In the Wake of Fukushima: Unpacking the Impact of an Exogenous Shock on the Exploration-Exploitation Balance

(1)

Master Thesis ∙ February 18th_{, 2019 ∙ JEL: O31, O33, Q40 ∙ Word Count: 14.782}

Joris Jansen ∙ S3272052 ∙ Supervised by dr. P.J. Steinberg ∙ Co-assessed by dr. Q. Dong

In the Wake of Fukushima: Unpacking the Impact of an Exogenous Shock

on the Exploration-Exploitation Balance

Joris Jansen𝐚,∗ | Supervised by dr. P.J. Steinberg𝐚 | Co-assessed by dr. Q. Dong𝐚 a_{| Faculty of Economics and Business | University of Groningen | The Netherlands |}

∗_{| Corresponding author | MSc BA Strategic Innovation Management |}

| Student Nr: S3272052 | Email: j.w.b.jansen@student.rug.nl |

Keywords: Exploration ∙ Exploitation ∙ Organizational Ambidexterity ∙ Organizational Learning ∙ Organizational Adaptation ∙ Exogenous Shocks

Abstract: Viewing exploration and exploitation through the lens of

organizational learning theory, this research investigates the impact of an exogenous shock on the exploration-exploitation balance and traces the strategic behaviour of ambidextrous firms. The hypotheses were tested on a longitudinal dataset over a time period from 2006 to 2014 using a sample of 84 electric utilities whereby the Fukushima nuclear disaster is considered as an exogenous shock posing challenges and opportunities in the energy sector. I used computer-aided text analysis to measure exploration and exploitation and deployed a Difference-in-Differences methodology to observe differences between nuclear and green firms. Despite not finding direct support for the hypotheses, additional analyses suggest that firms intensify their exploration activities at the cost of levels of exploitation in response to an exogenous shock that threatens existing knowledge and business models. I further find empirical evidence for the self-reinforcing effect that is associated with deploying an ambidextrous strategy which in consequence alleviates response behaviour. My research thereby contributes to the understanding of the environmental antecedents of exploration and exploitation, organizational ambidexterity literature, and the advancements in the field of computer-aided text analysis.

“The greatest danger in times of turbulence is not turbulence itself, but to act with yesterday’s logic.”

- Peter Drucker

(2)

List of Figures and Tables

Figure 1: The Paradoxical Association between Exploration and Exploitation……….8

Figure 2: Conceptual Model………..…..14

Figure 3: Interaction Plot………...28

Table 1: Word-lists Exploration and Exploitation Applied in CATA………..19

Table 2: Descriptive Statistics………...24

Table 3: Correlations………...24

Table 4: Regression Results Related to Hypothesis 1 and Hypothesis 2………..26

Figure A.1: Nuclear and Clean Energy Indices Values Pre-Fukushima and Post-Fukushima…………..….48

Figure A.2: Global Projections for Nuclear Power (2008 – 2016) ………....…48

Figure A.3: Global Nuclear Electricity Production (1970 – 2015) ………..…..…48

Figure D.1: Common-Trend Visualization………...……...53

Table A.1: Overview of Countries and Sample………..…….47

Table A.2: Responses of Nuclear Firms to the Fukushima Nuclear Disaster……….…………....49

Table A.3: Responses of Green Firms to the Fukushima Nuclear Disaster………..………50

Table B.1: Clean Energy Indices………...51

Table C.1: Description of Variables………...…...52

Table E.2: Variance Inflation Factors………...54

Table F.1: Post-hoc and Robustness Checks Related to Hypothesis 1………..……..….55

Table F.2: Robustness Checks Related to Hypothesis 2……….….56

List of Abbreviations and Acronyms

CATA : Computer-Aided Text Analysis

DiD : Difference-in-Differences

IAEA : International Atomic Energy Agency IEA : International Energy Agency

ISIN : International Securities Identification Number MD&A : Management Discussion & Analysis

NACE : Statistical Classification of Economic Activities in the European Community NEA : Nuclear Energy Agency

OA : Organizational Ambidexterity

OECD : Organisation for Economic Co-operation and Development OLS : Ordinary Least Squares

R&D : Research and Development ROA : Return on Assets

UN : United Nations

(4)

1. Introduction

Today’s global environment is marked by an increasingly complex, dynamic and uncertain nature, where firms are faced with challenges imposed by exogenous shocks, regulatory shifts, fierce global competition, and disruptive innovations (Crossan et al., 2008; D’Aveni, 1994; Sirmon et al., 2007). In light of these developments, shifting paradigms continually confront existing business models and technologies (Ansari & Krop, 2012). As such, Foster & Kaplan (2001, p. 15) remarked that “the

assumption of continuity, on which most of our leading corporations have been based for years, no longer holds. Discontinuity dominates.” Research henceforth particularly points to the prevalence of

organizational failure when the nature of change is discontinuous instead of incremental, which challenges firms to reconfigure and rethink their current way of doing business and causes uncertainties for success (Birkinshaw et al., 2016). Organizational scholars posit that firms survive under such circumstances by reconfiguring their assets and knowledge base (Eisenhardt & Martin, 2000; O’Reilly & Tushman, 2008; Teece et al., 1997). Accordingly, organizational learning has been proposed as a fundamental element to provide and sustain competitive advantage (de Geus, 1988; Stata, 1989).

In this regard, March (1991) argues that the fundamental adaptive challenge is to exploit existing assets and capabilities whilst engaging in sufficient exploration of new possibilities to maintain relevant in changing environmental conditions. Prior research hence suggests that engaging in both explorative and exploitative processes is crucial to survive, prosper, and succeed in the long-term (e.g. Gibson & Birkinshaw, 2004; He & Wong, 2004; March, 1991; O’Reilly & Tushman, 2008). Developed in the context of organizational learning, March (1991, p. 85) described that “the essence of exploitation is the

refinement and extension of existing competences, technologies, and paradigms. Its returns are positive, proximate, and predictable. At the other end of the spectrum, March (1991, p. 85) noted that “the essence of exploration is experimentation with new alternatives. Its returns are uncertain, distant, and often negative”. Since March (1991) established these dual concepts in the literature they have become

a dominant theme in organizational analyses of organizational learning, organizational adaptation and survival, technological innovation, organization design, and competitive advantage (e.g. Benner & Tushman, 2003; Gupta et al., 2006; Holmqvist, 2004; Katila & Ahuja, 2002; McGrath, 2001; Siggelkow & Levinthal, 2003). Despite the wide recognition in the literature with respect to the merits of engaging in both learning mechanisms, March (1991) argued that exploration and exploitation compete for scarce resources and are in fact, mutually exclusive activities. In contrast to earlier research regarding the trade-off as inevitable, recent research proposed several modes enabling firms to operate in the middle of the exploration-exploitation continuum (e.g. O’Reilly & Tushman, 2013). The simultaneous pursuit of exploration and exploitation is referred to as organizational ambidexterity (OA; Gupta et al., 2006; Simsek, 2009; Uotila, 2017). For instance, ambidextrous firms harmonize conflicting demands of exploration and exploitation through organizational separation (Benner & Tushman, 2003).

