How does Outward Foreign Direct Investment Affect Enterprise Green Technology Innovation? — Based on the Moderating effect of Resources

with the increasing strict environmental regulations in the green transition process, outward foreign direct investment is considered to be an effective approach to promote enterprises’ green technology innovation. Thus, this paper establishes a comprehensive research framework that integrates OFDI and green technology innovation from the micro level of the enterprise to analyze it. The �ndings show that: First, OFDI will positively affect corporate green-tech innovation as expected; Government subsidies have a U-shaped regulation on the relation between OFDI and green-tech innovation; Absorbed slack plays an inverted U-shaped moderating effect on the relation between OFDI and green-tech innovation, and the unabsorbed slack positively affect this process. As for the heterogeneity of property rights, the test results of non-state-owned enterprises and state-owned enterprises are basically consistent with the baseline results, except for the following two points: the unabsorbed slack of state-owned enterprises has no regulatory effect between OFDI and enterprise green technology innovation, and the absorbed slack of non-state-owned enterprises has no regulatory effect between OFDI and enterprise green technology innovation.


Introduction
Outward foreign direct investment (OFDI) is or has become an effective way for enterprises in developing countries to seek resources and catch up with cutting-edge technology (Seyoum et al., 2015;Wu et al., 2016;Li et al., 2016a;Piperopoulos et al., 2018;Javorcik et al., 2018).In China, the development scale of OFDI has been growing in recent years, providing an unprecedented opportunity for enterprise technology development.Meanwhile, the practice of Green Economic Transition and Development Strategy (such as strict environmental regulation policy) has been bring great pressure to enterprises' green / cleaner production in China (Li et al., 2017;Wu et al., 2020), which consequently makes appropriate green technology development a signi cant aspect for enterprises to promote production quality and realize sustainable development (Leonidou et al., 2013;Peng and Liu, 2016;Ebrahimi and Mirbargkar, 2017).
Therefore, given the particular period of green and sustainable economic transition, OFDI is likely to have an dramatic impact on upgrading enterprises' green-tech level.
Given the increasingly strict environmental regulation during the green economic transition period, in addition to production e ciency and quality, green / cleaner production capacity has also become an important contents of enterprise development.Therefore, the technologies, knowledge, management experience, and other resources gained through OFDI are likely to support the innovation of new technology, and whether it is able to further improve enterprises' green technology innovation capability, and thereby satisfying the requirements of economic green transition? Green technology innovation refers to the innovation of product design, green materials, green technology, green equipment, green recycling, green packaging and other technologies that can effectively deal with environmental problems and reduce environmental pollution (Kemp and Pearson, 2007; Aguilera-Caracuel and Ortizde-Mandojana, 2013).That is, green technology innovation is considered to be characterized by higher cost, higher risk and relatively exclusive resource support compare to the general counterpart, which makes the development of green technology needs more and more internal and external resources support.
Resource-based view (RBV) shows that resources are recognized as assets, capabilities, and knowledge of an enterprise, and are the basis for companies to improve production e ciency and pro t through effectively developing and implementing speci c development strategies (Porter, 1981;Barney, 1991;Li et al., 2017).When green technologies have not effectively released their market value potential, particularly in those brand new green technology elds with high risks, high value and underdeveloped characteristics, effective support of internal and external resources is able to drive enterprises to actively explore access to (such as OFDI) key resources to acquire and absorb new technology related resources, so as to improve the R&D capability of green technology.Government subsidies and slack resources are signi cant and typical external and internal resource to support enterprises to carry out green technology innovation (Peng and Liu, 2018 Therefore, when enterprises expect to effectively seek knowledge, technology and other resources for green technology innovation through OFDI, the joint of government subsidies and slack resources is able to provide su cient support for enterprises' OFDI and consequently promote the positive correlation mechanism between OFDI and enterprises' green new technology R&D.However, the moderating effects of both government subsidies and slack resources between OFDI and green technology innovation has not been theoretically and empirically tested in the context of China (Bai et al., 2020;Yang et al.,2020;Song et al.,2021).Additionally, the particularity of enterprise property right attribute will lead to signi cant differences between state-owned enterprises and private enterprises in resource acquisition, internal governance and development objectives in the practice of China's enterprises (Choi et al., 2011;Aghion et al., 2013;Wu et al., 2021).Thus, this paper will give an insight into the micro-mechanism of OFDI on corporate green technology innovation based on the moderating effects of both government subsidies and slack resources in the developing context of China's enterprises.
Based on the sample of the manufacturing enterprises in Shanghai and Shenzhen stock markets in 2011-2018, this paper establishes a comprehensive research framework that integrates OFDI and enterprise green technology innovation, focuses on the effects of OFDI on the green technology innovation, discusses the moderating effect of government subsidies and slack resources on the role of this relationship.In addition, this paper also analyzes the applicable boundary of basic conclusions from the perspective of heterogeneity of property rights.There are three main theoretical contributions to present literature.First, from the micro view of the enterprise, this paper explores the causal effect between OFDI and green technology innovation, and further analyzes the mechanism between them from the perspective of resource-based view, which will further enrich the literature on the OFDI and enterprise green technology innovation at the micro level; Second, this paper comprehensively discusses the moderating effect of government subsidies and slack resources on the role of OFDI in corporate green innovation, which will support to clarify the micro-action mechanism of OFDI on corporate green innovation, and to establish the important guarantee mechanism for effective use of resources, and to further improves the theory of resource-based views.Third, under the circumstances of China's vigorously implementing the strategy of "high quality development" and "going global", this study can provide a valuable reference for the enterprises to contribute to green development by optimizing the relationship between OFDI and green technology innovation from the perspective of resource support.

