The Double-Edged Sword of Entrepreneurial Orientation: Product Recalls and the Role of COO Power

Abstract

This study examines how entrepreneurial orientation (EO) influences firms’ likelihood of product recalls. Integrating EO and upper echelons theory, we first argue that EO’s bold, variance-enhancing actions increase a firm’s product recall risk by diverting managerial attention away from quality control. Using 23 years of data on U.S. public firms, we find that EO increases recall likelihood. Second, we argue that this relationship is moderated by chief operating officer (COO) power, and the effectiveness of COO power is contingent on the firm’s product life cycle context. Empirical analyses support our theory and offer new insights about EO’s potential downsides.

Keywords

entrepreneurial orientation upper echelons theory product recalls product life cycle quantitative methods

Introduction

Scholars have long extolled the virtues of possessing an entrepreneurial orientation (EO), defined as a firm’s strategic posture toward new product-market entry and value creation (Anderson et al., 2015; Patel et al., 2015). Firms with a high level of EO, characterized by their emphasis on innovativeness, risk-taking, proactiveness, autonomy, and competitive aggressiveness (Lumpkin & Dess, 1996; Wales et al., 2021), are frequently associated with positive outcomes such as enhanced performance and competitive advantage (Rauch et al., 2009; Wales, Gupta, & Mousa, 2013). Yet, these positive effects, which have been described as “over-replicated” and explored to the point of “ad nauseam” (Wales et al., 2021, p. 575), have long been acknowledged to depend on contingencies (e.g., Covin & Slevin, 1989; Wiklund, 1999). Despite this acknowledgement, EO research continues to privilege its beneficial outcomes, leaving the theorization and empirical examination of EO’s potential downsides substantially underdeveloped (Covin & Wales, 2019; Wales, Gupta, & Mousa, 2013; Wales, Patel, & Lumpkin, 2013).

This omission is surprising, as the EO-as-experimentation perspective (Wiklund & Shepherd, 2011) explicitly frames EO as a variance-enhancing strategy—one that deliberately increases the dispersion of outcomes by encouraging exploration, bold action, and new market entry, rather than a reliable pathway to superior performance. Complementing this view, the more recent EO-as-new-value-creation theory (Wales et al., 2023, p. 1756) argues that EO’s emphasis on the continuous pursuit of new product-market entry inherently “creates a distribution of outcomes, both wins and losses.” Taken together, these perspectives suggest that downside outcomes are not anomalies but an intrinsic feature of entrepreneurial action. Yet, the EO literature has devoted far more effort to theorizing its “wins” than to specifying the concrete forms of loss it may produce.

One particularly salient downside outcome is product recalls—where firms voluntarily or involuntarily withdraw defective or potentially harmful products from the market due to safety, regulatory, or performance concerns. Product recalls are an observable, operational manifestation of bold experimentation, where the same variance-enhancing logic that drives discovery also exposes high-EO firms to potential failure (Kim et al., 2022). In this sense, product recalls arguably represent a breakdown in quality control within the experimentation process—a vulnerability heightened by the increased velocity of new market entries characteristic of high-EO firms (Wales et al., 2023). Despite their conceptual alignment with the logic of outcome variance, however, we still lack a coherent theoretical explanation that links EO to product recalls.

The practical relevance of this relationship is equally compelling. Product recalls result in a cascade of negative consequences, including revenue losses, diminished customer satisfaction, tarnished brand reputation, and legal liabilities stemming from product-related claims (e.g., Chen et al., 2009; Eilert et al., 2017; Liu et al., 2017). Consider Samsung’s recall of their Galaxy Note 7 smartphones in August 2016, after reports emerged of batteries overheating, catching fire, and exploding. This recall resulted in losses exceeding $5 billion, substantial reputational damage, and personal injury to consumers (Reisinger, 2016). In the United States, the Consumer Product Safety Commission (CPSC) reviews about 8,000 death certificates each year relating to unintentional product-related deaths (CPSC, 2024).

Our study, therefore, seeks to extend our understanding of the EO-as-experimentation perspective by developing and testing a theory that directly links EO to product recalls. Drawing on the EO and product failure literatures, we first argue that firms with a higher EO are more exposed to product recall risk. Although high-EO firms have the potential to be more proactive in addressing product-related problems pre-launch (Sahoo & Yadav, 2017), their emphasis on bold experimentation and continuous new market-entry moves (Anderson et al., 2015; Wales, Gupta, & Mousa, 2013) will more likely increase the risk that products are commercialized before quality and safety issues are fully resolved. This quality-based mechanism reflects a fundamental tension in entrepreneurial experimentation: the same action logic that enables rapid discovery also compresses development and testing cycles, heightening exposure to post-launch defects (Kim et al., 2022).

Second, we advance contingency arguments about when and why EO is more or less likely to pose a recall risk to firms. To do so, we integrate insights from upper echelons theorizing into our theory (e.g., Hambrick, 2007; Hambrick & Mason, 1984), which argues that organizational outcomes reflect the experiences, values, and cognitive frames of top executives, and the influence that they exert within the top management team (TMT). We argue that the power of the chief operating officer (COO)—the executive responsible for internal operations and quality assurance (Hambrick & Cannella, 2004; Krause et al., 2013; Menz, 2012)—represents a critical contingency shaping how EO’s experimentation logic unfolds. As the firm’s quality advocate, a powerful COO is better positioned to embed operational discipline into strategic decision-making. Thus, we theorize that COO power weakens the positive relationship between EO and product recalls. Further, given upper echelons theory (UET) asserts that organizational context influences how executive characteristics translate into outcomes (Finkelstein et al., 2009), we also argue that the effectiveness of COO power depends on the firm’s concentration of innovation initiatives across the product life cycle (i.e., the proportion of activities at early- versus late-stage positions in the life cycle). When innovation portfolios are concentrated in the early stages—where design uncertainty and quality risks are greater—the demands on COO attention increase, reducing the extent to which COO power can effectively mitigate EO-related recall risk.

We test our hypotheses using a large sample of U.S. firms dealing with physical consumer products between 1998 and 2021. Our study contributes to the EO-as-experimentation perspective in two fundamental ways. First, we extend the theorized consequences of EO beyond performance variation to examine a specific and costly form of failure—product recalls (Hersel et al., 2019). Although prior research has recognized that EO increases variance in firm outcomes, most work has focused on the upper tail of the distribution or treated failure in only abstract terms as exit or underperformance (e.g., Gali et al., 2024). By directly linking EO to product recall risk, we show how variance-enhancing strategies can also generate tangible operational failures due to weakened quality control, thereby expanding our understanding of how EO can produce lower tail outcomes. Second, drawing on UET, we identify the conditions under which EO’s downside consequences are amplified or mitigated. Our study suggests that COO power, which proxies a TMT’s strategic focus on operational excellence and quality, attenuates the positive association between EO and recalls. Moreover, we demonstrate that this buffering effect is contingent on the firm’s product life cycle context. Together, these insights illuminate how the composition and power structure of the TMT shape the realization of entrepreneurial experimentation.

Theory and Hypotheses

EO-as-Experimentation

EO is commonly viewed as a performance-enhancing strategic orientation, enabling firms to discover and exploit opportunities through innovation, risk-taking, proactivity, autonomy, and competitive aggressiveness (Lumpkin & Dess, 1996; Rauch et al., 2009), which in turn improves firm performance. This view—which has been termed the EO-as-advantage perspective—“implicitly assumes that EO somehow provides an advantage to firms” (Wiklund & Shepherd, 2011, p. 929), typically by raising average returns, such as revenue growth, profitability, or market share. That is, EO is seen as shifting the firm’s performance distribution upward, improving expected outcomes. Importantly, EO scholarship has long recognized that such entrepreneurial postures are enacted by executives whose experiences, values, and cognitive frames guide the firm’s pursuit of opportunity (Covin & Slevin, 1989; Miller, 1983)—an insight rooted in UET (Hambrick & Mason, 1984). More recently, EO-as-new-value-creation theory extends this executive-centric foundation by asserting that new value only arises from EO when top managers convert entrepreneurial postures into bold and frequent product-market entry, a pattern of behavior that systematically increases product-market variance in firms (Wales et al., 2023). This view affirms that EO’s consequences depend on how executives choose to direct and enact entrepreneurial action.

Notwithstanding the benefits, EO-as-new-value-creation theory also highlights that the very mechanism through which value is generated—bold, frequent market entry enacted by managers—necessarily widens the distribution of firm outcomes (Wales et al., 2023). This insight is consistent with the variance-enhancing logic at the core of the EO-as-experimentation perspective (Wiklund & Shepherd, 2011), which is grounded in theories of organizational learning and evolutionary economics (e.g., March, 1991; Nelson & Winter, 1982). EO-as-experimentation emphasizes that EO widens the distribution of outcomes by propelling firms into novel and uncertainty-laden activities, such as launching unproven products, entering unfamiliar markets, or recombining knowledge in new ways. These actions resemble experiments, since firms often lack reliable ex-ante information about the associated risks or outcomes (Denrell, 2003; Levinthal & March, 1993). As a result, firms that embrace EO are more likely to experience extreme outcomes—either positively, through innovation-driven growth, or negatively, through strategic missteps or costly errors.

