Long-term and short-term effects of green strategy on corporate performance: evidence from Chinese listed companies

Building on the resource-based view (RBV) theory, this paper aims to shed light on how does the implementation of green strategies affect enterprises’ performance. To distinguish the evolution of strategy implementation effect, we adopt a panel estimation strategy and gather data from 3869 listed companies in China from 2008 to 2019. Furthermore, we innovatively use the semi-supervised clustering algorithm to classify the companies according to whether they implement green strategies or not and then discuss long-term and short-term financial effects of implementing green strategies. Our study finds that the implementation of green strategy facilitates a company’s long-term performance but hampers its short-term performance. According to the moderating analysis, a green strategy could negatively impact a company’s financial performance by increasing debt ratios. The findings highlight the importance of implementing green strategies and the obstacles in the process of transforming enterprises to be green.


Introduction
Economics assumes that enterprises aim to maximize profits. However, the global pollution and environmental problems brought about by large-scale industrialization force enterprises to bear corresponding environmental responsibilities. The requirements of regulatory authorities and public pressure require enterprises to adjust their existing production mode, strengthen environmental management, and adopt a new kind of strategy, namely the green strategy. The concept of green strategy is first presented by Porter (1996), who emphasizes that enterprises no longer achieve economic gains at the expense of the external environment. Specifically, green strategy refers to the incorporation of environmental management into a firm's production and operation fields, such as establishing environmental goals, controlling pollution in production process, and developing green products (Lin et al. 2021b). In addition to aligning with the national sustainable development strategy, green strategies also allow enterprises to gain relative competitive advantages (Lioui and Sharma 2012;Bassetti et al. 2021).
Due to the limited production resources (the RBV theory), a firm needs to make a trade-off between profit strategy and non-profit strategy (such as green strategy). Extensive literature has been conducted to investigate the linkage between green strategy and corporate performance, as shown in Table 1. Part of the literature draws contradictory conclusions. A stream of studies has found a negative, win-lose relationship between green strategy and financial performance in the short term. The main reason for this is that implementing a green strategy will occupy production resources and exert a crowding-out effect on corporate investment (Ramanathan et al. 2010;Busch and Hoffman 2011;Lin et al. 2021b). Additionally, environmental management (e.g., ISO14001 certification) requires additional capital and support, which further increases the cost of production and operation for firms (Wagner 2005;López et al. 2007;Trumpp and Guenther 2017). By comparison, another stream of studies has found a positive, win-win relationship between green strategy and financial performance in the long run. When an organization's products and services are compatible with the environment changes, it can achieve Communicated by Eyup Dogan. long-term benefits over its opponents (Sharma 2017). Such organizations may also benefit from the first-mover advantage in the long run by cultivating superior brand recognition and customer loyalty. In general, green strategy enhances organizational performance as it (i) promotes efficient use of resources (Shu et al. 2016), (ii) converts wastes into recyclable outputs (Chen and Chang 2013), (iii) reduces emissions (Menguc et al. 2010), and (iv) helps to develop a competitive advantage (Chakrabarty and Wang 2012).
In conclusion, literature on the impact of green strategy on corporate performance mainly focuses on short-term analyses, the results of which are largely inconsistent due to differences in research settings and sample data structures (Wagner 2005;Ramanathan et al. 2010). Specifically, most relevant research is conducted in developed countries, and data structures are cross-sectional based on survey methods. Therefore, further exploration is needed to provide empirically sound evidence about the relationship between green strategy and corporate performance. On this basis, we make extensions by investigating long-term and short-term effects of green strategy on corporate performance based on the sample of Chinese listed companies.
To reconcile the seemingly contradictory opinions listed in Table 1, this study aims to provide a comprehensive illustration of how green strategy impacts firm performance. Specifically, we try to answer the following questions: What is the accurate relationship between implementing green strategy and firm performance? Are the long-term and short-term effects of green strategy consistent? When implementing green strategy is not conducive to enterprise performance, how to alleviate the negative effects? To investigate above issues, we focus on the impact of green strategy implementation among Chinese enterprises, and utilize panel data including 3869 firms with 32,468 firm-year observations from 2008 to 2019. On this basis, we adopt the semi-supervised clustering procedure of double-test to distinguish whether the companies implement green strategies, then obtain an effective estimate of green strategies. Subsequently, we deploy various empirical strategies, including a long-term and short-term performance analysis, a moderating effect analysis, and further study, which support our arguments. The empirical results show a short-term decline in firm performance as a result of the implementation of green strategies, but then a positive relationship is observed between green strategies and firm performance in the long run.
The current research seeks to make the following contributions. First, although there is some evidence on the relationship between green strategy and corporate performance, this paper is the first to evaluate both long-term and short-term effects on corporate performance using panel data. Second, this study probes into the underlying mechanisms of how a green strategy exerts inhibitory effect on the short-term financial performance for a company. Third, we propose a double-check procedure to obtain an effective estimate of green strategies. The first procedure is based on semi-supervised clustering algorithms, and the second procedure is to test the validity of clustering results based on the data of ISO4001 certification and environmental scores in corporate social responsibility (CSR).
The rest of the paper is organized as follows. In the "Literature review and hypotheses" section, we develop the conceptual framework and hypotheses based on extant literature. The "Methodology and variables" section presents the related methodology and data. In the "Further study" section, we test the hypotheses and further examine the validity of the results. The "Conclusion and implications" section concludes the paper.