Literature And Hypothesis
Page 4/23

OFDI on green technology innovation
As an important resource acquisition channel, OFDI refers to the investment that made through direct establishment and operation of enterprises in foreign countries, including participating in capital, establishing joint ventures, buying existing enterprises and opening subsidiaries (or branches), which enables companies have access to the critical productive resources, such as advanced technology, management experience, research and development results, and cutting-edge information resources (Du et al., 2012;Huang, 2013;Feenstra et al., 2014;Bauer et al, 2016;Huang and Zhang, 2017;Wu et al, 2017;Hamida, 2017;Zhou et al., 2019).Green technology innovation is considered to have additional restrictions compared with the developing of general new technology, and thereby brings up more requirements on company's resource allocation, strategic deployment, and internal management model, which is likely to hurt their pro tability.Therefore, given the increasing strict environmental regulations, OFDI will probably polish up enterprises' green technology innovation capability through reverse technology spillover effect, investment income feedback effect, learning effect, and other approaches.Speci cally, reverse technology spillover effect is widely recognized as the knowledge spillover that enterprises seek outward advanced technology by embedding into the technological innovation network of the host country through the establishment of international subsidiaries, international joint ventures and foreign R&D centers, and is likely to enrich knowledge stock and improve the technological level of domestic enterprises in terms of the technology diffusion effect, imitation follow effect, talent ow effect and information platform effect (Pradhan and Singh, 2009;Du et al., 2012;Hamida 2017).The feedback effect of investment income refers to the income generated by enterprises' investment in the overseas branches, and is able to bene t feedback to the R&D investment and innovation activities of parent companies (Yang et al., 2013;Li et al., 2016a;Piperopoulos et al., 2017).In detail, the parent company is able to bypass trade barriers and achieve large-scale economy, thereby reducing the unit cost and increasing sales revenue through OFDI.Meanwhile, as an effective approach to expand the market scale, OFDI is expected to share parent company's R&D expenditure pressure by remaining the investment income in their oversea branches.In addition, learning effect emphasize that OFDI is also an effective approach for enterprises to update their knowledge base, including advanced technical knowledge and management knowledge, through a wider range of international communication and cooperation, which involves continuous exchange and learning with overseas Given the critical period of green transition of China's enterprises as well as their relatively weak green-tech innovation capacity compared with developed countries (Piperopoulos et al., 2018;Bai et al., 2020), OFDI is expected to quickly enhance enterprises' green-tech level and gain competitive advantage by seeking outward knowledge.Thus, hypothesis 1 is proposed as follow: H1: OFDI has a positive effect on corporate green technology innovation.That is to say, the OFDI from the perspective of corporate green technology innovation will not induce the green escape effect.

