Quality of Mandatory Social Responsibility Disclosure and Total Factor Productivity of Enterprises: Evidence from Chinese Listed Companies

Abstract

This study attempts to determine whether improving the quality of mandatory corporate social responsibility (CSR) information disclosure can have a positive impact on a firm’s development. To this end, an empirical analysis is conducted to establish a relationship between the quality of mandatory CSR information disclosure and a firm’s total factor productivity (TFP), using data from A-share listed companies in China over the period 2009–2020. The results show that: (1) Improving the quality of mandatory CSR disclosure leads to a significant increase in a firm’s TFP. (2) By alleviating a firm’s financing constraints and agency costs, the improved quality of mandatory CSR disclosure effectively enhances the firm’s investment efficiency and innovation capacity, leading to higher TFP. This research extends the influence of CSR disclosure on the economic development outcomes of Chinese firms, and provides theoretical guidance for the development and improvement of CSR disclosure mechanisms in emerging economies.

1. Introduction

China’s remarkable economic growth over the past four decades has been achieved by exploiting abundant resources and cheap inputs. However, this development model, which relies heavily on investment in production factors, has led to excessive resource consumption and compromised the economy’s potential for further development. Despite China’s per capita income reaching nearly US $12,400, akin to that of high-income economies, the country faces significant downward pressure on its economy due to issues such as imbalanced and insufficient economic development resulting from previous haphazard economic growth. It has, therefore, become a pressing necessity for China to enhance the quality of its overall economic development to overcome the current economic predicament. To this end, the reports of the 19th and 20th National Congresses of the Communist Party of China both proposed improving total factor productivity (TFP) and promoting changes in the quality, efficiency and dynamism of economic development. In this context, enterprises, as micro-units of economic activity, have undoubtedly emerged as a key strategy in China’s aspirations to get rid of the middle-income trap and join the ranks of developed economies by boosting their TFP to drive economic development.

Concurrently, as the concept of sustainable development has become more widely accepted around the world, demands to conserve natural resources, safeguard social welfare and improve the quality of economic growth have increased sharply in recent years. As primary agents of social production and services, companies are now under increased scrutiny with regard to their corporate social responsibility (CSR) obligations, making the disclosure of CSR information a universally recognized mandate. In order to promote the relatively nascent CSR process in China and improve the quality of corporate development, the Chinese government issued the Notice on the Annual Reports of Listed Companies in 2008 (the Notice, hereinafter) at the end of 2008. The Notice required companies in the Corporate Governance Sector, overseas-listed companies, financial companies and companies belonging to the SZSE 100 Index to publish CSR reports alongside their annual reports from 2008. Unlike developed economies such as Europe and the US, where voluntary disclosure of CSR information prevails, this policy in China not only mandates disclosure but also prescribes specific disclosure standards and content and covers more than 20% of the listed Chinese companies. Therefore, as Chen et al. [1] argued, the implementation of the Notice can be seen as the formal beginning of CSR disclosure in China.

Nevertheless, it remains uncertain whether mandatory CSR disclosure can substantially enhance the overall quality of China’s economic development. More specifically, the question arises as to whether higher-quality mandated CSR disclosure leads to a positive impact on a firm’s TFP. Relevantly, several scholars have constructed quasi-natural experiments by exploiting exogenous shocks generated by the publication of the Notice and have investigated the economic outcomes of mandatory disclosure of CSR information by applying the difference-in-difference approach [1,2,3,4]. For instance, a few scholars have discovered that mandatory CSR disclosure may escalate costs unrelated to business operations, potentially hindering normal business activities of firms and damaging their economic performance [1]. In contrast, others have argued that the Notice-driven improvement in CSR information disclosure quality can significantly boost corporate financing efficiency, effectively minimize principal–agent conflicts, and thereby enhance a firm’s innovation capacity and investment efficiency [2,3,4]. As a result, the association of the quality of mandatory CSR disclosure with the TFP of firms has not been directly examined in the literature. Based on an extension of existing research findings, we intuitively expect that the effects of mandatory CSR disclosure quality on the business performance, innovative R&D activities and investment efficiency of firms will be further transmitted to their TFP. Additionally, the improvement in TFP as an important indicator of the quality of corporate development is in line with the Chinese government’s interest in encouraging companies to disclose CSR information, both of which serve to promote the quality of economic development. Therefore, based on the current academic research, we further explore the influence of the mandatory CSR disclosure quality on a firm’s TFP. Our analysis is intended to provide a clearer understanding of the relationship between CSR disclosure and TFP, which has significant practical implications for the development and transformation of China, the second largest economy in the world.

The contribution of this paper to the existing body of literature is in the following three respects. Firstly, prior studies on mandatory CSR disclosure have predominantly focused on the short-term policy impact resulting from the release of the 2008 Notice, leading to an unrepresentative sample selection. Instead, this study utilizes data accumulated over the last decade that encompass a larger and more representative sample of companies, providing insights into the long-term effects of mandatory CSR disclosure on firms. Secondly, this study constructs indicators that better quantify the quality of information disclosure, allowing for a more accurate quantification of the influence of CSR information disclosure on a firm’s TFP. Thirdly, when it comes to research on CSR information disclosure, scholars generally consider principal–agent conflicts and external financing constraints to be the primary mechanisms through which CSR information disclosure affects a firm’s economic activities [3,4,5]. However, this paper not only verifies the role of those mechanisms but also examines the influence of two key factors, namely corporate innovation capability and investment efficiency, in the relationship between the quality of mandatory CSR disclosure and a firm’s TFP, thereby expanding the pathways through which CSR disclosure might influence company economic growth.

2. Literature Review, Theoretical Analysis and Hypothesis Development

2.1. Determinants of Firm’s TFP

In the field of corporate finance research, financing constraints and agency conflicts are often considered to be two important influences that hinder a firm’s R&D and innovation activities and capital allocation efficiency [3,4]. In turn, research has shown that a firm’s innovation capabilities and capital allocation efficiency are precisely the driving sources that allow for sustained progress in their productive efficiency [6]. Thus, a large body of empirical evidence suggests that alleviating a firm’s financing constraints and principal–agent conflict problems is an effective means of promoting the firm’s TFP [7]. On the one hand, research has shown that reducing financing constraints can increase the effectiveness of a firm’s business strategies, promote their investment activities in technological change and ultimately contribute to improvements in their productivity. For example, R&D and innovation activities are often characterized by high risks and long payback periods, which require adequate financing for the long-term continuity of innovative projects. Thus, it has been suggested that alleviating financing constraints can increase a firm’s willingness to take risks in R&D and innovation activities [4], which is highly beneficial for enhancing the firm’s innovation capabilities and TFP [7]. On the other hand, there is also empirical evidence that reducing financing constraints can increase the efficiency of a firm’s resource allocation, contributing to improvements in the firm’s TFP [8]. For example, Caggese and Cunat [9] show that well-financed firms are more likely to take advantage of fleeting investment opportunities than firms with financing constraints, ensuring the timeliness and effectiveness of their investment decisions. Moreover, mitigating financing constraints allows firms to achieve higher levels of factor productivity by expanding the scale of production [10], thereby unlocking higher productivity. Consequently, easing a firm’s financing constraints is undoubtedly an important means of increasing their TFP.

Moreover, in practice, the principal–agent conflict problem is widely recognized as the most prevalent friction between firms and external investors. This problem not only causes firms to deviate from their optimal resource allocation efficiency [11], but can also discourage R&D and innovation activities [4], leading to a decrease in a firm’s TFP [7]. For example, Biddle et al. [11] conducted a study which showed that the problem of agency conflict may cause firms to miss out on numerous projects with positive net present values (NPVs), thereby reducing their investment efficiency. Moreover, it has been found that firms with high information asymmetries and principal–agent conflicts may prioritize managers’ risk-averse behaviors at the costs of shareholders’ interests [12], thereby hindering R&D and innovation activities for projects with inherently high-risk characteristics, subsequently reducing the firm’s innovation capability and productivity. Consequently, mitigating principal–agent conflict problems in firms is considered an effective means of improving their TFP.

