Re-examining Strategic Flexibility: A Meta-Analysis of its Antecedents, Consequences and Contingencies

Dimensionality of strategic flexibility

The concept of SF has been addressed by scholars across the strategy, management, marketing, innovation, entrepreneurship and operations disciplines. This diversity, combined with a proliferation in both the theoretical approaches and the empirical methods employed, has resulted in a ‘conceptual schizophrenia’ characterizing the SF literature (Bahrami and Evans, 2011: 37), as evident in the lack of cumulative theory development, multiple theoretical tensions and the need to pursue further attempts to consolidate the field. In reviewing the literature, we find that SF has been conceptualized in various ways, as displayed in Table 1. By summarizing the common ground among the definitions, one can attempt to extrapolate SF's dimensions and offer implications regarding the means by which it is performed. A firm can exhibit SF by being reactive (responding to change) and/or proactive (creating new opportunities) in terms of the variety of available strategic options and/or the speed (timely response) of pursuing a strategic option; it can also react internally through resource deployment and/or externally via competitive actions.

Taken together, and similar to more recent conceptual overviews (Brozovic, 2018; Combe, 2012), we thus consider SF a multidimensional construct.2 The dimensionality of SF has been emphasized previously (see e.g. Bahrami and Evans, 2011; Brozovic, 2018; Evans, 1991), together with its antecedents (i.e. enablers, inhibitors and triggers), as has its consequences and contingencies. We emphasize the inherent theoretical tensions present within every aspect when describing the theoretical background of our study.

Antecedents of strategic flexibility

The antecedents of SF encompass enablers, inhibitors and triggers of SF, as identified by Brozovic (2018). Previous studies stemming from different disciplines have identified a plethora of antecedents affecting SF and a number of tensions related to these antecedents. Based on existing empirical studies, Table 2 summarizes the tensions related to various antecedents and provides the main arguments for their expected influence on SF.3 These studies have distinguished between two categories of enablers: the strategic orientations of a firm and its organizational design. In terms of a firm's strategic orientation, the literature frequently relates entrepreneurial orientation (Nadkarni and Narayanan, 2007), market orientation (Johnson et al., 2003) and learning orientation (Santos-Vijande, Sanchez and Trespalacios, 2012) to SF. Moreover, in terms of organizational design, decentralization (Volberda, 1998), slack resources (Barker and Barr, 2002) and formal routines (Shimizu and Hitt, 2004) have been related to SF. However, there has been disagreement regarding these enablers of SF. For example, while formal routines have been posited to result in more rapid responses to environmental changes (Worren, Moore and Cardona, 2002), they have also been identified as a means for advancing efficiency rather than flexibility (Schreyögg and Sydow, 2010).

Compared to the enablers, the inhibitors of SF are largely under-researched within the literature (Brozovic, 2018). We identify firm size, firm age and past success as potential inhibitors. However, in terms of firm size, some scholars ascribe SF as a characteristic of large firms because of the potential availability of slack resources (Pauwels and Matthyssens, 2004). In contrast, SF can be exhibited by smaller firms specifically because their resource adaptability is higher (Ebben and Johnson, 2005). Additionally, firm age (Barker and Barr, 2002) and past success (Lant, Milliken and Batra, 1992) have also been claimed to encourage SF.

Finally, the literature asserts that changes in the business environment may not only affect the SF–performance relationship, but also trigger SF. Environmental dynamism (Perez-Valls et al., 2016), demand uncertainty (Aaker and Mascarenhas, 1984) and competitive intensity (Das and Elango, 1995) may all act as external triggers of a firm's SF. Nevertheless, other scholars question whether the business environment has these triggering effects on SF. For example, Shimizu and Hitt (2004) caution that organizations tend to maintain the status quo in high environmental uncertainty, thereby disapproving of SF. Indeed, resistance to change as a reaction to pursuing SF in a firm is a practically implicit challenge that must be overcome to instil SF throughout the organizational hierarchy (Skordoulis, 2004).

To summarize, the extant literature offers a plethora of potential enablers, inhibitors and triggers of SF, but there are disagreements and controversies regarding how much (if at all) some of these factors actually affect SF. Another significant tension encompassing SF relates to the differential effects of SF, each of which are aligned with different theoretical paradigms.

