Lost in translation: The effects of incentive compensation on strategy surrogation

Transcription

1 Lost in translation: The effects of incentive compensation on strategy surrogation Jongwoon (Willie) Choi Ph.D. Student Emory University Gary Hecht Assistant Professor Emory University William Tayler Assistant Professor Emory University July 21, 2009 The authors appreciate the helpful comments and suggestions from Rob Bloomfield, Joel Demski, Ron Guymon, Kathryn Kadous, Steve Kachelmeier, Bill Kinney, Lisa Koonce, Joan Luft, Lil Mills, Nick Seybert, Mike Shields, Kristy Towry, Greg Waymire, Michael Williamson, an anonymous AAA Annual Meeting reviewer, and workshop participants at the Brigham Young University Accounting Research Symposium, Emory University, Northeastern University, and the University of Texas at Austin. We also appreciate the research assistance of Daniel Baum and Megan Buckley. This research benefited from a Dean s Research Grant from the Goizueta Business School at Emory University.

2 Lost in translation: The effects of incentive compensation on strategy surrogation Abstract To facilitate managers decision-making, firms develop strategic performance measurement systems that translate strategy into performance measures. Ideally, managers see measures for what they are: imperfect proxies for intangible strategic constructs. However, managers may increase their focus on measures to the detriment of their consideration of the strategic constructs the measures are intended to represent. As such, managers may underappreciate the representational role of measures, ultimately losing sight of the global framework in which the measures reside a phenomenon we label surrogation. In this paper, we investigate whether and how managers propensity to exhibit surrogation is affected by the use of strategically-linked performance measures for compensation purposes. We predict that this tendency is most prevalent when managers are compensated on a single measure of a strategic construct, and that this tendency is less prevalent when managers are compensated on multiple measures of a strategic construct. Via an experiment, we find support for these hypotheses. Our paper contributes to the literature on strategic performance measurement systems by highlighting the tendency of managers to use measures as surrogates for strategy. Further, our study speaks to an on-going academic debate regarding the trade-off between the decision-facilitating and decisioninfluencing roles of performance measures.

3 I. Introduction Strategic performance measurement systems facilitate managers decision-making by translating strategy into performance measures. A critical feature of these measurement systems is that strategically-linked performance measures are transparent (Kaplan and Norton 1996), allowing managers to see through the measures back to the strategy. This transparency allows managers to use measures to gauge the appropriateness of the strategy and make adjustments to the strategy as necessary. Kaplan and Norton refer to this process as "strategic learning (Kaplan and Norton 1996). Ideally, managers see measures for what they are: imperfect proxies for intangible strategic constructs. However, if managers do not fully appreciate the representational role of measures, then their ability to make appropriate judgments regarding the implementation of strategy will be hindered. Familiar examples of this form of myopia abound. For example, managers charged with "delighting the customer" who use customer satisfaction surveys to gauge strategic success may begin to see maximizing survey results as the strategy, and behave accordingly. University administrators accustomed to focusing on teacher evaluations during faculty reviews may lose sight of the strategic construct of "excellence in teaching" and behave as though the evaluations are the construct of interest. Investors, through repeated use of a firm s financial statements, may lose sight of the construct(s) of interest that these statements represent (as suggested by Ijiri (1967, 1975)). The behavior underlying these examples is consistent with increased attention to measures and corresponding decreased attention to the constructs for which the measures are intended to proxy. This suggests that managers may underappreciate the representational role of measures, ultimately losing sight of the global framework in which the measures reside (i.e., as proxies for strategic constructs) a phenomenon we label surrogation. 1

4 Many factors germane to the strategic performance measurement system context likely influence surrogation. This paper focuses on one such factor: the use of measures as a basis of compensation. Relying on extant psychology and managerial accounting literature, we develop two hypotheses related to the influence of incentive compensation on managers propensity to exhibit surrogation. First, we predict that basing a portion of managers compensation on a single measure of a strategic construct exacerbates surrogation of that construct. Specifically, we argue that the attention-directing effects of incentive compensation increase managers focus on that measure as a means of achieving higher compensation, at the expense of attention given to the measure s role as a representation of the strategic construct. Second, we predict that tying managers compensation to multiple measures of a strategic construct leads to less surrogation of that construct (relative to a scenario in which compensation is based on a single measure). In particular, we posit that managers, when compensated on multiple measures of a strategic construct, consider the relationships among the compensated measures. Such considerations highlight the limitations of any one measure as a representation of the strategic construct, thereby increasing attention to the representational role of the measures. We conduct an experiment to test our hypotheses. Graduate business students use the computer game Spore (Electronic Arts 2008; Lawton 2008) to design a virtual creature in line with a strategy statement provided to all participants. We provide four measures of a single strategic construct to participants in real-time as they design their creatures. We manipulate compensation tied to these measures at three levels: flat-wage compensation, performance-based compensation based on a single measure of the creature s abilities, and performance-based compensation based on multiple measures of the creature s abilities. 2

5 After designing his or her creature, each participant then views a series of 18 virtual creatures and makes choices regarding the design of these creatures. We designed these choices to allow us to evaluate the extent to which earlier incentive compensation tied to measures influenced participants tendency to surrogate strategy with compensated measures. Importantly, these 18 choices are not tied to compensation in any way, allowing us to view behavior absent the influence of incentive compensation. These choices are the basis of our key dependent variables. We find evidence consistent with our hypotheses. Participants compensated on a single measure of a strategic construct were more likely to exhibit strategy surrogation than participants who received a flat-wage. This result is robust even when choices consistent with surrogation entail an opportunity cost. Further, we find evidence that participants compensated on multiple measures of a strategic construct were less likely to exhibit surrogation than single-measurecompensation participants. Lastly, we also explore the relative surrogation propensity of multiple-measure-compensation participants and flat-wage-compensation participants, and find evidence that the propensity to surrogate is not significantly different across these compensation conditions. Our study contributes to the literature on strategic performance measurement systems by highlighting the tendency of individuals to surrogate strategy with performance measures. To the extent that surrogation is detrimental to the implementation of strategy, this propensity represents a cost inherent to the use of measures as strategic proxies. Our study helps develop academic and practitioner understanding of factors that influence the extent to which firms incur this potential cost (Ittner and Larcker 1998). 3

6 Further, our study is important to practicing managers and managerial accountants responsible for developing, implementing, and maintaining strategic performance measurement systems. Though losing sight of strategic constructs that underlie performance measures may be acceptable for low-level employees, managers responsible for the development and implementation of strategy bear a greater responsibility for recognizing measures representational role. Our results suggest that the difficulty of this responsibility is heightened when the measures are used for both performance measurement and compensation purposes. Related to this dual-purpose use of measures, our results also speak to an on-going academic debate regarding the trade-off involved in using performance measures for both decision-facilitating and decision-influencing purposes (Feltham et al. 2006; Ittner and Larcker 2001; Narayanan and Davila 1998; Otley 1994). A primary focus of this debate relates to managers incentives to intentionally manipulate measures to increase their pay, diminishing the usefulness of these measures for decision-facilitating purposes. We view managers propensity to surrogate as complementary to, but fundamentally distinct from, such intentional behavior. Specifically, our study speaks to a by-product of tying compensation to measures that arises absent managers manipulative actions. The remainder of this paper proceeds as follows. Section II provides the background and hypotheses. Section III describes our experiment and related procedures. Section IV presents results. Section V concludes. 4