2.2. Quality of Mandatory CSR Disclosure and TFP of Enterprises

In the field of CSR disclosure research, CSR reports are widely regarded as an important mechanism for external stakeholders to obtain information about a firm [2]. Several studies have highlighted that, as an effective supplement to financial data, CSR information can be used by investors to verify the authenticity of a company’s accounting information [2,13]. In addition, other information included in CSR reports, such as environmental emissions and charitable donations, can demonstrate a firm’s social responsibility performance [1,14]. As a result, some scholars have argued that disclosing high-quality CSR information can help firms reduce information asymmetry with investors and mitigate issues related to principal–agent conflicts [15]. Moreover, effectively communicating a firm’s commitment to social responsibility to external stakeholders through CSR reports can cultivate social trust capital and enhance the firm’s reputation [4]. For example, based on the principal–agent theory and information asymmetry theory, Liu and Tian [3] conducted a study showing that the mandatory disclosure of CSR information forces firms to interact with external stakeholders, which means that it is harder for managers to hide unfavorable information about their company. Consequently, it can effectively alleviate the problem of agency conflicts in firms. In addition, research has shown that moral hazards arising from the information asymmetries of borrowing firms and the adverse selection problems they create are the main factors affecting banks’ credit risk [16]. Therefore, based on the signaling theory, high-quality CSR disclosure has also been found to mitigate banks’ risk concerns about firms, thereby enhancing a firm’s ability to access external credit financing [5]. For example, Goss and Roberts [17] conducted a study that showed that high-quality CSR disclosure improves the debt relationship between firms and banks. As a result, firms can access credit facilities from banks at more favorable interest rates. Similarly, Hu et al. [18] argued that by disclosing comprehensive CSR information, firms are able to send a friendly and environmentally conscious signal to investors and the public. This contributes to broadening a firm’s external financing channels and reducing their financing costs.

On the other hand, research has shown that increasing short-term operating costs due to the pursuit of high-quality CSR disclosure could impede a firm’s operating and share price performance [1,15]. This may create disincentives for the firm’s share price and economic performance. Nonetheless, from a sustainability perspective, TFP is a better indicator of a firm’s technological progress, sustainability prospects, and long-term value than its short-term economic performance and share price performance. Consequently, TFP is anticipated to possess stronger synergies with a firm’s innovation capacity and capital allocation efficiency over the long term. In summary, we predict that a higher quality of mandatory CSR disclosure will have a beneficial effect on a firm’s external financing constraints and agency conflicts, as well as a driving effect on firms’ innovation capability and capital allocation efficiency, thus significantly increasing the firm’s TFP (see Figure 1). Based on the above logical analysis, this paper proposes Hypothesis 1:

Hypothesis 1 (H1).

A company’s TFP can be significantly increased by improving the quality of mandatory CSR disclosures.

3. Research Design

3.1. Data Sources

The research sample in this article is derived from non-financial A-share companies that were listed on the Shenzhen Stock Exchange and the Shanghai Stock Exchange from 2009 to 2020. On the basis of previous studies [1,19], we have a selection of companies that are also in line with the requirements below: (1) Excluding enterprises with abnormal financial conditions such as ST or *ST from the sample; (2) Excluding enterprises with missing key variables from the sample; (3) In order to separately examine the impact of the quality of mandatory CSR disclosure, in this study we also exclude the sample of firms that do not disclose CSR information and the sample of firms that voluntarily disclose CSR information; (4) Winsorize all continuous variables by 1%. After the above treatments, the final sample of this paper comprises 12,908 firm–year observations. The data for CSR disclosures were taken from the companies’ annual reports and stand-alone CSR reports available in the RSK or Hexun databases, while some corporate data that cannot be found in the database were collected manually through the company’s annual report and CSR report. Other firm-level data comes from CSMAR.

3.2. Variable Design

1.

Measurement of TFP of firms

In our study, TFP of firms is used as an explanatory variable. Most of the commonly used measures of TFP of firms are constructed based on the Cobb–Douglas production function:

$$Y_{it}=A_{it}{L}_{it}^{\alpha }{K}_{it}^{\beta }$$

(1)

In Equation (1), $Y_{it}$ represents output, and $L_{it}^{\alpha }$ and $K_{it}^{\beta }$ represent labor and capital inputs, respectively. $A_{it}$ is commonly known as TFP, which increases the marginal output of all factors simultaneously. By logarithmizing Equation (1), it can be transformed into the following linear form:

$$y_{it}=\alpha l_{i,t}+\beta k_{i,t}+{\mu }_{i,t}$$

(2)

In Equation (2), $y_{it}$, $l_{i,t}$ and $k_{i,t}$ represent the logarithmic forms of $Y_{it}$, $L_{i,t}$ and $K_{i,t}$, respectively, while the residual term ${\mu }_{i,t}$ contains information on the logarithmic form of the firm’s TFP ($A_{it}$). An OLS estimation of Equation (2) is sufficient to obtain an estimate of TFP. However, in the actual production process, a portion of ${\mu }_{i,t}$ is usually observable in the current period, which makes the OLS estimates biased. To solve this problem, the residual term ${\mu }_{i,t}$ in Equation (2) can be split in the following form, which then gives Equation (3):

$$y_{it}=\alpha l_{i,t}+\beta k_{i,t}+{\omega }_{i,t}+{\epsilon }_{i,t}$$

(3)

In Equation (3), ${\omega }_{i,t}$ is the part of ${\mu }_{i,t}$ that can be observed and affects the firm’s factor selection in the current period. ${\epsilon }_{i,t}$ is the true residual term, which contains unobservable technology shocks and measurement errors. Based on the OLS method, the fixed-effects approach to measuring TFP is based on Equation (3) by introducing individual firm dummy variables, thus solving the endogeneity problem due to the presence of ${\omega }_{i,t}$ and thus obtaining a consistent and unbiased estimate of the production function.

However, the methods described above still suffer from insurmountable problems, for example, the inability to identify information arising from changes in ${\omega }_{i,t}$ over time. To address this problem, Olley and Pakes [20] developed a consistent semi-parametric-based approach to estimating a firm’s TFP, which assumes that firms will make investment decisions based on current productivity conditions and therefore uses the firms’ current investments as a proxy variable for unobservable productivity shocks, thus addressing the simultaneity bias and selectivity bias of OLS. However, the OP method has a strict restriction in its use, that is, it guarantees that a firm’s investment is strictly greater than zero, which undoubtedly excludes those firms whose investment is less than or equal to zero, making the sample of firms in the study smaller. Therefore, based on the OP method, Levinsohn and Petrin [21] proposed using intermediate inputs as a proxy for investment, which overcomes the restriction of the OP method that firms must have positive investment, minimizes the loss of sample size, and has a lower adjustment cost for intermediate inputs, allowing the researcher to flexibly adjust the choice of proxies according to the characteristics of the data.

Since the fixed-effects method and the LP method are improvements of the OLS and OP methods, respectively, TFPs measured by the fixed-effects method and the LP method were chosen as the explanatory variables in the baseline regressions of this paper. In this paper, the financial data of firms required to calculate TFP were obtained from the CSMAR database.

2.

The quality of mandatory CSR disclosures

The core explanatory variable in this paper is the quality of mandatory CSR disclosure, so constructing and quantifying the quality of CSR disclosure is the focus of our study. When conducting the evaluation of CSR disclosure quality, content analysis and index methods are acknowledged as relatively more reasonable methods and have been applied more often by scholars in the field of CSR disclosure [2]. With reference to existing studies [2,22], we first quantitatively assess the quality of CSR disclosure by means of a content analysis. In this paper, the construction of indicators of the quality of mandatory CSR disclosure relies on the classification of CSR research data in the CSMAR database. Specifically, the CSMAR database records CSR information in ten categories, including information on the company’s relationship with upstream and downstream suppliers, its relationship with customers, shareholders, employees and creditors, as well as information on environmental protection, philanthropy, working conditions and deficiency in CSR performance. Following the research by Wang et al. [2], we define each category as a dummy variable based on whether the CSR information reported by the enterprise contains information in that category, and if the CSR information disclosed includes information on a category, then the corresponding dummy variable is equal to 1; otherwise, it is equal to 0. Finally, the sum of these ten dummy variables is used to quantify the quality of mandatory CSR disclosure (MCSRscore). In addition, third-party authorities on CSR disclosure quality ratings include the RKS and the Hexun, of which the RKS provides statistics only on firms voluntarily reporting CSR information and therefore does not meet the requirements of this paper for the sample of firms. Hexun has established an index-based scoring scale of 30%, 15%, 15%, 20% and 20% weighting of CSR information in five categories: employees, suppliers, environment, shareholders and social contribution, which can accurately and objectively measure the quality of CSR information disclosure. Therefore, following the existing literature [23], we also adopt the Hexun CSR disclosure index (HXMCSRscore), constructed based on the index method, as a substitute explanatory variable to be used for robustness testing.