Consequences of strategic flexibility

The theoretical tensions surrounding the consequences of SF are summarized in Table 3. When balancing the competing theoretical positions and considering the typical stance of the managerial literature, most researchers argue that SF is beneficial for firms (Brozovic, 2018). When theorizing on its positive performance effect, some scholars espouse SF as a dynamic capability because it emphasizes the flexible use of resources and the reconfiguration of processes (Johnson et al., 2003). Others consider SF to be a means of increasing control by maintaining a portfolio of strategic options (Dalziel, 2009). Adopting an organizational design perspective, other scholars have emphasized the benefits that arise from the degree to which an organization maintains a repertoire of managerial capabilities and the speed at which they can be activated (Wang, Senaratne and Rafiq, 2015).

However, in contrast to these viewpoints, other researchers argue that SF does not improve desired outcomes or may even reduce a firm's performance. For example, the flexibility–efficiency trade-off, which describes the distinction between efficient and flexible strategies (Ebben and Johnson, 2005), illustrates a long-standing dilemma in organization theory. Thompson (1967) calls this the ‘paradox of administration’ (p. 15), whereby managers need to reconcile, for example, the organization of routine tasks with tasks that are non-routine. Efficiencies in delivering strategy require the firm to maximize its output from existing resources, while flexibility requires it to reconfigure resources (Ghemawat and Ricart i Costa, 1993). Pursuing efficiency will therefore instil replicable routines that fundamentally oppose flexibility (Weigelt and Sarkar, 2009).

The suggestion here is not necessarily that efficiency is superior to flexibility, but that mixing both strategies can lead to underperformance (Ebben and Johnson, 2005). Firms may make a trade-off between flexibility and efficiency, but find that they achieve only mediocre performance as a result (Eisenhardt, Furr and Bingham, 2010). Moreover, firms may vacillate between strategic options, thereby wasting resources and risking failure (Shimizu and Hitt, 2004).

To conclude, the SF–performance relationship has been described as complex, and recent studies called for more research illuminating these tensions (Claussen, Essling and Peukert, 2018). Moreover, despite the multiplicity of different contingencies in previous studies, there is limited empirical evidence of the relative extent to which various distinct facets of the environment determine the performance effects of SF (Brozovic, 2018).

Contingencies of strategic flexibility

We have previously emphasized the tensions surrounding environmental dynamism, demand uncertainty and competitive intensity as triggers of SF. However, the extant literature has also regarded these three variables as contingencies in the SF–performance relationship, adding to the ambiguity surrounding these variables. Treating them as moderators of the strength of the relationship between SF and performance (Brozovic, 2018; Combe, 2012), the consolidating literature on SF has recommended even more research about them, despite some valuable insights (e.g. Grewal and Tansuhaj, 2001; Nadkarni and Narayanan, 2007).

Thus, SF is likely to become more important when an organization is faced with a high rate of change in the environment (Nadkarni and Narayanan, 2007). Similarly, high levels of environmental dynamism are suggested to favour flexibility (Johnson et al., 2003). As for demand uncertainty, Harrigan (1985) finds that SF is a means to manage demand volatilities. More specifically, demand uncertainty creates difficulty in devising strategic plans; SF should be more useful in these uncertain markets (Claussen, Essling and Peukert, 2018; Lee and Makhija, 2009). In other words, demand uncertainty moderates the relationship between SF and performance (Grewal and Tansuhaj, 2001). The same understanding is advanced for competitive intensity, whereby it is recommended to develop SF in order for firms to adapt to highly competitive environments (Hitt et al., 1998; Nadkarni and Narayanan, 2007). Indeed, the relationship between SF and firm performance is extensively moderated by competitive intensity (Guo and Cao, 2014). In addition, previous research has further indicated that firm size and firm age may determine the performance effects of SF, with smaller firms and younger firms benefitting more from SF (Segaro, Larimo and Jones, 2014). Consequently, we consider these contingencies in our analyses.

Our meta-analytical research framework is presented in Figure 1. While the framework itself is grounded in theory, the articulation of SF's expected antecedents, consequences and contingencies is driven by a combination of the portrayed theoretical tensions and patterns in existing empirical findings. Thus, the goal of this meta-analysis is to empirically determine the nomological network that encircles SF.