Moderating role of Government subsidies
In general, the developing of new technology is basically characterized by relatively large-scale investment and high risk (Kang and Park, 2012;Catozzella and Vivarelli, 2016;Shang et al.,2021).More speci cally, large-scale investment indicates that enterprise's R&D of new technology usually involves a large quantity of supportive resource, namely, a large amount of initial nancial support and high-tech R&D personnel participation.High risk suggests that there are technical risks, competitive risks and objective environmental risks in the development of enterprises' new technologies.Therefore, government subsidy is likely to stimulate the motivation of enterprises' innovative activities by directly reducing the cost and risk in new technology R&D progress (Kang and Park, 2012;Cappelen et al., 2012;Huergo and Moreno, 2017;Bellucci et al., 2019).In particular, the positive effect of government subsidies will be more signi cant on the developing of green new technologies that characterized with higher public goods spillover (Peng and Liu, 2018;Wu and Hu, 2020).Thus, as an essential external supportive resources for enterprises' R&D activities, government subsidy is likely to cultivate enterprises' green technology innovation motivation in terms of the direct and indirect approach (or moderate effect).
However, there is still no consensus on the speci c role of government subsidy in enterprises' technology innovation process.Recent efforts to explore the relation between government subsidies and technology innovation have gradually formed four main points, namely, positive effects, negative effect, non-linear relationship, heterogeneity of effect.As for the positive effects, government subsidy is recognized to be an effective approach to alleviate the nancing pressure and R&D risks in new technology innovation process, enhancing the return on investment, and consequently improving the motivation of rms' technology innovation (Bernini and Pellegrini, 2011;Cappelen et al., 2012;Bellucci et al., 2019;Guo et al., 2016;Huergo and Moreno, 2017).The negative side points out that government subsidies are likely to induce rent-seeking behavior of enterprises' decision-makers and policy-makers due to the imperfection of subsidy distribution system and supervision mechanism, resulting in deviation from the original purpose of subsidy policy (Catozzella and Vivarelli, 2016;Boeing, 2016;Wang et al., 2017).The non-linear relationship viewpoint holds that government subsidy has a threshold effect as well as an effective range of support intensity (Mao and Xu ,2015;Yu et al.,2016).In terms of the heterogeneity of effect, the speci c effect of government subsidies will change accordingly with the variations of situations (Howell, 2017).
Therefore, a positive U-shaped moderating role is likely to be played between OFDI and enterprise green technology innovation, suggesting that the increase of government subsidies will probably weaken the positive effect of OFDI on corporate green technology innovation when it fall below a certain threshold, while strengthen their positive relation when it climb above this threshold.This is possibly caused by the power of policy signal on speci c supportive elds that effectively regulate enterprises' green-tech innovative behaviors as well as the resources acquired from OFDI.Thus, hypothesis 2 is proposed as follow: H2: The government subsidies has a positive U-shaped moderating effect in the positive relationship between OFDI and enterprises green technological innovation.

Moderating role of slack resources
Slack resources usually refers to the corporate excess available resources in a given time (Voss et al., 2008), and are expected to bene t the development of new technologies (Li and Gao, 2014), which is considered to be divided into unabsorbed slack resources and absorbed slack resources respectively (Sharfmam, 1988).Absorbed slack resources are characterized by poor liquidity and exibility, and existing in the form of private costs, which is often arranged in speci c projects but not easy to be recon gured.Accordingly, enterprises are likely to be bloated within the organization, lower e ciency of resource allocation, less sensitive to environmental changes, and thereby damaging the progress of new development strategies and investment projects when the absorbed slack resource continuously growing (Huang and Li, 2012; Argiles-Bosch et al., 2016; Wu and Hu, 2020).Thus, it is easy to form an inverted Ushaped adjustment effect in the relationship between OFDI and corporate green technology innovation.In terms of a appropriate scale of absorbed slack resource, enterprises are likely to be adaptive to the variation of external context, and exibly con gured according to the OFDI project requirements as well (e.g.international capital participation, international joint ventures, international enterprises buyouts and overseas subsidiaries).While the relative higher or lower scale of absorbed slack resource are both unfavourable for positively moderate the relation between OFDI and enterprises' green-tech innovation, for a higher scale will probably reduce the exibility of resource allocation, and a lower scale will also lead to the problem of insu cient resources.
Unabsorbed slack resources refer to resources that are available and exible to be allocated to any other projects with relative less constraints compare with the counterparts.Owing to the higher liquidity and stronger conversion capabilities, unabsorbed slack resources are often allocated in the company's new development strategy, such as new product research and development, new market expansion, and other context (Klingebiel and Rammer, 2014).Therefore, unabsorbed slack resources are likely to positively moderate the relation between OFDI and enterprises' green technology innovation, that is, the more storage of unabsorbed slack resources (e.g.argiles Bosch et al., 2016 for nancial resources), the more resources that enterprises can reconstruct and use for recon guration, and the more powerful support and guarantee can be provided for OFDI, so as to strengthen the role of green new technology research and development of enterprises.Thus, the hypothesis 3 is proposed as follow: H3a: The absorbed slack resources have an inverted U-shaped moderating effect in the positive relationship between OFDI and corporate green technology innovation.
H3b: Unabsorbed slack resources will strengthen the positive effect of OFDI on corporate green technology innovation.

