Executives’ postgraduate education and corporate ambidextrous innovation: evidence from China’s listed companies

Descriptive Statistics

Descriptive statistics were conducted on the collected data, and the results are presented in Table 2. This table includes the number of observations, mean, standard deviation, minimum, and maximum values for each variable. The key findings are as follows:

From Table 2, it is evident that the mean and standard deviation for exploratory innovation are 2.573 and 1.276, respectively, with a range from 0 to 6.217. This indicates considerable variation in the levels of exploratory innovation among the listed companies. In comparison, exploitative innovation exhibits a lower mean of 2.400 but a higher standard deviation of 1.336, suggesting that the firms in the sample do not demonstrate a clear preference for either exploratory or exploitative innovation.

The mean value for CEO postgraduate education is 0.366, indicating that 36.6% of the CEOs in the sample have completed postgraduate education. The mean and standard deviation for the postgraduate education level of the executive team are 0.415 and 0.232, respectively, with values ranging from 0 to 1. This reflects some variation in the level of postgraduate education among executive teams across firms.

For the moderating variables, the mean values for gender and age are 0.935 and 49.611, respectively. This suggests a high proportion of male CEOs in the sample, as well as an overall older age distribution, predominantly consisting of middle-aged executives. The gender and age characteristics of the executive teams are similar.

Regarding the control variables, the mean and standard deviation for firm size are 22.099 and 1.315, respectively, which fall within a reasonable range.

The results in the table indicate that the statistical analysis of each indicator did not reveal any significant anomalies. To further explore the relationships between variables and address potential multicollinearity issues in the regression analysis, a correlation coefficient test was conducted. The correlation coefficients are summarized in Table 3, where the upper triangle presents the Spearman correlation coefficients along with their significance levels, while the lower triangle displays the Pearson correlation coefficients and their significance levels.

From the Pearson coefficient matrix, it can be observed that the correlation coefficients between CEO postgraduate education and both exploratory and exploitative innovation are 0.133 and 0.067, respectively, both significant at the 1% level. This indicates a significant positive correlation between CEO postgraduate education and corporate ambidextrous innovation strategy. Similarly, the postgraduate education level of the executive team is significantly positively correlated with both exploratory and exploitative innovation at the 1% level, with correlation coefficients of 0.237 and 0.094, respectively. In comparison, the Spearman correlation coefficient matrix yields similar results, with no significant differences. Additionally, the correlation coefficients between the key explanatory variables and moderating variables are generally below 0.4, suggesting no significant multicollinearity issues.

Benchmark Regression Analysis

Using formulas (1) and (2), regression analysis was conducted on the relevant variables using a hierarchical regression method, controlling for industry and year fixed effects. The analysis was performed in two stages: first, the regression results of the key explanatory variables on the dependent variables were explored. Then, the full regression model, including all explanatory and control variables, was examined. The specific regression results are presented in Table 4.

Columns (1) and (2) of Table 4 present the impact of the CEO’s postgraduate education on exploratory innovation. The results show that, regardless of the inclusion of control variables, the regression coefficients for the CEO’s postgraduate education are significantly positive, with values of 0.130 and 0.082, respectively. The former is significant at the 1% level, while the latter is significant at the 5% level. These findings suggest that CEOs with postgraduate education can indeed foster exploratory innovation within firms, thus supporting Hypothesis 1.

Columns (3) and (4) examine the relationship between the executive team’s postgraduate education level and exploratory innovation. The regression results indicate a strong positive correlation between the executive team’s postgraduate education level and exploratory innovation, with correlation coefficients of 0.779 and 0.410, both significant at the 1% level, before and after controlling for other variables. This suggests that a higher level of postgraduate education within the executive team significantly promotes exploratory innovation, supporting Hypothesis 3.

Columns (5) to (8) present the relationship between the CEO’s postgraduate education, the executive team’s postgraduate education level, and exploitative innovation, both before and after the inclusion of control variables. The results indicate that neither the CEO’s postgraduate education nor the executive team’s postgraduate education level shows a significant positive effect on exploitative innovation. In comparison to exploitative innovation, the impact of executives’ postgraduate education on exploratory innovation is more pronounced.

At the level of control variables, firm size, ownership type, solvency, equity concentration, dual role status, and R&D expenditure all positively impact both exploratory and exploitative innovation. In contrast, firm age since listing and growth capability have a negative inhibitory effect on both types of innovation.

