the association between activity based costing and improvement in financial performance 2002...

Available online at http://www.idealibrary.com ondoi: 10.1006/mare.2001.0175Management Accounting Research, 2002, 13, 139

The association between activity-based costingand improvement in financial performance

Douglass Cagwin* and Marinus J. Bouwman

This study investigates the improvement in financial performance that is associatedwith the use of activity-based costing (ABC), and the conditions under which suchimprovement is achieved. Internal auditors furnish information regarding companyfinancial performance, extent of ABC usage, and enabling conditions that have beenidentified in the literature as affecting ABC efficacy. Confirmatory factor analysis andstructural equation modelling are used to investigate the relationship between ABC andfinancial performance.Results show that there indeed is a positive association betweenABC and improvement

in ROIwhenABC is used concurrentlywith other strategic initiatives, when implementedin complex and diverse firms, when used in environments where costs are relativelyimportant, and when there are limited numbers of intra-company transactions. Inaddition, measures of success of ABC used in prior research appear to be predictors ofimprovement in financial performance.

c 2002 Elsevier Science Ltd. All rights reserved.

Key words: activity-based costing; new business initiatives; ABC success; structural equationmodels.

1. Introduction

Activity-based costing1 (ABC) has been promoted and adopted as a basis formaking strategic decisions and for improving profit performance for over a decade

*Assistant Professor of Accounting, School of Business, University of Texas at Brownsville, Brownsville,TX 78520, U.S.A. Tel: (956) 983-7300. E-mail: [email protected] L. McQueen Associate Professor of Accounting, SamM.Walton College of Business, University ofArkansas, Fayetteville, AR 72701, U.S.A. Tel: (501) 575-6117. E-mail: [email protected]

Received 28 December 1999; accepted 4 October 2001.1The terms activity-based costing (ABC) and activity-based management (ABM) are sometimes usedinterchangeably. Strictly speaking, ABC refers only to determining the costs of activities and the outputsthat those activities produce. Some researchers and practitioners prefer to use the term activity-basedmanagement (ABM) when describing how the activity information is used to support operating decisions.

10445005/02/$ - see front matter c 2002 Elsevier Science Ltd. All rights reserved.

2 D. Cagwin and M. J. Bouwman

(Bjornenak andMitchell, 1999). In addition, ABC information is now alsowidely usedto assess continuous improvement and to monitor process performance. AlthoughABC has found rapid and wide acceptance, there is significant diversity of opinionsregarding the efficacy of ABC (McGowan and Klammer, 1997). Several reservationshave been expressed concerning its usefulness, relevance, and practicality (Inneset al., 2000). Despite managers insistence that management accounting systems passthe costbenefit test (Foster and Young, 1997), there still is no significant body ofempirical evidence to validate the alleged benefits of ABC (Shim and Stagliano,1997; McGowan and Klammer, 1997). Empirical research is needed to documentthe (financial) consequences of ABC implementation (Kennedy and Bull, 2000;McGowan, 1998).Furthermore, previous research has suggested that the benefits of ABC are

more readily realized under enabling conditions such as sophisticated informationtechnology, highly competitive environment, complex firm processes, relativelyhigh importance of costs, and relatively low unused capacity and intra-companytransactions. Variables representing these conditions are appropriately incorporatedinto a model testing the efficacy of ABC.The purpose of this study is to measure the improvement in financial performance

that is associated with ABC use and the conditions under which such improvementis achieved. The research instrument is a cross-sectional mail survey of 1058 internalauditors, claimed to be knowledgeable and unbiased in the assessment of costsystems (Tanju and Helmi, 1991; Ray and Gupta, 1992). Confirmatory factor analysisand structural equation modelling using LISREL8 (Joreskog and Sorbom, 1993) areused to test a model hypothesizing the conditions under which there is a positiveassociation between time-impacted use of ABC and change in financial performance.Control is provided for the moderating effects of concurrent use of other strategicmanagement initiatives (e.g. TQM, JIT) and enabling conditions identified by priorresearch. In addition, the study tests the association between improvement infinancial performance and previously used measures of ABC efficacy, as suggestedby Foster and Swenson (1997).This study enhances previous research on ABC in three ways. One, it employs

internal auditors who constitute an objective and knowledgeable source of ABCperformance. Prior research has used respondents with a personal stake in ABC,such as controllers or ABC project managers (e.g. Shields, 1995; Swenson, 1995;Krumwiede, 1996, 1998). Second, this study tests a specific measure of improvementin financial performance, as opposed to unobservable general constructs such asperceptions of success, satisfaction, or financial benefit. Finally, it tests a structuralmodel synthesized from prior theoretical research, describing the conditions underwhich ABC should be successful.Results show that positive synergies are obtained from concurrent use of ABCwith

other initiatives. In addition, positive associations between ABC and improvementin ROI are reported when ABC is implemented in complex and diverse firms, inenvironments where costs are relatively important, and when there are limitednumbers of intra-company transactions to constrain benefits. There is also someindication that other enabling conditions (information technology sophistication,

As in Innes et al. (2000), this study defines ABC broadly to include both activity-based costing and activity-based management.

ABC and Improvement in Financial Performance 3

absence of excess capacity, and a competitive environment) affect the efficacy ofABC as expected, and that manufacturing firms may obtain greater benefits thannon-manufacturing firms. Finally, there is some evidence that previously usedmeasures of ABC success (Shields, 1995; Swenson, 1995; Krumwiede, 1996, 1998) arepredictors of improvement in financial performance.The remainder of the paper is organized as follows. The next section relates

background regarding ABC and situates this study in the context of prior research.Subsequent sections develop hypotheses, describe research methodology includingmodel specification, discuss data collection procedures, and present the results.

2. Prior research on ABC and financial performance

Activity based costingABC has received a great deal of attention as a cost management innovation.Numerous proponents of ABC argue that its methods are necessary to trace overheadcosts to cost objects, and thus to properly account for batch and product-levelcosts (Cooper, 1990). Researchers also argue that ABC is more effective in specificenvironmental conditions (enabling conditions) such as manufacturing complexity(Jones, 1991), environments with specialty product costs (Srinidhi, 1992) and diverse(multiple different) business environments (Cooper and Kaplan, 1988). Many alsorecommend using ABC to support process improvement (Turney, 1991) and todevelop cost-effective product designs (Cooper and Turney, 1989). Although ABCsystems are most often associated with manufacturing companies, they can beapplied in all types of organizations (Rotch, 1990; Tanju and Helmi, 1991).The theories of diffusion of innovations (Kwon and Zmud, 1987), transaction cost

economics (Roberts and Silvester, 1996), and information technology (Dixon, 1996)suggest that organizations adopt an innovation such as ABC to obtain benefits thatdirectly or indirectly impact financial performance measures. Evidence regarding thebenefits of ABC, however, is largely restricted to theoretical models and anecdotalinformation obtained from case studies and often related by practitioners (e.g.Barnes, 1991; Brimson, 1991; Bruns and Kaplan, 1987; Harris, 1990). Empiricalresearch on the efficacy of ABC has generally concentrated on identifying successfulABC systems and modelling the factors that lead to this success. Success has beendefined as use for decision-making (e.g. Innes andMitchell, 1995; Krumwiede, 1996,1998; Anderson and Young, 1999), satisfaction with the costing system (Shields,1995; Swenson, 1995; McGowan and Klammer, 1997), perceived financial benefita dichotomous measure with no reference to the criteria of benefit (Shields, 1995;Krumwiede, 1996, 1998), or other non-financial benefits (McGowan, 1998). Therealso has been no empirical evidence that demonstrates that ABC improves financialperformance nor has the link between successful ABC systems and improvement inbottom-line financial performance been tested.To the contrary, in an experimental setting, Drake et al. (1999) found that innovative

activity can produce a higher or lower level of firm profit when workers have ABCinformation. Workers can use the better information for their own gain, i.e. noinformation is better than good information used contrary to company goals.


3. Development of hypotheses

Direct association of ABC with change in financial performanceThe arguments in support of ABC are generally based on the comparative advantagethat firms can obtain from the superior information generated through ABC.Although ABC has strong theoretical underpinnings, Kaplan (1993) and otherresearchers caution practitioners that not every ABC system will produce thesebenefits. In addition, many researchers (e.g. Bjornenak, 1997; Gosselin, 1997b; Malmi,1997, 1999; Bromwich and Hong, 1999) and practitioners (e.g. Smith, 2000) questionthe inherent value of an ABC system. The issue of whether increasing use of ABCis directly associated with improvement in financial performance, without regardto firm- and industry-specific environmental conditions, has not been empiricallytested. This leads to the following hypothesis (in alternative form).

H1: There is a positive association between the extent of use of ABC and relativeimprovement in financial performance (compared with other firms in theindustry).

Financial performance is measured relative to other firms in the industry whilelevels of the variables of interest and other independent variables are tested. Theevaluation of this hypothesis provides a baseline for this research. Confirmationwould be expected only if ABC provided a comparative advantage, on average,for every firm, regardless of its circumstances. On the other hand, if, as expected,realization of the benefits of ABC requires the presence of specific firm conditions,then this hypothesis will not be confirmed and the focus will shift to hypothesis two,which identifies specific enabling conditions.

Enabling conditionsIn an event study, Gordon and Silvester (1999) could not find a significant stockmarket reaction to the installation of an ABC system. They note that it is possible thatABC may have differential impacts across firms depending on certain firm-specificfactors. They conclude that there is strong reason to believe that the benefits of ABCare contingent upon various behavioural and organizational factors.Previous research (e.g. Pattison and Arendt, 1994; Estrin et al., 1994; Cooper and

Kaplan, 1991) has identified specific environmental conditions (e.g. complexity andcompetition) that affect the potential benefits from the use of ABC. The theorysupports the proposition that under appropriate enabling conditions the improvedcosting information provided by ABC leads to improved decision-making, andtherefore should be associated with improved performance. This leads to thefollowing hypothesis (in alternative form).

