The traditional view of cost behavior assumes a simple mechanistic relation between cost drivers and costs. In contrast, contemporary cost management research recognizes that costs are caused by managers' operating decisions subject to various constraints, incentives, and psychological biases. This conceptual innovation opens up the “black box” of cost behavior and gives researchers a powerful new way to use observed cost behavior as a lens to study the determinants and the consequences of managers' operating decisions. Banker and Byzalov (2014) presented an overview of the economic theory of cost behavior and major estimation issues. The research literature on cost management has grown rapidly in the past few years and has enhanced the understanding of how managerial decisions influence observed costs. In this study, we provide a comprehensive review of recent findings and insights, with a particular emphasis on the implications of cost management for understanding issues in cost, managerial, and financial accounting, and challenges and opportunities for future research.
Cost behavior and cost management have emerged as major new research areas in accounting. We interpret “cost management” broadly as all deliberate operating decisions by managers that influence observed cost behavior.1 The traditional textbook view of cost behavior assumed a mechanistic relation between a cost driver and costs (e.g., Garrison, Noreen, and Brewer 2015). Refinements of the traditional view recognized that costs are caused by resources and that these resources are required to perform multiple activities (e.g., Kaplan and Cooper 1987; Cooper and Kaplan 1992). Research on these refinements examined the explanatory power of various activity cost drivers (e.g., Miller and Vollmann 1985; Foster and Gupta 1990; Banker and Johnston 1993; Banker, Potter, and Schroeder 1995; and many others); however, it continued to assume a mechanistic relation between activities and costs and did not attempt to examine the role of managerial decisions in cost behavior.2 In contrast, contemporary cost management research introduces managerial decisions as the fundamental driver of costs. This conceptual innovation opens up the “black box” of cost behavior, giving researchers a powerful new way of thinking about managerial decisions through the lens of observed cost behavior. For example, a researcher can use this decision-based approach to analyze how cost management interacts with demand characteristics, industry structure, strategic positioning, managers' incentives and psychological biases, other stakeholders' incentives and biases, earnings management, corporate governance, government regulation, and national culture, among other factors.
We review the rapidly growing literature in cost, managerial, and financial accounting on cost management and its implications, outline the main themes that tie together multiple findings in this literature, and identify challenges and opportunities for future research. Because costs arise from managerial decisions, many recent studies examine cost behavior as a means to gain insight into managerial incentives, psychological biases, and institutional factors. Because earnings = sales – costs, cost management has implications for various research areas that use earnings. For example, recent research shows that a better understanding of cost management decisions leads to new findings on fundamental analysis signals, earnings prediction, conditional conservatism, earnings management, and analysts' earnings forecasts, among other topics.
Managers choose resource levels subject to various constraints (e.g., resource adjustment costs, demand conditions, governance), incentives (e.g., performance compensation, stakeholder activism), and biases (e.g., overconfidence, hyperbolic discounting). All of these factors have been shown to influence costs. For example, when the resource adjustment costs are substantial, managers prefer to limit the magnitude of both upward and downward resource adjustments to save on the expected adjustment costs. Therefore, managers likely choose different resource levels for the same activity level depending on whether they are adjusting resources upward in response to an activity increase or downward in response to an activity decrease (e.g., Anderson, Banker, and Janakiraman 2003). As predicted, Anderson et al. (2003) and Banker, Byzalov, and Chen (2013b) find that adjustment costs influence managers' resource commitment decisions.3 Studies of managerial incentives show that managers choose a lower resource level for the same activity level when they have strong incentives to avoid reporting a loss (e.g., Dierynck, Landsman, and Renders 2012; Kama and Weiss 2013), and that they choose a higher resource level for the same activity level when they have an incentive to engage in empire building (e.g., Chen, Lu, and Sougiannis 2012). Studies of behavioral biases find that managers' resource commitment decisions are influenced by factors such as overconfidence (Chen, Gores, and Nasev 2015b; Qin, Mohan, and Kuang 2015) and hubris (Yang 2015). Many studies use observed cost behavior to learn about managerial responses to various other constraints, circumstances, and phenomena, thus gaining insights into the fundamental properties of managerial decisions.
Because earnings = sales – costs, cost management decisions that manifest in observed cost behavior affect various earnings properties, such as earnings predictability, asymmetric timeliness, and persistence. This link between cost management and earnings properties affects inferences in research areas that rely on earnings-based metrics. For example, Weiss (2010) shows that cost management decisions affect analysts' forecast accuracy through the unpredictable component of future earnings driven by future sales shocks. Banker, Basu, Byzalov, and Chen (2016a) show that cost management decisions impact inferences on conditional conservatism that are based on asymmetric timeliness of earnings. Because cost management decisions affect time-series properties of earnings, observed cost behavior can have predictive value in earnings forecasts (e.g., Banker and Chen 2006). If analysts and investors do not fully take advantage of this predictive value, then their earnings expectations will have a bias that varies predictably with sales changes and other determinants of costs. This bias can lead to systematic errors in analysts' earnings forecasts and systematic abnormal returns around earnings announcements.
Managers' operating decisions that manifest in cost behavior drive a broad range of additional outcomes, including (but not limited to) fundamental analysis signals, operating accruals, efficiency measures, firm boundaries, and economy-wide unemployment rates. Therefore, insights developed in the empirical context of cost behavior have implications for various research areas that require a broader understanding of managers' operating decisions. For example, cost management research finds that managers who are optimistic about future sales are more willing to retain unused resources during sales decreases (e.g., Anderson et al. 2003). The retention of unused resources increases costs (relative to revenue) and reduces earnings for the current period. Thus, if managerial optimism is justified, then higher costs and lower earnings during a sales decrease can convey good news about future demand. Conversely, lower costs and higher earnings during a sales decrease can convey bad news. Consistent with this argument, recent studies show that the interpretation of the SG&A cost ratio and other standard signals in fundamental analysis is reversed during sales decreases (Anderson, Banker, Huang, and Janakiraman 2007; Banker, Fang, and Mehta 2016d). Other recent studies leverage the insights on managers' operating decisions from cost management research to analyze accrual-based earnings management, efficiency measurement, outsourcing decisions, and macroeconomic outcomes, among other phenomena (e.g., Banker, Byzalov, Fang, and Jin 2016b; Atasoy and Banker 2014; Atasoy, Banker, and Byzalov 2016a; Rouxelin, Wongsunwai, and Yehuda 2017).
The article by Banker and Byzalov (2014) primarily sought to establish the economic theory of cost behavior and clarify major empirical implementation issues, and did not attempt to provide a detailed literature review or examine the broad range of implications resulting from opening the “black box” of cost behavior. We adopt the economic theory from Banker and Byzalov (2014) as the conceptual foundation and provide a comprehensive, systematic review of the recent findings and insights in the rapidly growing literature on cost management, with a particular emphasis on the implications of cost management and future research opportunities in addressing questions ranging from financial analysis and earnings management to boundaries of the firm.
In the next section, we discuss the conceptual foundations of cost management. We then review the main research findings on cost management and its determinants and discuss some of the methodological issues in this research. We next review the research on the implications of cost management for various topics in financial and managerial accounting and for some topics outside accounting. The final section concludes and provides suggestions for future research.
CONCEPTUAL FOUNDATIONS OF COST BEHAVIOR AND COST MANAGEMENT
The traditional view of cost behavior in accounting research and textbooks is based on the “black-box” model of fixed and variable costs (e.g., Garrison et al. 2015; Horngren, Datar, and Rajan 2015). This model describes a mechanistic linear relation between a cost driver, such as sales or production volume, and concurrent costs. In the short run, variable costs are proportional to the cost driver, and fixed costs are constant.4 The traditional model recognizes that not all costs are strictly fixed or strictly variable and allows for “mixed” costs that combine fixed and variable components. Extensions of the basic textbook model can incorporate multiple cost drivers as well as nonlinear effects such as congestion costs, economies of scale, and/or learning-by-doing (e.g., Horngren et al. 2015, Chap. 10).5 However, these extensions continue to assume a mechanistic relation between cost driver quantities and costs, leaving no room for deliberate managerial decisions motivated by various economic constraints, incentives, and biases. These extensions do not attempt to open the “black box” of cost behavior to understand how costs arise.
The activity-based costing (ABC) literature refines the traditional model by recognizing that (1) costs are caused by resources, such as buildings, equipment, direct and indirect labor, and (2) these resources are used to perform activities such as assembly of the finished product, machine setups, processing of customer orders, and product design (e.g., Kaplan and Cooper 1987; Cooper and Kaplan 1992, 1999). By analyzing how costs arise from committed resources and how resource requirements are linked to activities, this approach gains a better understanding of cost behavior. Textbook descriptions of ABC typically assume a mechanistic relation between activities and the associated resources, in which activity changes lead to proportional resource changes (e.g., Garrison et al. 2015; Horngren et al. 2015). Cooper and Kaplan (1992) point out that an activity decrease does not automatically remove unused resources and an activity increase does not automatically add needed resources. Instead, many resources change only if managers make a decision to adjust these resources (e.g., Cooper and Kaplan 1992; Banker and Hughes 1994). For example, managers need to decide whether to fire workers when activity decreases and whether to hire workers when activity increases, and they need to implement these decisions. The ABC literature does not attempt to model how managers make and implement these decisions. In other words, the economic link from activities to committed resources remains a “black box”, which is loosely approximated by assuming a proportional linear relation.
