Sunteți pe pagina 1din 28

Effects of management control mechanisms: towards a more

comprehensive analysis

Sebastian Goebel . Barbara E. Weißenberger

Abstract Even though it is widely accepted that management control systems consist of formal (results
and/or action controls) and informal mechanisms (personnel and/or cultural controls), empirical
research analyzing the effects of these management control mechanisms in a more comprehensive way
is still scarce. Based on a cross-sectional survey among 295 senior management accountants, we
examine how different management control mechanisms are related to control system effectiveness and
organizational commitment and how these two outcomes subsequently affect overall organizational
performance. Our results suggest that particularly more informal control mechanisms are strongly
associated with beneficial outcomes. Our study contributes to extant literature by identifying distinct
relative effects of alternative management control mechanisms on organizational- level outcomes.
Overall, our results support the growing relevance of more informal control mechanisms compared to
prevailing formal control instruments in contemporary organizations.

Keywords Management control . Management control systems . Formal control . Informal control .
Structural equation modeling

JEL Classification M10 . M40

1
1. Introduction

Achieving and sustaining management control is a critical issue in virtually all firms. Any loss of
management control can lead to reputation damage, financial distress, or even insolvency and overall
organizational failure. Today, it is generally accepted that management control is achieved by a broad
array of formal as well as informal instruments, comprising, e.g., performance measurement systems,
budgets, corporate policies and regulations, employee selection and training, codes of conduct, or even
the ‘tone at the top’ (e.g., Merchant/Otley 2007; Collier 2005; Abernethy/Chua 1996). Still, with respect
to the effectiveness of these instruments, literature in the field of management accounting and control
has traditionally analyzed only individual management control mechanisms in isolation with a particular
focus on accounting-based controls (Malmi/Brown 2008; Widener 2007). In a similar vein, management
control practice in many organizations also primarily relies on results-oriented controls to coordinate
and evaluate employees (Merchant/van der Stede 2000), even though such a largely ‘cybernetic’
command-and- control approach often is deemed not flexible enough to accommodate for accelerated
changes in today’s corporate environment (Nixon/Burns 2005; Otley et al. 1995). As a consequence,
there is only a limited understanding of the comprehensive effects of alternative control system
elements, in particular with respect to rather non-traditional and more informal control mechanisms that
have continually gained importance in the last decades (Tucker 2010).
Our study therefore addresses calls for research by applying a broader control conceptualization
and thus studying how different management controls relate to each other and operate as a package (e.g.,
Malmi/Brown 2008; Chenhall 2003; Abernethy/Chua 1996; Flamholtz et al. 1985; Otley 1980). More
specifically, we focus on the relative effects of formal and informal modes of control. In the last decades,
firms have increasingly refined their central ‘control philosophy’ towards an intensified use of informal,
i.e., more organic, controls to influence their employees’ behaviors and to reduce the reliance on
accounting-based performance measures as primary means of control (Frow et al. 2010; Stringer/Carey
2002). These changes have reinforced the growing importance of rather soft factors like, e.g., the
consideration of human capital as a key strategic resource (Widener 2004; Youndt et al. 1996) and it is
nowadays increasingly recognized that a set of well- designed formal and informal control system
elements constitutes a necessary prerequisite for sustainable organizational success (Epstein 2008).
Still, concurrent effects of more informal control mechanisms like an adequate employee training
and development approach or the impact of a coherent corporate culture on individuals’ behaviors have
been exposed to limited academic scrutiny and constitute a considerable research gap (Sandelin 2008;
Collier 2005). Most management accounting research thus continues to follow a rather partial approach
and addresses only partial models of organizational control structures (Berry et al. 2009; Speckle ́ 2001).

2
However, fragmentary analyses of isolated management control mechanisms can seriously constrain
the validity and reliability of empirical research results. Assessing only specific elements of an overall
management control system separately from other control mechanisms bears the risk of ‘‘serious model
underspecification’’ (Chenhall 2003, p. 131) and may cause unclear and conflicting results (Sandelin
2008; Fisher 1998). The consideration of a wider control definition that includes effects of formal as
well as informal control mechanisms may therefore provide explanations for inconclusive research
results by identifying potential complementary or substitution effects (Malmi/Brown 2008; Sandelin
2008; Bisbe/Otley 2004; Langfield-Smith 1997; Fisher 1995). This emphasizes the imperative need for
an integrated analysis of management control systems.
Recent research has started to recognize the growing importance of a more holistic analysis. For
example, Sandelin (2008) uses Merchant/Van der Stede object-of-control framework (Merchant/Van
der Stede 2012; Merchant 1985) to identify different management control packages formed by formal
as well as informal controls in a European technology firm. In a similar vein, Mundy (2010) conducts a
field study in a European multinational financial services organization and uses Simons’ (1995) levers
of control framework to provide evidence on the balancing of different control elements within the
firm’s management control system. Chenhall et al. (2010) conclude from a field study of a non-
government organization that formal management control mechanisms can either add to or impede the
effectiveness of cultural controls. Finally, Chenhall et al. (2011) address the relationship between
strategy, innovation and management control systems by focusing on formal vs. cultural controls.
Surveying 100 Russian CEOs, they find, among others, that both types of control mechanisms have an
impact on innovation, but that managers in firms following a product differentiation strategy rely more
heavily on the latter.
Summing up, the results of existing studies consistently show that organizations generally rely
on a variety of different control mechanisms that should be analyzed simultaneously in order to examine
their respective relative effects. Our paper deals with this research gap by addressing the research
question how different formal and informal management control mechanisms are related to control
system effectiveness and organizational commitment on an organizational level and how these two
outcomes subsequently affect overall organizational performance. For this purpose, we develop a
theoretical model based on the object-of-control framework (Merchant/Van der Stede 2012; Merchant
1985) with respect to distinct effects of alternative management control mechanisms. We test this model
empirically using PLS-based structural equation modeling with survey data gathered from 295 senior
management accountants of large- and medium-sized German companies. Our research approach is
thereby consistent with Otley’s (1999, p. 377) suggestions that ‘‘a more holistic approach is clearly

3
appropriate, with the unit of analysis being the organization’’.
Overall, our findings indicate that especially more informal management control techniques are
associated with beneficial effects. Our findings also indicate that a simple theory of ‘the more, the better’
does not hold, especially with respect to the use of results controls like, e.g., performance measurement
systems. Rather, the contribution of management control systems to organizational performance results
from a combined impact of formal and informal management control elements.
Our study contributes to management control research in a threefold way. First, we provide novel
empirical evidence to an emerging research area by analyzing control systems from an integrated point
of view that addresses a comprehensive array of alternative control mechanisms and analyzes their
respective effects in a more holistic fashion. We thus contribute to the understanding of other types of
control mechanisms besides traditional accounting-based controls. Ultimately, this more comprehensive
approach serves as a step towards the development of improved theories of how to design effective
management control systems (Malmi/ Brown 2008; Otley et al. 1995). Second, we follow Merchant et
al. (2003) who have called for more cross-disciplinary research that bridges single paradigms by
combining economic and behavioral theories. By relying on insights from agency and stewardship
theory, we are able to confirm that both theories offer complementary explanations for differing effects
of various management control mechanisms which may contribute to an improved understanding of the
overall effects of management control system elements. Third, in addition to recent empirical studies,
which have primarily applied qualitative, case-based research approaches in a single organization or a
small number of organizations (e.g., Chenhall et al. 2010; Mundy 2010; Sandelin 2008), we contribute
to existing literature by relying on a comparatively large cross-sectional data set that provides a
broadened empirical basis to study the effects of management control mechanisms.
The remainder of this paper is structured as follows. First, the relevant literature streams will be
reviewed in Sect. 2. In Sect. 3, we develop several distinct hypotheses regarding the effects of alternative
management control mechanisms. Section 4 provides a discussion of our research strategy as well as
methodological issues like, e.g., the identification of appropriate measurement instruments and the
characteristics of the chosen statistical data analysis technique. Section 5 describes the results of our
analyses. We conclude the paper with a discussion of the results and an illustration of the contributions
and potential limitations in Sect. 6.

