The philanthropic response of substantive and symbolic corporate social responsibility strategies to COVID-19 crisis: Evidence from China.

The COVID-19 epidemic broke out in China in January 2020, which triggered the largest wave of corporate philanthropic donations since the 2008 Sichuan earthquake. Based on A-share listed firms in the Shanghai and Shenzhen stock exchanges in 2020, we study whether substantive and symbolic corporate social responsibility (CSR) strategies affect corporate philanthropic responses during the COVID-19 crisis. We use the lagged annual donation and technical dimension scores (T scores) of rankins ratings (RKS) as proxies of CSR performance and CSR disclosure and then define the CSR gap as the gap between the two. The results show that substantive and symbolic strategies cause firms to have material differential responses in the COVID-19 crisis. Specifically, the CSR gap is negatively related to the possibility and the level of crisis donation. In addition, (1) this difference is more pronounced in the earlier period of the COVID-19 crisis; (2) the negative correlation is more pronounced in private firms; and (3) the crisis donation of firms with either strategy obtains no different response from the capital market. Our evidence suggests that the established CSR strategy influences the substantive response of Chinese firms to public emergencies, but their substantive response does not result in different reactions in the capital market.

INTRODUCTION

The sudden outbreak of the COVID‐19 epidemic in January 2020 caused a severe humanitarian crisis in Chinese regions, especially Wuhan. To help alleviate the crisis, a large number of firms made cash and in‐kind donations. According to SynTao's report, approximately 780 firms announced cash and in‐kind donation to the Wuhan area within 10 days (as of January 29) after the epidemic outbreak. The cash donation amounted to 8.11 billion RMB, and the value of the in‐kind donation reached 1.68 billion RMB, making this the largest wave of corporate philanthropy since the 2008 Sichuan earthquake.

This COVID‐19 crisis brings forth two important yet unexplored research questions. First, are there any material differences in the philanthropic response to the COVID‐19 crisis between Chinese firms that implement substantive and symbolic corporate social responsibility (CSR) strategies? Second, does the Chinese capital market react differently to philanthropic responses under substantive and symbolic strategies? In this study, we attempt to contribute toward the ongoing literature on COVID‐19 crises by answering the above two potential research questions.

In relevance to the first question, prior literature highlights that firms adopt two types of CSR strategies: substantive and symbolic strategies to maintain legality (Hyatt & Berente, 2017; Marquis & Qian, 2014; Perez‐Batres et al., 2012). Specifically, it is not clear whether firms that implement symbolic strategies respond timely and adequately for impression management in the COVID‐19 crisis. One possible argument can be that such firms were unable to appropriately respond to the crisis because of the lack of a substantive response mechanism. To the best of our knowledge, existing research lacks evidence on the above question. For example, in the analysis of the donation wave in the 2008 Sichuan earthquake, studies (e.g., Jia & Zhang, 2013a; Zhang et al., 2010, among others) did not consider the role of established substantive and symbolic CSR strategies in the firms' response to disasters. Moreover, previous research on events such as the 2008 Sichuan earthquake and the Sanlu milk crisis (Kong, 2012) was conducted over a decade ago, when the development of Chinese CSR regulation was just at the initial stage. This study is based on the current COVID‐19 crisis, which has occurred under the new regulatory environment (e.g., the new corporate governance guidance launched in 2018), and can thus provide a more meaningful reference for managers and academics, enabling them to obtain a better understanding of the current development of Chinese CSR regulation. Consequently, while investigating the first research question we contribute by providing new insights, on the philanthropic activities of Chinese listed firms in public emergencies for example, the COVID‐19 crisis, for the practitioners and academicians.

In relevance to the second research question, it is important to comprehend the “CSR‐Firm value” relationship in the Chinese capital market. In particular, if the market has an aversive attitude toward greenwashing activities, then it is very likely that only CSR activities under substantive strategies will enhance corporate value. The evidence provided by García‐Sánchez et al. (2021) supports this view in mature markets to some extent; only firms with a lower CSR gap between disclosure and performance obtain stronger financial access. However, prior literature lacks sufficient evidence from the Chinese market. Du (2015a) emphasized that the Chinese market punishes the decoupling of corporate environmental responsibility (green washing), but their evidence only involves exposed activities that have been reported on news media and lacks further support for the Chinese market's aversive attitude to symbolic strategy. Consequently, this study contributes by providing a meaningful reference for understanding the Chinese “CSR‐Firm value” relationship. Further, we enrich the understanding of the role of CSR in the corporate's fight against the risk of the COVID‐19 crisis. While extending the relevant literature (see e.g., Bose et al., 2021; Cheema‐Fox et al., 2020; Huang et al., 2020), our original findings reveal that substantive CSR activities help firms obtain higher legitimacy support. In particular, no matter what strategy it is based on, philanthropic donation in the crisis helps firms obtain higher abnormal returns, which supports substantive CSR activities' positive effects on helping firms combat the risks caused by the COVID‐19 epidemic in some ways.

Our sample comprises of Chinese firms listed on both Shanghai and Shenzhen Stock Exchanges in 2020. Further, referring to García‐Sánchez et al. (2021), we distinguish CSR strategy and the level of implementation according to the difference between CSR performance and disclosure. Then, we observe that there are material differences in the philanthropic response of firms that implemented substantive and symbolic CSR strategies, and the differences are even more pronounced in private firms. Furthermore, we follow the short‐term event study method of Krüger (2015) to observe whether the Chinese capital market has different feedback on the philanthropic donations of substantive and symbolic firms. We find that the capital market has material positive feedback on the COVID‐19 donation of substantive and symbolic firms, and the feedback for them shows no material difference.

The rest of the study is organized as follows. The second section discusses the institutional background in China. Section 3 outlines the relevant literature. Section 4 presents the hypotheses of the study. Section 5 discusses the research design and section 6 provides detailed empirical findings. Section 7 concludes the paper.

INSTITUTIONAL BACKGROUND

Development of CSR system in China

In the early Chinese market, CSR pressures mainly come from overseas stakeholders, such as international buyers and consumers and foreign investors. (Shahab et al., 2018; Shahab et al., 2020; Zhu & Zhang, 2015). However, these pressures are easy to relieve. For example, it is difficult for foreign investors to thoroughly understand a firm's CSR performance in China. Chinese firms can easily avoid these pressures by “purchasing” SA8000, ISO26000, and other qualification certifications (Boiral et al., 2017). In 2009, the development of the Chinese CSR regulation system entered a “semi mandatory” era. Regulators force important large firms (Shanghai and Shenzhen 100 Index, overseas listed firms and so on) to disclose CSR information. However, there are no specific requirements for the disclosure scope (e.g., lack of detailed reporting guidelines) and information quality (e.g., lack of mandatory assurance). The regulation system has not been substantively improved for a long time in recent years. For example, in the latest “Guidelines for Disclosure of Environmental, Social, and Corporate Governance of Listed Companies” released by the Shanghai stock exchange in 2018, yet there were still no requirements for specific information scopes or “comply or explain.” Firms still had considerable freedom in the breadth, depth, and quality of their CSR disclosures.

