Virtuous cycles : an empirical examination and extension of cyclical ambidexterity theory

(1)

Author: M. S. D. Thomas – 11392274

MSc in Business Administration – Strategy Track

Amsterdam Business School, University of Amsterdam

Final Version – June 23, 2017

Supervisor: M. P. Tempelaar

Virtuous Cycles

AN EMPIRICAL EXAMINATION AND EXTENSION OF

CYCLICAL AMBIDEXTERITY THEORY

(2)

MATTHEW THOMAS I

Statement of Originality

This document is written by Student Matthew Thomas, 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 the supervision of completion of the work, not for the contents.

Abstract

Prior research on the topic of organizational ambidexterity has focused nearly exclusively on cases of the phenomenon where the organization engages in explorative and exploitative activities simultaneously. While these studies have merit in their own right, our knowledge of how ambidexterity behaves in a longitudinal sense remains limited. This study seeks to address this knowledge gap by assessing the performance implications of cyclical

ambidexterity, a time-variant version of the concept based on the punctuated equilibrium

(PE) mechanism. It is argued that firms who cycle between exploration and exploitation will experience superior results that those who do not, impacted by the degree of slack and environmental dynamism. Findings indicate that cyclical ambidexterity presents a narrow path of superior performance that firms must stay within to benefit, with significant positive effect from slack. Interestingly, this effect is the opposite from what was predicted.

Environmental dynamism was also found to have a negative impact on the frequency of exploration-exploitation cycles.

(3)

MATTHEW THOMAS II

Introduction

The modern business landscape is more disrupted than ever before, and presents firms with a significant challenge: how to adapt to constantly shifting conditions, in order to maintain or grow demand for goods and services? A common clarion call in organizational scholarship is that firms should become ambidextrous - that is, aligning themselves

simultaneously with both the need to deliver efficiency in the short-run, while also pursuing new business opportunities so they can sustain strong performance in the long-run (O'Reilly & Tushman, 2013). Such a recommendation is innately paradoxical, in that balancing short-term needs via exploitative activities (within current business areas) and long-short-term needs with explorative activities (in new products and markets) requires firms to make trade-offs.

However, successful organizations are able to reconcile them to a significant degree, and as a result enjoy superior performance (Junni, et al., 2013).

Authors have typically analyzed ambidexterity as a static construct, in the form of the “traditional” types of structural and contextual ambidexterity, also referred to here as the simultaneous types. However, as noted by Piao (2010), researchers assume time-invariance, and place exploration and exploitation in perpetual conflict. This allows for cross-sectional study; however, this assumption may not always hold. This begs the question, what about time? There is evidence to suggest, especially in Burgelman’s studies of the Intel

Corporation, that firms may also pursue ambidexterity in the temporal dimension, oscillating between periods of explorative and exploitative activity. Such a possibility is an attractive proposition to managers, as the paradoxical nature of ambidexterity achieved in the

simultaneous sense incurs significant costs to the organization. If the firm is able to achieve similar results, or the conditions of the business environment are satisfied with the degree of adaptation provided by such oscillation, this type of ambidexterity would be more attractive

(5)

MATTHEW THOMAS 2

to managers when conditions allow. Indeed, such a type of ambidexterity would be consistent with the underlying aim of ambidexterity to achieve a high level of fit with the environment by balancing explorative and exploitative activities, thus achieving superior performance. However, this goal combined with the time-invariant assumption suggests that firms should be pursuing the same balance at all points in time. Maintaining a predetermined balance will make the firm less flexible, making a simultaneous approach detrimental to performance.

Scholars have begun to theorize about a type of ambidexterity governed by this oscillation between exploration and exploitation, which will be termed cyclical ambidexterity – ambidexterity governed by a mechanism of punctuated equilibrium (Boumgarden, et al., 2012; Simsek, et al., 2009; Gupta, et al., 2006). Despite the significant managerial

implications of such a phenomenon, the research agenda in this area is underdeveloped, with many critical questions remaining unanswered. Whereas the link between cross-sectional types of ambidexterity and performance are well-established, limited empirical work has been done with regard to cyclical ambidexterity. Thus, in the first instance, this study builds on the pioneering works on cyclical ambidexterity, in order to evaluate the impact that the use of such a mechanism has on the performance of the firm. In the second instance, this study delves further into contingent relationships with performance, specifically environmental dynamism and slack. High levels of environmental dynamism are one situation under which scholars generally expect to see traditional simultaneous types of ambidexterity.

Investigation in this area will determine if there are situations with particularly high levels of dynamism where cycling leads to performance improvements over a rigid balance of

exploration and exploitation. With regard to slack, organizational cycling is also frequently connected to a firm’s availability of resources, with further research contributing to our

(6)

MATTHEW THOMAS 3

understanding of under what levels of resource availability firms tend to engage in such behaviour. The combination of internal (slack) and external (environmental dynamism) measures of available resources, in conjunction with the direct effects of cyclical

ambidexterity, produce a more complete model of firm behaviour in making investment decisions into explorative and exploitative activities in resource-constrained situations.

Literature Review

Since the publication of March’s seminal paper in the area, organizational ambidexterity has become a topic of significant discussion amongst organizational

researchers. These studies have varied significantly in methodology and type, and generally find evidence that organizational ambidexterity results in superior performance over time (Gibson & Birkinshaw, 2004; O'Reilly & Tushman, 2013). A meta-analysis has further concluded that though the relationship between ambidexterity and performance is strongly influenced by contextual factors and methodological choices, the relationship is, overall, positive (Junni, et al., 2013). The primary benefit of ambidexterity is the balance of exploration and exploitation prevents the organization falling into a success trap from overusing its exploitative capabilities (Levinthal & March, 1993). This causes structural inertia, detrimental to explorative capabilities. It simultaneously prevents a failure trap, whereby the organization successively creates many highly innovative but failure products. Broadly, an explorative innovation relies on new products and new markets, whereas

exploitative innovation relies on refinement of existing products in existing markets (March, 1991).

(7)

MATTHEW THOMAS 4

Theoretical Groundwork

Ambidexterity in the simultaneous sense has been theorized to occur in two ways: structural, with separate explorative and exploitative units within the firm, and contextual, with the two occurring within the same unit. (Lavie, et al., 2010; O'Reilly & Tushman, 2013). Regardless of which of the two conventional types of ambidexterity is discussed, it has

become common to conceptualize ambidexterity as a dynamic capability, building off the work of O’Reilly & Tushman (2008). Such a definition poses ambidexterity as allowing the organization to mobilize, coordinate and integrate these conflicting goals while

simultaneously combining the resources produced from both independent structures of the organization (Raisch & Birkinshaw, 2008). However, this conceptualization of ambidexterity presents a conundrum, representative of the underlying tensions in the research agenda. Dynamic capabilities derive from evolutionary economics and the “low church” of the Resource-Based View, and time is an explicit consideration. In the specific sense of

ambidexterity, it illustrates the ongoing process of adaptation undertaken by an ambidextrous firm in order to achieve congruence with the business environment as per Teece, et al. (1997). However, the overarching orientation of traditional ambidexterity literature, even those

studies using this dynamic capability conceptualization, is cross-sectional.

