Introduction

Investors increasingly expect consistent, reliable, and comparable sustainability disclosure for more efficient risk pricing and capital allocation. Thus, sustainability reporting has become standard practice for many companies, with steady growth over the past decade (KPMG, 2022). Under pressure from governments, regulators, investors, and other stakeholders, many publicly listed companies are adopting CSR reporting as a crucial method for disclosing sustainability information (Perez-Batres et al., 2012). However, companies may exaggerate publicly disclosed CSR information while being stingy in investing resources and otherwise not living up to their CSR images (Tashman et al., 2019). This leads to the inconsistency between CSR disclosures and actual CSR activities, academically termed CSR decoupling (Sauerwald and Su, 2019; Chen et al., 2024; Hawn and Ioannou, 2016). Inevitably, CSR decoupling diminishes the quality of sustainability information and damages the company’s value and reputation (Schons and Steinmeier, 2006; Ioannou et al., 2023).

The volatility of China’s stock market, characterized by frequent fluctuations and sharp rises and falls, often undermines investor confidence and financial market stability. Many scholars have selected stock price crash risk (SPCR) as their research subject to explore its determining factors (Kim and Zhang, 2016; Ball, 2009; Kim et al., 2014; Hutton et al., 2009). However, a wealth of academic research investigating CSR performance’s efficacy on SPCR has obtained mixed evidence. Previous scholars argue that CSR disclosure effectively enhances information transparency, diminishes information asymmetry, and mitigates adverse news backlogs, reducing vulnerability to stock price crashes (Wang et al., 2023; Bouslah et al., 2018; Kim et al., 2014). The opposite view is that CSR is merely an instrument of self-interest management employs to cover up unfavorable corporate news, and using such a tool worsens corporate information transparency and reinforces management’s cover-up behavior, resulting in increased SPCR (Quan et al., 2015; Hao et al., 2018).

A possible reason for the inconsistent evidence is the existence of the CSR decoupling phenomenon. CSR has two dimensions: disclosure and actual performance. Previous studies implicitly assume that the two are consistent and use only CSR disclosure or CSR performance score as the proxy of CSR. However, many studies have shown that there is likely a difference between a company’s CSR disclosure and its actual CSR practices (Velte, 2023), which is defined as “CSR decoupling” (Sauerwald and Su, 2019; Hawn and Ioannou, 2016). Therefore, to help clarify mixed evidence in existing research, this paper explores CSR decoupling’s efficacy on future SPCR.

Previous literature concentrates mainly on CSR decoupling’s firm- and country-related determinants, whereas few studies explore the consequences of CSR decoupling (Velte, 2023). Previous scholars find that CSR decoupling is widespread and can conceal corporate failure, which has a “masking effect” on CSR issues. Hawn and Ioannou (2016) discover that market value can be negatively affected if external behavior is inconsistent with internal behavior. García-Sánchez et al. (2021) find that CSR decoupling elevates firms’ capital costs and reduces their finance capacities. In response to Velte’s (2023) call for more studies on the outcomes of CSR decoupling, this paper selects SPCR as the object of study.

CSR decoupling may stem from the imperfect nature of the CSR reporting framework, where methods and content lack uniformity, offering managers broad discretion to selective disclosure (Patten, 1991; Xu and Liu, 2018). Thus, managers may utilize information asymmetry with external stakeholders for opportunistic impression management to establish legitimacy, enhance reputations, or even deceive stakeholders (Marquis et al., 2016; Delmas and Burbano, 2011). This behavior undermines CSR information quality, reducing the reliability of reports and corporate transparency. Consequently, management may conceal negative CSR news until too late, leading to stock price plunges (Jin and Myers, 2006).

We use textual analysis to construct an indicator that intuitively embodies the inconsistency between CSR disclosures and actual CSR engagements. Taking Chinese listed companies over the 2010–2019 period as a sample, we find that CSR decoupling exacerbates SPCR, and our finding remains valid after endogeneity and robustness examinations. Mechanism focuses on subsample testing to analyze the effect of agency risk. We examine the relationship between CSR decoupling and SPCR from the perspectives of different levels of agency risks, focusing on executives’ experiences, monitoring pressures they encounter, and the institutional environment in which they operate. We find that in contexts with higher agency risk, the effect of CSR decoupling on SPCR is more pronounced, suggesting the role of information asymmetry as a crucial mechanism through which CSR decoupling influences SPCR.

This study makes several possible contributions. First, the association between CSR and SPCR explored in this article provides new empirical evidence for the currently debated topic. Certain studies indicate that CSR disclosures alleviate information asymmetry, enhance information transparency, and thus diminish the likelihood of shock price collapses (Kim et al., 2014; Bouslah et al., 2018). Other studies hold the opposite view that executives use CSR solely as a self-interest tool, reinforcing cover-up behavior and resulting in increased SPCR (Quan et al., 2015; Hao et al., 2018). Unlike previous literature, which only considers CSR disclosure or performance score as the proxy for CSR, we focus on the inconsistency between the two dimensions of CSR (disclosure and actual activities) and examine CSR decoupling’s efficacy on the crash risks of future stock prices, hence helping to clarify existing research’s mixed evidence.

