Opportunity/Threat Perception and Inertia in Response to Discontinuous Change: Replicating and Extending Gilbert (2005)

Generally speaking, discontinuous change is “[external change that requires] internal adaptation along a path that is nonlinear relative to a firm’s traditional innovation trajectory” (Gilbert, 2005: 742). Specifically, Gilbert (2005) builds on literature that views discontinuous change as challenging the established paradigm regarding how value is created and captured in a given industry (Christensen & Bower, 1996; Dosi, 1982; Tushman & Anderson, 1986), which comprises three critical aspects (Kammerlander et al., 2018). First, discontinuous change introduces a new mix of benefit dimensions so that, at least initially, innovations may underperform existing approaches in terms of traditional benefits while offering new, and new bundles of, benefit dimensions (Christensen, 1997). Second, discontinuous change involves fundamentally new processes and structures of transforming inputs into benefit (Christensen & Bower, 1996) and is, thus, competence destroying (Anderson & Tushman, 1990; Sosa, 2011) as it requires “fundamentally new skills and competences” (Tushman & Anderson, 1986: 444). Third, discontinuous change involves a new approach to transforming value into profit, especially in terms of revenue and pricing structures (Christensen, 2006; Markides, 2006).

As discontinuous change challenges existing paradigms—that is, mental models and activity patterns that are rooted in deeply embedded shared principles, beliefs, and norms (König et al., 2012)—the literature on organizational adaptation to discontinuous change has particularly benefited from the perspective of MOC (Weber, Lehmann, Graf-Vlachy, & König, 2019). MOC research illuminates how cognitive and emotional structures and processes affect the ways in which members of organizations sense, interpret, decide, and act in response to the environment (Daft & Weick, 1984; Eggers & Kaplan, 2013; Hodgkinson & Healey, 2008; Walsh, 1995). Studies investigating adaptation to discontinuous change from a MOC perspective have primarily highlighted that decision makers’ knowledge structures are bounded and difficult to change (Barr, Stimpert, & Huff, 1992; Kiesler & Sproull, 1982) and, in turn, can make decision makers overlook discontinuous change, interpret it in ways that confirm their schemas, decide against adopting it, and implement responses that reflect the established rather than the new mind-set (Kaplan, 2008a; T. Levitt, 1960). Tripsas and Gavetti (2000), for example, showed how managers at Polaroid interpreted digital imaging through the outdated prism of Polaroid’s traditional “razor-and-blade” model of selling film cartridges at margins that were high compared to those of camera hardware; Kammerlander and colleagues (2018) revealed how decision makers struggle with discontinuous change as it challenges what they perceive as their organization’s identity; Vuori and Huy (2016) revealed how managers at Nokia, depending on their hierarchical level, developed different kinds of fears when Nokia was attacked by Apple’s iPhone. Notably, these and other MOC-based studies help explain not only why incumbents succumb to discontinuous change even when they possess complementary assets but also why, sometimes, incumbents overcome deeply anchored tendencies of organizational inertia (e.g., Gerstner et al., 2013).

Gilbert (2005) also applies an MOC perspective to discontinuous change and adaptation, although he sheds light on an element of cognition that played no central role in research on discontinuous change prior to his own: decision makers’ perception of the discontinuous change as a threat or an opportunity (Dutton & Jackson, 1987).¹ Specifically, analyzing the responses of eight newspaper companies to the rise of online news, he induces two key ideas. First, he proposes that firms faced with discontinuous change may exhibit resource rigidity and routine rigidity—that is, two distinct types of organizational inertia. Gilbert argues that resource rigidity is driven by resource dependence (Pfeffer & Salancik, 1978), especially from investors and high-end customers (Christensen & Bower, 1996), and “incumbent position reinvestment incentives” (2005: 747)—that is, the tendency of incumbents to reject innovations that cannibalize the established market position (Chandy & Tellis, 1998). As for routine rigidity, he refers to research showing that routines, over time, become anchored in organizational structures and managers’ minds and activities (Feldman & Pentland, 2003; Teece, Pisano, & Shuen, 1997) in ways that render firms less likely to depart from the proven innovation paradigm (March & Simon, 1958).

Second, Gilbert (2005) induces that resource rigidity and routine rigidity are influenced in intriguing ways by decision makers’ perception of a discontinuous change as a threat rather than as an opportunity. Following Dutton and Jackson (1987) and Jackson and Dutton (1988), he conceptualizes opportunity/threat perception as the aggregate of three covariant attribute pairs: positive/negative, gain/loss, and high/low control. Thus, he interprets any instance in which decision makers describe the discontinuity as something positive and/or as a gain and/or as being under the focal firm’s control as opportunity perception, whereas he treats decision makers’ descriptions of the discontinuity as something negative and/or tied to a loss and/or beyond their and their firm’s control as indicative of threat perceptions of the discontinuity.

