IS resources, however,may result in longer response time for queriesrequiring integration of information from vari-ous locations. However, this may not pose amajor problem in an environment where inter-site data dependencies are relatively low.

DRM-related autonomy

The term autonomy has been used quiteextensively in administrative sciences to indi-cate how much power an organization has rel-ative to its environment (Price and Mueller1986; Selznick 1953; Van de Ven and Ferry1980). Whereas centralization of IS decisionsrelates to the extent to which the authority tomake IS decisions is located at the apex of theorganization, DRM-related autonomy is


Managing Dist~buted Environments

defined as the extent to which local DRM unitsindependently perform their data resourcemanagement functions.

Motivated by popular themes such as downsiz-ing, empowerment, total quality management,flexibility, and quick response to customers,organizations all over the world increasinglyhave realized that trust in and empowermentof employees hold the key to business suc-cess (Davenport and Short 1990; Hammer1990). The theme underlying the autonomouswork group is that the motivation and satisfac-tion of group members can be increased byproviding them with a significant level of auton-omy in deciding how to perform a given job(Hackman 1976; Nadler et al. 1979).

It is expected that more autonomy andempowerment will motivate the local DRMgroups to work with greater zeal to achieveDRM objectives. Additionally, greater DRM-related autonomy can facilitate a number oftechnically desirable features in a distributedenvironment. For example, each site can oper-ate independently and continuously with itsown data without requiring other sites to agreeon global data structures or definitions (Ceriand Pelagatti 1984). For organizations with lowlevels of intersite data dependence, lowdegrees of centralization of IS decisions, andlow degrees of concentration of IS resources,high levels of DRM-related autonomy may bemost appropriate. In such an environment, lowintersite data dependence and decentralizationof IS decisionsallow for specific hardware andsoftware platforms that are customized to bestmeet local needs. High levels of DRM-relatedautonomy will further add to such flexibility andmay result in greater DRM success.

However, high levels of DRM-related autono-my also can be a serious and contentiousissue in a distributed environment due topotential biases. The danger is in favoring theinterests of local sites above the overall corpo-rate interests. Furthermore, if too much DRM-related autonomy is granted to sites, it canresult in systems with incompatible databases,data structures, and data naming. It also couldlead to important corporate data being lockedin smaller, often incompatible, systems leading

to a chaotic operating environment. However,in an environment with low intersite datadependence, biases favoring local sites andpotential incompatibility of systems may notpose a major problem.

Low levels of DRM-related autonomy may bebeneficial in instances where data resourceshave to be managed to ensure a high degreeof standardization, security, and integrity. Fororganizations with high levels of intersite datadependence, high degrees of centralization ofIS decisions, and a high degree of concentra-tion of IS resources, low levels of DRM-relatedautonomy may be most appropriate. This willensure a high degree of standardization,security, minimum redundancy, and greatercompatibility leading to increased DRM suc-c, ess. However, too little autonomy may bedysfunctional as it is incompatible with thethemes of empowerment, local ownership/control, flexibility, and responsiveness in con-temporary organizations.

¯ Congruent configurations

Based on the above discussion, it is clear thatthe four IS factors (intersite data dependence,c.entralization of IS decisions, concentration ofIS resources, and DRM-related autonomy) areintricately tied to the success of DRM. Thispaper proposes that the success of DRM indistributed environments depends upon theproper alignmentJfit among these four IS fac-t,ors. A model depicting such congruence isshown in Figure 1.

"]’he notion of congruence reflected in the con-ceptual model suggests that an organization’sposition within this four-dimensional spaceinfluences the extent of DRM success. Forinstance, when an organization has a highlevel of intersite data dependence, the needfor coordination and control of DRM activitiesis much higher. In such a context, it appearsprudent to locate most of the IS resources at acentral site. Concentration of IS resources attlhe central site enhances coordination ofresources, enables closer control, ensuresgreater security, and results in greater efficien-

6 MIS Quarter/y/March 1998

Centralizationof IS Decisions

Concentration ofIS Resources

FIT

Intersite DataDependence

DRM-RelatedAutonomy

DRM SUCCESS

Figure 1. A Model of Fit Among IS Factors and DRM Success


cies in resource utilization. Also, centralizationof strategic and management control of ISdecisions will facilitate development of morecomprehensive long-term plans for resourceacquisition and usage in an environmentwhere most of the IS resources are concen-trated at the central site and the need for datacoordination is high. Furthermore, in such anenvironment, it appears logical to permit lowlevels of DRM-related autonomy to local sitesin order to avoid proliferation of incompatibledatabases, minimize .data redundancy, andpermit greater sharing of common data acrossmultiple sites.

On the other hand, if the local units/sites areindependent of each other, as in the case ofan unrelated diversified firm, then intersite datadependence is relatively low. In such an envi-ronment, the need for local responsiveness ofDRM is much higher. Greater distribution ofthe IS resources to local organizational units isin order, as it permits greater flexibility in pro-cessing information. It also allows more timelyresponses to users’ information needs. In suchan environment, decentralization of IS deci-sions will facilitate dm(elopment of hardwareand application architectures that are cus-tomized to the needs of the local organization-al units. Additionally, in these conditions, high-er levels of DRM-related autonomy granted tolocal sites will allow them to respond quicklyand effectively to the unique informationalneeds of the site.

Not all organizations achieve congruenceamong all of the four IS factors. For instance,consider an organization with high centraliza-tion of IS decisions accompanied by high con-centration of IS resources at the central siteand low levels of DRM-related autonomy tolocal sites. If such a firm is a fairly large andunrelated diversified corporation operating inmultiple markets without much market or tech-nological similarities, as already noted, there isbound to be little, if any, intersite data depen-dence. While there is congruency among thefirst three IS factors, this configuration is non-congruent with low intersite data dependence.Although this firm will be able to function, it iseasy to visualize that the non-congruenceamong the IS factors will lead to considerable

delays and inefficiencies in responding to bothinternal and external information needs. In thesame vein, consider an organization with lowcentralization of IS decisions accompanied bylow concentration of IS resources at the cen-tral site and high levels of DRM-related auton-omy to local sites. If this firm is operating inproduct/market domains that have high marketor technological similarities, then there isbound to be a high level of intersite datadependence resulting in a significant amountof data flow between these sites. In such ascenario, the lack of congruence can lead to aduplication of efforts, difficulty in coordinatingand integrating information, inaccurate infor-mation, and wasted resources.

Based on the previous discussions, in well-designed distributed processing environments,the configuration characterized by high inter-site data dependence, high centralization of ISdecisions, high concentration of IS resources,and low DRM-related autonomy is likely toresult in greater DRM success. Likewise, theopposite configuration---characterized by lowintersite data dependence, low centralizationof IS decisions, low concentration of ISresources, and high DRM-related autonomy--is also likely to lead to greater DRM success.Between these two extremes there may beother congruent configurations where these ISfactors are at moderate levels. Various typesand levels of non-congruence may also exist,which can hamper DRM success.

rhus, we posit that only the congruent combi-nations of the four IS factors are expected tobe associated with greater DRM success.Buch a view is in accordance with the notion offit as a gestalt (Venkatraman 1989). The fol-lowing hypothesis is proposed:

Organizations with internally congruent com-binations of the four IS factors--intersite datadependence, centralization of IS decisions,concentration of IS resources, and DRM-related--autonomy will display greater DRMsuccess than those without such congruence.



Research Methodology

Data

Data were collected from firms across themainland of the United States through a sur-vey instrument. An initial version of the surveyinstrument was developed based on the theo-ry-grounded operationalization of the variousconstructs. This version was subsequentlyrefined through extensive pretesting with acad-emics having significant expertise in databas-es, distributed systems, data administration,and management. The instrument was furtherpilot tested with ten data administrators andchief information officers from different firms ina major metropolitan area in the Midwest. Themultiple phases of instrument developmentand testing resulted in a significant degree ofrefinement and restructuring of the surveyinstrument as well as establishing the initialcontent validity (Nunnally 1978).