4
2. Literature review

Our research is nested in two different streams of literature. First, we draw from the broad literature on
management control systems to develop our theoretical framework. Second, we use agency theory as
well as stewardship theory as underpinning for our hypotheses regarding the effects on management
control system effectiveness and organizational commitment and how these two outcomes subsequently
affect overall organizational performance.

2.1 Management control systems

Anthony’s (1965) seminal work on management control systems is generally referred to as starting point
and seminal publication in the field of management control systems research and it is still associated
with a persistent effect on this discipline (Merchant and Otley 2007; Henri 2004). Anthony (1965, p.
17) defined management control as ‘‘the process by which managers assure that resources are obtained
and used effectively and efficiently in the accomplishment of the organization’s objectives’’ and
characterized it as the process that connects strategic planning (setting of long-term organizational
objectives) and operational control (evaluation of particular tasks). This fundamental definition has
strongly influenced subsequent researchers and has led to a predominant focus on financial
accountability and accounting-based performance measures to conduct adequate control in
organizations (Merchant/Otley 2007; Otley 1999; Langfield-Smith 1997). However, this rather narrow
conceptualization has been critically challenged during the last years (e.g., Johnson/Kaplan 1987) and
the traditional management control conceptualization has been refined from a primarily accounting-
based view to a more comprehensive or holistic perspective that comprises formal as well as informal
management controls. Collections of these individual management control mechanisms are generally
referred to as a ‘management control system’ (Merchant/ Van der Stede 2012; Bisbe and Otley 2004;
Flamholtz 1996; Flamholtz et al. 1985) or a ‘management control package’ (Malmi/Brown 2008;
Sandelin 2008; Otley 1980, 1999).
This broadened perspective implies that proper control can be achieved by applying two distinct
control strategies (Ouchi 1979). On the one hand, control can be conducted by relying on formal, i.e.,
‘mechanistic’ (Chenhall 2003), performance evaluation processes, extensive budgetary controls and
incentive compensation systems, detailed rules and standard operating procedures, etc.
(Abernethy/Chua 1996; Falkenberg/Herremans 1995). These instruments are designed to ensure
efficient and effective work processes and to maintain financial viability (Chenhall et al. 2010). On the
other hand, organizations can influence their employees’ behaviors by utilizing informal, i.e., ‘organic’
(Chenhall 2003), control mechanisms to promote an understanding of the overall organizational

5
objectives and to minimize divergent individual preferences (Eisenhardt 1985). Thus, informal
management control instruments like, e.g., selection and training approaches or the design of an
integrative corporate culture that builds on shared values and beliefs are considered as important
management control system components that affect employees’ perceptions and actions (Sandelin 2008;
Flamholtz 1996). Although informal management control mechanisms occasionally also encompass
individual formal elements, e.g., explicitly codified statements of an organization’s mission and core
values, they are, nevertheless, considered to be relatively more informal than formal due to their rather
indirect focus.
The importance of informal control mechanisms has been increasingly recognized in research
and practice and they are described as means to regain the relevance of management accounting (Otley
2001; Johnson 1992; Johnson/Kaplan 1987). In contrast to formal controls, which are based on explicit
and quantifiable standards, informal controls are less obtrusive and thus associated with fewer
disadvantages (e.g., Kerr 1975) because they affect behaviors rather indirectly (Merchant/Van der Stede
2012). A further distinction between both control alternatives is that formal controls are mainly
associated with the allocation of extrinsic rewards (e.g., base salary increases or higher levels of
performance-contingent incentive compensation), while informal controls are closely related to the
provision of intrinsic rewards (e.g., increased levels of autonomy or opportunities for personal
development) that can have a more sustainable effect on behaviors by fostering employees’ intrinsic
motivation (Rost/Osterloh 2009; Epstein 2008; Manzoni 2008). However, formal and informal control
mechanisms do not exist in isolation and it is generally acknowledged that companies rely on both
means of control in any setting (Collier 2005; Abernethy/Brownell 1997). A thorough understanding of
appropriate combinations of formal and informal controls is therefore considered to be of high
importance (e.g., Epstein 2008) since a well-designed management control system can ultimately lead
to a long-term competitive advantage (Flamholtz 1995).
Nevertheless, despite regular calls to study this phenomenon (e.g., Malmi/Brown 2008; Chenhall
2003; Abernethy/Chua 1996; Flamholtz et al. 1985; Otley 1980), relatively little is known about entire
management control configurations in contemporary organizations (Malmi/Brown 2008; Laitinen et al.
2004; Subramaniam/Mia 2003; Abernethy/Chua 1996). Consequently, Kruis (2008, p. 7) emphasizes
that ‘‘there are hardly any studies that focus on MCSs in their entirety’’ which leads to a ‘‘rich, but
deceptive’’ (Cardinal et al. 2010, p. 51) base of control literature. According to Mundy (2010), the
overall effects of control approaches can therefore only be evaluated by applying a holistic view that
simultaneously considers an array of different management control system elements.
Usually, management accounting research has applied a very restrictive approach and considered

6
only overall financial performance as a direct outcome or as a proxy for desired effects of management
control systems (Kruis 2008; Henri 2006; Otley 1999; Fisher 1995). However, Otley (1980) argued that
control mechanisms are likely to have only a small effect on organizational performance and recent
research has shown that there is no compelling evidence for a direct relationship between alternative
management controls and financial performance (e.g., Widener 2007; Henri 2006; Bisbe/Otley 2004).
Nevertheless, only a limited number of studies has investigated other outcome variables than
performance or considered indirect effects of control system components on organizational performance
(Shields et al. 2000; Fisher 1995). Particularly, there is a lack of studies that explicitly consider the
effectiveness of management control systems as an outcome, although there are many claims in theory
about this issue (Kruis 2008). For example, the effectiveness of management control systems can be
assessed in comparison to the realization of control objectives, i.e., the avoidance of common control
problems in organizations (e.g., Ferreira/Otley 2010; Kruis/Widener 2009; Kruis 2008; Flamholtz
1996).
To empirically assess management control systems in business practice, several conceptual
typologies have been introduced in literature. These include Simons’ (1995) levers of control
framework, Otley’s (1999) performance management system framework, Malmi/Brown’s (2008)
management control package framework, and Merchant/Van der Stede object-of-control framework
(Merchant/Van der Stede 2012; Merchant 1985).
The object-of-control framework specifically classifies management controls according to their
focus and allows for a differentiation between formal (results and action controls) and more informal
mechanisms (personnel and cultural controls). While results controls, which consist of financial and/or
non-financial target indicators, are used to define expected results and to monitor and evaluate
employees’ performance, action controls ensure that employees conduct only those activities that are
presumably beneficial to an organization. Examples of typical action controls are behavioral constraints
or pre-action reviews, e.g., in the form of detailed checklists and extensive standard operating
procedures. In contrast to these more formal controls, personnel controls are applied to design
prerequisites to foster individual motivation, e.g., by thorough selection and placement processes or by
appropriate employee training and development programs. Finally, cultural controls can be used to
define expected norms and values. Thus, they shape an organization’s internal climate and influence
employees’ mindsets by defining social conventions.
For our research, we will rely on the object-of-control framework for the following reasons. First
of all, the performance management system framework (Otley 1999) is intended to be used primarily in
case-based research (Merchant/ Otley 2007) and it largely ignores more informal means of control

