Stakeholder Protection, Public Trust, and Corporate Social Responsibility: Evidence from Listed SMEs in China

Abstract

In this paper, we investigate the effects of stakeholder protection and public trust on the corporate social responsibility (CSR) activities of listed enterprises on the Chinese Small and Medium Enterprise (SME) Board. We find that the degree of stakeholder protection has a significantly positive impact on SME CSR activities. The public trust is not associated with SME CSR disclosure significantly; it has a significantly negative impact on the SME implementation levels of CSR activities. Furthermore, the moderating effect of public trust on the relationship between the degree of stakeholder protection and SME CSR activities is not supported by our empirical study.

1. Introduction

Corporate social responsibility (CSR) concerns the managerial consideration of non-market forces or the social aspects of corporate activity outside of a market or regulatory framework, including issues such as community programs, employee welfare, charitable donations, and environmental protection [1]. It is also considered an enterprise’s obligation to protect and improve social welfare, ensuring equitable and sustainable benefits for the relevant stakeholders [2]. Over the past three decades, CSR has attracted much attention in both academic research and enterprise management practice [3].

In developed countries, such as the US and Western European countries, communications about CSR have emerged as a vital part of organizational marketing to enhance enterprises’ images [4]. Moreover, CSR has become increasingly important for the management practices of enterprises in emerging economies. China, for instance, has suffered from product safety, staff welfare, and environmental pollution issues, such as its milk and toy scandals in recent years [5]. Nevertheless, aside from the many studies on CSR in developed countries, research on CSR in emerging economies, particularly on the CSR of Small and Medium Enterprises (SMEs) in large emerging economies, such as China’s, remains relatively scarce. From an academic perspective, CSR scholars are faced with the challenge of identifying the determinants of enterprises in an emerging economy to behave in a socially responsible way [6].

As China becomes more globally integrated, CSR is becoming increasingly important [7]. The Chinese government also encourages domestic companies to learn how to conduct CSR to better address the increasingly serious problems of environmental pollution, ecological crises, labor abuse, and the absence of community service. Nevertheless, how to implement CSR remains a large challenge for enterprises in China. Chinese enterprises’ serious challenges in CSR performance can be partially attributed to the high costs of CSR, as well as the lack of these enterprises’ resources to implement CSR [8]. Since 2006, China’s central government has issued a number of CSR reporting guidelines for large firms as a political strategy to help balance extensive economic growth with the environmental and social impacts of that growth, despite the optional nature of these governmental issuances. In contrast, the CSR activities of China’s SMEs still have no authoritative guidelines. Although the “CSR Guide for China’s SMEs” was jointly issued by the China Center for SME Cooperation Development and Promotion and the National Council of China’s SMEs on 19 December 2013, it is largely symbolic and non-substantive. In fact, according to Marquis and Qian [9], even the CSR reporting guidelines for large firms issued by the central government do not contain specific laws or mandates but are instead examples of the government signaling what is considered to be an important field of enterprise focus.

Furthermore, Small and Medium Enterprises (SMEs) in China have achieved rapid and sustainable growth in the past three decades. Such growth has increasingly contributed to China’s economic development. Zheng and Chen, as well as Liu, argued that SMEs play a highly important role in creating employment opportunities [10,11]. SMEs, both in urban and rural areas, contribute more than 50% of the tax revenue and 65% of patents and inventions. However, under the financial system of China, SMEs are characterized by tight cash flow and narrow financial bases. It is thus difficult for them to meet their increasing funding demands for corporate growth [12]. Meanwhile, the inefficient property protections and legal systems of China increase the uncertainty of SMEs’ future operating activities [13]. China’s SMEs will particularly suffer from their lack of financial resources and enthusiastic goals when they seek to improve their CSR performance. The increasingly serious problems of environmental pollution, ecological crises, labor abuse, and the negative externality of enterprises’ productions in their communities are largely blamed for the lack of CSR actions by SMEs in China. For more details, readers can refer to the relevant data and reports on the China SME Service Platform: http://www.eme2000.com/html/redian/2013/1223/40581.html.

Existing studies on the determinants of CSR in China primarily focus on the empirical relationships between enterprise CSR and other corporate characteristics or political factors [9,14,15,16,17,18,19,20,21]. Nonetheless, CSR activities are affected not only by the economic environment but also by the formal and informal “rules of the game” in the local institutional environment [22]. These “rules of the game” may foster an environment where CSR is promoted actively or discouraged silently. The rule of law is the formal institutional foundation of a local economy. The regional legal infrastructure and its implementation efficiency will influence enterprises’ social behavior and determine the degree of stakeholder protection for an enterprise [23]. Stakeholders are any group or individual who can affect or be affected by the achievements of the organization’s objectives, including employees, customers, suppliers, communities, and other interested parties of an enterprise [24]. Stakeholders’ interests are directly related to the enterprise’s CSR activities and protected by the legal system. Therefore, the degree of stakeholder protection derived from the legal environment may affect the CSR activities of enterprises.

In addition, an emerging line of research has shown that public trust as an informal institutional factor plays an important role in economic development and related transactions [25,26,27]. Akerlof and Shiller note that trust represents the cornerstone of economic theory [28]; Guiso, Sapienza, and Zingales observe that the more trust there is in a country, the more its citizens will invest in the stock market [29]. Jin, Wang, Wang, and Yin found that social trust embedded in the regional environment is an important factor for the investment decisions of foreign institutional investors [30]. Nevertheless, economists and management scientists have not shown much interest in trust [27]. The implications of public trust for the enterprises’ non-market strategies, such as CSR activities, should be investigated thoroughly and deeply.

SMEs always have regional characteristics because of their smaller sizes and localized business organization formations and production factors [31,32], which together make stakeholder protection and social trust—embedded in the regional law and culture environment—important factors for SME CSR activities.

In this study, to extend the perspective of the existing literature on the CSR activities of SMEs, we investigate the effects of stakeholder protection and public trust on the CSR activities of listed enterprises of the Chinese SME Board. Because there are few entrepreneurial firms listed on the ChiNext market of the Shenzhen Stock Exchange, and the overwhelming majority of listed entrepreneurial firms belongs to the high-tech industry, this study does not focus on the CSR activities of entrepreneurial firms listed in the ChiNext market. Moreover, we employ a sample of enterprises listed on the Chinese SME Board primarily because only the CSR information of the listed SMEs is available; unlisted SME CSR information is scarcely disclosed in China. We further extend our research to examine whether the public trust playing an informal institutional role has a moderating effect on the relationship between the degree of stakeholder protection and SME CSR activities.

We find that the degree of stakeholder protection derived from the local legal environment has a significantly positive impact on SME CSR activities. Public trust is not significantly associated with SME CSR disclosure, and public trust has a significantly negative impact on the SME implementation levels of CSR activities. Furthermore, the moderating effect of public trust on the relationship between the degree of stakeholder protection and SME CSR activities is not supported by our empirical study.

The rest of this paper is organized as follows. Section 2 reviews the related literature and develops our hypotheses. Section 3 details our sample data, variable definitions, and identification strategy. Section 4 reports and explains our empirical results, while Section 5 concludes the research.

2. Related Literature and Hypothesis Development

2.1. Institutional Background of the SME Board

The Chinese government established its capital market in the early 1990s. When the Security Law was implemented, the China Securities Regulatory Commission (CSRC) was in charge of guiding the Initial Public Offering (IPO) market. Only a few SMEs could be selected to go public in the capital market because China’s capital market was designed to serve the reform of large state-controlled firms at that time. Before long, the CSRC assigned channels directly to sponsors based on their size, and operations under the channel system were adopted in 2000. The government evaluated and judged IPO applications after sponsors recommended prospective firms. In early 2004, this system was substituted for a sponsorship system that made the listing regulations more market-oriented and provided opportunities for a large number of SMEs to go public.