Indeed, the downside potential of EO is implicit across its five main dimensions (Lumpkin & Dess, 1996; McKenny et al., 2018). Innovativeness increases exposure to untested technologies and novel combinations, heightening the chance of technical or market failure (Singh & Fleming, 2010). Risk-taking amplifies exposure to uncertain returns, while proactiveness can lead firms to pre-emptively commit to trends that do not materialize (Patel et al., 2015). Autonomy decentralizes initiative and empowers independent action, which can increase variation in strategic choices and reduce coordination, thereby widening the range of potential outcomes that can emerge within the firm (March, 1991). Finally, competitive aggressiveness intensifies market confrontation and provokes reciprocal responses from rivals, escalating both the potential for exceptional wins and costly losses (Chen & Miller, 1994).

Yet, most EO research remains anchored in the EO-as-advantage perspective, emphasizing its performance-enhancing benefits while overlooking its downsides (e.g., Wales, Gupta, & Mousa, 2013; Wales, Patel, & Lumpkin, 2013). Even within the smaller body of research anchored in the EO-as-experimentation perspective, scholars have largely examined EO’s impact on the variance of performance outcomes—as increased dispersion around the mean—rather than examining the specific types of failure EO may produce (e.g., Covin & Wales, 2019; Wales et al., 2023; Wiklund & Shepherd, 2011). Although this was necessary to establish the variance-enhancing logic underpinning the EO-as-experimentation perspective and has significantly advanced our understanding of EO’s consequences, we know little about the concrete downside manifestations it can produce.

This limitation leaves our understanding of EO’s risks theoretically underdeveloped and empirically underexplored. Without theorizing how EO manifests as failure and the role of executives in shaping how experimentation unfolds, we lack insight into the mechanisms and contingencies through which EO’s exploratory tendencies can produce harmful outcomes. Indeed, the variance-enhancing logic underpinning EO-as-experimentation implies that failure is not incidental but intrinsic to EO. To address this omission, we focus on product recalls.

EO-as-Experimentation and Product Recalls

A product recall occurs when a firm withdraws a product from the market because it fails to meet safety or quality standards, or poses potential harm to consumers. Such events—often referred to as product-harm crises—are highly visible indicators of operational and design failures under uncertainty (Dawar & Pillutla, 2000; van Heerde et al., 2007). These events carry severe implications for firms and stakeholders alike: consumers may experience harm, investors face financial loss, and regulators often impose sanctions or corrective actions (Cleeren et al., 2017; Wowak & Boone, 2015). For firms, recalls generate substantial direct costs from remediation and operational disruption, lost revenues, and long-term reputational damage that undermines consumer trust (Chen et al., 2009; Eilert et al., 2017). Because they are externally verified, systematically recorded, and economically consequential, recalls provide an observable and theoretically meaningful manifestation of failure—precisely the kind of lower-tail outcome predicted by the variance-enhancing logic underpinning EO-as-experimentation.

Thus, we argue that being more entrepreneurially oriented will increase a firm’s product recall risk. This expectation rests on several considerations. First, entrepreneurially oriented firms are inherently predisposed to pursue novel opportunities, untested technologies, and new markets under incomplete information (March, 1991). Second, the very qualities that make EO valuable for discovery—risk-taking, innovativeness, and proactiveness—may also erode the safeguards that prevent defective products from reaching the market (Kim et al., 2022). Even when firms are vigilant, the speed and scope of their innovation activities heighten the likelihood that design flaws or process weaknesses will escape detection before market entry. Third, entrepreneurial firms’ tolerance for ambiguity and desire to lead markets can diminish managerial attention to process discipline, quality controls, and operational buffers—the organizational mechanisms that ordinarily intercept defects before they culminate in recalls (Hora et al., 2011; Kim et al., 2022).

Supporting this, Anderson et al. (2015) argued that firms with a strong EO favor actions with uncertain outcomes and commit significant resources to projects characterized by ambiguity and risk. This commitment can explain the greater variance observed among high-EO firms (Wales, Patel, & Lumpkin, 2013) and, we suggest, simultaneously heightens exposure to product recalls by redirecting managerial attention away from process reliability and quality assurance. While we do not necessarily view this as a deliberate “de-prioritization” of quality, the pace and breadth of change associated with possessing a high EO can outstrip the organization’s ability to maintain consistent reliability checks and controls. Thus, EO’s variance-enhancing nature implies that failures such as recalls are not anomalies but rather predictable risks when rapid experimentation exceeds an organization’s quality-assurance capacity. We therefore expect that firms with a higher EO are more likely to experience product recalls than firms with a lower EO. Formally stated:

Hypothesis 1 (H1): EO will be positively associated with the likelihood of product recall.

COO Power, EO, and Product Recalls

While we theorized in H1 that EO will increase a firm’s exposure to product recall risk due to diminished managerial attention to process reliability and quality concerns, we argue that this relationship will be moderated by the power of the COO. Although prior research has long acknowledged the TMT’s influence on firms’ EO (e.g., Covin & Slevin, 1989; Short et al., 2010; Wales et al., 2020), direct theoretical integration of upper echelons theorizing into EO research remains limited. We therefore draw on UET (Hambrick & Mason, 1984) to explain how variation in COO power shapes the enactment of EO’s experimentation.

UET asserts that strategic outcomes reflect how executives perceive and interpret their environments through boundedly rational filters, informed by their values and cognitive frames. Each functional executive (e.g., COO, cheif financial officer (CFO)) brings a distinct set of experiences and interpretive schemas that shape the field of vision (what the TMT notices), selective perception (what it attends to), and interpretation (how it construes risk and opportunity) in collective decision-making (Finkelstein et al., 2009; Hambrick, 2007; Ocasio, 1997). The power of these executives determines which cognitive frames are more likely to receive attention and which issues are prioritized in allocating resources. In this sense, COO power represents an attentional governance contingency that shapes the TMT’s interpretive focus (Joseph et al., 2014; Krause et al., 2015)—conditioning the extent to which entrepreneurial experimentation is accompanied by operational discipline. In the context of product recalls, this means that the effect of EO-induced risk will likely depend on the salience of the operational and quality focus of the COO.

The COO embodies the firm’s operational and quality assurance logic (Bendig, 2022). This functional executive is responsible for designing and overseeing systems that ensure process reliability, product integrity, and compliance with safety and performance standards (Hambrick & Cannella, 2004; Marcel, 2009; Menz, 2012). Conceptually, therefore, the COO represents the firm’s quality advocate—as these executives are “primarily devoted to operational issues, and develop specific expertise in managing internal operations” (Krause et al., 2013, p. 1629). Accordingly, we argue that COO power signals how strongly this operational logic shapes the TMT’s collective interpretation of uncertainty and the extent to which EO’s variance-enhancing tendencies are channeled through structures that emphasize reliability and control.

As COO power increases, operational and quality considerations gain greater prominence in the TMT, directing the attention of the TMT toward reliability, iterative testing, and process control. Consequently, EO’s variance-enhancing tendencies are more likely to be expressed through disciplined, learning-oriented experimentation when COO power is high. This aligns with broader work on organizational learning, which emphasizes the dual need for exploration and exploitation capabilities in innovation contexts (Lavie et al., 2010; Raisch & Birkinshaw, 2008). The COO’s operational expertise and lower-level focus embody exploitation—discipline, coordination, and quality control—that complements the broader entrepreneurial (exploratory) tendencies within high-EO firms. When these logics are effectively integrated, EO is enacted as structured, feedback-rich experimentation rather than trial-and-error improvisation.

Furthermore, increasing COO power generates symbolic and behavioral spillovers across the organization. It signals to middle managers and employees that quality and reliability are strategic imperatives. This ensures that quality systems, oversight processes, and early warning mechanisms are adequately resourced and embedded within the organization. Indeed, empirical studies of product failure show that firms emphasizing internal coordination and operational alignment are better able to detect and mitigate quality problems early (e.g., Haunschild & Sullivan, 2002; Hora et al., 2011). Thus, we hypothesize:

Hypothesis 2 (H2): COO power weakens the positive association between EO and the likelihood of product recall.

In addition to executive roles, UET also highlights that organizational context acts as a critical contingency influencing how executive characteristics translate into firm outcomes (Finkelstein et al., 2009; Hambrick, 2007). As Hambrick and Mason (1984, p. 197) noted, “the situation, upper echelon characteristics, and strategic choices interact to determine organizational performance levels.” This suggests that the moderating effect of COO power on EO’s downside risks will vary across contextual conditions. In the case of product recalls, a particularly important factor is the firm’s product life cycle, which captures the distribution of innovation efforts across early- and late-stage development activities (Utterback & Abernathy, 1975).