Particularly, Gilbert (2005) distills five nomological propositions, which we refer to as Propositions G1 through G5. Proposition G1 states, “The perception of an imminent threat in the face of discontinuous change enables managers to overcome sources of resource rigidity that stem from resource dependence [G1a] [and those that] stem from incumbent position reinvestment incentives [G1b]” (Gilbert, 2005: 747). In fact, he notes that “[under the conditions of discontinuous technological change] anticipation of a threat to the core business appears necessary to motivate response [italics added]” (Gilbert, 2006: 162). Gilbert (2005) theoretically undergirds this proposition by referring to prospect theory (Kahneman & Tversky, 1979), which suggests that decision makers engage in stronger risk seeking and externally directed reaction to an event if they perceive it as negative and as a loss (Audia & Greve, 2006; Ocasio, 1995).

Proposition G2 predicts that “[decision makers’] perception of [a discontinuous change as] an imminent threat leads to a contraction of authority [G2a], a reduced level of experimentation [G2b, and] a focus on existing resources [G2c] that amplif[y] routine rigidity” (Gilbert, 2005: 749–751). G2 echoes threat rigidity theory (Staw et al., 1981), especially the idea that opportunity/threat perceptions are evaluative appraisals (Lazarus, 1991), which influence both decision makers’ “cold,” reflective processing of information and their “hot,” reflexive and affective processing and reactions (Dutton & Jackson, 1987: 79). As Dutton and Jackson (1987: 82) note, “Evaluative appraisals are the affective components of cognitions; they make cognitions ‘hot.’” This might be crucial in the context of discontinuous change as “visceral (felt) reactions to choice alternatives often overpower evaluations based on subjective probability assessments” (Hodgkinson & Healey, 2014: 5). Dutton (1993: 200), for instance, specifically suggests that “opportunity frames are almost irresistible because of the positive ‘charge’ or emotion . . . that [they] evoke.” Conversely, threat is “a deep sense of vulnerability” (Gilbert, 2005: 742) that typically leads decision makers to constrict control and to exclude other organization members from participating in decision making (Staw et al., 1981). Additionally, faced with a threat—and the unpleasant emotions associated with it—subordinates are inclined to avoid responsibility for strategic decisions (Nutt, 1984). Especially interesting in this regard is Gilbert’s (2005) observation that threat framing reduces experimentation because it relaxes resource rigidity (see G1): The more resources managers possess to implement discontinuous change, the more difficult it becomes for them “to step back and change behavior” (Gilbert, 2005: 751).

Gilbert’s (2005) three final propositions distill his observations of how newspaper companies varied in their abilities to overcome resource rigidity and routine rigidity. Proposition G3 states, “Involving outside influence [e.g., consultants] when deciding how to respond to discontinuous change will increase the likelihood that managers will structurally differentiate a new venture from its parent organization” (Gilbert, 2005: 755). Proposition G4 predicts that “structural differentiation can help decouple threat perception in a parent from an opportunity perception in a new venture” (Gilbert, 2005: 755). Proposition G5 notes that “outside influence, structural differentiation, and opportunity framing combine to relax routine rigidity in a new venture” (Gilbert, 2005: 757). These propositions relate to research on the benefits of external networks (Geletkanycz & Hambrick, 1997; Stuart & Podolny, 1996) and studies on the role of structural ambidexterity in the context of discontinuous change (e.g., O’Reilly & Tushman, 2008).

Gilbert’s research is widely regarded as a substantial contribution to management research.² His notions of resource rigidity and routine rigidity are relevant to MOC-focused research on discontinuous change, particularly as they highlight that managers not only need to recognize discontinuous change and gather specific capabilities but also must “mobilize those capabilities in taking strategic action” (Eggers & Kaplan, 2013: 312). Gilbert’s idea to disentangle inertia has spurred other researchers to refine adaptation to discontinuous change even further, allowing them to develop more precise theories (König et al., 2013). Generally, his work has added substantially to scholars’ shift “away from viewing the inertia of incumbent firms as an inevitability” (Gerstner et al., 2013: 258) toward a focus on the heterogeneity of incumbent adaptation—that is, the standard perspective today (Eggers & Park, 2018).

Gilbert has also received attention in the wider MOC literature (e.g., Parmigiani & Howard-Grenville, 2011; Voss, Sirdeshmukh, & Voss, 2008), primarily because he adds to the ample body of studies investigating the impact of managerial appraisals on a diverse set of organizational outcomes (e.g., Audia & Greve, 2006; Bockmühl, König, Enders, Hungenberg, & Puck, 2011; Chattopadhyay et al., 2001; Fiegenbaum & Thomas, 1988; Ocasio, 1995; Shimizu, 2007). In particular, he offers an innovative reconciliation of the implications of threat rigidity theory (Staw et al., 1981) with those of prospect theory (Huff et al., 1992; Kahneman & Tversky, 1979; Lant, Milliken, & Batra, 1992). He suggests that both theories hold, although one explains the relaxation of resource rigidity while the other explains the exacerbation of routine rigidity. More broadly, Gilbert’s research has resonated with scholars studying MOC and the overall “microfoundations” of management (e.g., Aggarwal, Posen, & Workiewicz, 2017; Felin, Foss, & Ployhart, 2015). These authors have long noted the role of reflexive cognitive processes (Hodgkinson & Healey, 2011) in contexts of highly ambiguous, unstructured, and uncertain situations, including radical change (Huy, 2002) and crises (König, Graf-Vlachy, Bundy, & Little, 2020). Vuori and Huy (2016), for example, refer to Gilbert in their study of the combined implications of “cold,” deliberative cognition and affect-infused, “hot” cognition, such as fear, in Nokia’s battle against Apple’s iPhone.