Measurement of main variables
(1) Dependent variable.Enterprise green technology innovation(Gpatent) is measured by the green patent data applied by the enterprise.According to Li et al.,(2017) and Wu and Hu(2021), the green patent data were collected by screening the patents applied by the sample enterprises through the application of 14 keywords that are likely to re ect the speci c concept and meaning of green technology innovation.These keywords include environmental protection, energy saving, pollution control, water conservation, electricity saving, recycling, sustainable, clean, economical, emission reduction, green, low carbon, environmental protection and ecology.
(2) Independent variables.OFDI is expressed by the dummy variable of whether an enterprise has outward foreign direct investment behavior.It is assigned a value of 1 when an enterprise has an OFDI behavior in a certain year, and it is assigned a value of 0 when it has no OFDI behavior.
(3) Moderator.Government subsidies are measured by the natural logarithm of the total amount of government subsidies (in millions of RMB, lngov).Absorbed slack resources(asr) is the ratio of the sum of administrative expenses and sales expenses to operating income.Unabsorbed slack resources(uasr) are measured by the method of Wu and Hu (2020), which is the quick ratio of enterprise.
(4) Control variables.Control variables in the environmental eld.This paper uses the metrics of marketization(market) and economic development in the region where the company is located(econo).The marketisation degree in the region where the company is located is measured by Wang et al.'s (2018) China Marketisation Index.In this paper, the provinces with a marketisation index higher than the mean value are 1, and the provinces with a lower than average value are 0, thereby obtaining a virtual variable of the degree of marketisation of each province.For economic development, the value of the enterprise located in the eastern region is 1, otherwise it is 0 (Li et al., 2017).The control variables in the enterprise eld are expressed as ownership structure, enterprise size, Tobin's Q, leverage ratio, R&D expenses, corporate social responsibilities and equity incentive.Ownership structure that is the proportion of shares held by the largest shareholder(shr1).The enterprise size is measured by the total assets(in units of 100 million yuan, ).Equity incentive is the shareholding ratio of core technicians(tepo).Leverage ratio (lev) is expressed as the ratio of total liabilities to total assets.R&D expenditure (exp) is calculated by the ratio of R&D investment expenses to operating revenue.Corporate social responsibility (crs) is measured with the performance of enterprises in ful lling social responsibilities: whether to disclose the protection of rights, and interests of shareholders, creditors, employees, suppliers, and etc.If one of the above content disclosures is assigned a value of 1, the value is otherwise 0, and the resulting total score is 11 and the lowest value is 0.

Research model
Counting model is more proper than the linear model in our econometric analysis in terms of the dependent variable (green patent applications) characterized by discrete non-negative integers (Hausman, et al., 1984).Poisson regression and negative binomial regression are both considered to be appropriate econometric models for studying the knowledge production function based on patent data (Hausman, et al., 1984;Hall and Ziedonis, 2001;Wu and Hu, et al., 2020).The negative binomial regression model is selected according to the signi cant inconsistent between expectations and variance (see table 1) in this paper.Moreover, a standard negative binomial regression model is favored in this paper, for there are more than 75% numerical values of the dependent variable are non-zero integers.
Accordingly, the baseline model is structured as follows: Where, Moderators it refers to the intersection terms between OFDI and government subsidies (lngov or lngovsq), absorbed slack resources (asr or asrsq) and unabsorbed slack resources (uasr or uasrsq) respectively, which is used to test the moderating effect among government subsidies, redundant resources and their square terms.Controls it represents the control variables.and represent year xed effect and individual (industrial) xed effect respectively. is the random disturbance term.