7
(Sandelin 2008). Similarly, Simon’s (1995) levers of control framework focuses exclusively on formal,
i.e., explicitly codified, control system elements and it is also associated with several conceptual
ambiguities (Tessier/Otley 2012; Frow et al. 2010; Collier 2005). Although the framework proposed by
Malmi/Brown (2008) includes a variety of formal and informal control elements, it is a relatively new
framework which has not yet developed towards non-overlapping categories for different control
mechanisms. In contrast, the object-of-control framework is theoretically well-grounded in a long line
of empirical research that builds on Ouchi’s (1977, 1979) seminal control classifications
(Merchant/Otley 2007). Hutzschenreuter (2009) emphasizes that pre-testers found it easily accessible
and adequate to identify the full extent of various control forms in their respective organizations. The
object-of-control framework thus combines a broad scope of control mechanisms with sufficient rigidity
(Sandelin 2008) which supports its suitability for our study.

2.2 Agency and stewardship theory

From a theoretical point of view, the necessity to implement management control systems stems from
the fact that decision rights are usually widespread in organizations (Foss/Laursen 2005; Fama/Jensen
1983). According to the economic perspective of agency theory (Arrow 1985; Holmstrom 1979;
Jensen/Meckling 1976; Ross 1973), management control systems provide an instrument to align
diverging interests in delegation relationships between superiors and subordinates (Chenhall/Langfield-
Smith 2003; Falkenberg/Herremans 1995) and are able to ‘‘accommodate many alternative behavioral
or economic factors’’ (Lambert 2007, p. 264). Agency theory therefore provides a rather general
theoretical mechanism to understand the use of results controls, i.e., the use of performance measures
combined with rewards or sanctions as a part of management control systems. Specifically, agency
theory posits that these delegation relationships are characterized by information asymmetries and
diverging objectives (see, e.g., Baiman 1982, 1990; Eisenhardt 1985, 1989; for a detailed review).
Furthermore, it assumes that subordinates are fully rational, risk-averse, and opportunistically striving
for the realization of individual self-interests taking their potential costs and benefits into account
(Merchant et al. 2003). Institutional mechanisms are therefore required to align conflicting objectives
between superiors and subordinates whereas the choice among different control alternatives is
determined by the associated costs of measuring behaviors vs. measuring outcomes (Eisenhardt 1985;
Jensen/Meckling 1976). However, recommendations derived by agency theory primarily affect the
design of incentive compensation contracts or the specification of detailed operating procedures as
means of behavioral control and they mainly refer to the provision of purely extrinsic rewards

8
(Merchant/Otley 2007; Merchant et al. 2003; Eisenhardt 1985). Despite delivering important insights
with respect to the design of these control components, agency theory has been criticized for relying on
rather simplistic behavioral assumptions and for ignoring other elements of management control
systems. For example, Eisenhardt (1989, p. 71) stresses that agency theory ‘‘presents a partial view of
the world that, although it is valid, also ignores a good bit of complexity in organizations’’. Similarly,
Otley (1999, p. 363) emphasizes that the ‘‘discipline of economics does not provide a sufficiently rich
picture of the internal activities of organizations to provide reliable guidance to the designers of
management control systems’’. Consequently, conclusions from purely economics-based models need
to be interpreted cautiously (Frederickson 1992). It is therefore important to take complementary
behavioral theories into account in order to examine the effects of alternative management control
mechanisms. This integration of economic and behavioral theories is also an essential prerequisite to
understand complex organizational phenomena and to enrich theory development (Birnberg et al. 2007;
Atkinson et al. 1997).
Recently, stewardship theory, which builds on insights from sociology and psychology, has been
proposed as a complementary theory to assess the effects of different management controls (Hernandez
2012; Grundei 2008; Donaldson 2008; Tosi et al. 2003; Davis et al. 1997). Stewardship theory is based
on the assumption that ‘‘organizational participants are intrinsically motivated to achieve their tasks in
a pro-organizational manner’’ (Grundei 2008, p. 148). In contrast to agency theory, it presumes that at
least some employees show collectivistic and organization- centered behaviors instead of purely self-
serving motives (Merchant et al. 2003; Davis et al. 1997). These behaviors are considered rational since
they are expected to be associated with greater utility than the realization of individual self-interests,
i.e., employees expect to benefit most from the realization of collective organizational objectives
(Martynov 2009; Hernandez 2008; Grundey 2008; Davis et al. 1997). The underlying assumptions of
stewardship theory therefore correspond strongly to the foundations of ‘Theory Y’ which posits that
employees will act in an organization’s interest if the specific work context is designed properly (Tosi
et al. 2003; McGregor 1960). Stewardship behaviors are, however, not fostered by formal management
control mechanisms but rather facilitated through informal control mechanisms that strongly affect
employees’ intrinsic motivation (Hernandez 2008; Tosi et al. 2003). For example, desired employee
attitudes might be enhanced through opportunities for personal growth and achievement or greater levels
of autonomy (Hernandez 2012; Hernandez 2008; Davis et al. 1997; Donaldson and Davis 1991).
Accordingly, stewardship behaviors can be promoted by organizational designs that empower
employees instead of relying on strict monitoring (Merchant et al. 2003; Davis et al. 1997). Since both
agency theory and stewardship theory provide complementary explanations for the effects of different

9
control instruments, these two theories will be used as our main theoretical basis for the subsequent
hypotheses development.

