2.1 TMT–board intergroup interactions and innovation

When both TMT and BOD gender diversity are low, we predict less organizational innovation (Miller & Triana, 2009; Triandis et al., 1965), because the underrepresented members' voices and opinions in both TMT and BOD often times are not considered by the overrepresented members (Kanter, 1977; Ridgeway, 2011, 2014). Such difficulty communicating across demographic subgroups is known to be an issue for upper echelons demographic diversity (Ancona & Caldwell, 1992). The same communication issue may even be magnified when gender diversity in both the TMT and BOD are low. For example, a homogenous TMT tends to prefer investment projects that build on familiar concepts, ideas, and technologies, and is reluctant to pursue projects that build on new knowledge or change the firm's strategic direction (Ahuja & Lampert, 2001). Such rigid ideas can be furthered by a homogenous BOD, because the latter's low receptivity toward new investments limits its ability to recognize the value of new initiatives (Cox & Blake, 1991; Hambrick & Mason, 1984), reducing organizational innovation. Therefore, we expect the lowest level of innovation to be present when gender diversity is low within the entire strategic leadership team (i.e., both TMT and BOD).

When either TMT or BOD gender diversity is low in one group while high in the other, potential information-sharing benefits offered by a diverse group are likely to be poorly utilized by the group with homogenous members. For example, when the TMT is gender-diverse but the BOD is gender-homogenous, ideas coming from the TMT may be discouraged by the BOD in its advice-giving capacity if the board does not share the same perspective. Also, a gender-diverse TMT may propose innovations that are unrelated to the prior knowledge and experiences of gender homogenous BODs (R. J. Ely & Thomas, 2001). In a study about TMTs and chief executive officers (CEOs), Georgakakis et al. (2017) demonstrated that some level of demographic background and/or experience similarity between two different entities in the upper echelons facilitates understanding between those organizational members at different organizational levels. Compared to TMTs high in gender diversity, TMTs with low gender diversity are more likely to have a similar body of knowledge and perspectives (Byrne, 1971). In such cases, TMTs are less likely to search for, grasp, and consider innovative ideas presented by a gender-diversity BOD and are more likely to hold on to their established mental model of how things should be done (Thompson, 2018). Homogenous TMTs are more likely to overlook potential opportunities generated in mutual exchanges with gender-diverse BODs, which could provide sources of unconventional thinking and innovation.

By the same token, a gender-homogeneous BOD may not possess either a broad enough knowledge base or a diverse enough network of contacts to support frequent TMT innovative initiatives (Ibarra, 1992, 1993). Although a gender heterogeneous TMT might be open to and more likely to initiate innovation, a gender homogeneous BOD would prefer less proactive strategic behavior and change. Similarly, when TMT gender diversity is low, the TMT will be less receptive toward implementing organizational innovations that a gender-diverse BOD suggests. Although strategy is primarily formulated by the TMT, the BOD does serve in an advisory and resource provision capacity (Johnson et al., 1996) which also shapes strategy (Deutsch, 2005). Even if board members present multiple perspectives and suggestions in their advice role, these potential innovations may fall on deaf ears in the TMT if the TMT is not open to such innovations.

In contrast, when high TMT–BOD diversity is present, this abundance of perspectives will amplify the knowledge benefits, thereby overshadowing the potential for conflict and communication difficulties (Ancona & Caldwell, 1992), resulting in more innovation (Gruenfeld et al., 1996; Phillips et al., 2004). A gender-diverse BOD is more likely to approve a variety of strategic proposals put forward by the TMT if the BOD has a broader knowledge base and is more flexible in its own thinking (Cox & Blake, 1991; Goodstein et al., 1994; Post & Byron, 2015; Ruderman et al., 2002; Ward & Forker, 2017). The resource provision function of boards involves activities such as providing legitimacy, expertise, advice, counsel, information, and other resources to firm executives, which can facilitate innovation (Johnson et al., 1996). Different genders provide a range of experiences, perspectives, and social resources that can enhance the firm's ability to innovate (Miller & Triana, 2009).

Access to diverse resources via gender-diverse BODs is especially important for gender-diverse TMTs. Diverse management groups have an increased need for external resources in order to support intensive competitive activity (Andrevski et al., 2014), since they generate a variety of new competitive moves. As such, a gender-heterogeneous TMT can integrate a gender-diverse BOD's knowledge and use these connections to provide a competitive advantage (Barney et al., 2001). While gender heterogeneous TMTs might be equipped to acquire and accumulate various resources, the BOD can have a notable impact on the TMT's priorities by ensuring they stay knowledgeable regarding the industry, competition, and strategic options (Zahra et al., 2009). Gender-diverse BODs are therefore more likely to complement and expand the TMT's innovative capacity by providing access to external resources. BOD diversity can generate ideas for new strategic actions and provide resources supporting the TMT's innovative actions (Carpenter et al., 2004; Kim et al., 2009).