Descriptive analysis
In this paper, we use STATA15 to conduct empirical tests.Table 1 presents the descriptive analysis of the main variables as well as their multicollinearity test.The rst to third quantile statistics value and mean value of green patent applications are 9, 81, 92, 73.297, respectively, implying that the willingness and output of green technology innovation among different high-tech companies are signi cantly different.The second quantile, third quantile and mean value of OFDI are 1, 1, 0.59, respectively, suggesting that more than half of the sample enterprises have OFDI behavior.In addition, the variance in ation factor (VIF) of each variable is much lower than 10, indicating that there is no obvious multicollinearity between the variables.Accordingly, relevant assumptions will be analyzed and explored based on the results provided by model 8.
As for H1, model 8 presents that OFDI has a signi cant positive impact on enterprise green technology innovation (α=73.29,P < 0.01), which is also supported by the other 7 models.Thus, hypothesis 1 is strongly veri ed.While government subsidies have a positive U-shaped regulation on the relationship between OFDI and enterprise' green technology innovation in model 8 (α=69.02,p < 0.1 for the interaction between the OFDI and government subsidies; α=25.01,p <0.05 for the interaction between the OFDI and squared term of government subsidies), which is supported by model 3 as well.Thus, H2 is also veri ed, and suggesting that the government subsidies have a moderating effect threshold in the positive action of OFDI, which below this threshold is the weakening effect of regulation, above this threshold is the strengthening effect of moderating.In terms of H3, model 8 presents that the absorbed slack plays a positive inverted U-shaped moderating role on the relationship between OFDI and enterprise' green technology innovation (α= -197.2, p < 0.05 for the interaction between the OFDI and absorbed slack; α= -271.1, p < 0.1 for the interaction between the OFDI and squared term of absorbed slack).Thus, H3a is supported.There is a moderate interval for the adjustment of absorbed slack resources.In this interval, the enterprise can not only maintain the organization's resilience, but also transform it into exible resources in the necessary scenarios to support the investment and development of the enterprise.Additionally, unabsorbed slack plays a positive regulating role on this process (α= 0.805, p < 0.1 for the interaction between the OFDI and unabsorbed slack; α= 0.0261, p > 0.1 for the interaction between the OFDI and squared term of unabsorbed slack).
Thus, H3b is supported as well.