Although the “top‐down” development of the Chinese CSR system has stagnated, with the development of nonprofit organizations (NGOs) and socially responsible investors focusing on the CSR issue, the accountability pressure of Chinese citizens for CSR has been increasing. For example, on the eve of Xiaomi's IPO in Hong Kong in May 2018, NGOs such as friends of nature and the institute of public and environmental affairs reported the illegal pollution of Xiaomi's manufacturing subcontractor, Ichia Technologies, to the Hong Kong stock exchange, forcing Xiaomi to spend heavy cost on this environmental risk exposure from its supply chain. According to a questionnaire survey of Chinese public investors by Syntao (2020), 86% of respondents responded that they consider environmental protection, labor rights, and business ethics when investing firms.

Under the new circumstance, Chinese firms need to adopt a more cautious attitude and more sophisticated means to cope with CSR pressures. On the one hand, some firms choose to strengthen impression management, such as implementing detailed and rigorous information disclosure and comprehensive practices involving CSR compliance issues. According to GoldenBee (2019), the proportion of Chinese firms using the global reporting initiative standards in their CSR reports increased from 16.09% in 2010 to 26.02% in 2019, and the percentage of firms implementing the ISO2600 standard increased from 2.63% to 10.52% at the same time. On the other hand, some firms choose to strengthen substantive input, such as increasing social welfare spending. According to the report of the China charity alliance (CCA, 2019), Chinese firms donated as much as 93.147 billion RMB in 2019, accounting for 61.71% of the total amount of social donations.

COVID‐19 epidemic and corporate donation wave

The disease first broke out in Wuhan in mid‐January 2020, and Chinese authorities confirmed person‐to‐person transmission on January 20, 2020. To prevent the spread of the epidemic, the Wuhan local government began a traffic blockade on January 23, and then the scope of the blockade was gradually expanded. Hubei Province launched a major “public health emergency level I" response on January 24. By January 29, all provinces across the country had achieved a Level I response. Due to the Chinese government's timely control measures after the outbreak, China effectively controlled the epidemic within 1 month. On February 21, the Standing Committee of the Political Bureau of the Central Committee announced that prevention and control work had achieved initial results, with the number of newly confirmed and suspected cases across the country generally showing a downward trend. At the same time, it sent a signal to speed up the resumption of work and production, indicating a turning point of the Chinese epidemic disaster. On February 24, the World Health Organization declared the peak of the epidemic had passed in China. Subsequently, Chinese provinces began to gradually implement the deregulation, and Hubei became the last province to lift the provincial blockade on March 25.

Due to the high transmissibility of the COVID‐19 virus, in the early stage of the outbreak, several cities represented by Wuhan Hubei began to face serious problems of medical and living supplies. For example, on January 25, the Hubei Red Cross Society announced to the media that there was a huge shortage of medical materials in Hubei Province. In response, Chinese firms set off the largest wave of corporate philanthropic donations since the 2008 Sichuan earthquake. According to SynTao, in the early stage of the outbreak (as of January 29), approximately 780 firms had announced cash and in‐kind donation, with cash donations amounting to 8.11 billion RMB, and the value of the in‐kind donations reached 1.68 billion RMB. On March 7, the total cash donations amounted to approximately 22.5 billion RMB, the converted value of in‐kind donations reached 7.8 billion RMB, and the total philanthropic amount reached 30.3 billion RMB, which is nearly half of the 66 billion RMB that was donated a year after the 2008 Sichuan earthquake.

According to SynTao, cash donation is still the most important way in this philanthropic wave, with 68% of firms choosing it. Private firms account for 57.9% of the total donation, while state‐owned firms account for 28.1% of the total donation. Corporate philanthropic donation shows strong industrial characteristics. For example, the pharmaceutical and biological industries are more inclined to donate medical supplies, and the financial industry is more inclined to donate financial products such as insurance. Finally, we find that there is also a huge gap in the response speed of firms. Our samples show that the earliest and latest firm's first donation dates are January 20 and March 2, respectively, with a difference of 43 days, and the median date of donation is January 29.

Figure 1 records the timeline of the major events involved in our research.

FIGURE 1

The timeline of the major event

LITERATURE REVIEW

Substantive and symbolic CSR strategies

Scholars (Hyatt & Berente, 2017; Marquis & Qian, 2014; Schons & Steinmeier, 2016) have defined substantive strategies as “high priority assigned to criteria when making decisions”, which is related to the concepts of “proactive commitment” and “internal actions” (Hawn & Ioannou, 2016); meanwhile, the definition of symbolic strategies pertains to the strategy used to enhance reputation”, which is related to the concepts of “impression management” and “reactive commitment”.

In practice, symbolic firms are more inclined to use impression management in information disclosure to disguise their actual CSR performance levels than substantive firms, which is one way of “greenwashing” (Boiral et al., 2017; Delmas & Burbano, 2011; Lyon & Montgomery, 2015; Pope & Wæraas, 2016). Another related concept is CSR decoupling (García‐Sánchez et al., 2021; Meyer & Rowan, 1977; Sauerwald & Su, 2019), which refers to the gap between CSR disclosure and CSR performance. A firm that is highly involved in symbolic strategy should have a larger positive CSR decoupling level between its CSR disclosure and performance. Discussion on substantive and symbolic strategies also frequently appears in the research on corporate environmental responsibility, especially on the topic of the “greenwashing” issue (Bazillier & Vauday, 2009; Delmas & Burbano, 2011; Furlow, 2010; Laufer, 2003; Parguel et al., 2011; Ramus & Montiel, 2005; Testa et al., 2018; Torelli et al., 2019). These studies reveal that if firms with ‘symbolic strategy success’ avoid or reduce the impact of social and environmental risks, then it can make practical social and environmental inputs (which are costly) meaningless. Eventually, it can hurt firms that truly invest and have social and environmental advantages. Kassinis and Panayiotou (2017) provide an example: after the 2010 Deepwater Horizon disaster, BP successfully used impression management based on visual tools, reducing the negative influence when decoupling was exposed.

However, a symbolic strategy is not always successful. It depends on the idiosyncrasies of the market, for example, discrimination ability of participants and level of information barriers. When the level of information barriers is low and market participants have strong discrimination ability, the effect of the symbolic strategy is reduced. Walker and Wan (2012) find that although substantive environmental activities have no correlation with corporate financial performance, symbolic activities have a material negative correlation with performance. García‐Sánchez et al. (2021) further shows that symbolic strategy is not conducive to corporate financing; specifically, CSR decoupling increases analyst error and financing costs and reduces access to finance. In addition to direct harm, symbolic strategies may also bring indirect risks. Berrone et al. (2017) show that greenwashing has certain adverse effects. When facing deterioration in environmental performance, greenwashing firms face stricter scrutiny by nongovernmental organizations than other firms.

