«Speed, coordination and individualistic behaviors: a pilot NK modeling study to investigate the moderating effects of organizational structure on ...»
Esic Market Economics and Business Journal
Vol. 44, Issue 2, May-August 2013, 109-126
Speed, coordination and individualistic
behaviors: a pilot NK modeling study
to investigate the moderating effects of
organizational structure on performance
in individual firms
Pablo E. Pinto Cornejo*, Jorge G. Montecinos A., Damian Hine,
Peter Knights and Felipe Valdebenito Pedrero
The University of Queensland ** and Universidad Católica del Norte
A central concern within the field of organizational design is the study of the relationship between structure, innovation and performance. The basic understanding is that decentralized structures appear best suited to promoting innovation and change. Yet this comes at the cost of efficiency. Conversely, organizations that centralize power in the hands of a few appear well-suited functionally to achieving high levels of efficiency, but at the cost of generating inertia. Taking both forms as polar types at the opposite ends of a continuum, the managerial question of how much of each form is necessary to create a good configuration to perform satisfactorily in the long term remains unknown. In this pilot study, we attempt to answer this question by using an NK model. Our results show that exploration is facilitated by higher levels of decentralization, where a structure that combines centralization and decentralization features delivers the highest long-term performance. We also find that complex organizational forms achieve low performance. This suggests that problems with coordination are costly to manage. Assumptions of game-theory are introduced to quantify the risks of decentralization.
Keywords: Organizational structure, NK Model, exploration and exploitation, firm performance.
JEL codes: M21, C6, D21.
* Corresponding author. Email: firstname.lastname@example.org ** A member of CLADEA (Latin American Council of Management Schools).
ISSN 0212-1867 / e-ISSN 1989-3558 © ESIC Editorial, ESIC Business & Marketing School DOI: 10.7200/esicm.145.0442.3i http://www.esic.edu/esicmarket 110 Pablo E. Pinto Cornejo et al.
1. Introduction A central concern within the field of organizational design and configuration is the study of the relationship between structure, innovation and organizational performance (Cyert & March, 1963; Lawrence & Lorsch, 1967; Mintzberg, 1979). This concern goes back at least to the work of Burns and Stalker (1961) on mechanistic and organic forms and prior to that with Weber’s (1947) consideration of corporate bureaucracy. A major driver has been the discovery of the most appropriate organizational structure to foster innovation and survival in the long run (Greenwood & Hinings, 1988; Lam, 2005). This has resulted in a number of organizational typologies (Galán et al., 2012) from the multidivisional structure (Chandler, 1962), to the adhocracy (Mintzberg, 1979), to the radical, and very unstable spaghetti organization (Foss, 2003). However, and despite the recently renewed interest in organizational design (Greenwood & Miller, 2010; Schreyögg & Sydow, 2010), the literature is, with few exceptions (see, for example, Siggelkow & Levinthal, 2003), still inconsistent in answering the question of what dimensions of the structure have the most significant impact on both innovation and performance in the long run (Badir et al., 2009).
In this pilot study we seek to expand the categories of analysis to organizational types that have been overlooked in recent simulation studies of organizational design. To address this problem, we employ an experimental design - an NK modeling technique to simulate the behavior of organizations composed of many interacting elements. This technique proposed by Kauffman (1969; 1993) for understanding mutational processes in biological systems has proven more than adequate in other contexts, including in organizational design (Levinthal, 1997; Rivkin & Siggelkow, 2007). We aim to understand how search activity in organizations is best structured to achieve performance improvements, and discuss some benefits and costs linked to mixed forms of organization not considered in previous simulation models. The model is built around two dimensions of organizational structure: the degree of centralization of decision-making and the interdependency among decisions. The degree of centralization reflects the locus of decision-making power and refers to the extent to which authority is distributed among different units in an organization (Miller & Dröge, 1986). The interdependency among decisions reflects the level of complexity of the system (Kauffman, 1993; Simon, 1962) and describes the many parts and processes that a system must coordinate to achieve some measure of overall success (Kauffman & Levin, 1987).