H2: The association between the extent of use of ABC and relative improvement infinancial performance is impacted by specific enabling factors.

The specific enabling factors identified in this study (and the predicteddirection of impact) are as follows.

Importance of costs (positive).


Information technology sophistication (positive). Business unit complexity (positive). Level of intra-company transactions (negative). Unused capacity (negative). Competition (positive).

Measures of ABC success and change in financial performancePrevious researchers have developed and tested theories regarding the determinantsof ABC success. Success has been operationalized by survey items asking whetherrespondents believed that the system had been successful (Shields, 1995), whetherthey were satisfied with their cost systems (Swenson, 1995), whether ABC had beenworth implementing (Krumwiede, 1996, 1998), or by asking respondents to identifytheir level of satisfaction with the implementation of ABC (McGowan and Klammer,1997). Researchers have implicitly assumed that successful ABC systems lead to im-proved financial performance. However, this relationship between perceived successand specific measures of financial performance has not been tested. Moreover, whenexamining alternative measures of ABC success, Foster and Swenson (1997) foundthat pairwise correlations between alternative success measures were sizably lessthan 1.00 (0.45 to 0.75). Consequently, the current study tests the relationship be-tween several reported measures of success with improvement in financial perfor-mance through the following hypothesis (in alternative form).

H3: A firms relative improvement in financial performance (compared with otherfirms in the industry) is positively associatedwith the level of success of ABC.

The specific measures of success examined in this study include the following.

Perceived success of the ABC implementation. Satisfaction with the cost system. Expressed belief that ABC has been worth implementing.

Rejection of the null hypotheses would support the appropriateness of successconstructs used in prior ABC studies, and would enhance the credibility of both thisstudy and previous research by providing a tie between ABC success and financialperformance.

4. Model developmentresearch design

The impact of ABC on financial performance is examined using the following model:

1ROI = f (ABC, enabling variables, control variables)where 1ROI is the change in return on investment measured over the prior threeor five years. The relationships between the variables are presented graphicallyin Figure 1. The figure shows that ABC (the extent of ABC use) is a constructthat consists of four components: the organizational functions using ABC, theapplications for which ABC is used, the level of integration of ABC into firm strategic


Survey Questions Dependent Survey Questions(Observable X) Independent Constructs Constructs Observable Y

ns

--

?

7 F16a-g

3 F3,F6,F7

9 F17a-l

8 I12a-j

6 D1-D6

7 E1-E7

5 A1-A5

2 I10a, b

ABCUse (ABC)

Intra-Company

Transactions

OtherInitiatives

(INIT)

Complexity-Diversity

(COMPLEX)

Importance ofCosts

(IMPORT)

Enabling Conditions

(ENABLERS)

Unused Capacity (CAPAC) I11

Size(SIZE) 18Competition

(Comp) A6

InformationTechnology

(INFO)

--

ROI 3YrC6

ROI 5YrC7

?

?

Key

Construct confirmed with factor analysis

Additive index

Observed variable

2 F3, F4 Number of items and survey questionnumbers used to measure variables--Some questions have multiple parts(F16, F17, I10, I12)

Direct Effect

Moderating Effect (interaction)+, --, ? Expected sign of coefficient

(ROI)

Change inFinancial

Performance

?

Type(TYPE) 16Time

(TIME) H1PerformanceEvaluation

(EVAL)

Applications(APPLIC)

FunctionsUsing ABC (FUNCTION)

Figure 1. H1: Association of ABC with improved financial performance.

and performance evaluation systems, and length of time since implementation.The figure also identifies six specific enabling conditions: information technology,complexity/diversity, importance of costs, intra-company transactions, unusedcapacity, and competition. Control variables include firm size and type. In addition,themodel includes the use of other strategic initiatives, such as JIT and TQM, in orderto capture possible synergies between ABC and those initiatives.The remainder of this section describes the variables contained in this model.

Variable names are capitalized (see Figure 1). Figure 1 also identifies the specificquestionnaire items measuring each construct. The complete questionnaire isincluded in the appendix.Most constructs are latent constructs composed of two or more manifest variables.

Composite scores of multiple variables have the advantage of capturing more ofa constructs multi-dimensionality than individual questions (Foster and Swenson,1997). Use of multi-item measures also reduces the effect of random and measure-ment errors, and structural coefficients obtained are less biased than those obtainedusing manifest variables alone (Libby and Tan, 1994).

Dependent variableChange in return on investment (1ROI). The most common investment centreperformance measure is return on investment (Hilton, 1994). Furthermore, previousresearch shows a high correlation between ROI and other profitability measures(Prescott et al., 1986) and suggests that ROI is more readily available in business unitsthan other measures (Jacobson, 1987).


Using the change in ROI (1ROI) allows control for the level of ROI prior to thetest period.1ROI is also industry median adjusted to control for differences betweenindustries. Table 1 summarizes the definitions of this and subsequent variables.

Self-reported versus archival measures of performance. This study, like most researchattempting to link strategic initiatives to financial performance, relies on a self-reported measure of performance. Govindarajan (1988), Govindarajan and Fisher(1990), and Chenhall and Langfield-Smith (1998) all asked respondents to assess theirbusinesss performance relative to competitors over the last three years. However, asnoted by Young (1996), self-reported performance is not necessarily a valid proxy foractual performance.2 Management accounting research has been silent on this issue(Young, 1996).In order to evaluate the accuracy of self-reported measures, a subset of the

responses in this study was compared with actual financial statement informationretrieved from Compustat. Fifty-four respondents reported company-wide informa-tion for firms that were included in the Compustat database. For those companieswith complete information (ranging from 47 to 52 for an individual test), actualchange in ROI, industry adjusted by subtracting the median performance of thesubjects primary three-digit SIC code, was compared with the applicable five-pointLikert scale survey response. As shown in Table 2, the survey responses exhibit ahigh degree of reliability. Spearman correlation coefficients range from 0.71 for 5 yearROI change to 0.78 for 3 year ROI change. When the continuous measures obtainedfrom Compustat are converted to ranks and assigned values of one to five with thesame frequencies as the survey responses, correlations are 0.76 for 5 year ROI changeand 0.86 for 3 year ROI change. The majority of responses are identical (66.3 percent), and 99 per cent of responses are within one value (e.g. report 4 on the surveyand compute 5 from Compustat data).3

Independent variablesABC use (ABC). Unless a system is used extensively, it seems unlikely thatit can be significantly associated with financial benefit. One would expect thebenefits received from an innovation to depend on the extent to which it becomesincorporated into organizational subsystems. Shields (1995) found that ABC successis significantly correlated with several categories of use: performance measurement,activity analysis, product costing, and reengineering. He also found significantcorrelation of success with the percentage of costs processed through ABC.There is widespread agreement that incentive structures play a critical role in

the success of ABC (Drake et al., 1999). Employees pay more attention to thosemeasures of performance that affect their personal welfare. Banker and Datar (1987)

2On the other hand, using archival data sources is not problem free either. For example, there aresignificant discrepancies in financial data between the Compustat and Value Line databases (Kern andMorris, 1994), and in SIC codes between CRSP and Compustat, limiting the ability to compute accurateindustry mean-adjusted variables (Ong and Jensen, 1994).3Variances can occur for reasons other than lack of knowledge by the internal auditor. For example,choosing 4 (agree) versus 5 (strongly agree) requires a value judgement that can vary between subjects.Also, subjects could be reporting their belief in true unobservable financial performance rather thanreported financial performance. Finally, the lack of complete congruence may have more to do with therespondents confidence in his/her answer than the strength of the financial performance.


Table 1Definition of variables

Abbreviation Name Definition

1ROI Change in return ininvestment

Industry median-adjusted ROI, measured by theself-reported five-point Likert responses provided bycompany internal auditors to the survey questionsOver the last three (five) years, the ROI of yourbusiness unit has improved relative to other businessunits in your industry.

FUNCTION Breadth of ABC use Use of ABC by organizational functions (e.g.manufacturing, engineering, top management). Basedon seven survey items (F16a-g), adapted from Swenson(1995).

APPLIC Depth of ABC use Use of ABC for specific applications, activities anddecisions, such as product costing and pricing decisions(F17a-i), adapted from Krumwiede (1996).

EVAL Integration inevaluation systems

Level of integration of ABC into firm strategic andperformance evaluation systems (F3, F6, and F7),adapted from Krumwiede (1996) and Shields (1995).

TIME Time sinceimplementation

Length of time since ABC implementation took place.

ABC ABC use The extent and depth of ABC use. Composite of thevariables FUNCTION, APPLIC, EVAL, and TIME.

INIT Use of other initiatives Use of the practices TQM, JIT, BPR, CIM, JIT, FMS,theory of constraints (TOC), and Value Chain Analysis(VCA). Single index developed from binary responses tosurvey item I12. (As recommended by Babbie (1990), inthe absence of compelling reasons for differentialweighting, the practices are weighted equally.)

INFO Information technology Operationalized through the six items of section D ofthe survey instrument. The items were developed basedon Reeve (1996) as modified by Krumwiede (1996, 1998).

COMPLEX Complexity anddiversity of business

Operationalized through seven items developed byEstrin et al. (1994) and used by Krumwiede (1996, 1998),which measure each type of complexity and diversity(section E of survey).