The firm’s distribution of innovation activities across the product life cycle fundamentally alters the firm’s exposure to design uncertainty, technological novelty, and process complexity—factors that directly influence product quality and, consequently, recall risk. In early-stage activities, products remain unproven, technologies evolve rapidly, and iteration cycles tend to be fast (Hoberg & Maksimovic, 2022). These conditions heighten design volatility and make product failures more likely, thereby intensifying the importance of operational discipline and quality oversight (Thirumalai & Sinha, 2011). Under such conditions, COOs must distribute limited attentional resources to a broad set of uncertain and fast-moving initiatives. This dilutes the COO’s ability to monitor each initiative closely and weakens their capacity to offset EO-induced recall risks. As a result, the weakening effect of COO power on the positive relationship between EO and product recalls will be comparatively smaller. By contrast, when a firm’s innovation portfolio is concentrated in late-stage activities, where products and technologies are more stable and design architectures well-established, the COO faces fewer highly uncertain initiatives that require intensive monitoring. With a narrower set of risky (early-stage) projects demanding attention, COO power can be applied more effectively towards reliability safeguards, process discipline, and quality control. Under such conditions, the weakening effect of COO power on the positive relationship between EO and product recalls will be comparatively stronger. Formally stated:

Hypothesis 3 (H3): The extent to which COO power weakens the positive association between EO and product recalls will be stronger (weaker) when the firm’s innovation portfolio is concentrated in late-stage (early-stage) product life cycle activities.

Methods

Data and Sample

To test our hypotheses, we constructed a panel dataset spanning 1998 through 2021 by integrating data from multiple sources. We began by identifying all U.S. public firms operating in industries regulated by the CPSC. This yielded a comprehensive sample of firms exposed to product recall risk, irrespective of whether a recall occurred during the study period. We then matched CPSC recall data, which contains information on all consumer product recalls in the United States, to these firms to identify recall events. Our final sample includes both recall and non-recall observations across a range of CPSC-regulated industries. Further details about industry distribution are detailed in Table A4 of the Supplemental Appendix.

To measure EO, we derived a text-based indicator from annual 10-K filings submitted to the Securities and Exchange Commission’s (SEC) Electronic Data Gathering, Analysis, and Retrieval (EDGAR) system. Additional executive- and firm-level data were drawn from Execucomp and Compustat for our main analyses. We obtained information on firms’ product life cycle portfolios from Hoberg and Maksimovic (2022), who developed a novel text-based classification scheme. For supplementary analyses probing firms’ emphasis on quality as our core theoretical mechanism, we incorporated firm-level quality management certification and product quality controversy data from the London Stock Exchange Group (LSEG) databases. After merging these data, and removing observations outside the sample window or missing product recall data, the final panel comprised 9,068 firm-year observations across 901 firms.

Variables

Dependent Variable

Recalls are rare events. Of the 9,068 firm-year observations in our sample, 93.9% (8,508 observations) experienced no recall, 3.6% (331) experienced one recall, and 2.5% (229) experienced multiple recalls. Thus, 97.5% of our observations fall within the 0 to 1 recall range, indicating most variation occurs at the extensive margin—whether a recall occurs—rather than the intensive margin—how many recalls occur conditional on one taking place. Accordingly, our primary analysis uses panel conditional logit models, consistent with prior studies (e.g., Hoffmann et al., 2024; Kini et al., 2017). We code the dependent variable, Product Recall, as a binary indicator equal to [1] if the firm experienced at least one product recall in a given year, and [0] otherwise. This specification aligns with our theoretical focus on whether EO increases the likelihood of observable product quality failures.

Independent Variable

We measured EO using an established text-analytic approach that draws on validated lexical methods (see McKenny et al., 2018; Short et al., 2010). Specifically, we analyzed firms’ annual 10-K filings—mandatory, standardized disclosures submitted to the SEC via EDGAR to extract EO-related language. This approach has several advantages over survey-based alternatives. First, it allows for coverage of a large, representative sample of public firms. Second, 10-Ks serve as key strategic disclosures authored by senior executives. Third, 10-Ks reflect input from multiple functional areas, thus providing a comprehensive view of the firm’s core business—outlining where it operates and how it competes, its operations, and financial state. Finally, the mandatory nature and regular issuance of these reports allow for longitudinal and cross-comparative analysis.

Following McKenny et al. (2018), we use word dictionaries aligned to five behavioral dimensions of EO, originally proposed by Lumpkin and Dess (1996), to construct our measure. These include innovativeness, risk-taking, proactiveness, autonomy, and competitive aggressiveness. They refine the framework by removing words frequently used in non-entrepreneurial contexts and incorporating construct-relevant terms validated through manual coding.¹ This refinement process substantially enhances construct validity by reducing false positives and false negatives. Details about the dictionaries used are provided in Table A1 of the Supplemental Appendix.

To address contextual accuracy concerns inherent in traditional word-counting approaches, we employ sentence-based analysis rather than simple word frequency counts. Following recent advances in textual analysis (see Carlson & Burbano, 2025; de Kok, 2025), we utilized generative large language models (GLLMs) to verify that EO-related terms appear in genuinely entrepreneurial contexts, eliminating sentences where terms were used negatively. For example, “we do not take unnecessary risks in uncertain markets” uses risk terminology in a negative context; whereas “we take calculated risks to pursue innovative opportunities” represents a positive EO. Our validation process involves examining all sentences containing at least one word or phrase from the lexicon, with each sentence individually reviewed by the GLLM. Overall, a total of 43.5 million sentences were examined, of which 7.2% were identified as false positives and subsequently excluded from the analysis.²

To facilitate comparison across firms and years, we normalize the raw sentence count by the total number of sentences in the 10-K report and standardize the ratio within each industry-year. Normalization addresses the variation in 10-K report lengths across firms, while standardization controls for industry (at the 4-digit Standard Industrial Classification (SIC) level) and temporal differences in entrepreneurial discourse. This approach yields a measure indicating each firm’s relative EO compared to industry peers in the same year, capturing genuine strategic orientation rather than document length artifacts or industry conventions.³

Moderating Variables

To test our moderating argument in H2 regarding COO power, we follow prior research and use compensation as an observable indicator of executive power (e.g., Finkelstein, 1992). Indeed, the link between executive power and compensation is unequivocal in the literature and therefore represents a valid structural proxy (Ozgen et al., 2025). First, we identified whether firms in our sample had an executive in the TMT with a COO title. If there is a COO, we calculated their power by dividing the COO’s total remuneration (TDC1 in Execucomp) by the total remuneration of the CEO. Firms without a COO were coded as [0]—as COO power was not present inside the organization. While firms typically had COOs that were less powerful than the CEO (i.e., COO power < 1), 4.5% of COOs in our sample had power equal to or greater than the CEO.

To test our contingency arguments relating to product life cycle in H3, we follow Hoberg and Maksimovic (2022) and textually analyze firms’ 10-K reports to derive the variable Life Cycle as a continuous measure. The measure captures firms’ product portfolio positioning across life cycle stages by identifying the co-occurrence of terms in paragraphs such as “product” and “launch” (for early stages), and “product” and “discontinued” (for late stages; see Hoberg & Maksimovic, 2022). Using this approach, the authors developed four probability measures for each firm-year representing distinct stages: Life1 (product innovation), Life2 (process innovation), Life3 (maturity), and Life4 (decline). These probabilities sum to [1], providing a comprehensive view of a firm’s innovation portfolio composition.

Our operationalization of Life Cycle sums Life3 and Life4 probabilities, creating a continuous measure from [0] to [1]. Higher values indicate portfolios concentrated in late stages (maturity or decline), whereas lower values indicate portfolios concentrated in early stages (innovation or development).

Control Variables

We included several TMT-, firm-, and industry-level control variables to rule out alternative explanations. At the individual-level, we control for TMT compensation to proxy for the risk-taking attitude of the TMT directly attributable to executive pay. For this purpose, we include TMT Stock Option Pay, calculated as the average percentage of stock options relative to the total compensation of the top executives in the TMT. We also controlled for other TMT characteristics, including gender by measuring TMT Female, which is the proportion of female executives in the TMT, and TMT Age, which is the average age of top executives in the TMT.

At the firm-level, we include Firm Age (years since founding) and Firm Size (log of total assets). Firm age captures differences in organizational maturity that may affect recall practices, while firm size accounts for scale, visibility, and process capabilities that may influence recall likelihood. We also control for financial characteristics potentially correlated with both EO and recall likelihood: Leverage (long-term debt to total assets), return on assets (ROA: net income to total assets), Revenue, and Market-to-Book ratio, which is the ratio of market value to equity book value. In addition, we include R&D Intensity (R&D expenditure to revenue) and Labor Intensity (employees to revenue) to account for differences in innovation activity and labor use, respectively (Thirumalai & Sinha, 2011).

At the industry-level, we include Munificence, calculated as the coefficient of a regression of industry sales against time, and Dynamism, computed as the standard error of the slope in a regression of industry sales against time. As competition can also play a role (Ball et al., 2018), we capture industry competitiveness by measuring its Herfindahl–Hirschman index, calculated as the sum of the squared market share percentages of all firms operating within the same industry. For all these environmental variables, we classify industries by their 4-digit SIC codes.

In the mechanism analyses presented later in the paper and robustness checks presented in the Supplemental Appendix, we discuss additional variables not defined here. Supplemental Appendix Table A2 provides a complete list of all the variables used, along with information on their source and how each measure was operationalized.