China’s capital market has long been dominated by the exchanges in Shanghai and Shenzhen, without much diversification. Many enterprises were excluded because of the stringent listing requirements of size and profitability. To expand the capital market, the Shenzhen Stock Exchange (SZSE) began to build an SME listing market and constructed the SME Board in 2004. As of early 2017, more than 800 SMEs were listed on this board. With the launch of the SME Board, China moved a step closer to developing a multilayered capital market. The development of capital markets has been made possible by China’s economic and financial reforms.

2.2. Literature Review and Development of Our Hypotheses

Harjoto and Jo found that firms with a higher proportion of outside independent directors are more likely to be socially responsible [33]. Mallin and Michelon also confirmed that better social performance is associated with independent boards [34]. Walls, Berrone, and Phan noted that board characteristics play a particularly important role in a firm’s CSR performance; firms have more environmental concerns when their boards are larger and more independent [35].

Arora and Dharwadkar showed that the availability of free financial resources significantly moderates the relationship between board composition and CSR activities, such that greater board independence is associated with improved CSR only under the condition that spare financial resources are available [36]. Li, Luo, Wang, and Wu noted that better-performing firms are more likely than worse-performing ones to disclose CSR information and provide high-quality CSR reports [19].

Dam and Scholtens provided empirical support for the relationship between ownership type and CSR policies in Europe [37]. This empirical research has addressed a number of themes related to how the ownership structure and identity of firm ownership influence patterns of engagement with CSR. Anderson, Mansi, and Reeb indicated that large shareholders who are unable to sell their substantial shareholdings without eroding their value might thus be motivated to promote CSR because of CSR’s capacity to enhance a firm’s reputation and survival [38]. Marquis and Qian further integrated institutional theory with research on corporate political strategy to develop a political dependence model that explains how different types of dependency on the government lead enterprises to issue CSR reports [9].

Moreover, firms must continuously legitimize their activities to retain congruence between social and organizational objectives [39]. According to legitimacy theory [40,41], large companies and companies whose products are primarily exported to developed countries are particularly active in terms of their CSR activities and reporting because they are more visible and open to public scrutiny and hence have greater legitimacy needs. Several studies additionally document the relationship between the level of economic development and CSR [42]. A study by Li and Zhang revealed that firm size, profitability, leverage, and growth opportunities also affect the CSR in China [16]. In addition, Li, Fetscherin, Alon, Lattemann, and Yeh identified industry and firm-level factors to further explain cross-national CSR practices [6].

Aside from the abovementioned literature on CSR research, Yin and Zhang posited that institutional infrastructure and cultural ethics may exert an abiding influence on CSR approaches in emerging economies [18]. Frynas and Yamahaki stated that stakeholder theory and institutional theory dominate the theorization of external drivers of CSR [43]; the stakeholder theory predicts corporate actions to be a direct result of the pressures from different stakeholders’ appeals. In

Table 1. Summary of closely linked literature.

Subjects	Literature References
Board characteristics play a particularly important role in a firm’s CSR performance.	Harjoto and Jo [33]; Mallin and Michelon [34]; and Walls, Berrone, and Phan [35].
Financial resources have significant impacts on firm CSR activities.	Li, Luo, Wang, and Wu [19]; and Arora and Dharwadkar [36].
Ownership structure and identity of firm ownership influence patterns of engagement with CSR.	Marquis and Qian [9]; Dam and Scholtens [37]; and Anderson, Mansi, and Reeb [38]
Greater legitimacy needs influence corporate CSR activities.	Deegan [39]; Ashford and Gibbs [40]; and Branco and Rodrigues [41]
Studies additionally document the relationship between the level of economic development, or growth opportunities, and CSR.	Li, Fetscherin, Alon, Lattemann, and Yeh [6]; Li and Zhang [16]; and Baughn, Bodie, and McIntosh [42].
Institutional infrastructure and cultural ethics may exert an abiding influence on CSR approaches.	Yin and Zhang [18]; and Frynas and Yamahaki [43].

Note: CSR = corporate social responsibility.

, we summarize some closely linked literature to display what has been done within this particular topic.

Aside from the many studies on CSR in developed countries, research on CSR in emerging economies, particularly on the CSR of SMEs in large emerging economies, such as China’s, remains relatively scarce. Additionally, existing studies on the determinants of CSR in China primarily focus on the empirical relationships between enterprise CSR and other corporate characteristics. The impacts of regional legal infrastructure and informal institutional factors on SME CSR activities are always overlooked in the literature. In this study, to fill the gap of the existing literature on the CSR activities of SMEs, we investigate the effects of stakeholder protection and public trust on the CSR activities of listed enterprises of the Chinese SME Board.

Particularly, empirical studies have confirmed the impact of stakeholder pressures on CSR-related activities, such as the influence of stakeholder pressures on environmental policies and corporate philanthropy [44,45]. Very few stakeholder studies implicitly demonstrate variability in terms of individual interests and attitudes both within and between stakeholder groups [46]. Most studies provide rich evidence related to the impact of stakeholder pressure on firms’ CSR performance. Moreover, most stakeholder studies have largely failed to paint the overall stakeholder relationship as a multifaceted and multi-objective phenomenon [43]. Additionally, institutional theory suggests that enterprises have to conform to stakeholder appeals in a given business environment [47]. This also emphasizes the regulatory role of institutions that underpin economic activities [22]. Thus, the degree of stakeholder protection derived from the local legal environment may affect the CSR activities of enterprises because stakeholders’ appeals for CSR activities and pressures on CSR-related activities need to be guaranteed by a local legal environment that accordingly determines the degree of stakeholder protection. Although the conflict of interest between firm owners and managers with respect to pursuing social and environmental objectives was a concern in earlier CSR-related studies, a number of recent agency studies view CSR as conductive to financial and non-financial performance and find that large shareholders support firms’ CSR initiatives because they prefer to invest in responsible firms to avoid financial risks [48]. Thus, improvement in the degree of stakeholder protection in general will yield more CSR activities.

Cross and Prentice also emphasized the implications of legal protections on the behaviors of players in economic activity [49]. The opportunism of an enterprise’s managers will be decreased significantly with an increase in the degree of stakeholder protection, which provides the key to breaking the prisoner’s dilemma of CSR activities. SMEs in China are characterized by a tight cash flow, narrow financial bases, and a higher uncertainty of future development prospects. Considering the relatively scarce external and internal resources of SMEs and the regionality of SMEs’ business activities (according to the “Civil Procedure Law” in China, the disputes involving stakeholders and a corporation primarily come under the jurisdiction of the People’s Court where the corporation is located), the aforesaid mechanism is likely to work well. These arguments lead us to offer the following hypothesis.

Hypothesis 1. 

The degree of stakeholder protection derived from the local legal environment has a positive impact on SME CSR activities.

Moreover, Akerlof and Shiller noted that trust is the cornerstone of economic theory [28]. Social trust embedded in the regional environment is also an important factor for SMEs’ CSR activities. Public trust, as an informal institution, is defined as the subjective probability that individuals will consider the possibility of being cheated [29]. Without public trust, people may become stuck in an undesirable equilibrium, and with the help of trust, a better outcome can be attained.

Compared to formal institutions, such as laws and long-term government policies, public trust usually plays an implicit and non-mandatory role in economic behaviors. As the CSR disclosure of SMEs is conducted primarily under the obvious pressures from stakeholder appeals, government intentions, and legal constraints, public trust may have no significant impact on SME CSR disclosures. Su, Peng, Tan, and Cheung reported that, in an emerging economy, firms that adopt CSR practices positively signal to investors that their firms have superior capabilities for filling institutional voids [21]. For SMEs in China, these firms especially need to send this positive signal to obtain financial resources and local market recognition. In sum, CSR disclosure is more likely to be affected by the degree of stakeholder protection coming from the stakeholders’ pressures boosted by the local legal environment rather than by public trust embedded in the regional environment.