IMPORT Importance of costs Operationalized through the six items of section A,adapted from Estrin et al. (1994) and used byKrumwiede (1996, 1998).

INTRA Intra-companytransactions

Operationalized as the sum of two five-pointquantitative measures of percentage of intra-companypurchases and sales (items I10a and b). Expected to varynegatively with perceived benefit of ABC.

CAPAC Unused capacity A quantitative five-point measure ranging from 90 per cent (item I11). Sixteennon-manufacturers did not complete this survey item;the overall mean response is used for missing data.Unused capacity is expected to be negatively associatedwith improvement in performance.

COMP Level of competition Operationalized through survey item A6, adapted fromSwenson (1995).


Table 1Continued

Abbreviation Name Definition

SIZE Size of business unit The natural logarithm of the mid-point of an eight-pointself-reported sales category from a survey item adaptedfrom Krumwiede (1996). Measured as a naturallogarithm because research suggests a curvilinearrelationshipas size increases, innovation increases,but at a decreasing rate (Ettlie, 1983; Kimberly andEvanisko, 1981; Moch and Morse, 1977). The sign of theassociation is not predicted.

TYPE Type of company Binary variable, distinguishing manufacturing fromnon- manufacturing firms. No prediction is made as tothe sign of the association.

SUCCESS Success of ABCimplementation

Survey item H5, from Shields (1995).

SATISFACTION Satisfaction with thecost system

Survey item C1, from Swenson (1995).

BENEFIT Financial benefitobtained from ABC

Survey item H4, from Krumwiede (1996, 1998).

Table 2Correlations of self-reported dependent measures with actual reported (Compustat) performance measures adjustedfor industry performance (three-digit SIC)

Likert dependent measure with

Continuous measure Ranked measureMeasure n Pearson Spearman Pearson Spearman

ROI change3 years 51 0.77 0.78 0.86 0.86ROI change5 years 47 0.62 0.71 0.75 0.76

Likert dependent measure with ranked Compustat measureRanks Ranks Ranksidentical differ by 1 differ by >1

n Number % Number % Number %ROI change3 years 51 37 72.5 14 27.5 0 0.0ROI change5 years 47 28 59.6 18 38.3 1 2.1

98 65 66.3 32 32.7 1 1.0

Subject firms actual reported performance is adjusted by the median performance of firms in the subjectfirms primary three-digit SIC code. The number of industry firms ranges from four (SICs 376 and 799) to226 (SIC 131). Compustat firms are ranked 15 by ROI change with ranks assigned in the samefrequencies as the Likert scale dependent variable obtained from the survey.

demonstrated that lack of coordination between incentive systems and performancemeasures can wreak havoc on a firms performance. Swenson (1997) confirmed thatuse of ABC for both decision-making and performance measurement are typical ofthe best practices firms studied.Therefore, the construct ABC use (ABC) used in this study contains the following

three aspects of use: breadth of use (FUNCTION), depth of use (i.e. applicationsfor which ABC is used: APPLIC), and the level of integration into strategic and


performance evaluation systems (EVAL), as defined in Table 1. These three aspectscapture the extent of ABC use.Extent of (current) ABC use must be modified by the length of time since

implementation occurred (TIME). Realization of the advantages of an innovationoccurs when it is implemented on a widespread basis. This takes time. Evidencesuggests that plant-level implementation does not move in lock step with corporateimplementation (Swenson, 1995). In addition, accounting data has a historical focus;the benefits from use of ABC may not be measurable for several years.According to the theories of diffusion of innovation, diffusion of initiatives like

ABC in an organization is likely to occur in a non-linear manner (Kwon and Zmud,1987). Rogers (1983) suggests that diffusion of an innovation follows an S-shapedcurve. As an organization moves up the curve, a greater number of individuals andunits adopt the components of the innovation until a saturation point is reached onthe upper plateau of the S.To determine the change in performance attributable to the ABC initiative over

the measurement period, allowance must be made for the period of time duringwhich benefits were or were not received. A simple interaction between time sinceimplementation and current extent of ABC use would imply assumption of a linear,rather than a more appropriate S-shaped diffusion curve. Transformation of data tothe form of the cumulative probability function of a normal distribution (cdf)4 allowssimulation of the hypothesized S-shaped curve of diffusion. Therefore, the compositemeasure of ABC use (ABC) is constructed from the above three aspects capturingthe extent of ABC use, multiplied by the time diffusion percentage as simulated bythe cdf.

Other initiatives (INIT). Researchers have often noted that ABC and other strategicbusiness initiatives complement and enhance each other, rather than being individ-ually necessary and sufficient conditions for improvement (e.g. Cooper and Kaplan,1991; Anderson, 1995a; Evans and Ashworth, 1995; Shaw, 1998). Because ABC oftenprovides more and better information about processes, ABC may be most beneficialif other initiatives are employed concurrently. This linkage provides direction to theABC implementation and a ready application for the ABC information once it be-comes available (Swenson, 1998). For example, Carrier Corporation implemented JITand other improvement initiatives and used ABC as an enabler to support the de-velopment of cost effective product designs and manufacturing processes (Swenson,1998). Also, many companies have found that ABC fits well with the cost of qualityframework (Anderson and Sedatole, 1998). Krumwiede (1998) provided additionalweight to this argument by reporting that all 15 best practice firms had linked ABCto another improvement initiative. The variable INIT captures a firms use of otherstrategic business initiatives.

Enabling conditions. Prior research has suggested that the benefits of ABC aremore readily realized under specific environmental conditions. Evidence of ABCimplementation failures (e.g. Anderson, 1995a; Malmi, 1997) has caused researchers

41/(

2pi)

exp[1/2 2(x )2] where x = number of years since beginning of use of an initiative, = 5 years and = 2 years. Setting the mean of the probability distribution function as 5 years and thestandard deviation as 2 years allows assumption of a strongly sloping S over the 37 year interval and aplateau exceeding 10 years that is consistent with prior research (Husan and Nanda, 1995).


to suggest that achieving the systems objectives depends critically on organiza-tional and technical factors (Anderson and Young, 1999). Malmi (1997) states thatimplementation failures are related more to exogenous contextual factors than to theprocess of implementationthat even good implementation processes fall on barrenground. For example, Karmarkar et al. (1990) argued that complexity, importanceof costs and competition require more elaborate costing systems. Chenhall andLangfield-Smith (1998) recognize the potential moderating effects of environmentaland organizational variables and call for further research that considers the role ofadditional relevant firm-specific variables.The following variables are incorporated into the model testing the efficacy of ABC

(see Table 1 for variable specifications).

Information technology (INFO). Cooper (1988) suggests that ABC becomes morebeneficial as the cost of data collection and processing is reduced, which requireshigher levels of information technology. An information system providingdetailed historical data and easy access to users may provide much of the driverinformation needed by ABC. Reeve (1996) suggests that an integrated ABCsystem pre-supposes a relatively high level of INFO sophistication with extensiveand flexible information stratification and real-time activity driver information.SAP, PeopleSoft and Oracle have recently acquired ABC software companies orpartnered to develop ABC modules (Shaw, 1998).

Complexity and diversity (COMPLEX). A companys complexity increases asthe breadth of its product line expands, as each product uses more uniquecomponents, and as more process options are available to manufacture theproduct (or provide the service) (Swenson, 1998). Complexity factors are thebiggest single driver of cost (Gonzalves and Eiler, 1996). Product diversity andproduction process complexity have been shown to be important determinants ofthe need for reexamining cost allocation procedures (Gosselin, 1997a). Previousstudies have confirmed that ABC data are most likely to differ from traditionalcost data in settings with high coordination and control costs, such as those withdiverse products, processes, customer demands, or vendors (Foster and Gupta,1990; Cooper and Kaplan, 1991; Pattison and Arendt, 1994; Estrin et al., 1994;Anderson, 1995b; Banker et al., 1995).

Importance of costs (IMPORT). Even if ABC could substantially reduce product costdistortions, it is not likely to be helpful unless a firm can actually utilize better costinformation in its decision-making process. Besides competitive environment,other factors affecting the decision usefulness of cost information include thefirms use of cost data in pricing decisions, cost reduction efforts, need for specialcost studies, strategic focus, and average profit margin (Estrin et al., 1994).

Intra-company transactions (INTRA). When companies have a large number ofintra-company transactions, the financial performance of individual businessunits may bemisleading because of transfer pricingmethodology, and constraintson decision-making regarding source of supply and customer selection (Swenson,1995).5 Therefore, intra-company transactions are a potentially confoundingvariable to this study.

5Some companies with many intra-company transactions may use ABC costs to set prices, the positivevalue of ABC for these companies offsetting the negative effect of having large numbers of intra-companytransactions.


Unused capacity (CAPAC). ABC theory predicts that due to improvements inresource usage or cost-reduction programs, unused capacity will be created,particularly with respect to batch and product-sustaining activities. If managershave acted to eliminate these unused capacities, then the effects of ABC wouldshow up through lower costs. If, however, managers have not eliminated theseunused capacities, possibly because significant joint and indivisible costs arepresent, then the non-valued added costs identified byABCmay not translate intocost reductions or profit improvements (Kaplan, 1993). Maher and Marais (1998)demonstrated that ABC is likely to give erroneous cost estimates when there isa discontinuous relation between the demand for and provision of resources.Therefore the more unused capacity the less helpful is ABC.

Competition (COMP). As early as 1972, Khandwalla found that output marketcompetition is associatedwith greater use of management controls. More recently,Mia and Clarke (1999) argued that management accounting systems (MAS)can provide information used to identify, evaluate, and implement appropriatestrategies and found that level of competition is a determinant of the use of MAS.