If ABC is more empirically accurate than the traditional model, then the relevant activity cost drivers should have incremental explanatory power for overhead costs conditional on volume. Miller and Vollmann (1985), Kaplan and Cooper (1987), Foster and Gupta (1990), Banker and Johnston (1993), Datar, Kekre, Mukhopadhyay, and Srinivasan (1993), Banker, Potter, and Schroeder (1995), and others examine non-volume-based activity cost drivers. They generally find that these activity cost drivers have considerable incremental explanatory power relative to traditional volume-based drivers. This evidence supports ABC over the traditional model. Because these studies are rooted in the mechanistic view of cost behavior, they do not attempt to examine the role of deliberate managerial decisions in observed cost behavior. Using detailed activity and cost data for hospitals, Noreen and Soderstrom (1994, 1997) test whether overhead costs are strictly proportional to activity, which is a key “black-box” assumption in ABC. They reject this assumption, which suggests that managers' resource adjustments are not well approximated by a proportional linear relation with activity. These findings challenge the standard “black-box” approach and suggest that a different approach is needed.
Noreen (1991), Banker and Hughes (1994), Banker and Hansen (2002), Banker, Hwang, and Mishra (2002), Balakrishnan and Sivaramakrishnan (1996), Balachandran, Balakrishnan, and Sivaramakrishnan (1997), Göx (2000), Degraeve, Labro, and Roodhooft (2005), and others analyze whether allocated costs from ABC provide useful information for managers' pricing and product mix decisions. These analytical studies focus on decision making conditional on the cost structure information from ABC. Most papers treat the cost structure itself as given and do not explicitly explore how the cost structure arises from managerial decisions. Further, the resource commitments that determine costs at a given activity level might vary with the selling prices and the product mix. Therefore, it might be conceptually important to incorporate cost structure decisions in the analysis of pricing and product mix decisions.6
The most important conceptual foundation of contemporary cost management research is the insight that costs arise from resource commitment decisions by managers (Cooper and Kaplan 1992). To understand the economic nature of observed cost behavior, one must focus on the determinants and the consequences of these decisions. Managers choose resource levels subject to various constraints (e.g., demand conditions, production technology, resource adjustment costs, strength of corporate governance, debt covenants, government regulation), incentives (e.g., performance compensation, earnings targets, ownership type, stakeholder activism), and biases (e.g., overconfidence), which we review in detail in the next section. The shift in focus from mechanical cost drivers to deliberate cost management decisions gives rise to a powerful new way of thinking about cost behavior, which enables researchers to use observed variation in cost behavior to analyze a very broad range of phenomena that affect managerial decisions.
It is important to recognize that managers do not directly choose fixed, variable, or sticky costs.7 Instead, these cost behavior patterns arise from managers' decisions to commit resources subject to context-specific constraints. For example, direct labor costs are not fundamentally fixed or variable. They might be fixed in Western Europe and variable in the U.S. or China, depending on whether a country's labor laws prevent managers from laying off workers at will (e.g., Garrison et al. 2015, 33). Further, the behavior of direct labor costs in a given country might vary with firm-specific factors such as training costs for new hires or union contracts (e.g., Banker et al. 2013b).
Modeling the economic foundations of cost management overturns some of the major prior predictions. For example, the established traditional intuition in accounting research and textbooks has been that when managers face a more uncertain demand, they choose a more flexible cost structure with lower fixed and higher variable costs (e.g., Balakrishnan, Sivaramakrishnan, and Sprinkle 2008, 171). However, the proportion of fixed costs is not a direct decision variable for managers. Banker, Byzalov, and Plehn-Dujowich (2014b) model how the mix of fixed and variable costs arises from optimal capacity commitments by managers under demand uncertainty. They assume a standard translog production technology with a fixed capacity resource that is chosen in advance and a variable resource that is chosen after the demand is realized. When demand is high relative to capacity, this production technology has high congestion costs due to strained capacity (Banker, Datar, and Kekre 1988). Therefore, managers prefer to commit sufficient capacity in advance to avoid excessive congestion costs for high demand realizations. When demand uncertainty is greater, both unusually low and unusually high demand realizations become more likely. The greater likelihood of unusually high demand realizations increases the expected congestion costs, leading managers to choose higher capacity to mitigate these costs. Higher capacity is associated with higher fixed costs. Therefore, the proportion of fixed costs should increase with demand uncertainty, contrary to the traditional intuition. Banker et al.'s (2014b) empirical results for the manufacturing sector in the U.S. support their “counterintuitive” prediction.
Notably, additional economic factors can reverse this prediction. For example, Holzhacker, Krishnan, and Mahlendorf (2015a) examine the relation between demand uncertainty and cost behavior in hospitals. Because most hospitals are non-profits, hospital managers are less concerned about mitigating congestion costs to increase profit for favorable demand realizations and are more concerned about avoiding large losses for unfavorable demand realizations. Therefore, Banker et al.'s (2014b) congestion cost argument is much less applicable to hospitals and other non-profits, and the traditional loss-avoidance argument likely dominates. As expected, Holzhacker et al.'s (2015a) findings are consistent with the traditional intuition.8 These results demonstrate that a researcher should tailor the theory to match the specific incentives and constraints in his/her research context, recognizing that these details can change the predictions.
Another important conceptual foundation of contemporary cost management research is the resource adjustment costs. In the traditional view, variable costs represent resources that can be adjusted flexibly in the short run, while fixed costs represent resources that cannot be adjusted in the short run (Cooper and Kaplan 1992). In other words, the resource adjustment costs are either negligible or prohibitive, respectively. However, many resources fall between these two extremes. For example, managers can hire and fire skilled workers relatively quickly, but they need to incur significant adjustment costs such as search, screening, and training costs for new hires and severance payments to laid-off workers.9,10 The resource adjustment costs give rise to dynamic effects in the relation between activity and resources. Managers' choice of resource level depends not only on the concurrent activity, as in the traditional model, but also on the prior resource level (which affects the adjustment costs incurred in the current period) and expected future activity (which affects future adjustment costs), as well as managers' incentives and behavioral biases.
These dynamic effects of adjustment costs give rise to asymmetric cost behavior (i.e., sticky and anti-sticky costs), among other consequences, with predictable variation in both the direction and the magnitude of cost asymmetry. We briefly explain the intuition for these results in the next section. We refer the reader to Banker and Byzalov (2014) for an in-depth review of the economic theory of asymmetric cost behavior, and Noreen (2016) for an analytical model of cost asymmetry. However, we emphasize that asymmetric cost behavior is just one of many interesting manifestations of cost management decisions.
WHAT HAVE WE LEARNED ABOUT COST MANAGEMENT?
In this section, we review the main predictions and empirical findings on cost management and its determinants. We use the seminal article on sticky costs by Anderson et al. (2003) as a starting point because it laid the foundation for most of the recent cost management research (some of the notable earlier studies are discussed in the previous section). Although a large majority of published and working papers in this research area use empirical measures of asymmetric cost behavior, we note that Anderson et al.'s (2003) decision-based paradigm is much broader than just “sticky costs” or “asymmetric cost behavior.” Therefore, we incorporate cost management studies that adopt this fundamental decision-based paradigm but do not focus on asymmetric cost behavior; and we encourage future research that strays from the prevailing asymmetric cost behavior approach and examines other interesting facets of cost management.
Anderson et al. (2003) predict that many major cost categories are sticky on average, i.e., these costs fall less for sales decreases than they rise for equal sales increases. When sales volume decreases, managers keep some unused resources to avoid the resource adjustment costs. In contrast, when sales volume increases, managers cannot supply the required volume unless they add the needed resources. Therefore, on average, costs should be less sensitive to sales decreases than to sales increases, using sales revenue as a proxy for physical sales volume. Because this prediction arises from managerial decisions with resource adjustment costs, it enables a researcher to detect deliberate cost management decisions through observed cost behavior.
As predicted, empirical research finds cost stickiness on average for SG&A costs (Anderson et al. 2003), COGS and total operating costs (e.g., Subramaniam and Weidenmier 2003; Weiss 2010; Kama and Weiss 2013), operating cash flow (Banker et al. 2016d), and cash expenses (Shust and Weiss 2014) in U.S. Compustat data; for operating costs in broad-based cross-country samples (Calleja, Steliaros, and Thomas 2006; Banker and Byzalov 2014); and for various cost categories in country-specific samples for Australia (Bugeja, Lu, and Shan 2015), Brazil (Richartz and Borgert 2014), China (Bu, Wen, and Banker 2015; Liang, Chen, and Hu 2014; Xue and Hong 2016; Yang and Zhao 2016), Egypt (Ezat 2014; Ibrahim 2015), Indonesia (Warganegara and Tamara 2014), Italy (Dalla Via and Perego 2014), Japan (Yasukata and Kajiwara 2011), Jordan (Magheed 2016), Korea (Lee 2015), Philippines (Uy 2014, 2016), Spain (Werbin, Vinuesa, and Porporato 2012), and Turkey (Yükçü and Özkaya 2011).11 Industry-specific studies of the healthcare sector find cost stickiness in physical therapy clinics (Balakrishnan, Petersen, and Soderstrom 2004) and hospitals (Balakrishnan and Gruca 2008; Balakrishnan and Soderstrom 2009; Holzhacker, Krishnan, and Mahlendorf 2015b).
Several studies document cost stickiness for physical measures that are less susceptible to a potential confounding effect of changes in prices and product mix, such as total employees and total hours (e.g., Dierynck et al. 2012), detailed measures of individual activities (Banker and Liu 2015), and mix of employees (Kong, Liu, and Shen 2015). Using detailed data on airlines' physical capacity and activity, Cannon (2014) shows that cost stickiness for airlines arises because they add capacity when activity increases and keep unused capacity when activity decreases. This evidence directly validates Anderson et al.'s (2003) intuition that cost stickiness reflects an asymmetry in physical resource adjustment by managers.