7 II. Background and Hypotheses Strategic Performance Measurement Systems and Surrogation Our context of interest is firms strategic performance measurement systems. Prior research uses various labels for strategic performance measurement systems, such as performance measurement models (Malina et al. 2007) and balanced scorecards (Kaplan and Norton 1992). In this paper, we refer to any system with an explicit link between strategy and performance measurement as a strategic performance measurement system (Atkinson et al. 1997; Epstein and Manzoni 1997; Nanni, Dixon, and Vollman 1992; Webb 2004; ). A key feature of these systems is the mapping of relationships among strategic constructs and performance measures. 1 These outlines are commonly referred to as strategy maps (Kaplan and Norton 2000), value driver maps (Ittner and Larcker 2003), or causal chains of performance (Tayler 2009) (see Figure 1). [Insert Figure 1 here] Prior research highlights the importance of linking firm strategy and performance measures, noting at least three key functions of such alignment (see Chenhall (2003), Ittner and Larcker (1998, 2001), and Langfield-Smith (1997) for reviews). First, translation of strategy into a set of measures facilitates communication of strategy throughout the firm. (Malina and Selto 2001, 2004). That is, performance measures signal to managers a (potentially incomplete) set of actions to take, and inform them of how their individual actions affect others within the firm, as well as the firm as a whole. Second, the aligning of strategy with performance measures enables firm management to evaluate strategy. Performance measures provide tangible evidence that can 1 Strategic constructs, which define the purpose of actions taken, are often referred to as "strategic objectives" (Kaplan and Norton 1996). A collection of causally-connected strategic constructs define the strategy for an organization. 5

8 be used to determine not only the extent to which the firm is accomplishing its strategy (Bromwich 1990), but also to test and validate the strategy (Campbell et al. 2008). Third, tying measures to strategy allows for the development of new strategies (Simons 1991, 1994). Via performance measure information, managers may discover new strategic constructs and/or new strategies to pursue. Key to these functions is the transparency of measures, which allows managers to see through the measures back to the strategy (Kaplan and Norton 1996). Notwithstanding the obvious benefits of aligning managers' actions with a firm s strategic constructs, there exists a potential by-product of representing strategy with performance measures. Specifically, managers may focus on performance measures to the point that they lose focus on the strategic construct(s) for which the measures are intended to proxy. Ijiri (1967, 1975) discusses the implications of this phenomenon for investors. Ijiri (1975) suggests that the process of measurement is intended to create and disseminate information pertaining to some underlying construct of interest. For example, stakeholders use a firm s financial statements (i.e., accounting measures) to learn about the firm s economic viability, the construct of interest. Although measures are designed to be proxies for higher-level constructs, Ijiri (1967, 1975) identifies the potential for a measure to actually replace the construct as the focus of interest. The implications of this phenomenon are not trivial for stakeholders or the firm. 2 In a strategic performance measurement system setting, this suggests managers may underappreciate the representational role of the measure, ultimately losing sight of the global framework in which the 2 Another example is the relationship between Hicksian income, defined as the amount that can be paid out in dividends during a period, while leaving the firm equally well off at the beginning and end of the period, and accounting earnings (Hicks 1939, 176; Schipper and Vincent 2003). A firm s stakeholders rely on accounting earnings reported on the income statement to learn more information about the firm s (Hicksian) income. However, through repeated use of the earnings measure, investors may begin to act as though earnings (and not Hicksian income) is the construct of interest. Evidence on the market response to firms just missing earnings benchmarks (e.g., Burgstahler and Dichev 1997) may reflect this phenomenon at the market level. (See Camerer (1987) for a review of psychological phenomena reflected at the market level.) 6

9 measure resides (i.e., as a proxy for a strategic construct). We label this phenomenon as surrogation. 3 While our interest in surrogation is motivated by the costs and negative implications associated with such behavior, it is important to note that we do not adopt the view that surrogation and its implications are exclusively negative. 4 However, surrogation is potentially detrimental in many scenarios. For instance, if managers have surrogated strategic constructs with performance measures, and the performance measures suggest a failing strategy, managers may recommend abandoning the strategy, rather than first evaluating whether strategic constructs were appropriately operationalized (i.e., whether the appropriate performance measures were chosen). Moreover, a manager who defines a strategic construct via an inherently imperfect measure likely is less able to recognize when a measure becomes obsolete, or may not perceive the benefits of a new measure that better captures a strategic construct. Indeed, implications of surrogation are likely related to a potentially sub-optimal locking in of measures, business models, and/or incentive compensation schemes. The potential for these costs to mitigate the effectiveness of the strategic performance measurement system (e.g., the successful communication of strategy) suggests a need for a deeper understanding of the surrogation phenomenon. This paper offers an initial exploration of surrogation via an investigation of the effect of incentive compensation on managers propensity to exhibit surrogation. In the next two sub-sections, we consider how the attention-directing 3 The term surrogation is inspired by Ijiri (1967, 1975). 4 One class of scenarios in which surrogation and its implications are of little consequence relates to the extent to which measures capture the strategic construct. If a construct is perfectly represented by a single measure, or set of measures (see Farrell (2009) and Luft and Shields (2003)), then the notion of surrogation is moot. However, because strategic constructs are typically ill-defined and complex, it follows that attempts to measure constructs will fall short to some degree, ultimately allowing for some degree of surrogation. 7

10 effects of incentive compensation influence managers strategy implementation experiences, and thus, their propensity to surrogate. Incentive Compensation on a Single Performance Measure of a Strategic Construct A significant body of literature addressing the incentives-performance relationship suggests that a variety of interacting mechanisms determine how incentives affect performance (see Bonner and Sprinkle (2002) for a review). While no one theory provides a comprehensive explanation for how incentives ultimately influence performance, a key take-away from this body of research is that incentive compensation has attention-directing effects, which influence an individual s experience with performance measures (see Atkinson et al. (1997) and Lambert (2001)). Prior research shows that firms often use multiple performance measures to represent a single strategic construct (Banker et al. 2000; Chen et al. 2008) and often must decide on the number of strategically-linked performance measures on which to provide financial incentives. This decision is non-trivial; when designing compensation contracts, firms must balance potentially conflicting factors (e.g., contract completeness vs. contract complexity see Ittner and Larcker (1998) for a discussion). As such, firms may choose to provide incentives only on a subset of the available measures. 5 Given these findings, our paper examines whether and how incentive compensation affects managers tendency to surrogate strategic constructs with compensated measures. We first consider the most basic version of this context incentive compensation on a single measure of a strategic construct. Prior work in psychology demonstrates that humans have 5 For example, Banker et al. (2000) report that the hotel chain examined in their study collected information on three measures of customer satisfaction, but only included two of them in managers incentive plans. 8

