The Nonlinear Effect of Economic Policy Uncertainty on Corporate Social Responsibility

Abstract

The debate on how corporate social responsibility (CSR) responds to and reacts to the increasing economic policy uncertainty (EPU) is still unsettled. The present study sheds light on the curvilinear relationship between EPU and CSR, employing the provincial EPU index in China for the period of 2010–2016. More precisely, when EPU is in an appropriate range, as EPU grows, its CSR engagement experiences an initial increase; however, once the threshold value is reached, the additional increments in EPU result in a reduction in CSR engagement. Additionally, the inflection point of the state-owned firms, large-sized firms, and firms in their mature stage are large, indicating that economic policy uncertainty influences corporate social responsibility positively in a rather wider range. We also show that the inverted U-shaped effect is negatively moderated by financial constraints and government interventions. The findings of this study hold significance for policy-makers, offering valuable insights related to the structuring of efficient policies.

1. Introduction

The importance of being environmentally and socially responsible has been realized once again with the COVID-19 pandemic. China is presently in the process of refining and exploring the system for corporate social responsibility (CSR), and enterprises have greater discretion in whether to fulfill social responsibility and the degree of fulfillment [1]. Corporate social responsibility has transformed from a voluntary attribute to an obligatory duty [2], and it is a strategic tool that enhances competitiveness [3,4,5] and promotes corporate contributions to sustainable development [6]. The existing studies have identified the factors that are relevant to CSR engagement pertaining to internal characteristics and the external environment [7], among which changes in macroeconomic policies are an important factor. Uncertainty is thus created when policy changes occur that will affect the business environment [8]. According to the Special Survey Report on the Growth and Development of Chinese Entrepreneurs in 2019 released at the Boao Forum for Asia Annual Meeting in 2019, 55.5% of entrepreneurs stated that they were facing increasing uncertainty and operating pressure.

Economic policy uncertainty (EPU) implies that economic agents cannot accurately predict if, when, and how the government will modify the current economic policies [9]. Due to the seminal work of Bloom [10] and the subsequent development of the Baker index [11], studies on EPU have sparked a boom. The impact of EPU on firms’ behaviors has been extensively examined, such as reducing M&A activities [12] and promoting enterprise innovation [13]. The recent literature has shifted to examining the nonlinear effect of EPU, for instance, the U-shaped relationship between EPU and firms’ precautionary cash holdings and financialization [14,15]. There are two opposing views on the impact of EPU on CSR, i.e., it either promotes or inhibits CSR. In a period of high EPU, firms engage in less CSR activities due to cost concerns [16], future uncertainty [17], and risk aversion [18]. Firms may be reluctant to invest scarce resources in CSR [19]. Meanwhile, under the precautionary saving theory, firms may increase their precautionary cash holdings to prevent themselves from experiencing financing constraints [14]. A contrasting view suggests that improving social responsibility performance can satisfy the needs of stakeholders and increase firm value by bringing new market opportunities, reducing information asymmetry risks, and positively transmitting a signal of financial health [20,21,22]. Interestingly, when EPU is in an appropriate range, the economic prospects are promising. Therefore, for the sake of competition, firms may take more CSR actions in periods when EPU is in the desired range.

Due to the increasing economic policy uncertainty, it is worth examining how CSR is practiced as an important strategic management behavior. As far as we know, there is no literature examining the nonlinear relationships between EPU and CSR in China. At the same time, most of the existing research focuses on economic policy uncertainty at the national level and ignores the variability present at the geographical level. China is an ideal place to study this issue. First, the Chinese economy is currently in a transition phase with an increasing downward pressure. Changes in the government’s economic policies are important weathervanes for economic development. And, China’s transitional economy lacks coherent economic policies. Thus, firms face a turbulent business environment [23]. Economic policies are almost exogenous to Chinese firms, as firms usually do not have the opportunity to convince the government to implement favorable economic policies. This ensures that we can establish causal channels regardless of the endogenous factors between uncertainty and investment in CSR. Second, firms are subjected to a higher number of government interventions [24], where the central government executes a top-level design at the policy level, but in terms of specific implementations, the local government has higher decision-making powers. Since the reform and opening up of China, the central government has delegated some of its affairs and financial powers to local governments to stimulate local economic development. This has given local governments considerable autonomy in decision-making on resource allocation and policy formulation [25]. Local government policies may exert additional uncertainty [26]. Third, following the approach of Baker [11], Yu et al. [27] selected the daily newspapers from 31 provinces as the source for news media reports and performed keyword searches to construct a set of EPU indexes for China by province and over time. The provincial EPU has demonstrated success in identifying regional heterogeneity. And, regional heterogeneity has been demonstrated to be a characteristic that has a significant impact on CSR [7,28].

To deal with the topic, using 16,016 firm-year observations across 2907 unique firms between the year 2010–2016, we investigate the impact of EPU on firms’ CSR engagement from a nonlinear perspective. On this basis, this study further analyzes the heterogeneous impact of EPU on CSR from the perspectives of firm characteristics based on the equity nature, firm size, and the life-cycle stage. Lastly, we also explore the moderating role of financial constraints and government interventions.

