The socio-cognitive antecedents of ambidexterity : the role of paradoxical cognition and TMT reflexivity

The Socio-cognitive Antecedents of Ambidexterity

The role of Paradoxical Cognition and TMT Reflexivity

Author: Marlies de Kock

Student number: 10868887

Date August 2015

Qualification: MSc. In Business Administration – Strategy Track

Institution: Amsterdam Business School, University of Amsterdam

2

Statement of Originality

This document is written by Student: Marlies de Kock who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for supervision of completion of the work, nor for the contents.

3

Table of Contents

Abstract 4

Introduction 5

Literature Review and Hypotheses 10

Methodology 25

Analysis and Results 31

Discussion and Conclusion 37

References 48

4

Abstract

Top management teams (TMTs) play a key role in achieving organizational ambidexterity. Yet, it remains unclear how the TMT is able to deal with contradicting forces posed by explorative and exploitative knowledge resources. We explore the mechanism by which TMTs might successfully manage the contradicting pressures they face. This study suggests that paradoxical cognition and TMT reflexivity allow a TMT to reframe tensions and transcend the exploration-exploitation paradox and thereby facilitate organizational ambidexterity. It is furthermore proposed that benchmarking and lower level management team (LLMT) attention moderate the relation between TMT reflexivity and paradoxical cognition. These moderators provide more insight in the role and influence of internal and external information flows on achieving paradoxical cognition. We conducted research among 276 Dutch and German organizations and show that paradoxical cognition indeed fully mediates the relationship between TMT reflexivity and organizational ambidexterity. The moderating effects turn out to be not significant.

Key words: Ambidexterity; Exploration and Exploitation; Paradoxical Cognition, Reflexivity;

5

Introduction

“Paradoxes are paradoxical,- common and surprising, confusing and understandable”

(Cameron & Quinn, 1988, p. 13).

One of the more salient paradoxes in organizational life becomes apparent by the tension between exploration and exploitation (Lewis, 2000; Smith & Lewis, 2011). The contradicting pressures of this paradox arise as it is important for organizations to refine their existing products and procedures, but at the same time they need to invest resources in the development and discovery of new opportunities as well (March, 1991; Jansen, Tempelaar, van den Bosch, & Volberda, 2009; O’Reilly & Tushman, 2008). Organizations need to become ambidextrous, this means that they have to be able to explore and exploit simultaneously (Cao, Gedajlovic, & Zhang, 2009; Jansen et al, 2009; O’Reilly & Tushman, 2008). Achieving ambidexterity is fundamental for organizational survival, as this should enable organizations to continuously adapt to their environment (March, 1991) and prevent them from failure or success traps (Levinthal & March, 1993; Levitt & March, 1988; Ahuja & Lampert, 2001). Moreover, organizations that are able to overcome the polarities (of exploring and exploiting) have been shown to achieve superior organizational performance (Cao et al, 2009; He & Wong, 2004; Lubatkin, Simsek, Ling, & Veiga, 2006).

Top management teams (TMTs) play a key role in achieving organizational ambidexterity, as strategic decisions are generally made at the apex of the organization (Eisenhardt & Zbaracki, 1992). To overcome the exploration-exploitation paradox, managers have to be able to tap into different thought worlds effectively (Raisch, Birkinshaw, Probst, & Tushman, 2009). To be able to do this, managers must delve into the seemingly contradicting and sometimes irrational or even absurd (enacted) coexistence of opposites to be able to reframe their situation and transcend the apparent contradiction (Lewis, 2000). The ability of managers to capture both extremes is associated with the ability of actors to process and make sense of contradicting information flows or cues from the environment (Davis, Maranville, & Obloj, 1997). They make sense of these information flows based on the combined and negotiated perceptions of its members (Davis et al, 1997). This makes the transcendence of paradox a

6 socially constructed process (Van Neerijnen, 2015), that connotes the activities of perceiving and interpreting an organization’s internal and external environment (Davis et al, 1997).

Handling the joint pursuit of exploratory and exploitative orientations poses a great challenge for TMTs. Organizational members are naturally inclined against the exploration-exploitation paradox, as perceived uncertainties (March, 1991) and risks (Kahneman & Tversky, 1979) make them averse to the combination of both activities. TMTs are committed to current mental models due to cognitive and emotional biases, and therefore favor exploitation over exploration (e.g. Levinthal & March, 1993; March, 1991; Smith & Tushman, 2005). This causes actors to reduce and simplify the amount of cognitive information processing (Davis et al, 1997). Following Schippers, Edmondson, and West (2014), we recognize two main sources of information processing failures: on the one hand, there are failures attributed to breakdowns in interpersonal contact in teams, and on the other hand, there are failures related to the shortcomings of the cognitive frame. Both kinds of information processing failures leave actors blind to the exploration-exploitation paradox and complicate the reframing process (Cameron & Quinn, 1988).

Previous academic literature reveals multiple factors that play a supportive role in the TMTs’ ability to make sense of paradox and spearhead organizational ambidexterity. The literature provides organizational factors that could support individuals’ and teams’ abilities to act ambidextrously. According to Ghoshal and Bartlett (1997), socialization, recognition, and team-building practices help individuals to think and act ambidextrously (Raisch et al, 2009). Gibson and Birkinshaw (2004) further note that certain contexts, characterized by a combination of stretch, discipline, support, and trust, could allow managers to divide their time and resources between exploration and exploitation (Raisch et al, 2009). On the TMT level, Lubatkin et al, (2006) focus on TMTs in small to medium sized firms and suggest that a TMTs behavioral integration, including TMT wholeness and unity of effort, helps a TMT to process disparate demands of exploration and exploitation. Jansen et al, (2009) conclude that formal TMT contingency rewards and informal senior team social integration mechanisms are important for enabling TMTs to allow for contradictions. On the individual level, Mom, Van Den Bosch and Volberda (2007) found that managers that acquire high levels of top-down and horizontal knowledge flows, or top-down and bottom-up knowledge flows, may undertake high levels of exploration and exploitation. Moreover, successful individuals’ and TMTs’ are supported by

7 certain personal characteristics. Smith and Tushman (2005) note that individuals’ and TMTs’ ability to engage in paradoxical thinking may be vital for effectively managing contradictions. In addition, Cohen and Levinthal (1990) argue that actors need prior related knowledge of topics under consideration; individuals with a substantial breadth of prior knowledge categories, as well as various linkages across them, may be better able to perform contradicting tasks.