Finally, Gilbert’s work has left traces in the methodological and practitioner-oriented literature. Eisenhardt and Graebner (2007: 29) lauded it for the rich display of qualitative evidence, calling it an “excellent example . . . of blending construct tables with selected text descriptions,” and it has been showcased in a number of other influential methodological articles (e.g., Christensen, 2006; Kouamé & Langley, 2018). Gilbert himself translated his findings into practitioner-oriented publications (e.g., S. Anthony, Gilbert, & Johnson, 2017; Gilbert, 2003; Gilbert & Bower, 2002), providing oft-cited recommendations on how to overcome organizational inertia.

Our study is motivated not only by the general interest in replication of empirical work in management (Brandt et al., 2014; Tsang & Kwan, 1999) and the replication logic innate to qualitative research (Glaser & Strauss, 1967) but also by specific limitations of Gilbert’s (2005) work. In particular, limiting its generalizability, Gilbert analyzes relatively similarly sized quasi-monopolists that possess substantial slack resources. However, such firms can be expected to show idiosyncratic strategic practices (Voss et al., 2008) including especially strong incumbent position reinvestment behavior (R. Gilbert & Newbery, 1984; Reinganum, 1983). Moreover, their strategic position and available resources might influence the nature and intensity of organizational decision makers’ perception of and reaction to organizational challenges (e.g., Audia & Greve, 2006; Barr & Glynn, 2004; Lehner, 2000; Nohria & Gulati, 1996). The same is true for characteristics of the specific cultural context of the United States (House, Hanges, Javidan, Dorfman, & Gupta, 2004), which could interfere with mechanisms central to Gilbert’s model.

More importantly, Gilbert’s (2005) three-dimensional conceptualization of opportunity/threat perceptions contradicts other related theories and evidence, raising concerns about the internal validity of his research. In particular, in contrast to his definition (Dutton & Jackson, 1987), theory on appraisal—one of the most influential bodies of psychology—envisions perceived control as a distinct part of situation appraisal (Lazarus, 1991, 1993). Studies on managerial sensemaking published before Gilbert’s have also considered gain/loss framing and perceptions of control as discriminant dimensions of interpretation (Thomas et al., 1993).

We deem these contradictions especially important in light of the fact that Gilbert’s (2005) propositions—although they echo some prior findings (Huff et al., 1992; Lant et al., 1992)—contradict several extant theories and empirical evidence. Particularly, a substantial body of literature indicates that threat perceptions regarding a given strategic issue, including perceptions of lacking control, can also taper organizational resource commitment in response to this issue (e.g., Brown & Starkey, 2000; Milburn, Schuler, & Watman, 1983; Staw et al., 1981). Relatedly, prospect theory suggests that, at extreme points, when survival is threatened, organization members are likely to resign and allocate progressively fewer resources (Fiegenbaum & Thomas, 1988; Lehner, 2000; March & Shapira, 1987). Also counter to Gilbert, studies suggest that opportunity framing may not only induce cognitive flexibility but can also create positive “gloss” (Dutton, 1993: 199-200) and emotions (Dutton, 1988). Given that positive emotions can induce risk taking (Isen & Geva, 1987), opportunity frames could lead to higher commitment and top management involvement in response to discontinuous change—a conjecture that is supported by other studies observing increased resource commitment to new technologies under opportunity perceptions (e.g., Sharma, 2000; White, Varadarajan, & Dacin, 2003).

We undertook 10 longitudinal case studies in two industries. For the empirical generalization, we analyzed six incumbents in the German book-retailing industry; for the generalization and extension, we revisited these cases and added four incumbents from the European mobile telecommunication industry. We chose these two settings, first, because both industries were affected by discontinuous changes akin to that studied by Gilbert (2005): the emergence of online retailing in the book-retailing industry and the emergence of MVNOs in the telecommunication industry. Second, these settings simultaneously diverged to different degrees from the U.S. newspaper industry studied by Gilbert. The German book-retailing industry was relatively fragmented, whereas U.S. newspapers were quasi-monopolists (Gilbert, 2005). The European mobile telecommunication industry, while being an oligopolistic market, had been a traditionally highly dynamic environment, whereas the established newspaper market was rather stable prior to the advent of the Internet (Gilbert, 2001).

For physical book retailers, online retailing presented a discontinuity that was bound to especially high levels of uncertainty and required nonlinear internal adaptation, as defined in our Theory Background section. First, online book retailing initially fell short on customers’ established performance criteria but introduced new features, such as convenience and customized recommendations (Schrape, 2011). Second, online retailing required sophisticated information technology competencies, which were less relevant in the established business. As Jeff Bezos, founder and CEO of Amazon, noted, “[Online book retailing] is a technology business, and [physical book retailing is] a real estate business” (Beunza & Garud, 2007; Collura & Applegate, 2000: 27). Third, online retailing also introduced new revenue categories, such as banners and online marketplaces for secondary sellers (Altman & Tripsas, 2013).