As competition increases, there is a greater chance that a competitor will exploit anycosting errors made. In addition, research has shown that non-competitive situationssuch as oligopoly can lead to optimal, strategic costing systems that have more incommon with traditional mark-ups than with ABC (Alles, 1990; Banker and Hughes,1991; Banker and Potter, 1991). Thus, more reliable cost information may be neededas competition increases (Cooper, 1988).

Control variablesBusiness unit size (SIZE). Hicks (1999) argues that ABC works just as well in smallfirms as it does in large ones. The literature proposes two conflicting effects for theinteraction of firm size with ABC. Anderson (1995a) concluded that implementationis most likely to be disruptive if it occurs over a protracted period. Large, verticallyintegrated firms are more likely to have lengthy implementation processes that causesignificant organizational disruption. However, Selto and Jasinski (1996) reason that,other than in large companies that are well staffed and well trained, ABC is notsufficiently understood to be implemented successfully as a stand-alone system, letalone being integrated within a companys strategy. The combination of the time andfirm size variables provide control for the organizational disruption anticipated byAnderson (1995a). The identification of breadth and depth of use of ABC providescontrol for the small company resource problems noted by Selto and Jasinski (1996).

Type of company (TYPE). Georgantzas and Shapiro (1993) and Schroeder (1990)analytically demonstrated that industry type moderates the relationship betweeninnovation and performance. In this study, macro-economic differences betweenindustries are controlled through the use of industry-adjusted dependent variables,eliminating the need to model a direct effect. Firm-specific conditions affectingABC are measured through enabling condition variables. However, ABC researchalso suggests that the efficacy of initiatives may fundamentally differ betweenmanufacturing and service companies (Rotch, 1990; Cooper, 1988, 1989). Therefore,a binary variable differentiates the 106 manufacturing firms from the 98 non-manufacturing firms and is interacted with ABC.


Survey Questions Dependent Survey Questions(Observable X) Independent Constructs Constructs Observable Y

ns

?

1 H5

1 H4

1 C1

8 I12a-j

6 D1-D6

7 E1-E7

5 A1-A5

2 I10a, b

ABCSuccess

Intra-Company

Transactions

OtherInitiatives

(INIT)

Complexity-Diversity

(COMPLEX)

Importance ofCosts

(IMPORT)

Enabling Conditions

(ENABLERS)

Unused Capacity (CAPAC) I11

Size(SIZE) 18Competition

(Comp) A6

InformationTechnology

(INFO)

--

--

ROI 3YrC6

ROI 5YrC7

?

?

Key

Construct confirmed with factor analysis

Additive index

Observed variable

2 F3, F4 Number of items and survey questionnumbers used to measure variables--Some questions have multiple parts(F16, F17, I10, I12)

Direct Effect

Moderating Effect (interaction)+, --, ? Expected sign of coefficient

(ROI)

Change inFinancial

Performance

?

Type(TYPE) 16

Figure 2. H2: Association of SUCCESS with improved financial performance.

ABC SuccessThe variables of interest in testing Hypothesis 3, which associates previouslyreported measures of ABC success with improvement in ROI, are those developedby Shields (1995); Swenson (1995), and Krumwiede (1996, 1998). Figure 2 showsthe changes that need to be made to the ABC model described thus far. Thesection that modelled ABC use is replaced by single item measures of SUCCESS,SATISFACTION, and financial BENEFIT, respectively (Table 1).

5. Sample selection and survey instrument

SubjectsAs noted by Shields (1995), one of the limitations of previous research is thatresults may have been weakened because of being based on responses providedby potentially biased subjects, those responsible for design, implementation, andoperation of the innovation.6 For example,McGowan andKlammer (1997) and Fosterand Swenson (1997) found that perceptions of ABC vary depending on the role of

6As with other studies, because this research relies on self-reported data, it is potentially subject toreporting biases and measurement error called common-method bias (Johnson et al., 1995). However,Miller and Roth (1994) suggest that care in the selection of respondents can contribute to overcomingcommon-method bias. The selection of unbiased, objective, and knowledgeable internal auditors isbelieved to eliminate most, if not all potential effects from common-method bias that may be presentin other research.


the individuals involved. Preparers reported more favourable attitudes toward ABCthan users, with project leaders or champions being most favourable.The current studymitigates this limitation by utilizing internal auditors as subjects.

The Statements of Responsibilities in Internal Auditing (IIA, 1990), and Section 100 of theStandards of Practice for Internal Auditors (IIA, 1998) require that internal auditors beindependent of the activities they audit. Independence permits internal auditors torender impartial and unbiased judgments (Standards, Section 100.01). In addition totheir independence and objectivity, internal auditors are appropriate subjects becausethey are knowledgeable, possess varied talents and expertise, and have access torelevant information (Tatikonda and Tatikonda, 1993; Stoner and Werner, 1995).

Population and sampling proceduresThe firms studied are for profit firms that employ internal auditors who aremembers of the Institute of Internal Auditors (IIA). The sample is drawn fromthe population of those practicing members of ten geographically diverse USchapters of the IIA7 where information was available to the researchers. Auditorsemployed in the banking industry were excluded because they often have highlyspecialized responsibilities, limiting their exposure to new business initiatives.Auditors employed by governmental and non-profit organizations were excludedas well since those organizations do not measure improved financial performanceas improvement in profitability (ROI). Sample size is further limited to fiverandomly drawn subjects per organization. A mail survey was used to collect theinformation.To permit generalization to the broader population of firms, the sampling design

included entities with a broad range of applications in ABC ranging from noimplementation to extensive applications. As stated by Chenhall (1997), the samplingensured that the study included a proportion of firms that employed varying levels ofuse of other strategic business initiatives and varying levels of enabling conditions.This increased the comprehensiveness of the study, and permitted the identificationof the effects on performance of the amplifying effect of enabling variables across arange of ABC observations.In order to investigate the effect of firm characteristics on implementation project

outcomes a large sample of firms is needed (Anderson and Young, 1999). Accord-ingly, the questionnaire was distributed to 1058 internal auditing professionals.This sample was reduced by 68 that were returned unopened because of incorrectaddress or change of employment with no forwarding address (see Table 3). Inaddition, 28 uncompleted or partially completed surveys were returned because thesubjects were not knowledgeable about their companys systems, company policiesagainst response to surveys, or other reasons, leaving an adjusted sample size of962. 210 completed responses were received. In six instances, there were multipleresponses from the same business unit. Differences were minor, and responses werecombined into a single observation by averaging scores. Of the remaining 204 usableresponses, 137 are from the first and 67 from the secondmailings, yielding a response

7The IIA serves as the internal auditing professions authority on significant issues affecting internalauditors, and is the only organization dedicated solely to the advancement of the internal auditor andthe profession on a worldwide basis. The IIA is the worlds leader in research and educational issues forinternal auditors and is the standard-setting body for the profession. It has approximately 70 000 membersin 230 local chapters, national institutes, and audit clubs in more than 120 countries (IIA, 2001).


Table 3Summary of sample

Questionnaires mailed 1058

Less: undeliverable 68

Net questionnaires delivered 990

Less: incomplete responses:Company does not use cost allocation methods 12Company policy against responding to surveys 6Respondent is consultant 1Respondent is no longer employed at subject firm 7Respondent is not knowledgeable about cost systems 2 28Net responses possible 962

Responses receivedFirst mailing 141Second mailing 69 210

Response rate 21.8%Less: responses from same business unit averaged intosingle response 6

Useable responses 204 21.2%

Responses reportingNo use of ABC 139Implementation stage of ABC 18Use of ABC for decision-making 47

204

Generally because the position is extremely specialized such as railroad rate auditor.

rate of 21.2 per cent. Sixty-five responses (31.8 per cent) indicate some use of ABC.The remaining 139 respondents serve as a non-using control group. The number ofresponses is adequate for purposes of statistical analysis.8

8For effective analysis, the sample covariance matrix must be reasonably stable and approximate thepattern of covariances in the population. In general, ceteris paribus, the larger the sample size themore likely this will be the case. Guadagnoli and Velicer (1998) reviewed the literature on sample sizeconsiderations in factor analysis and principal components analysis and conducted an extensive MonteCarlo study on sample size effects. Consistent with other Monte Carlo studies, they found no support foroften used rules of thumb based on respondents-to-variables criteria (e.g. 5 : 1). As quoted from Jaccardand Wan (1996),The most important factors influencing the stability of the sample covariance matrix were the absolutesample size and the magnitudes of the path coefficients from the latent constructs to the observedindicators (referred to as saturation). When such standardized path coefficients were low (i.e. near 0.40),sample size was quite important. At moderate to high saturation levels (e.g. standardized path coefficientsof 0.60 to 0.80), once a certain sample size was achieved, further improvements in stability were small withincreasing N . When saturation was high (standardized path coefficients of 0.80), sample sizes as low as 50performed well, even when the number of variables in the covariance matrix was large.Jaccard and Wan then recommend a sample size of 75100 in conditions of high saturation, and 150 formoderate saturation levels. The saturation levels obtained in this study are high for 76 per cent of themulti-item variables used to test H1 and H2 and moderate to high or high for 80 per cent of those usedto test H3. These levels are adequate to expect a stable covariance matrix.As a check, a sensitivity test is performed whereby the 19 ABC manifest variables are reduced to six,reducing the number of manifest variables to 25 and increasing the sample size/variable ratio to 8 : 1 from5 : 1. Results are not impacted.