Secondly, this paper complements existing research on CSR decoupling’s economic consequences. Previous literature mainly refers to country- and firm-related determinants of CSR decoupling, while exploring its outcomes is relatively scarce (Velte, 2023). Velte’s (2023) literature review encourages forthcoming scholars to pay more attention to how CSR decoupling influences shareholders’ and other stakeholders’ decisions. In response to this call, we take asymmetric risk as an entry point and examine CSR decoupling’s economic consequences from the perspective of stock price crashes. By reviewing asymmetric fluctuations in the capital market, we provide insights into the behavioral strategies of internal corporate managers and potential market impacts.

Finally, this paper enriches empirical evidence on the determinants of SPCR. While existing literature explores SPCR’s determinants from the perspectives of perfect rationality and bounded rationality of managers (Ball, 2009; Dang et al., 2019; Kothari et al., 2009; LaFond and Watts, 2008; Kim et al., 2016), this paper views CSR decoupling as managers’ opportunistic impression management and explores the underlying mechanism through which CSR decoupling affects future SPCR, which complements the existing literature on SPCR’s determinants.

Literature review and hypothesis

Literature review

CSR

Scholars have extensively investigated the impact of a firm’s CSR on various stakeholders. However, there is no consensus on whether the impact of CSR is positive or negative. With the advocacy of sustainable development, scholars increasingly view CSR as beneficial (Tsang et al., 2023). Scholars who support the beneficial effects of CSR argue that it is an investment in customer satisfaction, reputation, and stakeholder management, which can ultimately lead to financial performance (Radhakrishnan et al., 2018). Independent CSR reports can better attract investors and analysts (Bucaro et al., 2020; Dhaliwal et al., 2012; Khan et al., 2016). The non-financial information disclosed in CSR reports can also improve corporate transparency, reduce information asymmetry, and stabilize stock prices (Wang et al., 2023; Kim et al., 2014; Bouslah et al., 2018). In contrast, Friedman (1970) contends that socially responsible behavior yields few quantifiable economic benefits, yet it incurs numerous costs that diminish profits and shareholder wealth. Some scholars believe that CSR is merely a self-serving tool management employs to conceal unfavorable corporate news and reinforce cover-up behavior (Quan et al., 2015; Hao et al., 2018). Thus, investors may unintentionally overestimate the fundamental value of positive CSR stocks (Elliott et al., 2014).

CSR decoupling

Empirical research on CSR has provided numerous inconsistent empirical evidence. To pursue a more profound understanding of CSR, scholars are increasingly focusing on the research of CSR decoupling. More literature focuses on CSR decoupling’s determinants than its consequences (Velte, 2023). Existing studies include the following three factors that influence CSR decoupling: first, external environments, such as institutional factors like laws, international market regimes, political systems, and GRI guidelines, and external stakeholder factors like institutional investors and analyst coverage (Yu et al., 2020; García-Sánchez et al., 2021; Tashman et al., 2019; Zhang, 2022); second, corporate characteristics, such as financial constraints, board size, board gender diversity, board network, CSR committee, and family firm (Eliwa et al., 2023; Gull et al., 2023; Parra-Domínguez et al., 2021; Xia et al., 2023; Yu et al., 2020; Zhao et al., 2022); third, manager traits, such as CEO overconfidence, CEO power, CEO narcissism, and political connections (Shahab et al., 2021; Al-Shammari et al., 2019; Sauerwald and Su, 2019; Marquis and Qian, 2014).

Research on CSR decoupling’s firm-level consequences appears to be relatively scarce. Previous studies find that the symbolic use of CSR has a “masking effect” on CSR issues and may damage a company’s image, which eventually reduces the firm’s stock price and market value (Du, 2015; Hawn and Ioannou, 2016; Chen and Dagestani, 2023). Some studies examine stakeholders’ reactions to CSR decoupling and find it significantly reduces customer satisfaction (Ioannou et al., 2023), enlarges firms’ analyst forecast errors (García-Sánchez et al., 2021), and increases actual employee turnover (Robertson et al., 2023).

Stock price crash risk

Scholars have mainly explored the causes of SPCR from two perspectives: perfect rationality and bounded rationality of managers. Studies based on the ideal rationality of managers show that because of the principal-agent problem arising from the separation of powers, managers maintain their positions and salaries (Ball, 2009; LaFond and Watts, 2008), avoid taxes (Kim et al., 2011), build corporate empires (Kothari et al., 2009), and seek other self-interests by concealing negative information within their companies. When accumulated adverse information reaches a critical threshold and is sensed by the market in a concentrated manner at some point, stock prices can collapse (Kim and Zhang, 2016).