Our observation period spanned the years from 1993, the year the German book retailer Lehmanns launched the country’s first online book retailing site, to 2008, when most players had come to view Amazon’s model of online retailing as the dominant design. We selected the cases based on a theoretical and “purposeful” (Patton, 1987: 51) sampling logic. Specifically, the organizations had to be, first, in the industry’s top 100 (as listed in the trade magazine Börsenblatt) to ensure a comparable size to that of the organizations studied by Gilbert (2005); second, at least 10 years old at the time of the discontinuity to ensure residual fit with the traditional business model (Gilbert, 2006); and third, sufficiently heterogeneous regarding the theoretical variables of interest to enable us to juxtapose polar types (Eisenhardt, 1989). The top left of Table 1 describes the six book retailers we studied. As all companies were promised strict confidentiality, we disguised firms’ names and precise figures. With the exception of Il Libro, which was a large division of a conglomerate, all book retailers were stand-alone businesses.

As part of the generalization and extension, we additionally explored the responses of four mobile network operators (MNOs) from three European countries to the emergence of no-frills MVNOs between 2000 and 2006. MVNOs provide mobile telecom services to end customers by reselling wholesale minutes they purchase from MNOs. In stark contrast to MNOs, MVNOs do not own infrastructure and focus on basic services, such as voice and text messaging. MVNOs constituted a discontinuous change as per our definition. First, the performance metrics stressed by MVNOs—low prices, simple tariff structures, and basic services—diverged considerably from those pursued by traditional MNOs. Second, transitioning to an MVNO business model would render existing resources and capabilities obsolete. For instance, MVNOs did not need to use the best possible network, and a reputation for service and reliability was worth much less in the MVNO context than in the MNO context. On the contrary, MVNOs needed to run a lean organization with a low cost base (Dahlström, Deprez, & Steil, 2004). Third, MVNOs had a very different approach to transforming value into profit because they sold their offerings primarily via the Internet, had a very simple tariff structure with little price differentiation, and, at least initially, did not sell subscriptions bundled with subsidized devices. Thus, as late as 2005, industry observers wondered whether MVNOs “could have the same impact as no-frills airlines have had on the aviation industry” (Marketing Week, 2005: 26) and highlighted the fact that MVNOs, if implemented by MNOs, might cannibalize their revenue, shift profit pools, and ultimately, change the profitability of the incumbent technology (Christensen, Kaufman, & Shih, 2008). The bottom portion of Table 1 provides an overview of the four MNOs, all of them national subsidiaries of European telecom firms. Three were market leaders in their respective countries, and one was in third place. Each firm launched its own MVNO in 2005 or 2006.

Like Gilbert (2005), we developed longitudinal, embedded case accounts (Gibbert, Ruigrok, & Wicki, 2008; Yin, 1994) of perceptions, decision making, and decision implementation among the senior decision makers at the levels of the parent organization as well as the venture unit, that is, in the unit responsible for developing a response to the discontinuity. We collected rich data as shown in Table 1. Our most important source was a total of 45 in-depth interviews with company representatives. Each interview lasted on average 80 min and was conducted between 2006 and 2009. We interviewed top executives and middle managers because both are crucial for the allocation of monetary and attentional resources (Bower, 1970; Burgelman, 1983) and the implementation of routines as well as change processes (Huy, 2001, 2002, 2011). In particular, middle managers might have a lower commitment to the status quo and exhibit greater awareness of discontinuities than top managers (Hill & Rothaermel, 2003), and corresponding cognitive states may propagate not only top down but also bottom up in an organization (Hodgkinson & Healey, 2008; Kammerlander et al., 2018). All interviews followed an interview guideline based on Gilbert (2001, 2005) and included problem-focused questions as well as narrative questions to elicit rich, anecdotal evidence (Daft, 1983; Kvale, 1996). To reduce possible retrospective bias, we (a) ensured interviewees’ confidentiality and gave them the opportunity to report their recollections freely, (b) worded questions to allow interviewees to not answer whenever they felt they could not make reliable statements, (c) compared information from multiple informants, and (d) repeatedly triangulated interview data with contemporary archival sources (Golden, 1992; Huber, 1985; Huber & Power, 1985; Miller, Cardinal, & Glick, 1997). Most interviews were conducted by two investigators (Eisenhardt, 1989), who took and then compared individual notes (Nadin & Cassell, 2006). Interviews were transcribed verbatim within 24 hr.

We collected extensive additional data for triangulation. Company-internal documents, company visits, follow-up conversations, visits of industry fairs, 23 expert interviews, and supplementary data collected through, for example, a student project with the German Publishers and Booksellers Association, a supplementary survey, and workshops with executives filled gaps in our historical accounts and provided a more ethnographic, intuitive understanding (Tsoukas & Chia, 2002). We also collected about 400 pieces of archival public data, including historical interviews with top managers, historical websites, analyst reports, and expert comments mentioning one or more of the companies during discussions on the industries’ development. Notably, at no time did we influence managers’ perceptions of the discontinuous change, nor did we interfere in any strategic decision of the firms we observed as part of our study.