We argue that majority of the existing literature is from mature capital markets, especially from Europe and the United States. These markets have formed a mature stakeholder protection system, including a stable socially responsible investment market and public opinion supervision system. The cross‐border research evidence of Dhaliwal et al. (2014) shows that the effect of CSR disclosure on financing cost reduction is not consistent in different markets and this effect is more obvious in the stakeholder orientation and more financially opaque markets. The Chinese market lacks direct evidence on the effects of their strategies. Du (2015a) finds that exposure to greenwashing would cause a material decline in the market value of firms. However, it is still unknown whether market participants have the ability to discriminate unexposed greenwashing under a symbolic strategy. Du et al. (2018) further find that auditors will refer to corporate environmental performance when issuing audit opinions, but when firms are involved in greenwashing, the negative correlation between environmental performance and nonstandard opinions is weakened. We consider that auditors are professional stakeholders, and whether other unprofessional stakeholders have the ability to discriminate symbolic activities is unknown. Moreover, the above research evidence is limited to the environmental dimension of CSR and lacks support for the broader dimension.

Corporate donation and public emergencies

Carroll (1999) argues that CSR has economic, legal, ethical, and discretionary components. Following primary insights (Brammer & Millington, 2008; Gautier & Pache, 2015; Godfrey, 2005; Porter & Kramer, 2002), the most important element of the discretionary component is a corporate donation. In China, corporate donation also plays the most crucial role in CSR practices and has become a popular means for firms to manifest their social responsibility concerns. For instance, Lin et al. (2015) find that Chinese firms use philanthropic donations as the main way to respond to an abrupt termination of political connections. Du (2015b) finds that Chinese family‐owned firms use philanthropic donations to cover up for poor environmental performance.

When a major public emergency occurs, the philanthropic donation is an important way for firms to quickly and effectively meet social legitimacy expectations. The existing literature on the Chinese market shows that the 2008 Sichuan earthquake triggered a huge wave of corporate philanthropic donations before the COVID‐19 crisis. Based on this philanthropic wave, research mainly analyzes institutional economics theory. For example, Zhang et al. (2010) and Hoi et al. (2020) provide evidence that private firms have a higher probability and level of corporate philanthropic donations for an earthquake. Jia and Zhang (2013a) further find that private firms with higher levels of managerial shareholding are more inclined to donate for the Sichuan earthquake and board size and political connections can strengthen this positive nexus. Gao and Hafsi (2017) show that firms that rely more on government support are more likely to donate after an earthquake, while firms that are more scrutinized by the government and the public, as well as firms with more resources, are more likely to donate more. Jia and Zhang (2013b) focus on the influence of female directors in this philanthropic wave. They find that when a board has at least three women, the philanthropic response of their firm is more pronounced. In terms of economic consequences, relevant conclusions support the positive correlation between the 2008 Sichuan earthquake donation and corporate performance. Song et al. (2012) and Hoi et al. (2020) both support the evidence that disaster donation brings stronger abnormal returns. Gao et al. (2012) further show that donating firms that directly provide products or services to final consumers receive higher abnormal stock returns.

However, we argue that the 2008 Sichuan earthquake occurred more than a decade ago. At that time, China had not carried out important sustainable reforms, including mandatory CSR reporting systems and green credit support systems. In the past decade, the Chinese capital market has made much progress in green and sustainable development. Moreover, historical studies do not take the different effects that symbolic and substantive strategies may have on a firm's response to public emergencies into account.

HYPOTHESES DEVELOPMENT

In relevance to the first research question, legitimacy theory presents that the survival and development of a firm must be consistent with public expectations of its legitimacy (Dowling & Pfeffer, 1975; Mahadeo et al., 2011; Milne & Patten, 2002; Patten, 1992; Suchman, 1995). Under the framework of legitimacy theory, to meet public concerns about environmental and social issues, firms need to maintain their legitimacy through CSR activities. The legitimacy framework is also applicable in the Chinese market. Dai et al. (2018) find that Chinese firms that issue high‐quality CSR reports have a higher level of government and media legitimacy, which ultimately translates to higher financial performance.

When a major public disaster breaks out and the resources controlled by the government cannot meet the needs of the public, the publics expectation may transfer to firms that are considered to have resource advantages, and the public may transfer this expectation into stronger legitimacy pressure on firms. At this time, some firms choose philanthropic donations to respond to incremental pressure. However, symbolic and substantive CSR strategies can have impacts on the response's decision‐making and process.

First, there are impacts on response range. The original intention of symbolic strategy is to make firms reduce costs as much as possible in CSR activities. Therefore, after major disasters, the institutional inertia brought by symbolic strategy makes it more difficult for the actual controllers of firms to accept substantive philanthropic donation, which needs more cost. Even if they realize that substantive philanthropic donation at this time can make greater benefits on impression management, it is difficult for them to make a sufficient philanthropic response like substantive firms. Tan and Tang (2014) find that there is an effect of institutional inertia on the philanthropic donation of the 2008 Sichuan earthquake: firms with a lower degree of separation of voting rights and cash flow rights are more reluctant to donate. They argue that controlling shareholders stingily use the resources under their control for disaster donation; even though there is extensive evidence that disaster donation can bring greater legitimacy benefits to their firms.

Second, there are impacts on response speed. Symbolic strategy requires firms to focus on nonsubstantive activities. When disasters occur, they lack the basis for substantive activities. For example, they lack a rapid communication mechanism with charitable organizations and lack experience liaising with local government departments in severely crisis areas to carry out substantive activities. This leads to the difference in response speed between symbolic and substantive firms. Accordingly, we propose the first hypothesis as:

Firms implementing symbolic and substantive CSR strategies have material differences in the philanthropic response to the COVID‐19 crisis.

A large number of studies suggest that firm ownership has an important influence on CSR activities in China (Byun et al., 2017; Hoi et al., 2020; Li et al., 2013; Li & Lu, 2015; Li & Zhang, 2010; Tang et al., 2018; Xu et al., 2015). Zhang et al. (2010) find that private firms show a stronger disaster philanthropic response in the 2008 Sichuan earthquake. Ownership may also influence the response under different strategies in the COVID‐19 crisis. First, private firms tend to have higher governance efficiency (Estrin and Pérotin, 1991; Li et al., 2020; Zhang et al., 2001), which provides a stronger basis for the implementation of substantive strategies, enables substantive private firms to respond to disasters greatly, and increases the crisis response gap between symbolic strategies and substantive strategies. Second, compared with state‐owned firms, Chinese private firms face greater competitive pressure (Ahlstrom & Bruton, 2001) and have a stronger motivation to build a reputation and political connections through CSR activities (Lin et al., 2015; Zhang et al., 2010). Therefore, stronger motivation makes private firms that implement substantive strategies increase their response to the COVID‐19 crisis. Consequently, we propose the second hypothesis as:

There is a greater difference in philanthropic response to the COVID‐19 crisis in private firms with symbolic and substantive strategies.

