4.1. Variable Description
The results of the variable description are shown in
Table 2, where Panel A reports descriptive statistics and Panel B reports the results of mean-difference tests. Panel A shows that: (1) The standard deviation of the variable
Growth was up to 1.8696, indicating that there was a certain difference in the growth of the sample enterprises. (2) The coefficient of variation of the variable
BHC was 243.71%, indicating that the sample had strong volatility. This was because the standard deviation of the variable
BHC was large but its mean value was small, which indicated that the board of directors of the listed companies in China had highly different levels of human capital. However, due to the level of human capital being generally low, once listed companies raise the level of board human capital through various methods, such as recruitment of talents and organization of training, they will make significant differences relative to most listed companies, which causes the high volatility of board human capital. (3) The mean value of the variable
DEI was greater than 2.5 (a postgraduate degree was worth 3), and the coefficient of variation was 20.34%, indicating that a board’s education level of Chinese listed companies was relatively concentrated, and the average education level was close to a graduate degree. (4) The mean value of
BAC was 0.2742, which indicated that about one-third of the directors in the sample enterprises had experience in R&D, design, or marketing. (5) The standard deviation of the variable
SEA was 0.1129, but the coefficient of variation was 124.2%, indicating that the sample was highly volatile. This was caused by the small mean value of the variable
SEA, which indicates that the overseas experience of directors of Chinese-listed companies is relatively concentrated. However, due to its low average proportion, once a company had a large number of returnees, it was significantly different from most companies and its volatility was greater. (6) The coefficient of variation of the variable
IOTTOU was 233.15%, which was much larger than the coefficient of variation of the variable
IOTTOE, which was 110.77%. That is to say, although the standard deviation of the variable
IOTTOE was large, the variable
IOTTOU had a large variability with respect to its mean value, indicating that there were some differences in the level of exploitative innovation of listed companies in China compared with exploratory innovation, the average level of exploitative innovation R&D investment was low, and enterprises were more likely to catch up with innovation by increasing the investment; as such, the volatility of exploitative innovation was higher. (7) The standard deviation of the variable
SIZE was 1.2852, but its coefficient of variation was only 0.0582, which was related to the large mean value of the variable
SIZE. This indicates that there were large differences in the size of sample companies, and due to the generally high average size of enterprises, it was difficult for smaller enterprises to expand their size in a short time; as such, their volatility was relatively small. Panel B divides the samples into two groups according to the average value of board human capital, and tests the differences between them. The mean-difference test results show that the higher the level of board human capital is, the greater the enterprise growth was, consistent with the hypotheses developed earlier.
4.3. Multiple Regression Results and Discussion
OLS regression analysis was used to analyze the relationship between board human capital, ambidextrous innovation, and enterprise growth (
Table 4 and
Table 5).
The Relationship between Board Human Capital and Enterprise Growth. First, we examined the impact of board human capital on enterprise growth (Model 1) and report the results in
Table 4. In columns (1) and (2) of
Table 4, we show that the variable
BHC was positively correlated with
Growth (
t = 10.0344,
p < 0.01). That is, board human capital positively affected enterprise growth at a significant level of 1%, indicating that the higher the level of board human capital, the more beneficial it was to enterprise growth. From an economic perspective, every standard deviation increase of board human capital (1.9780) resulted in an enterprise growth improvement equivalent to 6.42% (= 0.0607 × 1.9780/1.8696) of the sample standard deviation. It can be seen that board human capital had a significant positive relationship with enterprise growth in both a statistical and economic sense. Therefore, H1 was supported. Further, when examining the impact of the sub-index of board human capital on enterprise growth, we found that the estimated coefficient of the variable
DEI was positive (
t = 9.4364,
p < 0.01), and a board’s education level had a positive impact on enterprise growth at a significance level of 1%. From an economic perspective, every standard deviation increase of a board’s education level (0.5178) resulted in an enterprise growth improvement equivalent to 6.01% (0.2170 × 0.5178/1.8696) of the sample standard deviation. Therefore, a board’s education level had a significantly positive impact on the enterprise growth, and H1a was supported. The estimated coefficient of variable
BAC was significantly negative (
t = −3.7042,
p < 0.01), indicating that the more directors with an R&D, design, or marketing background on the board, the less likely an enterprise grew. This is inconsistent with the theoretical derivation, which may be due to the fact that the board of directors is composed of members from different professional fields that can hardly reach a unified opinion on the various opinions and suggestions provided during the meeting [
47]. Moreover, the confusion and problems, such as communication, will appear frequently, and the emergence of these problems is not conducive to the formulation and implementation of enterprise strategic decision-making, thus threatening the survival and development of enterprises. Furthermore, the increasingly fierce global economic integration and market competition put forward higher requirements for the standardized management of enterprises, which not only requires senior executives to have a certain professional background, but also requires them to have professional management knowledge, management art, and organizational ability. When the number of directors with a background in “output function” in the board was more, it also meant the number of directors with professional management experience was less, which was not conducive to the development of enterprises in the long run. H1b was therefore rejected. The variable
SEA was positively correlated with
Growth (
t = 7.1741,
p < 0.01), where a board’s overseas experience had a positive impact on enterprise growth at a significance level of 1%. From an economic perspective, every standard deviation increase of a board’s overseas experience (0.1129) resulted in an enterprise growth improvement equivalent to 4.60% (= 0.7625 × 0.1129/1.8696) of the sample standard deviation. To sum up, the more directors with overseas educational and work experience in the board, the more conducive it was to enterprise growth. H1c was therefore supported.
