Do Corporate Social Responsibility Categories Distinctly Influence Innovation? A Resource-Based Theory Perspective

Abstract

Despite the existing vast literature on corporate social responsibility (CSR), there is a lack of research on the influences of CSR categories (i.e., employees; suppliers, customers, and consumers; environment; and social welfare). The objective of this paper is to investigate the influences of distinct CSR categories on firm innovation from a resource-based theory perspective. Based on a sample of Chinese A-share listed firms from 2010 to 2017, we find that employee-, supplier, customer, and consumer-, and environment-related CSR promotes innovation, while society welfare-related CSR reduces firm innovation. We also examine the distinct mechanisms of distinct CSR categories to influence innovation. Our findings mainly contribute to the literature on the association between CSR and innovation by considering the different influences of distinct CSR categories.

1. Introduction

With the development of the Internet economy, artificial intelligence, and big data technology, innovation has become the supporting strategy for the sustainable growth of firms. At the same time, firms have become increasingly aware of the importance of corporate social responsibility (CSR) because of public attention to environmental, social, and sustainable development issues. CSR and innovation are playing a more and more important role in the modern economy and have become two important issues in the era of the knowledge economy. Previous studies discuss the association between CSR and firm innovation but fail to draw consistent conclusions. Some studies show that CSR can promote innovation [1] by accelerating valuable external information exchange between stakeholders [2], identifying new business opportunities from social needs [3], and increasing employee participation and confidence [4]. However, other studies find that CSR can cause financial burdens, weaken firms’ competitiveness, and negatively affect innovation [5]. Gallego-Álvarez et al. [6] show that research and development (R&D) investment may be incompatible with encouraging firms’ sustainable behavior. Mithani et al. [7] also believe that CSR and innovation cannot easily be combined. The causal relationship between CSR and innovation is complex and empirically questionable. Despite these vast findings, there is a lack of research on the influences of CSR categories on innovation. According to the resource-based theory, firm development depends on the quantity and quality of resources. Firms conduct distinct CSR categories for different reasons which may bring different resources to firms. This directly influences firms’ innovation performance. Therefore, this study contributes to previous literature by exploring the association between distinct CSR activities and firm innovation from the perspective of resource-based theory.

According to the degree of resources that stakeholders bring to the firm, this paper defines the key stakeholders as follows: employees, suppliers, customers, and consumers (SCC), environment, and social welfare. Therefore, we distinguish four categories of CSR and test the relationship between four distinct CSR categories and innovation based on a sample of Chinese A-share listed firms from 2010 to 2017. In addition, we also test the mechanism of different categories of CSR on innovation. We use the HeXun index to measure the performance of four categories of CSR activities.

Our results show that employee-, SCC-, and environment-related CSR positively promote innovation, while social welfare-related CSR reduces firm innovation. We also find that employee-related CSR promotes innovation by helping firms increase human resources; SCC-related CSR promotes innovation by helping firms obtain cooperative resources from SCC; environment-related CSR promotes innovation by helping firms get government-related resources such as subsides. The results are still valid to a series of robustness tests, such as fixed effect regression, Tobit regression, instrumental variable method estimation, and exclusion tests (excluding the impact of economic recession and political uncertainty).

This study makes several contributions. First, it contributes to the stream of literature on CSR influences. This study distinguishes the impacts of four distinct CSR activities on innovation based on the resources-based theory. CSR activities towards different stakeholders bring different resources to firms and thus influence innovation differently. Our findings, to a certain extent, help to explain the contradiction in existing studies. Second, this study contributes to the literature that explores the drivers of innovative activities by discussing the driving mechanisms of innovation from a resource-based perspective. This study implies that the driving force of innovation is the acquisition of direct resources, which helps firms to make innovation decisions. Finally, this paper also contributes to the studies on the relationship between CSR and innovation in developing countries. Differing from previous findings of no relationship between CSR and innovation in developing countries [5], our study shows that certain categories of CSR have a significant positive impact on innovation in China. This provides different empirical evidence for understanding the relationship between CSR and innovation in developing countries.

The remainder of this study is organized as follows. Section 2 reviews the previous studies and provides research hypotheses. Section 3 describes the research design and sample selection. Section 4 reports the results of empirical analysis. Section 5 concludes the paper.

2. Literature Review and Research Hypotheses

Most studies of the relationship between CSR and innovation indicate a positive impact. Based on the resource-based theory, McWilliams and Siegel [8] argue that CSR creates process or product innovation because firms have experienced the upgrading of systems, products, and services in the process of fulfilling CSR. Bansal [9] also proves the positive correlation between CSR and innovation. He believes CSR sets high standards for products, production processes, and marketing means, so firms will take the initiative to increase R&D investment. Little [10] argues that CSR can promote firms’ innovation transfer to other organizations in their supply chain using information transmission. Moreover, MacGregor and Fontrodona [11] use the case of Metalquimia to show that CSR diffusion and innovation promote each other and are intertwined value-added mechanisms. Wagner [3] also finds a positive association between CSR and innovation, which is more significant in large family firms. Luo and Du [12] find that the CSR projects of firms can establish a good interactive relationship with stakeholders, which is conducive to the sharing and exchange of external knowledge and internal knowledge, thus having a positive impact on innovation. This positive relationship is stronger for firms with higher R&D investment and firms operating in more competitive markets. Using Spanish small and medium enterprise data, Martinez-Conesa et al. [13] show that CSR improves innovation performance. Using agribusiness data from Murcia, Spain, Briones Peñalver et al. [1] find that CSR positively influences innovation in agribusiness. Furthermore, Ueki et al. [14] prove that skilled employees are key elements in the positive relationship between safety-oriented CSR and innovation. Afridi et al. [15] find that authenticity and employee volunteerism play intermediary roles in the positive relationship between perceived CSR and innovative work behavior. Santos-Jaén et al. [16] show that the positive effect of CSR on innovation is mediated by debt terms and by good behavior. Bacinello et al. [17] unpack CSR into three dimensions, i.e., social, economic, and environmental, according to the triple bottom line perspective and measures sustainable innovation as a single construct, also finding a positive link between CSR and sustainable innovation.

However, some studies suggest no relationship or even a negative correlation between CSR and innovation. Gallego-Álvarez et al. [6] analyze the bidirectional relationship between CSR and innovation according to the resource-based theory and find that CSR has a negative impact on innovation and vice versa. The reason is that both CSR and innovation need capital investment. CSR increases the operation cost of firms, resulting in the reduction of innovation investment. Bocquetr et al. [18] prove that CSR is negatively related to innovation based on agency theory. There are management speculations in the performance of CSR, which affects the firm’s internal operation efficiency and inhibits innovation. Moreover, Mithani [7] finds that R&D has a larger impact on economic performance than its contribution to the socio-ecological environment. Ecological investments in the environment weaken the R&D effect because managers’ attention to innovation can be undermined by their emphasis on environment-related CSR. Furthermore, using data from 150 companies in 12 developing countries, Ullah and Sun [5] find no significant relationship between CSR and innovation in developing countries.

Other studies show that the relationship between CSR and innovation is not homogeneous; it depends on the type of innovation and CSR dimension. According to Lankoski [19], the impact of CSR depends on passive or active responses, which can usually be distinguished into responsive and strategic CSR. Using a survey of Luxembourg firms, Bocque et al. [20] reveal the differentiated effects of strategic versus responsive CSR on technological innovation. The results show that strategic CSR has a positive effect on innovation. Poussing et al. [21] come to the same conclusion using 2018 survey data from Luxembourg firms. Moreover, Costa et al. [22] find that CSR has a positive impact on exploratory innovation and a nonsignificant effect on exploitative innovation. Furthermore, Pan et al. [23] unpack CSR into three dimensions, i.e., finance, society, and environment, and divide ecological innovation into two kinds, i.e., pollution prevention and sustainable environmental innovation, to study the relationship between CSR and ecological innovation. The results show that environmental CSR is positively related to pollution prevention innovation and that social and financial CSR have no effect on their relationship. There is a U-shaped relationship between environmental CSR and sustainable environmental innovation. Social and financial CSR has a positive moderating effect on the relationship between them. Using survey data from Spanish firms, García-Piqueres and García-Ramos [24] divide CSR into three dimensions, i.e., economy, society, and environment, and innovation into three types, i.e., product, process, and organizational, and find that the relationship between CSR and innovation is not homogeneous.

