To Compete From Within: The Competitive Dynamics of Resource Bundling

Resource Structuring and Factor-Market Competition

Resource structuring involves the acquisition or development and divestiture of resources. Thus, it is the first stage of the resource management process and has important performance implications. Because a firm’s resource endowments determine the breadth of actions it can subsequently take (Ndofor et al., 2011), resource structuring can also create path dependencies. Excluding non-tradable resources, such as know-how, social capital, and reputation, which are developed internally (Dierickx & Cool, 1989), firms procure strategic resources in strategic factor markets (Barney, 1986). Firms also divest their non-productive assets in factor markets, which is particularly salient for those experiencing or anticipating poor performance (Morrow, Sirmon, Hitt, & Holcomb, 2007).

Factor markets, where rivals vie for the same resources, constitute competitive domains. Due to the inherently scarce nature of the traded—strategic—resources, strategic factor markets (Barney, 1986) are particularly prone to “competition over resource positions” (Markman et al., 2009: 423), in the form of bidding wars (Makadok & Barney, 2001). The more scarce, nonimitable, and nonsubstitutable a resource is, the more intense the rivalry it is likely to incite (Peteraf, 1993). Moreover, assuming that factor market overlap engenders awareness among rivals of each other’s factor market activity, factor market competition tends to escalate reciprocally as rivals acquire scarce resources (Theeke & Lee, 2017), such as patents (Gianiodis, Markman, & Panagopoulos, 2019) and supply chain resources (Ellram, Tate, & Feitzinger, 2013). A firm might also choose to acquire strategic resources to pre-empt its rivals (Capron & Chatain, 2008). In any case, intensifying factor market competition benefits factor owners, who are able to appropriate more value as a result (Asmussen, 2015).

Resource Leveraging and Product-Market Competition

The final stage in the resource management process—leveraging—involves the utilization of existing firm capabilities to execute a chosen strategy and, ultimately, create value (Sirmon et al., 2011). Leveraging activity relates to demand, and it is product-market oriented. It involves identifying opportunities in the marketplace, matching them with the capabilities needed to capture them, integrating those capabilities, and deploying them in product markets (Sirmon et al., 2007). Product markets are the “flipside” of factor markets. As noted, rivalry within and across product markets has been the main focus of the competitive dynamics research (Hughes-Morgan, Kolev, & Mcnamara, 2018). The actions herein, such as new product introductions (e.g., Lee, Smith, Grimm, & Schomburg, 2000), product market entry (e.g., Baum & Korn, 1996) and exit (e.g., Upson, Ketchen, Connelly, & Ranft, 2012), the launch of marketing campaigns (e.g., Yu & Cannella, 2007), and changes in pricing (e.g., Ferrier, Smith, & Grimm, 1999), largely correspond to resource deployment activities in the resource management framework (Costa, Cool, & Dierickx, 2013). Likewise, the resource orchestration literature has emphasized the relationship between product market-oriented activity and performance outcomes (D’Oria, Crook, Ketchen, Sirmon, & Wright, 2021).

In competitive dynamics, “firms act, and rivals react” (Smith et al., 2005: 318), resulting in escalating patterns of market competition and rivalry. Accordingly, a focal firm can be driven to engage in competitive activity when rivals’ activity threatens its market position, leading to dyadic exchanges of competitive moves (Chen & Miller, 2015). Alternatively, a firm can be driven to display competitive behavior by the intensity of competition in a given market or industry (D’Aveni, 1994; Young, Smith, & Grimm, 1996) in order to improve or maintain its “relative competitive position” (Smith et al., 2005: 321). As a strategic choice, rivals may also refrain from escalating competition within and across markets (Andrevski & Miller, 2022), as this can benefit customers at the expense of competing firms (Andrevski & Ferrier, 2019).

Enhanced Awareness

In competitive dynamics, awareness of rival actions is a prerequisite of response and escalation, which hinges upon the assumption that market overlap among rivals results in awareness of each other’s market-oriented actions (Chen, 1996). Conversely, because bundling activity is internal, it is assumed to be less visible to outsiders (Bridoux et al., 2013), which can limit rivals’ awareness and therefore likelihood of response. Further, because “awareness” in competitive dynamics implies an understanding of the cause-and-effect relationships between rival actions and their outcomes (Andrevski, Miller, Le Breton-Miller, & Ferrier, 2022), the internal nature of bundling activity is also associated with causal ambiguity (Reed & DeFillippi, 1990). We argue that, although bundling activity is relatively less visible, as explained below, three factors contingent on factor and product market overlap can enhance rivals’ awareness of each other’s bundling activity, thereby facilitating reaction and escalation.

Resource Similarity

In competitive dynamics, rivalry is more likely to escalate among rivals with similar resource profiles (Chen, Su, & Tsai, 2007; Smith et al., 2005). Due to the literature’s emphasis on product markets (Chen et al., 2021), however, resource similarity is considered independent of market overlap. ³ Distinguishing the competitive implications of factor market overlap from those of product market overlap helps clarify the relationship between market commonality and resource similarity. Because resources traded in a given factor market are likely to be similar in function and characteristics (Felin, Kauffman, Mastrogiorgio, & Mastrogiorgio, 2016), rivals overlapping in factor markets are likely to have similar resource profiles, enhancing their capability and likelihood of reacting to each other’s activity. As resources determine the breadth of actions a firm can take (Ndofor et al., 2011), firms that possess similar resources are more likely to be capable of engaging in similar types of activities, including bundling. Resource similarity among rivals can also eliminate or minimize the need for additional resource acquisition when rivals react to each other with conforming bundling activity. These effects are particularly salient for firms that overlap in strategic factor markets (Barney, 1986). Strategic resources are scarcer and costlier than others. As a result, firms prioritize their strategic resources by building their organizational capabilities “around” them through bundling actions that accentuate such resources to maximize their value-creating potential (Makadok, 2001).