Estimation Strategy

We employ a two-way fixed effects panel model to examine how EO influences the likelihood of product recall, controlling for TMT, firm, and industry characteristics. This approach considers the possibility that a firm experiencing a product recall at time t may change its EO later in the same period. Additionally, it is plausible that a product recall occurring at time t could subsequently impact the firm’s financial performance or value. We lag all independent variables by 1 year to account for potential reverse causality. Our conditional logit model is expressed in Equation (1):

\Pr (Product Recal l_{it} = 1) = F [β E O_{it - 1} + γ^{'} X_{it - 1} + δ^{'} Z_{jt - 1} + α_{j} + θ_{t}]

(1)

where i indexes the firm, j indexes the industry (4-digit SIC), and t represents year. The binary outcome variable is Product Recall_it, while EO_it-1 represents our measure of EO. X_it-1 is a vector of TMT- and firm-level controls, while Z_it-1 captures industry-level covariates. The parameters $α_{j}$ and $θ_{t}$ represent our industry and year fixed effects, respectively.

Analysis and Results

Main Analyses

Table 1 provides the means, standard deviations, and pairwise correlations for the variables used in our main analyses. The mean of Product Recall is 0.06, indicating that approximately 6% of observations in our analysis involve firms experiencing a recall. Visual inspection of the correlation coefficients and examination of variance inflation factors (VIFs) indicate that multicollinearity is not an issue (all VIF values are below the threshold level of 10; average VIF = 1.61). Further descriptive details of the variables used are provided in Supplemental Appendix Table A3.

Table 1.

Descriptive Statistics and Pairwise Correlations.

	Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)	(16)	(17)	(18)
(1)	Product recall	1.00
(2)	EO	0.03*	1.00
(3)	Firm age	0.11*	0.02*	1.00
(4)	Firm size	0.19*	0.04*	0.48*	1.00
(5)	Leverage	0.01	−0.02	0.16*	0.24*	1.00
(6)	ROA	0.04*	0.04*	0.07*	0.16*	−0.01	1.00
(7)	Market-to-book	0.01	−0.01	0.00	0.02*	−0.00	0.02*	1.00
(8)	Revenue	0.26*	0.03*	0.25*	0.52*	0.07*	0.05*	0.02	1.00
(9)	R&D intensity	−0.05*	0.04*	−0.10*	−0.11*	−0.07*	−0.19*	−0.00	−0.04*	1.00
(10)	Labor intensity	−0.01	0.02	−0.07*	−0.16*	0.04*	−0.05*	−0.00	−0.08*	0.28*	1.00
(11)	Munificence	0.05*	0.01	−0.13*	−0.04*	−0.09*	−0.03*	−0.00	−0.00	0.03*	0.02*	1.00
(12)	Dynamism	−0.04*	0.00	−0.04*	−0.10*	−0.05*	−0.04*	0.00	−0.08*	0.05*	−0.13*	−0.16*	1.00
(13)	HHI	0.16*	−0.01	0.21*	0.18*	−0.01	0.03*	0.01	0.20*	−0.07*	−0.24*	0.04*	0.07*	1.00
(14)	TMT age	0.04*	−0.04*	0.31*	0.24*	0.07*	0.06*	−0.00	0.14*	−0.05*	−0.02	−0.08*	−0.04*	0.13*	1.00
(15)	TMT female	0.04*	0.06*	−0.01	0.00	−0.01	0.04*	0.00	0.01	−0.06*	0.13*	−0.03*	−0.10*	−0.06*	−0.05*	1.00
(16)	TMT stock option pay	0.00	0.02	0.05*	0.06*	0.02	0.05*	−0.02	−0.03*	0.02*	−0.06*	−0.01	0.01	0.01	−0.07*	0.02	1.00
(17)	COO power	−0.00	−0.02	−0.09*	−0.00	−0.03*	−0.02	0.00	−0.01	−0.01	0.05*	0.03*	−0.01	−0.01	−0.02*	−0.01	−0.04*	1.00
(18)	Life cycle	−0.00	0.00	0.09*	0.11*	−0.07*	−0.01	−0.00	0.11*	0.01	−0.18*	0.01	0.10*	0.07*	0.06*	−0.15*	−0.05*	−0.01	1.00
	Mean	0.06	0.16	25.37	7.25	0.17	0.04	3.76	0.07	6.58	5.69	1.08	1.02	0.30	52.82	0.10	2.16	0.22	0.35
	S.D.	0.24	0.93	17.29	1.72	0.22	0.18	5.77	0.26	9.78	5.30	0.07	0.02	0.20	5.19	0.14	3.45	0.46	0.13

Note. N = 9,068. Pairwise correlations based on observations that are used in our baseline and moderator regression models.

COO: chief operating officer; EO: entrepreneurial orientation; HHI: Herfindahl–Hirschman index; ROA: return on assets; TMT: top management team.

Denotes 5% level of significance.

To test H1, we examine whether EO is associated with an increased likelihood of recall by estimating Equation (1). The results of the conditional logistic regressions are provided in the first two columns of Table 2. Column 1 includes only controls. In column 2, we include EO. In support of H1, we find that EO is positively and significantly associated with the likelihood of product recall (β = .194, p < .01). Following the method suggested by Kitazawa (2012), we calculate the average semi-elasticity of product recall likelihood with respect to a unit change in EO to determine effect size. Our results indicate that a one standard deviation increase in EO increases the probability of experiencing a recall by 16.90%.

Table 2.

Regression Results of the Effects of EO and Moderating Role of COO Power.

DV: Product recall	(1)	(2)	(3)	(4)
EO		0.194*** (0.068)	0.267*** (0.077)	0.538*** (0.147)
COO power			−0.036 (0.133)	−0.061 (0.128)
Life cycle				0.091 (0.852)
EO × COO power			−0.458** (0.210)	−0.461** (0.198)
EO × Life cycle				−0.840** (0.371)
COO power × Life cycle				0.028 (0.030)
EO × COO power × Life cycle				−0.059** (0.030)
Firm age	0.005 (0.008)	0.003 (0.008)	0.003 (0.007)	0.001 (0.008)
Firm size	0.469*** (0.106)	0.466*** (0.105)	0.472*** (0.103)	0.429*** (0.134)
Leverage	−0.365 (0.472)	−0.376 (0.473)	−0.412 (0.467)	−0.125 (0.621)
ROA	0.476 (0.884)	0.408 (0.884)	0.364 (0.861)	0.316 (1.166)
Market-to-book	0.001 (0.001)	0.001 (0.001)	0.001 (0.001)	0.041** (0.016)
Revenue	0.797** (0.351)	0.820** (0.346)	0.846** (0.343)	1.713* (1.001)
R&D intensity	−0.015 (0.024)	−0.021 (0.024)	−0.022 (0.025)	−0.031 (0.029)
Labor intensity	−0.014 (0.027)	−0.014 (0.027)	−0.015 (0.027)	−0.008 (0.030)
Munificence	1.262 (0.835)	1.219 (0.849)	1.299 (0.833)	1.449 (0.893)
Dynamism	−2.592 (7.067)	−2.733 (7.048)	−2.491 (6.993)	−.732 (7.461)
HHI	−0.006 (0.685)	0.008 (0.689)	−0.069 (0.680)	0.087 (0.674)
TMT age	−0.018 (0.021)	−0.015 (0.021)	−0.014 (0.021)	−0.012 (0.021)
TMT female	0.592 (0.648)	0.508 (0.638)	0.475 (0.648)	0.463 (0.669)
TMT stock option pay	0.030 (0.025)	0.029 (0.024)	0.031 (0.024)	0.024 (0.024)
Observations	9,068	9,068	9,068	9,068
Chi-squared	355.6***	406.3***	422.1***	359.7***
Pseudo-R²	0.130	0.135	0.137	0.137

Note. All independent variables in the models are lagged by 1 year. The models are estimated using conditional logit and include both year and industry fixed effects. Robust standard errors are reported in parentheses.

COO: chief operating officer; EO: entrepreneurial orientation; HHI: Herfindahl–Hirschman index; ROA: return on assets; TMT: top management team.

p < .1. **p < .05. ***p < .01.

In H2, we predicted that the positive relationship between EO and product recalls would be weakened by COO power. To test this hypothesis, we included an interaction term between EO and COO power in Model 3 of Table 2. In support of H2, we find that COO power negatively moderates the positive EO–recall relationship (β = .458, p < .05). We then estimated the marginal effects of EO on recall likelihood at different levels of COO power to examine whether the nature of the EO-recall relationship changes as a function of COO power. We focus on values of COO power between [0] and [1], which represented approximately 95% of firm-year observations in our sample. The results are presented in Table 3. Interestingly, the results suggest that COO power not only weakens the positive association between EO and product recalls, but also reverses the relationship from positive to negative at high levels of COO power. Specifically, we find that the valence of the effect switches from positive to negative and is marginally significant when COO power is approximately 0.8.

Table 3.

Marginal Effects Analysis.

COO power	Average marginal effects of EO
0	0.027*** (0.003)
0.2	0.019*** (0.003)
0.4	0.010*** (0.003)
0.6	0.002 (0.004)
0.8	−0.007* (0.004)
1	−0.015*** (0.005)

Note. Robust standard errors are reported in parentheses.

Estimates based on full model results in Table 2 (Model 3).

COO: chief operating officer; EO: entrepreneurial orientation.

p < .1. **p < .05. ***p < .01.