Literature review and hypotheses
RBV assumes that firms are the integration of resources and that resources, especially non-substitutable resources, determine the performance of firms (Barney 1991;Hart 1995). Despite the widespread discussion regarding the relationship between resources and firms' performance, RBV has been criticized for its inability to uncover firms' resource management processes due to insufficient data. However, the green activities of a firm will shape its tangible and intangible resources, which in turn affect its performance (He and Shen 2019). Following the relevant literature, we ground our research on the RBV, summarize various green activities of enterprises as the implementation of green strategy, and investigate how green strategy affects enterprises' short-term and long-term performance. Figure 1 illustrates Long-term corporate performance the conceptual framework and the hypotheses in the study are presented in the following sections.

Green strategy
Initially outlined by Porter (1996), the concept of green strategy has evolved over time, and currently consists of three components: environmental behavior, green innovation, and pollution control (Bansal 2005;Montiel and Delgado-Ceballos 2014;Duque-Grisales et al. 2020). However, it is more important to determine whether enterprises implement green strategies. The evaluation of a green strategy involves two steps. The first step is to build an evaluation index system according to the definition of green strategy, and the second step is to analyze it by comprehensive evaluation methods. Comprehensive evaluation methods commonly used include questionnaire surveys, fuzzy comprehensive evaluation, and cluster analysis. Among them, the questionnaire survey aggregates all the same types of index data and uses factor analysis to ensure its accuracy; a fuzzy comprehensive evaluation blends subjective judgments and objective index weightings for a comprehensive analysis; and cluster analysis assigns classification rules to the output indices based on the input index data. Table 2 lists the research methods related to green strategy assessment.
As shown in Table 2, the majority of green strategy evaluation data is collected through questionnaires, and the measured items primarily address two aspects: environmental management and pollution control. While indicator effectiveness analysis based on questionnaire data can fully retain the data information, the non-public and non-verifiable characteristics of data collection pose major obstacles to questionnaire surveys. Similarly, the fuzzy comprehensive evaluation method lacks further verification of the evaluation results (Wang and Wang 2007). Nevertheless, semi-supervised clustering can compensate for these shortcomings. Specifically, cluster analysis can verify the effectiveness of the clustering results in the training set through semi-supervised learning and further expanding to extensive samples, which can indirectly verify the reliability of the analysis (Schmiedeberg 2010;Bynen 2012). K-means cluster analysis