Bivariate relationships

First, we estimate the bivariate relationships among all key constructs. As suggested by Aguinis et al. (2011), we use Hunter and Schmidt's (2004) random effects approach at the effect-size level, which allows for the correction of statistical artefacts and thus provides a relatively accurate estimate of the average strength and variance of a relationship in the population of interest. We use artefact-corrected effect sizes, transformed into Fisher's z coefficients and weighted by the estimated inverse of their variance. This correction gives more weight to more precise estimates before converting them back to correlation coefficients. For each bivariate relationship, we report a 95% confidence interval of the sample-weighted reliability-adjusted averaged correlation. We also calculate the ‘fail-safe N’, which indicates the number of non-significant and unavailable studies that would be needed to make the cumulative effect size non-significant to further evaluate a potential publication bias. We test the homogeneity of the distribution of effect size using the Q-statistic. Heterogeneity for all relationships was significant. However, the standard deviations are all below the common cut-off value for an acceptable standard deviation in a population (Cortina, 2003). Thus, analyses of the main effects are warranted. For each focal relationship, we apply the trim-and-fill method to assess publication bias and impute the missing values. None of the imputed relationships changed in direction or significance, and all bivariate relationships are reported in Table 3.

Antecedents of strategic flexibility

The bivariate relationships between antecedents and SF in Table 3 are largely in line with our theoretical expectations. However, while primary studies investigating the antecedents of SF focus only on subsets of potential enablers, inhibitors and triggers of SF, the aggregated study effects can be used to assess simultaneously the effects of variables that prior research have not considered jointly (Bergh et al., 2016). Therefore, we use MASEM, a two-stage procedure, to estimate the unique contribution of each of the antecedents of SF (Cheung and Chan, 2005). During the first stage, we conduct several meta-analyses, relating all constructs of interest to each other, using data from the primary studies in our database. We thus calculate a complete correlation matrix, including the effect sizes of all relationships (see Appendix 5 in the online Supporting Information). In the second stage, we apply structural equation modelling on the correlation matrix.

Given the variability in sample sizes associated with each correlation coefficient in the meta-analytic correlation matrix, we base this analysis on the harmonic mean of the sample sizes comprising each entry in the meta-analytic correlation matrix (Viswesvaran and Ones, 1995). The fit indexes for the MASEM model in Table 5 suggest a good fit for the model (χ²/d.f. = 23.66/26; RMSEA = 0.00; AGFI = 0.99; SRMR = 0.02). All of the relationships hold when the statistically non-significant antecedents are removed. Moreover, using trim-and-fill-corrected effect sizes, the parameter estimates do not change.

Regarding enablers of SF, we find that, consistent with Schweiger et al. (2019), entrepreneurial orientation (β = 0.33, p < 0.01), market orientation (β = 0.22, p < 0.01) and learning orientation (β = 0.30, p < 0.01) all are positively associated with SF, indicating the importance of the strategic orientation of a firm. Turning to aspects of organizational design, we find that decentralization (β = 0.54, p < 0.01) and formal routines (β = 0.28, p < 0.01) are positively associated with SF, while slack resources (β = −0.22, p < 0.01) is negatively associated with SF. This is an unexpected finding, considering that previous studies have emphasized that firms need uncommitted, liquid resources to enable SF (Barker and Barr, 2002).

When considering the inhibitors of SF, we find that past success (β = −0.21, p < 0.01) is indeed negatively associated with SF, and there is no significant relationship between firm size and SF (β = −0.01, n.s.) and even a positive relationship between firm age and SF (β = 0.06, p < 0.01). These findings highlight that not all commonly acknowledged inhibitors of SF can be traced in existing empirical studies. Finally, the findings related to the triggers of SF reveal that competitive intensity is positively associated with SF (β = 0.38, p < 0.01). Contrary to our expectations derived from the literature, environmental dynamism (β = −0.24, p < 0.01) and demand uncertainty (β = −0.06, p < 0.01) are negatively associated with SF. Discovering that both environmental dynamism and demand uncertainty do not trigger SF highlights an important gap between theory and empirical findings in existing studies.

