stock market performance than greater measurement on an absolute scale. Finally, the balanced scorecard process, economic value measurement, and causal business modeling are associated with higher measurement system satisfaction, but exhibit almost no association with economic performance.<br><br> # 2003 Elsevier Ltd. All rights reserved. Introduction Managerial accounting is evolving to encompass a more strategic approach that emphasizes the identi 0cation, measurement, and management of the key 0nancial and non- 0nancial drivers of stra- tegic success and shareholder value ( Institute of Management Accountants, 1999; International Federation of Accountants, 1998 ).<br><br> In response, many 0rms are adopting strategic performance measurement (SPM) systems that (1) provide information that allows the 0rm to identify the strategies o*ering the highest potential for achiev- ingthe 0rm 9s objectives, and (2) align management processes, such as target setting, decision-making, and performance evaluation, with the achievement of the chosen strategic objectives (e.g., Gates, 1999; Otley, 1999 ). Proponentsofstrategicperformancemeasurement advocate two general approaches for developing SPM systems. The simplest approach calls for 0rms to measure and use a diverse set of 0nancial and non- 0nancial measures.<br><br> Advocates of this 8 8measurement diversity 9 9 approach argue that a 0361-3682/03/$ - see front matter # 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0361-3682(03)00033-3 Accounting, Organizations and Society 28 (2003) 715 3741 www.elsevier.com/locate/aos * Corresponding author.<br><br> Tel.: +1-215-898-7786; fax: +1- 215-573-2054. E-mail address: ittner@wharton.upenn.edu (C.D. Ittner).<br><br> broad set of measures keeps managers from sub- optimizing by ignoring relevant performance dimensions or improving one measure at the expense of others. As a result, these advocates claim that 0rms achieve higher performance when they place greater emphasis on a broad set of 0nancial and non- 0nancial performance measures (e.g., Lingle & Schiemann, 1996 ). A second approach is based on contingency theory, which argues that strategic performance measures must be aligned with the 0rm 9s strategy and/or value drivers ( Fisher, 1995b; Lang 0eld-Smith, 1997 ).<br><br> Under this approach, performance theoretically is enhanced when 8 8measurement gaps 9 9 between the 0rm 9s strategicprioritiesandmeasurementpractices are minimized. Thus, performance is expected to be lower when the SPM system places either less or more emphasis on a measurement practice than the level required by the 0rm 9s strategy and value drivers. Closely related to the contingency perspective is the use of measurement techniques such as the balanced scorecard process, causal business model- ing, and economic value measurement.<br><br> Advocates argue that these techniques help companies improve the alignment between their performance measurement systems and their organizational objectives ( Gates, 1999; Kaplan & Norton, 1992, 1996, 2001; Stewart, 1991; Young & O 9Byrne, 2001 ). Despite these arguments, the extent to which 0rms claiming to use these techniques actu- ally link their performance measures more closely to strategic priorities is unknown. Using data from 140 US 0nancial services 0rms, we extend prior research on the performance implications of strategic performance measure- ment along several dimensions.<br><br> First, we examine a broader set of measurement system uses (goal setting, capital investment decisions, identi 0cation of improvement opportunities and development of action plans, performance evaluation, and external disclosure) and measurement capabilities than prior studies that typically focus only on perfor- mance evaluation and compensation. Second, we investigate the relation between SPM practices and actual 0nancial outcomes (accounting and stock returns) rather than relying exclusively on self-reported measurement satisfaction or 0rm performance. Third, we examine each of the SPM approaches and compare their relative ability to explain 0rm performance.<br><br> Fourth, we extend prior contingency research by looking at the alignment between speci 0c value drivers and measurement, in addition to the traditional alignment with 0rm or manufacturing strategy. Fifth, we provide evi- dence on the use and performance consequences of the three measurement alignment techniques (balanced scorecard, economic value measure- ment, and business modeling), an area that has received surprisingly little attention in the research literature. Finally, we examine potential lags between the implementation of performance measurement systems and economic results.<br><br> We 0nd consistent evidence that SPM practices are associated with 1- and 3-year stock returns, but not with our two accounting measures (return on assets and sales growth). In particular, 0nancial services 0rms that make more extensive use of a broad set of 0nancial and (particularly) non- 0nancial measures than those with similar strategies or value drivers earn higher stock returns. These results are even stronger in the subsample of 0rms with more mature performance measurement sys- tems, suggesting that these measurement practices yield economic results with some lag.<br><br> We 0nd little support for the hypothesis that more or less extensive measurement than predicted by the 0rm 9s strategy or value drivers adversely a*ect performance. Instead, our results indicate that greater measurement emphasis and diversity than predicted by our benchmark model is asso- ciated with higher satisfaction and stock market performance. These 0ndings suggest that the average measurement practices of 0rms pursuing similar strategies or value drivers currently are not optimal in this industry.<br><br> We also 0nd that greater measurement diversity compared with 0rms with similar strategies or value drivers has a stronger relationship with stock market performance than greater overall measurement. This evidence suggests that the appropriate benchmark for assessing measurement diversity is greater measurement relative to com- petitors with similar strategies or value drivers rather than greater measurement on an absolute scale. 716 C.D.<br><br> Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 On average, 0rms claiming to use a balanced scorecard exhibit few di*erences in their emphasis on non- 0nancial performance categories than non- users, and make little use of the causal 8 8business models 9 9 of leading and lagging indicators that balanced scorecardadvocatesclaim isafoundation of the scorecard process. In contrast, economic value and business model users place signi 0cantly greater emphasis on non- 0nancial value drivers and measures than 0rms that do not use these practices. Although balanced scorecard, economic value, and causal business model users all rate satisfaction with their measurement systems higher than non-users, we 0nd almost no evidence that these techniques are associated with accounting or stock market performance.<br><br> The remainder of the paper is organized as fol- lows. The following section reviews related litera- ture and develops our hypotheses. We then discuss our sample selection and measurement methods.<br><br> The next section reports our contemporaneous performance results, followed by our analysis of lagged performance e*ects in the subset of more mature systems. A summary of our results and limitations to our study conclude. Literature review and hypotheses A study by the Conference Board de 0nes strate- gic performance measurement (SPM) as a system that 8 8translates business strategies into deliverable results.<br><br> SPM systems combine 0nancial, strategic, and operating business measures to gauge how well a company meets its targets 9 9 ( Gates, 1999 , p. 4). Case studies by Gates (1999) and Ruddle and Feeny (2000) 0nd that 0rms have adopted two general approaches to designing SPM systems.