We contribute in three ways to the existing literature. First, based on the TMGT (too much of a good thing) effect, we propose and validate the existence of a nonlinear (inverted U-shaped) relationship between EPU and CSR behaviors, thereby offering a well-supported explanation for the inconsistencies observed in the existing research. Previous studies on this issue have ignored the nonlinear effect displayed by uncertainty. EPU exhibits an effect of first promoting and then suppressing firms’ CSR. This study provides a comprehensive and precise description of the effect of EPU on CSR to fill the gaps in related research fields. Second, the study complements the literature on factors influencing CSR. Our study uses a newly developed EPU index for Chinese provinces, which explores the reasons behind this phenomenon from the perspective of provincial macroeconomic policy changes. In addition, unlike previous studies that have focused more on developed countries, this study examines the significance of EPU in China, an emerging country that relies heavily on government regulations. Finally, our research reveals that EPU has divergent effects on firms with varying property rights, firm sizes, and life cycles. Differences in the inflection points of the U-shaped curve imply how much economic policy uncertainty positively affects CSR behavior. At the same time, the financial constraint and government intervention faced by enterprises can weaken the impact of EPU on CSR. This can provide new clues for related research.

The rest parts are organized as follows. In Section 2, we introduce related literature and develop research hypotheses. Section 3 sets out the methodology and data. Section 4 reports and discusses the results of our empirical analysis as well as robustness tests. Section 5 presents the heterogeneity analyses and the moderating effects, and the final section contains the conclusion, policy recommendations, and limitations of our study.

2. Literature Review and Research Hypotheses

2.1. Literature on CSR under EPU

1.

CSR is positively associated with EPU

The intuition is that, with an increase in EPU, firm risk also escalates. Managers actively participate in a higher level of CSR to decrease systematic risks [3], offset the negative impacts [29], and strive to maintain sustainability [30]. The stakeholder theory advocates investing in socially responsible activities to meet the expectations of stakeholders. CSR activities generate goodwill and moral capital for the company, serving as an insurance mechanism to mitigate stakeholders’ negative judgments during periods of uncertainty [31,32,33]. Scholars demonstrated that the valuation of firms with higher CSR engagement experiences less impact from EPU [34,35]. Furthermore, the signal transmission theory posits that companies interpret CSR performance as a channel to transmit unobservable quality, aiming at enhancing their credibility with suppliers, customers, employees, and other stakeholders [36,37], especially in emerging economies [38]. The theory of strategic growth option also contends with the fact that companies facing intense competition are inclined to adopt preemptive strategies, including investing in CSR, to enhance product differentiation, particularly when external uncertainty is on the rise [39].

2.

CSR is negatively associated with EPU

Higher EPU is generally accompanied by a steep economic decline, such as cutbacks on consumer and government spending, and upward pressure on the firm’s cost of capital. Wang et al. found that managers are risk-averse and will invest less when uncertainty increases [18]. According to the waiting option theory, uncertainty might increase the waiting value and firms might be extra cautious when making investment and financial decisions. Some scholars [8,40,41,42,43,44] generally believe that firms respond to high EPU by postponing their investment. Guedhami et al. indicated that companies allocate resources inefficiently and prefer to hold more cash during high EPU [45]. Based on the theory of information asymmetry, excessive EPU will make it more difficult for investors to search for information, which will weaken corporate investment [18]. Scholars revealed that risk-averse executives tend not to invest in a politically risky environment [46]. Wang et al. showed that, when EPU increases, the expected future social responsibility rating gradually declines, in turn, highly increasing the information asymmetry [47]. Zhao et al. also found that the effect of EPU on various dimensions of CSR is still significantly negative [16]. Firms tend to reduce their charitable giving to conserve resources as a precautionary move in an environment with high EPU [17,48]. Su et al. observed the negative effect between EPU and CSR in food enterprises, which are a part of a sensitive industry [49].

2.2. Hypotheses on EPU and CSR

In summary, most studies have investigated the positive or negative impacts of EPU on CSR from a singular viewpoint, such as EPU promoting CSR in developed European countries [22], EPU promoting CSR in the USA [41], and EPU inhibiting CSR in China [16]. EPU can serve as both a risk and an opportunity, influencing diverse aspects of both macroeconomic dynamics and micro-level firm behaviors. Some authors believe that economic policy uncertainty and enterprise social responsibilities are jointly affected by “promotional effects” and “inhibition effects”. Hoang et al. investigated the nonlinear relationship between corporate social performance and EPU in the UK [50]. The existing research on this topic mostly pertains to a linear perspective and does not consider regional EPU heterogeneity. Firms cannot have enough access to external resources to deal with excessive uncertainty. The TMGT effect [51] has pointed out that a “positive” variable exceeding the threshold cancels any positive effect on firms or even produces negative effects.

In fact, moderate levels of EPU tends to bring opportunities. Specifically, when the business environment is stable and easily predictable, firms will simply respond to the incidents that have already occurred rather than undertaking any additional CSR activities. In contrast, EPU creates an unfavorable business environment, forcing firms to change and transform themselves, thereby requiring more support from their stakeholders [31]. For example, firms could encourage R&D personnel to add social responsibility attributes to the products and processes as important means to achieve product differentiation, which in turn help companies to achieve competitive advantages in the market [32,33]. Moreover, investment in CSR is an active behavior that helps a firm to enhance its corporate image, signal its finance health, and obtain scarce resources, especially in a turbulent environment [20].