Green strategy and performance
As shown in Table 1, most studies examining the relationship between green strategy and firm performance have yielded contradictory results. It is important to discuss the reasons behind the positive and negative correlations. When implementing a green strategy, the production and operation costs for enterprises will inevitably increase, especially in the initial stage. In detail, implementing a green strategy requires additional investment, such as purchasing green production equipment, introducing energy-saving and emission reduction technologies, and recruiting relevant technical talents (Hashai 2015;Leonidou et al. 2017). Due to the limitation and redeployment of production resources, this kind of investment increases production costs and limits current production capacity. Consequently, the cost of implementing a green strategy is greater than the benefit in the short term (Lin et al. 2021a). Based on arguments above, we propose the following hypothesis1: Hypothesis1: There is a negative relationship between green strategy and corporate performance in the short term.
However, green strategy's negative effects on firm performance are likely to decrease over time. In the long run, green strategy provides firms with differentiated products/ services and lower production risks, which then leads to greater competitive advantage and improved revenues. For instance, by developing green products and services that meet the needs of consumers, enterprises attain the first-mover advantage (Leonidou et al. 2017). Moreover, green initiatives facilitate firms' establishment of good reputations among environmental stakeholders, including environmentalists and regulatory authorities, which are conducive to market expansion and can reduce the risk of environmental punishment (Lin et al. 2021b). Taken together, the benefits of implementing green strategies will rapidly exceed the costs of redeploying production resources, resulting in increased performance for companies in the long run (Trumpp and Guenther, 2017), thus leading to the second hypothesis: Hypothesis 2: There is a positive relationship between green strategy and corporate performance in the long term.

The moderating role of debt ratio
The implementation of a green strategy will squeeze the capital and labor originally used for production, thereby increasing the short-term debt burden of the company (Chang et al. 2021;Li et al. 2021a, b). In this regard, green strategy implementation drives firms to set environmental goals and adjust the resource allocation plans of the production and non-production departments. According to the RBV theory, the principal production resource (the capital used for production) will decrease, and more capital investment will be required in order to maintain the current output level, resulting in an increase in the debt ratio of enterprises.
Taken together, the company's financial performance is restrained by the increase in its debt ratio (Yao et al. 2021). Specifically, high-debt companies are not only subject to higher funding rates and interest rates (Binsbergen et al. 2010), but also have higher financial risks, which negatively affect their solvency, corporate credit, and refinancing capabilities (Allayannis et al. 2003). In summary, high debt ratios associated with green strategies could negatively affect enterprise performance. Therefore, the level of debt will have a negative moderating effect on the relationship between green strategy and firm performance. Based on these arguments, we propose the following the third hypothesis: Hypothesis 3: A firm's debt ratio negatively moderates the relationship between green strategy and corporate performance in the short term.

Measurement of green strategy
Existing literature shows that the evaluation of green strategy implementation mostly takes developed countries as the research background. Among the few studies on the evaluation of green strategy of Chinese enterprises, most of them use a single index method or questionnaire survey method. For example, He and Shen (2019) utilize the ISO 4001 certification to measure a company's environmental management capabilities; Pan et al. (2020) survey 524 Chinese manufacturing enterprises to examine the implementation of positive environmental strategies from the perspective of green products and environmental protection. Therefore, comprehensive evaluation methods and standardized evaluation processes are required. This section adopts a doublecheck evaluation procedure to analyze the implementation of green strategies by Chinese enterprises. First, we introduce a clustering algorithm and then construct an index system for the implementation of green strategies. On this basis, we cluster large samples using the semi-supervised algorithm, and finally, evaluate the clustering results.