These findings reject the traditional mechanistic model of fixed and variable costs and provide evidence of deliberate cost management decisions that manifest in observed cost stickiness. However, it is important to recognize that not all costs must be sticky. For example, when the adjustment costs are negligible, one should not expect cost stickiness (e.g., Banker and Byzalov 2014). Further, costs that are sticky on average are not always sticky. Using a firm-year measure of cost asymmetry, Weiss (2010) finds that SG&A costs are anti-sticky (i.e., more sensitive to sales decreases than to sales increases) in 45 percent of his sample. Banker, Byzalov, Ciftci, and Mashruwala (2014a) show that SG&A costs, COGS, and total employees are sticky following a prior sales increase and are anti-sticky following a prior sales decrease. These patterns are primarily due to managerial optimism and pessimism, respectively, which we discuss later. Because sales increases outnumber decreases 68 percent to 32 percent in Banker et al.'s (2014a) sample, on average cost stickiness for prior increases dominates, thus reproducing the standard findings of cost stickiness on average.12 These findings illustrate that a researcher should not mechanically adopt a “black-box model of sticky costs” as a generic substitute for the traditional “black-box” model; instead, we encourage researchers to look inside the “black box” and think about the underlying managerial decisions in the specific research context. The same should apply to teaching these fundamental concepts of cost behavior to undergraduate or graduate students. Instead of mechanically (and incorrectly) memorizing cost behavior based on account names, students should be taught to think of the trade-offs involved and deduce how different costs would likely behave under different circumstances.
The observed degree of cost stickiness and anti-stickiness varies predictably with the economic determinants of managers' cost management decisions. Some of the papers in this area seek to test and refine the theory of asymmetric cost behavior by finding good empirical proxies for these economic determinants. Many recent studies leverage the established theory of asymmetric cost behavior to study correlates of cost stickiness that are interesting in their own right. In other words, the main objective of these studies is not to provide insights into cost behavior but rather to learn about the economic nature of these particular variables and their role in managerial decisions, using observed cost behavior as a lens to examine the underlying economic mechanisms.
The first major determinant of cost management decisions is the resource adjustment costs. When the adjustment costs are larger, managers are more willing to retain unused resources during sales decreases to avoid these costs. Therefore, cost stickiness should increase with empirical proxies for the magnitude of resource adjustment costs (Anderson et al. 2003; Banker et al. 2013b). Anderson et al. (2003) use asset intensity and employee intensity as firm-level proxies, and Banker et al. (2013b) use the strictness of employment protection laws in OECD countries as a country-level proxy. As predicted, these proxies significantly affect cost stickiness, which confirms that resource adjustment costs play a role in cost management decisions.
Multiple studies leverage this understanding of how the difficulty in resource adjustment manifests in observed cost behavior to examine how various institutional and economic factors influence adjustment costs. Using hospital data, Balakrishnan and Gruca (2008) find that cost stickiness is greater in patient services (a core activity for hospitals) than in support services, which suggests that core activities have larger adjustment costs. Kim and Wang (2016) argue that state-provided unemployment benefits mitigate the negative consequences of layoffs for workers and thus reduce the adjustment costs of layoffs from managers' perspective. As expected, firms in U.S. states with more generous unemployment benefits have lower cost stickiness. Banker, Flasher, and Zhang (2014c) show that firms pursuing the product differentiation strategy have greater cost stickiness than firms pursuing the cost leadership strategy, consistent with greater resource adjustment costs under the product differentiation strategy. Ballas, Naoum, and Vlismas (2015) find similar results for the prospector-defender topology of Miles and Snow (2003). Venieris, Naoum, and Vlismas (2015) argue that firms with high organization capital (using high intangible assets as a proxy) have higher resource adjustment costs, resulting in greater cost stickiness.
The second major determinant of cost management decisions is managers' expectations for future sales. When managers are optimistic about future demand, they are more willing to retain unused resources during sales decreases because they expect to use these resources after the demand rebounds. This tendency to retain resources should increase cost stickiness (when costs are sticky) or reduce cost anti-stickiness (when costs are anti-sticky). When managers are pessimistic about future demand, they are likely to aggressively cut unused resources during sales decreases because they expect these resources to remain unused. This tendency to cut resources should reduce cost stickiness or increase cost anti-stickiness.
Using GDP growth and a dummy variable for successive sales decreases as proxies for managerial expectations, Anderson et al. (2003) find that managerial optimism increases cost stickiness. Banker et al. (2014a) proxy for expectations using different combinations of sales increases and decreases over two consecutive periods. As expected, they find significant cost stickiness in the optimistic scenario (two successive sales increases), significant cost anti-stickiness in the pessimistic scenario (two successive sales decreases), and a more moderate degree of asymmetry in the mixed scenarios (an increase followed by a decrease and vice versa). Banker et al. (2016d) show that costs become less sticky on average during economic slowdowns because of managerial pessimism. Chen et al. (2015b) argue that managerial overconfidence leads to excessive optimism about future sales. As expected, they find a positive association between overconfidence proxies and cost stickiness. In additional tests, they show that cost stickiness attributable to the overconfidence proxies is associated with lower future performance, which confirms that these proxies capture excessive (rather than rational) optimism, consistent with overconfidence. Qin et al. (2015) also find a positive relation between overconfidence proxies and cost stickiness. Anderson and Lee (2016) argue that a firm's life cycle stage is associated with managers' expectations for future growth. As expected, firms in the introduction and growth stages have greater cost stickiness than mature firms. Silge and Wohrmann (2016) find lower cost stickiness for firms in the decline stage. Banker, Hwang, and Oh (2016e) use capacity utilization, which is reported by all listed manufacturing firms in Korea, as a proxy for optimism, and find that high capacity utilization is associated with greater cost stickiness. Lee, Pittman, and Saffar (2016) find that cost stickiness increases during election years, which they attribute to political uncertainty that reduces the precision of managers' expectations.
Several studies gauge expectations using information disclosed by managers. Yasukata and Kajiwara (2011) use managers' sales forecasts, which are disclosed by almost all listed firms in Japan. Atasoy, Banker, and Nasev (2016b) use a confidential dataset of German firms that includes managers' self-reported expectations. Chen, Kama, and Lehavy (2016) use the tone of forward-looking statements in companies' 10-K forms. These studies find that managerial optimism increases cost stickiness, as expected. Thus, the empirical evidence consistently shows that cost management decisions in the current period are influenced not only by the concurrent demand changes but also by managers' forward-looking demand expectations. This evidence also suggests that a researcher can use observed cost behavior to extract information about managers' expectations (we review several studies that use this approach in the next section).
The third major determinant of cost management decisions is managerial incentives and their interactions with governance, regulation, and ownership. For example, if managers derive personal utility from empire building, then they will likely add excessive resources when sales increase and will be reluctant to cut unneeded resources when sales decrease. These resource choices can lead to excessive cost stickiness, relative to the value-maximizing level of cost stickiness for the firm, because of the agency problem between empire-building managers and shareholders (Anderson et al. 2003; Chen et al. 2012). Consistent with this agency cost argument, Chen et al. (2012) find a positive association between free cash flow (a proxy for managers' ability to overspend) and cost stickiness, and find that this association is weaker in firms with good corporate governance. Chen, Ni, and Wu (2014) use the adoption of anti-takeover laws in U.S. states as a proxy for changes in corporate governance, but do not find a significant association with cost stickiness. Cannon, Hu, Lee, and Yang (2016) focus on international merger and acquisition laws that increase the takeover threat, and find that these laws reduce cost stickiness, especially for firms with more severe agency problems. Overall, the results in these papers suggest the existence of “bad” cost stickiness that likely reduces firm value through wasteful overspending by managers, as opposed to “good” cost stickiness that contributes to firm value through more efficient resource planning in the presence of resource adjustment costs and expectations about future uncertainty.
Firms can shape managerial incentives through performance compensation. Brüggen and Zehnder (2014) argue that because equity-based executive compensation aligns managers' and shareholders' interests, it encourages “good” cost stickiness and discourages “bad” cost stickiness. They find that cost stickiness increases with the proportion of equity-based compensation, consistent with “good” cost stickiness. Banker, Jin, and Mehta (2016f) find that short-term bonus incentives reduce cost stickiness, whereas long-term equity incentives increase cost stickiness. These results support the value-increasing interpretation of cost stickiness. Further, the results suggest that excessively low cost stickiness constitutes undesirable myopic behavior that temporarily boosts earnings but reduces firm value in the long run. In conjunction with the evidence on the agency problem in Chen et al. (2012) and related papers, these findings suggest that “good” and “bad” cost stickiness co-exist. Some firms likely have high cost stickiness for the wrong reasons (e.g., poor governance), while others have high cost stickiness for the right reasons (e.g., well-designed incentive compensation in a high-growth firm).13 We further note that each firm likely has a value-maximizing level of cost stickiness that is determined by its economic characteristics; both “too much” and “too little” cost stickiness relative to this firm-specific optimum can destroy value. The literature has not yet resolved this issue.