Since larger organizations are more likely tohave well-established DRM functions and arealso likely to be operating in a distributed sys-tems environment, Business Weeks top 1,000companies in the U.S. (Business Week 1989)served as the target population. G~ven theimportance of this contemporary topic, thequestionnaire was personally addressed to theCIOs of all the 1,000 firms after verifying theirnames with the Directory of Top ComputerExecutives (Applied Computer Research1989). A total of 220 usable responses werereturned, providing a response rate of 22%.Given that the survey was unsolicited and theinstrument quite complex, this response ratecan be considered satisfactory and compara-ble to other studies in IS reseamh (Raho et al.1987).

The responding firms were generally large insize (mean = 18,217; median = 6,000 employ-ees) and represented a wide variety of indus-tries in manufacturing (automotive products,chemicals, pharmaceutical, electronics, etc.)as well as in service (banking, insurance, utili-ties, transportation, food services, retail) sec-tors. There was an even distribution betweenmanufacturing and service sectors (51.4% in

manufacturing and 48.6% in service). Thesefirms operated a fairly large number of facilities(mean = 88) as well as a number of DRM sites(mean = 11), suggesting that a distributedenvironment is indeed a fair representation ofthe organizations at the time of the study.2

Operationalization, vafidity, andrefiability of research constructs

An exploratory factor analysis using principalcomponents factor analysis with varimax rota-tion was performed to examine the unidimen-sionality/convergent validity (Nunnally 1978) each predefined multi-item construct. A jointfactor analysis (using all the indicator items ofall the antecedent variables) employing thesame factor extraction and rotation approachwas employed to determine discriminant validi-ty (Price and Mueller 1986). The four common-ly employed decision rules (Hair et al. 1979)were applied to identify the factors: (1) mini-mum eigen value of 1; (2) minimum factorloading of 0.40 for each indicator item; (3) sim-plicity of factor structure; and (4) exclusion single item factors from the standpoint of parsi-mony. Reliability was evaluated by assessingthe internal consistency of the indicator itemsrepresenting each construct using Cronbach’salpha (Cronbach 1951). Previous researchsuggests a value of 0.60 to 0.70 to be accept-able in exploratory research (Yoon et al. 1995,p. 92). As will be detailed, the results affirmthat all of the scales display satisfactory levelsof reliability with alpha values much higherthan the minimum threshold.

The results of factor analysis relating to unidi-mensionality/convergent validity are shown inAppendix A. Details of the indicator items usedfor operationalization of variables, factor load-

2An analysis of these demographic variables across the fiveclusters of firms derived in this study is discussed later. Inlight of the fact that the largest 1,000 firms in the U.S. weresampled, that there was an even distribution of both manu-facturing and service sectors, and, furthermore, that all ofthe industry segments in both manufacturing and servicesectors were fairly well represented, the results of thisstudy can be generalized to large firms in practically alltypes of industries.



ings, reliability evaluation, and descriptive sta-tistics for intersite data dependence, central-ization of IS decisions, concentration of ISresources, DRM-related autonomy, and thetwo measures of DRM success are also pro-vided in Appendix A. These measures arebriefly described below.

Intersite Data Dependence

Intersite data dependence captures the extentof interdependence among various businesslocations (sites) in terms of the intensity information flows among the sites. Informationflow was classified into four categories: onlinetransactions, batch updates, regular reportgeneration, and ad hoc queries. Respondentswere asked to indicate the proportion of datainvolved in each category as a percentage ofthe total amount of data within the organization.For each data category, the intensity of dataflow between sites was measured using aseven-point scale ranging from (1) very little (7) great extent. The proportion of data flowwithin each category was multiplied by itsintensity. Factor analysis of these four mea-sures resulted in a single factor explaining54.7% of the total variance. The compositemeasure of intersite data dependence is con-structed as the sum of the proportion of dataflow within each category weighted by its inten-sity. The extent of intersite data dependence(mean = 4.304) appears to be moderate.

Centralization of IS Decisions

Centralization of IS decisions is defined as thedegree to which the authority to make IS deci-sions is located at the apex of the IS organiza-tion. The measure of centralization of IS deci-sions uses 24 of 37 decision items originallydeveloped by Pugh et al. (1968). These itemswere selected based on their appropriatenessand were adapted to represent a wide range ofIS decisions. Respondents were asked to indi-cate which level of management had theauthority to make each decision. Managementlevels were represented on four-point scales

ranging from (1)lower management, (2) dle management, (3) chief information officeror CIO, and (4) above the CIO. Respondentswere asked to leave out items that were notrelevant to them.

Factor analysis of the 24 items revealed fourfactors explaining 66.9% of the total variance.As shown in Appendix A, these factors repre-sent centralization of IS decisions pertaining tomanagement of IS personnel, selection ofhardware and software, IS operational control,and IS project planning. Mean values of itemswith highest loading on each factor are usedas’, measures of the degree of IS centralization.Factor loading and reliability coefficients areshown in Appendix A. As can be seen, the reli-ability coefficients are satisfactory. The meanvalues for these four measures of centraliza-tion ranged from 1.77 to 2.48, indicating that,on average, the locus of decision making fallsbetween CIO and middle management.

Concentration of IS Resources

Concentration of IS resources was measuredas a percentage of each of the four categoriesof IS resources: hardware, application soft-ware, data, and IS personnel located at centralvis-a-vis local sites. Factor analysis of the fouritems yielded a single factor explaining 83.4%of the total variance. The factor loading rangedfrom 0.88 to 0.93. The extent of concentrationof IS resources is represented by the meanscore of these four items. Higher values onthis measure denote higher resource concen-tration at the central site. As can be seen fromAppendix A, the reliability of this construct is ata satisfactory level. The sample mean of72..9% indicates that, on average, a very highpercentage of IS resources is concentrated atthe, central site.

DR’M-Related Autonomy

The measure of DRM-related autonomy cap-tures the degree to which local DRM unitsindependently perform the data resource man-

10 MIS Quarter/y/March 1998


agement functions for their sites. Twenty-eightitems were used to represent the DRM activi-ties (Durell 1985; Gillenson 1982; GUIDE1977; Kahn 1983; Leong-Hong and Marron1978; Tillmann 1984). The respondents wereasked to indicate whether each of these activi-ties was predominantly carried out at the locallevel, central/corporate level, or jointly at bothlevels. Each item was scored as (1) if theactivity was entirely performed at the locallevel, (0) if the responsibility for the activityresided at the central/corporate level, and (0.5)if the responsibility was shared by both.Respondents also had an option to choose"not applicable" if a particular DRM activity wasnot performed.

Factor analysis of the 28 DRM-related autono-my items resulted in three factors explaining67.6% of the total variance. These three fac-tors correspond with the paradigm of threemanagement levels (Anthony 1965). As can seen from Appendix A, the three factors repre-sent the levels of autonomy with respect tooperational control DRM functions, manage-ment control DRM functions, and strategicplanning DRM functions. DRM-related autono-my for each of the three levels was evaluatedby the mean score of the individual items withhighest loading on each factor. The reliabilityof these three measures of DRM-relatedautonomy is satisfactory. The autonomyscores range from 0 to 1 with higher scoresindicating greater autonomy. The mean valuesof these three measures range from 0.17 to0.23 indicating that, on average, the local siteshave low levels of DRM-related autonomy.3

DRM Success

While there is a vast body of research on defin-ing and measuring IS success, there is no sin-gle measure that covers all aspects (seeDeLone and McLean (1992) for a comprehen-sive summary). In light of the difficulties in

~The low values appear to be reflective of the nature of dis-tributed database environment prevailing at the time of thestudy. However, in view of the skewed values, caution mustbe exercised while interpreting the results.

establishing a direct link between IS successand organizational performance, user satisfac-tion, or user information satisfaction (Bailey andPearson 1983; Ives et al. 1983) has emergedas a good surrogate measure. Even thoughDRM is a subset of the IS function, a directapplication of IS success measures is likely toresult in inappropriate scope and lack of preci-sion. More specific and detailed measurementis necessary to evaluate DRM success.

In past research, the performance of dataadministration usually has been operational-ized as a self-reported single-item scale mea-suring the degree of success (successful, par-tially successful, or unsuccessful) of dataadministration in an organization (Gillenson1982; Kahn 1983). Problems with relying onlyon such single-item scales are quite wellknown (Galletta and Lederer 1989). In light these inadequacies, DRM success was evalu-ated by two different measures: (1) extent which preset DRM objectives are achievedand (2) the overall success in implementationof the DRM function.