<br><br> In the following sections, we review each of these approaches and develop hypotheses regarding their performance implications. Greater 8 8Measurement Diversity 9 9 One approach to strategic performance measurementis supplementing traditional 0nancial measures with a diverse mix of non- 0nancial mea- sures that are expected to capture key strategic performance dimensions that are not accurately reKected in short-term accounting measures. Brancato (1995) and Fisher (1995a) indicate that many 0rms believe that 0nancial measures are too historical and 8 8backward-looking, 9 9 lack pre- dictive ability to explain future performance, reward short-term or incorrect behavior, provide little information on root causes or solutions to problems, and give inadequate consideration to diLcult to quantify 8 8intangible 9 9 assets such as intellectual capital.<br><br> As a result, many 0rms are supplementing 0nancial metrics with a diverse set of non- 0nancial performance measures that are believed to provide better information on strategic progress and success. 1 Early writings on the balanced scorecard pro- vided considerable impetus behind this 8 8measure- ment diversity 9 9 approach to SPM. Kaplan and Norton 9s (1992) original balanced scorecard arti- cle argues that 0rms should supplement 0nancial measures with non- 0nancial measures focused on three other perspectives: customers, internal busi- ness processes, and learning and growth.<br><br> Although the authors note that the speci:c mea- sures within these categories should be tailored to the 0rm 9s strategy, they also claim that these broad perspectives are common across (and should be used with) all strategic choices. Other writers have extended these categories by adding additional perspectives focused on employees, partners and suppliers, and the environment (e.g., Edvinsson & Malone, 1997; Schiemann & Lingle, 1999 ). An assumption in these classi 0cation schemes is that measurement across all these cate- gories is necessary regardless of the 0rm 9s strategy.<br><br> Schiemann and Lingle (1999 , p. 8), for example, write: For us, measuring the 8 8right things 9 9 entails measuring results in the six performance areas 1 Consistentwiththeseclaims,anumberofaccountingstudies provide evidence that non- 0nancial measures can be leading indicators of 0nancial performance (e.g., Banker, Potter, & Sri- nivasan,2000; Behn&Riley,1999;Ittner&Larcker,1997;Ittner, Larcker, & Meyer, in press; Nagar & Rajan, 2001 ). However, only Banker et al.<br><br> (2000) and Ittner et al. (in press) examine the use of these measures for performance evaluation or decision- making purposes. C.D.<br><br> Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 717 that are key to strategic success. And when we use the term 8 8strategic measurement, 9 9 we mean measurement focused on these six per- spectives or areas of performance. Several empirical studies implicitly or explicitly draw on the measurement diversity approach in their tests.<br><br> A widely-cited practitioner-oriented study by Lingle and Schiemann (1996) reports that 8 8measurement-managed 9 9 0rms (de 0ned as those in which management updates and reviews semi- annual performance measures in three or more of their sixprimary performancecategories,and where senior management reports being in agreement on measurable criteria for determining strategic suc- cess) achieve statistically higher self-reported industry standing, 0nancial performance relative to competitors, and progress in managing change e*orts than 0rms that are not 8 8measurement- managed. 2 Scott and Tiessen 9s (1999) academic study indicates that work teams having more diverse performance measures (i.e., both 0nancial and non- 0nancial measures or more categories of measures) achieve higher self-assessed perfor- mance (relative to expectations). Hoque and James (2000) also 0nd a signi 0cant positive rela- tion between perceived organizational perfor- mance and the use of a diverse set of performance measures related to the four balanced scorecard categories.<br><br> However, all of these studies place heavyreliance onperceptualresultsindicatorsand/ or simple univariate tests, making it diLcult to place substantive interpretations on their results. Theoretical research on the economic bene 0ts from greater measurement diversity is ambiguous. Nearly all of this work focuses on reward systems.<br><br> Holmstrom (1979) and Banker and Datar (1989) show that more (costless) measures are preferred to feweriftheadditionalmeasuresprovideincremental information on dimensions of the agent 9s actions that the principal wishes to motivate. In addition, Feltham and Xie (1994), Hauser, Siemester, and Wernerfelt (1994), and Hemmer (1996) demon- strate how incentives based on non- 0nancial mea- sures can improve contracting by incorporating information on managerial actions that is not fully captured in contemporaneous 0nancial results. However, these theories also imply that a performance category (e.g., the four balanced scorecard perspectives) may not improve incentive contracting if it provides no incremental infor- mation on the manager 9s action, imposes too much risk on the agent, or is too costly to measure, indicating that the measurement diver- sity approach to SPM may have no impact on economic performance.<br><br> 3 The limited and mixed evidence on the perfor- mance implications of this SPM approach leads us to examine our 0rst hypothesis: Hypothesis 1. Organizational performance is posi- tively associated with the extent to which the 0rm measures and uses information related to a diverse set of 0nancial and non- 0nancial performance measures. Alignment with strategy and value drivers A second general approach emphasizes the implementation of performance measurement sys- tems that are more closely linked to the 0rm 9s speci:c strategy and value drivers.<br><br> Contingency theory has long held that control systems must be aligned with organizational characteristics such as 0rm strategy (see Fisher, 1995b for a review). Similarly, economic theories contend that the optimal design of a 0rm 9s information and reward systems is a function of the 0rm 9s business strategy (e.g., Brickley, Smith, & Zimmerman, 1997; Mil- grom & Roberts, 1992 ). More recently, advocates of this SPM approach have extended these theories to argue that a key 2 Although Lingle and Schiemann (1996) provide evidence on the extent of 8 8measurement gaps 9 9 in their sample, they do not use these gaps in their performance tests, instead relying on the level of measurement in their performance categories.<br><br> 3 Other researchers argue that greater measurement diver- sity can have a negative e*ect on performance by making the systems too complex and diLcult to understand, promoting information overload, spreading agents 9 e*orts over too many objectives, reducing motivation by including multiple goals that are inconsistent in the short-term, and increasing adminis- trative costs relative to simpler systems (e.g., Heneman, Led- ford, & Gresham, 1999; Schick, Gordon, & Haka, 1990 ). 718 C.D. Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 element in managing the links between strategy and performance is identifying and measuring the speci 0c factors, or 8 8value drivers, 9 9 that actually lead to strategic success or 0rm value ( Ittner & Larcker, 2001 ).