(5)

various scholars call for empirical evidence and longitudinal perspectives on how firms adapt the balance between exploration and exploitation in the face of environmental change (Raisch et al., 2009; Koryak et al., 2018; Luger et al., 2018). In light of this, a nascent research stream has yet highlighted the role of environmental dynamism in the context of exploration and exploitation (e.g. Jansen et al., 2005; Sidhu et al., 2004). Sidhu et al. (2004) report for instance that dynamic environments induce the pursuit of greater exploration. Notwithstanding, notably absent in the literature is the role of exogenous shocks (Lavie et al., 2010). Whereas environmental dynamism entails a certain degree of predictability (Lavie et al., 2010), exogenous shocks exert a disruptive impact and refer to sudden and unexpected environmental jolts outside the limits of the organization’s control (Meyer, 1982). Further research is needed that draws insights from realistic environmental conditions to understand the firms’ tendencies to explore versus exploit in the face of an exogenous shock (Lavie et al., 2010).

Drawing on these gaps in the literature, I aim to depart from preceding research by studying exploration and exploitation over time while introducing a realistic exogenous shock. In line with this goal, this research is guided by the following research question:

RQ1: How do firms adapt the exploration-exploitation balance in response to an exogenous shock?

To provide a richer understanding this research will additionally explore the role of OA in moderating the relationship between an exogenous shock and the exploration-exploitation balance. Research has emphasized the contingent role of strategic orientations in examining the effect of the environment on decisions between exploration and exploitation (Auh & Menguc, 2005; Chattopadhyay et al., 2001; Voss et al., 2008). Whereas former research mainly considered the strategies of firms deploying a one-sided focus, this research traces the actions of firms following an ambidextrous strategy, which thus far has remained rather elusive in these discussions. To examine this, I formulate a secondary research question:

RQ2: How does organizational ambidexterity moderate the relationship between an exogenous shock

and the exploration-exploitation balance?

(6)

the self-reinforcing nature that is associated with deploying an ambidextrous strategy, which in consequence, alleviates response behaviour.

The contributions of this research are thereby fourfold. First, it adds the effects of an exogenous shock to the emergent dialogue on environmental antecedents of exploration and exploitation. Second, it contributes to the burgeoning literature on OA by tracing the strategic behaviour of ambidextrous firms. Third, this research constitutes methodological contributions by providing insights into applying computer-aided text analysis (CATA). Fourth, it offers practical insights that may assist managers when confronted with unanticipated shocks and provide policymakers with insights to incentive innovation.

The remainder of this study is structured as follows. Section 2 provides a brief overview of the core theory underlying organizational learning and digs deeper into the notion of exploration, exploitation, and OA, acting as the foundation for hypotheses development. Section 3 outlines the research methodology used in this study. Section 4 presents the empirical results, as well as the post-hoc -and robustness analyses. Section 5 concludes by discussing the findings, the implications for research and practice, and finally the limitations and suggestions for future research.

2. Theoretical Background and Hypotheses

To investigate how firms deal with sudden disruption in their business environment, this research integrates perspectives of organizational learning and March’s (1991) framework of exploration and exploitation. I first root my research in the theory of organizational learning, to further elaborate on the concepts of exploration, exploitation, and OA, leading to the development of my hypotheses.

2.1 Organizational Learning

(7)

ones (March, 1991; Levinthal & March, 1993).1_{In this research, I view organizations as boundedly}

rational actors engaged in a process of search and learning (Denrell & March, 2001) and posit that the learning components previously discussed provide insights in patterns of exploration and exploitation.

2.2 Exploration and Exploitation in Organizational Learning

In the pioneering work of James March (1991), the concepts of exploration and exploitation were introduced, where organizations are perceived as adaptive systems. Grounded in organizational learning, March (1991) distinguished two patterns of learning behaviours, whereas “exploration includes things

captured by terms such as search, variation, risk-taking, experimentation, play, flexibility, discovery, innovation” (March, 1991, p. 71). Exploitation activities include “such things as refinement, choice, production, efficiency, selection, implementation, execution” (March, 1991, p. 71). Levinthal and March

(1993) augmented this by stating that exploration involves "a pursuit of new knowledge," (p. 105) whereas exploitation involves "the use and development of things already known" (p. 105). Exploratory actions build on broad and diverse knowledge that involves searching for novel knowledge in domains relatively distant from the firm’s core. Exploitative actions build on deep and familiar knowledge, and leverages the firm’s existing knowledge base (Baum et al., 2000; Benner & Tushman, 2003; March, 1991; Rosenkopf & Nerkar, 2001; Lavie et al., 2010).

The Quest for Balance

It has long been recognized that engaging in both exploitative and explorative processes is crucial for success and long-term survival of firms (Gibson & Birkinshaw, 2004; March, 1991; O’Reilly & Tushman, 2008). Firms deploying exploitation-oriented strategies focus on current processes and conform to existing demands, and thereby increase the chance of short-term success. Firms may however thereby reduce the firm’s capacity to respond radically to future changes and new opportunities (Benner & Tushman, 2003; March, 1991), whereby they trade-off immediate reliability while risking obsolescence in the future (Holmqvist, 2004; Leonard-Barton, 1992). In contrast, exploration activities increase the odds of long-term survival at the cost of certain returns, efficiency and refinement of existing capabilities. They trade-off short-term productivity for long-term innovation. Focusing on exploration, while excluding activities of exploitation, implies risking the costs of experimentation without reaping its benefits. On the contrary, focusing on exploitation while excluding explorative activities are likely to trap themselves in suboptimal stable equilibria (Benner & Tushman, 2003; March, 1991). As such, Levinthal and March (1993, p. 105) noted that “the basic problem confronting an

organization is to engage in sufficient exploitation to ensure its current viability and, at the same time, to devote enough energy to exploration to ensure its future viability.” An appropriate balance between

the two learning strategies is therefore a crucial factor for firm survival and prosperity (March, 1991; Levinthal & March, 1993).

Recent research acknowledges that a balanced strategy yields increased benefits in contrast to the exclusive commitment to either one strategy (e.g. He & Wong, 2004; Lavie et al., 2010; Lavie &

1_{Other scholars have explored organizational learning describing similar types of learning: double loop vs. single loop learning}

(8)

Rosenkopf, 2006; O’Reilly & Tushman, 2013). Combining these dual learning strategies is however not without caveats. Firms are strongly biased towards exploitation due to its certainty, short-term success and beneficial feedback from pursuing and augmenting current processes and competences (Gupta et al., 2006; Benner & Tushman, 2003; Levinthal & March, 1993). Moreover, they form a paradoxical relationship (March, 1991), which implies that exploration and exploitation are “contradictory yet

interrelated elements that exist simultaneously and persist over time” (Smith & Lewis, 2011, p. 382). A Paradoxical Association

Despite the broad recognition in literature of the merits of balance, March (1991, 1996, 2006) consistently argued that the two strategies are incompatible. Central to his framework is the argument that maintaining an appropriate balance between exploration and exploitation is largely complicated by trade-offs due to constraints in resource allocation, attention, and organizational routines. Exploration and exploitation compete for scarce resources, implying that resources devoted to exploration results in fewer resources available to exploitation, and vice versa. Furthermore, the two activities are self-reinforcing in a sense that exploration often leads to failure, thereby spurring the search for more novel ideas resulting in a ‘failure trap’. In contrast, exploitation may lead to a ‘success trap’ as exploitation is associated with short-term success and thereby it reinforces itself along the exploitation trajectory. Additional tensions arise due to the diverging mind-sets and organizational routines needed for the contradictory activities. Thus, however March (1991, 1996, 2006) maintains clear in his argumentation that exploration and exploitation are fundamental for long-run adaptation, the simultaneous pursuit seems rather complicated and in fact, points to a zero-sum game. Logic hence dictates that exploration and exploitation lie on two opposite ends of the same continuum (Gupta et al., 2006; Lavie et al., 2010).