Managers are likely to contain costs more aggressively when they need to meet a particular performance benchmark. Dierynck et al. (2012) and Kama and Weiss (2013) show that cost stickiness is lower when managers have a strong incentive to manage earnings to avoid a loss or an earnings decrease. Banker and Fang (2016) examine firms that obtain new loan financing. These firms reduce cost stickiness prior to loan approval to improve financial performance. After the loan has been approved, cost stickiness increases but is attenuated for firms that face more financial covenants on earnings.14 Homburg, Nasev, and Richter (2015a) show that financial distress is associated with lower cost stickiness because managers have a strong incentive to reduce costs when the firm's survival is at stake. Homburg, Nasev, and Swam (2015b) argue that credit ratings give managers a strong incentive to contain costs when a firm's rating is right above the investment grade cutoff. They find lower cost stickiness above this cutoff, as expected. Thus, cost management decisions vary predictably with specific performance benchmarks faced by managers.
Managerial incentives interact with regulation and ownership. Holzhacker et al. (2015b) study how the transition from cost-based reimbursement to fixed-price reimbursement in German hospitals affected cost behavior. This regulatory change gave hospitals a stronger incentive to contain costs during demand decreases, reducing cost stickiness. Holzhacker et al. (2015b) also predict and find that this decrease in cost stickiness is more pronounced in for-profit hospitals than in non-profit and government hospitals, because the latter two ownership types face stronger institutional pressures against cost cutting. Bu et al. (2015) and Gu, Tang, and Wu (2016) show that state ownership increases cost stickiness in China. Prabowo, Hooghiemstra, and Van Veen-Dirks (2016) find similar results for European firms. Gu et al. (2016) and Prabowo et al. (2016) find that the effect of state ownership on cost stickiness is stronger when managers face greater political pressure to avoid layoffs, such as when government officials in China are up for a promotion or politicians in Europe are up for re-election. Ezat (2014) and Ibrahim (2015) find that institutional and state ownership increase cost stickiness in Egypt. Hall (2016) examines how the ownership structure in U.S. banks affects cost behavior. Publicly owned banks face greater investor pressure to report favorable earnings than privately owned banks, and they are better able to raise regulatory capital through equity sales. Therefore, publicly owned banks reduce cost stickiness to a greater extent than private banks when they need to meet an earnings benchmark, but they reduce cost stickiness to a lesser extent than private banks when they need to comply with regulatory capital requirements, consistent with the distinct incentives for these ownership types. As Hall's (2016) paper illustrates, the effect of ownership and other institutional factors cannot be reduced to a generic one-size-fits-all recipe; instead, a researcher should analyze the incentive effects of these institutional factors in the specific research context.
Managerial incentives are influenced by managers' personal characteristics and behavioral biases. As mentioned above, Chen et al. (2015b) and Qin et al. (2015) argue that overconfident managers suffer from an optimistic bias. Therefore, these managers have greater psychological incentives to keep unused resources during sales decreases, resulting in greater cost stickiness. Yang (2015) finds that CEO hubris is associated with greater cost stickiness. Liang, Zhao, and Wang (2015) argue that male managers and younger managers are more aggressive and less risk averse, and find that these managers' firms tend to have greater cost stickiness. Gores, Homburg, and Nasev (2015) show that managers' resource commitments are influenced by the stock market sentiment (i.e., investors' psychological biases that can include both irrational optimism and irrational pessimism in different periods). Using cross-country data, Kitching, Mashruwala, and Pevzner (2016) predict and find that cost stickiness varies with the national culture attributes such as uncertainty avoidance, masculinity, and long-term orientation. For example, because future demand is uncertain, managers in uncertainty-avoidant cultures likely adopt a shorter-term view. Therefore, they are less willing to retain unused resources during a sales decrease to accommodate expected future demand, resulting in lower cost stickiness. Thus, observed cost behavior enables researchers to gain broader insight into managers' personality and psychological biases.
Managerial incentives could interact with additional firm characteristics. K. Liu, X. Liu, and Reid (2017) focus on a firm's stakeholder orientation and find that both customer-oriented firms and employee-oriented firms exhibit greater cost stickiness. These results could reflect either higher resource adjustment costs or a more severe agency problem for high stakeholder-orientation firms. In additional tests, the effect of customer orientation is more consistent with the adjustment cost explanation while the effect of employee orientation is more consistent with the agency problem explanation, thus revealing the economic nature of stakeholder orientation. Golden and Rezaee (2016) find that corporate sustainability is positively associated with cost stickiness. Habib and Hasan (2016) find a positive association between corporate social responsibility and cost stickiness, while Paek, T. Kim, and H. Kim (2016) find a negative association. Xu and Zheng (2016) examine the relation between tax avoidance and cost stickiness.
In addition to the research literature in English that we focus on, there is also a rapidly growing international literature in other languages, which reflects a growing international interest in cost management research. As an example, the online supplement lists 152 papers written in Chinese and 81 papers written in Korean (see Appendix A for the link to the downloadable file). As this international literature matures, some of the studies go beyond just documenting generic determinants of cost stickiness in country X and instead leverage distinctive local institutions and regulations to learn about managerial decisions, economic institutions, and psychological biases.
While most of the recent cost management research uses the cost asymmetry framework, several recent studies examine cost management decisions in the context of fixed and variable costs. As mentioned earlier, Banker et al. (2014b) and Holzhacker et al. (2015a) analyze how managerial decisions in response to demand uncertainty affect a firm's cost structure. These papers use log-linear models that measure cost elasticity (i.e., the average sensitivity of costs to sales changes of both signs) and do not include cost asymmetry in the main tests. Both papers tackle the established textbook intuition about the effect of demand uncertainty on the mix of fixed and variable costs, which is best captured empirically using cost elasticity. Given this research context, these papers do not need to focus on asymmetric cost behavior; furthermore, incorporating cost asymmetry would have diluted these papers' impact. Aboody, Levi, and Weiss (2016) examine how option-based compensation affects managers' choice of operating leverage (the ratio of fixed to variable costs) through managers' risk-taking incentives. Operating leverage directly reflects the risk embedded in the cost structure and is thus an important cost structure characteristic in the context of risk-taking incentives.15 Holzhacker et al. (2015b), Hall (2016), and others conduct large parts of the empirical analysis using cost elasticity and add cost asymmetry only when it is pertinent to the research question. Thus, a researcher should not feel compelled to use empirical measures of asymmetric cost behavior if these measures are not congruent with the research context. Further, when the cost asymmetry framework is appropriate, the degree of cost stickiness or anti-stickiness is not necessarily the only measure of interest. In many applications, the cost response to sales increases or the implications for context-specific financial ratios could be either as relevant or more relevant than the degree of cost asymmetry. More important, and regardless of the cost behavior measures used, the impact of empirical findings is enhanced considerably when the researcher is able to leverage the findings to learn about the fundamental features of managerial decision making or generate new insights about institutional, economic, or behavioral factors that are interesting in their own right.
Alternative Methodologies in Cost Management Research
Empirical research typically measures cost asymmetry using Anderson et al.'s (2003) cost stickiness model and its direct extensions such as Banker et al.'s (2014a) two-period model. Several studies propose alternative measures. Weiss (2010) develops a firm-level measure of cost asymmetry that is based on a rolling window of four observations.16 Banker, Basu, and Byzalov (2013a) and Kaspereit and Lopatta (2016) propose regression-based firm-year cost stickiness scores. These measures are based on the predicted degree of asymmetry in a regression model that incorporates higher-order interactions with the standard empirical determinants of cost stickiness, such as asset and employee intensity. This measurement approach is analogous to the Khan and Watts (2009) firm-year conservatism score (C_Score). We caution that these firm-year scores only capture variation associated with the standard determinants of cost stickiness. Therefore, these scores should not be used as dependent variables in studies that examine new determinants of cost management decisions, because by definition new determinants must be incremental to the standard explanatory variables.17 A more appropriate way to study new determinants is to include them as additional interaction terms in a model that also includes interactions with the standard explanatory variables. A researcher can potentially use the firm-year scores to examine the effect of cost stickiness on other outcomes. However, this approach requires a strong assumption that the component variables of the firm-year score, such as asset and employee intensity, do not directly affect the outcome of interest and are not correlated with its omitted determinants, similar to the standard exclusion restriction for instrumental variables. If this assumption does not hold, then the effect of the cost stickiness score cannot be reliably identified.18
Banker et al. (2013a) propose an asymmetric cost behavior model for the levels of sales and costs. Because earnings = sales – costs, this levels specification can be directly applied to earnings, unlike the standard log-change specification of Anderson et al. (2003), thus facilitating the analysis of the implications of cost management decisions for earnings behavior. Anderson, Lee, and Mashruwala (2016) propose a two-driver model of asymmetric cost behavior that combines cost stickiness (an asymmetry with respect to activity changes) with cost inertia (an asymmetry due to the adjustment of major assets in place such as property, plant, and equipment). They show that this extension improves the explanatory power considerably.
Shust and Weiss (2014) use cash costs as a proxy for economic costs and find that both economic choices (as reflected in cash costs) and financial reporting choices (as reflected in the accrual component of reported costs) contribute to the asymmetry in reported costs. However, it is difficult to disentangle economic and reporting choices because the accrual component of costs likely conveys useful economic information (e.g., Ball 2013). More generally, managers' reporting decisions and their interaction with accounting rules could confound the estimates. For example, conservatism in revenue recognition could cause nonlinear measurement error in reported sales revenue as a proxy for activity, while conservatism in expense recognition could cause nonlinear measurement error in reported expenses as a proxy for economic costs (e.g., Banker et al. 2013a). Incomplete matching between revenue and expenses could add further measurement error (e.g., Folsom and Paek 2016). For capital resources, acquisition adjustment costs incurred to bring the resource to the condition and location of its intended use (e.g., shipping and installation costs) are capitalized, whereas disposal adjustment costs are expensed when incurred. These accounting-related sources of systematic measurement error in reported sales and costs could distort empirical estimates. Understanding the interplay between financial reporting decisions and cost management decisions is an interesting topic for future research.