Consequences of strategic flexibility

The bivariate relationships in Table 4 support the claim that SF is positively related to performance consequences. The correlations between SF and firm performance range from −0.59 to 0.87, and the correlation frequency is normally distributed (Z_skewness = −0.30, Z_kurtosis = 0.92, p_{Kolmogorov–Smirnov} < 0.05). Most of the SF–performance correlations in the database are positive (85%), and the adjusted mean correlation of the focal performance relationship is 0.24 (95% CI = 0.23–0.27).

We next investigate the specific conditions under which the performance effect of SF might be larger or smaller. We consider measurement of constructs, as well as methodological and sample characteristics, as potential influencers of the SF–performance relationship (e.g. Karna, Richter and Riesenkampff, 2016). Regarding the measurement of SF, we code all operationalizations according to the six dimensions displayed in Table 1. In line with other meta-analyses (e.g. Rubera and Kirca, 2012), we group the various operationalizations of firm performance into innovation outcomes, market outcomes and financial outcomes (see Appendix 6 in the online Supporting Information). We further code the year of data collection, whether the study is published, the ISI impact factor and secondary data versus survey data collection, as well as the geographic region of the sample and whether the sample consists of large firms or SMEs.

We then use MARA (Van den Noortgate et al., 2013), a special type of weighted least squares regression analysis designed to assess the relationship between effect size and potential moderator variables by modelling heterogeneity in the effect size distribution. The homogeneity tests for the correlations involving the SF–performance relationship indicate that moderator variables may explain heterogeneity in the effect sizes (χ²₁₈₁ = 2,495.87, p < 0.01). We model the reliability-corrected correlations in the moderation analysis as a linear function of the determinants and perform a three-level analysis to account for within-study error correlation between effect sizes and dependence of effect sizes within studies (Van den Noortgate et al., 2013). Therefore, we specify the following model:

(1)

where Y_ijk represents the reliability-corrected correlations of effect j in study k, γ₀₀₀ is a constant, γ_0l0 represents parameter estimates of the determinants, X_l,jk represents variable matrices of the determinants (moderators), r_00k is the study-level residual error term, u_0jk is the effect-level residual error term and e_ijk is the measurement-level residual error term. The results of the hierarchical linear regression appear in Table 6.

The positive SF–performance relationship is stronger when SF is measured in terms of proactively creating new opportunities (γ = 0.05, p < 0.01), internal resource deployment (γ = 0.05, p < 0.05) and external competitive actions (γ = 0.04, p < 0.01), compared to when it is not. Measuring SF in terms of reactivity (responding to change), variety (increasing options) and speed (timely response) does not affect the SF–performance relationship. We further find that the performance effect of SF is stronger when performance is measured in terms of innovation outcomes (γ = 0.05, p < 0.01) or market outcomes (γ = 0.07, p < 0.01) rather than financial outcomes. Finally, the SF–performance relationship is stronger both in newer studies (γ = 0.06, p < 0.01) and in published studies (γ = 0.03, p < 0.05).

Contingencies of strategic flexibility

The lack of available correlations involving interaction terms precludes moderation analyses of the theoretically relevant contingencies with MARA. Therefore, consistent with other meta-analyses (e.g. Blut et al., 2015), we report the partial correlation between the moderator and the effect size while controlling for all significant effects from the MARA analysis. Table 7 shows that only environmental dynamism significantly and positively moderates the SF–performance relationship (r_corrected = 0.46, p < 0.01). While we find directional support for the moderating effects of demand uncertainty and competitive intensity, these effects are not statistically significant (the significance tests in all analyses take the sample size k into account). Moreover, neither firm size nor firm age moderate the SF–performance relationship.

Post-hoc analysis: from strategic flexibility to financial outcomes

Both the bivariate results in Table 4 and the MARA results in Table 6 reveal that SF has a higher performance effect when firm performance is measured as innovation or market outcomes rather than as financial outcomes. In line with other meta-analyses (e.g. Rubera and Kirca, 2012), we thus explore the potential chain of effects, from strategic flexibility to financial outcomes, via innovation and market outcomes. On the one hand, SF enables firms to anticipate market demands and to develop innovative products and services (Neill and Rose, 2007), thereby increasing performance in terms of innovation outcomes (Kortmann et al., 2014). On the other hand, SF enables firms to respond and adapt to changing market demands (Johnson et al., 2003), thereby increasing performance in terms of market outcomes (Segaro, Larimo and Jones, 2014). Finally, improvements in both innovation and market outcomes should subsequently increase financial outcomes (Rubera and Kirca, 2012).