More recent research recognizes the paradoxical association between exploration and exploitation (e.g. Andriopoulos & Lewis, 2009; Smith & Lewis, 2011; Koryak et al., 2018). The activities of exploration and exploitation require fundamentally different structures, processes, strategies and capabilities. While exploration has been related to flexibility, decentralization, and loose cultures, exploitation has been linked to efficiency, centralization, and tight cultures (Benner & Tushman, 2003; Chang et al., 2009; McGrath, 2001; Siggelkow & Levinthal, 2003). As they draw on different structures and processes, firms need to choose on which they emphasize in the allocation of their resources (Christensen & Overdorf, 2000; Gupta et al., 2006). Firms that choose for exploitation trade flexibility for stability and thereby foster structural inertia, limiting their ability to adapt to environmental threats (Hannan & Freeman, 1984). Organizations that are largely committed to exploitation have difficulties to engage in exploration activities, and vice versa (Sørensen & Stuart, 2000). Despite their positive relationship when they co-occur over time, they still generate tensions (Lavie et al., 2010; Lubatkin et al., 2006; Smith & Lewis, 2011). Figure 1 depicts the paradoxical association between exploration and exploitation (Lavie et al., 2010).

Figure 1: The Paradoxical Association between Exploration and Exploitation (Lavie et al., 2010)

(9)

In sum, it has been made clear that engaging in both learning strategies is crucial. Yet, exploration and exploitation create tensions and firms need to manage trade-offs between them. But how do firms manage these inherently contradicting activities? Earlier studies regarded the balance dilemma as insurmountable, however, more recent research suggests several solutions to reconcile demands of exploration and exploitation simultaneously (e.g. O’Reilly & Tushman, 2013; Raisch & Birkinshaw, 2008; Simsek, 2009).

2.3 Balancing Modes: Organizational Ambidexterity and Punctuated Equilibrium

Ambidexterity, the ability of humans to use both hands with equal skill, is increasingly used as a metaphor to refer to the capability of a firm to explore and exploit with equal dexterity (Simsek, 2009). While Duncan (1976) was the first to introduce the idea of the ambidextrous organization, Tushman & O’Reilly (1996) advanced a theory by proposing dual structures for exploration and exploitation as an enabler for OA. In light of the growing research on this topic, scholars have conceptualized and defined OA in different ways.2_{Nonetheless, there is a consensus that OA is in some way related to the}

simultaneous pursuit of exploration and exploitation (Cao et al., 2009; Lavie et al., 2010). Despite the diverging conceptualizations and measurements of OA, its positive relationship to performance is robust (Junni et al., 2013; O’Reilly & Tushman, 2013). Studies have demonstrated for example that OA contributes to the survivability of the firm (Piao, 2010; Cottrell & Nault, 2004) and significantly improves performance. It is positively related to sales growth (Auh & Menguc, 2005; He & Wong, 2004), innovation (Jansen et al., 2009; Katila & Ahuja, 2002), and subjective ratings of performance (Cao et al., 2009; Gibson & Birkinshaw, 2004; Lubatkin et al., 2009).

Research has identified several means by which firms resolve the inherent tensions between exploration and exploitation. First, the most studied and practiced mode is structural ambidexterity (Stettner & Lavie, 2014), referring to the creation of dual structures by making use of dedicated teams or organizational units for exploration and exploitation separately (O’Reilly & Tushman, 2004; Benner & Tushman, 2003; Raisch et al., 2009). Second, Gibson & Birkinshaw (2004) proposed the concept of

contextual ambidexterity, which describes the supportive behavioural context achieved by designing the

appropriate systems, processes, and culture that enables individuals to make judgements about the division of their time to manage conflicting demands of alignment and adaptability. Third, domain

separation refers to harmonizing exploration and exploitation activities across different domains (e.g.

through strategic alliances), thereby enabling organizations to avoid trade-offs in each single domain (Lavie & Rosenkopf, 2006; Stettner & Lavie, 2014).

An alternative approach to the quest of balancing exploration and exploitation is rooted in the theory of punctuated equilibrium (Tushman & Romanelli, 1985). Gupta et al. (2006, p. 698) define punctuated equilibrium3 as “temporal cycling between long periods of exploitation and short bursts of

exploration.” In more practical terms, Mudambi & Swift (2011, 2014) for instance demonstrate how

firms undertake sequential moves between exploration and exploitation by pro-actively managing R&D expenses. Important to note is that punctuated equilibrium is conceptually distinct from OA, as it does not refer to the simultaneous pursuit of exploration and exploitation, but rather denotes patterns of

2_{OA has for instance been defined as: “the ability to simultaneously pursue both incremental and discontinuous innovation”}

(Tushman & O’Reilly, 1996, p. 24). Or, Lubatkin et al. (2006, p.2) describe OA as being “capable of exploiting existing

competencies as well as exploring new opportunities with equal dexterity.’’

3_{Scholars also refer to sequential ambidexterity (O’Reilly & Tushman, 2013), or organizational vacillation (Boumgarden et}

(10)

temporal cycling between them (Gupta et al., 2006; Uotila, 2017). In accordance with this distinction, I follow prior research and define OA as the simultaneous pursuit of exploration and exploitation (Gupta et al., 2006; Simsek, 2009; Uotila, 2017). Thereby in particular, viewing the ambidextrous organization as maintaining high levels of balance along the exploration-exploitation continuum (Simsek, 2009). I continue with elaborating on how not only firm activities relate to punctuated equilibrium, but also how environmental shocks can account for similar patterns associated with this phenomenon.

2.4 Hypotheses Development

2.4.1 Organizational Learning in the Face of Environmental Change

The topic of how firms adapt to changes in their external environment is an ongoing topic of interest in the business literature (Birkinshaw et al., 2016). Shifting environmental conditions entails conforming alternations to the exploration-exploitation balance (Auh & Menguc, 2005). Although the previous section has described several modes that enable firms to reconcile conflicting demands of exploration and exploitation, their coexistence does not negate the inherent trade-offs between them (Lavie et al., 2010). They lie on the same continuum, implying that firms must choose on which they emphasize in the face of environmental change.