As discussed by Piao (2010), the time-invariant assumption that is widely used is based on the idea that exploration and exploitation are in constant conflict. However, managerial anecdotes suggest that this is not always the case – many firms, especially small ones, pursue each type of activity in discrete timespans. Burgelman’s 1991 and 2002 studies of the Intel Corporation evidences a similar method of adaptation, whereby the firm

maintains a relatively stagnant mix of products and changes its focus periodically with bursts of exploratory activity. This is likely a result of the costs associated with a firm becoming ambidextrous – resolving the tensions between exploration and exploitation is critical to a

(8)

MATTHEW THOMAS 5

conventionally ambidextrous organization, but imposes financial and organizational penalties by requiring specific structures or individually ambidextrous employees (Birkinshaw & Gibson, 2004). In theory, being ambidextrous in the aggregate over a period of time would reduce these costs, as the internal conflicts created would not be as extreme. Such a type of ambidexterity could be an effective substitute to the traditional types in some cases.

A further implication of the use of the time-invariant assumption relates to the

concept of balance, whereby an organization arrives at an appropriate mix of exploration and exploitation suited to its environment. Studies of ambidexterity cross-sectionally imply that such a balance, once achieved, is to be enshrined within the firm and vigorously maintained. However, with the modern trend of increasing market dynamism, it is absurd to claim that a single balance level is appropriate indefinitely. Such a situation is a paradox for time-invariant ambidexterity, which purports to allow a business to achieve long-term fit with its environment, but could be, in and of itself, a source of rigidity should the organization’s balance of exploration and exploitation not adapt over time to external requirements. In this sense, pursuing conventional ambidexterity can have a negative impact on overall firm performance, the exact opposite of the benefits espoused by its advocates. Once again, a time-variant form of ambidexterity is a solution to this issue, allowing an organization to achieve ambidexterity in the long run, not just at one specific point in time.

From Statics Types to Dynamism

Given the above issue with pursuing a wholly static approach to ambidexterity and the usefulness of a dynamic type, it is clear a dynamic approach has merit. Given the

well-developed understanding of traditional types of ambidexterity, a significant number of scholars have called for research to explore the longitudinal issue (O'Reilly & Tushman, 2013; Gibson & Birkinshaw, 2004). This has led to dynamic models of organizational

(9)

MATTHEW THOMAS 6

adaptation underpinned by a mechanism of punctuated equilibrium (PE). Under PE, a firm is hypothesized to undertake long periods with high levels of exploitative activity, which are punctuated with brief bursts of explorative, revolutionary change (Simsek, et al., 2009). Similar processes have been suggested to help small electronics manufacturers to move back and forth between exploration and exploitation (Brown & Eisenhardt, 1997), as firms can more easily change their organizational structure than they can change soft factors such as culture (Boumgarden, et al., 2012). This has resulted in two broad streams of thought - cyclical ambidexterity, focused on here, and vacillation theory.

Gupta, et al. (2006) began to explore the dynamic area referred to here as cyclical ambidexterity by applying PE to the traditional ambidexterity literature. The authors

conceptualize exploration and exploitation as polar opposites on a continuum. They conclude that under certain circumstances, traditional conceptualizations of ambidexterity may be challenging or impractical to achieve, given the costs imposed by the paradoxical nature of simultaneously pursuing both exploration and exploitation. In the case of Intel Corporation, Burgelman (2002) appears to similarly conclude that the PE mechanism best explains firm behaviour. Intel had developed a series of organizational processes and an organizational culture that functioned as a dynamic capability, allowing it to manage punctuations

effectively. Such “high-order capabilities and routines for change”, as called by Kang, et al. (2016), are critical to an organization’s ability to understand when to cycle, as well as conduct the cycling event in an organized manner. Without this dynamic capability,

organizations managing punctuations have limited ability to control how events unfold, and are unable to undertake them in a deliberate, sustainable manner. These events are essentially random. Conversely, with the dynamic capability required, firms are able to harness these events to their own benefit.

(10)

MATTHEW THOMAS 7

Simsek, et al. (2009) provide one of the clearest frameworks to look at this type of ambidexterity, with a typology of organizational ambidexterity divided on structural and temporal dimensions. The authors further postulate on the antecedents of the four types of ambidexterity outlined: harmonic (contextual), partitional (structural), cyclical, and

reciprocal. They reconceptualise the PE construct as a “temporally sequential” type of ambidexterity, in other words a dynamic type. The resulting construct has had many names, including temporal (Gibson & Birkinshaw, 2004), sequential (O'Reilly & Tushman, 2013), or cyclical ambidexterity (Simsek, et al., 2009) – all refer to a type of ambidexterity governed by PE. While far from a consensus name, indeed the lack of a common title is indicative of the weakness of the literature in this area, cyclical ambidexterity best captures the underlying PE mechanism at play.

The alternative to the cyclical ambidexterity concept is vacillation, stemming from Nickerson & Zenger’s (2002) vacillation theory. Typical organizational theories (e.g. contingency theory, transaction cost economics, ambidexterity), use a contingent-fit model, where the organization is expected to change to maintain a degree of fit with its environment. However, organizations will not change beyond the degree required to maintain fit.

Vacillation theory, by comparison, proposes that there may be benefits to changes in organizational structure even though they are not required to maintain fit with the

environment, as this may allow the organization to temporarily achieve a higher level of functionality than if it remained static within one structure. This is accomplished by swings in directions of organizational choices under the same PE mechanism (e.g. from

centralization to decentralization). Boumgarden, et al. (2012) further investigate the relationship between this theory and traditional simultaneous types of ambidexterity, and

(11)

MATTHEW THOMAS 8

conclude that under some circumstances ambidexterity alone may not provide the highest level of performance, though the two are generally complimentary.

These two competing mechanisms present a significant definitional challenge. Both utilize the same underlying mechanism, but approach the issue of adaptation from different theoretical perspectives. The differences can be summarized into two major points. First, all types of ambidexterity assume a balance between exploration and exploitation, either fixed (traditional types) or varying to maintain fit (cyclical ambidexterity). Vacillation assumes no such balance, and suggests change will happen regardless as it is the act of undertaking the change that enables superior performance. Second, vacillation suggests that only in some cases do organizations develop “high-order capabilities and routines for change” (Kang, et al., 2016). While some organizations that vacillate may go on to develop such dynamic capabilities, the conceptualization of ambidexterity as a dynamic capability explicitly requires the firm to have such capabilities (O'Reilly & Tushman, 2008). Thus, it is best (though imperfect) to understand vacillation as a lower-order variant of cyclical

ambidexterity. Organizations that vacillate but are not cyclically ambidextrous are less purposeful in their swings between exploration and exploitation, and do not have the high-level routines for change that make changes more than random.