The coding guidelines for our studies are provided in Table A2 in the online supplement. The empirical generalization followed Gilbert’s (2005) design as faithfully (Easley, Madden, & Dunn, 2000) as possible. We reconstructed a coding scheme (Mayring, 2003) from Gilbert (2001, 2005) as the basis of our structural content analysis (Miles & Huberman, 1994). As part of interpreting the findings of the empirical generalization and, even more so, during the generalization and extension, we carefully monitored the reliability of existing coding guidelines and added new codes, much in the spirit of an informant-based, inductive analysis (Gioia, Corley, & Hamilton, 2013). In particular, in the generalization and extension, we dropped the coding of “opportunity” versus “threat” perception and instead differentiated between the two dimensions of “gain/loss framing” and “high/medium/low perceived control” (Brockner et al., 2004). Notably, we added a code for “relevance” perceptions, which captured the degree to which interviewees perceived the respective discontinuity as a high-priority situation (Jackson & Dutton, 1988)—in other words, the intensity of attention or recognition (Kaplan, Murray, & Henderson, 2003). Relevance is orthogonal to the direction of perception, that is, gain or loss, but, as defined by research on appraisal and coping, equally essential to appraisal (Tomaka, Blascovich, Kibler, & Ernst, 1997). As described later, considering relevance perception was crucial to recognizing some instances that, at first, appeared to constitute anomalies to Gilbert’s model as theoretical replications (Yin, 1994). Throughout the various stages of the analysis, we aimed to maximize reliability. Cases were first coded independently by up to four coders. Intercoder agreement (Krippendorff, 2004; Landis & Koch, 1977) was high from the beginning of the analysis and increased to quasiperfect coherence toward the end. Remaining incongruent codings were dropped from the final analysis.

Both the empirical generalization and the generalization and extension include in-depth single case studies. In the empirical generalization, each case study was treated as a microreplication (Yin, 1994) of Gilbert’s (2005) research. Throughout the generalization and extension, each new case served as a replication of the evolving theoretical conclusions that we derived from previous cases (Eisenhardt, 1989). For each case, we compiled 10- to 50-page case write-ups reconstructing its “story” (Stake, 2006) along the chronological structure of Gilbert’s model.

Very early in our research, we noted that we had to extend Gilbert’s (2005) analytical approach to account for the iterative process of resource allocation that we observed in the book-retailing companies. Specifically, in line with Gilbert’s notion of resource allocation as an “iterative process” (Gilbert, 2006: 152; see also Bower, 1970; Bower & Gilbert, 2005), we induced that the decision makers of the studied companies did not perform the resource allocation process only once but went through the process repeatedly. As such, similar to Gilbert (2006) and prior related case studies (Benner, 2010; Kaplan, 2008b; Tripsas, 2009), we partitioned the companies’ responses into discrete “temporal brackets” (Langley, 1999: 703), or phases, each of which covered one cycle of resource allocation. We induced the phases in vivo (Glaser & Strauss, 1967) from our informants’ statements about “turning points” (Kaplan, 2008b: 733) in their own perceptions of the discontinuity as well as their accounts of changes in the shared perceptions among and the resource allocation behavior of decision makers.³ Following Langley (1999), we further ensured that each phase is clearly distinct from the preceding and following phases and characterized by a certain level of phase-internal coherence and specificity in the activities. For each phase, we then evaluated whether the evidence corroborated or contradicted Gilbert’s model (in the empirical generalization) or our own revised theory (in the generalization and extension).

Our observations strengthen Gilbert’s (2005) propositions about threat-induced routine rigidity and those about the effects of structural separation and external influence as relaxants of such manifestations of threat rigidity.⁴ First and foremost, all phases for which we had sufficient data represent literal or theoretical replications of propositions G2a, G2b, and G2c, which argue that threat perception—mediated by contraction of authority (G2a), a reduced level of experimentation (G2b), and a focus on existing resources (G2c)—amplifies routine rigidity, whereas opportunity perception relaxes it. In fact, as illustrated in the third column in Table 3, no firm in our sample whose managers perceived online book retailing as an opportunity showed either of these three intermediate types of response, and all cases where managers perceived online book retailing as an imminent threat showed at least two of the three indicators of routine rigidity. For instance, managers at Il Libro (Phase 2) began to increasingly perceive online retailing as a loss and getting out of control, leading to a contraction of authority in that the top management installed its own managers, including a new CEO, in the venture unit and eventually even decided to integrate most of the online store into the established organization. Such responses were especially remarkable because they were uncharacteristic of Il Libro’s decentralistic corporate culture, as emphasized by a number of informants. We observed reduced experimentation as a consequence of threat perception in the third phase at Book 2000, when managers began to (re)focus on its retail outlets and increasingly neglected the online store. Similarly, Bookies (Phase 2) stopped engaging in innovative marketing campaigns when threat perceptions increased. Finally, threat perception co-emerged with a focus on existing resources in many cases. Readme (Phase 2) began making investments in large stores, Book 2000 (Phase 3) invested in a new inventory management system to improve store operations, and Bookies (Phase 2) expanded the physical store’s catalog to defend the existing customer base.