3.

Innovation capacity and investment efficiency

A firm’s innovation capacity and investment efficiency are the key mediating variables used to validate the impact channel. In particular, we select a firm’s investment in R&D as the proxy variable of a firm’s innovation capability. This is because, compared to indicators such as the number of innovation patents of a firm, a firm’s R&D investment is not only an effective measure of a firm’s investment in intangible assets, but also an immediate response to a firm’s willingness to innovate [5]. In contrast, the output of a firm’s innovation patents requires a long period of research and development, which is usually difficult to estimate accurately. Second, referring to Chen et al. [24], this paper measures the efficiency of a firm’s capital allocation in terms of its investment efficiency; specifically, investment efficiency is calculated by constructing the following model:

$${Invest}_{it}={\beta }_0+{\beta }_1{NEG}_{it-1}+{\beta }_2{Growth}_{it-1}+{\beta }_{3}NEG\ast {Growth}_{it-1}+{\epsilon }_{i,t}$$

(4)

In Equation (4), the dependent variable ${Invest}_{it}$ is equal to the sum of the firm’s cash outlay for fixed assets, intangible assets and other long-term assets, minus the cash proceeds from the sale of assets, divided by the firm’s total assets at the beginning of the year; ${NEG}_{it-1}$ is a dummy variable that takes the value of 1 if the firm’s sales revenue growth rate is less than 0 in year t−1, and 0 otherwise; ${Growth}_{it-1}$ is the growth rate of the enterprise’s revenue from year t−1 to year t. Subsequently, Equation (4) is estimated by year and by industry, and the values of a firm’s observations for each industry in each year of the regression are required to be greater than 20. Finally, we use the absolute value of the estimated residuals to measure the firm’s investment efficiency (InvEff), where the closer the value is to zero, the more efficient the firm’s investment.

4.

The agency costs and financing constraints of firms

In the section on mechanism testing, a firm’s agency costs are measures of the firm’s management expense ratio (Mfee) and management plus selling expense ratio (MSfee) [5], where the management expense ratio (Mfee) is the ratio of management expenses to operating income, and the management plus selling expense ratio (MSfee) is the ratio of the sum of management plus selling expenses to operating income. A firm’s financing constraint is measured by the firm’s cost of debt (Cost) and the KZ index (KZ). The cost of debt is the ratio of a company’s interest expense to the company’s debt. The KZ index (KZ) is a common measure of financing constraints of a firm, which can be calculated from financial information including operating cash flow, cash holdings, cash payout ratio, leverage and Tobin’s Q [25]. In particular, a higher value of the KZ index is an indication of a higher degree of the financing constraint that the company is facing.

5.

Control variables

Referring to existing studies related to the TFP of enterprises [8,9,25], the control variables in this paper are selected as follows: firm size (Size), revenue growth (Growth), years listed (Age), cash holdings (Cash), leverage (Lev), profitability (Roa), percentage of shares held by major shareholders (Top1) and Tobin’s Q (TQ). Studies have suggested that larger firms have more resources at their disposal, thus strongly supporting activities such as innovation and investment, which are more conducive to a firm’s TFP [8]. Revenue growth is seen as the main driver of a company’s long-term sustainability and reflects, to some extent, the company’s growth opportunities and ability to compete in the industry in which it operates [26]. Debt ratios (Lev) are a double-edged sword for firms: Moderate corporate leverage facilitates expansion and reproduction, allowing firms to match their higher levels of factor productivity with a larger scale of production [10]. However, excessive leverage not only increases agency costs and reduces the efficiency of corporate investment, but also leads to financial distress. In addition, companies with better profitability tend to be more competitive, so their drive for productivity is stronger. The cash holdings, which is the ratio of a company’s liquid assets to its total assets, largely reflects the stability of a company’s operations and is an important indicator that cannot be ignored in corporate finance research. The age of a firm is usually positively correlated with its sense of innovation [27], which may be beneficial to the total factor productivity of a firm. Finally, majority shareholder ownership (Top1) and Tobin’s Q are related to a firm’s managerial efficiency and the replacement cost of firm assets, respectively, which may affect firm productivity to some extent [25].

Table 1. Description of variables used in this study.

Variable	Symbol	Definition
TFP (LP method)	TFP_LP	TFP of enterprises calculated by the LP method
TFP (Fixed effect method)	TFP_FE	TFP of enterprises calculated by the fixed-effects method
Mandatory CSR Disclosure Quality	MCSRscore	The scores for mandatory disclosure of CSR information calculated according to the CSMAR database.
Ratio of management expenses	Mfee	Management expense divided by operating income
Management and selling expense ratio	MSfee	The sum of management and sales expenses divided by operating income
Cost of debt	Cost	Interest expense divided by total liabilities
KZ index	KZ	Financing constraints of enterprises
Innovation capability	RD	The logarithm of enterprise R&D investment
Investment efficiency	InvEff	Absolute value of inefficient investment of enterprises
Enterprise size	Size	The logarithm of total assets.
Age of listing	Age	The logarithm of the company’s listing years
Profitability	Roa	Net income divided by total assets
Growth	Growth	Operating profit annual growth rate
Cash holdings	Cash	The ratio of cash holdings to total assets
Tobin’s Q	TQ	Ratio of company’s market value to replacement cost of capital
Asset–liability ratio	Lev	The ratio of total liabilities to total assets
Percentage of shares held by major shareholders	Top1	The fraction of shares held by the largest shareholders

shows the symbols for all variables and their definitions.

3.3. Model Design

To explore the impact of the quality of mandatory CSR disclosure on a firm’s TFP, this paper performs a regression test by establishing Equation (5). Since there may be a time lag in the impact of mandatory CSR disclosure quality on the TFP of enterprises, in Equation (5), we choose TFP in year t + 1 as the dependent variable.

$${TFP}_{it+1}={\beta }_0+{\beta }_1{MCSRscore}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it+1}$$

(5)

In addition, in the mechanism test section, we construct regression Equations (6)–(9) to investigate the association between the mandatory CSR disclosure quality and a firm’s agency costs and financing constraints. Specifically, in Equations (6) and (7), we take the firm’s management fee rate (Mfee) and management plus sales fee rate (MSfee) as proxy variables for the firm’s agency costs, respectively. In Equations (8) and (9), we adopt the firm’s interest cost (Cost) and KZ index, respectively, as proxies for the firm’s financing constraints.

$${Mfee}_{it}={\beta }_0+{\beta }_1{CSRscore}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it}$$

(6)

$${MSfee}_{it}={\beta }_0+{\beta }_1{CSRscore}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it}$$

(7)

$${Cost}_{it}={\beta }_0+{\beta }_1{CSRscore}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it}$$

(8)

$${KZ}_{it}={\beta }_0+{\beta }_1{CSRscore}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it}$$

(9)

To further test whether a firm’s ability to innovate and investment efficiency are the key pathways by which the quality of mandatory CSR disclosure affects their TFP, we construct a mediating-effect model using corporate innovation capability (RD) and investment efficiency (InvEff) as mediating variables, respectively, as shown in Equations (10)–(13):

$${RD}_{it}={\beta }_0+{\beta }_1{MCSRscore}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it}$$

(10)

$${TFP}_{it+1}={\beta }_0+{\beta }_1{MCSRscore}_{it}+{\beta }_2{RD}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it}$$

(11)

$${InvEff}_{it}={\beta }_0+{\beta }_1{MCSRscore}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it}$$

(12)

$${TFP}_{it+1}={\beta }_0+{\beta }_1{MCSRscore}_{it}+{\beta }_2{InvEff}_{it}+\sum {\beta }_i{Controls}_{it}+\alpha +\gamma +{\epsilon }_{it}$$

(13)

In Equations (5)–(13), β is the estimated coefficient of interest in this paper. $\alpha$ and $\gamma$ represent a dummy industrial variable and the virtual time variable to control for fixed industry effects and fixed year effects, respectively. Control is a set of firm-level control variables related to a firm’s TFP. ε is a random-disturbance term.

4. Test Results

4.1. Statistical Description Results

The descriptive statistics of all variables are presented in

Table 2. Statistical description of all variables.