A significant challenge to both theories, stemming from their PE mechanism, is Levinthal & March’s (1993) myopias of learning. Organizations have to be cognizant to avoid myopias specific to learning, such as the success trap and the failure trap. The success trap is of particular concern in a firm engaging in some sort of cycling, given the returns to exploitation tend to be closer temporally (March, 1991). This creates an incentive to the organization to exploit more than explore, with the organization excelling at exploitation but losing the ability to explore as it increases its exploitation. The failure trap is the reverse

(12)

MATTHEW THOMAS 9

situation, where the organization begins producing a large amount of very innovative new products, but loses the ability to make them successful as it loses its exploitative ability. Simultaneous types of ambidexterity are frequently prescribed as a solution to this problem, as they build a balance of exploration and exploitation within the organization that becomes difficult to change. The introduction of cycling or vacillation increases the risk of falling into one of these traps, and while vacillation theory suggests larger swings are more desirable to a point (Kang, et al., 2016), larger swings also increase the risk of the return swing not

occurring. Thus, “higher order routines” (i.e. a dynamic capability for change) are key to enable firms to undertake increasingly large swings, as the dynamic capabilities developed enable the organization to resist the attractive but dangerous proposition of remaining either extremely exploitative or extremely explorative, making cycling back and forth a

performance-enhancing endeavour.

This combination of ambidexterity, vacillation theory, and myopias of learning forms the crux of the cyclical ambidexterity argument: cyclical ambidexterity presents a narrow balance, that if maintained, promises improved performance and thus a competitive

advantage. However, straying from the optimal point may make the firm worse off. As an analogy, this is akin to piloting a boat in a winding river with a strong current in the centre. As one progresses down the river, to maximize speed, it is necessary to continually adjust course to stay within the current. One must also be diligent, and learn how to avoid the current sweeping the vessel aground on either side. This risk is enhanced when a course correction is too strong relative to the captain’s level of skill, with some corrections being too strong even for the best captains. An alternative would be to navigate in part of the channel with little or no current. However, by doing so, you will be passed by those who are more efficiently using the current resources available to them. Some may still enter the current

(13)

MATTHEW THOMAS 10

accidentally, and be momentarily swept away, but skill is required to make this more than a chance event.

Overall, the literature is severely constrained by the number of studies including a temporal aspect, and the ongoing theoretical debates in the field are likely impeding the progress of empirical work. While Simsek, et al. (2009) provides a unified foundation on which further empirical research can be built, it is wholly conceptual, and few scholars have taken to the research agenda of extending the understanding of cyclical ambidexterity. Boumgarden, et al. (2012) continues to theorize about the linkages between ambidexterity, the PE mechanism, and performance. Piao (2010) invokes a temporal dimension, necessary to study the relationship between exploration and organizational lifespan, however the results are only applicable to organizational longevity. Though there is significant theorization on the topic, laying necessary groundwork, empirical studies are nearly non-existent.

A natural place to begin empirical work is with performance, which in the context of traditional types of ambidexterity has received continual attention, to the point of requiring meta-analysis to sift through the findings. Performance is top-of-mind to management, given how important it is to firm survival, valuation, and, in many cases, compensation. The results of examining such a relationship has clear managerial implications, and is a critical aspect of the phenomenon. Thus, the lack of knowledge in this area should be rectified. Geerts, et al. (2010) conducted an initial foray with a narrow study, examining the relationship in a comparative context between service and manufacturing firms. However, this study is of limited usefulness given its emphasis on the comparison between the two industries rather than firms that are and are not cyclically ambidextrous. Recently, Kang, et al. (2016) have conducted the first large sample empirical study on performance connected to vacillation theory, finding a positive relationship. However, a straightforward connection between

(14)

MATTHEW THOMAS 11

vacillation and performance was tested in this study, and cyclical ambidexterity was not directly considered. This leaves much room for corroboration and providing further nuance to the linkage. As such, this paper will seek to answer the question of “What is the

implication of cyclical ambidexterity on firm performance?”

Contributions

The primary contribution of this study is elaborating on the conceptualization of ambidexterity as a dynamic capability. As previously discussed, while other authors have conceptualized ambidexterity as a dynamic capability (e.g. Raisch & Birkinshaw, 2008; Jansen, et al., 2009), they have not studied it in a dynamic sense. O’Reilly & Tushman’s (2008) definition of ambidexterity as a dynamic capability being “a set of actions taken by senior management that permit the enterprise to identify opportunities and threats and reconfigure assets to adapt to these” is well-suited to cyclical ambidexterity. Should an organization demonstrate that it is capable of periodically reconfiguring itself with successive swings between exploration and exploitation with above-average performance, it is ultimately demonstrating it possesses a dynamic capability that allows it to do so. Such a discovery would cement the conceptualization of ambidexterity as a dynamic capability in the longitudinal sense, and lend significant credence to the claim in a cross-sectional sense.

Building on Kang, et al. (2016) and Geerts, et al. (2010) (who only considered the direct connection of cycling on performance) this study will also provide additional credence to the direct connection. Further testing and refinement of theory in this area is critical to our understanding of this poorly explored phenomenon. This study will also contribute a more nuanced understanding of cycling through the inclusion of contingencies, these being environmental dynamism and slack. Examination here will improve our understanding of under what circumstances cyclical ambidexterity performs best, and in what cases do firms

(15)

MATTHEW THOMAS 12

exhibit the phenomenon. This combination of an internal and an external effect have commonality in that they are measures of the availability of resources to the firm, and will allow exploration into how firms make investment decisions under resource scarcity. Specifically, environmental dynamism, which is derived from Aldrich (1979), will detail the adaptive effectiveness of the phenomenon by demonstrating under what market conditions firms decide to be cyclically ambidextrous. This will show the limits of simultaneous ambidexterity’s ability to keep pace with environmental change, and demonstrate whether or not cyclical ambidexterity is a useful substitute for other adaptation methods. Slack,

originating in Cyert & March (1963), will elaborate under what resource conditions firms turn to cyclical ambidexterity – managerial anecdote would suggest this to be in cases with more constrained resources. Such knowledge will help understand how firms optimize their levels of slack to maximize performance.

Hypothesis Development

Performance

An organization that is cyclically ambidextrous would have a significant advantage over its non-ambidextrous counterparts with regards to performance. As per Burgelman (1991), organizations must maintain a requisite degree of fit with their environment, but it is also important how the organization maintains this fit. Strategic reorientation is the worst possible situation, as the value of the firm’s competencies and assets becomes impaired. A cyclically ambidextrous organization would instead conduct periodic strategic renewal as a result of the equilibrium punctuation, maintaining fit without impairment of its assets and competencies.

Dynamic capabilities are firm processes that integrate, reconfigure, and release resources to match market change (Eisenhardt & Martin, 2000). Cyclical ambidexterity

(16)

MATTHEW THOMAS 13

draws upon two of these processes, in the areas of knowledge and reconfiguration, to function. Organizations have been characterized as being internally “sticky”, that is significant effort is required to force change (Szulanski, 1996). An organization with a dynamic capability to reconfigure itself has to overcome this stickiness, allowing it to change more freely. However, cyclical ambidexterity cannot be accomplished by simply making the organization flexible to change, it must also learn how to select an appropriate balance of exploration and exploitation best suited for environmental conditions. This is dependent on the firm’s ability to absorb knowledge and experience and subsequently leverage it, or its absorptive capacity (ACAP; Zahra & George, 2002). When the firm has both dynamic capabilities, those being the ability to be flexible and select an optimal balance, the

organization will be able to oscillate between exploration and exploitation effectively. As oscillations become more frequent, the organization will be able to undertake these changes with increasing ease and productiveness as a result of firm learning through ACAP,

producing an increasingly strong dynamic capability. The cyclically ambidextrous

organization is then able to maintain congruence with the business environment (Teece, et al., 1997). Subsequently, such a dynamic capability enables the firm to mobilize and coordinate resources superior to its non-ambidextrous competitors, creating a competitive advantage and hence superior performance.