<br><br> By linking strategies to their underlying value drivers, and tying information systems, goals and objectives, resource allocation, and performance evaluation to these drivers, SPM systems are expected to improve communication of the speci:c actions required to achieve the cho- sen strategy, motivate performance against stra- tegic value driver goals, and provide more rapid feedback on whether the strategy is achieving its objectives. In addition, the SPM literature increasingly argues that the value driver analysis should not only inKuence the design and use of measurement systems, but should also a*ect external disclosure requirements (e.g., Black, Wright, Bachman, Makall, & Wright, 1998; Eccles, Herz, Keegan, & Phillips, 2001; Gates, 1999 ). This use of SPM systems is consistent with calls in the 0nancial accounting community for greater disclosure of information on key value drivers (e.g., American Institute of Certi- 0ed Public Accountants, 1994; Financial Accounting Standards Board, 2001; Wallman, 1995 ).<br><br> Academic research on contingency approaches to strategic performance measurement falls into three categories. Early studies focused on the inKuence of perceived environmental uncertainty (PEU) on management accounting systems. According to this literature, PEU relates to the extent to which the 0rm 9s competitive environ- ment is highly dynamic and unpredictable, factors that are likely to be highly correlated with the extent to which the 0rm 9s strategy is focused on innovation and growth ( Dent, 1990; Lang 0eld- Smith, 1997 ).<br><br> Larcker (1981) and Gordon and Nar- ayanan (1984) , for example, examine the relation between PEU and three performance measurement system attributes: focus (internal vs. external mea- sures), quanti 0cation ( 0nancial vs. non- 0nancial measures), and time horizon (historical vs.<br><br> future- oriented), with mixed results. Larcker 9s (1981) study of strategic capital budgeting decisions 0nds no association between environmental characteristics and variations in the perceived importance of these measurement attributes. In contrast, Gordon and Narayanan (1984) 0nd a signi 0cant positive association between PEU and the perceived importance of externally-oriented, non- 0nancial, and ex ante information.<br><br> Chenhall and Morris (1986) examine the association between PEU and the perceived usefulness of four management accounting system attributes: scope (i.e., the external, non- 0nancial, and future-orien- ted attributes examined in the two previous stud- ies), timeliness, aggregation (i.e., the level of aggregation by time period and functional area and the use of analytical or decision models), and integration (i.e., the setting of precise tar- gets for activities and their interdependencies and the reporting of intra-sub-unit-interactions). They 0nd signi 0cant positive associations between PEU and preferences for broader scope and more timely information, but no associ- ations with aggregation and integration pre- ferences. None of these studies investigates performance consequences.<br><br> More recent studies directly examine the e*ects of organizational strategy on performance measurement choices, and the relation between these choices and organizational performance. These studies typically measure strategy as a con- tinuum between 0rms following a 8 8defender, 9 9 8 8harvest, 9 9 or 8 8cost leadership 9 9 strategy and 0rms following a 8 8prospector, 9 9 8 8build, 9 9 or innovation 9 9 strategy. The majority of these studies 0nd sig- ni 0cantrelationsbetweentheorganization 9sstrategy and performance measurement system, with a smaller set of studies also 0nding higher organiza- tional performance when measurement is more closely aligned with the chosen strategy.<br><br> Simons (1987) and Govindarajan (1988) , for example, 0nd higher performance in organizations following defender or low cost strategies when bonuses are based on budget targets. Govindarajan and Gupta (1985) also 0nd that greater reliance on non- 0nancial compensation criteria has a stronger positive impact in organizations following a build strategy than in those following a harvest strategy. Studies by Abernethy and Guthrie (1994), Chong and Chong (1997), and Bouwens and Abernethy (2000) , among others, generally support the hypothesis that broad scope performance C.D.<br><br> Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 719 measurement systems are associated with higher (self-reported) performance in companies follow- ing prospector or di*erentiation strategies. A third set of studies provides evidence on the associations between speci 0c manufacturing stra- tegies or value drivers (such as quality and Kex- ibility), the choice of performance measures, and manufacturing performance. This research 0nds systematic links among these choices, with an organization 9s emphasis on inventory reduction or just-in-time production, quality, or manufacturing Kexibility positively associated with the provision of non- 0nancial measures and goals and greater emphasis on non- 0nancial measures in reward systems.<br><br> However, empirical support for the hypothesized performance bene 0ts from these measurement practices is mixed, with some studies 0nding positive performance e*ects when manu- facturing strategies are aligned with measurement systems (e.g., Abernethy & Lillis, 1995 ), others 0nding mixed results depending upon the type of performance measurement attribute being exam- ined or the extent to which other manufacturing improvement practices are implemented (e.g., Itt- ner and Larcker, 1995; Sim and Killough, 1998 ), and others 0nding no signi 0cant associations (e.g., Pererra, Harrison, & Poole, 1997; Young and Selto, 1991 ). Although the contingency-based performance measurement literature is relatively extensive, it generally looks at only one or a few strategies or value drivers at a time, does not compare results across various SPM approaches (e.g., measure- ment diversity vs. strategy alignment vs.<br><br> value driver alignment), and relies quite heavily on respondents 9 self-reported performance. As a result, we extend these studies by providing fur- ther evidence on the following hypotheses: Hypothesis 2. Organizational performance is posi- tively associated with the extent to which perfor- mance measurement practices are aligned with the 0rm 9s strategy.<br><br> Hypothesis 2. Organizational performance is posi- tively associated with the extent to which the per- formance measurement practices are aligned with the 0rm 9s value drivers. Measurement alignment techniques The performance measurement literature pro- poses several techniques that are claimed to improve the alignment between performance measurement systems and the 0rm 9s organiza- tional objectives.<br><br> These techniques include the balanced scorecard process, economic value measurement, and causal business models. Kaplan and Norton 9s (1996, 2001) recent writings on the balanced scorecard, for example, de 0ne this pro- cess as a method for using 0nancial and non- 0nancial measures to communicate the multiple, linked objectives that a 0rm must achieve to satisfy its mission and reach its long-term strategic goals. According to these authors, balanced scorecard systems improve performance by translating strategy into speci 0c objectives and measures that are linked in a causal chain of leading and lagging indicators covering the four scorecard perspectives ( 0nancial, customer, internal business process, and learning and growth).<br><br> Economic value-based SPM systems focus on the creation of long-term shareholder value through the use of residual income or cashKow- related measures. Advocates of economic value techniques (often referred to as value-based man- agement approaches) argue that performance measurement systems should be aligned with the 0rm 9s ultimate organizational objective: improved economic performance. Stern, Stewart, and Chew (1995) , for example, maintain that e*ective performance measurement requires the 0rm to make economic value measures the cor- nerstone of a total management system that focuses on shareholder value enhancement for capital budgeting, goal setting, investor communication, and compensation.