In summary, the postgraduate education of executives has a significant positive effect on exploratory innovation, while its impact on exploitative innovation is statistically insignificant. A comparison of the coefficients reveals that the postgraduate education level of the TMT has a stronger influence on ambidextrous innovation within firms than the CEO’s education alone. This may be due to executives with postgraduate training possessing greater psychological resilience and risk tolerance, which makes them more inclined toward high-risk, high-reward exploratory innovation. Moreover, the TMT’s diversity can further enhance cognitive variety, creating a cumulative effect that amplifies the innovation-promoting influence of postgraduate education.

Moderating Effects Analysis

The benchmark regression results show that CEOs with postgraduate education, as well as executive teams with higher levels of postgraduate education, are more effective in promoting exploratory innovation. However, this effect may be influenced by factors such as executive gender and age. To further investigate the impact of these factors on the innovation outcomes associated with postgraduate education, the study generates interaction terms by multiplying CEO postgraduate education with CEO age, executive team postgraduate education with average team age, CEO postgraduate education with CEO gender, and executive team postgraduate education with team gender composition. These interaction terms are incorporated into the models, which control for industry and year fixed effects. The regression results, presented in Tables 5 and 6, provide insights into the moderating effects of age and gender on the relationship between postgraduate education and corporate ambidextrous innovation.

Table 5 presents the regression results for models (3) and (4). Columns (1), (2), (5), and (6) validate the moderating effect of CEO gender on the relationship between CEO postgraduate education and corporate ambidextrous innovation. The coefficients of the interaction term Postgra_CEOgender* are positive, with values of 0.069 and 0.047 after including control variables. The former is significant at the 10% level, suggesting that when the CEO is male, the positive impact of the CEO’s educational background on the firm’s ambidextrous innovation is more pronounced, particularly in terms of exploratory innovation. However, exploitative innovation does not show significant effects in the interaction between CEO gender and postgraduate education.

Columns (3), (4), (7), and (8) explore the role of the gender composition of the TMT in moderating the relationship between the team’s postgraduate education level and ambidextrous innovation. The interaction term between the TMT’s postgraduate education level and its gender composition is positive and statistically significant at the 1% level for exploratory innovation. The coefficients for this interaction term are 0.893 and 0.448 before and after the inclusion of control variables, respectively, supporting Hypothesis 6a. For exploitative innovation, however, the moderating effect of the TMT’s average age is only significant at the 10% level when control variables are excluded, with a coefficient of 0.309, which is much lower than the effect observed for exploratory innovation.

This result suggests that gender plays a stronger moderating role in enhancing exploratory innovation through the postgraduate education of the TMT. In firms led by male CEOs or those with a higher proportion of male members in the TMT, the positive impact of postgraduate education on exploratory innovation is more pronounced, fostering a stronger drive towards exploratory innovation within the organization.

Based on regression models (5) and (6), Table 6 presents the moderating effect of executive age on the relationship between CEO postgraduate education and corporate ambidextrous innovation. Columns (1) to (4) focus on the effect of age on the relationship between postgraduate education and exploratory innovation, while columns (5) to (8) examine the effect of age on the relationship with exploitative innovation.

The results show that for exploratory innovation, both the interaction terms Postgra_CEO*age and Postgra*mage have significantly positive coefficients of 0.001 and 0.008, respectively, after including control variables, with a significance level of 10% or higher. This suggests that the CEO’s age positively moderates the effect of postgraduate education on exploratory innovation. Specifically, as the age of executives increases, the positive impact of their postgraduate education on exploratory innovation becomes stronger. Similarly, the average age of the TMT also positively moderates the relationship between the team’s postgraduate education level and exploratory innovation. The older the TMT, the more pronounced the effect of their postgraduate education on enhancing ambidextrous innovation.

However, for exploitative innovation, the coefficients of Postgra_CEO*age and Postgra*mage are 0.000 and −0.004, respectively, after including control variables, but neither is statistically significant. This indicates that age does not significantly moderate the relationship between postgraduate education and exploitative innovation.

Comparing the moderating effects of age on the relationship between postgraduate education and exploratory versus exploitative innovation, it is evident that age significantly moderates the relationship with exploratory innovation, both from the CEO and executive team perspectives. This result can be partly explained by the “overconfidence” hypothesis (Steinberg et al. 2022; Mehraein et al. 2023). As executives age, their interpersonal networks and social capital expand, which can enhance their confidence. In some cases, this heightened confidence may even lead to overconfidence, making them more inclined to favor high-risk, high-reward exploratory innovation strategies.