Variable	Obs	Mean	Min	Median	Max	SD
TFP_LP	12,908	9.084	5.830	9.010	13.375	0.966
TFP_FE	12,908	11.523	7.896	11.429	15.770	1.140
MCSRscore	12,908	5.886	1.000	7.000	10.000	2.328
Mfee	12,908	0.088	0.009	0.073	0.535	0.066
MSfee	12,908	0.165	0.015	0.130	0.843	0.125
Cost	12,908	0.059	−0.031	0.052	0.541	0.048
KZ	12,908	0.949	−10.786	1.189	11.201	2.037
RD	12,908	17.762	0.000	17.908	23.647	1.611
InvEff	12,908	0.040	0.000	0.030	0.396	0.043
Size	12,908	22.148	19.305	22.043	25.936	1.047
Age	12,908	1.458	0.000	1.609	2.996	0.938
Roa	12,908	0.032	−0.247	0.034	0.229	0.068
Growth	12,908	0.160	−0.580	0.099	2.412	0.395
Lev	12,908	0.433	0.057	0.424	0.892	0.190
Top1	12,908	0.326	0.091	0.305	0.752	0.141
Cash	12,908	0.136	0.001	0.110	0.798	0.101
TQ	12,908	1.989	0.875	1.625	8.136	1.168

. The means and standard deviations for TFP_LP (TFP_FE) are 9.084 (11.523) and 0.966 (1.140), respectively, indicating that TFP varies to some extent between the firms in our sample, but within a reasonable range of values. The mean (median) value of the independent variable mandatory CSR disclosure quality (MCSRscore) is 5.886 (7.000), the maximum value is 10.000, the minimum value is 1.000 and the standard deviation is 2.328, which indicates that the quality of CSR disclosure varies widely among the sample companies and that the overall CSR disclosure quality of the sample is poor and still has considerable room for improvement. The mean value of management fee rate (Mfee) and management and sales fee rate (MSfee) is 0.088 (0.165) with a standard deviation of 0.066 (0.125), while the mean of the interest expense (Cost) and KZ index (KZ) is 0.059 (0.949) with a standard deviation of 0.048 (2.037), which indicates considerable variation in agency costs and financing constraints across firms. The mean value of the firms’ R&D expenditure (RD) is 17.762 with a standard deviation of 1611, and the mean value of the firms’ investment efficiency (InvEff) is 0.040 with a standard deviation of 0.043, both of which are within a reasonable range of values.

The range of values for the control variables does not differ significantly from the existing literature [5]. In particular, the mean and median values for size are 22.148 and 22.043, respectively, indicating that more than half of the sample are medium and large enterprises. The mean of Roa is 0.032, which suggests that the return on assets of the sample firms is around 3%. The mean and median values of the Top1 variable are 0.326 and 0.305, respectively, which implies that most firms have a high proportion of large shareholders. The mean and the median of the Lev are, respectively, 0.433 and 0.424, which means that most companies have total debt of around 40% of their total assets. The mean and median of Growth are 0.160 and 0.099, which indicates that the sample companies have good growth. Tobin’s Q has a mean and median of 1.989 and 1.625, indicating that most companies have a market value above the replacement cost of their capital. The control variables selected in this paper are within a reasonable range of values.

4.2. Correlation Analysis

Before carrying out the basic regression analysis, it is first necessary to test the correlation between the variables in order to eliminate the impact of multicollinearity among the variables on the results of this study.

Table 3. The correlation matrix.

Variable	TFP_LP	TFP_FE	MCSRscore	Size	Age	Roa	InvEff	Growth	Lev	Top1	TQ
TFP_LP	1
TFP_FE	0.972 ***	1
MCSRscore	0.058 ***	0.068 ***	1
Size	0.787 ***	0.863 ***	0.044 ***	1
Age	−0.078 ***	−0.080 ***	0.074 ***	−0.071 ***	1
Roa	0.124 ***	0.112 ***	0.064 ***	0.020 **	−0.008	1
Growth	0.134 ***	0.115 ***	−0.019 **	0.057 ***	−0.003	0.295 ***	1
Lev	0.424 ***	0.448 ***	−0.041	0.448 ***	−0.139 ***	−0.357 ***	−0.013	1
Top1	0.168 ***	0.193 ***	0.012	0.169 ***	−0.104 ***	0.138 ***	0.020 **	0.052 ***	1
Cash	−0.084 ***	−0.137 ***	0.001	−0.165 ***	−0.017 *	0.250 ***	0.036 ***	−0.345 ***	0.052 ***	1
TQ	−0.328 ***	−0.366 ***	0.004	−0.415 ***	0.048 ***	0.159 ***	0.030 ***	−0.260 ***	−0.056 ***	0.195 ***	1

***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively.

shows the Pearson correlation coefficients between the variables in the base regression. It should be noted that the correlation coefficients between the MCSRscore and the dependent variables (TFP_LP and TFP_FE) appear to be significant at the 1% level, which means that the MCSRscore is significantly correlated with the TFP of firms. Meanwhile, the coefficients of correlation of each control variable with TFP obtained are also significant at the 10% level, which suggests that the selected control variables in this paper are reasonable. In addition, we can see that the absolute values of the coefficients of correlation among the control variables are less than 0.5, with the exception of the coefficients of correlation between the dependent variables TFP_LP and TFP_FE, which indicates that there is no serious problem of multicollinearity in the regression of the basic model.

4.3. Preliminary Regression Results

To test Hypothesis 1, an OLS regression model with fixed effects (Equation (5)) is adopted in this study. The regression results are shown in

Table 4. Benchmark regression results.

Variable	(1)	(2)	(3)	(4)
	TFP_LP	TFP_LP	TFP_FE	TFP_FE
MCSRscore	0.006 ***	0.006 ***	0.007 ***	0.007 ***
	(3.18)	(3.07)	(3.87)	(3.41)
Size	0.645 ***	0.651 ***	0.856 ***	0.868 ***
	(109.36)	(106.60)	(148.82)	(145.79)
Age	0.768 ***	0.757 ***	0.001	0.775 ***
	(21.18)	(20.78)	(0.25)	(21.59)
Roa	−0.002	−0.006	2.103 ***	0.002
	(−0.40)	(−1.12)	(21.16)	(0.31)
Growth	2.083 ***	2.053 ***	0.112 ***	2.043 ***
	(20.17)	(20.02)	(7.51)	(20.69)
Lev	0.129 ***	0.143 ***	0.803 ***	0.126 ***
	(8.35)	(9.08)	(22.50)	(8.34)
Top1	0.001 ***	0.001 ***	0.002 ***	0.002 ***
	(4.18)	(4.36)	(7.11)	(6.76)
Cash	0.437 ***	0.439 ***	0.204 ***	0.186 ***
	(8.48)	(8.45)	(4.03)	(3.64)
TQ	−0.021 ***	−0.005	−0.022 ***	−0.006
	(−4.81)	(−0.99)	(−5.17)	(−1.36)
Constant	−5.720 ***	−5.886 ***	−7.973 ***	−8.252 ***
	(−45.31)	(−44.89)	(−64.76)	(−64.64)
Industry	Yes	Yes	Yes	Yes
Time		Yes		Yes
Obs	12,908	12,908	12,908	12,908
R2	0.733	0.736	0.821	0.823

Figures in () are t-values; *** denotes significance at the 1% level.

, where the dependent variable is TFP_LP in the first two columns and TFP_FE in the second two columns. In Table 4, columns (1) and (3), the coefficients on the regression of the explanatory variable MCSRscore appear to be significantly positive (β = 0.006, p < 1%; β = 0.007, p < 1%) when only industry fixed effects are included. In both columns (2) and (4) of Table 4, the coefficients on the explanatory variable MCSRscore are still significantly positive (β = 0.006, p < 1%; β = 0.007, p < 1%) when both industry and year fixed effects are included. In summary, the findings in Table 4 provide support for Hypothesis 1, suggesting that a better quality of mandatory CSR disclosure can significantly contribute to a firm’s TFP.

In terms of the control variable regression results, size, growth, the proportion of large shareholders and cash holdings play a positive role in a firm’s TFP to some extent, which is in line with the previous analysis when selecting the control variables for this study. The remaining controlling variables are not found to be significant in influencing TFP.