There is no test to ensure that the study is free of non-response bias. Two separateprocedures were performed to help assess the possibility of bias. As in (Gosselin,1997a) and Krumwiede (1998), a reason for non-response section was included atthe bottom of the transmittal letter. In addition, the median responses of the firstmailing were compared with those of the second mailing. Wilcoxon 2-sample signedrank tests (Hollander and Wolfe, 1973) and Pearson chi-square tests of proportions(Feinberg, 1983) on both the raw data and the additive indexes revealed significantdifferences (p< 0.05) on five of the 75 variables tested, or 6.7 per cent. This is slightlymore than would be expected by chance. Second mailing respondents tended toreport at a somewhat higher level of aggregation (e.g. company versus division;median 4.82 versus 4.29p< 0.0385), have less tendency to be manufacturers(0.42 versus 0.57p< 0.044), be less likely to use CIM (0.10 versus 0.22p< 0.047),and be less satisfied with their business unit cost (3.20 versus 2.91p< 0.041) andperformance measurement (3.18 versus 2.87p< 0.033) systems. It is not suprisingthat the test revealed some differences. For example, a possible explanation for slowerresponses by internal auditors with company-wide responsibilities is that they tendto travel more often, and are thus likely to have delayed responses.

Survey instrumentData were extracted from a 96-item survey instrument. As in Kaynak (1996), Shields(1995), Swenson (1995), Grandzol and Gershon (1997), McGowan and Klammer(1997) and Krumwiede (1996, 1998) the instrument was constructed so that analysiscould be conducted at the appropriate level of knowledge of the individualrespondents (plant, division, region, subsidiary, country, or entire company).9 Thisreduced measurement error associated with differing levels of ABC use in differentsegments of firms. As described in the variable descriptions, most of the survey itemswere adapted from previous research.10 The survey instrument is included in theappendix.

6. Results

Descriptive statisticsAs described in Table 3, 204 completed questionnaires were returned. Fifty per centof the respondents reported for their entire company, 17.2 per cent reported for theirdivision, with the remainder spread among plant, group, subsidiary, and countrybusiness units. 46.6 per cent reported that their business unit revenues exceed $1billion, while, as is not suprising for firms employing internal auditors, only 8.3 percent reported for business units with revenues under $50 million. Manufacturingfirms constituted 52 per cent of the responses.

9Multiple tests were run whereby performance and ABC use were regressed against level of business unit.In no cases was a business unit significant at conventional levels.10Procedures prescribed by Dillman (1999) for maximizing response rates were followed. Specific stepstaken included (1) sending a second mailing, (2) promising confidentiality of responses, (3) includingdeadline dates for reply, (4) including personalized cover letters, (5) including a postage-paid, self-addressed envelope for reply, and (6) promising to send a summary of results on request.


As reported in Table 4, all but 45 (22.1 per cent) of the respondents indicated thattheir business unit was significantly using at least one business initiative. Themedianfirm used two practices (range from zero to six) with JIT and TQM the most oftenreferenced at 46 per cent. Manufacturers had more mean use (2.56 versus 1.32) thannon-manufacturers and companies over $1 billion in revenues had higher use thansmaller companies (2.2 versus 1.5). Discounting the purely manufacturing initiativesCIM and FMS, the difference in use between manufacturers and non-manufacturersreduced to 2.06 initiatives versus 1.32 initiatives.Forty-seven respondents, 23 per cent, reported that they were significant users

of ABC. Another 18 respondents indicated that they were implementing ABC, butthat the system was not yet in significant use. This rate is somewhat lower thanprior research (Shim and Stagliano, 1997; Geishecker, 1996), which reported that 27to 44 per cent of respondents were using ABC.11 Manufacturers reported higheruse than non-manufacturers (31.1 per cent versus 14.3 per cent). Significant ABCusers generally also used other initiatives (mean of 2.2 other initiatives). Only sevenof 47 ABC firms were using ABC in isolation.As reported in Table 5, significant users tended to use ABC in several applications,

with, as expected, cost reduction and product costing representing the highest uses(4.37 and 4.13 out of 5, respectively). The majority, 34, had been using ABC forover two years. 72 per cent felt that the implementation had been successful, 66 percent felt that ABC had been worth implementing, and 64 per cent felt that benefitsexceeded costs. The correlation of SUCCESS with BENEFIT is 0.60, statisticallysignificant and consistent with that found by Shields (1995) of 0.53.The correlation matrix portraying the univariate relationships between new

business initiatives is presented as Table 6. For the full sample, 39 per cent of therelationships are significant at the = 0.05 level, and all significant relationshipsare positive except that of business process reengineering (BPR) and the theoryof constraints (TOC). JIT exhibits the strongest relationship with other initiatives,with significant correlations between it and all other initiatives except TOC. ABCis significantly correlated with JIT at the = 0.05 level, and BPR at = 0.10.Somewhat suprisingly, the relationships are qualitatively similar for the partitionincluding manufacturing firms only.

Content validity and reliabilityConfirmatory factor analysis is used to test the unidimensionality of each of thesix multi-item constructs FUNCTION, APPLIC, EVAL, INFO, COMPLEX, andIMPORT.12 One indicator of fit is the chi-square statistic (2). A good fitting modelmay be indicated when the ratio of 2 to the degrees of freedom is less than two(Tabachnick and Fidell, 1996). However, this statistic is sensitive to sample sizeand violations of assumptions of multivariate normality (Bentler, 1983; Joreskogand Sorbom, 1989), which can lead to rejections of the model even when the fit isreasonable. Therefore, it is useful to supplement the 2 with other indicators of fit.

11Prior research has generally used samples from populations that consisted of manufacturers exclusively.This is a likely explanation for the lower use of ABC found in this study.12To use confirmatory factor analysis for verifying unidimensionality, a measurement model is specifiedfor each construct. Individual items constituting the construct are examined to see how closely theyrepresent the same construct.


Table4

Useofinnovative

businesspractices,

n=

204

Percentage

ofSu

rvey

Num

berPercentage

ofManufacturers

Non

-manufacturers

userswith$1

item

Businessinitiative

using

total

Num

ber

Percent

Num

ber

Percent

billion

(n=

106)

(n=

98)

I12

Activity-basedcosting(A

BC)

4723.0

3331.1

1414.3

40.4

Totalqualitymanagem

ent(TQ

M)

9446.1

6561.3

2929.6

58.5

Just-in-time(JIT)

9546.6

6662.3

2929.6

56.8

Com

puter-integrated

manufacturing

(CIM

)37

18.1

3734.9

00.0

54.1

Businessprocessreengineering(BPR

)79

38.7

4239.6

3737.8

49.4

Value

chainanalysis(VCA)

2311.3

1615.1

77.1

43.5

Flexiblemanufacturing

system

s(FMS)

167.8

1615.1

00.0

68.8

Theory

ofconstraints(TOC)

94.4

65.7

33.1

66.7

Number

ofinitiativesin

use

01

23

45

6Total

Respo

nses

(total)

4545

3447

236

4204

Respo

nses

(manufacturing

)14

1912

3021

64

106

Respo

nses

(non

-manufacturing

)31

2622

172

00

98%(Total)

22.1

22.1

16.7

23.0

11.3

2.9

2.0

100.0

%(M

anufacturing

)13.2

17.9

11.3

28.3

19.8

5.7

3.8

100.0

%(N

on-m

anufacturing

)31.6

26.5

22.4

17.3

2.0

0.0

0.0

100.0

Allinitiatives

Omitting

purelymanufacturing

initiatives

Total

Manufacturers

Non

-manufacturers

Total

Manufacturers

Non

-manufacturers

MeanNum

berof

Initiatives

1.96

2.56

1.32

1.70

2.06

1.32

Median

23

12

21

46.6%

ofrespon

seswerefrom

businessun

itsov

er$1

billion

. O

mitting

CIM

andFM

S.

ABC and Improvement in Financial Performance 19Table5

UseofABCSign

ificant

Users,n

=47

Survey

Extent

ofuse

Num

berof

respon

ses

item

App

lication

12

34

5Mean

Median

4&5

Percent

F17

Prod

uctcosting

04

518

194.13

437

78.7

Costreduction

00

225

194.37

444

93.6

Pricingdecision

s0

912

1114

3.65

425

53.2

Prod

uctm

ixdecision

s0

87

1813

3.78

431

66.0

Determinecustom

erprofi

tability

06

1022

83.70

430

63.8

Budg

eting

06

822

103.78

432

68.1

Off-lineanalytictool

02

1520

93.70

429

61.7

Outsourcing

decision

s0

1011

196

3.46

425

53.2

Performance

measurement

07

624

93.76

433

70.2

Survey

Num

berof

years

item

Timesince

5

H1a

Implem

entation

ofABC

49

713

59

H1b

Use

fordecision

-making

713

99

45

H4

ABC

hasbeen

worth

implem

enting

?

Value

Num

berPercent

Yes

3166.0

Will

be4

8.5

Tooearlyto

tell

1123.4

No

12.1

H5

ABC

hasbeen

successful?

F15

bene

fit>

cost?

Value

Num

ber

Percent

Num

ber

Percent

Strong

lyagree

919.1

1327.7

Agree

2553.2

1736.2

Noop

inion

817.0

1225.5

Disagree

36.4

48.5

Strong

lydisagree

24.3

12.1

65respon

dentsindicatedsomeuseof

ABC

.47respon

dentsindicateduseto

asign

ificant

extent

indecision

-making.