On the other hand, Jin, Wang, Wang, and Yin posited that social trust can function as an information-processing mechanism to cope with informational barriers [30]. When local public trust increases, the subjective probability that individuals will consider the possibility of being cheated falls, and SMEs will have the opportunity to break their undesirable equilibrium. The detrimental effects from the informational asymmetry between the insiders and outsiders of an enterprise will also be reduced. Therefore, although CSR disclosure is a necessary signal sent by China’s SMEs to seek financial resources and local market recognition, SMEs will no longer need to maintain a higher implementation level of CSR activities, particularly considering that CSR activities are bound to use an enterprise’s resources. Social trust embedded in the regional environment still has a potentially negative influence on the implementation level of SME CSR. Notably, CSR decisions are made by SME managers who are profoundly influenced by local social trust, because, as per the theory of social and economic networks, the ties between members and their social environments will have important impacts on members’ behaviors due to their efforts to maintain social capital and minimize information asymmetry [50]. Thus, we make the following hypothesis:

Hypothesis 2A. 

Public trust is likely to have no significant relation to SME CSR disclosure, perhaps because the necessity of sending out a positive signal via CSR initiatives dominates the potential influence of public trust in CSR disclosures.

Hypothesis 2B. 

Public trust is expected to have a negative impact on the SME implementation levels of CSR activities.

Furthermore, the impacts of the local legal environment and public trust on SME CSR may or may not interact with each other. SMEs usually relate very closely to their regional environments. Individuals with different beliefs, embedded in the regional environment, will take different actions despite facing identical business environments [51]. All else being equal, when the managers of SMEs engage in CSR strategies, they are fettered by the local legal environment and public trust.

On the one hand, the positive effects of local legal protection for stakeholders’ interests on CSR activities may be weakened by an increase in public trust as the detrimental effects of information asymmetry between the insiders and outsiders of an enterprise become reduced. In this scenario, SMEs will no longer need to send out redundant signals through CSR activities, but the local public trust may have no significant moderating effects on the relationship between the degree of stakeholder protection and SME CSR activities because the role of public trust as an informal institutional factor is auxiliary and soft [27], which implies that the moderating effect of public trust should not be overestimated. In particular, unlike Reference [51], when the public trust does not attain a certain threshold, its level is low overall, and the effects of legal protection on SME CSR activities may have no significant differences at relatively higher or lower levels of local public trust. Our arguments yield the following hypotheses.

Hypothesis 3A. 

The improvement of local public trust will weaken the positive effects of legal protections for stakeholders on SME CSR activities.

Hypothesis 3B. 

The moderating effect of public trust on the relationship between the degree of stakeholder protection and SME CSR activities is not expected in the context of China.

3. Data and Methodology

3.1. Sample Selection

This paper focuses on the CSR activities of enterprises listed on the SME Board in China. Information on the CSR performance of SMEs is available only for enterprises listed on the SME Board. Unlisted SMEs virtually never disclose this information. Information about SME CSR disclosures was collected from the CSR research database provided via the China Stock Market and Accounting Research (CSMAR) dataset. Information on SME CSR scores reflecting the conduct level of CSR activities was collected from Rankins CSR Ratings, RKS. Rankins CSR Ratings, RKS, was established in 2007. RKS is now one of the most authoritative third-party CSR rating agencies. Its report rating system is in accordance with the international social responsibility standard ISO26000, which includes 15 first-class indices, such as labor rights, environment, fair operation, community participation, etc., as well as 63 second-class indices, including social investment and climate change mitigation, based on four zero class indices, namely macrocosm, content, technique, and industry. The maximum score in this rating system is 100: Macrocosm has an assigned weight of 30%, content has an assigned weight of 45%, technique has an assigned weight of 15%, and industry has an assigned weight of 10%. The degree of stakeholder protection is represented by the index of regional market development and legal institutional environment proposed by Wang, Fan, and Yu [52]. This index is constructed based on three first-class scores: the development of the market intermediary organization, the legal environment, and property protection. We employ a proxy of the number of non-government organizations (NGOs) per million people for public trusts at the provincial level. This variable is calculated by using information from the National NGO administration Bureau of China. According to Chenhall, Hall, and Smith [53], as well as Wu, Firth, and Rui [54], NGOs are an important component of social capital and foster social trust. Jin, Wang, Wang, and Yin also employed NGOs as a proxy for the social trust level across provinces in China [30].

The data on variables reflecting corporate governance, finance, and business information were collected or calculated from the China Stock Market and Accounting Research (CSMAR) dataset. Data on managers’ political connections, as well as board political connections, were collected manually from the firms’ prospectuses, as well as from the website of the Shenzhen Stock Exchange. Macroeconomic variable data were found in the China Statistical Yearbooks for various years. We adopted the industrial classification criterion “Listed Companies Classification and Code”, which was issued by China’s Securities Regulatory Commission.

Because the time period of the regional legal environment index proposed by Wang, Fan, and Yu [52] is between 2008 and 2014, our study also spans 2008 to 2014. To remove the influence of outliers on our empirical results, we winsorized all continuous variables in the interval of (1%, 99%). Finally, all data on the value variables were inflation-adjusted to eliminate the impact of price fluctuations.

3.2. Variable Definition and Summary Statistics

The dependent variables in this study related to SME CSR activities include CSR disclosure (Disclosure), which is a dummy variable that equals 1 if the enterprise disclosed the CSR report in a given year (0 otherwise) and CSR performance level (Scores). Our main explanatory variables are the degree of stakeholder protection (Stakeprotect) and public trust (Ptrust). In line with Chenhall, Hall, and Smith [53], Wu, Firth, and Rui [54], as well as Jin, Wang, Wang, and Yin [30], Ptrust is a logarithm of the number of non-government organizations (NGOs) per million people.

Following Anderson, Mansi, and Reeb [38]; Branco and Rodrigues [41]; Li and Zhang [16]; Li, Fetscherin, Alon, Lattemann, and Yeh [6]; Harjoto and Jo [33]; Mallin and Michelon [34]; Arora and Dharwadkar [36]; Dam and Scholtens [37]; Walls, Berrone, and Phan [35]; Li, Luo, Wang, and Wu [19]; and Marquis and Qian [9], several control variables are included in our empirical equations, to control for other factors that have been identified in the current literature and that may affect firms’ CSR performance.

In particular, we control for manager’s political connection (Mpconnection) as a dummy variable that equals 1 if the executive was formerly an officer of the government or a manager of a state-owned firm, and 0 otherwise; and the board political connection (Bpconnection) is used as a dummy variable that equals 1 if one member of the board of directors (excluding the independent directors) was formerly an officer of the government or a manager of a state-owned firm, and 0 otherwise.

This study controls for the following factors: the logarithm of the number of board directors (Boardscale), the proportion of independent directors in a directorate (Indirector), the large shareholder’s ownership (Largeshare), which is a ratio of stock holdings by the top ten shareholders to the total equity when the enterprise goes public; firm age (Firmage), calculated by the logarithm of the number of years since the firm was established, firms’ return on assets (Roa); firm size (Firmsize), measured by the logarithm of the SMEs’ total assets; firm slack resources (Slackres), calculated as a ratio of the sum of cash flow from a firm’s operating, financing, and investing activities compared to the SMEs’ total assets; the firm’s debt-to-asset ratio (Debtoasset); the firm’s Tobin’s Q (Tobinq); firm foreign sales (Fsales), which is the firm’s foreign sales as a percentage of total sales; the GDP per capita (Gdppc), which is a logarithm of the annual GDP per capita of the province where the firm is located; and the high-tech industry dummy variable (Hdummy), which equals 1 if the enterprise belongs to a high-tech industry, and 0 otherwise. High-tech industries are identified according to the “High-Tech Fields with the Government’s Primary Support” promulgated by China’s State Council in 2016, which includes eight fields: electronic information, bio-medicine, new materials, aerospace, high-tech services, new energy and energy savings, resource and environment, and advanced manufacturing and automation. Hdummy can capture the industry cluster effects because the number of SMEs that belong to high-tech industries accounts for 62.80% of the total in our sample data. Moreover, due to our limited sample size, particularly for Scores, including redundant industry dummies in the regressions would reduce the degree of freedom too greatly. Finally, state ownership is also not considered here. After the implementation of the policy “Invigorating Large Enterprises and Relaxing Control over Small Ones” by the Chinese government, in the industry sector, almost all SMEs in China no longer have a nominal relationship with China’s government, officially at least. In addition, dummy variables for year are also included in our regressions, to control for the macroeconomic fluctuations from other external shocks.