In relevance to the second research question, the discrimination ability of market participants for activities under symbolic strategy is closely related to the current development level of CSR information systems. Prior studies (Tashman et al., 2019; Yu et al., 2020) find that the market systems of different countries have an important influence on activities under symbolic strategy. When the CSR information system is under higher development, for example, information users can more accurately evaluate the actual CSR performance with a good CSR assurance system (Ackers & Eccles, 2015); meanwhile, symbolic firms may have difficulty extracting legal benefits in impression management from information activities (García‐Sánchez et al., 2021; Walker & Wan, 2012). In contrast, when the information system is underdeveloped, the symbolic strategy may be more effective and lower‐cost suitable. At this time, it is difficult for information users to evaluate the firm's actual CSR performance, and “good talking” firms with insufficient actual CSR investment may be deemed to have a “good walk” (Schons & Steinmeier, 2016).

In the recent Chinese market, although the CSR information system has experienced a long period of stagnation, market reforms based on the concept of green and sustainable development are still ongoing along with the development of government and citizen awareness, including socially responsible institutional investors, green finance credit policies, etc. (Du et al., 2015; Liu et al., 2019; Xing et al., 2020; Xu & Li, 2020; Ye & Zhang, 2011; Zhang et al., 2011). Several studies provide evidence that some participants in the Chinese market have differential responses to certain symbolic activities (Du, 2015a; Du et al., 2018). However, the Chinese market still lacks direct supporting evidence for the ability to discriminate CSR activities under substantive and symbolic strategies. It is hard to determine whether the response of substantive and symbolic corporate philanthropic donation in the COVID‐19 crisis may cause different consequences in the market. Accordingly, we propose a group of adversarial hypotheses:

The market has a differential response to the crisis donation of firms that implement symbolic and substantive CSR strategies.

The market has no differential response to the crisis donation of firms that implement symbolic and substantive CSR strategies.

Figure 2 shows the main theoretical framework in our research:

FIGURE 2

Theoretical framework

RESEARCH DESIGN

Data and sample selection

The original sample is Hushen A‐share listed firms in 2020. The illustration of COVID‐19 donation data is as follows: (1) Donation amount. The original data come from the records by SynTao's report and the CSMAR Public Health Emergency of International Concern (PHEIC) database. Among them, SynTao's records include the COVID‐19 donation as of January 29, and CSMAR PHEIC's record deadline is March 20, with a total of 1884 corporate donations recorded. These firms include listed and unlisted firms in A‐share and other markets. There are 38 donating group firms making donations in the name of firm groups, so it is difficult to determine the proportion of donations distributed among the listed subsidiaries. Therefore, we eliminated the 203 listed subsidiaries involved in these 38 group firms. For in‐kind donations that do not indicate the value amount, we manually convert them based on fair value. In addition, we also check the donation data of SynTao and CSMAR PHEIC with online news item by item to correct errors. (2) Donation date. These data come from our manual collation of online news. If the relevant news report does not mention the firm's specific donation date, we use the earliest news report day as the donation date. The sources of other data are as follows: capital market data and firm characteristics data are from CSMAR, and rankins ratings (RKS) data are from the Rankins database.

Our sample screening process is as follows. First, we eliminate firms from financial industry and firms‐year observations with missing data. Second, to avoid the impact of extreme values, we winsorize all the continuous variables at the 1% and 99% levels. After screening, our final sample is comprised of 3227 firm‐year observations.

Measurements

According to Hyatt and Berente (2017), an important basis for identifying the firms that employ substantial and symbolic strategies is the level of decoupling between their CSR performance and disclosure. Symbolic firms have a higher positive level of CSR decoupling. Referring to García‐Sánchez et al. (2021), we first quantify the levels of CSR performance and disclosure, then divide them into quantiles, and finally calculate the difference between them to determine the decoupling score; this score is then used as a proxy variable for CSR decoupling. The specific steps are as follows:

First, for CSR performance, third‐party CSR ratings (such as KLD, etc.) are widely used as proxy variables in mature capital markets. Unfortunately, the Chinese capital market lacks reliable and effective commercial CSR ratings. Although widely used by previous studies, RKS ratings are based only on corporate self‐selective CSR disclosures and do not consider any third‐party information, and Hexun ratings have serious flaws in terms of their indicator design (Zhong et al., 2019). A relatively limited number of listed firms are covered by other ESG ratings. For instance, the Syntao ESG ratings include only listed firms with approximately the top 1300 market values in 2019. MSCI ESG Ratings provides ratings for only 244 listed firms in 2019.

Therefore, we refer to Lin et al. (2015) and use corporate donations as a proxy for CSR performance. Corporate philanthropic donations are an important part of CSR (Carroll, 1979), as they can more effectively reflect substantial corporate CSR input. Furthermore, in the Chinese market, the information disclosure of corporate donations is mandated, and it is subject to the supervision of external assurance systems (e.g., public auditing). Compared with other unassured CSR information, this information quality is higher. We perform 5‐quantile processing according to corporate donation in year t‐1 ; specifically, we sort the firms from high to low according to five intervals and assign scores to them that range from 1 to 5. For those without donations, this value is replaced with zero, and the interval is assigned after processing that is, [0, 5].

Second, according to Liao et al. (2019), T scores of RKS ratings are used as a proxy for CSR disclosure. We also conduct 5‐quantile processing for the T scores. The disclosure level of firms not covered by RKS is defined as zero; thus, the score interval is [0, 5], and a higher score indicates better CSR disclosure.

Third, we define the variable Gap, which is equal to each firm's CSR disclosure score minus its CSR performance score. The score interval of Gap is [−5, 5]; specifically, the higher this score is for a given firm, the more decoupled the CSR of that firm is.

Finally, to examine symbolic CSR strategy more intuitively, we define a dummy variable, Symbolic. If the value of the variable Gap is greater than zero, Symbolic is equal to one; otherwise, it is zero. Then, we define a continuous variable, Symbolic_rank, to measure the degree to which a firm implements symbolic strategy.

When Symbolic is equal to one, Symbolic_rank is equal to the corresponding Gap score. When CSR_symbolic is equal to zero, Symbolic_rank is equal to zero. Thus, the range of Symbolic_rank is [0, 5]; specifically, the higher this value is for a firm, the more symbolic strategy is implemented in it.

Event study

For hypotheses 3a and 3b, we employ an event study. We set the event date as the day of the firm's first donation; if the donation date is not available, we use the news reporting date for replacement. Among our samples, the earliest and latest first donation dates are January 20 and March 2, respectively. The median event day is January 29. According to Krüger (2015), we set the pre‐event window evaluation period to (−250, −50) and the event window period to (−5, +5) and (−10, +10). To calculate abnormal returns (AR), we use the return of firm “‘on day t minus the market return, so the model is: where the market return is the value‐weighted market return.