However, excessive EPU dramatically increases firms’ business risk levels [52] and managers’ risk perceptions [53], amplifies the information asymmetry between financing institutions and firms, and makes it impossible for firms to accurately predict the outcomes of CSR activities or future development trends [16]. In this situation, its marginal benefits gradually decrease, and firms enact strict retrenchments. Moreover, resource constraints during high EPU may cause some enterprises to engage in speculative behaviors, such as using low-priced raw materials that harm the environment or cutting costs. Firms then perform CSR passively or to a lesser degree.

There are significant structural differences between regional economies in China, such as the policy intensity and economic and financial development conditions of each region [54,55], and these differences may affect the magnitude of CSR activity as a response to EPU shocks. The provincial-level EPU in different regions varies and undoubtedly has an impact on the behavior of economic entities [56]. Yu et al. showed that provincial EPU imposes a significantly positive impact on firms’ carbon emission intensity in China [27].

Based on the TMGT effect, we suggest that EPU first promotes CSR, but when EPU exceeds a certain point, it inhibits CSR. Therefore, the hypothesis is as follows:

H1: 

There is an inverted U-shaped relationship between EPU and CSR.

The conceptual model of this study is summarized in Figure 1.

3. Empirical Specification and Data

(1)

The empirical model

We construct the following baseline regression model to investigate the relationship between provincial economic policy uncertainty and corporate social responsibility:

$${CSR}_{ijth}=\rho +\alpha {EPU}_{ijth}+\beta {EPU}_{ijth}^2+\gamma {Control}_{ijth}+{\epsilon }_{ijth}$$

(1)

The subscripts i, j, h, and t are for firms, provinces, industries, and years, respectively. ${CSR}_{ijth}$ is the corporate social responsibility score of firm i. The core explanatory variables are the primary term ${EPU}_{ijth}$ and the secondary term ${EPU}_{ijth}^2$ which measure the standardized EPU index in province j at year t. ${Controls}_{ijth}$ is a set of control variables that concerns firm-level characteristics and the level of economic development. We also include the province, the industry and firm fixed effects to control for unobserved heterogeneity, and use the clustering robust standard errors at the industry levels in all models. ${\epsilon }_{irth}$ denotes the unobserved exogenous error term.

(2)

Measures of CSR and EPU

CSR is used as an explained variable. The measurement of CSR is based on the ratings provided by third-party organizations [16]. In this study, we mainly examine the actual level of CSR implementation; thus, we choose the Hexun evaluation index (https://www.hexun.com/ (accessed on 21 November 2023)). The evaluation system applies a comprehensive rating based on firms’ CSR activities using the following five categories: firms’ responsibility to shareholders, to employees, to suppliers, customers and consumer rights, to the environment, and to society. The 5 categories contain 13 subcategories and 37 indicators, which are based on public information such as annual reports, CSR reports, official websites, and media reports. The higher the score, the higher the firm’s level of CSR involvement.

The explanatory variable is EPU, which is province specific. Following the methodology of Baker [11], Yu et al. selected the daily newspapers of 31 provinces in China that are regarded as the primary source of official information and used Chinese keywords for screening and index construction [27]. This methodology is more reliable, accurate, and comprehensive. The provincial EPU not only accurately interprets the central government’s policies but also successfully reflects the variations among the different provinces in China and is suitable for panel data analysis. The higher the provincial EPU index, the higher the provincial EPU faced by a company.

(3)

Sample selection and data collection

Our study uses a large sample of A-share listed companies on China’s SSE and SZSE. We eventually narrow our sample to the years between 2010 and 2016 due to CSR and provincial EPU data availability. The information on the company background and financial statistics used in this study is mainly obtained from the China Stock Market and Accounting Research Database (CSMAR). The data of the professional CSR evaluation system, Hexun.com (https://www.hexun.com/ (accessed on 21 November 2023)), are adopted to measure the CSR behavior of Chinese listed firms. The EPU index was developed by Yu et al. [27], which can be downloaded from the website (http://cedcdata.cufe.edu.cn/cedc/metadata/list.html (accessed on 21 November 2023)). We define a firm’s location as the province of its headquarters. According to the province–year code, we merge the provincial EPU index with each firm’s CSR. Like the common practice, we delete financial firms, financially distressed firms, and firms with incomplete data. Our final sample consists of 16,016 firm-year observations with 2907 unique firms. To reduce the influence of extreme values, we winsorize the continuous variables at the 1% and 99% quantiles. We conduct a variance inflation factor (VIF) test on these variables. The VIF test results (1.44) show that there is no multicollinearity.

Table 1. Variables definition.

Variables	Definition
CSR	The actual score of each company in Hexun.
EPU	The standardized EPU index at the provincial level in China.
ROE	Ratio of net income to net assets.
LEV	The ratio of total debt over total assets.
Firm size	Log value of total assets.
Board	The board size of the firm.
Share1	The percentage of the largest holder on the board.
Share10	The percentage of top ten holders on the board.
Daul	A dummy variable is defined as 1 if duality of chairman and general manager, otherwise 0.
Firm age	Log value of firm age.
Gdp	Per capita GDP of each province.

provides a list of all used variables and their details.

