You are on page 1of 14


4, NOVEMBER 2017

Middle Managers’ Engagement in Autonomous
Strategic Actions: Does It Really Matter How Top
Managers Use Budgets?
Stefan Linder and Simon S. Torp

Abstract—Fostering middle managers’ entrepreneurial behav- developments and outcomes of alternative courses of actions,
ior is a key concern for established firms. Budgets are one of the coordinating activities, and tracking status and performance
most widely used management tools in organizations, and several [18]–in either an “answer machine” manner or in a “dialogue
conceptual articles claim that the way in which they are used,
most notably in an interactive way, affects middle managers’ au- or idea creation” way. Janssen et al. [22] distinguish a con-
tonomous strategic actions (ASA). This type of entrepreneurial ceptual use from an instrumental and a symbolic use and find
behavior in established firms focuses on opportunities outside the that their effects differ. Abernethy and Brownell [15], in turn,
firms’ currently served product-market domains. Using structural find that management control systems can be used in an en-
equation modeling and a formative measurement instrument for abling and a coercive manner, again with differing effects. The
interactive use of budgets, we test these claims advanced in concep-
tual literature within a broad sample of large firms. We find that the arguably most influential typology of different ways of using
way in which budgets are used does not significantly impact middle management control systems, however, is from Simons [18].
managers’ engagement in ASA. In contrast, firms’ boundary sys- Drawing on case research, he submits that firms differ in how
tems and middle managers’ autonomy seem important antecedents strongly they use management control systems “interactively,”
to ASA. respectively “diagnostically.” According to Simons [18], [19],
Index Terms—Autonomous strategic action (ASA), budgets, in- using a management control system interactively means that
teractive use, middle managers. the information generated by the system is “an important and
recurring agenda addressed by the highest levels of manage-
I. INTRODUCTION ment” [18, p. 97], attracts “frequent and regular attention from
operating managers at all levels of the organization” [18, p.
UDGETS, “the quantitative expression of a proposed plan
B of action by management” (see [1, p. 206]), are one of
the most widely used management tools in organizations (e.g.,
97], is “interpreted and discussed in face-to-face meetings of
superiors, subordinates, and peers” [18, p. 97], and is “a cata-
lyst for the continual challenge and debate of underlying data,
[2]–[4]) and thus regularly attract scholarly interest. Much of assumptions, and action plans” [18, p. 97]. Using a system “di-
this significant body of work studied the existence of bud- agnostically,” in turn, corresponds to drawing on its information
gets, the role of participation in setting budget targets, and their so to check that “things are on track”–meaning that managers
effects on individual behavior and firm-level performance (e.g., devote little attention to it and simply use the system like a
[5]–[14]). These efforts have greatly enhanced our understand- thermostat [18]. Simons goes on to submit that using budgets
ing of budgets. Yet, scholars increasingly agree that budgets–and interactively fosters middle managers’ intrapreneurial behav-
management control systems generally–can be used in different ior and, in particular, “autonomous strategic actions” (ASA)
ways, and that the way in which they are used is what should [18], [19].
largely determine their effects (e.g., [15]–[20]). ASA are bottom-up entrepreneurial efforts of middle man-
For instance, as early as in 1980, Burchell et al. [21] suggested agers within established firms, which recognize and explore op-
that managers can use management control systems–that is, portunities outside the firm’s currently served product-market
formal information-based systems designed to influence pat- domains [24]. By focusing on new products for markets so far
terns of behavior within an organization through forecasting unserved by the firm, they are a particular type of intrapreneurial
activity, which may, if successful, entail important changes in
Manuscript received June 8, 2016; revised November 20, 2016 and February
18, 2017; accepted April 2, 2017. Date of publication May 9, 2017; date of the markets served and products offered by a firm. They have,
current version October 18, 2017. This work was supported by the ESSEC therefore, been suggested to be one source of strategic renewal
Business School under Grant 043-263-1-2-05-P-1. Review of this manuscript [25], [26]. Yet, little is known so far on what role budgets play in
was arranged by Department Editor P. E. D. Love. (Corresponding author:
Stefan Linder.) fostering (or hampering) this type of intrapreneurial behavior-
S. Linder is with the Department of Accounting and Management Control, ES- and especially whether interactively used budgets increase ASA
SEC Business School, Cergy Pontoise 95021, France (e-mail: or not.
S.S. Torp is with the Department of Business and Technology, Aarhus
University, Herning 7400, Denmark (e-mail: Simons’ claim has been reiterated in the literature and is con-
Digital Object Identifier 10.1109/TEM.2017.2693295 ceptually appealing, yet it contrasts sharply with the (implicitly)

0018-9391 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See standards/publications/rights/index.html for more information.

we consider the various ways of using budgets plans” as examples of management tools that managers often use alongside the so-called boundary and belief systems. They focus . the current strategy of the business” [18. and development of new organizational II. and firm adaptation (see. therefore. systems has often been linked to enhanced current financial ef- we discuss the knowledge about the antecedents of ASA and ficiency and implementation of cost strategies (see. [26] descriptions of ASA. while Bellora-Bienengräber and Günther [29] ad- tively they use a control system. [50] specify the hypotheses to test. In contrast to surement instrument to capture interactive use of a control sys. Within Simons’ der empirical scrutiny. hamper employee initiative. in if deviations from standards are spotted. Whereas Belief systems relate to an organization’s core values and ASA are related to product innovations and may eventually af- are “used to inspire and direct search for new opportunities” fect performance. budgets serve top Simons’ [18] concept of interactive versus diagnostic use. By drawing the attention of managers on all levels contribute to the sparse number of empirical studies that test of an organization to these strategic uncertainties interactively all the “levers of control” in Simons’ [18] framework within a used management tools “stimulate search and learning. [16] and [20]). p. Simons [18. and the degree of diagnostic dress product development and organizational performance. 40].LINDER AND TORP: MIDDLE MANAGERS’ ENGAGEMENT IN AUTONOMOUS STRATEGIC ACTIONS 451 held view in much of the literature that management control [18. p. its management as “levers for implementing intended strategies” relation to his “levers of control” framework. [46]. and hence. In fact. They are used in a sim- contrast to extant research about the control systems-ASA rela. specifically budgets. on opportunity-seeking behavior” [18. a diagnostic use of control the impacts of different ways of using budgets. Moreover. [4]. 7]. this manner of using control systems. 91]. We. allowing single study. We focus on annual budgets and middle levers of control framework. systems.g. e. e. [19]. Yet. However. Control Framework While the available case and anecdotal evidence is illustra- Whether intended or not. p. and strategic operating and financial behaviors. studies on interactive use of con.g. the degree of how interac. p. e. the resources and energies of the firm” [18. performance.g.g.. 3]. which actual temperature deviates from the preset standard [18]. in. we draw on a multi-item operationalization of ASA. [26]. e.. The third lever is the degree to which budgets interactively or diagnostically?” top managers use a formal information-based management tool Our study contributes to the literature in several ways. creativity. yet-to-be realized innovations (see. they are a distinct phenomenon. belief systems and boundary sys- managers’ engagement in ASA because middle managers have tems complement each other since the “use of imprecise belief been identified as pivotal for such actions [24]. learning. ilar manner as a thermostat. and [51]). processes. standards. e.g. 109] highlights “annual profit plans or bud- thereby avoiding confounding ASA with other intrapreneurial gets. our results underscore the importance of new strategies to emerge as participants throughout the organiza- considering boundary systems for ASA. We thus interactively. this study is the first to use a formative mea. In these cases. formal information-based construct (see. Bisbe and Otley [17] concentrate on the ef- at their disposal for shaping these contributions: the firm’s be. but also to lower innovation. [31]). development of new capabilities. 7]. management systems serve primarily as means for confirming trol systems have relied on reflective measurement instruments that everything is going according to preset goals.. in systems inspires unfocused search efforts that risk dissipating the following. evidence testing the proposed link between interactively used tions to an organization [48]. and use. Not surprisingly. and literature on [18. p. fects of an interactive use of control systems on innovation and lief systems. Boundary result when budgets are used diagnostically. [16]).. Subsequently. [52]).g. formal. he means “the managers use annual budgets–a relation that seldom attracts uncertainties and contingencies that could threaten or invalidate attention. In his “levers of control” frame. Simons equally notes that annual budgets and profit plans are frequently used as di- In order to answer our research question. 39] and serve “to set limits activities of the kind of ASA. Such shared values pull individuals’ actions to the systems. little empirical are used shapes the nature and extent of individual contribu. Different Ways of Using Budgets Within Simons’ Levers of (see. of a “lever” tion respond to perceived opportunities and threats” [18. Similarly. its boundary systems. Simons [18] posits that top managers have four levers ples is available. that so far has attracted little systematic research. such an effect should only be the ing to the strategic choices (see. p. which triggers adaptive action if the tion. First. influence managers’ perception of reality. put these claims un. they trapreneurship. interactively used tools should enhance innova- tion. to draw attention to the strategic uncertainties facing the or- we provide rare evidence on the relation of ASA with how ganization [17]–[19]. By strategic uncertainties.g. thus potentially suf.. [9]). Third. Hence. information-based systems. and longer term performance A. e. p. we answer the question “does it matter for mid. 94]..g. creativity. “delineate the acceptable domain of activity for tive use of budgets should heighten the level of intrapreneurial organizational participants” [18. e. whereas an interac. Whereas literature agrees that interactive use is a formative tive use. Consequently. a danger that dle managers’ engagement in ASA whether top managers use boundary systems curb. we first discuss agnostic control systems [18]. [52] and [53]). and ASA in larger sam- work. adaptation. e.g. second year forecasts. in turn. p. [16]. DEGREE OF USING BUDGETS INTERACTIVELY AND ASA capabilities (see. (see. and procedures and for triggering corrective actions fering from misspecification bias (see. thus giving a mean- following Simons [18]. which he terms “interac- tem.. the way in which management tools tive of the theoretical points made by Simons.” when used diagnostically. [64]). Bedford [46] concentrates on the performance effects. Second. and [27]–[30]). is in line with Burgelman’s [24]. [20].. e.