Further research (1) Heterogeneity of property rights
Due to the different nature of property rights, there are signi cant differences in the technological innovation behaviors, objectives and business environment of enterprises (Aghion et al.,2013;Wu and Hu, 2020).Thus it is necessary to divide the total sample into state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs) for expansion testing.As shown in table 3, the Wald test suggests that the model is tted in this paper for all 4 models passed the joint signi cance level test of 0.001.While the 2 nd and 4 th model is favored in this paper for they have the smaller log likelihood value and Akaike information criterion (AIC) value compare with their counterparts.
Accordingly, relevant assumptions will be analyzed and explored based on the results provided by this two models.According to model 2 and model 4, in terms of the results for SOEs, OFDI of state-owned enterprises has a signi cant positive effect on their green technology innovation (α= 51.04, p < 0.01).Government subsidies play a positive Ushaped regulatory role in the positive effect of OFDI on enterprise green technology innovation (α=33.18,p < 0.01 for the interaction between the OFDI and government subsidies; α=13.04,p <0.01 for the interaction between the OFDI and squared term of government subsidies).Absorbed slack has an inverted U-shaped regulation on the relation between OFDI and enterprise green technology innovation (α= -108.0,p < 0.1 for the interaction between the OFDI and absorbed slack; α=-191.7,p < 0.1 for the interaction between the OFDI and squared term of absorbed slack), while unabsorbed slack has no signi cant moderating effect in this process (α= 2.170, p > 0.1 for the interaction between the OFDI and unabsorbed slack; α= 0.0421, p > 0.1 for the interaction between the OFDI and squared term of unabsorbed slack).
As for Non-SOEs, OFDI of non-state-owned enterprises has a signi cant positive effect on their green technology innovation (α= 117.3, p < 0.01).Government subsidies in such enterprises play a positive U-shaped regulatory role in the process of OFDI positively affect their green technology innovation (α=81.75,p < 0.01 for the interaction between the OFDI and government subsidies; α=24.17,p <0.01 for the interaction between the OFDI and squared term of government subsidies).Absorbed slack has no signi cant moderating effect on the relation between OFDI and rms' green-tech innovation (α=-238.3,p > 0.1 for the interaction between the OFDI and absorbed slack; α= -1216.0,p > 0.1 for the interaction between the OFDI and squared term of absorbed slack), while unabsorbed slack positively moderates this process (α= 0.312, p< 0.1 for the interaction between the OFDI and unabsorbed slack; α= 0.118, p > 0.1 for the interaction between the OFDI and squared term of unabsorbed slack).
(2) Robustness test In order to ensure the more robust conclusions, this paper has the following two scenarios to test the robustness that are re ected in the above research process: First, the whole samples are divided into two respective sub-samples for robustness test, namely, enterprises that above the average assets are classi ed as large-scale enterprises and the else are classi ed as small-scale enterprises.Second, this paper will replace control variables, which replaces the control variable indicator at the enterprise level with equity balance, ratio of independent directors, Salary incentives.equity balance is the sum of the shareholding proportion in the second to tenth largest shareholders (shr2-10); Ratio of independent directors (indir) is the proportion of independent directors on the board; Salary incentives which are expressed as the total annual salary of the directors, supervisors and senior managers (chsas, in millions of yuan).The results as shown in Table 4. (3) Endogenous analysis Enterprise green technology innovation behaviors are recognized as the strategic behavior of enterprises, which is likely to introduce self selection deviation due to the in uence of its own characteristics, such as enterprise scale and leverage ratio.Propensity Score Matching (PSM) is one of the important methods to solve the problem of selfselection bias problem (Rosenbaum and Rubin,1985).Thus, this paper will use this method to resolve potential estimation errors.This paper divides the total sample into two categories: enterprises with OFDI behavior and enterprises without OFDI behavior, and use the moderators and control variables in the basic regression as matching variables, use the nearest-neighbor one-to-one matching method to match the processing group and the control group.Also, the logit model is applied to estimate the propensity score, and evaluating the effect of OFDI on green technology innovation based on the average treatment effect ATT.  5. Discussion And Implications

Conclusions and discussion
This paper establishes a comprehensive research framework that integrates OFDI and enterprise green technology innovation, focuses on the effects of OFDI on the green technology innovation, discusses the moderating effect of government subsidies and slack resources on the role of this relationship and makes an in-depth comparative analysis of the above relationship from the property heterogeneity.The main conclusions of this paper are as follows: First, OFDI will positively affect corporate green-tech innovation as expected, and is likely to act as an effective approach to improve their technological level and innovation ability (Oscar and Carmen, 2015;Piperopoulos et al., 2017;Javorcik et al., 2018).Second, government subsidies have a U-shaped regulation on the relation between OFDI and green-tech innovation.Absorbed slack plays an inverted U-shaped moderating effect on the relation between OFDI and green-tech innovation, and the unabsorbed slack positively affect this process, which is consistent with the contribution of Wu and Hu (2020), Argiles-Bosch et al. (2016), that is, the absorbed slack is easier to be con gured for enterprise green technology R&D than the unabsorbed slack.
As for the heterogeneity of property rights, the test results of non-state-owned enterprises and state-owned enterprises are basically consistent with the baseline results, except for the following two points: the unabsorbed slack of stateowned enterprises has no regulatory effect between OFDI and enterprise green technology innovation, and the absorbed slack of non-state-owned enterprises has no regulatory effect between OFDI and enterprise green technology innovation.

Public policy implications
The implications of this article are: First, the positive effect of OFDI shows that the "going global" strategy that the government has implemented can signi cantly promote the green development of enterprises.Thus, the government can further explore the long-term guarantee mechanism of the positive effect of OFDI on the green technology innovation performance, and improve the mechanism for the resolution and prevention of a series of risks or obstacles that enterprises may face in the process of OFDI.Second, from the results of the moderating effect of resources, companies need to focus on the differences in the role of various resources, and reasonably construct e cient guarantee mechanisms for different types of resources based on the characteristics of each resource.Third, from the results of the heterogeneity of property rights, we need pay attention to the differences in property rights between enterprises, when strengthening the resource base in the positive role of OFDI.All resources should be allocated reasonably and effectively according to the property characteristics of enterprises.
There are some limitations of this framework.First, with limited data available, this paper only discusses the role of OFDI on corporate green technology innovation, which does not further examination of the role of different forms of OFDI on corporate green technology innovation.Second, due to data limitations, the data used in this paper is unbalanced panel data, which makes it impossible to discuss the long-term effects on the green technology innovation performance.This needs to be in-depth analysis after accumulating su cient data. Figures consumers, suppliers and scienti c research institutions and consequently promote their new technology R&D capability by both learning-by-doing and learning-by-searching effect (DeLocker 2013; Bena and Li 2014; Ye et al., 2018; Yi et al.,2020).