Studies based on managers’ bounded rationality show that because of different cultural backgrounds, cognitive levels, and values, managers can suffer from psychological and behavioral biases (Dang et al., 2019); they may have difficulties thoroughly and objectively assessing their abilities and the benefits of risky investments, and they cannot properly handle negative feedback. When the poor performances of investments and negative information accumulate to certain levels and are released suddenly, companies’ stock prices will be adversely affected. For example, Kim et al. (2016) reveal that overconfident managers often overestimate companies’ future cash flows and their ability to achieve positive outcomes, thus ignoring negative feedback to continue projects likely to yield unfavorable financial outcomes. Consequently, this behavior accumulates poorly performing projects, and eventually, stock prices crash.

Research hypothesis

Information asymmetry within the agency framework is considered a primary factor that leads to SPCR (Habib et al., 2017; Chen et al., 2001). Asymmetric information enables managers to withhold unfavorable information over an extended duration, aiming to optimize compensation, safeguard employment, and mitigate potential litigation arising from disclosures of adverse news (Kothari et al., 2009). Previous research suggests that alleviating agency conflicts and enhancing corporate information transparency can help mitigate the SPCR. While information transparency decreases in the CSR reporting documents, decoupling strategies may reduce CSR reporting quality, and the information function of CSR disclosure is impaired (Velte, 2023). Thus, we argue that CSR decoupling can be considered as management taking advantage of information asymmetry to realize an inflated description of CSR performance, hindering negative information from being detected by external investors, and elevating SPCR. In the above analysis, management’s improper utilization of information asymmetry is a critical determinant exacerbating SPCR. Then, when the company faces higher agency risks, where managers have higher discretionary power and lower self-discipline, it is probable that this could result in a more pronounced exacerbating effect of CSR decoupling on SPCR. We then analyze the underlying logic of CSR decoupling and its relationship with stock price collapse from the perspective of different agency risks.

CSR decoupling and SPCR

According to the asymmetric information theory, senior executives have information advantages over external investors, and management can exaggerate their contributions in CSR reports to cultivate favorable company images and mislead investors about company fundamentals. For instance, Hemingway and Maclagan (2004) reveal that firms may use CSR to conceal their unethical behaviors. This can be seen as the managerial behavior of using CSR reports for self-serving impression management, inevitably leading to the decoupling of CSR.

A wide range of impression management phenomena can occur in CSR information disclosures. First, there is room for manipulation in CSR disclosures. Relative to obligatory financial disclosures, corporate management possesses increased discretion in preparing CSR reports, given that CSR disclosure is voluntary and lacks specific reporting guidelines and strict audit or assurance standards (Tsang et al., 2023). The core content of CSR reports is mainly qualitative, descriptive information. Compared to quantitative data, managers are more likely to manipulate descriptive qualitative details (Bloomfield, 2002). Textual expressions are often more ambiguous, subtle, and flexible, making strict standardization and legal constraints difficult. Managers have greater discretion in textual disclosures, allowing manipulation of market behavior (Brockman et al., 2013). At the same time, companies have strong motivations to engage in impression management when reporting CSR information. One of the primary purposes of disclosing CSR information lies in shaping stakeholders’ perceptions of companies (Guthrie and Parker, 1990; Hooghiemstra, 2000). While CSR does not directly improve firms’ profitability, it may be valuable to firms, as it can mitigate the impact of unfavorable news and exert an insurance effect (Bhattacharya and Sen, 2004; Tucker and Melewar, 2005). Because of CSR’s various benefits, companies may embellish their CSR disclosures and thus engage in impression management.

Managers’ impression management within CSR reporting directly weakens the quality of CSR information, significantly reduces its reliability and corporate transparency, and increases information asymmetry between companies and external investors, thus providing opportunities for management to hide unfavorable news for longer until the critical point when unfavorable information is released, which eventually leads to stock price plunges (Jin and Myers, 2006). Hence, the more CSR disclosure deviates from actual performance, the higher the degree of CSR decoupling, the less truthful and less credible the disclosure is, and the more likely managers will whitewash poor performance and hide lousy news that risks crashes. Meanwhile, because CSR decoupling distorts the original CSR information, external investors cannot accurately assess the veracity of reported information. Based on positive disclosures, they will form optimistic predictions, resulting in stock prices deviating from firms’ fundamental values and creating stock price bubbles. However, when a company’s CSR performance is unsatisfactory, and the market realizes its CSR report is exaggerated, it will reduce forecasted future corporate earnings and sell off stock, hence increasing SPCR.

To conclude, management has both the opportunity and the motivation to exploit information asymmetry for opportunistic impression management in CSR reports. SPCR may be exacerbated whenever a firm’s outwardly presented CSR image is inconsistent with its actual performance, preventing investors from receiving negative news promptly. Thus, we propose our first hypothesis as follows:

H1: CSR decoupling increases a firm’s SPCR.