Despite explanations, the current literature still insufficiently explains how organizational exploration, exploitation, and ambidexterity are achieved. Specifically, there is a lack of attention regarding the socio-cognitive antecedents of organizational ambidexterity (Van Neerijnen, 2015). This is problematic as the resolution of the exploration-exploitation paradox poses a socio-cognitive challenge that demands the willful and able contribution of a TMT to transcend paradoxical tensions (Van Neerijnen, 2015). As a result, current scholarly studies lack a profound understanding of how a TMT actually copes with contrasting forces (Van Neerijnen, 2015; O’Reilly & Tushman, 2013; Raisch et al, 2009). With information processing as a central aspect of these activities, it is important to identify factors that influence the effectiveness of the information processing activities in teams (Schippers et al, 2014). However, little is known about the precise nature of TMT practices that affect these information streams, and how this could enhance paradoxical cognition. In particular, the current literature fails to explain how TMT reflexivity leads to paradoxical cognition by countering information processing failures attributed to breakdowns in interpersonal contact in teams. Also, the current literature fails to explain how information flows from the internal and external environment lead to paradoxical cognition by countering information processing bias raised by shortcomings of a TMT’s mental frame. The lack of attention paid to the information flows that enable a TMT to perceive the paradoxical tensions restricts us from achieving a profound understanding of the tools that a TMT could use to transcend a contradiction and come to ambidextrous solutions.

This study applies a socio-cognitive approach and highlights the supportive role of TMTs, and especially their role in sensing and seizing opportunities for the creation of novel combinations of knowledge between exploration and exploitation activities. We conduct empirical research among 267 Dutch and German small and medium-sized enterprises (SMEs), with a headcount between 25 and 250 employees, and we aim to assess how TMTs could realize organizational ambidexterity by analyzing related socio-cognitive processes. This group is the subject of this analysis, as particularly TMTs in SMEs have to face a unique set of cognitive

8 challenges. This is due to the fact that TMTs in SMEs are likely to play both a strategic and operational role and therefore, they directly experience the dissonance of competing knowledge demands inherent in the pursuit of an ambidextrous orientation (Lubatkin et al, 2006).

Moreover, this study pursues a conceptual framework with antecedents of organizational ambidexterity. TMT reflexivity is associated with the learning behaviors incorporated within a TMT (Schippers, West, & Dawson, 2015). TMT reflexivity counters the sort of information processing biases within a TMT that arise due to breakdowns in interpersonal interaction (Schippers et al, 2014). Paradoxical cognition thus enables TMTs to make sense of the input from TMT reflexivity and determines whether this will lead to organizational ambidexterity (Smith & Tushman, 2005). In this framework, paradoxical cognition refers to a shared mental model for information processing. Paradoxical cognition enables the effective handling of tensions between exploration and exploitation by providing a platform for cognitive differentiation and integration processes (Smith & Tushman, 2005). It is argued that the relationship between TMT reflexivity and paradoxical cognition is moderated by cues from the internal and external environment. These cues enable a TMT to counter the sort of information failures related to the shortcomings of their TMT cognitive frame. Benchmarking is seen as a tool TMTs can use to acquire new and external information flows in the form of reports. By means of lower level management team (LLMT) attention, in the form of conversations between the LLMT and TMT, the TMT can acquire internal bottom-up flows of information. As such, we propose the following conceptual framework:

Figure 1. Conceptual Model

TMT Reflexivity LLMT Attention Benchmarking Paradoxical Cognition Organizational Ambidexterity

9 Based on this framework, we endeavour to make five contributions to the current knowledge base in two distinct ways. To start off, we analyse successful information processing within a TMT. As such, we first contribute by empirically verifying the construct of paradoxical cognition (Smith & Tushman, 2005) and its effect on organizational ambidexterity (Cao et al, 2009; Jansen et al, 2009; O’Reilly & Tushman, 2008). By doing so, we add to the understanding of the socio-cognitive antecedents of organizational ambidexterity (Van Neerijnen, 2015). Second, we introduce TMT reflexivity as an antecedent of paradoxical cognition. We contribute to the theoretical field of the antecedents of paradoxical cognition by arguing that TMT reflexivity could counter the information processing failures TMTs are somewhat aware of themselves. Furthermore, we contribute by arguing that the single and double loop learning behaviors associated with TMT reflexivity both positively affect paradoxical cognition (Van Neerijnen, 2015). Third, we argue that paradoxical cognition mediates the TMT reflexivity-organizational ambidexterity relationship (Van Neerijnen, 2015). The ability of the TMT to engage in cognitive differentiation as well as in cognitive integration processes will determine how well they can embrace the exploration-exploitation paradox and whether or not TMT reflexivity leads to organizational ambidexterity. We argue for the capacity of TMT reflexivity and paradoxical cognition to inspire organizational participants to engage in ‘Janusian thinking’, i.e. the ability to combine antithetical elements into a new mental model that transcends previous explorative and exploitative schemata (Lewis, 2000).

Subsequently, we analyze the processing of information flows from the internal and external environments within the TMT. We further enhance the current knowledge base by arguing that cues extracted from internal and external information flows could support TMTs in their ability to reach paradoxical cognition and counter ‘Schismogenic thinking’, i.e. the tendency of managers to forcefully exclude the consideration of undesired information (Cameron & Quinn, 1988). TMT reflexive practices that consider information from benchmarks and LLMT attention could bring imperceptible blind spots to light and could counter information-processing biases TMTs are not aware of. As such, our fourth contribution relates to our claim that benchmarking, as an independent variable from the external environment, positively moderates the relation between TMT reflexivity and paradoxical cognition. Benchmarks provide a TMT with contradicting information that aids paradoxical tensions in a TMT. TMT reflexivity enables

10 the TMT to process this information in a way that enables the TMT to contrast alternative ways of thinking with existing ways of thinking, leading to paradoxical cognition. Fifth, we contribute by arguing that LLMT attention, as an independent variable from the internal environment, provides the TMT with contradicting information as well, and therefore moderates the relationship between TMT reflexivity and paradoxical cognition.

Literature Review and Hypotheses

Achieving Ambidexterity and the Exploration-Exploitation Paradox

For securing organizational survival, organizations need to become ambidextrous - that is to say, they have to explore and exploit simultaneously (Cao et al, 2009; Jansen et al, 2009; O’Reilly & Tushman, 2008). In academic literature, ambidexterity is seen as a dynamic capability, as it allows organizations to adapt continuously and adequately to changes in the environment (O’Reilly & Tushman, 2008). This phenomenon is dialectical in nature as an organization is required to continuously differentiate between and integrate explorative and exploitative knowledge resources (O’Reilly & Tushman, 2008; Smith & Lewis, 2011).

Successfully dealing with the two distinct forces is challenging as exploration and exploitation activities need to be structurally differentiated and safeguarded from each other’s potential negative influences. Yet at the same time, to truly benefit from the value enhancing effects of both and reach synergistic outcomes they need to be combined as well (Jansen et al, 2009; O’Reilly & Tushman, 2008). This is exceptionally hard to do because explorative and exploitative activities are contradictory in nature; they are associated with different structures and processes, they are situated in distinct contexts in managerial logics and often compete for organizational resources (Gavetti & Levinthal, 2000; March, 1991).