The use of a simultaneous type of ambidexterity may also be detrimental to

performance. With regard to environmental fit, traditional types of ambidexterity advocate for an organization to achieve a balance of exploration and exploitation (a firm’s “balance of activities”) to be maintained in both the short-run and long-run (March, 1991). However, achieving balance at one point in time (as assumed by research under a time-invariant

(17)

MATTHEW THOMAS 14

balanced in the long run. This is instead determined by the ability of the organization to vary, and thus maintain fit (Burgelman, 1991). The ability of a fixed balance between exploration and exploitation to maintain fit over the long term implies there is an invariant natural balance of activities for a given company or industry. Moreover, the probability of correctly choosing such a value the first time, in a one-shot game approach, is infinitesimally small. Because of this, even in a simultaneous type of ambidexterity, some degree of adaptation is required.

The level of stability proposed by simultaneous ambidexterity is far from that encountered in modern markets, which are instead characterized by increasing levels of disruption and demand for innovation. Such disruption increases the level of environmental dynamism in the market, i.e. the amount of uncertainty from the external environment (Baum & Wally, 2003). Specifically, increasing levels of demand uncertainty (Wernerfelt &

Karnani, 1987) and effect uncertainty (Milliken, 1987) will result, changing the natural balance of the firm’s activities. This induces a gap to appear between the firm’s current balance of activities and that which would be ideal given the environment, depressing

performance as theorized by Burgelman (1991). In order to continually align the balance, the organization needs to vary the balance at every point in time by being cyclically

ambidextrous. This allows the organization to continually maintain an internal balance of activities that matches that mandated by the environment, maintaining congruency and improving performance. In some cases, a well-established level of balance may in fact have the opposite effect of hindering organizational adaptation, rather than encouraging it. Instead of being a vehicle to maintain fit, the existing level of balance becomes embedded as a distinctive competence (Levinthal & March, 1993). If the firm does not learn to adapt and

(18)

MATTHEW THOMAS 15

improve the fit of its balance of activities to bring it closer to that required by the environment, it will face increasingly poor returns as the gap grows.

Under certain circumstances, it may be possible for a cyclically ambidextrous firm to perform better than either a contextually or structurally ambidextrous firm when high levels of dynamism are not present. Being ambidextrous is fundamentally a costly endeavour, and requires creating units or structures that are paradoxical. To overcome the paradox,

integration mechanisms are required, and such mechanisms result in higher management costs to the organization in the form of coordination costs than one devoid of paradox. An organization that is cyclically ambidextrous incurs costs, but rather than coordination costs these are switching costs. As the organization oscillates between exploration and

exploitation, structural changes and common vision are necessary. However, organizations can change their structure more easily than they can change their cultures (Boumgarden, et al., 2012), thus cyclical ambidexterity offers an absolute cost advantage over both

simultaneous types by eliminating the need for costly integration mechanisms.

The aforementioned myopias of learning (Levinthal & March, 1993), however, present a challenge to cyclically ambidextrous organizations. When an organization deviates from balance as one undertaking cyclical ambidexterity does, there is a chance of falling into either the success or failure trap, preventing the natural backswing of the pendulum towards balance. The larger the deviation, the larger the potential benefit, but it becomes more difficult to return to balance (Lavie & Rosenkopf, 2006) and risk is elevated. Under vacillation theory, where organizations do not necessarily have high-level subroutines necessary to control this process, large swings should not be undertaken lightly, as they are more likely to result in the loss of competencies (Kang, et al., 2016). In a cyclically

(19)

MATTHEW THOMAS 16

ability to cycle without running afoul of the myopias. It does, however, imply an upper limit to the degree of swing. Swings that are beyond the organization’s ability to conduct in an orderly manner, as determined by the strength of the dynamic capability, remain a danger. It is conceivable that some swings are so large that regardless of how strong the dynamic capability is that myopia cannot be avoided, leading to a theoretical maximum ability of the mechanism to accommodate environmental fit.

Despite the danger outlined above, cyclical ambidexterity presents several methods by which it can provide performance above that of other ambidextrous and non-ambidextrous firms. Thus, it is hypothesized that:

H1: A cyclically ambidextrous firm will perform better than a non-cyclically ambidextrous firm.

Slack

Slack is the excess resources present in an organization given a level of output (Nohria & Gulati, 1996). The concept is attributed as responsible for a wide variety of phenomena, including innovation. Three main varieties are hypothesized: absorbed, unabsorbed, and potential slack. Absorbed slack is the excess resources employed in

operations above and beyond what is required for short-term maintenance (Greve, 2003), and manifests itself as benefits from free coffee to personal development time. Unabsorbed slack is present as excess financial resources such as cash and short-term securities, whereas potential slack is accessible excess financial resources outside the firm, such as available credit. While not available directly to staff, excess financial resources reduce the strictness of criteria for investment decisions, R&D projects, and general monitoring.

Slack in ambidexterity research has been found to vary with firm size and

environmental munificence (Cao, et al., 2009), which relates to the resources available to a firm. Firm size has been related to strategy choice (Hofer, 1975), and amount of resources,

(20)

MATTHEW THOMAS 17

with larger firms commanding significant resources (Chen & Hambrick, 1995). Small organizations are resource-constrained, unlike their large counterparts, which impacts the method in which they compete. Limited resources challenge an organization’s ability to simultaneously engage in exploration and exploitation, forcing them away from simultaneous types of ambidexterity and to search for alternative methods of adaptation. However,

maintaining congruence with the environment is even more important for small firms than large firms. The latter may have a desirable position within a protected strategic group, command market power, or otherwise be able to raise barriers to entry into its markets to protect market share (Porter, 1979), whereas a small firm has neither the method or excess resources to influence market forces. However, small organizations have an advantage of being less bureaucratic and less prone to inertia than their larger counterparts (Haveman, 1993), allowing them to enact strategic change more easily, and hence cycle.

This combination of factors makes cyclical ambidexterity ideally suited to small firms. It is a relatively cheap adaptation mechanism, within the resource budgets of these firms, as the organization does not have to support the costs of simultaneous types of

ambidexterity. The organization also incurs minimal costs during the cycling process as the firm is flexible, and the mechanism allows the firm to maintain a high level of congruence with the environment quickly. This is not the case with larger firms, where their higher level of slack resources does not necessitate cycling to begin with, but also will incur additional costs from cycling due to their more rigid structure and have less of a need to quickly maintain congruence due to market power. In summary, it is expected that:

H2A: As slack decreases, the performance benefit from cyclical ambidexterity increases

(21)

MATTHEW THOMAS 18

Environmental munificence, defined by Dess & Beard (1984) as the extent to which the environment supports the sustained growth of a firm, also impacts cyclical

ambidexterity via absorbed slack. In highly munificent environments, growth opportunities and the resources required to pursue them are generally more plentiful (Cao, et al., 2009). Easy firm growth results in simultaneous increase of the organization’s slack resources, initially via unabsorbed slack. Over time, high availability of resources tends to relax management control, with more opportunities for experimenting and less rigorous

performance management (Greve, 2003), including some slack being converted to absorbed slack in the form of suboptimal structural choices and other areas of the firm’s cost structure (Yasai-Ardekani, 1986). This effect is particularly acute in large, mature organizations where accumulation has had significant time to occur.