4. Empirical Results and Analysis

4.1. Descriptive Statistics

Table 2. Descriptive statistics for the variables.

Variables	N	Mean	S.D.	Min	Median	Max
CSR	16,016	27.40	18.25	−2.870	22.61	76.63
EPU	16,016	26.81	19.31	2.183	20.63	86.25
EPU2	16,016	1092	1576	4.766	425.6	7440
ROE	16,016	0.0475	0.0551	−0.137	0.0421	0.219
LEV	16,016	0.423	0.216	0.0445	0.414	0.893
Firm size	16,016	21.98	1.279	19.60	21.80	25.89
Board	16,016	2.261	0.179	0	2.303	2.944
Share1	16,016	0.361	0.153	0.0893	0.344	0.760
Share10	16,016	0.593	0.162	0.221	0.605	0.966
Daul	16,016	0.254	0.435	0	0	1
Firm age	16,016	2.733	0.372	1.386	2.773	3.401
Gdp	16,016	10.25	0.701	7.833	10.23	11.30

presents the descriptive statistics of the key variables in our dataset. The mean value of CSR is 27.40, indicating that the overall level of CSR fulfillment is relatively low and still has great room for improvement. The minimum and maximum values of CSR are −2.870 and 76.63, respectively. The firms with the worst CSR have a negative score, while the firms with the best social responsibility performance have a score of 76.63. The standard deviation of CSR is 18.25, suggesting that the social responsibility performance of sample enterprises varies greatly. Since there are no widely accepted or mandatory CSR reporting standards in the Chinese institutional background, the capacity and willingness of Chinese companies regarding CSR implementation is still relatively limited. Some firms still neglect the significance of CSR and much disreputable information on CSR activities continues to be disseminated. The CSR practice of Chinese listed companies is still in its initial stage.

The average value and standard deviation of the provincial EPU are 26.81 and 19.31, respectively. This table shows that the EPU of 31 provinces demonstrates a huge difference, which could also be verified from the minimum value (2.183) and the maximum value (86.25). We still portray the trend of changes in EPU in 31 provinces (excluding Hong Kong, Macao, and Taiwan) from 2010 to 2016 in Figure 2. Figure 2 shows that there is a wide-ranging disparity in EPU across provinces. And the EPU of the 31 provinces has obvious dynamic characteristics that changed over time. The EPU index of Inner Mongolia, Heilongjiang Province Liaoning Province, Ningxia Hui Autonomous Region, Jiangxi Province, Zhejiang Province, Jiangsu Province, Shanghai, and Guangdong Province exhibits rapid growth, indicating that their economic policies have undergone major changes. Overall, the EPU index values of the northeast and eastern coastal areas are generally higher than those of western inland areas.

4.2. Baseline Results

Table 3. Effect of EPU on CSR.

	(1)	(2)	(3)
EPU	0.1237 ***	0.0918 ***	0.0918 ***
	(0.0259)	(0.0249)	(0.0291)
EPU2	−0.0017 ***	−0.0013 ***	−0.0013 ***
	(0.0003)	(0.0003)	(0.0003)
ROE		93.1019 ***	93.1019 ***
		(3.0728)	(4.6337)
LEV		−6.6288 ***	−6.6288 ***
		(1.1242)	(1.1673)
Firm size		5.1559 ***	5.1559 ***
		(0.3155)	(0.4288)
Board		2.8437 **	2.8437
		(1.2158)	(1.7160)
Share 1		7.3081 ***	7.3081 ***
		(2.3271)	(2.6702)
Share 10		−9.9885 ***	−9.9885 ***
		(1.5801)	(2.0401)
Daul		−0.8571 **	−0.8571 **
		(0.3765)	(0.3759)
Firm age		−9.3362 ***	−9.3362 ***
		(1.5814)	(3.2195)
Gdp		−4.6596 ***	−4.6596 ***
		(1.1987)	(1.7185)
_cons	25.9704 ***	−18.2146 *	−18.2146
	(0.3993)	(10.1295)	(13.2088)
Province Effect	YES	YES	YES
Industry Effect	YES	YES	YES
Firm Effect	YES	YES	YES
N	15,727	15,727	15,727
R2	0.628	0.673	0.673

* p < 0.1, ** p < 0.05, *** p < 0.01.

indicates the baseline regression results. Column (1) is the simplest case in which we do not control for firm-level characteristics and the data are without clustering standard errors. After controlling for fixed effects and firm-level characteristics, Columns (2) and (3) show the empirical results without and with clustering standard errors, respectively. The subsequent regression in this study uses Model (3).