Furthermore, if gender diversity is high on both the TMT and the BOD, then there will be more women represented in the upper echelons and it will be normative to involve women in strategic decision-making. Research on board gender diversity shows that when women reach a critical mass of three or more women on the board, firms are more innovative (Torchia et al., 2011). Having high gender diversity on both the TMT and the BOD makes it more likely that the upper echelons has a critical mass of women, which can bolster innovation. A gender-diverse BOD can then encourage the TMT to explore various options because they are willing to embrace those options. A gender-diverse TMT would be more likely to accept alternative perspectives from a gender-diverse BOD, which would enhance rather than simply echo the TMT's voice. In sum, we suggest that high gender diversity on both the TMT and BOD amplifies the advantages and mitigates the disadvantages of gender diversity, resulting in more innovation. We predict that a gender diverse TMT is more likely to echo a gender diverse BOD as both teams are more likely to introduce new ideas, perspectives, and knowledge. The BOD will also be better able to leverage the TMT's innovative capacity by providing a wide range of knowledge-based resources.

2.2 The moderating role of environmental dynamism

Research highlights that more stable environments constrain strategic decision-making processes (Yamak et al., 2014), while dynamic environments demand a broader range of knowledge from TMTs and BODs in order to handle ever-changing situations (Gordon et al., 2000). We propose that TMTs and BODs will share an overarching goal for more organizational innovation in a dynamic environment, because the context demands that for firm competitiveness (Luciano et al., 2020; Zaccaro et al., 2012). In a volatile environment, it is important for the firm to possess diverse perspectives and have the capability to integrate knowledge across members of the strategic leadership team (Judge & Miller, 1991). Therefore, in a dynamic environment, we expect the TMT and BOD will benefit from gender diversity and from more interdependent rather than independent work which will ultimately maximize innovation. Hence, we propose that the positive effect of TMT–BOD gender diversity on innovation is moderated by environmental dynamism such that a gender heterogeneous TMT in conjunction with a gender heterogeneous BOD will be more willing to undertake necessary innovative changes in dynamic environments than stable ones for the following reasons.

First, the benefits associated with gender diversity-based knowledge contained within the TMT–BOD match will be stimulated by a dynamic environment. A dynamic environment motivates members of TMTs and BODs to adopt a coherent and consistent schema for strategic action to achieve continuous innovation (Gordon et al., 2000; Rajagopalan & Spreitzer, 1997). As such, both the TMT and BOD should be more proactive and cooperative regarding strategic actions. A wealth of useful information offered by TMT–BOD gender diversity is useful for innovative initiatives and more likely to be obtained when cooperation and coordination are high, which we argue, will be most demanded in a dynamic environment compared to a stable environment. This is consistent with the theory by Luciano et al. (2020), who stated that dynamic environments require rapid responses and realignments that necessitate strong interdependence between the TMT and the BOD while also maintaining BOD independence required for good governance. In stable environments, a broad range of knowledge is less necessary, because while BOD and TMT members may possess diverse perspectives based on their gender differences, they are less motivated to work together interdependently. A dynamic environment will provide a context that demands greater coordination, and when both the TMT and the BOD possess gender diversity-based information, the two teams may be better equipped to innovate.

Second, a dynamic environment requires more creative decision-making than a stable environment (Hambrick et al., 2015; Hambrick & Mason, 1984), which stimulates the information-sharing benefits of diverse groups. In dynamic environments, gender diversity should be seen as an opportunity rather than a threat, because of the need for innovation. Further, research on team dynamics shows that when diverse teams approach their work with the mindset that diversity is an advantage, the teams perform better (Homan et al., 2007, 2008). Research demonstrates that the need for knowledge-based assets that heterogeneous TMTs and BODs possess is indeed evident within dynamic environmental contexts (Cannella et al., 2008). The gender diverse nature of the TMT and the BOD joint influence is amplified by the dynamic environmental context, which will trigger more of a need for TMT–BOD intergroup interdependence through improved coordination/cooperation (Cooper et al., 2014; Luciano et al., 2020) and a greater demand for innovative outcomes. In environments that are stable rather than dynamic, we expect less impetus for TMT–BOD coordination and less emphasis on maximizing innovation efforts, especially if either (or both) the TMT and BOD are also low on gender diversity. In sum, because dynamic environments require more creativity from strategic leadership teams to meet the demands for greater innovation needed for competitiveness, we propose a gender-diverse TMT will (a) be better able to coordinate with a gender diverse BOD, and (b) be better equipped to contribute interdependently with a gender diverse BOD.

