Companies listed on China's A-share stock market (traded on the Shanghai Stock Exchange and Shenzhen Stock Exchange) from 2008 to 2020 were selected as the research sample. The sample was preprocessed as follows: (1) Listed companies in the finance, insurance, and real estate sectors were excluded. (2) ST and *ST shares were excluded. (3) Companies with missing data were excluded. (4) Companies that had not released information about investment in patent development and application were excluded. (5) To avoid extreme values, the sample data were winsorized at 1%. A total of 18497 companies were obtained for the final sample. Data on patents were manually collected from the official website of the China National Intellectual Property Administration, while other data were obtained from the China Stock Market & Accounting Research (CSMAR) database. Some missing data on returnee CEOs were manually collected from Baidu, Sina, and individual enterprises’ official websites.

Following previous authors, a new model is constructed to evaluate the impact of a CEO's overseas experience on corporate innovation performance:

To verify the mediating effects of disclosure quality and R&D investment, we adopt a three-pronged approach: First, we evaluate whether a CEO's overseas experience can significantly increase an enterprise's innovation output. Second, we assess whether a CEO's overseas experience can promote R&D investment and disclosure quality. Third, we identify how the interactions between a CEO's overseas experience and R&D investment and disclosure quality simultaneously affect corporate innovation performance. Therefore, we construct the following models:

Table 2 presents a descriptive statistical analysis of the variables. The sample companies' average innovation output is 1.865, indicating an increased emphasis on corporate innovation; however, the gap between the maximum of 6.068 and minimum of 0.000 is large, which reveals varying degrees of engagement in innovation and a differentiated level of innovation output among the sample enterprises as well as a lack of motivation for innovation among some enterprises. The average value for a CEO's overseas experience is 0.137, which means that only a few CEOs among the sample enterprises have been abroad. The average LEV is 0.465, which indicates a high asset-liability ratio for the sample enterprises; the average INBOARD is 2.162. The average number of members on a board of directors for the sample enterprises is eight, with a maximum of 2.708 and a minimum of 1.609, which means that board size can reach a maximum of 14 members and a minimum of eight members. This reflects a large difference in board size among the sample enterprises. The average FEMALE is 0.143, which means that most sample enterprises have a small proportion of female executives. However, the maximum value of FEMALE is 0.600, while the minimum is 0.000, which shows a substantially differentiated degree of emphasis on female executives among the sample enterprises.

Table 3 shows the correlations among the major variables; all the correlation coefficients are <0.5, which rules out the problem of multilinearity. A CEO's overseas experience (CEOsea) is positively and significantly correlated with corporate innovation performance (PA1) at the 1% level, which supports H1. The correlation coefficients between CEOsea and disclosure quality (KV) and R&D investment (RD1) are 0.060 and 0.048, respectively, which are significant at the 1% level. The correlation coefficients between PA1 and KVand RD1 are 0.173 and 0.426, respectively, and are significant at the 1% level, which supports H2. Nonetheless, further regression analysis is required to validate the correlation analysis outcomes.

Table 4 shows the regression results for H1. Column (1) in Table 4 lists only CEOsea, and shows that the correlation between CEOsea and PA1 is positive and significant at the 1% level; Column (2) displays the regression results after the introduction of the control variables, and the correlation coefficient is positive and significant at the 1% level; Column (3) displays the regression outcome after the introduction of the dummy variables for industry and year: CEOsea and PA1 are positively and significantly correlated at the 1% level, which supports H1 and indicates the positive role of CEOs’ overseas experience in promoting corporate innovation performance.

In addition, many factors contribute to corporate innovation performance. As shown in Column (3) in Table 4, the variable, LEV, is negatively and significantly correlated with PA1 at the 1% level, which means that corporation innovation performance decreases with an increasing asset-liability ratio. With a high asset-liability ratio, an enterprise will attach less importance to innovation and thus show poorer innovation performance. SIZEis positively and significantly correlated with PA1 at the 1% level, whereas AGEis negatively and significantly correlated with PA1 at the 1% level. This may be because younger enterprises have more room for trial and error in innovation attempts, whereas older companies are more financially capable of investing in innovative projects. INDENP is positively and significantly correlated with PA1 at the 10% level. In enterprises with a more independent board of directors, management have more power, while there will be less hindrance in starting innovative projects; meanwhile, the social connections of a more independent board of directors are more complex, while such directors have more domain-specific expertise to solve problems in innovation attempts. EXCUHLDN is positively and significantly correlated with PA1 at the 1% level: in companies with more shareholding executives, the top management have stronger control over the company and hence suffer fewer obstacles in their innovation attempts.

There might be an endogeneity problem in the correlation between CEOs’ overseas experience and corporate innovation performance because innovative companies prefer job applicants with overseas experiences. To test for the endogeneity, we selected a variable for treaty port cities in the late Qing Dynasty (Local) as the instrumental variable. Local=1 when an enterprise is headquartered in a treaty port city (Shanghai, Beijing, Tianjin, Fujian, Hubei, Shandong, Jiangsu, Guangdong, or Jiangxi); otherwise, Local=0 (Ang et al., 2014). In Table 5, Column (1) shows the result of the first-step regression. This instrumental variable (Local) is positively and significantly correlated with PA1 at the 1% level, which is consistent with our assumption. Column (2) in Table 5 presents the regression result when the endogeneity problem is accounted for, and reveals that CEOsea remains positively correlated with PA1 at the 1% significance level; this means that CEOs’ overseas experience contributes positively to corporate innovation performance, and that H1 still holds.