Based on Model (3), it can be seen that the coefficients of EPU and EPU² are 0.0918 and −0.0013, all of which are significant at the 1% level, verifying the nonlinear relationship between EPU and CSR. It can be seen from the estimated coefficients of EPU and EPU² in Column (3) that the inflection point of the inverted U-shaped curve falls at 34.2039. It shows that the impact of EPU on CSR transitions from a positive effect to a negative effect as the level of EPU crosses the threshold value. Economic policy uncertainty is not necessarily a good thing; it creates both opportunities and risks. Governments and businesses should be wary of inflection points, which are the top of the inverted U-shaped curves and represent the highest level of CSR performance under uncertainty. This phenomenon also indirectly reflects the reasons for the low level of CSR in China. Overall, in all samples, observations less than the inflection point account for 73.60% of the sample size, which means that the impact of EPU becomes negative until the 74th percentiles. The possible explanation is that for most Chinese companies, the ability to obtain competitive superiority from CSR performance is relatively strong, leading to the benefits outweighing the costs. Therefore, the inflection point of the inverted U-shaped curve becomes too large and EPU positively influences CSR in a rather wider range.

To test whether the U-shaped relationship between provincial EPU and CSR is valid, this study uses the “utest” command developed by Haans for reference [57]. The U-test results are shown in

Table 4. Effect of EPU on CSR: U-test testing.

	Lower Bound	Upper Bound
Variables	EPU	EPU2
Interval	2.1832	86.2528
Slope	0.0859	−0.1397
t-value	2.8950	−4.1517
P > |t|	0.00131	0.00002
Extreme point: 34.2039 P > |t| = 0.00131
95% Fieller interv24.9616 [24.9616; 38.3575]

. The inflection point is about 34.2039 where the curve attains its maximum. The 95% confidence interval of the inflection point is [24.9616; 38.3575], suggesting that threshold value lies well within the data range and removing the doubts that the data only reveal one half of the curve. The slope can take negative values in the data interval and we reject the original hypothesis at the 1% statistical level. The inverted U-shaped relationship exists.

We still plot the relationship between EPU and CSR in Figure 3. When EPU is low and does not exceed the threshold value, the increased EPU promotes CSR; when EPU exceeds the threshold value, the increase in EPU actually suppresses CSR. One possible explanation is that the increasing rate of marginal social responsibility benefit is less than that of the marginal economic policy uncertainty cost. Therefore, Hypothesis 1 is validated.

4.3. Robustness Check

We conduct a wide range of robustness checks for baseline analysis. The results are shown in

Table 5. The robustness results.

	(1)	(2)	(3)	(4)	(5)	(6)
	CSR	CSR	RCSR	Huazheng ESG	Time Trend	Balanced Panel
EPU1	0.0146 ***
	(0.0046)
EPU12	−0.0001 ***
	(0.0001)
EPU2		7.1388 ***
		(2.2629)
EPU22		−1.1377 ***
		(0.3389)
EPU			0.0036 ***	0.0032 **	0.0587 **	0.1407 ***
			(0.0013)	(0.0014)	(0.0293)	(0.0384)
EPU2			−0.0001 ***	−0.0001 *	−0.0010 ***	−0.0020 ***
			(0.0001)	(0.0001)	(0.0003)	(0.0005)
Trend					−2.2323 ***
					(0.2766)
Controls	YES	YES	YES	YES		YES
Province Effect	YES	YES	YES	YES		YES
Industry Effect	YES	YES	YES	YES		YES
Firm Effect	YES	YES	YES	YES		YES
N	15,727	15,727	15,727	15,267		11,389
R2	0.673	0.673	0.582	0.600	0.676	0.658

* p < 0.1, ** p < 0.05, *** p < 0.01.

. First, we use several alternative proxies of EPU and CSR as the explanatory and dependent variables in baseline regression instead of the original ones. Alternative measures of EPU consist of the unstandardized annualized EPU index (EPU1) and the natural logarithm of the annualized EPU index (EPU2). We assign the corporate social responsibility rating (RCSR) levels of A, B, C, D, and E to 5, 4, 3, 2, and 1, respectively, for the alternative CSR measure. We also use the ESG rating from Huazheng rating agencies for an alternative CSR measure [58]. The results are shown in Columns (1)–(4). The coefficient of EPU is significantly positive at the 10% level, and the coefficient of EPU² is significantly negative at the 1% level. These are consistent with our main findings.

Second, drawing on the existing research [21], the study adds a linear time trend variable (Trend) into the regression to disentangle the time effect and policy uncertainty. The nonlinearity of the EPU–CSR relationship still holds true after including Trend in the Column (5).

Third, we use balanced panel data to examine the impact of EPU on the CSR of continuously existing enterprises. Column (6) indicates that EPU has a significantly inverted U-shaped relationship with firms’ CSR.

Finally, we use the instrumental variable method to further solve the missing variables and endogenous problems existing in the estimation. We use the previous year’ s EPU of the province where the enterprise was located as the instrumental variable of the current year’s EPU [27]. The 2SLS regression estimation results are listed in

Table 6. Instrumental variable regression results.

	First Stage	Second Stage
Variables	EPU	CSR
IV	0.0741 ***
	(0.0260)
EPU		0.1088
		(0.2605)
EPU2		−0.0084 **
		(0.0043)
Fixed Effects?	YES	YES
Clustering SE?	YES	YES
Observations	12,788	12,788
Cragg-Donald Wald F statistic)		31.683
Anderson canon. corr. LM statistic)		63.049
R2	0.4292	0.6820

** p < 0.05, *** p < 0.01.

. From the regression results of the first stage, there is a significant correlation between the instrumental variable and EPU. At the same time, all the F statistics are greater than 10, indicating that the underidentification test and weak instrumental variable test have been passed. Moreover, the coefficient of EPU and EPU² in the second stage regression is also consistent with that reported in our baseline regression in Table 3, confirming that the inverted-U relationship between EPU and CSR is still valid, thus also indicating that our findings are not driven by endogeneity.