Table 2. Sample distribution of Small and Medium Enterprise (SME) CSR disclosure.

Industry Classification	Total Enterprises	CSR Firms	Non-CSR Firms
Agriculture	16	3	13
Forestry	16	2	14
Livestock farming	64	17	47
Fishery	24	0	24
Coal mining and preparation	8	5	3
Nonferrous metal mining	8	8	0
Mining auxiliary activities	32	0	32
Agro-food processing	136	24	112
Food manufacturing	104	11	93
Liquor, beverages, and tea	40	6	34
Textiles	120	21	99
Clothes	152	34	118
Leather, fur, feathers, and footwear	16	4	12
Wood, bamboo, rattan, palm, and grass products	32	0	32
Furniture	24	0	24
Paper making	72	12	60
Printing and recording media reproduction	32	8	24
Education, art, sports, and entertainment	56	12	44
Petroleum, coking, and nuclear fuel processing	16	0	16
Chemical raw materials and chemical products	504	54	450
Medical products	376	48	328
Chemical fiber manufacturing	48	13	35
Rubber and plastic	152	19	133
Nonmetallic mineral products	160	29	131
Ferrous metal smelting and rolling	24	8	16
Non-ferrous metal smelting and rolling	152	22	130
Metal products	232	10	222
General equipment manufacturing	312	35	277
Special equipment manufacturing	344	39	305
Automobile manufacturing	256	7	249
Railway, shipbuilding, and aerospace	24	1	23
Electrical machinery and equipment	560	47	513
Computer and communications	648	77	571
Instrument and apparatus manufacturing	48	7	41
Other manufacturing	72	9	63
Comprehensive utilization of waste resources	16	7	9
Electric power, heat production, and supply	16	7	9
Gas production and supply	24	0	24
Civil engineering	128	22	106
Architectural decoration and other construction	88	8	80
Wholesale trade	56	5	51
Retail sales	104	13	91
Road transportation	24	0	24
Water transportation	16	0	16
Handling and forwarding agents	8	0	8
Warehousing industry	16	0	16
Catering industry	8	0	8
Telecommunications, radio, and television	8	0	8
Internet and related services	104	16	88
Software and information technology services	248	61	187
Estate industry	88	15	73
Business services	96	3	93
Professional technical services	48	1	47
Ecological and environmental protection	16	0	16
Public facility management	24	7	17
Public health	8	1	7
Film and video recording	32	1	31
Culture and art	8	0	8

presents a sample distribution of SME CSR disclosure. Industry classification is based on the “Listed Companies Classification and Code” issued by China’s Securities Regulatory Commission. In this study, 648 enterprises participate in the computer and communications industry, which is the largest industry in our sample data. In this industry classification, the number of enterprises disclosing CSR reports (CSR firms) is 77, and the number of those not disclosing CSR reports (NCSR firms) is 571. In sum, the number of NCSR firms is greater than that of the CSR firms.

Table 3. Descriptive statistics of the main variables.

Variable	Observations	Mean	Std. Dev.	Minimum	Maximum
Disclosure	6064	0.125	0.331	0.000	1.000
Scores	429	34.683	9.538	18.770	63.610
Stakeprotect	5306	8.103	4.117	1.480	16.190
Ptrust	6064	6.022	0.350	4.652	6.915
Mpconnection	4645	0.675	0.468	0.000	1.000
Bpconnection	4645	0.745	0.436	0.000	1.000
Boardscale	4627	2.133	0.174	1.610	2.480
Indirector	4600	37.398	8.087	22.220	60.000
Largeshare	4658	63.804	13.210	31.060	90.040
Firmage	6060	2.556	0.467	1.100	3.660
Roa	4646	5.163	4.934	−13.230	21.060
Firmsize	4651	7.629	0.845	5.890	10.060
Slackres	4641	2.151	14.740	−29.530	59.140
Debtoasset	4646	35.640	18.933	3.320	80.920
Tobinq	4443	2.930	1.764	1.040	10.200
Fsales	4575	16.988	24.219	0.000	93.300
Gdppc	6064	1.587	0.449	0.370	2.440

presents a summary of the statistics for our sample. All ratio variables are displayed as percentages (%). All level variables are in millions of CNY. Before taking the logarithm, the GDP per capita is represented in units of ten thousand CNY. Moreover, a large number of SMEs that are reported to have disclosed CSR information in the CSMAR dataset have no CSR scores in the dataset of Rankins CSR Ratings. Nevertheless, this is an inevitable occurrence when studying SME CSR activities in China because of the lack of detailed data on China’s SMEs. The difference between the extreme values reflects the comparative results across enterprises over different years. The difference between the extreme values in individual years is not as large as it appears here. We tested for multi-collinearity between the explanatory and main control variables, including the continuous variables and dummy variables, to guard against the impacts of highly multi-collinear regressors on our parameter estimates. The condition number obtained from the test is 3.737, and the mean VIF is 1.88, which suggests that high multi-collinearity was not found.

Furthermore, Figure 1 and Figure 2 show the changes for the annual average degree of stakeholder protection and the annual average public trust across China’s provinces over the sample years. To summarize, these figures show that the indexes of stakeholder protection and public trust, in general, have an upward trend, despite fluctuating during some sample periods.

3.3. Estimation Methods

We apply various methods to test our hypotheses. To examine the relationship between the degree of stakeholder protection and public trust and the likelihood of SME CSR disclosure, as well as the moderating effect of public trust on the association of stakeholder protection with CSR disclosure, panel Probit and Logit regressions are employed.

Additionally, we employ panel data model clustering in two dimensions simultaneously (e.g., firm and time), as proposed by Petersen [55], to test Hypotheses 1, 2B, 3A, and 3B. This approach allows for correlations among different firms in the same year and for different years in the same firm. Petersen argues that, in empirical financial work, researchers are often confronted with panel data. The residuals may be correlated either across firms or across time, and Ordinary Least Squares (OLS) standard errors can be biased [55]. The best method for estimating standard errors in a panel dataset depends on the source of dependence in the data. For datasets with only a firm effect, standard errors clustered by firm produce unbiased standard errors. If the data have only a time effect, Fama–MacBeth estimates are better than standard errors clustered by time when there are few years and are equally good when the number of years is sufficiently large.

Nevertheless, if we cluster by firm, we must assume there is no time effect (no cross-sectional correlation) and vice versa. Because these assumptions can be incorrect in many situations in empirical works on finance [56,57], we should consider a data structure with both a firm and a time effect [55]. One way that researchers can address two sources of correlation is to estimate one of the dimensions by including dummy variables. However, this approach only works when the dependence is correctly specified. Because we do not always know the precise form of this dependence, a less parametric approach is preferred. The solution is to simultaneously cluster in two dimensions, namely firm and time. In this study, we cluster by using the enterprise and time dimensions; standard errors clustered in two dimensions are unbiased and produce correctly sized confidence intervals.

The two-step estimator of the Heckman selection model is also used to test Hypotheses 1, 2B, 3A, and 3B. The selection equation includes Disclosure, the main explanatory variables, and all other control variables in this study. The outcome equation includes the dependent variable of Scores and the same covariates used in the selection equation. Cameron and Trivedi state that an investigator usually uses the same set of regressors in both equations because it is often hard to determine an excluded variable that does not directly affect the outcome but does affect selection [58]. Two-step estimation relies on a univariate assumption and is expected to be relatively more robust than one-step estimation. The Heckman selection model considers the possibility of selection bias by allowing for possible dependence in two parts: the selection equation and the outcome equation of the entire model.