An important methodological challenge for both theoretical and empirical cost management research is to incorporate the output pricing decisions by managers and the resource market factors that affect the input prices. When demand increases (i.e., the demand curve shifts to the right, increasing the quantity demanded at any given price), managers can either increase the physical sales volume or increase the selling price, or both. Many cost management determinants from the previous subsection could influence managers' pricing decisions. For example, when the resource adjustment costs are large, managers might prefer to adjust the selling price while maintaining stable resource levels and sales volume. This scenario would manifest as a weak association between sales revenue (price × volume) and costs in the data. When managers view a demand increase as temporary, they might increase the price without adding resources. Contrarily, when they view a demand increase as permanent, they might keep the original price to gain market share and add needed resources. Thus, a given sales revenue increase might correspond to a larger activity increase and thus higher costs in the latter case. Further, cost management could cause an asymmetry in managers' pricing decisions. For example, in many theoretical and empirical models in the industrial organization literature, the optimal selling price equals the marginal cost plus a markup percentage that is determined by the demand parameters (e.g., Shy 1995, Chaps. 5–7; Berry, Levinsohn, and Pakes 2004). If activity changes have a nonlinear effect on marginal costs through asymmetric resource adjustment, then the optimal price is a nonlinear function of demand changes. This price nonlinearity could confound cost asymmetry estimates (e.g., Cannon 2014).
The marginal costs of resources are influenced by the demand and supply conditions in the resource markets. Assuming an upward-sloping supply curve for a resource, the resource price varies with the aggregate demand for the resource. For example, the wage (i.e., resource price) of skilled production workers varies with the total demand for skilled production workers from all manufacturing firms. The implications for firm-level cost behavior are sensitive to the correlation structure of demand across firms. If the demand for different firms' output is positively correlated (e.g., all firms face high demand during macroeconomic booms and low demand during recessions), then each firms' demand in the output market is positively correlated with the aggregate demand in the resource market. Therefore, the firms will tend to face higher resource prices during firm-level demand increases and lower resource prices during firm-level demand decreases. Contrarily, if the demand for different firms' output is uncorrelated, then each firm's demand in the output market is uncorrelated with the aggregate demand in the resource market. Therefore, the firms will face similar resource prices on average during firm-level demand increases and decreases. For some resources, managers could also enter a long-term sourcing arrangement with a supplier to prevent resource price fluctuations, or could combine multiple supply sources to take advantage of price fluctuations across markets.
Resource market economics could also influence the adjustment costs for resources such as capital equipment. If the demand for equipment is strongly positively correlated across firms, then the firms will want to add new equipment all at the same time, causing unusually high equipment purchase prices, and will want to dispose of unused equipment all at the same time, causing unusually low equipment resale prices. This correlation can increase both the purchase adjustment costs and the disposal adjustment costs for equipment. Understanding the implications of output pricing decisions and resource market economics is an interesting challenge for future research.
The empirical cost management research could benefit from research designs based on natural experiments, especially in studies of the implications of observed cost behavior for other outcomes. An acceptable natural experiment must satisfy the standard exclusion restrictions for instrumental variables, i.e., it must not directly affect the outcome of interest and must not be correlated with its omitted determinants. We strongly caution against the mechanical use of instrumental variable estimation, or the equivalent estimation of a system of equations, as a silver-bullet solution for endogeneity. If an instrument does not satisfy the exclusion restrictions, then instrumental variable estimation can amplify the endogeneity problem and lead to much more severe bias than ordinary least squares (e.g., Larcker and Rusticus 2010; Lennox, Francis, and Wang 2012).19
The standard theory of asymmetric cost behavior describes resource adjustment in for-profit firms. It can—and should—be modified to fit some other research contexts, such as non-profit and government organizations. For example, because sales revenue = price × volume, revenue of for-profit firms is directly associated with the physical activity volume that consumes resources. However, revenue is a much less appropriate proxy of activity in non-profits that generate revenue through donations and in government organizations that generate revenue through taxes (e.g., local and state governments) or get a direct budget allocation (e.g., public schools). For these entities, revenue does not directly reflect activity (e.g., tax revenue for a local government does not represent sales of municipal services to individual taxpayers). Instead, revenue only has an indirect association with activity through the organization's budget constraint. For example, when a local government faces a tax revenue shortfall, it might be forced to cut municipal services. However, it could also choose to maintain the original service level and run a temporary budget deficit. Thus, the relation between revenue and activity for non-profit and government organizations involves major additional decisions that are not covered by the standard theory for for-profit firms. These organizations also have distinct incentives and institutional pressures. Therefore, studies of non-profits (e.g., Banker and Harris 2016; Wu, Young, Yu, and Hsu 2016) and government organizations (e.g., Bradbury and Scott 2015; Cohen, Karatzimas, and Naoum 2017) could benefit from tailoring the theory to fit the specific context and incentives in their research settings.
The standard theory of asymmetric cost behavior describes the adjustment of resources that are consumed based on concurrent activity levels. A sales decrease directly reduces the resource requirements, resulting in resource slack. Therefore, managers need to choose between keeping and eliminating the slack resources, giving rise to asymmetric cost behavior. This link between sales decreases and resource slack occupies a central place in the theory. However, some major resources such as R&D employees and skilled marketing employees do not fit this description. A sales decrease does not free up R&D and marketing resources; on the contrary, the demands on these resources might increase. Therefore, the standard resource slack argument does not apply. Further, R&D and marketing resources often constitute a strategic long-term investment to generate future revenue. When managers cut these resources, they do not just incur adjustment costs such as severance payments; more important, they sacrifice future revenue. Therefore, a researcher should not mechanically apply the standard theory of asymmetric cost behavior “as is” to analyze these resources.
Instead, the researcher should modify the theory to incorporate the trade-off between the current costs and the future benefits of strategic investment in R&D and marketing resources, with a careful consideration of how this trade-off changes during sales decreases. Many of the standard predictions will likely continue to hold in this modified theory. For example, managers' investment in R&D and marketing resources will likely vary with resource adjustment costs, managerial expectations, governance, performance incentives, and behavioral factors, similar to the standard theory. However, because the resource requirements are not physically tied to concurrent activity, the dynamic relation between sales changes and resource changes could potentially be much more complex. For example, if a firm's sales decrease because the firm is falling behind competitors in new product development, then the optimal response to this sales decrease could be either a large increase in R&D investment that will improve future sales with a lag of several years, or a large decrease in R&D investment that will be followed by gradual sales decreases as the firm prepares to exit the market. Therefore, while many of the standard empirical predictions might continue to hold, the empirical tests should be extended to incorporate the dynamic role of resource investment as a driver of future revenue. This extension might require major theoretical and empirical modifications. Developing this theory and the associated empirical tests could be an interesting challenge for future research.
WHAT HAVE WE LEARNED ABOUT THE IMPLICATIONS OF COST MANAGEMENT?
A very interesting research direction is exploring the implications of cost management for various topics in financial and managerial accounting that require interpreting observed costs or earnings, predicting future costs or earnings, or understanding managers' operating decisions more generally.
Implications of Cost Management Decisions for Earnings Properties
Because earnings = sales – costs, cost management decisions that manifest in cost behavior directly affect earnings properties such as earnings predictability, asymmetric timeliness, and persistence. Weiss (2010) argues that because cost stickiness increases the sensitivity of earnings to sales changes, future earnings in sticky cost firms are more sensitive to unexpected future sales changes. Therefore, for the same conditional variance of future sales changes, based on the information available in the current period, future earnings in these firms have higher conditional variance, i.e., lower predictability. This decrease in earnings predictability should reduce analysts' forecast accuracy. As predicted, Weiss (2010) finds that firms with greater cost stickiness tend to have less accurate analyst forecasts. Banker et al. (2016a) argue that cost management decisions that give rise to cost stickiness (an asymmetry in operations) can be mistaken for conditional conservatism (an asymmetry in accounting recognition, e.g., Basu ) because both phenomena affect earnings asymmetrically. As predicted, they find that cost stickiness biases standard conservatism measures and distorts inferences about various sources of variation in conservatism. Oded and Weiss (2013) find that real economic choices lead to variation in asymmetric timeliness of earnings. Hartlieb and Loy (2016) document that cost stickiness is negatively associated with standard measures of earnings smoothing. Banker and Liang (2017) find that managers' operating decisions in response to indicators such as losses, sales decreases, earnings decreases, and negative operating cash flows affect earnings persistence, earnings volatility, and earnings forecast accuracy. Homburg, Nasev, Reimer, and Uhrig-Homburg (2016) show that cost stickiness increases a firm's credit risk because it increases earnings volatility. Ciftci and Salama (2016) argue that investors and analysts demand more information from sticky cost firms because cost stickiness reduces earnings predictability, and they find that these firms are more likely to issue management earnings forecasts.