WKFCM algorithm
The process of dividing a collection of physical or abstract objects into multiple classes composed of similar objects is called clustering. The cluster generated by clustering is a set of data objects similar to the objects in the same cluster and different from the objects in other clusters. The first time the clustering method is used for strategic evaluation is by Harrigan (1985), who employs factor analysis and simple clustering to distinguish strategic groups. In subsequent studies, Fuzzy-C-Means (FCM) in clustering algorithms is further expanded and studied (e.g., Budayan et al. (2009).
However, the traditional FCM clustering algorithm has two defects: 1) the constraint condition that the sum of the membership value must be 1 makes it sensitive to outliers and noises; 2) it is an iterative descent algorithm, which makes the initial clustering center sensitive and difficult to converge to the global optimum. Therefore, we improve the FCM algorithm.
With the improved FCM algorithm based on the kernel, the original points are mapped into the feature space via kernel functions. Since the original space cannot be divided by a linear function, it can be transformed into a higherdimensional space where a linear function can be found, so the original data can easily be divided. This high latitude space is called feature space, and the inner product of the mapping function from low latitude to high dimensional space is called kernel function. Furthermore, introducing kernel functions into machine learning is a meaningful way to increase computing performance when the number of features in feature space is large, and the computation of kernel functions is relatively tiny compared to the inner product of feature space. Therefore, the kernel-based FCM (KFCM) algorithm improves the clustering performance and makes the algorithm robust to noises and outliers.
Consider t he nonlinear mapping function ϕ ∶ x → ϕ(x) ∈ R D k , where D k is the dimension of the eigenvector x . We do not need to know x 's transformation explicitly, just its dot product instead, which is ϕ(x) • ϕ(x) = κ(x, x). Kernel functions can take many forms, including polynomials and radial basis functions (RBF). Given a kernel function κ , KFCM is usually defined as a constraint minimization problem, and the objective function is as follows: is the distance based on a kernel between the i eigenvector kernel and the k eigenvector, which can be calculated by kernel matrix, defined as We can get the following equation after simplification: Similar to FCM algorithm, clustering centers can be linearly combined by eigenvectors: where u ij is defined as: Traditional KFCM believes that every object is equally important, but this is not the case in reality. Therefore, to make sure that each object's relative importance is shown, eigenvector weights are used to serve as clustering centers, as calculated by: where, w l is a set of pre-determined weights used to define the weight influence of each eigenvector. Furthermore, cluster centers are more influenced by objects with a high weight. Therefore, a weighted kernel-based FCM (WKFCM) algorithm is proposed.

Indicators for green strategy
Green strategy indicators are primarily selected from two directions: existing survey databases and questionnaire surveys based on specific purposes. In specific, Thomson Reuters' (Duque-Grisales et al. 2020;Lartey et al. 2020;Lin et al. 2021b) database, which relates to enterprise green development strategy, environment and resource management, and green product production, forms the basis for researching enterprise green strategy indicators. In addition, De Mendonca and Zhou (2019) and Lin et al. (2021b) expand the scope of the green strategy indicator system to varying degrees. By combining the environmental disclosure content of listed companies in China using the CSMAR database, we develop 25 indicators that can be used to evaluate green strategies for Chinese listed companies. Table 3 shows the content and definition of each evaluation indicator.
There are four categories of green strategies discussed in this section. Environmental management macroscopically describes the company's institution construction on environment; pollution control concerns whether pollution control systems have been implemented by the enterprise; cleaner production focuses on the environmental behavior of the enterprise in the production process; and pollution discharge records the pollutant emissions of the enterprise. The sample contains 3869 listed firms with 32,468 firm-year observations, and the descriptive statistics of each indicator are shown in Appendix Table 11. If the company's past environmental protection targets are fulfilled and future environmental targets are disclosed, it is assigned a value of 1; otherwise it is 0 3. Does the company have an environmental management system?
If it is disclosed that the company has formulated a series of management systems such as related environmental management systems, regulations, and responsibilities, it is assigned a value of 1; otherwise, it is 0 4. Has the company organized environmental education and training?
If the company's participation in environmental protection education and training is disclosed, it is assigned a value of 1; otherwise, it is 0 5. Does the company have special environmental protection expenditure?
If the company's participation in environmental protection related special activities, and other social welfare activities is disclosed, it is assigned a value of 1; otherwise, it is 0 6. Has the company established an emergency mechanism for environmental incidents?
If the company's establishment of a response mechanism for major environmental-related emergencies, the emergency measures it has taken, and the handling of pollutants are disclosed, it is assigned a value of 1; otherwise, it is 0 7. Has the company implemented the "Three Simultaneities" system?
If the company's implementation of the "Three Simultaneities" system is disclosed, it is assigned a value of 1; otherwise, it is 0 8. Has the company established the environmental information disclosure system?
If the company separately discloses environmental reports or discloses environmental-related information in the social responsibility report, it is assigned a value of 1; otherwise, it is 0 Pollution control 9. Has the company established a waste gas treatment system? 0 = no description; 1 = qualitative description; 2 = quantitative description (currency / numerical description) 10. Has the company established a wastewater treatment system? 0 = no description; 1 = qualitative description; 2 = quantitative description (currency / numerical description) 11. Has the company established a dust treatment system? 0 = no description; 1 = qualitative description; 2 = quantitative description (currency / numerical description) 12. Has the company established a solid waste recycling system? 0 = no description; 1 = qualitative description; 2 = quantitative description (currency / numerical description) 13. Has the company established a noise, light pollution, and radiation treatment system? 0 = no description; 1 = qualitative description; 2 = quantitative description (currency / numerical description) Cleaner production 14. Has the company established a cleaner production system? 0 = no description; 1 = qualitative description; 2 = quantitative description 15. Is the company up to standard in pollutant discharge?
If the pollutant discharge reaches the standard, it is assigned a value of 1; otherwise, it is 0 16. Is there any sudden environmental accident in the production process of the company?
If there is no sudden major environmental pollution incident, it is assigned a value of 1; otherwise, it is 0 17. Has the company committed any environmental violation in the production process?
If there is no environmental violation, it is assigned a value of 1; otherwise, it is 0 18. Does the company have environmental petition cases in the production process?
If there is no environmental petition event, it is assigned a value of 1, otherwise; it is 0 19. Is the company a key pollution monitoring unit?
If the company is a key monitoring unit, it is assigned a value of 1, otherwise; it is 0

