FDI Spillovers on Corporate Social Responsibility: The Channel of Labor Mobility

: This paper examines the spillover effects of foreign direct investment on corporate social responsibility of domestic ﬁrms, with a focus on the channel of labor mobility. With a corporate social responsibility measure based on corporate responsibility toward the environment, the public, the employees, and the consumers, we test whether and how foreign direct investment affects corporate social responsibility of Chinese ﬁrms. Our results show that ﬁrms run by entrepreneurs who have work experience and, especially, management experience in multinational enterprises investing in China conduct more corporate social responsibility activities compared with ﬁrms run by entrepreneurs without such experience. We further ﬁnd that multinational enterprise management experience may enhance the entrepreneurial self-assessment of social status, broaden the international vision of entrepreneurs, and reduce the level of ﬁrm family control, which help increase ﬁrms’ uptake of corporate social responsibility activities. Our results suggest that foreign direct investment can serve as a vehicle for sustainable development, through the transfer of corporate social responsibility-related managerial knowledge from foreign to local ﬁrms.


Introduction
Corporate social responsibility (CSR) activities can help increase social welfare and sustainability by contributing to various social issues such as human capital development, environment protection, and social cohesion [1]. Compared with the level of CSR in developed countries, however, CSR activities in developing economies are not pervasive and in the early stage of development [2][3][4]. According to ASSET4, the leading provider of environmental, social, and corporate governance (ESG) data, the overall CSR ratings of developed countries are generally much higher than those of developing countries. For example, the CSR ratings for United Kingdom, United States, Germany, and France are 64.32, 51.91, 58.25 and 71.45, respectively, while the ratings for China, Russia, India, and Zimbabwe are 25.59, 37.52, 47.16 and 11.75, respectively. Therefore, to achieve a more sustainable development, an important question to be addressed in developing countries is how to effectively promote CSR engagement of domestic firms. Some recent observations appear to show that CSR activities in developing countries seem to increase with the rise of foreign direct investment (FDI) inflows [5][6][7].
One explanation for such a positive relationship is that FDI spillovers on CSR activities may take place through knowledge transfer from foreign firms to domestic firms [7], since multinational enterprises (MNEs) usually put CSR high on their corporate agenda [5,[8][9][10]. However, as previous studies on spillovers have been focused on domestic productivity [11][12][13], there is still a research gap regarding the effect of FDI on local CSR and the mechanisms through which spillovers occur. This paper aims country [76]. For instance, CSR practices may help MNEs obtain licenses and contracts [77]. Second, there exists external pressure on MNEs to conduct CSR. For example, as required by the Chinese government, MNEs investing in China need to comply with international labor standards which are generally higher than the local ones [9]. Third, some parents of MNEs may extend their CSR policies to their overseas subsidiaries and require subsidiaries to carry out CSR activities [8].
Recent studies have shown that multinational presence is associated with CSR of domestic firms in developing countries. For example, regarding CSR as a signaling device for foreign firms to reveal their types, Goyal [5] theoretically shows that CSR in developing countries increases with the rise of inward FDI flows. Based on questionnaires to managers of both local and foreign firms, Kuada and Hinson [6] find that local firms may increase CSR activities due to pressure from foreign firms. Taking Ghana for example, Nyuur, Ofori and Debrah [7] empirically show that FDI inflows into developing countries have a positive impact on CSR activities of local firms through knowledge transfer.
FDI spillovers may arise through several channels: competition [11,78], demonstration [11,79], linkages [28,80,81], and labor mobility [82,83]. Among them, empirical evidence on labor mobility is still scarce compared with other channels. Theoretically, spillovers can take place through the labor mobility channel when workers and managers with MNE work experience are hired by domestic firms, or start their own business. As MNEs usually devote great efforts to staff training [84,85], their employees are able to accumulate knowledge and skills which can be transferred to domestic firms with labor mobility. However, the labor mobility channel has not until recently been empirically tested probably due to data constraints. For example, Görg and Strobl [86] show that firms run by entrepreneurs who once worked in MNEs in the same industry are more productive than other domestic firms in Ghana. Poole [87] provides indirect evidence that in Brazil the wage rates of local workers increase with the presence of former MNE-trained workers. Liu et al. [88] find that private firms in China run by entrepreneurs with MNE experience have higher return to equity (ROE) and return to assets (ROA) than those run by entrepreneurs without such experience. Nevertheless, there has been no empirical analysis of the labor mobility channel of FDI spillovers on domestic CSR.
Taken together, considering the significant FDI spillovers on various aspects of the host country as well as the high CSR engagement of MNEs, we propose that the CSR performance of domestic firms in developing countries can be promoted by the presence of MNEs. Specifically, we regard labor mobility as a potential way to accumulate and transfer CSR related knowledge, skills, and attitudes to domestic firms. This leads to our hypothesis of how FDI may affect local CSR performance: Hypothesis 1. Firms run by entrepreneurs who have worked in MNEs are more likely to engage in CSR activities.