Through literature review, we find that scholars draw inconsistent conclusions on the relationship between CSR and innovation according to different theories. Even using the same theory, scholars have different empirical results on the relationship between CSR and innovation. For example, based on resource-based theory, McWilliams and Siegel [8] prove that CSR is positively related to innovation, while Gallego-Á lvarez et al. [6] draw the opposite conclusion. Scholars have realized that the relationship between CSR and innovation is complex because there are different classifications of CSR, and the economic consequences of fulfilling different types of CSR are different. Scholars have classified CSR from the following perspectives: (i) responsive and strategic. (ii) triple bottom line (economic, social, and environmental). Considering that the existing literature is more based on resource-based theory to study the relationship between CSR and innovation, the present study examines the relationship between CSR and innovation by identifying key stakeholders according to their degree of resource acquisition. Specifically, the present study analyzes the relationships between employee-related CSR and innovation, SCC-related CSR and innovation, environment-related CSR and innovation, social welfare-related CSR and innovation, and total CSR and innovation.

2.1. Employee-Related CSR and Innovation

Firm innovation depends on continuous resources input and efficiency improvements [25]. Human resources is a key link for innovation [26] because employees are the source of many innovative ideas [27], and their efforts and collaboration are related to the implementation efficiency of innovation decisions [28]. How to stimulate employees’ willingness to innovate [29] and how to maintain a stable innovation team [30] are key factors to improve firms’ innovation efficiency. Employee-related CSR is helpful to achieve the above goals.

Employee-related CSR helps to reduce the employee turnover rate, accumulate human capital, and form a stable innovation team. First, employee-related CSR enables employees to experience mutual benefits [31]. According to the theory of social exchange, the interaction between people is essentially an exchange relationship [32]. Exchanged resources include not only material resources but also nonmaterial resources such as emotion, information, and reputation [33]. Firms provide employees with a rich salary, good welfare, and a positive organizational atmosphere, among other resources. Such employees will correspondingly repay their firms with their work and innovation [34]. Moreover, employee-related CSR enables employees to identify with their organization [35]. According to the organizational identity theory, individuals always put themselves into a certain group and connect their success or failure with the organization [36]. Employee-related CSR is regarded as a significant signal to evaluate the attractiveness and uniqueness of an organization [37], which is an important driving force for employees’ identification with the organization [35]. As a result, employee-related CSR improves employees’ job satisfaction and positively affects their organizational commitment, which reduces their turnover rate and contributes to the formation of a stable innovation team [38].

A stable innovation team helps to improve innovation efficiency. First, the main challenges to firm innovation are the unpredictability of results and the high probability of failure [26]. Thus, tolerance for failure is a key driver of innovation [39]. A stable innovation team helps its employees to identify with the team [40], overcome difficulties and failures in the innovation process, and find positive significance in their failures [41]. In addition, a stable innovation team is conducive to the transfer and accumulation of knowledge. Identification with the organization encourages employees to show positive work attitudes and behavior to maintain consistency within the organization [42,43], which helps them to generate potentially valuable novel ideas from a long-term perspective [44]. Employees in stable innovation teams are also more willing to share their knowledge with team members [45,46], which enhances the learning and innovation abilities of the team as a whole. Thus, the first hypothesis to test is as follows:

Hypothesis 1 (H1).

Employee-related CSR is positively associated with firm innovation.

2.2. SCC-Related CSR and Innovation

SCC-related CSR enables firms to establish good relationships with their partners, who provide necessary external resources for firm innovation [47]. First, suppliers and customers need innovation in the value chain [48], but it is difficult for them to understand firms’ innovation ability as external stakeholders. Supplier- and customer-related CSR enables firms’ partners to observe and participate in the firms’ internal operations, which helps them to understand the potential risks and uncertainties of firm innovation [49], judge the firms’ innovation ability, and improve their willingness to provide resource support [50]. Moreover, consumer-related CSR stimulates consumers’ positive attitudes, which helps firm innovation [51]. Some studies show that consumer-related CSR can increase consumers’ satisfaction and generate positive responses [52], leading to improvements in the firms’ brand reputation and higher product premiums [53], which provide necessary financial support for firm innovation [54]. In addition, consumer-related CSR cultivates loyal consumers with a high tolerance for products, which provides a buffer for firm innovation and supports firms’ continuous product development.

SCC-related CSR produces a knowledge spillover effect and promotes firm innovation. First, supplier- and customer-related CSR helps to establish close relationships, which stimulates knowledge exchange [55], mutual learning, creative recombination of knowledge, and new perspectives for solving problems. Consumer-related CSR also encourages consumers to become a key source of innovative ideas. Loyal consumers bring more product feedback to firms [56] and actively participate in the companies’ new product development and improvement [57,58]. Accordingly, Chu et al. [47] also find that the firms’ degree of dependence on stakeholders positively relates to firm innovation because such dependence is conducive to the transmission of specific information and knowledge. Thus, the second hypothesis to test is as follows:

Hypothesis 2 (H2).

SCC-related CSR is positively associated with firm innovation.

2.3. Environment-Related CSR and Innovation

Firms produce innovation patents and obtain innovation knowledge in the process of creating an environmental CSR to protect the natural environment. First, firms must improve their production techniques and develop energy-saving materials to meet environmental protection standards [59], which requires technological innovation [60]. Jaffe and Palmer [61] show that DuPont produced many innovative patents in the process of coping with global warming, which promoted its innovation ability. Moreover, firms establish good relationships with environmental protection organizations by using environment-related CSR to protect the natural environment [62], which brings firms external environment knowledge. The combination of internal and external knowledge promotes firm innovation [63].

Environment-related CSR enables firms to reduce their environmental fines and obtain government subsidies, which provide additional financial support for firm innovation [64]. First, firms face pressure from the government and laws [9,65] against environmental degradation, with stricter legal penalties if they destroy the ecological environment. In this regulatory environment, environment-related CSR recognizes a stronger need to avoid high fines, which may consume firms’ resources for innovation. Second, environment-related CSR enables firms to receive government subsidies, which reduce firms’ marginal costs under the problems of performance decline and innovation uncertainty [66]. In April 2014, China expanded its Environmental Protection Law from 47 articles to 70 articles to promote green development. Subsequently, firms received more government subsidies for environment-related CSR. Government-provided subsidies also require firms to obtain returns, such as patents and technology upgrades, which encourage firm innovation [67]. Thus, the third hypothesis to test is as follows:

Hypothesis 3 (H3).

Environment-related CSR is positively associated with firm innovation.

2.4. Social Welfare-Related CSR and Innovation

Social welfare-related CSR includes donations, taxes, and other social responsibilities, where the stakeholder is the public. The difference between social welfare-related CSR and employee-, environment-, or SCC-related (e.g., government-related) CSR is the indirect and uncertain resources that social welfare brings. Some studies show that social welfare-related CSR enables firms to obtain government support [68], strengthen their cooperative relationships with suppliers [69], and attract job seekers with the same values [70]. If firms undertake social welfare-related CSR, can they also fulfill their employee-, SCC-, and government-related CSR? The answers remain unverified. In other words, employee-, SCC-, and government-related CSR may be able to bring more resources for innovation to firms.

It is uncertain whether social welfare-related CSR can bring more resources for innovation to firms, but the crowding-out effect on firm innovation is certain. Firms’ tax burden reduces their available surplus, which decreases their innovation investment [71] and inhibits firm innovation [72,73]. Especially in China and other emerging economies, firms face greater resource constraints [74]; thus, the crowding-out effect of social welfare-related CSR on firm innovation will be more significant there. Hence, the fourth hypothesis to test is as follows:

Hypothesis 4 (H4).

Social welfare-related CSR is negatively associated with firm innovation.

2.5. Total CSR and Innovation

Based on the above analysis, we believe that the closer the relationship between CSR and the firm, the greater the positive effect of CSR on firm innovation. Although social welfare-related CSR may be negatively associated with firm innovation, generally, CSR should play a positive role in firm innovation. Thus, the fifth hypothesis to test is as follows:

Hypothesis 5 (H5).

Total CSR is positively associated with firm innovation.

3. Methodology

3.1. Sample and Data

We take A-share listed firms in China from the Shanghai and Shenzhen stock exchanges from 2010 to 2017 as the research sample. The CSR data are from the social responsibility reports of listed companies in the HeXun Net CSR rating index, and the innovation variables are from the China Stock Market and Accounting Research Database (CSMAR). Other corporate financial and governance data are from the annual reports of the listed companies and the CSMAR database.

The reason for choosing 2010 as the research starting point is that HeXun Net began disclosing CSR scores in 2010. Moreover, there are two reasons for choosing Chinese firm data as the sample. One reason is because China can be representative of developing countries. Firms in developing countries are unique in fulfilling CSR because they prefer to use CSR as a tool to achieve their economic goals and are concerned about the relationship between the hypothetical “tool” and the results (Jones et al., 2018). Developing countries also better match the resource-based theory because they have fewer resources and pay greater attention to the resources that fulfilling CSR can bring. The second reason is the availability of indicators. HeXun Net’s CSR classification scores can identify the types of CSR, including the advantages and disadvantages of various CSR behaviors.