To ease interpretation, Figure 1 plots the average effect of EO on recall likelihood at different values of COO power. Estimates above (below) the horizontal zero-line indicate a positive (negative) effect on recall risk. As can be seen in this graph, the association between EO and recall risk weakens as COO power increases. The valence of the association between EO and recall risk shifts from positive to negative and is marginally significant at high levels of COO power (approximately > 0.8). This finding suggests that operationally focused executives not only help balance entrepreneurial ambitions with quality control considerations but can also transform EO into disciplined experimentation to enable firms to become more proactive in identifying and rectifying product risks before they manifest in recalls.

Figure 1.

Moderating effect of COO power on the EO–recall relationship.

In H3, we argued that the weakening influence of COO power on the positive relationship between EO and product recalls would be contingent on the firm’s product life cycle context. To test this, we included an interaction term between EO and Life Cycle, COO Power and Life Cycle, and a three-way interaction term between EO, COO Power, and Life Cycle in Model 4 of Table 2. In support of H3, we find a significant negative three-way interaction term (β = −.059, p < .05), which suggests that when firms have a higher concentration of late-stage innovation activities, the weakening effect of COO power is further amplified. To aid interpretation, Figure 2 plots the average marginal effect of EO on product recall likelihood across varying levels of COO power, distinguishing between firms whose portfolios are concentrated in early versus late life cycle stages. The results indicate clear contextual differences in the strength of COO power’s moderating role, consistent with H3.

Figure 2.

Interaction effects of COO power and life cycle on the EO–recall relationship.

For firms concentrated in early life cycle stages (solid line), EO exhibits a strong positive association with recall risk when COO power is low. However, this marginal effect steadily declines as COO power increases, eventually becoming non-significant. This suggests that in early-stage contexts, COO power substantially weakens—but does not overturn—the positive EO–recall relationship. In short, while COO power can meaningfully temper the risks associated with entrepreneurial aggressiveness, it cannot overturn EO’s underlying tendency to elevate recall likelihood in innovation-intensive, early-stage contexts.

By contrast, for firms with portfolios concentrated in late stages (dashed line), COO power attenuates EO’s effect more quickly and more strongly. As COO power increases, the marginal positive effect of EO on recall risks declines sharply, becoming non-significant at a COO power threshold of roughly 0.13—approximately half the sample mean. At above-mean levels of COO power, the association between EO and recall risk becomes negative and statistically significant, indicating that in mature product contexts, COO power can effectively offset EO-related recall risks and eventually reverse the effect as COO power increases. The steeper decline of the dashed line demonstrates that mature product contexts require substantially less operational oversight to achieve equivalent recall risk mitigation compared to early-stage contexts. These findings suggest that EO only becomes proactive in identifying product issues before they manifest in recalls under specific contextual circumstances: when COO power is high and innovation activities are concentrated in late stages.

Finally, Figure 3 plots the COO power threshold required to neutralize EO’s effect on product recall likelihood across different product life cycle stages. As 95% of COOs receive remuneration below that of the CEO, we focus the y-axis on COO power between 0 and 1. The downward-sloping curve shows that as product portfolios mature, the COO power required to offset EO-related risks declines substantially. For firms with more early-stage portfolios (life cycle = 0.30), the neutralization threshold is approximately 0.77 (3.5 times the sample mean). In contrast, for firms with more mature portfolios (life cycle = 0.70), the threshold falls to around 0.13, well below the sample mean, suggesting that even modest levels of COO power are sufficient to neutralize EO-related recall risk in late-stage contexts.

Figure 3.

COO power threshold given product life cycle.

Addressing Sampling Concerns and Endogeneity

To address potential concerns about our sampling strategy and strengthen causal inference, we conducted a permutation test—a robust counterfactual approach (Rosenbaum, 2002). This test examines whether our observed relationship between EO and recall likelihood could have emerged by chance or from industry-specific factors rather than firm-level EO. In this test, we maintained the structure of our data but disrupted the hypothesized causal mechanism by randomly reassigning each firm’s EO score to another firm within the same industry. We then re-estimated our baseline model with these permuted values and recorded the resulting coefficient. This process was repeated 1,000 times to generate a null distribution, which is displayed in Figure 4, representing what would be observed if firm-level EO had no true relationship with recall likelihood. Our actual coefficient of 0.194 (from column 2 of Table 2) is represented by the dashed vertical line in Figure 4. It exceeds all 1,000 permuted coefficients, placing it at approximately the 99.9th percentile of the distribution. The contrast between our observed effect and this null distribution provides evidence that the relationship between EO and recall likelihood is not an artifact of our sampling approach or industry-level confounds, but instead represents a firm-level association.⁴

Figure 4.

Null distribution of EO coefficients on product recall likelihood (1,000 simulations).

Endogeneity also remains an important consideration for establishing causal inference. When estimating our model, EO may be endogenous for several reasons. First, there may be unobserved factors associated with EO that affect the likelihood of product recall, such as the firm’s other management practices or technological advancements within the industry, which evolve over time and are not easily observed. Ignoring these variables could result in an omitted variable problem. Second, measurement error may arise when constructing the EO measure. Specifically, the words or phrases captured in the 10-K-report may not accurately reflect the firm’s EO, affecting the reliability of our findings. In addition, there may be reverse causality concerns, despite the fact we have lagged all the independent variables in our model. Estimating Equation (1) when EO is endogenous can therefore lead to inconsistent estimates.

Following recent recommendations for strengthening causal inference in management research (Frake et al., 2025), we first employ the impact threshold for a confounding variable (ITCV) test (Busenbark et al., 2022; Frank et al., 2013) to assess how sensitive our results are to potential omitted variables. In our analysis, we calculated an ITCV value of 0.078 (α = .10), which indicates the strength of the correlation that an omitted variable would need to have with both our independent and dependent variables to invalidate our findings. Following Busenbark et al. (2022), we also evaluate the robustness of this value by comparing it to the square root of the product of partial correlations between our independent variable, dependent variable, and each control variable. Results show that all control variables fell well below 59% of the ITCV. This suggests that our findings are unlikely to be substantially biased by omitted variables.

As an additional precaution, we also employ a system Generalized Method of Moments (GMMs) estimation approach to address potential dynamic endogeneity concerns. Dynamic endogeneity occurs when today’s independent variables are influenced by yesterday’s dependent variables (Li et al., 2021), a situation particularly relevant in our context where firms may strategically adjust their TMT composition (hiring COOs). Additionally, unobserved firm characteristics that influence recall likelihood may drive both EO and TMT structural decisions simultaneously, creating potential simultaneity bias that could affect our results. A GMM estimation addresses these concerns by simultaneously handling multiple endogenous variables while controlling for unobserved heterogeneity, using lagged levels and differences of the endogenous variables as instruments to exploit the panel structure of our data. In our implementation, we treat EO, COO power, and life cycle, as well as their two- and three-way interaction terms, as potentially endogenous variables.

Table 4 presents the system GMM estimation results. Column 1 includes our baseline model testing H1, column 2 incorporates the interaction term to test H2, and column 3 incorporates the three-way interaction term and additional interactions to test H3. In support of H1, we find that the coefficient of EO remains positively and significantly associated with product recall likelihood in both the baseline model (β = .027, p < .05) and the full specification (β = .042, p < .01). Similarly, the coefficient of the interaction term between EO and COO power is negative and significant (β = −.111, p < .05). In support of H3, the three-way interaction among EO, COO power, and life cycle is negative and statistically significant (β =−.024, p < .01), indicating that the attenuating effect of COO power on the EO–product recall relationship becomes stronger as firms move into late stages of the life cycle.

Table 4.

GMM Regression Results.

	(1)	(2)	(3)
	System GMM	System GMM	System GMM
DV:	Product recall	Product recall	Product recall
EO	0.027** (0.013)	0.042*** (0.011)	0.084*** (0.010)
COO power	0.028 (0.027)	0.029 (0.032)	0.038 (0.032)
Life cycle			−0.275*** (0.051)
EO × COO power		−0.111** (0.050)	−0.128*** (0.017)
EO × Life cycle			−0.138*** (0.025)
COO power × Life cycle			−0.011 (0.034)
EO × COO power × Life cycle			−0.024*** (0.005)
Firm age	0.000 (0.000)	0.001** (0.000)	0.001* (0.001)
Firm size	0.004* (0.002)	0.004 (0.003)	−0.000 (0.005)
Leverage	0.004 (0.009)	0.011 (0.015)	0.017 (0.019)
ROA	−0.020 (0.021)	−0.026 (0.021)	−0.054*** (0.020)
Market-to-book	0.001* (0.001)	0.001 (0.001)	0.001* (0.001)
Revenue	0.162** (0.080)	0.225*** (0.061)	0.313*** (0.065)
R&D intensity	−0.000 (0.000)	−0.000 (0.000)	−0.001 (0.000)
Labor intensity	−0.001 (0.000)	−0.001 (0.001)	−0.002** (0.001)
Munificence	0.006 (0.031)	0.016 (0.031)	−0.022 (0.059)
Dynamism	−0.017 (0.121)	0.142 (0.147)	0.216 (0.225)
HHI	−0.027 (0.026)	−0.041 (0.025)	0.003 (0.066)
TMT age	−0.000 (0.001)	0.000 (0.001)	−0.001 (0.001)
TMT female	−0.012 (0.017)	−0.030* (0.018)	−0.035* (0.021)
TMT stock option pay	0.001 (0.001)	0.001 (0.001)	−0.001 (0.001)
Constant	0.035 (0.127)	−0.097 (0.156)	0.266 (0.411)
Arellano–Bond test for AR(1)	−4.12***	−7.51***	−7.91***
Arellano–Bond test for AR(2)	1.510	1.530	−0.020
Hansen test χ² (p-value)	0.639	0.635	0.557
Number of instruments	205	222	293
Observations	7533	7533	7533

Note. All models are estimated using robust two-step Arellano–Bond system GMM dynamic panel estimation. Each model includes a lagged dependent variable. All models include both year and industry fixed effects. The Hansen test assesses the validity of the instruments under the null hypothesis of joint validity of all instruments.