Performance Pressure

Despite its internal and non-zero-sum nature, bundling activity can constitute performance pressure and escalate among rivals that overlap in product markets. However, this process differs from the conventional conceptualization of product market rivalry in both the motivation behind firm activity and the mechanism through which this activity escalates. Firms are in a constant race to achieve and sustain competitive advantage, which requires creation of greater economic value than rivals. Economic value creation is a function of both cost and consumer value (Peteraf & Barney, 2003). The latter is about resource leveraging through which firms seek and exploit market opportunities (Sirmon et al., 2008); leveraging and deployment activity can escalate into product market rivalry via dyadic exchanges, where firms “attack” rivals through market actions to gain customers or, threatened by rivals’ attacks, respond with their own market activity to defend their positions (Chen & Miller, 1994). Resource bundling, on the other hand, is about optimal utilization of the firm’s existing resources to maximize value creation (Sirmon et al., 2011). Firms “performing particular activities more efficiently than competitors” can achieve cost-driven competitive advantages (Porter, 1996: 62). Thus, bundling can enable a firm to improve its relative performance without engaging in further factor or product market activity (Barney & Arikan, 2001).

On the flip side, rivals’ bundling actions—and the implied efficiency advantages—can serve as a salient competitive cue for a focal firm. Because value creation materializes ultimately in product markets, performance pressure is conditioned by overlap in product markets. As markets evolve, rivals’ efforts to maximize value creation through bundling inevitably affect each other’s relative performance. In turn, a focal firm may become motivated to “act” in order to improve or maintain its own standing (Barnett, 1997; D’Aveni et al., 2010). Notably, such activity reflects a firm’s reaction to the broader competitive environment rather than retaliation to a specific rival (Derfus, Maggitti, Grimm, & Smith, 2008), as an initial “attack” is not present in this case. Accordingly, firms often utilize market-level benchmarking (Panagiotou, 2007; Porter, 1980), where rivals’ capability levels serve as reference points for bundling. As competition becomes a series of races in operational effectiveness (Porter, 1996), firms tend to engage in escalating series of actions “to maintain pace with opponents, which leads to an environment of great competitive intensity” (Chen, Lin, & Michel, 2010: 1410-1411); those that fail to do so face performance declines (Andrevski & Ferrier, 2019). For successful firms, such competition driven by dynamism in crafting and constant fine-tuning capabilities yields “a series of temporary competitive advantages” rather than a sustained one based on scarce resources (Morrow et al., 2007: 281).

Bundling can be a salient reaction to such pressure—one that does not constitute direct “attack” on rivals. Unlike resource structuring and leveraging (market-oriented activities), bundling is less likely to escalate into zero-sum rivalry, which primarily benefits customers and factor owners. Moreover, bundling does not necessarily involve additional investment in resource acquisition and can be executed relatively quickly, yet it may yield enduring positive impact on firm performance (Bridoux et al., 2013). Empirical evidence also suggests that performance pressure motivates bundling activity (Ketchen & Palmer, 1999). The benefit a firm can derive from an activity depends on the intensity of competition in the context where the activity occurs (Nadkarni, Chen, & Chen, 2016). Accordingly, as firms react to the performance pressure driven by their rivals’ bundling efforts by intensifying their own bundling activity, bundling is likely to escalate at the market or industry level. A salient example is the recent widespread race to integrate artificial intelligence (AI) into existing organizational capabilities. Accordingly, across a multitude of industries, ranging from retail to manufacturing, transportation, and healthcare, “peer competitive pressure is the largest influencer of the decision to adopt AI and make it work across all enterprise functions” at the “industry level” (Seong & Bughin, 2018). Figure 1 illustrates our model of the competitive drivers of bundling activity among rivals that overlap in both factor and product markets.

OPEN IN VIEWER

Figure 1

The Competitive Drivers of Resource Bundling

Building on these theoretical reflections, we first outline a baseline hypothesis on the relationship between the bundling intensity of a focal firm and the preceding aggregate bundling intensity of the rivals it overlaps with both in product and factor markets. To account for the sequential nature of the resource management process (Sirmon et al., 2007), we hold constant focal firm resource structuring and deployment activity.

As firms compete, they take a variety of actions in combination (D’Aveni, 1994). In this vein, the competitive dynamics literature analyzes firm competitive activity in repertoires—the entire set of actions a firm takes in a given period (Ferrier et al., 1999; Miller & Chen, 1996b). Nonconformity reflects how much a firm’s repertoire of actions differs from the industry average (Miller & Chen, 1996a). The literature has used various terms to describe such nonconformity, including “competitive deviance” (Ndofor et al., 2011), “action dissimilarity” (Gimeno & Woo, 1996), and “competitive action heterogeneity” (Hughes-Morgan, Ferrier, & Labianca, 2011). Action nonconformity is one of the critical aspects of competitive behavior, because a repertoire that does not conform to market or industry norms can be difficult for rivals to imitate, impeding their ability to respond (Chen & Miller, 1994) or delaying their response (Chen, Smith, & Grimm, 1992). Thus, nonconforming activity can lead to distinctive and more enduring competitive advantages.

Nonconformity in resource management has salient implications for competitive advantage (Ndofor et al., 2011). Nonconformity in bundling activity can lead to performance heterogeneity among rivals under resource parity. Unlike internally developed strategic resources (Dierickx & Cool, 1989), resources acquired in factor markets are not necessarily unique, and their value depends on how they complement firm capabilities (Makadok, 2010). By adopting a repertoire of bundling actions that differs from those of its rivals, a focal firm can achieve unique capability configurations and competitive advantage, even in the absence of superior resources (Peteraf, 1993). Bundling may also facilitate novel resource combinations. For example, Baker and Nelson (2005) found that via innovative extraction methods, depleted coal mines can become sources of methane gas. Indeed, research indicates that competitive nonconformity improves relative performance (Caves & Ghemawat, 1992; Ferrier & Lee, 2002) and that firms tend to experiment with nonconforming action combinations when faced with poor performance (Hughes-Morgan & Ferrier, 2017). Further, as performance pressure can drive rivals to simultaneously invest in resource acquisition, costs of inputs—both scarce and generic—tend to increase, making bundling activity more costly (Barnett, 1997; Derfus et al., 2008). Firms that follow best practices and conform to proven capability configurations are particularly vulnerable to this effect as they might need to simultaneously compete with each other for the same inputs. Under these conditions, nonconformity in bundling activity can also be cost-saving by enabling the utilization of alternative inputs. Overall, rivals’ bundling activity can impose competitive pressure on a focal firm to not only intensify its bundling activity but also differentiate it.