Later, in our description of the results of the empirical generalization, we elaborate on the fact that our data confirm the overall effect of threat perception on reduced experimentation observed by Gilbert (2005; Proposition G2b) but intriguingly contradict him in terms of the underlying causal mechanism. In this regard, we did not observe what Gilbert had noted, namely, that “because the expansion [of the online units] occurred so quickly, the resources invested [in response to the threat perception] reinforced rather than reshaped established routines” (Gilbert 2005: 751). In contrast, we observed that all four instances where threat coincided with reduced experimentation that it was constrained resources and low perceived control that led managers and employees to refrain from pursuing experimental approaches to online book retailing. Thus, our data seem to indicate that perception of an imminent threat leads to a reduced level of experimentation not necessarily due to escalated threat-induced resource spending.

Furthermore, our findings support G3, which posits that outside influence fosters the structural separation of the venture unit. The only two cases we regard as merely theoretical replications are the first phase of the response of Book 2000 and the first phase of Bookies’ response. In both cases, outside influence was present, but structural differentiation did not take place. Two mechanisms explain this apparent contradiction of Gilbert’s (2005) proposition. First, in both cases, external influence was present only at the venture level and not at the parent units that made the decision on the ventures’ organizational setup. Second, in both cases, the external influencers were either information technology providers or traditional book wholesalers that had limited expertise in implementing organizational structures to address discontinuous change. In our other cases, as well as in Gilbert’s, the external influence was of a more strategic nature, for example, from external consultants (e.g., Il Libro, Phases 1 and 2; see Gilbert, 2005: 752).

Gilbert’s (2005) Proposition G4 is also consistently supported by our evidence. Gilbert suggests that under threat framing at the parent level, structural differentiation helps to decouple perception in the venture unit from that in the parent unit. This, in turn, allows those working in the venture unit to develop an opportunity frame. In contrast, if the venture unit is structurally integrated, perception in the venture unit remains coupled to the threat perception at the parent level. All phases for which we had sufficient data and to which the proposition was applicable—that is, threat perception was prevalent at the parent level—confirmed Gilbert’s proposition.

Finally, our evidence supports G5. All our cases provide literal replications of the proposition that outside influence, structural differentiation, and opportunity perception in the venture unit combine to relax routine rigidity (O’Reilly & Tushman, 2013).

Our data were substantially more heterogeneous regarding Propositions G1a and G1b, which suggest that the perception of a discontinuous change as an imminent threat enables managers to overcome resource rigidity stemming from resource dependence and incumbent position reinvestment incentives. Crucially important—because it legitimizes our choice of context for the replication—is the fact that the resource rigidity we observed was indeed driven by these two very causes, as laid out in Table A3 in the online supplement. For instance, regarding resource dependence, many of our interviewees argued like the CEO of Readme, who remarked, “I thought our customers wanted us to be a fine book retailer—nothing else.” Regarding incumbent position reinvestment, we found that most branch managers highlighted the danger of cannibalization, just like Gilbert (2005) had observed.

However, as highlighted in Table 3, we detected two types of anomaly that we could not discard as theoretical replications of G1a and G1b.

As suggested by Tsang and Kwan (1999), we began the generalization and extension by reflecting on our insights in light of other relevant literature. When we considered the anomalies to G1a and G1b, we noted that they might potentially be related to Gilbert’s (2005) conceptualization of opportunity/threat perception (Dutton & Jackson, 1987; Jackson & Dutton, 1988), which made him code any statement that denoted the discontinuity as negative, likely to result in loss, and/or largely out of one’s control as threat perception and any statement denoting it as positive, likely to result in gain, and/or largely within of one’s control as opportunity perception. Specifically, it became apparent that findings from other domains suggest that perceptions of “positive/negative” and “gain/loss” are conceptually distinct from perceptions of control (Thomas et al., 1993). In particular, research on appraisal and coping (Lazarus, 1966, 1991; Lazarus & Folkman, 1984) contradicts Gilbert’s assumption that positive/negative, gain/loss, and control/no control are all covariant indicators of opportunity/threat perception that have similar consequences. This research instead suggests that people generally appraise events by first assessing “its significance for the person’s well-being (primary appraisal) [and then] the available coping resources and options (secondary appraisal)” (Lazarus & Launier, 1978: 302). In other words, there is substantial evidence for a need to subsume the attribute pairs of “positive/negative” and “gain/loss” under one dimension of appraisal, and treat appraisals of “high/low perceived control” as a separate dimension. This notion has also received empirical support from research in the field of organization science (Brockner et al., 2004; Thomas et al., 1993; White et al., 2003). Moreover, it resonates with well-established theories in the realm of psychology, all of which highlight the relatively independent effect of control perceptions (e.g., Ajzen, 1991; Bandura, 1977; DeCharms, 1968; Rotter, 1966).

