Can accountants use what they know

Activating vs. Using Accounting Causal Relationships: Experimental Evidence on the Decision Performance Effects of Problem Similarity and Comparison

Ella Mae Matsumura 4250C Grainger Hall School of Business University of Wisconsin—Madison Madison, WI 53706-1323 Phone: 608-262-9731 Email: [email protected]

Sandra C. Vera-Muñoz 248 Mendoza College of Business University of Notre Dame Notre Dame, IN 46556-5646 Phone: 574-631-9041 Email: [email protected]

March 2006

We gratefully acknowledge the helpful comments and suggestions of Vicky Arnold, Jon Davis, Karla Johnstone, Bill Kinney, Marlys Lipe, Brian Mayhew, Jeff Miller, Pam Murphy, Dave Ricchiute, Steven Salterio, Lisa Sedor, Tom Stober, Erick Valentine, Alan Webb, and workshop participants at the 2006 Management Accounting Research Conference, the 2005 AAA Annual Meeting, the ABO/AAA Research Conference, and the University of Notre Dame. The authors are indebted to Rafael Muñoz, David Tsui, Bob Williamson, and the undergraduate and graduate accounting students at the University of Notre Dame who participated in the experiment. We are grateful to John Belisle, Mark Bellantoni, Charlie Goines, Lori Lewalski, and Mark Smeraglinolo for their able research assistance. Professor Matsumura acknowledges financial support from the Wisconsin Alumni Research Foundation, and Professor Vera-Muñoz acknowledges financial support by KPMG Peat Marwick LLP through its Faculty Fellowship program.

Activating vs. Using Accounting Causal Relationships: Experimental Evidence on the Decision Performance Effects of Problem Similarity and Comparison ABSTRACT: Accountants tend to overlook relevant principles and causal relationships that are commonly taught in education and experienced in practice, thus hindering their decision performance. We develop and test predictions about the tendency of accountants to omit key causal relationships in a challenging decision setting: recommending revenue-increasing strategies to a client in the midst of a business-model change. We first elicit comparisons of various source problems from some of our participants (the comparison condition), but not from the others (the advice condition). Next, the participants assume the role of a business consultant and recommend ways to improve a client’s revenue in a new setting. The new setting shares with the source problems either (1) a key causal relationship and several literal similarities (the literal similarity condition), or (2) a key causal relationship but no literal similarities (the relational similarity condition). We predict and find that comparison enhances the quality of accountants’ recommendations when the source problems are literally similar to the new setting, and that depriving participants of literal similarities enhances the quality of recommendations when participants do not make comparisons. Our findings show that similarity and comparison help reduce the gap between activating and appropriately using accounting causal relationships in new settings. Additional tests confirm that the findings are not explained by whether a participant is equipped with the key accounting causal relationship. The study has implications for decision making in seemingly familiar but complex accounting settings, such as revenueenhancement or cost-containment decisions following business-model changes. Keywords: causal-knowledge; literal vs. relational similarity; comparison and advice; accounting causal relationships. Data availability: The data are available from the authors upon request

I. INTRODUCTION Accountants possess a collection of accounting principles and “tools” developed from their formal training and experience. Yet, real-world business cases (Gavetti and Rivkin 2005, 2003; Magretta 2002) and empirical accounting research (e.g., Luft and Shields 2001; VeraMuñoz et al. 2001; Vera-Muñoz 1998) show that individuals often fail to spontaneously use relevant principles or causal relationships in new settings, even when they are able to activate (or access) the relevant knowledge from memory. The goals of this study are to develop and test predictions about the tendency of accountants to omit key accounting causal relationships from their analysis of a particular problem, examine the effectiveness of two mechanisms that might reduce this tendency, and, in the end, help decision-makers improve the quality of their decisions. The real-world accounting problem that we examine (providing revenue-increasing recommendations for a client in the midst of a business-model change) is inherently difficult and requires considerable cognitive effort to address. Furthermore, there are no standards or templates to guide problem-solving for our complex accounting problem. Against this background, reasoning by analogy emerges as a powerful cognitive mechanism to enhance decision performance (Holyoak and Thagard 1995; Reeves and Weisberg 1994). Yet prior research shows that it is extremely easy to reason poorly through analogies, and individuals rarely consider how to use them well (Gavetti and Rivkin 2005, 2003; Magretta 2002). In this study we draw upon analogical reasoning research and theories of cognition to examine whether comparison of recalled experiences (Gentner and Namy 1999) and similarity between current and recalled experiences (Gentner and Markman 1997; Goldstone et al. 1991) affect accountants’ consideration of key accounting causal relationships in new and complex accounting settings. The importance of studying the decision-performance effects of comparison and problem similarity in accounting settings is twofold. First, contemporary business cases provide evidence that experienced decision-makers—including accountants—characteristically exhibit difficulty in 1

appropriately using relevant abstract causal relationships in new settings, even when they have acquired such knowledge from previous experiences. For example, mergers and acquisitions sometimes force accountants to deal with expansion of distribution channels and associated changes in cost structures. To illustrate, as a result of General Mills Incorporated (GMI)’s $6 billion acquisition of Colombo Yogurt in 1994, they added a second distribution channel— independent yogurt shops—to their existing impulse location merchants (Horngren et al. 2005, 179; Guy and Saly 2000). 1 Emphasis on the many abstract attributes shared by the two distribution channels (e.g., both sold yogurt, both served similar geographical areas and clientele, and both had salespeople experienced in the foodservice sector) initially led GMI to conclude that they could manage both channels similarly. Thus, they charged the same prices, provided the same promotions and merchandising, and used the same sales force for both channels. However, distribution costs per case of yogurt differed dramatically between the channels because independent yogurt shops typically purchased full pallets of yogurt, while impulselocation merchants bought in less than full pallets. Thus, emphasis on the many literal similarities shared by the two distribution channels initially obscured a relevant causal relationship, namely, that the same product required different marketing and distribution support costs when sold in different channels. Consequently, even though sales dollars for frozen yogurt products were relatively constant, key structural differences between the two distribution channels that initially went unnoticed by GMI’s accountants and managers led to falling profits. Only after GMI’s accountants looked beyond the literal similarities between the two distribution channels and adopted a new activity-based costing approach for marketing and distribution costs were they able to quantify the impact on profits.

1

Examples of impulse location merchants include hospital cafeterias, college dining halls, and convenience stores, among others.

2

Second, studying the decision-performance effects of comparison and problem similarity is important because these factors co-exist naturally in accounting decision-making settings. For instance, accountants make inferences about newly encountered transactions, people, or objects based on their perceived similarity to transactions, people, and objects with which they are familiar. To illustrate, when authoritative guidance for the treatment of an accounting issue does not exist, or when available standards have few or no precise rules, accountants often compare available precedents in accounting situations to identify key relations shared by the precedent and the accounting issue at hand (Nelson 2003, 95; Salterio 1996; Merchant et al. 1993). Yet, current problem-solving theories assume that comparisons are non-automatic, and the evidence supports this assumption (e.g., Holyoak 1985; Lewis and Anderson 1985). Cognitive scientists paint a relatively simple picture of analogical reasoning. An individual starts with a situation to be handled—the new setting. The individual then invokes other problems that she knows well from direct or vicarious experience and, through a process of similarity mapping, identifies one or more settings that, she believes, display similar characteristics. These settings are the source problems. From the sources emerge a specific and well-understood solution strategy that was or should have been adopted for the source problems. The specific solution is then applied to the new setting (e.g., Bassok et al. 1998; Gentner et al. 1997; Sherwin 1999; Kolodner 1993). In contrast to the simplified picture portrayed above, the objective of our study is not to demonstrate how best to teach accountants to use analogical reasoning to learn from source problems a new specific solution that could then be applied to a similar problem. 2 Instead, the objective of our study is to demonstrate how comparison and problem similarity affect accountants’ ability to reason by analogy to facilitate development of a mental model—for a new setting—that incorporates a relevant accounting cause-and-effect relationship (hereafter causal 2

For example, see Hanson and Phillips 2006; Gentner et al. 2003; Thompson et al. 2000; Loewenstein et al. 1999.

3

relation) which they already know. 3 Thus, we simulate a real-world accounting setting where accountants are prompted with partially understood source problems (i.e., their solution strategies are neither provided nor elicited) that share with the new setting a relevant accounting causal relation (discussed below) (Gentner et al. 2003, 394; Kurtz et al. 2001; Ferguson and Forbus 1998). To our knowledge, this is the first accounting study to document the effects of comparison and problem similarity on accountants’ recommendation quality in a complex accounting decision setting. Prior to the experiment our participants had received formal cost accounting training on the relevant subject matter of our setting, activity-based costing and management (ABC/ABM). Thus, by choosing an accounting issue our participants are familiar with, we are able to disentangle two different components of performance: 1) activation of the relevant causal relation; and 2) appropriate use of the relevant causal relation. Such separation is needed because successful complex decision performance requires a number of components, which are not necessarily acquired simultaneously (Ross and Kennedy 1990). For instance, the relevant knowledge might have been acquired from prior multiple exposures to similar experiences. Importantly, in our study, we use accountants’ recommendations in a complex managerial accounting setting to gauge whether they appropriately use the relevant causal relation. To our knowledge, this is the first accounting study to empirically disentangle activation of the relevant causal relation, which we assume will occur independent of our experimental treatments, and use of the relevant causal relation in a new setting. In our experiment we first either explicitly elicit comparison of various source problems, or do not elicit such comparison and instead ask participants to provide advice for each source problem separately. Eliciting comparison of the source problems involves requiring participants

3

A mental model of a problem is a framework or knowledge representation that contains relevant abstract principles or relationships (which may be causal) to guide problem-solving (Holland et al. 1986, 12).