11 limited attention, and that task structure can affect the attention given to certain features of the environment (James 1890; Shiffrin 1988). Consistent with this research, we posit that incentive compensation focuses managers attention on the compensated measure as the means by which their actions affect their compensation. Moreover, we argue that such selective attention comes at the expense of managers attention to the role of the measure as an imperfect proxy for a strategic construct. That is, a manager who focuses on a measure in its role as a determinant of compensation likely de-emphasizes the global picture in which the measure was originally communicated (i.e., the measure as a representation of a strategic construct, as communicated via the strategic performance measurement system). Ultimately, the attention-directing effects of a single-measure compensation scheme influence the manager s strategy-implementation experience (Frederickson et al. 1999). Relative to a scenario in which measures are not used for compensation purposes, incentive compensation on a single measure of a strategic construct increases a manager s propensity to underappreciate the representational role of that measure (i.e., increases the manager propensity to exhibit surrogation). This leads to the following hypothesis: H1: Compensating on a single measure of a strategic construct leads managers to surrogate the strategic construct more than they would under a flat-wage compensation scheme. Incentive Compensation on Multiple Performance Measures of a Strategic Construct Like compensation on a single measure of a strategic construct, compensation on multiple measures of a single strategic construct also leads managers to view those measures as 9

12 a means by which their actions determine their compensation. 6 Thus, a manager s attention is directed toward compensated measures (regardless of the number of compensated measures), and the nature of the strategy implementation experience is influenced by the direction of this attention. However, managers attention towards, and experience with, multiple measures is markedly different than that under compensation on a single measure of a strategic construct. Specifically, when managers face a multiple-measure compensation scheme, their task is inherently more complex. With multiple compensated measures, a manager likely compares and contrasts the compensated measures on a variety of dimensions. This process leads managers to maintain greater awareness of the measures representational role for two reasons. First, comparing the measures leads to (or reminds the manager of) the realization that all of the compensated measures represent the same strategic construct. Second, contrasting the compensated measures highlights the limited ability of any one measure to completely represent the corresponding strategic construct. Thus, relative to a single-measure compensation scheme, managers will attend more to the representational role of the measures (i.e., their role as proxies for a related strategic construct). Thus, the strategy-implementation experience is framed not only by the compensation scheme, but also by the representational role of measures in the strategic performance measurement system. Therefore, relative to managers compensated on a single measure of a strategic construct, managers compensated on multiple measures of a strategic construct are less likely to exhibit surrogation. This leads to the following hypothesis: 6 In this paper, we are interested in examining surrogation in a simple environment to maintain internal validity. Thus, we consider a single strategic construct. While the implications of other sources of complexity (i.e., multiple strategic constructs, ambiguity of the categorization of measures, the nature of causal relationships among measures and constructs, etc.) relate to our phenomenon of interest, they are beyond the scope of this initial investigation of surrogation. 10

13 H2: Compensating on multiple performance measures of a single strategic construct leads managers to surrogate a related strategic construct less than they would under a single-measure compensation scheme. We do not explicitly hypothesize managers propensity to surrogate when they are compensated on multiple measures relative to the flat-wage compensation benchmark. However, our experiment design, as described in the next section, allows us to make such a comparison. We provide related results in Section IV. III. Method We tested our hypotheses in an experimental lab where participants used computers to work through self-guided instructions and perform various tasks (described below). Participants were 79 students recruited from graduate business courses at a highly ranked business school in the southeastern United States. On average, participants were 29 years old. Forty-eight (61%) of the participants were male. Strategy Implementation via Creature Design After an introduction to the study (described below), participants designed a virtual creature using the Creature Creator module of the computer game Spore (Electronic Arts 2008; Lawton 2008). Instructions indicated that, following the experiment session, each participant s creature would be placed into an environment where it would interact with other creatures from the computer game Spore. Instructions further indicated that: 11

14 The purpose of the game of Spore is to create a creature that interacts with other life forms in a way that enables it to become the dominant species. You can become the dominant species by making friends with other species, or by overpowering them. In accordance with this purpose, participants were provided with the following strategy for designing their creature: Design a creature that can socialize with other life forms such that it will become the dominant species on the planet. Figure 2 provides a sample screen similar to that viewed by participants in our study. At the outset of this phase of the experiment, participants received an endowment of 500 DNA points, which could be used to purchase a variety of body parts for their creature. 7 Each part added one or more abilities to the creature. For example, adding a mouth allowed a creature to eat food, sing (useful for making friends), and bite (to attack enemies) with varying degrees of effectiveness. Participants could see what abilities each part added before purchase, as well as measures of each ability in real-time as they added each part. When a part was added to the creature, the cost of the part was automatically deducted from the DNA point endowment. If a part was removed from the creature, ability measures were immediately adjusted and the associated DNA points were fully refunded. [Insert Figure 2 here] While designing their creature, participants saw numerous measures of their creature s abilities, including four measures tied to their creature s ability to socialize, the strategic construct of interest. These quantitative measures reflected the creature s ability to sing, dance, charm, and pose (on a scale of 0 to 5). Participants were told that, in the game of Spore, each of these abilities was equally important to their creature s ability to socialize. 7 This budget limit (500 DNA points) serves to ensure that participants have enough resources to design a creature in line with the provided strategy, but not without making trade-offs. That is, given this resource constraint, participants have to choose between multiple ways of achieving the stated strategy. 12

15 Compensation Some participants received incentive compensation tied directly to measures of their creature s social ability. The nature of this source of compensation is our independent variable, which we manipulated at three levels: no social-ability compensation (flat-wage compensation), performance-based compensation based on a single measure of the creature s social-ability (single-measure compensation), and performance-based compensation based on multiple measures of the creature s social-ability (multiple-measure compensation). In the flat-wage compensation condition, participants received $10 regardless of the state of their creature s social-ability measures. In the single-measure compensation condition, participants received $2 per level of their creature s sing measure. For example, if a participant designed a creature with sing level 4, then he received $8. In the multiple-measure compensation condition, participants received $1 per level of their creature s sing, dance, and charm measures. For example, if a participant designed a creature with sing level 3, dance level 2, and charm level 4, then he received $9. In addition to this (manipulated) form of compensation, all participants received compensation based on a post-experiment assessment of their success at achieving the provided strategic construct of designing a creature that can socialize effectively. Specifically, after the completion of the experiment session, each participant s creature was individually placed in a random Spore environment. 8 A research assistant then performed simple commands to allow each participant s creature to befriend neighboring creatures. The time to befriend two 8 Though the environment was randomly determined, the same environment was used for each participant to reduce noise in the post-experiment measurement of creatures social abilities. 13