e. and this study suffers. very little “induced strategic action” than to measure ASA. Allocation of decision rights. several studies highlight the perceived interests of the strategic actors at the operational top managers’ support for entrepreneurial behavior–including a and middle levels in the organization” (see [24. Thus. NOVEMBER 2017 on middle managers’ recognition and exploration of business notably.g. budgeting.g. in. 64. Therefore. only to verify teractive use of budgets on ASA.g. The study of this evidence speaks to the diagnostic vs. from multiple [24]. the degree of formalization or innovation more broadly. yet not for ASA and interactive use of budgets were measured puts tight fostering ASA. and [42]). 1350]). VOL.” can serve top management as et al. [41]. ASA are a of years now and have been reiterated in the literature. if ex post signed-off and implemented. In fact. [46]. and create politi- its strategic context. In other words. the small sample size and the way lever to implement intended strategies [18]. project screening criteria. p. all tive systems on ASA. Thus. [35] and Burgelman and Grove [26] describe ASA. in turn. tool more generally–in a “diagnostic” way. [30] is informative. In contrast.” This evidence can be Simons’ claims based on the study seems highly problematic. cost accounting. ASA ture and allocation of decision rights. [20]. The first necessarily apply to ASA. So far only one other study besides Simons’ work has tried to ent ways of using budgets for ASA. i.. research) to foster ASA provided that they are used by top Yet. cal leeway to trigger top managers to retroactively rationalize an The structural context “refers to the various administrative autonomous initiative to become part of the (renewed) official mechanisms which top management can manipulate to influence strategy (see. [43]. e. Evidence is scarcer with respect to an important measurement-related issue: While literature agrees the impact of planning. the situation is even more problematic within the litera. By strategic context. organization’s structural context. found to matter for the level of ASA [45]. For these reasons.. hence. Obviously. there is some anecdotal Moreover. based on the evidence. more likely to capture what Burgelman [24] calls empirical literature on change and innovation. there is a broad and rich body of are. Only one he submits that using budgets–or any other management control small study so far has tried to shed light on the impact of in. e. [35]. and Annual budgets. yet which may. e. thus potentially suffering from a misspecification them have been suggested by Simons [18] (drawing on case bias (see. formalization firm may not translate to ASA.. [24] and [35]). seems to inhibit ASA dence on the impact of different manners of using budgets on engagement [32] as well as bottom-up driven innovation and innovation or related constructs is very limited and suffers from change more generally [36]. and incen- that interactive use is a formative construct (see.. of positions and relationships. based on case evidence. interactive use of thus very likely picked up other behaviors. However. especially middle managers’ decision-making autonomy. which stands in contrast the impact of management practices that focus on controlling to the descriptions given in [18] and which may have led subordinates or imposing senior managers’ will (see. While this study by Hofmann whether “things are on track. [37]. Thus insights for the broad to empirically test whether it truly matters how top managers phenomena of innovation or change more generally may not use budgets for the level of middle managers’ ASA. For example. The available evidence consistently points sights about the role of budgets or their different ways of use to autonomy being an important explanatory variable for the for product innovation within the markets already served by a level of ASA. truly affect the level of ASA and. [19] claims have been around for a number a diagnostic use for ASA. clear signaling of support for bottom-up initiatives for and ex- . it is time particular type of intrapreneurship.g. They are thus likely to lead to what is some. Firms’ incentive systems have not been studies on interactive use of control systems have relied on re. They shortcomings. Budgets are thus one element of an which they do not ex ante have the sign-off from top manage. e.g. it step for doing so is to review the knowledge on the antecedents is not possible to assess whether different forms of using budgets to ASA to clarify the relation. in fact. the measures shed light on the impact of interactively used budgets on ASA used to capture ASA are quite remote from how Burgelman [30]. [41]. ASA relates to the of positions and relationships. 4. is the entail important changes in the markets served and products element within the structural context that so far has received offered by a firm. most of the attention in empirical research on ASA (e.e. NO.. the study treats interactive use as a reflective evidence in the literature on innovation and change regarding rather than a formative construct. to judge the validity of Simons’ [18] claims. limits on the conclusions that can be drawn for the role of differ. e. [36])– to a misspecification bias in findings (see. This even more so as the evi. and hence. [16].” In contrast. as already described.. Yet. as planning and performance measurement systems (see. Antecedents to ASA those factors affecting how managers perceive the strategic situ- Extant research highlights two groups of factors as especially ation of their firm. budgets-ASA relations. Thus. as outlined. for which man. agers in the focal firm face substantial knowledge gaps and for [24]. times called “radical innovation” [63].g. in turn. [4]. In particular. whether and how managers can question the important for the level of ASA: an organization’s structural and strategic direction. a use of management practices which resembles drawing any conclusions about the empirical veracity of what Simons’ termed “diagnostic use. and notably different ways of using [27]–[30]). managers in a way that he terms “interactive use. extant literature on ASA refers to all B. [17]). flective measurement instruments (see. voice ideas for alternatives. a diagnostic use may even hamper ASA..g. resource availability and are not the same as top-down driven corporate entrepreneurship allocation to lower-level members. the structural context comprises organizational struc- opportunities in new product-market domains–and hence. ture focusing on the link between budgets and ASA. viewed to support Simons’ view about a detrimental effect of As Simons’ [18].. ment. and [44]).g. e. [31]).452 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT. as well front-end of a group of product innovations. but not ASA. [31]).

survey evidence failed to trol systems at General Motors as a historical example of such a find such a relation [37]. such as precise explanations of outcomes expected from organizational work as another important antecedent to middle managers’ entrepreneurial behavior (see. ] about performance gaps and rienced by their firms. [32]. inter- middle managers’ engagement in ASA. In the following. such an exchange implied by an interactively used valid.LINDER AND TORP: MIDDLE MANAGERS’ ENGAGEMENT IN AUTONOMOUS STRATEGIC ACTIONS 453 perimentation with new products or service offerings. organization” [18. 39].. e. [40] suggests that declining performance of an organiza. and peers. An organization’s performance has been sug- gested as another element of the strategic context. Hypotheses ily and thus develop a political base for the proposal” [47. change. p. we consider these two active use resembles collaborative problem-solving processes. Theoretical model and hypotheses. They influence managers’ perception of real- ity. Mar- Overall. So far. p. and [38]) and bottom-up driven innovation and change (see. As Aiken (see. Hence. internal verbal communication or collaborative problem-solving tally different effects on ASA than an interactive use.. 1 summarizes the hypothesized conversations between middle and top managers that have been relations. Thus. e. Yet. it seems likely that interactive use should . [2]–[4]) and since scholars increasingly agree that the way according to Simons [18. using formal information-based control sys- ditional factors taken from work by Simons on his “levers of tems in an interactive way may provide for the type of strategic control” framework [18].g. factors alongside the different uses of budgets in the model which have been found to foster innovation in organizations that we develop. [16]–[20]). 1. we enrich this model by two ad. Since research on ASA has already identified top man. he notes that “interactive control systems guide vation and change. belief systems and boundary systems as described by Simons [18] seem to refer to two additional aspects of an organization’s strategic context. drawing on Simons Similarly.. Marginson’s [48] call to explore promotes a cross-fertilization of ideas about possible solutions. 107]. however. and innovation [49]. . thus giving a meaning to the strategic choices (see. [35]. [63]. which fosters lower.” the role of different uses of control systems for ASA remains Moreover. face-to-face meetings implies continual challenging and debate agement support and autonomy as important for understanding of underlying data. [32] and [54]). surprisingly little is known about the role of different ginson [48] points in a similar direction by stressing that the ways of using budgets for ASA. p.. communication “facilitates the rapid diffusion of information ers interested in enhancing the level of intrapreneurship expe. [37]. new mar- kets or market segments and their exhibited openness to discuss new business ideas–as a core element of the strategic context. Simons [18] relates his concept of the interactive use of con. Fig. p. .g. He cites Alfred Sloan’s use of con- tion may be an antecedent to ASA. neither Simons’ (1995) boundary systems Fig. our present knowledge is unsatisfactory as et al. Moreover. we present the relationships between differ. effect seems plausible. which have been studied in the literatures on inno- for example. extensive internal verbal it hampers the formulation of recommendations for practition. while the experimentation and learning that are necessary for new au- some case evidence as well as organizational learning theory tonomous strategic initiatives to emerge and be tested in the [39]. subordinates. and change more generally as it to “locate supporters more eas- C. [49]. throughout the organization [ . ent ways of using annual budgets and ASA. [52] and [53]). since an interactive use in addition to the regular [18].g. [47. nor the “organizational boundaries” highlighted in the strategic entrepreneurship literature have been studied empirically in their relation to ASA. As annual budgets are one of way control systems are used should influence ASA. 97].. e. processes. [24]. e. While the phenomenon of interactive use as defined by Simons trols to Burgelman’s work on ASA and submits that diagnostic [18] is thus more comprehensive than the constructs of extensive use of information-based management systems has fundamen.g. renewal induced by an interactive use of control systems.g. since a key feature of an interactive use. 637].and middle-level members’ ASA (see. In fact. p. control system may promote middle managers’ engagement in developing and promoting new ideas. new products or services..g. assumptions. found important for adaptation. e. 261]. and “delineate the acceptable domain of activity for organizational participants” [18. and action plans [18]. Such an the most widely used management tools in organizations (see. p. e. In fact. Simons’ idea of boundary systems resonates well with the emphasis put in strate- gic entrepreneurship literature on “organizational boundaries” [55. [36]). Besides top management support. 637] have already noted. is the intensive frequent discus- in which management tools are used is decisive for their effects sion and interpretation of the respective information in face-to- on individual behaviors as well as organization-level outcomes face meetings of superiors.