In addition, we exclude financial and special treatment firms because those firms are likely to be trapped in financial distress with abnormal operations. To eliminate the influence of extreme values, we winsorize all continuous variables at the 1% and 99% quantiles; finally, 17,647 valid samples are obtained.

3.2. Variables

3.2.1. Dependent Variable

In line with previous research [75,76,77], this study measures innovation using the following items: (i) the number of applied patents (APP); (ii) the number of applied patents for inventions (IAPP); (iii) the number of patents granted in the current year and the following 3 years (GRT); and (iv) the number of patents for inventions granted in the current year and the following 3 years (IGRT). Patents include inventions, utility models, and designs. Of these, invention patents are the most innovative. Thus, we use the total number of patents and the number of invention patents separately to measure innovation. Since firms have many abnormal patent applications, we measure innovation by both the number of patents applied for and the number of patents granted. Considering there are many observed values of 0 in the number of patents, the natural logarithm is taken after adding 1 to the number of patents with reference to the usual practice in the literature.

3.2.2. Independent Variable

This study researches the relationship between CSR and innovation based on both stakeholder theory and resource-based theory. The resource-based theory considers that firm development depends on the quantity and quality of resources. In order to maintain a sustainable competitive advantage, the firm must have valuable, rare, irreplaceable, and irreplaceable assets [78]. According to our hypothesis, CSR towards different stakeholders influences firm resources in different ways [79,80] and thus tends to affect innovation negatively or positively. Therefore, we classify CSR based on the categories of stakeholders (investors, employees, customers and suppliers, environment, and the public) defined by Papa-solomou et al. [81] to explore whether different types of CSR can influence innovation differently through the different resources they bring.

The HeXun Net CSR rating index is used to measure the extent of CSR activities. The HeXun Net data are taken from the social responsibility and annual reports published by listed firms in the Shanghai and Shenzhen stock exchanges. The data are evaluated using five responsibility dimensions: i.e., shareholder, employee, SCC, environmental, and social welfare responsibilities. The weight proportions of the five dimensions are adjusted according to the industry and then summed to obtain the CSR score. The higher the total score, the better the CSR performance. We intend to measure CSR for key stakeholders; thus, we use the scores of the above five dimensions to measure responsibilities toward employees (REMP), SCC (RSCC), the environment (RENV), and social welfare (RSOC), and the total CSR (TCSR).

3.2.3. Control Variables

Following the literature, we introduce several controls that might influence innovation. First, firm size (SIZE) affects innovation [82,83]; we thus control for firm size, measured as the log of the total assets of the firm. Then, we control for several financial variables [23]. Asset liability ratio (LEV) is measured as the percentage of total debts to total assets. Return on assets (ROA) is defined as net profits divided by total assets. Cash flow (OCF) is measured as cash flow from operating activities divided by total assets. Company growth (TOBINQ) is defined as the total market value divided by total assets. In addition, we include several corporate governance variables [13]. The firms’ capital structure (SOE) is measured as a dummy variable, which equals 1 when firms are state-owned firms and 0 otherwise. Ownership concentration (OWN) is calculated using the shareholding ratio of the largest shareholder. The separation between ownership and control (SPE) is defined as the difference between control and ownership. The integration of chairperson and chief executive officer (CEO) (DUAL) is measured as a dummy variable, which equals 1 when the chairperson of the firm takes the position of CEO and 0 otherwise. Board size (BOARD) is defined as the natural logarithm of the number of board directors. Independent director (INDRCT) is measured as the percentage of independent board directors.

Table 1. Variables and definitions.

Variables	Definition
dependent variable
lnAPP	The natural logarithm of one plus the number of applied patents
lnIAPP	The natural logarithm of one plus the number of applied patents for inventions.
lnGRT	The natural logarithm of one plus the number of patents, which are granted in the current year and next three years.
lnIGRT	The natural logarithm of one plus the number of patents for inventions, which are granted in the current year and next three years.
Independent variable
REMP	The scores of the corporate responsibility toward employees
RSCC	The scores of the corporate responsibility toward suppliers, customers, and consumers.
RENV	The scores of the corporate responsibility toward the environment.
RSOC	The scores of the corporate responsibility toward the social welfare.
TCSR	Total scores of CSR.
Control variable
SIZE	The natural logarithm of total assets.
LEV	The percentage of total debts to total assets.
OCF	Cash flow from operating activities divided by total assets
TBQ	Total market value divided by total asset
SOE	An indicator, which equals one when firms are state-owned enterprises (SOEs), and 0 otherwise
OWN	Shareholding ratio of the largest shareholder
SEP	Difference between control and ownership
DUAL	An indicator, which equals one when the Chairman of the firm takes the position of CEO, and 0 otherwise.
BOARD	The natural logarithm of the number of directors of the board.
INDRCT	The percentage of inside directors of the board.

reports the index selection and measurement methods for the core variables.

3.3. Empirical Models

To test the hypotheses presented earlier and using a list of variables (Table 1), we test the following OLS econometric model:

$$\ln (1+{\mathrm{patent}}_{\mathrm{it}}{)=\mathsf{\beta }}_0{+\mathsf{\beta }}_1{\mathrm{CSR}\_\mathrm{item}}_{\mathrm{it}}+{\displaystyle \sum _{\mathrm{j}}{\mathsf{\beta }}_{\mathrm{j}}{\mathrm{Control}}_{\mathrm{j},\mathrm{it}}}{+\mathrm{IndustryFE}+\mathrm{YearFE}+\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

where the subscripts i and t represent the firm and period, respectively. The dependent variable ln(1 + patent_it) is a series of indicators to measure enterprise innovation activities, which are characterized by identifying the total number of patents and the number of invention patents from the applications and grants dimensions. The independent variable CSR_item_it is CSR, which is described using five dimensions: employee-, SCC-, environment-, and social welfare-related CSR and the total CSR. Control_j,it is a control variable matrix at the firm level to describe the impact of firm heterogeneity on innovation. This paper also controls for the industry fixed effect (IndustryFE) and time fixed effect (YearFE).

4. Empirical Results

4.1. Descriptive Statistics

The descriptive statistics are reported in

Table 2. Descriptive statistics of main variables.

Variables	N	Mean	SD	Min.	Median	Max.
APP	17,647	30.090	213.385	0	2	422
IAPP	17,647	14.784	140.713	0	1	175
GRT	17,647	17.346	108.972	0	1	278
IGRT	17,647	3.562	33.244	0	0	52
REMP	17,647	3.126	3.499	0	1.810	15
RSCC	17,647	2.272	5.274	0	0	20
RENV	17,647	2.366	5.733	0	0	23
RSOC	17,647	4.885	4.557	−7.340	4.390	16.470
TCSR	17,647	26.406	17.929	−3.210	22.180	76.060
SIZE	17,647	21.964	1.257	19.661	21.786	25.863
LEV	17,647	0.425	0.214	0.045	0.419	0.876
ROA	17,647	0.043	0.046	−0.106	0.038	0.188
OCF	17,647	0.042	0.074	−0.190	0.042	0.246
TOBINQ	17,647	2.216	1.423	0.937	1.752	9.207
SOE	17,647	0.412	0.492	0	0.000	1
OWN	17,647	0.361	0.151	0.093	0.342	0.752
SEP	17,647	0.050	0.076	0	0	0.284
DUAL	17,647	0.272	0.445	0	0	1
BOARD	17,647	2.151	0.200	1.609	2.197	2.708

. The standard deviations of APP, IAPP, GRT, and IGRT are 213.385, 140.713, 108.972, and 33.244, respectively, indicating considerable differences in innovation performance among the sampled firms. According to the mean of the dependent variables, the average number of APP is 30, half of which are invention patent applications (IAPP). We also note that the average number of patents granted (GRT) is 17, of which only 4 are invention patents (IGRT). The result shows that the grant rate for invention patents is lower than that for other patents. The medians of the dependent variables (APP, IAPP, GRT, IGRT) are 2, 1, 1, and 0, respectively, indicating a right-skewness distribution in the patent application and grant data.

The standard deviations of REMP, RSCC, RENV, RSOC, and TCSR are 3.499, 5.274, 5.733, 4.557, and 17.929, respectively, showing that the differences in CSR and its four components are also large among the sampled firms. Furthermore, comparing the max, mean, and median, employee- and social welfare-related CSR have better results than SCC- and environment-related CSR. At least half of the firms fail to fulfill their SCC- and environment-related CSR (median is 0). The lowest score for RSOC and TCSR is <0 because some firms are punished for their CSR misconduct, such as serious production accidents and administrative sanctions. The statistics for the other control variables are within a reasonable range, which implies that outliers are not an issue.

4.2. Basic Regression

To examine our hypotheses, we test the relationship between CSR and firm innovation. First, we test the impacts of four CSR components on firm innovation performance to examine H1–H4. Second, we test the influence of the total CSR on firm innovation performance to examine H5. The results are as follows.