16 computer-based creatures was recorded. 9 Participants within the same experiment condition were randomly assigned to 10-person groups and ranked according to the befriending-speed measure. Participants relative ranking within these groups determined this portion of their compensation. 10 The top performer in each group earned $20, the 2 nd best performer received $18, etc. Dependent Variables After participants finished designing their virtual creature, they viewed 18 successive pictures of other virtual creatures. With each of these creatures, participants saw two sets of potential realizations (i.e., modification packages) of the four measures related to the creature s social ability (sing, dance, charm, and pose). As they viewed each creature, participants were told to select the modification package that they believed, if added to the creature, would best enable them to implement the stated strategy (i.e., Design a creature that can socialize with other life forms such that it will become the dominant species on the planet ). In addition, participants were told to assume that the DNA cost was equal across each pair of modification packages. Participants viewed each of the 18 creatures one at a time, and made their modification-package selections as they viewed each creature. Once a modification package was selected and participants moved on to the next creature, they could not go back and change 9 A creature s total social-ability level (i.e., the sum of a creature s sing, dance, charm and pose levels) is a key determinant of this measure of performance, as creatures with lower measure levels require more attempts (i.e., take longer) to befriend other creatures. If applicable, failure to befriend two creatures was also documented. 10 We used 10-person subsets to increase the salience, perceived attainability, and relevance of this source of compensation. However, given that the number of participants in our experiment conditions was not a multiple of 10, one 10-person group in each experiment condition included randomly selected participants already part of another group. This allowed all participants to be ranked in groups of 10. These filler participants were not compensated based on their relative ranking in this second group. 14

17 earlier modification-package selections. See Figure 3 for an example of what participants saw as they made these selections. 11 [Insert Figure 3 here] Participants 18 modification-package selections were used to form our key dependent variables. We designed modification-package pairings ex ante to allow us to evaluate, via participants' selections, the extent to which our incentive compensation manipulation influenced participants tendency to surrogate strategy with compensated measures. Importantly, participants were not compensated for their performance in this phase of the experiment in any way, allowing us to view behavior absent the influence of incentive compensation. 12 To construct our primary dependent variables for the study, we focus on the social-ability measure(s) for which participants received compensation while designing their virtual creature at the beginning of the experiment. For single-measure-compensation participants, we focus on their behavior as it pertains to the sing measure. Accordingly, for participants in this condition, we code each of their individual modification-package choices as either consistent or inconsistent with surrogation with sing. To illustrate, consider the sample pairing presented in Figure 3. This screen displays the following two modification-package choices: Modification #1: sing = 2, dance = 4, charm = 1, and pose = 5 Modification #2: sing = 3, dance = 2, charm = 4, and pose = 2 11 To control for an order effect, we created four orders for the 18 choices (randomly determined) and fully crossed these four order conditions with our three compensation conditions. Results are unaffected by the order in which choices are made; thus, we ignore this variable in subsequent analyses and discussion. 12 The focus of this paper, in part, is to provide empirical evidence of surrogation (as opposed to focusing on potential debiasers ). Compensation tied to this dependent variable may have precluded us from observing the phenomenon of interest in this simple setting. Thus, at the extreme, our results speak specifically to settings where managers must make strategic decisions absent compensation tied directly to optimal decision-making (e.g., when results lag decisions sufficiently to preclude complete contracting). 15

18 For this pairing, a choice of modification package #2 would be coded as 1, because this option maximizes the compensated measure of sing, and is thus consistent with surrogation with sing. In contrast, a choice of modification package #1 would be coded as 0, because this option appears to forego maximizing the compensated measure of sing. For multiple-measure-compensation participants, we focus on behavior as it pertains to the sing, dance, and charm measures. Accordingly, for participants in this condition, we code each of their individual modification-package choices as either consistent or inconsistent with surrogation with sing, dance, and charm. As can be seen in Figure 3, modification package #2 is the surrogation-consistent choice for participants in the multiple-measure compensation condition. The combined levels of sing, dance, and charm are higher for modification package #2 (combined level = 9) than for modification package #1 (combined level = 7). Thus, a choice of modification package #2 (modification package #1) would be coded as 1 (0). Although all participants saw the same 18 creatures with their associated modificationpackage pairings, not all pairings are used to calculate participants surrogation scores. To compute these scores, we sum the coded modification-package selections across the appropriate subset of the 18 pairings (based on the experiment condition), then divide by the total number of relevant pairings. Twelve of the 18 pairings are relevant for evaluating participants tendency to surrogate with sing (i.e., for 6 of the 18 pairings, both modification packages have the same sing level, and thus are not relevant for an analysis of surrogation with sing). A different set of 12 pairings are relevant for evaluating participants tendency to surrogate with sing, dance, and charm. We label the surrogation score based on all applicable pairings the comprehensive surrogation score. Table 1 presents the 18 modification-package pairings, along with a breakdown of the pairings used to compute our dependent variables. 16

19 [Insert Table 1 here] In addition to categorizing modification-package pairings in terms of their relevance for each type of surrogation (surrogation with sing and surrogation with sing, dance, and charm), we further categorize them according to the extent to which participants incur an opportunity cost in making a surrogation-consistent choice. Low opportunity cost pairings are those in which the total level of the four social-ability measures (sing, dance, charm, and pose) is equal between the two modification packages. Thus, a participant who chooses the modification package consistent with surrogation incurs little cost to do so. In contrast, high opportunity cost pairings are those in which the surrogation-consistent modification package has a lower total social-ability level. Thus, a participant who chooses the modification package consistent with surrogation incurs a relatively high cost, because a creature s total social ability level is a key determinant of its ability to achieve the strategic objective of socializing with other creatures. Again, consider the example choice displayed in Figure 3. While modification choice #2 maximizes the sing level (i.e., sing level 2 for modification package #1 vs. sing level 3 for modification package #2), modification package #1 has a higher total social-ability level (i.e., sing, dance, charm, and pose levels sum to 12 for modification package #1, but only sum to 11 for modification package #2). Thus, choosing the surrogation-consistent modification package in this example entails incurring a cost (i.e., foregoes the higher total social-ability). The last column on Table 1 indicates whether there is a low or high opportunity cost to selecting the surrogation-consistent modification package. In summary, we compute three measures of surrogation: comprehensive, low opportunity cost, and high opportunity cost. The calculation of these measures is based on the expectation that if participants surrogate, they will do so on compensated measures. Thus, as described 17