The effect of agency risk

We expect the CSR decoupling results from management leveraging their information advantage over external stakeholders to distort genuine CSR information. This inevitably diminishes the quality of CSR reporting, leading to reduced corporate transparency and increased risk of SPCR. Many scholars have adopted agency theory frameworks to study the determinants of CSR decoupling under various agent risks (Velte, 2023). They discovered many internal and external governance factors that can constrain CSR decoupling, such as institutional investor shareholding (Yu et al., 2020), analyst and media attention (Zhang, 2022; Yue and Li, 2023), as well as the institutional environment in which the corporation operates (Tashman et al., 2019). Therefore, for corporate managers facing more constraints and lower agency risks, their motivation and discretion to implement symbolic CSR disclosure strategies are both constrained.

Extant research has also examined the influence of agency conflicts and information transparency on SPCR across various agent contexts (Habib et al., 2017). The executives’ experiences (Cao et al., 2019; Jin et al., 2022) and the supervision of other capital market participants all influence the company’s agency relationships (Callen and Fang, 2013; Boubaker et al., 2014). Management is less inclined to withhold bad news in companies with lower agency risks, and the company’s information transparency is superior (Habib et al., 2017; Chen et al., 2001). In this context, even if the CSR efforts are exaggerated by CSR disclosure, companies with lower agency risks are more likely to achieve more transparent financial disclosures or engage in timely and effective communication with investors. SPCR may not necessarily increase significantly with CSR decoupling. Conversely, in companies with higher agency risks, the negative impact of CSR decoupling is more likely to increase SPCR. Drawing from the preceding analysis, we posit the second hypothesis:

H2: When firms face higher agency risks, CSR decoupling has a more significant exacerbating effect on SPCR.

Methodology

Data and Sample

This paper uses Chinese A-share listed companies from 2010 to 2019 as the research sample. The sampling period starting from 2010 is due to the global financial crisis that erupted in 2008. The exclusion of years beyond 2020 is because the COVID-19 pandemic broke out in China in January 2020. Listed companies’ philanthropic response to the pandemic could potentially disrupt the measurement of CSR decoupling. Evidence from another country indicates that companies during the pandemic are more likely to implement symbolic CSR reporting strategies (Khanchel et al., 2023). Then, we clean the sample as follows: (1) exclude observations with less than 30 weeks of annual weekly stock returns; (2) delete companies in the financial sector; (3) exclude ST and PT companies; and (4) remove observations with missing relevant variables. After the above screening, 5297 company-annual observations were obtained. The CSR performance-rating data in this paper were obtained from the Hexun website, and the data on variables, including stock data, financial data, and corporate governance data, were gathered from the CSMAR database. All continuous variables below 1% or above 99% are winsorized to control for the effect of outliers.

Model and variables

Independent variable

Referring to Sauerwald and Su (2019) and Zhang (2022), we define CSR decoupling as the gap between CSR disclosure and actual CSR performance and use the following methodological measures.

Firstly, a textual analysis was conducted on each CSR report to construct the Optimistic Tone of Listed Companies’ CSR Reports (Optimistic Tone). The optimistic tone is an inverse measure of CSR reporting quality, and we use it as a proxy variable for CSR disclosure. Specifically, we first searched the company’s CSR report from Juchao Information Network (www.cninfo.com.cn), which is a website for the disclosure of information pertaining to publicly listed companies in China. Some firms issued standalone CSR reports under diverse titles, including “ESG reports” or “Sustainability reports.” As long as the report’s content includes social and environmental issues, the report can be considered a standalone CSR report and included in our sample (Tsang et al., 2023; Lys et al., 2015; Sauerwald and Su, 2019). Then, each CSR report undergoes detailed textual analysis to construct the Optimistic Tone index (Arslan-Ayaydin et al., 2016; Tetlock et al., 2008). We calculated the frequency of positive and negative tone words in reports from a certain year and defined them as POSPCT and NEGPCT, respectively. The calculation of frequency relies on word categorization and utilizes a lexicon established and verified by Loughran and Mcdonald (2011). The Optimistic Tone is determined by Eq. (1):

$${\rm{Optimistic\; Tone}}=\frac{{\rm{POSPCT}}-{\rm{NEGPCT}}}{{\rm{POSPCT}}+{\rm{NEGPCT}}}$$
(1)

Next, the rating data from Hexun was used to assess companies’ CSR performance (CSR Performance). To ensure numerical comparability, Optimistic Tone and CSR Performance were transformed into Z-scores. Finally, we define optimistic tone minus CSR performance as the degree of CSR decoupling:

$${{\rm{CSR}}}{{\_}}{{\rm{GAP}}}={\rm{Optimistic}}\,{\rm{Tone}}-{\rm{CSR}}\,{\rm{Performance}}$$
(2)

Dependent variable

Following Hutton et al. (2009) and Kim et al. (2011), we measure SPCR with the following indicators: negative conditional return skewness (NCSKEW) and down-to-up volatility (DUVOL).