[49]. [46].p. autonomy has been proposed as an important A diagnostic use of budgets draws managerial attention to motivator that should enhance performance in creative and spotting deviations from the plan and to developing corrective innovative activities (see.. the lower c. This should facilitate the expect decision-making autonomy to have a positive effect on realization of an intended strategy as it leads to a focus on op. Middle Not surprisingly. 4. [38]). by fostering open discussion knowledge sharing. [36]. NO. a firm’s financial and nonfinancial (e. .p. Such behavior by top managers of the firm” [18. edge and facts are decisive rather than hierarchical positions tion of actions among individuals and facilitate cooperation and (see. [35] and [36]). the delegation of decision-making rights to middle managers is often portrayed as an antecedent to innovation. e. [18] claims.g.. adaptation. [44]. and change literature posits that tous information-based managerial tools. for instance. Hence. it is unlikely that ASA. management. [32]). on middle and lower-level managers’ ASA. the middle managers’ engagement in ASA. H5: The larger middle managers’ autonomy in market and product- Without such questioning of the organization’s intended strat. Therefore. Top man- tainties. With. a core roadblock to innovation and change. however.g. level of middle managers’ engagement in ASA. level of middle managers’ engagement in ASA. we actions to close potential gaps [18].g.p.e. e. the higher c. serious and interested search efforts that risk dissipating the resources and energies manner [24].g. [42]. Our third and fourth hypothesis thus expect that the signal of support by top management “is a read as follows. [36]. strong antecedent of autonomous strategic behavior on the part of middle-level managers’ behavior as well as others in the H3: The stronger the belief systems. new tems play an important role in influencing managers’ behaviors markets or market segments. VOL. our first hypothesis thus is as follows. our next hypothesis reads as follows. [59]. 40]. facilitating experimentation and risk-taking through their own our second hypothesis reads as follows. it implies evidence (see. [35]. e. and [60]). p. Since budgets are one of the most ubiqui. H1: The greater the interactive use of budgets.g. Therefore. e. belief systems can be considered to create the desire for that they have confidence in the abilities of the members of their middle... [24]. e. but also Thus. In other words.and lower-level managers engage in ASA. and experimentation with new products or service offerings. [24]. the higher c. [32]. [30]. while boundary systems motivate them to do so as they challenging jobs and projects [38]. thus inhibiting ASA [24]. the and entrepreneurship in established firms (see. 703]...and lower-level managers can scan the environment suf- ficiently broadly and intensively for new business opportunities 3) Top Management Support: Top management support for related to new product or service offerings for markets that are experimentation and entrepreneurship is another commonly not yet served by the firm. the insufficient support for experimentation and entrepreneurship as level of middle managers’ engagement in ASA. [32]. 64.454 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT. economic viability and thus to lead to strategic renewal.g. Yet. in line with literature. and [37]). the extent to which bureaucratic decision-making processes can impede new they are used interactively is likely to have pronounced effects organizational ideas and initiatives (see. [32]. Thus. It implies that knowl- provide guidance to them. [49] stage due to a lack of resources likely will not engage in such and [62]) and ASA [37]. p. e. it 2) Middle Managers’ Autonomy: Much of the strategic should foster ASA. Management support 1) Boundary and Belief Systems: According to Simon’s denotes a clear signaling of support for bottom-up initiatives for levers of control framework.. H4: The stronger the boundary systems. the level of middle firm” [32.g. the level of H6: The stronger the top management support. ASA–an expectation also supported by (albeit quite limited) erational challenges [18]. that many (if not most) of such initiatives will out boundaries. i. perimentation. rather than to the strategic uncer. belief systems and boundary sys. managers’ engagement in ASA. [24]. by diverting attention suggested antecedent of ASA and related concepts such as in- to attaining existing plans. novation (see.p. the higher c. Thus. thereby ultimately rendering discovery and of new ideas and by acknowledging the inherent risks with ex- development of new business opportunities more likely. the higher c. [57].p.g. related decisions. e.g. [56] and [57]). we actions in the first place. middle. and a readiness for and active en- since the “use of imprecise belief systems inspires unfocused gagement in discussing ideas in an honest. e. behavior at both the individual and team levels (see. questioning the plans and the firm’s current business model.. [32]. Moreover. top management support for experimenta- managers fearing that their engagement in ASA may never bear tion and entrepreneurship has been shown in studies to enhance fruit in the sense of their initiative reaching a “proof-of-concept” individuals’ creativity and innovation behavior (see. top and attention) resources may be spread too broadly to allow management support reduces the (career) risks of engaging in any individual initiative to grow sufficiently so as to prove its ASA (see. diagnostically used budgets reduce the likelihood that agement can be instrumental in encouraging such behaviors by middle. e. Moreover. the higher c..and lower-level members of an organization to engage organization to perform effectively and to succeed in working on in ASA...p. NOVEMBER 2017 have similar effects–and according to Simons’ [18] claims. They thus heighten the coordina. [33]. innovation. time not turn out to be very successful economically [24]. Kanter [36] identified H2: The greater the diagnostic use of budgets. the level of their engagement in egy and the firm’s business model. This leads to our fifth that attention is diverted from strategic uncertainties and from hypothesis. at the same time. not only shows that they trust middle managers [49]. and [55]). In line with Simons’ [47]. [61]. and [58]).g. a danger that boundary systems curb. Hence.

These two waves of mailing produced 141 latent constructs meet the convergent criteria with each loading responses. Measurement Instruments sonalized letters and each was asked to fill in a two-page survey instrument.. [67]. including the primary industry affiliation. Harman’s one-factor test best suited for validly testing these claims. number of employees. Two weeks after the initial letter. responses or a response rate of 62. comparing the responding companies with the non- There are several ways of testing the claims advanced in con. The 500 companies cover a broad set of industries Core independent variables of interest to us are not directly and have at least 275 full-time employees. following the recommendations by Craighead et atively large response rates. error. we conducted a confirmative factor analysis compar- nonresponse bias. thereby reducing the dangers of al. The survey was comple. We drew on a set of several items our survey instrument. (Δχ2 = 2015. The questionnaire observable. and analytical procedure ings. correspond- ing to all items loading onto their conceptualized factors. The results [68]. we sional formative construct [64]. the extent to which budgets are used in an interactive between the constructs. Sample and Data Collection Procedure clearly indicated that the multifactor model better fits the data Whenever possible. e. among tion existed for the interactive use of budgets when we collected others. a survey We conducted three statistical tests to check for the pres- collecting data from firms in a broad range of industries seems ence of common method bias. All three tests thereby Subsequent to these pretests. and our data.4%. tors. we conducted struments.. [4]. these. we describe the research design. but are latent constructs. a new measurement instrument had to be developed. sample. a test tacted by phone and asked to participate in the survey. we relied on multiple observed items for each of these tonomy. the constructs display high levels of reliability. Since no such operationaliza- conducted tests on information from KOB regarding. Therefore.6%). thus confirming the findings of the two previous tests that These pretests raised no concerns. au. [16]. size. yielding a total of 312 The test indicated no reason for concern.g. To reduce measurement included items on budgets. tions of interactive way of use rather point toward a multidimen- To assess the potential presence of a nonresponse bias. As an open economy [68] on all items underlying the latent factors suggests that with a variety of industries. and ASA in addition to several constructs. Nevertheless. all items loaded on their conceptualized fac- quite willing to participate in surveys. Procedural remedies com- prised the use of a cover story that avoided a link to ASA. subsequently. Moreover. Denmark seems a suitable site for no single factor accounted for most of the variance explained such a survey as Danish firms should be representative of firms (variance explained of individual factors ranged from 5. The Appendix gives the exact wording. firm age. or KOB). METHOD legal form. being significantly related to its underlying factor. data for testing the hypotheses indicate that common method variance is unlikely to bias our were collected as part of a larger data collection effort. 171 answered the questionnaire. as indicated by compar. relying on the existing scales whenever possible. [27]. management support. ASA. p < 0. Research Design and Population of the confidentiality of their responses. dangers of common method bias [70]. As reported control variables. 1) Interactive Use of Budgets: Several measures for interac- mented by additional data collected from a national database of tive use of control systems have been used in the literature in Danish firms (Koebmandsstandens oplysningsbureau. data To address the risk of common method bias tainting our find- collection.59 to 0. heads of marketing of the 500 largest companies in Denmark (in terms of the number of employees) were contacted with per- C. . Third.81 [66]. as indicated by composite reliabilities above 0. no common method variance is present. Second. was pretested in face-to-face meetings with three managers. assuring respondents A. whereas Simons’ [18] descrip- core financial data. capital structure. we engaged in extensive pretesting of the marker variable test [69]. and the firm’s founding year. ing a one-factor model with a multifactor model. [68]–[70]). ASA. The instrument used for collecting data that should not be related to ASA based on theoretical grounds. the literature (see. Danish firms in general are 36. The to the first letter. the remaining respondents were con. Our hypothesized relationships seem way–affect the engagement of middle managers in ASA.001). a second letter with the ques. the questionnaire was pretested on a sample of None of these variables was statistically significantly linked to 92 managers from 62 firms (not included in the main dataset). boundary and belief systems.LINDER AND TORP: MIDDLE MANAGERS’ ENGAGEMENT IN AUTONOMOUS STRATEGIC ACTIONS 455 III. Since unlikely to be part of respondents’ theory-in-use. The lat- conceptual literature on how different ways of using budgets– ter reduces the likelihood of respondents making associations especially. sector. Finally.g. df = 10. which reduces literature does not limit these claims to a particular industry. measurement instruments. turnover. [69]. e. First. Chi- square difference testing (Δχ2 ) comparing these two models B. among others the firms’ formal risk management procedures. Most of them relied on reflective measures (see. Moreover. Out of of normality was conducted as recommended by Hult et al. responding ones. and [28]). None of the tests suggested that the nonre- ceptual literature about the role of interactively used budgets for sponding firms differ from the responding ones.3% to in many other countries.83 and average tionnaire was sent to the respondents who had not responded variance extracted (AVE) ranging from 0. the recent past. but maintains them to apply to all kinds of industries. we relied on existing measurement in. In a third step.63. in the Appendix. and including the scales We are interested in empirically testing claims voiced in in a larger set of scales dealing with quite diverse topics. we used procedural and statistical means recommended in used in the following subsections.