20 above, we compute these three measures based on single-measure-compensation participants propensity to surrogate with sing. Similarly, for multiple-measure-compensation participants, we compute these same three measures based on sing, dance, and charm. Additionally, we compute these measures for flat-wage-compensation participants. When comparing flat-wagecompensation participants behavior to participants in the single-measure compensation condition, we calculate their surrogation scores using the same set of modification-package pairings used to calculate scores for single-measure-compensation participants. Likewise, when comparing flat-wage participants behavior to that of participants in the multiple-measure compensation condition, we calculate their surrogation scores using the same set of modification-package pairings used to calculate scores for multiple-measure-compensation participants. Session Timeline As participants arrived at a session, they were randomly assigned to one of the three compensation conditions. Participants read computer-based instructions and watched a video tutorial on using the Creature Creator module of Spore. Participants then practiced building a virtual creature for 12 minutes. After practicing, participants read additional instructions (including instructions regarding compensation for the study, and the strategy to be implemented in the study), took a short quiz to ensure their understanding of the instructions, and then proceeded to design their virtual creatures in line with the strategy provided. Participants were then presented with the 18 modification-package pairings, and made their selections for each pairing. Finally, participants completed a questionnaire to elicit demographic and additional process-related information. Upon completion of all experiment sessions, a research assistant 18

21 recorded the social-ability measure levels of each creature, and embedded participants creatures into a virtual Spore environment to determine participants final compensation (by timing the creatures ability to socialize with others). Participants were debriefed and compensated after all sessions were completed. Figure 4 provides a timeline for the experiment. [Insert Figure 4 here] Additional Discussion of Experiment Setting and Design Our use of the computer game Spore for our experiment instrument helps accomplish three important goals related to our ability to examine surrogation. First, the setting in Spore is far removed from most participants practical and educational experiences, mitigating the potential impact of participants priors regarding appropriate surrogates for strategic constructs. For example, a setting involving the measurement of customer satisfaction using customer satisfaction survey data would likely have induced participants prior beliefs (or personal experiences) regarding such scenarios. Importantly, responses to a post-experimental debriefing question indicates that none of the 79 participants in this study had ever played the game of Spore before. 13 Second, the portion of Spore we utilize in our study is simple, but still allows participants to get their hands dirty implementing a provided strategy and dealing with various associated measures. Third, in this particular setting there is a clear distinction between the strategic construct (social ability) and related measures (sing, dance, charm, and pose). 14 Therefore, we are able to measure the extent to which participants surrogate the strategic 13 We also asked participants the extent to which they had played video games in general. Generally, participants exhibited very little recent experience playing video games. Experience playing video games is not a significant predictor of any of the results in our paper. 14 Each of the four social-ability measures in Spore captures non-overlapping dimensions of the social-ability construct. This feature allows for a clean test of our hypotheses. That is, relations among measures would introduce other factors to our research design (e.g., a hierarchy among measures, a differential extent of representativeness, etc.) that are beyond the scope of our study. 19

22 construct of social ability with the related measures. In summary, Spore provides us with an ideal task and setting in which to examine our phenomenon of interest. Despite these benefits of using Spore for our study, the internal validity of our measure of surrogation is still potentially threatened. In particular, participants focus on compensated social-ability measures occurs by construction if (1) the strategy statement lacks salience, (2) the compensation scheme signals greater importance of the compensated measure relative to the uncompensated measures, or (3) participants have differential access to social-ability measure information across conditions. We used multiple techniques to rule out these possibilities. First, we explicitly stated the strategy on two occasions, the latter of which occurred immediately before participants began designing their creature. Second, we included in the task instructions an explicit statement that all four measures of social-ability sing, dance, charm, and pose are equally important in implementing the strategy and achieving the goal of global domination. Third, participants completed a quiz on the information from the instructions before designing their creatures. Two questions on this quiz address the statement of strategy and the equal importance of each ability measure. One multiple-choice question reads, "Which of the following will help your creature achieve global domination most rapidly?" Options are "The ability to fly," "Social abilities," "Attack abilities," and "The ability to jump." Participants answering incorrectly had to try again until they gave the correct response ("Social abilities"). A second question reads, "In Spore, which of the following is most important in determining your creature's ability to make friends with other creatures?" Options include sing, dance, charm, pose, and "All socializing abilities are equally important in determining a creature's ability to make friends." Again, participants could not continue until they provided the correct response ("All socializing abilities are equally 20

23 important in determining a creature's ability to make friends"). 15 Finally, we provided all participants with performance measure information for all four social-ability measures in realtime while they designed their creature. Thus, information for all four social-ability measures was equally available to all participants. Before discussing our results, we note that finding evidence supporting both of our hypotheses helps rule out alternative explanations. Recall that H1 predicts an increase in surrogation with compensation on a single measure. Support for this hypothesis is consistent with the possibility of participants being universally unaware of strategy, or that the compensation scheme signaled greater importance of compensated measures relative to uncompensated measures. However, recall that H2 predicts a decrease in surrogation with compensation on multiple measures. Thus, support for this hypothesis directly conflicts with these alternative explanations. IV. Results We begin with an analysis related to the effectiveness of our incentive compensation manipulation. To assess whether or not participants performed the initial design task consistent with the incentives provided, we asked participants to "indicate how much attention you gave to the [sing, dance, charm, and pose] abilities, by allocating 100 points among the four abilities. Allocate more points to abilities to which you gave more attention." We compared the mean number of points allocated either to sing, or to sing, dance, and charm across conditions (reported in Table 2). As expected, single-measure-compensation participants allocated more points (i.e., paid more attention) to the sing measure (mean = 35.1%) than the flat-wagecompensation participants (mean = 22.4%) (t = 3.44; p < 0.01). In addition, multiple-measure- 15 Once a correct response was given on the quiz, the answer was briefly reaffirmed via a pop-up window. 21

24 compensation participants allocated more points to the sing, dance, and charm measures (mean = 86.2%) than flat-wage-compensation participants (mean 81.1%) (t = 1.98; p = 0.03). 16 These findings support the conclusion that our compensation manipulation was successful. 17 We now turn to our hypothesis tests and related additional analyses. [Insert Table 2 here] Tests of Hypothesis 1 Hypothesis 1 predicts that compensation on a single measure of a strategic construct increases managers propensity to surrogate that strategic construct. To test this hypothesis, we examine flat-wage-compensation and single-measure-compensation participants modificationpackage choices. H1 predicts that, relative to flat-wage-compensation participants, singlemeasure-compensation participants will choose modification packages with higher levels of the sing measure, at the expense of other, equally important, social-ability measures. Such behavior would be consistent with surrogation. Further, a strategy of wealth-maximization would not explain results consistent with H1 because participants are not compensated for their modification-package choices. 16 In addition, we examine the social-ability levels of the initial creatures that participants designed. Participants in the single-measure compensation condition designed creatures with a higher level of sing (mean = 3.81) than participants in the flat-wage compensation condition (mean = 2.96) (t = 2.98, p < 0.01). However, participants in the multiple-measure compensation condition (mean = 8.88) did not appear to design creatures with higher levels of sing, dance, and charm (combined) than participants in the flat-wage compensation condition (mean = 8.41) (t = 0.80, NS), although the means are directionally consistent with the expected influence of incentives. This latter result likely stems from parameters inherent to the Spore software which would lead to a small difference between the sum of the sing, dance, and charm levels for a participant maximizing on sing, dance, and charm and a participant maximizing on all four social-ability measures. This constraint is, in part, the reason why we elicited participants self-reported point (i.e., attention) allocations. 17 With respect to participants compensation, total compensation earned by participants did not differ across conditions (F = 0.97, p = 0.39, two-tailed). However, compensation based on participants creature design did differ across conditions (F = 15.01, p < 0.01, two-tailed). Notably, the mean creature-design compensation for singlemeasure-compensation participants (mean = $7.62) was lower than that for the multiple-measure-compensation participants (mean = $8.88) and the flat-wage-compensation participants (mean = $10.00). The rank order of compensation across conditions works against finding support for our hypotheses (i.e., single-measurecompensation participants greater propensity to surrogate is not a result of higher measure-related compensation). 22