Table 3. Impact of employees related CSR on firm innovation.

Variable	lnAPP	lnIAPP	lnGRT	lnIGRT
REMP	0.027 ***	0.028 ***	0.023 ***	0.019 ***
	(8.43)	(10.00)	(7.85)	(9.56)
SIZE	0.271 ***	0.263 ***	0.241 ***	0.166 ***
	(21.64)	(24.68)	(21.26)	(22.15)
LEV	−0.455 ***	−0.306 ***	−0.414 ***	−0.265 ***
	(−6.83)	(−5.41)	(−6.88)	(−6.65)
ROA	3.137 ***	2.510 ***	2.381 ***	1.015 ***
	(11.55)	(10.86)	(9.68)	(6.24)
OCF	0.822 ***	0.599 ***	0.711 ***	0.279 ***
	(5.39)	(4.62)	(5.15)	(3.05)
TOBINQ	−0.033 ***	0.004	−0.028 ***	0.015 ***
	(−3.72)	(0.49)	(−3.47)	(2.86)
SOE	−0.102 ***	−0.016	−0.086 ***	−0.003
	(−4.02)	(−0.73)	(−3.72)	(−0.18)
OWN	0.069	−0.102 *	0.119 *	−0.025
	(0.95)	(−1.65)	(1.81)	(−0.59)
SEP	0.063	0.097	0.064	0.004
	(0.46)	(0.83)	(0.51)	(0.05)
DUAL	0.059 **	0.073 ***	0.055 **	0.059 ***
	(2.46)	(3.58)	(2.57)	(4.12)
BOARD	0.325 ***	0.290 ***	0.245 ***	0.204 ***
	(5.96)	(6.24)	(4.96)	(6.24)
Constant	−5.947 ***	−5.997 ***	−5.149 ***	−3.725 ***
	(−21.23)	(−25.16)	(−20.31)	(−22.22)
Ind and Year	Yes	Yes	Yes	Yes
Observations	17,647	17,647	17,647	17,647
Adjusted R2	0.347	0.287	0.349	0.262
F	147.8	111.8	149.0	99.00

Note: t statistics are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the results of the test for H1 from the regression analysis based on model (1). The dependent variable of innovation is measured using four methods (lnAPP, lnIAPP, lnGRT, and lnIGRT). If REMP has a positive regression coefficient, H1 is supported, i.e., firms with superior employee-related CSR activities will increase their innovation. In the sample, the regression coefficients for REMP are 0.027, 0.028, 0.023, and 0.019, respectively, which are significantly positive values (p < 0.01), as expected. This result suggests that the more firms perform employee-related CSR, the greater their innovation performance. Thus, as employees are key internal stakeholders, employee-related CSR can explain changes in innovation performance.

Table 4. Impact of SCC related CSR on firm innovation.

Variable	lnAPP	lnIAPP	lnGRT	lnIGRT
RSCC	0.019 ***	0.014 ***	0.017 ***	0.009 ***
	(9.07)	(7.91)	(8.95)	(7.15)
SIZE	0.272 ***	0.272 ***	0.240 ***	0.173 ***
	(21.87)	(25.65)	(21.36)	(23.17)
LEV	−0.456 ***	−0.314 ***	−0.414 ***	−0.271 ***
	(−6.86)	(−5.54)	(−6.88)	(−6.80)
ROA	3.149 ***	2.560 ***	2.385 ***	1.052 ***
	(11.61)	(11.07)	(9.71)	(6.46)
OCF	0.811 ***	0.594 ***	0.700 ***	0.276 ***
	(5.32)	(4.58)	(5.07)	(3.02)
TOBINQ	−0.033 ***	0.006	−0.028 ***	0.017 ***
	(−3.67)	(0.86)	(−3.46)	(3.25)
SOE	−0.094 ***	−0.004	−0.079 ***	0.005
	(−3.69)	(−0.20)	(−3.44)	(0.35)
OWN	0.077	−0.098	0.126 *	−0.024
	(1.06)	(−1.59)	(1.92)	(−0.54)
SEP	0.056	0.096	0.056	0.004
	(0.41)	(0.82)	(0.45)	(0.05)
DUAL	0.058 **	0.072 ***	0.055 **	0.058 ***
	(2.44)	(3.54)	(2.55)	(4.08)
BOARD	0.322 ***	0.289 ***	0.242 ***	0.203 ***
	(5.89)	(6.20)	(4.89)	(6.20)
Constant	−5.949 ***	−6.155 ***	−5.121 ***	−3.846 ***
	(−21.34)	(−25.91)	(−20.30)	(−23.02)
Ind and Year	Yes	Yes	Yes	Yes
Observations	17,647	17,647	17,647	17,647
Adjusted R2	0.348	0.285	0.350	0.261
F	148.0	111.0	149.4	98.15

Note: t statistics are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the results of the test for H2. The independent variable is CSR toward SCC (RSCC). We find that the four coefficients of RSCC (0.019, 0.014, 0.017, and 0.009) are significantly positive (p < 0.01). This result suggests that firms with superior SCC-related CSR activities increase their innovation, proving H2. Thus, as key external stakeholders, SCC-related CSR can explain changes in innovation performance.

Table 5. Impact of environment-related CSR on firm innovation.

Variable	lnAPP	lnIAPP	lnGRT	lnIGRT
RENV	0.020 ***	0.016 ***	0.019 ***	0.012 ***
	(10.19)	(9.62)	(10.81)	(10.34)
SIZE	0.268 ***	0.267 ***	0.234 ***	0.166 ***
	(21.51)	(25.14)	(20.82)	(22.30)
LEV	−0.449 ***	−0.307 ***	−0.406 ***	−0.264 ***
	(−6.76)	(−5.42)	(−6.75)	(−6.62)
ROA	3.187 ***	2.581 ***	2.411 ***	1.056 ***
	(11.76)	(11.18)	(9.84)	(6.51)
OCF	0.813 ***	0.595 ***	0.701 ***	0.274 ***
	(5.34)	(4.58)	(5.08)	(3.01)
TOBINQ	−0.033 ***	0.005	−0.029 ***	0.016 ***
	(−3.75)	(0.72)	(−3.62)	(2.97)
SOE	−0.099 ***	−0.009	−0.085 ***	0.000
	(−3.91)	(−0.43)	(−3.70)	(0.03)
OWN	0.072	−0.101	0.123 *	−0.024
	(1.00)	(−1.64)	(1.87)	(−0.56)
SEP	0.063	0.100	0.062	0.005
	(0.46)	(0.86)	(0.50)	(0.06)
DUAL	0.059 **	0.073 ***	0.056 **	0.059 ***
	(2.46)	(3.57)	(2.57)	(4.11)
BOARD	0.324 ***	0.290 ***	0.243 ***	0.203 ***
	(5.93)	(6.23)	(4.93)	(6.22)
Constant	−5.853 ***	−6.042 ***	−4.991 ***	−3.710 ***
	(−20.97)	(−25.41)	(−19.77)	(−22.19)
Ind and Year	Yes	Yes	Yes	Yes
Observations	17,647	17,647	17,647	17,647
Adjusted R2	0.349	0.286	0.351	0.263
F	148.5	111.6	150.3	99.33

Note: t statistics are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the results of the test for H3. The independent variable is environment-related CSR (RENV). We find that the four coefficients of RENV (0.020, 0.016, 0.019, and 0.012) are significantly positive (p < 0.01). This result suggests that firms with superior environment-related CSR increase innovation, proving H3. Thus, similar to employee- and SCC-related CSR, performing environmental-related CSR can also bring direct resources for firm innovation against the background of the government’s rewards for environmental protection and punishments for environmental damage.

Table 6. Impact of social welfare-related CSR on firm innovation.