4

to identify similarities and differences between the problems, and to state an overall main principle that captures the essence of the problems taken together. Next, participants assume the role of a business consultant and are prompted with a new accounting setting that requires them to recommend revenue-increasing strategies to a client in the midst of a business-model change. We vary the similarity between the source problems and the new setting as either literal or relational. We preface our experiment in order to describe literal and relational similarities in the context of our study. To illustrate a literal similarity, consider two different hypothetical companies that share a cost attribute: each manufactures a single, “low-cost-to-make,” standard product. Assume further that the companies share an abstract causal relation: their “low-cost-tomake” products cause low manufacturing overhead expenses. As described, these companies are said to be literally similar because they share both an abstract cost attribute and an abstract causal relation. To illustrate a relational similarity, consider the following two companies. One is a manufacturing company with a product mix consisting of a high proportion of “unique” products relative to their “standard” products, and the other is an online apparel retailer with a customer mix consisting of a high proportion of “high-cost-to-serve” customers relative to their “low-costto-serve” customers. These two companies share a relevant abstract causal relation: the manufacturer’s high ratio of unique products relative to its standard products is causally related to high manufacturing overhead expenses in the same way that the online retailer’s high ratio of high-cost-to-serve customers relative to low-cost-to-serve customers is causally related to high selling and administrative expenses. 4 However, the companies share only few or no concrete or

4

This description is extracted from real-world examples of companies that are increasingly discriminating between their profitable and unprofitable customers. For instance, a recent Associated Press article refers to unprofitable customers as those who “tie up a sales worker but never buy anything, or who buy only during big sales. Or customers who file for a rebate, then return the items.” Brad Anderson, chief executive of Best Buy, indicates that “What we are trying to do is not to eliminate those customers, but just diminish the number of offers we make to them.” The article also describes an investment firm that found that one customer with a

5

abstract attributes. Thus, these companies are said to be relationally similar (Chen 1996; Gentner et al. 1993). We first obtain a measure of activation of the relevant causal relation by eliciting participants’ responses to two questions presented after the participants are exposed to the experimental treatments but before they provide their recommendations for the client in the new setting. We expect that activation of the relevant causal relation will occur independent of our experimental treatments, and we conduct tests to validate this assumption. Next, we use accountants’ recommendations for the client in the new setting to gauge their use of the relevant causal relation. We predict and find that accountants who are first asked to compare the source accounting problems before making their recommendations for the client in the new setting use more correctly the relevant causal relation when the source problems and the new setting are literally similar. Further, we predict and find that depriving participants of literal similarities between the source problems and the new setting enhances recommendation quality when they do not first compare the source problems. Additional tests confirm that the findings are not explained by whether a participant is equipped with the key accounting causal relationship. By examining the decision performance effects of problem similarity and comparison, we show that these mechanisms help reduce the gap between activating accounting causal relationships and appropriately using them in new accounting settings. We organize the remainder of this paper as follows. In Section II we use cognitive psychology research to derive predictions for the effects of comparison of source problems and problem similarity between source problems and a new setting on accountants’ recommendation quality. We describe our experiment and report our results in Sections III and IV, respectively.

portfolio of $500,000 was tying up three financial advisers almost full-time with requests for help and information. In this case, the firm asked him to go elsewhere (Freed 2004).

6

Section V concludes, discusses implications, and provides some directions for future research. II. HYPOTHESES DEVELOPMENT Comparison and Problem Similarity in Analogical Reasoning Experimental work and field evidence strongly suggest that the effectiveness of analogical reasoning depends largely on the individual’s ability to capture deep structural features and causal relations—as opposed to superficial features—from previous problem situations (Thagard 1996; Holyoak and Thagard 1995; Reeves and Weisberg 1994; Ross and Kennedy 1990). Two factors that are critical to the analogical reasoning process are explicit comparison of the source problems, and similarity between the source problems and the new setting. Comparison of the Source Problems Psychology research shows that explicit comparison of source problems—including those that are partially understood—facilitates deeper understanding of the problems, thus making key components of meaning more salient and relations shared by the problems more evident (Gentner et al. 2003, 394; Kurtz et al. 2001; Ferguson and Forbus 1998). For instance, recent negotiation studies with graduate business students, business executives, and consultants (Gentner et al. 2003; Gentner et al. 1997; Thompson et al. 2000; Loewenstein et al. 1999) find that participants who are asked to compare two short cases (i.e., the source problems), and to derive from them an overall principle are three times more likely to successfully apply a common solution strategy (e.g., contingent contracts) to a subsequent face-to-face negotiation task, relative to participants who received the same cases separately and were not asked to compare them. Other studies find that explicit comparison helps develop deeper understanding of principles or relations (Catrambone and Holyoak 1989) and more differentiated knowledge

7

structures (Schwartz and Bransford 1998). 5 To ensure that the comparison process enables capturing deep structural features and causal relations, it must include a search of both similarities and differences. While similarities usually spring to mind quickly, the process of searching actively for differences rarely comes naturally—it is often thwarted by the confirmation bias (Hamilton 1979; Hastie and Kumar 1979; Einhorn and Hogarth 1978; Wason and Johnson-Laird 1972). That is, people are better equipped to seek out information that confirms their beliefs than to ignore contradictory data that challenge them, even when they have no vested interest in the beliefs. Confirmation bias affects decision-makers after they have adopted an analogy, especially a superficial one. For instance, as illustrated earlier with the GMI example, decision-makers may continue to act on superficial analogies for a long time, ultimately hurting the company’s profits. Similarity Between the Source Problems and the New Setting When the decision-maker encounters a new problem situation to be addressed (i.e., the new setting), she develops a mental model that captures the problem’s salient features. She then mentally scours other settings with which she is familiar, due either to direct or vicarious experience, and identifies source problems that display similar salient features. This step represents the starting point for a local search process that requires the decision-maker to pay attention to select features and abstract causal relations of the source problems for potential application to the new setting. In general, however, individuals tend to not spontaneously remember previous experiences that share with the new setting only relevant abstract principles or relations which

5

For further examples, see Hanson and Phillips (2006); Gregan-Paxton and Cote (2000); Gentner and Namy (1999); Gentner and Medina (1998); and Forbus et al. (1995). All of these studies, however, illustrate the standard source-to-new setting matching paradigm. That is, participants in these studies receive intentional instructions that make them explicitly aware (whether intentionally or unintentionally) that their memory for the source problems and their common solution strategy would be subsequently tested in a new setting.

8

they would value in hindsight (Reeves and Weisberg 1994). Instead, individuals tend to remember previous experiences that may share irrelevant problem features (Gentner et al. 1993). Thus, central to the analogical reasoning process is the role played by the similarity, literal or relational, between the source problems and the new setting. Literal similarity is said to occur when the source problems and the new setting match in terms of both attributes (i.e., properties of objects, both concrete and abstract) and relations (i.e., links that define the causal relationship between two or more attributes when considered in combination). In contrast, relational similarity is said to occur when the source problems share a relevant abstract principle or relation with the new setting, but few or no concrete or abstract attributes (Medin et al. 1993; Tsoukas 1993, 337; Holoyoak and Thagard 1989; Gentner 1983, 159). Research Hypotheses Recall that our objective is not to demonstrate how best to teach accountants to use analogical reasoning to learn from source problems a specific solution that could then be applied to a similar problem. Instead, the objective of our study is to show how comparison and problem similarity affect accountants’ ability to develop a mental model for a new setting that incorporates a relevant accounting causal relation that they already know. To guide and organize our discussion, we summarize graphically in Figure 1 the constructs examined and our research hypotheses. [Insert Figure 1 here] Based on findings from the lines of research on comparison and problem similarity, we argue that the incremental benefit of comparison of source problems on accountants’ recommendation quality is better understood when examined in juxtaposition to the similarity between the source problems and the new setting, and vice-versa (Ross and Kennedy 1990). On one hand, literal similarity matches between source problems and the new setting are highly accessible, as they can be indexed in memory by attributes or object descriptions, by relations, or 9

by both. However, literal similarity is not very useful for developing a mental model of the new setting that includes a relevant causal relation because there is too much overlap between and among the problems’ attributes and relations to discern what is relevant. On the other hand, source problems and new settings that are relationally similar facilitate the development of mental models that include relevant common principles or relations (Gentner et al. 1993; Clement and Gentner 1991; Holyoak and Koh 1987, experiment 2). This is because the shared features of relationally-similar problems are sparse enough to allow discerning what is relevant, and thus, make more salient the embedded structural relations common to a class of problems (Ross and Kennedy 1990). When prompted with a new setting, accountants who are first asked explicitly to compare various source accounting problems will develop an appropriate mental model of the new setting that incorporates a relevant accounting causal relation, leading to enhanced recommendation quality. We argue that this comparison effect should occur only when the source problems and the new setting are literally similar (i.e., the source problems share a relevant causal relation with the new setting and maintain a one-to-one attribute match). That is, there should be no effect of comparison on accountants’ recommendation quality when the source problems and the new setting are relationally similar (i.e., the source problems share a relevant causal relation with the new setting but few or no attributes). This is because, as discussed above, relationally-similar problems should make the relevant causal relation salient, so that even accountants who are not asked to compare the source problems (e.g., instead they are asked to provide advice for each problem separately) should be able to develop an appropriate mental model of the new setting that incorporates the relevant accounting causal relation. To test these arguments, we propose the following hypotheses:

10

H1a:

When prompted with a new accounting setting that is literally similar to source accounting problems (i.e., they share a relevant causal relation and maintain a one-to-one attribute match), accountants who are first asked to compare the source problems will use more correctly a relevant causal relation in the new setting when making accounting recommendations, relative to accountants who are not asked to compare the source problems.

H1b:

When prompted with a new accounting setting that is relationally similar to source accounting problems (i.e., they share a relevant causal relation but few or no attributes), explicit elicitation of a comparison of the source problems will not be related to accountants’ ability to correctly use a relevant causal relation in the new setting when making accounting recommendations.