First, the following regression model (3) is run on weekly returns of stock i:

$${r}_{i,t}={\alpha }_{i}+{\beta }_{1,i}{r}_{{ {m}},t-2}+{\beta }_{2,i}{r}_{{ {m}},t-1}+{\beta }_{3,i}{r}_{{ {m}},t}+{\beta }_{4,i}{r}_{{ {m}},t+1}+{\beta }_{5,i}{r}_{{ {m}},t+2}+{{\varepsilon }}_{{\rm{i}},{\rm{t}}}$$
(3)

where \({r}_{i,t}\) denotes weekly individual stock returns considering reinvestment of cash dividends, and \({r}_{{\rm {m}},t}\) represents the average return of all stocks weighted by market capitalization.

The market-adjusted return for stock i in week t is then calculated as

$${\mathrm{W}}_{{i},{t}}=\mathrm{ln}(1+{{\rm{\varepsilon }}}_{{i},{t}})$$
(4)

where \({{\varepsilon }}_{{i},{t}}\) is the residual estimated by Eq. (3).

We construct the measure of negative stock weekly return bias (NCSKEW) based on the firm’s weekly idiosyncratic return. This statistic represents the distribution’s asymmetry, as Eq. (5) expresses:

$${{{ {NCSKEW}}}}_{i,t}=-\left[n{(n-1)}^{\frac{3}{2}}\sum {W}_{i,t}^{3}\right]/\left[(n-1)(n-2){(\sum {W}_{i,t}^{2})}^{\frac{3}{2}}\right]$$
(5)

Subscript t represents year t; n represents the trading weeks for firm i in year t. Greater NCSKEW values denote more significant negative return skew factors and increased risks of stock price collapse.

The second indicator is the weekly return up/down volatility ratio (DUVOL), which compares the dispersions of above-average and below-average returns and is calculated by Eq. (6):

$${{{{DUVOL}}}}_{i,t}={\mathrm{ln}}\left\{\left({n}_{{ {u}}}-1\right)\sum _{{{{DOWN}}}}{W}_{i,t}^{2}/\left[\left({n}_{{ {d}}}-1\right)\sum _{{{{UP}}}}{W}_{i,t}^{2}\right]\right\}$$
(6)

where t denotes year t; nu and nd, respectively, denote the number of weeks in which Wi,t was above and below its annual average return. Larger DUVOL values indicate distributions of returns that skew to the left, which are associated with greater stock price vulnerability.

Baseline model

Referring to existing literature (An et al., 2020; Xu et al., 2014; Kim et al., 2011), we estimated model (7) to test our hypothesis:

$$\begin{array}{l}{{\rm{CrashRisk}}}_{i,t+1}={\alpha }_{0}+{\alpha }_{1}{{{{CSR}}}{{\_}}{{{GAP}}}}_{i,t}+{\alpha }_{2}\sum {{{{Controls}}}}_{i,t}\\\qquad\qquad\qquad\qquad+\,{\alpha }_{3}\sum {{{{Year}}}}_{i,t}+{\alpha }_{4}\sum {{{{Indus}}}}_{i,t}+{\varepsilon }_{i,t}\end{array}$$
(7)

As shown in model (7), we orient the explanatory variables’ relevant indicators to the latter year (CrashRiski,t+1). CrashRiski,t+1 represents the risk of stock price crash of firm i in year t + 1 (measured by NCSKEWi,t+1 and DUVOLi,t+1, respectively, and CRSAHi,t+1 in the robustness test), while the core explanatory variable CSR_GAPi,t represents the degree of CSR decoupling of firm i in year t. Controls are a set of control variables: market volatility (Sigma), market returns (Ret), financial leverage (Lev), book-to-market ratio (BM), firm size (Size), return on total assets (ROA), information opacity (Opaque), average monthly excess turnover (Dturn), year-fixed effects, and industry-fixed effects (Indus). Table 1 represents detailed definitions of the control variables. According to our research hypothesis, if the coefficient \({\alpha }_{1}\) is statistically significant and positive, CSR decoupling likely elevates SPCR.

Table 1 Variable definitions.
Full size table

The above content outlines the principles of sample selection, measurement methods for the main variables, and the model for the baseline regression. Next, we will provide descriptive statistics for each variable and apply the established model for multiple regression testing to identify the causal relationship between CSR decoupling and SPCR, thus validating our hypothesis H1. Furthermore, we will conduct a series of robustness tests to enhance the reliability of our conclusions. Specifically, we use the Heckman two-stage model, further controlling for fixed effects and altering the measurement methods of the main variables. Finally, based on our main findings, we will analyze the association between CSR decoupling and SPCR in different agency risk contexts to verify our hypothesis H2 through sub-sample tests in the mechanism examination section.