Considering the direct effect of SF on financial outcomes, SF may enable firms to possess a variety of different plans, decisions and strategies unrelated to innovation or the firm's market position (e.g. Evans 1991). This variety, however, introduces costs and potential benefits needed to exceed the necessary investments, as highlighted by the opponents of SF in Table 3. For example, when assets are committed for the purpose of SF, the firm forgoes its short-term earning potential, introducing the additional burden of opportunity costs (Shimizu and Hitt, 2004). Thus, one may expect no direct effect from SF on financial outcomes; rather, one may expect that the innovation and market outcomes will mediate the financial effects of SF.

We use MASEM to gain insights into the chain of effects between SF and financial performance (Bergh et al., 2016). As Figure 2a summarizes, an analysis of the expected mediation model did not result in an adequate model fit. Given this, and consistent with the modification indices, we revise the model, as shown in Figure 2c. This model fits the data well (χ²/d.f. = 19.44/44; RMSEA = 0.03; AGFI = 0.99; SRMR = 0.02) and demonstrates that SF has a direct positive effect on innovation outcomes (β = 0.35, p < 0.01), a small direct positive effect on market outcomes (β = 0.17, p < 0.01) and a small direct negative effect on financial outcomes (β = −0.06, p < 0.01). We use a bootstrapping procedure with 1,000 samples to test for indirect effects. The results show a positive effect of SF on market outcomes through innovation outcomes (β = 0.17, p < 0.01; 95% CI = 0.16–0.19) and a positive effect of SF on financial outcomes through innovation outcomes and market outcomes (β = 0.21, p < 0.01; 95% CI = 0.19–0.23). The overall effect of SF on financial outcomes, which combines both the direct and indirect effects, is positive (β = 0.15, p < 0.01; 95% CI = 0.13–0.18). These results suggest that SF is indeed an important part of a firm's competitive armoury, but its effect on financial outcomes might be more nuanced than previously theorized.

Research implications

Table 8 summarizes our core findings, theoretical and managerial implications and future research directions. Several potential antecedents of SF have proved to be controversial in the literature (see Table 2). We address these tensions and offer insights into the key enablers, inhibitors and triggers of SF. Our multivariate results reveal that both the strategic orientation and the organizational design of a firm enable SF. Interestingly, decentralization has a higher impact on SF than any other enabler (χ²_diff = 8.75, p < 0.01). Thus, shared decision-making across an organization appears to be crucial to promote SF. We further find that slack resources negatively affect SF. While this finding contradicts the common claim that organizations without uncommitted, liquid resources are less likely to adapt to environmental changes (Barker and Barr, 2002), it is in line with Sanchez (1995), who caution that resources should be flexible rather than slack. Following his logic, resource surplus implies they are not utilized optimally and efficiently, while resource flexibility implies that resources can swiftly be re-utilized to underpin new strategic options.

The analyses also reveal that past success inhibits SF, while firm size and firm age do not inhibit SF. These insights extend the knowledge on the inhibitors of SF, previously an under-researched area, and provide resolution to tensions on whether firm size and firm age inhibit SF. Finally, competitive intensity triggers SF, while environmental dynamism and demand uncertainty exhibit a negative relationship with SF. Discovering that neither environmental dynamism nor demand uncertainty trigger SF challenges the existing literature and raises some potential controversies, especially in suggesting that SF is achieved more easily in highly competitive environments. However, taking into consideration the multidimensionality of the environment and the results of the contingency analyses, the whole picture is complex. Evidently, more research on the triggers and contingencies of SF, especially on environmental dynamism and competitive intensity, is needed.

Our findings also address the conflicting empirical evidence regarding the relationship between SF and performance (see Table 3). We find a general moderately positive effect size of 0.24 for the SF–performance relationship, thereby resolving the ongoing debate over whether SF has performance-enhancing benefits for firms or whether it can even harm firms. Our findings support the arguments submitted by proponents of SF, namely that, in today's competitive landscape, firms’ success depends on their ability to address changing conditions and to constantly adapt their strategies accordingly.