<br><br> The result- ing performance measurement system is expec- ted to improve alignment between performance measures and strategic objectives by requiring 0rms to choose internal objectives that lead to shareholder value enhancement, select strategies to achieve these objectives, identify the value dri- vers that actually create value for the 0rm, and select action plans, lower-level performance mea- sures, and targets based on the priorities identi- 0ed in the value driver analysis ( Copeland, 720 C.D. Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 Koller, & Murrin, 1996; Dixon and Hedley, 1997; Stern et al., 1995 ). 4 A third alignment technique that is stressed in both the balanced scorecard and economic value literatures is the formal development of causal 8 8business models. 9 9 Eccles (1991) , for example, argues that e*ective performance measurement system design requires 0rms to 0rst understand the causal model linking key success factors to the ultimate objectives of the 0rm.<br><br> Kaplan and Nor- ton (1996, 2001) now state that a critical element of the balanced scorecard process is the develop- ment of 8 8strategic maps 9 9 that embed strategy in a system of cause-and-e*ect relations that connects desired strategic outcomes with the drivers expec- ted to lead to these outcomes. Similarly, Dixon and Hedley (1997), Copeland et al. (1996), and Young and O 9Byrne (2001) stress the importance of linking 0nancial performance measures and their non- 0nancial value drivers to achieve the bene 0ts from economic value measurement pro- grams and promote value-creating behavior in the 0rm.<br><br> 5 Despite this emphasis on causal business models, Gates 9 (1999) study indicates that most 0rms adopt SPM systems without articulating these causal models or maps. Surprisingly little research has been conducted on the performance implications of the proposed measurement alignment techniques. Surveys on the use of balanced scorecards ( Banker, Janakiraman, & Konstans, 2001a; Chenhall & Lang 0eld-Smith, 1998; Kaplan & Norton, 2001; Rigby, 2001; Sandt, Schae*er, & Weber, 2001; Towers Perrin, 1996 , p.<br><br> 357) and economic value measures ( Ittner & Larcker, 1998 ) typically 0nd moderately greater satisfaction with or perceived performance from these techniques than from other performance measurement practices. In contrast, Ittner et al. 9s (in press) examination of a balanced scorecard bonus plan in a large bank indicates that the scorecard plan was deemed unsuccessful, leading it to be abandoned in favor of a revenue-based incentive plan. Banker, Janakiraman, Konstans, and Pizzini (2001b) and Sandt et al.<br><br> (2001) also 0nd greater satisfaction with performance measurement systems when systematic linkages among performance measures are understood and articulated. However, none of these studies directly examines the association between these techniques and actual 0rm performance. Wallace(1997)andHoganandLewis(1999) reach conKictingconclusionsregardingtheperformanceof 0rms adopting compensation plans based on eco- nomic value measures.<br><br> Wallace (1997) 0nds eco- nomic value adopters, relativeto a controlsample of non-adopters, decrease new investments, increase payouts to shareholders through share repurchases, and utilize assets more intensively, leading to sig- ni 0cantly greater change in residual income. In contrast, HoganandLewis(1999) 0ndnosigni 0cant di*erence in economic value users and non-users after matching control 0rms on past performance to control for mean reversion in performance levels. The claimed bene 0ts from the measurement align- ment techniques, together with the limited evidence to support these claims, lead to our 0nal hypothesis: Hypothesis 4.<br><br> Organizational performance is posi- tively associated with the use of balanced scor- ecards, economic value measures, and causal business models. Data and variables Sample We test these hypotheses using a sample of US 0nancial services 0rms that responded to a survey conducted by the authors in conjunction with the Cap Gemini Ernst & Young Center for Business 4 Forty-one percent of the 0rms in Gates 9 (1999) study describe their SPM systems as value-based, while 40 percent describe their systems as following a balanced scorecard approach. Although many 0rms view their economic value measurement systems as strategic, others argue that these sys- tems do not actually represent SPM systems because they focus on a single outcome (shareholder value) rather than on the strategies used to achieve this outcome.<br><br> In addition, some crit- ics of economic value measures charge that these metrics drive managers to focus on short-term operational and 0nancial performance, to the detriment of longer-term investments in customers, innovation, and employee capabilities (e.g., Kaplan & Norton, 2001 , pp. 378 3379). We examine these assertions later in the paper.<br><br> 5 See Heskett (1994) for examples of causal business models linking employee measures, customer measures, and 0nancial performance. C.D. Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 721 Innovation.<br><br> In contrast to most prior research, we restrict our sample to a single industry. An important advantage of this choice is that we can implicitly control for the myriad of confounding variables that can substantively impact any results from a multi-industry, cross-sectional study. The 0nancial services industry was selected because our 0eld research on performance measurement innovations indicates that 0nancial service 0rms are actively debating their choice of value drivers and performance measures.<br><br> Although restricting the sample to a single industry limits our ability to generalize the results, we believe that a single industry analysis has substantially higher internal validity than a multi-industry analysis. A random sample of 600 0rms was solicited to participate in the survey. A marketing research 0rm telephoned senior executives from each of these 0rms to request participation.<br><br> Those agree- ing to participate were sent a survey or guided to a web site containing the questionnaire. Executives from 140 0rms (23.3%) completed usable surveys during November of 1999. The respondents represent a variety of 0nancial service sectors, including regional banks (33.3% of the sample), insurance companies (21.4%), diversi 0ed 0nancial 0rms (17.1%), savings and loans (14.3%), money center banks (5.7%), and others (e.g., consumer 0nance, investment banking, etc.) (7.1%).<br><br> Relative to the population of 0nancial services 0rms on Compustat, the resulting sample contains a larger percentage of diversi 0ed 0nancial 0rms, life insur- ance companies, and banks, and a smaller percen- tage of investment banks and savings and loans. Strategy We asked respondents to evaluate 12 aspects of the company 9s organizational strategy and corpo- rate environment that are commonly used to measure strategy and perceived environmental uncertainty. Principal components analysis of these questions (with oblique rotation) reveals three factors with eigenvalues greater than one.<br><br> The factors capture the extent to which the 0rm 9s strategy focuses on (1) innovation , (2) =exibility in changing its product and service o*erings and responding to market demands, and (3) maintaining current relationships and product o*erings by pursuing existing customers and markets in stable environments. Thethreestrategyconstructsrepresenttheaverage standardized response to each question loading greater than 0.40 on these factors. 