4.4. Endogenous Test

First, our research may be subject to reverse causality, as companies with a higher level of TFP may be more likely to disclose CSR information of a higher quality; Second, endogeneity may also be introduced by a firm’s motives for disclosing CSR information. For instance, firms that are less productive may need to provide more detailed CSR information in response to industry and regional shocks, while firms with a higher TFP may provide less detailed CSR information depending on their individual financial situation [2,5].

Regarding the first issue, we adopt the instrumental-variables method to mitigate endogeneity. Based on existing studies [28,29], we use the firm’s MCSRscore in year t−1 (LMCSRscore) and the mean of the MCSRscore of the other firms in the same industry (MCSRscore_ind) as the instrumental variables (IV) and conduct a two-stage OLS (2SLS) regression to verify the robustness of the baseline regression results. After examining the instrumental variables for the rationality, the following is found: the Kleibergen-Paap rk LM statistic is significant at the 1% level (p = 0.000), which means that the instrumental variables are identifiable; The F value of Kleibergen–Paap rk Wald (F = 1116.23) is much greater than the 10% level of the Stock–Yogo test, which means that the instrumental variables correlate significantly with the endogenous explanatory variables. The first three columns of

Table 5. Endogeneity test.

Variable	(1)	(2)	(3)	(4)	(5)
	IV-2SLS			PSM
	MCSRscore	TFP_LP	TFP_FE	TFP_LP	TFP_FE
MCSRscore		0.009 ***	0.011 ***	0.008 **	0.007 *
		(2.99)	(4.06)	(2.02)	(1.94)
MCSRscore_ind	0.225 ***
	(6.47)
MLCSRscore	0.579 ***
	(103.08)
Size	0.077 ***	0.651 ***	0.867 ***	0.663 ***	0.878 ***
	(3.92)	(110.71)	(152.75)	(56.27)	(78.29)
Age	0.090 ***	−0.007	0.001	−0.004	0.003
	(5.33)	(−1.32)	(0.11)	(−0.34)	(0.33)
Roa	0.710 ***	2.053 ***	2.041 ***	2.478 ***	2.584 ***
	(2.70)	(26.17)	(26.95)	(11.66)	(12.57)
Growth	0.085 **	0.144 ***	0.127 ***	0.109 ***	0.072 ***
	(2.10)	(11.98)	(10.91)	(3.98)	(2.80)
Lev	−0.218 **	0.757 ***	0.777 ***	0.783 ***	0.818 ***
	(−2.10)	(24.36)	(25.88)	(11.51)	(12.49)
Top1	−0.001	0.001 ***	0.002 ***	0.001 **	0.002 ***
	(−1.04)	(4.27)	(6.51)	(2.20)	(3.10)
Cash	0.054	0.438 ***	0.186 ***	0.398 ***	0.056
	(0.33)	(8.87)	(3.90)	(4.33)	(0.63)
TQ	0.023	−0.005	−0.006	0.001	−0.006
	(1.44)	(−1.01)	(−1.42)	(0.07)	(−0.68)
Constant				−6.176 ***	−10.541 ***
				(−24.25)	(−29.14)
Chi-sq (1) P	0.000
F-value	1116.23
Industry	Yes	Yes	Yes	Yes	Yes
Time	Yes	Yes	Yes	Yes	Yes
Obs	12,908	12,908	12,908	3347	3347
R2	0.466	0.653	0.770	0.749	0.835

Figures in () are t-values; ***, ** and * denote significance at the 1%, 5% and 10% levels, respectively.

show the results of the IV-2SLS regressions. In column (1), we report the first stage of regression results, where the coefficients of MCSRscor_ind and LMCSRscore are 0.225 and 0.579, respectively, both significantly positive at the 1% level, suggesting a significant correlation between the instrumental and explanatory variables. In the second stage of regressions in columns (2) and (3), the coefficients of the explanatory variable MCSRscore are 0.009 and 0.011, respectively, both significantly positive at the 1% level, which suggests that, after mitigating the potential endogeneity problem, improving the quality of mandatory CSR disclosure is still a significant driver of a firm’s TFP. Moreover, the results for the control variables are similar to those in the benchmark regression in Table 4. In summary, the results of this paper appear to be robust to the IV-2SLS tests in the first three columns of Table 5.

To address the second issue, we use propensity score matching (PSM) to match firms with better MCSR scores to firms with worse MCSR scores in terms of their industry, region and financial status. In particular, we partition the sample into an experimental group and a control group according to the median MCSR score of firms belonging to the same industry and region and match the two groups of firms using the PSM method. Following previous studies [1,5], we include Size, Lev, Age, Roa, Growth, Cash, Analyst and Media as the covariates for a 1:1 neighborhood matching with caliper = 0.25. In particular, Analysts and Media are the natural logarithms of the number of analysts and media-tracking firms, respectively [5]. After the PSM matching, a total of 3347 samples are obtained in which the differences in each covariate between the experimental and control groups are significantly lower, and their differences are all less than 5%, thus effectively mitigating the effect of industry, regional and individual financial status differences on the results of this paper. The results of the regressions on the basis of the PSM sample are reported in Table 5, columns (4) and (5). In both columns (4) and (5), we can see that the coefficients of the MCSRscore remain significant and positive (β = 0.008, p < 5%; β = 0.007, p < 10%), suggesting that our results stay robust when controlling for the differences in industry, region and financial status of the firms.

4.5. Robustness Test

In order to further strengthen the reliability of the results of the research, this study adopts the following methods to test the robustness of the regression results. (1) The increase in the quality of CSR disclosure on a firm’s TFP may be achieved through the firm’s innovation path, while the firm’s technology development and the application of innovation patents usually take a long time, and this time lag is always difficult to observe [5]. Therefore, referring to Comin and Hobijn [30], we also run regressions using the TFP in period T + 2 as the dependent variable in order to more intuitively examine the impact of improvements in the quality of mandatory CSR disclosure on a firm’s TFP over the next two years, although this will result in the loss of sample observations in 2020. (2) Referring to Ackerberg et al. [31], this paper also uses TFP, as measured by the ACF method, as the dependent variable for robustness testing, which is able to overcome the implausible assumption that the adjustment cost of factor inputs is zero in the LP method. (3) Third, we replace the independent variables. As mentioned earlier, we also use Hexun’s CSR index (HXMCSRscore) as an independent variable for the robustness test [23].

The robustness results are reported in

Table 6. Robustness tests.

Variable	(1)	(2)	(3)	(4)	(5)
	TFP_LP (T + 2)	TFP_FE (T + 2)	TFP (ACF)	TFP_LP	TFP_FE
MCSRscore	0.007 ***	0.007 ***	0.005 ***
	(2.84)	(3.16)	(3.25)
HXMCSRscore				0.071 ***	0.057 ***
				(6.44)	(5.32)
Size	0.648 ***	0.861 ***	0.820 ***	0.646 ***	0.864 ***
	(84.93)	(112.86)	(139.29)	(103.46)	(142.37)
Age	0.697 ***	0.721 ***	0.762 ***	0.833 ***	0.848 ***
	(16.24)	(16.74)	(21.52)	(22.02)	(22.70)
Roa	−0.007	0.000	0.001	−0.007	−0.000
	(−1.09)	(0.01)	(0.13)	(−1.33)	(−0.07)
Growth	2.239 ***	2.420 ***	2.020 ***	2.230 ***	2.359 ***
	(17.89)	(19.58)	(20.67)	(15.10)	(16.57)
Lev	0.183 ***	0.167 ***	0.130 ***	0.114 ***	0.096 ***
	(10.30)	(9.66)	(8.63)	(7.07)	(6.20)
Top1	0.002 ***	0.003 ***	0.002 ***	0.001 ***	0.002 ***
	(4.93)	(6.99)	(6.31)	(2.79)	(5.31)
Cash	0.338 ***	0.121 *	0.221 ***	0.367 ***	0.114 **
	(5.42)	(1.92)	(4.41)	(7.01)	(2.21)
TQ	0.010	0.007	−0.006	−0.004	−0.006
	(1.52)	(0.98)	(−1.36)	(−0.74)	(−1.32)
Constant	−5.767 ***	−8.047 ***	−7.692 ***	−5.946 ***	−8.301 ***
	(−35.14)	(−48.97)	(−60.99)	(−44.57)	(−63.82)
Industry	Yes	Yes	Yes	Yes	Yes
Time	Yes	Yes	Yes	Yes	Yes
Obs	10,356	10,356	12,908	12,908	12,908
R2	0.696	0.784	0.812	0.746	0.831

Figures in () are t-values; ***, ** and * denote significance at the 1%, 5% and 10% levels, respectively.