Table 6Correlation matrix of new business initiatives. Manufacturers, n = 106, to the lower left of the diagonal; all firms,n = 204, to the upper right (Spearman correlations)

ABC JIT CIM BPR VCA FMS TOC TQM

Activity-based costing (ABC) 1.000 0.202 0.041 0.125 0.030 0.027 0.059 0.090Just-in-time (JIT) 0.252 1.000 0.249 0.267 0.164 0.276 0.087 0.340Computer integrated manufacturing (CIM) 0.058 0.241 1.000 0.018 0.074 0.147 0.023 0.126Business process reengineering (BPR) 0.126 0.273 0.074 1.000 0.130 0.030 0.171 0.254Value chain analysis (VCA) 0.056 0.234 0.040 0.052 1.000 0.300 0.301 0.044Flexible manufacturing systems (FMS) 0.004 0.229 0.129 0.093 0.349 1.000 0.295 0.014Theory of constraints (TOC) 0.049 0.115 0.002 0.019 0.467 0.428 1.000 0.055Total quality management (TQM) 0.174 0.313 0.049 0.485 0.081 0.069 0.130 1.000Bold = significant at the 0.05 levelUnderlined = significant at the 0.10 level

Total Full sample ManufacturersPossible Number % Number %

Significant at 5% 28 11 39.3 10 35.7Significant at 10% 28 14 50.0 11 39.3

JIT CIM BPR VCA FMS TOC TQMNumber significantly using ABC with 30 7 23 6 3 1 25Per cent (47 firms) 63.8 14.9 48.9 12.8 6.4 2.1 53.2

Number of initiatives

Mean 0 1 2 3 4 5 6Number significantly using ABC with 2.2 7 6 17 10 4 2 1Per cent 14.9 12.8 36.2 21.3 8.5 4.3 2.1

A goodness of fit index (GFI) of 0.90 or higher for the model suggests that there is noevidence of a lack of unidimensionality (Joreskog and Sorbom, 1989), and an adjustedgoodness of fit index (AGFI) of 0.80 and a root-mean-square residual (RMR) under 0.10are generally regarded as indications of good fit (Libby and Tan, 1994).The 2 statistics, and GFI, AGFI, and RMR indices for the six constructs, are

reported in Table 7. After deletion of five of 37 survey items, 2 tests that the modelsfit the data are not rejected (p< 0.01). Furthermore, all GFI and AGFI values areabove 0.90 and 0.80, respectively, indicating that there is no evidence of a lack ofuni-dimensionality.A scale exhibits discriminant validity if its constituent items estimate only one

construct (Bagozzi et al., 1991). Lack of discriminant validity usually results inan over-estimation of correlation among constructs. To test scales for discriminantvalidity a 2 difference test is used (Ahire et al., 1996). A set of confirmatory factoranalyses is run on each multi-item pair of scales, first allowing for correlationbetween the two constructs and then fixing the correlation between the two scalesat one. A statistical significant difference in 2 statistics demonstrates that the twoconstructs under consideration are distinct (Venkatraman, 1989).For the six multi-item scales in the instrument, a total of 15 discriminant validity

checks were run. The three ABC scales (FUNCTION, APPLIC, and EVAL) failedto yield statistically significant 2 differences (the 2 difference is under two).


Table7

Constructun

idim

ensionality,conv

ergent

valid

ity,andreliability

Adjusted

Rootm

ean

Bentler-

Goodn

ess

good

ness

square

Bonn

ett

Cronb

ach

Instrument

Num

berof

item

sChi-squ

are

offit

indexof

fitindex

residu

alcoefficient

alph

aCon

struct

Description

item

sOriginalDeleted

degreesof

freedo

mp-value

(GFI)

(AGFI)

(RMR)

(N

FI)

(A

lpha)

FUNCTION

Function

susingABC

F16a-g

70

2.96/7

0.89

0.99

0.95

0.026

0.99

0.90

APP

LIC

App

lications

ABC

used

for

F17a-i

90

16.60/19

0.62

0.95

0.88

0.044

0.96

0.92

EVAL

Use

forperformance

evaluation

F3,6,7

30

2.23/1

0.14

0.98

0.87

0.027

0.98

0.87

INFO

Inform

ationtechno

logy

soph

istication

D1D6

61

3.93/4

0.42

0.99

0.97

0.016

0.99

0.84

COMPL

EXCom

plexitydiversity

E1E

77

21.81/4

0.77

1.00

0.99

0.012

0.99

0.79

IMPO

RT

Impo

rtance

ofcosts

A1-A5

52

10.71/1

0.00

0.97

0.80

0.082

0.91

0.54

375

ABC

AllABC

variables

190

131.99/127

0.36

0.94

0.91

0.012

0.98

0.94

GFI

valueof

0.90

sugg

eststhatthereisno

lack

ofun

idim

ension

ality(JoreskogandSorbom

,1989); A

GFI

valueof

0.80

indicatesagood

fitting

mod

el(Tabachn

ickandFidell,1996); R

MRvalueof

under0.10

indicatesagood

fitting

mod

el(Tabachn

ickandFidell,1996);NFI

valueof

0.90

sugg

estsstrong

conv

ergent

valid

ity(Tabachn

ickandFidell,1996); A

lpha

valueof

0.50

indicatesacceptablereliability(N

unnaly,1978).


Therefore, after confirming unidimensionality, the 19 variables from those constructswere combined into a single construct for testing.An NFI value of 0.90 or above demonstrates strong convergent validity (Tabachnick

and Fidell, 1996). The NFI values for all of the constructs are reported in Table 7. Allof the scales had values over 0.90, demonstrating strong convergent validity.Reliability refers to the degree of dependability, consistency, or stability of a scale

(Gatewood and Field, 1990). Cronbachs coefficient alpha () is a widely usedmeasure of scale reliability (Cronbach, 1951). The Cronbachs alpha values for eachconstruct are shown in Table 7. All scales have acceptable reliability.

Preliminary test of efficacyPrior to formal hypothesis testing a rough approximation of the primary model istested. This (regression) model does not include the refinements and advantagesobtained from use of structural equation modelling (as specified in Figure 1).However, it does yield information regarding the overall efficacy of the enablers thatis not obtained with the LISREL model.A construct composed of the two ROI variables is regressed against constructs

for ABC use, other initiative use, size, a composite construct composed of the sixenablers, and an interaction variable (composed of ABC, other initiative use, andenablers). Survey items are weighted equally within constructs and constructs areweighted equally within composite constructs. The regression model is

1ROI = + 1ABC+ 2INIT+ 3ENABLE+ 4ABCINITENABLE+ 5SIZEWhere Expected

sign1ROI = the average of five-point measures of +

industry-adjusted improvement of ROIover three and five years

ABC = the average of 19 five-point Likert +measures of ABC use

INIT = the sum of eight binary measures of +significant initiative use

ENABLE = the average of six measures of enabling ?variables which in turn are composed ofthe average of individual survey items

ABCINITENABLE = an interaction term +SIZE = the log of the mid-point of a five-point ?

Likert sales category

The results of this regression are presented as Table 8. Use of other initiatives issignificant at the = 0.05 level, and the interactive term is significant at 0.081. Thereappears to be an overall effect of enabling variables and use of initiatives combinedwith ABC. It is noteworthy that this effect is not present when ABC is dropped fromthe interaction term. The contrast between effects with and without inclusion of ABCis an indicator of probable efficacy of the use of ABC under favourable enablingconditions.


Table 8Exploratory multiple regression analysis of overall effect of enabling conditions on improvement in performance,n = 204

Model F 6.666Model p-value 0.0001R square 0.1441Adjusted R square 0.1225

StandardizedParameter parameter Standard

Variable estimate estimate error t-statistic p-value

Intercept 1.018 0.000 0.950 1.072 0.285ABC 0.038 0.064 0.066 0.576 0.565Other INITiatives 0.152 0.221 0.056 2.697 0.004ENABLErs 0.235 0.096 0.171 1.372 0.172ABCINITENABLE 0.009 0.175 0.007 1.405 0.081SIZE 0.049 0.105 0.322 1.532 0.127

Bold = significant at the 0.05 level; Underlined = significant at the 0.10 level; Intercept and size tested withtwo-tailed test; other variables with one-tailed tests.

Hypothesis testingThe purpose of the first two hypotheses is to test whether ABC is directly associatedwith improvement in ROI (H1) and to identify the enabling conditions under whichABC results in an improvement in ROI (H2). To perform these tests, the conceptualmodel presented previously as Figure 1 is modified to that shown in Figure 3.Figure 3 also reports the results of testing.13 Product terms are created for theinteractions between ABC and each of the enabling variables, other initiatives, andsize. Positive significance of the ABC variable would indicate a direct effect on changein performance, regardless of environmental conditions. Positive significance of aproduct term indicates that ABC is positively associated with an improvement inperformance when used in the environment described by the product term.14

The fit of the model is good: 2 (1017 df) = 911, p< 0.99, GFI = 0.96, AGFI = 0.92,RMR = 0.075. The model explains 36% of the variance of the dependent construct.Many of the variables have significant direct effects: INFOrmation technology,IMPORTance of costs, SIZE, and other INITiatives have positive direct effectsat the 0.05 level. Number of INTRA-company transactions and COMPetitiveenvironment have negative direct effects at the 0.05 level. COMPLEXity (positive)and unused CAPACity (negative) are not significant at conventional levels. It islogical that INFO, SIZE, INIT and COMP are significant predictors of change in

13Factor loadings and structural coefficients are obtained using the maximum likelihood estimationmethod. Estimation involves finding the values of the coefficients that produce an estimated covariancematrix that is as close as possible to the sample covariance structure of the manifest variables (Libby andTan, 1994).14A potential problem with this approach is that the measurement error for a given product indicatormust be a function of the measurement error of the component parts of the product terms (Jaccard andWan, 1996). Joreskog and Yang (1996) developed an approach to address this problem, which requires theformation of four new matrices and the imposition of nine constraints per product term. The resultingmodel requires estimation of a number of parameters that is larger than the sample size, resultingin unstable parameters. However, the parameters and t-statistics derived are nearly identical to thosepreviously reported. All variables retain their signs and significance levels are stable within 0.05 and 0.10boundaries.