Further, as discussed earlier, extracting relevant common features requires a step beyond creating a temporary match between seemingly different problems: individuals must develop a mental model for the new setting that incorporates only relevant abstract principles or relations, and that is devoid of specific attributes and idiosyncratic details (Gentner et al. 2003, 394; Boland et al. 2001; Chen 1996; Collins and Burstein 1989; Johnson-Laird 1989; Gentner 1983, 1989; Gentner and Toupin 1986). Simply stated, abstract mental models should offer more flexible and encompassing ways of addressing new problems (Whittlesea 1997; Ross and Kennedy 1990; Vallacher and Wegner 1987). Application of problem similarity research to our accounting setting suggests that source problems that are relationally similar to the new setting (i.e., the source problems share a relevant causal relation with the new setting but few or no attributes) offer an effective mechanism for the development of a mental model of the new setting that incorporates a relevant causal relation. This relational similarity effect should occur only when a comparison of the source problems is not elicited. There should be no effect of problem similarity on recommendation quality when comparison of the source problems is elicited first. This is because, as discussed above, comparison “actively” informs accountants as to which deep structural features are causally relevant, so that even accountants who are prompted with source problems that are literally similar to the new setting should be able to develop an appropriate mental model of the new 11

setting that captures the relevant causal relation. To test these arguments, we propose the following hypotheses: H2a:

When a comparison of source accounting problems is not explicitly elicited, accountants who are prompted with a new accounting setting that is relationally similar to the source problems (i.e., they share a relevant causal relation but few or no attributes) will use more correctly the causal relation in the new setting when making accounting recommendations, relative to accountants who are prompted with source problems that are literally similar to the new setting (i.e., they share a relevant causal relation and maintain a one-to-one attribute match).

H2b:

When a comparison of source accounting problems is explicitly elicited prior to making a recommendation for the new accounting setting, the similarity between the source problems and the new setting will not be related to accountants’ ability to correctly use the causal relation in the new setting when making accounting recommendations. III. METHOD

Overview Consistent with our theoretical development, we constructed a between-subjects experiment to test our hypotheses (depicted graphically in Figure 1) in order to focus on how the following two factors affect the quality of accounting recommendations in the new setting: (1) whether accountants are explicitly asked (or not asked) to compare the source accounting problems before making a recommendation for the new accounting setting; and (2) the similarity (literal vs. relational) between the source problems and the new setting. We hold constant the total amount of information provided to rule out information overload as a potential explanation for our results (Turetken and Sharda 2004; Hwang and Lin 1999; Chewning and Harrell 1990; Schick et al. 1990). We validate our assumption that, owing to their formal training on the underlying accounting issue of the experimental setting (i.e., ABC/ABM), participants’ activation of the relevant causal relation upon exposure to the source problems is independent of our treatment variables.

12

Participants Twenty-three Master of Science in Accountancy students enrolled in a graduate strategic cost management course in a Public Accounting Report Top 5 graduate accounting program at a private university, and ninety upper-level undergraduate business students enrolled in a cost accounting course in a Public Accounting Report Top 5 undergraduate accounting program at the same university (PAR 2004, 5) participated in the experiment in their classrooms. We assigned the graduate and undergraduate students randomly to the four experimental conditions. No one experimental condition contained disproportionately more undergraduate or more graduate accounting students. We purposely chose to use these subjects as participants because prior to the experiment, they had completed in-depth discussions (of homework problems, cases, and current business articles) and formal testing (i.e., exams and quizzes) of the key underlying subject matter of our experiment, namely, activity-based costing and management (ABC/ABM). This is important to ensure a homogeneous group of participants that share a common body of declarative (factual) ABC/ABM knowledge. Other potential subject groups include employees of a particular firm that uses ABC/ABM, employees of a particular cross-section of ABC/ABM firms, or people who have experience in the use of ABC/ABM. However, because these groups’ own experiences with specific ABC/ABM applications may differ from the situation presented in our experimental materials, focusing on any of these groups from firms that use different ABC/ABM applications would compromise experimental control. 6

6

See also Lipe and Salterio (2000) for a discussion of advantages and disadvantages of using experienced participants for experiments.

13

The total sample used for our analyses consists of 110 participants. 7 Participation in the study was voluntary. We conducted the study in two consecutive semesters during a regularly scheduled 75-minute class period, with one of the experimenters present in the classroom during the entire duration of the experiment. One of the researchers was the instructor for 87 (79 percent) of the participants, and a second instructor was in charge of 23 (21 percent) of the participants. Out of the 110 participants, 93 (84 percent) were accounting majors, 15 (14 percent) were finance majors, and 2 (2 percent) were completing majors in marketing or business-science. Seventy-four participants (67 percent) were males, and 36 (33 percent) were females. The average age of the participants was 21 years (s.d. = 1.07), and their accounting experience ranged from 0 to 60 months, with a mean of 4.31 months (s.d. = 8.00). To rule out systematic differences in the participants’ accounting program membership (graduate vs. undergraduate), instructor, major, gender, age, and accounting experience, we included these variables as covariates in our planned contrasts test (discussed below). Our results show that none of these variables affected the dependent variable at conventional levels; thus, they are not discussed further. The participants received a flat compensation consisting of 10 credit points (10 percent) towards their total final exam points for the cost-related course they were enrolled in. In addition, to encourage participants to exert high cognitive effort, they were told they could earn a monetary bonus whose magnitude hinged on their responses to the materials. The average monetary compensation was $7.13 (s.d. = $1.24), ranging from a minimum of $4.00 to a maximum of $9.00.

7

The responses of two undergraduate students to a post-experimental recall exercise were not sensible, and were dropped from all analyses. We also dropped a participant who did not complete an activity-based-costing test. The results remain qualitatively the same when these participants are included in the sample.

14

Procedures, Case Materials, and Tasks To guide and organize our discussion, we summarize graphically in Figure 2 our experimental procedures and tasks. We collected the data in two parts: the experiment (part one), and the post-experiment (part two). In part one, participants first received various source accounting problems that were used for the comparison variable manipulation, and later, for the problem similarity manipulation (discussed below). Participants assumed the role of a business consultant and were prompted with a new accounting setting that required them to recommend revenue-increasing strategies for a client in the midst of a business-model change. We also collected data to construct our dependent variable, in addition to a moderator variable for hypotheses testing (discussed below). [Insert Figure 2 here] We provided the experimental materials in three serially numbered envelopes, two for part one (the experiment), and the third one for the post-experimental questionnaire. We instructed the participants to open one envelope at a time, write down their starting and ending times for each task, proceed through the tasks at their own pace, and put all the materials back in the envelope before proceeding to the next envelope. A key feature of the experiment is that participants examined the source problems without being made aware, either explicitly or implicitly, of the nature of the subsequent task (related to the new setting) or of any relationship between the two tasks. The first envelope contained three pages, including a cover sheet with instructions, a sheet containing three short case narratives (one for each of three companies), and a response sheet for either the comparison or advice manipulation. Each case narrative consisted of a threeparagraph description of a company’s business, products, customers, and operating results for a one-year period; the narratives appeared in three adjacent columns of equal length (one column for each company). To avoid potential order effects, we counterbalanced the order of 15

presentation of the three companies on the page so that each company appeared in each of the three possible column positions (i.e., left, middle, or right) with equal frequency. The counterbalancing made use of six different patterns for each of the four experimental conditions, for a total of 24 unique combinations. The patterns yielded no significant differences, and are not discussed further. The second envelope contained two pages that included a three-paragraph case narrative about the new setting, Business Forms Printing (hereafter BFP), and three questions related to the case (discussed further below). On average, the participants completed part one (materials in the first two envelopes) in 34.45 minutes (s.d. = 5.60). 8 In part two (the post-experimental part) we asked participants to complete a debriefing questionnaire eliciting their perceptions of task difficulty and realism, familiarity with the companies portrayed in part one, and questions regarding academic background and accounting-related work experience. In addition, this section included questions intended for validation checks (described further below). Independent Variables Comparison versus Advice We first manipulated whether participants are explicitly asked (or not asked) to compare various source problems. We use three companies as source problems because prior studies suggest that principle (or causal relation) abstraction generally does not occur unless several source exemplars are provided (Reeves and Weisberg 1994; Ahn et al. 1992; Gick and Holyoak 1983, Experiments 1-3). One of the source companies (Mega Outfitting Company, MOC) operates in the mail order clothing business; the second company (University T-shirt Printing, UTP) is a college T-shirt printing shop, and the third one is either a tailgate-party catering business (Tailgate Party Catering, TPC) or a corporate catering business (Lambert’s Corporate

8

We included time to complete part one as a covariate in our planned contrasts test (discussed below). The results are not significant at conventional levels; thus, they are not discussed further.

16

Catering, LCC), depending on the problem similarity manipulation (discussed next). As shown in panel A of Appendix 1, we asked participants in the comparison condition to list as many similarities, differences, and combinations of similarities-differences as they could identify between and among the three source companies. To avoid ambiguities in interpreting the instructions, we provided participants with an example of a comparison in a non-accounting context. In addition, we asked these participants to articulate in one sentence the overall main principle that captures the essence of the three cases taken together. As shown in panel B of Appendix 1, we asked participants in the advice condition to recommend, for each of the three source companies separately, two possible ways in which the companies could increase their profits next year relative to the current year. Because we purposely designed the case narratives of the source companies such that they would be either literally- or relationally-similar to the new setting, we provide further details of the relevant accounting causal relation next. Literal versus Relational Similarity Our participants assume the role of a business consultant and are prompted with a new accounting setting that requires them to recommend revenue-increasing strategies for a client in the midst of a business-model change. As shown in the narrative in Appendix 2, the company (BFP) in the new setting is described as a printing shop that provides business forms for various domestic and international clients, and has recently introduced a new inventory management service in an effort to increase profits. Recall that the three source problems are either literally similar or relationally similar to the company in the new setting. Panels A and B of Appendix 3 outline the literal and relational similarity manipulations, respectively. In the literal similarity condition, the source problems share (implicitly) with the new setting a relevant causal relation, and maintain a one-to-one match of several abstract attributes. Recall that an attribute could be either concrete or abstract. As shown in panel A of Appendix 3, in the literal similarity condition the three source problems and the new setting match one-to-one 17

regarding four abstract attributes: (1) homogeneous products; (2) one-year increase in the number of customers; (3) increase in the ratio of “high-cost-to-serve” to “low-cost-to-serve” customers (hereafter high-cost and low-cost customers, respectively); and (4) a one-year decrease in profits. For all the source problems and for the new setting, sales revenues increased and prices did not change. The terms “homogeneous,” “heterogeneous,” “high-cost,” and “lowcost” customers were not used in the case narratives. Instead, we designed the customers’ descriptions to make them implicitly consistent with these terms. The high-cost customers impose more demands upon the companies’ resources— ultimately reflected in the selling and administrative (S&A) expenses—than the low-cost customers. For instance, an excerpt of the narrative for UTP describes its high-cost customers as follows: 9 UTP’s customers cannot accurately estimate the annual demand for shirts. UTP has therefore agreed to print and deliver additional reorders (min. = 2,500; max = 10,000) within 10 days. These reorders require UTP employees to handle extra production scheduling and expedited deliveries. UTP accepts returns of defective shirts, and of shirts that their customers are unable to sell in a given year. Although most of the orders are delivered by truck twice a year, UTP delivers reorders several times during the year. UTP keeps in storage orders for their smaller clients until they are needed.