Empirical results

Descriptive statistics

Table 2 reports the descriptive statistics for the main variables. In particular, the means (medians) of the explanatory variables NCSKEW and DUVOL are −0.358 (−0.239) and −0.296 (−0.233), correspondingly, consistent with previous literature (Xu et al., 2014). In addition, the standard deviation of CSR decoupling CSR_GAP is 1.507, indicating that the extent of CSR decoupling varies significantly among different companies.

Table 2 Descriptive statistics.
Full size table

Baseline regression results

Table 3 shows the baseline analysis results, investigating the nexus between CSR decoupling and SPCR. Specifically, columns (1) and (2) include only relevant control variables. Columns (3) and (4) further control industry- and year-fixed effects. We observe that the coefficient on CSR decoupling (CSR_GAP) is notably positive at a significant level of 1% regardless of whether using NCSKEW or DUVOL as the explanatory variable. Regarding economic significance, as column (3) shows, every 1 percent increase in CSR decoupling increases crash risk by ~2.09%. In conclusion, CSR decoupling exacerbates the risk of stock price collapse, supporting our hypothesis H1.

Table 3 Baseline regression.
Full size table

Endogeneity

The findings may face potential endogeneity problems because companies more likely to experience stock price collapses might be more motivated to use CSR decoupling to conceal their negative internal information, which may lead to biased results. In addition, as CSR reporting is primarily voluntary, self-selection bias may exist in observations. To alleviate it, we use the Heckman two-stage analysis and control for firm-fixed effects to support our baseline findings further.

Heckman two-stage model

Initially, we construct a Probit regression model to estimate the likelihood of a company disclosing a CSR report. CSRD is binary, denoted as 1 if a firm published a CSR report in a given year and 0 if it did not. The independent variables include the following ones that influence whether a company chooses to disclose a report: book-to-market ratio (BM), firm size (Size), financial leverage (Lev), return on total assets (ROA), ownership concentration (Top1), proportion of independent directors (Indep), CEO duality (DUAL), and market value (TobinQ). In addition, following Zhang (2022), we include the mandatory disclosure variable (Mandatory), which equals one if the firm is obligated to release CSR reports and 0 for voluntary disclosure. In the subsequent stage, the inverse Mills ratio obtained from the first stage is introduced as a control variable.

Table 4 presents the regression results of the Heckman two-stage model. In the first stage regression, the coefficient on the mandatory disclosure variable (Mandatory) is significantly positive at the 1% level, indicating that firms that are required to disclose CSR reports are more likely to publish their social responsibility reports (Li et al., 2021). After controlling IMR to our model (7), the results in columns (2) and (3) reveal significant positive correlations between CSR decoupling and the two indicators of SPCR at the 1% and 5% levels, respectively, consistent with the baseline regression results.

Table 4 Heckman two-stage model.
Full size table

Controlling for firm-fixed effects

Following Shahab et al. (2021), we control firm-level fixed effects to alleviate concerns related to unaccounted-for firm-level time-invariant missing variables in the model (7). As Table 5 shows, the coefficients of CSR decoupling are significantly positive at the 10% level, suggesting our baseline findings are still robust.

Table 5 Firm fixed effects.
Full size table

Additional robustness tests

Omitted variables

First, the lagged stock price crash risk indicator is added as a control variable. In addition, referring to Xu et al. (2014), Yu et al. (2023), and Wan et al. (2024), corporate governance variables such as the proportion of independent directors, board size, and separation of duties are added to the model (7). Table 6 reveals the regression results, and the baseline relationship remains robust after adding the two sets of control variables.

Table 6 Regression analysis with omitted variables.
Full size table

Alternative samples

Given the anomalous phenomenon of thousands of stocks rising and falling together in our stock market in 2015, this extreme situation may negatively impact stock market risk. Therefore, we excluded the observations from 2015 and re-regressed the model (7). As Table 7 shows, the coefficients of CSR_GAP on the risks of stock price crashes are 0.0193 and 0.0098, respectively. The significant positive correlation between CSR decoupling and two indicators of crash risk persists.

Table 7 Using the alternative sample excluding outliers.
Full size table

Alternative measurements for SPCR

Following Kim et al. (2011), the distribution of extreme weekly returns (CRASH) is a suitable SPCR indicator. The variable equals one if the following inequality is satisfied one or more times during the year for stock i, indicating that the stock has experienced a crash event, and 0 otherwise. Equation (8) explicitly expresses the inequality on which the determination is based:

$${W}_{i,\tau }\, <\, {{{Average}}}\left({W}_{i,\tau }\right)-3.09\sigma$$
(8)

where \({{\rm {Average}}}\left({W}_{i,\tau }\right)\) is the mean of weekly returns, σ is the standard deviation of weekly returns, and 3.09σ corresponds to an interval with a normal distribution probability of <1%. We use logistical regression to re-run our model (7) because of a dummy dependent variable. As reported in column (1) of Table 8, the coefficients of CSR_GAP are significantly positive at p < 0.05, supporting our hypothesis H1.