Achievement of DRM Objectives: Twenty-two items were used to represent various DRMobjectives. Respondents were asked to indi-cate on five-point scales ranging from (1) notsuccessful to (5) very successful the extent success in achieving each of the 22 DRMobjectives. Factor analysis of these 22 itemsresulted in five factors explaining 69.8% of thetotal variance (Appendix A). These five factorsrepresent achievement of DRM objectivesrelated to database administration efficiencyand effectiveness, strategic planning and infor-mation requirement analysis, training and con-suiting, standard setting and enforcement, andimproved data sharing and reduced dataredundancy. As can be seen from Appendix A,the reliability of these five measures is at a sat-isfactory level. Mean values of items with high-est loading on each of the five factors are usedas measures of success in achieving DRMobjectives. The mean values of these DRMsuccess measures range from 3.0 to 3.8, sug-gesting that, on average, the respondentsbelieve that they have been able to achieve anabove average degree of DRM success. Acomposite measure of overall achievement of



DRM objectives was also developed by aggre-gating the five individual scales.

Overall Success: This measure of DRM suc-cess is analogous to the measures used byKahn and by Gillenson. Respondents wereasked to indicate on a seven-point scale rang-ing from (1) not successful to (7) very success-ful the overall success in implementation ofDRM functions. The mean value of 4.2 for thisconstruct indicates that, on average, therespondents believe that they have realized abetter than average level of success in imple-mentation of DRM functions.

Discriminant Validity of Antecedent andOutcome Variables

A joint domain factor analysis was performed,including all of the items used to develop thefour original antecedent research constructs.The result provides significant support for fac-torial/discriminant validity of the measurementscales (see Appendix B). Seven factorsemerged in this analysis. Intersite data depen-dence, concentration of IS resources andDRM-related autonomy emerged as single fac-tors. Centralization of IS decisions emerged asfour subfactors. All 28 items of DRMorelatedautonomy loaded on a single factor withoutany cross-loading onto the four subfactors ofcentralization of IS decisions.4 This result reit-erates the conceptual distinction across thesetwo constructs. The factor Ioadings were quitehigh and ranged from 0.43 to 0.89; the sevenfactors explained 70.8% of the total variance.

Unlike in the antecedent variable domain,achievement of DRM objectives was the onlymulti-indicator variable of DRM success, withthe other variable, overall success, being asingle-item scale. Hence, there was no need toperform joint factor analysis to assess discrimi-nant validity in the outcome variable domain.As already noted earlier in unidimensionaiity

’~Note, however, that the three subfactors of autonomy hatemerged during unidimensionality assessment were usedin the subsequent statistical tests so as to not lose the finergranularity of analysis it affords.

checks, factor analysis of 28 DRM objectiveitems resulted in a five-factor solution with sat-isfactory factor loading that ranged from 0.54to 0.85 and explained 69.8% of the total vari-ance. Overall, the various factor analysesresults demonstrate satisfactory support forboth convergent and discriminant validity ofthe scales developed to measure theantecedent variables as well as the scalesdeveloped to measure DRM success.

Zero-order correlation among all the IS vari-ables are shown in Table 1. Descriptive statis-tics as well as interitem reliability coefficients(Cronbach alpha) are also displayed in thistable. The results of zero-order correlationbetween the DRM success measures indicatesignificant correlation (r = 0.37-0.57;p < 0.001) between the five dimensions ofac, hievement of DRM objectives and the mea-sure of overall success. This also provides evi-dence for criterion validity of the measure ofachievement of DRM objectives (Nunnally1978). Of the two measures of DRM success,the measure of objective achievement is morecomprehensive and focuses on specific DRMfunctions within an organization. This measureis therefore examined in detail in subsequentstatistical analyses presented in this paper.The single item measure, overall success, isused as another indicator of DRM success toprovide a more complete picture.

Methods of Analysis

The fit or congruence among the four IS fac-tors is viewed within a multivariate perspectiveas one of gestalt rather than one of a bivariatefit between each pair of factors. As set forth inthe conceptual model, the congruence amongthe four IS factors is expected to be related toenhanced DRM success. As elucidated in priorresearch, cluster analysis is recommended asthe most appropriate statistical technique forexamining such a fit (Venkatraman 1989).Cluster analysis is a technique for groupingindividuals, cases, objects, or entities (firms inthis study) into groups. It differs from othertechniques such as discriminant analysis pri-marily because of the data inductive approach


Table 1. Correlation Matrix for Antecedent Variablesa

No. Variables Mean (S.D.) VAR-1 VAR-2 VAR-3 VAR-4 VAR-5 VAR-6 VAR-7 VAR-8 VAR-9

1. Intersite Data Dependence 4.304 (1.648) (0.724)Centralization of Decisions on:

2. Management of IS Personnel __

3. Hardware and Software4. IS Operational Functions __5. IS Project Planning __

6. Concentration: IS Resources 0.196"** (0.937)

DRM Related Autonomy on:7. Strategic DRM functions - 0.137"8. Mgt, Control DRM Functions - 0.183"* - 0,120"9. Operational DRM Functions

2.484 (0.617) (0.914)b

2.338 (0.515) 0.520 .... (0.880)1.769(0.588) 0.508 .... 0.549 .... (0.851)2.413(0.429) 0.537 .... 0.532 .... 0.647 .... (0.789)

72.900(30.8)% __ 0.144"* 0.186"**

0.184 (0.281) - 0.132" - 0.131" - 0.226*** - 0.719 .... (0.965)0.232 (0.288) - 0.150"* - 0.257 .... - 0.683 .... 0.894 .... (0.942)0.171 (0.285) - 0.142" - 0.587 .... 0.688 .... 0.797 .... (0.802)

Correlation Matrix for DRM Success Variablesa

No. Variables Mean (S.D.) VAR-1 VAR-2 VAR-3 VAR-4 VAR-5 VAR-6

Achievement of DRM Objectives:1. DBA Efficiency/Effeciveness2. Str. Planning/Req. Analysis3. Training/Consulting4. Standard Setting/Enforcement5. Data Sharing/Redundancy6. Overall SUCCESSc

3.787 (0.698) (0.800)3.011 (0.846) 0.378 .... (0.881)3.317 (0.759) 0.528 .... 0.421 ....3.196 (0.966) 0.454 .... 0.430 ....3.214 (0.854) 0.382 .... 0.452 ....4.158 (1.229) 0.571 .... 0.509 ....

(0.751)0.394 .... (0.816)0.367 .... 0.454 .... (0.736)0.497 .... 0.494 .... 0.395 .... (N/A)

Note: Only significant correlation coefficients reported; .... p < 0.001; ***p < 0.01; **p < 0.05; *p < 0.10.aSample sizes vary from 185 to 218 due to missing values.bCronbach Alpha values within parentheses for each factor.cScale range is 1 to 7 for this variable while it is 1 to 5 for all others in the table.


in deriving the number and characteristics ofthe groups/clusters which are not known priorto the analysis (Afifi and Clark 1990). it mustbe noted that this technique is exploratory innature. Primarily, this approach seeks to drawout groups whose internal membership is high-ly coherent in terms of the variousattributes/characteristics of objects/entities thatare of interest while simultaneously distin-guishing each group from other groups. Thisusually is achieved by maximizing theEuchdean distance (of ong=nal variables ortheir transformation to standardized form)and/or Mahalanobis distance (Afifi and Azen1979). This study made use of the K-Meansclustering technique using the Quick Clusterroutine available in SPSS software.

Results

Given the exploratory nature of the researchstudy, we experimented with culling out a dif-ferent.set of clusters consisting of two, three,four, five, and six groups, and used differentoptions (Euclidean and Mahalanobis distance).Regardless of the option, it was imperative tomeet the final goal of deriving a parsimoniousset of clusters that could be clearly distin-guished from one another. To evaluate the dis-tinctiveness of each derived cluster, equality ofvariable means across the clusters was tested,using the F-test. A five-cluster solution waschosen based on meaningfulness of the pat-tern of relationships among the variables(Hambrick 1983). Table 2 shows variablemeans and standard deviations related to eachof the five clusters (columns 1 through 5). values and significance levels associated withthe test of equality of variable means acrossthe five groups are shown in column 6. Thelast column shows pairs of clusters where thevariable means are significantly different.