Comparison of clustering methods
Compared with factor analysis based on questionnaire data and fuzzy comprehensive evaluation, semi-supervised learning has the following characteristics. We train classifiers that consider the relationship between input and output variables based on the small samples with labels in semi-supervised learning and then apply them to other large samples, thus ensuring that the clustering process and results are reliable. The first step of the semi-supervised clustering algorithm is to select the algorithm with the highest clustering accuracy based on the data with the existing labels. Our training set is the IRIS data set, consisting of 150 iris characteristics with five dimensions: calyx length, calyx width, petal length, petal width, and each iris variety (Setosa, Versicolor, Virginia). A good clustering effect for the WKFCM algorithm means that the difference between four-dimensional feature data and natural varieties is the smallest, so the accuracy of clustering is high. After calculation, the accuracy of the FCM algorithm is 89.33%, and that of the WKFCM is 91.33%. Therefore, the validity of WKFCM clustering algorithm is verified.

Clustering results
Subsequently, the WKFCM algorithm is applied to the cluster analysis of large samples. Based on the green strategy indicator data from Chinese listed companies, Figs. 2 and 3 illustrate how the WKFCM clustering algorithm distinguishes whether companies implement green strategies. It can be seen from Fig. 2 that the number of companies implementing green strategies has increased at a rising rate. Specifically, compared to that in the 12th Five-Year Plan period (2011)(2012)(2013)(2014)(2015), the number has grown significantly in the 13th Five-Year Plan (2016-2020). Due to the apparent time effect of the number of companies implementing green strategies, it is appropriate to analyze the companylevel green strategy issues using the panel data structure.  The distribution of companies implementing green strategies in Fig. 3 shows that most of these companies are in the manufacturing industry. It is, therefore, essential to take into account the heterogeneity of industries when discussing green strategy and company performance.