Relative Resource Endowments: Attenuated and Conforming Reaction

Both resource orchestration and competitive dynamics perspectives imply a potential tradeoff between a firm’s resource position advantages and dynamism in bundling activity.

Value creation is maximized when bundling activity and resources complement each other (Black & Boal, 1994). Yet, there are limitations to the synergy and the competitive improvement a firm can achieve through resource bundling and capability building (Makadok, 2001; Ray, Barney, & Muhanna, 2004). Resource and action heterogeneity represent two distinct but related paths toward performance heterogeneity. Capturing this interplay requires an examination of the “independent effects” of “relative or comparative levels of rivals’ resource management and resources” (Sirmon et al., 2008: 920). Accordingly, at high levels of “relative resource endowments” (Ndofor et al., 2011), the marginal benefit of bundling activity diminishes, as resources themselves might become the main drivers of performance heterogeneity. Conversely, “managerial ability becomes increasingly important in firms at lower levels of resource quality” (Holcomb et al., 2009: 478), as “large differential in resources [in favor of the focal firm] overwhelms the influence of managerial bundling” (Sirmon et al., 2008: 923-924). Further, bundling actions also entail both direct and indirect costs, including investments in complementary assets (Ellram et al., 2013), learning (Winter, 2003), administrative costs and complexity (Helfat et al., 2007), as well as attention and cognitive effort—another limited resource (Helfat & Peteraf, 2015). Accordingly, from a rational point of view, firms with high relative resource endowments can choose to react to rivals’ bundling activity to a lesser extent than those with smaller resource endowments, as the marginal positive effect of bundling on value-creation diminishes at higher resource levels.

A firm’s resource advantages can also mitigate the perceived “pressure to bundle” that is driven by rivals’ bundling activity. In competitive dynamics, a reactor’s perception of rivalry depends on not only the action and the actor, but also its own relative position; two firms might react differently to the same competitive cue. From a behavioral point of view, “competitive asymmetry” explains why rivals often do not pose an “equal threat” to each other (Chen, 1996: 102; Miller, 2003); “powerful firms,” those with greater market share and resource endowments may not be motivated by the moves of “weaker firms.” In this vein, the variance among rivals in terms of resource endowments can also result in competitive asymmetries when reacting to the performance pressure driven by the intensity of bundling at the market level. Accordingly, a focal firm with superior resource endowments is less apt to react to the bundling activity of its rivals, which it might perceive as “weaker” in terms of resource endowments or factor market share.

The competitive asymmetry caused by the variance in resource endowments among rivals might affect not only a firm’s likelihood and intensity of reaction to rivals’ bundling, but also how it might react. There is evidence that larger firms with superior resource endowments compete differently than smaller ones with limited resources (Chen & Hambrick, 1995). To compete, in addition to intensive resource bundling, a firm can differentiate bundling activity to compensate for “comparative resource disadvantages” (Sirmon et al., 2008: 922). As unique capability configurations have the potential to drive relative performance without scarce resources that are acquired in strategic factor markets (Schmidt & Keil, 2013), by pursuing nonconformity in bundling, a firm with resource disadvantages can also mitigate its exposure to factor-market rivalry, which could further accentuate its “weakness.” Conversely, firms with high relative resource endowments, which can potentially achieve superior performance by exploiting their existing resource advantages, might not have the motivation to pursue experimentation and nonconformity in bundling. For instance, Shuen, Feiler, and Teece (2014) point to the scarcity of hydrocarbon reserves as a main driver of innovation and action heterogeneity in the upstream oil and gas industry. Rather than competing head-on with rivals, firms can be driven to develop or adopt new technologies, build new routines, and invest in experimentation. By contrast, when faced with competition, larger firms with superior resource endowments are likely to utilize more conventional and conforming capability configurations.

Multi-Factor Market Contact: Imitative and Intensified Reaction

Based on the assumption that firms overlapping in multiple markets are likely to be similar in scale, scope, and market opportunities (Edwards, 1955), multimarket contact has primarily been studied as an antecedent of anticompetitive behavior. Accordingly, the strategic similarity among multimarket rivals drives them to “forbear”—refrain from aggressive competition—to prevent rivalry from spreading to other joint markets (Gimeno & Woo, 1996), and, at the same time, retaliate against those who “defect” (Young, Smith, Grimm, & Simon, 2000). Although multimarket contact has most frequently been associated with forbearance across product markets (Yu & Cannella, 2013), it also has salient implications for imitative behavior (e.g., Hsieh & Vermeulen, 2014; Scott, 2001). Analyzing imitative behavior among rivals, Lieberman and Asaba (2006: 376) posit that institutional and competitive forces “are not mutually exclusive; both types of imitation can occur simultaneously.” In this view, (a) ambiguity surrounding rivals’ activity (such as bundling) and environmental uncertainty drive mimetic behavior through information-seeking, and (b) market commonality drives multimarket rivals to be vigilant and reactive to each other (Lieberman & Asaba, 2006).

In line with these premises, multi-factor market contact (MFMC)—overlap across multiple factor markets (Markman et al., 2009)—can foster imitative bundling behavior among rivals. First, as noted, because factor market overlap implies potential resource similarity, multi-factor market rivals are even more likely to have similar resources than others. Second, multi-factor market rivals also have more opportunities to observe and learn from each other’s bundling activity across multiple factor markets. Rivals’ actions can provide information to each other (Fiegenbaum, Hart, & Schendel, 1996), which can mitigate the causal ambiguity that tends to surround bundling activity (Bridoux et al., 2013). Typically, the exact values of strategic resources traded in factor markets are ex-ante unknown and dependent on how they are bundled (Schmidt & Keil, 2013). Accordingly, rivals’ bundling activity can mitigate the ambiguity about the routines and complementary assets needed to deploy the traded resources and reveal their actual value-creating potential. Contact across multiple factor markets further bolsters such mutual learning among rivals (Phillips & Mason, 2001).