3. 1
Data and samplesAccording to the availability of relevant data set, this paper uses the high-tech enterprises listed in Shanghai and Shenzhen stock markets over the period of 2011 to 2018 as research samples.With reference to the industry categories identi ed by the classi cation of high tech industries (manufacturing industry) (2013) issued by the China Bureau of statistics, and based on the main business scope and industry types of Listed Companies in Shanghai and Shenzhen, 5572 observations were preliminarily obtained.The observed values are screened according to the following criteria: excluding enterprises listed less than 1 year; excluding special treatment (ST) rms, for their net pro t of two consecutive scal years is negative; excluding the enterprises with missing or zero value of government subsidies and redundant resources, for their incapability in re ecting the government support effect; the tail reduction treatment is carried out for the main continuous variables at 1% and 99% to eliminate the extreme values.Thus, a total of 4219 observations were obtained, corresponding to 693 enterprises.Based on the list of overseas investment enterprises (institutions) announced by theMinistry of Commerce, the China Stock Market & Accounting Research Database (CSMAR) of listed companies 'mergers and acquisitions and reorganization events, and searched the sample companies' o cial websites, annual reports, and announcements to obtain the years of enterprises and enterprises that had OFDI behavior during the inspection period.All other data comes from Wind Information Financial Terminal, the CSMAR Database, the State Intellectual Property O ce's o cial website, and Baiten's o cial website.The green patent data are from the State Intellectual Property O ce website and Baiten's o cial website.Other data are mainly from the Wind Information Financial Terminal and the CSMAR database.
; Huergo and Moreno, 2017;Argiles-Bosch et al., 2016; Leyva-de la Hiz et al., 2019; Wu and Hu, 2020).As external supporting resources of enterprises, government subsidy is expected to effectively alleviate the resource constraints in new technology R&D and new project investment process of enterprise (Bernini and Pellegrini, 2011; Cappelen et al., 2012; Huergo and Moreno, 2017; Bellucci et al., 2019).As available internal resources of enterprises, slack resources are likely to be redeployed to promote enterprises to develop new capabilities which enable rms to be adapted to unexpected internal and external uctuations and simultaneously seize potential opportunities (Lin and Liu, 2012; Huang and Li, 2012; Argiles-Bosch et al., 2016; Leyvade la Hiz et al., 2019).

Table 2
presents the hypothesis test results.The Wald test implies that all 8 models passed the joint signi cance level test of 0.001, suggesting that the model setting in this paper is appropriate.Models 1 to 8 provide baseline results of the casual effect that OFDI have on rms' green-tech innovation.While the 8 th model is more preferable in this paper for it has the smallest log likelihood value and Akaike information criterion (AIC) value among all the eight models.

Table 3
Basic Results of SOEs and Non-SOEs

Table 4
Results of robustness test t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01 From table4, model 1 and model 2 present the sub-sample results of robustness test.Model 3 and model 4 present the results of substituted control variables.From models 1 to 4, the robustness test results under the two scenarios are basically consistent with the benchmark results, except that unabsorbed slack has no signi cant positive regulatory effect in both scenarios.Overall, the baseline results are robust.

Table 5
shows the balance test of the matching variables in this paper of unmatched and matched.The absolute values of the bias of matched are less than 10%, and the t-test results are not signi cant, indicating that the samples matched the analytical requirements.

Table 6
ATT effect of OFDI on Green Technology Innovation Table6shows the standard errors of ATT, ATU and ATE and their corresponding signi cance test results.From table7, the coe cient of ATT is 91.612, with statistically signi cant, which indicates that after controlling other in uencing factors, the performance of green technology innovation of enterprises with OFDI behavior is 91.612 higher than that of enterprises without OFDI behavior.Thus, OFDI can signi cantly improve the performance of green technology innovation.