2.3 The mediating role of innovation

Hypothesis 2 suggests that the positive effect of TMT–BOD gender diversity fit on innovation is moderated by environmental dynamism. We move beyond that prediction by examining mediation and considering a mechanism that links intergroup TMT–BOD gender diversity to firm performance (Hoobler et al., 2018; Lawrence, 1997). We focus on the role of innovation as the mediating mechanism between the interaction of TMT–BOD gender diversity and environmental dynamism on firm performance.

Our reasoning is that innovation can transmit the interactive effect of the benefits associated with high TMT–BOD gender diversity and environmental dynamism to firm performance. This is because innovation modifies a firm's current market status, preventing the organization from entering an inflexible state of inertia (Klarner & Raisch, 2013; Levinthal & March, 1993). In this sense, innovation can be viewed as continuous change and adjustments by a focal firm in order to outperform its rivals (Brown & Eisenhardt, 1997). Although innovations do not always result in better firm performance, firms making purposeful changes are more likely to have better performance than those firms that do not (D. Miller, 1991). As such, the innovation benefits accrued from gender diversity in both TMT and BOD should ultimately result in better performance, particularly in dynamic environments.

Taken together, we propose a comprehensive moderated mediation model such that the positive joint effect of TMT gender diversity and BOD gender diversity within our multiteam system framework has an indirect effect on firm performance via innovation. This TMT gender diversity-BOD gender diversity positive interaction effect will be more salient in a dynamic environment because (a) coordination between the TMT and BOD will be possible with both having gender diversity (Georgakakis et al., 2017), and (b) creativity will be maximized because both the TMT and the BOD will seek to utilize their knowledge bases to meet the superordinate goal of increasing organization innovation (Sherif, 1966). In contrast, in a stable environment where innovation is often not a prerequisite for long-term competitiveness, we posit that TMT to BOD gender diversity within the multiteam system results in less coordination between the TMT and BOD and fewer innovative outputs based on additive independent effects of the TMT or the BOD rather than multiplicative interdependent effects, both of which diminish organizational innovation and subsequent firm performance. The following prediction is made:

3.1 Data and sample

We test our hypotheses by constructing panel data on the TMTs and boards of publicly listed Chinese companies from multiple sources. Publicly listed Chinese companies are posted on both the Shanghai and Shenzhen Stock Exchanges. We extracted TMT and board characteristics (e.g., gender) from the Corporate Governance Database provided by the China Stock Market and Accounting Research (CSMAR) database, a prominent provider of data from publicly listed Chinese firms. We extracted firm-level financial information and other data (e.g., R&D expenditure, number of employees and ownership structure) from the annual reports of each publicly listed Chinese company. We then matched different databases based on a unique stock code identifier. The final sample contained 944 firm-year observations consisting of 328 firms from 2008 to 2013.

3.2 Dependent variables

3.3 Independent variables

3.4 Moderating variable

3.5 Control variables

We included several control variables that can influence the dependent variables. First, we controlled for Past Innovation (innovation during year t). Second, prior studies suggest that foreign ownership structure affects firm performance so we controlled for a firm's foreign ownership (Tuschke & Sanders, 2003). Foreign Ownership was measured as the ratio of foreign-owned capital divided by the total capital. We also included Firm Age measured as number of years since establishment (Barron et al., 1994) and Firm Size measured as the log of the number of employees (Jayaraman et al., 2000).

We also controlled for CEO Power. Following prior studies (Haynes & Hillman, 2010; Krause et al., 2016), we operationalized CEO power as a summed index of three frequently employed measures consistent with previous approaches to measuring CEO power (e.g., Cannella & Shen, 2001; Krause et al., 2015). The three elements included: (1) CEO duality measured as a dichotomous variable with a value of 1 reflecting a combined CEO/board chair, and a value of 0 reflecting a separate CEO and board chair (Finkelstein & D'Aveni, 1994); (2) CEO stock ownership measured as the percentage of outstanding firm stock owned by the CEO (Boeker, 1997); and (3) CEO compensation measured as the ratio of the CEO's salary divided by the total salary of all TMT members. We obtained the three variables from CSMAR's Corporate Governance sub-database. Following prior studies (Haynes & Hillman, 2010) we first standardized the two non-dichotomous variables of CEO Stock Ownership and CEO Compensation to have the same range as the dichotomous variable, CEO Duality. We then aggregated these three variables and standardized them to form an index of CEO power. We also controlled for the Ratio of Outsiders on the Board by considering that outside directors may steer the firm into new businesses. We measured the ratio of outsiders on the board by calculating the number of outside directors on the board divided by total board size. The average value is 0.32 in our sample.