As can be seen from Table 2, both the F-testsand the results of tests of significance of pair-wise contrasts indicate that the group meansof these five clusters are significantly differentexcept in the case of autonomy on strategicDRM functions. Clusters one and two appearto be internally congruent while the remaining

three clusters are non-congruent. Cluster one,relative to the other clusters, represents agroup of firms with high intersite data depen-dence, high centralization of IS decisions, ahigh concentration of IS resources, and lowlevels of DRM-related autonomy. Just theopposite characteristics are depicted by clustertwo firms. This cluster represents a group offirms With low to moderate intersite datadependence, a low centralization of IS deci-sions, low concentration of IS resources, andhigh levels of DRM-related autonomy. Table 2further shows that clusters three, four, and fivedepict firms with varying levels/types of non-congruence among IS factors. Table 3 pre-sents the same results in a summary form cat-egorizing the group means of each antecedentvariable as "high" (H), "moderate" (M), "low" (L).

As can be seen from Table 3, cluster threefirms display moderate levels of intersite datadependence, a high degree of centralization ofIS decisions, a moderate degree of resourceconcentration, and low levels of DRM-relatedautonomy. Whereas high degree of centraliza-tion of IS decisions and low levels of DRM-related autonomy are congruent, this pattern isnoncongruent with moderate levels of intersitedata dependence and moderate degree of ISresource concentration. Cluster four firmsexhibit a moderate degree of centralization ofIS; decisions, which fits with the moderate lev-els of DRM-related autonomy. However, thesefirms display the lowest levels of intersite datadependence and also low degree of concen-tration of IS resources leading to a fair degreeof non-congruence. Cluster five further illus-trates a condition of non-congruence amongthe IS factors. The firms in this cluster exhibithigh levels of intersite data dependence, a lowdegree of centralization of IS decisions, mod-erate levels of DRM-related autonomy, and amoderate degree of resource concentration.The source of non-congruency for firms in thiscluster is predominantly their high intersitedata dependence and low centralization of ISdecisions.

The central thesis of this study was that organi-zations with internally congruent IS factors willoutperform (in terms of DRM success) those


Table 2. Cluster Analyses: Five Clusters~

Variables

Intersite Data Dependence

Centralization of IS Decisions:Management of IS personnel

Mean (S.D.) of Cluster Groups

Congruent GroupsGroup I I Group 2 ,Group 3(n = 24) (n = (n = 29)

~i:i:::::::::::::::::::::::::::::::::::::::::::::::::::::::~i:i:i:i:i:i:i:i:i:i:i:i:i:i:!:!:!:!:i:i:i:i:i:i:!:!:!:]:i:i:~

~:!~i~!~!~!~i~i~i~i~i~i~!~i~i~i~i~i~i~i~i~i~i:i~i~i~i~i:i:i~i:i:i:i~i~i:i:i:i:i:i~i:i~i~i:i~

i:~i~:~i(0i!~!~)~: ~i~i~i~::ii~i;i~~ 3.007 (0.44)

Non-congruent GroupsGroup 4(n = 22)

1.547 (0.55)

2.573 (0.59)

Group 5(n = 38)

5.787 (0.61

2.168 (0.45)

F (d.fJsig.)b

) 177.1(4,149) .....

26.3 (4, 149) ....

Significant ContrastValues

:::::::::::::::::::::::::::::::::::::::::::::::::::::::::: :~::::::7F:::i:i:i i i~ ~i:i::..:.:.: ;k;k:k~:.:.:.:.’.:.:.: =================================================

Selection of hardware/software

IS operational control functions

IS project planning

2.717 (0.43)

2.179 (0.48)

2.945 (0.38)

2.418 (0.37)

1.918 (0.51)

2.527 (0.52)

!2.084 (0.41)

1.450 (0.33)

2.138 (0.31)

28.1 (4, 149) .... ,::i~t:~2~i::;~::~{ili~i~iiiiiiiiiiiiiiii.

Concentration of IS Resources ::~ii!~iiiii!!::ii~!!iiii!::::~!~!iii!ii!~!!i!i!!::77.8% (30.7)

DRM-Related Autonomy on: ~i::i::~::iii::ii~::~::i::iii::i::iii::iii::i::i:~iii::i::{::i~:i::iii:~i~iii~i:~i:~i::iii~iii~iii::i::i~i:~i:~i::i:~i::i::i::i::i~i::i:~i~i~i::i::iii::i~i~Operational DRM functions ~i~ili(~!~)! i~i~iiii~(~)il 0.135 (0.22)Management control DRM functions ~p~(~)~ ~p~;~(~ 0.150 (0.20)Strategic DRM functions ..~9~(~5)::. ~~ 0.121 (0.22)

65.7% (32.6)

0.219 (0.24)0.217 (0.25)0.163 (0.20)

aCluster groups derived from K-Means orocedure using Quick Cluster analysis of SPSS.~Test of significant differences across Cluster Groups using one-way ANOVA...... p < 0.0001; .... p < 0.001; ***p < 0.01; **p < 0.05; *p < 0.1.

78.6% (28.6)

0.153 (0.25)0.232 (0.30)o. 175 (0.29)

2.41 (4, 149)**2.42 (4, 149)**

Managing Dist~buted Environments

Table 3. Summary of Cluster Analyses Results: Five Clustersa

VariablesIntersite Data DependenceCentralization of IS Decisions:Management of IS personnelSelection of hardware/softwareIS operational control functionsIS project planning

Congruent GroupsGroup I I Group2(n=24)I (n=41)

Group 3(n = 29)

Non-Congruent Groups

I

Group4 I Group5

(n = 22) (n =

MMMMM

H

Operational DRM functions , ..........................~ .............................., .........................iBi ..........................L M MManagementcontrolDRMfunctions I

~ I ~ IL M M

Strategic DRM functions .... M H M

aCluster groups derived from K-Means ~rocedure using Quick Cluster analysis of SPSS. H (High), (Moderate), and L (Low) indicate relative magnitude of the group means on each variable across clusters.

organizations that lack such a congruence.One-way ANOVA was used to test for perfor-mance differences across the five clustersidentified above. As noted earlier, achievementof DRM objectives and overall success areused as indicators of DRM success. Theresults are shown in Table 4. Columns 1through 5 show the mean values and the stan-dard deviations of respective DRM successmeasures for each of the five clusters. Column6 presents the F-values, degrees of freedom,and significance levels. For each DRM successmeasure, clusters where the mean values aresignificantly different from each other are iden-tified and presented in column 7.

As can be seen from Table 4, the resultsappear to partially uphold the central thesis ofthe study. The F-tests indicate that the groupmeans of these five clusters are significantlydifferent on the composite measure ofachievement of DRM objectives and two of itssubfactors: (1) achievement of DRM objectiveswith respect to training and support to IS andend user and (2) enhanced data sharing andreduced data redundancy. The group meansalso are significantly different on the measureof overall success. Cluster one firms, whichdisplayed a high level of congruency among

the four IS factors, appear to have realizedhigh levels of DRM success. These firms dis-play the highest mean ratings on all measuresof achievement of DRM objectives: databaseadministration efficiency and effectiveness;strategic planning and information requirementanalysis; training and support to IS and end-user; standard setting and enforcement; andenhanced data sharing and reduced dataredundancy. The mean ratings on these mea-sures (maximum score of 5.0) represent highlevels of success in achievement of DRMobjectives. Furthermore, cluster one firmsachieved the highest mean rating on the comoposite measure of achievement of DRM objecotires and also on the seven-point scale repre-senting overall DRM success. These meanw]lues are significantly higher than those ofother non-congruent clusters as shown in col-umn 7 of Table 4. Firms represented by clustertwo, another congruent group, also displayhigh levels.of DRM success. This group alsoattained higher mean ratings than the non-con-gruent clusters on almost all measures of DRMsuccess. Firms represented by the remainingthree non-congruent clusters (clusters three,four, and five) appear to generally displaylower levels of DRM success.