Variable	lnAPP	lnIAPP	lnGRT	lnIGRT
RSOC	−0.008 ***	−0.010 ***	−0.008 ***	−0.008 ***
	(−3.11)	(−4.43)	(−3.47)	(−4.94)
SIZE	0.305 ***	0.298 ***	0.270 ***	0.190 ***
	(25.22)	(28.94)	(24.70)	(26.26)
LEV	−0.476 ***	−0.327 ***	−0.431 ***	−0.278 ***
	(−7.15)	(−5.77)	(−7.16)	(−6.98)
ROA	3.364 ***	2.754 ***	2.585 ***	1.189 ***
	(12.34)	(11.87)	(10.48)	(7.29)
OCF	0.848 ***	0.627 ***	0.734 ***	0.299 ***
	(5.55)	(4.82)	(5.31)	(3.27)
TOBINQ	−0.025 ***	0.012	−0.021 ***	0.021 ***
	(−2.81)	(1.59)	(−2.62)	(3.91)
SOE	−0.082 ***	0.005	−0.068 ***	0.011
	(−3.22)	(0.23)	(−2.98)	(0.74)
OWN	0.063	−0.107 *	0.114 *	−0.029
	(0.87)	(−1.73)	(1.74)	(−0.66)
SEP	0.086	0.121	0.084	0.021
	(0.63)	(1.03)	(0.67)	(0.26)
DUAL	0.057 **	0.071 ***	0.054 **	0.057 ***
	(2.39)	(3.48)	(2.49)	(4.01)
BOARD	0.325 ***	0.289 ***	0.244 ***	0.203 ***
	(5.94)	(6.20)	(4.93)	(6.18)
Constant	−6.578 ***	−6.644 ***	−5.688 ***	−4.166 ***
	(−24.17)	(−28.67)	(−23.10)	(−25.56)
Ind and Year	Yes	Yes	Yes	Yes
Observations	17,647	17,647	17,647	17,647
Adjusted R2	0.345	0.283	0.347	0.259
F	146.3	110.0	147.8	97.58

Note: t statistics are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the results of the test for H4. The independent variable is social welfare-related CSR (RSOC). If RSOC has a negative regression coefficient, H4 is supported, i.e., firms with superior social welfare-related CSR activities will decrease innovation. In Table 6, the regression coefficients of RSOC are −0.008, −0.010, −0.008, and −0.008, respectively, which are significantly negative values (p < 0.01), as expected. This result suggests that the more firms perform social welfare-related CSR, the lower their innovation performance. This may be because social welfare cannot bring direct resources to firms.

Table 7. Impact of total CSR on firm innovation.

Variable	lnAPP	lnIAPP	lnGRT	lnIGRT
TCSR	0.007 ***	0.006 ***	0.006 ***	0.004 ***
	(10.51)	(10.03)	(9.85)	(9.09)
SIZE	0.261 ***	0.260 ***	0.232 ***	0.165 ***
	(20.65)	(24.23)	(20.31)	(21.88)
LEV	−0.422 ***	−0.285 ***	−0.386 ***	−0.252 ***
	(−6.33)	(−5.02)	(−6.40)	(−6.32)
ROA	2.623 ***	2.122 ***	1.944 ***	0.772 ***
	(9.43)	(8.96)	(7.72)	(4.63)
OCF	0.753 ***	0.546 ***	0.653 ***	0.245 ***
	(4.94)	(4.20)	(4.73)	(2.68)
TOBINQ	−0.030 ***	0.008	−0.025 ***	0.018 ***
	(−3.39)	(1.06)	(−3.16)	(3.44)
SOE	−0.095 ***	−0.006	−0.080 ***	0.004
	(−3.75)	(−0.28)	(−3.47)	(0.27)
OWN	0.052	−0.118 *	0.104	−0.036
	(0.72)	(−1.90)	(1.59)	(−0.82)
SEP	0.055	0.093	0.056	0.002
	(0.40)	(0.80)	(0.45)	(0.03)
DUAL	0.059 **	0.072 ***	0.055 **	0.059 ***
	(2.45)	(3.56)	(2.56)	(4.10)
BOARD	0.322 ***	0.288 ***	0.242 ***	0.203 ***
	(5.90)	(6.19)	(4.90)	(6.19)
Constant	−5.785 ***	−5.980 ***	−5.007 ***	−3.734 ***
	(−20.65)	(−25.05)	(−19.74)	(−22.23)
Ind and Year	Yes	Yes	Yes	Yes
Observations	17,647	17,647	17,647	17,647
Adjusted R2	0.349	0.287	0.351	0.262
F	148.7	111.8	149.8	98.82

Note: t statistics are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

shows the results of the test for H5. The independent variable is total CSR (TCSR). We find that the four coefficients of TCSR (0.007, 0.006, 0.006, and 0.004) are significantly positive (p < 0.01). This result suggests that firms with superior total CSR activities increase their innovation, proving H5. In general, although the implementation of CSR for different stakeholders has different impacts on firm innovation, the implementation of CSR generally promotes firm innovation.

4.3. Robustness Checks

We also perform numerous robustness tests. For simplicity, we only report results with the variable lnIGRT as a measure of firm innovation. Invention patents represent the firms’ most original innovations [84]. Patent grants, rather than patent applications, capture actual innovation performance because patent applications are easily manipulated by managers with no administrative approval [85]. Therefore, we choose the natural logarithm of 1 plus the number of invention patents granted (lnIGRT) as a proxy for firm innovation in our robustness tests. In addition, the results of the robustness tests with other proxies for firm innovation remain consistent with those reported in the following subsections.

4.3.1. Firm Fixed Effects Specification

Some missing variables could potentially cause spurious correlations between CSR and firm innovation. To alleviate these concerns, we include firm fixed effects to capture time-invariant factors that may lead to the spurious correlation. The results are reported in

Table 8. Firm fixed-effect specification.

Variable	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT
REMP	0.006 ***
	(2.73)
RSCC		0.003 ***
		(2.61)
RENV			0.005 ***
			(4.20)
RSOC				0.000
				(0.16)
TCSR					0.001 ***
					(3.65)
SIZE	0.040 **	0.041 **	0.038 **	0.045 ***	0.038 **
	(2.50)	(2.56)	(2.39)	(2.81)	(2.38)
LEV	−0.322 ***	−0.323 ***	−0.321 ***	−0.323 ***	−0.317 ***
	(−5.65)	(−5.66)	(−5.62)	(−5.67)	(−5.55)
ROA	0.485 ***	0.489 ***	0.482 ***	0.497 ***	0.435 ***
	(2.95)	(2.97)	(2.93)	(3.02)	(2.63)
OCF	−0.100	−0.100	−0.100	−0.100	−0.108
	(−1.22)	(−1.21)	(−1.21)	(−1.21)	(−1.31)
TOBINQ	0.011 *	0.011 *	0.010 *	0.012 **	0.011 *
	(1.88)	(1.91)	(1.77)	(2.14)	(1.85)
SOE	−0.033	−0.034	−0.033	−0.034	−0.032
	(−0.67)	(−0.69)	(−0.68)	(−0.70)	(−0.64)
OWN	−0.050	−0.053	−0.049	−0.055	−0.060
	(−0.52)	(−0.55)	(−0.50)	(−0.57)	(−0.62)
SEP	−0.046	−0.049	−0.043	−0.053	−0.045
	(−0.30)	(−0.32)	(−0.28)	(−0.35)	(−0.30)
DUAL	0.054 ***	0.054 ***	0.054 ***	0.053 ***	0.054 ***
	(2.94)	(2.95)	(2.96)	(2.91)	(2.94)
BOARD	0.115 **	0.116 **	0.114 **	0.117 **	0.114 **
	(2.24)	(2.25)	(2.21)	(2.27)	(2.23)
Constant	−0.468	−0.477	−0.422	−0.556	−0.442
	(−1.32)	(−1.35)	(−1.19)	(−1.58)	(−1.25)
Firm and Year	Yes	Yes	Yes	Yes	Yes
Observations	17,647	17,647	17,647	17,647	17,647
Adjusted R2	0.599	0.599	0.599	0.599	0.599
F	216.2	216.2	217.0	215.7	216.7

Note: t statistics are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

. The coefficients of REMP, RSCC, RENV, and TCSR (0.006, 0.003, 0.005, and 0.001, respectively) are all significantly positive (p < 0.01), supporting H1–H3 and H5. The coefficient of RSOC is not significant, which indicates that social welfare-related CSR differs from other CSR components because it does not improve firm innovation performance. This is also consistent with H4 and the results of the OLS model.

4.3.2. Instrumental Variable Method

We use the instrumental variable method to further mitigate endogeneity concerns in this study. We select the degree of religiousness as our instrumental variable, which is measured by the natural logarithm of 1 plus the number of Buddhist monasteries in the province where a firm is registered. The religions in a district affect corporate behaviors, including CSR activities [86]. In China, Buddhism is the most prevalent religion to positively influence firms’ CSR decisions [87]. However, the degree of religiousness has no direct effect on firms’ innovation performance. Therefore, it is suitable to use religiousness as an instrumental variable in our research setting.

In the first-stage regression, CSR is the dependent variable and TEMP is the independent variable. In this stage, we predict the CSR value. In the second-stage regression, we regress firm innovation on the predicted CSR from the first-stage regression and the other control variables used in the model (1). With this approach, the predicted value from the first-stage regression is no longer correlated with the error term of the second-stage regression and the estimated coefficient is consistent.

Table 9. Instrumental variable method.