Specifically, equal treatment of explorative and exploitative knowledge resources is challenging as they are identified as two fundamentally different learning activities (March, 1991). Exploration is associated with radical innovation, discovery, and experimentation (Katila & Ahuja, 2002; Rosenkopf & Nerkar, 2001), while exploitation is associated with incremental innovation, refinement, and implementation (Beckman, Haunschild, Phillips, 2004; Benner & Tushman, 2002). As such, exploration and exploitation emerge from contradictory

knowledge-11 processing capabilities (Floyd & Lane, 2000; Lubatkin et al, 2006). Exploitation emerges from the use of explicit knowledge bases, such that by internalizing and combining these knowledge bases, incremental refinements to existing trajectories can be made (Nonake, 1994; Lubatkin et al, 2006). This is associated with convergent thinking and reducing variance (Flynn & Chatman, 2001; Smith & Tushman, 2005). On the contrary, exploration emerges from the use of tacit knowledge bases, such that by externalizing and combining these bases, new trajectories are developed (Nonaka, 1994; Lubatkin et al, 2006). This is associated with divergent thinking and increasing variance (Flynn & Chatman, 2001; Smith & Tushman, 2005). In this study, we explore the mechanism that enables TMTs to successfully manage the contradicting knowledge processes that lead to exploration and exploitation.

Reframing the Paradox: a Socially Constructed Process

Previous ambidexterity literature contends that achieving organizational ambidexterity is primarily a cognitive challenge (Eisenhardt, Furr, & Bingham, 2010; Gilbert, 2005; Smith & Tushman, 2005; Van Neerijnen, 2015). This is due to the fact that a paradox is created as the oppositional tendencies between exploration and exploitation are juxtaposed through a TMT’s cognition (Smith & Tushman, 2005; Van Neerijnen, 2015). As paradoxical tensions are perceptual in nature, organizational members must cognitively reframe this sometimes irrational or even absurd (enacted) coexistence of opposites in order to transcend their apparent contradictions and realize ambidextrous solutions (Lewis, 2000). Reframing entails “moving to a different kind of comprehension” (Cameron & Quinn, 1988 p. 21). In light of this, Rothenberg (1979) introduced the concept of ‘Janusian thinking’. This is a common thought pattern that enables creative individuals to see the integrative function of anti-ethical elements. According to this concept, ingenious thinking occurs when the opposites of existing logics are brought together (Cameron & Quinn, 1988; Lewis, 2000).

Paradoxical pressures serve as a notable trigger mechanism as they could produce the tensions that lead to such thinking. Contradictions cause dissonance that necessitates action to resolve the contradictions (Ford & Backoff, 1988; Davis et al, 1997). For this, actors need to be able to process explicitly contradictory cues within the existing flow of information. As such, they must be able to keep looking for cues and triggers beyond their existing cognitive frame (Bartunek, 1988; Nystrom & Starbuck, 1984; Davis et al, 1997). In this way, they could find the

12 trigger of paradoxical tensions in both the internal and external environment of their organization or by contrasting both (Van de Ven & Poole, 1988; Davis et al, 1997).

A TMT’s understanding of the trigger posed by paradoxes is fundamentally a social constructionist process that determines how TMTs are able to create new and valuable configurations of exploratory and exploitative knowledge resources (Lewis, 2000; Smith & Lewis, 2011; Van Neerijnen, 2015). Actors first construct a unique and individual interpretation of an experience. This is because actors cannot occupy exactly the same position in the world or share the exact same history (Neisser, 1976; Davis et al, 1997). Organizational members reach convergence of frames as they create and maintain an inter-subjective world by engaging in conversations that result in certain mental models (e.g. Johnson-Laird, 1989; Van Neerijnen, 2015), or cognitive frames, so to speak (e.g. Walsh, 1995; Van Neerijnen, 2015). Lewis (2000) refers to ‘mixed messages’ as the creator of paradoxical tensions that lead to reframing and shared cognitive frames. His work implies that paradoxical tensions occur due to inconsistencies between statements or between verbal and nonverbal responses that appear during social interactions. Thus, information processing is a central aspect of these socio-cognitive activities (Schippers et al, 2014).

The Challenge of Reframing: Information Processing Failures

The social constructionist process by which cognitive frames are maintained render reframing challenging, as the character of this process leads to the possibility of information processing failures to occur. An ‘information-processing failure’ is defined as “a distortion in the exchange of, communication about, or elaboration on information due to either an omission or error in information sampling or biased elaboration of the information” (Schippers et al, 2014, p. 722). In line with Schippers et al, this study acknowledges information processing failures of two kinds: on the one hand, these failures could be attributed to the nature of the cognitive frame and the associated cognitive and emotional biases (e.g. Davis et al, 1997; Smith & Tushman, 2005), while on the other hand, they stem from confusion, misunderstanding, or withholding of information in teams that occurs due to ruptures in interpersonal interactions (Schippers et al, 2014). This makes it challenging for TMTs to be able to pursue one set of goals while looking for appropriate information and cues that will signal the need to shift to another (Cameron & Quinn, 1988).

13

The failures could be attributed to breakdowns in interpersonal contact in teams, as teams

often process information in ways that produce systematic errors (Heath, Larrick, & Klayman, 1998; Schippers et al, 2014). Schippers et al, (2014) organize these failures in three categories: the failure to elaborate and examine implications of shared information; the failure to share relevant information; and the failure to update or alter current behaviors or prior conclusions. Teams should be aware of these biases in order to recognize, counter, and learn from them (Schippers et al, 2014). However, teams will not always be reasonably accurate in recognizing effective and ineffective information processing, making it challenging to do so (Cameron & Quinn, 1988).

The failures could also be related to shortcomings of the cognitive frame; specifically, these failures are attributed to numerous manifestations of bounded rationality (Kahneman, 2007; Simon, 1947; Lewis, 2000; Schippers et al, 2014). The cognitive frame acts as a subtle filtering device, as managers tend to ignore cues located outside the cognitive frame. This makes a TMT resistant to the incorporation of new information (Tripsas & Gavetti, 2000). As TMTs make sense of experiences through their existing cognitive frame, they choose interpretations that support rather than challenge their existing mental frames (Westenholz, 1993; Lewis, 2000; Cameron & Quinn, 1988; Davis et al, 1997). The cognitive bias could be caused by TMTs values, motives, and problem solving styles, which tend to reflect preferences to certain information processing inferences (Starbuck & Milliken, 1988; Levit & March, 1988; Cameron & Quinn, 1988). In lieu of this, Bateson (1979) introduced ‘Schismogenic thinking’; this refers to purposeful thinking that forcefully excludes undesired information. The assumptions that underlie such thinking prevent managers from seeing the information flow or feedback loop that connects them to the environment (Cameron & Quinn, 1988).