Closely related to this study is the line of research that investigates exploration and exploitation in the context of dynamic environments (Jansen et al., 2005; Sidhu et al., 2004; Posen & Levinthal, 2012). Environmental dynamism is typically defined as the degree of unpredictability and the rate of environmental change (Dess & Beard, 1984). The preceding research demonstrates that firms are inclined to focus more on an exploration orientation in dynamic environments. Because dynamic environments result in the obsolescence of current processes and products, they require the development of new ones (Jansen et al., 2005; Sørensen & Stuart, 2000). Sidhu et al. (2004), suggest that firms reduce uncertainty by expanding the scope of information acquisition to develop novel approaches and adapt to external developments. In addition, Posen & Levinthal (2012) demonstrate in a simulation study4_that

an action biases arises when moving from a stable to a dynamic environment, causing existing beliefs within the organization to be undermined and thereby resulting in a higher probability to explore (versus exploit). This bias is exaggerated when the environment is munificence-reducing. On the contrary, munificence-enhancing environments are subject to an in-action bias, thereby promoting actions in the area of exploitation.5

Although this preceding research has provided relevant insights into how firms respond to changes in their environments, exogenous shocks are different in the dimension of predictability and magnitude (Lavie et al., 2010). Exogenous shocks exert a disruptive impact and refer to sudden and unexpected environmental jolts outside the limits of the organization’s control (Meyer, 1982). In contrast to environmental dynamism, exogenous shocks are prompted by unanticipated events such as regulatory shifts and call for direct organizational response (Meyer et al., 1990). According to Li & Tallman (2011), exogenous shocks can be conceptualized by leveraging frameworks of disruptive change. The notion of punctuated equilibrium describes a process of long periods of industry stability that is interrupted with challenging periods of disruptive change (Miller & Friesen, 1984; Tushman &

4_{Posen & Levinthal (2012) applied the bandit-model and refer to frequency and unpredictability of environmental change.}

They model the level of turbulence as the probability of a shock to the payoff values of alternatives.

5_{Munificence-enhancing implies an increase to the expected returns to the set of alternatives available to organizations in the}

(11)

Romanelli, 1985). Even weak forces may trigger such change by releasing the accrued pressure of prior political, technological, and social changes (Meyer et al., 1990).

In line with this idea, firms experience enduring periods of ‘convergence’ punctuated with periods of ‘reorientation’. From the organizational perspective, convergence is described as “relatively long

time spans of incremental change and adaptation which elaborate structures, systems, controls, and resources toward increased coalignment" (Tushman & Romanelli 1985, p. 173). Reorientations are

characterized by "simultaneous and discontinuous shifts in strategy, the distribution of power, the firm's

core structure, and the nature and permissiveness of control systems" (Tushman & Romanelli, 1985, p.

179). According to Li & Tallman (2011), exogenous shocks can account for reorienting effects as they require adaptation and the revision of strategies, structures, and controls to fit with a sudden change in the environment.

In the face of such disruptive forces, scholars of punctuated equilibrium have suggested that firms survive by engaging in organization-wide change instead of undertaking continuous or incremental changes(Miller & Friesen, 1984; Romanelli & Tushman, 1994). In turn, research has suggested that moving from exploitation, where firms build on existing knowledge and competences, to exploration, where firms discover novel resources to renew competitive positions, is a form of such profound organizational change (Benner & Tushman, 2003; Gupta et al., 2006; Holmqvist, 2004; Katila & Ahuja, 2002; McGrath, 2001). According to Mudambi & Swift (2011), this suggests that during periods of stability, firms focus on exploitation, and move to exploration during periods of punctuated change. Synthesizing these perspectives, an exogenous shock is likely to trigger exploration activities as it places a novel demand on firms, while on the contrary, firms seem to be more focused towards exploitation during periods of stability. But what are the underlying mechanisms triggering these activities?

March (1991, p. 80) concluded that “exogenous environmental change makes adaptation

essential,” but he also alleviated this conclusion by remarking that it also “makes learning from experience difficult.” He continued by stating that knowledge degenerates under conditions of

exogenous turbulence (March, 1991). In this regard, as exploitation leverages the firm’s existing knowledge base (Baum et al., 2000; Benner & Tushman, 2003; Levinthal & March, 1993), logic dictates that exploration of novel knowledge is needed when an exogenous shock threatens the existing knowledge base of a firm. Nonetheless, deeper insights of organizational learning are required to obtain a more nuanced understanding of what steers firms to either exploration or exploitation.

(12)

domains to explore for more beneficial alternatives. On the contrary, success narrows search down close to the status quo (Billinger et al., 2016; Greve, 2003; Madsen & Desai, 2010; Maslach, 2016).

Considering that the majority of firms’ activities are routine driven, firms enact the environment in line with the functioning of their routines (March, 1981). Applying routine-responses to changing environmental conditions leads to unsatisfying performance (Lant & Mezias, 1992). As a consequence, the organization’s current routines, practices, and processes are pulled into doubt (Levitt & March, 1988), thereby making consequent change more likely (March & Simon, 1958). In light of this, March (1981, p. 564) pointed out: "Most change in organizations results neither from extraordinary

organizational processes nor forces, nor from uncommon imagination, persistence or skill, but from relatively stable, routine processes that relate organizations to their environments." Building on this

line of argumentation, Lant & Mezias (1992) proposed that environmental change increases the likelihood of firms experimenting and exploring for alternative routines, rules and technologies to boost performance, instead of amplifying current routines. Because exploitation essentially implies refining and extending existing competences, technologies and paradigms, and exploration essentially implies experimentation with new alternatives (March, 1991), these arguments support the idea that environmental stability and success reinforce exploitation activities, and environmental change related to failure induce exploration activities.

In addition, as the environment changes, there might be a gap between what the organizational has learned to do and what it needs to do. Accordingly, the institutionalized learning may no longer fit the context (Crossan et al., 1999). As a result of the observation that current belief systems or theories in use do not suffice to interpret certain experiences, exploratory learning may be triggered, as it requires a change of the organization’s core assumptions (Argyris & Schön, 1978). Such a break from the past may, for example, involve experimentation with alternative decision-making and the creation of a diverse experience base that nurtures new understandings (Fiol & Lyles, 1985; Virany et al., 1992). Broad and diverse knowledge in turn, form the foundation for exploratory actions (Baum et al., 2000; Benner & Tushman, 2003; March, 1991). In a similar vein, firms engage in learning from other’s experiences when own experiences are inadequate to deal with novel challenges or opportunities (March 1991, Miner & Haunschild, 1995). According to Baum & Dahlin (2007) this more exploratory learning mode entails learning diverse and new practices from other firms and is also likely to be prevalent in the occurrence of failure. Firms intensify search activities to learn from the failure of similar organizations to prevent harmful outcomes for themselves (Baum & Dahlin, 2007).