Variable	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT
REMP	0.308 ***
	(2.98)
RSCC		0.364 *
		(1.90)
RENV			0.240 **
			(2.53)
RSOC				−4.960
				(−0.18)
TCSR					0.068 ***
					(2.85)
SIZE	−0.150	−0.371	−0.216	2.424	−0.209
	(−1.32)	(−1.26)	(−1.35)	(0.20)	(−1.49)
LEV	−0.001	0.180	0.090	2.353	0.278
	(−0.01)	(0.68)	(0.55)	(0.16)	(1.33)
ROA	−0.612	−1.545	−0.114	47.608	−5.359 **
	(−0.98)	(−1.07)	(−0.20)	(0.19)	(−2.34)
OCF	0.096	−0.252	−0.033	6.814	−0.550 *
	(0.64)	(−0.71)	(−0.16)	(0.19)	(−1.69)
TOBINQ	−0.076 **	−0.133	−0.085 **	−0.290	−0.032
	(−2.26)	(−1.62)	(−1.97)	(−0.17)	(−1.57)
SOE	−0.206 ***	−0.196 *	−0.187 **	0.541	−0.107 **
	(−2.70)	(−1.71)	(−2.26)	(0.19)	(−2.23)
OWN	0.094	0.320	0.146	1.717	−0.086
	(1.24)	(1.56)	(1.44)	(0.18)	(−1.19)
SEP	−0.177	−0.458	−0.198	4.638	−0.246
	(−1.29)	(−1.48)	(−1.19)	(0.18)	(−1.58)
DUAL	0.068 ***	0.066 *	0.068 ***	−0.376	0.064 ***
	(3.12)	(1.94)	(2.65)	(−0.16)	(2.84)
BOARD	0.160 ***	0.059	0.131 **	−2.464	0.133 **
	(3.09)	(0.53)	(2.00)	(−0.17)	(2.30)
Constant	2.418	6.971	4.026	−31.934	3.097
	(1.09)	(1.19)	(1.24)	(−0.21)	(1.21)
Ind and Year	Yes	Yes	Yes	Yes	Yes
Observations	17,533	17,533	17,533	17,533	17,533
Adjusted R2	−0.670	−3.091	−1.323	−438.011	−0.832
F	43.04	17.57	30.93	0.164	39.22

Note: Z statistics are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

reports the results from the second-stage regression. Due to space limitations, we do not report the results from the first-stage regression. After we use the instrumental variable to mitigate the endogeneity concerns, the coefficients of REMP, RSCC, RENV, and TCSR (0.308, 0.364, 0.240, and 0.068, respectively) are still significantly positive (p < 0.01), which is consistent with H1–H3 and H5. The RSOC coefficient is not significant, but it is consistent with our argument that different CSR components influence firm innovation differently. Thus, the results in Table 9 support our view that some CSR activities causally improve firm innovation performance, but others do not.

4.3.3. Alternative Methodology

To provide robust evidence, we use the negative binomial model to re-estimate model (1) because of the count-based nature of the innovation variable (patents granted). Panel A in

Table 10. Alternative methodology.

Panel A Negative Binomial Regression
Variable	IGRT	IGRT	IGRT	IGRT	IGRT
REMP	0.029 ***
	(5.32)
RSCC		0.012 ***
		(3.33)
RENV			0.013 ***
			(4.16)
RSOC				−0.032 ***
				(−5.72)
TCSR					0.006 ***
					(5.06)
SIZE	0.730 ***	0.743 ***	0.739 ***	0.771 ***	0.730 ***
	(30.09)	(30.78)	(30.67)	(32.71)	(30.00)
LEV	−0.913 ***	−0.923 ***	−0.919 ***	−0.879 ***	−0.905 ***
	(−6.90)	(−6.97)	(−6.94)	(−6.62)	(−6.84)
ROA	3.583 ***	3.718 ***	3.722 ***	4.158 ***	3.183 ***
	(6.17)	(6.41)	(6.42)	(7.16)	(5.36)
OCF	0.290	0.308	0.304	0.432	0.232
	(0.90)	(0.95)	(0.94)	(1.34)	(0.72)
TOBINQ	0.006	0.011	0.011	0.016	0.010
	(0.29)	(0.55)	(0.54)	(0.82)	(0.53)
SOE	−0.061	−0.045	−0.053	−0.033	−0.051
	(−1.24)	(−0.93)	(−1.09)	(−0.68)	(−1.05)
OWN	0.152	0.123	0.125	0.105	0.117
	(1.09)	(0.88)	(0.89)	(0.75)	(0.84)
SEP	−0.114	−0.072	−0.073	−0.014	−0.087
	(−0.45)	(−0.29)	(−0.29)	(−0.05)	(−0.34)
DUAL	0.142 ***	0.143 ***	0.142 ***	0.135 ***	0.143 ***
	(3.16)	(3.19)	(3.17)	(3.01)	(3.19)
BOARD	0.746 ***	0.741 ***	0.749 ***	0.751 ***	0.741 ***
	(7.25)	(7.20)	(7.29)	(7.30)	(7.21)
Intercept	−17.671 ***	−17.919 ***	−17.836 ***	−18.420 ***	−17.695 ***
	(−33.40)	(−33.89)	(−33.80)	(−35.54)	(−33.45)
Ind and Year	Yes	Yes	Yes	Yes	Yes
Observations	17,647	17,647	17,647	17,647	17,647
Pseudo R2	0.170	0.170	0.170	0.170	0.170
Panel B Tobit Regression
Variable	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT
REMP	0.033 ***
	(6.16)
RSCC		0.018 ***
		(5.21)
RENV			0.018 ***
			(5.81)
RSOC				−0.022 ***
				(−4.36)
TCSR					0.007 ***
					(6.78)
SIZE	0.421 ***	0.431 ***	0.426 ***	0.472 ***	0.414 ***
	(18.45)	(18.96)	(18.74)	(21.43)	(18.02)
LEV	−0.880 ***	−0.888 ***	−0.881 ***	−0.900 ***	−0.846 ***
	(−7.17)	(−7.24)	(−7.18)	(−7.33)	(−6.88)
ROA	3.566 ***	3.618 ***	3.661 ***	3.965 ***	3.029 ***
	(7.08)	(7.18)	(7.27)	(7.85)	(5.89)
OCF	0.714 **	0.722 **	0.724 **	0.799 ***	0.648 **
	(2.44)	(2.46)	(2.47)	(2.73)	(2.21)
TOBINQ	−0.069 ***	−0.067 ***	−0.066 ***	−0.059 ***	−0.065 ***
	(−3.71)	(−3.60)	(−3.60)	(−3.18)	(−3.51)
SOE	−0.054	−0.040	−0.046	−0.023	−0.044
	(−1.17)	(−0.87)	(−1.00)	(−0.49)	(−0.94)
OWN	−0.049	−0.050	−0.052	−0.065	−0.064
	(−0.38)	(−0.38)	(−0.40)	(−0.50)	(−0.49)
SEP	0.029	0.025	0.042	0.095	0.020
	(0.12)	(0.10)	(0.18)	(0.40)	(0.08)
DUAL	0.108 ***	0.109 ***	0.109 ***	0.108 **	0.109 ***
	(2.58)	(2.60)	(2.60)	(2.57)	(2.59)
BOARD	0.647 ***	0.645 ***	0.648 ***	0.646 ***	0.643 ***
	(6.50)	(6.46)	(6.50)	(6.47)	(6.45)
Constant	−11.488 ***	−11.665 ***	−11.558 ***	−12.425 ***	−11.397 ***
	(−22.40)	(−22.81)	(−22.57)	(−24.96)	(−22.22)
Ind and Year	Yes	Yes	Yes	Yes	Yes
Observations	17,647	17,647	17,647	17,647	17,647
Pseudo R2	0.213	0.213	0.213	0.213	0.213

Note: Z statistics are in parentheses. ** p < 0.05, *** p < 0.01.

reports the results. The dependent variable is the number of invention patents granted (IGRT). As shown in Panel A of Table 10, we find that the coefficients of REMP, RSCC, RENV, and TCSR (0.029, 0.012, 0.013, and 0.006, respectively) are significantly positive (p < 0.01), and the coefficients of RSOC (−0.032) are significantly negative (p < 0.01). The results are consistent with our OLS model.

In our sample, the 25% quantile for all four proxies of firm innovation is 0, indicating that a large proportion of firms have no patent applications or grants. Therefore, we use the Tobit regression model to re-estimate model (1). Panel B in Table 10 reports the results. The coefficients of REMP, RSCC, RENV, and TCSR (0.033, 0.018, 0.018, and 0.007, respectively) are significantly positive (p < 0.01) and the coefficient of RSOC (−0.022) is significantly negative (p < 0.01). The results are consistent with our basic model.

4.4. Mechanism Test

Having established a positive causal link between four types of CSR activities and firm innovation, we aim to further understand the underlying mechanisms through which these CSR activities encourage firm innovation. For each type of CSR activity, we discuss the underlying mechanism.