COO: chief operating officer; EO: entrepreneurial orientation; HHI: Herfindahl–Hirschman index; ROA: return on assets; TMT: top management team.

p < .1. **p < .05. ***p < .01.

We conducted several diagnostic tests to validate our GMM estimation approach. The Arellano–Bond test for first-order serial correlation yields significant results (AR(1) = −4.12, p < .01 for the baseline model; AR(1) = −7.51, p < .01 for the model including two-way interaction; and AR(1) = −7.91, p < .01 for the full model with three-way interaction), which is expected and indicates that the first-differenced errors are serially correlated. Critically, the Arellano–Bond test for second-order serial correlation, AR(2), shows no evidence of remaining correlation confirming that our moment conditions are valid. The Hansen test for instrument validity fails to reject the null hypothesis of joint instrument validity in both specifications (χ²p-values of .639 for the baseline model, 0.635 for the model with two-way interaction, and 0.557 for the model with three-way interaction), providing additional evidence that our instruments are appropriate.

Testing of Quality Mechanism

Given our theorizing suggests that EO dampens a firm’s focus on quality and operational effectiveness, we empirically examine this mechanism. Specifically, we investigate whether EO influences firms’ adoption of formal quality management systems (QMSs). Given that a firm’s underlying approach to quality control is not directly observable, we use QMS certifications as a measurable indicator of formal quality control commitment. Using the LSEG (formerly Refinitiv) database, we construct a binary variable, QMS Adoption, indicating whether a firm has obtained certification in established quality management frameworks relating to Six Sigma, Lean Manufacturing, TQM, or ISO 9000.⁵ As shown in Table 5, our analysis reveals a negative and significant relationship between EO and QMS adoption (β = −.01, p < .05), suggesting that entrepreneurially oriented firms are less likely to implement formal QMS. This finding aligns with our general argument that with greater innovation, proactiveness, and risk-taking, EO may lead firms to operate with less structured quality control processes.

Table 5.

Mechanism Analysis.

Variable	(1)
	DV: QMS adoption
EO	−0.010** (0.005)
Firm age	0.002*** (0.000)
Firm size	0.085*** (0.005)
Leverage	−0.123*** (0.028)
ROA	0.151*** (0.054)
Market-to-book	0.000 (0.001)
Revenue	−0.213*** (0.036)
R&D intensity	−0.001 (0.001)
Labor intensity	0.005*** (0.002)
Munificence	0.031 (0.090)
Dynamism	−0.758* (0.425)
HHI	−0.011 (0.067)
TMT age	−0.003*** (0.001)
TMT female	−0.063* (0.037)
TMT stock option pay	−0.004** (0.001)
Observations	4983
Adjusted R²	0.398

Note. The dependent variable (DV) is QMS Adoption that equals 1 if a firm implements QMSs or similar quality principles and otherwise zero. The model includes year and industry fixed effects with all independent variables being lagged by 1 year. Robust standard errors are reported in parentheses.

COO: chief operating officer; EO: entrepreneurial orientation; HHI: Herfindahl–Hirschman index; ROA: return on assets; TMT: top management team.

p < .1. **p < .05. ***p < .01.

Robustness Checks and Additional Analysis

To provide further robustness to our results, we also performed a battery of additional tests that we present in Supplemental Appendix. These include using: (1) a linear probability model (LPM) as an alternative estimation approach (columns 1 and 2, Supplemental Appendix Table A7); (2) alternative measures of recall risk (columns 3 to 6, Supplemental Appendix Table A7); (3) an alternative measure of EO that uses the three original dimensions instead of five (column 7, Supplemental Appendix Table A7); (4) a model that includes CEO characteristics rather than TMT characteristics as controls (column 8, Supplemental Appendix Table A7); (5) the value of the recall as an alternate independent variable (Supplemental Appendix Table A8); and (6) performing analyses that examine the impact of substantial shifts in EO (Supplemental Appendix Table A9). The results support our main analyses.

Discussion

Our findings shed light on a potential downside of EO as an experimentation logic, observing both direct and contingent harmful implications for firms. Directly, we show that the variance-enhancing nature of EO that is critical for discovering and exploiting new opportunities (Wales et al., 2023) also elevates firms’ exposure to product recalls. Conditionally, this downside is shaped by executive governance and the firm’s product life cycle. COO power weakens the EO–recall relationship by embedding reliability and discipline into the experimentation process, and this buffering effect is strongest when firms’ innovation portfolios are concentrated in late stages, where oversight can be focused and quality issues more predictable. Next, we discuss the theoretical implications of our findings.

Theoretical Implications and Future Research Directions

Our study contributes to the EO-as-experimentation perspective in two main ways. First, we develop theory that identifies a novel mechanism connecting EO’s experimentation logic to a concrete downside outcome. While most research has emphasized EO’s positive influence on innovation, growth, and performance (Atuahene-Gima & Ko, 2001; Rauch et al., 2009), only recently have scholars begun to theorize its downsides. For example, Gali et al. (2024) demonstrated that EO can increase the likelihood of firm failure by sustaining costly exploratory behaviors, echoing arguments from organizational learning theory about the “failure trap” (Levinthal & March, 1993; March, 1991). Yet, this work remains largely aggregate and distal.

We extend this emerging line of inquiry by theorizing a new quality-based mechanism that explicitly links EO’s experimentation logic to a specific operational failure—namely product recalls. Drawing on the variance-enhancing logic central to the EO-as-experimentation perspective (Wiklund & Shepherd, 2011) and the more recent EO-as-new-value-creation theory (Wales et al., 2023), our study suggests that EO’s emphasis on speed, novelty, and bold market entry systematically reallocates organizational attention and resources away from quality assurance and process reliability (as indicated by a reduced focus on QMS adoption). This bias represents an inherent trade-off within entrepreneurial action: high-EO firms privilege discovery over discipline, and early market feedback over internal validation. We do not believe this prioritization of novelty over discipline is willful neglect of quality control or standards, rather EO’s emphasis on frequent and bold market entry, and the velocity at which such actions occur (Wales et al., 2023), naturally “run faster” than the firm’s ability to maintain high-quality standards and controls. The result is an endogenous vulnerability produced when experimentation demands exceed the firm’s operational bandwidth.

By theorizing and empirically showing this quality mechanism, we extend EO theory beyond abstract notions of performance variance to specify how variance materializes through a specific behavioral and organizational pathway. Product recalls capture a lower tail expression of EO’s variance logic—failures that emerge during experimentation when rapid, EO-induced change strains managerial attention and feedback cycles, allowing product defects and quality issues to escape detection. This conceptualization translates variance logic from outcome dispersion to process dynamics (i.e., how EO is enacted), answering calls in the literature for EO research to go beyond the status quo (Wales et al., 2021).

In doing so, our study introduces the idea of process variance as a complement to outcome variance. We demonstrate that the process through which EO is enacted internally (by managers and in what context) also alters what results firms achieve and how they are produced. This theoretical extension yields a more complete understanding of EO as a double-edged strategic posture—one that amplifies both learning and liability because the very mechanisms that accelerate experimentation can also compromise quality controls.

Second, and related, our study offers more nuanced theorizing about when and why being entrepreneurially oriented is more or less likely to pose a recall risk. By embedding UET logic into EO research, we extend recent calls in the literature to examine how top managers shape how EO is enacted inside organizations (Wales et al., 2020, 2021). Although the role of the TMT has long been acknowledged in the EO literature (e.g., Covin & Slevin, 1989; Wiklund, 1999), UET’s central premise—that strategic outcomes reflect the bounded rationality, cognitive frames, and influence of top executives—has rarely been explicitly integrated into theorizing about EO’s consequences (Wales et al., 2020). Thus, our study extends EO theory by introducing COO power as an important attentional governance contingency that explains when and why EO’s experimentation logic unfolds either as costly failure or disciplined learning.

When COO power is high, the quality-oriented logic of the COO gains salience within the TMT, channeling EO toward more disciplined experimentation (but does not reverse recall risk). When COO power is low, quality concerns receive less attention, leaving experimentation to proceed unchecked and increasing the likelihood that defects escape early detection. Thus, COO power weakens the positive EO–recall relationship by stabilizing the experimentation process. We further show that the magnitude of this mitigating effect depends on the firm’s innovation portfolio. Firms concentrated in early-stage product development face greater design uncertainty, faster iteration cycles, and more volatile feedback loops—conditions in which bounded attention is particularly strained and operational oversight more crucial. In such environments, COO power can meaningfully reduce (but not eliminate) the risks associated with EO’s experimentation logic. By contrast, in late-stage portfolios where technologies are more stable and quality issues more predictable, COO have greater opportunity to oversee risky, early-stage initiatives. Interestingly, under these marginal conditions—where COO power is high (greater than mean-levels) and firms’ innovation activities concentrated in late stages—EO leads firms to become more proactive in identifying product quality concerns before they manifest in recalls. Thus, by integrating UET into EO theory, our study positions EO as a managerially configured process shaped by attention, influence, and executive governance. This perspective moves the field beyond generic claims about “more EO” toward a richer understanding of how EO is enacted and why its outcomes diverge across firms.