Model design

To empirically test our hypothesis 1, we employ the following logit model to observe the relationships between the different CSR strategies and the probability of COVID‐19 donation: where the dependent variable Dona is a dummy variable that indicates whether a firm made donations during the COVID‐19 crisis (as of March 20). If firm i donated during the crisis, this value equals one; otherwise, it equals zero. To further distinguish the differences in donation response speeds of different types of firms, we set the dependent variable Early, which is a dummy variable that indicates whether the firm will donate before the median donation date (January 29) during the COVID‐19 crisis. When the firm makes a donation before the median donation date (January 29), the value is one; otherwise, it is zero. The independent variables include the continuous variable Gap, which represents the decoupling level between the CSR disclosure and CSR performance of the firms. The dummy variable Symbolic is used to indicate whether a firm employs the symbolic CSR strategy; if yes, the value is one; otherwise, it is zero. The continuous variable Symbolic_rank represents the extent to which a firm adopts the symbolic CSR strategy.

Referring to Jia and Zhang (2018), Lin et al. (2015), Luo et al. (2017) and Wu et al. (2018), our model includes the following control variables: the ratio of cash and equivalents to total assets (CASH), return on assets (ROA), the asset‐to‐liability ratio (LEV), the shareholding ratio of the largest shareholder (TOP1), the book‐to‐market value ratio (BTM), the natural logarithm of total assets (SIZE) and the natural logarithm of listing time (AGE). All the above control variables lag in t‐1, specifically, the value of 2019. Finally, we control industry effects in the model.

For hypothesis 1, we also design the following Tobit model to observe the relationship between the different CSR strategies and the level of crisis donations: where the dependent variable ln_Dona is the natural logarithm of the donation in the whole COVID‐19 crisis (as of March 20). The dependent variable ln_Early is the donation amount announced by the firm before the median donation date (January 29), which is used to observe the level of the firm's rapid response. The independent variables and control variables are consistent as in Equation (2).

To empirically test hypothesis 2, we employ the following model: where the dependent variables are the dummy variables Dona and Early and the continuous variables ln_Dona and ln_Early. For dummy variables and continuous variables, we use logit and Tobit methods, respectively, for estimation. The dummy variable Private is used to identify the private firm; when it is private, the value is one; otherwise, it is zero. The interaction term Gap × Private is our prime interest in this equation. Other control variables and effects are consistent with Equations (2) and (3). Similarly, except for variables related to donations, we use all the lagged variables in t‐1 (2019).

To empirically test hypotheses 3a and 3b, we employ the following Model and examine the market response to crisis donations under different CSR strategies: where CAR5 and CAR10 are the cumulative abnormal returns calculated during windows (−5, 5) and (−10, 10). The independent variables include continuous variable Gap, dummy variable Symbolic and continuous variable Symbolic_rank. Referring to (Chen et al., 2017), we include the natural logarithm of total assets (SIZE), the ratio of book to market value (BTM) and the asset‐to‐liability ratio (LEV) as control variables. The independent variables and control variables are lagged in t‐1 (2019).

EMPIRICAL RESULTS

Descriptive statistics and correlations

Table 1 provides a summary of the descriptive statistics of the main variables. The average of the dummy variables Dona and Early are 0.146 and 0.0657, indicating that as of March 20 and January 29, approximately 14.6%, and 6.57% of the sample firms donated during the COVID‐19 crisis. The continuous variable Donation has many values equal to zero, which means there is a serious problem with its tail distribution. To deal with this issue, we use Tobit regression. The variables CAR5 and CAR10 correspond to the cumulative abnormal returns of the window periods (−5, +5) and (−10, +10), respectively. Their mean values are 0.033 and 0.0285, indicating that the cumulative returns of all the samples in these two window periods are 3.3% and 2.85%, respectively. The average of the dummy variable Symbolic is 0.0806, indicating that in our setting; approximately 8.6% of the samples implement symbolic strategy. The average value of the variable Private is 0.726, indicating that approximately 72.6% of the sample firms are private. The descriptive statistics of the control variables are generally consistent with those of previous studies.

TABLE 1

Descriptive statistics

Variables	N	Mean	SD	Min	P25	Median	P75	Max
Dona	3227	0.146	0.353	0	0	0	0	1
ln_Dona	3227	0.810	2.085	0	0	0	0	9.904
Early	3227	0.0657	0.248	0	0	0	0	1
ln_ Early	3227	0.377	1.479	0	0	0	0	9.21
CAR5	458	0.033	0.155	−0.343	−0.0633	0.00024	0.0958	0.954
CAR10	458	0.0285	0.15	−0.477	−0.0679	−0.006	0.0866	0.822
Gap	3227	−1.579	1.948	−5	−3	−1	0	5
Symbolic	3227	0.0806	0.272	0	0	0	0	1
Symbolic_rank	3227	0.174	0.686	0	0	0	0	5
CASH	3227	20.23	1.417	14.55	19.32	20.13	21.03	26.4
ROA	3227	0.0247	0.121	−1.575	0.0133	0.0381	0.0693	0.526
LEV	3227	0.416	0.209	0.00836	0.255	0.405	0.554	1.712
TOP1	3227	33.25	14.4	4.08	22.17	30.66	42.49	88.24
BTM	3227	0.669	0.251	0.0238	0.489	0.668	0.851	1.442
Size	3227	22.25	1.312	18.29	21.32	22.07	22.95	28.64
Age	3227	2.088	0.9	0	1.386	2.197	2.944	3.401
Private	3227	0. 726	0.446	0	0	1	1	1

Table 2 reports the Pearson correlations of the main variables. The correlation coefficients between the donation variables Dona and ln_Dona and the decoupling variable Gap are −0.059 and − 0.054, respectively, and both of these are significant at 1% level. The correlation coefficients between the donation variables early and ln_Early and the variable Gap are similar. These results show that the incidence and level of philanthropic donations are lower in symbolic firms, which preliminarily supports Hypothesis 1. The correlation coefficients between the donation variables Dona, ln_Dona, early and ln_Early between the variables Symbolic and Symbolic_rank are weakly positive and nonsignificant. Thus, additional evidence is required.

TABLE 2

Bivariate correlations

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)
(1) Dona	1
(2) ln_Dona	0.962 a	1
(3) Early	0.641 a	0.617 a	1
(4) ln_ Early	0.664 a	0.680 a	0.940 a	1
(5) Gap	−0.059 a	−0.054 a	−0.086 a	−0.076 a	1
(6) Symbolic	0.004	0.003	0.016	0.023	0.567 a	1
(7) Symbolic_rank	0.006	0.008	0.009	0.013	0.580 a	0.855 a	1
(8) CASH	0.230 a	0.261 a	0.246 a	0.287 a	−0.016	0.158 a	0.174 a	1
(9) ROA	0.031 b	0.044 c	0.075 a	0.078 a	−0.109 a	−0.008	−0.005	0.198 a	1
(10) LEV	0.062 a	0.064 a	0.049 a	0.058 a	0.026	0.057 a	0.071 a	0.265 a	−0.359 a	1
(11) TOP1	0.01	0.02	−0.004	0.023	0.030 b	0.049 a	0.068 a	0.204 a	0.165 a	0.004	1
(12) BTM	0.001	0.004	0.007	0.009	−0.011	0.103 a	0.103 a	0.444 a	−0.02	0.335 a	0.112 a	1
(13) SIZE	0.232 a	0.265 a	0.256 a	0.300 a	0.001	0.181 a	0.195 a	0.846 a	0.067 a	0.463 a	0.176 a	0.576 a	1
(14) AGE	0.057 a	0.068 a	0.055 a	0.080 a	0.112 a	0.161 a	0.135 a	0.266 a	−0.176 a	0.320 a	−0.089 a	0.222 a	0.412 a	1
(15) Private	0.039 c	0.027	0.037 c	0.019	−0.165 a	−0.158 a	−0.162 a	−0.339 a	−0.004	−0.201 a	−0.197 a	−0.285 a	−0.371 a	−0.409 a	1

Note: Significance level.