Implications of the Predictive Value of Cost Management Decisions for Future Earnings
Cost management decisions affect the time-series properties of earnings, i.e., the association between current period information and future earnings. Therefore, observed cost behavior can have predictive value for future earnings. Banker and Chen (2006) develop an earnings prediction model that incorporates cost variability and cost stickiness and show that this model outperforms other standard earnings prediction models.20 Aboody, Levi, and Weiss (2014) show that firm-level operating leverage has asymmetric predictive value for future earnings. Atasoy et al. (2016b) find that managerial expectations have asymmetric predictive value for future earnings. Banker, Chen, and Park (2016c) show that analysts' earnings forecasts are more consistent with the cost variability and cost stickiness model of Banker and Chen (2006) than with the traditional model of fixed and variable costs or a simpler proportional cost model. This finding suggests that analysts incorporate some (but not necessarily full) understanding of asymmetric cost behavior in their earnings forecasts. Yasukata (2013) examines managers' sales and earnings forecasts in Japan, where almost all of the publicly traded firms disclose both of these forecasts. He finds that managers' earnings forecasts are an asymmetric function of the concurrent sales forecast, which indicates that managers incorporate cost stickiness in their earnings forecasts.
If managers and analysts do not fully exploit the predictive value of observed cost behavior in forecasting earnings, then their reported forecasts will have a systematic bias that can vary with sales changes. Yasukata (2013) finds that managers underestimate future costs for both sales increases and sales decreases, which leads to a consistent optimistic bias in their earnings forecasts. Banker, Park, and Zhong (2016h) show that analysts' relative forecast accuracy is persistent over time, which they attribute to systematic differences in analysts' ability to understand cost behavior. Ciftci, Mashruwala, and Weiss (2016) document an asymmetric relation between analysts' ex post sales forecast errors and ex post earnings forecast errors. They claim that this relation between ex post forecast errors indicates that analysts do not fully understand asymmetric cost behavior.21
If investors do not fully incorporate the predictive value of cost behavior in their earnings expectations, then their expectation errors will lead to systematic abnormal returns around earnings announcements. Banker, Kama, and Weiss (2012) find that stock prices do not fully incorporate cost asymmetry information in a timely manner. This incomplete stock price response gives rise to abnormal stock returns that are consistent with the post-earnings-announcement drift (PEAD) anomaly. Huang, Jiang, Tu, and Zhou (2016) report that firms with high growth in operating costs have significantly lower future abnormal returns than other firms. They attribute this finding to investors' inattention to asymmetric cost behavior and argue that it is consistent with PEAD.
Implications of Cost Management Decisions for the Interpretation of Financial Ratios in Fundamental Analysis
Better understanding of cost management affects the interpretation of standard financial ratios in fundamental analysis because these ratios are directly influenced by managers' cost management decisions. Anderson et al. (2007) focus on the SG&A cost ratio. An increase in this ratio is interpreted as an unfavorable signal about future earnings in fundamental analysis (Lev and Thiagarajan 1993). However, Anderson et al. (2007) argue that this signal conveys favorable information during sales decreases because managers are more likely to retain slack SG&A resources when they are optimistic about future demand. As predicted, they find that SG&A cost ratio increases constitute a negative signal during sales increases but constitute a positive signal during sales decreases. Baumgarten, Bonenkamp, and Homburg (2010) use the industry-average SG&A cost ratio to classify firms with an increase in the ratio into intended and unintended. They find that intended increases in the SG&A ratio significantly increase future profitability. Johnson (2016) finds that the interpretation of the SG&A cost ratio varies across finer-grained combinations of changes in sales and SG&A costs. Anderson and Yu (2016) show that the interpretation of additional fundamental analysis signals differs between sales increases and decreases. Hwang, Lee, and Yang (2016) find that when managers increase inventory during a sales decrease, this inventory decision predicts subsequent sales increases and is consistent with managerial optimism. Banker et al. (2016e) report that the operating performance ratios of sales-decrease firms improve during recessions because managers are more pessimistic and are thus willing to cut resources more aggressively for a given sales decrease. In other words, a current profitability increase can signal bad news as a proxy for managerial pessimism, while a current profitability decrease can signal good news as a proxy for optimism. Banker and Park (2016) show that investors misinterpret how changes in profitability among sales-decrease firms relate to managerial optimism and pessimism, which results in positive abnormal returns for sales-decrease firms with both the lowest and the highest changes in profitability.
Implications of Cost Management Decisions for Additional Phenomena in Accounting and Other Fields
Managers' operating decisions manifest in additional facets of firm performance beyond costs and earnings. Therefore, a researcher can leverage the insights from the cost management research to analyze a broad range of phenomena that are not directly related to costs and earnings. For example, Banker et al. (2016b) predict and find that managers' operating decisions in response to sales decreases have an asymmetric effect on major accrual components such as accounts receivable, inventory, and accounts payable. These operating asymmetries lead to predictable nonlinear bias in standard discretionary accrual models (e.g., Jones 1991; Dechow, Sloan, and Sweeney 1995; McNichols 2002). Banker et al. (2016b) show that alleviating this bias changes some of the major findings in earnings management research. Hwang et al. (2016) find that the asymmetry in inventory adjustment during sales decreases varies with managerial optimism and pessimism, operating cycle length, and sales volatility, consistent with forward-looking inventory planning decisions by managers.22 Gupta, Pevzner, and Seethamraju (2015) show that manufacturing firms' cost structure prior to the economic crisis of 2008–2009 influenced their inventory management during the crisis.
Cost management decisions have implications for various topics in managerial accounting. For example, Caylor and Lopez (2013) show that compensation committees take asymmetric cost behavior into account when designing executive compensation contracts. If a compensation committee recognizes that sticky costs disproportionately reduce earnings during sales decreases, then the CEO is less likely to be penalized for ROA decreases attributable to normal cost stickiness. Banker et al. (2016f) find that cost stickiness affects the choice of short-term versus long-term incentive compensation.23 Banker et al. (2013a) show that asymmetric cost behavior affects the major applications of the cost-volume-profit (CVP) analysis.
Holzhacker et al. (2015a), Atasoy et al. (2016a), and Banker et al. (2016g) examine how demand uncertainty affects managers' outsourcing decisions that shape the boundaries of the firm. Managerial and cost accounting textbooks typically discuss outsourcing in the context of the make-or-buy decision for production components, reducing this decision to a simple comparison of total costs under “make” and “buy” at a given demand level. Outsourcing changes a firm's cost structure by replacing the costs of internally committed resources with payments to a supplier (e.g., Holzhacker et al. 2015a). It can also affect managers' ability to react to demand changes. For example, outsourcing can increase the production lead time, reducing managers' ability to cope with unexpected demand fluctuations (e.g., Atasoy et al. 2016a). It can also remove some of the institutional impediments to quick resource adjustment, enabling managers to better cope with demand changes (e.g., Banker et al. 2016g). Thus, a researcher can use the economic insights from cost management research to better understand the outsourcing decisions and the boundaries of the firm, and can use recent results on outsourcing to gain insight into forces that affect short-term cost management decisions.
There is a large analytical literature on various supply chain issues in the operations management research. One of the key topics in this literature is the bullwhip effect (Lee, Padmanabhan, and Whang 1997), or amplification of demand shocks as they move upstream through the supply chain, and various supply chain arrangements that could mitigate this effect (e.g., Cachon and Fisher 2000; Aviv 2007; Bray and Mendelson 2012; and many others). This research does not consider resource adjustment costs and how they interact with bullwhip-related variability, while the cost management research does not consider the downstream-upstream interactions that cause the bullwhip effect, suggesting opportunities for new insights in both areas. Another major issue in this literature is strategic sourcing decisions, such as the choice between an “efficient” supplier with low costs and a “responsive” supplier with a short lead time (e.g., Song and Zipkin 2009; Allon and Van Mieghem 2010; Janakiraman, Seshadri, and Sheopuri 2015; Boute and Van Mieghem 2015), the choice between long-term contracts and spot market purchases (e.g., Lee and Whang 2002; Mendelson and Tunca 2007), and the strategic consequences of outsourcing (e.g., Feng and Lu 2012). While these studies use sophisticated analytical models of transaction costs (Williamson 1979), they ignore managerial incentives and biases studied in the cost management research. Therefore, the insights on managerial behavior from cost management research could make a major conceptual contribution to the supply chain literature, while the insights on the supply chain issues could enrich cost management research.24
The resource-based view in the strategy literature analyzes the firm as a bundle of resources that provide a competitive advantage (e.g., Wernerfelt 1984; Peteraf 1993). Strategy research shows that firms engage in mergers and acquisitions and form alliances to gain access to other firms' valuable resources (e.g., Hennart and Reddy 1997; Kogut 1988; Rouse and Daellenbach 1999). Understanding how resource adjustment costs and cost management decisions interact with the strategic role of resources in creating a competitive advantage could generate interesting new findings in both accounting research and strategy research.
Market structure models in the empirical industrial organization literature (e.g., Berry et al. 2004; Einav and Levin 2010) and in the structural marketing literature (e.g., Chintagunta, Erdem, Rossi, and Wedel 2006; Reiss 2011) combine state-of-the-art estimates of consumer demand with a very simple supply specification with constant marginal costs. This research uses the demand and marginal cost estimates to evaluate various counterfactuals, such as horizontal mergers or the introduction of new products. In many cases, the predictions for prices play a central role in the counterfactual analysis (e.g., will a merger significantly increase prices?). Prices are predicted as the marginal cost plus a markup percentage that is determined by the demand elasticity estimates. The findings of asymmetric cost behavior in accounting research suggest that the marginal costs in these counterfactuals should vary with the direction of sales changes, managerial expectations, and other factors that influence cost management decisions. Thus, integrating the insights from cost management research could change the results of many counterfactual analyses in industrial organization and marketing, while insights on market structure and demand characteristics from these areas could advance cost management research.