Table 4. One-Way ANOVA Across Five Cluster Groups for DRM Success Variablesa

Mean (S.D.) of Cluster Groups

Congruent Groups Non-congruent GroupsDependent Variables Group 1 Group 2 Group 3 I Group 4

(n=24) (n=41) (n=29)I (n=22)

Objectivesb.ii~i~iiit~!i~)ii_ii~iiii~(~i 3.268 (0.68) 3.330 (0.57)

Achievement of DRM Objectives .........................................................., .............................................................

1. Database Administration .........................................................................................................................

Info. Req. Analysis ~(~)~ ;:~~)~ 2.776 (0.86) 3.167 (0.97)3. Training and Suppo~ to ,::::3~5~::(~)~,~5~(~ 3.152 (0.73) 3.258 (0.69)

4. Standard Setting and .......................................................... , ...........................................................Enforcement .~::~~)~.~;~:~; 3.184 (1.05) 3.227 (0.90)

Data Redundancy .~:~(~.~~ 3.190 (0.97) 2.886 (0.94)

Overall SUCCESS= .~:~(~;~)~.~;~(~ 3.714 (1.24) 4.000 (1.15)

Group 5(n = 38)

3.379 (0.55)

3.773 (0.77)

3.141 (0.78)3.243 (0.78)

3.175 (1.09)

3.118 (0.87)4.579 (1.08)

F (d.fJsig.) Significant ContrastValues

2.448 (4, 148)** :::::::*~::::::::::==============================================

aOne-way ANOVA to evaluate significance of differences on the success variables across the dedved Cluster Groups.~Composite of the (five) subfactors that emerged in factor analyses.=Scale range is 1 to 7 for this variable while it is 1 to 5 for all others in the table.***p < 0.01; **p < 0.05; *p < 0.1.


Conventional wisdom would suggest that larg-er firms may devote a greater amount ofresources to the DRM function and pursueDRM activities more systematically, therebyrealizing greater success in their DRM efforts.It is also possible that the nature of industrymight influence the level of importance and theamount of attention devoted to the DRM func-tion. For instance, firms in service industriesthat are usually more information intensive,such as banks and~insurance companies,might be expected to pay greater attention todata management than firms in other indus-tries. It is, therefore, prudent to test for thepotential confounding effects of such factorson the relationship of the fit among the IS fac-tors and DRM success. Hence, the study test-ed for differences among the five clusters interms of industry representation (service vs.manufacturing) and organizational size asreflected by number of DRM sites and numberof facilities operated by the firms. The resultsindicate that there are no major differencesbetween the firms in the congruent and non-congruent clusters on these three demograph-ic characteristics. Thus, the findings are notconfounded by these demographics.

Discussion andImplications

Cluster analysis technique proved to be usefulin deriving distinct and meaningful clustersfrom the data. Clusters one and two wereinternally coherent and meaningful. Clusterone represented firms with high intersite datadependence. The IS organization within thesefirms is characterized by high centralization ofboth operational and strategic IS decisions.These organizations have also managed toretain and concentrate practically all of the ISresources such as hardware, software, data,telecommunications, and personnel at theircentral site. This configuration allows forgreater control over IS resources and ensureshardware and software compatibility acrossthe organization. Furthermore, the low level ofautonomy granted to local sites on DRM func-tions facilitates data standardization, reduces

data redundancy, and helps maintain dataintegrity. This results in a smooth flow of accu-rate and quality data/information between unitsof the organization which is most critical toorganizations with high intersite data depen-dencies. The findings presented here are con-sistent with the arguments on the appropriate-ness of centralization of IS decisions for orga-nizations with high degrees of functional inter-dependencies (Goodhue et al. 1992b).

Organizations represented by cluster two haveIS characteristics opposite those organizationsin cluster one. These firms have low intersitedata dependence and have most of their ISresources distributed to local sites. Theseorganizations have further decentralizedauthority to make operational and strategic ISdecisions to lower levels of management.Moreover, local sites within these organiza-tions also enjoy the highest levels of autonomya~nd control over their DRM functions. Thisconfiguration permits greater adaptability andfaster response to local needs that is essentialfor firms that operate in distinctly differentp, roduct/market domains where intersite dataClependencies are relatively low.

Overall, cluster one and two firms appear to behighly congruent and yet different (statistically)from each other onall of the IS characteristicsexcept on the level of autonomy on strategicF)RM functions, as indicated in Table 2. Thissuggests that, regardless of the degree ofautonomy designed for the local sites, theremay be a tendency to retain the authority tomanage strategic DRM functions at the centralsite. This is because strategic DRM decisionsmay best be handled at the central site inorder to provide unity of strategic direction. Asr~oted earlier, firms represented by clustersthree, four, and five display varying degrees ofnon-congruency among the IS factors. Thecluster five firms have perhaps the worst inter-hal fit.

The results further indicate that organizationswithin cluster one have achieved the highestlevel of DRM success. As noted above, thedesign of the IS function in these organizationsis characterized by high intersite data depen-dence, high centralization of IS decisions, high



concentration of IS resoumes, and low DRM-related autonomy. The results appear to rein-force the concept that the fit or congruenceamong a set of key organizational variablesyields greater effectiveness~ On their own, highdegrees of centralization and low DRM-relatedautonomy may be stifling and tie up the organi-zation in excessive bureaucracy. This couldrender the organization inflexible and non-responsive. However, the internal congruencethat is achieved when these IS factors arematched with conditions of high intersite datadependence and high concentration of ISresources enables achievement of greaterDRM success.

While the results appear to suggest unequivo-cally the relevance and importance of gestaltfit as indicated by the superior performance ofthe internally congruent cluster one, the sameargument cannot be forcefully made for theother equally internally congruent cluster two.At this stage, we can only speculate that tech-nological changes such as powerful/cost-effec-tive workstations and LANs have enticed someorganizations to migrate to distributed environ-ments earlier than they otherwise might havedone. Requisite organizational changes suchas formulating DRM policies, creating datastandards, and educating end-user developerson data hygiene should precede or at leastaccompany such a move. However, consider-able time and organizational learning isrequired for effective implementation of thesechanges. It may also be noted that the studywas performed in 1989, when the DRM con-cept was just being accepted as a reality incompanies. Consequently, it may be too earlyfor the distributed DRM environment to yieldbenefits.

Implications for managementpractice

With the increasing trend toward distributedsystems, it is imperative that IS managersevaluate the importance of the key IS factorsin designing effective IS organizations. Thefindings of this study emphasize the need todesign IS organizations that are internally con-

gruent in order to realize greater effectiveness.Specifically, the focus of this study on DRMunderscores the importance of managing dataresources in a distributed environment. Thecongruent structures identified in this studyprovide IS managers with a benchmarkagainst which they can compare the design oftheir own DRM functions.

Implications for future research

Rapid changes occurring in the informationsystems area often require study of the prob-lems and issues for which a well establishedtheory or conceptual model does not exist. Theencouraging results of this study indicate thatthe concept of fit is useful for studying complexand messy organizational issues such as dataadministration.

Although there was no evidence of differencesacross the cluster of firms that emerged in thisstudy in terms of some of the demographicfactors (e.g., size, industry type), it may necessary to fine-tune the analysis by a moreextensive study of specific industries/environ-ments. This may serve to validate the salienceof the clusters that emerged in this study. Forinstance, it is possible that concerns of datastandardization and sharing may be moreimportant in service industries such as bankingand insurance than for manufacturing compa-nies with multiple plant sites. Replications ofthe study need to be conducted across differ-ent environmental settings to further validatethe existence of such gestalt fit among the ISfactors and the success of the DRM function.Furthermore, this study considered only four ISfactors. There may be a number of other rele-vant factors that may influence the effective-ness of the DRM function in distributed envi-ronments, such as overall organizational struc-ture,- leadership style/culture, role of IS, datasecurity requirements, and constraintsimposed by the migration/evolution path of thedistributed environment. These factors can beconsidered for inclusion in future studies.

Finally, data resource management is only oneof the many responsibilities of the IS function



within an organization. It would be insightful toapply the concept of fit and the method of clus-ter analysis to study other complex IS-relatedissues such as new application developmentand data center operation.