In line with previous studies (cf. Rivkin & Siggelkow, 2007), we simulate two organizational types which have been previously studied: a “Block-Diagonal” structure (Siggelkow & Levinthal, 2003), in which decisions are made independently in two divisions; and a “fully interdependent” structure (Kauffman, 1993), which is modeled using a matrix where all decisions interact with each other in a single firm with no divisions.
We also establish a different set of relationships within a firm structure with two divisions, so as to study two new designs. One termed “semi-decentralized” strucSpeed, coordination and individualistic behaviors: a pilot NK modeling study… 111 ture, in which each division makes proposals to their headquarters which selects the combination of decisions with the highest overall payoff for the firm (Wall, 2010).
The second we refer to as an “individualistic” structure (Press, 2007), in which divisions seek to maximize their own partial fitness (utility), where the winning division forces the other to follow its lead, resulting in a potential win-lose situation, which cannibalizes the competencies of the other division. Game theory assumptions are employed to simulate the individualistic structure. This simulation design, and its concomitant results, provide an extension of previous research on organizational design.
2. Theoretical Foundations
The creation of the new ideas and innovations that support competitiveness and growth (Tripsas, 2009) rely on the ability of an organization to explore and find new opportunities, or high peaks in existing or new landscapes (March, 1991). In accordance with its ecological derivation, a landscape is an “area that is spatially heterogeneous in at least one factor of interest” (Turner et al., 2001, p. 3). The purpose of exploration; “is to find and occupy a high spot on this landscape, i.e., to select a combination of choices [i.e., decisions, activities] that, together, are highly successful.” (Siggelkow & Rivkin, 2005, p. 104).
The achievement of successful exploration however poses an organizational dilemma as managers seek to unleash the power of exploration at the lower levels of an organization in order to promote innovation, while maintaining strategic focus requires managers to preserve organizational unity in decision-making to bring to fruition future developments (Siggelkow & Rivkin, 2006). Failure to strike a balance between these two agendas can lead to incompatible and internally competitive actions (Nickerson & Zenger, 2002) hampering performance (O’Reilly & Tushman, 2007).
Part of this challenge is that organizational choices made to bring efficiency gains can also hinder a firm’s ability to develop new knowledge (Tripsas & Gavetti, 2000).
Prior research on organizational design has shown that several elements found in more mechanistic forms of organization, including hierarchical structures, centralized decision-making and formal controls and communication channels, are likely to enhance operational efficiency but also produce risk aversion as firms avoid uncertain possibilities in favor of actions that have produced positive results in the past (Lam, 2005; Miller et al., 2006). Conversely, more organic structures include several elements that appear to foster creativity, complexity and adaptability, such as decentralized decision making, a lack of formally defined tasks and loose coupling systems, but at the cost of efficiency (Sheremata, 2000). The latter suggests that the coherence of an organizational design “is not accidental” (Greenwood & Hinings, 1988, p. 295); but that “there are several models of organization with differential efficiencies depending on the nature of the work and the types of tasks to be performed” (Litwak, 1961, p. 181).
112 Pablo E. Pinto Cornejo et al.
While formal structure has been measured in a variety of dimensions (Price & Mueller, 1986), there appears to be a consensus that centralization (the distribution of authority within an organization), formalization (the degree of work standardization) and complexity (the degree of specialization, and number of hierarchical levels within an organization) are the basic dimensions of structure (Tsai, 2002; Van de Ven, 1976). Thus, we draw on the work of Van de Ven et al., making centralization and complexity the focal aspects of this paper.
Centralization refers to the concentration of authority or decision-making power (Miller & Dröge, 1986), and points to whether the locus of authority and decisionmaking lies in the higher or lower levels of the organizational hierarchy (Jansen et al., 2006). In research on multidivisional organizations, centralization has focused on the dichotomous relationship between the relative degree of influence or control exercised by the corporate headquarters and the individual organizational units in relevant decision-making processes (Tsai, 2002).