X VARIABLES

7 F16a-g

9 F17a-I ABC 0.050.39

3 F3, 6,7 (0.348)

0.10 8 I12a-j Add INIT 0.30 1.89 ABC X INIT

4.78 (0.003)(0.000)

5 D1-D5 INFO 0.35 0.02 ABC X INFO4.57 0.31

(0.000) (0.388)

5 E1-E5 COMPLEX 0.02 0.12 ABC X COMPLEX0.31 2.26

(0.757) (0.012)

4 A1-A4 IMPORT 0.20 0.07 ABC X IMPORT2.05 1.30

(0.040) (0.097)

-0.16 -0.09 2 I10a,b Add INTRA -1.41 ABC X INTRA

-3.14 (0.079)(0.002)

-0.02 1 I11 1 CAPAC -0.00 -0.14 ABC X CAPAC

-0.05 ROI-3 YR (0.444)(0.960) 1 C6

ROI-5 YR1 A6 1 COMP -0.21 1 C7 0.03 ABC X COMP

-4.66 0.66 (0.000) (0.255)

1 I6 1 SIZE 0.22 -0.10 ABC X SIZE2.94 -1.69

(0.003) (0.093)

0.04 1 I8 1 TYPE 0.54 ABC X TYPE

(0.295)

KeyConstruct 1 coefficient constrained to 1

Observed variable 0.01 Standardized coefficient0.08 t-statistic

2 F3,F4 Number of items and survey question -0.936 p-valuenumbers used to measure variables

INIT Bold & Filled = significant at 0.05Direct Effect

ABC X SIZE Underlined = significant at 0.10Add Construct obtained by summing responses

Change inFinancial

Performance

INTERACTIONCONSTRUCTS

DIRECTCONSTRUCTS

Figure 3. LISREL models of ABC and financial performance with control for enabling conditions andother initiatives.

financial performance. The significance of IMPORT may indicate cost awareness,presumably leading to cost control efforts. The significance of INTRA may indicate apreponderance of pricing at other than market prices.The effect of ABC, although positive, is not significant (p< 0.3483). This means that

there is no direct affect associated with use of ABC. As expected, H1 is not confirmed.More importantly, however, the interactions of ABC with COMPLEXity (p< 0.012)

and other INITiatives (p< 0.003) are positive and significant, which means that H2 isconfirmed for those conditions. The interactions of ABC with IMPORTance of costs,INTRA-company transactions, and SIZE are significant at 0.10 (p< 0.097, p< 0.079,and p< 0.093, respectively). The signs of the other enabling variable interactions areas expected, although not significant. It is very possible that use of a larger samplesize would have increased statistical power sufficiently to result in significance. Also,results for ABCCAPAC (unused capacity) may be weakened in twoways: (1) surveyresponses indicated that non-manufacturers had difficulty assessing their capacityutilization, and (2) for some companies, ABC may become less relevant as they


approach capacity constraints. Opportunity costs become more important than costallocation.The results of the three tests of H3 are presented in Table 9. As expected with use

of single item variables of interest, model fit is not as good as that of the previousmodel. 2 generally approaches three times degrees of freedom rather than thedesired two. GFIs range in the lower 0.80s and AGFIs in the upper 0.70s, althoughthe RMR for all three models are under 0.08. Variable significance is consistent forthe three models. INFOrmation technology, IMPORTance of costs (except againstSATISFACTION), and other INITiatives are positive and significant at 0.05. INTRA-company transactions and COMPetitive environment are negative and significant.SATISFACTION (p< 0.104), SUCCESS (p< 0.059), and financial BENEFIT (p< 0.174)are positively signed and SUCCESS is marginally significant. Although no firmstatistical conclusions can be reached regarding H3, it appears that the variables arerelatively good proxies for improvement in performance associated with use of ABC.

Sensitivity analysisAs additional checks on the specifications of the models, the analysis was re-estimated with (1) limited, and (2) substantial error correlation allowed betweenthe independent manifest variables, (3) restriction of the error correlation of thedependent variables, (4) change in ROI over separate 3 and 5 year periods ratherthan a construct derived from the combination of the two periods, (5) all correlationsbetween the latent constructs allowed rather than only those statistically significantat the 0.10 level, (6) a direct effect of industry TYPE on change in ROI (even thoughthe ROI variable is industry adjusted), (7) a reduction in the number of manifestABC variables from 19 to six (H1 and H2 model only), and (8) for the precedingexploratory regression analysis, change in ROI measured over separate 3 and 5year periods. Although there is some change in fit statistics of the models, there islittle change in significance levels of the independent variables, with the followingexceptions.When correlated errors of the manifest independent variables are estimated,

t-statistics of all variables tend to increase. If a large number of correlations areestimated, the interaction terms and the SUCCESS and SATISFACTION variablesbecome significant at = 0.05. However, in the absence of an error theory to explainthese correlations, no inference can be made from these results.Since correlated errors were expected for the dependent variables ROI3 and ROI5,

the results have been reported with correlated error terms. When the errors betweenROI3 and ROI5 are not allowed to correlate, significance levels of the independentvariables are generally weakened somewhat. Although signs remain as expected,ABCIMPORTance of costs and ABCSIZE lose their significance (p< 0.176 andp< 0.142 versus p< 0.097 and p< 0.093). ABCINIT also loses some significance (p 50%

b. Your purchases < 10% 1025% 2550% > 50%

11. At what percentage of capacity does your business unit usually operate?

< 50% 5065% 6580% 8090% > 90%

12. Check if the following is used to a significant extent in your business unit:

a. Activity-based costing (ABC) b. Just-in-time (JIT)

c. Computer-integrated d. Business process

manufacturing (CIM) engineeringf. Value chain analysis g. Flexible manufacturing

systemsh. Theory of constraints (TOC) i. Total quality management

(TQM)j. Lean manufacturing techniques j. Other (describe)

Please comment on any refinements that can be made to the survey (questionsneeded, unnecessary, or those that should be changed):

Thank you for participating!

References

Ahire, S. L., Golhar, D. Y. and Waller, M. A., 1996. Development and validation of TQMimplementation constructs, Decision Sciences,Winter, 2356.

Alles, M., 1990. A Strategic Model of Costing, Working Paper, Stanford University.Anderson, S. W., 1995a. A framework for assessing cost management system changes: thecase of activity-based costing implementation at General Motors, 19861993, Journal ofManagement Accounting Research, Fall, 151.

Anderson, S. W., 1995b. Measuring the impact of product mix heterogeneity on manufacturingoverhead cost, The Accounting Review (3rd Qtr) 363388.

Anderson, S. W. and Young, S. M., 1999. The impact of contextual and process factors onthe evaluation of activity-based costing systems, Accounting, Organizations and Society, 24,525559.


Anderson, S. W. and Sedatole, K., 1998. Designing quality into products: the use ofaccounting data in new product development, Accounting Horizons, September, 12, 213233.

Bagozzi, R. P., Yi, Y. and Phillips, L. W., 1991. Assessing construct validity in organizationalresearch, Administrative Science Quarterly, 36, 421458.

Banker, R. D. and Datar, S. M., 1987. Accounting for Labor Productivity in Manufac-turing Operations: an Application, in W. Bruns, R. S. Kaplan (eds), Accounting andManagementField Study Perspectives, Boston, MA, Harvard Business School Press, 169203.

Banker, R. D. and Hughes, J., 1991. Activity-Based Costing and Equilibrium Pricing, Working Pa-per, University of Minnesota.

Banker, R. D. and Potter, G., 1991. Economic Comparisons of Single Cost Driver and ABC Systems,Working Paper, University of Minnesota.

Banker, R. D., Potter, G. and Schroeder, R. G., 1995. An empirical analysis of manufacturingoverhead cost drivers, Journal of Accounting and Economics, 19, 115137.

Barnes, F. C., 1991. IES can improve management decisions using activity-based costing,Industrial Engineering, September, 4450.

Bentler, M., 1983. Some contributions of efficient statistics in structural models: specificationand estimation of moment structures, Psychometrica, 493517.

Bjornenak, T., 1997. Diffusion and accounting: the case of ABC in Norway, ManagementAccounting Research, 8, 317.

Bjornenak, T. and Mitchell, F., 1999. A Study of the Development of the Activity-Based Cost-ing journal Literature 19871998, Working Paper, University of Pittsburgh.

Brimson, J. A., 1991. Activity Costing, an Activity-Based Costing Approach, New York, Wiley.Bromwich, M. and Hong, C., 1999. Activity-based costing systems and incremental costs,Management Accounting Research, 10, 3960.

Bruns,W. J. and Kaplan, R. S., 1987.Accounting andManagement Field Study Perspectives, Boston,MA, Harvard Business School Press.

Chenhall, R. H., 1997. Reliance on manufacturing performance measures, total quality mana-gement and organizational performance,Management Accounting Research, June, 8, 187206.

Chenhall, R. H. and Langfield-Smith, K., 1998. The relationship between strategic priori-ties, management techniques and management accounting: an empirical investigation usinga systems approach, Accounting, Organizations and Society, 23, 243264.

Cooper, R., 1988. The rise of activity-based costingpart two: when do I need an activity-basedsystem? Journal of Cost Management, Fall, 4148.