Moreover, we controlled for other TMT demographic-heterogeneity indices as well as board demographic-heterogeneity indices such as TMT (board) age heterogeneity and TMT (board) tenure heterogeneity. Following prior studies (Hillman et al., 2007; Tsui & O'Reilly, 1989; Westphal & Zajac, 1995) we measured TMT (board) age heterogeneity with an analog of the Euclidean distance measure (i.e., the coefficient of variation) commonly used in research on organizational demography. The following formula was used for this calculation: , where is the individual age and is the value of every other individual's age. We operationalized TMT (board) tenure heterogeneity by following prior studies (e.g., Joshi & Knight, 2015) which classified TMT (board) tenure into groups. Specifically, we had three groups designated as having tenure of less than 3 years, having tenure of 3 to 6 years, and finally having tenure of more than 6 years. We then calculated the TMT tenure heterogeneity and board tenure heterogeneity separately. We also included a set of industry dummy variables in order to control for industry effects, and we controlled for the year effect by generating year dummy variables to include in the analyses.

3.6 Econometric model

We utilize the panel data econometric model by organizing data into firm-year observation units. The fixed and random effects are the most common models that can control unobserved effects and partially solve the endogeneity problem. The econometric model is written as: , where is the dependent variable, is the vector of explanatory variables that are independent of each other for i and t, is the vector of coefficients to be estimated, is the firm specific unobservable effects, and is the random error following the normal distribution. Hausman's (1978) specification test was used to detect whether or not a fixed- and random-effects model is most appropriate. The Hausman test was not significant for our models, suggesting that the fixed-effects models were less reliable and efficient than random-effects models.

Moreover, time-specific factors such as regulatory changes or economic downturns may also affect innovation and firm performance (Certo & Semadeni, 2006); these omitted variables may cause endogeneity problems. Including time dummy variables in panel data models having a large N (total firms) and a relatively small T (time periods) is useful in reducing the effects of contemporaneous correlation (Certo & Semadeni, 2006). We therefore also included a set of dummy variables for each year in addition to implementing the random-effects model handling the omitted variables (Baltagi, 2013). Innovation can also affect TMT characteristics. For example, a firm that is more likely to accept greater innovation may also be more likely to add TMT members who differ from the current cohort and affect TMT heterogeneity. All of the explanatory variables are accordingly lagged before the dependent variable in order to account for reverse-causality effects. In testing the main effects of TMT gender heterogeneity on innovation, the dependent variable, innovation was predicted at year t + 1 (e.g., 2011), and the independent variables (i.e., TMT gender heterogeneity) and control variables were predicted at year t (e.g., 2010). To test for mediation, we account for the independent variables and control variables at year t (e.g., 2010), and lag innovation to time t + 1 (e.g., one-year change between 2010 and 2011) and firm performance to year t + 2 (e.g., 2012).

3.7 Correcting for selection effects and endogeneity

The endogeneity problem occurs when there are errors (i.e., the true value of independent variables is not observed), an omitted variable (a variable that correlates to both the outcome and one or more explanatory variables is not included in the model), or simultaneous causality (when outcome variables affect explanatory variables and vice versa). Since the explanatory variables are directly observable in annual report logs and TMT information is available in CSMAR's Corporate Governance Database, the errors-in-variables issue is less problematic. Omitted variable bias is less likely because we control for several alternative explanatory variables within a longitudinal setting. Nevertheless, our model could still suffer from simultaneous causality.

In order to alleviate the endogeneity concern in TMT gender diversity and board gender diversity, we include the following instrumental variables addressing endogeneity concerns below. State Ownership. The firm's governance structure may affect the gender heterogeneity in the TMT (Cooper et al., 2014). TMT Size. Prior studies (Hart & Van Vugt, 2006; Thatcher & Patel, 2012) suggest that team size influences the likelihood of attribute alignment (e.g., the homogeneity within a group or between groups will be changed,) so that TMT size influences the likelihood of TMT gender heterogeneity. Board Size. Similarly, a large board size is important for developing more transparent and rigorous corporate governance structures. BOD members with a large board size may need to work closely regardless of the presence of gender heterogeneity, resulting in less conflict, with positive implications for innovation and performance. CEO Gender. P. M. Lee and James (2007) suggest that the CEO's gender would affect the appointments for other TMT members. Board Chair's Gender. The gender of the chair of the board may also affect appointments for board members. We additionally used three-digit industry codes and year dummies. These proposed instrumental variables potentially increase or decrease heterogeneity strength but may be less correlated with yearly performance fluctuations and industry variations. We therefore include Industry Dummies and Year Dummies as instrumental variables.