<0.01.

<0.10.

<0.05.

Analysis for crisis philanthropic response

For Hypothesis 1, we analyze the philanthropic donations in the full sample period (as of March 20) and the sample period before the median day (as of January 29). The analysis results are reported in Table 3 and Table 4.

TABLE 3

Regression results of COVID‐19 crisis donation as of march 20, 2020

Dependent variable=	Dona	ln_Dona	Dona	ln_Dona	Dona	ln_Dona
	(1)	(2)	(3)	(4)	(5)	(6)
Gap	−0.088 a	−0.383 a
	(−3.107)	(−3.445)
Symbolic			−0.154	−0.298
			(−0.722)	(−0.357)
Symbolic_rank					−0.144	−0.554
					(−1.618)	(−1.578)
CASH	0.095	0.419	0.100	0.431	0.101	0.434
	(1.110)	(1.248)	(1.175)	(1.286)	(1.180)	(1.289)
ROA	0.361	0.095	0.594	1.141	0.566	0.992
	(0.401)	(0.028)	(0.655)	(0.335)	(0.625)	(0.292)
LEV	−0.711 b	−3.113**	−0.635	−2.809 b	−0.644	−2.847 b
	(−1.794)	(−1.987)	(−1.606)	(−1.791)	(−1.635)	(−1.823)
TOP1	−0.016 a	−0.056 a	−0.017 a	−0.059 a	−0.017 a	−0.058 a
	(−3.749)	(−3.329)	(−3.897)	(−3.501)	(−3.867)	(−3.462)
BTM	−1.665 a	−7.170 a	−1.632 a	−7.035 a	−1.642 a	−7.087 a
	(−5.475)	(−6.120)	(−5.367)	(−6.012)	(−5.403)	(−6.049)
SIZE	0.891 a	3.533 a	0.881 a	3.500 a	0.889 a	3.534 a
	(8.242)	(8.714)	(8.206)	(8.679)	(8.253)	(8.732)
AGE	−0.265 a	−0.842 a	−0.283 a	−0.941 a	−0.281 a	−0.921 a
	(−3.554)	(−2.921)	(−3.793)	(−3.260)	(−3.778)	(−3.200)
Constant	−20.652 a	−84.544 a	−20.337 a	−83.270 a	−20.533 a	−84.058 a
	(−14.810)	(−17.362)	(−14.518)	(−17.143)	(−14.617)	(−17.284)
Constant		7.752 a		7.780 a		7.771 a
		(44.840)		(45.116)		(44.994)
Industry effect	Yes	Yes	Yes	Yes	Yes	Yes
N	3220	3227	3220	3227	3220	3227
Pseudo R2	0.188	0.101	0.185	0.099	0.186	0.099

Note: t‐test values are reported in parentheses. Significance level.

<0.005.

<0.05, <0.1.

** <0.05.

TABLE 4

Regression results of crisis donation as of January 29, 2020

Dependent variable=	Early	ln_ early	Early	ln_ early	Early	ln_ early
	(1)	(2)	(3)	(4)	(5)	(6)
Gap	−0.085**	−0.361 a
	(−2.121)	(−8.605)
Symbolic			−0.336	−1.443 a
			(−1.071)	(−8.839)
Symbolic_rank					−0.175	−0.778 a
					(−1.361)	(−12.880)
CASH	0.293**	1.248 a	0.295**	1.244 a	0.297**	1.251 a
	(2.465)	(105.101)	(2.508)	(105.679)	(2.519)	(106.329)
ROA	−2.076 b	−9.333 a	−1.883 b	−8.429 a	−1.894 b	−8.505 a
	(−1.896)	(−20.119)	(−1.716)	(−19.003)	(−1.726)	(−19.127)
LEV	−0.756	−3.386 a	−0.694	−3.202 a	−0.693	−3.205 a
	(−1.261)	(−7.781)	(−1.161)	(−7.387)	(−1.167)	(−7.400)
TOP1	−0.012 b	−0.044 a	−0.012**	−0.045 a	−0.012**	−0.045 a
	(−1.936)	(−7.844)	(−2.001)	(−8.068)	(−2.000)	(−7.989)
BTM	−2.143 a	−9.185 a	−2.116 a	−9.075 a	−2.132 a	−9.143 a
	(−4.922)	(−29.943)	(−4.891)	(−29.727)	(−4.933)	(−29.989)
SIZE	0.893 a	3.817 a	0.887 a	3.813 a	0.890 a	3.832 a
	(6.153)	(350.623)	(6.094)	(353.373)	(6.124)	(355.407)
AGE	−0.273**	−1.273 a	−0.285 a	−1.313 a	−0.286 a	−1.314 a
	(−2.535)	(−13.985)	(−2.683)	(−14.442)	(−2.689)	(−14.483)
Constant	−40.284 a	−160.891 a	−39.720 a	−160.891 a	−39.840 a	−161.493 a
	(−20.486)	(−646.418)	(−19.799)	(−653.781)	(−19.899)	(−656.641)
Constant		8.439 a		8.454 a		8.449 a
		(77.419)		(78.004)		(78.078)
Industry effect	Yes	Yes	Yes	Yes	Yes	Yes
N	3061	3227	3061	3227	3061	3227
Pseudo R2	0.247	0.155	0.245	0.154	0.245	0.154

Note: t‐test values are reported in parentheses. Significance test.

<0.005.

<0.05, <0.1.

** <0.05.

Table 3 reports the regression results of crisis donations for the full period (as of March 20). Among them, Columns (1), (3), and (5) are logit regression results based on the dummy variable Dona. Since there is no variation in the value of dummy variables in some industries, they cannot participate in logit regression, so the sample size drops to 3220. Columns (1) and (2) are the analysis results based on the dummy variable Dona and the continuous variable ln_Dona, respectively. The coefficients of the variable Gap are −0.088 and − 0.383, respectively, which are significant above 1%. They show that the level of the CSR gap is significantly and negatively correlated with the probability of crisis donation and the level of donation. Columns (3) and (4) are the analysis results based on the dummy variable Symbolic. The coefficients are −0.154 and − 0.298, respectively, but they are not significant. This shows that in the analysis of the whole period, the relationship between symbolic strategy and crisis donations is not clear. Columns (5) and (6) are regression results based on the continuous variable symbolic. The variable coefficients are −0.144 and − 0.554 (both nonsignificant), indicating that the relationship between the level of implementation of symbolic strategy and crisis donation is also unclear.