Our study opens several promising directions for future research, summarized in Table 6. First, scholars can investigate other EO-related failures. While we examined product recalls, EO is likely to produce other downside outcomes, such as ethical lapses, regulatory violations, and safety incidents. Identifying such failures will advance our understanding of the range of possible downside outcomes that EO may plausibly influence and how. Second, future work could examine the organizational controls and governance mechanisms that mitigate EO’s downside potential. Our findings suggest that EO reduces firms’ focus on quality control, but COO power mitigates this influence. Future work can explore other controls and governance levers that may shape the distribution of outcomes (Wiklund & Shepherd, 2011), such as board characteristics and composition. Third, the relationship between EO and new value creation merits further attention. EO-as-new-value-creation theory highlights that bold market entry increases outcome variance, yet little is known about how firms can preserve upside potential while minimizing downside spillovers. Fourth, because EO is enacted by top managers, future studies can examine other TMT roles, functional backgrounds, and power distributions (beyond our focus on COOs). Finally, future research could explore further nuances in how portfolio composition, technological maturity, and product life cycle dynamics interact with EO’s experimentation logic. For instance, scholars could examine how firms’ breadth and depth of search across product-market domains impact downside outcomes.

Table 6.

Future Research Directions.

Theme	Opportunity for future research	Indicative research questions
EO-related failures	Extend the downside logic to other specific lower tail outcomes	• What other types of failures (ethical lapses, safety incidents, customer-harm events) arise from EO’s variance logic?• Do different EO dimensions or combinations of dimensions (risk-taking, innovativeness, proactiveness) generate distinct failure profiles?
Controls that shift outcome distributions	Identify organizational controls that reconfigure experimentation outcomes	• Which internal controls (quality systems, risk routines, digital monitoring) most effectively constrain EO’s lower tail risks?• How do board governance, TMT composition, and incentive systems alter EO’s failure likelihood?
EO and value creation mechanisms	Understand how experimentation creates—and sometimes destroys—value	• How can firms preserve EO’s upside while reducing lower tail risks?• When does experimentation generate learning versus costly variance?
Executive governance contingencies	Expand beyond COO power to other executive roles	• How do CFOs, CTOs, CMOs, or chief risk officers shape EO’s expression and consequences?• How do distributed TMT attention systems influence experimentation quality?
Innovation portfolio, cycle dynamics, and organizational capacity	Explore when experimentation outpaces managerial and organizational capacity	• Under what innovation portfolio configurations (e.g., breadth, depth, stage mix) does EO overwhelm quality systems?• How do life cycle stages, technological novelty, or uncertainty domains interact with EO to produce failure?

CFO: cheif financial officer; CMO: chief marketing officer; COO: chief operating officer; CTO: chief technology officer; EO: entrepreneurial orientation; TMT: top management team.

Practical Implications and Limitations

Our findings also have important implications for managers. Our study highlights the need for managers to recognize the recall risks that accompany highly entrepreneurial strategic postures. Given the substantial costs of recalls to the firm, failing to manage this risk can erode the very value EO seeks to create. Our findings suggest that strengthening quality oversight is critical to mitigating—and potentially reversing—this risk. In particular, powerful COOs can help counterbalance the liabilities of high-EO by embedding operational excellence and quality assurance into decision-making. This balance is especially vital for firms whose innovation activities are concentrated in the early product life cycle stages, where the risk of recall is higher.

Our study is not without limitations. First, we rely on publicly available archival data to assess both product recalls and EO. Future research could complement this approach with surveys or interview data that capture executives’ perceptions. Second, while we theorize and identify an important quality mechanism linking EO to product recalls, our measures are proxy-based. Richer data—such as detailed case studies or access to firm’s quality control records—could offer deeper insight into how high-EO firms structure and monitor quality assurance. Third, we focus on one outcome—product recalls. Future studies could examine whether similar mechanisms explain EO’s influence on other manifestations of operational failure, such as warranty claims, safety violations, or customer complaints, to expand the generality of our findings.

Despite these limitations, our study advances our understanding of how EO can translate into a concrete operational failure and how executive governance conditions this process. We hope our study stimulates future research to further explore EO’s downsides.

Supplemental Material

sj-pdf-1-etp-10.1177_10422587261419465 – Supplemental material for The Double-Edged Sword of Entrepreneurial Orientation: Product Recalls and the Role of COO Power

Supplemental material, sj-pdf-1-etp-10.1177_10422587261419465 for The Double-Edged Sword of Entrepreneurial Orientation: Product Recalls and the Role of COO Power by Pide Lun, Ralf Zurbruegg, Matthew P. Mount and Chee Seng Cheong in Entrepreneurship Theory and Practice

Footnotes

Authors’ Note

The first three authors contributed equally to the development of this article. The final author played a role in curating some of the data used.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Matthew P. Mount

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Pide Lun is a Lecturer in Business Analytics at Adelaide University, Australia. His research focuses on entrepreneurial orientation, corporate governance, and firm performance.

Ralf Zurbruegg is Professor of Finance and Business Analytics at Adelaide University, Australia. His research interest is in sustainability - broadly defined - and his work has appeared in several top tier outlets across finance, marketing, and management.

Matthew P. Mount is Professor of Strategy at Deakin Business School, Australia. His work centres on understanding how organizations and managers make strategic decisions and has been published in several of the field’s top journals.

Chee Seng Cheong is Associate Professor of Finance at Adelaide University. His primary research interests are in financial analysts’ forecast accuracy, real estate investment and accounting quality.

References

Anderson

B. S.

Kreiser

P. M.

Kuratko

D. F.

Hornsby

J. S.

Eshima

(2015). Reconceptualizing entrepreneurial orientation. Strategic Management Journal, 36(10), 1579–1596.

Atuahene-Gima

(2001). An empirical investigation of the effect of market orientation and entrepreneurship orientation alignment on product innovation. Organization Science, 12(1), 54–74.

Ball

G. P.

Shah

Wowak

K. D.

(2018). Product competition, managerial discretion, and manufacturing recalls in the US pharmaceutical industry. Journal of Operations Management, 58(1), 59–72.

Bendig

(2022). Chief operating officer characteristics and how they relate to exploration via patenting versus venturing. Journal of Business Research, 140, 297–309.

Busenbark

J. R.

Yoon

Gamache

D. L.

Withers

M. C.

(2022). Omitted variable bias: Examining management research with the impact threshold of a confounding variable (ITCV). Journal of Management, 48(1), 17–48.

Carlson

N. A.

Burbano

(2025). The use of LLMs to annotate data in management research. Foundational guidelines and warnings. Strategic Management Journal. Advance online publication. https://doi.org/10.1002/smj.70023

Chen

Ganesan

Liu

(2009). Does a firm’s product-recall strategy affect its financial value? An examination of strategic alternatives during product-harm crises. Journal of Marketing, 73(6), 214–226.

Chen

Miller

(1994). Competitive attack, retaliation and performance: An expectancy-valence framework. Strategic Management Journal, 15(2), 85–102.

Cleeren

Dekimpe

M. G.

van Heerde

H. J.

(2017). Marketing research on product-harm crises: A review, managerial implications, and an agenda for future research. Journal of the Academy of Marketing Science, 45(5), 593–615.

10.

Covin

J. G.

Slevin

D. P.

(1989). Strategic management of small firms in hostile and benign environments. Strategic Management Journal, 10(1), 75–87.

11.

Covin

J. G.

Wales

W. J.

(2019). Crafting high-impact entrepreneurial orientation research: Some suggested guidelines. Entrepreneurship Theory and Practice, 43(1), 3–18.

12.

CPSC. (2024). United States Consumer Product Safety Commission 2024 Annual Report. https://www.cpsc.gov/s3fs-public/CPSC-FY24-Annual-Report_0.pdf

13.

Dawar

Pillutla

M. M.

(2000). Impact of product-harm crises on brand equity: The moderating role of consumer expectations. Journal of Marketing Research, 37(2), 215–226.

14.

de Kok

. (2025). ChatGPT for textual analysis? How to use generative LLMs in accounting research. Management Science, 71(9), 7888–7906.

15.

Denrell

(2003). Vicarious learning, undersampling of failure, and the myths of management. Organization Science, 14(3), 227–243.

16.

Eilert

Jayachandran

Kalaignanam

Swartz

T. A.

(2017). Does it pay to recall your product early? An empirical investigation in the automobile industry. Journal of Marketing, 81(3), 111–126.

17.

Finkelstein

(1992). Power in top management teams: Dimensions, measurement, and validation. Academy of Management Journal, 35(3), 505–538.

18.

Finkelstein

Hambrick

D. C.

Cannella

A. A.

(2009). Strategic leadership: Theory and research on executives, top management teams, and boards. Oxford University Press.