(13)

that are believed to be affected differently (negatively affected and positively affected or unaffected)6_,

which leads to the following hypothesis:

Hypothesis 1: In response to an exogenous shock, there will be a difference in strategic

behaviour with respect to the relative exploration orientation of firms, such that firms that are negatively affected by the shock will intensify exploration (versus exploitation) levels, and firms that are positively or not affected by the shock will decrease or maintain exploration (versus exploitation) levels

2.4.2 The Moderating Effect of Organizational Ambidexterity

The effects of environmental threats and opportunities on the exploration-exploitation balance may be contingent on a firm’s strategic orientation (Auh & Menguc, 2005; Chattopadhyay et al., 2001; Voss et al., 2008). The ambidextrous strategy may therefore influence strategic decision-making in allocating scarce resources and attention to exploration and exploitation in the face of an exogenous shock.

While the argumentation of March (1991) points to a zero-sum game if the two-learning strategies are combined, scholars have argued that firms following an ambidextrous strategy can deal with the exploration-exploitation tension and create synergies that promote complementary returns (Smith & Lewis, 2011; Jansen et al., 2006). As has been pointed out, actions that are successful tend to be repeated (e.g. Lant & Mezias, 1992; Levitt & March, 1988). In considering the robust relationship of OA with performance (Junni et al., 2013), it seems highly likely that firms associate a balanced allocation with success. From an organizational learning perspective firms may therefore be less motivated to promote changes to this balance. Even in the face of a changing environment that may decrease a firm’s coalignment returns, the complementary returns could mask or even compensate for the performance gap, which thereby motivates the preservation of balance (Luger et al., 2018).

In addition, scholars have argued that OA is not simply a short-term decision that can be taken at any time. Rather, it demands a long-term vision (O’Reilly & Tushman, 2008, 2011), and enduring resource commitment to achieve it (Lavie et al., 2010). Organizational vision represents a mental image of the future of the firm and thereby guides actions (Ruvio et al., 2010; Koryak et al., 2018). The implementation of an ambidextrous vision demands fundamental changes to the organization in terms of organizational architecture, strategy, and culture (Gibson & Birkinshaw, 2004), and requires a unity of purpose across organizational members (Jansen et al., 2006; Lubatkin et al., 2006; Sidhu et al., 2004). Inherent to such a long-term vision is that managers learn to resist short-term pressures and thus strive to preserve the exploration-exploitation balance (O’Reilly & Tushman, 2011). Similarly, as OA constitutes a source of competitive advantage and can be regarded as a rare, valuable, and difficult to imitate capability that requires a significant amount of effort and time to build, managers want to protect it and act as sentinels of the ambidextrous vision (O’Reilly & Tushman, 2008). Moving away from this balance would imply a contradicting decision that does not reflect the firm’s strategy, vision, and established structures (Lavie et al., 2010).

Furthermore, the self-reinforcing nature of learning makes it appealing for organizations to amplify its current focus (Levinthal & March, 1993). Whereas exploration and exploitation are characterized by a self-reinforcing nature (March, 1991, Leonard-Barton, 1992), ambidextrous firms

6_{This research studies two sets of firms whereby: (1) nuclear firms are treated as being negatively affected by the shock; (2)}

(14)

with a balanced allocation on the exploration-exploitation continuum similarly are prone self-reinforcing effects (Luger et al., 2018). Moreover, investments in the overall structure creates stability and continuity in actions, which thereby gives rise to organizational inertia (Hannan & Freeman, 1984). As the organizational arrangements of ambidextrous firms are substantial (Gibson & Birkinshaw, 2004; Lavie et al, 2010), they are likely to promote continuity in stabilizing the balanced resource-allocation (Luger et al., 2018).

Concluding, just as exploration-oriented firms and exploitation-oriented firms may have the tendency to increasingly move to either end on the exploration-exploitation continuum, firms with a higher balanced allocation are more inclined to gradually reinforce their balance, they become ‘stuck in the middle’. This self-reinforcing effect is explicated by its association with success, protective managerial actions, and inertial tendencies. Firms following an ambidextrous strategy are therefore expected to be less likely to give up this balance in the face of short-term pressures as exerted by exogenous shocks. I therefore hypothesize the following:

Hypothesis 2: The degree of influence of the exogenous shock on the relative exploration

(versus exploitation) is moderated by high levels of organizational ambidexterity, such that it alleviates response behaviour

Figure 2 presents the conceptual model, thereby displaying the hypotheses and the variables involved in the analyses. These constructs will be elaborated on in the following methodological section.

Figure 2: Conceptual Model

Exogenous Shock Organizational Ambidexterity Relative Exploration Orientation Control Variables: - Firm Size - Firm Age - Return on Assets - Unabsorbed Slack - R&D Investments - R&D Missing Dummy - Year Dummies - Geo. Group Dummies

(15)

3. Methodology

This section describes and justifies the methodological choices made in this study. First, the empirical context of this research is described. Second, the stepwise process of sample selection and data collection is explained. Third, the variables and corresponding measurement items will be presented. Finally, the analytical method describes the strategy for hypotheses testing and the reliability of this approach.

3.1 Empirical Context

The empirical context of this study is the energy industry, where the 2011 Fukushima nuclear disaster is regarded as an exogenous shock posing significant challenges and opportunities for energy utilities. This disaster resulted in an enormous setback for the nuclear industry worldwide, intensifying discussions regarding the risks of nuclear power and the transition towards alternative, clean energy sources (Ferstl et al., 2012; Wittneben, 2012). Kungl (2015) thereby described Fukushima as an exogenous shock destabilizing the energy sector, marked by high intensity and the power to disrupt a multiplicity of fields including international electricity markets. The nuclear industry may specifically relate to accumulated stress as previously discussed by Meyer et al. (1990) due to previous nuclear accidents at the Three Mile Island (e.g. Bowen et al., 1983), Chernobyl (e.g. Fields & Janjigian, 1989), and the reignition of the international debate on the future of nuclear energy (e.g. Wittneben, 2012; Ming et al., 2016), thereby increasing the likelihood of Fukushima having reorienting effects on nuclear firms. In sum, the Fukushima nuclear disaster provides a unique opportunity to unpack the impact of an exogenous shock on the exploration-exploitation balance (see Appendix A for an overview of the Fukushima Nuclear Disaster).

The particular relevance of the energy industry for this study is further motivated by the following factors: Uotila et al. (2009) explicitly suggested to research areas where there is a potential trade-off between exploration and exploitation; the energy industry is in this regard an interesting setting, as nowadays electric utilities are facing great challenges related to the mitigation of climate change and energy efficiency while having to guarantee energy security (Costa-Campi et al., 2010). As such, incumbent utilities are confronted with business model challenges and must explore new ways of creating, delivering, and capturing value from renewable energy technologies, while simultaneously continuing to exploit conventional power stations to secure a stable supply of electricity (Richter, 2013). Research has in this regard highlighted the importance of OA (Richter, 2013), and innovation in this industry (Anadón, 2012; Costa-Campi et al., 2010).

3.2 Sample and Data Collection

I used a cross-national longitudinal data set (2006 – 2014) of publicly listed electric utilities in the energy industry to test my hypotheses. I applied a stepwise process for sample selection and data collection, as will be explained below.