It thus seems ideally 4) Belief Systems: Similar to boundary systems.94. 7 = fully agree).g. df = 48. on three items per dimension. Each approach has its Responses to these four items were collected using seven-point strengths. e. which are used to form the formative cision autonomy and way of budget use are all perceptual by higher-order construct of interactive use of budgets. p < 0. CFI = 0. p. To control for size.80. Likert scales (1 = fully disagree.08) fits the data substantially better than natural logarithm to the total amount of assets. mensionality of the measure.. To control for these various effects in structural equation ferred to budgeting and budgets and was measured on 7-point modeling (SEM). organization.88. Responses were collected on a seven-point Likert scale than the hypothesized model. considering measurement error [71].10). Firm size and age have been linked to bureaucracy and factor model (χ2 = 204. to provide accurate responses. the four. The Appendix gives the exact wording of the four items. global comparative firms’ entrepreneurial and innovative activities (see. fit is an advantage of covariance-based approaches.001). then this pro- (1 = fully disagree. The Appendix provides de. Such indi- 5) Top Management Support: Assessment of top manage. NOVEMBER 2017 Drawing on Simons [18]. The four items were measured using seven-point Likert scales (1 = fully disagree. and openness to ideas and exchange with their subordinates. 8) Control Variables: Several control variables serve to bet- tor analysis corroborates this finding: As indicated by the ter distill the effects of different ways of using budgets for Chi-square difference test (df = 6. we followed the recom. The items focus on top managers’ coefficients and their respective statistical significances. df = 54. vides additional comfort to rejecting the hypotheses related to (The Appendix provides details on the items and their measure. A confirmative fac. nature. Analytical Procedure 7 = fully agree).’ agnostic use of performance measurement systems from Henri SIC-codes).93. Finally. Cronbach’s alpha is 0.. we drew on a subset of three a broader set of such indicators is available is less likely to . The Appendix reports the details. root-mean-square error of approxima. their four conceptualized dimensions. 1) by testing the full hypothesized model and assessing path sion of a scale by Choi [57].’s equivalent to the U. Since all known indica- 6) Middle Managers’ Autonomy: To measure middle man. the 12 items loaded onto of 0. studies (see. 7 = fully agree). [30] and [46]). late to middle managers’ market and product-related autonomy mons [18. D. If this is the case. cators permit testing hypothesized relations in two ways [74]: ment’s support for entrepreneurship relied on an adjusted ver. suggesting tails on the scale (original measures in Danish. [16]. 3) Boundary Systems: Boundary systems are captured by a set of four items developed by Widener [27]. ASA. 64. e. pendix). See the Appendix SEM facilitates simultaneous testing and allows explicitly for details. as indi- of budgets. 4. we regressed each item of the ASA scale on Likert scales (1 = fully disagree.09. The average tenure was 13. [37]). Factor analysis confirmed the unidi- Steenkamp and van Trijp [75] and Gerbing and Hamilton [76]. dependence is a prerequisite for a valid measurement of a we follow Larsen [63] and collect respondents’ tenure with their higher-order formative construct. NO.g. The actual alpha is 0. an analytical procedure for which agers’ market-related autonomy. to control for potential 2) Diagnostic Use of Budgets: For measuring diagnostic use sector effects. which exhibits a Cronbach’s alpha In an exploratory factor analysis. The four-item reflective scale lends itself to being trans. we adapted an established and validated scale for di.g. Firm age draws on the founding year of a firm RMSEA = 0. Such in.91. Given that levels of ASA. The items used in the present study re- istics of an interactive use of a control system as defined by Si. Two alternative procedures of the firm’s use of belief systems draws on an existing scale are common: a covariance-based and a variance-based [par- [27]. which were saved in new variables. Cronbach’s Availability of a broad range of indicators for assessing model alpha of the scale in our sample is 0. After extensive pretesting. we thus formulated items to measure items out of a broader reflective scale initially developed and interactive use of annual budgets as a four dimensional formative validated by Andersen [43]. we relied on firms’ primary industries. Cronbach’s the various controls prior to running the SEM models. tors have their limitations. Cronbach’s alpha is 0. public database.) through systematic model trimming. e.87. tial least squares (PLS)] approach. To assess the measure how often middle managers engage in ASA (see the Ap- measurement of interactive budget use. [38]). All items were measured on seven-point Likert scales mendations for assessing higher-order formative constructs by (1 = never. translated by the that they were sufficiently acquainted with their organizations authors).. de- lower-level constructs. SEM models then drew on the residuals of these regressions. fit index CFI = 0. 7 = always). measure as a set of four lower-level constructs with each measur. which has since been used in several construct.U. Therefore. we follow common practice and apply a tion RMSEA = 0.86. measurement of the in the past three years (2010–2013). VOL. as listed in the a one-factor model (χ2 = 335.456 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT. assessment suited for testing our hypotheses.85. cated by their NACE-codes (the E. those constructs removed when creating these nested models ment properties.89. We relied on seven-point construct of interactive use of budgets in the main sample drew Likert scales ranging from 1 (fully disagree) to 7 (fully agree). We implemented the formative Cronbach’s alpha is 0. 7) Autonomous Strategic Action: Seven items from [37] ing one of the four dimensions outlined by Simons. similar to the results in other studies (see. leav. 97]. This supports the independence of the four reported in the same database. 2) by testing the fit of alternative nested models within the ing aside other aspects of (a more generally defined) leadership full hypothesized model for whether they offer better model fit climate. The four dimensions correspond to the four character. experience with the pivotal firm is important.S.07 years. 7 = fully agree).