25 To test H1, we first calculate the percentage of each participant s modification-package choices that are consistent with surrogation with sing (as described in Section III). Next, we compare this average surrogation score between flat-wage compensation and single-measure compensation conditions. As shown in Panel A of Table 3, participants mean comprehensive surrogation score is 0.59 in the single-measure compensation condition, but is only 0.39 in the flat-wage compensation condition. As shown in Panel B of Table 3, the difference of 0.20 is significant (t = 4.56, p < 0.01). [Insert Table 3 here] These results are robust to differentiating modification-package pairings in terms of the opportunity cost of selecting a surrogation-consistent modification package. Panel A of Table 3 presents participants' high opportunity cost and low opportunity cost surrogation scores. As discussed in Section III, a high opportunity cost pairing is one in which the surrogationconsistent modification package (i.e. the package which maximizes the previously compensated social-ability measure(s)) has a lower total social-ability level than the alternative modification package. This conflict does not exist in a low opportunity cost pairing (i.e., the total socialability level does not differ across modification packages). As noted in Panel B of Table 3, mean high and low opportunity cost surrogation scores are higher in the single-measure compensation condition than in the flat-wage compensation condition (low opportunity cost: t = 4.15, p < 0.01; high opportunity cost: t = 2.49, p = 0.01) Our consideration of high opportunity cost pairings refers to sacrificing a higher overall social-ability (i.e. the total of sing, dance, charm, and pose levels) for higher levels of a particular aspect of social-ability (either the sing level, or the sum of the sing, dance, and charm levels, depending on the compensation condition and hypothesis). However, participants could also have viewed low variance among social abilities as desirable. As such, high opportunity cost surrogation may also be demonstrated by sacrificing low social-ability variance to maximize a particular social-ability (sing, or the sum of the sing, dance, and charm levels). For both hypotheses, untabulated results investigating this possibility are qualitatively similar to those focusing on opportunity cost framed in terms of the total level of social-ability. 23

26 Collectively, these results are consistent with our hypothesis that compensation on a single measure of a strategic construct increases managers' propensity to surrogate the strategic construct with the compensated performance measure. Especially interesting is the robustness of the result when considering choices where surrogating with sing entails incurring an opportunity cost. Tests of Hypothesis 2 Hypothesis 2 predicts that compensation on multiple measures of a strategic construct leads managers to surrogate that strategic construct less than if they are compensated on a single measure. Operationally, this suggests that participants in the multiple-measure compensation condition will exhibit lower surrogation scores than participants in the single-measure compensation condition. 19 In calculating surrogation scores, we focus on the social-ability measures for which participants received compensation while designing their initial virtual creature (see Section III). Thus, for single-measure-compensation participants, we use the same surrogation scores used to test H1, which are computed based on participants surrogation with sing. For multiple-measurecompensation participants, we compute surrogation scores based on choices related to the sum of the sing, dance, and charm measure levels. As shown in Panel A of Table 3, participants mean comprehensive surrogation score is 0.59 in the single-measure compensation condition, but is only 0.46 in the multiple-measure compensation condition. As shown in Panel B of Table 3, the difference of 0.13 is significant (t 19 Note that H2 does not speak directly to the effectiveness with which participants in the multiple-measure setting will implement strategy. Surrogation aside, with attention focused on a more complete picture of the strategy, we would expect managers to be better able to implement the strategy. However, our design allows us to test for surrogation independent of performance implications. 24

27 = 2.15, p = 0.02). These results are robust when considering the low opportunity cost surrogation score (mean single = 0.84; mean multiple = 0.63; t = 3.38, p < 0.01). With respect to the high opportunity cost surrogation score, however, the results are directionally consistent with our hypothesis (mean single = 0.34 vs. mean multiple = 0.29), but not statistically significant (t = 0.58, NS). Importantly, the low opportunity cost choices provide the most conservative test of H2. That is, choices consistent with surrogation are most likely to be made when surrogation is least costly. Thus, our evidence is generally consistent with H2. A concern regarding our tests of H2 is that these tests entail using a different subset of modification-package choices across compensation conditions. That is, the comprehensive, low opportunity cost, and high opportunity cost surrogation scores are computed using subsets of modification-package pairings that differ between the single-measure and multiple-measure compensation conditions. Given this lack of consistency, we ran alternative analyses of H2 using pairings that are used to compute surrogation scores for both of these compensation conditions (Pairings #1-8 in Table 1). Based on these alternative measures, results for the comprehensive and low opportunity cost surrogation scores are inferentially similar to those initially reported (untabulated). For the alternative high opportunity cost surrogation scores, however, we find that the difference between multiple-measure-compensation participants (mean = 0.28) and single-measure-compensation participants (mean = 0.40) is marginally significant (t = 1.35, p = 0.09). Comparison of Flat-Wage Compensation and Multiple-Measure Compensation Conditions As a follow-up to tests of H2, we compare the surrogation scores of participants in the flat-wage compensation and the multiple-measure compensation conditions. To serve as a 25

28 relevant comparison with multiple-measure-compensation participants, we calculate the flatwage compensation condition surrogation scores based on modification-package choices that capture surrogation with sing, dance, and charm (i.e., the compensated measures in the multiplemeasure compensation condition). [Insert Table 4 here] As shown in Panel A of Table 4, participants mean comprehensive surrogation score is 0.46 in the multiple-measure compensation condition, and is 0.40 in the flat-wage compensation condition. As shown in Panel B of Table 4, the difference of 0.06 is not significant (t = 1.16, NS). This result is similar for the low opportunity cost surrogation score (mean flat-wage = 0.63; mean multiple = 0.63; t = 0.02, NS). With respect to the high opportunity cost surrogation score, the difference between the flat-wage compensation condition (mean flat-wage = 0.16) and the multiplemeasure compensation condition (mean multiple = 0.29) is only marginally significant (t = 1.70, p = 0.10). These results are consistent with the tests related to H2. Analysis of Point Allocations To further explore the phenomenon of surrogation, we investigate participants responses to a question designed to elicit their allocation of attention among the four measures of socialability while making modification-package choices. Specifically, participants were asked to "indicate how much attention you gave to the [sing, dance, charm, and pose] abilities, by allocating 100 points among the four abilities, [allocating] more points to abilities to which you gave more attention." Our findings, reported in Table 5, suggest that single-measurecompensation participants tended to allocate more attention to the measure on which they were compensated when designing their creature. Specifically, single-measure-compensation 26