6 Flexible equals the average answer to four questions asking the respondent 9s agreement with the statements, 8 8We respondrapidlytoearlysignalsofopportunityinour market, 9 9 8 8We have greater Kexibility to respond to changes in our environment than our competitors, 9 9 8 8We have the ability to adjust capacity within a short period of time, 9 9 and 8 8We have the ability to change product or service o*erings rapidly 9 9 (scales ranging from 1=strongly disagree to 6=strongly agree).<br><br> Innovate equals the average standardized response to four questions asking the respondent 9s agreement with the statements, 8 8We o*er a more expanded range of products and services than our competitors, 9 9 8 8We are 0rst to market with new products or services, 9 9 8 8We respond rapidly to early signals of opportunity in our market, 9 9 and 8 8We expect most of our future growth in pro 0ts to come from our new product and service o*erings. 9 9 Main- tain equals the average standardized response to three questions asking the respondent 9s agreement with the statements, 8 8We are most active in devel- oping the markets we currently serve, rather than entering new markets with our products or ser- vices, 9 9 8 8We operate in markets for our products or services that are highly predictable, 9 9 and 8 8It is easy toforecasthowactionsofcompetitorswilla*ectthe performance of our organization. 9 9 CoeLcient a 9s are 0.66 for Innovate , 0.75 for Flexible , and 0.46 for Maintain . The Pearson (Spearman) correlation between Innovate and Flexible is 0.49 (0.50), between Innovate and Maintain is 0.18 (0.15), and between Flexible and Maintain is 0.17 (0.15). 7 6 One question asking whether the 0rm is more cost eLcient than its competitors did not load greater than 0.40 on any of the factors.<br><br> This question is dropped from the analysis. 7 Despite the signi 0cant correlations, in no case does the Variance InKation Factor (VIF) exceed 2.5 in any of our regression models, suggesting no serious problems with multi- collinearity in our tests. Since one question cross-loaded on the Innovate and Flexible factors, we repeated our analyses after dropping this question from the strategy constructs.<br><br> This change had no substantive impact on our results. 722 C.D. Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 Value drivers Respondents were also asked the extent to which various performance categories are impor- tant drivers of their 0rms 9 long-term organiza- tional success, on a scale ranging from 1 (not at all important or not applicable to their organization) to 6 (extremely important).<br><br> The 10 value driver categories include short-term 0nancial perfor- mance (e.g., annual earnings, return on assets, cost reduction), customer relations (e.g., market share, customersatisfaction,customerretention),employee relations (e.g., employee satisfaction, turnover, workforce capabilities), supplier relations (e.g., on-time delivery, input into product/service design), operational performance (e.g., productiv- ity, safety, cycle time), product and service quality (e.g., defect rates, quality awards), alliances with otherorganizations(e.g.,jointmarketingorproduct design, joint ventures), environmental perfor- mance (e.g., government citations, environmental compliance or certi 0cation), product and service innovation(e.g.,newproductorservicedevelopment success, development cycle time), and community (e.g., public image, community involvement). These 10 categories are drawn from value driver discussions in the balanced scorecard, intangible asset, intellectual capital, and value-based man- agement literatures (e.g., Edvinsson & Malone, 1997; Kaplan & Norton, 1996, 2001; Schiemann & Lingle, 1999 ). Table 1 provides the mean importance score given to each of the value driver categories.<br><br> Short- term 0nancial performance ranks only fourth Table 1 Mean survey responses on the importance of performance categories to long-term organizational success and their use in performa nce measurement and decision-making Importance to long-term success a Extent goals set b Extent measures related to these categories are used for the following purposes: Measurement quality e Problem identi 0cation c Capital investments c Performance evaluation c External disclosure d Performance category: f Short-term 0nancial results 4.572 5.369 5.190 4.581 5.138 5.428 5.465 Customer relations 5.511 4.285 4.378 3.917 4.236 2.722 3.724 Employee relations 3.862 2.853 2.775 2.109 2.913 1.871 2.465 Operational performance 5.020 4.600 4.573 4.431 4.747 3.889 4.453 Quality 5.031 4.018 3.992 3.876 3.723 2.941 3.588 Alliances 3.060 2.223 2.204 2.122 1.938 2.262 1.979 Supplier relations 2.875 2.051 2.063 2.150 1.904 1.664 2.001 Environmental performance 2.079 1.740 1.594 1.695 1.485 1.572 1.634 Innovation 4.114 3.407 3.294 3.598 2.941 3.005 2.706 Community 4.066 3.492 3.053 2.641 2.814 3.127 2.804 a Scale: 1=measure not applicable or not at all important to 6=extremely important. b Scale: 1=measure not applicable or no goals established to 6=explicit goals established. c Scale: 1=measure not applicable or not used at all to 6=used extensively.<br><br> d Scale: 1=measure not applicable or no external disclosure to 6=complete external disclosure. e Scale: 1=measure not applicable or extremely poor quality of measurement to 6=high quality of measurement. f Performance categories are de 0ned as short-term 0nancial performance (e.g., annual earnings, return on assets, cost reduction), customer relations (e.g., market share, customer satisfaction, customer retention), employee relations (e.g., employee satisfac tion, turnover, workforce capabilities), supplier relations (e.g., on-time delivery, input into product/service design), operational perfor- mance (e.g., productivity, safety, cycle time), product and service quality (e.g., defect rates, quality awards), alliances wit h other organizations (e.g., joint marketing or product design, joint ventures), environmental performance (e.g., government citations, envir- onmental compliance or certi 0cation), product and service innovation (e.g., new product or service development success, develop ment cycle time), and community (e.g., public image, community involvement).<br><br> C.D. Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 723 most important, behind customer relations, pro- duct and service quality, and operational perfor- mance. Innovation, community relations, and employee relations also receive relatively high importance scores, with environmental perfor- mance and supplier relations believed to be rela- tively unimportant in this industry.<br><br> Spearman correlations among the strategy and value driver measures are shown in Table 2 . Correlations between the strategy constructs and individual value drivers are generally small (mean=0.132, median=0.125). Two value driver categories do not vary signi 0cantly with any of the strategies: short-term 0nancial performance and operational results.<br><br> Innovate is positively corre- lated ( P < 0.10, two-tailed) with all of the remain- ing value drivers except community, with a mean (median) correlation of 0.188 (0.145). Flexibl e is positively correlated with the importance placed on quality, alliances, suppliers, and innovation, but not with the other performance categories. Firms following a Maintain strategy place greater emphasis on community and the environment, but this strategy is not signi 0cantly associated with the other value drivers categories.<br><br> Overall, the rela- tively small correlations in Table 2 suggest that the 0rm 9s chosen strategies are not synonymous with the value drivers used to achieve these strategies. For example, the evidence suggests that some 0rms attempt to implement an Innovate strategy by placing greater reliance on supplier expertise, while others focus on improving innovation by developing their own, internal employee cap- abilities. These results suggest that measuring a 0rm 9s overall strategy without considering its chosen value drivers provides an incomplete representation of strategic attributes.