. Columns (1) and (2) show that coefficient of MCSRscore remains both positive and significant (β = 0.007, p < 1%; β = 0.007, p < 1%), indicating that a better quality of mandatory CSR disclosure contributes significantly to a firm’s TFP over the next two years. In column (3), when TFP is measured by the ACF method, the coefficient for MCSRscore is still significantly positive (β = 0.005, p < 1%), indicating that the results of the benchmark regression remain robust after replacing the TFP measure. The coefficients on HXCSRscore are significantly positive in both column (4) and column (5) (β = 0.071, p < 1%; β = 0.057, p < 1%), suggesting that the results are still robust to the replacement of the quality of mandatory CSR disclosure indicator.

4.6. Mechanism Analysis

On the basis of the findings of the empirical analysis above, we have tentatively shown that improving the quality of mandatory CSR disclosure can effectively increase a firm’s TFP. Next, we further discuss the internal mechanism of this positive effect between the quality of mandatory CSR disclosure and TFP based on previous theoretical analysis. Specifically, by regressing Equations (6)–(9), we further examine whether the agency costs of firms and the alleviation of external financing constraints are the influential mechanisms behind the quality of mandatory CSR disclosure driving a firm’s TFP.

The mechanism test results are presented in

Table 7. Mechanism tests.

Variable	(1)	(2)	(3)	(4)
	Mfee	MSfee	Cost	KZ
MCSRscore	−0.001 ***	−0.023 **	−0.003 ***	−0.008 **
	(−2.61)	(−1.97)	(−4.63)	(−2.30)
Size	−0.005 ***	−0.062 **	0.394 ***	0.591 ***
	(−11.02)	(−2.10)	(59.73)	(66.18)
Age	−0.002 ***	−0.012	0.083 ***	−0.084 ***
	(−5.05)	(−0.44)	(13.20)	(−10.09)
Roa	−0.300 ***	−4.846 ***	5.094 ***	1.521 ***
	(−23.97)	(−8.26)	(35.13)	(10.58)
Growth	−0.017 ***	0.512 ***	0.041 **	−0.101 ***
	(−11.82)	(5.82)	(2.19)	(−4.76)
Lev	−0.067 ***	0.826 ***	−0.253 ***	0.247 ***
	(−16.23)	(3.75)	(−5.52)	(4.62)
Top1	−0.023 ***	−1.219 ***	−0.365 ***	−0.109 *
	(−7.77)	(−6.57)	(−8.44)	(−1.94)
Cash	0.020 ***	−3.054 ***	0.136 **	0.356 ***
	(3.83)	(−9.11)	(2.39)	(4.85)
TQ	0.015 ***	0.203 ***	0.145 ***	0.166 ***
	(19.58)	(5.05)	(21.78)	(20.78)
Constant	0.235 ***	7.131 ***	−7.315 ***	−10.229 ***
	(23.10)	(11.18)	(−52.82)	(−53.48)
Industry	Yes	Yes	Yes	Yes
Time	Yes	Yes	Yes	Yes
Obs	12,908	12,908	12,908	12,908
R2	0.369	0.078	0.391	0.407

Figures in () are t-values; ***, ** and * denote significance at the 1%, 5% and 10% levels, respectively.

. Columns (1) and (2) of Table 7 show the regression tests of Equations (6) and (7). We can see that the regression coefficients of the MCSRscore are significantly negative in both columns (β = −0.001, p < 1%; β = −0.023, p < 5%), implying that the agency costs of firms decrease significantly with the increasing quality of mandatory CSR disclosure. This result confirms our previous hypothesis that mandatory CSR information disclosure significantly alleviates information asymmetry issues, thereby effectively improving corporate agency conflicts.

Column (3) of Table 7 shows the results of Equation (8). The regression coefficient for MCSRscore also shows a significantly negative value (β = −0.003, p < 1%), indicating that a better mandatory CSR disclosure score can effectively decrease the costs of debt and enhance the firm’s ability to raise external finance. Column (4) presents the results of Equation (9), showing that the coefficient of regression of MCSRscore is −0.008 and significantly negative at the 5% level, which indicates that increasing the MCSRscore can be effective in alleviating the financing constraints of firms. The results in columns (3) and (4) confirm our previous hypothesis that firms effectively increase their external financing efficiency by transmitting friendly signals to the outside world through CSR information disclosure.

It has been extensively documented, on the basis of the information asymmetry theory, agency cost theory and signaling theory, that mitigating financing constraints and agency conflicts can effectively stimulate a firm’s TFP [8,9,25]. Thus, the results in Table 7 provide a preliminary validation of the mechanism by which the quality of mandatory CSR disclosure increases a firm’s total factor productivity. This mechanism of influence is further investigated in the heterogeneity analysis.

5. Heterogeneity Analysis

5.1. External Financing for SOEs and Non-SOEs

Studies have shown that state-owned banks hold a relatively high share of the total assets of the Chinese banking system and that there is therefore an inevitable ownership bias in lending to enterprises, which generally results in non-SOEs having less access to bank finance than SOEs [5]. As a result, non-SOEs in China tend to suffer from more serious financing constraints compared to SOEs. Combined with previous analyses, we assume that better quality mandatory CSR disclosure should be more likely to alleviate the funding constraints faced by non-SOEs relative to SOEs, and should thus have a greater driving effect on their TFP. In order to test the above hypothesis, we sub-divide the sample firms based on their ownership status into SOE and non-SOE groups and run separate regressions.

The regression results for the sub-groups are reported in

Table 8. Analysis of the heterogeneity of SOEs and non-SOEs.

Variable	(1)	(2)	(3)	(4)
	Non-SOEs	SOEs	Non-SOEs	SOEs
	TFP_LP	TFP_LP	TFP_FE	TFP_FE
MCSRscore	0.011 ***	−0.003	0.012 ***	−0.001
	(2.96)	(−1.42)	(3.34)	(−0.37)
Size	0.577 ***	0.538 ***	0.763 ***	0.692 ***
	(23.18)	(29.38)	(32.08)	(40.74)
Age	0.064 *	0.009	0.098 ***	0.069 ***
	(1.88)	(0.34)	(3.01)	(2.73)
Roa	1.202 ***	0.854 ***	1.093 ***	0.710 ***
	(7.12)	(9.06)	(7.57)	(7.72)
Growth	0.174 ***	0.194 ***	0.183 ***	0.199 ***
	(9.04)	(14.23)	(10.32)	(14.67)
Lev	0.103	0.332 ***	0.069	0.321 ***
	(1.18)	(5.83)	(0.80)	(5.67)
Top1	0.001	−0.000	0.001	0.001
	(1.21)	(−0.36)	(0.85)	(0.96)
Cash	−0.024	0.111 *	−0.247 ***	−0.092
	(−0.26)	(1.85)	(−2.80)	(−1.60)
TQ	0.015	0.025 ***	0.013	0.027 ***
	(1.64)	(4.56)	(1.55)	(4.89)
Constant	−3.916 ***	−3.122 ***	−5.504 ***	−4.233 ***
	(−7.20)	(−7.95)	(−10.60)	(−11.53)
Industry	Yes	Yes	Yes	Yes
Time	Yes	Yes	Yes	Yes
Obs	8822	3726	8822	3726
R2	0.954	0.935	0.970	0.951
p-value	0.000 ***		0.000 ***

Figures in () are t-values; *** and * denote significance at the 1% and 10% levels, respectively. “p-values” were used to test for differences in CSRscore coefficients between groups and were obtained by bootstrap sampling 500 times.

. In columns (1) and (2) of Table 8, for the dependent variable TFP_LP, the coefficients of the MCSRscore in the two regressions are 0.011 and −0.003, respectively, the former being significantly positive at the 1% level, while the latter is insignificant. A further comparison revealed that the two regression coefficients are statistically significantly different from each other at the 1% level (0.011 is significantly greater than −0.003, p-value = 0.000). This is an indication that the enhancement of the quality of mandatory CSR disclosure is a better driver of TFP for non-SOEs. Columns (3) and (4) of Table 8 show similar results when TFP_FE is used. The coefficient for MCSRscore in column (3) is significant and positive, but the coefficient for MCSRscore in column (4) is not significant, and a further comparison shows that the two regression coefficients are also statistically significantly different at the 1% level (0.012 is significantly greater than −0.001, p-value = 0.00). Once again, this supports the view that the improved quality of mandatory CSR disclosure is a better driver of TFP for non-SOEs than it is for SOEs. The findings of Table 8 provide further evidence that the improvement in mandatory CSR information disclosure quality is a crucial mechanism driving the alleviation of financing constraints and the enhancement of a firm’s TFP.