5. Additional Analysis

In this section, we investigate the heterogeneity impacts of provincial EPU on CSR and test the moderating effects of financial constraints and government interventions.

5.1. Heterogeneity Analysis

Comprehensive information on firm characteristics enables us to disentangle the heterogeneity of estimated effects. This study disaggregates the sample by firm ownership, firm size, and firm life cycle to the uncertainty shock. Meanwhile, we use the Chow test to test the coefficient difference between groups after grouping regression. We present the results in

Table 7. The results of heterogeneity analysis.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	SOEs	Non-SOEs	Large Size	Small Size	Growth	Mature	Declining
EPU	0.1241 *	0.1016 ***	0.1322 **	0.0623 ***	0.1084 *	0.0669 *	0.1418 **
	(0.0622)	(0.0299)	(0.0510)	(0.0224)	(0.05783)	(0.0447)	(0.0606)
EPU2	−0.0017 **	−0.0015 ***	−0.0019 ***	−0.0010 ***	−0.0015 **	−0.0009 *	−0.0021 ***
	(0.0008)	(0.0004)	(0.0006)	(0.0003)	(0.0007)	(0.0005)	(0.0007)
Controls	YES	YES	YES	YES	YES	YES	YES
Province Effect	YES	YES	YES	YES	YES	YES	YES
Industry Effect	YES	YES	YES	YES	YES	YES	YES
Firm Effect	YES	YES	YES	YES	YES	YES	YES
N	6322	9352	7624	7726	5393	4588	3851
R2	0.6674	0.6795	0.6495	0.7126	0.7265	0.7309	0.7566
Difference	F(108, 15,791) = 5.69 Prob > F = 0.0000		F(114, 15,785) = 5.13 Prob > F = 0.0000		F(228, 15,671) = 1.50 Prob > F = 0.0000

* p < 0.1, ** p < 0.05, *** p < 0.01.

and focus on the differences in the inflection points and vertices of inverted U-shaped curves in different contexts. The inflection point of the inverted U-shaped curve becomes too large, and economic policy uncertainty influences corporate social responsibility positively in a rather wider range.

5.1.1. Firm Ownership

Prior studies demonstrate that firms possessing political advantages have strong incentives to meet the needs of stakeholders. Does EPU have a different impact on CSR depending on the form of ownership? The study further divides the total sample into state-owned firms (SOEs) and other firms (non-SOEs). We assign the dummy variable value of 1 if the enterprise is an SOE and 0 otherwise. We then examine the heterogeneous effects of provincial EPU on CSR fulfillment under different firm ownership structures using OLS. As shown in Columns (1)–(2) of

Table 8. The results of the moderating effect.

	(1)	(2)
	Financial Constraint	Government Intervention
EPU	0.0959 ***	−0.6235 ***
	(0.0293)	(0.0.1382)
EPU2	−0.0014 ***	0.0079 ***
	(0.0003)	(0.0019)
EPU ∗ FC	−0.2180
	(0.1475)
EPU2 ∗ FC	0.0029 *
	(0.0016)
FC	−2.7227
	(3.9158)
EPU ∗ Gov		0.0762 ***
		(0.1383)
EPU2 ∗ Gov		−0.0010 ***
		(0.0002)
Gov		−2.2989 ***
		(0.1382)
Province Effect	YES	YES
Industry Effect	YES	YES
Firm Effect	YES	YES
N	15,727	15,727
R2	0.6735	0.6770

* p < 0.1, *** p < 0.01.

, the coefficient of EPU² is significantly negative in the sample of SOEs and non-SOEs and their relationships pass the U-test. It indicates that the inverted U-shaped relationship between economic policy uncertainty and corporate social responsibility exists for both SOEs and non-SOEs. However, the results of the “Chow test” for the differences between SOEs and non-SOEs indicate that the impact is more potent for firms that are SOEs.

Compare the two curves shown in Figure 4. The overall trends are relatively consistent, CSR first increases and then decreases. However, in terms of the inflection point of the two groups, the difference is significant. For the non-state-owned firm group, the inflection point is about 34.5422, while for the state-owned firm group, the inflection point falls at 37.4372, which means that the inflection point of non-state-owned firms arrives earlier than that of the state-owned firms. Specifically, for the state-owned firms, economic policy uncertainty can promote social responsibility performance with a wider uncertainty range compared to non-state-owned firms, while economic policy uncertainty inhibits their social responsibility performance with a narrower uncertainty range compared to the non-state-owned firms, meaning that state-owned firms can withstand higher EPU. From the graph, it can also be seen that at lower EPU, there is a clear “scissor fork”. That is, at the beginning, as EPU increases, the social responsibility performance of non-state-owned firms is better than that of state-owned firms; after the critical value is exceeded, the curve of state-owned firms is overall above that of non-state-owned firms.