We followed Cooper et al.'s (2014) approach to assess the exogeneity of our TMT gender-heterogeneity measure using the Durbin–Wu–Hausman test in the “ivreg2” STATA package. First, it is critical to check if some of the excluded instruments are redundant (Hall & Peixe, 2000). An instrument is redundant if its inclusion in the excluded instruments set has no impact on the asymptotic variance covariance matrix of the estimators. We performed the redundant test, which suggests two instrumental variables satisfy the test: CEO Gender and Board Chair's Gender. We then performed an overidentification test. The Hansen J statistic is 0.00 which means these instrumental variables in the test are valid and exactly identified. We further checked for weak identification tests using the Cragg–Donald Wald F test, where the null hypothesis means the instruments are weak. If the Cragg–Donald statistic is above the critical value at significance level of 5% (Stock & Yogo, 2005), then one should reject the null hypothesis that the instrumental variables are not relevant. In our test the Cragg–Donald statistic was 10.51, which is greater than the Stock–Yogo critical value of 7.12 at 10% maximal relative bias. The instruments are therefore relevant and could effectively reduce the threat of endogeneity. Finally, the Endogeneity test of endogenous regressors is 2.35 with p-value = 0.426, which means the specified endogenous regressors can actually be treated as exogenous. Therefore, the TMT gender diversity and board gender diversity measure is exogenous and the unbiased estimates can be reported (Davidson & Mackinnon, 1993).

4.1 Post hoc analyses

We adopted several steps in order to test the robustness of the results. First, we provide two alternative measures of TMT gender diversity and board gender diversity using Women Number and Blau's heterogeneity index. For Hypothesis 1, Table 3 shows that the coefficient for the interaction term TMT Women Number × Board Women Number is also positive and statistically significant (β = 0.210, p = 0.008 in Model 2 of Table 3), and the coefficient of the interaction term TMT Gender Heterogeneity × Board Gender Heterogeneity is positive and statistically significant (β = 0.209, p = 0.008 in Model 5), providing support. For Hypothesis 2, the coefficient for the interaction term TMT Women Number × Board Women Number × Environmental Dynamism is positive and statistically significant (β = 0.214, p = 0.028 in Model 3), and the coefficient of the interaction term TMT Gender Heterogeneity × Board Gender Heterogeneity × Environmental Dynamism is positive and statistically significant (β = 0.212, p = 0.038 in Model 6 of Table 3). For Hypothesis 3, we also perform nonparametric bootstrap estimations (with 1000 replications) using a 95% confidence interval. We found positive mediation evidence on performance for TMT women number when board women number and environmental dynamism were both high (effect = 0.02; lower level of the bias-corrected confidence interval = 0.00 and upper level of the bias-corrected confidence interval = 0.05). Similarly, we found positive mediation evidence for TMT gender heterogeneity when board gender heterogeneity and environmental dynamism were both high (effect = 0.23; lower level of the bias-corrected confidence interval = 0.18 and upper level of the bias-corrected confidence interval = 0.29), providing support.

Second, we conducted an additional robustness test by using innovation at t + 1 but lagging firm performance at year t + 3. The results are consistent with those using a one-year lag (positive mediation evidence for TMT women ratio when board women ratio and environmental dynamism were both high with effect = 0.23; lower level of the bias-corrected confidence interval = 0.19 and upper level of the bias-corrected confidence interval = 0.28). Finally, while testing the moderated mediation effect we performed a nonparametric bootstrap estimation (with 1000 replications and 10,000 replications) using a 95% confidence interval. The results provide similar support for Hypothesis 3. Thus, we find consistent support for Hypotheses 1, 2, and 3.

6.1 Data and sample

We constructed our sample by referring to FTSE350 UK public companies. We extracted TMT and BOD characteristic data (e.g., gender, age, tenure, etc.) from BoardEx and obtained the corresponding financial data from the Worldscope database of Datastream. We also constructed a series of control variables that are identical to those control variables used in Study 1. The final UK sample contained 1356 firm-year observations consisting of 245 firms from 2008 to 2015. We then reran the same analyses conducted in Study 1, using the UK data, and reported the reestimated results in Table 4. As shown in Model 3, the coefficient of the interaction term, TMT Women Ratio × Board Women Ratio, is positive and statistically significant (β = 1.280, p = 0.000). We used the estimated coefficients reported in Model 3 to plot the relationship in a similar vein (see Figure 3). The positive association between TMT gender diversity and innovation is more positive under high levels of board gender diversity versus low levels of board gender diversity. This positive difference is statistically significant because the t-value of the simple slope difference between two lines is 5.52 (p = 0.000). These results using the UK data support Hypothesis 1.

The coefficient of the interaction term, TMT Women Ratio × Board Women Ratio × Environmental Dynamism, is positive and statistically significant (β = 1.915, p = 0.051 in Model 6). This supports Hypothesis 2. We then plot the three-way interaction of the UK data in Figure 4. Similar to that of China in Study 1, Study 2 shows that the positive association between TMT gender diversity and innovation is also more positive under high levels of board gender diversity and high levels of environmental dynamism (Line 1) versus low levels of board gender diversity and low levels of environmental dynamism (Line 4). This is supported with the t-value of simple slopes difference of 4.98 (p = 0.000) between Line 1 and Line 4. Both the estimated results and the interaction plot using the UK data support Hypothesis 2 predicting that the positive effect of TMT–board gender diversity on innovation is moderated by environmental dynamism such that the effect is stronger in dynamic environments than stable ones.