From a contingency point of view, our results also indicate that neither the aspects of the environment nor the firm characteristics affect the SF–performance relationship in the way we might expect based on previous studies conducted in particular settings (e.g. Grewal and Tansuhaj, 2001). While environmental dynamism positively moderates the SF–performance relationship, no significant effects were found for either demand uncertainty or competitive intensity. These findings suggest a more nuanced view on the nature of environmental contingencies affecting the consequences of SF than previously considered.

Moreover, we demonstrate that conflicting findings in the literature can be at least partly attributed to the conceptualizations of SF employed in earlier studies. In order to appreciate the wider contribution of SF, it must be acknowledged that it is a multidimensional concept, something that many earlier studies failed to recognize sufficiently. We find that the performance effect is larger when SF is measured in terms of proactively creating new opportunities, internal resource deployment and external competitive actions. It is surprising to find that neither variety nor speed (timely response) affect the SF–performance relationship. Our analyses suggest that for many firms, the cost of developing and maintaining strategic options may outweigh the benefits.

We also identified the avenues through which SF influences firm performance. We find that SF has a negative direct effect on financial performance, therefore providing prima facie support for the opponents of SF when the ‘bottom-line’ performance of firms is considered. However, when the indirect positive effect of SF through innovation outcomes and market outcomes is considered, SF's overall impact on financial outcomes is positive, with an effect size of 0.15. This overall positive effect suggests that, all else being equal, the innovation- and market-related outcomes of SF (positive indirect effect) outweigh the costs associated with SF (negative direct effect), leading to a positive SF effect on financial performance. However, it should be stated that the small-to-moderate effect size of the SF–performance relationship does not justify the overly enthusiastic appraisal of SF in the normative literature. Thus, while our findings attest to the overall importance of SF, we caution against regarding SF as a panacea for all kinds of competitive threats.

Although we were able to study only firm-level effects, there is a dearth of literature devoted to the micro foundations and micro-level variables in the study of SF. Individual-level phenomena with regard to C-suite executives (such as leadership variables) should be examined in order to identify the individual- and also group/team-level dynamics that enable SF to flourish (such as motivations and engagement). Moreover, future studies in SF should adopt multi-level approaches in which micro-, meso- and macro-level phenomena can be examined simultaneously, along with their interactions and interdependencies. Equally, studies should consider the extent to which SF can be differentially geared within organizations. By this we mean that SF can be treated in an ambidextrous manner whereby parts of the organization are exposed to the demands of SF whereas other parts are not. From this form of intra-organizational research, modest forms of flexibility can be considered exploitative whereas substantive forms can be described as exploratory. This is apparent in the ambidexterity literature but not explicitly from a SF perspective.

Managerial implications

In 1964 the average tenure of firms on the S&P 500 index was 33 years. By 2016 this reduced to 24 years and is projected to be a mere 12 years by 2027 (Anthony et al., 2018). This equates to half of these large US$6bn or more firms disappearing by 2027. Various reasons account for this attrition, not least of which is firms’ lack of strategic flexibility in aligning their offerings to the market. Our results provide important insights for managers faced with questions regarding SF to remain competitively vital, in order to strategically flourish over time.

Table 8 provides a series of actionable managerial practices that are available for strategists to engage in seeking to realize SF. Our results from the multivariate analysis of antecedents of SF demonstrate the importance of entrepreneurial orientation, market orientation, learning orientation, decentralization and formal routines in enabling a firm to become flexible. Thus, engaging in these firm-wide orientations demands the ability to engage in ambidextrous and polydextrous behaviours that allow these different plates to be consistently and simultaneously spun. However, managers also need to be aware that slack resources can have unintended consequences with regard to SF and that past success further inhibits SF, while firm size and firm age do not have opposing effects. Finally, the positive effect of competitive intensity on SF (but not of environmental dynamism and demand uncertainty) highlights that competitive pressure will motivate firms to become more flexible. These different findings indicate a repertoire of nuanced managerial interventions and not a ‘one-size-fits-all’ approach. Clearly, managers need to be aware of the complacency effect that is often characterized by exemplary past performance. The appetite firms have to engage in success needs to remain once they become successful. Ensuring that structural, leadership and cultural transformation remains aligned with the evolving marketplace is critical, and maintaining a focus on the competitive pressures will keep this intensity to perform high.