10
3. Hypotheses development and research model

Our first set of hypotheses relates to the relative effects of formal mechanisms, i.e., results and action
controls, on management control system effectiveness and the level of organizational commitment.
Results controls constitute important and widespread control mechanisms that ‘‘form the backbone of
the management control systems of many organizations’’ (Otley 2006, p. 303). These control
mechanisms can be used to define expectations, to monitor goal-attainment, and to provide feedback on
employees performance (Van der Stede et al. 2006). In addition, they often form the basis for
performance-contingent extrinsic rewards that might affect employees’ motivation (Kominis and
Emmanuel 2007). Hence, results controls can particularly alleviate control problems caused by a lack
of direction or a lack of motivation (Merchant and Van der Stede 2012). Similarly, action controls like,
e.g., preaction reviews or standard operating procedures can reduce a lack of direction by specifying
desirable behaviors (Merchant/Van der Stede 2012). In addition, they can also mitigate potential
negative effects caused by a lack of motivation or personal limitations through the recourse to superiors’
expertise in regular interactions (Merchant and Van der Stede 2012).
According to theoretical propositions of agency theory, both control mechanisms can contribute
to an increased alignment of individual and organizational interests in an incomplete information setting
(Gjesdal 1982; Holmstrom 1979), whereas the choice between a primary emphasis on results or action
controls depends to a large degree on their relative costs and the level of measurement uncertainty
(Falkenberg and Herremans 1995; Anderson and Oliver 1987; Eisenhardt 1985). The costs of particular
control mechanisms can be assessed in different ways and it might be helpful to integrate suggestions
of organizational theory to specify control-related costs (Eisenhardt 1985). For example, results controls
are particularly valuable if the desired results can be measured precisely and if the standards of desirable
performance can be clearly defined (Snell 1992; Eisenhardt 1985; Ouchi 1977). This necessitates that
results controls ‘‘must be stable, with low noise and variation’’ (Widener 2007, p. 764), i.e., low
measurement uncertainty. In such situations, measurement costs are rather low. However, even in
situations that are characterized by high levels of uncertainty, more instead of less information is
frequently needed, i.e., results controls are still supposed to be important and to provide relevant
information (Kruis and Widener 2009; Galbraith 1973). Ouchi and Maguire (1975, p. 568) therefore
conclude that ‘‘paradoxically, output measures are used most when they are least appropriate’’ which
might be caused by a demand for quantifiable measures. On the other hand, organizations can increase
their reliance on action controls if the availability of adequate results measures is restricted but sufficient
knowledge of cause-effect relationships available. In contrast to organizational theory, agency theory
suggests that action controls can also be employed if specific cause-effect knowledge is imperfect.

11
However, additional costs may arise in these situations, e.g., due to the need for increased personal
observations, evaluative meetings, or additional layers of management (Kirsch 1996;
Govindarajan/Fischer 1990). While action controls might therefore not be well suited to guide non-
routine tasks, they nevertheless provide structure for routine tasks that can relatively easy be specified
(Merchant 1985). A non-contingent and positive direct effect has also been found in several empirical
studies (e.g., Kihn 2007; Cardinal 2001; Snell/Youndt 1995) which might provide a rationale for the
common use of action controls in many organizations (Merchant/Van der Stede 2012). In line with
theoretical propositions of the object-of-control framework (Merchant/Van der Stede 2012; Merchant
1985), we thus state our first hypotheses as follows:

H1a Higher levels of results controls lead to an increased effectiveness of management control systems.

H2a Higher levels of action controls lead to an increased effectiveness of management control systems.

Nevertheless, these formal control mechanisms might be associated with quite different effects
on the overall level of organizational commitment which is usually not explicitly considered in purely
accountability-oriented formal control systems (Merchant/Otley 2007; Davis et al. 1997). Several
scholars have thus stressed that results and action controls only lead to behavioral or temporary
compliance at best, i.e., behavior in conformance with mandated activities and performance levels
instead of true goal congruence which is reflected in high levels of organizational commitment
(Bouillon et al. 2006; Pfeffer 1998; Kohn 1993; Ouchi 1979). Even worse, these controls might even be
counterproductive and suppress employees’ commitment as a result of restricted levels of autonomy
and mutual trust (Donaldson 2008; Walton 1985; Ouchi 1979). Following this line of reasoning we
hypothesize that:

H1b Higher levels of results controls lead to a decrease of organizational commitment.

H2b Higher levels of action controls lead to a decrease of organizational commitment.

In contrast to recommendations of agency theory, stewardship theory underlines the importance


of more informal control mechanisms to establish effective control. Thus, particularly personnel and
cultural controls can be employed to foster employees’ intrinsic motivation and to ensure appropriate
behaviors (Merchant and Otley 2007; Merchant 1985). According to Merchant and Van der Stede
(2012), both control mechanisms are capable of mitigating control problems in organizations. Most
notably, they are able to reduce problems caused by a lack of direction or personal limitations. For
example, through well-developed staffing and training practices, personnel controls can help to assure
that employees are appropriately qualified and possess the necessary knowledge for a particular job

12
(Campbell 2012; Snell 1992). Similarly, cultural controls can provide guidance by stressing the
importance of organizational norms and values and by indirectly communicating expected behaviors
that might contribute to more effective management control. Both control mechanisms are furthermore
rather unobtrusive and thus associated with fewer negative side effects compared to results and action
controls (Merchant and Van der Stede 2012; Free/ Macintosh 2009). In addition, they are characterized
by comparatively low costs and they can be applied irrespective of particular task characteristics
(Merchant and Van der Stede 2012; Merchant 1985). This leads to our following hypotheses:

H3a Higher levels of personnel controls lead to an increased effectiveness of management control
systems.

H4a Higher levels of cultural controls lead to an increased effectiveness of management control systems.

Besides their expected positive effect on management control system effectiveness, personnel
and cultural controls can also contribute to higher levels of organizational commitment by evoking
stewardship behaviors among employees. In contrast to formal controls, these control mechanisms
foster individual autonomy and create an environment that is based on a shared understanding of the
importance of organizational objectives. This shared understanding constitutes an important factor that
influences employees’ commitment by reinforcing a sense of organizational purpose (Hernandez 2008).
It might therefore be reasonable to expect that personnel and cultural controls lead to ‘‘high commitment
as a result of internalized values’’ (Ouchi 1979, p. 841). Based on this reasoning we hypothesize that:

H3b Higher levels of personnel controls lead to an increase of organizational commitment.

H4b Higher levels of cultural controls lead to an increase of organizational commitment.

Finally, extant literature suggests that higher levels of management control system effectiveness
and organizational commitment also increase organizational performance. We thus follow recent studies
and assume an indirect rather than a direct effect of management control mechanisms on organizational
performance for which there is insufficient prior theoretical and empirical evidence (Widener 2007;
Henri 2006; Bisbe/Otley 2004). According to Merchant/Van der Stede (2012), effective management
control systems can prevent common control problems and undesired outcomes like, e.g., losses due to
excessive costs or inadequate decision- making processes that may negatively affect organizational
performance. Hence, more effective management control systems ‘‘increase the probability that the
organization will achieve its goals’’ (Merchant/Van der Stede 2012, p. 6). In addition, higher levels of
organizational commitment can also lead to superior organizational performance. While previous
studies have primarily identified effects of individual commitment on managerial performance (e.g.,

13
Nouri/Parker 1998), recent research suggests that this relationship holds also on organizational levels
(Conway/Briner 2012; Winkler et al. 2012; Allen/Grisaffe 2001). Employees who identify strongly with
their organizations and show high levels of commitment are therefore more likely to contribute to
organizational objectives (Hernandez 2008; Davis et al. 1997). Consequently, we specify our final
hypotheses as follows:
H5 Higher levels of management control system effectiveness lead to an increase of organizational
performance.
H6 Higher levels of organizational commitment lead to an increase of organizational performance.

Our theoretical research model, which is based on the hypotheses derived above, is outlined in Fig. 1.