Paradoxical Cognition as an antecedent of Organizational Ambidexterity

Smith and Tushman (2005) published a seminal study in which they introduced paradoxical cognition as a necessity for paradoxical thinking and as an antecedent of organizational ambidexterity. This is because paradoxical cognition provides TMTs with the means to escape cognitive biases, and therefore they can process seemingly contradicting information; this in turn enables them to conceive of ambidextrous solutions (Smith & Tushman, 2005). Paradoxical cognition is defined as the “managerial frames and processes that recognize and embrace

14 contradiction” (Smith & Tushman, 2005, p. 523). Shared managerial frames create a template through which the TMT filters knowledge, controls attention, and directs action. This defines a TMT’s interpretation of the issues it faces, and therefore how TMTs interpret tensions between exploitation and exploration (Weick, 1979; Smith & Tushman, 2005). Importantly, cognitive frames provide the foundation for cognitive processes, defined as “behavioural routines and ways that managers use to think about and respond to information” (Smith & Tushman, 2005, p. 526). Cognitive processes are divided in two distinct processes of differentiation and integration that enable TMTs to cope with inconsistent strategic agendas (Smith & Tushman, 2005). The suggested impact of this construct still requires empirical confirmation.

Organizational ambidexterity is dialectical in nature as an organization is required to continuously differentiate and integrate explorative and exploitative knowledge resources (O’Reilly & Tushman, 2008; Smith & Lewis, 2011). To achieve organizational ambidexterity, exploration and exploitation must both succeed, and therefore managers should confront the relation between these two products by cognitively differentiating and integrating related contradicting tensions (Van Neerijnen, 2015). This results in two main reasons why paradoxical cognition leads to ambidexterity.

First, paradoxical cognition enables processes of cognitive differentiation. These

processes involve “the recognition and articulation of distinctions” (Smith & Tushman, 2005, p. 527). This enables TMTs to generate new categories and classifications that allow them to shift between agendas of exploration and exploitation in multiple ways (Smith & Tushman, 2005). To begin with, TMTs are enabled to shift between agendas as differentiations dampen commitments to existing logics. As such, managers are encouraged to explore new opportunities, unburdened by the context of the prevalent logic (Smith & Tushman, 2005; Gilbert, 2005). Furthermore, accessing and using more information helps decision makers to respond effectively to challenging situations (Weick, Sutcliffe, & Obstfeld, 1999; Smith & Tushman, 2005). Also, by means of cognitive differentiation, team members constantly contrast existing ways of thinking with innovative ways of thinking and evaluate the differences (Smith & Tushman, 2005). Finally, a crucial function of cognitive differentiation processes is that they open up the possibility for TMTs to identify more targeted and confused synergies between existing perspectives. The more differentiated knowledge is, the deeper the understanding a TMT has of

15 the different parts of the products, thus the more likely it is that TMTs come to points of convergence and synergy (Smith & Tushman, 2005).

Second, paradoxical cognition enables cognitive integration processes as well. These processes involve “ the shifting of levels of analysis to identify potential linkages” (Smith & Tushman, 2005, p. 527). These processes are important, as only making distinctions is an insufficient process for establishing synergies. When knowledge is differentiated, managers recognize new linkages they were previously unaware of (Tsang & Zahra, 2008; Nystrom & Starbuck, 1984; Van Neerijnen, 2015). By applying processes of integration, conceptual connections could be conceived among differentiated perspectives. This involves shifting levels of analysis to a meta-level of thinking to identify possible synergies (Smith & Tushman, 2005; Van Neerijnen, 2015). Consequently, cognitive integration processes enable ‘Janusian thinking’ (Lewis, 2000).

In light of these arguments, TMTs need to shift to the organizational level, dismantle existing logics, and link the overarching frame of the separate logic categories, to reinforce the cooperation between inconsistent strategic agendas. This is enabled through paradoxical TMT cognition. This is because ultimately, diverging forces could trigger the usage of targeted integration mechanisms that lead to organizational ambidexterity (Smith & Lewis, 2011; Andriopoulos & Lewis, 2009). As such, we predict that:

“H1: TMT paradoxical cognition positively relates to ambidexterity.

TMT Reflexivity as an antecedent of Paradoxical Cognition

The ability of TMTs to achieve paradoxical cognition is likely to be conditioned by TMT reflexivity. Existing conceptual literature has proposed factors that could be antecedents of paradoxical cognition, such as team design, task interdependence, and leadership style (Smith & Tushman, 2005). However, even though it is likely that these factors are of great value, they do not provide a complete explanation of the construct, as we know very little about how TMT learning behaviors, such as reflexivity, affect paradoxical cognition (Van Neerijnen, 2015). Based on the results of previous studies, we know that reflexivity has a strong impact on the ability of managers to successfully deal with paradoxes (Bledow, Frese, Anderson, Erez, & Farr, 2009; Hirst, Van Knippenberg, & Zhou, 2009; Lüscher & Lewis, 2008; Poole & Van de Ven,

16 1989; Van Neerijnen, 2015; Lewis, 2000). A form of practice based on open discussion helps to enable paradoxical thoughts and social reframing (Westenholz, 1993). In past academic literature, attention has been granted to self and social reflection or interaction (Lewis, 2000) and reflective questioning (Lüscher & Lewis, 2008), which enable actors to process paradoxical tensions.

We draw on organizational behavior literature and define TMT reflexivity as “a

discussion-based process in which teams assess current information and past or planned actions, decisions or conclusions, with respect to goals, processes or outcomes” (Schippers et al, 2014, p. 735). Reflexivity is seen as a relational activity, as TMTs continuously make sense of situations by communicating and questioning views and ideas within the teams (Schippers, West, & Dawson, 2010). The related practices greatly impact the way managers enact on their environment (Weick, 1979). During reflective practices, questions are raised and answered. As such, TMT reflexivity creates a situation in which TMT members will engage in more systematic information processing (De Dreu, 2007). In line with Smith and Tushman (2005), we argue that low levels of cognitive bias and high levels of cognitive understanding lead to paradoxical cognition. We argue that TMT reflexivity counters cognitive biases and adds to the TMT understanding in three main ways.

First, TMT reflexivity could restore distortions in interpretations of experiences between

individual members of a team. This is because the elaboration of abstract models triggers attention that determines a managers understanding of the issues they face (Dutton & Ashford, 1993; Nigam & Ocasio, 2010; Ocasio, 1977). Additionally, members become aware of the confines of their cognitive frame as unarticulated preferences and meanings are made explicit through reflexivity (Cameron & Quinn, 1988; Levitt & March, 1988). Also, reflexivity renders actors more open-minded, this makes them more likely to process different sorts of information (Schippers, Den Hartog, Koopman, & Wienk, 2003). Actors are less likely to stick to ingrained ways of thinking as reflexivity stimulates organizational participants to consider and value new ways of thinking (Schippers et al, 2003; West, 2000). Furthermore, reflexivity aids to the reconstruction of interpretations as related practices create a sense of commitment among team members that enable them to successfully process information (Cameron & Quinn, 1988; Huber & Lewis, 2010; Schippers et al, 2014; Martin & Eisenhardt, 2010). However, a condition for this

17 kind of repairs, is the ability of at least one team member to detect distortions in the TMT cognitive frame (Schippers et al, 2014).