Validity tests of clustering results
Although we have found that the WKFCM algorithm has a higher classification accuracy, and completed the first procedure of double-checking, we still need to test the effectiveness of classifying whether Chinese listed companies implement green strategies based on the WKFCM algorithm, and to complete the second procedure. To measure the effectiveness of clustering, the idea of Li et al. (2021a, b) is referred to and the correlation between clustering results and other proxy variables of green strategy is tested. We choose the ISO4001 certification and environmental score in CSR as the proxy variables of green strategy for three reasons. First, the ISO 4001 environmental certification measures the corporate environmental management level, whereas the CSR environmental scores indicate the impact of corporate environmental management. Accordingly, these variables are reasonable indicators of green strategy. Second, a few studies have used the ISO4001 certification to determine if enterprise green strategies are implemented, including He and Shen (2019) and Kraus et al. (2020). Third, a large number of Chinese enterprises and organizations have responded positively to the ISO14001 environmental certification standard since its release in 1996. According to the data of the National Certification and Accreditation Supervision and Administration Commission in 2021, 1 there are 335,841 organizations that have passed the ISO14001 environmental management system certification and remain effective in China. Table 4 presents the results of the validity tests concerning the clustering results. In columns (1) and (2), we find that the green strategy obtained by the clustering method is positively and significantly associated with the ISO 14001 certification, and this positive association is retained even when industry and year fixed effects are taken into account. Similarly, columns (3) and (4) show that green strategy is also positively and significantly associated with environmental scores in CSR, and the positive correlation is robust after the industry and year fixed effects are controlled. In summary, the validity tests reported in Table 4 verify that our cluster estimation is correlated with the concept of green strategy in existing literature and has performed as expected.

Data and variables
The cluster analysis has allowed us to effectively categorize Chinese companies according to whether they implement green strategies. Next, using the CSMAR database, we select related performance indicators, control variables, and moderating variables to examine the long-term and short-term effects of implementing green strategies for companies. The accounting-based and market-based indicators pertain to company performance. To go into detail, accounting-based indicators reflect owners' interests based on capital return, which can be further divided into return on assets and return on equity; market-based indicators reflect managers' performance and indicate profitability, characterized by return on sale and net profit. With the core explanatory variable of this paper being green strategy, we obtain the effective estimation of implementing green strategies by the WKFCM algorithm based on four categories including 25 indicators. In addition to the core explanatory variable, the company's size, age, investment, R&D, and degree of competition in its industry will also affect its financial performance. Hence, there are selected as the control variables. To analyze how a green strategy implementation impacts enterprise financial performance, we select the debt ratio as the moderating variable. All variables are defined and described in Table 5, and Appendix Table 12 provides descriptive statistics and bivariate correlation coefficients for all variables examined in the study.

Empirical models for hypotheses
We predict that the implementation of a green strategy may inhibit shor t-ter m cor porate per for mance (Hypothesis 1), but that it may promote long-term corporate performance (Hypothesis 2). Additionally, we propose that the debt ratio has a moderate effect on the short-term performance of enterprises, meaning that enterprises implementing green strategies will incur higher debt, which will negatively affect their short-term performance (Hypothesis 3). To sum up, we establish the following empirical model. The logarithm of net operating revenue (total operating incomeoperating costs-related taxes) The ratio of R&D expenditure to the compan's sale revenue CSMAR He and Shen (2019) HHI It represents the Herfndahl-Hirschman Index, which is calculated by the proportion of the company's total revenue to the industry's total revenue CSMAR He and Shen (2019)

Debt_Ratio
The ratio of total debts to the company's total assets CSMAR Chang et al. (2021); Lin et al. (2021b) where Y it is a dependent variable, representing ROA, ROE, ROS, and net profit;Moderator it stands for the moderating variable;X it is a vector of control variables for firm i in year t . In addition, we include a complete set of firm and year dummies to address fixed, unobserved firm heterogeneity. Especially, t is a set of year fixed effects, and i is a set of firm fixed effects. As the firm fixed effects already capture fixed industry differences, industry dummies cannot be included (He and Shen, 2019).

Cross-sectional dependence
Before carrying out the main regression, we refer to Pesaran (2021) and employ the Pesaran cross-sectional dependency test (Pesaran's CD-test) to determine whether cross-sectional dependence exists within the data.
The Pesaran's CD-test statistic is computed as: (1) where ̂ ij is the correlation coefficient between the timeseries for each panel variable. The null hypothesis of this test is cross-sectional independence. Table 6 displays the results of the Pesaran's CD-test for all variables. The null hypothesis of cross-sectional independence is rejected since the p-values are less than 0.01, indicating that the variables studied have cross-sectional correlation. Therefore, when performing the above empirical models, it is necessary to solve the cross-section correlation problem.