Limitations

Due to limited resources, the study wasrestricted to only four key antecedent variablesand two aspects of DRM success. These arenot necessarily the only variables of impor-tance. Other variables, as noted in the previ-ous section, may be important. In order toattain a high level of responses, only the CIOof each organization was solicited. While hav-ing a high level of confidence in the quality ofinformation gathered--because the respon-dents were very senior executives in IS func-tion-there is still potential for method/report-ing bias since the same executive providedinformation on both the antecedent and depen-dent variables.

Conclusions

Lack of an established theory on design ofdata resource management for distributed pro-cessing environments motivated us to under-take this exploratory study. In light of increas-ing recommendations from recent studies(Niederman et al. 1991) to treat data as a cor-porate resource and the importance of theDRM function in distributed environments, thisstudy is both timely and significant. However,being exploratory in nature, it should be con-sidered as setting the stage for further work inthis domain. While many organizationalresearchers have made assertions on theimportance of fit among key organizationalvariables and have promoted the need to gobeyond viewing fit from bivariate to a multivari-ate perspective (Venkatraman 1989), there a lack of empirical studies. This is one of theearliest attempts to conceptualize fit as gestaltand empirically validate such a view throughcluster analysis. This study further examinesperformance differences among the groups of

firms identified through the cluster analysis.The results of this study provide significantsupport for the concept of gestalt fit among thevariables that were examined.

References

Afifi, A. A., and Azen, S. P. Statistical Analysis:A Computer Oriented Approach, 2nd ed.,Academic Press, New York 1979.

Afifi, A. A., and Clark, V. Computer-AidedMultivariate Analysis, 2nd ed., VanNostrand Reinhold, New York 1990.

Anthony, R. N. Planning and Control Systems:A Framework for Analysis, Division ofResearch, Harvard Business School,Boston, MA, 1965.

Applied Computer Research. Directory of TopComputer Executives, (1/2), AppliedComputer Research, Inc., Phoenix, AZ, Fall1989.

Auerbach. "Data Base Administrator," Part 1,Doc. No. 3-06-04, Auerbach PublishingCompany, Philadelphia, 1975.

Bailey, J. E., and Pearson, S. W."Development of a Tool for Measuring andAnalyzing Computer User Satisfaction,"Management Science, (29:5), May 1983,pp. 530-545.

Bray, O. H. Distributed Database ManagementSystems, Lexington Books, Toronto, 1982.

Business Week. "The Business Week Top1000," Business Week--Special 1989Bonus Issue, New York, 1989.

[]urns, T., and Stalker, G. M. The Managementof Innovation, Tavistock, London, 1961.

Cash, J. I., McFarlan, W. F., McKenney, J. L.,and Applegate, L. M. Corporate InformationSystems Management: Text and Cases(3rd ed.), Irwin, Homewood, IL 1992.

Ceri, S., and Pelagatti, G. DistributedDatabases: Principles and Systems,McGraw-Hill, New York 1984.

Chandler, Jr., A. D. Strategy and Structure,Doubleday, Garden City, N J, 1962.

Charan, R. "How Networks ReshapeOrganizations--For Results," HarvardBusiness Review (69:5), September/October 1991, pp. 104-115.



Cronbach, L. J. "Coefficient Alpha and theInternal Structure of Tests," Psychometrika,(16:3), 1951, pp. 297-334.

Davenport, T. H., and Short, J. Eo "The NewIndustrial Engineering: InformationTechnology and Business ProcessRedesign," Sloan Management Review(31:4), Summer 1990, pp. 11-27.

DeLone, W. R., and McLean, E. R.=Information Systems Success: The Questfor the Dependent Variable," InformationSystems Research (3:1), March 1992,pp. 60-95.

Durell, W. Data Administration, McGraw-Hill,New York 1985.

Galletta, D., and Lederer, A. L. "SomeCautions in the Measurement of UserInformation Satisfaction," DecisionSciences (29:3), 1989, pp. 419-439.

Gillenson, M. L. "The State of Practice of DataAdministration--1981 ," Communications ofthe ACM (25:10), 1982, pp. 699-706.

Goodhue, D. L., Kirsch, L. J., Quillard, J. A.,and Wybo, M. D. "Strategic Data Planning:Lessons From the Field," MIS Quarterly(16:1), 1992a, pp. 11-34.

Goodhue, D. L., Quillard, J. A., and Rockart, J.F. "Managing the Data Resource: AContingency Perspective," MIS Quarterly(12:3), 1988, pp. 373-391.

Goodhue, D. L., Wybo, M. D., and Kirsch, L. J."The Impact of Data Integration on theCosts and Benefits of InformationSystems," MIS Quarterly (16:3), 1992b, pp.293-311.

Grinyer, P. H., Yasai-Ardekani, M., and AI-Bazzaz, S. "Strategy, Structure, TheEnvironment, and Financial Performance in48 U.Ko Companies," Academy ofManagement Journal (23:2), 1980,pp. 193-220.

GUIDE Interantional Corporation. "Establishingthe Data Administration Function," Chicago,June 1977.

Hackman, J. R. "Group Influences onIndividuals in Organizations," in Handbookof Industrial and OrganizationalPsychology, M. D.Dunnette (ed.) RandMcNally, Chicago, 1976.

Hair, J. F., Anderson, R. E., Tatham, R. L., andGrablowsky, B. J. Multivariate DataAnalysis, PPC Books, Tulsa, OK, 1979.

Hambrick, D. C., "High Profit Strategies inMature Capital Goods Industries: AContingency Approach," Academy ofManagement Journal (26:4), 1983,pp. 687-707.

Hambrick, D. C., and Mason, P. A., "UpperEchelons: The Organization as a Reflectionof its Top Managers," Academy ofManagement Review (9:2), 1984, pp. 193-206.

Hammer, M. "Reengineering Work: Don’tAutomate, Obliterate," Harvard BusinessReview (68:4), July-August 1990,pp. 104-112.

Ives, B., Olson, M. H., and Baroudi, J. J., "TheMeasurement of User InformationSatisfaction," Communications of the ACM(26:10), 1983, pp. 785-793.

Jain, H. K., and Ryu, H. "The Issue of SiteAutonomy in Distributed DatabaseAdministration," Proceedings of the NinthInternational Conference on InformationSystems, J. I. DeGross and M. H. Olson(eds.), Minneapolis, MN, 1988, pp. 203-212.

Kahn, B. K. "Some Realities of DataAdministration," Communications of theACM (26:10), October 1983, pp. 794-799.

Laudon, K. C., and Laudon, J. P. ManagementInformation Systems: A ContemporaryPerspective (2nd ed.), Macmillan PublishingCompany, New York 1991.

Leifer, R. "Matching Computer-BasedInformation Systems with OrganizationalStructures," M/S Quarterly (12:1), March1988, pp. 63-73.

Leong-Hong, B., and Marron, B. "DatabaseAdministration: Concepts, Tools,Experiences, and Problems," NationalBureau of Standards Special Publication500-28, U.S. Government Printing Office,Washington, DC, March 1978.

Nadler, D. A., Hackman, J. Ro, and Lawler,E. E. Managing Organizational Behavior,Little, Brown & Company, Boston 1979.

Niederman, F., Brancheu, Jo C., andWetherbe, J. C. "Information SystemsManagement Issues for the 1990s," MISQuarterly (15:4), December 1991,pp. 475-502.

Norburn, D., and Miller, P. "Strategy andExecutive Reward: The Mismatch in theStrategic Process," Journal of GeneralManagement (6:4), 1981, pp. 7-27.

M/S Quarterly/March 1998 21


Nunnally, J. C. Psychometric Theory, McGraw-Hill, New York, 1978.

O’Brien, J. Management Information Systems:An End User Perspective, Irwin,Homewood, IL, 1993.

Price, J. L., and Mueller, C. W. Handbook ofOrganizational Measurement, PitmanPublishing Inc., Marshfield, MA, 1986.