Second, TMT Reflexivity could counter conscious information processing biases that occur in teams due to failures in interpersonal contact (Schippers et al, 2014). Schippers et al, demonstrate that this happens in three distinct ways. First, TMT reflexivity may mitigate the failure to search for and share information, as team reflexivity increases the probability that a team will identify useful information (Schulz-hardt, Brodbeck, Mojzisch, Kerschreiter, & Frey, 2006; Schippers, 2014). Second, team reflexivity helps to remit failures to elaborate and derive conclusions from knowledge resources; as such, reflexivity enables teams to carefully consider new information (Lubatkin et al, 2006; Schippers et al, 2014). Third, team reflexivity can help mitigate the failure to revise and update conclusions by enabling teams to pay explicit attention to the decision making process (Schippers et al, 2014). This may also entail considering threats and opportunities in the broader internal or external environment.

Third, TMT reflective practices increase a TMTs information processing capabilities, as it stimulates the single- and double loop learning processes that are associated with processes of cognitive differentiation and integration. The cognitive differentiation process is stimulated as both single- and double loop learning behaviors associated with TMT reflexivity play a distinctive role (Van Neerijnen, 2015). Single-loop learning enables actors to frame lessons learned from previous actions in existing mental models; this leads to small differentiations in existing logics. Double-loop learning involves a deeper assessment of a situation by reflecting on values and norms (Argyris, 1991; Lewis, 2000). Importantly, double-loop learning frees organizational participants from perspective-limiting assumptions (Gray, 2007). Therefore, organizational members can actively question ingrained mental models and move beyond entrenched assumptions (Argyris & Schön, 1997; Lewis, 2000). As such, it gives TMTs the impetus to explore new inferences or encourages them to use other terms to make cognitive differentiations (Smith & Tushman, 2005).

The cognitive integration process is stimulated by both single- and double-loop learning behaviors associated with TMT reflexivity (Van Neerijnen, 2015). Only if a TMT is willing to criticize and reconsider the manner in which distinct constructions relate to one another, new connections between the different constructs can be developed. Single-loop learning aids a TMT in developing more sophisticated knowledge structures; this leads to a greater understanding and

18 helps managers to establish alterations in prevalent logics. At the same time, double-loop learning facilitates the integration process of TMTs, as it supports them in their effort to ‘break their lenses’, dismiss past tensions, and recognize linkages they did not see before (Dunbar, Garud, & Raghuram, 1996; Van Neerijnen, 2015). These processes are associated with the unlearning of schemata, which facilitates the rise of new cognitive integrations and the construction of a new-shared mental model (Tsang & Zahra, 2008; Nystrom and Starbuck, 1984; Van Neerijnen, 2015).

As such, TMT reflexivity counters information processing biases. This may help a TMT to delve more deeply into the seemingly contradicting coexistence of opposites that lead TMTs to embrace contradicting forces. Therefore we argue:

H2: TMT reflexivity positively relates to paradoxical thinking.

The mediating role of Paradoxical Cognition concerning the relation between TMT Reflexivity and Organizational Ambidexterity

Our arguments suggest that the ability of reflective TMTs to achieve organizational ambidexterity is likely to be mediated by paradoxical cognition. Prior literature indicates that shock, open communication, experimentation, and paradoxical leadership are antecedents of social reframing that enables organizational ambidexterity (Davis et al, 1997; Lewis, 2000). However, these instruments might be insufficient for social reframing on their own (Davis et al, 1997). Reflexivity could support these instruments, as this enables managers with a different frame of reference to draw on divergent insights that might help actors to embrace paradoxical tensions (Davis et al, 1997). We further argue that TMT reflexivity influences organizational ambidexterity through TMT paradoxical cognition due to two main reasons.

First, TMT reflexivity provides a TMT with the input that enables them to deal with the constraints that are inherent to human cognition. However, without a mental model that enables cognitive differentiation and integration, the outcomes of reflexivity are likely to be ignored (Van Neerijnen, 2015; Smith & Tushman, 2005). TMTs need a mental frame that enables them to make sense of the input from reflexivity (Van Neerijnen, 2015). Thus, while TMT reflexivity is likely to increase the chances that TMT members recognize linkages between exploration and exploitation activities, it is through paradoxical TMT cognition that such combinations are realized.

19 Second, by making sense of the exploration-exploitation paradox on the cognitive and surface level, actors can go through a dialectical process in which the subjective and objective world alternate. The former refers to making sense of the exploration-exploitation paradox based on an abstract, cognitive, and emotional level (Van Neerijnen, 2015; Lewis, 2000; Smith & Lewis, 2011; Weick, 1979). The latter encounters contradictory agendas in the organizational system in the form of material tensions that could be the outcomes of previous sense making experiences (Van Neerijnen, 2015). Complex TMTs constantly dismantle logics and search for linear cause and effect relationships (Cameron & Quinn, 1988). Thus, we argue that paradoxical cognition enables a TMT to constantly shift between the polarity on the cognitive and surface level, and thereby increase the complexity of a TMT. The more complex a TMT is, the more dimensions and relationships are used, the more able TMTs are to transcend paradoxical forces and come to creative, synergistic results.

Polaroid’s story provides an example in which a TMT was unable to engage in processes of cognitive differentiation and integration (Tripsas & Gavetti, 2000). Polaroid’s TMT firmly believed that customers valued a physical print due to their past success. However, a transition to digital imaging was occurring in the competitive landscape. Polaroid’s TMT viewed this transition through a technology focused cognitive frame and therefore failed to cognitively differentiate their extant analog cameras from their newly developed digital cameras. They conceptualized digital innovation as fundamentally similar to its existing strategy. When new TMT members were hired that articulated a different perspective, neither group considered the other’s viewpoint – there was a cognitive dissonance between the old TMT and the newly hired members. A greater ability for cognitive integration in Polaroid’s TMT might have resulted in a very different course of the company’s history. This would have increased the chances that the TMT would have recognized and acted on the new differentiated knowledge domains to construct synergistic ideas. As such, we argue that:

H3: TMT paradoxical cognition fully mediates the relationship between reflexivity and ambidexterity.

The moderating role of Benchmarking in the relation between TMT Reflexivity and Paradoxical Cognition

20 increased ability of a TMT to attain paradoxical cognition. This is mainly because information from benchmarks contains new and seemingly oppositional information that provoke tensions next to existing logics. It seems needless to say that TMT reflexivity enables the TMT to see the value of this information in a way that this leads to social reframing (Lewis, 2000; Tushman & Smith, 2005). We explore this relationship, as prior literature suggests that understanding how organizations handle contradiction requires paying attention to the organizational environment (Jansen et al, 2006). In particular, environmental conditions of plurality, change, and scarcity could render latent tensions salient (Smith & Lewis, 2011). Despite these suggestions, the nature and influences of the external environment on a TMT’s ability to deal with cognitive contradiction remains somewhat underdeveloped.