Data
The firm-level data in this study come from nationwide surveys of privately owned enterprises in China conducted in 2008, 2010 and 2012. The surveys were jointly conducted by the United Front Work Department of the Central Committee of the Communist Party of China, All-China Federation of Industry and Commence, the State Administration of Industry and Commerce, and the Private Economy Research Institute of China. The surveys had been conducted every two years since early 1990s. We used the surveys in 2008, 2010 and 2012, as information on CSR is available for these years. We used repeated cross-sectional data as firms were re-sampled for each survey. The sample was drawn from 31 provinces that cover all areas in mainland China. The surveys were conducted by detailed face-to-face interviews with firm owners or entrepreneurs. The questionnaires covered information regarding entrepreneurial characteristics, firm characteristics, and a wide range of management practices of firms. Most relevant to our study, the surveys collected detailed information on entrepreneurs' occupational careers prior to establishing their own businesses, which provides us the opportunity to investigate the spillover effects of FDI via labor mobility on corporate CSR policies. We excluded observations with missing values on some key variables and end up with 6747 private firms, mainly small-and medium-sized. Table 1 presents the distribution of our sample by province,  industry, and year, respectively. It reflects the fact that in order to achieve a balanced representation of the population of the private firms in China, a multi-stage stratified random sampling procedure based upon administrative regions and industries was used in the survey. In our sample, on average the share of equity owned by the entrepreneur was about 70.3 percent in the survey year, which suggests that entrepreneurs in our sample play a leading role in corporate decision making process and thus their personal experience can have crucial impact on corporate policies. Detailed variable definitions are provided in Table 2. The dependent variable in our study is a self-constructed index measuring CSR activities. Following the literature on CSR, our index is based on four most commonly acknowledged components of CSR including environmental protection, monetary donation, employee welfare, and product quality improvement. These four components explicitly reflect corporate responsibility for the environment, the public, the employees, and the consumers, respectively.

CSR
An index based on corporate responsibility for the environment, the public, the employees, and the consumers. Detailed information is provided in the Appendix A Table A1.
Corporate Social Responsibility (CSR) components env_prot A sub-index based on firm investment for environmental protection and environmental protection fees. donation The standardized value of firm donation amount.

empl_welf
A sub-index based on the average wage, insurance provision, and whether the firm has labor union, labor welfare committee, and Communist Party organization.
prod_qual A sub-index based on the number of well-known trademarks, self-designed products, patents, research and development (R&D) expenditure, and employee training expenditure.

Multinational Enterprise (MNE) Experience
MNE_exp A dummy variable that equals 1 if an entrepreneur has ever worked in MNEs, and 0 otherwise.

MNE_manager
A dummy variable that equals 1 if an entrepreneur has ever worked in MNEs at the management level, and 0 otherwise.

Entrepreneurial characteristics
male A dummy variable that equals 1 if an entrepreneur is a male, and 0 otherwise.

age
The age of an entrepreneur in the survey year.
edu A dummy variable that equals 1 if an entrepreneur has education higher than high school, and 0 otherwise.
party A dummy variable that equals 1 if an entrepreneur is a member of the Communist Party of China, and 0 otherwise.
pc An index measuring an entrepreneur's political connections following Zhao and Lu [89], which is based on whether an entrepreneur is a deputy of the People's Congress or the Chinese People's Political Consultative Conference and the deputy level.
soe_exp A dummy variable that equals 1 if an entrepreneur has ever worked in state-owned or collective enterprises, and 0 otherwise.
private_exp A dummy variable that equals 1 if an entrepreneur has ever worked in private enterprises, and 0 otherwise.

Firm Characteristics firm_age
The number of years since a firm was established in the survey year.

sales
The monetary sales of a firm in the survey year (thousand Yuan). profit The profit of a firm in the survey year (thousand Yuan). leverage The ratio of firm bank loans to assets (percent).