Table 8 Alternative measurements.
Full size table

Alternative measurements for CSR decoupling

When measuring optimistic tone, the difference between the numbers of words with positive tones and negative tones was calculated by dividing by the total number of words in the CSR report, expressed in Eq. (9) below.

$${\rm{Alternative\; Optimistic\; Tone}}=\frac{{\rm{POSPCT}}-{\rm{NEGPCT}}}{{\rm{N}}}$$
(9)

We then subtract the value of the external performance indicator from the optimistic tone and denote this new measurement as CSR_GAP2. As columns (2) and (3) show, the coefficients of CSR_GAP2 are still significantly positive for NCSKEW and DUVOL, respectively, indicating that this paper’s findings continue to hold.

Further analysis

This section focuses on subsample testing to analyze the mechanism behind the relationship between CSR decoupling and SPCR, thus validating our second hypothesis and strengthening the reliability of the previous findings. We examine the relationship between CSR decoupling and SPCR from the perspectives of different levels of agency risks, focusing on executives’ experiences, monitoring pressures they encounter, and the institutional environment in which they operate.

Executives experience

Empirical research based on imprinting theory has confirmed that executives with specific experiences have particular influences on corporate behavior (Marquis and Tilcsik, 2013). Specifically, we limit executives’ personal experiences to academic experience and overseas experience. Academic training cultivates rigorous visionary thinking, making executives more prudent in logic, self-disciplined, and conservative in behavior (Francis et al., 2015). Therefore, executives with academic backgrounds are less likely to engage in behaviors harmful to investors. Hiring executives with academic experience can reduce agency risks. Moreover, they possess elevated social ethics and heightened senses of social responsibility (Cho et al., 2017), so they may be more inclined to “practice what they preach” when expressing CSR. Meanwhile, overseas experience indicates good education and expertise, and executives with overseas experience improve corporate governance (Giannetti et al., 2015). In the Chinese context, the concept of a professional agent is not as prevalent as in foreign capital markets. Executives with overseas experience tend to have a deeper understanding of agency relationships and are more inclined to consider the interests of investors (Cao et al., 2019). Therefore, we hypothesize that when executives have academic or overseas experience, the probability that they also have rational and objective philosophies is higher. Companies with executives lacking academic or overseas experience face relatively higher agency risks. Hence, CSR decoupling’s efficacy on SPCR is more significant.

Referring to Zhao et al. (2022), we define an executive as having academic experience if they have full-time teaching experience in a university, full-time research experience in a scientific research institution, or related research experience. Following Xu and Hou (2021), we identify a manager who has worked in overseas postings or studied in overseas universities as having overseas experience. This paper distinguishes between executives’ academic and overseas backgrounds for subsample regressions. As Panel A of Table 9 shows, CSR decoupling’s efficacy on SPCR is only significant in the subgroup of executives without overseas or academic experience, which is in line with our expectations and validates Hypothesis 2.

Table 9 Mechanism tests.
Full size table

Monitoring forces

From the perspective of external monitoring forces, this paper examines the impacts of institutional investors and media attention on mitigating CSR decoupling’s consequences. Institutional investors in the Chinese capital market tend to possess more professional expertise and proactivity. They are also particularly conscious of and driven to oversee managers’ tendencies to withhold unfavorable news, which lets stock prices more accurately mirror firms’ intrinsic values (Callen and Fang, 2013). The media, an effective information intermediary, can also exercise a public supervisory function with its coverage, allowing investors to obtain more information and thus significantly reducing future SPCR (An et al., 2020). Therefore, we believe that higher institutional investor holdings and more media attention are conducive to improving information transparency and reducing asymmetry between internal and external aspects of the companies, resulting in reduced agency risks. Companies with higher institutional investor ownership and greater media attention tend to exhibit weaker relationships between CSR decoupling and SPCR.

Following Liu and Wang (2022), we use the sum of reports from both newspapers and online media to measure media attention. Then, we calculate institutional ownership based on the shareholding proportion of institutional investors and use the industry annual median to divide the sample into high and low external monitoring subgroups. Panel B of Table 9 reveals that decoupling’s efficacy on SPCR is significant in the subsample with lower institutional investor shareholding and media attention. The results also validate Hypothesis 2, indicating that the impact of CSR decoupling on SPCR in contexts with higher agency risks is more significant.

External governance environment

Marquis and Qian (2014) revealed that the external governance environment can positively influence the accessibility and credibility of information concerning a company’s greenwashing activities and its performance available to internal and external stakeholders. Areas with high marketization and generally sound law environments have better investor protection and higher levels of investor protection negatively affect stock price synchronization (Jin and Myers, 2006). These findings suggest that markets with more well-developed institutional environments can convey more information to investors about firm characteristics and increase information transparency, thus reducing agency conflicts.