Moreover, to control for potential endogeneity, we include the lagged main explanatory variables instead of their simultaneous forms in the regression, partially to reduce the potential for reverse causation. Structural equation modeling (SEM) is used to further conduct a robustness test against the possible influence of the potential endogeneity problem on our empirical results.

4. Empirical Results

4.1. The Effects of Stakeholder Protection and Public Trust on SME CSR Activities

Table 4. Estimates from the panel Probit and Logit regressions of SME CSR disclosure.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Stakeprotect	0.081 **	0.081 **	0.085 **	0.094 **	0.084 **	0.179 **	0.180 **	0.165 *	0.226 ***	0.183 **
	(0.039)	(0.039)	(0.040)	(0.039)	(0.039)	(0.087)	(0.087)	(0.087)	(0.086)	(0.087)
Ptrust	−0.186	−0.278	−0.184	−0.573	−0.178	0.088	−0.228	0.384	−1.041	0.098
	(0.393)	(0.387)	(0.392)	(0.408)	(0.390)	(1.112)	(1.099)	(1.085)	(1.030)	(1.114)
Mpconnection	0.682 ***	0.080	0.357	0.566 **		0.129	0.130	0.127	0.131
	(0.254)	(0.184)	(0.270)	(0.247)		(9.118)	(9.078)	(8.797)	(8.552)
Bpconnection	−0.948 ***		−0.488	−0.696 **		--		0.153	0.131
	(0.278)		(0.311)	(0.275)				(12.837)	(17.197)
Boardscale	0.137		0.285	−0.059	0.108	−0.140		−0.246	−0.370	−0.167
	(0.718)		(0.729)	(0.699)	(0.721)	(1.462)		(1.424)	(1.409)	(1.463)
Indirector	−0.006		−0.003	−0.007	−0.006	−0.029		−0.024	−0.026	−0.029
	(0.015)		(0.015)	(0.015)	(0.015)	(0.030)		(0.029)	(0.030)	(0.030)
Largeshare	−0.008	−0.007	−0.004	−0.009	−0.008	−0.011	−0.010	−0.006	−0.009	−0.009
	(0.007)	(0.007)	(0.007)	(0.007)	(0.007)	(0.018)	(0.018)	(0.017)	(0.017)	(0.018)
Firmage	0.848 ***	0.829 ***	0.695 ***	1.017 ***	0.834 ***	−3.905 *	−3.793 *	−3.849 *	−0.653	−4.033 *
	(0.242)	(0.234)	(0.236)	(0.246)	(0.237)	(2.227)	(2.191)	(2.128)	(2.025)	(2.225)
Roa	0.032 **	0.035 **		0.033 **	0.033 **	0.028	0.034		0.030	0.029
	(0.016)	(0.016)		(0.016)	(0.016)	(0.032)	(0.032)		(0.031)	(0.032)
Firmsize	1.060 ***	1.041 ***	1.043 ***	1.095 ***	1.045 ***	1.567 ***	1.518 ***	1.639 ***	1.444 ***	1.579 ***
	(0.174)	(0.170)	(0.169)	(0.165)	(0.171)	(0.498)	(0.489)	(0.448)	(0.442)	(0.498)
Slackres	0.002	0.002		0.002	0.002	0.004	0.005		0.005	0.004
	(0.004)	(0.004)		(0.004)	(0.004)	(0.007)	(0.006)		(0.007)	(0.006)
Debtoasset	−0.002	−0.0004		−0.003	−0.002	0.008	0.011		0.002	0.008
	(0.006)	(0.006)		(0.006)	(0.006)	(0.014)	(0.014)		(0.013)	(0.014)
Tobinq	0.014	0.019		0.005	0.014	−0.060	−0.057		−0.070	−0.061
	(0.063)	(0.063)		(0.060)	(0.064)	(0.114)	(0.113)		(0.108)	(0.114)
Fsales	−0.006	−0.007	−0.009 *		−0.007	−0.015	−0.015	−0.016		−0.015
	(0.005)	(0.004)	(0.005)		(0.004)	(0.018)	(0.017)	(0.017)		(0.017)
Gdppc	−1.350 ***	−1.241 ***	−1.302 ***	−1.301 ***	−1.384 ***	−4.001 *	−2.957	−4.060 *	−3.203	−4.079 *
	(0.462)	(0.448)	(0.463)	(0.451)	(0.461)	(2.249)	(2.184)	(2.205)	(2.147)	(2.250)
Hdummy	−0.214	−0.240	−0.174		−0.241	--	--	--		--
	(0.190)	(0.189)	(0.180)		(0.189)
Yeardummies	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
p-value of χ2 test	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Number of obs.	3642	3675	3787	3701	3642	625	641	641	673	625

Note: *, **, and *** denote statistical significance at 10%, 5%, and 1% levels, respectively. The corresponding clustered sandwich standard errors are displayed in parentheses.

presents the panel Probit and Logit regression estimates. The estimates in Columns (1)–(5) are obtained from random-effects Probit regressions. The Likelihood Ratio test results strongly reject the pooled panel Probit regressions, which justify using random-effects panel Probit regressions. It is worth mentioning that the fixed-effects panel Probit model always yields inconsistent estimates [58].

Additionally, the values in Columns (6)–(10) are estimated with fixed-effects panel Logit models. The pooled Logit model would lead to inconsistent parameter estimates because it ignores the possibility of unobserved individual heterogeneity. Here, even if the unobserved individual heterogeneity is correlated with the regressors, and if the link function for the distribution of the dependent variable is assumed to have a logistic distribution instead of a standard normal distribution, the signs and significance of the parameter estimates for Stakeprotect and Ptrust are very consistent. We could obtain similar results by using population-averaged panel Probit and Logit estimators. To save space, however, these results are not reported here.

Table 4 shows that our Hypotheses 1 and 2A are supported by the empirical evidence. The coefficient estimates of Stakeprotect, from 0.081 to 0.226 in different specifications of our regressions, are all significantly positive. The signs for the coefficient estimates of Ptrust are also positive and negative, and they are all insignificant. The empirical results confirm that the degree of stakeholder protection derived from the local legal environment has a positive impact on the SMEs’ CSR activities (i.e., CSR disclosures). Moreover, public trust is likely to have no significant relationship to the SMEs’ CSR disclosures, perhaps because the necessity to send a positive signal via CSR initiatives dominates the potential influence of public trust on CSR disclosures.