Cooper, R., 1989. You need a new cost system when. . . . Harvard Business Review,January/February, 7782.

Cooper, R., 1990. Cost classification in unit-based and activity-based manufacturing costsystems, Journal of Cost Management, Fall, 414.

Cooper, R., 1996. Activity-based costing and the lean enterprise, Journal of Cost Management,Winter, 614.

Cooper, R. and Kaplan, R. S., 1988. Measure costs right: make the right decisions, HarvardBusiness Review, September/October, 96105.

Cooper, R. and Kaplan, R. S., 1991. The Design of Cost Management Systems: Text, Cases,and Readings, Englewood Cliffs, NJ, Prentice-Hall.

Cooper, R. and Turney, P. B., 1989. Hewlett-Packard: the Roseville Network Division, Boston, MA,Harvard Business School, (Case No. 9-189-117).

Cronbach, L., 1951. Coefficient alpha and the internal structure of tests, Psychometrica, 16,297334.

Dillman, D. A., 1999.Mail and Telephone Surveys: the Total DesignMethod, 2nd edition, NewYork,Wiley.

Dixon, J. M., 1996. Total Quality Management in ISO-9000 Registered Organizations: an Em-pirical Examination of the Critical Characteristics Associated with Levels of Financial Per-formance, Dissertation, Florida State University.


Drake, A. R., Haka, S. F. and Ravenscroft, S. P., 1999. Cost System and Incentive StructureEffects on Innovation, Efficiency and Profitability in Teams.

Estrin, T. L., Kantor, J. and Albers, D., 1994. Is ABC suitable for your company? ManagementAccounting, April, 4045.

Evans, H. and Ashworth, G., 1995. Activity-based management: moving beyond adolescence,Management AccountingLondon, December, 2630.

Feinberg, S., 1983. The Analysis of Cross-Classified Categorical Data, Cambridge, MA, MIT Press.Foster, G. and Gupta, M., 1990. Manufacturing overhead cost driver analysis, Journal ofAccounting and Economics, 12, 309337.

Foster, G. and Swenson, D. W., 1997. Measuring the success of activity-based cost manage-ment and its determinants, Journal of Management Accounting Research, 9, 109141.

Foster, G. and Young, S. M., 1997. Frontiers of management accounting research, Journal ofManagement Accounting Research, 9, 6377.

Gatewood, R. D. and Field, H. S., 1990. Human Resource Selection, 2nd edition, Chicago, IL,Dryden.

Geishecker, M. L., 1996. New Technologies Support ABC, Management Accounting, March,4248.

Georgantzas, N. C. and Shapiro, H. J., 1993. Viable theoretical forms of synchronous productioninnovation, Journal of Operations Management, 11, 161183.

Gonzalves, F. A. and Eiler, R. G., 1996. Managing complexity through performance measure-ment,Management Accounting, August, 35.

Gordon, L. A. and Silvester, K. J., 1999. Stock market reactions to activity-based costing adop-tions, Journal of Accounting and Public Policy, September, 18, 229251.

Gosselin, M., 1997a. The effect of strategy and organizational structure on the adoption andimplementation of activity-based costing, Accounting, Organizations and Society, 22, 105122.

Gosselin, M., 1997b. Bandwagon Theories: Some Explanations for the Activity-Based CostingParadox, paper presented at the EIASM Workshop on Manufacturing Accounting, Edin-burgh, 57 June.

Govindarajan, V., 1988. A contingency approach to strategy implementation at the business-unit level integrating administrative mechanisms with strategy, Academy of ManagementJournal, 31, 828853.

Govindarajan, V. and Fisher, J., 1990. Strategy, control system, and resource sharing: effects onbusiness-unit performance, Academy of Management Journal, 33, 259285.

Grandzol, J. R. and Gershon, M., 1997. Which TQM practices really matter: an empirical inves-tigation, Quality Management Journal, 4, 4359.

Granlund, M., 1997. The Challenge of Management Accounting Change. Doctoral Manuscript,Turku School of Economics and Business Administration.

Granlund, M. and Lukka, K., 1998. Its a small world of management accounting practices,Journal of Management Accounting Research, 10, 151179.

Guadagnoli, E. and Velicer, W. F., 1998. Relation of sample size to the stability of compo-nent patterns, Psychological Bulletin, 103, 265275.

Harris, E., 1990. The impact of JIT production on product costing information systems,Production and Inventory Management Journal (1st Qtr) 4448.

Hicks, D. T., 1999. Yes, ABC is for small business too, Journal of Accountancy,August, 188, 4143.Hilton, R. W., 1994.Managerial Accounting, 2nd edition, New York, McGraw-Hill.Hollander, M. and Wolfe, D., 1973. Nonparametric Statistical Methods, New York, CambridgeUniversity Press.

Husan, M. and Nanda, D., 1995. The impact of just-in-time manufacturing on firm perfor-mance, Journal of Operations Management, January, 514.

Innes, J., Mitchell, F. and Sinclair, D., 2000. Activity-based costing in the U.K.s largestcompanies: a comparison of 1994 and 1999 survey results, Management Accounting Research,11, 349362.


Innes, J. and Mitchell, F., 1995. A survey of activity-based costing in the U.K.s largest compa-nies,Management Accounting Research, June, 137153.

Institute of Internal Auditors, 1990. Statement of Responsibilities of Internal Auditing, AltamonteSprings, FL, Institute of Internal Auditors.

Institute of Internal Auditors, 1998. Standards for the Professional Practice of Internal Auditing,Altamonte Springs, FL, Institute of Internal Auditors.

Institute of Internal Auditors, 2001. www.TheIIA.org/ecm/iiacni.cfm?doc_id=256.Jaccard, J. and Wan, C. K., 1996. LISREL Approaches to Interaction Effects in Multiple Regression,Thousand Oaks, CA, Sage.

Jacobson, R., 1987. The validity of ROI as a measure of business performance, The AmericanEconomic Review, 77, 470478.

Johnson, E. N., Walker, K. B. and Westergaard, E., 1995. Supplier concentration and pricing ofaudit services in New Zealand, Auditing: a Journal of Practice and Theory, Fall, 7489.

Jones, L. F., 1991. Product costing at Caterpillar,Management Accounting, February, 3442.Joreskog, K. G. and Sorbom, D., 1989. LISREL 7: a Guide to the Program and Application, 2ndedition, Chicago, IL, Scientific Software.

Joreskog, K. G. and Sorbom, D., 1993. LISREL 8: Structural Equation Modeling with theSIMPLIS Command Language, Chicago, IL, Scientific Software.

Joreskog, K. G. and Yang, F., 1996. Non-Linear Structural Equation Models: the Kenny-Judd Model with Interaction Effects, in G. Marcoulides, R. Schumacker (eds), AdvancedStructural Equation Modeling, Hillside, NJ, Erlbaum, 5788.

Kaplan, R. S., 1993. Research opportunities in management accounting, Journal of ManagementAccounting Research, Fall, 114.

Karmarkar, U. S., Lederer, P. J. and Zimmerman, J. L., 1990. Choosing ManufacturingProduction Control and Cost Accounting Systems, in R. S. Kaplan (ed), Measures ofManufacturing Excellence.

Kaynak, H., 1996. The Relationship between Just-in-Time Purchasing and Total QualityManagement and Their Effects on the Performance of Firms Operating in the U.S.: anEmpirical Investigation, Dissertation, University of North Texas.

Kennedy, T. and Bull, R., 2000. The great debate,Management Accounting,May, 78, 3233.Kern, B. B. and Morris, M. H., 1994. Differences in the COMPUSTAT and expanded valueline data bases and the potential impact on empirical research, The Accounting Review, 69,274284.

Khandwalla, P. N., 1972. The effect of different types of competition on the use of managementcontrols, Journal of Accounting Research, Autumn, 10, 275285.

Krumwiede, K. R., 1996. An Empirical Examination of Factors Affecting the Adoption andInfusion of Activity-Based Costing, Dissertation, University of Tennessee.

Krumwiede, K. R., 1998. The implementation stages of activity-based costing and the impactof contextual and organizational factors, Journal of Management Accounting Research, 10,239277.

Kwon, T. H. and Zmud, R. W., 1987. Unifying the Fragmented Models of InformationSystems and Implementation, in R. J. Boland, R. Hirscheim (eds), Critical Issues in InformationSystems Research, New York, Wiley.

Libby, R. and Tan, H., 1994. Modeling the determinants of audit expertise, Accounting.Organizations and Society, 19, 701716.

Maher, M.W. andMarais, M. L., 1998. A field study on the limitations of activity-based costingwhen resources are provided on a joint and indivisible basis, Journal of Accounting Research,Spring, 36, 129142.

Malmi, T., 1999. Activity-based costing diffusion across organizations: an exploratory empiri-cal analysis of Finnish, Accounting, Organizations and Society, November, 24, 649672.

Malmi, T., 1997. Towards explaining activity-based costing failure: accounting and control, ina decentralised organisation,Management Accounting Research, 8, 459480.


McGowan, A. S., 1998. Perceived benefits of ABCM implementation, Accounting Horizons,March, 3150.

McGowan, A. S. and Klammer, T. P., 1997. Satisfaction with activity-based cost manage-ment implementation, Journal of Management Accounting Research, 9, 217237.

Mia, L. and Clarke, B., 1999. Market competition, management accounting systems and busi-ness unit performance,Management Accounting Research, June, 10, 137158.

Miller, J. G. and Roth, A. V., 1994. A taxonomy of manufacturing strategies, ManagementScience, 40, 285304.

Nunnaly, J., 1978. Psyc

the association between activity based costing and improvement in financial performance 2002...

Documents