The excess of slack resources relaxes the need of a firm to be efficient. One result of this may be giving rise to needless ambidexterity, where an organization achieves balance but has no long-run need to be ambidextrous. The excess of resources in the firm have allowed it to adapt an organizational structure that is beyond the needs of the firm, but is convenient as it contributes to stability by avoiding cycling. Conversely, in a resource-constrained

environment, such excess becomes the subject of cost cutting, or was never within the bounds of the firm’s resources to begin with. In both cases, we should expect that:

H2B: As slack decreases, a firm is more likely to be cyclically ambidextrous.

Environmental Dynamism

A turbulent market is one of the major reasons why a firm would seek to become ambidextrous. Environmental dynamism has two major dimensions – the degree to which change is predictable, and how quickly the market moves (Dess & Beard, 1984).

(22)

MATTHEW THOMAS 19

contracts and vertically integrating, but these efforts do not assist with managing the so-called “clockspeed” of the market (Masini, et al., 2004). This is an area where firms must turn to adaptation mechanisms such as ambidexterity.

As cyclical ambidexterity maintains environmental fit by the firm cycling its structure to achieve an appropriate balance of activities at any given point in time, there is an upper limit as to how quickly a firm can adapt itself to changing market conditions (Birkinshaw & Gibson, 2004). Organizational realities related to variation of structure and culture have a fixed threshold that constrains fast responses, and responses that push this boundary are at risk of running afoul of myopias. When market velocity is low, this is a nonissue –

adaptation occurs sufficiently quickly to meet the needs of the environment. However, when market velocity is high, a gap will open between the organization’s current balance of

activities and the ideal balance (Burgelman, 1991). If left unaddressed, the presence of the gap will depress returns and make the organization less profitable. The organization will not only waste resources by using them in an inefficient way by undertaking the wrong activities, but also will set the firm back from its competitors. Unsurprisingly, the firm will become less profitable.

Given the limits imposed by the mechanism, it is expected that:

H3A: As environmental dynamism increases, the performance of a cyclically ambidextrous

firm decreases (negative interaction effect).

The business strategy literature, as well as many domains in economics, generally conceptualize managers as being boundedly rational (Puranam, et al., 2015). Though having matured over the years, the conceptualization is built upon Simon (1955) where managers are assumed to be rational actors who make decisions with imperfect information, with an aim to “satisfice” – achieve targets while satisfying constraints. The two other major components of

(23)

MATTHEW THOMAS 20

management decision theory are aspiration levels and problemistic search, also originating from Simon (1955) and further developed in Cyert & March (1963), now used in a variety of contexts such as R&D impacts and internationalization theory (e.g. Madsen, 2005; Wennberg & Holmquist, 2008; Greve, 2003). Should a firm’s performance fall below management’s aspiration level, it is theorized they will engage in problemistic search in an attempt to return performance to at least meet expectations (Lant, 1992).

Under a condition where performance is falling due to poor environmental fit as a result of the constraints of the PE mechanism, performance will fall below aspiration levels. Management is then expected to engage in problemistic search to reverse the fall in

performance, resulting in a prescription of a new adaptation mechanism that will improve firm response time. Simultaneous types of ambidexterity are prime candidates for this

position, as they do not require the firm to reconstruct organizational structures and culture in order to adapt, improving adaptation speed. However, any adaptation mechanism that is faster than one governed by the PE mechanism will satisfice management, and lead to the firm moving away from cyclical ambidexterity. Thus, it is expected that:

H3B: As environmental dynamism increases, a firm is less likely to be cyclically

ambidextrous.

Fig. 1 – Full Model

Environmental Dynamism Cyclical Ambidexterity Performance (Q) H3B H1 H3A Slack H2B H2A

(24)

MATTHEW THOMAS 21

Methods

Data and Sample

To test these hypotheses a panel of financial and patent data is used from US public firms. Financial data was drawn from the North American Annual Fundamentals database from Standard & Poor’s Compustat. Patent data was drawn from the US Patent Office via USPO PatentsView, with additional data from the American National Bureau of Economic Research Patent Data Project being used to link firms together in the two databases. Economic indicators were drawn from the World Bank.

Data was collected from 1976 through 2006. However, due to patent citation

completeness issues prior to 1980, the years used for analysis are 1980-2006. The 1976-1979 data are used as an input to the analysis in the calculation of some variables but not directly included. To ensure generalizability of the findings of the study, the initial sample of firms drawn upon was the S&P 500 index, as it stood at December 31, 2006. This sample

construction, although it potentially introduces survivor bias, is necessary for data availability reasons as many of the firms in the index in 1980 do not have sufficient data available. After removing firms with insufficient patent filings to construct variables (initially set at five patents in consecutive years) and discontinuous data, 181 firms expanded into 3004 firm-years of data are included in the analysis. This includes firms from 37 industries as denoted by their two-digit Standard Industrial Classification (SIC) codes, a wide range of industries enabling a high level of generalization of the results. These firms together filed 211.026 patents comprising 2.993.416 citations in the years under investigation.

(25)

MATTHEW THOMAS 22

Operationalization of Constructs

Cyclical Ambidexterity

The trade-off between exploration and exploitation is central to ambidexterity, and is used to measure the latter construct by way of patent citation data, building on the approach of Wang & Li (2008) and Benner & Tushman (2002). Exploration is operationalized as the number of patent citations in a given year that the focal firm has not previously cited and does not own the patent in question. Conversely, exploitation is operationalized as the number of patent citations in a given year that have been historically cited by the given firm, or cite patents that the firm owns. This approach has the benefit of providing significantly more granularity to the issue of balance versus the competing technical classification

approach, in that each individual patent has its own internal exploration/exploitation balance, rather than the blunter categorization done by technical classification of determining each patent to be one or the other

To achieve this operationalization, all patents cited by the sample firms within the study years were extracted from the USPO database. Subsequently, all the citations made by these patents were also extracted from the database, and sorted by firm, patent grant date, and patent number. This produces a time-ordered set of citations for each firm. Each citation was determined to be explorative or exploitative using a query function that referenced the firm’s prior activity, and stored the result in a dummy variable. The number of citations was then aggregated at the firm level for each year, providing a count of explorative citations, exploitative citations, and total number of citations.

Subsequent to enumerating the number of each type of citation for the focal firm in a given year, the balance is re-expressed as a ratio using Kang, et al.’s (2016) focus ratio, which is computed as:

(26)

MATTHEW THOMAS 23

𝑓_𝑖𝑡 = 𝑅𝑖𝑡 𝑇𝐶_𝑖𝑡−

𝑇𝑖𝑡

𝑇𝐶_𝑖𝑡

Here fit denotes the focus ratio of firm i at time t. Rit is the number of explorative citations

made by firm i at time t, Tit being the number of exploitative citations made by firm i at time

t, and TCit being the total number of patent citations made by firm i at time t. This produces a

zero-centred balance variable with positive values indicating the firm is engaging in more explorative activity, and negative values indicating more exploitative activity. A zero value indicates perfect balance.