which exhibits. hence. the fit of the models was tested using the RM- about the underlying distributions of the variables [72]. which is the most comprehensive model of these nested a set of nested models as a second test of the hypothesized rela. The structural model provides path coefficients to rejecting certain hypotheses. and autonomy as independent vari- ternative nested models [71]. we first estimated the measurement model before analyz. and RMSEA = 0. models.” making model with a good fit. Since several variables in Model 1 do not show statis- beliefs systems. This lends further comfort to our preliminary conclusion that some of the hypothesized relations cannot be supported empir- IV. as recommended by Gerbing and we rely on covariance-based structural equation modelling us. the Tucker-Lewis index (TLI). In comparison to As recommended by Hult et al.001).81. management support. Like in the original Model 1.79 and the goodness-of-fit index (GFI) and the RMSEA. for example. relations in SEM. The CFI [77] takes into con- based SEM. PLS additionally affords the advantage of being a sideration sample size. if in contrast. reliance on multiple fit indices use of budgets (IB) nor firms’ belief systems are significantly rather than on the chi-square test alone is recommended. having a broad set of model.08 indicate a Our main objective is to test established theorizing from Si. the two variables do not exhibit a statistically significant indication that hypothesis 2 may not receive empirical support pairwise correlation.90 or higher indicate a procedure that allows a kind of “explorative testing. [67] and Gerbing and covariance-based SEM. and we draw on a comparatively large sample for a gest that the model fits the data well (NFI = 0. The first stage involves estimat. neither interactive known and criticized [71]. we compare model.g. boundary systems and belief systems. PLS makes less restrictive assumptions Anderson [74]. RESULTS ically in our sample. Therefore.86. More- common thresholds for statistical significance. CFI = 0. neither the possibility for exploration CFI = 0.LINDER AND TORP: MIDDLE MANAGERS’ ENGAGEMENT IN AUTONOMOUS STRATEGIC ACTIONS 457 lead to an erroneous rejection of hypotheses. approximation in the population. As shown in Table II. and their significances and. the fit indicators for this Model tionships by drawing on model fit indicators like.79. In contrast. In contrast to covariance. pothesized model and comparing the model fits of these nested ing the measurement model using confirmatory factor analysis models among each other and with the hypothesized model. different model fit indicators seems important to us as it offers a Based on the hypothesized relations. TLI = 0. Yet. this suggests sized relationships of diagnostic use of budgets. to create a first nested model (see Table I). better fit to the data ing the structural model which we use to test the hypothesized than Model 1 (CFI = 0. our structural model particularly robust way of testing nested models.88. If the model with the best fit ables. 2. Following a stepwise approach. and the measures used thus nor the suitability of the analytical approach for smaller samples are suited to testing the hypothesized relations in a structural is of primary importance to us. allows testing our hypotheses. It allows basing comprises ASA as a dependent variable and interactive use of the testing of hypotheses on two assessments: the significance budgets. interactive use accepting all hypothesized relationships. p < 0. The path coefficient of diagnostic use of budgets with ASA In particular. Consequently. The chi-square test of the measurement model df = 137. ing hypotheses 4 and 5. 3. Yet. This suggests that model fit might further the proportional improvement of the model fit by comparing the improve when dropping them (which would indicate that their hypothesized model with a restricted baseline model. hypotheses should be rejected). and belief systems with ASA nested submodels exhibits the best model fit. It to determine convergent and discriminant validity.05). by the Chi-square difference test (Δχ2 ). RMSEA = 0. diagnostic use. ter than the hypothesized model–thus giving additional comfort surement model. as also indicated [73]. before testing such further . It is thus SEA and the CFI in addition to the normed fit index (NFI) and also well suited to smaller samples. Thus. any of its of budgets. and linked to ASA (at common threshold levels) in this more par- we proceeded to inspect several comparative GFIs that measure simonious Model 2. good fit. Anderson [74] and Kline [71]. middle managers’ decision-making autonomy tically significant relations with ASA. Δχ2 = 666. of path coefficients in the hypothesized model and the fit of al. we test As shown in Table I. diagnostic use does not correlate with ASA at is furthest away from attaining statistical significance. this indicates that fail to attain statistical significance at common threshold-levels. several alternative nested and top management support for experimentation show greater models of Model 1 are conceivable. The second likely results in a more parsimonious model that fits the data bet- stage compares the theoretical structural model with the mea. This gives additional comfort to rejecting was significant. survey study. its sensitivity to sample size is well hypothesis 2. among the core constructs that are whether the model fit improves when removing relations from of interest for testing the hypothesized relations boundary and Model 1. as it considers the error of known. This yields the Following the two-step procedure of Anderson and Gerbing more parsimonious Model 2. and values of 0. top management support.. values below 0. the hypotheses corresponding to the relationships excluded from This suggests rejecting hypotheses 1. The RMSEA is sensitive to the number it well suited to researching phenomena about which little is of estimated parameters in the model. 1 suggest that it badly fits our data (e. Calculating this model in AMOS shows that the hypothe- corresponds to the hypothesized structural model. This gives a first over.06). This suggests removing diagnostic use so in our sample. while accept- the respective nested model should be rejected.85. however. RMSEA = 0. we conduct a second test of the ing AMOS 22 in a two-stage procedure recommended by hypotheses by calculating a set of models nested within our hy- Anderson and Gerbing [73]. Starting point of these nested models is the hypothesized As recommended by Gerbing and Anderson [73]. and 6. The fit characteristics of the measurement model sug- mons [18].06. correlations with ASA than the way in which budgets are used. [74].

33 ∗∗∗ 0.89 0.91 0. Like Model 2.88 0.051 5 Model 5 (BC.89 3 Top Management Support (SUP) 0.001). [37]).81 0.79 ∗∗∗ 518 0. AUTO. which a firm experiences.24 ∗∗ 0. AUTO = Middle Managers’ Autonomy. trimmed models. NO.72 ∗∗∗ 0. SUP = Top Management Support.38 ∗∗∗ 0.08 ∗∗∗ 294 0.001).55 ∗∗∗ 0. ASA. hence. Table II summarizes these fits for the nested models. Consequently.050 4 Model 4 (BC.85 4 Boundary Systems (BC) 0. Instead.21 ∗ 0. This is corroborated by the χ2 -difference the extent of middle managers’ autonomy in market-related test showing a significant improvement of model fit when decisions are significant factors for explaining the level of ASA moving from model 3 to model 4 (Δχ2 = 621.059 3 Model 3 (IB.44. 64.97 −389.26 ∗∗∗ 0.064 2 Model 2 (IB. bound- calculated when removing belief systems from Model 2b.05 (two-tailed test).049 1 Model 1 (DB.05).001). In contrast. Our study does not find a statistically significant impact of but excludes the two budgetary uses along with the belief how interactively budgets are used for the level of middle man- systems.85 0. 1442.01 −142. df = 77. p < 0.001. This calls hypothesis 6 into question.92 and agers’ engagement in ASA.90 0.41 −666. one < 0. SUP) 1312.88 0. df = 391.41.001). AUTO. Model 5 boosts very good fits (CFI > 0.87 0.43 ∗∗∗ 0.85 0. attain statistical significance in either Model 1 or the more par.01.05).05) simonious Models 2 and 2b. BC. 4.85 0. BC.77 0. BS = Belief Systems. TABLE II ALTERNATIVE STRUCTURAL MODELS Model and description χ2 Δ χ2 df NFI Delta2 TLI CFI RMSEA 0 Measurement model 1394. hypothesis 3.55 ∗∗∗ 75 0. further sub.458 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT. ∗ p < 0. we cannot find support for explaining the level of ASA.93 0. Yet.81 0. This provides additional evidence that hypothesis 1 cannot be In fact. DB = Diagnostic Use of Budgets.055 2b Model 2b (DB. p < 0. SUP) 301.38 ∗∗∗ 0.05 (two-tailed test).001) when top management support Since the relation between beliefs systems and ASA did not is removed as a variable. SUP) 922.79 0. Fig. this alterna. BC = Boundary Systems.25 ∗∗ 0. BC. stantiate our initial conclusion that hypothesis 3 should not Overall. from model 3b to 4 (Δχ2 = 164.36 ∗∗∗ 0.65. a relation. and 5. df = 73.87 0.81 0. SUP) 466.21 −337. AUTO.051 3b Model 3b (DB. ∗ p < 0. AUTO.01.88 6 Middle Managers’ Autonomy (AUTO) 0. the analysis of the path coefficients in AMOS and be accepted. experienced by firms.001).80 7 Interactive Use of Budgets (IB) 0.20 584. Model 3b can be on ASA cannot be supported with our data.77 0.86 0.06 −0. e. Similarly. p < 0.047 Significance levels: ∗∗∗ p < 0.92 0.86 5 Diagnostic Use of Budgets (DB) 0. IB. .88 0. p of dropping the diagnostic use of budgets from Model 1.31 ∗∗∗ 0. Model 3 that results from dropping belief sys.86 0. respectively Model 2b (Δχ2 = 337.18 ∗∗ N = 295..84 0.91.80 0. the comparison of model fits for the nested models suggest that tems from Model 2 exhibits overall acceptable fit with the data hypotheses 1 and 2 about the impact of how budgets are used (CFI = 0. (hypothesis 5) seem important for understanding the level of icantly improves fits in comparison to Model 2 (Δχ2 = 389.34 ∗∗∗ 0. VOL.91 0.83 0.12 −1175.95 0. BC.95.91 2 Belief Systems (BS) 0. model fit improves (Δχ2 = supported for our sample. we calculated an alternative Model 2b: instead p < 0. Significance of correlations: ∗∗∗ p < 0. Model 4 does not suggest such relative to Model 1 (Δχ2 = 1175.88 0. BS. AUTO. To shed additional light on the hypothesized relations. Model 4 relates bound- ary systems.55.92 ∗∗∗ 655 0. BC. BS.88 and RMSEA = 0.87 0. as Model 5 illustrates. DISCUSSION engage in further model trimming.28 640 0.92 0. out of the six ones hypothesized that truly seem to matter for df = 89.56 −621. ∗∗ p < 0. SUP) 1979.08 0.90 0.g.27 ∗∗∗ 0. df = 265.08. AUTO. p < 0.001. dropping belief systems may allow and shows significant positive effects of both autonomy and an enhanced model fit. RMSEA < 0. IB = Interactive Use of Budgets. Cronbach’s alphas along the diagonal. This lends support to hypotheses 4 are not related to the level of ASA and. boundary systems and RMSEA = 0.91 ∗∗∗ 205 0.01 0. 2 depicts the two relations df = 121. and top management support to ASA. AUTO) 159.44 ∗∗∗ 397 0. BS.15 0. we V.55 ∗∗∗ 0. ∗∗ p < 0.41 ∗∗∗ 132 0.72 0.95 0. NOVEMBER 2017 TABLE I CORRELATIONS Variables 1 2 3 4 5 6 1 Autonomous Strategic Actions (ASA) 0. SUP) 804. whereas earlier research found top management could remove the interactive use. support for entrepreneurship to statistically significantly affect tive way of trimming Model 1 boosts significantly improved fits the level of ASA (see. autonomy. This would indicate that belief systems boundary systems on ASA.43 ∗∗∗ 0. It exhibits good fit with the data (CFI = 0. The ary systems (hypothesis 4) and middle managers’ autonomy Δχ2 statistic suggest that dropping belief-systems indeed signif.