<br><br> Performance measurement practices We asked respondents the extent to which their 0rms use each of the 10 value driver categories for: (1) identifying problems and improvement oppor- tunities and developing action plans, (2) evaluat- ing major capital investments, (3) evaluating managerial performance, and (4) disclosing infor- mation to external parties (e.g., via fact books, analyst meetings, conference calls, press releases, company Internet websites, and one-on-one meet- ings). Scales for these questions range from 1 (not at all important or not used at all) to 6 (used extensively or complete external disclosure). We also asked respondents to rate how well their organizations measure information in these cate- gories (1=extremely poor quality of measurement to 6=high quality of measurement) and the extent Table 2 Spearman correlations among strategy variables and value driver importance scores Flexible Innovate Maintain Financial Customer Employee Operations Quality Alliances Suppliers Environment Innovation Innovate 0.50 *** Maintain 0.15 * 0.16 * Financial 0.03 0.04 0.07 Customer 0.03 0.16 * 0.12 0.13 Employee 0.06 0.25 *** 0.13 À 0.04 0.32 *** Operations 0.06 0.13 0.10 0.12 0.32 *** 0.09 Quality 0.17 ** 0.20 ** 0.12 À 0.11 0.35 *** 0.10 0.40 *** Alliances 0.21 ** 0.28 *** 0.05 À 0.01 0.23 *** 0.17 ** 0.19 ** 0.33 *** Suppliers 0.15 * 0.28 *** 0.03 0.09 0.16 * 0.20 ** 0.11 0.25 *** 0.42 *** Environment 0.11 0.16 * 0.20 ** À 0.12 0.03 0.10 0.13 0.19 ** 0.26 *** 0.45 *** Innovation 0.16 * 0.31 *** 0.01 0.11 0.41 *** 0.13 0.25 *** 0.27 *** 0.31 *** 0.27 *** 0.22 ** Community 0.07 0.07 0.21 ** 0.03 0.44 *** 0.10 0.17 ** 0.30 *** 0.25 *** 0.10 0.10 0.32*** Flexible=the extent to which the 0rm follows a strategy focused on Kexibility; Innovate=the extent to which the 0rm follows an innovation-oriented strategy; Maintain=the extent to which the 0rm follows a strategy focused on maintaining existing relationships in stable and p redictable markets.<br><br> See Table 1 for de 0nitions of the remaining value driver performance categories. * P < 0.10 (2-tailed). ** P < 0.05 (2-tailed).<br><br> *** P < 0.01 (2-tailed). 724 C.D. Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 to which goals are set (1=not applicable or no goals set to 6=explicit goals set).<br><br> 8 Table 1 provides information on the consistency between the perceived importance of the indivi- dual value driver categories and the corresponding use and quality of performance measures related to these categories. With the exception of short- term 0nancial performance, the importance scores for each performance category are lower than the scores provided for the use and quality of related performance measures. These di*erences are con- sistent with the 8 8measurement gaps 9 9 identi 0ed in other studies (e.g., Dixon, Nanni, & Vollmann, 1990; Lingle & Schiemann, 1996; Stivers, Covin, Hall, & Smalt, 1998 ).<br><br> The di*erences vary across uses, indicating that extensive use of performance measures for one purpose does not necessarily imply that the measures are used for other pur- poses. The largest di*erences relate to the external disclosure of customer and quality information, the use of employee information for evaluating capital investments, and the quality of customer- related measures. Di*erences related to identifying problems and developing action plans generally are smaller than those associated with other uses.<br><br> Use of measurement alignment techniques The use of performance measurement alignment techniques is assessed using three questions on the implementation of balanced scorecards and eco- nomic value measures (e.g., economic value added or cash Kow return on investment) and the extent of formal reliance on a 8 8business model 9 9 or 8 8the- ory of business 9 9 that causally links performance drivers to performance outcomes. Following Krumwiede (1998) , a six-point scale is used to measure the implementation of balanced scor- ecards or economic value measures, where 1=not considered, 2=implemented and abandoned, 3=considering, 4=implementing now, 5=used, and 6=used extensively. For our analyses, we code scorecard or economic value measure use one if the respondent stated that the 0rm uses or extensively uses that method, and zero otherwise.<br><br> 9 As shown in Panel A of Table 3 , implemen- tation of economic value measures is more fre- quent than balanced scorecard implementation. More than one-third of the 0rms (36.7%) use economic value measures to some extent and another 7.9% are implementing them now. In contrast, only 20% use balanced scorecards, with an additional 10.7% of scorecard systems under- going implementation.<br><br> Roughly 50% of the 0rms have not considered implementing a balanced scorecard, vs. 31.4% for economic value measures. Business model reliance ranges from 1=not at all to 6=completely.<br><br> Nearly 30% of the 0rms place no reliance on a formal, causal business model or theory of the business ( Table 3 , Panel B), and only 34.7% make substantial to complete use of business models (four or greater on the six- point scale). Extensive reliance on business models is coded one for responses of four or greater to this question, and zero otherwise. Although Kaplan and Norton (1996, 2001) now argue that causal business models are an integral component of the balanced scorecard concept, 76.9% of the 0rms claiming to use a balanced scorecard place little or no reliance on business models (not reported in the table).<br><br> Similarly, 8 Although the survey did not ask respondents the factors they considered when assessing measurement quality, Caval- luzzo and Ittner (in press) 0nd that managers 9 perceptions of measurement system quality are a function of data limitations (i.e., the ability of existing information systems to provide valid, reliable, and timely data in a cost e*ective manner) and diLculties selecting and interpreting appropriate performance measures. Thus, responses to this question are likely to reKect performance measurement attributes other than scope, such as timeliness and information system e*ectiveness, that have been identi 0ed in prior studies (e.g., Chenhall & Morrris, 1986; Gates, 1999 ). The goal setting question, in turn, is consistent with the target setting component of Chenhall and Morris 9 (1986) system integration construct, as well as Simons 9 (1987) examination of the emphasis on budgetary and output goals in di*erent strategic contexts.<br><br> 9 We do not distinguish between 0rms that have not con- sidered, are considering, or are implementing these techniques because none of these 0rms actually used these techniques at the time of the survey, making it impossible to detect any per- formance di*erences due to these di*erent responses. Very few 0rms have implemented and abandoned these practices, pre- venting us from examining the performance consequences of this decision. We also repeated the analyses using separate dichotomous variables for 0rms making 8 8extensive 9 9 use of these practices and 0rms making 8 8very extensive 8 8 use.<br><br> These results are reported later in the paper. C.D. Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 725 79.2% of economic value users make little or no use of business models, despite claims that greater understanding of the causal model linking 0nan- cial and non- 0nancial measures to economic results can enhance the bene 0ts from the adoption of economic value measures.