Consequently, in our generalization and extension, we adapted and expanded our coding scheme and the analysis in a theory-informed fashion. Specifically, we recoded all previously collected material for opportunity/threat perceptions to separately account for two dimensions: The first is “gain/loss framing,” under which we also subsumed the perceptions of positive/negative, as suggested by Thomas and McDaniel (1990). We use the term “framing” to account for the fact that gain/loss frames are the labels used in prospect theory (Kahneman & Tversky, 1979), which plays an important part in Gilbert’s theorizing, especially as Gilbert (2005) uses the term in his Proposition G5. The second is “perceived control,” which notably covers all instances of managers’ social construction of control in the currently perceived as well as the anticipated competitive context (see Table A2 in the online supplement for coding guidelines).

In keeping with Tsang and Kwan’s (1999) notion of generalization and extension, we also collected extensive additional data in our second empirical context, that is, the telecommunication industry. Doing so allowed us to observe not only more variance in framing but also additional combinations of perceptions that were missing from our initial data set. We compared the responses of the four European MNOs in our sample to the predictions of Gilbert’s (2005) entire model. However, for two reasons, we focused our further inquiry, that is, the generalization and extension, on the effects of frames and perceptions on resource rigidity (i.e., G1a and G1b). First, the empirical generalization did not reveal substantial anomalies regarding Propositions G2 through G5; and second, in our analyses of the MNO cases, we did not stumble upon further anomalies—neither to G2 through G5 nor to our subsequently revised theorizing presented in Propositions P3 and P4 (see Additional Extensions section). Tables 4 and 5 summarize our extended observations in the book-retailing and telecommunication industries, respectively, ordered alphabetically by case and response phase.

Our analyses revealed various instances that, first, violated Gilbert’s (2005) premise of inherently covariant perceptions of “gain/loss” and “perceived control” and, second, demonstrated that one can reliably distinguish three (as opposed to Gilbert’s two) distinct levels of perceived control. First, in the case of Readme (Phase 2), managers at the parent level predominantly perceived online book retailing not only as highly relevant for their organization but also clearly as a gain. However, at the same time, they also perceived low levels of control over the future of that business.⁵ For example, Readme’s CEO explained,

The Internet is a great opportunity, also for book retailers. Yet, to my eyes, we just don’t have the same resources as Amazon or Bol.de and all those big ones. This has huge implications. For instance, we can’t persuade other companies . . . to link their home page with ours, and so forth.

Our adapted coding scheme allowed for greater nuance than Gilbert’s original coding scheme. According to Gilbert, the preceding statement should have been coded as an indication of threat, given the low level of perceived control it reflects. Conversely, our new guidelines suggest that it expresses a gain frame (“great opportunity”) coupled with low perceived control.

Second, our data include co-occurrences of imminent loss framing with high levels of perceived control in the mobile telecommunication industry (parent units of Alpha and Cell4U, both Phase 1). For instance, decision makers of Alpha noted that the emergence of no-frills MVNOs would lead to economic losses: “It will cause a lasting change in the market, and you cannot recover [the losses associated with] that.” However, they simultaneously perceived high levels of control over the situation: “We knew everything that needed to be done if that market development were to take place.”

Third, we observed cases in which the decision makers of the respective companies unanimously framed the discontinuity either as a substantial gain or as an imminent loss, respectively, but perceived only moderate levels of control (parent units of Bookies, Phase 1; Il Libro and Jubilados, Phase 2; Alpha, Cell4U, Moby, and Ring Me, all Phase 2). For instance, in the book-retailing industry, the CEO of Jubilados’ online specialist business framed online book retailing as a gain and something eventually positive, while noting a moderate level of control:

We know that Amazon is extremely strong, and that limits our potential, definitely. However, there is a small niche where we can thrive and grow profitably. . . . Today, all of us know that it is possible to make money on the Internet, and that’s what we do.

Unbundling opportunity/threat perception into two distinct dimensions enabled us to induce three thus-far undescribed patterns of resource allocation behavior in response to discontinuous change under loss framing.⁶

Treating perceived control as a distinct dimension of appraisal is also instrumental for explaining high resource commitment to discontinuous change under what Gilbert (2005) coded as opportunity perception—an anomaly to G1 we observed in the empirical generalization. The pattern we induced here relies on considering the additional dimension of the perceived “relevance” of the discontinuity—that is, as conceptualized in our Method section, the degree to which the change is on top managers’ cognitive agendas (Gerstner et al., 2013; Kaplan et al., 2003; Ocasio, 1997) and perceived as a high-priority situation (Jackson & Dutton, 1988). Notably, whenever decision makers framed the discontinuous change as a loss, they also perceived it to be a relevant event. In contrast, gain framing co-emerged with varying levels of perceived relevance.