Differences in ownership types lead to different attitudes towards economic policy uncertainty. Specifically, in China’s special economic institutional environment, when economic policy uncertainty increases, the task of maintaining stability and development for the government becomes more urgent. Typically, the government is the major shareholder of state-owned firms and holds abundant social resources [59]. State-owned firms are the important material and political foundation of socialism with Chinese characteristics, the fulfillment of social responsibility is an intrinsic attribute, and they need to play a leading role. However, there is greater discretion in whether and to what extent non-state-owned firms fulfill their social responsibility. They are more vulnerable to the negative effects of economic policy uncertainty and then reducing socially responsible investment.

5.1.2. Firm Size

Firms’ business decisions may be influenced by the scale of resources. This study calculates the median firm size by year and industry and further divides enterprises into large and small size groups. As shown in Columns (3)–(4), Table 8, the coefficients of EPU² are all significantly negative in the two subsamples and their relationships pass the U-test, indicating that the inverted U-shaped relationship between economic policy uncertainty and corporate social responsibility exists for both large-sized and small-sized firms. However, the results of the “Chow test” for the differences between large-sized and small-sized firms indicate that the impact is more potent for firms that are large-sized firms.

For the large-sized firms group, the inflection point is about 34.7453, while for small-sized firms group, the inflection point falls at 31.7857, which means that the inflection point of small-sized firms arrives earlier than that of large-sized firms. Therefore, economic policy uncertainty influences corporate social responsibility for large-sized firms positively in a rather wider range. As can be also seen in Figure 5, in the large-sized group, the curve vertex is higher. Well-funded enterprises can effectively allocate resources to improve CSR performance and send positive signals. In other words, small-sized firms have been under more economic pressure for a long time. With increased EPU, these smaller enterprises turn to reducing CSR investment earlier. The phenomenon is in line with our expectations. The firm size represents the abundance of resources. Large-sized firms have sufficient capital and a strong ability to resist risks and to meet the needs of various stakeholders. The volatile economic environment adds further difficulties to the realization of small firms’ investment and financing objectives. In order to effectively mitigate the potentially adverse effects of policy uncertainty, small firms are rapidly reducing their socially responsible investments. The empirical results regarding firm size and CSR in Table 3 also underscore the significance of economic strength in the context of uncertainty.

5.1.3. Firm Life Cycle

Depending on where a firm is in its life cycle, its structures, capabilities, strengths, strategies, and cash flow will vary. Faff et al. examined the importance of the life cycle in the determinant of corporate policies and found that different life cycles influence resource allocation [60]. In recessions, leverage tends to be higher than in other stages, and risk aversion increases significantly, which further affects corporate investment behaviors [61]. In line with life-cycle theory, we argue that the influence of economic policy uncertainty on CSR may be different in the life-cycle stage.

We further empirically test this conjecture. We adopt the coding method proposed by Dickinson and combine three types of cash flow information characteristics (positive or negative) to define the life cycle [62]. Considering that the quantity of listed companies in the introduction stage is relatively small, they are classified as the growth stage. A dummy variable is equal to 1, 2, or 3, respectively, if the firm is in the growth, mature, or declining life-cycle stage. The group regression results are shown in Columns (5)–(7) of Table 8. The results show that provincial EPU always has a nonlinear effect on CSR in any life-cycle stage. However, the results of the “Chow test” for the differences indicate that the impact is more potent for firms that are in the mature stage. Furthermore, from the coefficients of EPU and EPU², we can calculate that the inflection point is 36.1333, 36.1666, and 33.7619, respectively. The top of the inverted U-shaped curve represents the best level of CSR while under uncertainty. As can be also seen in Figure 6, the impact of EPU on CSR will be steeper when the firms are in the mature stage. A possible reason for this phenomenon is that firms in the mature stage are well capitalized and have a greater demand for maintaining their brand image, which compels them to engage more in CSR activities.

5.2. The Moderating Effect

We further draw on the moderating effect testing method developed by Edwards and Lambert for corresponding validation [63]. The main indicator for determining the existence of a moderating effect is the interaction term between the quadratic term of the explanatory variable and the moderating variable. We present the results in Table 8. Columns (1)–(2) show the moderating effect of financial constraints and government interventions on the U-shaped relationship, respectively.

5.2.1. Financial Constraints

Financial constraints (FC) largely determine the difficulty for companies to obtain funds from the external economic environment, thus affecting the firm’s operational activities. In general, firms with lower financial constraints have more access to discretionary funds. CSR performance can alleviate financial constraints [64]. Hadlock and Pierce proposed the SA index [65], arguing that the size and age of an enterprise greatly determine an enterprise’s financial constraints. A higher SA value means that financial constraint is more pronounced.

The coefficient of the interaction term (EPU² ∗ FC) in Column (1) is positive and statistically significant at the 1% level, stating that financial constraints do have a moderating effect on the U-shaped relationship. Compare the two curves shown in Figure 7. The financial constraints weaken the inverted U-shaped relationship between EPU and CSR.