We define benchmarking as the extent to which a TMT compares features from their organization with characteristics of organizations in the same or other industries. In the relevant literature, the process of benchmarking is described as “a continuous, systematic process for evaluating the products, services and work processes for organizations that are recognized as representing best practices for the purpose of organizational improvement” (Spendolini, 1992, p. 9; Costa, Formoso, Kagioglou, Alarcon, & Caldas, 2006). Information from benchmarks has been demonstrated to be a catalyst for general changes in organizational thinking (Drew, 1997; Cassell, Nadin, & Older Gray, 2001). This is because the lessons learned from competitors can be used to establish improvement targets (Costa et al, 2006). Benchmarking provides an influx of external knowledge in the form of reports; these kinds of inflows are rather unambiguous and provide the recipient managers with data in a standardized way (Brady & Davies, 2004; Mom et al, 2007). Consequently, this study contends that benchmarking moderates the relationship between TMT reflexivity and paradoxical cognition in numerous ways.

First, benchmarking provokes contradiction, as the phenomenon suggests the important responsibilities of the TMT to scan the environment and assess the differences and similarities of surrounding organizations (Aldrich, McKelvey, & Ulrich, 1984; Porac & Thomas, 1990). Through the practice of continuous comparison, events in the environment can trigger a change in the mind-sets of TMTs (Nigam & Ocasio, 2010). By identifying contrasts, paradoxical tensions arise that allow a TMT to contrast existing ways of thinking with alternative attitudes and approaches (Smith & Tushman, 2005; Noda & Bower, 1996); the comparisons that follow from this play a pivotal role, as Levitt and March (1988) state that organizational learning

21 depends on comparing aspiration levels with observed outcomes.

Second, benchmarking leads to contradicting information, as TMTs could use benchmarking as a tool to incorporate new external logics next to existing internal logics (Drew, 1997; Cassell et al, 2001). Benchmarking aids the plurality of information by providing the ‘variance in material’, as one of the main objectives of benchmarking is to introduce external ideas to an organization (Splendolini, 1992). Exploration beyond organizational boundaries leads to more mixed messages than internal exploration (Rosenkofs & Nerkar, 2001; Raisch et al, 2009; Ahuja, 2000; Mol & Birkinshaw, 2014). Hence, external knowledge could be used as a means to create contradiction because it is new by nature; it creates new insights and could thus be used as a source of new inspiration (Raisch et al, 2009; Eisenhardt & Martin, 2000). However, empirical research by Mol and Birkinsaw (2014) suggests the opposite. They argue that true differentiations transpire primarily internally through a TMT that acts on their own initiative. This could transpire under high levels of competitive rivalry (Jansen et al, 2006). This refers to the degree of competition reflected in the number of competitors and the number of areas in which there is competition. This could trap TMTs in endless improvements of existing ways of thinking (Levinthal & March, 1993; Jansen et al, 2006). This is because TMTs’ mental frames within organizational environments are socially constructed (Porac & Thomas, 1990). As a result, outcomes of new ideas tend to rapidly become diffused across the population of competitors, causing contradictions between different organisations to fade away (Levinthal & March, 1993; Jansen et al, 2006).

Third, benchmarking creates contradiction, as information about changing environments

makes current logics obsolete. Paradoxes emerge as managerial beliefs fail to keep up with external changes (Cameron & Quinn, 1988; Smith & Lewis, 2011). Specifically, a high level of environmental dynamism introduces the need for rationales that depart from existing logics (Jansen et al, 2006). Environmental dynamism entails the rate of change and the degree of instability in a environment (Levinthal & March, 1993). During periods of turbulence, external knowledge sourcing will stimulate second-order learning, while during convergent periods, such processes pass on information that improves existing competencies associated with single-loop learning (Virany, Tushman, & Romanelli, 1992).

Finally, competing values between managers towards the usability of information from benchmarks leads to mixed messages. The implicit values of managers range from an internal to

22 an external focus (Quinn & Rohrbaugh, 1983). Different values could lead individual managers to interpret information differently, giving rise to contradicting thoughts (Davis et al, 1997). To be exact, managers that are oriented towards externally generated facts and empiricisms tend to process information by analyzing the focal problem, cue, or message from a static, cross-functional view. Conversely, managers with an internal focus tend to analyze the problem, cues, or message from a dynamic, longitudinal perspective (Cameron & Quinn, 1988).

Consequently, we assert that TMT reflexivity is a necessary antecedent in our model. This is because benchmarking could only lead to new insights as TMTs are receptive to new ideas (Costa et al, 2006). Overall, TMT reflexivity improves the receptiveness of TMTs towards new information (refer to hypothesis 2). Prior literature suggests that information from the external environment even improves the effect of TMT reflexivity as an information processing structure (Schippers, West, & Dawson, 2012). This is because through environmental scanning, a richer array of information is collected and this enhances a TMT’s ability to construct interpretations that form a basis for debate (Thomas, Clark, & Gioia, 1993). Also, information-processing failures are reduced by information that is grounded in concrete or relatively concrete evidence to support or clarify observations (Schulz-Hardt et al, 2006; Schippers et al, 2014). However, there are circumstances in which the relationship between external output and high levels of information processing is negative. The academic literature suggests that knowledge categories could become so complex that additional inputs from the external environment could compromise or slow down information processing (Mol & Birkinshaw, 2014). We recognize the value of this argument, however, we believe that by introducing reflexivity as a condition, a TMT is able to select relevant benchmarks and can thus diminish complexity. As such, we contend:

H4: External benchmarking moderates the relation between TMT reflexivity and cognitive paradox, such that this relation is stronger for TMTs with high levels of external benchmarking, but weaker for TMTs with low levels of external benchmarking.

The moderating role of LLMT Attention in the relationship between TMT Reflexivity and Paradoxical Cognition

23 TMTs are likely to be supported in their ability to reach paradoxical cognition by including information acquired from LLMTs in their reflexive practices. This is because juxtaposing information from LLMTs with TMTs’ entrenched beliefs may provoke tensions that lead to social reframing (Andriopoulos & Lewis, 2009; Floyd & Lane, 2000). We consider the role of LLMT attention, because what managers interpret as a cue and how they interpret it may vary according to hierarchy, functional area, and division (Floyd & Lane, 2000). This is rooted in the idea that organizational learning unfolds on a basis of interpretations from the past (Levitt & March, 1988). As such, the literature leads us to suspect that LLMT attention influences the relation between TMT reflexivity and paradoxical cognition.

We define LLMT attention as “conversations between the TMT and lower level management teams about (future) market needs, strategy development, new-product developments and the alternative use of knowledge and capital”. This familiarizes the TMT with the LLMT and, next to formal integration, this also leads to increased informal integration (O’Reilly, Caldwell, & Barnett, 1989; Jansen et al, 2009). We see the TMT as the recipient of a bottom-up inflow of knowledge through face-to-face meetings. Importantly, the LLMTs themselves are not included in the TMT reflexive practices; a TMT member that acquires information has the responsibility to share this information during a TMT reflexive practice. As such, it is argued that LLMT attention improves the effect of TMT reflexivity on paradoxical cognition in a number of ways.