Mediation variables status
A variable based on an entrepreneur's self-assessment of social status.

export
A dummy that equals 1 if a firm is engaged in exports, and 0 otherwise. family A variable constructed by factor analysis based on whether the entrepreneur makes significant decisions by himself and whether the entrepreneur is responsible for firm's daily management.
Specifically, the component of environmental protection measures firm investment in environmental protection and environmental protection fees. The component of monetary donation is the standardized value of firm donation amount. The component of employee welfare includes information on the average wage, various insurance provisions, and whether the firm has the labor union, labor welfare committee, and Communist Party organization. The component of product quality improvement covers information on the number of well-known trademarks, the number of self-designed products, the number of patents, research and development (R&D) expenditure, and employee training expenditure. Using factor analysis, we first separately create these components (except for donation) and then construct the CSR index. The detailed information on the index is provided in the Appendix A Table A1.
The main independent variable is entrepreneurial work experience in multinational firms. We use two variables to capture different job position levels. MNE_exp is a dummy variable that equals one if the entrepreneur has ever worked in MNEs in China before he/she started his/her own private enterprise and zero otherwise. MNE_manager is also a dummy variable which equals one if the entrepreneur has worked in MNEs as a manager before starting his/her own business and zero otherwise. In our sample, 4.8 percent and 3.2 percent of entrepreneurs have MNE work experience and MNE management experience, respectively. As managerial personnel are more deeply involved in business operations of firms, we expect that entrepreneurs with MNE management experience are more likely to obtain knowledge and management skills related to CSR than those with MNE work experience in general.
In our regressions, we control for various entrepreneurial and firm characteristics. We include entrepreneurial age (age), gender (male), education (edu), membership of the Chinese Communist Party (party), degree of political connections (pc), work experience in state-owned or collective firms (soe_exp), and work experience in private firms (private_exp). The inclusion of entrepreneurs' human capital, political capital, and other work experiences in our specifications also helps to better isolate the impact of MNE experience on CSR activities. Variables on firm characteristics include firm age (firm_age), sales (sales), profit (profit), and the leverage ratio (leverage), which are shown to be important determinants of CSR [17,[90][91][92][93]. In addition, we include province, industry, and year dummies to control for variations across provinces, industries, and time. Table 3 provides the descriptive statistics of variables.  Table 4 presents the results of the mean-difference tests between entrepreneurs with and without MNE experience. The results indicate that firms run by entrepreneurs with MNE work experience (Panel A) and especially management experience (Panel B) conduct more CSR activities. Results of the mean-difference tests show that the differences are statistically significant at 1 percent level in both cases. This finding is supportive of our hypothesis. In addition, there are also significant differences in terms of entrepreneurial and firm characteristics between entrepreneurs with and without MNE experience. For example, entrepreneurs with MNE experience have more years of formal education, and their firms are more profitable. In the next section we will employ rigorous econometric regression analysis to better control for potential factors that might confound the relationship between MNE experience and CSR engagement. Table A2 in the Appendix A reports the correlations among variables. We also examine the variance inflation factors (VIFs) of all variables. All VIFs are below 2 and much lower than the commonly accepted threshold of 10. These results suggest that multicollinearity does not pose a serious concern in our regressions.