In summary, CEO gender positively enhances the interaction between the CEO’s postgraduate education and exploratory innovation. Similarly, the gender composition of the TMT plays a significant moderating role in the relationship between the team’s postgraduate education level and exploratory innovation, further strengthening the impact of postgraduate education on exploratory innovation. In contrast, neither CEO gender nor the gender composition of the TMT significantly moderates the effect of postgraduate education on exploitative innovation. Similar to gender, age also significantly interacts with postgraduate education to influence ambidextrous innovation. CEO age strengthens the positive relationship between the CEO’s postgraduate education and exploratory innovation, while the average age of the TMT positively moderates the effect of the team’s postgraduate education on exploratory innovation. As the average age of the TMT increases, the positive impact of postgraduate education on exploratory innovation becomes more pronounced.

Heterogeneity Analysis

Given the unique institutional context in China, executive structures and innovation outcomes may vary between state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs). SOEs typically carry greater political and social responsibilities and tend to prioritize operational stability, which makes them more risk-averse. To account for these differences, the sample was divided into SOEs and non-SOEs, with executive age and gender included as control variables at the executive level. Additionally, year and industry fixed effects were controlled for. The effect of executive postgraduate education on ambidextrous innovation was then analyzed separately for these two groups. The results of the grouped regression analysis are presented in Table 7.

Columns (1) and (2) of Table 7 present the regression results for the impact of the CEO’s postgraduate education on exploratory and exploitative innovation in SOEs. The coefficients for exploratory and exploitative innovation are 0.039 and 0.121, respectively, but neither is statistically significant. In contrast, Columns (5) and (6) show the relationship between the CEO’s postgraduate education and ambidextrous innovation in non-SOEs. Here, the coefficient for exploratory innovation is 0.103, which is significant at the 5% level. These results suggest that, in the context of SOEs, the CEO’s influence on the firm’s innovation decisions is limited. This may be due to the inherent stability and low-risk strategies that SOEs tend to prioritize. In non-SOEs, however, the CEO’s postgraduate education has a notable positive effect on exploratory innovation, while it shows no significant impact on exploitative innovation.

Columns (3) and (4) and Columns (7) and (8) further analyze the effect of the TMT’s postgraduate education level on ambidextrous innovation in both SOEs and non-SOEs. A comparison of the relevant coefficients and their significance levels reveals no substantial difference between the two types of enterprises. Regardless of ownership structure, the TMT’s postgraduate education level is positively correlated with exploratory innovation, with significance levels of 5% or higher. This suggests that a higher postgraduate education level within the TMT has a stronger stimulating effect on exploratory innovation. However, the regression coefficients for the TMT’s postgraduate education level in relation to exploitative innovation are statistically insignificant in both SOEs and non-SOEs.

One possible explanation for this pattern is that as the proportion of postgraduate-educated members in the TMT increases, their influence and decision-making power within the team are further consolidated. This consolidation allows the cognitive “imprints” shaped by postgraduate education to more effectively manifest in the firm’s innovation strategy, with their innovation mindset and risk-taking propensity becoming more prominently reflected in strategic decisions. In contrast, exploitative innovation typically requires lower levels of innovative thinking and technical skills, making it a more attainable goal for most firms. Due to its larger base and existing stock, exploitative innovation is less susceptible to influence from the firm’s ownership structure or the decision-making preferences of the TMT.

Robustness Checks

Instrumental Variables

Endogeneity is a common concern in studies examining corporate governance and performance (Schultz et al. 2010; Wintoki et al. 2012). Although this study controls for year and industry fixed effects to mitigate the influence of time-invariant omitted variables, potential reverse causality issues may still arise. Specifically, a firm’s anticipated ambidextrous innovation capabilities could influence the recruitment and composition of the TMT. This suggests that instead of innovation being solely driven by the characteristics of the management team, the firm’s strategic innovation goals might also shape the selection and structure of its leadership.

To address this endogeneity concern, this study adopts the instrumental variable (IV) approach used in prior research (Yang et al. 2023; You et al. 2023), aiming to provide more robust estimates by accounting for potential reverse causality and omitted variable biases, thereby strengthening the validity of the results. The chosen instrument must satisfy two conditions: it should be correlated with the endogenous explanatory variable and uncorrelated with the error term. The study uses the CEO’s prior experience in R&D (CEO_RD) as the instrumental variable, which is highly correlated with the CEO’s postgraduate education (Wang et al. 2019). According to the “China Science and Technology Statistics Yearbook,” 36% and 18% of personnel employed by R&D institutions hold master’s and doctoral degrees, respectively, significantly higher than their proportions in the general labor market. Moreover, the CEO’s past R&D experience is unlikely to directly affect the dual innovation outcomes of listed companies, making it a valid instrument.