Despite having an increasingly heterogeneous customer mix, the source problems and the company in the new setting use a traditional (volume-based) cost accounting system to allocate their S&A costs for pricing purposes. That is, the source problems as well as the company in the new setting compute the S&A rate based on estimated cost of goods sold at the beginning of the year. Although not mentioned in the case materials, using ABC would be more appropriate than using volume-based costing, because ABC would allow the source problems and the company in

9

This description mirrors many real-life examples. For instance, a recent Wall Street Journal article indicates, “[R]etailers accustomed to selling clothes in their stores have been vexed by the complexities of Internet distribution and customer merchandise returns. Wal-Mart Stores, Inc., which sells more apparel than any other retailer in the nation, pulled clothing off its Website in 2001. While it was one of the more popular online categories, high handling costs made it infeasible” (Merrick 2003, B2).

18

the new setting to account for the differential costs of serving their heterogeneous customer mix, and thus, price their products differentially (Atkinson et al. 2004, pp. 145-146; Hilton et al. 2003, pp. 220-221; Blocher et al. 2002, pp. 833-835; Kaplan and Cooper 1998, 191). This explains why, even though the companies’ customers and sales revenues have increased, their profits have decreased. Hence, the relevant accounting causal relation shared by the source problems and the new setting can be articulated as follows: even in the face of increasing customers and revenues, an increase in the ratio of high-cost to low-cost customers causes low profits if the company fails to price products differentially to reflect the differential S&A costs of serving an increasingly heterogeneous customer mix. Panel B of Appendix 3 outlines the relational similarity condition. In this condition the source problems share (implicitly) with the new setting the relevant causal relation discussed above. However, the source problems and the new setting do not maintain a one-to-one attribute match. As shown in panel B, we retained the industry categories of the source companies but changed some of the attributes as needed. Two of the source companies, Mega Outfitting Company (MOC) and University T-shirt Printing (UTP), had the same names as those in the literal similarity condition. The third source company is comparable to the literal similarity condition’s Tailgate Party Catering (TPC), but differs in name because appropriately changing the abstract attributes suggested a corporate catering business. Consequently, we renamed the company Lambert’s Corporate Catering (LCC). The three source companies use a traditional (volume-based) cost accounting system to allocate their S&A costs for pricing purposes, and LCC and UTP use a traditional cost accounting system to allocate production overhead. Panel B of Appendix 3 shows that the two source companies that experienced either an increase in either the ratio of high-cost to low-cost customers (MOC) or in the ratio of high-costto-make to low-cost-to-make products (LCC) experienced a one-year decrease in profits. It follows that ABC would be more appropriate for MOC’s customer costs and LCC’s product 19

costs, allowing these companies to account for the differential costs of serving (providing) their increasingly heterogeneous customers (products) and thus, price their services (products) differentially. This explains why, even though LCC’s and MOC’s sales revenues and/or customers have increased, their profits have decreased. In contrast, UTP is provided as a counterexample: because UTP has both homogeneous products and customers, it experienced a one-year profit increase due to increasing customers and revenues. For instance, an excerpt of the narrative for UTP describes its “high-profit” customers as follows: UTP’s two customers are able to estimate the annual demand for shirts with 99% accuracy. Therefore, UTP has not encountered requests of additional reorders or expedited (overnight) orders. UTP has agreed to deliver by truck half of the ordered shirts to both the University Bookstore and the retail chain store by the end of April, with the rest of the shirts delivered by the end of June. Finally, UTP’s two customers have been able to sell all the shirts ordered in a given year; thus, UTP has not encountered returned orders. However, UTP accepts returns of defective shirts (UTP pays actual shipping charges). Dependent Variable The dependent variable is constructed from participants’ recommendations as to how to enhance revenues for BFP (the company in the new setting). This is consistent with prior analogical research that uses a new-setting solution as an indirect method of gauging participants’ application of an abstract principle or relation (e.g., Thompson et al. 2000; Reeves and Weisberg 1994). In particular, we asked participants to recommend two possible ways in which BFP could increase its revenues, and to discuss the issues that BFP must consider as they attempt to increase revenues. We use the explanations provided by each participant to code each of their two recommendations as revenue-increasing: customer mix (coded = 2), revenueincreasing: other (coded = 1), or neither (coded = 0). A response was coded as “revenueincreasing: customer mix” if it suggested using differential pricing of products and/or services to address the increasingly heterogeneous customer mix. That is, these recommendations reflect appropriate use of the causal relation articulated earlier. For instance, these responses indicated 20

that BFP should modify the mix of customers (e.g., “BFP could attract more customers, but would want to make sure they attract the right customers (few, high-quantity orders, good prediction of needs) because increased revenue is meaningless if you increase costs by the same amount”); or BFP should increase their prices selectively (e.g., “Charge expensive customers more for the business forms”); or BFP should increase prices and/or quantities of business forms sold (e.g., “Charge customers a fee for non-standard quantities shipped;” “BFP can either increase the prices for small orders, or impose a minimum quantity per order”). A response was coded as “revenue-increasing: other” if it entailed recommendations that increase revenues or decrease costs, but otherwise do not address (directly or indirectly) BFP’s heterogeneous customer mix (e.g., “Expand the product line or services offered;” “Reduce manufacturing overhead”). Any recommendations that were unrelated to the main issues described above were coded as “neither” (e.g., “Hire a consultant;” “Increase advertising”). The minimum score for our dependent variable (based on the participants’ two recommendations) is 0 and the maximum score is 4. Two trained research assistants blind to the experimental conditions independently coded the participants’ responses, 10 then met to resolve differences in coding. Is Activation of the Relevant Causal Relation Independent of Experimental Treatments? We chose an accounting issue our participants are familiar with due to their formal cost accounting training on ABC/ABM. Thus, activation of the implicit key causal relation (i.e., changes in the ratio of a company’s heterogeneous to homogeneous customers or products may potentially cause changes in its revenues and profits) should be independent of our treatment variables. In this section we provide descriptive statistics and results of statistical tests to validate the assumption that participants activated the relevant causal relation upon exposure to the three

10

The Cronbach’s alpha coefficient, a measure for inter-rater agreement over and above that expected by chance, is 0.710, comfortably above the minimum 0.60 requirement (Nunnally 1978; Cronbach 1951).

21

source problems and to rule out that our experimental treatments played a role on this activation. We asked our participants two questions immediately after they read the case narrative for BFP (the company in the new setting), but before they provided their revenue-enhancing recommendations (see Figure 2). We elicited responses to these two questions ex post because doing so ex ante (i.e., before exposing the participants to our treatments) would have risked experimental demand effects. 11 The first question (“a” in Appendix 2) asked participants to indicate whether they expected BFP’s 2002 profit to increase (coded = 1), decrease (coded = –1), or remain unchanged (coded = 0) compared to BFP’s 2001 profits. We also asked participants to provide a rationale for their answers. The second question (“b” in Appendix 2) asked participants to evaluate BFP’s cost accounting system: “Is BFP’s cost system appropriate for decision making? If it is, then explain why. If it is not, then explain why not and how BFP should change it.” We coded the participants’ responses to question “b” as superior (coded = 2), average (coded = 1), or inferior (coded = 0). A response was coded as “superior” if it indicated that the cost system was not appropriate and suggested the need to use cost drivers to account for the costs of customers’ increasing activities, or if it pointed out that there was a problem with the current method of computing the S&A rate, and provided logical insights as to what the problem was. A response was coded as “average” if it indicated that the cost system was not appropriate and noted the increasing variety of customers, or the problems with the current S&A rate, but did not address specifically the need to use cost drivers, as described above. A response was coded as “inferior” if it indicated that the current system was appropriate for decision making and the answer focused only on the product (which was homogeneous, and thus, not the focus of the 11

This is similar to Vera-Muñoz (1998), who measures ex post a variable that captures the participants’ maximization objectives when making resource allocation decisions. Explicitly asking participants ex ante (i.e., before they were exposed to the context manipulation) whether their objective would be to maximize either cash flows or net income would have risked demand effects. Vera-Muñoz reports that the participants’ maximization objectives were not affected by either the treatment variable (context) or by their accounting knowledge.

22

problem), as opposed to focusing on the increasing variety of customers, or if the answer was vague and did not provide any meaningful insights. Our two trained research assistants independently coded the participants’ responses, 12 then met to resolve differences in coding. Table 1 shows the frequencies and means of participants’ cost accounting system assessments scores (question “b”) cross-classified by their profit trend forecasts (question “a”). The table shows that 88 out of the 110 participants (80 percent) correctly predicted that BFP’s profits would decrease. A correlation analysis reveals that participants’ responses to the two questions are marginally negatively correlated (Spearman’s rho = –0.176; p = 0.066, two-tailed). This result indicates that participants who correctly forecasted a decrease in BFP’s profits for the forthcoming year (question “a”) received higher scores, on average, in their response to question “b.” ANOVA results (not shown) support our assumption that participants’ profit trend forecasts are not affected by our experimental treatments (p = 0.188 for literal vs. relational similarity; and p = 0.335 for advice vs. comparison). Similarly, ANOVA results (not shown) support our assumption that participants’ cost accounting system assessment scores are not affected by our experimental treatments (p = 0.584 for literal vs. relational similarity; and p = 0.344 for advice vs. comparison). [Insert Table 1 here] Covariate In this study, our problem similarity manipulation refers to differences between the source and target companies. However, it is possible that within source-company differences may potentially affect decision performance. Recall that, across our two problem similarity conditions, participants were exposed to different sets of source companies. That is, the relational similarity condition includes two companies (MOC and LCC) that have heterogeneous customers or products, and experienced a one-year profit decrease. In contrast, the third source 12

The Cronbach’s alpha coefficient is 0.728.