From the focus ratio, cycling events can be identified. Such an event occurs when a firm switches between majority exploratory and majority exploitative activity between time t

– 1 and time t. In mathematical terms, when a sign change occurs between time t – 1 and time t (i.e. fit-1 * fit < 0), a cycling event occurs. To measure cyclical ambidexterity, a moving

five-year count of cycling events (from time t to time t – 4) is kept, known as cycling frequency, and used as the independent variable.

To delve further into the impacts of cycling on firm performance, an additional variable of cycling intensity was constructed, calculated as:

𝑠_𝑖𝑡 = |𝑓_𝑖𝑡− 𝑓_𝑖𝑡−1|

Where sit is cycling intensity, fit is the focal firm’s focus ratio in the current period, and fit-1 is

the focal firm’s focus ratio in the previous period. The absolute value of this value produces a measure of deviation ranging from zero to two, where two indicates a shift from 100% of either exploration or exploitation to 100% of the other.

(27)

MATTHEW THOMAS 24

Performance

The performance impacts of ambidexterity have been measured using a wide range of measures (O'Reilly & Tushman, 2013). However, a large sample study lends itself well to financial measures of performance. These do present challenges, as the effects of

exploitative activity are felt much quicker by a firm than exploitative activity (Uotila, et al., 2009). A market-based measure of performance resolves this issue to a significant extent, as per Fama’s (1970) Efficient Market Hypothesis, the effects of information are already “priced in” to the value of the security. Further, market measures avoid potential time lag as they contain this forward-looking aspect (Wang & Li, 2008). In this vein, Tobin’s Q is used. The simpler approximated Q value is used, computed as the firm’s market value divided by book value. As Wang & Li (2008) mentions, other studies such as Lindenberg & Ross (1981) have shown that this measure accounts for 96% of the explanatory power of the significantly more laborious full measure, and is computed as:

𝑄 =𝑀𝑎𝑟𝑘𝑒𝑡 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐸𝑞𝑢𝑖𝑡𝑦 + 𝐵𝑜𝑜𝑘 𝑉𝑎𝑙𝑢𝑒 𝑜𝑓 𝐷𝑒𝑏𝑡 𝑇𝑜𝑡𝑎𝑙 𝐴𝑠𝑠𝑒𝑡𝑠

This measure uses Compustat items CSHO, PRCC_F, DLTT, DLC, and AT.

Slack

A battery of three measures of slack, as employed by Greve (2003) and Bromiley (1991) is used here. Absorbed slack represents how lean the firm’s operations are, and is the ratio of selling, general, and administrative expenses to sales (items XSGA, SALE).

Unabsorbed slack, the ratio of cash and short-term marketable securities to liabilities, shows excess firm assets that could be employed (items CHE, LT). Potential slack is the ratio of debt to equity, which indicates the firm’s ability to raise funds through debt issuance and thereby relax constraints in its operating environment (items DT, SEQ). Combined, these three measures provide a holistic picture of the possible sources of slack for a firm.

(28)

MATTHEW THOMAS 25

Environmental Dynamism

Environmental dynamism has been measured by a variety of index methods over time. In this case, a multi-measure method as advocated by Howell, et al. (2016), built upon the original conception by Aldrich (1979) is used. This is comprised of two components: first, a measure of market velocity, which is “the speed or rate at which new opportunities emerge” (Eisenhardt, 1989). Secondly, market ambiguity, defined as “lack of clarity, such that it is difficult to interpret or distinguish opportunities” (March & Olsen, 1976). Both aspects are important challenges for firms to overcome – speed alone tests a firm’s ability to adapt, especially in accelerating situations, however a lack of clarity exacerbates this issue and can render a firm rudderless. Conversely, high ambiguity is easier to overcome in a slower moving market than a faster one. This underscores the need for a multifaceted approach.

Market velocity is calculated as the year-over-year percentage change in industry shipments, with industry shipments calculated as the summation of the SALE data item for all firms in the Compustat database sharing the same two-digit SIC code for a given year. Market ambiguity is calculated as the standard deviation of the SALE data item for all firms in the database with the same two-digit SIC code for the five preceding years (t – 1 to t – 5).

Control Variables

A pressing issue with regard to the research design is the potential for false positives for the presence of cyclical ambidexterity caused by other phenomena. An additional concern is the difference in intensity of cyclical ambidexterity may be skewed by other factors,

unrelated to the hypotheses. Thus, five control variables are added to account for these situations:

1. Industry – The large number of ambidexterity related studies specific to individual industries suggests that the phenomenon is significantly affected in practice by

(29)

MATTHEW THOMAS 26

industry economics. Firms are assigned a dummy variable representing their two-digit SIC code to account for industry effects.

2. Firm Size – It is widely theorized that a firm’s size has a significant impact on the way it goes about managing ambidexterity, with larger firms relying more on structural ambidexterity while smaller ones rely on contextual ambidexterity

(Lubatkin, et al., 2006; Brown & Eisenhardt, 1997). A firm’s size, as determined by it’s number of employees (data item EMP), as well as market capitalization (CSHO * PRCC_F) will be used to control for these effects.

3. Financial Distress – As explained by Levinthal & March (1993), firms in their death throes may enter into the failure trap, whereby they begin producing large numbers of highly novel but failure innovations. Such a situation may produce a trend similar to cyclical ambidexterity’s PE mechanism. To control for this and other situations of poor financial performance, the current ratio is used, calculated as current assets divided by current liabilities (items ACT and LCT).

4. R&D Intensity – While R&D expenses are required as a direct input towards the creation of patents and hence the citations used as the independent variable, R&D expenses may also be correlated with cyclical ambidexterity. However, there is no causal relationship as R&D expenses can go towards either explorative or exploitative activity, thus making it necessary to control for this factor, which is measured as the firm’s reported R&D expenditures divided by revenue (items XRD and SALE) as per Stettner & Lavie (2014).

5. Economic Cycles – The ebb and flow of markets and macroeconomic factors has wide ranging impacts on firms, and to what extent they focus on exploitative and explorative activities. Thus, the change in North American GDP is used to control for changes over the business cycle. North America was specifically selected, as the S&P

(30)

MATTHEW THOMAS 27

500 is an index of American firms that also tend to have a presence in Canada and Mexico. Preliminary studies confirmed this choice has the strongest relationship with performance.

Results

The analysis begins with descriptive statistics and Pearson bivariate correlations, as displayed in table 1. A large number of correlations are significant, especially with Cycling Frequency. As would be expected, Cycling Event and Cycling Frequency are significantly correlated due to the latter variable being a summation of the former. There is minimal evidence of multicollinearity amongst the battery of slack and industry dynamism variables, though the correlation between Absorbed Slack and Unabsorbed slack is relatively high. As an assurance against multicollinearity here, regression models were run with and without these variables, and it was found that the results are similar.

Impact of Cyclical Ambidexterity on Performance

The analysis proceeds to test the direct effect of Cycling Frequency on Tobin’s Q. Further investigation and preliminary analysis showed significant rightward skew of the Q variable, in addition to a distinct shape in the standardized residuals of regression analyses. A Box-Cox transformation was applied to the raw Q values, settling on a natural logarithm transformation which significantly resolved these issues. Unless otherwise noted, all further analysis was performed using log-transformed Q values.