14
4. Research method
4.1 Sample selection and survey design

The data for our study was collected by means of a questionnaire-based (online) survey from Sep. 2011
to Dec. 2011 as part of a larger research project. Potential participants were selected from a non-public
database that included 2273 large- and medium-sized companies.1 All organizations are located in
Germany which limits potential biases caused by cultural differences (Hartmann 2005). In line with
standard practice in empirical management accounting research, financial institutions, insurance
companies, and real estate organizations were excluded due to their specific business models and
divergent regulatory environments. Furthermore, we had to exclude several other companies for various
reasons (e.g., a lack of a separate management accounting department, double counts due to changed
legal forms, or recent insolvencies) which resulted in a final sample of 1757 organizations.
Senior management accountants (heads of management accounting departments) of these
organizations constituted the target population of our study because they are expected to possess detailed
knowledge about their company’s management control system (Auzair and Langfield-Smith 2005;
Shields and Young 1993). With respect to the design and the conceptualization of the questionnaire,
recommendations from Dillman et al. (2009) and Van der Stede et al. (2005) were closely followed.
The questionnaire was also extensively pre-tested and subsequently discussed with several academics
and management accountants in order to assess its layout, clarity, and understandability as well as the
overall content validity of the survey questions. Moreover, a translation/retranslation procedure was
conducted to ensure inter- language validity since the survey was conducted in German language
whereas most of the survey questions were derived from English-language publications (Hartmann
2005; Daniel and Reitsperger 1991).
In total, 311 management accountants participated in our survey which results in a very
satisfactory response rate of 17.70 %. However, responses from 12 organizations had to be discarded
due to missing values. Additionally, we had to exclude 4 small companies with less than 50 employees
to ensure a reasonable size of the participating organizations and the existence of formal management
control mechanisms (Van der Stede et al. 2006). These adjustments resulted in a final sample of 295
organizations. Our final sample contains large- and medium-sized companies in terms of revenues
(mean = 2315 million EUR; median = 600 million EUR) and in terms of the number of employees
(mean = 8786; median = 1900). The sample furthermore comprises organizations from a variety of
different sectors with manufacturing/engineering (12.88 %), chemicals/health care (12.54 %), consumer
goods (11.53 %), and wholesale/retail (11.53 %) being the predominant industries (Table 1).

15
We also checked for potential sample biases. First, we conducted a non-response bias test by
splitting the sample in four equal groups and comparing the (latent) variable scores between early (first
quarter) and late (fourth quarter) respondents by means of non-parametric Mann-Whitney-U-tests. The
rationale of this approach is that late respondents are expected to be similar to non-respondents that
chose not to participate in the survey (Van der Stede et al. 2005; Armstrong and Overton 1977). We
found no statistically significant differences (p\0.05) between early and late respondents which supports
the representativeness of our sample. Second, we assessed the existence of a common method bias using
Harman’s single factor test (Harman 1967). The results yielded no single or general factor that accounts
for the majority of covariance among the individual indicators. This indicates the non- existence of a
substantial common method bias that can be caused, e.g., by common rater effects, consistency motives,
or social desirability issues (Podsakoff et al. 2003; Podsakoff and Organ 1986). Taken together, the
absence of potential biases supports the suitability of our sample for further analyses.

4.2. Variable measurement

Most variables of our research model were measured as latent variables with multiple indicators.
Particular attention was paid to the conceptual specification and the selection of appropriate variables
from existing literature (Bisbe et al. 2007). For example, only those measurement instruments were
considered acceptable that have consistently shown sufficient levels of reliability and validity. In some
cases, however, the wording of existing scales had to be slightly adjusted in order to better match the
underlying conceptual typology as well as the organizational-level perspective of our study and to
consider comments and suggestions raised by pre-testers. In general, we applied six-point rating scales
ranging from ‘does not apply at all’ to ‘does completely apply’ as anchor labels (divergent scale labels
of main constructs are indicated below). With respect to to the even number of points, we did not provide
an ‘in between’- category to reduce acquiescence behavior and to increase operational validity as such
a category is oftentimes misinterpreted by survey respondents as ‘no opinion’ or ‘no answer’ which
impairs data quality (Bartram 2007).
The design of management control mechanisms was assessed by the following constructs.
Results controls were measured by relying on a construct that was originally developed by
Jaworski/MacInnes (1989) to evaluate the extent of ‘output controls’ and that was subsequently used in
a variety of empirical studies (e.g., Jaworski et al. 1993; Cravens et al. 2004; Hutschenreuther 2009).
Overall, this measure reflects the definition and subsequent evaluation of employees’ performance
goals. Similarly, the conceptualization of action controls builds on constructs to measure ‘behavior
control’ used by Jaworski/MacInnes (1989), Jaworski et al. (1993), and Hutzschenreuter (2009).

16
However, we added an additional indicator from Kren/ Kerr (1993) to quantify the reliance on corporate
policies and procedures manuals in order to create a more comprehensive measure that is in line with
theoretical underpinnings of the object-of-control framework (Merchant/Van der Stede 2012; Merchant
1985). The extent of personnel controls was assessed by a measurement instrument from
Hutzschenreuter (2009) that builds on the conceptualization of ‘input control’ from Snell (1992). In
addition, we included an indicator used by Wargitsch (2010) to further evaluate specific objectives of
employee selection processes. Similar items were also used by Widener (2004), although she focuses
solely on staffing procedures and pays less attention to personal development activities. Thus, we
decided to use a more balanced conceptualization of personnel controls that corresponds more strongly
to our underlying conceptual typology. Finally, cultural controls were operationalized by combining
two indicators from Wargitsch (2010), which build on Ouchi’s (1979) ‘clan control’ definition, and four
additional indicators from a construct used by Widener (2007) to assess an organization’s ‘belief
system’. This measurement instrument thus captures the scope of shared norms, beliefs, and values that
can influence employees’ behavior.
With respect to the effects of management control mechanisms, we measured management
control system effectiveness by relying on constructs used by Kruis (2008), Kruis/Widener (2009), and
Ferreira/Otley (2010). We adapted and refined ideas from these authors to develop a construct that is
closely related to the theoretical propositions of the object-of-control framework (Merchant/Van der
Stede 2012; Merchant 1985). Management control system effectiveness is measured in terms of a
control system’s capability to address the three primary control problems in organizations, i.e., lack of
direction, lack of motivation, and personal limitations. This measure thus focuses on the attainment of
overall control goals (Kruis 2008). In contrast, the level of collective organizational commitment was
assessed by relying on a construct used by Johnson et al. (2002) and Hutzschenreuter (2009) to evaluate
the perceived scope of employees’ commitment and identification with a company from an
organizational-level perspective. This construct essentially constitutes a modified version of a
measurement instrument proposed by Mowday et al. (1979) which Meyer et al. (1989, p. 152) denote
as the ‘‘most widely used measure of affective commitment’’. To measure organizational performance,
we followed Henri’s (2006) conceptualization and added an additional indicator that quantifies a
company’s market share in order to create a more comprehensive reflection of performance that equally
captures financial and product market performance. Consistent with previous studies (e.g., Grafton et
al. 2010), organizational performance was not assessed as an absolute but as a relative measure that
compares key performance dimensions over the last year to those of relevant competitors using ‘much
worse’ and ‘much better’ as anchor labels.

17
In addition, we included control variables that have been identified as particularly influential with
respect to their contextual effects on organizational outcomes. First, we controlled for organizational
size in order to account for effects caused by increasing complexity within organizations (Chenhall
2003). We measured organizational size using the natural logarithm of the number of employees
(Libby/Waterhouse 1996; Snell/Youndt 1995; Snell 1992). Second, we included environmental
uncertainty as a contextual variable to control for effects caused by dynamic environments.
Environmental uncertainty was measured by an adapted three-item scale from Hartmann et al. (2010)
which builds on conceptualizations from Govindarajan (1984) and Merchant (1990).
All variables were measured on an organizational level, which is also why we surveyed only one
respondent per organization. Our level of analysis are therefore not specific types of controls that may
be chosen under given contingencies, but rather the average universe of controls used in organizations
and their general impact. In this respect, our research is also in line with Brown et al. (2008, p. 1) who
state that the ‘‘impact control elements have on performance may be more dependent on the existing
elements of the control package than on traditional contingency factors’’.