23

company (UTP) is included as a counterexample. Because UTP has both homogeneous products and customers, it experienced a one-year profit increase due to increasing customers and revenues. To rule out within source-company differences as an alternative explanation for our results, we constructed a covariate from participants’ responses to a post-experimental question that elicited which of the three companies, if any, was the most helpful to them in providing BFP’s profit trend forecast, or whether all three companies were similarly helpful. We coded the participants’ responses as 1 = MOC was the most helpful; 0 = otherwise (and similarly for the other two source companies); 1 = all three companies were similarly helpful; 0 = otherwise; 1 = none of the companies was helpful; 0 = otherwise. We ran five separate ANCOVA models to include these responses as a covariate in our ANCOVA model. The ANCOVA model that includes the covariate related to UTP (i.e., the counterexample) is significantly associated with our dependent variable (p = 0.032). 13 Therefore, we include this covariate, referred to as UTP_Source, in our ANCOVA model for our planned contrasts (discussed below). Although declarative (factual) knowledge of ABC is assumed in this study to be above the minimum necessary for the experimental task, we measure the participants’ knowledge using a series of seven multiple-choice questions. This ABC-knowledge quiz was administered in the classroom several weeks prior to the experiment. The mean score is 6.07 (s.d. = 0.90). Reliability test results yield a Cronbach’s alpha of 0.11. 14 To rule out systematic differences in the participants’ declarative ABC knowledge as alternative explanations for our results, we included this variable as a covariate in our planned contrasts tests. Our results show that this variable did not affect the dependent variable at conventional levels (p > 0.10); thus, it is not discussed 13

14

The proportion of participants in the literal similarity condition that indicated that UTP was the most helpful in providing BFP’s profit trend forecast (7 out of 55 participants, or 13 percent) is the same as the proportion of participants who provided the same response in the relational similarity condition. Cronbach’s alpha is sensitive to test length, with longer tests receiving a higher score; thus, the low score is likely due to the test’s short length.

24

further. IV. RESULTS Validation Checks We used post-experimental data to assess participants’ perceptions of whether the products and customers portrayed in the source problems and the new setting were “homogeneous” or “heterogeneous” (as shown in Panels A and B of Appendix 3 for the literal and relational similarity conditions, respectively). We asked participants the following question: “For each of the four companies you read about, please indicate below whether the company’s products generate similar or different demands for activities (and therefore, activity costs) within the company. Also, indicate below whether the company’s customers generate similar or different demands for activities (and therefore, activity costs) within the company. Use an S for similar, and a D for different.” We coded each of the participants’ eight responses as 1 if correct, and 0 otherwise. We ran four separate binary logistic regression models (one for each of the three source companies and one for the new setting) to regress the participants’ responses regarding the nature of the products against the problem similarity variable (literal vs. relational). We ran another four binary logistic regression models to regress the participants’ responses regarding the nature of the companies’ customers against the problem similarity variable. We find that, in general, participants perceived the companies’ products as “similar” when they were implicitly described as homogeneous in the narratives, and as “dissimilar” when they were implicitly described as heterogeneous in the narratives. 15 Taken together, these results provide support for the

15

For example, in the literal similarity condition, the narrative implicitly depicts UTP’s customers as being “heterogeneous,” while in the relational similarity condition, the narrative implicitly depicts UTP’s customers as being “homogeneous.” Our binary logistic regression results show that the proportion of participants in the relational similarity condition who identified UTP’s customers as being “similar” (82.1 percent) is significantly higher than the proportion of participants in the literal similarity condition who identified UTP’s customers as being “similar” (47.3 percent) (Wald statistic = 13.735; p = 0.000). In the interest of brevity, the logistic regression results for the other source companies are not shown here.

25

effectiveness of our problem similarity manipulation. 16 Descriptive Statistics and Tests of Research Hypotheses Figure 3 shows graphically the mean recommendation scores (out of a maximum of 4) for each combination of the comparison and problem similarity variables, and Panel A of Table 2 shows the mean and marginal recommendation scores in tabular form by cross-classification condition. Univariate tests of the marginal means show that the mean recommendation score of participants in the comparison condition (mean = 2.79; s.d. = 0.986) is significantly higher than the mean score of participants in the advice condition (mean = 2.33; s.d. = 1.116) (t = 2.256; p = 0.026, two-tailed). The mean recommendation score of participants in the literal similarity condition (mean = 2.40; s.d. = 1.180) is not significantly different from the mean score of participants in the relational similarity condition (mean = 2.73; s.d. = 0.932) (t = 1.614; p = 0.110, two-tailed). [Insert Figure 3 and Table 2 here] As a basis for our planned contrast tests, we ran an ANCOVA model (not shown) that regresses the participants’ recommendation scores against our two independent variables and the UTP_Source covariate. The model shows significant main effects for Comparison (F = 6.360; p = 0.013), a marginally significant main effect for Problem Similarity (F = 2.799; p = 0.097), and a significant effect for UTP_Source (F = 4.707; p = 0.032) (all p-values are two-tailed). Panel B of Table 2 presents the results of our planned contrast to test our hypotheses. Recall that H1a predicts that when prompted with a new accounting setting that is literally 16

Also, we used a post-experimental question to elicit participants’ perceptions of task difficulty because it is possible that these perceptions are related to task performance. The question asked, “How difficult was it to provide a recommendation for Business Forms Printing?,” using a 9-point scale with endpoints labeled “Not difficult” and “Very difficult,” and a midpoint labeled “Moderately difficult.” The average difficulty score in the literal similarity condition (4.89; s.d. = 1.80) is not significantly different from the average score in the relational similarity condition (5.25; s.d. = 1.60) (p > 0.10). Also, the average difficulty score in the comparison condition (4.82; s.d. = 1.59) is not significantly different from the average score in the advice condition (5.33; s.d. = 1.80) (p > 0.10). Adding difficulty as a control variable in our planned contrasts tests does not affect the dependent variable at conventional levels (p > 0.10), and is not discussed further.

26

similar to source accounting problems, accountants who are first asked to compare the source problems will use more correctly a relevant causal relation in the new setting when making accounting recommendations, relative to accountants who are not asked to compare the source problems. For participants in the literal-similarity condition, the mean recommendation score of those who were first asked to compare the source problems (i.e., the comparison condition: cell 2 mean = 2.68) is significantly higher than the mean recommendation score of participants who were not asked to compare the source problems (i.e., the advice condition: cell 1 mean = 2.11;) (F = 4.524; p = 0.019, one-tailed). This result supports H1a. H1b predicts that when prompted with a new accounting setting that is relationally similar to source accounting problems, explicit elicitation of a comparison of the source problems will not be related to the accountants’ ability to correctly use a relevant causal relation in the new setting when making accounting recommendations. For participants in the relationalsimilarity condition, the mean recommendation score of those who were first asked to compare the source problems (i.e., the comparison condition: cell 4 mean = 2.89) is not significantly different from the mean score of those who were not first asked to compare the source problems (i.e., the advice condition: cell 3 mean = 2.56) (F = 1.782; p = 0.188, two-tailed). This result supports H1b. H2a predicts that when a comparison of source accounting problems is not explicitly elicited (i.e., the advice condition), accountants who are prompted with a new accounting setting that is relationally-similar to source accounting problems will use more correctly the relevant causal relation in the new setting when making accounting recommendations, relative to accountants who are prompted with source problems that are literally-similar to the new setting. For participants in the advice condition, the mean recommendation score of those in the relational-similarity condition (cell 3 mean = 2.56) is significantly higher than the mean recommendation score of participants in the literal-similarity condition (cell 1 mean = 2.11) (F = 27

2.899; p = 0.047, one-tailed). This result supports H2a. Finally, H2b predicts that when a comparison of source accounting problems is explicitly elicited (i.e., the comparison condition) prior to making a recommendation for the new accounting setting, the similarity between the source problems and the new setting will not be related to the participants’ ability to correctly use the relevant causal relation in the new setting when making accounting recommendations. For participants in the comparison condition, the mean recommendation score of those in the relational-similarity condition (cell 4 mean = 2.89) is not significantly different from the mean recommendation score of participants in the literalsimilarity condition (cell 2 mean = 2.68) (F = 0.508; p = 0.479, two-tailed). This result supports H2b. Our results show that eliciting comparison of source problems before participants make recommendations in a new setting is a powerful mechanism to enhance the quality of the accountants’ decision performance. At the same time, our results also suggest that relational similarity between the source problems and the new setting also enhances accountants’ recommendation performance, but only when a comparison of the source problems is not explicitly elicited first. As discussed earlier, additional tests confirm that the findings are not explained by whether a participant was able to activate the key accounting causal relationship. Taken together, our results provide support for the effectiveness of problem similarity and explicit elicitation of a comparison of source problems as mechanisms that help reduce the gap between activating and using accounting causal relationships in new accounting settings. V. CONCLUSION This study addresses a key question that remains unresolved in accounting research: Will providing mechanisms that reduce accountants’ tendency to omit key accounting causal relationships from their analysis of problems help improve their decision performance in new settings? We examine the effects of two candidate mechanisms—comparison of recalled 28

experiences and similarity between current and recalled experiences—which co-exist naturally in accounting decision-making settings, and are expected to facilitate application of a key accounting causal relation to a new complex accounting setting. Although we draw on analogical reasoning research, our study’s purpose differs from prior research. Our study is not designed to demonstrate how best to teach accountants to use analogical reasoning to learn from source problems a new specific solution to be applied to a similar problem (e.g., see Gentner et al. 2003; Thompson et al. 2000). Instead, our study examines how comparison and problem similarity affect accountants’ ability to reason by analogy to facilitate development of a mental model—for a new setting—that incorporates a relevant accounting causal relation which they already know. Our experimental inquiry was motivated by real-life business cases that provide evidence that experienced decision-makers—including accountants—characteristically make little effort to distinguish deep, structural similarities from superficial similarities in seemingly familiar but complex settings. Several contemporary examples illustrate companies that experienced financial disappointments following business-model changes, such as introduction of new distribution channels. These examples include GMI’s profitability problems associated with acquiring Colombo Yogurt. This case involved cost issues related to new distribution channels, exemplifying that it is critical for accountants and managers to distinguish deep, structural similarities from superficial similarities as companies expand channels or change their business models otherwise after entering into mergers or acquisitions. Recall that in the GMI-Colombo situation, emphasis on the many literal similarities shared by the two distribution channels initially obscured a relevant causal relationship, namely, that the same product required different marketing and distribution support costs when sold in different channels. Much like in some of our experimental scenarios, GMI experienced substantial heterogeneity in serving the two yogurt distribution channels, with associated effects on costs and profits. Only after General Mills recognized the structural differences and adopted a 29