respectively. boundary systems seem to as lagged survey data or data from experiments. The core constructs of interest to us are of percep- Moreover. more research linking ASA to alternative functional relationships. important for motivating middle managers to engage in ASA. but seem ASA experienced by the firms in our sample. This can be expected to render tainting our results and while all three statistical tests which we finding and successfully developing new business opportuni. First. Different ways of using budgets may have a a covariance-based or a variance-based SEM. Our directional claims can thus only be drawn from the not received attention in the literature on ASA. Our study cannot for parsimony. it calls some of the claims voiced in literature test the hypotheses with data suited to causal inferences. future research should same time. the results of ASA and their consequences for performance tionships assumed. Regardless of whether one uses dle managers’ ASA. that ASA are an important source of strategic renewal. our study captured middle managers’ engagement in importance than whether budgets are used interactively or ASA. top management support. [32]. while scholars agree these forms of using budgets to explain firms’ ASA. re. whether one considers performance as an antecedent or Since our data were collected recently. our study highlights that boundary systems we cannot empirically test causality. the consideration of interaction or mediation requires data which allows models with substantial time lags. In contrast. who submitted that interactive use of budgets fosters mid. Even though our results seem fairly robust. Results for Model 5. Yet. diag. underlying theory. Finally. interactive use. Yet. we ran models that allowed such outcomes as enhanced performance or strategic renewal is for interaction effects between autonomy. such into question. This lends additional com- have–as hypothesized-a positive impact on middle managers’ fort to rejecting hypotheses 1 and 2 and accepting hypotheses engagement in ASA. Given the nature of ASA. several limitations As outlined earlier. our study relied on a single key informant given. Results–not reported here for parsimony. and [44]). tions for firm performance. spotting Our findings seem robust with respect to the functional rela. further studies with multiple respondents per firm seem when contemplating an investment of time and effort into ASA. we conducted the full Fig. carried out do not suggest common method bias being a prob- ties more likely–something that middle managers may consider lem.4 software.LINDER AND TORP: MIDDLE MANAGERS’ ENGAGEMENT IN AUTONOMOUS STRATEGIC ACTIONS 459 nostic use. Second. None of these models offered a superior fit with our data. like Linder and Bothello [37]. tional nature and members within one organization may per- financial resources too broadly to allow any individual initiative ceive them differently. To verify that our results for the acceptance/rejection of the hypotheses about the role of the different ways of using bud- gets are not due to the choice of our analytical procedure– covariance-based SEM–and given the growing popularity of the PLS method in management research. but not the results of these efforts and their implica- diagnostically. several explanations for such an effect can be merit attention. Moreover. we did not find such a relation. but merely show associa- are an antecedent of ASA–an antecedent which so far has tions. boundary systems tonomy and boundary systems are. Boundary systems provide guidance and thus heighten the per organization. a systematic bias in responses. 2. Whereas we followed common procedural rec- coordination of actions among individuals and facilitate coop. At the that these directional claims are invalid. and firm-longevity (see [24]. At least for ASA. Middle man. effects. explicit assessment of hetero- agers fearing that their engagement in ASA may never bear fruit geneous perceptions when measuring the variables may help in the sense of their initiative reaching a “proof of concept” stage reduce noise in the data and thus reduce the risk of accidentally due to a lack of resources are unlikely to engage in such actions rejecting a hypothesis. Therefore. [35]. However. the way in which plethora of effects–functional or dysfunctional ones (see Parker budgets are used does not seem to be related to the level of [65] for an overview of effects suggested in literature). or the analytical procedure used: We ran the models with able to expect that the differences in middle managers’ en- the squared values for the interactive way of using budgets and gagement in ASA would already have translated into perfor- for the diagnostic use to allow for exponential relationships of mance differentials between firms. warranted. in contrast. be the lever within Simons’ [18] framework that is of greater Third. While we do not have a reason to suspect to grow sufficiently to prove its economic viability. analysis in variance-based SEM (PLS) using the SmartPLS 3. there which assumed linear relationships (details not reported here is little empirical evidence on this matter. This suggests that boundary systems are 4 and 5. we tested whether past performance of firms explains the level of ASA-as anecdotal evidence suggests. ommendations to minimize the risk of common method bias eration and knowledge sharing. but available upon request). but available from the authors upon request–yield the same conclusions for These findings stand in contrast to the predictions of Simons the impact of the two different ways of using budgets than when [18]. in the first place. using covariance-based SEM. While we do not have a reason to suspect it underscores the importance of autonomy for ASA. adapta- sults are the same as the ones described in the results section tion. au- not to affect the level of ASA. the cross-sectional nature of our data implies that Consequently. Besides testing for address this issue. it would be unreason- not. needed. guidance avoids spreading a firm’s financial and non.2. .

NO.90 0.83 0. [20]. Focus on Strategic Uncertainties – Budgets and budget controls help top management and middle managers to discuss risks and opportunities that can 0.66 – Budgets are used to ensure that everything is on track 0.75 Face-to-Face Interpretation and Discussion – There is a lot of interaction between top managers and department/unit managers in the budget process. considering the different ways of using budgets decisions seem significant factors in explaining ASA. e.79 – The budgeting process creates discussions between top management and middle managers on the expectations of the 0.85 0. we test the impact of diagnostic and interactive use middle managers’ engagement in ASA. 0.89 0. [15]. prior anecdotal and case-based evidence (see. These alongside boundary and belief systems. First. we are the first to use a formative measurement in- [26]).89 0. seem to matter.89 0. This lends some comfort to accepting the other parts of strument to capture interactive use of a control system. . studies on interactive use of control systems have relied on reflective measurement instruments (see. tion of ASAs with how managers use a ubiquitous managerial it is important to note that our results for middle managers’ practice–annual budgeting–a relation that seldom attracts at- autonomy mirror those obtained in an earlier study by Linder tention from scholars in strategic entrepreneurship or business and Bothello [37] on data collected in 2009 from CFOs and innovation. To our best knowledge.72 ∗ Original measures in Danish. which budgets are used is decisive for their behavioral and firm. VOL. Burgelman’s [24]. We thus contribute to results are important for ongoing theory-building efforts and for the still sparse empirical literature that tests all the levers in advising practitioners.90 0. e. 0. Simons [18].65 – The fulfillment of budget goals is monitored closely 0.87 0.88 0.95 0.80 – In case of a negative deviation. [26] descriptions of ASA. an instrument which corresponds to his seminal work on the levers of control. little is known about its empirical validity. [31]). [64]).81 – We analyze deviations from the budget closely 0. Simons’ framework within a single study.85 0. while his claim is research about the control systems-ASA relation. e.89 0. Besides autonomy. boundary systems of budgets together and within Simons’ [18] levers of con- and the extent of middle managers’ autonomy in market-related trol framework. NOVEMBER 2017 While these limitations call for caution in interpreting results. we are the first to use a formative measurement instrument of level consequences (see..64 – Top management often uses budget information as a means of questioning and debating the ongoing decisions and 0.79 actions of departments Intensive Use by Operating Managers – The budget information forces both top management and managers at all levels of the firm to continuously question and 0.76 affect the strategic goals of the company – The budgeting process unites top management and middle managers in open discussions about the company’s strategy 0. we draw on compelling and has been reiterated in literature over the past a multi-item operationalization of ASA which is in line with two decades. hence. e. in interactive use and.85 0. thereby avoiding Our results do not allow us to support Simons’ [18] claim that confounding ASA with other intrapreneurial behaviors. While our findings. Yet. 0.85 Intensive Use by Top Management – Budget controls are an important and continuous part of top management tasks 0. Thus. we provide rare evidence on the rela- APPENDIX 1: MEASURES∗ Dimensions and Variables Construct AVE Factor Indicator Reliability Loadings Reliability Interactive Use of Budgets (IB) 0.79 actions of department/managers – The budgeting process is continuous and it demands regular and frequent attention from managers at all levels 0. translated into English by the authors. and [27]–[30]). [24] and Second. literature agrees that interactive use is a formative construct (see. Similarly.g.68 Diagnostic Use of Budgets (DB) 0.87 0.g.72 – All managers in the organization are involved in the budgeting process 0..80 0.g. in contrast to explicitly links interactive use to ASA.. e. thus potentially suffering from An increasing number of scholars agrees that the way in a misspecification bias (see. using annual budgets in an interactive way enhances the level of Third. and [19]). hence.. our results Business practitioners can take away from this study that del. VI. Our study contributes to the literature in at least three ways.g. 64..69 – Top management uses the budgeting process to discuss the competitive actions of the company with middle managers. [18].460 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT.72 revise their assumptions – Top management often use budgeting information as a means of questioning and debating the ongoing decisions and 0. In fact. CONCLUSION [4]. underscore the importance of considering boundary systems for egating decision-rights to middle managers promises to foster ASA and.82 0. corrective action is initiated 0.80 0.85 0. 4.77 company plan.g. of a “lever” that so far has attracted little ASA. Yet. an organization’s boundary systems systematic research. [19] Simons’ [18] conceptualization [64].72 and activities – Top managers continuously discuss the budget and budget controls with middle managers. [16].