Green strategy and short-term corporate performance
Hypothesis 1 is examined first, which predicts that a green strategy will inhibit corporate performance in the short term.
To solve the problems of cross-section correlation, possible autocorrelation, and heteroscedasticity, and to obtain an effective estimation, we use Driscoll-Kraay standard errors. The results are reported in Table 7. With regard to the variable of green strategy, the estimated coefficient in Column (1) is negative and statistically significant at the p < 0.01 level, indicating that green strategy significantly restrains corporate's ROA. Similarly, Columns (2)-(4) also verify the negative correlation between green strategy and ROE, ROS, and net profit, thus supporting our prediction in Hypothesis 1, which is consistent with Hashai (2015) and   Leonidou et al. (2017). The short-term negative impact of green strategies is due to two possible reasons, first, the cost of implementing the strategies outweighs their benefit, and second, the expectation of expanded market demand through the implementation of the strategies has not yet materialized.

Green strategy and long-term corporate performance
Second, we verify whether implementing a green strategy will improve the company's financial performance over time (Hypothesis 2). To visualize the time effect of green strategy implementation, we distinguish between the current effect, the medium-term effect (2 years after the implementation), and the long-term effect (4 years after the implementation). We take ROA, ROE, ROS, and net profit as the explanatory variables to test the dynamic effect of green strategy. The results are reported in Table 8. Taking ROA as an example, the first column represents the impact of implementing a green strategy on enterprises' current financial performance. In column (2), we estimate the medium-term impact of implementing green strategy on enterprises' ROA after 2 years, whose estimated coefficients are negative at the p < 0.05 level and smaller than the current impact, suggesting that the negative impact will weaken with time. A similar conclusion can be drawn from Column (3), which shows that 4 years after implementing the green strategy, the company has improved its financial performance, thus verifying Hypothesis 2. Additionally, hypothesis 2 is further supported by using ROE, ROS, and net profit as the explained variables. One possible explanation is that the implementation of green strategies has brought enterprises long-term dividends, such as resource savings, efficiency improvements and technological upgrades, which consequently led to increased profits (Brulhart ei al. 2019;Leyvade la Hiz et al. 2019).

Moderating analysis
As suggested by Hypothesis 1, implementing green strategies will not be beneficial for the short-term financial performance of enterprises, which negatively affects their desire to implement green strategies. However, the sustainable development and green revolution of the whole industry are inevitable, so exploring the critical obstacles to implementing a green strategy is essential.
In the discussion of moderating effects, we claim that implementing a green strategy leads to higher debts, thereby negatively affecting the short-term performance of enterprises. To test Hypothesis 3, we employ the analytical framework of Formula (3) to verify the moderating effect. The results are reported in Table 9. As illustrated in Columns (1)-(4), high debt ratios have a detrimental impact on business performance, and the cross-term of green strategy and debt ratio negatively impacts enterprise performance at the same time. However, the implementation of green strategies promotes the financial performance of enterprises if debt-related moderating effects are controlled, indicating that debt problem is a significant obstacle for companies to implement green strategies. To encourage enterprises to implement green strategies, government subsidies and incentives policies for green loans are needed (Bi et al. 2017;Chang et al. 2021). In addition, we take the explanatory variable of ROA as an example to examine the interactive effects, as illustrated in Fig. 4, which supports Hypothesis 3.