The first stage of the regression involves control variables such as dual role, firm size, firm age, ownership type, ownership concentration, leverage, growth capability, and corporate R&D expenditure, along with industry and year fixed effects. In this stage, the CEO_RD is regressed on the CEO’s postgraduate education. The predicted values from this first stage are then incorporated into the second stage, where they are used to regress the dependent variables—exploratory and exploitative innovation—yielding the Two-Stage Least Squares (2SLS) regression results.

Columns (1) and (3) of Table 8 report the effects of the CEO’s postgraduate education on exploratory and exploitative innovation using robust standard errors. The regression coefficients are 0.082 and 0.039, respectively, with the coefficient for exploratory innovation being significant at the 5% level. These results indicate that the CEO’s postgraduate education has a statistically significant positive impact on exploratory innovation, while its effect on exploitative innovation is statistically insignificant. Columns (2) and (4) present the second-stage 2SLS regression results. In the first stage (not reported), the coefficient of CEO_RD is significantly positive at the 1% level. The second-stage results show that the CEO’s postgraduate education continues to have a significantly positive impact on exploratory innovation, with the coefficient increasing to 0.843 compared to Column (1). This suggests the potential presence of measurement error (Shabir et al. 2022). In contrast, the effect of the CEO’s postgraduate education on exploitative innovation remains negative and statistically insignificant.

Overall, these findings confirm the robustness of the initial results. Additionally, weak instrument tests for CEO_RD show F-statistics (F = 21.751 for exploratory innovation and F = 15.830 for exploitative innovation) that exceed the usual threshold, confirming that CEO_RD is a strong instrument and further reinforcing the robustness of the conclusions. The instrumental variable regression results corroborate the previous conclusion that highly educated talents in the TMT contribute more to exploratory innovation than to exploitative innovation, strengthening the credibility of this finding.

Replacing Control Variables

In the earlier analysis, R&D expenditure was selected as a key control variable to represent the firm’s R&D efforts. Another commonly used indicator for measuring a firm’s R&D activities is the number of R&D personnel. Given the high correlation between these two variables in the context of innovation decision-making, they were not included in the model simultaneously. Therefore, we replaced the original control variable, R&D expenditure (design1), with the number of R&D personnel (design2) and re-conducted the regression analysis.

After replacing the measure of R&D input, the CEO’s postgraduate education continues to have a significant positive effect on exploratory innovation (β = 0.081, p < 0.1). Similarly, the postgraduate education level of the TMT remains significantly positively correlated with exploratory innovation (β = 0.548, p < 0.01). The effect on exploitative innovation remains insignificant, consistent with the previous regression results (see Table S2 for details).

Replacing the Measurement of the Dependent Variable

To ensure the validity and reliability of the research conclusions, this study follows the approach of Dosi et al. (2006) by using the number of granted patents for robustness testing. Exploratory innovation is quantified as the natural logarithm of the number of invention patent grants plus one, denoted as Lninno_3. Exploitative innovation is defined as the natural logarithm of the number of utility model and design patent grants plus one, denoted as Lninno_4. The results are presented in Table 9.

The findings indicate that, even after replacing the dependent variables, the postgraduate education level of the TMT continues to significantly promote exploratory innovation at the 1% level. However, its effect on exploitative innovation remains statistically insignificant. Overall, the previous regression results are robust.

Lagged Time Periods

Considering the complexity involved in the TMT’s process of receiving and processing dynamic market information, as well as the inherent lag associated with using corporate patent data to represent ambidextrous innovation capabilities, this study introduces a one-period lag in the patent data. This adjustment allows for a re-examination of the effects of the CEO’s postgraduate education and the TMT’s postgraduate education level on corporate ambidextrous innovation.

The regression results for lagged patent application data are presented in Columns (1) to (4) of Table 10, while those for lagged patent grant data appear in Columns (5) to (8). The findings show that when the dependent variable is the number of patent applications, the regression coefficients for the CEO’s postgraduate education and the TMT’s postgraduate education level on lagged exploratory innovation are 0.103 and 0.362, respectively, both significant at the 1% level. However, their impact on exploitative innovation is not significant, indicating that the effect of the CEO’s and TMT’s postgraduate education is more pronounced for exploratory innovation than for exploitative innovation. This result aligns with the previous baseline regression findings.

When the dependent variable is the number of patent grants, the TMT’s postgraduate education level continues to have a significant positive impact on exploratory innovation, while its influence on exploitative innovation remains limited. These findings highlight the greater relevance of the CEO’s and TMT’s educational background in promoting exploratory innovation over exploitative innovation, further reinforcing the robustness of the study’s results.