The notion of centralization has long been a consideration in organizational design theorizing, from Weber’s notion of bureaucracy, to the work of the Aston group (Pugh et al., 1969), and to contingency theory (Lawrence & Lorsch, 1967).
Most of the work within this tradition argues that organizations must centralize to attain superior efficiency (Adler & Borys, 1996). Centralized power in the hands of a few reduces diversity in decision-making (Siggelkow & Rivkin, 2005), thereby increasing speed, and control (Sheremata, 2000).
Centralization may however constrain a firm’s ability to experiment as there is less latitude for new strategic choices (Jansen et al., 2006). Centralization may also induce conformity with rules and established routines, making people “less receptive and supportive of ideas that might deviate from the status quo” (Lubatkin et al., 2006, p. 652), whereas the associated vertical decision process may reduce the felt need for interactions to solve problems collaboratively (Miller, 1987). The result may generate structural inertia (Hannan & Freeman, 1984), constraining an organization’s ability to respond to new opportunities, and thereby to change, grow, and compete (Oldham & Cummings, 1996).
Decentralization, on the other hand, appears best suited to igniting creativity (Nonaka & Konno, 1998; Zammuto & O’Connor, 1992), since it eliminates organizational constraints to freedom in the conduct of work (Amabile et al., 1996).
Decentralization is beneficial to promote innovation and thus adaptability to unstable and more turbulent environments (Benner & Tushman, 2003). Yet this comes at the cost of efficiency.
Accordingly, several studies have observed that extreme degrees of decentralization may have negative effects on innovation (Van Looy et al., 2005). Extreme decentralization may lead to chaotic resource allocation (Demsetz, 1988) and the creation Speed, coordination and individualistic behaviors: a pilot NK modeling study… 113 of knowledge in areas so dissonant that top management cannot coordinate these efforts and effectively integrate them with the core business of the firm (Siggelkow & Rivkin, 2006). This can also lead independent divisions to develop products that are both incompatible and internally competitive, and firm performance can suffer as a result (Nickerson & Zenger, 2002).
The dichotomy between centralization and decentralization, however, is an oversimplification of the reality. The challenge remains not that a firm centralize or decentralize per se, but to strike a balance “that permits speedier improvement without completely sacrificing diversity of search” (Siggelkow & Rivkin, 2005, p. 117).
Based largely on this idea, numerous studies have turned to the notion of structural separation of activities or ambidexterity (Jansen et al., 2006). Unfortunately, there remains much confusion and contradiction in the literature, and it is still not clear how organizations can cope with countervailing functions, pattern maintenance, and adaptation (Schreyögg & Sydow, 2010) to successfully achieve ambidexterity (Raisch & Birkinshaw, 2008).
Incorporating complexity into modeling
Complexity refers to the characteristics of a system “of being intricate and compounded” (Olausson & Berggren, 2010, p. 384). A complex system is, according to Simon (1962, p. 468), “made up of a large number of parts that interact in a nonsimple way”. Complex systems are structured in levels or hierarchies, with each level resting on the one below it (Simon, 1995). This property shared by hierarchically organized systems is called nearly complete decomposability NCD (Simon, 2002).
Systems that display NCD properties tend to evolve faster than non-hierarchic systems of comparable size (Simon, 1962).
In this paper, we model designs that are built around the dimensions of centralization and complexity. The organizational structural modeling reflects, but also complements, that of Siggelkow and Levinthal (2003), Rivkin and Siggelkow (2007) and Wall (2010), with each organizational structural form varying according to the degree to which authority is distributed among different divisional units in the decision-making process (degree of centralization) and the level of complexity (interactions among decisions).
3. Research Objectives and Methodology
In this study, we aim to understand how search activity in organizations is best structured to achieve performance improvements, and discuss some benefits and costs linked to mixed forms of organization not considered in previous simulation models. In addressing this problem, we use an NK agent-based single firm level computational model.
114 Pablo E. Pinto Cornejo et al.