A large literature in finance examines the implications of operating leverage (i.e., the mix of fixed and variable costs) for various outcomes, including firm risk (e.g., Lev 1974), the trade-off between operating and financial leverage (e.g., Mandelker and Rhee 1984; Kahl, Lunn, and Nilsson 2014; Simintzi et al. 2015; Kumar and Yerramilli 2016), stock return properties (Sagi and Seasholes 2007; Garcia-Feijoo and Jorgensen 2010; Gulen, Xing, and Zhang 2011; Novy-Marx 2011; Donangelo 2014), and cost of equity (e.g., Chen et al. 2011), among others. This finance literature is based on a simple mechanistic view of fixed and variable costs. Therefore, researchers have an opportunity to leverage the understanding of cost management from accounting research to gain interesting new insights on these major issues in finance research.
An important question in the management information systems (MIS) literature is how the advancement of information technology (IT) impacts the business models of IT companies. Many software vendors have changed their licensing models from an on-premises model into an on-demand model, also known as software-as-a-service (Benlian and Hess 2011). With the help of high-capacity networks, not only software but also hardware can be provided on demand, which enables cloud computing (Armbrust et al. 2010). Theoretical research in this area examines how on-demand services influence IT companies' dynamic pricing and quality management strategies (e.g., Choudhary 2007; Fan, Kumar, and Whinston 2009). However, the client companies' cost management incentive is largely ignored. For example, if the manager of a client company is more worried about high resource commitment, then on-demand service can be more attractive, which can further affect the optimal pricing of the service. Hence, understanding the clients' cost management decisions can lead to better software pricing strategies in the MIS literature.
An emerging stream of marketing and MIS literature studies the economic impact of the sharing economy firms, such as Uber Technologies Inc., Airbnb Inc., and cloud computing vendors (e.g., Jiang and Tian 2016). The sharing economy likely interacts with cost management decisions. Because companies can avoid many adjustment costs when they access resources through the sharing economy, they can better manage costs. However, the sharing economy also hurts the incumbents by increasing competition. For example, Uber reduces the demand for taxis (Wallsten 2015) and Airbnb reduces hotel sales (Zervas, Proserpio, and Byers 2017). Therefore, managers in the incumbent firms need to adjust resources to cope with competition from the sharing economy firms, which often use a very different business model. Understanding these competitive effects is an interesting challenge both for cost management research in accounting and for marketing and MIS research.
Insight from cost management research could contribute to the literature on organizational forms. For example, franchising is an important but controversial form of vertical integration that has been studied in economics, finance, and marketing. The literature suggests that companies expand by franchising instead of adding company-owned outlets because of the difficulty in raising capital (e.g., Mathewson and Winter 1985), high monitoring costs for outlets in remote locations (e.g., Brickley and Dark 1987; Lafontaine 1992), and the risk sharing between franchisors and franchisees (e.g., Martin 1988). However, this literature does not consider the cost management effects of franchising. Because franchisors can generate sales without owning the resources, managers have greater flexibility in cost management. This flexibility could change the nature of competition between franchisors and other firms, thus influencing managers' choice of organizational form.
Atasoy and Banker (2014) show that cost management decisions confound firm efficiency scores derived from data envelopment analysis (DEA; Charnes, Cooper, and Rhodes 1978). DEA efficiency scores have diverse applications. For example, Demerjian, Lev, and McVay (2012) use these scores to estimate managerial ability, while Chen, Delmas, and Lieberman (2015) and Zhou, Ang, and Poh (2008) review numerous studies related to strategic management and operations management, respectively, that use DEA efficiency scores. Therefore, the confounding effect of cost management decisions on the DEA efficiency scores could have implications for research areas far beyond cost management research in accounting.
Jang, Radhakrishnan, and Yehuda (2016) find that acquirer firms that have lower cost stickiness prior to an acquisition tend to have better post-acquisition performance. This finding suggests that a firm's cost behavior, which reflects managers' short-term cost management decisions, reveals useful information about these managers' long-term investment decisions such as mergers and acquisitions. Rouxelin et al. (2017) show that because firm-level cost stickiness reflects managers' expectations of future demand, aggregate cost stickiness (i.e., average cost stickiness across all firms in a given year) helps predict future macroeconomic outcomes, such as the unemployment rate, providing relevant information for macroeconomic policy. Thus, because cost behavior reflects managerial actions, we can use the observed cost behavior to extract useful information for various other areas that rely on managerial decisions and expectations.
Contemporary cost management research views managers' operating decisions as the fundamental driving force behind observed costs. This conceptual innovation gives researchers a powerful new way of using observed cost behavior as a lens to study a wide range of phenomena that influence, or are influenced by, managerial decisions. We reviewed the rapidly growing literature in English on cost management and its implications in cost and financial accounting, outlined the main themes in this research, and discussed challenges and opportunities for future research. Many other studies have appeared in other languages. As an illustration, we list in the online supplement 152 studies written in Chinese and 81 studies written in Korean.
Managers make operating decisions subject to various constraints, incentives, and biases. All of these factors systematically affect observed costs. For example, recent research shows that observed cost behavior is influenced by demand uncertainty, resource adjustment costs, managerial expectations, corporate governance, strategic positioning, earnings targets, overconfidence, government regulation, and national culture, among other factors. Some of these studies severely limit their own contribution by positioning the analysis narrowly as just another generic study of yet another “determinant of cost stickiness.” A much more fruitful approach is to use the empirical findings on cost behavior as a means to learn about managerial decision making or about interesting institutional or economic factors, gaining insights that are not confined to just “cost stickiness” or “cost behavior.” We recommend that researchers in cost management adopt this broader view.
Because earnings = sales – costs, cost management has implications for many areas of financial accounting. First, cost management decisions directly affect earnings properties such as asymmetric timeliness, persistence, and predictability. Second, because cost management decisions affect the time-series properties of earnings, they have predictive value for future earnings. Third, if managers, analysts, and investors do not fully understand this predictive value, then this incomplete understanding can lead to systematic earnings forecast errors and systematic abnormal returns. Recent research shows that these effects of cost management change inferences on financial accounting topics that rely on understanding or forecasting earnings. Many of these financial accounting implications have not been explored yet. Managers' operating decisions influence additional outcomes beyond costs and earnings. Therefore, insights gained in the context of cost management have implications for broader research areas, including research on fundamental analysis signals, accrual-based earnings management, supply chain management, market structure models, efficiency measurement, firm boundaries, and macroeconomic forecasts, among others. Many of these implications remain to be explored.
We encourage researchers to look inside the “black box” and modify the theory to fit the economic details of their specific research context. We caution that these modifications must be carefully rooted in the economic foundations of managerial decision making, recognizing the systematic interactions of managerial decisions with the context-specific economic factors. Some of these modifications could provide an important theoretical contribution. For example, the literature does not yet have a theory of cost management in non-profit and government organizations. The accounting literature also does not have a theory of cost management for strategic resources that drive future sales, such as R&D and marketing resources. The research literature thus far does not provide a complete understanding of whether observed cost stickiness primarily reflects rational resource management to avoid adjustment costs (i.e., “good” cost stickiness) or wasteful overspending due to an agency problem (i.e., “bad” cost stickiness). This fundamental question affects the interpretation of many findings in the literature and can set the agenda for future research on cost management. In a related vein, understanding how “good” or “bad” cost management affects benchmarking or the forecasting of earnings can provide valuable insights into many important questions in financial accounting research.
Contemporary cost management research has great potential because the insights learned through the empirical lens of cost behavior generalize to inform our understanding of the broad economic and behavioral foundations of managerial decisions. For example, by combining empirical findings on asymmetric cost behavior with the economic theory of cost management decisions, recent research has learned about managers' forward-looking expectations, agency conflicts, incentives to manage earnings, and behavioral biases such as overconfidence. Many exciting—and eminently feasible—research opportunities in this area are yet to be explored.
“Cost behavior” refers to the observed association between activity and costs, while “cost management” emphasizes the underlying decisions. Contemporary research on how costs behave is primarily interested in understanding the managerial decisions, i.e., it focuses on cost management rather than just a mechanistic relation between costs and their drivers. Some sources use narrower definitions of cost management. For example, some strategic cost management textbooks define cost management as “the development and use of cost management information” (Blocher, Stout, Juras, and Cokins 2016, 3) and as “expansion of management accounting to simultaneously focus on reducing costs and strengthening an organization's strategic position” (Eldenburg and Wolcott 2005, 9), while some vendors of project cost management software define cost management as “the process of planning and controlling the budget of a project or business” (source: http://www.costmanagement.eu/blog-article/198-cost-management-explained-in-4-steps). We do not use these narrow definitions.
Banker and Johnston (1993) propose that costs are related to managerial operating strategy, although the focus of the paper is to identify the most relevant cost drivers.
Cost asymmetry (i.e., cost stickiness and/or cost anti-stickiness) is just one of many manifestations of managerial decisions affecting cost behavior.
The economics literature often states that “all costs are variable in the long run” (e.g., Varian 1992, 65). This statement only means that all costs can be adjusted in the long run; it does not imply that they must be proportional to the cost driver (e.g., Noreen and Soderstrom 1994). The standard cost allocation procedure in accounting (e.g., Garrison et al. 2015; Horngren et al. 2015) implicitly assumes that fixed overhead costs are strictly proportional to the cost allocation base in the long run.