Next, we reexamined Hypothesis 3 using the UK data. The results show a positive mediation effect of innovation on the Tobin's Q performance measure with respect to TMT gender diversity when board gender diversity and environmental dynamism were both high because the mediation effect = 0.19 with the lower level of the bias-corrected confidence interval = 0.15 and the upper level of the bias-corrected confidence interval = 0.23. Values are positive and the confidence interval does not contain zero. Thus, the UK data further support Hypothesis 3 which also received support in Study 1 using the Chinese data.

Furthermore, we reran the analyses by using women number and Blau's index as two alternative measures of TMT gender diversity and board gender diversity (see Table 5). The results in Table 5 are consistent with those in Table 4, further supporting our hypotheses. In conclusion, the results obtained from the UK data (i.e., Study 2) are consistent with that of the Chinese data (i.e., Study 1), suggesting that our theory is generalizable across country context.

7.1 Theoretical implications

The preceding analysis demonstrates clear support for the proposed theory. All other things being equal, TMT–BOD gender diversity (i.e., high gender diversity in both TMT and BOD) has a positive effect on organizational innovation. This positive effect of TMT–BOD gender diversity on innovation is moderated by environmental dynamism such that the effect is stronger in dynamic environments than in stable ones. Furthermore, innovation mediates the interaction effect between TMT–board gender diversity and environmental dynamism on firm performance. These analyses offer lessons for several fields of study: the research on gender diversity, organizational research on diversity and innovation, and research on the upper echelons, such as the multiteam perspective on strategic leadership. We discuss implications for each of these areas.

We advance research on gender diversity in the upper echelons by testing the interaction between diversity on the TMT and BOD to influence firm innovation. Whether analyzing executive characteristics and their organizational impact (Bader et al., 2019) or gender in the boardroom and board performance (Kakabadse et al., 2015), extant literature does not include the nuance of gender diversity among the TMT and BOD and its effect on the organization's innovation. Only an understanding of different degrees of TMT and BOD gender diversity could quantify how gender diversity in the multiteam strategic context relates to organizational innovation and performance. By highlighting what prior studies leave implicit or ignore—the role of other relevant groups intimately interacting with the focal group—the proposed framework lends itself to an analysis of the impact of upper echelons multiteam diversity interaction effects that had previously been assumed or ignored.

We begin to address an important but unresolved issue in upper echelons research with respect to the nature of the relationship between TMT and BOD joint effects on organizational outcomes. Recent theorizing by Luciano et al. (2020) taking a multiteam approach to strategic leadership proposes that TMT–BOD interdependence and BOD independence are both especially critical in dynamic environments. Although we do not directly test their theory, our results do have implications for research on TMT–BOD joint effects on organizations. Analyzing the effect of group–group (i.e., TMT–BOD) gender diversity pays great dividends, as it allows us to simultaneously consider how these effects may change relative to each other (e.g., one may increase while the other declines), depending on the different type of combination of TMT to BOD gender diversity. Appreciating the significance of how simultaneous changes in these effects vary across different TMT–BOD diversity combinations is necessary for moving beyond the conventional focus on gender diversity within a single group, thereby highlighting the important role of other relevant groups that intimately interact with the focal group, producing the joint effect of multiteam gender diversity on organizational innovation and subsequent firm performance. Our research answers calls to conduct TMT–BOD multiteam research, and it also supports the theory that environmental dynamism creates a context where TMT–BOD integration is especially important (Luciano et al., 2020). Indeed, the findings show that high environmental dynamism is a catalyst for innovation when both TMT and BOD gender diversity are high. However, our findings also suggest that multiteam work examining TMT–BOD joint effects needs to consider the diversity of the TMT and the BOD as an important factor that can influence group decision-making outcomes like innovation.

We make important contributions to the multiteam approach to strategic leadership by shedding light on whether and how BOD–TMT gender diversity stimulates organizational innovation. Previous research on when board diversity helps or hurts innovation focused on the degree of gender diversity (Miller & Triana, 2009) or how prior firm performance moderates the effect of board diversity (Triana et al., 2014), while other research suggests the importance of the environment in studying TMT demography (Hambrick et al., 2015). Instead, the multiteam approach to strategic leadership pinpoints that environmental dynamism challenges a firm's ability to sustain superior performance, and therefore triggers both the TMT and BOD to work interdependently to meet this environmental demand (Luciano et al., 2020). However, the interactive effects of environmental dynamism with TMT and BOD gender diversity to influence innovation and firm performance have not been theoretically developed or empirically tested. We advance this stream of research by (a) developing arguments for how the effect of BOD diversity depends on TMT diversity and environmental dynamism, (b) stating how their joint effects influence proximal innovation and distal performance, and (c) demonstrating that proximal organizational innovation mediates the effect of TMT–BOD gender diversity on distal firm performance, especially in a dynamic environment. Our findings advance this stream of research by empirically testing what Luciano et al. (2020) believed might happen and replacing that speculation with empirical evidence that supports the multiteam approach to strategic leadership. Findings imply that diversity can be harnessed for innovation and that the match between TMT–BOD diversity and environmental dynamism are equally important.