29 participants allocated more points to the sing measure (mean = 33.5%) than the flat-wage compensation participants (mean = 23.7%) (t = 2.52; p < 0.01). In contrast, multiple-measurecompensation participants did not allocate significantly more points to the sing, dance, and charm measures (mean = 83.7%) than the flat-wage-compensation participants (mean 80.0%) (t = 1.57; NS). This latter finding is particularly notable given that participants in the multiplemeasure compensation condition reported allocating more attention to the sing, dance, and charm measures than flat-wage-compensation participants when designing their creatures (i.e., when multiple-measure-compensation participants were provided incentive compensation). Collectively, these results are consistent with our theory. [Insert Table 5 here] To complement this analysis, we also compare participants propensity to allocate equal attention-allocation points to all four social-ability measures. Such behavior would be consistent with the information received regarding the equal importance of each of the four social-ability measures in achieving the strategy of global domination via socialization. The proportion of participants in the single-measure compensation condition who allocated an equal number of points to each measure is 15.4%. Consistent with H1, this proportion is less than the proportion of participants who allocated an equal number of points to each measure in the flat-wage compensation condition (55.6%). The difference of 40.2% is significant (t = 3.29; p = 0.01). Consistent with H2, the proportion of participants in the single-measure compensation condition who allocated an equal number of points to each measure (15.4%) is less than the proportion of participants who allocated an equal number of points to each measure in the multiple-measure compensation condition (34.6%). The difference of 19.2% is marginally significant (t = 1.61; p = 0.06). Finally, the proportion of participants who allocated an equal number of points to each 27

30 measure in the multiple-measure compensation condition is not significantly different than that of the flat-wage compensation condition (t = 1.54; NS). V. Conclusion This paper investigates the influence of incentive compensation on managers' propensity to surrogate strategic constructs with compensated measures. Our findings suggest that compensating managers on a single measure of a strategic construct increases their propensity to surrogate that construct with the compensated measure. Our results further suggest that this propensity to surrogate strategy with compensated measures is less pronounced when compensating managers on multiple measures of a strategic construct. Note that this second finding (related to H2) helps to rule out a number of possible alternative explanations for the first finding (related to H1). For example, results from H1 are consistent with participants believing that, despite their acknowledgement to the contrary during the instructions quiz, compensation tied to social-ability measures implies a greater importance of one measure of social ability relative to another measure of social ability. H1 results could also be driven by participants' desire to be consistent with their earlier behavior (Abelson et al. 1968). Motivated reasoning (see Kunda (1990) and Tayler (2009)), among other psychology-based explanations, could also drive H1 results; participants look for (and find) reasons why surrogation-consistent modification packages, which also happen to be consistent with participants' emphasis when designing their creature, is superior. However, none of these explanations for H1 can explain why participants in the multiple-measure compensation setting are less likely to make surrogation-consistent choices than participants in the single-measure compensation setting. Thus, the results for H2 add to our understanding of the surrogation-related forces behind H1 results. 28

31 Our paper contributes to a variety of research streams, including literature on firms use of strategic performance measurement systems. We highlight and investigate a potentially detrimental response to performance measure information managers tendency to use measures as surrogates of strategic constructs. Our study helps develop academic and practitioner understanding of factors that influence the extent to which firms incur this potential cost (Ittner and Larcker 1998). Further, our study is important to practicing managers and managerial accountants responsible for developing and implementing strategic performance measurement systems. Our study also speaks to an on-going debate regarding the trade-off involved in using performance measures for both decision-facilitating and decision-influencing purposes (Feltham et al. 2006; Ittner and Larcker 2001; Narayanan and Davila 1998; Otley 1994). While the arguments at the center of this debate relate to managers intentional manipulation of performance measure information, our paper focuses on a source of this trade-off that is less strategic in nature. That is, our paper develops an understanding of how basing compensation on measures potentially detracts from the effective use of such information for strategy-related purposes. Further, our paper suggests that basing compensation on a broader set of measures diminishes this problem. Our paper is subject to several limitations. First, while we view the rich, but nonbusiness oriented, task setting as a strength of our design, our choice invites questions of boundary conditions and generalizability. For instance, it may be easier for managers to keep strategy in mind (despite the influence of measure-based compensation) in more familiar environments. Second, we have developed and tested our hypotheses referring to compensation on a single measure or multiple measures. While we operationalize these conditions using 29

32 specific numbers of measures, we appreciate the continuous nature of the number of possible measures that may be used in a strategic performance measurement system. The point of inflection (i.e., the number of measures at which surrogation begins to be mitigated) is likely influenced by a variety of factors, the investigation of which is beyond the scope of this paper. Finally, though our paper draws on a vast and well-established literature in psychology on attention, our measures of attention are self-reported, and thus, potentially very noisy (relative to measures used in psychology). Though we believe the goals of our study necessitated this sacrifice, future research may benefit through more precise measures of attention. While our focus in this paper has been the influence of basing compensation on performance measures, future research could investigate other institutional factors that may inhibit or exacerbate surrogation. Prior research discusses how optimal performance measure weighting is influenced by various measure attributes (e.g., congruence, precision, and sensitivity), which may also influence managers propensity to surrogate a strategic construct with that measure (Feltham and Xie 1994; Krishnan et al. 2005). Further, the extent to which surrogation under multiple-measure compensation is lower than under single-measure compensation may depend on the relationship among the multiple measures (e.g., extent of overlap, causal relationships, etc.). Finally, future research could consider specific implications of surrogation. For instance, to what extent does strategy surrogation influence managers evaluations of employees? Research that addresses such questions would contribute greatly to our understanding of the surrogation phenomenon, as well as the influence of surrogation on the effectiveness of strategic performance measurement systems. 30