31. 15. considerably enhance the quality of the manuscript. 1975.” Manage..” J.. “Managing budget emphasis through the ex. “Budget use and managerial performance. Account. 27.89 0.63 – Wonder how to improve business and the firm’s strategic position by introducing new products to new markets/customer 0. Lambert. 274–284.79 Top Management Support (SUP) 0. doi: 10.77 0. product innovation. Young. 2004. vol.2016. 4..73 – Develop plans on how new business opportunities in new markets can be implemented into the firm’s activities.64 a new market ACKNOWLEDGMENT [8] S. direction and purpose 0. pp. T.. pp. Libby and R. 2013. Brownell. 2005. [11] D.” Account. 13. The authors would like to thank the editor and the four anony. 177–203. “Budgeting in times of economic crisis. 5. Foster. Bruns and J. pp. Askholt vol. Soc. Bisbe and D. no. MA. “Is the annual budget really dead?” Eur. Cost Accounting: A Managerial istics. W.79 – Our mission statement inspires our managers and employees 0. and firm performance: Empirical evidence from [7] B.” J. and the moderating effect of strategic uncertainty. Soc. J. control systems on product innovation. “Budgetary control and organization School Press. A.68 – Our mission statement clearly communicates the company’s core values to our workforce 0. T.86 0. vol. structure. pp. [20] E.. 33. Boston. 29. Soc.. D.69 – Our workforce knows the firm’s code of business conduct 0. “The effects of the interactive use of management pp. ment control systems research. 2012. Res.. Account. Datar.” Manage. Manage. 2006. 1993.1109/TP. 2016. 127–143.” J. of the 2016 ACMAR and EURAM conferences for their com.” Asia-Pacific Manage. D. D.” Account.82 0. 24.LINDER AND TORP: MIDDLE MANAGERS’ ENGAGEMENT IN AUTONOMOUS STRATEGIC ACTIONS 461 APPENDIX 1: (CONTINUED) Dimensions and Variables Construct AVE Factor Indicator Reliability Loadings Reliability Middle Managers’ Autonomy (AUTO) 0. Levers of Control. no. F. 127–154. G. pp.81 0. [19] R. Otley.85 0. 1999. T. Res.80 0. Shields and S. 50. “Diagnostic and interactive uses of budgets pp. Otley. 1995. 6. Milani. A.2572160. WRD.72 – Top managers act according to the company’s core values.. and to J. 15.” Account. C.. M. 3.82 0. thus creating a common direction 0. F. M. S.” Manage. Control. vol. 47–61.” Account. “Antecedents of participative budgeting. Account..1109/TP. Salonen. Horngren. 2003.63 – Seek support on the part of top management for ideas to start sales/production of new products for new markets 0. “Practice developments in budgeting: An overview and research perspective. [16] J.. 56–75. Wouters. Org. Schäffer. K. 10. 10. 1975. Org. for his support in running the PLS analyses in SmartPLS. Org. Otley.. M. vol. and W. 2010. Mahlendorf.77 – Sales to new customer segments or markets 0. vol. vol. Eckholm and J. Waterhouse.79 0. pp.” Account. Manage. “The relationship of participation in budget-setting to indus. “The role of management control systems in creating com- WRD. vol. vol.. Q. 9. Systems to Drive Strategic Renewal. no.” Account.61 – Activities aiming at increasing market share 0. “Interactive budgeting. Account. no. 13th ed. vol. Org. 293–322. pp.66 – They appreciate it when middle managers experiment with new ideas and products 0. pp.58 Boundary Systems (BC) 0. 21. also like to thank the respondents who provided information on [10] M.90 0. They would Res.59 segments – Seek information on business areas (products. USA: Pearson. Account. Emphasis.83 0.66 – They listen to middle managers when significant decisions are made 0. “Performance measurement: A framework for manage.88 0. 265–280.85 0. Laitinen. 0. Lindsay. 49–76. Shields and M..80 0. Wallin.” Account. T.82 0. [12] J.85 0. D. vol.77 – Our firm has a system that communicates risks and activities that should be avoided 0.74 – Our employees know the company’s core values. markets) outside the firm’s current activities 0. Henri. pp. “Antecedents and consequences of partic- their firm’s practices. Org. and M. vol..69 – Our code of business conduct and other guidelines specify behaviors that are off-limits 0. 363–382. 16. pp. A.92 0. 122–149. Shields. [4] S. pp. and S. 1978. Res. vol. vol. vol.. 23. 519–539. [15] M.87 0.64 markets outside the normal budget procedure – Form coalitions with other managers in the firm in order to rally support and approval for going with a new product into 0. Account. 709–737.76 Autonomous Strategic Actions 0. D. Hansen. 189–204. and G.” J. 2016. Davila and M. doi: 10. Account.69 – They are open to new ideas and initiatives from all employees 0. [17] J.83 0. 1489– REFERENCES 1517. “Beyond budgeting or budgeting reconsid- ered? A survey of North-American budgeting practice. pp. vol. Sponem and C.86 0. Länsiluoto. plicit design of conditional budgetary slack.83 0. 529–558. [13] S. MA. Shen and S. [14] S. vol. 1999. 1998.. 95–116. perspective. 7. M. Thaten. 30.61 – Our firm relies on formal guidelines for specifying appropriate behavior 0. Otley. pp. pp. Finnish firms. [18] R. doi: Account.” Contemp. Simons. 30. J.63 – Look for ideas on new business opportunities for the firm by selling new products to new markets 0.59 – They actively seek middle managers’ opinions and ideas about strategic issues 0. Abernethy and P. Van der Stede. U. Soc. trial supervisor performance and attitudes: A field study.2618601. Org. J. 2016. Res. vol. USA: Harvard Business [5] W. H. S. “Management control systems and strategy: A resource-based [3] T. Boston. “The role of budgets in organisa- [2] D. Perera. ments on earlier versions of the manuscript.81 – Development of significant new products 0.1007/s00187-016-0237-2. petitive advantage: New perspectives. They are also indebted to the participants ipative budgeting: Evidence on the effects of asymmetrical information.76 0. roles and satisfaction: A configural approach. ... Res. mous reviewers for the many helpful comments that helped to [9] T. How Managers Use Innovative Control Account. 1990. Res.80 0. pp. 587–608. 2. 4.81 0. Becker. Rev. Manage. [6] K. tions facing strategic change: An exploratory study. Soc..67 Belief System (BS) 0.67 – Search for more resources outside regular budgeting processes for projects aiming at introducing new products to new 0.89 0. “Exploring differences in budget character- [1] C. Simons..88 0. no.. Account. Rev. pp.89 0. Soc..2016. 2000. Res.