Country Selection

(16)

Co-operation and Development (OECD), an intergovernmental organization of industrialized countries. These countries account for approximately 82% of the global nuclear energy production and nuclear power is an important source of energy in these countries with about one-fifth of the electricity produced by its utilities (NEA, 2012). The rationale underlying the choice of these members is justified as it results in communality across the sample and reduces structural differences in a cross-national context (Mama & Bassen, 2013). Besides, members of the NEA have a shared knowledge pool and common regulatory guidance (NEA, 2012), which may influence the strategic choices of firms. I further reduced heterogeneity in the sample by removing countries that are not economically developed (e.g. Brasil, Mexico) according to the ‘World Economic Situation and Prospects’ report issued by the United Nations (UN, 2006).

Company Selection

Next, I determined nuclear power operators by investigating the International Atomic Energy Agency's (IAEA) Power Reactor Information System (2011) and examining listing statuses in Orbis. This led to the identification of 52 utilities domestic to their respective market. I excluded 7 utilities due to firms’ delisting or unavailability of appropriate annual reports (either missing or language discrepancies), resulting in a final sample of 45 nuclear utilities within the 35.17_{sector of the Nace Rev. 2 classification}

system. For the green energy utilities,8_{I drew on different sources as the yield per source was relatively}

low after accounting for industry code and data requirements (e.g. annual reports, financial data). I first followed Ferstl et al. (2012), and selected firms that are classified as Alternative Energy in Thomson Reuters Datastream. Next, I selected several indices (see appendix B) based on prior research (Betzer et al., 2013; Reboredo, 2015; Sadorsky, 2012; Ortas & Moneva, 2013). As the Nace Rev. 2 classification does not distinguish between types of utilities I complemented the sample by adding renewable utilities according to the Thomson Reuters Business Classification using Eikon. I compared the industry codes by loading these companies into Orbis and included only utilities within the industry code 35.1 (NACE Rev. 2 Classification) to maintain homogeneity in the sample. In total I identified approximately 150 clean energy utilities but again had to exclude firms due to the lack of appropriate annual reports or firms’ delisting, resulting in a sample of 41 green firms. The final sample consists of 84 publicly listed utilities (see Appendix A for an overview of utilities per country) and contains one group that is expected to be threatened (nuclear firms), and one comparison group that is expected to be positively affected or unaffected (green firms). This distinction of the impact of the shock was based on prior research (Lopatta & Kaspereit, 2014; Ming et al., 2016; NEA, 2017), where additional insights were drawn from firm annual reports (See Table A.2 for nuclear firm responses, and Table A.3 for green firm responses).

Although a larger sample was highly desirable, I did not include firms with a broader range of activities because it gives rise to the limitation of a potential lack of generalizability (Benner, 2009). The sample size is further justified by the inclusion of near to the universe of nuclear utilities and all clean energy utilities that fulfilled the data requirements. Additionally, the focused industry setting, yet internationally oriented, mitigates unobserved effects in my variables that could for example emanate from differences in discretion levels in other sectors (Crossland & Hambrick, 2011).

7_{Electric power generation, transmission and distribution; This classification is comparable to the ‘491’ sector of the Standard}

Industrial Classification (SIC) of the United States.

(17)

Data Collection

In continuation, to collect data I used Orbis and the Thomson Reuters Datastream database which contain comprehensive global firm data such as key financial indicators. To provide flexible time-windows for analysis I collected data ranging in years 2005 – 2015. Unique company identifiers (ticker, ISIN) were used to extract data (firm size, firm age, return on assets, unabsorbed slack, R&D investments) to control for confounding factors. I followed Lopatta & Kaspereit (2014) and retrieved missing datapoints from firm annual reports (the collection of annual reports will be explained in the following section). I subsequently merged all data in Excel to allow for further analyses in Stata. Appendix C presents an overview of the variables and the corresponding measurement items.

3.3 Measurements

3.3.1 Dependent Variable

Exploration – Exploitation. While there are various measures of exploration and exploitation (e.g. He

& Wong, 2004; Lavie & Rosenkopf, 2006), I followed recent research (Heyden et al., 2015; McKenny et al., 2018; Luger et al., 2018), and adapted the approach of Uotila et al. (2009), who were the first to develop a measure of exploration and exploitation using computer-aided-text analysis (CATA). CATA is a form of content analysis that enables the measurement of constructs by converting text into quantitative data based on the frequency of specific words (Short et al., 2010). CATA draws on the assumption that “groups of words reveal underlying themes, and that, for instance, co-occurrences of

keywords can be interpreted as reflecting association between the underlying concepts” (Duriau et al.,

2007, p. 6).

I chose the approach of Uotila et al. (2009) for various reasons. Scholars have noted that the diverging conceptualizations and measurements of exploration and exploitation have led to inconsistent findings and limited comparability across studies (Cao et al., 2009; Lavie et al., 2010). Following Lavie et al. (2010), research should relate back to March’s (1991) original definitions of exploration and exploitation, which is consistent with the approach of Uotila et al. (2009). According to Uotila et al. (2009), this operationalization has an advantage over other operationalizations (e.g. depth versus breadth of search activities, see Katila & Ahuja, 2002), due to the increased generalizability outside the respective contexts. Furthermore, because organizational learning occurs over time, studying it requires longitudinal data (Argote & Miron-Spektor, 2011). This approach is the only established measure for exploration and exploitation that enables the collection of longitudinal data on large-scale and organizational-level (Luger et al., 2018). Finally, CATA has been proposed as a novel measurement tool with the potential to realize important advancements in theory (McKenny et al., 2018).

Textual Source: Annual Reports

(18)

I hand-collected a total of 1054 annual reports that were retrieved from a variety of web-based sources such as the investor relations page on company websites or other specialist archival depositories like www.marketscreener.com. In case the required reports were unavailable I e-mailed the company. Despite four responses from Swiss and German firms, there were no responses from Japanese and Spanish firms. For the US sample I used Form 10-K, which are similar to annual reports, filed by publicly traded companies on a yearly basis (McKenny et al., 2018). While McKenny et al. (2018) used only the Management Discussion & Analysis (MD&A) section of Form 10-K, I included the entire report. Because for non-US firms the full input of reports was used, which contain sections such as financial statements, I considered taking merely the MD&A section as inadequate and not corresponding to other annual reports. This choice is thus motivated by the preservation of commonality across the sample and enhancement of generalizability in a cross-national context.

Calculating Exploration-Exploitation Scores

I adopted word lists directly from the article of Uotila et al. (2009) and the German validated translation of this list by Heyden et al. (2015) for German and Swiss annual reports (see Table 1 for an overview). Thereby exploration is defined as “things captured by terms such as search, variation, risk-taking,

experimentation, play, flexibility, discovery, innovation” (March 1991, p.71), and exploitation as “such things as refinement, choice, production, efficiency, selection, implementation, execution” (March,

1991, p. 71). I used LIWC 2015 as a software package to process the textual inputs and analyze the number of words related to exploration and exploitation in each report. In line with the conceptual premise of exploration and exploitation reflecting two ends on the same continuum, I measured the relative degree of exploration in each period by dividing the total of exploration words by the total of the exploration plus exploitation words (Heyden et al., 2015; Luger et al., 2018; Uotila et al., 2009). The scores can by construction range from zero to one, whereby one reflects the maximum relative exploration orientation and zero reflects the minimum relative exploration orientation and thus implies an orientation towards exploitation.