Pugh, D. S., Hickson, D. J., Hinings, C. R., andTurner, C. "Dimer~sions of OrganizationStructure," Administrative Science Quarterly(13:1), 1968, pp. 65-105~

Raho, L. E., Belohav, J. A., and Fiedler, K. D."Assimilating New Technology into theOrganization: An Assessment of McFarlanand McKenney Model," MIS Quarterly(11:1), 1987, pp. 47-57.

Schlack, M. "Client/Server’s Ho{ in NewApplications," Datamation (37:5), May1991, pp. 56-57.

Selznick, P. TVA and the Grass Roots,University of California Press, Berkeley,CA, 1953.

Thomas, R. C., and Burns, A. "The Case forDistributed Decision Making System,"Computer Journal (25:1), February 1982,pp. 147-152.

Tilmann, G. D. "Data Administration vs. DataBase Administration: There is a Difference"Infosystems (31:2), February 1984, pp. 98-102.

Van de Ven, A. H., and Ferry, D. L. Measuringand Assessing Organizations, Wiley-Interscience, New York 1980.

Venkatraman, N. "The Concept of Fit inStrategy Research: Toward Verbal andStatistical Correspondence," Academy ofManagement Review (14:3), 1989,pp. 423-444.

Venkatraman, N., and Camilus, J. C."Exploring the Concept of ’Fit’ in StrategicManagement," Academy of ManagementReview(9:3), 1984, pp. 513-525.

Weldon, J. L. Data Base Administration,Plenum Press, New York, 1981.

Yoon, Y., Guimaraes, T. and O’Neal, Q."Exploring the Factors Associated withExpert Systems Success," MIS Quarterly(19:1), March 1995, pp. 83-106.

About the Authors

Hemant Jain is Lawrence G. RegnerResearch Professor of ManagementInformation Systems and the director of UWMMIS Consortium in the School of BusinessAdministration at the University of Wisconsin-Milwaukee. Jain received his Ph.D. in industri-al engineering and information systems fromLehigh University in 1981. His current reseamhinterests are in the area of distributed andcooperative computing systems, data model-ing and database design, object orientedanalysis and design, decision support sys-tems, model management systems, groupdecision support systems, and information sys-tems design. He has published a large numberof articles in prestigious journals such as IEEETransactions on Software Engineering,Information Systems Research, DecisionSupport Systems, Decision Sciences, andh~formation & Management. professor Jain ison the editorial board of the Journal ofDatabase Management as well as a consultantto a number of large corporations. He is cur-rently developing (along with three of his Ph.D.students) an integrated knowledge-basedapproach for the design of co-operative com-puting environments (hardware, software, andnetwork architectures), strategies for objectpartitioning and distribution in co-operativecomputing environments, and approaches foroperation and control of co-operative comput-ing environments.

K. Ramamurthy is an associate professor ofMIS at the University of Wisconsin-Milwaukee.He received his bachelor’s degree in mechani-cal engineering and a graduate diploma in sta-tistical quality control and operations reseamhin India, an M.B.A. from Canada, and a Ph.D.degree in management information systemsfrom the University of Pittsburgh. He has over19 years of industry experience and has heldseveral senior technical and executive posi-tions. His current research interests includeadoption, implementation, and diffusion ofmodern information technologies; strategic ISplanning; electronic commerce includinginterorganizational systems/EDI and theInternet; data resoume management in distrib-



uted environments; decision systems for indi-vidual and group support; self-directed teams;business process reengineering; TQM; andmanagement of computer integrated manufac-turing technologies. He has published quiteextensively in major journals including M/SQuarterly, Journal of Management InformationSystems, Decision Sciences, InternationalJournal of Man-Machine Studies, IEEETransactions on Engineering Management,International Journal of Production Research,OMEGA, Transportation Journal, and INFOR,and in a number of conference proceedings.He is a member of INFORMS, AiS, and theBeta Gamma and Sigma Honor Society.

Hwa-Suk Ryu received his Ph.D. in manage-ment science/MIS from the University ofWisconsin-Milwaukee in 1990. He has under-graduate and master’s degrees in businessadministration. He is currently the generalmanager of Samsung Data Systems in Se0ul,Korea. His reseamh interests are in the areasof data administration and database manage-ment systems. A 1988 paper appeared in theProceedings of the /nternationa/ Conferenceon/nformation Systems. He has been a mem-ber in TIMS, Decision Sciences Institute, ACM,and SIGBDP.

Masoud Yasai-Ardekani is an associate pro-fessor of strategic management at the Schoolof Business Administration, University ofWisconsin-Milwaukee. He holds a bachelor’sdegree in electrical engineering from theImperial College of Science and Technology,University of London, and a Ph.D. in manage-ment studies from the Graduate BusinessCenter, the City University, London. He con-ducts research in the area of strategic man-agement. His research interests includedesigns for strategic planning processes, andcompetitive intelligence systems; industry envi-ronment-strategy and firm performance; orga-nizational decline, strategic change and turn-around. He is also engaged in multi-discipli-nary research on: expert systems and cogni-tive mapping for strategic decisions, designs ofdecision processes and structures for effectivemanagement of information systems; quantifi-cation of uncertainty in organizations; andstrategic management of public sector organi-zations. He has published several articles inleading management journals such as TheAcademy of Management Journal, Academy ofManagement Review, Strategic ManagementJournal, Management Science OrganizationStudies, Journal of Management Studies, and/EEE Transactions on EngineeringManagement.

MIS Quarterly/March 1998 23


Appendi:x A

Individual Factor Analysis for Undimensionality/ConvergentValidity Check

The following indicator variables were suitably reworded, randomly assigned and cast as questionnaireitems. Operationalization of each construct, the results o1~ factor analyses of the study’s variables, fac-tor Ioadings, interitem reliability (Cronbach’s alpha), and descriptive statistics of each of the factors aredisplayed.

1. Intersite Data Dependence (DEPEND): Responses were sought on the proportion of each of thefollowing four classes of data w~ (out of the total data within the responding firm) and the extent data flow across sites x; on a 1 to 7 scale (1 = very little and 7 = significant extent). A weightedaggregate would represent the extent of the intersite data dependence. A single factor emerged infactor analysis.

1. Online Transactions.2. Batch Updates.3. Regular Report Generation.4. Ad hoc/Unscheduled Queries.

Intersite Data Dependence = Z w~ x~ for i = 1 to 4

FactorLoadings

0.6240.6940.8530.766

where w~ = proportion of i th category of data out of total amount of data processedx; = extent of data flow of i th category across the sites (on 1 to 7 scale)

Cronbach’s ~ = 0.7237; Interitem Correlation = 0.396; Mean (S.D.) = 4.304 (1.648)

2. Centralization of IS Decisions: Responses were sought as to the level of management to whomthe decision making on each of the indicated 24 information systems (IS) activities is delegated evenif others have to confirm the decision. Value of 1 assigned if locus of decision making is lower ISmanagement, 2 for middle level IS management, 3 for CIO, and 4 for above the CIO level. Four (outof 24) indicator variables were dropped out after multiple runs of factor analyses (with varimax rota-tion) as they failed to meet preset decision criteria. Finally, four factors of IS decision-making struc-ture emerged, which were labeled as:

A. Centralization of decisions related to management of IS personnel (CENTR-I: Five indicatorvariables; Cronbach’s (x = 0.9140; Interitem Correlation = 0.680; Mean (S.D.) = 2,484 (0.617))

1. Determining the number of IS supervisors required.2. Appointment of IS supervisory staff.3. Promotion of IS supervisory staff.4. Salary increase of IS supervisory staff.5. Dismissal of IS supervisor.

FactorLoadings

0.7020.8520.8510.7840.819



B. Centralization of decisions related to selection of hardware and software (CENTR-2: Fiveindicator variables; Cronbach’s ~ = 0.8797; Interitem Correlation = 0.594; Mean (S.D.) = 2.338(0.515))

1. Selection of type er brand for new main-frame computer.2. Selection of type or brand for new PCs/VVorkstations.3. Selection of type or brand for DBMS software.4. Selection of type or brand for data dictionary.5. Selection of type for data communication service.