Subsequently, a hierarchical multiple regression analysis was performed. In the first model, the control variables of firm size, current ratio, market capitalization, R&D intensity, GDP change, and industry effects are entered. This model was significant with F (31, 2116) = 30, p < .000, and adjusted R2 = .292. Secondly, the independent variable of Cycling Frequency was entered, and this model was also significant with F (32, 2115) = 30, p < .000,

(31)

MATTHEW THOMAS 28

Table 1: Descriptive Statistics and Correlation Matrix

Mean SD Max Min 1 2 3 4 5 6 7 8 9 10 11 12

1. Log-Transformed Q .19 .34 1.60 -1.82 2. Cycling Event .14 .35 1.00 .00 .07** 3. Cycling Frequency .42 .89 .00 5.00 .11** .48** 4. Market Velocity .07 .10 1.37 -.36 -.11** -.07** -.11** 5. Market Ambiguity 46319.14 37498.21 334033.46 384.87 .09** .08** .09** .01 6. Unabsorbed Slack .61 1.77 34.02 .00 .28** .01 -.02 -.03 .05* 7. Potential Slack .71 2.93 66.97 -35.28 -.13** -.07** -.07** .04* .07** -.07** 8. Absorbed Slack .26 .37 9.07 .00 .22** .00 .00 -.01 .03 .32** -.06** 9. Number of Employees 51.09 95.82 876.80 .00 -.27** -.05* -.04* .03 .09** -.14** .13** -.12** 10. Market Capitalization 14545.20 39278.01 508329.45 .00 .17** -.03 .01 -.07* .21** -.02 .06** -.03 .30** 11. R&D Intensity .12 .68 15.49 .00 .11** .01 -.03 -.01 .05* .57** -.03 .33** -.06** -.03 12. Current Ratio 2.31 2.23 44.01 .00 .30** .01 -.05* -.02 .00 .89** -.07** .29** -.23** -.10** .49** 13. Economic Cycles 0.03 0.02 0.07 -0.02 .078** -0.01 -0.01 .15** .09** 0.01 -0.01 0.00 -0.01 0.01 0.01 0.02 *p < .05, **p < .01

(32)

MATTHEW THOMAS 29

and adjusted R2 of .303, a 1.1% increase over the model with only controls entered. Cycling Frequency itself is significant at the 99.9% confidence level, with a standardized β = .113, making it more influential than most individual industry effects (save SICs 29, 37, 42) but approximately half as influential as the other controls. Given these results, hypothesis 1 is supported.

Table 2: Direct Effect Hierarchical Multiple Regression Model

Model 1 Model 2 B SE B β B SE B β (Constant) 0.10*** 0.02 0.07*** 0.02 Number of Employees 0.00*** 0 -0.23 -0.01*** 0 -0.23 Market Capitalization 0.00*** 0 0.34 0.00*** 0 0.34 R&D Intensity -0.01 0.01 -0.03 -0.11 0.01 -0.02 Current Ratio 0.04*** 0 0.24 0.04*** 0 0.25 GDP Growth 1.36*** 0.35 0.07 1.34*** 0.35 0.07

Industry Effects (Summary) -2.35 1.74 -0.71 -1.83 1.73 -0.52

Cycling Frequency 0.037*** 0.01 0.11

Adjusted R2 _0.34*** _0.325***

F 33.94 33.37

Δ Adjusted R2 _0.33 _0.011

(33)

MATTHEW THOMAS 30

Additional Analysis: Cycling Frequency and Intensity

With the main analysis completed, further attention was paid to the frequency and intensity of cycling events, and the implications this has on performance. The descriptive statistics in table 3 (significant with ANOVA, p = .002) demonstrate the variation in performance achieved through increasing levels of cycling, indicating a curvilinear relationship with a peak at around three cycling events per five years.

With regard to cycling intensity, a similar curvilinear relationship arises when average

Q values are plotted against thresholds of cycling intensity. As previously discussed, the

cycling intensity variable takes a value between zero and two. The continuum was divided into eight 0.25 segments and the average Q at each point was taken, resulting in the following figure. To some extent, cycling intensity improves performance, although overlarge cycles incur performance penalties.

Fig. 2 – Cycling Intensity v. Performance

-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.25 0.5 0.75 1 1.25 1.5 1.75 2 A ve ra ge Q Cycling Intensity

Table 3: Cycling Frequency Descriptives

Dependent Variable: Tobin’s Q* Cycling Frequency Mean N SD 0 2.03 1638 2.29 1 2.16 225 1.78 2 2.41 235 1.74 3 2.92 109 4.26 4 1.88 30 1.10 5 2.51 5 1.66 Total 2.13 2242 2.33 *Untransformed **ANOVA significant at p < .05

(34)

MATTHEW THOMAS 31

To extend this analysis, a contour plot was used to visualize the contingent relationship between the three variables. Unfortunately, the chart is somewhat “lumpy” due to sample size, but a string of high points forming a diagonal between a frequency of four and intensity of one is suggestive of the prime place for a firm to position itself on this chart.

Impact of Slack

Given the setup of the model, slack is tested in two parts.

Fig. 3 – Contour Plot of Cycling Frequency, Intensity, and Performance

***χ2_{test significant at p < .000} Intensity Fr eq ue nc y 2.0 1.5 1.0 0.5 5 4 3 2 1 > – – – – < 0.0 0.0 0.1 0.1 0.2 0.2 0.3 0.3 0.4 0.4 Q

Fig. 4 – Slack Hypotheses

Cyclical

Ambidexterity Performance (Q)

Slack

H2B

(35)

MATTHEW THOMAS 32

The first step in assessing the relationship is moderation analysis, using the

PROCESS macro for SPSS. A multiple additive moderation (MAM, PROCESS model 2) is used, however software limitations restrict MAM in PROCESS to two moderator variables. To skirt this issue, a compositional approach of three discrete MAM models is used to derive the second-order interactions between all three moderator variables. Number of employees, market capitalization, R&D intensity, current ratio, and North American GDP growth are entered as controls as per the initial regression analysis. Variables were mean centred.

In model 1, unabsorbed slack is not a significant predictor of Q, although it has a significant positive interaction effect. Potential slack is a significant predictor, but the

interaction effect is only weakly significant (.1 > p > .05). The combined interaction effect is significant (p = .001), however it only has limited explanatory power (Δ R2 = 0.5%). In model 2, unabsorbed slack continues to not be a significant predictor, and the interaction

effect becomes weakly significant. Absorbed slack is very significant with relatively large

positive coefficients for both the predictor variable and interaction effect. The combined

interaction effect provides significantly more explanatory power than the two variables alone

(1.7% Δ R2 vs. 0.1% and 1.35% respectively). Finally, in model 3, both slack variables are

significant predictors, however only absorbed slack has a significant interaction effect.

Though the combined effect is again larger, it is not notably so. Overall, absorbed slack has

the largest impact, with a strong positive moderation effect present when modeled with both

other variables. Unabsorbed slack further contributes to a positive moderation effect,

(36)

MATTHEW THOMAS 33

H2A is unsupported, as while the results of the analysis are significant, the interaction

effect is positive, contrary to the negative relationship hypothesized.