This nonlinear cost function varies with the concurrent cost driver levels, but it does not depend on whether the cost drivers increased or decreased from the prior period. This function differs qualitatively from asymmetric cost behavior, in which total costs vary with both the level of and the change in the cost drivers (e.g., Banker and Byzalov 2014).
For example, when the selling price per unit is higher, managers are more likely to keep excess capacity during demand decreases to prevent future stock-outs. This excess capacity can weaken the cost response to concurrent demand decreases. Thus, managers' pricing decisions can directly affect the cost structure. The cost function approach in the economics textbooks is based on the assumption that managers' cost-minimizing input choices conditional on output can be fully separated from other facets of profit maximization (e.g., Varian 1992, Chap. 2).
Cost and managerial accounting textbooks typically discuss the choice between alternative cost structures as one of the major applications of cost-volume-profit (CVP) analysis, viewing the mix of fixed and variable costs as a direct decision variable for managers (e.g., Horngren et al. 2015, 80; Garrison et al. 2015, 205).
This reversal of findings for non-profit versus for-profit firms generalizes to other contexts. For example, Atasoy et al. (2016a) predict and find that greater demand uncertainty is associated with less outsourcing in for-profit manufacturing firms because outsourcing can increase the production lead time and cause excessive production variability due to coordination problems in the supply chain. In contrast, Holzhacker et al. (2015a) and Banker, Lee, and Park (2016g) show that demand uncertainty is associated with more outsourcing in hospitals and state-owned utilities, respectively, which is likely attributable to the non-profit objectives of these entities. The results might also vary with the type of uncertainty. For example, while Atasoy et al. (2016a), Holzhacker et al. (2015a), and Banker et al. (2016g) analyze demand uncertainty, which is a general demand characteristic for all industries, Kallapur and Eldenburg (2005) study uncertainty in the specific context of a reimbursement policy change for hospitals that increased contribution margin uncertainty but did not directly affect demand uncertainty. They find that this form of contribution margin uncertainty is associated with a more flexible cost structure, consistent with the traditional intuition.
Adjustment costs only refer to the one-time costs incurred to implement a change in resource levels. They do not include routine costs incurred every period based on resource levels. For example, a firm with a more skilled work force likely has higher search and training costs for new hires, which are adjustment costs, as well as higher wages and benefits, which are not adjustment costs. Adjustment costs can include both explicit monetary costs (e.g., a fee to a contractor to train new workers) and implicit costs (e.g., the opportunity cost of lost output when experienced workers are assigned to train new hires, or reduced productivity due to low employee morale following layoffs).
Adjustment costs vary with the time horizon of decisions. For example, because the screening and recruitment of skilled workers require considerable time and effort, the adjustment costs of hiring skilled workers on very short notice (e.g., a few days) are likely prohibitive. Therefore, skilled indirect labor can be viewed as a fixed cost in the extreme short run. For monthly, quarterly, or annual decision horizons, the labor adjustment costs likely remain significant but are not prohibitive. Traditional cost allocations and ABC assume that the fixed overhead costs are proportional to the relevant cost driver(s) in the long run, similar to variable costs in the short run. This approximation is appropriate only if the long-run adjustment costs are negligible. However, many resources have substantial adjustment costs even in the long run. For example, managers likely invest substantial time and effort in the hiring decisions for skilled workers, regardless of the time horizon of these decisions. Research in psychology and behavioral economics, for instance, documents biases, such as hyperbolic discounting, in how managers may trade-off the short term against the long term (e.g., Laibson 1997; Frederick, Loewenstein, and O'Donoghue 2002). Such biases may distort the resource decisions and hence the observed costs. Understanding the time horizon effects could be an interesting topic for future research.
The impact of such country-specific studies is greater when they are able to leverage distinctive local institutions, economic factors, or disclosure requirements to shed light on fundamental properties of managerial decision making.
Even cost categories that do not appear sticky in Anderson et al.'s (2003) single-period specification can manifest asymmetric cost behavior in Banker et al.'s (2014a) two-period specification. For example, Banker et al. (2014a) report that the number of employees in their Compustat sample does not exhibit significant cost stickiness on average, which might seem to be consistent with the traditional model of fixed and variable costs. However, in the two-period specification, the number of employees exhibits significant cost stickiness conditional on a prior sales increase and significant cost anti-stickiness conditional on a prior sales decrease. Thus, the single-period model does not detect an asymmetry only because these two conditional asymmetries cancel each other out, on average. The conditional asymmetries reject the traditional model in favor of asymmetric cost behavior.
Additionally, some of the empirical tests could have multiple interpretations. For example, high free cash flow could increase cost stickiness either because it facilitates overspending (i.e., “bad” cost stickiness) or because it gives managers more latitude to focus on long-term value creation (i.e., “good” cost stickiness).
The results for financial covenants could reflect an additional agency problem between creditors and shareholders. Firm value for the shareholders is based on expected profits over the entire demand distribution. The optimal level of cost stickiness for the shareholders increases the expected profits, but it also increases losses for unfavorable demand realizations. Because creditors are exposed to the downside risk but do not fully participate in the upside, they likely prefer a lower level of cost stickiness than the shareholders.
Studies in finance examine the implications of operating leverage for outcomes such as firm risk (e.g., Lev 1974), financial leverage (e.g., Mandelker and Rhee 1984), stock returns (e.g., Sagi and Seasholes 2007; Novy-Marx 2011), and cost of equity (e.g., Chen, Kacperczyk, and Ortiz-Molina 2011). We discuss these implications in the next section. Some of these papers also present descriptive evidence on some of the determinants of operating leverage, such as unionization (Chen et al. 2011), labor mobility (Donangelo 2014), employment protection (Simintzi, Vig, and Volpin 2015), labor adjustment costs (Bhattacharjee, Higson, and Holly 2015), and customer concentration (Irvine, Park, and Yildizhan 2016); however, these studies do not seek to understand the underlying cost management decisions.
Weiss (2010) notes that this measure is defined only for firms that have both a sales increase and a sales decrease during the estimation window, which reduces sample size and possibly introduces selection bias. The computation requires that sales and costs move in consistent directions (i.e., both variables increase or both variables decrease). This selection on the dependent variable can distort inferences (e.g., Banker and Byzalov 2014).
Similarly, the Khan and Watts (2009) C_Score should not be used as a dependent variable in studies of new determinants of conservatism.
A similar caveat applies to other measures that could be used to study the effect of cost stickiness on other outcomes. For example, if a researcher uses industry-specific cost stickiness estimates from the Anderson et al. (2003) model as an explanatory variable, then the required assumption is that industry is not directly associated with the relevant outcome. For the Weiss (2010) firm-year measure, the required assumption is that the determinants of realized sales and costs during the rolling window used to compute this measure are not directly associated with the outcome. Additionally, because the degree of cost stickiness is not a direct decision variable for managers, “the effect of cost stickiness on outcome X” might not have a meaningful interpretation regardless of the estimation approach used.
Propensity score matching cannot be used as a remedy for endogeneity because by design it only addresses functional form misspecification with respect to the included observables (Shipman, Swanquist, and Whited 2017).
The earnings prediction literature typically uses absolute forecast errors (AFE) to assess the out-of-sample accuracy of the prediction models and uses ordinary least squares (OLS) to estimate these models. However, OLS is designed to minimize squared errors rather than absolute errors (e.g., Greene 2012). When model performance is assessed using AFE, a more appropriate estimation approach is median regression, also known as least-absolute-deviation (LAD) regression, because it is designed to minimize absolute errors (e.g., Koenker 2000; Basu and Markov 2004). Alternatively, if a researcher prefers to use OLS, then it is more appropriate to use metrics based on squared errors, such as the root-mean-square-error (RMSE).
A more rigorous way to test analysts' understanding of cost behavior would be to examine whether their ex post forecast errors are systematically associated with various ex ante predictors of sales changes that analysts could observe when they formed these forecasts. For example, suppose that these ex ante variables are not systematically associated with analysts' sales forecast errors but are systematically associated with analysts' earnings forecast errors. These associations would indicate that analysts understand the sales time series but do not fully understand cost behavior conditional on sales. We note that the empirical implementation of such a test must be combined with specific assumptions about the analysts' objective function. For example, if a researcher assumes that analysts aim to minimize squared forecast errors, then the association should be measured using ordinary least squares; and if analysts aim to minimize absolute forecast errors, then the association should be measured using median regression (Basu and Markov 2004).
Many of these results resemble the standard predictions for cost stickiness. However, inventory adjustment involves additional economic considerations that do not arise in the standard theory of sticky costs. Suppose that managers retained slack resources during a sales decrease to avoid resource adjustment costs. Conditional on this resource decision, they can maintain the original production level, using the slack resources to build up inventory, or they can cut production, idling these resources. This production decision involves a trade-off between the incremental costs and the expected future benefits of inventory, in addition to the standard adjustment cost considerations. It is important to recognize that the standard theory of asymmetric cost behavior can—and should—be modified to incorporate relevant economic considerations, such as production decisions in the context of inventory management.
A few papers recognize that there might be two-way relations between cost management and other phenomena. For example, Banker et al. (2016f) point out that compensation contracts can affect cost stickiness through managerial incentives, while cost stickiness can affect the design of managerial incentives. They attempt to address this issue through simultaneous estimation of the cost stickiness equation and the compensation equation using two-stage least squares (2SLS).
Additionally, because the adjustment costs of resources are associated with firms' operating and investment risk, high adjustment costs could increase holdup problems (i.e., ex post rent extraction by the customer after the supplier has invested in relationship-specific assets), leading to underinvestment by the suppliers (e.g., Grossman and Hart 1986).