Specifically, as illustrated in Tables 4 and 5, under gain framing, three patterns emerged from our analysis. The first pattern is gain framing, low relevance perception, moderate/high perceived control, and resource rigidity. We identified gain framing accompanied by moderate or high levels of perceived control in the first phases of book retailers Book 2000 and Readme and MNOs Ring Me and Moby.⁸ However, our coding revealed that the decision makers in the mentioned book-retailing cases perceived online retailing as a fairly irrelevant issue. Not surprisingly, this resulted in limited resource commitment to the discontinuous change. During the first phase of Book 2000’s response, for example, gain framing and high control perceptions were dominant across all hierarchical levels. The CEO reminisced, “Of course, in those moments you think you can shape the future.” Yet, top management also felt that online book retailing was not yet strategically relevant. The CEO stated, “In the beginning, we did not take [it] too seriously.” Other managers confirmed that he did not pay much attention to online retailing because he had not recognized its importance. As a result, the online store did not receive financial and operational support from the organizational apex.

The second pattern is gain framing, high relevance perception, low perceived control, and resource rigidity. In one case, Readme’s second response phase, managers perceived the discontinuity as a highly relevant gain but, as indicated already, simultaneously sensed low levels of control. In response to this perception of control-reducing constraint, Readme’s decision makers not only cut investments in online book retailing but also shut down its entire self-operated online business.

The third pattern is gain framing, high relevance perception, moderate/high perceived control, and relaxation of resource rigidity. In two of the relevant cases, Book 2000’s second response phase and Bookies’ first response phase, managers perceived online retailing not only as a highly relevant issue that could have a decidedly positive impact on their business but also as an issue over which they had moderate or high control. Together, these perceptions led to a high level of resource commitment to the discontinuous change. For example, in Book 2000’s second response phase, decision makers framed the discontinuity as a controllable gain, especially when their company had the opportunity to further develop the online business in cooperation with an external nonprofit, a state-funded institution, and a publishing company. The CEO noted,

We [thought we could] gain a real competitive advantage, which proved to be right, because we were one of the early ones in the online business.

Moreover, in contrast to the view adopted in the company’s first response phase, Book 2000’s CEO and his team now perceived online retailing as a strategically relevant issue and, in turn, invested heavily in the development of their online specialist business. The floor manager who had founded the online business explained,

Once you could see that there was a real business, of course the CEO scheduled a meeting with us. . . . I think [top management became engaged] when we reached the first 100,000 Deutsche Mark. If sales exceed such a level, our CEO personally takes care of the business.

In the first phase of Bookies’ reaction, decision makers viewed online book retailing as just as relevant and only slightly less positive and controllable than did the managers at Book 2000 (Phase 2). In addition, our data indicate that because of these perceptions, Bookies decided to invest substantial financial and operational resources in its online business. The CEO explained,

We saw that everyone was going online. Many of my competitors suddenly had an online store. I also found it compelling and somehow fascinating, although I had my doubts. In the end, we said, “We need to open such a store, too.”

The first of the three patterns resembles the one observed by Gilbert (2005): The combination of gain framing and high levels of perceived control—that is, the combination Gilbert labels “opportunity perception”—does not suffice to overcome barriers to resource commitment to the discontinuous change (note, however, that Gilbert does not consider relevance perception). However, the second and third patterns contradict Gilbert’s observations. The second pattern involves framing the discontinuity as a substantial but hardly controllable gain, therefore violating Gilbert’s unidimensional conceptualization of opportunity/threat perception. The third pattern contradicts Gilbert in that incumbents overcome sources of resource rigidity even in the absence of threat perception, which Gilbert portrays as a necessary precursor of resource commitment to discontinuous change (see Gilbert, 2006: 162).

These findings echo various streams of the extant literature that envision gain or opportunity perceptions together with substantial perceived control as catalysts for change (Ford & Baucus, 1987; Shimizu, 2007; Thomas et al., 1993; White et al., 2003). Dutton and Jackson (1987), for instance, refer to the threat rigidity thesis, that is, the conjecture that threat perceptions cause organizational rigidity, and emphasize that in contrast, “when decision makers label issues as opportunities, involvement in the process of resolving the issue will be greater” (p. 83). Dutton (1993), although she does not explicitly discuss the separate role of perceived control, argues that this effect of gain framing might be rooted in the positive activating emotions it evokes. Similarly, the concept of self-efficacy (Bandura, 1977) strongly supports the notion that gain perceptions in combination with perceived control can effectuate individual, and thus ultimately organizational, action. Synthesizing all our insights, we formally posit the following:

Proposition P2: Gain framing relaxes resource rigidity in response to a discontinuous change if organizational decision makers perceive the change as a highly relevant issue and perceive at least moderate levels of control.

Table 6 provides a comprehensive overview of the various configurations of gain/loss framing and perceived control we have described in Propositions P1 and P2, respectively, and their consequences, that is, whether or not resource rigidity relaxed. Note that the table assumes that decision makers perceive the focal discontinuous change as a relevant, that is, high-priority, situation. In accordance with our observations (see Proposition P2) and given the abundant evidence in the literature (e.g., Eggers & Kaplan, 2013; Gerstner et al., 2013; Kammerlander et al., 2018; Kaplan et al., 2003), we expect, but do not formally propose, that resource rigidity does not relax when the discontinuous change is perceived as irrelevant.

A key objective of a generalization and extension is to uncover important novel nuances of the focal phenomenon (Tsang & Kwan, 1999). In this vein, we induced two additional propositions (P3 and P4) from our data, both of which substantially extend Gilbert’s (2005) work.