When these two factors—resource similarity and the availability of information on rivals’ capability configurations across multiple markets—are combined with pressure to bundle driven by rivals’ activity, imitation of multi-factor market rivals in bundling becomes a viable option for a focal firm. Regarding the motivation, Theeke and Lee (2017) suggest that although multimarket contact can engender forbearance in product market activities, it actually intensifies competitive tendencies in “upstream” activities through which retaliation is not possible. Because bundling is an internal activity which does not constitute an “attack,” rivals lack the motivation to forbear in their bundling across markets. Imitation of rival activity is a common response to heightened competitive pressure (Derfus et al., 2008; Major, Maggitti, Smith, Grimm, & Derfus, 2016) and uncertainty (Posen, Ross, Wu, Benigni, & Cao, 2023; Sirmon & Hitt, 2009). A firm might choose not to adopt certain technologies, routines, and capability configurations before observing rival adoption (Endenich, Hahn, Reimsbach, & Wickert, 2023). From an economic standpoint, too, imitation of “closest rivals” in bundling can be more feasible than nonconforming activity because it does not involve the potential time and costs associated with learning and experimentation (Boyd & Bresser, 2008; Giachetti, Lampel, & Pira, 2017). The efficient, “quick and vigorous” reaction enabled by imitation (Porter, 1985: 498) could further escalate bundling activity. Thus, multi-factor market rivals are more likely to react to the performance pressure driven by each other’s bundling activity with their own, albeit in a more conforming manner. Figure 2 summarizes our hypotheses.

OPEN IN VIEWER

Figure 2

Hypotheses

Context: Upstream Oil and Gas

We tested our hypotheses in the context of the upstream oil and gas business. Consistent with our hypotheses, firms that are primarily occupied with crude oil and raw natural gas production tend to overlap in both product and factor markets. Within and across oil and gas basins, these firms compete for land and leases granting them exclusive rights to explore for and exploit oil and gas reserves, highly scarce and strategic resources. Oil and gas basins are essentially large depressions in the earth’s crust that contain hydrocarbon-rich sediment; they vary in their geological characteristics, depth, and pressure of reservoirs, as well as the type and quality of hydrocarbons (International Association of Drilling Contractors, 2023). These characteristics determine how operators can produce hydrocarbons. Thus, the basins constitute geographic factor markets for strategic resources (Lim, Celly, Morse, & Rowe, 2013), which are prone to bidding wars that benefit landowners. The efficiency of these factor markets accentuates the importance of resource bundling in this context. Other than engaging in bidding wars in factor markets or resource picking by estimating reserves more accurately than others (Le Breton-Miller & Miller, 2015; Makadok & Barney, 2001), firms can improve their production and relative performance by bundling their reserves via different routines, techniques, and complementary resources. In the United States, operators must get permission for every such major action they plan to undertake. The data related to these permits are published by state agencies, such as the Railroad Commission of Texas (RRC).

These firms also tend to overlap in product markets since they sell their output in the spot market or via contracts varying in length at the benchmark index prices, such as West Texas Intermediate for crude oil and Henry Hub for natural gas in the United States. Accordingly, this context is prone to escalating resource bundling activity. In fact, the recent shale boom that originated from Texas has been enabled by intense bundling activity. During this period, firms’ bundling activity (e.g., in the form of hydraulic fracturing) has gradually intensified. According to the U.S. Energy Information Administration (EIA), the total number of hydraulically fractured wells in the United States increased from 23,000 in 2000 to 300,000 in 2015 (EIA, 2016), making the United States the world’s largest crude oil and natural gas producer (EIA, 2024).

The tradeoff between resource endowments and the motivation to engage in resource bundling and pursue nonconformity in bundling activity is also evident in the upstream oil and gas industry. Whereas small, resource-strapped firms have been investing in “unconventional” methods to better utilize their limited reserves in Texas shales, which were once deemed “worthless,” upstream giants like Exxon and Chevron chose to ignore shale drilling until it became “undeniable” (Zuckerman, 2014). Local firms such as Mitchell Energy pioneered innovative extraction methods, including hydraulic fracturing, ushering in the recent Texas shale boom. On the other hand, there is evidence that upstream oil and gas companies that operate in proximity tend to follow one another in their activities (Covert, 2015; Décaire, Gilje, & Taillard, 2020). Thus, firms competing across multiple basins were largely responsible for the nationwide diffusion of these techniques (Tabuchi & Migliozzi, 2023).

Sample

We utilized a proprietary upstream oil and gas dataset, provided courtesy of WellDatabase, to build our sample, and we defined factor markets based on the major oil and gas basins in the United States. Accordingly, we chose the Barnett Shale in the Fort Worth Basin in North Texas as the focal factor market, which is considered the birthplace of the so-called shale revolution (Zuckerman, 2014). We limited our sample to operators that were active in the Texas portion of this basin and captured both dependent variables in this area. We captured MFMC of focal firms across the remaining 29 major oil and gas basins in the United States where some of the firms that have been active in the Barnett Shale also operated. The figure in Appendix A (see the online supplemental material) lists and illustrates the locations of the basins that are the geographic factor markets analyzed in this study. In our dataset, over 75% of all wells with production history in the Fort Worth Basin were directed at the Barnett Shale—the sub-basin. Therefore, we excluded wells outside the Barnett Shale for the focal basin. We do not utilize states or counties to define factor markets because these basins cover large areas and often extend across administrative divisions.

The final sample contains 1,068,830 firm-month observations from 5,415 unique operators (panels) spanning the period from 1981 to 2019. We excluded from the sample firms that did not have any production or development activity within this period. We also excluded observations in the data without operator information and panels with errors identified based on names as well as production and incorporation years.

Measures

Focal firm bundling nonconformity (Focal Firm BN)

Following the dissimilarity measure of Connelly, Lee, Tihanyi, Certo, and Johnson (2019), the dependent variable for Hypotheses 1b, 2b, and 3b was measured as the Euclidean distance between the action combination of the focal firm and the basin/factor-market average of 16 different types of bundling actions that are included in the first dependent variable (well completions, stimulations, injections, and tests), captured as dummies, in a given month. Aligned with our arguments about the competitive implications of resource bundling, instead of comparing rivals in dyads, our measure compares the focal firm to the market average.