5.2. Heterogeneity in Firm Size

It has also been found that firm size tends to have a negative relationship with financing constraints [32]. For example, it has been suggested that larger firms are usually able to have more assets as collateral to obtain credit resources from external sources compared to smaller firms [33], and therefore their activities such as innovation and investment are less constrained by financing constraints, which is undoubtedly more conducive to the enhancement of a firm’s TFP. Combined with previous analyses, we expect that improvements in the quality of mandatory CSR disclosure would have a greater effect in improving the external financing constraints of smaller firms, and thus make a greater contribution to their TFP, relative to larger firms with relatively abundant credit resources. To support this conjecture, the paper also divides the full sample of firms into a sub-sample group of large firms and a sub-sample group of small firms, using the median size of the sample firms as the criterion, and conducts a comparative regression analysis.

The sub-group regression results are presented in

Table 9. Heterogeneity analysis of firm size.

Variable	(1)	(2)	(3)	(4)
	Small Firms	Large Firms	Small Firms	Large Firms
	TFP_LP	TFP_LP	TFP_FE	TFP_FE
MCSRscore	0.008 ***	−0.002	0.011 ***	0.001
	(3.65)	(−0.90)	(3.51)	(0.10)
Size	0.555 ***	0.479 ***	0.641 ***	0.715 ***
	(24.87)	(18.47)	(27.53)	(32.45)
Age	0.063 **	0.127 ***	0.167 ***	0.119 ***
	(2.45)	(4.03)	(5.67)	(4.56)
Roa	1.178 ***	0.674 ***	0.603 ***	0.831 ***
	(9.94)	(6.01)	(5.54)	(7.32)
Growth	0.138 ***	0.255 ***	0.250 ***	0.160 ***
	(9.88)	(13.73)	(14.46)	(11.12)
Lev	0.254 ***	0.358 ***	0.303 ***	0.145 **
	(3.73)	(5.28)	(4.57)	(2.12)
Top1	−0.001 *	0.001	0.002	−0.001
	(−1.66)	(0.99)	(1.45)	(−0.74)
Cash	0.105	0.176 **	−0.030	−0.134 **
	(1.42)	(2.56)	(−0.46)	(−2.00)
TQ	0.051 ***	0.015 **	0.015 **	0.053 ***
	(7.11)	(2.31)	(2.48)	(8.37)
Constant	−3.413 ***	−2.175 ***	−3.388 ***	−4.507 ***
	(−6.85)	(−4.04)	(−7.05)	(−9.21)
Industry	Yes	Yes	Yes	Yes
Time	Yes	Yes	Yes	Yes
Obs	6447	6451	6447	6451
R2	0.945	0.919	0.951	0.928
p-value	0.000 ***		0.000 ***

Figures in () are t-values; ***, ** and * denote significance at the 1%, 5% and 10% levels, respectively. “p-values” were used to test for differences in CSRscore coefficients between groups and were obtained by bootstrap sampling 500 times.

. In columns (1) and (2) of Table 9, when the dependent variable is TFP_LP, the coefficients of the MCSRscore in the two regressions are 0.008 and −0.002, respectively, with the former being significantly positive at the 1% level, while the latter is not significant. A further comparison reveals that the two regression coefficients are statistically significantly different from each other at the 1% level (0.008 is significantly greater than −0.002, p-value = 0.000). This suggests that the improvement in the quality of mandatory CSR disclosure is a better driver of TFP among smaller firms. The results in columns (3) and (4) of

Table 10. Heterogeneity analysis of equity incentives.

Variable	(1)	(2)	(3)	(4)
	Low Mshare	High Mshare	Low Mshare	High Mshare
	TFP_LP	TFP_LP	TFP_FE	TFP_FE
MCSRscore	0.009 ***	0.001	0.010 ***	0.001
	(3.48)	(0.41)	(3.67)	(0.10)
Size	0.536 ***	0.551 ***	0.688 ***	0.720 ***
	(21.22)	(26.20)	(28.61)	(35.55)
Age	0.008	0.057 **	0.082 **	0.087 ***
	(0.22)	(2.17)	(2.35)	(3.38)
Roa	0.810 ***	0.900 ***	0.696 ***	0.743 ***
	(6.85)	(7.65)	(5.99)	(6.75)
Growth	0.202 ***	0.191 ***	0.204 ***	0.199 ***
	(12.35)	(12.20)	(12.51)	(13.28)
Lev	0.278 ***	0.294 ***	0.251 ***	0.252 ***
	(3.92)	(4.10)	(3.35)	(3.70)
Top1	−0.002 *	0.002 *	−0.002	0.003 ***
	(−1.80)	(1.90)	(−1.64)	(2.96)
Cash	−0.050	0.283 ***	−0.220 ***	0.078
	(−0.75)	(3.73)	(−3.29)	(1.13)
TQ	0.036 ***	0.020 ***	0.037 ***	0.023 ***
	(5.97)	(2.86)	(5.87)	(3.55)
Constant	−3.003 ***	−3.488 ***	−4.057 ***	−4.819 ***
	(−5.61)	(−7.62)	(−7.91)	(−10.83)
Industry	Yes	Yes	Yes	Yes
Time	Yes	Yes	Yes	Yes
Obs	6448	6450	6448	6450
R2	0.958	0.938	0.966	0.936
p-value	0.000 ***		0.000 ***

Figures in () are t-values; ***, ** and * denote significance at the 1%, 5% and 10% levels, respectively. “p-values” were used to test for differences in CSRscore coefficients between groups and were obtained by bootstrap sampling 500 times.

are similar when the dependent variable is TFP_FE, with the coefficient for MCSRscore being significantly positive in column (3), while the coefficient for MCSRscore is not significant in column (4). A further comparison reveals that the two regression coefficients are also statistically significantly different from each other at the 1% level (0.011 is significantly greater than 0.001, p-value = 0.00). This is again an indication that the promotion of mandatory CSR disclosure quality on TFP is better for smaller companies. Once again, the findings in Table 9 support the view that the mitigating role of the improved quality of mandatory CSR disclosure on the financing constraint problem is a key mechanism driving a firm’s TFP.

5.3. Heterogeneity in Equity Incentives

Equity incentives allow managers to participate in the allocation of a firm’s profits and to assume the same risks as shareholders. This helps to strengthen the alignment of interests between managers and shareholders and thus serves as an important approach to mitigate principal–agent conflicts. For example, Tzioumis [34] shows that US listed firms that engage in the distribution of residual claims among stakeholders through the introduction of management equity plans are better able to reduce principal–agent costs. Combined with previous analyses, we expect that an increase in the quality of mandatory CSR disclosure is more effective in mitigating the agency conflict problem in firms with lower management shareholdings, and therefore has a stronger contribution to their TFP. To prove this conjecture, we divide the sample into a sub-sample with high management shareholding and a sub-sample with low management shareholding using the median management shareholding (Mshare) of the firms and conduct a comparative regression analysis, where Mshare is the ratio of the number of shares held by management to the number of shares in the firm.

The estimated results of the sub-group regressions are presented In Table 10. In columns (1) and (2) of Table 10, when the dependent variable is TFP_LP, the coefficients of the MCSRscore in the two regressions are 0.009 and 0.001, respectively, with the former being significantly positive at the 1% level, while the latter is not significant. A further comparison reveals that the two regression coefficients are statistically significantly different at the 1% level (0.009 is significantly greater than 0.001, p-value = 0.000). This indicates that improving the quality of mandatory CSR disclosure is a better driver of TFP for firms with lower management ownership. The results in Table 10, columns (3) and (4) are similar when the dependent variable is TFP_FE. Column (3) reveals that the coefficient for MCSRscore is significantly positive, while column (4) shows that the coefficient for MCSRscore is not significant, and a further comparison reveals that the two regression coefficients are also statistically significantly different at the 1% level (0.010 is significantly greater than 0.001, p-value = 0.00). This again supports the view that the enhancement of the quality of mandatory CSR disclosure on TFP is greater in companies with lower management shareholding. Overall, the findings in Table 10 prove that the mitigating effect of increased quality of mandatory CSR disclosure on principal–agent problems is an important mechanism driving a firm’s TFP.