4.3 Data analysis technique

To analyze our research model, we employ the partial least squares (PLS) approach that constitutes a
variance-based structural equation modeling technique (Tenenhaus et al. 2005; Chin 1998; Wold 1982,
1985). Specifically, we use the software SmartPLS 2.0 M3 (Ringle et al. 2005). The use of structural
equation modeling as a ‘second generation of multivariate analysis’ (Fornell 1987) has been
increasingly proposed to overcome limitations of more traditional statistical analysis techniques due to
its advanced features (Smith/Langfield-Smith 2004; Baines/Langfield-Smith 2003; Chin/Newsted
1999). However, the application of structural equation modeling in management accounting studies has
been fairly modest (Smith/ Langfield-Smith 2004). Consequently, several scholars have called for a
more intense use of structural equation modeling in empirical management accounting research
(Smith/Langfield-Smith 2004; Chenhall 2003; Shields/Shields 1998).

The PLS algorithm simultaneously calculates several measurement models, i.e., the relationships
between indicators and their respective latent constructs, as well as a structural model which quantifies
the relationship between these constructs (Hulland 1999; Chin 1998). The objective of this modeling
approach is to maximize the explained variance of dependent constructs (Hair et al. 2011). In contrast
to covariance-based structural equation modeling techniques (Jo ̈reskog 1973; Jo ̈reskog 1970), the

18
variance-based PLS approach is associated with several distinctive characteristics that support its
appropriateness for our study. First of all, the PLS technique constitutes a non-parametric approach and
does therefore not rely on normally distributed data (Henseler et al. 2009; Chin/Newsted 1999; Chin
1998). In addition, it is also less restrictive in terms of necessary sample size requirements and it allows
for the analysis of rather complex research models (Hair et al. 2011). Finally, the PLS technique is also
generally preferred in research areas with less comprehensive theoretical foundations (Henseler et al.
2009; Reinartz et al. 2009) which is a pervasive characteristic of current management control systems
research (Malmi/Brown 2008; Sandelin 2008).
While the PLS approach has been extensively used in related disciplines like, e.g., marketing
(Hair et al. 2012b), management information systems (Ringle et al. 2012), or strategic management
(Hair et al. 2012a), it is also becoming increasingly popular in top-rated management accounting
journals (e.g., Hartmann/Slapnicˇar 2012; Chenhall et al. 2011; Elbashir et al. 2011; Hall 2011;
Hartmann et al. 2010; Abernethy et al. 2010). Conditional on the specific research context, the PLS
technique might therefore be more appropriate than conventional covariance-based structural equation
modeling approaches (Hair et al. 2011). Smith/Langfield-Smith (2004, p. 76) even emphasize that the
PLS approach is ‘‘tailor-made for management accounting research’’.
The evaluation and interpretation of our results in the subsequent section follows the two-stage
approach suggested by Hulland (1999). First, we assess the reliability and validity of our measurement
models. Second, we complement these initial analyses by an evaluation of the structural model. This
sequence ensures that the underlying constructs are valid and reliable before examining and interpreting
specific relationships among constructs.

19
5. Results

5.1 Results of the measurement models

We assess the appropriateness of our measurement models in terms of their individual indicator
reliability, internal consistency reliability, convergent validity, and discriminant validity (Hair et al.
2011, 2012b; Henseler et al. 2009).
Individual indicator reliability is evaluated based on the standardized factor loadings of
individual indicators on their respective constructs. Generally, standardized loadings should at least be
0.7 which implies that more than 50 % of an indicators variance is explained by the assigned construct
(Henseler et al. 2009; Hulland 1999). However, lower loadings are still acceptable if other indicators of
a construct show sufficient loadings (Chin 1998). Hence, while indicators with factor loadings below
0.4 should be eliminated from the measurement model, indicators that load below 0.7 do not need to be
removed mandatorily, except in cases where an elimination leads to an increase of internal consistency
measures above suggested critical levels (Hair et al. 2011; Hulland 1999). Almost all individual
indicators of our measurement models load above 0.7 on their respective constructs. Only three
indicators show loadings marginally below 0.7 but still above 0.6 which deems to be acceptable since
they belong to different constructs and other loadings of these constructs exhibit loadings well above
the recommended threshold (Barclay et al. 1995).
Internal consistency reliability is assessed by a construct’s Cronbach’s alpha (CA) and Dillon-
Goldstein’s rho as a measure of composite reliability (CR). In contrast to Cronbach’s alpha, Dillon-
Goldstein’s rho does not assume tau- equivalence, i.e., equal factor loadings of individual indicators,
which can prevent an underestimation of internal consistency reliability (Henseler et al. 2009; Werts et
al. 1974). Several scholars therefore recommend to present CA values only for informational purposes
and to use CR to evaluate individual constructs (e.g., Tenenhaus et al. 2005; Chin 1998). Proposed
minimum requirements range from 0.7 to 0.5 for CA (Nunnally 1967) and from 0.7 to 0.6 for CR
(Tenenhaus et al. 2005; Bagozzi/Yi 1988). Table 2 shows that all constructs used in our study display
sufficient levels of internal consistency reliability.
Convergent validity is analyzed by the average variance extracted (AVE) of constructs. AVE
values describe the average variance shared between a construct and its associated indicators
(Fornell/Larcker 1981). Generally, AVE values should exceed a level of 0.5 which implies that a
construct reflects more than half of its indicators variance and, thus, the absence of significant
measurement errors (Bagozzi/Yi 1988; Vandenbosch 1996). As can be seen in Table 2, all constructs
of our research model exhibit high levels of convergent validity.

20
Discriminant validity, i.e., the extent to which a particular construct conceptually differs from
other constructs, is finally warranted if the square root of a construct’s AVE exceeds the correlations
with other constructs of the research model. This is also referred to as the Fornell/Larcker-Criterion
(Fornell/Larcker 1981). In addition, we analyzed the cross-loadings of individual indicators which
should be highest with their associated constructs (Chin 1998). Tables 3 and 4 show that sufficient
discriminant validity of our measures can be confirmed as well.
Overall, the results of the measurement models indicate that all constructs of our research model
are characterized by sufficient levels of reliability and validity. This supports their suitability for the
analysis of structural relationships among constructs.

5.2 Results of the structural model

In order to assess the results of the structural model, we rely on interpretations of path coefficients,
multiple squared correlations (R2), effect sizes (f2), and the predictive relevance (Q2). Since the PLS
approach does not provide conventional goodness-of-fit indices like covariance-based structural
equation modeling techniques, sufficient evidence of model fit is warranted by reliable and valid
measurement instruments, an appropriate level of variance explained, and positive predictive relevance
(Hartmann/Slapnicˇar 2012; Tenenhaus et al. 2005; Vanden- bosch 1996). The results of the structural
model are illustrated in Fig. 2 and Table 5.