Column (1) of Table 7 verifies the impact of CEOs’ overseas experience (CEOsea) on disclosure quality (KV); the result shows that the correlation between CEOsea and PA1 is positive and significant at the 5% level. The regression results show that CEOs’ overseas experience can improve disclosure quality, which validates H2. Column (2) of Table 7 shows the results of the test for the impact of CEOs’ overseas experience and disclosure quality on corporate innovation performance. In this case, the correlation between CEOsea and PA1 is positive and significant at the 1% level, which suggests a direct impact of the former on corporate innovation performance, consistent with the result shown in Column (3) of Table 4. KV and PA1 are positively and significantly correlated at the 1% level, implying a positive impact of disclosure quality on corporate innovation performance. As shown in the aforementioned analyses, disclosure quality plays a mediating role in the relationship between CEO's overseas experience and corporate innovation performance.

In Table 7, Model (3) verifies the impact of CEOs’ overseas experience on R&D investment (RD1); the result shows that CEOsea is positively and significantly correlated with RD1 at the 1% level. The result reveals that CEOs’ overseas experience can increase R&D investment, which supports H2. Column (4) in Table 7 shows the result of the test for the impacts of CEOsea and RD1 on PA1. In this case, CEOsea is positively and significantly correlated with PA1 at the 1% level (with a regression coefficient of 0.006), which suggests a direct impact of the former on corporate innovation performance. This is consistent with the result in Column (3) in Table 4. RD1 and PA1 are positively and significantly correlated at the 1% level, which means that R&D investment positively affects an organization's innovation performance. Given the analyses above and the results of the mediating-effect tests, disclosure quality and R&D investment mediate the impact of CEOs’ overseas experience on organizational innovation performance.

Compared with those without overseas experiences, returnee CEOs can considerably increase the number of invention patents and patent applications. This is because organizations employing returnee CEOs attach more importance to technical innovation; a returnee CEO with both theoretical and technical expertise often has the final say in decision-making and can fulfill an organization's needs for innovation. Compared with UMPs and design patents, invention patents are more widely recognized and hence are more valued by returnee CEOs. In the present study, invention patents are distinguished from the other two types of patents. We use the natural logarithm of 1 plus the number of licensed invention patents (PA3) to measure the innovation output of invention patents, as shown in Column (1); the natural logarithm of 1 plus the number of licensed UMPs and design patents (PA4) is used to measure the innovation output of UMPs and design patents, as shown in Column (2). The control as well as year and industry dummy variables are introduced into the model.

Table 8 presents the verification results. Column (1) shows that the correlation between CEOsea and PA3 is significant at the 1% level (with a regression coefficient of 0.118); as Column (2) shows, the correlation between CEOsea and PA4 is significant at the 5% level (with a regression coefficient of 0.080). The considerable difference between the two coefficients indicates that returnee CEOs play a more salient role in promoting invention patents than in increasing UMPs and design patents, which is consistent with our assumption.

High-tech firms are the driving force for future innovation; the high-tech sector, which is financially powerful and promotes enterprise incubation and industry agglomeration, holds great appeal to returnees and is thus the cradle of innovation. In contrast, non-high-tech firms that boast cost advantages find it challenging to survive in a fiercely competitive environment; these firms are rather conservative and require governmental support; thus, they show little appeal to talents and have limited capacities for innovation. In the present study, whether a firm is a high-tech firm is used as a control variable in the regression model: if a firm is high-tech, the variable assumes a value of 1, as in Column (1); otherwise, the variable assumes a value of 0, as in Column (2). The control as well as year and industry dummy variables are controlled for.

As Column (1) in Table 9 shows, CEOsea shows a positive and significant correlation at the 1% level (with a regression coefficient of 0.131); however, in Column (2), the correlation coefficient of CEOsea is not significant, which means that the impact of returnee CEOs on corporate innovation performance is mainly felt in high-tech firms.

Internal control restrains and balances organizational behaviors in corporate management to ensure a stable operation. Internal control, as a procedure that influences every aspect of an enterprise, is often compulsory and requires deference from all members of the enterprise, from the chair to the general employees. Furthermore, well-organized internal control can boost investment in R&D in an organization and hence promote corporate innovation performance. Thus, in the present study, a dummy variable for internal control is introduced: sample enterprises whose values are larger than the median of the internal control variable are marked as 1, as in Column (1), whereas those with values lower than the median are marked as 0, as in Column (2). The other variables, including the year and industry dummies, are controlled for in the model.

Table 10 shows the test results. In Column (1), CEOsea shows a positive and significant correlation with corporate innovation performance at the 1% level (with a regression coefficient of 0.113); in Column (2), the correlation of CEOsea with corporate innovation performance is positive and significant at the 5% level (with a regression coefficient of 0.104). This means that as internal control strengthens, an organization will experience a higher demand for innovation and will be more likely to invest more in innovative projects.