5.2.2. Government Interventions

Government interventions (Gov), as a component of the institutional environment, can have an important impact on firms’ behaviors. We use the relationship between the government and the market to measure regional government intervention level, and the data are from “China’s Provincial Marketization Index Report” (https://cmi.ssap.com.cn (accessed on 21 November 2023)). The higher score of the relationship between the government and market, the lower degree of government intervention in the region, and the higher proportion of market allocation of economic resources. Markets at different degrees of intervention may have different responses to economic policy uncertainty.

The estimated coefficient of EPU² is 0.0079 (p < 0.001) in Column (2), stating that the nonlinear relationship between economic policy uncertainty and CSR remains after the inclusion of government intervention. And, the coefficient of the interaction term (EPU ∗ Gov) is significantly positive, and the coefficient of the interaction term (EPU² ∗ Gov) is significantly negative. It indicates that the government interventions have a negatively moderating effect on the inverted U-shaped relationship. When the firm is located in regions with more government interventions, as shown in Figure 8, the inverted U-shaped relationship between economic policy uncertainty and corporate social responsibility is significantly weakened until the positive U-shaped relationship appears. Therefore, in areas with a high degree of government intervention, when economic policy uncertainty increases, local governments have the ability to influence firms’ decisions for social responsibilities and provide implicit and explicit protection. The above analysis reveals the phenomenon of the government’s “visible hand” in China.

6. Conclusions and Discussion

Corporate social responsibility is vital to both the society and firms, but it may fluctuate due to uncertainty. Using China’s listed firms as the research setting, this study explores the relationship between provincial economic policy uncertainty and corporate social responsibility. We find an inverted U-shaped relationship between the two variables. Specifically, when EPU is low and does not exceed the threshold, the increase in EPU helps to promote CSR; when EPU exceeds the threshold, the increase in EPU suppresses CSR. Higher economic policy uncertainty inhibits CSR behavior, which explains why the overall level of social responsibility is low in China. The inverted U-shaped relationships are robust for alternative measures of variables, adding a time trend variable, endogeneity problems, and balanced panel data. We further examine this relationship under different firm characteristics. In terms of ownership structure, the inverted U-shaped relationships between EPU and CSR exist for both SOEs and non-SOEs, but the inflection point of non-state-owned firms arrives earlier. Therefore, economic policy uncertainty influences corporate social responsibility for state-owned firms positively in a rather wider range. In terms of firm size, the nonlinear effects of economic policy uncertainty on corporate social responsibility are statistically significant in both large-sized firms and small-sized firms, but the effects are stronger in large-sized firms. In terms of corporate life cycle, the firms are in the mature stage which engage more in CSR activities. In addition, our moderating effect analysis shows that firms’ financial constraints and government interventions weaken the inverted-U relationship.

Based on the aforementioned research findings, this study proposes the following policy insights from the perspectives of government and enterprises.

Firstly, the above research conclusions show that economic policy uncertainty is a “double-edged sword” for CSR. At a moderate level of EPU, its escalation helps to promote CSR engagement, and it should also be noted that, if policy adjustments are too frequent and the level of EPU is too high, its escalation reduces CSR investment. Neither invariable nor excessive economic policy changing is beneficial for firms to engage in CSR activities. Thus, the government should enhance the visibility of its macroeconomic policies and the fluidity of the economic policy transition to avoid the adverse effect of social responsibility default due to frequent changes in economic policies in times of economic weakness. In other words, providing stable economic circumstances can benefit corporate social responsibility.

Secondly, the impact of EPU on CSR is heterogeneous in firms with different characteristics, and financial constraints and government interventions have an interaction effect on the relationship between EPU and CSR. Therefore, macroeconomic decision-making departments should holistically consider the current business environment and corporate characteristics and implement different measures to deal with EPU to promote CSR and avoid the policy misunderstanding of “one size fits all”. Firms should develop effective and well-prepared plans to reduce the risk of uncertainty based on their own characteristics.

Thirdly, firms should dialectically view the risks and opportunities brought by EPU and strategically adjust CSR investment decisions. Managers should have a profound understanding of the core value of corporate social responsibility and should also be aware of the importance of resources in the relationship between EPU and CSR. For instance, managers may preempt their opponents’ moves and “go steady and go far” as soon as policies change. Instead, when EPU is extremely high, managers need to be vigilant because its negative impact on firm operations may far outweigh its potential benefits, and under such circumstances, a conservative strategy may be more appropriate. At the same time, firms are better able to meet shareholders’ needs and discover new market opportunities in an uncertain environment by promoting their business capacity.

7. Limitations and Future Directions

Our study has several limitations, which also provide several interesting ideas for future research. Firstly, our sample is confined to publicly listed firms in China, potentially restricting the applicability of our findings. Subsequent research scenarios can be shifted to other developing countries, especially developing countries with poor economic policy stability or regional heterogeneity. Secondly, our study’s time frame spans from 2010 to 2016 due to data constraints. Therefore, more data from longer time horizons should be supplemented in future research. Such investigations can better reveal the EPU–CSR linkage and help to verify the conclusions drawn in our research. Finally, our focus is primarily on the direct impact and heterogeneous effects of provincial EPU on CSR, with a limited exploration of the underlying mechanisms linking EPU to CSR. Subsequent studies should introduce intermediary variables to augment our theoretical framework and unravel the intricate pathways through which EPU influences CSR practices.