Further study
Our empirical analysis has examined the impact of implementing green strategies on the long-and short-term performance of companies. We assume that the anti-incentive effect on short-term performance will adversely impact the strategic decision-making of enterprise decision-makers. To draw reliable conclusions, this section will examine the robustness of the short-term effect of green strategy based on different data structures and sample ranges.
Using the pooled OLS model and the clustered robust standard error, we evaluate the overall effect of green strategy implementation on enterprise financial performance comprehensively. After controlling the fixed effects of time and individual, the pooled OLS method shows that the implementation of green strategies does not improve enterprise financial performance under the cross-sectional data structure, indicating that the negative effects of green strategy implementation is robust. According to Coles et al. (2001), extreme values of the sample have an impact on the estimation results, so we remove the outliers less than 1% and greater than 99%, then we verify the link between green strategy and short-term performance, and it is a robust negative relationship, as indicated by Table 10.

Conclusion
New development patterns require economic growth to harmonize with ecosystems, and the implementation of green strategies by enterprises is essential for sustainable development models (Olson, 2008). Additionally, quantifying the benefits of executing a green strategy is a significant factor in making a company's strategic choice . Based on the theory of RBV, we have investigated how the adoption of green strategies by firms affects their short-term and long-term financial performance based on a large short panel data structure composed of 3869 listed companies in China from 2008 to 2019. Furthermore, we innovatively use the semisupervised clustering algorithm to classify the companies according to whether they implement green strategies and adopt a double-test procedure to verify the effectiveness of the clustering results. Our study finds that the implementation of green strategies facilitates the company's long-term performance but hampers their short-term performance. To mitigate the short-term negative effects of green strategies implementation, the debt ratio of enterprises is a vital issue.

Implications for policymakers
This study provides significant implications for policymakers. Specifically, as hypothesized, the implementation of green strategies will impede enterprises' short-term financial performance, but improve their long-term financial performance. Focusing on corporate debt, we investigate the short-term barriers to green strategy implementation and find that firms implementing green strategies will have higher debt ratios, which means higher debt costs and financial risk, thus negatively impacting short-term performance (Leyva-de la Hiz et al. 2019; Yao et al. 2021). In view of this, policymakers are suggested to incentivize firms to implement green strategies by subsidizing green products and technologies or providing green credit support for their green development (Zhang et al. 2021).

Implications for firm managers
The current study conveys essential ethical and economic incentives for firm managers as well. Notably, a company's development is often constrained by limited resources and underdeveloped management skills, and implementing a nonmarket strategy is risky (Sharma 2017). Thus, firm managers tend to be hesitant about reallocating limited resources for green strategy implementation and ponder how it can benefit them (Chakrabarty and Wang 2012). We have removed some concerns of managers by showing that a green strategy will provide a long-term benefit. Moreover, sustainability-oriented strategic goals will shift firms' focus to competitive advantage, resource allocation efficiency, and sustainable development. Therefore, firm managers, especially those in the manufacturing industry, are suggested to actively formulate and implement green strategies.

Limitations and future research
Although our study provides new insights into the differences between long-term and short-term performance when implementing green strategies, the following limitations can reveal avenues for further research. First, validity tests for green strategy clustering need to be enhanced. To verify the effectiveness of the clustering algorithm in this paper, in addition to the validity tests shown in Table 4, text mining technology needs to be employed to analyze the annual reports of enterprises and gather information on their current implementation state. Second, the empirical strategy for the long-term performance analysis of green strategies needs to be improved. In this paper, the long-term analysis is limited by the short panel data structure and relies on lagging explanatory variable, which is relatively simple. Even though this method can intuitively evaluate the impact of implementing a green strategy on financial performance after two and four years, it does not take into account other factors that may affect the estimates, nor does it account for the time change of implementing the strategy. Therefore, further research is required to overcome the above shortcomings, including the adoption of long panel data structures and the introduction of a time-varying parameter model. Lastly, despite debt ratios moderate the interaction between green strategy and short-term performance, further investigation of debt structures, debt costs, and debt risks of firms is needed to understand the financial constraints on implementing green strategies.   Table 12 Author contribution Weihua Yu: conceptualization, supervision and writing-reviewing. Xin Jin: data curation, methodology, software, visualization, writing-original draft and editing. All authors read and approved the final manuscript.

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations
Ethics approval and consent to participate Not applicable.
Consent to publication Not applicable.

Competing interests
The authors declare no competing interests.