Binary identifiers represent action types: (1) well modification-related completion actions (reentry, plugback, sidetrack, deepening, and other/unclassified); (2) well stimulation actions (slickwater, surfactant, and acid); (3) well injection actions (water, gas, carbon dioxide, and other/unclassified injections); and (4) well tests (steam injection, flow, gas lift, and pump). For robustness checks, an alternative measure of action nonconformity was calculated under four broader action categories, namely (1) completions, (2) stimulations, (3) injections, and (4) tests, and similar results were obtained. Accordingly, the bundling nonconformity of firm i at time t is calculated based on the difference between the binary action indicator W i of the focal firm i and the average binary action indicator W j of rival firms j in the focal factor market, where R denotes the total number of rivals j at time t in the focal factor market, for each of the 16 action categories a, as specified in Equation 1 below:

F o c a l F i r m B N i t = l n [ ∑ a = 1 16 ( W i a t − ( 1 R ∑ j = 1 R W j a t ) ) 2 ]

(1)

Due to its non-normal distribution and continuous and exclusively positive nature (without zeros), we log-transformed this variable. This transformation allows us to interpret the results in terms of percentage change in the outcome variable. As reported in Appendix E, we obtained practically identical results using the untransformed version of this variable.

Firm size, age, and scope have critical implications for action nonconformity (Chen & Hambrick, 1995; Miller & Chen, 1996a; Smith et al., 2005) and could confound the relationships of interest. Firms that are larger in size and scope or older can have a greater variety of resources and capabilities or have accumulated more experience. To account for those, in addition to Firm Age (in years), we controlled for firm size through the count of Focal Firm Wells in the Area and a ratio of this measure to the firm’s overall well portfolio nationwide (Weight in Firm Well Portfolio). Boyd and Bresser (2008: 1087) suggest that “a firm’s total number of competitive moves is a proxy indicator for its size.” Accordingly, in subsequent robustness tests, we also included Focal Firm BI as a control variable in combination with Focal Firm Wells in the Area in the models predicting Focal Firm BN and obtained similar results.

To establish temporal precedence, all predictor variables, including controls, were lagged by 1 month. Continuous predictor variables were also standardized and mean-centered. In line with the recent work and best practices (Aguinis, Ramani, & Alabduljader, 2018), we did not drop observations or transform our dependent variables to deal with outliers. However, we conducted several additional tests to determine whether our analyses are sensitive to outlier observations (see Appendix E).

Multi-factor market contact

MFMC was measured at “firm-in-market level” (Gimeno & Jeong, 2001), following Gimeno and Woo (1996), as the average of multimarket contacts with all rivals that the focal firm overlaps with in the focal factor market in a given month. Thus, at time t, for focal firm i and rival firm j in the focal factor market f, MFMC is measured as the count of other factor markets m, where D is a binary variable that equals 1 if both i and j operate at time t, divided by the total number of rivals R firm i meets in the focal factor market f at time t (see Equation 2):

M F M C i f t = ∑ j ≠ i ∑ m ≠ f D i j m t R i f t

(2)

Analysis

As noted, Focal Firm BI, the dependent variable for Hypotheses 1a, 2a, and 3a (tested in Models 1–5), required a count estimator that allows for time and firm fixed effects and robust clustered standard errors (SEs). Our sample is an unbalanced panel, and our hypotheses address within-firm changes. Thus, we chose two-way fixed-effects models, allowing predictions net of time-constant factors and time-varying macroeconomic conditions. The mean value for Focal Firm BI was 1.325 with a variance of 18.905. Thus, we took several steps to account for potential overdispersion. First, once firm and year fixed effects were entered in a Poisson model, the Cameron and Trivedi (2005) test failed to reject the null hypothesis of no overdispersion (t = 1.60, p = .11). Further, even in the case of overdispersion, considering the nature of our data, what matters is conditional mean and variance. The Poisson pseudo maximum likelihood (PPML) and “conditional” Poisson fixed effects estimators do not require the variance to equal the mean, and remain consistent as long as the conditional mean is correctly specified (Wooldridge, 1999). Accordingly, we employed PPML models with two-way fixed effects whose assumptions are not violated by the specifications of the dependent variable (Silva & Tenreyro, 2011). Additionally, as a robustness test, using a fixed-effects negative binomial model—which does not allow the use of heteroscedasticity and autocorrelation robust “true” fixed effects (Guimaraes, 2008; Wooldridge, 1999)—we obtained similar results (see Model 18 in Table E1 in the Appendix file). We used the Stata package “ppmlhdfe” (Correia, Guimarães, & Zylkin, 2020) to implement the PPML estimator. Additionally, using the default “xtpoisson” module on Stata, we were able to run fixed-effects Poisson models, include year and firm fixed effects, and obtain identical results. In these models, as noted, Focal Firm Wells in the Area (at t − 1) was included as an exposure variable. Used in count models, an exposure variable is a covariate that is log-transformed and its coefficient fixed at 1. This variable enables us to capture Focal Firm BI at time t relative to firm assets at t − 1, consistent with the measurement of Rivals’ BI (at the factor-market level), albeit without any transformation.

Our second dependent variable, Focal Firm BN (i.e., Focal Firm Bundling Nonconformity), is continuous. A Hausman test confirmed the superiority of fixed effects over random effects (Δχ² = 2,365.22, p < .001). Accordingly, we used linear two-way fixed effects models (Models 6–10) with year and firm fixed effects and robust clustered SEs to deal with potential heteroscedasticity and autocorrelation. For both dependent variables, year and firm fixed effects also reduce omitted variable bias. All continuous variables were grand-mean centered and standardized to improve the interpretability of the results and facilitate the inclusion of the interaction terms. In PPML models, model fit was assessed using Akaike’s information criterion (AIC), with smaller figures indicating better fit, as well as pseudo R² and change in deviance statistics. In the linear models, we used both R² and AIC statistics to assess model fit.

Results

Table 1 presents descriptive statistics and the correlation matrix. Table 2 displays the regression results from fixed-effects PPML models (Models 1–5), testing Hypotheses 1a, 2a, and 3a. Table 3 displays the results from linear fixed-effects models (Models 6–10), testing Hypotheses 1b, 2b, and 3b. Model fit statistics, including ΔAIC (AIC_null – AIC_model), Δdeviance, and R² (McFadden’s Pseudo R² for count models), are also shown in Tables 2 and 3. Sample size varies across analyses; the count estimator drops observations that are singletons or separated by a fixed effect (162,352 observations in total). The logged version of our second dependent variable, Focal Firm BN, contains some negative values, which are the logs of fractions.