The Relationship between Board Human Capital and Ambidextrous Innovation. According to Model 2, the relationship was tested between board human capital and ambidextrous innovation, and the estimated results are shown in
Table 4. In columns (3)–(6) of
Table 4, we show that Model 2 contained the main effect of board human capital and ambidextrous innovation, which was mainly used to verify hypothesis H2. The results show that the variable
BHC was positively correlated with
IOTTOU and
IOTTOE (
t = 5.6093,
p < 0.01;
t = 4.0222,
p < 0.01), where board human capital positively affected ambidextrous innovation at a 1% significance level, indicating that the higher the level of board human capital, the more conducive it was toward improving the level of corporate binary innovation. Therefore, H2 and H3 were supported. Due to the difference between exploitation innovation and exploratory innovation in the demand for resources and capabilities, we further studied the impact of the sub-index of board human capital on ambidextrous innovation, and found that the variables
DEI and
SEA were positively correlated with
IOTTOU (
t = 4.5698,
p < 0.01;
t = 2.6023,
p < 0.01), but
BAC was not significantly correlated with
IOTTOU (
t = 0.6007,
p > 0.1). This suggests that a board’s education level and overseas experience have a positive impact on the exploitation innovation at a significance level of 1%. This shows that the higher the education level of directors, the more overseas experience directors have, the more conducive to improving the utilization of existing knowledge and technology, leading to achieving technological progress and product optimization, and promoting the level of exploitation innovation. However, a board’s professional background had no significant positive correlation with exploitation innovation, which is inconsistent with the theoretical derivation. This may be related to the nature of exploitation innovation. Exploitation innovation requires sufficient knowledge and skills to improve and optimize existing products, which is independent of whether directors have the background of "export function." It can be seen that H2a and H2c were supported and H2b was rejected. The variables
BAC and
SEA were positively correlated with
IOTTOE (
t = 4.3086,
p < 0.01;
t = 2.3088,
p < 0.05), while
DEI was negatively correlated with
IOTTOE (
t = −0.5592,
p > 0.1), suggesting that a board’s professional background and overseas experience had a positive impact on exploratory innovation at a significance level of 1% and 5%, respectively, indicating that the more background in “output function” and overseas experience directors had, the more conducive they were to acquiring new knowledge and skills and to achieving a breakthrough innovation. However, a board’s education level was negatively correlated with exploratory innovation and was not significant, which is inconsistent with the theoretical deduction. This may be related to the highly educated directors’ strong sense of risk aversion and lower participation in high-risk, high-return R&D and investment projects. Therefore, H3b and H3c were supported, and H3a was rejected.
The Relationship between Ambidextrous Innovation and Enterprise Growth. The relationship between ambidextrous innovation and enterprise growth was tested according to Models 3 and 4, and the estimated results are shown in
Table 5. In the columns (1)–(4) of
Table 5, we show that Model 3 mainly validated whether ambidextrous innovation can promote enterprise growth, and was used to test H4a and H4b. The regression results show that both
IOTTOU and
IOTTOE were positively correlated with
Growth (
t = 5.7919,
p < 0.01;
t = 4.0133,
p < 0.01), suggesting that both exploitation innovation and exploratory innovation had a positive impact on enterprise growth at a significance level of 1%. Therefore, H4a and H4b were supported. Model 4 mainly verified the influence of the matching mode of binary innovation on enterprise growth, which was used to test H5a and H5b. The regression results show that the balance effect of exploitation innovation and exploratory innovation had a weak but significantly positive impact on enterprise growth (
t = 2.0907,
p < 0.05), while the product effect had little significant association with enterprise growth (
t = 0.1715,
p > 0.1). Therefore, H5a was supported and H5b was rejected.