33 References Abelson, R. P., E. Aronson, W. J. McGuire, T. Newcomb, M. J. Rosenberg, and P. H. Tannenbaum, eds Theories of Cognitive Consistency: A Sourcebook. Chicago, IL: Rand McNally. Atkinson, A. A., J. H. Waterhouse, and R. Wells A stakeholder approach to strategic performance measurement. Sloan Management Review 38 (3): , R. D. Banker, R. S. Kaplan, and S. M. Young Management Accounting. Upper Saddle River, NJ: Prentice-Hall, Inc. Banker, R. D., G. Potter, and D. Srinivasan An empirical investigation of an incentive plan that includes nonfinancial performance measures. The Accounting Review 75 (1): Bonner, S. E., and G. B. Sprinkle The effect of monetary incentives on effort and task performance: Theories, evidence, and a framework for research. Accounting, Organizations and Society 27 (4-5): Bromwich, M The case for strategic management accounting: The role of accounting information for strategy in competitive markets. Accounting, Organizations and Society 15 (1-2): Burgstahler, D., and I. Dichev Earnings management to avoid earnings decreases and losses. Journal of Accounting & Economics 24 (1): Camerer, C. F Do biases in probability judgment matter in markets? Experimental Evidence. American Economic Review 77 (5): Campbell, D., S. M. Datar, S. L. Kulp, and V. G. Narayanan Testing strategy with multiple performance measures: Evidence from a Balanced Scorecard at Store24. Working paper, Harvard Business School. Chen, C. X., M. Martin, and K. A. Merchant The effects of measurement alternatives on the forward-looking properties of customer satisfaction measures. Working paper, The University of Illinois at Urbana-Champaign, Arizona State University, The University of Southern California. Chenhall, R. H Management control systems design within its organizational context: Findings from contingency-based research and directions for the future. Accounting, Organizations and Society 28: Electronic Arts Spore. Redwood City, CA. 31

34 Epstein, M., and J. F. Manzoni The balanced scorecard and tableau de bord: Translating strategy into action. Management Accounting 79 (2): Farrell, A. M Effects of multiple performance measures, multicollinearity, and task structure on individuals judgment performance. Working paper, The University of Illinois at Urbana-Champaign. Feltham, G. A., R. Indjejikian, and D. Nanda Dynamic incentives and dual-purpose accounting. Journal of Accounting and Economics 42 (3): , and J. Xie Performance measure congruity and diversity in multi-task principal/agent relations. The Accounting Review 69 (3): Hicks, J Value and Capital. Oxford, U.K.: University Press. Ijiri, Y The Foundations of Accounting Measurement: A Mathematical, Economic, and Behavioral Inquiry. Englewood Cliffs, NJ: Prentice-Hall, Inc Theory of Accounting Measurement. Sarasota, FL: American Accounting Association. Ittner, C. D., and D. F. Larcker Innovations in performance measurement: Trends and research implications. Journal of Management Accounting Research 10: , and Assessing empirical research in managerial accounting: A value-based management perspective. Journal of Accounting and Economics 32 (1-3): , and Coming up short on nonfinancial performance measurement. Harvard Business Review (November): James, W The Principle of Psychology. New York, NY: Henry Holt. Kaplan, R. S., and D. P. Norton The Balanced Scorecard Measures that drive performance. Harvard Business Review 70 (1): , and The Balanced Scorecard: Translating Strategy into Action. Boston, MA: Harvard Business School Press., and Having trouble with your strategy? Then map it. Harvard Business Review 78 (5): Krishnan, R., J. L. Luft, and M. D. Shields Effects of accounting-method choices on subjective performance-measure weighting decisions: Experimental evidence on precision and error covariance. The Accounting Review 80 (4):

35 Kunda, Z The case for motivated reasoning. Psychological Bulletin 108 (3): Lambert, R. A Contracting theory and accounting. Journal of Accounting and Economics, 32 (1-3): Langfield-Smith, K Management control systems and strategy: A critical review. Accounting, Organizations and Society 22 (2): Lawton, C Electronic Arts bets big on new game. Wall Street Journal (September 2): B1. Libby, R Accounting and Human Information Processing: Theory and Applications. Englewood Cliffs, NJ: Prentice-Hall, Inc. Luft, J., and M. D. Shields Mapping management accounting: Graphics and guidelines for theory-consistent empirical research. Accounting, Organizations, and Society 28 (2-3): Malina, M. A., and F. H. Selto Communicating and controlling strategy: An empirical study of the effectiveness of the Balanced Scorecard. Journal of Management Accounting Research 13: , and Choice and change of measures in performance measurement models. Management Accounting Research 15: , H. S. O. Nørreklit, and F. H. Selto Relations among measures, climate of control, and performance measurement models. Contemporary Accounting Research 24 (3): Nanni, A., J. Dixon, and T. Vollman Integrated performance measurement: Management accounting to support the new manufacturing realities. Journal of Management Accounting Research 4: Narayanan, V. G., and A. Davila Using delegation and control systems to mitigate the trade-off between the performance-evaluation and belief-revision uses of accounting signals. Journal of Accounting and Economics 25 (3): Otley, D Management control in contemporary organizations: Towards a wider framework. Management Accounting Research 5: Runkel, P. J., and J. E. McGrath Research on Human Behavior: A Systematic Guide to Method. New York, NY: Holt, Rinehart, and Winston, Inc. Schipper, K., and L. Vincent Earnings quality. Accounting Horizons 17 (Supplement):

36 Shiffrin, R. M Attention. In Stevens Handbook of Experimental Psychology, edited by R. C. Atkinson, R.J. Herrnstein, G. Lindzey, and R.D. Luce. New York, NY: Wiley. Simons, R Strategic orientation and top management attention to control systems. Strategic Management Journal 12: Simons, R How new top managers use control systems as levers of strategic renewal. Strategic Management Journal 15: Tayler, W. B The Balanced Scorecard as a strategy-evaluation tool: The effects of responsibility and causal-chain focus. Working paper, Emory University. Webb, R. A Managers commitment to the goals contained in a strategic performance measurement system. Contemporary Accounting Research 21 (4):

37 Figure 1 Linking Strategy to Measures of Performance Strategic performance measurement systems facilitate managers actions by mapping relationships among strategic constructs and performance measures. Above is a simplified example of this mapping based, in part, on the predictive validity framework (Runkel and McGrath 1972; Libby 1981). Horizontal linkages at the Conceptual Level represent the hypothesized cause-and-effect relationships between strategic constructs (also called strategic objectives (Kaplan and Norton 1996)). These linkages form strategy maps (Kaplan and Norton 2000) (also called value driver maps (Ittner and Larcker 2003) or causal chains of performance (Tayler 2009)). Downward links between Conceptual Level strategic constructs and Operational Level measures represent the role of measures as imperfect representations of higher-level strategic constructs. Horizontal linkages at the Operational Level represent testable relationships between measures (based on the hypothesized relationships of the associated strategic constructs at the Conceptual Level). Figure 1 also captures the potential effect of exogenous factors, such as economic conditions, that impact the realization of performance measures. 35

38 Figure 2 Screenshot of Initial Creature Design Above is a sample screenshot of the Creature Creator module of the computer game Spore. The left side of the screen displays various body parts that can be added to the creature. The upperright corner of the screen displays various measures of the creature s abilities. The third row of boxes on the upper right displays the four social-ability measures: sing (level = 2), dance (level = 3), charm (level = 2), and pose (level = 3). During the experiment, these measures were constantly available and were adjusted real-time as participants added and removed parts from their creatures. DNA points, used as currency to purchase parts, are displayed in the bottomleft corner of the screen (currently DNA = 45). Buttons at the top center of screen allow participants to paint their creature and see their creature move in a simple environment. 36