Econometric Model
To identify the FDI spillover effects on CSR, we set up the econometric model as follows: where the subscript i, j, p, t denote firm, industry, province, and year, respectively. CSR refers to the CSR index; MNE denote the main explanatory variables MNE_exp or MNE_manager; X is a vector of the control variables. We standardize explanatory variables (except for the dummy variables) to make them consistent with the indexed CSR measurement and to facilitate comparisons of the estimates of different variables. We also include industry (u j ), province (u p ), and year (u t ) fixed effects in the model. We first estimate the econometric model using the ordinary least squares (OLS) method, which can effectively isolate the effect of MNE experience on CSR engagement conditioning on the control variables as well as the industry, province, and year fixed effects. To explore the underlying mechanisms why MNE experience can enhance CSR activities, we then test the effects of three potential mechanism variables including the entrepreneurial self-assessment of social status, the international vision of the entrepreneur, and the degree of family control in a firm. To identify the impact of the mechanism/mediation variables, we estimate alternative models following the Baron and Kenny [94] method which has been widely applied [95][96][97]. We first test the effect of MNE experience on the mediation variables. We then add the mediation variables to the original model in the baseline analysis so as to compare the effects of MNE experience on CSR without and with the mediation variables. This helps us distinguish the total effect of MNE experience as well as its direct and indirect effects (through mediation variables) on CSR.
As for the identification issue in our model specification, it is noted that entrepreneurial MNE experience was prior to corporate CSR activities, which eliminates the reverse causality problem. Nonetheless, a potential concern in our regressions is that the omitted variables might bias our estimates. We pursue two strategies to assess the validity of our results. First, we include a comprehensive set of control variables in the model specification to reduce omitted variables bias. Second, considering that further unobservables may affect selection into the group of entrepreneurs with both MNE experience and CSR activities, we conduct an assessment of unobservable selection following the approach proposed by Oster [24].
At the core of the approach in Oster [24] is a proportional selection relationship on observed and unobserved variables. Different from the previously common method simply observing coefficient movements after the inclusion of controls, this formal test concurrently takes into account both coefficient and R-squared movements to identify omitted variables bias. For our main results, we calculate the coefficient of proportionality (δ), which indicates the degree of selection on unobservables, relative to the selection on observables, necessary to overturn the effect of MNE experience on CSR activities. The calculation of δ requires a hypothetical maximum R-squared (R max ), which would be attainable in a regression that includes both observed and unobserved controls. Following the recommendation of Oster [24], we assume the maximum R-squared as R max = min 1.3R F , 1 , where R F is the R-squared from a regression with the full set of observed controls. Table 5 presents the regression results using MNE work experience (MNE_exp) or MNE management experience (MNE_manager) as the main explanatory variable. In the table, we report the effect of MNE experience without and with control variables. As shown in Columns (1) and (2), the coefficient on MNE_exp is statistically significant at the 5 percent level, suggesting that MNE work experience has significantly positive effect on the index of CSR activities. Specifically, in Column (2) with all control variables, holding other factors constant, the CSR index is on average higher by 0.082 standard deviations for firms run by entrepreneurs with MNE work experience than those without such experience. When we use MNE management experience as the main explanatory variable in Columns (3) and (4), we find that the effect of MNE management experience is significantly positive at the 1 percent level. Furthermore, it is shown that the effect of MNE management experience on CSR is much more significant, both economically and statistically, than that of MNE work experience. One explanation for the larger and more significant effects of MNE management experience is that the potential to learn about CSR is larger with higher job position, as management staff is more deeply involved in critical corporate policies. Taken together, these results support our hypothesis that firms run by entrepreneurs with MNE experience conduct more CSR activities.

Main Results
We report the coefficients of proportionality (δ) in the bottom row of Table 5. The estimates of δ range from 4 to 8, which suggests that unobservables would need to be more than 4 to 8 times as important as all controls included to produce a treatment effect equal to zero. This provides strong support that it is unlikely that our results are exclusively driven by unobservable selection as a rich set of important observed controls, especially measures of entrepreneurial human and political capital, are included in regressions. Note: Robust standard errors are reported in parentheses. ***, ** indicate statistical significance at the 1%, 5% and 10% levels, respectively. δ indicates the coefficient of proportionality suggested by Oster [24].
As for the control variables, the results show that firms run by entrepreneurs who are more educated (edu), acquire membership of the Communist Party (party), with higher degree of political connections (pc), and with state-owned or collective enterprises work experience (soe_exp) tend to conduct more CSR activities. Also, CSR seems to be significantly positively correlated with firm age (firm_age), size (sales), profitability (profit), and access to external finance (leverage).

Robustness Checks
We perform several robustness checks regarding the relationship between MNE experience and the CSR index. Results are reported in Table 6. In Column (1), we exclude observations with any foreign or state-owned paid-in capital in the registration year. As foreign firms may have stronger motivation to conduct CSR activities than local firms, the CSR engagement of firms with paid-in capital by foreign investors may be higher anyway than other firms in the sample. Similarly, firms with state-owned capital may act differently from start-up private firms with respect to CSR activities, since the state-owned partners may impose political objectives on firms. Nevertheless, the main findings in the baseline regressions still hold. In Column (2), we exclude observations transformed from state-owned or collectively owned enterprises. Dropping these observations results in a smaller sample size from 6747 to 3066, as survey questions regarding information on firm transformation are contained in questionnaires only for years 2008 and 2010, but not for year 2012. Compared with private firms, state-owned or collectively owned firms may have different corporate policies and strategies which can have long-lasting influences on firm behaviors even after being transformed into private firms. The results of excluding these firms are consistent with those in the baseline regressions. We then in Column (3) exclude firms operating in finance and public utilities sectors where firms might be intrinsically different from other firms and the baseline results hold. These findings further support our hypothesis. Note: Control variables and fixed effects are the same as those in Table 5. Robust standard errors are reported in parentheses. ***, ** indicate statistical significance at the 1%, 5% and 10% levels, respectively.