4.4.1. Mechanism for Employee-Related CSR in Firm Innovation

The underlying mechanism through which employee-related CSR promotes firm innovation is stable teamwork. Mao and Weathers [88] observe that positive employee treatment is conducive to a corporate culture that tolerates failure, an engaging work environment, and a loyal, collaborative, and motivated research team, which enhances firm innovation [89,90]. Therefore, we argue that firms that conduct more employee-related CSR tend to have a stable creative research team, which benefits firm innovation. To test this conjecture, we use the probability of technological executives’ turnover as a proxy for stable teamwork and test the following logit econometric model (2).

$$\mathrm{TURNOVER}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\mathrm{REMP}+{\displaystyle \sum }_{\mathrm{j}}{\mathsf{\beta }}_{\mathrm{j}}{\mathrm{Control}}_{\mathrm{j},\mathrm{jt}}+\mathrm{IndustryFE}+\mathrm{YearFE}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(2)

where the dependent variable is the employee turnover rate (TURNOVER), which measures whether or not there is turnover in the technological executive. The independent variable is employee-related CSR (REMP), and the control variable is consistent with model (1). We remove observations without technological executives.

Column (1) in

Table 11. Mechanism for CSR to innovation.

	(1)	(2)	(3)	(4)	(5)	(6)
Group	Full-Sample	Low HHI	High HHI	Full-Sample	Non-SOE	SOE
Variables	TURNOVER	lnIGRT	lnIGRT	lnSUB	lnSUB	lnSUB
REMP	−0.015 **
	(−2.29)
RSCC		0.011 ***	0.006 ***
		(6.63)	(3.22)
RENV				0.007	0.023 ***	0.006
				(1.38)	(3.46)	(0.76)
SIZE	0.100 ***	0.171 ***	0.173 ***	0.894 ***	1.046 ***	0.864 ***
	(3.98)	(16.56)	(16.11)	(28.10)	(25.83)	(16.43)
LEV	0.081	−0.297 ***	−0.212 ***	0.635 ***	−0.353 *	1.502 ***
	(0.59)	(−5.45)	(−3.65)	(3.74)	(−1.81)	(4.80)
ROA	−2.537 ***	1.048 ***	1.118 ***	5.213 ***	5.956 ***	2.261 *
	(−4.54)	(4.66)	(4.75)	(7.53)	(7.77)	(1.73)
OCF	−0.458	0.421 ***	0.120	0.164	0.378	0.143
	(−1.40)	(3.39)	(0.89)	(0.42)	(0.87)	(0.20)
TOBINQ	0.030	0.016 **	0.019 ***	−0.281 ***	−0.267 ***	−0.201 ***
	(1.62)	(2.10)	(2.62)	(−12.42)	(−11.17)	(−4.12)
SOE	−0.124 **	0.000	0.012	−0.059
	(−2.35)	(0.00)	(0.51)	(−0.91)
OWN	−0.132	−0.049	−0.002	0.203	1.241 ***	−1.181 ***
	(−0.89)	(−0.82)	(−0.04)	(1.10)	(5.67)	(−3.59)
SEP	−0.117	−0.040	0.058	−0.616 *	−0.878 **	−1.770 ***
	(−0.42)	(−0.36)	(0.48)	(−1.76)	(−2.17)	(−2.77)
DUAL	−0.033	0.097 ***	0.011	0.121 **	0.111 *	0.159
	(−0.71)	(4.97)	(0.54)	(1.98)	(1.81)	(1.10)
BOARD	−0.062	0.268 ***	0.118 **	−0.029	0.372 **	−0.623 **
	(−0.56)	(5.92)	(2.51)	(−0.21)	(2.33)	(−2.52)
Intercept	−3.423 ***	−3.936 ***	−3.663 ***	−4.319 ***	−8.350 ***	−2.537 **
	(−6.07)	(−17.15)	(−14.66)	(−6.06)	(−9.23)	(−2.09)
Ind and Year	Yes	Yes	Yes	Yes	Yes	Yes
Observations	13,394	9601	8046	17,646	10,370	7276
F		57.13	46.97	94.11	69.26	41.21
Adj-R2		0.263	0.268	0.252	0.293	0.243
Pseudo R2	0.0782

Note: Column (1) examines employee loyalty as the mechanism for employee-related CSR to innovation. Column (2) and (3) examine resource and knowledge acquisition as the mechanism for customer- and supplier-related CSR to innovation. Column (4) to (6) examine government resource as the mechanism for environment-related CSR to innovation. Z statistics are in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.

reports the results. The coefficient of REMP (−0.015) is significantly negative at the 5% level, which is consistent with our conjecture. The results indicate that firms conducting more employee-related CSR have lower turnover rates in their technological executives who play an important role in firm innovation. Therefore, employee-related CSR enhances firm innovation through stable creative research teams.

4.4.2. Mechanism for SCC-Related CSR in Firm Innovation

The underlying mechanism through which SCC-related CSR encourages firm innovation is the acquisition of resources and knowledge. Studies show that a good and stable customer–supplier relationship is conducive to resource acquisition [50] and knowledge sharing [91], which benefit firm innovation [6]. Consumers and suppliers bring resources and knowledge, which play a significant role in innovation and competence for firms facing fierce competition [47,92]. We use the cross-sectional variation in the firms’ extent of competition to examine whether SCC-related CSR improves firm innovation through resource and knowledge acquisition. We expect that the positive effects of customer- and supplier-related CSR on firm innovation are more pronounced when firms face fierce competition.

To test this conjecture, we use the Hirschman–Herfindahl Index (HHI) as an inverse proxy for the extent of competition and partition our sample into two subsamples based on whether the HHI for a firm year is above the sample median. When the HHI of the industry is higher than the overall industry median, it is attributed to the HHI1 group; otherwise, it is in the HHI0 group. We re-estimate model (1) and expect the coefficient for SCC-related CSR (RSCC) in the low HHI subsample to be significantly greater than that in the high HHI subsample.

Columns (2) and (3) in Table 11 report the results. The dependent variable is lnIGRT and the independent variable is RSCC. The coefficient for RSCC in Column (2) (HHI0, 0.011) is greater than that in Column (3) (HHI1, 0.006), but both are significantly positive. We further find that the difference between these two coefficients is statistically significant (p < 0.1). Therefore, the results are consistent with our conjecture: i.e., SCC-related CSR encourages innovation through resources and knowledge acquisition.

4.4.3. Mechanism for Environment-Related CSR in Firm Innovation

The underlying mechanism through which environment-related CSR enhances firm innovation is to obtain government resources. Environmental protection is one of the Chinese government’s important aims; therefore, firms that conduct environment-related CSR are more likely to build a favorable impression in the eyes of the government [93] or obtain political connections [94] and directly influence the government’s resource allocation decisions [95]. According to the resource-based theory, firm resources are positively related to innovation performance. We use government subsidies as a proxy for government resources. The government subsidies are controlled and allocated by the government and can benefit firm innovation. Hence, we expect a positive relationship between environment-related CSR and government subsidies. To test this conjecture, we test the following OLS econometric model (3).

$$\mathrm{lnSUB}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\mathrm{RENV}+{\displaystyle \sum }_{\mathrm{j}}{\mathsf{\beta }}_{\mathrm{j}}{\mathrm{Control}}_{\mathrm{j},\mathrm{jt}}+\mathrm{IndustryFE}+\mathrm{YearFE}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(3)

where the dependent variable is government subsidies (lnSUB), the independent variable is CSR-related environment (RENV), and the control variable is the same as in model (1). Considering that the government subsidy amount for some firms in some years is 0, the measurement method for the government subsidy is to add 1 to the government subsidy amount and take the natural logarithm.

The results are reported in Column (4) of Table 11. However, the coefficient of RENV is statistically insignificant, which is inconsistent with our expectations. A possible explanation for this finding is that the positive influence of government subsidies on innovation mainly exists in non-state-owned enterprises (non-SOEs) because most non-SOEs have limited resources and their innovation is highly sensitive to their resources. To further support our conjecture, we partition our sample into two subsamples based on whether a firm is an SOE or not and re-estimate model (3). Columns (5) and (6) in Table 11 report the results. The coefficient of RENV is significantly positive (0.023, p < 0.01) in Column (5) (non-SOE subsamples), whereas it is statistically insignificant in Column (6) (SOE subsamples). Therefore, we argue that environment-related CSR promotes firm innovation through gaining resources from the government, but mainly for non-SOEs.