new activity-based costing approach for selling and distribution costs were they able to quantify the impact of their heterogeneous distribution channels on profits. Our results provide support for the effectiveness of problem similarity and explicit elicitation of a comparison of source problems as mechanisms that help reduce the gap between activating accounting causal relationships and appropriately using them in new accounting settings. We predict and find that comparison of similarities and differences enhances accountants’ recommendation quality when the source problems are literally similar to the new setting, and depriving participants of literal similarities provides incremental decisionperformance benefits when participants do not explicitly compare the source problems. The finding on the importance of comparison given literal similarity between the source problems and the new setting suggests guarding against individuals’ tendencies, when facing new problem settings, to look only for similarities with prior experiences. Our results suggest that actively comparing both similarities and differences between business problems, events, or opportunities can play a significant role in improving decision performance, especially when the source problems and the new setting are seemingly similar on the surface. Our finding on decision-performance impairment when the source problems are literally similar to the new setting (and recalled experiences are not explicitly compared) is especially surprising because intuition suggests that experience with problems and settings that are the most similar should be the most helpful. In particular, a casual observer might expect that the best decision performance would occur among participants in the advice condition (e.g., those who provided revenue-enhancing recommendations for the source problems), particularly when the source problems were literally similar to the new setting. At a practical level, our experiment produced the counter-intuitive result that experience with many source problems that are literally similar to new settings is not necessarily the most helpful, and in fact, interferes with a decision-

30

maker’s ability to appropriately use key causal relations in new settings, even when accountants are able to successfully activate the task-relevant causal relationship. Several possibilities for future research emerge from our study. First, one caveat in our study is that our subjects were graduate and undergraduate accounting students. While our choice of participants stemmed from the need to control for the participants’ factual (declarative) knowledge of ABC/ABM, as well as to ascertain that they possessed the task-relevant accounting causal relationship, future research might use other subject pools with more experience using ABC/ABM. These different subject pools have different advantages and disadvantages, and as argued by Lipe and Salterio (2000, 298), theory does not suggest an optimal choice. In our study, focusing on employees of a firm that uses a particular ABC/ABM implementation would compromise experimental control if ABC/ABM is used predominantly for product profitability analyses but not for customer profitability analyses, or vice-versa, or if the firm is manufacturing-oriented as opposed to service-oriented, or vice-versa. Similarly, focusing on employees from a cross-section of firms means that most or all participants would be presented with ABC/ABM implementations that may differ from the situations presented in our experimental materials. Finally, training experimental participants to use ABC/ABM can lead to significant demand effects, and would also compromise experimental control, since one of the critical features of our experiment is avoiding cluing participants about the nature of the revenue-increasing recommendation task. Thus, using subjects with ABC/ABM experience would require carefully controlling for the issues outlined above. The results of our study have implications for decision-making in seemingly familiar but complex contexts, including revenue-enhancement or cost-containment decisions following a business-model change. Future research could examine the effects of problem similarity and comparison in other contexts, such as financial reporting and auditing. For example, recent discussions of U.S. financial reporting recommend that U.S. standard setting move from a “rules31

based” approach toward a “principles-based” approach (e.g., see Nobes 2005; Nelson 2003; Schipper 2003). An important theme that has emerged from these discussions is that moving to a principles-based approach is desirable because such an approach allows the appropriate exercise of professional judgment (Schipper 2003). When available standards offer few or no precise rules, reasoning by analogy offers a powerful mechanism to enhance accountants’ professional judgment (Nelson 2003; Salterio 1996). However, as suggested by the results of our study, it is extremely easy to reason poorly through analogies without the help of other cognitive mechanisms. Thus, one broad question for accounting practice and future research that arises from our findings is: How do comparison and problem similarity affect accountants’ ability to map relations between elements of standards or examples to their own decision problems when standards offer few or no rules in matters related to financial reporting? One approach to this inquiry might be to compare the effects of these mechanisms on accountants’ professional judgments using a principles-based versus a rulesbased financial reporting approach. This research inquiry could also be extended to members of audit committees, who are charged with the oversight of the financial reporting process. In view of the Sarbanes-Oxley Act of 2002, audit committees are likely to seek more detailed guidance as to precisely what is expected in terms of their role in financial reporting (Schipper 2003, 71). Furthermore, in our experiment, we examined “pure” mechanisms: comparison vs. advice. In real-world accounting situations, appropriate use of a key accounting principle or causal relation can also occur with “hybrid” mechanisms. However, hybrid models often develop in the absence of knowledge of the effectiveness of their components. While examining “pure” types is an important precursor to examination of hybrid models, future research could examine the generality of our findings using “hybrid” models (see also Nadler et al. 2003 for other models of knowledge creation and transfer).

32

Finally, an important issue addressed in our study is the knowledge that participants are expected to bring to the task. In our context, because of their formal accounting training, the participants had knowledge of the causal relation that was needed for decision-making in the new setting. Future research could examine the generality of our findings when participants do not have such knowledge, or when they have acquired the knowledge not from training (e.g., explicit knowledge, or “know-what”), but from experience (e.g., tacit knowledge, or “know-how”).

33

REFERENCES Ahn, W., W. Brewer, and R. Mooney. 1992. Schema acquisition from a single example. Journal of Experimental Psychology: Learning, Memory, and Cognition 18: 391-412. Atkinson, A., R. Kaplan, and S.M. Young. 2004. Activity-Based Cost Systems (Chapter 4, pp. 121-179). In Management Accounting (4th Edition). Upper Saddle River, NJ: Pearson/Prentice Hall. Baron, R., and D. Kenny. 1986. The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology 51 (December): 1173-1182. Bassok, M., V. Chase, and S. Martin. 1998. Adding apples and oranges: Alignment of semantic and formal knowledge. Cognitive Psychology 35 (March): 99-134. Blocher, E., K. Chen, and T. Lin. 2002. Managing marketing effectiveness, productivity, and customer profitability (Chapter 17, pp. 804-853). In Cost Management—A Strategic Emphasis (2nd Edition). New York, NY: McGraw-Hill/Irwin. Boland, R., J. Singh, P. Salipante, J. Aram, S. Fay, and P. Kanawattanachai. 2001. Knowledge representations and knowledge transfer. Academy of Management Journal 44 (2): 393417. Catrambone, R., and K. Holyoak. 1989. Overcoming contextual limitations on problem-solving transfer. Journal of Experimental Psychology: Learning, Memory, & Cognition 15: 11471156. Chen, Z. 1996. Generating suggestions through document structure mapping. Decision Support Systems 16: 297-314. Chewning, E., and A. Harrell. 1990. The effect of information overload on decision makers’ cue utilization levels and decision quality in a financial distress decision task. Accounting, Organizations and Society 15: 527-542. Clement, C., and D. Gentner. 1991. Systematicity as a selection constraint in analogical mapping. Cognitive Science 15: 89-132. Collins, A. and M. Burstein. 1989. Afterword: A framework for a theory of comparison and mapping. In Similarity and Analogical Reasoning (S. Vosniadou and A. Ornoty, eds.). New York: Cambridge University Press. Cronbach, L. 1951. Coefficient alpha and the internal structure of tests. Psychometrika 16: 297334. Einhorn, H., and R. Hogarth. 1978. Confidence in judgment: Persistence in the illusion of validity. Psychological Review 85: 395-416. 34

Ferguson, R.W., and K. Forbus. 1998. Telling juxtapositions: Using repetition and alignable difference in diagram understanding. In K. Holyoak, D. Gentner, and B. Kokinov (Eds.), Advances in Analogical Research (pp. 109-117). Sofia, Bulgaria: New Burlingame University. Forbus, K., D. Gentner, and K. Law. 1995. MAC/FAC: A model of similarity-based retrieval. Cognitive Science 19: 141-205. Freed, J. 2004. Warding off “Demon Customers.” The Associated Press. Available at www.CBSNews.com (July 6). Gavetti, G., and J. Rivkin. 2005. How strategists really think—Tapping the power of analogy. Harvard Business Review (April): 1-10. Gavetti, G., and J. Rivkin. 2003. The use and abuse of analogies. Case 9-703-429 (February 4). Harvard Business School Publications: 1-12. Gentner, D. 1983. Structure mapping: A theoretical framework for analogy. Cognitive Science 7: 155-170. ——–, and C. Toupin. 1986. Systematicity and surface similarity in the development of analogy. Cognitive Science 10: 277-300. ——–. 1989. The mechanisms of analogical learning. In S. Vosniadou and A. Ortony (Eds.), Similarity, Analogy, and Thought (pp. 199-241). Cambridge, England: Cambridge University Press. ——–, M.J. Ratterman, and K. Forbus. 1993. The roles of similarity in transfer: Separating retrievability from inferential soundness. Cognitive Psychology 25: 524-575. —–——–, S. Brem, R. Ferguson, A. Markman, B. Levidow, P. Wolff, and K. Forbus. 1997.

Analogical reasoning and conceptual change: A case study of Johannes Kepler. Journal of the Learning Sciences 6: 3-40. ——–—–, and A. Markman. 1997. Structure mapping in analogy and similarity. American

Psychologist 52: 45-56. ——–, and J. Medina. 1998. Similarity and the development of rules. Cognition 65: 263-297. ——–, and L. Namy. 1999. Comparison in the development of categories. Cognitive Development 14: 487-513. ——–, J. Loewenstein, and L. Thompson. 2003. Learning and transfer: A general role for analogical encoding. Journal of Educational Psychology 95: 393-408. Gick, M., and K. Holyoak. 1983. Schema induction and analogical transfer. Cognitive Psychology 15: 1-38.