Table 5. Estimates from the panel data models clustering in two dimensions and the Heckman selection models.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Stakeprotect	0.542 **	0.476 **	0.556 ***	0.596 ***	0.485 **	0.501 ***	0.681 ***	0.510 ***	0.486	0.619 ***
	(0.223)	(0.218)	(0.208)	(0.217)	(0.216)	(0.173)	(0.244)	(0.170)	(1.348)	(0.234)
Ptrust	−8.664 ***	−8.539 ***	−8.839 ***	−8.454 ***	−8.486 ***	−8.491 ***	−10.133 ***	−8.770 ***	−0.742	−9.578 ***
	(2.606)	(2.715)	(2.774)	(2.714)	(2.649)	(1.921)	(2.348)	(1.807)	(11.550)	(2.257)
Mpconnection	4.386 ***	1.711	3.405 ***	4.406 ***		4.665 **	1.796	3.498 **	14.718
	(1.503)	(1.543)	(1.307)	(1.489)		(1.838)	(1.113)	(1.746)	(13.481)
Bpconnection	−3.150 ***		−2.603 **	−3.170 ***		−3.512 *		−2.800	−16.368
	(1.131)		(1.103)	(1.219)		(1.877)		(1.784)	(16.446)
Boardscale	7.283 *		7.143	8.648 **	7.883 *	6.298 *		6.417 *	10.090	6.285 *
	(4.077)		(4.550)	(4.219)	(4.065)	(3.563)		(3.577)	(13.854)	(3.788)
Indirector	0.065		0.086	0.093	0.083	0.048		0.074	0.098	0.059
	(0.099)		(0.111)	(0.096)	(0.097)	(0.081)		(0.081)	(0.298)	(0.085)
Largeshare	−0.164 ***	−0.167 **	−0.170 ***	−0.152 ***	−0.172 ***	−0.172 ***	−0.146 ***	−0.169 ***	−0.283	−0.160 ***
	(0.063)	(0.065)	(0.062)	(0.056)	(0.063)	(0.039)	(0.045)	(0.038)	(0.190)	(0.042)
Firmage	0.309	−0.162	0.157	0.314	−0.029	0.216	−3.208	0.243	16.516	−2.178
	(1.500)	(1.564)	(1.489)	(1.549)	(1.565)	(1.397)	(2.451)	(1.362)	(19.513)	(2.391)
Roa	−0.284 *	−0.236		−0.295 **	−0.254	−0.219 *	−0.378 **		0.705	−0.330 *
	(0.154)	(0.157)		(0.149)	(0.157)	(0.127)	(0.180)		(1.181)	(0.171)
Firmsize	3.001 **	3.043 **	1.860 **	2.771 **	2.805 **	3.487 ***	0.624	2.379 ***	17.666	1.274
	(1.353)	(1.388)	(0.866)	(1.375)	(1.324)	(1.141)	(1.977)	(0.894)	(17.053)	(1.905)
Slackres	−0.026	−0.022		−0.029 *	−0.026	−0.024	−0.010		−0.082	−0.017
	(0.017)	(0.015)		(0.016)	(0.017)	(0.030)	(0.033)		(0.126)	(0.032)
Debtoasset	−0.074	−0.061		−0.073	−0.047	−0.074 **	−0.085 **		0.009	−0.063 *
	(0.068)	(0.066)		(0.065)	(0.061)	(0.034)	(0.039)		(0.157)	(0.036)
Tobinq	0.645	0.526		0.678 *	0.658	0.497	−0.265		3.617	0.045
	(0.402)	(0.405)		(0.348)	(0.427)	(0.410)	(0.578)		(3.977)	(0.561)
Fsales	0.027	0.032	0.028		0.026	0.027	0.037 *	0.029		0.031
	(0.033)	(0.032)	(0.033)		(0.034)	(0.019)	(0.020)	(0.019)		(0.019)
Gdppc	0.856	0.748	0.709	0.325	1.175	0.589	0.670	0.374	−0.781	0.954
	(1.871)	(1.887)	(1.934)	(1.918)	(1.862)	(1.823)	(1.967)	(1.816)	(7.031)	(1.884)
Hdummy	0.094	0.275	0.731		0.469	0.167	0.573	0.609		0.752
	(1.371)	(1.345)	(1.347)		(1.401)	(0.972)	(1.063)	(0.957)		(0.992)
Constant	52.077 ***	66.049 ***	59.964 ***	47.587 **	46.846 **	46.070 *	132.242 ***	53.181 **	−311.278	95.247 *
	(18.907)	(18.761)	(21.169)	(21.315)	(22.402)	(26.283)	(51.195)	(23.685)	(421.277)	(52.489)
Yeardummies	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
p value of F test	0.000	0.000	0.000	0.000	0.000
p value of χ2 test						0.000	0.000	0.000	0.932	0.000
Number of obs.	415	416	420	422	415	3448	3448	3448	3448	3448

Note: *, **, and *** denote statistical significance at 10%, 5%, and 1% levels, respectively. For (1)–(5), the corresponding 2D clustered standard errors are displayed in parentheses. The corresponding robust standard errors are reported in parentheses in (6)–(10).

shows empirical evidence for the effects of stakeholder protection and public trust on the SME implementation levels of CSR activities employing panel data model clustering in two dimensions simultaneously (e.g., firm and time), as proposed by Petersen [55] and the Heckman selection models. The former is used to estimate Specifications (1)–(5); the latter approach is employed to estimate Specifications (6)–(10) in Table 5. The evidence in Table 5 shows that the degree of stakeholder protection has a significantly positive effect on the SME implementation levels of CSR activities—the estimates for the coefficient of Stakeprotect are all approximately 0.500, and public trust has a significantly negative impact on the SME implementation levels of CSR activities; the estimates of the coefficient of Ptrust fluctuate from −10.133 to −8.454. Although the estimates in Specification (9) are inconsistent with those in the other specifications, this difference can be ignored due to the poor performance of the fitness test for this specification—the p-value of the χ² test is 0.932. We retain this specification only to maintain the congruent relationship between Table 4 and Table 5.

Hypotheses 1 and 2B are supported by the empirical evidence. The degree of stakeholder protection derived from the local legal environment affects the CSR activities of enterprises because the stakeholders’ appeals for CSR activities and pressures on CSR-related activities need to be guaranteed by the local legal environment, which accordingly determines the degree of stakeholder protection.

Furthermore, the opportunism of the enterprise’s managers is refrained from and relieved significantly, providing the key to overcoming the prisoner’s dilemma of CSR activities. SMEs in China are characterized by a tight cash flow, narrow financial bases, and a higher uncertainty of future development prospects. Because of the relatively scarce external and internal resources of SMEs and the regionality of SME businesses, the aforesaid mechanism works well. In addition, when the local public trust increases, the subjective probability that individuals will consider the possibility of being cheated falls, and SMEs will then have the opportunity to break the undesirable equilibrium. The detrimental effects of the information asymmetry between the insiders and outsiders of an enterprise are thus reduced. Therefore, although CSR disclosure is a necessary signal sent by China’s SMEs to strive for financial resources and market recognition, SMEs no longer need to maintain a higher implementation level of CSR activities.

For the estimates of the coefficients of the control variables, considering the mixed results in existing studies [6,9,16,19,33,34,35,36,37,41], our estimated results in Table 4 and Table 5 largely agree with the literature.

Furthermore, we tested the moderating effect of public trust on the relationship between stakeholder protection and SME CSR activities. The estimates in Columns (1)–(5) of

Table 6. Testing results for the moderating effect of Ptrust on the relationship between Stakeprotect and CSR activities.