4.5. Exclusion of Alternative Explanations

Flammer and Ioannou [96] suggest that firms tend to engage in both innovation and CSR in response to financial crises. Therefore, our results may be driven by the adverse macro environment, which enhances firms’ propensity to engage in both innovation and CSR [96]. Thus, our results may only be found in an adverse macro environment period and disappear in other periods. To rule out this alternative explanation, we partition our sample into two subsamples based on whether or not the macro environment is adverse and re-estimate our main results. We measure the adverse macro environment from two perspectives: economic downturn and political uncertainty (Besides the economic recession, political uncertainty also largely influences corporate decisions in China ).

An economic downturn is relatively high for firm years after 2012 because of the sharp decrease in the growth rate of the Chinese Gross Domestic Product to less than 10% after 2012. Thus, we partition our sample into two subsamples based on whether or not the economic downturn is high and re-estimate our main results, which are reported in Panel A of

Table 12. Exclusion for alternative mechanism.

Panel A Exclusion for Economy Downturn
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Economy downturn	Low	High	Low	High	Low	High	Low	High	Low	High
Variables.	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT
REMP	0.030 ***	0.033 ***
	(6.20)	(14.27)
RSCC			0.013 ***	0.018 ***
			(4.29)	(11.95)
RENV					0.021 ***	0.024 ***
					(7.67)	(17.26)
RSOC							−0.029 ***	−0.023 ***
							(−8.38)	(−13.32)
TCSR									0.004 ***	0.005 ***
									(3.77)	(9.68)
Intercept	−2.744 ***	−0.406 **	−3.158 ***	−0.516 ***	−2.505 ***	−0.371 **	−3.994 ***	−0.989 ***	−3.183 ***	−0.558 ***
	(−6.51)	(−2.20)	(−7.61)	(−2.80)	(−5.97)	(−2.02)	(−10.39)	(−5.44)	(−7.57)	(−3.01)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Ind and Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	3238	14,409	3238	14,409	3238	14,409	3238	14,409	3238	14,409
F	23.91	76.85	21.97	71.07	25.90	85.79	27.01	74.37	21.57	66.41
Adj-R2	0.072	0.055	0.067	0.051	0.078	0.061	0.081	0.053	0.065	0.048
Panel B Exclusion for political uncertainty
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Political uncertainty	Low	High	Low	High	Low	High	Low	High	Low	High
Variables.	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT	lnIGRT
REMP	0.024 ***	0.028 ***
	(5.48)	(12.00)
RSCC			0.010 ***	0.015 ***
			(3.40)	(9.55)
RENV					0.019 ***	0.021 ***
					(7.37)	(14.50)
RSOC							−0.035 ***	−0.021 ***
							(−10.16)	(−12.75)
TCSR									0.003 ***	0.003 ***
									(2.87)	(6.94)
Intercept	−3.518 ***	−0.677 ***	−3.875 ***	−0.794 ***	−3.257 ***	−0.641 ***	−4.643 ***	−1.161 ***	−3.945 ***	−0.849 ***
	(−8.21)	(−3.78)	(−9.09)	(−4.43)	(−7.66)	(−3.59)	(−11.75)	(−6.59)	(−9.22)	(−4.71)
Control	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Ind and Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	4334	13,313	4334	13,313	4334	13,313	4334	13,313	4334	13,313
F	27.26	53.17	25.48	48.23	29.60	59.40	34.34	54.92	25.16	44.18
Adj-R2	0.062	0.041	0.059	0.038	0.068	0.046	0.078	0.043	0.058	0.034

Note: Panel A examines subsamples based on economy downturn. Panel B examines subsamples based on political uncertainty. t-statistics in parentheses ** p < 0.05, *** p < 0.01.

. We find results for both subsamples (low and high economic downturn) consistent with those reported in earlier tables.

Political uncertainty is high for firm-years from 2013 to 2014, when the 18th National Congress of the Communist Party of China was held. Therefore, we partition our sample into two subsamples based on whether or not political uncertainty is high and re-estimate our main results. Panel B in Table 12 reports the results for both subsamples (low and high political uncertainty), which are also consistent with those reported in earlier tables. In summary, Table 12 suggests that our results are not driven by the adverse macro environment.

5. Discussion and Conclusions

This study examined the relationship between CSR and innovation by identifying distinct CSR categories. Previous studies report inconsistent results for the association between CSR and firm innovation. We document that studies of the relationship between CSR and innovation must consider that the derived resources for innovation from fulfilling CSR for different stakeholders differ. Therefore, according to the degree of resources firms invest in innovation, we divide CSR into employee-, SCC-, environment-, and social welfare-related CSR and use the OLS model to investigate the impact of performing different types of CSR on firm innovation. We also analyze the mechanism by which different types of CSR affect firm innovation.

Based on a resource-based theory perspective, this study argues that total CSR is positively associated with firm innovation, but the impact of CSR to different types of stakeholders on innovation is different. The empirical evidence from a sample of Chinese A-share listed firms supports our hypothesis. Consistent with the resource-based theory, the evidence shows that fulfilling CSR has brought valuable, rare, unrepeatable, and irreplaceable assets to firms and promotes firm innovation. It is also consistent with the observation of McWilliams and Siegel [8]. Specifically, this study finds employee-, SCC-, and environment-related CSR positively promote innovation, while social welfare-related CSR negatively affects innovation. The present study extends findings of previous research [23,24] by classifying CSR into five categories according to the closeness between stakeholders and firms, while they classify CSR according to triple bottom line theory (economy, society, and environment). The classification is more detailed enough to describe the relationship between CSR and corporate innovation.

This study also shows that employee-related CSR reduces the employee turnover rate, thus maintaining human resources and promoting firm innovation. It is consistent with Flammer and Kacperczy [97] observation that employee-related CSR is positively associated with firm innovation. The present study provides empirical support for prior qualitative studies (see, e.g., [98,99]) on employee-related CSR promoting employees’ innovation enthusiasm and identification with the organization.

This study finds SCC-related CSR strengthens cooperation between firms, which brings cooperation resources and promotes firm innovation. Although some studies have theoretically expounded that a good and stable SCC relationship is conducive to resource acquisition and knowledge sharing, which benefit firm innovation (see, e.g., [6,47]), previous literature does not prove it empirically. The present study uses the cross-sectional variation in the firms’ extent of competition to examine SCC-related CSR improvement in firm innovation through resource and knowledge acquisition.

This study argues environment-related CSR increases government subsidies, which brings government resources and promotes firm innovation. Different from Pan et al. [23] research, the present study discusses the relationship between environment-related CSR and firm overall innovation, while previous literature researched the relationship between environment-related CSR and environmental innovation. Based on the resource-based theory, the present study emphasizes that the underlying mechanism through which environment-related CSR enhances firm innovation is to obtain government resources. The empirical evidence supports a positive relationship between environment-related CSR and government subsidies in non-SOEs subsamples because most non-SOEs have limited resources and their innovation is highly sensitive to their resources.

This study shows social welfare-related CSR is negatively associated with firm innovation. There is no literature to directly study the relationship between social welfare-related CSR and innovation. Some studies on the economic consequences of social welfare-related CSR (such as donation and taxation) have reached inconsistent conclusions. Some literature concludes that social welfare-related CSR has positive economic consequences (see, e.g., [68,69,70]), but others have reached the opposite conclusion (see, e.g., [71,72,73]). Based on the data of China, the present study draws the conclusion that the crowding-out effect of social welfare-related CSR on firm innovation is more significant. In addition, the results remain significant to a series of robustness tests, such as fixed effect regression, Tobit regression, and the instrumental variable method. We also exclude the alternative explanations of the adverse macro-environment which enhance a firm’s propensity to engage in both innovation and CSR.

The study findings enrich the research literature on CSR and firm innovation. From the perspective of providing resources, we distinguish the responsibilities for relevant stakeholders and clarify the CSR mechanisms affecting firm innovation. The findings have various managerial implications. First, it enlightens managers to realize the importance of CSR to innovation. Managers should upgrade CSR to the level of organizational strategy and fulfill CSR from the perspective of system theory. Second, managers should consider the different impacts of these four categories of CSR on innovation when making CSR decisions, since there is an optimal level of CSR investment, and exceeding it may have a resource crowding effect. This optimal level depends on the resource level that different types of CSR bring to firms. The closer the relationship between stakeholders and firms, the more innovative resources can be brought to firms. Of course, with the change of economic environment, the relationship between stakeholders and firms will change. For example, in the period of the traditional economy, employees are important resources for innovation, but in the era of the digital economy, the importance of SCC has increased. With environmental problems are increasingly concerned, the resources brought by environment-related CSR have increased. Third, the relationship between CSR and innovation is affected by state institutions, which affect firms’ decision-making. This study finds that the relationship between environment-related CSR and firm innovation is due to the influence of government subsidies, but mainly for non-SOEs. The conclusion shows that government subsidies have a greater impact on CSR for firms with few resources. Firms operating in developing countries can perform more environment-related CSR activities to obtain government support and thus benefit innovation, but this maybe not work for firms operating in developed countries.