To validate the above hypothesis, we construct a subgroup test. Following Wang et al. (2008) and Yu and Pan (2008), we use marketization and legal environment levels to measure the external governance environments where firms operate. Specifically, our measure of marketization level is Wang et al. (2019) marketization index. It is a composite index that integrates government roles, market efficiency, property rights protection, financial system, and corporate competitiveness. It is calculated through weighting and reflects the level of marketization in the province where the company operates. The measurement of the legal environment level is based on the completion rate of economic-legal cases (i.e., the fraction of cases finished compared to the total number of cases received) in each province. We divide the entire sample into high and low-marketization/legal environment subsamples. Specifically, suppose the marketization or legal environment level of the province where the company is located is higher than (less than or equal to) the median value of all the observations grouped by province and year. In that case, the company falls into the high (low) marketization or legal environment subsample.

As shown in Panel C of Table 9, CSR decoupling’s impact on SPCR is significant only in the group with low marketization and legal environment levels. The regression result aligns with our expectations and further validates our earlier hypothesis that the relationship between CSR decoupling and SPCR is more pronounced in contexts with higher agency risks.

Discussion and conclusion

A growing number of public firms are issuing CSR reports as a crucial method to obtain ethical and sustainable development reputations (Perez-Batres et al., 2012). However, management may exaggerate publicly disclosed CSR information and use CSR reporting to conceal their unethical behaviors. This self-serving impression management, namely CSR decoupling (Hemingway and Maclagan, 2004), increases information asymmetry between firms and external investors, thus providing opportunities for management to hide unfavorable news for longer until the critical point when the firm’s actual CSR activities don’t meet the investor’s expectations arising from previous exaggerated CSR reporting, which eventually leads to stock price plunges (Jin and Myers, 2006). Hence, the higher the degree of CSR decoupling, the more likely stock price crashes.

This paper utilizes textual analysis to measure a firm’s CSR decoupling and examine CSR decoupling’s impact on firms’ SPCR and the corresponding mechanism. The findings indicate that CSR decoupling exacerbates SPCR. This conclusion remains valid across various robustness examinations. Based on the perspective of information asymmetry, the mechanism analysis explores the impact of agency risk on the relationship between CSR decoupling and SPCR and finds that the exacerbating effect of CSR decoupling on SPCR in situations of higher agency risk is more significant. Specifically, through sub-sample analysis, we find that the relationship between CSR decoupling and SPCR is only significant in companies without executives’ academic or overseas experience. Similar conclusions are also being discovered in subsamples with lower institutional investor ownership, less media attention, and weaker marketization and legal environments.

Previous studies ignore the two dimensions of CSR (disclosure and actual activities), leading to a confused relationship between CSR and SPCR (Kim et al., 2014; Bouslah et al., 2018; Quan et al., 2015; Hao et al., 2018). Our paper addresses this issue from the CSR decoupling perspective, reconciling CSR disclosure with actual CSR behavior and proving the negative consequence of CSR decoupling. Furthermore, prior studies have primarily focused on country- and firm-specific determinants of CSR decoupling (Velte, 2023), and this paper enhances the existing literature on the economic consequences of CSR decoupling. Finally, this paper contributes to the determinants of SPCR by enhancing empirical insights into CSR decoupling.

This paper provides the following practical implications. First, firms should cultivate an enterprise culture with a high sense of social responsibility, improve the mechanism for disclosing information publicly, and strengthen internal governance to deliver more accurate information to the stakeholders. Second, investors should enhance their awareness of risk prevention, learn to identify the financial and non-financial information disclosed publicly by firms, judge its authenticity and validity, and diversify risks reasonably. Finally, regulatory authorities should establish and improve relevant regulations on CSR disclosure, build a scientific and comprehensive supervision system, clarify the subject of accountability and disciplinary mechanisms, and encourage firms to fulfill their social responsibilities and avoid CSR decoupling actively.

Future research can address several limitations inherent in our study. First, we characterize CSR decoupling as the inconsistency between the Optimistic Tone of CSR reports and the factual CSR performance. The construction of Optimistic Tone measurement based on text analysis can effectively alleviate concerns related to using discrepancies between two external databases to calculate CSR decoupling (Velte, 2023). However, relying solely on the measurement of Optimistic Tone in CSR disclosures overlooks the richness of the information content and decision usefulness of CSR reports. Moreover, effective and accurate text analysis relies on specialized textual dictionaries specifically built for CSR reports. Therefore, future research could attempt to establish specialized dictionaries for text analysis of CSR reports to achieve more precise measurements of CSR disclosures. Furthermore, the sample period of this study did not include samples after 2019. This is because the comparability of CSR strategies and performance during the COVID-19 pandemic among listed companies with other periods is questionable. Meanwhile, the pandemic also affects listed companies’ stock price crash risks. Future research could explore the relationship between CSR decoupling and SPCR in the context of public crisis events such as the pandemic.