First, LLMTs may carry conflicting messages because managers at different levels of the

firm perform different strategic roles, leading them to pay attention to different cues (Floyd & Lane, 2000; Tripsas & Gavetti, 2000). Floyd and Lane (2000) provide an overview of the literature that considers these different roles. They combine multiple important studies and conclude that TMTs perform a decision-making role that is associated with ratification, direction, and recognition. Conversely, LLMTs’ roles are concerned with experimenting, conforming, and responding; in other words, they react to information (Floyd & Lane, 2000). Also, from the literature on paradoxes, we know that managers have different roles between management layers. Moreover, TMTs set the context; they provide strategic leadership, determine a firm’s portfolio of projects, and make decisions, while LLMTs choose how firms should apply their discipline and passion to enhance product development (Andriopoulos & Lewis, 2009).

24

Second, contradicting information arise due to the different boundary spanning activities

LLMTs find themselves in, compared to TMTs; as such, they are exposed to different cues (Katz & Tushman, 1979; Floyd & Lane, 2000). Previous case studies in the field of strategy illustrate that LLMTs are directly confronted with new customer demands, new technical developments, and changing market conditions (Burgelman, 1983a; Van De Ven, 1980; Mom et al, 2007). Thus, they are in a good position to identify new opportunities ( Burgelman, 1983b; Van de Ven, 1980; Mom et al, 2007). In contrast, the boundary-spanning activities TMTs engage in primarily focus on the capital market for shareholder concerns (Pfeffer & Salancik, 2003; Floyd & Lane, 2000). These cues constrain the TMTs in their ability to develop new ideas, which makes them likely to prefer the established status quo strategy (Burgelman, 1994; Floyd & Lane, 2000). However, it should be kept in mind that such impediments tend to be minimized in SMEs, as compared to large firms, simply because TMTs in SMEs are closer to the operating core and to a firm’s existing competencies (Lubatkin et al, 2006).

Third, contradicting assumptions arise as managers interpret cues differently because of

differences in prior logic and expertise (Mom et al, 2007), and because managers might be oriented towards different core-values (Cameron & Quinn, 1988). TMTs tend to have expertise related to existing technologies and therefore they interpret cues from this perspective. On the contrary, LLMTs tend to have knowledge related to newer areas of expertise; as such, they are more likely to interpret cues from this perspective (Burgelman, 1991; Floyd & Lane, 2000).

Contradicting interpretations may also stem from variance in core values between

different management levels of a firm, such as variance in; motives, behavioral norms, beliefs, expectations and problem-solving styles (Nandram & Klandermans, 1993; Floyd & Lane, 2000; Cameron & Quinn, 1988). Motives vary as managers tend to be motivated by near-by situations (Locke, Saari, Shaw, & Latham, 1981; Martin & Eisenhardt, 2000). For example, LLMTs may put the goals and problems of their own team ahead of general firm aims and problems (Collis & Montgomery, 1997; Freeland, 1996; Martin & Eisenhardt, 2010), while TMTs put firm aims and problems ahead (Floyd & Lane, 2000; Martin & Eisenhardt, 2000). This leads to contradiction, as managers that tend to have a ‘multiple focus’ take the time to seek diverse opinions, search for solutions, and integrate various perspectives, while managers that have a purposive orientation make rapid decisions (Cameron & Quinn, 1988). As such, LLMTs could push for shorter timelines while TMTs are held back by timely processes leading to contradiction.

25

Accordingly, TMT reflexivity is a necessary condition. It is needless to say that this will enable the TMT to counter distortions in processing the information acquired by LLMT attention (refer to hypothesis 2). We consequently suggest that including LLMT information in the TMT reflexive practice may further improve information processing. Successful information processing is enabled as conversations between layers influence a TMT’s ability to build a realistic understanding of key preferences (Eisenhardt, Kahwajy, & Bourgeouis, 1997; Burgelman, 1983b; Floyd & Lane, 2000; Mom et al, 2007). Additionally, increased and enhanced information processing occurs as LLMT attention leads to information richness and accuracy that aids a TMT to recognize contradiction (Eisenhardt & Zabracki, 1992; Smith & Tushman, 2005). Nevertheless, abundant information could also increase difficulties for the recognition of paradoxical forces, as actors are confronted with an information overload and therefore experience chaos (Davis et al, 1997). However, we suggest that TMT reflexivity enables the TMT to select relevant information. Consequently, we believe that:

H4: Lower level interaction moderates the relation between reflexivity and cognitive paradox, such that this relation is stronger for TMTs with high levels of interaction, but weaker for TMTs with low levels of interaction with LLMT.

Methodology

Data and Sample

The data that is used in this study is retrieved from an ongoing research project designed to establish an innovation benchmark among Dutch and German firms. The data is obtained by means of an online survey among top managers from firms with a headcount between 25 and 250 employees. The 250 upper-level cut-off point is the generally accepted number for small and medium enterprise (SME) categorization (European Commission, 2003). TMTs of firms of the selected size have to deal with the same competitive pressures as TMTs in very large organizations. The lower-cut-off-point is taken because firms with less than 25 employees are omitted to account for some degree of organizational and coordinative challenges, as very small firms are not very likely to have the slack resources that enable TMTs in large organizations to

26 focus entirely on either exploration or exploitation. For these reasons, TMTs in SME’s are more likely to rely on the competences of the top management team to achieve ambidexterity compared to TMTs in very large firms.

After omitting the companies that fell outside the 25 to 250 employee headcount and screening the data on the absence of key information criteria, we retained data from 267 firms. Our final sample covers a wide range of industries, including metal and wood manufacturing (37,6%), administrative support, management of companies, professional, scientific and technical services (15,8%), retail and wholesale trade (15,4%), administrative support services (6,4%), utilities and construction (4,9%) and other industries (19,9%). On average, a TMT consisted of 6,72 team members (s.d. = 4.61). The mean age of the employees in the sample was 46,30 years (s.d. = 8,38). Furthermore, 54,9 percent of the organizations were located in the Netherlands against 45,1 percent in Germany. Further, the average age of the organizations in the sample was 45.55 years (s.d. = 7,98). Finally, the organizations in the sample employed on average 119.95 fulltime employees (s.d. = 107.56).

The data is tested for non-response bias. T-tests are conducted to examine the differences

between respondents and non-respondents on prior financial performance, the number of fulltime employees and total assets. T-tests reveal no significant differences. This shows that non-response bias not seems to impose a problem.

To assess the validity and reliability of the CEO response, each CEO is asked to let another senior manager fill in the survey. This resulted in 41 responses, representing 15,41 percent of the final sample of 267.