FDI Spillovers on Different CSR Components
The results documented to this point have shown that MNE experience is positively associated with the CSR index. Next, we turn to examine the impact of MNE experience on the four specific CSR components separately. This exploration would lend further support to our main results from the perspective of the factor construction. More importantly, it can shed some light on which of these CSR dimensions may be most influenced by MNE experience.
We report the effect of MNE experience without and with control variables for each CSR component in Table 7. As for environmental protection (env_prot) and donation amount (donation) in Columns (1) to (4), the results show that MNE work experience has no significant effect, while MNE management experience is marginally significant at the 10 percent level. In Columns (5) to (8), we find positive effects of both MNE work experience and management experience on employee welfare (empl_welf ) and product quality improvement (prod_qual), and all the coefficients are statistically significant at the 1 percent level. These results highlight that the effect of MNE experience on improving CSR is mainly focused on the employee welfare and product quality improvement dimensions. Such a finding is largely consistent with the superiority of MNEs in human resource management and technology innovation [72,73]. We also report the coefficients of proportionality (δ) in the bottom row of each panel (we report the coefficient of proportionality (δ) for each CSR component, even when the coefficient on MNE experience is insignificant, so as to provide information on how large the unobservable selection is in our research setting). All the absolute values of coefficients of proportionality are larger than 1, which can be considered indicative of robust effects [24]. Note: Control variables and fixed effects are the same as those in Table 5. Robust standard errors are reported in parentheses. ***, * indicate statistical significance at the 1%, 5% and 10% levels, respectively. δ indicates the coefficient of proportionality suggested by Oster [24].

Potential Mechanisms
To better understand how FDI spillovers occur through the labor mobility channel, we test three potential mechanisms that may link MNE experience with CSR performance: entrepreneurial self-assessment of social status, the international vision of the entrepreneur, and the level of firm family control. Considering that the estimated coefficients of MNE management experience (MNE_manager) are larger, in both magnitude and significance, than those of MNE work experience (MNE_exp), we will focus on examining the effect of MNE management experience in the following explorations of potential mechanisms.

Self-Assessment of Social Status
In developing countries like China, workers and especially managers in MNEs usually gain high social and economic status. By constructing an occupational prestige scale and socio-economic indexes for 161 professions, Li [98] analyzes the social prestige stratification in China. The results show that in China the management job in foreign firms ranks 10th in the occupation prestige analysis (top 9 mostly being government and research job) and it ranks second based on the socio-economic indexes. This is consistent with the finding that variations in prestige scales are attributable more to variation in education than in other factors [99], as managers in MNEs are generally better educated and even have overseas study experience.
It is noted that social prestige and the related social assessment system may have critical influence on the identity, values, and behaviors of individuals [100]. Previous studies have shown that individuals with high social status may perform more CSR activities. First, high social status usually means abundant economic resources and high capabilities [101], which can help people easily assimilate into society and actively take part in voluntary activities [102]. Second, it is possible that people in high social status may be faced with public pressure to conduct CSR activities. Therefore, we argue that entrepreneurs who gained high social status from their management experience in foreign firms may be more inclined to conduct CSR.
We test this mechanism and report the results in Columns (1) and (2) of Table 8. In the surveys, the entrepreneurs reported the self-assessment of their social status on a 1 to 10 scale. We create a measure of entrepreneurial self-assessment of social status (status) which takes on the value one if the entrepreneur has a score higher than the mean of the original variable and zero otherwise. In Column (1), we test the effect of MNE management experience (MNE_manager) on the self-assessment of the social status of entrepreneurs. The estimated coefficient of MNE_manager is 0.115 and significant at the 10 percent level, which supports the findings in previous studies on the positive relation between MNE experience and high social status. In Column (2), the mechanism variable status is added to the baseline regression in Table 5 and the effect of status on CSR is significantly positive at the 1 percent level. Moreover, compared with the baseline results, we find that after controlling for the mediation variable, the coefficient of MNE_manager declines in magnitude. To estimate the mediation effect, we employ the Sobel-Goodman mediation test. The result shows that the proportion of the total effect mediated by status is about 5.78 percent. This implies that improving entrepreneurial self-assessment of social status is a mechanism for MNE management experience to affect CSR activities.