The path coefficients of the structural model can be interpreted like standardized beta coefficients
of ordinary least squares (OLS) regressions (Henseler et al. 2009). However, unlike traditional
parametric OLS regressions, standard errors and significance levels of path coefficients (t-values) are
calculated by means of non- parametric resampling procedures like, e.g., bootstrapping or jack-knifing
(Hair et al. 2011; Chin 1998). In line with Chin (1998) and recent studies in management accounting
research (e.g., Chenhall et al. 2011; Hartmann et al. 2010; Hall 2008), we apply bootstrapping using
500 samples with replacement to evaluate the statistical significance of our results.

Our hypotheses H1a–H4a suggested a positive effect of different management control mechanisms
on management control system effectiveness. The empirical results confirm our hypotheses for action
controls (H2a: b = 0.281; t = 4.809; p\0.01), personnel controls (H3a: b = 0.306; t = 4.844; p\0.01), and
cultural controls (H4a: b = 0.195; t = 2.920; p\0.01). In line with our theoretical propositions, these
management control instruments thus influence management control system effectiveness by
successfully addressing common control problems in organizations. However, despite a positive path

21
coefficient, a significant effect of results controls cannot be confirmed (H1a: b = 0.030; t = 0.875;
p[0.10) which implies that results controls do not seem to affect management control system
effectiveness when controlling for other control mechanisms.

With respect to the effects of control mechanisms on the level of organizational commitment (H1b–
H4b), we find that personnel controls (H3b: b = 0.234; t = 3.229; p\0.01) and cultural controls (H4b: b
= 0.456; t = 7.065; p\0.01) are significantly associated with higher levels of organizational commitment.
Although both formal control mechanisms show a negative path coefficient, results controls (H1b: b =
-0.002; t = 0.056; p[0.10) and action controls (H2b: b = -0.063; t = 1.184; p[0.10) are not significantly
related to lower levels of organizational commitment. Thus, they do not seem to be characterized by
potential detrimental effects that might negatively influence employees’ commitment.

Furthermore, we suggested positive effects of management control system effectiveness and


organizational commitment on organizational performance. The results corroborate our hypotheses and
show that both management control system effectiveness (H5: b = 0.195; t = 2.768; p\0.01) and
organizational commitment (H6: b = 0.113; t = 1.711; p\0.10) are significantly related to organizational
performance. An analysis of the total effects furthermore indicates that, in contrast to formal control
mechanisms, more informal control mechanisms seem to constitute the most important drivers of overall
organizational performance which supports their fundamental importance.
In order to validate the robustness of our findings, we additionally test for potential problems that
might be caused by high levels multicollinearity. The variance inflation factors (VIFs) of all constructs
are less than 2.0 and thus well below commonly suggested critical thresholds (Hair et al. 2011; Widener
2007). This indicates that multicollinearity is not an issue that affects the interpretation of path
coefficients.
Multiple squared correlations (R2) are subsequently utilized to assess the explained variance of
our dependent variables. Although some scholars have proposed thresholds to interpret the extent of
variance explained (e.g., Chin 1998), there exists no general convention on how to evaluate different
values because R2 values are basically dependent on the research design and the characteristics of the
available data (Hair et al. 2011; Jain 1994). The interpretation of R2 values thus has to be conducted in
the context of the underlying research question. An analysis of R2 values shows that the different control
mechanisms explain management control system effectiveness (R2 = 45.73 %) and organizational
commitment (R2 = 40.68 %) very well. These two constructs subsequently explain a moderate amount
of variance of organizational performance (R2 = 9.46 %). This value can be considered satisfactory since
an organization’s performance generally depends on many additional factors that are outside the scope

22
of our research design.
The analysis of path coefficients and multiple squared correlations is complemented by an
assessment of effect sizes (f2) that are often not explicitly examined and reported in empirical
management accounting studies (Lee et al. 2011). A construct’s effect size describes the increase of a
dependent variables R2 value relative to the amount of unexplained variance (Henseler et al. 2009).
Cohen (1988) suggests labeling effect sizes of 0.02, 0.15, and 0.35 as small, medium, and large.
Nevertheless, it has to be taken into account that this interpretation is dependent on the overall level of
respective R2 values. If R2 values are rather small, effect sizes generally tend to be small as well. Similar
to the interpretation of R2 values, the magnitude of effect sizes therefore needs to be interpreted in
context. Table 5 indicates that most effect sizes that correspond to significant path coefficients can be
considered as moderate. Overall, these results are comparable to other studies that reported effect sizes
of different control mechanisms (e.g., Hutzschenreuter 2009).
Finally, the predictive relevance (Q2) of our research model is analyzed by means of the Stone-
Geisser-test that builds on cross-validated redundancy indices (Geisser 1974; Stone 1974). Q2 values
are calculated by a blindfolding procedure that excludes different parts of the underlying data and uses
the remaining data to reconstruct the excluded parts (Hair et al. 2011; Tenenhaus et al. 2005). Generally,
a structural model exhibits sufficient predictive relevance if Q2 values of dependent constructs are larger
than zero (Henseler et al. 2009; Chin 1998). As can be seen in Table 5, all Q2 values are well above zero
which confirms the predictive relevance of our model.

23
6. Discussion and conclusions

Our study provides empirical evidence of the relative effects of alternative management control
mechanisms. We will discuss our results in more detail below. Furthermore, we will outline
contributions to theory and practice as well as potential limitations of our research design.
While it is generally accepted that organizations use a variety of alternative control mechanisms
to guide and influence their employees’ behavior, the relative effects of these different instruments are
rather unclear because they have traditionally been examined only on a separate basis. Our analysis
indicates that on an organizational level particularly informal management control mechanisms, i.e.,
personnel and cultural controls, contribute strongly to both an increased effectiveness of management
control systems and higher levels of organizational commitment which, in turn, lead to superior
organizational performance. However, we find that the effect of organizational commitment on
organizational performance is not as strong as the effect of management control system effectiveness.
A reason for this weaker effect might be the existence of additional indirect relations between
commitment and organizational performance, e.g., mediating effects of different psychological factors
(Klein et al. 2012; Hernandez 2012).
Nevertheless, we can generally support the beneficial effects of an increased reliance on informal
controls in contemporary organizations. Our results are therefore in line with recent studies that
similarly emphasize the importance of more informal management control mechanisms (e.g., Chenhall
et al. 2010, 2011; Hutzschenreuter 2009; Busco et al. 2008; Sandelin 2008). With respect to formal
means of control, we can only corroborate the hypothesis that action controls are positively associated
with higher levels of management control system effectiveness which corresponds well with the results
of previous studies that have identified positive effects of these control mechanisms (e.g., Kihn 2007;
Cardinal 2001; Snell/ Youndt 1995).
Interestingly, our results show that formal results controls are not associated with an increased
effectiveness of management control systems which illustrates that these instruments might be overrated
in terms of their potential benefits. These results—albeit counter-intuitive at first sight—are consistent
with recent studies in empirical management accounting research that were also unable to confirm a
positive effect of results controls on organizational outcomes (e.g., Wargitsch 2010; Hutzschenreuter
2009; Snell and Youndt 1995). Although we find a positive and significant bivariate correlation between
results controls and management control system effectiveness, this relationship is not statistically
significant when including additional management control mechanisms in our structural model.
Similarly, Widener (2007) and Henri (2006) also found that certain management control elements show
significant bivariate correlations that became non-significant when controlling for alternative control