FactorLoadings

0.6670.6730.8190.8000.786

C. Centralization of decisions related to IS operational control (CENTR-3: Five indicator vari-ables; Cronbach’s ~ = 0.8511; Interitem Correlation = 0.588; Mean (S.D.) = 1.769 (0.588))

FactorLoadings

1. Scheduling development tasks. 0.5602. Setting standards for data definition. 0.7953. End-user training methods to be used. 0,8164. To determine data access control method. 0.7485. To (re)schedule responsibilities/areas of work of senior analyst/programmers. 0.546

D. Centralization of decisions related to IS project planning (CENTR-4: Five indicator variables;Cronbach’s ~ = 0.7894; Interitem Correlation = 0.429; Mean (S.D.) = 2.413 (0.429))

1. To determine a new systems development project.2. Setting end-user requirements priority.3. "Buy or development" decision for application software.4. Creation of a new information center.5. Creation of a new project team.

FactorLoadings

0.6890,6320.6820.6360.585

3. Concentration of IS Resources at Central vis-~,-vis Local Sites (RESOURCE): Responses weresought on the proportion (as a percentage) of the total of each of the following IS resources thatwere retained at central site vis-a-vis distribution to local sites. An aggregate would represent ISresource availability at central site. A single factor emerged in factor analysis.

1. Mainframe/Mini Computers.2. Data Sources.3. IS Personnel.4. Application Software,

Cronbach’s c~ = 0.9369; Interitem Correlation = 0,788; Mean (S.D.) = 72.9% (30.8%)

FactorLoadings

0.9280.9260,8750.922



4. DRM-Related Autonomy (AUTONOMY): Responses were sought as to whether each of the indi-cated 28 DRM activities were carried out predominantly at local/divisional, central/corporate, or bothlocal and central levels. A value of 0 assigned to central, 0.5 to both, and 1 to local level responses.These 28 indicator variables split into three factors in factor analysis (with varimax rotation) mappingAnthony’s (1965) paradigm.

A. Autonomy on operational control DRM functions (12 indicator variables; Cronbach’s o~ 0.9654; Interitem Correlation = 0.699; Mean (S.D.) = 0.184 (0.281))

FactorLoadings

1. Evaluate and select hardware and software including DBMS, data dictionary, etc. 0.6522. Logical and physical database (re)design. 0.6503. Develop and monitor database performance measure. 0.7364. Develop end-user support policies. 0.7845. Monitor and maintain data integrity and security. 0.7296. Resolve database operational problems. 0.7487. Maintain DBMS and related software. 0.8428. Maintain data dictionary/directory. 0.6729. Support application systems designers that use DBMS. 0.681

10. Enforce database operation standards. 0.78111. Database performance tuning. 0.79912. Data communication administration. 0.702

B. Autonomy on management control DRM functions (13 indicator variables; Cronbach’s ~ =0.9415; Interitem Correlation = 0,553; Mean (S.D.) = 0.232 (0,288))

FactorLoadings

1. Assess impact of changes in technology and information requirements on data. 0.5422. Develop global and local data dictionary. 0.5773. Analyze data requirements, both global user view and local user view. 0.7514. Data distribution (fragmentation and allocation of data to local sites). 0.5835. Set standards for data definition/naming convention. 0.5586. Set data security standard. 0.6067. Set policies and procedures for database backup and recovery. 0.6958. Enforce standards for data definition. 0.6199. Training and educating end users. 0.573

10. Maintain consistency and compatibility between local sites. 0.61711. Liaison with systems and application analyst. 0.66712. Liaison with programmers. 0.67413. Liaison with end users. 0.739

C. Autonomy on strategic planning DRM functions (three indicator variables; Cronbach’s e< 0.8022; Interitem Correlation = 0.575; Mean (S.D.) = 0.171 (0.285))

FactorLoadings

0.7320.7170.786

1. Development of strategic plans for data as corporate resource.2. Define database goals in support of DP and organizational goals.3. Development of long-range plans for DB to achieve database goals.



5. DRM Success: The degree to which DRM was successful was measured by multiple methods asfollows:

Extent to which DRM Objectives are Achieved (DRM-OBJECT): Responses were sought on theextent of success in achieving each of the indicated 22 DRM objectives on a 1 to 5 scale (1 = notsuccessful and 5 = very successful). Three (out of 22) indicator variables were dropped out aftermultiple runs of factor analyses (with varimax rotation) as they failed to meet standard decision crite-ria used in research. Finally, five factors of the extent of success in achieving DRM objectivesemerged. These were labeled as:

A. Database administration efficiency and effectiveness objectives (six indicator variables;Cronbach’s ~ = 0.8001; Interitem Correlation = 0.400; Mean (S.D.) = 3.787 (0.698))

1. Maintaining data integrity.2. Maintaining data accuracy.3. Maintaining data security.4. Providing timely data.5. Maintaining data availability.6. Efficient data distribution.

FactorLoadings

0.8310.7970.6240.7120.7590.542

B. Strategic planning and information requirement analysis objectives (four indicator variables;Cronbach’s ~ = 0.8806; Interitem Correlation = 0.648; Mean (S.D.) = 3.011 (0.846))

FactorLoadings

1. Developing strategic plans for data as corporate resource.2. Developing long-range plans for database.3. Identifying your company’s information needs.4. Establishing priorities of information needs.

0.8190.7130.8460.789

C. Training of/support to both IS personnel and end users’ objectives (four indicator variables;Cronbach’s (~ = 0.7508; Interitem Correlation = 0.430; Mean (S.D.) = 3.317 (0.759))

FactorLoadings

0.7530.7440.6050.602

1. Training end users.2. Providing consulting to end users.3. Training data processing personnel.4. Providing consulting to data processing personnel.

D. Standard setting and enforcement objectives (three indicator variables; Cronbach’s ~ =0.8161; Interitem Correlation = 0,597; Mean (S.D.) = 3.196 (0.966))

1. Maintaining data dictionary.2. Setting standards for data definition.3. Enforcing data standards.

FactorLoadings

0.7770.7630.697



E. Enhanced data sharing and reduced data redundancy objectives (two indicator variables;Cronbach’s ~ = 0.7363; Interitem Correlation = 0.583; Mean (S.D.) = 3.214 (0.854))

FactorLoadings

1. Reducing data redundancy. 0.8072. Promoting data sharing. 0.815

Overall Success (SUCCESS): Response was solicited on the extent to which DRM functions havebeen successfully implemented in the firm on a 1 to 7 scale (1 = not successful and 7 = very suc-cessful).

Cronbach’s e and Interitem Correlation N/A (being single-item scale); Mean (S.D.) = 4.158 (1.229).



Appendix BJoint Factor Analysis/Discriminant Validity Check for

Antecedent VariablesaIndicator

Items AUTONOMY CENTR-1 CENTR-2 CENTR-3 RESOURCE CENTR-4 DEPEND

AUT1AUT2AUT3AUT4AUT5AUT6AUT7AUT8AUT9AUT10AUT11AUT12AUT13AUT14AUT15AUT16AUT17AUT18AUT19AUT20AUT21AUT22AUT23AUT24AUT25AUT26AUT27AUT28CEN1CEN2CEN3CEN4CEN5CENTCEN8CEN9CEN10CEN11CEN16CEN19CEN20CEN21CEN22CEN12CEN15CEN17CEN18CEN23RESO1RESO2RESO3RESO4DEPND1DEPND2DEPND3DEPND4

.5991

.7050

.6837

.8039

.7328

.8848

.7959,7543.7605.8112.8544,7483.8136.8440.7380.8012.8517,8613.8868,8879,7296,6693,8731.8483,8485,7338,8682,7773

.7175

.8587.8564.7909.8140

.6819

.6571

.8007

.7902

.7731.5810.7907.7940.7481.5558

.4624

.7958

.8087

.6150

.7097

.4431

.6266

.5556

.6533

.4478

.5726

.5647.7280,8489.7566

Eigen Value 20.38 8.165 2.556 2.022 1.993 1.747 1.535Cum Variance 36.4% 51.0% 55.6% 59.2% 62.8% 65.9% 68.6%

aOnly factor Ioadings greater than 0.40 shown; Final factor analysis (CEN6, CEN13, CEN14, and CEN24 dropped out).


Success of Data Resource Management in Distributed Environments: An Empirical Investigation

Documents

Transcript of Success of Data Resource Management in Distributed Environments: An Empirical Investigation