Moreover, our meta-analytical findings demonstrate that SF has an overall positive impact on firm performance. However, not all firms can extract the highest gains from SF. As such, the findings on the contingencies provide useful managerial insights pertaining to which firms should invest more or less in SF. While SF appears to be especially valuable for firms that face high levels of environmental dynamism, firms facing lower levels of dynamism should be more cautious about whether their investment will pay off. However, demand uncertainty and competitive intensity did not affect the SF–performance relationship and are thus not helpful to managers in terms of determining what level of SF may be appropriate. We also found that firms should invest in SF regardless of their size and age, as these characteristics did not affect the SF–performance relationship.

Our findings further reveal that managers should be cognisant of the total impact that SF can have on firm performance, rather than focusing on its impact on financial outcomes alone. We examine a more comprehensive model of the mechanisms that mediate the SF–financial outcomes relationship than those tested in the extant literature. We thus suggest that measures of the mediating variables – innovation and market outcomes – may be useful for managers in terms of tracking the impact of SF on overall performance. Only if managers undertake such tracking can the full potential of SF be captured. A myopic focus on financial outcomes may underestimate the importance of SF.

Limitations and directions for future research

Despite its contributions, this study has some important limitations that should be considered when interpreting our findings. First, any meta-analysis is constrained by the nature and scope of the original studies. Thus, our study is de facto limited to examining the effects of variables that were available in existing studies. Moreover, because the primary-level studies were not experimental in nature, the reported results do not provide direct and unequivocal evidence regarding causality. Indeed, the vast majority of the studies have used cross-sectional methods to examine SF. However, most research questions concerning SF – even if not framed as such – address issues of change. Consequently, we recommend that future studies use longitudinal or experimental designs in order to overcome these methodological weaknesses and capture the causal effects of SF. Another important concern is that the relationships reported in the original studies may be positively biased because of an oversampling of successful firms; firms that failed to be flexible may well have exited the market and would, therefore, not have been captured by the studies in our sample.

In terms of measurement, many studies have employed a unidimensional measure of SF, thus ignoring the multidimensional composition of the construct. Their empirical findings may have underestimated the extent of SF because they did not consider that SF can be a combination of various factors: the ability to be both reactive and proactive and to have internal and external options, which enable a firm to take a variety of actions and to do so rapidly. Thus, multidimensional measures of SF should be employed in order to better capture the relationships under scrutiny. Moreover, most studies investigating the SF–performance relationship under different environmental conditions have employed a unidimensional measure of the environment. A shortcoming of the argument that more flexibility is needed in more dynamic markets is that this fails to consider the multidimensionality of the environment and the conflicting empirical findings. Therefore, further research should detail the effects of SF under varying environmental conditions.

Despite significant progress in understanding SF and its nomological network, several knowledge gaps remain, and future research directions arise from these gaps. Longitudinal studies that differentiate clearly between short-term and long-term outcomes would also be valuable – this would yield a greater understanding of SF's effects on firm performance. In addition, while the current focus in the strategy literature is on the external consequences of SF, it is important to note that SF may affect several internal organizational factors that, in turn, may affect innovation, market and financial outcomes. Further research should therefore investigate the mediating role of the costs of experimentation, inefficiency, stress among employees and a potential lack of focus in the SF–performance relationship. Due to data limitations, we could not examine individual- or team-level aspects of SF. Thus, we only focused on the firm level while neglecting the micro and meso level of SF such as leadership and motivation, commitment and focus among employees. Future studies should take a more multi-level approach, in which both the micro and meso level of SF are involved.

In summary, this meta-analysis develops an integrative framework of the antecedents, consequences and contingencies of SF by resolving several tensions and conflicting findings in the literature. By highlighting which enablers, inhibitors and triggers have the strongest impact on building SF, and by clarifying when and how SF influences the performance of firms, we provide managerial guidance to benefit from SF and conceptual clarity for future research on SF.