<br><br> Performance variables We assess the performance implications of stra- tegic performance measurement using two sets of variables: (1) managers 9 responses regarding their satisfaction with the performance measurement system, and (2) publicly-available information on the 0rm 9s accounting and stock market perfor- mance. We include the satisfaction measure to allow comparisons of our results to other perfor- mance measurement studies using satisfaction as their dependent variable (e.g., Banker et al., 2001a, 2001b; Rigby, 2001; Sandt et al., 2001 ). The accounting and market measures provide more direct tests of the inKuence of measurement practices on economic performance, which most SPM advocates argue is the ultimate objective of these systems.<br><br> Three questions are used to measure a 0rm 9s satisfaction with its measurement system: (1) how well the system meets expectations (1=has not met expectations to 6=exceeded expectations); (2) how well the system compares to the respondent 9s concept of an 8 8ideal 9 9 system (1=not at all ideal to 6=very close to ideal); and (3) overall satisfaction with the system (1=not at all satis 0ed to 6=com- pletely satis 0ed). Actual responses to all three questions range from 1 to 6, and indicate that these 0nancial services 0rms, on average, are moderately satis 0ed with their measurement sys- tems. Mean (median) scores are 3.54 (4.00) relative to expectations and 3.13 (3.00) relative to ideal.<br><br> Mean (median) overall satisfaction is 3.42 (4.00), with 20.0% of the respondents rating their satis- faction 5 or 6 and 37.2% rating it 1 or 2. The three satisfaction questions load on a single factor with a coeLcient alpha of 0.91. Our satisfaction con- struct represents the average standardized response to these three questions.<br><br> We evaluate economic performance using sev- eral measures that are commonly employed to assess 0nancial results. These include two publicly- available accounting measures (return on assets and 3-year sales growth) and two stock return measures (1-year and 3-year returns). The accounting and stock return performance mea- sures, obtained from Compustat and CRSP, are measured contemporaneously with the date of the survey.<br><br> The 1-year measures are for 0scal 1999 and the 3-year measures are for the time period covering 0scal 1997 3 0scal 1999. For 0rms in our sample, mean (median) sales growth is 22.5% (16.3%), and return on assets (ROA) is 1.8% (1.2%). Mean (median) 1-year stock returns are À 12.9% ( À 13.5%), while 3-year stock returns are 1.6% (1.9%).<br><br> 10 We include several additional variables in the eco- nomic performance tests to control for other poten- tial determinants of accounting and stock price performance. To account for the well-known e*ects of organizational size and growth opportunities on 10 Compared with all 0nancial services 0rms on Compustat, the mean 0rm in our sample has signi 0cantly lower sales growth and stock returns. Return on assets is not statistically di*erent.<br><br> Table 3 The use of performance measurement alignment techniques in 0nancial services 0rms Balanced scorecard (%) Economic value measures (%) Panel A. Use of balanced scorecard and economic value measures ( n =140) Not considered 50.7 31.4 Implemented and abandoned 1.4 2.1 Considering 17.1 22.1 Implementing now 10.7 7.9 Use 15.0 24.5 Use extensively 5.0 12.1 Panel B. Formal reliance on a 8 8Theory of the Business 9 9 or business model that causally links performance drivers to performance outcomes ( n =136) 1=Not at all 29.7% 2 16.7% 3 18.8% 4 19.6% 5 10.8% 6=Completely 4.3% 726 C.D.<br><br> Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 0rm performance, we include the log of assets (denoted Size ) and the ratio of the book value of assets to the market value of equity (denoted BTOM , an inverse measure of growth opportu- nities) in the models. We also include the median performance of other 0rms in the same four-digit SIC code to control for the e*ects of industry sec- tor on 0nancial performance. 11 Performance tests We test our hypotheses by examining the rela- tions between our 0rm performance variables and four di*erent measurement system characteristics: (1) overall measurement diversity, (2) alignment with 0rm strategy, (3) alignment with 0rm value drivers, and (4) use of measurement alignment techniques.<br><br> Overall measurement diversity Our 0rst set of tests investigates whether greater 8 8measurement diversity 9 9 is associated with super- ior performance ( Hypothesis 1 ). Consistent with claims in the performance measurement literature, we assume that greater measurement diversity is characterized by more extensive use of a broad set of 0nancial and non- 0nancial measures for per- formance evaluation and decision-making pur- poses, independent of the 0rm 9s strategy and value drivers. Overall measurement diversity equals the average standardized rating for each of the ten value driver categories across all uses (problem identi 0cation, capital investments, performance evaluation, and external disclosure), goal setting, and measurement quality.<br><br> 12 A higher value for this variablemeans thatthe 0rmusesall of the measures toagreaterextent,setsmoreextensivegoals,andhas greater measurement quality for these categories. Given the recent emphasis on the use of non- 0nancial measures, we also examine the relative importance of 0nancial vs. non- 0nancial measure- ment (denoted Financial measurement focus and Non-:nancial measurement focus , respectively).<br><br> 13 If greater non- 0nancial measurement is more bene- 0cial than greater 0nancial measurement, the association between performance and the use of non- 0nancial measures will be stronger than that with 0nancial measures. Evidence on the inKuence of Overall measure- ment diversity on performance is presented in Panel A of Table 4 . 14 When satisfaction with the measurement system is the dependent variable, the coeLcient on Over- all Measurement Diversity is positive and highly signi 0cant ( P < 0.01, two-tailed), with an adjusted R 2 of 0.215.<br><br> As seen in Panel B, this result is dri- ven by greater use and improved measurement of non- 0nancial performance measures, which is sig- ni 0cantly positive when overall measurement diversity is decomposed into 0nancial and non- 0nancialcomponents.Incontrast, greateremphasis on 0nancial measures is not signi 0cantly associated with measurement satisfaction. This evidence is consistent with prior studies that 0nd greater measurement satisfaction when more emphasis is placed on non- 0nancial measures (e.g., Sandt et al., 2001 ), but does not support claims that greater emphasis on traditional budgetary or 0nancial control uses of performance measures leads to lower satisfaction ( Banker et al., 2001b ). Although overall and non- 0nancial measure- ment focus are positively associated with satisfac- tion, their relation with economic performance is mixed.<br><br> Neither variable is signi 0cantly associated with ROA, sales growth, nor 3-year stock returns. 11 Since 0rm strategy can also a*ect economic performance, we repeated our performance tests using the three strategy constructs as additional control variables. This addition had virtually no impact on our other results.<br><br> 12 We use a single measurement focus construct rather than separate constructs for usage, goal setting, and measurement quality because of high correlations among these practices, which creates problems with multicollinearity. When separate constructs are computed for each of these three measurement characteristics, the correlation between usage and goal setting is 0.79, between usage and measurement quality is 0.83, and between goal setting and measurement quality is 0.74. 13 Financial focus is measured by the values assigned to short-term 0nancial results and non- 0nancial focus is measured by the average values assigned to the remaining nine perfor- mance categories.<br><br> 14 We delete observations from our performance tests when studentized residuals are greater than four standard deviations from the mean in order to mitigate the impact of outliers on our results. For each regression, fewer than three observations were removed. C.D.