Measurements

All of the study’s constructs are measured on a seven-point likert-scale ranging from (1) “strongly disagree” to (7) “strongly agree”. All items are included in the appendix at the end of this study.

Organizational ambidexterity. Prior research of Lubatkin et al, (2006) suggests to

develop a measure for organizational ambidexterity in two steps. First, we used a scale based on exploration and exploitation as measured by Jansen et al, (2006). Second, exploration and exploitation are added together, resulting in a measure for organizational ambidexterity as a whole (Cao et al, 2009; Gupta, Smith, & Shalley, 2006).

27 To operationalize the separate scales in the first step, exploratory factor analysis is carried out. The outcomes indicated that we had to drop one item of the exploration scale, this resulted in a six-item scale for exploration (α = .87). This scale indicates the extent to which firms depart from current knowledge and search for radical innovations for new clients or emerging markets. For the exploitation scale, exploratory factor analysis indicated we had to remove 2 items. This resulted in a six-item scale (α = .79). This scale indicates the extent to which firms build on current knowledge and pursue incremental innovations and are in search of incremental innovations that meet the need of pre-existing clients and markets (Benner & Tushman, 2003; Smith & Tushman, 2005). A sub sequential joint exploratory factor analysis showed the contemplated 2-factor structure in which each item loaded clearly on their intended factor. Furthermore, all factor loadings were above .56 with cross loadings under .30 and both factors have eigenvalues greater than one.

In the past literature, ambidexterity as a whole has been operationalized in multiple ways: as the multiplication of exploration and exploitation (Cao et al, 2009; Gibson & Birkinshaw 2004; He & Wong 2004), the absolute difference between exploration and exploitation (Cao et al, 2009; He & Wong, 2004), and the sum of exploration and exploitation (Jansen et al, 2009; Lubatkin et al, 2006). In consonance with Jansen et al, (2009) and others, we followed the conduct as suggested by Edwards (1994) to choose the most interpretable approach for the measurement of ambidexterity. First, we ran an unconstrained regression analysis with an subjective performance scale. For this we took corporate venturing as dependent and exploration and exploitation as independent variables (R2=.28). Then, we mean centered the exploitation and

exploration scaled before computing them as a whole to mitigate possible multicollinearity. Following this, we ran three constrained regression analysis. One in which exploitation and exploration where multiplied (R2=.20), one in which exploitation was subtracted from

exploration (R2=.086) and one in which they were added together (R2=.23). Based on this, the

additive model turned out to be superior compared to the other two models. An F-test was carried out to test whether the additive model showed a significant loss of information compared to the unconstrained model, outcomes indicate that this is not the case. As a final check of the convergent validity of our measurement of ambidexterity, we compared the fit of ambidexterity operationalized as a second-order factor structure to a model in which exploration and exploitation are treated as independent variables using AMOS 20. The first proposed factor

28 structure generated the closest fit to the data (χ²= 159,654, DF=37, CFI=0.941, RMSEA=0.085) compared to the alternative in which ambidexterity was specified as a multiplied model

(χ²=140.384, DF=44, CFI=0,612, RMSEA=0,091). Accordingly, ambidexterity is

operationalized as a second-order construct.

Top management team reflexivity. This study uses the scale developed by Schippers et al,

(2007) that is based on the items of Carter and West (1998). Information is obtained about the extent to which top management teams reflect upon and modify their functioning. The scale demonstrated high reliability (α=.93) and exploratory factor analysis showed high factor loadings (all factor loadings above .70). Confirmatory factor analysis provided further confirmation of the construct validity of reflexivity (χ²=45.064, DF=6, CFI=0,992, RMSEA=0,082).

Top management team paradoxical cognition. We are not aware of any published scales

for top management team paradoxical cognition. Therefore, a new scale is developed for which five academic experts and two corporate members where consulted. This scale is based on the work of Smith and Tushman (2005) and measures the degree to which differentiation and integration at a cognitive level occurs in top management teams. For differentiation we asked members of the top management team to what extent they contrast between existing and new products. For integration we asked to what extent they identify synergies between existing and new products. An exploratory factor analysis led to the exclusion of two items from the pre-existing survey, resulting in a 4-item scale (α = .84). An additional CFA revealed a close fitting model (χ²=8.54, DF=1, CFI=1.000, RMSEA= 0.015) further providing evidence of construct validity of the scale.

Benchmarking. An unawareness of a published scale that measures benchmarking led to

the development of a new scale for which the same academic experts corporate members were consulted. A scale was developed that measures the extent to what the top management team compares their organization with other organizations. This resulted in a three-item scale. However, an exploratory factor analysis indicated we had to drop one item of the benchmark scale. The resulting 2-item scale shows high reliability (α = .86). It was not possible to carry out an additional CFA on this two-factor scale.

LLMT attention. We are also not aware of any pre-existing, published scales for LLMT

attention (α = .88). Therefore, a new four-item scale is developed for which the same academic experts and corporate members were consulted. The scale measures the extent to which a TMT

29 interacts with and discusses certain topics with employees from lower levels of the organization. Confirmatory factor analysis provided further confirmation of the construct validity of the scale (χ²=9.589, DF=1, CFI=1,000, RMSEA=0,000).

Control variables. This study controls for several alternative explanations. It is controlled

for organizational size, as larger organizations may be better able to achieve ambidexterity than smaller ones. This is for the reason that they may have more resources to pursue exploration and exploitation simultaneously (Ahuja & Lampert, 2001). The number of fulltime employees measured as natural logarithm is taken as a control variable to account for organizational size. This study also controls for organizational age, as incumbent organizations are naturally more inclined to engage in exploitation (Gilbert, 2005). Therefore we added the natural logarithm of the years since founding to control for organizational age. This study controls for slack

resources, as Voss and Voss (2013) showed that slack resources allow organizations to incline

entirely towards explorative or exploitative efforts. Furthermore, Ashill and Jobber (2010), found that a careful balance between exploration and exploitation may be more challenging to sustain in uncertain environments. Therefore, we control for state uncertainty. Lastly, we included a dummy variable for the country. This is for the reason that differences between countries could influence the outcomes (Schwarts, 1999).

Aggregation and Measurement Analysis

Before aggregating the data, we examined the intra-class correlation coefficient (ICC2) for the data on the main variables (Lebreton & Senter, 2007). The intra-class correlations showed a strong level of inter rater reliability as scores above 0,7 indicate a strong level of inter rater reliability (Jones, Johnson, Butler, & Main, 1983). On average, scores were 0,93 for reflexivity, 0,84 for paradoxical cognition, 0,86 for benchmarking, 0,88 for LLMT attention, 0,87 for exploration, and 0,77 for exploitation.

To provide evidence for the measurements, a factor analysis using a principal component

analysis with varimax rotation was used (as discussed above). The items had to meet the following requirements: (1), commonality > 0,3, (2), dominant loading greater than 0,5, (3), cross-loading lower than 0,3. Based on this, some items were removed from the study but all remaining items met these criteria.