While the majority of the reports corresponded to one firm-year observation, the reports were occasionally divided over multiple files. In this case I processed all files and calculated the averages to determine the scores of exploration and exploitation in each specific period. Furthermore, two firms were subject to mergers. I accounted for this by determining the relative possession of shares by each firm. In both cases the shares were approximately equal, and I thus included both firms’ annual reports

ex-ante the merger and calculated the average of the aggregate scores. 28 files that were either corrupt

or did not process correctly were deleted from the sample.9_{Furthermore, in the process of investigating}

outliers, I detected incorrect observations due to different languages in reports,10_{and hence, deleted 16}

observations from my sample.

Additionally, as this study is the first to consider the Form 10-K in its entirety using CATA, I compared the results of the MD&A with the entire Form 10-K to provide insights for future research and contribute to the advancements in the field of applying CATA. I followed McKenny et al. (2018) and used 10% of the sample (representing 35% of the US sample) to obtain stable estimates. I deleted all text except for the MD&A section and processed the files accordingly. As expected, the relative

9_{The processing was first done with .doc files which did not process correctly. The files were then converted to pdf format}

which processed most, yet not all, files correctly.

10_{After running the first processing of texts, the software program announced errors due to different languages in the same}

(19)

exploration scores derived from the entire Form 10-K were on average lower (22%) than scores derived from the MD&A section. A plausible explanation for this is most likely the higher density of words related to exploration and exploitation in the MD&A section.

Table 1: Word-lists Exploration and Exploitation Applied in CATA Original English exploration words and word roots (Uotila et al., 2009):

Explor*, Search*, Variation*, Risk*, Experiment*, Play*, Flexib*, Discover*, Innovat*

Original English exploitation words and word roots (Uotila et al., 2009):

Exploit*, Refine*, Choice*, Production*, Efficien*, Select*, Implement*, Execut*

Translated German exploration words and word roots (Heyden et al., 2015):

Erkund*, Such*, Variation*, Experiment*, Spiel*, Flexib*, Entdeck*, Ausfind*, Ergründ*

Translated German exploitation words and word roots (Heyden et al., 2015):

Ausbeut*, Verfein*, Raffinier*, Wahl*, Auswahl*, Produktion*, Erzeug*, Effizien*, Selekt*, Ausführ*, Durchführ*

3.3.2 Independent Variable

Exogenous Shock. Operationalized as dummy variable whereby the pre-shock years (2006 – 2010) are

coded as ‘0’ and the post-shock years as ‘1’ (2012 – 2014). This approach is similar in principle to other research deploying DiD designs (Lima & Neto, 2018; Lu & Wang, 2018, Moser & Voena, 2012). The year of the shock (2011) was excluded to eliminate the threat of confounding factors due to containing both pre- and post-shock information.

3.3.3 Moderating Variable

Organizational Ambidexterity. In following prior research (Gibson & Birkinshaw, 2004; Luger et al.,

(20)

3.3.4 Control Variables

To control for potential confounding factors, I included several control variables based on prior research. Logarithmic transformation was applied to variables subject to skewness (de Jong & Freel, 2010).

Firm Size. Larger firms are associated with inertia and inadaptability to changing external

conditions (Lant & Mezias, 1992; Tushman & Romanelli, 1985). Moreover, larger firms are more likely to have deeper pools of slack resources (Bourgeios, 1981), and specialized organizational units (e.g. R&D departments) that focus on exploration or watch out for more disruptive changes that may help to respond to potentially damaging risks more effectively (Thompson, 1976; Vahlne & Jonsson, 2017). Firm size was measured as the natural logarithm of the number of employees (Cao et al., 2009; Oehmichen et al., 2017).

Firm Age. Firm age has been related to the institutional routines and norms that provoke inertia

(Tushman & Romanelli, 1985). I thus controlled for firm age by taking the natural logarithm of the number of years since foundation (Oehmichen et al., 2017; Lubatkin et al., 2006).

Return on Assets. Return on assets (ROA) is included as a common proxy for firm performance

(Greve, 2003). Differences in profitability may impact the strategic orientations of firms (Heyden et al., 2015), and investments in innovation (Greve, 2003). ROA was calculated by dividing net income by total assets in each period (Heyden et al., 2015).

Unabsorbed Slack. Organizational slack is a common explanatory factor in organizational

decision making and adaptability (Chen et al., 2007; Cyert & March, 1963). It acts as a cushion that buffers organizations from external shocks and allows to initiate changes in strategy (Meyer, 1982; Bourgeois, 1981), and facilitates experimentation and learning (Levinthal & March, 1981). While Bourgeois (1981) identified three types of organizational slack, research has stressed the importance of unabsorbed slack in the context of exploration and exploitation (Lavie et al., 2010; Luger et al., 2018), which denotes a firm’s uncommitted, ready-to-deploy assets (O’Reilly & Tushman, 2004). Consistent with this argumentation I operationalized unabsorbed slack as the current ratio. Calculated as the current assets divided by current liabilities (Iyer & Miller, 2008).

R&D Investments. I additionally controlled for R&D investments as a proxy for the firm’s

innovation capabilities by taking the natural logarithm of R&D expenses (Heyden et al., 2015). R&D investments may affect innovation activities as they promote the absorption, creation, exploitation and the transformation of knowledge (Cohen & Levinthal, 1990). R&D investments positively influence exploration through the development of new information (Koryak et al., 2018). As excluding all firms without reported R&D investments could introduce sample bias (Wagner, 2007), I follow prior research (Uotila et al., 2009; Wagner, 2007), and treat absent R&D investments as being zero and allocate an R&D missing dummy coded as ‘0’ when reported, and ‘1’ when missing.

3.4 Analytical Method

(21)

& Pischke, 2008). To estimate this effect, I asserted a dummy variable noting ‘1’ for negatively affected firms (nuclear firms) and ‘0’ for green firms, to subsequently create an interaction effect with the exogenous shock.

The data is thereby structured as a strongly balanced panel data set, implying that I observe two groups of firms over multiple points in time (2006 – 2014). In contrast to most prior research, which mostly implemented a cross-sectional research design (Junni et al., 2013; Koryak et al., 2018; Raisch et al., 2008), a longitudinal design decreases the risk of confounding correlation with causation (Luger et al., 2018). Similarly, Uotila et al. (2009) stressed the importance of longitudinal designs in the context of exploration and exploitation to control for endogeneity and unobserved heterogeneity (Uotila et al., 2009).