The International Vision
Given the cross-border nature of MNEs, entrepreneurs who have previously worked in MNEs may learn to have an international vision, which increases the involvement of their firms in international economic activities. Using data in 2000, 2002 and 2004 from the same surveys that we use in this study, Liu, Lu and Zhang [88] show that firms run by entrepreneurs with MNE experience are 42.68 percent more likely than their counterparts to have international involvement such as purchasing foreign assets, exporting, and investing abroad. Filatotchev, et al. [103] point out that entrepreneurial international background and global networks are important to determine a firm's export orientation and performance.
The international vision of the founders has been found to be a crucial factor for their firms' CSR performance. On the one hand, exporters in developing countries conduct CSR activities to increase their exports to developed countries, in which consumers care about the social and environmental implications of their consumption behavior and, therefore, have strong preference for products produced according to ethical standards [104]. On the other hand, exporters may have to comply with programs and policies related to CSR in their own countries. For example, since 2013, the Ministry of Commerce of China has required Chinese firms to perform CSR activities in their international business operations, such as active participation in local public services, donation, and environmental protection (http://www.mofcom.gov.cn/article/fbhfn/fbh2013/201302/20130200040491.shtml. Accessed on 16 March 2018). Therefore, we expect that entrepreneurs who broaden their international vision through MNE management experience may conduct more CSR activities.
We report the results of testing this mechanism in Columns (3) and (4) of Table 8. We use the export status of a firm to proxy its international involvement and therefore the international vision of the entrepreneur. The variable export is a dummy which equals one if a firm is engaged in export activities and zero otherwise. As shown in Column (3), the coefficient on MNE_manager is significantly positive at the 1 percent level, indicating that MNE management experience increases the probability of exporting. In Column (4), we add the mechanism variable export to the original model and compare the results with the baseline OLS results. It can be seen that the coefficient on export is significantly positive at the 1 percent level. In addition, the magnitude of the effect of MNE_manager on CSR decreases after controlling for export. The result of the Sobel-Goodman mediation test shows that the proportion of the total effect mediated by export is nearly 30.84 percent. These results therefore support entrepreneurial international vision as a mechanism through which MNE management experience affects CSR performance.

Family Control
Family control of business is prevalent in Chinese private firms [22,23]. As a critical dimension of corporate governance, family ownership has been shown to be closely related to firm CSR performance. Some argue that firms with high levels of family control are actively engaged in CSR activities, as they may have strong incentives to provide protection against future adverse events [105], or to accumulate socio-emotional wealth [106]. Nevertheless, some recent studies show that firms with high family control may exhibit low CSR performance. For example, Campopiano and De Massis [107] find that compared with non-family firms, family firms are less compliant with CSR standards and disseminate a greater variety of CSR reports. Using data of S&P 500 firms, Block and Wagner [108] find that family ownership can negatively affect community-relate CSR performance. It is possible that family firms face credit constraints to finance the costs of CSR activities and are reluctant to cover the costs with their family endowment [109]. Also, as information asymmetry is relatively low in family-controlled firms [110], it would be difficult for inefficient managers in these firms to cover themselves with socially responsible activities to gain stakeholders' support [111]. As multinationals usually follow modern corporate governance practices and have low levels of family control of business, we expect that multinational management experience of entrepreneurs may be able to change their views on family ownership and, therefore, affect their firms' CSR performance.
Columns (5) and (6) of Table 8 present the results of testing the mechanism of the level of family control. In the surveys, two questions were related to the degree of family control: (i) who makes the significant decisions of the firm, and (ii) who is responsible for the daily management of the firm. The alternative answers for the first question are: the entrepreneur himself, the board of the shareholders, the board of the directors, and others. The answers for the second question include the entrepreneur himself, the entrepreneur and other managers, the entrepreneurs and the party organizations, and others. We first generate two dummy variables, deciosn_self and admin_self, which take on the value one if the answer to the above two questions is "the entrepreneur himself" and zero otherwise. Based on these two dummies, we then construct an index to measure the level of family control (family) using factor analysis. It is shown in Column (5) that the coefficient on MNE_manager is significantly negative at the 1 percent level, which suggests that firms run by entrepreneurs who have MNE management experience have lower levels of family control. In Column (6), we add the mechanism variable family to the baseline model. We find that the effect of family on CSR is significantly negative at the 1 percent level, indicating that firms with higher levels of family control conduct lower levels of CSR. Compared with the baseline results, we find that the magnitude of the MNE_manager coefficient declines after adding the family variable. The Sobel-Goodman mediation test shows that the proportion of the total effect mediated by family is about 15.79 percent. These results provide some evidence that family control is a mechanism for MNE management experience affecting CSR performance. We further add the three mechanism variables simultaneously into the baseline regression and the results support the mentioned mechanisms connecting MNE management experience and CSR (results are available upon request).