Returning to hierarchical multiple regression, the analysis moves to test the effect of slack on cyclical ambidexterity. In model 1, the same battery of controls as used to test H1

are entered. Consideration was given to pruning insignificant predictors. However as there was minimal impact on R2 by leaving them in, no changes were made. This model is significant (p < .000), with F (31, 2114) = 8, with R2 = 7.6%. Slack variables are then entered in model 2. Of these variables, only unabsorbed and potential slack are significant, echoing results from the MAM analysis, with p = .001, .049 respectively), producing an increase in R2_{of 0.4% in an overall significant model where F (34, 2145) = 6, p < .000.}

Table 4: Slack Multiple Additive Moderation Models

Model 1 Model 2 Model 3

B SE B SE B SE

(Constant) 0.07*** 0.02 0.08*** 0.02 0.08*** 0.02

Cycling Frequency 0.03*** 0.01 0.04*** 0.01 0.03*** 0.01

Unabsorbed Slack 0.00 0.01 0.00 0.01

Unabsorbed Slack Interaction 0.02*** 0.01 0.01 0.01

Potential Slack -0.01*** 0.00 -0.01*** 0.00

Potential Slack Interaction -0.01* 0.00 0.00 0.00

Absorbed Slack 0.23*** 0.02 0.23*** 0.02

Absorbed Slack Interaction 0.23*** 0.04 0.24*** 0.04

Number of Employees 0.00*** 0.00 0.00*** 0.00 0.00*** 0.00

Market Capitalization 0.00*** 0.00 0.00*** 0.00 0.00*** 0.00

R&D Intensity -0.01 0.01 -0.03** 0.01 -0.03** 0.01

Current Ratio 0.04*** 0.01 0.04*** 0.01 0.04*** 0.00

GDP Growth North America 1.35*** 0.37 1.37*** 0.36 1.36*** 0.36

R2 _.256*** _.281*** _.284***

F 73.45 83.52 84.64

Δ R2_{from Interaction} _.005** _.017*** _.017***

(37)

MATTHEW THOMAS 34

In order to assess the net effects of slack on cycling frequency, the B and β values are compared for unabsorbed and potential slack. The two present slack variables have inverse polarity coefficients, presenting countervailing effects. While the coefficient for unabsorbed slack is approximately four times larger in both cases than that of potential slack, this is not in and of itself conclusive. To assess which effect dominated, the values of unabsorbed and potential slack in the dataset were multiplied by their respective coefficients and the difference taken to see which effect dominated the most frequently. The unabsorbed slack effect was stronger in 54% of cases, indicating the overall effect of slack on cycling frequency is positive. For this reason, H2B is not supported as it assumed a negative

correlation.

Table 5: Slack Multiple Regression Model

Model 1 Model 2 B SE B β B SE B β (Constant) 0.91*** 0.07 1.014*** 0.08 Number of Employees 0.00 0.00 -0.02 0.00 0.00 -0.02 Market Capitalization 0.00*** 0.00 0.05 1.24* 0.00 0.04 R&D Intensity -0.06* 0.03 -0.04 -0.10*** 0.04 -0.07 Current Ratio -0.02 0.01 -0.04 -0.07*** 0.02 -0.16 GDP Growth North America 0.54 1.20 0.01 0.54 1.20 0.01 Industry Effects -13.74 5.93 -1.72 -13.96 5.94 -1.77 Unabsorbed Slack 0.08*** 0.03 0.15 Potential Slack -0.02** 0.01 -0.04 Absorbed Slack 0.01 0.06 0.00 Adj. R2 _.076*** _.080*** F 6.519 6.328 Δ Adj. R2 _0.076 _0.04

(38)

MATTHEW THOMAS 35

Impact of Environmental Dynamism

Once again, testing environmental dynamism is a two-part test, and follows closely from the analysis of slack.

To test the moderation effect of environmental dynamism, the PROCESS macro is again used. Both market velocity and market ambiguity are entered as moderator variables into a MAM model in combination with cycling frequency as the independent variable, and the same battery of controls as used previously. Variables were mean-centred. No

significant interaction effects are found for either of the variables individually (p = .78, .79), or in combination, and subsequently no increase in R2_{is found from these effects. However,}

both variables are found to have a significant direct relationship with Q. To further probe the absence of an interaction effect, both constituent variables of the environmental dynamism construct were entered into separate simple moderation models (PROCESS model 1). Results are near-identical to that found with the MAM model, with no significant interaction effects found. This lack of evidence resoundingly rejects H3A.

Fig. 4 – Environmental Dynamism Hypotheses

Environmental Dynamism Cyclical Ambidexterity Performance (Q) H3B H3A

(39)

MATTHEW THOMAS 36

Hierarchical multiple regression is used one last time to test the effect of

environmental dynamism on cyclical ambidexterity. Model 1 is identical to the control model used to test H2B, and is significant (p < .000), with F (31, 2114) = 8, with R2 = 7.6%.

Market Velocity and Market Ambiguity are entered in model 2, producing an increase in R2 of 1.4% in a significant model where F (33, 2114) = 7, p < .000. Both market velocity and market ambiguity are significant at the 99.9% confidence level. However, the coefficient for market ambiguity is extremely small. The coefficient for market velocity is negative,

showing as market velocity increases firms cycle less frequently, and hence are less likely to be cyclically ambidextrous. Thus, H3B is supported.

Table 6: Environmental Dynamism Moderation Models

Model 1 Model 2 Model 3

B SE B SE B SE

(Constant) 0.07*** 0.02 0.07*** 0.02 0.07*** 0.02

Cycling Frequency 0.03*** 0.01 0.03*** 0.01 0.03*** 0.01

Market Velocity -0.22*** 0.07 -0.21*** 0.07

Market Velocity Interaction 0.02 0.07 0.02 0.07

Market Ambiguity 0.00** 0.00 0.00*** 0.00

Market Ambiguity Interaction 0.00 0.00 0.00 0.00

Number of Employees -0.01*** 0.00 -0.01*** 0.00 -0.01*** 0.00

Market Capitalization 0.00*** 0.00 0.00*** 0.00 0.00*** 0.00

R&D Intensity -0.02* 0.01 -0.02* 0.01 -0.02* 0.01

Current Ratio 0.04*** 0.00 0.04*** 0.00 0.04*** 0.00

GDP Growth North America 1.33*** 0.37 1.40*** 0.37 1.27*** 0.37

R2 _.255*** _.252*** _.251***

F 73.22 90.12 89.70

Δ R2_{from Interaction} _0.00 _0.00 _0.00

Virtuous cycles : an empirical examination and extension of cyclical ambidexterity theory

Author: M. S. D. Thomas – 11392274

MSc in Business Administration – Strategy Track

Amsterdam Business School, University of Amsterdam

Final Version – June 23, 2017

Supervisor: M. P. Tempelaar

Virtuous Cycles

AN EMPIRICAL EXAMINATION AND EXTENSION OF

CYCLICAL AMBIDEXTERITY THEORY

Statement of Originality

Abstract

Contents

Introduction

Literature Review

Theoretical Groundwork

From Statics Types to Dynamism

Contributions

Hypothesis Development

Performance

Slack

Environmental Dynamism

Methods

Data and Sample

Operationalization of Constructs

Cyclical Ambidexterity

Performance

Slack

Environmental Dynamism

Control Variables

Results

Impact of Cyclical Ambidexterity on Performance

Additional Analysis: Cycling Frequency and Intensity

Impact of Slack

Impact of Environmental Dynamism