6. Extended Analysis

In the theoretical analysis section, we predict that better quality mandatory CSR disclosure can increase a firm’s TFP through the two channels of improving the firm’s innovation capability and investment efficiency. Therefore, in this section, through the regression analysis of Equations (10)–(13), we further examine the mediating role played by a firm’s innovation capability and investment efficiency in the relationship between the quality of mandatory CSR disclosure and the firm’s TFP.

Table 11. Analysis of the intermediary effect of innovation capability and investment efficiency.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	RD	TFP_LP	TFP_FE	InvEff	TFP_LP	TFP_FE
MCSRscore	0.032 ***	0.004 ***	0.003 *	−0.001 ***	0.005 ***	0.006 ***
	(6.75)	(3.60)	(1.72)	(−2.81)	(2.83)	(3.20)
RD		0.090 ***	0.102 ***
		(14.13)	(15.82)
InvEff					−1.813 ***	−1.598 ***
					(−15.17)	(−13.11)
Size	0.867 ***	0.573 ***	0.780 ***	−0.001 ***	0.649 ***	0.866 ***
	(58.10)	(69.03)	(96.10)	(−2.71)	(107.17)	(146.50)
Age	−0.435 ***	0.796 ***	0.819 ***	0.019 ***	0.791 ***	0.806 ***
	(−5.85)	(22.62)	(23.87)	(7.65)	(22.04)	(22.74)
Roa	0.136 ***	−0.018 ***	−0.012 **	−0.002	−0.007	0.001
	(9.71)	(−3.42)	(−2.38)	(−0.96)	(−1.25)	(0.20)
Growth	1.565 ***	1.911 ***	1.884 ***	0.031 ***	2.109 ***	2.093 ***
	(7.52)	(19.11)	(19.67)	(5.08)	(20.67)	(21.29)
Lev	0.025	0.141 ***	0.123 ***	0.019 ***	0.178 ***	0.156 ***
	(0.78)	(9.09)	(8.37)	(10.52)	(11.29)	(10.25)
Top1	0.000	0.001 ***	0.002 ***	0.000 ***	0.002 ***	0.002 ***
	(0.23)	(4.35)	(6.76)	(5.00)	(5.15)	(7.48)
Cash	0.493 ***	0.394 ***	0.136 ***	−0.012 ***	0.417 ***	0.167 ***
	(3.15)	(7.53)	(2.67)	(−3.00)	(8.12)	(3.30)
TQ	0.027 **	−0.007	−0.009 **	0.001	−0.004	−0.006
	(2.23)	(−1.56)	(−2.04)	(0.71)	(−0.89)	(−1.26)
Constant	−1.826 ***	−5.721 ***	−8.066 ***	0.057 ***	−5.784 ***	−8.162 ***
	(−5.49)	(−44.45)	(−65.10)	(5.09)	(−44.44)	(−64.30)
Industry	Yes	Yes	Yes	Yes	Yes	Yes
Time	Yes	Yes	Yes	Yes	Yes	Yes
Obs	12,908	12,908	12,908	12,908	12,908	12,908
R2	0.476	0.748	0.834	0.094	0.742	0.502

Figures in () are t-values; ***, ** and * denote significance at the 1%, 5% and 10% levels, respectively.

reports the regression results of Equations (10)–(13), where column (1) corresponds to the regression results of Equation (10). The regression coefficient of MCSRscore can be seen to be significantly positive at the 1% level (β = 0.032, p < 1%) in column (1) when the firm’s R&D expenditure (RD) is used as the dependent variable, implying that a firm’s innovation capability improves significantly with the quality of mandatory CSR disclosure. Columns (2) and (3) correspond to the test results of Equation (11) and show that the regression coefficients of the explanatory variable MCSRscore are significantly positive in both columns (β = 0.004, p < 1%; β = 0.003, p < 10%), and the regression coefficients of the mediating variable RD are also significantly positive (β = 0.090, p < 1%; β = 0.102, p < 1%). This suggests a positive mediating effect of corporate innovation capability between the quality of mandatory CSR disclosure and TFP. Column (4) of Table 11 corresponds to the regression results of Equation (12). In column (4), when the firm’s investment efficiency (InvEff) is adopted as the dependent variable, it can be seen that the regression coefficient of MCSRscore is significantly negative at the 1% level (β = −0.001, p < 1%), which is an indication that the firm’s investment efficiency significantly improves with the better quality of mandatory CSR disclosure. Columns (5) and (6) correspond to the test results of Equation (13). In both columns (5) and (6), the regression coefficients of the explanatory variable MCSRscore are significantly positive (β = 0.005, p < 1%; β = 0.006, p < 1%), while the regression coefficients of the mediating variable InvEff are significantly negative (β = −1.813, p < 1%; β = −1.598, p < 1%). This suggests a positive mediating effect of a firm’s investment efficiency between the quality of mandatory CSR disclosure and TFP.

In summary, the findings in Table 11 show that improving the quality of mandatory CSR disclosure improves a firm’s TFP by increasing its innovativeness and investment efficiency, thereby validating our previous hypothesis regarding the two important impact channels of firm innovation capability and capital allocation efficiency.

7. Discussion

This study makes several significant theoretical contributions. Firstly, although many studies have discussed the role of mandatory CSR disclosure in the development of Chinese firms [1,2], none of them have adequately quantified the impact of the quality of CSR disclosure on the development of the company. In contrast, this paper establishes a quality assessment index for CSR information disclosure and investigates its correlation with a firm’s TFP, leading to an enhanced understanding of the consequences of CSR information disclosure on a firm’s development transformation. Secondly, our study is the first to include innovation capability and investment efficiency in the study of the impact of CSR disclosure on TFP and finds that improvements in a firm’s innovation capability and investment efficiency are important channels through which CSR disclosure can promote a firm’s TFP. This further expands the channels of influence in the field of CSR research for promoting firm economic development.

In addition, this paper has certain limitations and future research could potentially address these areas of improvement. First, this study used only content analysis and indicator methods to measure the quality of CSR information disclosure; therefore, future research could develop and implement superior approaches to assess CSR information quality. Secondly, our study adopted TFP, calculated using the LP and ACF methodologies, as an indicator to measure the quality of firm development. However, despite their widespread use, these techniques still have several limitations. Therefore, future research may seek alternative approaches to calculate the TFP of firms. Third, besides examining channels such as innovation capacity and investment efficiency that affect a firm’s TFP, future studies could also explore other external factors such as industry competitiveness and environmental pressures.

8. Conclusions and Policy Implications

Using data on Chinese A-share listed companies from 2009 to 2020, this article empirically examines how the quality of mandatory CSR information disclosure affects a firm’s TFP. Our findings are as follows: (1) An improvement of the quality of mandatory CSR information can lead to an enhancement of a firm’s TFP by reducing the firm’s principal–agent problems and easing their external financing constraints; (2) The driving effect of improvements in the quality of mandatory CSR information on a firm’s TFP is stronger for non-SOEs, large firms, and firms with weaker equity incentives; (3) A firm’s innovation capability and investment efficiency are two key channels through which the quality of mandatory CSR information disclosure promotes a firm’s TFP. We derive the following policy implications from the above findings:

(1)

External financing constraints and principal–agent conflicts are significant constraints on enterprise productivity. Therefore, it is imperative for the Chinese government to appropriately accelerate the development of financial markets to improve the overall external financing environment for enterprises. In addition, companies should proactively disclose their information and send positive signals to external stakeholders. This will not only help reduce information asymmetry between firms and the public, but also widen financing channels for firms. Such measures are crucial for reducing information friction and improving access to finance, which are essential for firms to achieve higher productivity.

(2)

Innovation and investment are two important channels through which mandatory CSR information disclosure affects a firm’s TFP, but improvements in a firm’s innovation capability and investment efficiency are difficult to achieve in a short period of time. Therefore, the government should focus on guiding enterprises to treat CSR activities and related information disclosure as a long-term investment, as it is beneficial to the long-term development of enterprises.

(3)

Given the different ownership characteristics, scale of operation and equity incentive structure of enterprises, the effect of CSR disclosure quality on a firm’s innovation capability, investment efficiency and TFP may differ. Therefore, the government should formulate and implement tailored CSR policies based on the unique characteristics of enterprises to better promote economic and business development.