The Mediating Effect of Ambidextrous Innovation. Using the methods of Wen et al. [
48] and combining Models 1, 2, and 5, we tested whether ambidextrous innovation had an intermediary effect on the relationship between board human capital and enterprise growth according to the following procedures. The first step was to test whether the coefficient
of Model 1 was statistically significant. If the subsequent test was established, it was a mediating effect; otherwise, it was a masking effect. In the second step, the coefficient
of Model 2 and the coefficient
of Model 5 were tested successively. If both were significant, it was an indirect effect, and the fourth step was carried out; otherwise, the next step was carried out. In the third step, we used the bootstrap method to test
. If the result was significant, it was an indirect effect, and then proceeded to the next step; otherwise, the indirect effect was not significant and we stopped the analysis. The fourth step was to test the coefficient
of Model 5. If it was significant, the direct effect was significant and the next step was carried out; otherwise, the direct effect was not significant and only the intermediary effect existed. The fifth step was to compare
and
. If the sign was the same, they belonged to a partial mediation effect; if not, it was a masking effect. Based on the above research methods and the regression results in columns (1)–(9) of
Table 4, we found that in the previous analysis, it was known that
and
were significant. After adding exploitation innovation into the regression equation, the regression coefficient of exploitation innovation on enterprise growth in column (7) was positively correlated at the 1% significance level (
t = 5.4379,
p < 0.01) where
was significant. Moreover, by comparing columns (1) and (7), the regression coefficient of the board human capital was reduced to some degree, but it was still positively correlated at the significant level of 1% (
t = 5.3094,
p < 0.01); that is,
was significant. Since the signs of
and
were the same, exploitation innovation played an intermediary role in the relationship between board human capital and enterprise growth. Next, we considered the mediating effects of exploratory innovation. Similar to the analysis of the mediating effect of exploitation innovation, exploratory innovation also played a partial mediating role in the relationship between board human capital and enterprise growth. Therefore, ambidextrous innovation played an intermediary role in board human capital and enterprise growth, and H6 was supported.
Further, we examined the mediating effect of ambidextrous innovation on the index of board human capital and enterprise growth, and the results are shown in columns (2)–(10) of
Table 4. We found that exploitation innovation played part of the mediating effect on a board’s education level and enterprise growth, while exploratory innovation had no significant indirect effect on a board’s education level and enterprise growth. The indirect effect of exploitation innovation on a board’s professional background and enterprise growth was not significant, and exploratory innovation played a covering effect on a board’s professional background and enterprise growth. Ambidextrous innovation played part of the mediating effect between a board’s overseas experience and enterprise growth (Considering the mediating effect of
IOTTOE on
DEI and
Growth, since
, the bootstrap method was used to test and
p = 0.441 was obtained. Therefore, the indirect effect was not significant. Considering the mediating effect of
IOTTOE between
BAC and
Growth, we adopted the bootstrap test and obtained
p = 0.000; that is, the indirect effect was significant, but there was a masking effect due to the difference between
and
. Similarly, the indirect effect of
IOTTOU on
BAC and
Growth was not significant). In conclusion, H6c was supported, H6a was partially supported, and H6b was rejected.
4.4. Robustness Test
In order to verify the reliability of the regression results, we conducted the robustness tests from the following aspects.
Problem with Measurement Error. The growth rate of business income was adopted as a substitute indicator for growth. Considering the cumulative effect of board human capital, we take t+1 period and t+2 period for enterprise growth, and then conducted the regression of the main effect of board human capital on enterprise growth. The specific results of the indicator replacement are shown in columns (1) and (2) of
Table 6. We found no matter whether t+1 period or t+2 period was applied, board human capital still had a significantly positive impact on enterprise growth (
t = 2.6780,
p < 0.01;
t =2.4669,
p < 0.05).
Problems with Legacy Variables. Since the characteristics of enterprises also affected enterprise growth, there could be a problem of legacy variables; therefore, we adopted the placebo test to examine this possibility. Specifically, we randomly sorted board human capital in the sample before the regression analysis. If the result after regression was no longer significant, it indicated that board human capital was related to enterprise growth. On the contrary, if the result was still significant, it indicated that the basic regression result may have been caused by omitted variables. The results of the placebo test are shown in column (3) of
Table 6, with little association between board human capital after reordering and enterprise growth, indicating that board human capital promoted enterprise growth, and the results were not caused by other omitted variables.
Problems with Reverse Causality. In the theoretical derivation part, it was concluded that the higher the board human capital is, the better it is at improving enterprise growth. However, enterprises with higher growth may also tend to choose directors with a higher human capital level to form the board of directors. There may be a causal endogeneity problem between board human capital and enterprise growth. In order to reduce the interference of endogeneity problems, we chose the average board human capital grouped by year and industry as the instrumental variable (IV). The average board human capital in the same industry and year was related to board human capital but did not affect enterprise growth. Subsequently, we used a two-stage least squares regression (
2SLS) to test the weak instrumental variables. The results showed that
Shea’s Partial R2 was less than 0.01, but the
F statistic was 85.7294, and the
p-value of the
F statistic is 0. Therefore, the average board human capital was not a weak instrumental variable. Column (4) in
Table 6 gives the regression results of the first stage, showing that the estimated coefficient of the instrumental variable (IV) was significantly positive (
t = 0.1000,
p < 0.01). Column (5) is the second-stage regression result, showing that even after considering endogeneity problems, board human capital was still significantly positively associated with enterprise growth at significance the level of 1% (
t = 0.0790,
p < 0.01), suggesting that board human capital had a significantly positive impact on enterprise growth.