Bus. “Using performance measures [48] D. C.” Manage. 11. Abrahamsson. Kuratko. Eng. 47–69. on the use of budgets. ture. “Individual and contextual predictors of creative performance: [31] W. Hartmann. performance. NY. Cummings. R. Rodgers and A. J. J. F. [44] T. 2009. no. [54] J. A. 327–339. Moulang. “Sources and consequences of competitive 10. pp. “Twentieth-century textbook budgetary discourse: Formal- [39] J. S. vol. 76. “How to kill creativity. 81. 1349–1364. Burgelman. 1981. 2000.” Entrepreneurship literature review and a conceptual framework. “Antecedents to autonomous strategic action: 1993. Hopwood. T..” Accounting. 5. vol. then rein in chaos— mation systems for strategy implementation in hospitals. Fornell and D.” U. 17. vol. C.” Long Range Planning. Sarpong. Control. vol. and S. 2010. B. Widener. 385–415. pp.” Acad. “Employee creativity: Personal and [36] R. no. “Corporate entrepreneurship and strategic management: Acad. line]. 17. vol. W. USA: Wiley. vol. G.-J. 1996. [45] S. 2007. international business research. Rev.. a measurement scale. [58] J. A. 1996. Man- Theory Practice. Manage. 39.. K. 4. and J. 305–327. L. “Organizational struc- roles of accounting in organizations and society. vol. [22] S. vol. “Intraorganizational ecology of strategy making and [61] S. 2. Moeller. van Witteloostuijn. J. T. mature organizations: Overcoming innovation-to-organization problems. 2012. [35] R.. “From the editors: Com- Control. NOVEMBER 2017 [21] S. C. 389–405. [67] T. A. NY. 41. doi: 10.. Manage. Zahra. Rev. 226– [64] Bisbe. Noda and J. and measuring autonomy: An entrepreneurial orientation perspective. 28. 5. Zahra. accounting constructs: a methodological note on the risks of conceptual [38] R. Strategy Manage. Manage. Günther.” Account. Manage. 29. 4. 23. Cogliser. no. 2007. vol. pp. Hornsby. Kanter. pp.. and M.. 239–262. J. merit-based Aug. unobservable variables and measurement error. no. [55] D... 159–192.” Strategic Manage.” Strate. and S. Syst. Linder and J. Hult et al. vol. S.” Acad. [62] G. pp. vol. and supply chain man- 1996. 153–182. vol. USA: Prentice-Hall. 1980. Middle managers’ contribution to implemented information 1983. Manage. A Behavioral Theory of the Firm. 3.. 7. 77–87. pp. no. 2011. Lumpkin. Covin. 4. M. May 2015. Larsen.. 29–41. J.” J. Bothello. and proposal making in administrative bureaucracies. 33. pp. pp. 148–172. Nahapiet.1007/s00187-015-0227-9 mon method variance in international business research. Inf. Hult. Guiral. J. [On. Maclean.K. 2015. Account. 2011.. What about decline?” IEEE Trans. 2011. no. 1019–1031. 29. A. Grow from Within: Mastering Corporate misspecification. 5. Simon. 2006. vol. 2004. 30. New York. vol. London. 2013. [46] D. Bellora-Bienengräber and T. no. [40] R. Agostino and M. S.. Frishammar. L. 63. Hobo- chological empowerment and individual creativity.” IEEE Trans. no. NY. S. 32. Org. agement. Burns and G. J. 689–706. Hardy. and D. Larcker. 178–184. S. NJ.” J. Scenarios: The Art of Strategic Conversation. 699–716. vol. vol. 1–23. S. Bedford.” J. Psychol. L. 2. Soc. innovation: Implications for firm performance. Ireland. vol. Chang. Airline industry. methods bias in social science research and recommendations on how to [43] T. 18. 184–200.. Aiken. no. and Chenhall.” Annu. 25–50. 2000. J. vol.. [65] L. Wolcott and M. Linder. Bower. D. K. [57] J. [51] D. pp. Entrepreneurship and Innovation. F..” [24] R. Man- Account. Englewood [66] C. Andersen. Burgelman. 607–334. van der Heijden. G. J. pp. [42] S. Sci. Craighead. 1. 2012. 1983. M. ken. pp. A. Arnaboldi. G.” IEEE Trans. Manage.. pp. pp. pp. “Potential model misspecification bias: For. Insights from a process study. J. Kurkkio. 2016. Available: http://ssrn. Rev. vol.” Manage. 2005.. Hughes. Choi.” Health Policy. J. Eng. 2008. New York. J. [63] T. Hart. no.S. vol. cor. Burgelman and A. vol. pp. vol. 35. and L. Clubb. Org. 155–176. pp. Quart. Wearing.” Eur. “The interplay of tion of the internal environment for corporate entrepreneurship: Assessing the levers of control in product development. 32. Day. vol. F. Organization. 4. Oldham and A. Rev. pp. no. “A matter of foresight: How [27] S.” Account. gic Manage. Manage. M.. and N. Marketing Res. Control. 3. USA: McGraw-Hill. “Understanding entrepreneurship strategy. 39. C. “Addressing common method variance: Guidelines for survey re- source allocation. pp. vol.” Res.” Administ. 7.” vol. vol.” J. vol. vol. “Management control systems across different modes of vol. no. mative indicators enhancing theory for accounting researchers.” J.. Manage. 613–625. K. Manage. J. vol. M. 2010. “A model tecedents and consequences of firms’ process innovation capability: A of middle-level managers’ entrepreneurial behavior. Organizations and Society. Stalker. Sci.2055412 inertia: A study of the U. pp. 2009. E. 1996.. Ketchen. Acad. Manage. “An- [32] D. R. 519–529. Dougherty and C. Hornsby. pp. Dunn. 1980. J.” J. vol. Widener. French. trepreneurial opportunities. “Conceptualizing corporate [60] G. and D.. 327–351. vol. [33] T. D. pp. A. The mediating role of psychological processes.” Strategic Manage. and D. 33. 39–50. 33.. and L. W. [23] D. J. Mann. Manage. L.. Jr. New York. “Sources of 257. [53] K. F. J. vative corporate ventures. Soc. 39. 22. [28] C. 578–588. Soc. [37] S. pp. Amabile.. “Let chaos reign.” Eur. 8.: Tavistock Publications. 46.” J. 61–70. Rev. A. vol.. March. promotions. no. “Determinants and effects of the pp. pp. normalization and rebuttal in an anglo-saxon environment. [71] R. no. K. 187–199. Kuratko. [59] S. 2007. pp. Design issues in Balanced Scorecards: The 2002. Chen.. Entrepreneurship Theory Practice. NJ.. 757–788. 1994. 539–569. “Raising radicals: Different processes for championing inno. VOL. 58.. [69] C. Grafton. Grove. vol. Principles and Practice of Structural Equation Modelling.” J. Kuratko. Account. “An empirical analysis of the levers of control framework. Eden. “An integrative framework for strategy-making processes.” Harvard Bus. NO. 2. 1991.” Entrepreneurship Theory Practice. Methodol. 519–544. M. 62. 23. Lichtenthaler. M. repeatedly: Managing strategic dynamics for corporate longevity. 3. NY. pp. 14. M. 17. [41] R. [70] S. in the American Corporation. vol. 28. “Performance measurement system use in generating psy. Cyert and J. 1958. vol. The Management of Innovation. technology innovation. Sep. 1994. D. 1963. age. B. or http://dx. pp. 3. P. pp. M. [52] D. 27. Covin. Lillis. Parker. [30] S. Miller and M. M. Venturing. Acad. rate autonomy. Burgelman. Ireland. R. 6. Janssen. “The role of performance 1983. [50] J. 2012. vol./Oct. 55. 2002.. A.. 251–280. D.” Int. Sci. mational leadership. S. a U. Sci. Carless.doi. Hofmann. Account. M.. pp. pp. 5–27. “Strategy making as iterated processes of re. March and H. “A short measure of transfor- organizational adaptation: Theory and field research. The Change Masters: Innovation and Entrepreneurship contextual factors at work. “A model of the interaction of strategic behavior. 631–652. “Defining management 236.. and G. work process. Burchell. pp. A. 243– [68] P.” pp. 1992. Gleich. Manage. 107– on strategy formation at middle-management levels: Evidence from 128. diagnostic and interactive use of control systems: An empirical analysis [56] T. B. “Middle managers’ percep- [29] L.” Org. Schlaefke. “The virtuous circle of discovery and creation of en. age. Eng. J. and engagement in autonomous strategic action. 64. Bus. 965–979. no. Bacharach. vol. 10. [34] R.. 3.. Organizations. 31. 1961. 2012. “How CEOs use management infor- [26] R. N. Stud. 2005. J. A. control it. vol.” practices enable (or impede) organizational foresightfulness. Res. J. [49] D.. “Management control systems and their effects conceptually in innovation control. “Fostering strategic renewal: monetary incentives.” Account. search on information technology. USA: Simon & Schuster. Podsakoff. Psychol. J. MacKenzie. 789–820.. pp. Lippitz. . and J. 2015. 2007.” Org. Nov. Clayton. G. and the concept of strategy. 12–30. Podsakoff. 16. W. “what” and “how” of control. pp. ization.” Strategic Entrepreneurship J. Hoboken. USA: Wiley. Finance.” Creativity Res. “Sustained product innovation in large. 23.2139/ssrn. vol. pp. pp. Bus. operations. 2. USA: Guilford. Schneider. “An assessment of the use of structural equation modeling in pp. Batista-Foguet. vol. pp. 59. no. L. Wald. 253–273. 1998. vol. Org. 2002.” 2012. and U. 19–46. “The [47] M. 1120–1153.462 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT. pp. measurement and evaluation in building organizational capabilities and [25] D. and J. “The performance effect of strategic planning and corpo. “Evaluating structural equation models with Cliffs. Naranjo-Gil and F. A. 2010. A.K. Int. D.” Eur. Kline. and R. S1. pp. 2. J. Marginson. porate context..

vol.. degree in eco- nomics and management from Copenhagen Business School. Torp received the M. J. terplay between management. 419–456. M. . “An updated paradigm for scale develop. [77] P. and capital expendi- ture management. “The use of LISREL in validating tional structures. pp.” J. motivation and incentives. Van Trijp.” Soc. Steenkamp and H. Anderson and D. 1980. 1996. Market. Frederiksberg. Res. “Multivariate analysis with latent variables: Causal model- ing. 132–160. Denmark. agement. 2. Anderson. Bentler. His research has been published in marketing constructs. Res.D. planning and budgeting. vol. His research interests cover strategic risk fit indices for structural equation models.. vol. Stefan Linder received the Ph. “The use of partial least Simon S. Psychol. pp. He is an Associate Professor with Aarhus Uni- [74] D. His recent research articles were featured in. management. Gerbing and J. no. and organiza- [75] J. Hamilton. Aarhus.” Annu. W. 1988. Gerbing. Frederiksberg. Paris. R. France. and R. 31. Ringle. no. 1.” Int. C. and the Journal of Banking and Finance. degree in corporate finance [73] J. ing and auditing from Aarhus University. Henseler. Denmark. and Economic and Industrial Democracy. Methods Res.. vol. 20. Equation Model. B. 283–299. among others. the IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT. 4. G. and the in- pp. vol. middle managers’ autonomous strategic ac- tions. degree in account- squares in path modeling in international marketing. 186–192.D. Gerbing and J. C. Management Decision.. Rev. in 2011. Denmark. Market. pp. 8. E. Market.” Adv. His research interests include organizational control. 25. 277–319. [76] D. pp. Int. 62–72.LINDER AND TORP: MIDDLE MANAGERS’ ENGAGEMENT IN AUTONOMOUS STRATEGIC ACTIONS 463 [72] J. and the Ph. 3. “Monte Carlo evaluation of goodness of versity.. the International Journal of Strategic Change Man- 1991.Sc. vol. incentive design. 1992. pp. in 2011.. “Viability of exploratory factor analysis as a precursor to confirmatory factor analysis. 2009. M. the European Accounting Review. finance. He is currently an Assistant Professor with the ESSEC Business School. and strategy from the Copenhagen Business School.” Struct. Sinkovics. ment incorporating unidimensionality and its assessment.