24
mechanisms.
A possible reason for this might be that there exists only a very weak relationship between formal
control mechanisms and the respective organizational outcome variables after a given minimum level
of formal controls has been established. If we assume that all surveyed firms have implemented this
minimum level, our results might simply suggest that increasing the level of results controls beyond that
threshold will only have a very weak impact on management control system effectiveness as well.
Another reason might be that different ways of implementing results controls in firms have strongly
diverging positive and negative effects which are not captured by measuring the level of results control
usage. Our results therefore do not support any simplistic theories assuming that increasing the level of
formal control mechanisms will lead to better organizational performance in a sense of ‘the more, the
better’. Instead, our results underline the importance of analyzing effects of different control
mechanisms simultaneously because an individual analysis of results controls would have documented
a positive effect. By applying a more comprehensive approach, we can therefore conclude that
alternative management controls seem to be much more important than results controls. This reasoning
is also in line with research on the relations of extrinsic and other forms of motivations (e.g., Gneezy et
al. 2011; Rost and Osterloh 2009).
Our study contributes to management control literature in several ways. In general, it addresses
the limited body of knowledge regarding different management control mechanisms in contemporary
organizations and their relative effects on organizational-level outcomes. Our analyses explicitly
respond to calls to study the diverse effects of management control mechanisms and to examine the
relationship between formal and informal controls by adopting a more holistic perspective (e.g., Malmi
and Brown 2008; Bisbe and Otley 2004). Thus, our approach goes beyond previous studies that have
mostly focused only on different formal management control techniques. By applying a theoretically
well-grounded framework that captures an array of different management controls, we contribute to a
growing stream of recent literature that comprehensively analyzes effects of different control system
elements in order to avoid potential biases due to partial and incomplete research models and which has,
by now, rather researched single organizations and/ or rather small samples (e.g., Chenhall et al. 2010,
2011; Mundy 2010; Sandelin 2008). Furthermore, drawing on agency theory and stewardship theory as
theoretical framework, we follow recent suggestions to combine economic and behavioral theories
(Merchant et al. 2003) and confirm the suitability of both theories to explain diverse effects of
management control mechanisms.
Since management control constitutes a major concern for almost all organizations (Berry et al.
2009), our findings also have important implications for business practice. According to Epstein (2002),

25
companies have to be aware of the relevant key drivers of organizational success. Our joint
consideration of formal and informal control mechanisms provides evidence which control system
elements seem to be more important in order to foster desired employee behaviors and sustainable long-
term value creation. Instead of stressing the importance of comprehensive formal management control
techniques, organizations might benefit from shifting their focus to more informal means of control.
Hence, informal control mechanisms should not be regarded as mere premises but rather as important
control system elements that strongly affect employees’ behaviors and subsequent organizational
outcomes (Sandelin 2008).
Similar to most empirical studies, our study is also subject to several limitations that may
stimulate further research. For example, we rely on perceptual data from senior management
accountants because information regarding an organization’s control mechanisms is not publicly
available. Hence, our results might be associated with commonly mentioned shortcomings of
questionnaire-based survey studies like, e.g., individual respondents’ subjectivity and social desirability
issues. Notwithstanding this potential caveat, Venkatraman and Ramanujam (1987) emphasize that self-
ratings might be less biased than commonly expected because they tend to be highly correlated with
objective measures.
Moreover, our structural equation modeling technique is subject to all limitations related to the
choice of a variance-based method of analysis, mainly the lack of overall goodness-of-fit measures
compared to covariance-based methods. Nevertheless, PLS allows for a much more intuitive
interpretation of results and does not rely on normally-distributed data. As it is also increasingly used
in accounting research as well as related fields, e.g., marketing, information systems, or strategic
management, we consider the choice of this method, even in the light of existing limitations, as
appropriate. Furthermore, additional tests with other linear models (e.g., OLS regressions), which were
conducted as robustness checks, did not yield any different inferences from the empirical results.
Another limitation relates to our findings regarding the non-significant effect of results controls
on management control system effectiveness which should be interpreted cautiously for several reasons.
First, results controls constitute an important part of an organization’s management control system and
they are often used to a considerable extent. Their widespread adoption in organizations might therefore
curtail differentiating effects with respect to organizational outcomes. Second, our used measurement
instrument largely corresponds to what has been labeled as a ‘diagnostic’ (Simons 1995) or ‘coercive’
control mechanism (Adler and Borys 1996). However, recent studies showed that results controls can
also be used to encourage innovation or organizational learning which reflects a more ‘interactive’ or
‘enabling’ role of these control instruments that might lead to distinct effects on organizational

26
outcomes (e.g., Widener 2007; Henri 2006). Subsequent studies might therefore not only consider if
results controls are implemented but also how they are applied depending on different context factors.
This also includes identifying and empirically confirming conditions that allow for a more effective
utilization of formal controls as part of a comprehensive management control system.
In addition, the object-of-control framework does not explicitly differentiate between the use of
financial and non-financial results controls. Further research might therefore investigate the role of
differently specified results control measures while controlling for alternative management control
mechanisms. Another issue might be the differing impact of formal vs. informal management control
mechanisms in specific contingencies which has not been addressed by our research design but is a
matter for further analysis. Whereas we assessed an ‘overall control style’ on an organizational level,
control approaches might also differ between particular organizational departments and functions which
represents an area for further research.
Another limitation of our study is the number of control variables. We chose to survey
‘environmental uncertainty’ and ‘size’ as major relevant control variables in management control
systems research from a theoretical perspective. Still, the impact of these control variables is rather
weak. This result is nevertheless in line with, e.g., Chenhall et al. (2011), who use ‘size’ and ‘business’
as control variables which neither have a strong control path in their model. Even though further research
should search for and survey additional control variables and/or explore the effectiveness of
management control instruments in more specific settings, our results do not indicate that relevant
control variables have been neglected in our general setting referring to organizations as a whole.
Moreover, analyzing effects of alternative control mechanisms on an individual level with respect to
specific control situations may provide additional insights. For example, it seems reasonable to expect
that the extent of particular control problems and the effects of control mechanisms depend, among
other aspects, on employees’ individual characteristics, e.g., psychological traits and attitudes,
hierarchical level, gender, age, or other demographic variables.
In this regard, potential endogeneity may be a limitation; not only due to such omitted variables
and their possible impact on population heterogeneity but also due to simultaneity. Taken strictly, our
model hypothesizes causality but our measurement relates only to correlations between variables
(Antonakis et al. 2010; van Lent 2007). For example, even though we assume that control mechanisms
have an impact on organizational performance, there might be a reverse relation as firms with high
organizational performance may simply be able to afford more costly controls than firms with low
organizational performance. Still, both concerns call for different research designs, e.g., in the case of
simultaneity for long-term panel studies or experimental designs.

27
Taken together, although considerable care was devoted to an a priori theoretical specification of
the relationships between our constructs, it is not possible to distinguish cause-and-effect relations
empirically in cross-sectional studies (Widener 2007; Van der Stede et al. 2005).
Despite these potential limitations, we believe that our findings provide important insights for
theory and practice with respect to the design of management control systems in contemporary
organizations. In particular, we hope that our results might contribute to the development of more
comprehensive theories of control that equally consider both formal as well as more informal
mechanisms as fundamental means of control which might avoid spurious findings of individual
analyses.

28

S-ar putea să vă placă și