Furthermore, to test whether there is a difference in the response speed of the symbolic strategy firms, we generated two dependent variables early and ln_Early according to the median donation date (January 29). The relevant results are reported in Table 4. In columns (1) and (2), the coefficients of the variable Gap are −0.085 (p < 0.05) and − 0.361 (p < 0.01), respectively. They are consistent with the results of the analysis in the whole period, indicating that the CSR gap is negatively related to the possibility of crisis donations and the level of donations. Columns (3) and (5) are the logit analysis results based on the dummy variable Early. The coefficients of the dummy variable Symbolic and the continuous variable Symbolic_rank are −0.336 and − 0.175, respectively, but not significant. This evidence suggests that even in the early period of crisis donation, there is no clear relationship between symbolic strategy and the possibility of crisis donation behavior. However, in the analysis of columns (4) and (6), the analysis based on Tobit regression shows that the coefficients of the variables Symbolic and Symbolic_rank are −1.443 and − 0.778, respectively, and both are significant above 1%. This indicates that in the earlier period of crisis donation, there is a significant and negative correlation between the level of implementation of symbolic strategy and the level of crisis donations.

Based on the analysis results of the two periods, we find that (1) the gap between CSR performance and disclosure is significantly and negatively correlated with the possibility and level of crisis donations, regardless of the early stage or the whole period of the COVID‐19 crisis. (2) Regardless of the period, there is no significant evidence that symbolic strategy and substantive strategy will cause differences in the occurrence of crisis donations. However, in the early stages of the crisis, firms that implemented symbolic strategy show a lower level of philanthropic response. These results together suggest that, along with the crisis, the response difference between substantive and symbolic firms weakens. In general, we find that this evidence can support Hypothesis 1; that is, firms implementing symbolic and substantive CSR strategies have material differences in the philanthropic response to the COVID‐19 crisis.

Analysis based on ownerships

For Hypothesis 2, we design model (5) for analysis, and the results are reported in Table 5. Our interest is the interaction term Gap×Private. Columns (1) and (2) are the analysis for the whole period as of March 20. The coefficients of the interaction term Gap × Private are −0.092 (not significant) and − 0.470 (p < 0.01), respectively. This shows that private ownership strengthens the negative correlation between the gap and the level of donation, but it has no effect on the relationship between the gap and the occurrence of donation. Columns (3) and (4) are the analysis of the earlier stage as of the donation median date (January 29). The results are consistent with the whole period. The interaction terms Gap × Private are −0.081 (not significant) and − 0.522 (p < 0.01), respectively. The above results support Hypothesis 2. There is a greater difference in philanthropic response to the COVID‐19 crisis in private firms with symbolic and substantive strategies.

TABLE 5

Results of mechanism analysis of ownerships

Dependent variable=	Dona	ln_Dona	Early	ln_ early
	(1)	(2)	(3)	(4)
Gap × private	−0.092	−0.470 a	−0.081	−0.522**
	(−1.038)	(−8.197)	(−1.322)	(−2.135)
Gap	0.009	0.117**	−0.014	0.065
	(0.122)	(2.208)	(−0.276)	(0.325)
Private	0.937 a	3.932 a	0.581 a	2.122 a
	(3.358)	(17.594)	(3.141)	(2.962)
CASH	0.368 a	1.460 a	0.145 b	0.623 b
	(2.966)	(121.202)	(1.675)	(1.866)
ROA	−1.876 b	−8.209 a	0.432	0.561
	(−1.720)	(−17.230)	(0.488)	(0.169)
LEV	−0.637	−2.769 a	−0.630	−2.638 b
	(−1.042)	(−6.347)	(−1.577)	(−1.694)
TOP1	−0.009	−0.030 a	−0.014 a	−0.046 a
	(−1.348)	(−5.198)	(−3.120)	(−2.692)
BTM	−1.899 a	−8.003 a	−1.500 a	−6.541 a
	(−4.287)	(−26.041)	(−4.884)	(−5.555)
SIZE	0.853 a	3.680 a	0.858 a	3.347 a
	(5.726)	(333.158)	(7.839)	(8.290)
AGE	−0.149	−0.682 a	−0.171**	−0.397
	(−1.276)	(−7.459)	(−2.145)	(−1.304)
_cons	−41.944 a	−167.394 a	−21.727 a	−87.478 a
	(−20.394)	(−659.323)	(−14.769)	(−17.468)
Constant		8.272 a		7.655 a
		(76.256)		(44.064)
Industry effect	Yes	Yes	Yes	Yes
N	3061	3227	3220	3227
r2_p	0.262	0.165	0.197	0.107

Note: t‐test values are reported in parentheses. Significance test.

<0.005.

<0.05, <0.1.

** <0.05.

Market reaction

Univariate tests

For hypotheses 3a and 3b, we first apply a univariate test to the firms' abnormal returns. In Table 6, we report the comparison results of the cumulative excess return (CAR) between the symbolic group and the non‐symbolic group based on the donation sample under different window periods. The event date is the first donation date or the first donation news reporting date. In the event windows (−5, +5) and (−10, +10), the sample size of the non‐symbolic group is 415, and the mean CAR values are 0.0320 and 0.0270, which are significant at the 1% level. This means that donating firms that do not employ symbolic strategies can obtain approximately 3% abnormal returns. In the symbolic group, the sample size was 43, and the average value of CAR was higher, 0.0390 (not significant) and 0.0470 (significant above the level of 10%). There is a slight difference in the mean between the symbolic group and the control group, which are −0.007 and − 0.0200, respectively, but not significantly. Therefore, the univariate results initially support Hypothesis 3b, and the Chinese capital market does not have a differentiated response to the crisis donation of firms that implement different strategies.

TABLE 6

Univariate tests of cumulative abnormal return

Event window		Symbolic = 0		Symbolic = 1
	N	Mean	N	Mean	Mean difference
(−5, +5)	415	0.0320 a (4.343)	43	0.0390 (1.366)	−0.007 (−0.274)
(−10, +10)	415	0.0270 a (3.644)	43	0.0470 b (1.925)	−0.0200 (−0.841)

Note: t‐test values are reported in parentheses. Significance test.

<0.005.

<0.05, <0.1.

0.0320

0.0270

0.0470

Regression results

For Hypotheses 3a and 3b, we run an OLS regression based on model (5), and the results are reported in Table 7. The samples participating in the regression are firms that make crisis donations, and the dependent variables CAR5 and CAR10 are the cumulative abnormal returns calculated based on windows (−5, 5) and (−10, 10), respectively. Columns (1)–(3) are the regression results based on the dependent variable CAR5. The coefficients of the variables Gap, Symbolic, and Symbolic_rank are −0.05, 0.011 and 0.003, respectively, which are not significant and support Hypothesis 3b. Columns (4)–(6) are the regression results based on the dependent variable CAR10. Similar to the results under CAR5, the coefficients of the variables Gap, symbolic and Symbolic_rank are all insignificant, and Hypothesis 3b is supported. In summary, we find that Hypothesis 3b establishes that the Chinese market does not have a differential response to the crisis philanthropy response of firms that implement different strategies.