35

Gilovich, T. 1981. Seeing the past in the present: The effect of associations to familiar events on judgments and decisions. Journal of Personality and Social Psychology: 797-808. Goldstone, R., D. Medin, and D. Gentner. 1991. Relational similarity and non-independence of features in similarity judgments. Cognitive Psychology 23: 222-262. Gregan-Paxton, J., and J. Cote. 2000. How do investors make predictions? Insights from analogical reasoning. Journal of Behavioral Decision Making 13: 307-327. Guy, J. and J. Saly 2000. Colombo frozen yogurt. Cases from Management Accounting Practice, Vol. 15, Institute of Management Accountants, Montvale, NJ, www.imanet.org. Hamilton, D. 1979. A cognitive attributional analysis of stereotyping. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 12). New York: Academic Press. Hanson, E., and F. Phillips. 2006. Teaching financial accounting with analogies: Improving initial comprehension and enhancing subsequent learning. Issues in Accounting Education (Vol. 21/1): 1-14. Hastie, R., and P. Kumar. 1979. Person memory: Personality traits as organizing principles in memory for behavior. Journal of Personality and Social Psychology 37: 25-38. Hilton, R., M. Maher, and F. Selto. 2003. Activity-Based Costing Systems (Chapter 4, pp. 140183. In Cost Management—Strategies for Business Decisions (2nd Edition). New York, NY: McGraw-Hill/Irwin. Holland, J., K. Holyoak, R. Nisbett, and P. Thagard. 1986. Induction—Processes of Inference, Learning, and Discovery. Cambridge, MASS: The MIT Press. Holyoak, K. 1985. The pragmatics of analogical transfer. In G. H. Bower (Ed.), The Psychology of Learning and Motivation (Vol. 19, pp. 59-87). New York: Academic Press. Holyoak, K., and K. Koh. 1987. Surface and structural similarity in analogical transfer. Memory and Cognition 15: 332-340. ——–, and P. Thagard. 1995. Mental Leaps: Analogy in Creative Thought. Cambridge, MA: MIT Press. ——–, and ______. 1989. Analogical mapping by constraint satisfaction. Cognitive Science 13: 295-355. Horngren, C., S. Datar, and G. Foster. 2005. Chapter 5 Case—Colombo Frozen Yogurt: Activity-Based Costing, in Cost Accounting—A Managerial Emphasis (12th Edition): Upper Saddle River, NJ: Pearson/Prentice Hall: 179. Hwang, M., and J. Lin. 1999. Information dimension, information overload, and decision quality. Journal of Information Science 25: 213-218. Johnson-Laird, P. 1989. Analogy and the exercise of creativity. In Similarity and Analogical 36

Reasoning (S. Vosniadou and A. Ornoty, eds.). New York: Cambridge University Press. Kaplan, R., and R. Cooper. 1998. ABC in Service Industries (Chapter 12), in Cost & Effect— Using Integrated Cost Systems to Drive Profitability and Performance. Harvard Business School Press. Boston, MA: 228-251. Kolodner, J. 1993. Case-based reasoning. Morgan Kaufmann Publishers. San Mateo, CA. Kurtz, K., C-H. Miao, and D. Gentner. 2001. Learning by analogical bootstrapping. The Journal of the Learning Sciences 10(4): 417-446. Lewis, M., and J. Anderson. 1985. Discrimination of operator schemata in problem solving: Learning from examples. Cognitive Psychology 17: 26-65. Lipe, M., and S. Salterio. 2000. The balanced scorecard: Judgmental effects of common and unique performance measures. The Accounting Review (July): 283-298. Loewenstein, J., L. Thompson, and D. Gentner. 1999. Analogical encoding facilitates knowledge transfer in negotiation. Psychonomic Bulletin & Review 6: 586-597. Luft, J., and M. Shields. 2001. Why does fixation persist? Experimental evidence on the judgment performance effects of expensing intangibles. The Accounting Review (October): 561-587. Magretta, J. 2002. Why business models matter. Harvard Business Review (May): 3-8. Marchant, G., J. Robinson, U. Anderson, and M. Schadewald. 1993. The use of analogy in legal argument: Problem similarity, precedent, and expertise. Organizational Behavior & Human Decision Processes: 95-119. Medin, D., R. Goldstone, and D. Gentner. 1993. Respects for similarity. Psychological Review 100: 254-278. Merrick, A. 2003. Sears to sell clothing on its website. The Wall Street Journal (September 26): B2. Nadler, J., L. Thompson, and L. VanBoven. 2003. Learning negotiation skills: Four models of knowledge creation and transfer. Management Science 49: 529-540. Nelson, M. 2003. Behavioral evidence on the effects of principles- and rules-based standards. Accounting Horizons (March): 91-104. Nobes, C. 2005. Rules-based standards and the lack of principles in accounting. Accounting Horizons (March): 25-34. Nunnally, J. 1978. Psychometric Theory (2nd Edition). New York, NY: McGraw-Hill. Public Accounting Report (PAR). 2004. 23rd Annual Professors’ Survey School Rankings. PAR (November 30): 1-7. 37

Reeves, L., and R. Weisberg. 1994. The role of content and abstract information in analogical transfer. Psychological Bulletin 115: 381-400. Ross, B., and P. Kennedy. 1990. Generalizing from the use of earlier examples in problemsolving. Journal of Experimental Psychology: Learning, Memory and Cognition 16(1): 42-55. Salterio, S. 1996. The effects of precendents and client position on auditors’ financial accounting policy judgments. Accounting, Organizations and Society 21: 467-486. Schick, A., L. Gordon, and S. Haka. 1990. Information overload: A temporal approach. Accounting, Organizations and Society 15: 199-220. Schipper, K. 2003. Principles-based accounting standards. Accounting Horizons (March): 61-72. Schwartz, D., and J. Bransford. 1998. A time for telling. Cognition & Instruction 16: 475-522. Sherwin, E. 1999. A defense of analogical reasoning in law. The University of Chicago Law Review (Fall): 1179-1197. Thagard, P. 1996. Mind: Introduction to Cognitive Science. Cambridge, MA: MIT Press. Thompson, L., D. Gentner, and J. Loewenstein. 2000. Avoiding missed opportunities in managerial life: Analogical training more powerful than individual case training. Organizational Behavior and Human Decision Processes 82: 60-75. Tsoukas, H. 1993. Analogical reasoning and knowledge generation in organization theory. Organization Studies 14(3): 323-346. Turetken, O., and R. Sharda. 2004. Development of a fisheye-based information search processing aid (FISPA) for managing information overload in the web environment. Decision Support Systems (June): 415-434. Vallacher, R., and D. Wegner. 1987. What do people think they’re doing? Action identification and human behavior. Psychological Review 94: 3-15. Vera-Muñoz, S. 1998. The effects of accounting knowledge and context on the omission of opportunity costs in resource allocation decisions. The Accounting Review (January): 4772. ——–, W.R. Kinney, Jr., and S. Bonner. 2001. The effects of domain experience and task presentation format on accountants’ information relevance assurance. The Accounting Review (July): 405-429. Wason, P., and P. Johnson-Laird. 1972. Psychology of Reasoning: Structure and Content. London: Batsford.

38

Whittlesea, B. 1997. The representation of general and particular knowledge. In K. Lamberts and D. Shanks (Eds.), Knowledge, Concepts and Categories. Cambridge, MA: MIT Press: 335-370.

39

Recommendation Scores

FIGURE 1 Predicted Effects of Comparison (H1a and H1b) and Problem Similarity (H2a and H2b) on Accounting Recommendation Qualitya

H2b

H1b H2a

H1a

Relational similarity Literal similarity

Not elicited (Advice) Elicited (Comparison) Comparison

a

The dependent variable is constructed from participants’ revenue-enhancing recommendations in the new setting. Participants in the comparison not elicited (advice) condition were asked to provide advice for each of three companies (the source problems) as to how to improve the companies’ profits in year t, relative to year t – 1. Participants in the comparison elicited condition were asked to write down similarities and differences between and among the three source companies, combinations of similarities and differences, and to articulate the overall main principle that captures the essence of what they learned from the three source companies taken together. Participants in the literal similarity condition received three source problems that share (implicitly) with a new setting a relevant causal relation and maintain a one-to-one attribute match. Participants in the relational similarity condition received three source problems that share (implicitly) with a new setting the relevant causal relation but do not maintain the one-to-one attribute match.

40

FIGURE 2 Experimental Procedures and Tasks Part One

Part Two

Provide advice for each source a problem (Appendix 1)

Read narratives for source problems

Read narrative of new accounting c setting (Appendix 2)

Profit trend forecast elicited (Appendix 2)

Assessment of cost accounting system evaluation elicited (Appendix 2)

Revenueenhancing recommendation elicited (Appendix 2)

Compare three b source problems (Appendix 1) a

Explicit comparison of three source accounting problems is not elicited. Explicit comparison of three source accounting problems is elicited. c The source accounting problems are either literally similar or relationally similar to the new accounting setting. (See variable definitions in Figure 1.) b

41

Complete postexperimental questionnaire

Recommendation Scores

FIGURE 3 Observed Effects of Comparison (H1a and H1b) and Problem Similarity (H2a and H2b) on Accounting Recommendation Qualitya

2.89

H1b

2.56

H2a

2.68

H1a

2.11

Not elicited (Advice)

H2b

Elicited (Comparison)

Comparison

See variable definitions in Figure 1.

42

Relational similarity Literal similarity

TABLE 1 Frequencies of Participants’ Profit Trend Forecasts and Mean Scores on Cost Accounting System Assessment Profit Trend a Forecast

Frequencies

Mean Score on Cost Accounting System Assessments c (s.d. in parentheses)

Decrease

88 (80%)

Inferior ( = 0) Average ( = 1) Superior (= 2)

39 22 27

0.86 (0.86)

No change

6 (5%)

Inferior ( = 0) Average ( = 1) Superior (= 2)

4 1 1

0.50 (0.84)

Increase

16 (15%)

Inferior ( = 0) Average ( = 1) Superior (= 2)

11 2 3 110

0.50 (0.816)

Total a

Frequencies (%)

Score on Cost Accounting System b Assessments

110

After participants read the narrative for the new setting but before they provided their recommendations, they answered the following question: “Based on the above information, you expect BFP’s 2002 profit to (choose one): ____ increase ____ decrease ____ remain unchanged compared to BFP’s 2001 profit. Explain the rationale to your answer.” (See also Appendix 2). We coded the participants’ responses as follows: increase = +1, decrease = –1, and remains unchanged = 0.

b

c

After participants answered the question above but before they provided their recommendations for the company in the new setting, they answered the following question: “Is BFP’s cost system appropriate for decision making? If it is, then explain why. If it is not, then explain why not and how BFP should change it.” (See also Appendix 2). We coded the participants’ responses as superior ( = 2), average ( = 1), or inferior ( = 0). The mean scores of participants’ profit trend forecasts are marginally negatively correlated with the scores on their cost accounting system assessments (Spearman’s rho = –0.176; p = 0.066).