OPEN IN VIEWER

Table 1

Summary Statistics and Bivariate Correlations

	Variables	M	SD	Min	Max	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)
(1)	Focal firm BI	1.325	4.348	0.00	298.00
(2)	Focal firm BN a	−1.059	0.831	−5.35	1.06	0.34
(3)	Natural gas index price b	9.302	3.627	3.99	20.77	0.04	0.21
(4)	Price shock b	0.150	0.357	0.00	1.00	0.01	−0.02	0.16
(5)	All wells in the area b c	24.609	7.143	10.66	36.37	0.07	0.42	0.64	0.22
(6)	Firm age (in years)	39.626	20.223	0.00	113.00	0.24	0.24	0.34	0.13	0.34
(7)	Well accidents	0.000	0.008	0.00	1.00	0.01	0.01	0.00	0.00	0.00	0.00
(8)	Weight in firm well portfolio	0.388	0.377	0.00	1.00	0.09	0.07	−0.15	−0.05	−0.14	−0.25	0.00
(9)	Per well production	64.059	290.168	0.00	76,726.59	0.02	0.01	0.02	0.00	0.01	−0.07	0.04	0.09
(10)	New well drills	0.012	0.435	0.00	67.00	0.13	0.04	0.02	0.01	0.04	0.03	0.00	0.00	0.11
(11)	Focal firm wells in the area	18.764	92.940	1.00	6,166.00	0.35	0.12	0.05	0.02	0.06	0.15	0.03	0.00	0.13	0.42
(12)	Focal firm RRE	1,212.540	721.905	106.00	3,189.00	0.34	0.31	−0.06	−0.06	0.04	0.14	0.01	0.39	0.11	0.03	0.20
(13)	MFMC	0.029	0.069	0.00	0.51	0.12	0.09	0.07	0.01	0.10	0.18	0.01	−0.26	0.13	0.08	0.19	0.13
(14)	Rivals’ BI	131.336	65.220	8.02	323.11	0.09	0.60	−0.15	−0.13	0.01	−0.09	0.00	0.04	−0.03	−0.01	−0.02	0.17	0.02

Note. N = 1,068,830. |r| > .003 is significant (p < .05). Year and firm fixed effects not shown.

^aLog-transformed. ^bFactor-market- or industry-level controls. ^cRescaled by 1,000.

OPEN IN VIEWER

Table 2

Fixed-Effects PPML Models Predicting Focal Firm Bundling Intensity

Predictors (t − 1)	Controls	Main Effects	Interactions		Full Model
	Model 1	Model 2	Model 3	Model 4	Model 5
Natural gas index price	−0.029	−0.026	−0.025	−0.026	−0.025
	(0.001)	(0.003)	(0.004)	(0.003)	(0.005)
	[0.009]	[0.009]	[0.009]	[0.009]	[0.009]
Price shock	−0.009	−0.012	−0.012	−0.012	−0.012
	(0.192)	(0.096)	(0.089)	(0.093)	(0.083)
	[0.007]	[0.007]	[0.007]	[0.007]	[0.007]
All wells in the area	0.229	0.317	0.313	0.318	0.313
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.030]	[0.029]	[0.028]	[0.029]	[0.028]
Firm age (in years)	0.121	0.057	0.049	0.063	0.054
	(0.488)	(0.745)	(0.787)	(0.721)	(0.764)
	[0.175]	[0.176]	[0.180]	[0.175]	[0.181]
Well accidents	−0.023	−0.020	−0.014	−0.043	−0.041
	(0.849)	(0.870)	(0.908)	(0.705)	(0.715)
	[0.122]	[0.121]	[0.122]	[0.112]	[0.113]
Weight in firm well portfolio	0.605	0.602	0.593	0.609	0.600
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.024]	[0.024]	[0.025]	[0.024]	[0.024]
Per well production	−0.084	−0.083	−0.079	−0.086	−0.083
	(0.369)	(0.377)	(0.396)	(0.351)	(0.368)
	[0.094]	[0.094]	[0.093]	[0.092]	[0.092]
New well drills	0.004	0.004	0.004	0.004	0.004
	(0.159)	(0.162)	(0.183)	(0.081)	(0.084)
	[0.003]	[0.003]	[0.003]	[0.002]	[0.002]
Focal firm RRE	−0.219	−0.220	−0.178	−0.223	−0.175
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.021]	[0.021]	[0.024]	[0.020]	[0.024]
MFMC	−0.088	−0.087	−0.091	−0.078	−0.081
	(0.003)	(0.003)	(0.002)	(0.008)	(0.005)
	[0.030]	[0.030]	[0.029]	[0.029]	[0.029]
Rivals’ BI		0.200	0.268	0.200	0.277
		(0.000)	(0.000)	(0.000)	(0.000)
		[0.009]	[0.017]	[0.009]	[0.017]
Rivals’ BI * Focal firm RRE			−0.081		−0.092
			(0.000)		(0.000)
			[0.016]		[0.016]
Rivals’ BI * MFMC				0.042	0.051
				(0.044)	(0.012)
				[0.021]	[0.021]
AICnull − AICmodel	108,822	112,897	115,426	114,539	117,792
Pseudo R2	0.6287	0.6294	0.6299	0.6297	0.6303
Δdeviance a	108,816	4,076	2,532	1,645	3,255

Note. N = 906,478; N_panels = 3,686; 162,352 singleton/FE-separated cases were dropped. Exposure term: Focal firm wells in the area. Robust SEs clustered by unit are shown in square brackets; p values are shown in parentheses; two-tailed tests. Year and firm fixed effects are not reported. All continuous predictors are standardized.

a

Δdeviance (Null−M1, M1−M2, M2−M3, M2−M4, M4−M5) χ² were all significant (p < .001).