We further advance the multiteam approach to strategic leadership of Luciano et al. (2020) by testing this theory across two different country contexts. In doing so, we both support and extend the theory. Support is found for the theory by demonstrating that in both China and the United Kingdom there is a strong positive relationship between women in the TMT and firm innovation when women are well represented on the BOD and environmental dynamism is also high. However, the shapes of the three-way interactions also differ somewhat in ways that suggest the multiteam strategic leadership theory may need to be refined to account for important country differences. For example, all other slopes in Figure 2 for the Chinese data are either flat (and not significantly different from zero) or negative in the case of a low women ratio on the BOD and low environmental dynamism. This means that in China, having a high ratio of women on the TMT makes no difference (or may be slightly counterproductive) without a high ratio of women on the BOD or environmental dynamism. By contrast, in Figure 4 for the UK data, all other slopes are either positive or flat. Specifically, in the United Kingdom there is still a slight positive slope of having a high ratio of women on the TMT positively predicting firm innovation when either environmental dynamism is high and ratio of women on the BOD is low, or vice versa, when environmental dynamism is low but the ratio of women on the BOD is high. The only time the slope is flat and a high ratio of women on the TMT makes no difference in the United Kingdom is when everything in the environment is stacked against innovation (low environmental dynamism combined with low women representation on the BOD). This implies some important country differences in the context (Johns, 2006) which must be considered in future theorizing applying the multiteam strategic leadership model of Luciano et al. (2020) across countries. We describe this further in the future research suggestions.

We conclude that the diversity-performance relationship should not only account for the various thresholds of diversity within the TMT and board as well as the boundaries within which the teams operate, but also potential mediating mechanisms. We go beyond the contingency arguments that are often explored within the diversity arena (Joshi & Roh, 2009; Richard et al., 2007) by simultaneously beginning to uncover intervening variables (Lawrence, 1997) and linking TMT and BOD gender diversity to firm performance via a relevant mechanism (Andrevski et al., 2014). This is an endeavor rarely undertaken in the upper echelons literature (Hoobler et al., 2018).

7.2 Practical implications

Although a 2016 study revealed that out of approximately 22,000 companies in over 91 countries, 11,000 of the organizations in the study did not have women representation in top leadership positions (Kossek & Buzzanell, 2018; Noland et al., 2016), the global trend toward a workplace composed of more gender diversity in the upper echelons continues. Therefore, there is a need to understand the effect of such demographic differences in the upper echelons (Glass & Cook, 2018). The present study reveals that cracks in the glass ceiling pertaining to gender diversity in both the TMT and board allow firms to realize a competitive advantage from their diversity through innovation. This research answers calls from academics and practitioners regarding the complicated and seemingly contradictory impacts of gender diversity in the upper echelons on organizational performance. The present study implies that managers competing in dynamic environmental contexts are more likely to gain innovation benefits if they are able to develop an appropriate degree of BOD and TMT gender diversity in the upper echelons of the organization.

An implication of our study is that managers should not expect to observe the innovation effectiveness they desire if they place all their efforts in only diversifying their TMT without the BOD or vice versa. This study goes beyond the naive approach to understanding diversity effects, which assumes only the TMT is instrumental in initiating innovation. The presence of interaction effects for TMT gender diversity and BOD gender diversity suggests that neglecting to diversify corporate boards where TMT gender diversity already exists (or vice versa) may lead to an incomplete understanding of innovation. The TMT–BOD gender diversity interface provides managers with a richer understanding of how gender diversity impacts innovation as opposed to studying the TMT or the BOD in isolation.

Results suggest that the most innovative firms are those that have higher levels of gender diversity among both the TMT and the BOD, and this is especially true when the firm operates under environmental dynamism. The association between BOD gender diversity and firm innovation has been demonstrated before (e.g., Miller & Triana, 2009), but in the present study we go beyond that finding by showing that gender diversity in the upper echelons of organizations has the greatest impact on firm innovation when the TMT and BOD are both diverse. Research on diversity in teams shows that groups of people who share demographic characteristics often develop trust (Dyer & Singh, 1998; Yli-Renko et al., 2001), identification (R. Ely, 1994), and liking between people (Tsui & O'Reilly, 1989). Therefore, it is likely that having more women (i.e., a more gender diverse group) in both the TMT and the BOD can facilitate social interaction between the two leadership teams which can spur knowledge exchange and facilitate innovation. For organizations, this means that firms with higher levels of gender diversity in both major leadership teams (TMT and BOD) can reap the most benefits from their employees' knowledge as it pertains to innovation.