43

TABLE 2 Descriptive Statistics and Planned Contrasts (n = 110) Panel A. Dependent Variable is Mean Recommendation Scores (out of a maximum of 4); Std. Dev. in Parentheses Comparisonb Not Elicited Elicited a Similarity (Advice) (Comparison) Total 1 2 2.11 2.68 2.40 Literal (1.219) (1.090) (1.180) [27] [28] [55] 3 Relational

Total

4 2.56 (0.974) [27]

2.89 (0.875) [28]

2.73 (0.932) [55]

2.33 (1.116) [54]

2.79 (0.986) [56]

2.56 (1.071) [110]

Panel B. Planned Contrasts Contrast H1a: Effects of comparison in the literal similarity condition Not elicited (Advice) vs. Elicited (Comparison) (cell 1 vs. cell 2, Panel A)

Mean Square

F-statistic

p-valuec

4.759

4.524

0.019

1.875

1.782

0.188

3.050

2.899

0.047

0.534

0.508

0.479

H1b: Effects of comparison in the relational similarity condition Not elicited (Advice) vs. Elicited (Comparison) (cell 3 vs. cell 4, Panel A) H2a: Effects of problem similarity in the comparison not elicited (advice) condition Literal vs. Relational Similarity (cell 1 vs. cell 3, Panel A) H2b: Effects of problem similarity in the comparison elicited condition Literal vs. Relational Similarity (cell 2 vs. cell 4, Panel A)

44

__ TABLE 2 (Continued) Descriptive Statistics and Planned Contrasts (n = 110) a

b

c

Participants in the literal similarity condition received three source problems that share (implicitly) with a new setting a relevant causal relation and maintain a one-to-one attribute match. Participants in the relational similarity condition received three source problems that share (implicitly) with a new setting the relevant causal relation but do not maintain the one-to-one attribute match. Participants who were not explicitly elicited to compare the source problems (i.e., the advice condition) were asked to provide advice to each of three source companies as to how to improve their profits in year t, relative to year t – 1. Participants who were explicitly elicited to compare the source problems (i.e., the comparison condition) were asked to write down similarities and differences between and among the three source companies, combinations of similarities and differences, and to articulate the overall main principle that captures the essence of what they learned from the three business cases taken together. One-tailed for H1a and H2a; two-tailed for H1b and H2b.

45

APPENDIX 1 Comparison Manipulation Panel A. Comparison Explicitly Elicited a) Think about the similarities and differences among the three preceding business cases. In the time allotted, list as many similarities, differences, and combinations of similarities-differences as you can identify between and among the cases. For example, consider the following list: hockey, football, and snowboarding. Similarities among the three items in the list include: (1) they are all sports; and (2) individuals use padding to protect their bodies from injury. Differences include: (1) hockey requires a puck, football requires a ball, and snowboarding requires a snowboard. Combinations of similaritiesdifferences include: (1) hockey and football are team sports, while snowboarding is an individual sport; and (2) hockey and snowboarding are Winter Olympic sports, while football is not. Similarities among the three business cases:

Differences among the three business cases:

Combinations of similarities-differences:

b) Using only one sentence, articulate the overall main principle that captures the essence of the three cases taken together.

Panel B. Comparison Not Elicited (Advice) For each of the three source companies, the participants received the following instructions: Based on the information provided above [in the preceding company description], recommend two possible ways in which [Company name] can increase their profits over the 2002 level. List first your recommendation with the highest potential for increasing profits: Recommendation #1 (first highest potential):

Recommendation #2 (second highest potential):

46

APPENDIX 2 Narrative for the Company in the New Setting Instructions: For this part please assume that you are a business consulting associate who works for a consulting firm. Your managing partner has asked you to assess the situation and provide recommendations for a new client, Business Forms Printing (BFP), described below. Business Forms Printing (BFP), a Midwest printing shop, provides business forms for various companies in the US and around the world. Until the end of 2001, the company printed and shipped the forms in standard batch sizes. Customers were required to purchase in standard quantities and therefore sometimes held large quantities of inventories of their business forms. At the beginning of 2002, in an effort to increase profits, BFP introduced a new policy of shipping only what the customer requests, as frequently as the customer wants. BFP therefore now provides value to its customers by managing the inventory of business forms. Customers do not pay for ordered forms until they receive them. Owing to the new policy, BFP asks its customers to provide a yearly forecast of the number of forms needed. However, BFP management has noticed that some customers forecast with up to 99% accuracy, while others (especially the new customers) do not forecast accurately. As a result, BFP holds rather large quantities of inventories of business forms for some of its customers, while it holds little inventory for other customers. In addition, some customers place small orders several times a year, while others place large orders a few times a year. Further, BFP encounters business form obsolescence when customers change their forms while BFP holds a large inventory of the old forms. Finally, some customers take advantage of BFP’s promise to ship only what the customer requires. Therefore, BFP sometimes ships less than a full carton of forms. This requires an employee to open the carton, count out the number of forms to ship, and then package the forms for shipping. BFP’s printing processes for the various companies’ forms are quite similar in terms of printing machines used and time requirements. For pricing purposes, BFP computes the cost of goods sold as direct materials plus direct labor plus applied manufacturing overhead cost. The manufacturing overhead rate is computed at the beginning of each year as estimated total manufacturing overhead costs divided by estimated machine hours. In addition, BFP computes a selling and administrative (S&A) rate to allocate selling and administrative costs for pricing purposes. This S&A rate is computed at the beginning of each year as estimated total selling and administrative costs divided by estimated total cost of goods sold. Customers pay actual shipping charges by an international shipping company. In 2002, the number of customers increased and sales revenue increased, although sales prices for the forms did not change. (a) Based on the above information, you expect BFP’s 2002 profit to (choose one): ____increase ____decrease ____remain unchanged Explain the rationale for your answer to part (a). (b) Is BFP’s cost system appropriate for decision making? If it is, then explain why. If it is not, then explain why not and how BFP should change it. (c) Recommend two possible ways in which BFP can increase its revenues. What issues should BFP consider as they attempt to increase revenues?

47

APPENDIX 3 Manipulation of Problem Similarity between the Source Companies and the Company in the New Setting Panel A. Literal Similarity Company Name [Industry] Mega Outfitting Co. (MOC) [Merchandising]

Nature of Product(s) Homogeneous: moderately priced business casual clothing and accessories

Nature of Customers Heterogeneous: vary in terms of number of orders placed over a one-year period; dollar amount of orders placed; how the order is placed (phone, mail, or website); number and dollar amount of merchandise returned; and shipping charges (regular vs. expedited delivery)

University T-shirt Printing (UTP) [Manufacturing]

Homogeneous: highquality adult-size football T-shirts with Midwestern university logo

Heterogeneous: University bookstore, national department store chain, local bookstores, local organizations; vary in terms of size of orders placed, storage costs, ability to accurately estimate demand of T-shirts, and shipping charges (regular vs. expedited delivery)

Tailgate Party Catering (TPC) [Catering]

Homogeneous: standard tailgate party catering with two similarly-priced menu options

Heterogeneous: Residential vs. non-residential (commercial) clients; vary in terms of number of parties catered per football season; location (intown vs. out-of-town); how the orders are placed (over the phone or using a website); rush orders vs. regular orders

Homogeneous: printing forms that use a similar printing process and machine time; inventory management service

Heterogeneous: vary in terms of number of orders placed over a one-year period; dollar amount of orders placed; ability to accurately estimate number of forms needed each year; dollar amount of inventory in storage; size of shipments (full carton vs. less than full cartons)

Company in the new setting: Business Forms Printing (BFP) [Manufacturing]

48

One-Year Change in Number of Customers

One-Year Change in Ratio of HighCost to Low-Cost Customers

One-Year Change in Profits

↑

↑

↓

↑

↑

↓

↑

↑

↓

↑

↑

↓

APPENDIX 3 (Continued) Manipulation of Problem Similarity between the Source Companies and the Company in the New Setting Panel B. Relational Similarity

Company Name [Industry] Mega Outfitting Co. (MOC) [Merchandising]

University T-shirt Printing (UTP) [Manufacturing]

Lambert’s Corporate Catering (LCC) [Catering]

Company in the new setting: Business Forms Printing (BFP) [Manufacturing]

Nature of Product(s) Homogeneous: moderately priced business casual clothing and accessories

Nature of Customers Heterogeneous: vary in terms of number of orders placed over a one-year period; dollar amount of orders placed; how the order is placed (phone, mail, or website); number and dollar amount of merchandise returned; and shipping charges (regular vs. expedited delivery)

Homogeneous: highquality adult-size football T-shirts with Midwestern university logo

Homogeneous: university bookstore and national department store chain; similar in terms of ability to accurately estimate demand of Tshirts, shipping charges, schedule of deliveries over the year, no merchandise returns necessary

Heterogeneous: menu includes both standard and complex selections for breakfast, lunch, dinner, or cocktail party receptions; differently-priced menu selections

Homogeneous: local professional services firms; similar in terms of location (in-town only); how the orders are placed (over the phone only); only regular orders accepted

Homogeneous: printing forms that use a similar printing process and machine time; inventory management service

Heterogeneous: vary in terms of number of orders placed over a one-year period; dollar amount of orders placed; ability to accurately estimate number of forms needed each year; dollar amount of inventory in storage; size of shipments (full carton vs. less than full cartons)

49

One-Year Change in Ratio of HighCost to Low-Cost Customers or Products

One-Year Change in Profits

↑

↑

↓

(customers)

(customers)

One-Year Change in Number of Customers or Products

↑

Ø

(customers)

(customers)

↑

↑

(products)

(products)

↑

↑

(customers)

(customers)

↑

↓

↓