OPEN IN VIEWER

Table 3

Fixed-Effects OLS Models Predicting Focal Firm Bundling Nonconformity

Predictors (t − 1)	Controls	Main Effects	Interactions		Full
	Model 6	Model 7	Model 8	Model 9	Model 10
Natural gas index price	−0.031	−0.023	−0.023	−0.023	−0.023
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.002]	[0.002]	[0.002]	[0.002]	[0.002]
Price shock	−0.043	−0.043	−0.043	−0.043	−0.043
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.001]	[0.001]	[0.001]	[0.001]	[0.001]
All wells in the area	0.655	0.635	0.635	0.634	0.634
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.010]	[0.009]	[0.009]	[0.009]	[0.009]
Firm age (in years)	−0.411	−0.471	−0.475	−0.471	−0.475
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.098]	[0.117]	[0.116]	[0.117]	[0.116]
Well accidents	0.297	0.305	0.307	0.310	0.311
	(0.004)	(0.002)	(0.001)	(0.002)	(0.001)
	[0.104]	[0.098]	[0.096]	[0.098]	[0.096]
Weight in firm well portfolio	0.043	0.041	0.037	0.040	0.037
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.004]	[0.004]	[0.004]	[0.004]	[0.004]
Per well production	0.004	0.004	0.004	0.004	0.004
	(0.116)	(0.077)	(0.105)	(0.080)	(0.108)
	[0.002]	[0.002]	[0.002]	[0.002]	[0.002]
New well drills	0.010	0.010	0.010	0.010	0.010
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.002]	[0.002]	[0.002]	[0.002]	[0.002]
Focal firm wells in the area	0.014	0.014	0.013	0.013	0.012
	(0.156)	(0.154)	(0.165)	(0.186)	(0.194)
	[0.010]	[0.010]	[0.009]	[0.010]	[0.009]
Focal firm RRE	0.089	0.089	0.096	0.089	0.097
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
	[0.004]	[0.004]	[0.005]	[0.004]	[0.005]
MFMC	0.003	0.005	0.004	0.006	0.005
	(0.547)	(0.422)	(0.525)	(0.247)	(0.350)
	[0.006]	[0.006]	[0.006]	[0.006]	[0.006]
Rivals’ BI		0.275	0.271	0.273	0.269
		(0.000)	(0.000)	(0.000)	(0.000)
		[0.002]	[0.002]	[0.002]	[0.002]
Rivals’ BI * Focal firm RRE			−0.023		−0.022
			(0.000)		(0.000)
			[0.002]		[0.002]
Rivals’ BI * MFMC				−0.011	−0.009
				(0.002)	(0.006)
				[0.004]	[0.003]
AICnull–AICmodel	135,724	175,094	177,034	175,463	177,289
R 2	0.6847	0.6961	0.6967	0.6962	0.6968

Note. N = 1,068,830; N_panels = 5,415. Robust SEs clustered by unit are shown in square brackets; p values are shown in parentheses; two-tailed tests. Year and firm fixed effects are not reported. All continuous predictors are standardized.

The results in Tables 2 and 3 support all Hypotheses (at p < .05). The hypothesized main effects and moderators were significant both individually (Models 2–4 and 7–9) and in the full models (Models 5 and 10). Per the incidence rate ratios, in Model 5, with each standard deviation (SD) increase in Rivals’ BI, the predicted rate of Focal Firm BI increases by 31.9%. Based on the interaction terms, with each SD increase in Focal Firm RRE, this effect decreases by 8.8%, and with each SD increase in MFMC, it increases by 5.3%. Following best practices to emphasize the economic magnitude of results in such models (Busenbark, Graffin, Campbell, & Lee, 2022; Zelner, 2009), we additionally interpret our results as percentage changes in the predicted count outcome based on marginal effects. Holding all other variables at their mean, increasing Rivals’ BI from its mean to its mean plus 1 SD is associated with an approximately 31% increase in the predicted count of Focal Firm BI. However, at a high level of (+1 SD) Focal Firm RRE, increasing Rivals’ BI from its mean to its mean +1 SD is associated with only a 19% increase in the predicted count of Focal Firm BI. Figure C1 (see Appendix C) illustrates this negative moderation effect based on predicted values. Conversely, at a high level of (+1 SD) MFMC, the same increase in Rivals’ BI corresponds to a 38% increase in the predicted count of Focal Firm BI. Figure C2 illustrates this positive moderation effect.

Given the difficulties in illustrating and interpreting interaction effects from nonlinear fixed-effects models (Wiersema & Bowen, 2009), we also calculated the “true interaction effects” to further assess these interactions (see Appendix C) using the technique of Ai and Norton (2003). ⁴ Accordingly, the “true interaction effect” values for Rivals’ BI and Focal Firm RRE were negative and significant (at p < .05) for around 76% of all observations (see Figure C3); firms with high relative resource endowments were less responsive to their rivals’ bundling activity, granting further support for Hypothesis 2a. In Figure C4, we present the “true interaction effects” for Rivals’ BI and MFMC, which were positive for all observations, with around 95% of them also significant (at p < .05). In line with Hypothesis 3a, MFMC intensifies the positive relationship between Rivals’ BI and Focal Firm BI.

Model 10—the full model predicting nonconformity—accounted for 69.68% of within panel variance. The dependent variable, Focal Firm BN, was logged, and the predictors were standardized. We found that holding all else constant, 1 SD increase in Rivals’ BI results in a 30.91% increase in Focal Firm BN. However, the coefficients of the interaction terms suggest that with every SD increase in Focal Firm RRE, the positive effect of Rivals’ BI on Focal Firm BN decreases by 2.20%, and with every SD increase in MFMC, it decreases by 0.93%.

Figure C5 illustrates the interaction plot for the predicted values of Focal Firm BN based on Rivals’ BI and Focal Firm RRE. In Figure C6, we plot the marginal effects of Rivals’ BI on Focal Firm BN at ±1 SD of Focal Firm RRE. These plots suggest that, although not large in its magnitude, Focal Firm RRE weakens the positive relationship between Rivals’ BI and Focal Firm BN. In Figure C7, we plot the interaction between Rivals’ BI and MFMC on Focal Firm BN. Figure C8 illustrates the marginal effects of Rivals’ BI on Focal Firm BN at ±1 SD of MFMC. Both plots suggest a similar pattern. Although, as hypothesized, we found a negative and significant moderation effect, on average, this effect was relatively small, which could stem from the narrow breadth of actions reported in our sample (see the Discussion section below).