Conclusions and Discussion
This paper investigates whether and how FDI affects corporate social responsibility of local firms through the channel of labor mobility. Using nationwide data of private firms in China, we find that firms run by entrepreneurs with MNE work experience, especially MNE management experience, significantly perform more CSR activities. We also examine the potential mechanisms behind the channel of labor mobility. The results show that MNE management experience may enhance entrepreneurial self-assessment of social status, broaden the international vision of entrepreneurs, and reduce the level of a firm's family control, which increases firm CSR performance. The results suggest that FDI can serve as a vehicle for sustainable development, through transfer of CSR-related managerial knowledge and skills from foreign to local firms.
Our findings are consistent with those in previous studies which suggest a positive relationship between FDI and local CSR performance [5][6][7]18]. By contrast with these studies, this paper contributes to the literature by systematically identifying labor mobility as a channel of FDI spillovers on CSR and further explains how this channel is realized. Our results are also consistent with earlier findings that FDI may have positive impact on local productivity or wage rates [11,12,86,87]. From the perspective of CSR, we add to the literature by providing new insights into the understanding of FDI spillovers. Our results have important implications for research on CSR, in terms of theoretical development and the design of future empirical studies. Researchers could apply our framework to other developing countries, such as Russia and India, which have also been witnessing large amounts of FDI inflows and paying increasing attention to issues related to sustainability such as corporate social activities.
This study also has some important policy and practical implications. First, our results support a positive effect of FDI on local CSR. Therefore, for developing countries with relatively low levels of CSR, one way to improve CSR activities is to attract FDI and extract CSR-related benefits from it. Related to this, the finding that the host countries of FDI may gain from FDI spillovers on CSR further rationalizes the pro-FDI policies in developing countries. Policy makers can attract and retain FDI by offering special incentives to foreign investors, such as tax reductions, subsidies to investment, favorable labor policies, and land-use fees, etc. This study also suggests that policies aiming at liberalizing the foreign investment regime could be helpful for achieving sustainable development in developing countries.
Second, we show the importance of labor mobility in conveying the impact of FDI on CSR. As knowledge is attached to workers and personnel, the labor turnover effect is closely associated with the degree of physical labor mobility. This implies that policy makers can design and implement policies facilitating labor force mobility so as to better reap the spillover benefits of FDI. One case in point is relaxing "Hukou" restrictions in China. Although the Chinese government has been gradually relaxing "Hukou" since the 1980s, labor mobility constraints due to this regime still exist [112]. Our results support policies further relaxing the "Hukou" system and other restrictions on labor mobility in China.
Third, our results provide some human resource management implications. On the one hand, the experience and commitment of top management to a large extent determine the acquisition of foreign management knowledge. This would provide some guidelines for issues with respect to the adoption of new management practices. On the other hand, considering that the distance in development between local and foreign firms can be significant in developing countries, more efforts should be made to understand the factors that affect the effectiveness of managerial knowledge transmission and acquisition. This paper has some limitations and raises several opportunities for future studies. First, to form a condensed research framework, we focus on the labor mobility channel of FDI spillovers and leave other potential spillover channels such as competition and demonstration out of the discussion. Although a full investigation of all the channels is beyond the scope of the current paper, it would be worth developing a unified framework to incorporate various channels of FDI spillovers on CSR. Second, due to data availability, our sample is restricted to privately owned firms. As institutional arrangements and incentives differ substantially for private and state-owned firms, they may have different CSR determinants and performance. In the future, it would be interesting to compare our analysis of private firms with the analysis examining state-owned firms.

Conflicts of Interest:
The authors declare no conflict of interest.