Variable	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)
Stakeprotect	0.042 *	0.039 *	0.038 *	0.058 *	0.043 *	0.520 *	0.574 *	0.510 *	0.580 *	0.559 *
	(0.023)	(0.021)	(0.022)	(0.031)	(0.025)	(0.310)	(0.311)	(0.307)	(0.310)	(0.309)
Ptrust	−0.218	−0.313	−0.222	−0.603	−0.212	−8.552 ***	−8.536 ***	−8.590 ***	−8.387 ***	−8.377 ***
	(0.392)	(0.388)	(0.395)	(0.409)	(0.391)	(2.862)	(2.975)	(2.984)	(2.913)	(2.870)
Stakeprotect × Ptrust	0.143	0.155	0.169	0.134	0.146	0.105	0.003	0.214	0.065	0.103
	(0.104)	(0.103)	(0.105)	(0.105)	(0.104)	(0.740)	(0.755)	(0.743)	(0.710)	(0.766)
Mpconnection	0.659 ***	0.063	0.345	0.543 **		4.387 ***	1.711	3.418 ***	4.410 ***
	(0.248)	(0.182)	(0.261)	(0.242)		(1.497)	(1.549)	(1.300)	(1.478)
Bpconnection	−0.937 ***		−0.500 *	−0.683 **		−3.155 ***		−2.622 **	−3.171 ***
	(0.274)		(0.301)	(0.273)		(1.123)		(1.106)	(1.214)
Boardscale	0.141		0.271	−0.057	0.112	7.327 *		7.193	8.673 **	7.920 **
	(0.719)		(0.730)	(0.702)	(0.722)	(4.008)		(4.531)	(4.145)	(3.994)
Indirector	−0.006		−0.004	−0.008	−0.006	0.065		0.086	0.094	0.083
	(0.015)		(0.015)	(0.015)	(0.015)	(0.099)		(0.111)	(0.096)	(0.096)
Largeshare	−0.009	−0.007	−0.004	−0.009	−0.008	−0.164 ***	−0.167 **	−0.171 ***	−0.152 ***	−0.173 ***
	(0.007)	(0.007)	(0.007)	(0.007)	(0.007)	(0.063)	(0.066)	(0.061)	(0.056)	(0.064)
Firmage	0.837 ***	0.818 ***	0.686 ***	1.004 ***	0.824 ***	0.130	−0.162	0.125	0.305	−0.044
	(0.242)	(0.235)	(0.237)	(0.246)	(0.237)	(1.647)	(1.567)	(1.489)	(1.553)	(1.572)
Roa	0.033 *	0.036 **		0.034 **	0.033 **	−0.285 *	−0.236		−0.297 **	−0.256 *
	(0.016)	(0.016)		(0.016)	(0.016)	(0.151)	(0.155)		(0.146)	(0.155)
Firmsize	1.060 ***	1.043 ***	1.063 ***	1.096 ***	1.045 ***	2.986 **	3.043 **	1.842 **	2.761 **	2.791 **
	(0.175)	(0.171)	(0.172)	(0.166)	(0.172)	(1.381)	(1.417)	(0.872)	(1.403)	(1.355)
Slackres	0.002	0.003		0.003	0.003	−0.026	−0.022		−0.029 *	−0.026
	(0.004)	(0.004)		(0.004)	(0.004)	(0.016)	(0.015)		(0.015)	(0.016)
Debtoasset	−0.002	0.0004		−0.002	−0.001	−0.074	−0.061		−0.073	−0.046
	(0.006)	(0.006)		(0.006)	(0.006)	(0.068)	(0.066)		(0.065)	(0.061)
Tobinq	0.011	0.016		0.002	0.010	0.640	0.526		0.676 *	0.652
	(0.064)	(0.064)		(0.061)	(0.064)	(0.408)	(0.411)		(0.351)	(0.434)
Fsales	−0.006	−0.007	−0.009 *		−0.007	0.027	0.031	0.027		0.026
	(0.004)	(0.004)	(0.005)		(0.004)	(0.033)	(0.032)	(0.033)		(0.034)
Gdppc	−0.967 **	−0.832 *	−0.851 *	−0.956 **	−0.988 **	1.125	0.756	1.231	0.490	1.438
	(0.460)	(0.446)	(0.458)	(0.451)	(0.457)	(2.072)	(2.247)	(2.346)	(2.072)	(2.132)
Hdummy	−0.207	−0.233	−0.169		−0.236	0.118	0.275	0.767		0.491
	(0.189)	(0.188)	(0.181)		(0.188)	(1.428)	(1.394)	(1.386)		(1.443)
Constant	−11.345 ***	−11.137 ***	−12.105 ***	−9.521 ***	−11.433 ***	51.110 ***	66.021 ***	58.025 ***	47.046 **	45.950 **
	(3.283)	(2.832)	(3.431)	(3.144)	(3.294)	(19.186)	(19.657)	(21.877)	(20.983)	(22.200)
Yeardummies	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
p-value of F test						0.000	0.000	0.000	0.000	0.000
p-value of χ2 test	0.000	0.000	0.000	0.000	0.000
Number of obs.	3642	3675	3787	3701	3642	415	416	420	422	415

Note: *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Corresponding clustered sandwich standard errors are displayed in parentheses in Specifications (1)–(5). In Specifications (6)–(10), corresponding 2D clustered standard errors are displayed in parentheses.

were obtained based on the random-effects Probit regressions whose dependent variable is Disclosure. The LR test results strongly reject the pooled panel Probit regressions, thereby justifying the use of random-effects panel Probit regressions. We also obtained similar results by using population-averaged panel Probit and Logit model estimators, as well as fixed-effects panel Logit model estimators. To save space, these results are not provided here, but they are available upon the reader’s request. The coefficients in Columns (6)–(10) of Table 5 were estimated by employing panel data model clustering in two dimensions simultaneously, where the dependent variable is Scores. When the Heckman selection model is applied, the main results still hold.

The interaction term Stakeprotect × Ptrust is included in our regressions to capture the potential moderating effect of public trust on the relationship between stakeholder protection and SME CSR activities. To eliminate the multi-collinearity among Stakeprotect, Ptrust, and their interaction terms, Stakeprotect and Ptrust are standardized before they are multiplied. Thus, Stakeprotect × Ptrust in our regressions does not suffer from multi-collinearity with Stakeprotect or Ptrust.

Table 6 shows that Hypothesis 3B is supported by our empirical evidence, which is inconsistent with Hypothesis 3A. The moderating effect of public trust should not be overestimated. When the public trust does not attain a certain threshold, its level will be low overall, and the effects of legal protection on SME CSR activities may have no significant difference with a relatively higher or lower level of local public trust. Because of the limited dataset size, it is very difficult for the threshold panel data model to obtain consistent estimates. Nevertheless, the inference derived from our argument is supported by the empirical results, which at least suggests that our argument cannot be rejected. Comparing the estimates of the coefficient of Stakeprotect in Table 6 with the corresponding estimates in Table 4 and Table 5, we find that the magnitude of these estimates in Table 6 all decrease by more than half, and their significance also drops substantially (Columns (1)–(5) in Table 6 are compared with Columns (1)–(5) in Table 4; and Columns (5)–(10) in Table 6 are compared to Columns (1)–(5) in Table 5. Notably, all estimates of the coefficient of Stakeprotect × Ptrust are insignificant in Table 6. The moderating effect of public trust on the relationship between the degree of stakeholder protection and SME CSR activities should not be expected in the context of China.

4.2. Robustness Tests

To test the robustness of our empirical results, we first used an alternative measure, blood donation per capita, to gauge public trust across provinces in China [59]. The “Blood Donation Law” indicates that blood donations are not compensated by money, which reflects the citizens’ civism. Thus, blood donations are an alternative indicator of public trust. Here, our empirical results still hold. Second, we included the lagged main explanatory variables instead of their simultaneous forms in the regressions, to partially reduce the potential for reverse causation. Here, the basic results also do not change.

Structural equation modeling was used to further conduct our robustness test against the possible influence of the potential endogeneity problem on our empirical results. In particular, we employed the structural component, called path analysis, of a structural equation model to conduct a further robustness test [60]. When the dependent variable is Disclosure, the endogenous variables are defined as Disclosure and Stakeprotect. This setup is based on the abovementioned empirical results in Table 4 and Table 6, which identify the significant effect of Stakeprotect rather than Ptrust, or the standardized interaction term of Stakeprotect and Ptrust, on SME CSR disclosures. Notably, the predictors for Stakeprotect are Ptrust, Gdppc, and Instrument. Instrument is defined as the number of mission primary school enrollments per one thousand inhabitants in 1919 across provinces in China. These data come from an investigative report released by the Continuation Committee in 1922. Fang and Zhao employed this indicator as an instrumental variable for institutions across the provinces in China [61].

Instrument reflects the historical degree of the influence of Western religious culture on different regions of China. Religious courses and other enlightening curricula such as law, politics, and science, courses taught in mission primary schools, helped to enhance civic law and moral consciousness, and thus they continue to affect the legal environment across the provinces in China, thereby providing a profound historical background. Additionally, on average, the regional levels of public trust and economic development play an important role in the regional legal environment. Finally, the predictors for Disclosure are Stakeprotect and all other control variables defined in this study.

On the other hand, when the dependent variable is Scores, the endogenous variables are defined as Scores, Stakeprotect, and Ptrust. The predictors for Scores are Stakeprotect, Ptrust, and all the control variables defined in this study; the predictors for Stakeprotect comprise Ptrust, Gdppc, and Instrument; and the predictors for Ptrust are Stakeprotect, Instrument, and Socialtrust.