19.

Frake

Gibbs

Goldfarb

Hiraiwa

Starr

Yamaguchi

(2025). From perfect to practical: Partial identification methods for causal inference in strategic management research. Strategic Management Journal, 46(8), 1894–1929.

20.

Frank

K. A.

Maroulis

S. J.

Duong

M. Q.

Kelcey

B. M.

(2013). What would it take to change an inference? Using Rubin’s causal model to interpret the robustness of causal inferences. Educational Evaluation and Policy Analysis, 35(4), 437–460.

21.

Gali

Hughes

Wang

C. L.

(2024). Entrepreneurial entropy: A resource exhaustion theory of firm failure from entrepreneurial orientation. Entrepreneurship Theory and Practice, 48(1), 141–170.

22.

Hambrick

D. C.

(2007). Upper echelons theory: An update. Academy of Management Review, 32(2), 334–343.

23.

Hambrick

D. C.

Cannella

A. A.

(2004). CEOs who have COOs: Contingency analysis of an unexplored structural form. Strategic Management Journal, 25(10), 959–979.

24.

Hambrick

D. C.

Mason

P. A.

(1984). Upper echelons: The organization as a reflection of its top managers. Academy of Management Review, 9(2), 193–206.

25.

Haunschild

P. R.

Sullivan

B. N.

(2002). Learning from complexity: Effects of prior accidents and incidents on airlines learning. Administrative Science Quarterly, 47(4), 609–643.

26.

Hersel

M. C.

Helmuth

C. A.

Zorn

M. L.

Shropshire

Ridge

J. W.

(2019). The corrective actions organizations pursue following misconduct: A review and research agenda. Academy of Management Annals, 13(2), 547–585.

27.

Hoberg

Maksimovic

(2022). Product life cycles in corporate finance. Review of Financial Studies, 35(9), 4249–4299.

28.

Hoffmann

Cheong

C. S.

Phan

Zurbruegg

(2024). So, sue me…if you can! How legal changes diminishing managers’ risk of being held liable by shareholders affect firms’ likelihood to recall products. Journal of Marketing, 88(5), 89–110.

29.

Hora

Bapuji

Roth

A. V.

(2011). Safety hazard and time to recall: The role of recall strategy, product defect type, and supply chain player in the U.S. toy industry. Journal of Operations Management, 29(7–8), 766–777.

30.

Joseph

Ocasio

McDonnell

M-H.

(2014). The structural elaboration of board independence: Executive power, institutional logics, and the adoption of CEO-only board structures in U.S. corporate governance. Academy of Management Journal, 57(6), 1834–1858.

31.

Kim

Koenigsberg

Ofek

(2022). I don’t “recall”: The decision to delay innovation launch to avoid costly product failure. Management Science, 68(12), 8889–8908.

32.

Kini

Shenoy

Subramaniam

(2017). Impact of financial leverage on the incidence and severity of product failures: Evidence from product recalls. Review of Financial Studies, 30(5), 1790–1829.

33.

Kitazawa

(2012). Hyperbolic transformation and average elasticity in the framework of the fixed effects logit model. Theoretical Economics Letters, 2(2), 192–199.

34.

Krause

Semadeni

Cannella

A. A.

(2013). External COO/presidents as expert directors: A new look at the service role of boards. Strategic Management Journal, 34(13), 1628–1641.

35.

Krause

Priem

Love

(2015). Who’s in charge here? Co-CEOs, power gaps, and firm performance. Strategic Management Journal, 36(13), 2099–2110.

36.

Lavie

Stettner

Tushman

M. L.

(2010). Exploration and exploitation within and across organizations. Academy of Management Annals, 4(1), 109–155.

37.

Levinthal

D. A.

March

J. G

. (1993). The myopia of learning. Strategic Management Journal, 14(S2), 95–112.

38.

Ding

Wan

(2021). Dealing with dynamic endogeneity in international business research. Journal of International Business Studies, 52, 339–362.

39.

Liu

Shankar

Yun

(2017). Crisis management strategies and the long-term effects of product recalls on firm value. Journal of Marketing, 81(5), 30–48.

40.

Lumpkin

G. T.

Dess

G. G.

(1996). Clarifying the entrepreneurial orientation construct and linking it to performance. Academy of Management Review, 21(1), 135–172.

41.

Marcel

J. J.

(2009). Why top management team characteristics matter when employing a chief operating officer: A strategic contingency perspective. Strategic Management Journal, 30(6), 647–658.

42.

March

J. G.

(1991). Exploration and exploitation in organizational learning. Organization Science, 2(1), 71–87.

43.

McKenny

A. F.

Aguinis

Short

J. C.

Anglin

A. H.

(2018). What doesn’t get measured does exist: Improving the accuracy of computer-aided text analysis. Journal of Management, 44(7), 2909–2933.

44.

Menz

(2012). Functional top management team members: A review, synthesis, and research agenda. Journal of Management, 38(1), 45–80.

45.

Miller

(1983). The correlates of entrepreneurship in three types of firms. Management Science, 29(7), 770–791.

46.

Nelson

R. R.

Winter

S. G.

(1982). An evolutionary theory of economic change. Harvard University Press.

47.

Ocasio

(1997). Towards an attention-based view of the firm. Strategic Management Journal, 18(SI), 187–206.

48.

Ozgen

Mooney

Zhou

(2025). CEO power: A review, critique, and future research directions. Journal of Management, 51(1), 132–171.

49.

Patel

P. C.

Kojtamäki

Parida

Wincent

(2015). Entrepreneurial orientation-as-experimentation and firm performance: The enabling role of absorptive capacity. Strategic Management Journal, 36(11), 1739–1749.

50.

Raisch

Birkinshaw

(2008). Organizational ambidexterity: Antecedents, outcomes, and moderators. Journal of Management, 34(3), 375–409.

51.

Rauch

Wiklund

Lumpkin

G. T.

Frese

(2009). Entrepreneurial orientation and business performance: An assessment of past research and suggestions for the future. Entrepreneurship Theory and Practice, 33(3), 761–787.

52.

Reisinger

(2016). Here’s how bad the Galaxy Note 7 fallout is for Samsung. Fortune. https://fortune.com/2016/10/11/samsung-galaxy-note-7-fallout/.

53.

Rosenbaum

P. R.

(2002). Observational Studies (2nd ed.). Springer.

54.

Sahoo

Yadav

(2017). Entrepreneurial orientation of SMEs, total quality management and firm performance. Journal of Manufacturing Technology Management, 28(7), 892–912.

55.

Singh

Fleming

(2010). Lone inventors as sources of breakthroughs: Myth or reality? Management Science, 56(1), 41–56.

56.

Short

J. C.

Broberg

J. C.

Cogliser

C. C.

Brigham

K. H.

(2010). Construct validation using computer-aided text analysis (CATA): An illustration using entrepreneurial orientation. Organizational Research Methods, 13(2), 320–347.

57.

Thirumalai

Sinha

K. K.

(2011). Product recalls in the medical device industry: An empirical exploration of the sources and financial consequences. Management Science, 57(2), 376–392.

58.

Titus

Jr Mbaraonye

Abdurakhmonov

Parker

(2023). Avoiding the ties that bind: Entrepreneurial orientation and the creative constraint of corporate political activity. Journal of Management, 50(7), 1–32.

59.

Utterback

J. M.

Abernathy

W. J.

(1975). A dynamic model of process and product innovation. Omega, 3(6), 639–656.

60.

van Heerde

Helsen

Dekimpe

M. G

. (2007). The impact of a product-harm crisis on marketing effectiveness. Marketing Science, 26(2), 230–245.

61.

Wales

W. J.

Covin

J. G.

Monsen

(2020). Entrepreneurial orientation: The necessity of a multilevel conceptualization. Strategic Entrepreneurship Journal, 14(4), 639–660.

62.

Wales

W. J.

Kraus

Filser

Stöckmann

Covin

J. G.

(2021). The status quo of research on entrepreneurial orientation: Conversational landmarks and theoretical scaffolding. Journal of Business Research, 128, 564–577.

63.

Wales

W. J.

Covin

J. G.

Schüler

Baum

(2023). Entrepreneurial orientation as a theory of new value creation. Journal of Technology Transfer, 48, 1752–1772.

64.

Wales

W. J.

Gupta

V. K.

Mousa

(2013). Empirical research on entrepreneurial orientation: An assessment and suggestions for future research. International Small Business Journal: Researching Entrepreneurship, 31(4), 357–383.

65.

Wales

W. J.

Patel

P. C.

Lumpkin

G. T.

(2013). In pursuit of greatness: CEO narcissism, entrepreneurial orientation, and firm performance variance. Journal of Management Studies, 50(6), 1041–1069.

66.

Wiklund

(1999). The sustainability of the entrepreneurial orientation—Performance relationship. Entrepreneurship Theory and Practice, 24(1), 37–48.

67.

Wiklund

Shepherd

(2011). Where to from here? EO as experimentation, failure, and distribution of outcomes. Entrepreneurship Theory and Practice, 35(5), 925–946.

68.

Wowak

K. D.

Boone

C. A.

(2015). So many recalls, so little research: A review of the literature and road map for future research. Journal of Supply Chain Management, 51(4), 54–72.