<br><br> Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 727 However, both variables are positively related to 1-year stock returns ( P < 0.05, two-tailed). These 0ndings lend weak support to the hypothesis that greater measurement diversity is associated with higher performance. With respect to the control variables, 0rm per- formance is strongly related to sector perfor- mance, while ROA is positively related to the 0rm 9s growth opportunities (i.e., negatively related to its book-to-market ratio).<br><br> Organizational size is not signi 0cant in any of the models. Alignment between strategy and measurement practices The preceding tests assume that greater measurement diversity a*ects performance, inde- pendent of the 0rm 9s strategies. Contingency and economic theories, on the other hand, contend that measurement practices must be aligned with the 0rm 9s strategy.<br><br> We investigate these theories by examining whether performance is enhanced when strategy and measurement are more closely Table 4 Association between measurement diversity and performance in 0nancial services 0rms; t-statistics in parentheses Satisfaction ROA Sales growth One-year stock returns Three-year stock returns Panel A: Association between overall measurement diversity and performance Intercept 0.012 0.012 0.150 ** 0.045 À 0.305 ** (0.176) (1.645) (1.892) (0.401) ( À 2.131) Overall measurement diversity 0.873 *** 0.001 À 0.031 0.087 ** 0.064 (6.255) (0.471) ( À 1.224) (2.033) (1.326) Sector performance 0.597 *** 0.674 *** 0.847 *** 1.219 *** (7.082) (4.609) (3.474) (3.852) SIZE 0.0006 0.0069 À 0.0117 0.0236 (0.800) (0.360) ( À 0.918) (1.565) BTOM À 0.012 *** À 0.176 *** 0.0096 0.0632 ( À 3.056) ( À 3.159) (0.160) (0.633) Adj. R 2 0.215 0.321 0.259 0.095 0.118 Panel B: Association between :nancial and non-:nancial measurement focus and performance Intercept 0.012 0.0118 0.150 * 0.045 À 0.310 ** (0.176) (1.634) (1.884) (0.395) ( À 2.153) Financial measurement focus 0.072 À 0.008 À 0.0041 À 0.0032 0.0213 (0.682) ( À 0.375) ( À 0.219) ( À 0.102) (0.595) Non- 0nancial measurement focus 0.793 *** 0.0018 À 0.0279 0.0843 ** 0.0505 (5.830) (0.607) ( À 1.119) (2.035) (1.093) Sector performance 0.603 *** 0.675 *** 0.853 *** 1.230 *** (7.051) (4.527) (3.479) (3.858) SIZE 0.006 0.0069 À 0.0115 0.0236 (0.431) (0.914) ( À 0.903) (1.562) BTOM À 0.0124 *** À 0.177 *** 0.0102 0.0695 ( À 3.041) ( À 3.132) (0.170) (0.686) Adj. R 2 0.210 0.317 0.252 0.088 0.111 Overall measurement diversity=the average extent to which measures related to the 10 performance categories are used for perfor - mance evaluation and decision-making, have goals set, and have high quality of measurement.<br><br> Financial measurement focus=the average extent to which short-term accounting measures are used for performance evaluation and decision-making, have goals set, and have high quality of measurement. Non- 0nancial measurement focus=the average extent to which measures related to the nine non- 0nancial performance categories are used for performance evaluation and decision-making, have goals set, and have high quality of measurement. Sector performance=the median performance of other 0rms in the same four-digit SIC code.<br><br> SIZE=the log of assets. BTOM=the ratio of book value of assets to the market value of the 0rm (an inverse measure of growth opportunities). * P < 0.10 (2-tailed).<br><br> ** P < 0.05 (2-tailed). *** P < 0.01 (2-tailed). 728 C.D.<br><br> Ittner et al./Accounting, Organizations and Society 28 (2003) 715 3741 aligned ( Hypothesis 2 ). We 0rst develop bench- mark models for assessing the extent of alignment by regressing each combination of measurement characteristics (use, measurement quality, and goals) and type of measure (e.g., short-term 0nancial, customer, employee, etc.) on the three strategy constructs ( Flexible , Innovate , and Main- tain ). This yields 60 individual benchmark regres- sions (i.e., 10 types of measures multiplied by six measurement characteristics).<br><br> Our proxies for measurement system alignment are then computed by averaging the standardized residuals from these models. This approach assumes that 0rms, on average, have correctly chosen their performance measurement systems, and that the estimated models capture the appro- priate benchmark for system characteristics given the 0rm 9s strategy ( Van de Vin & Drazin, 1985 ). If the benchmark models represent optimal measurement practices, then any deviations from the estimated models (i.e., either too little or too much measurement emphasis) should be negatively associated with satisfaction or performance.<br><br> These deviations are analogous to the 8 8measurement gaps 9 9 discussed in the performance measurement literature in that gaps (or deviations) between the perceived importance of a performance criteria and the criteria 9s use in the performance measurement system are assumed to be detrimental. To allow for potential di*erences between measurement practices that are less or more extensive than predicted, we compute separate variables for negative and positive residuals and then take their absolute values (denoted Negative overall measurement residual and Positive overall measurement residual , respectively). 15 We also subdivide each of these variables into 0nancial and non- 0nancial components to investigate whether this measurement distinction inKuences performance.<br><br> The resulting performance tests are presented in Table 5 . The adjusted R 2 s for the satisfaction models (0.117 and 0.124) are smaller than those using the measurement diversity scores in Table 4 (0.215 and 0.210). The di*erences in explanatory power imply that respondents are more likely to rate measurement satisfaction in terms of the overall diversity and extent of measurement rather than relative to the requirements of their chosen strategies.<br><br> Positive deviations from the benchmark model have a signi 0cant positive association with measurement satisfaction, indicating that 0rms with more extensive measurement than competi- tors following similar strategies rate their systems more highly. However, negative deviations from the benchmark models show no di*erences in satisfaction. When the overall residuals are subdivided into 0nancial and non- 0nancial components, Positive non-:nancial measurement residual has a highly signi 0cant positive association with satisfaction ( P < 0.01, two-tailed), again suggesting that satis- faction is related to greater use of non- 0nancial measures than 0rms with similar strategies.<br><br> Nega- tive non-:nancial measurement residual , in con- trast, is not signi 0cant in the satisfaction model. Turning to the economic performance tests, we continue to 0nd no evidence that measurement characteristics are associated with accounting per- formance. We also 0nd no evidence that lower than predicted measurement inKuences perfor- mance.<br><br> Greater than predicted measurement ( Positive overall measurement residual ), on the other hand, has positive and signi 0cant associ- ations with both 1-year stock returns ( P < 0.05, 2-tailed) and 3-year stock returns ( P < 0.10, 2-tailed). When we subdivide the overall residuals into 0nancial and non- 0nancial components, the individual 0nancial and non- 0nancial residual measures are not signi 0cantly associated with stock market performance at the 10% level.Thisis somewhat surprising given the earlier signi 0cant association between Positive overall measurement residual and the two stock return measures. Further 15 For example, assume that three 0rms have average resi- duals of 0, 3, and À 3 after averaging the residuals across all 60 benchmark models.<br><br> Negative overall measurement residual and Positive overall measurement residual equal (0, 0) for the 0rst 0rm, (0, 3) for the second 0rm, and (

less