Endogeneity

We conducted a battery of tests to address potential endogeneity in our models (see Appendix D). First, we predicted Focal Firm BN, our continuous outcome variable, using a two-stage least squares (2SLS) instrumental-variable approach. In the 2SLS models, we relied on lagged instrumental variables relevant to and highly correlated with the endogenous regressors but not with the outcome variable, Focal Firm BN. In Appendix D, we define these instruments, outline our identification strategy, and summarize the sensitivity analysis techniques employed. In Table D1, we report the second-stage results from our 2SLS fixed effects panel-data models predicting Focal Firm BN along with the statistics addressing the conditions of relevance, exogeneity, and exclusion for all models. In Table D2, we provide the first-stage results. In Table D3, we report the details and results of the sensitivity analyses and falsification tests conducted using three complementary techniques, which provide additional support for the validity of our instrumentation.

While the fixed effects included in our models account for unobserved heterogeneity at the firm and year levels (Certo & Semadeni, 2006), to assess the robustness of inferences to omitted variable bias—typically driven by unobservable heterogeneity, and a common source of endogeneity (Bascle, 2008)—we also implemented the impact threshold of a confounding variable (ITCV) and its extension, the robustness of inference to replacement (RIR) techniques (Busenbark, Yoon, Gamache, & Withers, 2022). In Appendix D, we discuss the details of these analyses and report our findings, which provide additional evidence for the robustness of our results from both the fixed-effects OLS and PPML models to omitted variable bias and the potential endogeneity it may cause. Taken together, these analyses provide substantive support for the robustness of our findings to endogeneity concerns.

Robustness Checks

To assess the robustness of our findings, we employed alternative model specifications using different estimators, variable operationalizations, lag structures, and sets of control variables. We also estimated alternative versions of both dependent variables. We conducted additional robustness and diagnostic checks to assess the influence of outliers and potential multicollinearity, among some other important considerations. Appendix E describes these analyses in detail and reports the associated results from Models 11–28 in Tables E1 and E2.

Contributions

Firms’ ability to compete “is rooted in firm resources” (Chen et al., 2021: 1824) and “much progress needs to be made to understand the complexity of resource competition” (Capron & Chatain, 2008: 115). Our study of the upstream oil and gas industry demonstrates how rivals’ bundling activity at the market level drives firm-level bundling activity and nonconformity. Whereas an advantage in strategic resource endowments weakens focal firm response via bundling and nonconformity, MFMC intensifies it, albeit in a more conforming and imitative manner. Our study contributes to the development and bridging of competitive dynamics and resource-based perspectives.

We posit that, although bundling is an internal activity, it nevertheless has competitive drivers and that it is prone to escalation. The RBV focuses on the identification, procurement, bundling, and deployment of strategic resources but neglects firms’ interactions outside market contexts. Especially in efficient factor markets, resource bundling and organizational capabilities can underlie effective competition, and firms do not take resource bundling actions in a vacuum. We address this neglected aspect of resource management, which warrants more attention both from the RBV and dynamic capability scholars. In doing so, we also distinguish the competitive implications of product and factor market overlap to suggest how the combination of the two can facilitate the escalation of bundling activity among rivals.

The competitive dynamics literature and most of the strategic analysis and market positioning literature have concentrated on product markets. Rivalry in factor markets has received less attention, and bundling activity has largely been overlooked. Our research demonstrates distinctive dynamics that drive the nature, evolution, and development of bundling processes in a competitive setting. Our findings that focal firm bundling intensity and differentiation are conditioned by rivals’ bundling activity suggest a new direction for competitive dynamics, particularly where market commonality and resource similarity are high (Chen, 1996). These are critical strategic processes worthy of further study and have the potential to bridge the resource-based and competitive analysis perspectives. If rivals react to each other’s bundling activity and bundling can be a salient response to macro competitive pressures, as we suggest, then bundling should also be incorporated into existing models of competition and rivalry, as its omission risks model misspecification.

Our study also contributes to the work on multimarket competition in two ways. First, we capture how MFMC affects firm bundling activity, which is not subject to the forbearance-retaliation mechanisms due to its internal nature. Our findings confirm the distinct nature of bundling activity and its competitive drivers across factor markets. Second, this is the first study to empirically examine the implications of MFMC for interfirm dynamics.

Limitations and Future Research Opportunities

Our study is just a beginning in what may prove an expansive sphere of inquiry and thus has limitations. For example, we analyzed bundling activity among firms that produce a relatively undifferentiated output. Although this was useful in minimizing the influence of product market rivalry on firm bundling behavior, we must take care when generalizing to other industries. Our study also examines one region of the United States, in an industry with specific commercial and technical competitive options, which may not apply to other regions or jurisdictions. In our sample, bundling activity is visible largely due to the public nature of permit information related to firms’ actions through the regulating agency. In different settings, rivals’ awareness may be driven by alternative mechanisms, such as discerning. Thus, future research could explore bundling interactions in other industries.

Our research examines firms at a distance and assesses objective indicators of competition. Thus, we infer motives from actual behavior, but we cannot know the motivations and rationalizations of the managers of these firms. Therefore, it is important to perform qualitative studies to understand why managers adopted the strategies they did and their reasoning for bundling intensively or relying on resource endowments and imitating or differentiating action repertoires. We recommend such studies in the future.

Next, due to the high “decay” rates of oil and gas reserves (Karadag & Poppo, 2023)—the strategic resources in our context—the bundling intensity of firms we examined may be higher than that observed in other contexts. Accordingly, future empirical research on the interfirm dynamics of bundling where firms rely on different types of strategic resources (e.g., human resources and patented technologies) can advance both the RBV and competitive dynamics. Given that multimarket contact with rivals can influence entry/exit behavior (Korn & Baum, 1999), future research could also examine how bundling interactions among rivals shape these dynamics across product and factor markets.

Finally, our study was subject to data limitations. First, we could not examine the financial implications of our findings due to a lack of data. Second, we found a relatively weak moderating effect of MFMC on the relationship between rivals’ bundling and focal firm bundling nonconformity; future research would benefit from examining this effect in contexts with greater variety in bundling activity. Third, because our sample contained lease data only for land developed for production, we used new well drills as a proxy for resource structuring activity. Finally, we lacked data on firms’ ownership structure, alliances, and merger and acquisition activity, which may also be relevant to resource bundling. Rivals may engage in cooperative actions such as strategic alliance formation (Chen & Miller, 2015), which fall outside the scope of this study. Subsequent researchers would do well to investigate these aspects of resource bundling and its competitive implications.