Moreover, our results further show that environmental dynamism is a critical contextual factor that firms should be aware of as they attempt to enhance the effects of TMT and BOD diversity on organizational innovation and performance. According to Steven Johnson, a science author and media theorist, “If you look at history, innovation doesn't come just from giving people incentives; it comes from creating environments where their ideas can connect” (Isomäki, 2019). Dynamic environments appear to be fertile ground for innovation to happen, especially when a company's TMT and BOD are also diverse enough to have different perspectives and ideas represented. The ultimate role of human resources (HR) in organizations is to support managers and employees so the workforce can be productive and the organization can succeed (Gómez-Mejía et al., 2020). This study's findings suggest that if firms in dynamic environments want to innovate, they should cultivate a gender-diverse TMT and BOD by making necessary efforts to break the glass ceiling and providing opportunity for women to advance to the TMT and BOD, with each team being important for innovation and performance.

7.3 Limitations and future research

This study has limitations that offer opportunities for future research. First, although we examined innovation as a mediator, we did not capture other underlying processes often mentioned in the literature (e.g., elaboration of task-related information, relationship conflict, and task conflict among others) that might intervene between heterogeneity and organizational outcomes (Jehn et al., 1999). We cite group-level research that has reported such processes and make no claims of having measured those processes ourselves. Future research should consider additional knowledge-sharing processes that account for heterogeneity effects. Although work group research as well as some upper echelons studies have begun documenting a connection between diversity and group processes such as problem-solving capabilities and better decision-making (Miller & Triana, 2009; Smith et al., 1994), most firm performance outcome studies do not measure mediators, leaving the intervening mechanism question unaddressed. We attempted to take a step forward by modeling innovation as a key mediating mechanism.

Second, although our research provides an important first step to future research that may further unpack the outcomes of a multiteam leadership perspective of the TMT and BOD across various contexts in evidence from both China and the United Kingdom, future research should investigate whether our findings, that the TMT–BOD gender diversity interaction effect on innovation improves subsequent firm performance, particularly in dynamic environments, generalizes to more countries. It is possible that the strength and intensity of effects might vary across country. For example, in our Study 1 Chinese sample, we observed more positive effects on innovation when there was low gender diversity in the TMT and low gender diversity in the BOD under stable environmental conditions. We did not observe this effect in the Study 2 UK sample. This could imply that in certain countries, being congruent across independent variables (e.g., low TMT diversity, low BOD diversity, low dynamism OR high TMT diversity, high BOD diversity, high dynamism) could lead to positive outcomes even at low, low, low levels. In other words, despite innovation being maximized at high TMT diversity, high BOD diversity, and high dynamism, in certain countries we might observe more modest effects on innovation even when the independent variables are aligned at low levels.

One possible explanation for the subtle differences between the China and UK findings which future research may test is the influence of patriarchy in a culture. When environmental dynamism is low, and the ratio of women on the TMT and BOD are also low, this may represent a very traditional workforce demographic where patriarchy is not challenged. In market situations where dynamism is low, this may reduce conflict and produce enough innovation to satisfy a firm's competitive needs in that environment. Chinese tradition enforces patriarchy, because in China one tenet of Confucian belief is the Three Cardinal Guides and the Five Constant Virtues, emphasizing that females should be unconditionally subordinate to males. For example, a wife should be unconditionally subordinate to her husband, and men possess the decision role in the family team. China has also never had a female Emperor or head of state in its history. By contrast, the United Kingdom has had two different women Prime Ministers, Margaret Thatcher (served 1979–1990) and Theresa May (served 2016–2019) as well as Queen Elizabeth who has been monarch over 68 years since 1952. Although this tenet of patriarchy (and others) may certainly affect the impact of TMT and BOD gender heterogeneity on innovation, we do not know if the effects observed would be stronger or weaker in other national cultures. Future research should employ our framework in other countries and examine what role cultural values and practices have on the nature of effects in the multiteam TMT–BOD leadership framework.

Third, while this study shed light on the effects of diversity in human capital, it does not capture the variations in the ways that companies manage diversity. Organizations vary in the presence and effectiveness of human resource diversity practices such as diversity training, development, and mentoring (Richard et al., 2013). Effectively implemented diversity practices could strengthen the positive effects we found, just as ineffective ones could strengthen any potential negative effects that might arise from TMT heterogeneity and its interaction with board heterogeneity. Future research would benefit from exploring the role of HRM practices and a climate of diversity on the human capital effects we found (Nishii, 2013; M. Triana et al., 2021).