TABLE 7

Regression results of cumulative abnormal return

Dependent variable=	CAR5	CAR5	CAR5	CAR10	CAR10	CAR10
	(1)	(2)	(3)	(4)	(5)	(6)
Gap	−0.005			−0.005
	(−1.368)			(−1.574)
Symbolic		0.011			0.021
		(0.362)			(0.836)
Symbolic_rank			0.003			0.001
			(0.271)			(0.120)
SIZE	−0.002	−0.004	−0.004	0.002	0.000	0.001
	(−0.354)	(−0.712)	(−0.702)	(0.428)	(0.058)	(0.123)
BTM	−0.024	−0.022	−0.022	−0.068**	−0.065**	−0.066**
	(−0.848)	(−0.754)	(−0.741)	(−2.483)	(−2.375)	(−2.397)
LEV	−0.096**	−0.095**	−0.096**	−0.055	−0.054	−0.054
	(−2.239)	(−2.198)	(−2.212)	(−1.320)	(−1.285)	(−1.295)
Constant	0.126	0.176	0.177	0.035	0.087	0.081
	(1.085)	(1.603)	(1.584)	(0.318)	(0.796)	(0.737)
N	458	458	458	458	458	458
Adj.R2	0.032	0.027	0.027	0.030	0.027	0.025

Note: t‐test values are reported in parentheses. Significance test.

** <0.05.

Robustness checks

In the main analysis, some firms adopt group donations; thus, we cannot distinguish the donations of these firms and are excluded. This may affect our results. In this round of analysis, we include the firms donated by firm groups. Since it is impossible to distinguish the number of donations, we only use the dummy variables Dona and Early to perform logit regression analysis. In the analysis of the dependent variables Dona and Early, the sample size rises to 3216 and 3399, respectively.

Table 8 reports the analysis results based on an alternative sample design. Among them, columns (1)–(3) are the analysis results based on the variable Dona, which is under the full sample period (as of March 20). The coefficient of the variable Gap is −0.098 (p < 0.05), and the coefficients of the symbolic dummy and continuous variables Symbolic and Symbolic_rank are negative but not significant. Columns (4)–(6) are the analysis results based on the dummy variable Early, that is, under the earlier period (as of January 29). The coefficients of the variables Gap, Symbolic and Symbolic_rank are −0.101 (p < 0.05), −0.249 (nonsignificant) and − 0.173 (p < 0.05), respectively. The above results are consistent with the results of our main analysis.

TABLE 8

Alternative samples selection

Dependent variable=	Dona	Dona	Dona	Early	Early	Early
	(1)	(2)	(3)	(4)	(5)	(6)
Gap	−0.098**			−0.101 a
	(−2.471)			(−3.632)
Symbolic		−0.425			−0.249
		(−1.365)			(−1.182)
Symbolic_rank			−0.207			−0.173**
			(−1.638)			(−1.966)
CASH	0.290**	0.291**	0.292**	0.083	0.086	0.086
	(2.493)	(2.528)	(2.523)	(0.965)	(1.017)	(1.009)
ROA	−2.046 b	−1.831 b	−1.845 b	0.387	0.642	0.619
	(−1.866)	(−1.665)	(−1.679)	(0.426)	(0.702)	(0.678)
LEV	−0.650	−0.586	−0.591	−0.642 b	−0.565	−0.573
	(−1.110)	(−1.005)	(−1.017)	(−1.653)	(−1.460)	(−1.482)
TOP1	−0.015**	−0.015**	−0.015**	−0.019 a	−0.020 a	−0.020 a
	(−2.413)	(−2.465)	(−2.475)	(−4.469)	(−4.603)	(−4.587)
BTM	−2.148 a	−2.116 a	−2.138 a	−1.648 a	−1.607 a	−1.622 a
	(−4.980)	(−4.942)	(−5.000)	(−5.475)	(−5.341)	(−5.393)
SIZE	0.857 a	0.849 a	0.856 a	0.859 a	0.848 a	0.858 a
	(6.119)	(6.068)	(6.093)	(8.080)	(8.057)	(8.099)
AGE	−0.325 a	−0.337 a	−0.340 a	−0.319 a	−0.336 a	−0.336 a
	(−3.096)	(−3.248)	(−3.273)	(−4.375)	(−4.616)	(−4.632)
Constant	−38.746 a	−39.630 a	−38.944 a	−19.528 a	−19.150 a	−19.371 a
	(−20.661)	(−20.294)	(−20.202)	(−14.517)	(−14.216)	(−14.300)
Industry effect	Yes	Yes	Yes	Yes	Yes	Yes
N	3216	3216	3216	3399	3399	3399
Pseudo R2	0.237	0.235	0.235	0.181	0.176	0.177

Note: t‐test values are reported in parentheses. Significance test.

<0.005.

<0.05, <0.1.

In addition, following Zhang et al. (2010), we also added the province control variable Hubei to regression models (2), (3), and (4) to control for the influence of corporate geographic location. When the place of registration of the firm belongs to Hubei, the dummy variable Hubei equals one; otherwise, it equals zero. The analysis conclusion is consistent with the main analysis. Due to brevity, we have not reported the results.

CONCLUSION

Based on the wave of corporate philanthropic donation in the COVID‐19 crisis in 2020, our study focuses on two research questions: (1) Are there any material differences in the philanthropic response to the COVID‐19 crisis between Chinese firms that implement substantive and symbolic CSR strategies? (2) Does the Chinese capital market react differently to philanthropic responses under substantive and symbolic strategies? The analysis results show that there are material differences in the philanthropic response of firms that implement symbolic and substantive CSR strategies, and private ownership exacerbates the response gap. Furthermore, we test whether the Chinese capital market reacts differently to firms' philanthropic responses under substantive and symbolic strategies, and empirical evidence does not support the existence of such differences.

This study offers some unique contributions. First, our study helps to enrich managers and academics to obtain the latest understanding of the philanthropic donation of Chinese firms in public emergencies. Second, our findings provide a meaningful reference for the Chinese “CSR‐Firm value” relationship mechanism. Finally, our findings enrich the understanding of the role of CSR in firms' fight against the risks posed by the COVID‐19 crisis.

Our study has some limitations that future research may address. Due to some objective factors, we cannot find a perfect CSR rating proxy for the CSR performance of Chinese firms; therefore, our use of corporate philanthropic donation as a proxy for CSR performance may result in a validity problem, which is the first important limitation of this study. Second, the in‐kind donation data come from corporate announcements or media reports, so it is difficult to estimate the fair value of some items accurately when converting. These figures are not supported by more accurate accounting information, which may affect the accuracy of the data. Moreover, our evidence of whether the market has the ability to discriminate between different strategies is derived from a single public event, which may introduce the problem of sampling deviations. These limitations offer new avenues for future research.