Socialtrust is measured based on the China General Social Survey (CGSS) conducted jointly by the Survey Research Center at Hong Kong University and the Sociology Department of the People’s University of China (the website of the China General Social Survey (CGSS) is http://www.chinagss.org/). The indicator is constructed according to the responses to the question, “How trustworthy are the people in your city?” The respondents can select one of five choices: “highly untrustworthy”, “untrustworthy”, “neutral”, “trustworthy”, or “highly trustworthy”. Respondents assign values from 1 to 5, where 1 reflects “highly untrustworthy”, and 5 represents “highly trustworthy”. Socialtrust is the average of the scores by province, which, along with Stakeprotect and Instrument, are highly associated with Ptrust in this study and can thus be treated as the predictors for Ptrust. Indeed, we find that our empirical results are fairly robust against changes in the predictors for potentially endogenous explanatory variables in this study, as long as Instrument and Gdppc are included in the structural models.

Table 7. Robustness tests based on the structural equation models.

	Endogenous Variables: Disclosure and Stakeprotect
Structural Disclosure
Variable	(1)	(2)	Variable	(1)	(2)
Stakeprotect	0.104 *	0.060 **	Roa	0.101 ***	0.103 ***
	(0.062)	(0.029)		(0.015)	(0.015)
Ptrust		0.005	Firmsize	0.206 ***	0.200 ***
		(0.018)		(0.015)	(0.015)
Stakeprotect × Ptrust		0.020	Slackres	−0.013	−0.011
		(0.026)		(0.013)	(0.013)
Mpconnection	0.071 ***	0.071 ***	Debtoasset	0.021	0.022
	(0.019)	(0.019)		(0.016)	(0.016)
Bpconnection	−0.073 ***	−0.072 ***	Tobinq	0.019	0.018
	(0.019)	(0.019)		(0.014)	(0.015)
Boardscale	0.007	0.008	Fsales	0.007	0.007
	(0.019)	(0.019)		(0.013)	(0.013)
Indirector	0.002	0.003	Gdppc	0.054	0.043
	(0.019)	(0.019)		(0.052)	(0.040)
Largeshare	−0.031 **	−0.029 **	Hdummy	−0.008	−0.007
	(0.013)	(0.013)		(0.013)	(0.013)
Firmage	0.114 ***	0.111 ***	Constant	−2.164 ***	−2.203 ***
	(0.013)	(0.013)		(0.342)	(0.438)
Structural Stakeprotect
Variable	(1)	(2)	Variable	(1)	(2)
Instrument	0.026 ***	0.019 **	Gdppc	0.619 ***	0.616 ***
	(0.009)	(0.008)		(0.009)	(0.009)
Ptrust	0.302 ***	0.303 ***	Constant	−5.328 ***	−5.340 ***
	(0.009)	(0.010)		(0.145)	(0.145)
p-value of χ2 test				0.000	0.000
Number of obs.				6064	6064
	Endogenous variables: Scores, Stakeprotect, and Ptrust
Structural Scores
Variable	(1)	(2)	Variable	(1)	(2)
Stakeprotect	0.246 ***	0.252 **	Roa	−0.151 **	−0.151 **
	(0.083)	(0.104)		(0.068)	(0.069)
Ptrust	−0.332 ***	−0.332 ***	Firmsize	0.342 ***	0.342 ***
	(0.074)	(0.074)		(0.071)	(0.072)
Stakeprotect × Ptrust		−0.007	Slackres	−0.054	−0.055
		(0.066)		(0.049)	(0.049)
Mpconnection	0.175 *	0.175 *	Debtoasset	−0.170 **	−0.171 **
	(0.091)	(0.091)		(0.078)	(0.079)
Bpconnection	−0.141	−0.140	Tobinq	0.001	0.001
	(0.087)	(0.087)		(0.062)	(0.062)
Boardscale	0.065	0.065	Fsales	0.097 *	0.097 *
	(0.079)	(0.079)		(0.051)	(0.051)
Indirector	−0.023	−0.023	Gdppc	0.088	0.083
	(0.079)	(0.079)		(0.079)	(0.096)
Largeshare	−0.209 ***	−0.208 ***	Hdummy	0.043	0.043
	(0.050)	(0.051)		(0.057)	(0.057)
Firmage	0.003	0.004	Constant	6.977 ***	6.998 ***
	(0.049)	(0.049)		(1.896)	(1.907)
Structural Stakeprotect
Variable	(1)	(2)	Variable	(1)	(2)
Instrument	0.356 ***	0.357 ***	Gdppc	0.525 ***	0.524 ***
	(0.035)	(0.035)		(0.066)	(0.066)
Ptrust	0.449 ***	0.449 ***	Constant	−8.285 ***	−8.285 ***
	(0.098)	(0.098)		(1.486)	(1.486)
Structural Ptrust
Variable	(1)	(2)	Variable	(1)	(2)
Instrument	0.493 ***	0.493 ***	Socialtrust	0.182 ***	0.181 ***
	(0.027)	(0.027)		(0.028)	(0.028)
Stakeprotect	0.824 ***	0.824 ***	Constant	12.039 ***	12.039 ***
	(0.032)	(0.032)		(0.834)	(0.835)
p-value of χ2 test				0.013	0.000
Number of obs.				332	332

Note: *, **, and *** denote statistical significance at 10%, 5%, and 1% levels, respectively. Corresponding robust standard errors are displayed in parentheses.

presents the empirical results from the structural equation modeling. The estimated standardized coefficients for our main explanatory variables show that Stakeprotect has a significantly positive effect on SME CSR disclosures, while both Ptrust and Stakeprotect × Ptrust have no significant impacts on SME CSR disclosures for the endogeneity problem. Additionally, Stakeprotect and Ptrust have a significantly positive and negative impact, respectively, on the level of SME CSR activities. However, the interaction term Stakeprotect × Ptrust is not significantly associated with Scores. Furthermore, the comparative fit index is greater than 0.95, and the root mean squared error of approximation is less than 0.08 when the interaction term is excluded from the structural equation model.

Therefore, we find that, even when the complicated relationships (including potential causality, endogeneity, or hierarchy) among our dependent variable and main explanatory variables are considered, the empirical results in this study are still maintained. Hypotheses 1, 2A, 2B, and 3B are also supported by our evidence. The significance and magnitude of the estimates of other control variables experience slight changes in this case. Signs of these estimates, however, are almost unchanged compared to the results in Table 5 and Table 6.

5. Conclusions

CSR has become increasingly more important for the management practices of enterprises in emerging economies. Apart from the many studies on CSR in developed countries, research on the CSR in emerging economies, particularly on the CSR of Small and Medium Enterprises (SMEs) in large emerging economies, such as China’s, remains relatively scarce. In this study, we investigated the effects of stakeholder protection and public trust on the CSR activities of listed enterprises on the Chinese SME Board. Our study makes an important contribution by broadening the scope of research on the determinants of SME CSR activities in emerging market economies. The main findings highlight the aspects of local legal protection for stakeholders and public trust embedded in the local environment for studies on SME CSR activities.

Our results suggest that the degree of stakeholder protection derived from the local legal environment has a significantly positive impact on SME CSR activities. Public trust is not significantly associated with SME CSR disclosure, and public trust has a significantly negative impact on the SME implementation levels of CSR activities. Furthermore, the moderating effect of public trust on the relationship between the degree of stakeholder protection and SME CSR activities is not supported by our empirical study.

Our results have significant implications for the literature and policy practices for the regulation of SME CSR activities. In emerging and transitional economies, the improvement and development of the legal system is one of the most important factors for promoting SME CSR activities. In the context of social and economic regime changes, the degree of stakeholder protection, rather than the level of local public trust, is closely associated with regional moral levels and should be fully assessed by the authorities for policy implementations to improve SME CSR activities.

Finally, our study only focuses on the CSR activities of enterprises listed on the SME Board in China. When the data limitation could be broken through, we should further pay close attention to the CSR activities of numerous unlisted SMEs to obtain a more general research conclusion. Additionally, Valeri and Baggio pointed out the potentially promising application of social network analysis and blockchain technology in studying the corporate behavior [62,63,64]. Thus, our work could be extended in future studies by examining the empirical impacts of social network effect and new technology applications on SMEs’ CSR activities, and investigating the interaction effects of these new influence factors on the relationship between the degree of stakeholder protection, or public trust, and SME CSR activities.