Although it is preferred to use narrow event windows to reduce the impacts of confounding events (De Jong & Naumovska, 2015), having only one observation per firm per year makes it necessary to broaden the time-frame to provide for stronger inference (Hausman & Kuersteiner, 2008; Donald & Lang, 2007). The data initially ranged from 2005 - 2015, but 2005 was excluded due to many missing data points,11_{and 2015 due to the signing of the Paris Climate Agreement (UN, 2015), which could be}

a potential confounding event. In addition, due to differences in reporting dates, time of policy announcements, and the fact that the annual reports in 2011 contain both pre- and post-shock information, I eliminated potential confounding factors by excluding the year 2011 from my analysis. Concluding, my analysis covers 2006 – 2010 as pre-shock years, and 2012 - 2014 as post-shock years. A post-shock measurement of three years is consistent with a standard approach for measuring organizational learning (Haunschild & Sullivan, 2002; Luger et al., 2018).

A crucial assumption of the DiD approach is the common-trend assumption, implying that it is imperative for the outcome variable trends to be similar in the treated and untreated groups before the shock (Angrist & Pischke, 2008). To validate the DiD estimator, I followed Lima & Neto (2018) and inspect the pre-shock trends of both groups graphically and conducted a formal test to confirm whether there are significant differences between the groups in the pre-shock period. I did so by including interaction effects of the time dummies and the treatment indicator for the pre-shock period. The common-trend assumption is violated if the differences in coefficients are statistically significant between the groups in the pre-shock period. The results and the graphical inspection (see Appendix D) however show no significant differences between the groups, and I thus infer that the common-trend assumption is validated.

In the process of defining my models, I encountered the problem of heteroscedasticity (Andrews, 1991). I detected a linear form of heteroscedasticity in my data by drawing inference based on visual inspection (Stata: rvfplot). To formally confirm this, I conducted the Breusch & Pagan (1979) test (Stata: hettest). Accordingly, I accounted for this issue by applying robust standard errors (Stata:

vce(robust); Huber/White/sandwich robust variance estimator; see White, 1980), which produces

heteroscedasticity-consistent standard errors for OLS regression coefficient estimates (Lu & White, 2014). I additionally guaranteed the reliability of my results by including year and group fixed effects allocating dummies for each individual effect respectively. Year fixed effects account for macroeconomic patterns, while group fixed effects capture unobserved heterogeneity between geographical groups (Lima & Nito, 2018; Lu & Wang, 2018; Schulz et al., 2016). Due to the fact that I rely on relatively low amounts of firms per country and their distribution is highly unequal, including

(22)

country fixed effects could result in biased standard errors and econometric problems.12_{I therefore}

prevent misspecification of the statistical models by accounting for geographic fixed effects by aggregating countries at the continental level according to the UN (2018) "Standard Country or Area Codes for Statistical Use".13

Although I do not expect endogeneity concerns to pose a major threat to the validity of the findings, I can eliminate this to a greater extent when testing for the first hypothesis. Because the primary interest here is to isolate the effect of the exogenous shock on the relative exploration orientation, and OA could potentially produce endogeneity issues due to its dynamic relationship with exploration and exploitation (Wintoki et al., 2012), I exclude it from the models related to the first hypothesis. The rationale underlying the choice is further explicated by the expected ambiguous relationships that OA exerts with respect to strategic responses and the type of firm. Testing for hypothesis 2 accounts for this to a greater extent by considering the temporal structure and firm type.

4. Results

The following section first presents descriptive statistics and correlations. Then, the regression results related to the hypotheses are displayed. This will be followed by a description of the post-hoc tests were performed to investigate intra-group differences. Lastly, procedures of robustness analyses will be clarified.

4.1 Descriptive Statistics and Correlations

Table 2 displays summary statistics for all firms and separately for nuclear and green firms, thereby covering the sample sizes, means, standard deviations, minimum and maximum values per variable. The sample comprises a total of 84 firms with 649 observations. A salient observation is that both nuclear and green firms have a correspondingly high relative exploration orientation (0,42) and thereby similarly high levels of organizational ambidexterity (0,22). The subsequent binary variable for OA indicates that approximately 30% of the observations are considered to be truly ambidextrous in both groups. To assess whether this is representative at the firm-level I investigated the averages per group per year and conclude that these values reflect the truly ambidextrous population. The R&D missing

dummy suggests that approximately 30% of the observations contain reported R&D investment data.

Further discernible are the overall differences in the control variables: the nuclear firms are on average larger, older, more profitable, invest more on R&D, but have less unabsorbed slack resources in comparison to green firms. Although these differences were overall anticipated due to the relative novelty of green firms, the latter contradicts reasoning of Bourgeios (1981), who argued that larger firms are more likely to have deeper pools of slack resources. Nevertheless, it could be that other existing types of slack as proposed by Bourgeios (1981) are present and unobserved in this study. Other observations are consistent with the notion that small firms tend to lack capital and extensive resources, thereby not able to invest as extensively in R&D in comparisons to their larger counterparts (Schumpeter, 1942; Koryak et al., 2018).

12_{The variation inflation factors rise to inappropriate levels that could confound my results. Excluding specific country}

dummies appeared not to be a solution as it would lead to the exclusion of too many countries. Creating groups on continental level resolved this issue.

(23)

(24)

Table 2: Descriptive Statistics

Notes: *Included continuous measure of OA for additional information; the subsequent dummy variable for organizational ambidexterity was used for further analyses; ln = natural logarithm Table 3: Correlations

Notes: Observations = 649; Significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1.

Variable All firms Nuclear Firms Green Firms

n = 84 | obs. = 649 n = 44 | obs. = 349 n = 40 | obs. = 300

Mean S.D. Mean S.D. Min. Max. Mean S.D. Min. Max.

1 Relative Exploration Orientation 0.424 0.153 0.420 0.148 0.11 0.76 0.428 0.160 0.06 0.75

2 Exogenous Shock 0.384 0.487 0.372 0.484 0.00 1.00 0.397 0.490 0.00 1.00

3 Organizational Ambidexterity Continuous* 0.221 0.038 0.222 0.031 0.10 0.25 0.220 0.045 0.05 0.25 4 Organizational Ambidexterity 0.314 0.465 0.281 0.450 0.00 1.00 0.353 0.479 0.00 1.00 5 Firm Size (ln) 8.147 2.237 9.596 1.022 7.15 12.4 6.461 2.080 2.20 10.63 6 Firm Age (ln) 3.156 0.986 3.609 0.833 1.61 4.85 2.629 0.884 0.00 4.68 7 Return on Assets 2.081 9.329 3.731 2.721 -6.86 15.26 0.161 13.158 -53.17 56.71 8 Unabsorbed Slack 1.766 3.667 1.011 0.421 0.32 3.36 2.644 5.244 0.13 71.80 9 R&D Investments (ln) 3.189 5.482 4.246 6.433 0.00 17.56 1.959 3.762 0.00 11.20

10 R&D Missing Dummy 0.724 0.447 0.685 0.465 0.00 1.00 0.770 0.422 0.00 1.00

Variable (1) (2) (3) (4) (5) (6) (7) (8)

(1) Relative Exploration Orientation 1

In the Wake of Fukushima: Unpacking the Impact of an Exogenous Shock on the Exploration-Exploitation Balance