Do CEO Rhetorical Strategies A ﬀ ect Corporate Social Performance? Evidence from China

: How can chief executive o ﬃ cers (CEOs) persuade employees to participate in corporate social responsibility (CSR) activities, so as to enhance ﬁrms’ corporate social performance (CSP)? The purpose of this study is to examine the relationship between CEO rhetorical strategies and ﬁrms’ CSP. According to Aristotle’s classiﬁcation, we divide CEO rhetorical strategies into three categories: pathos, ethos, and logos, using the text analysis method. We apply a Probit model to predict whether CEOs use rhetorical strategies and then adopt ﬁxed-e ﬀ ect multiple regression models to measure the impact of various rhetorical strategies on CSP. An empirical analysis based on data on the listed manufacturing companies in the Shanghai Stock Exchange and Shenzhen Stock Exchange from 2014 to 2016 shows that both CEO pathos strategy and CEO logos strategy have positive e ﬀ ects on CSP; however, the relationship between the CEO ethos strategy and CSP is not signiﬁcant. Our ﬁndings contribute to upper echelons theory and CSR research and provide suggestions for CEOs to apply proper rhetorical strategies.


Introduction
Corporate social performance (CSP) is defined as a configuration of principles of social responsibility, social responsiveness procedures, policies, programs, and observable results for a business organization, as they relate to the firms' societal relationships [1]. A high standard of CSP can improve the relationship between companies and stakeholders; enhance the reputation of the company among clients, staff, regulators, and vendors [2][3][4][5][6][7]; and, ultimately, transfer into important strategic resources and capabilities [8]. Strategic decision-making and implementation of companies are influenced by top managers [9]. While research on the effect of chief executive officers (CEOs) on CSP has been widely discussed, the existing literature mainly focuses on CEOs' demographic characteristics [10][11][12], and how CEOs persuade employees to engage in CSR activities has been less illuminated.
To fill this gap, this study investigated how CEOs affect CSP from a new perspective, that is to say, how CEO rhetorical strategies toward CSR affect employee engagement and thereby improve CSP. Correct communication is one of the fundamental components of corporate social responsibility development [13]. Existing studies on CSR communication have particularly focused on external communication [14][15][16][17], which is designed to search for legitimacy [18,19]. However, firms have especially sought to make their CSR activities visible and accessible to both internal and external stakeholders [20]. Different from previous studies, this paper focuses on the CSR internal communication, which is also important, as CSR strategy requires acknowledgement within the company and CSR implementation relies on all employees' joint attempts.
Based on the persuasion model [21], this study aimed to fill the existing research gap by focusing on the following question: How can CEOs persuade employees to conduct CSR activities to enhance reasoning producing rationality [39,[49][50][51][52], the more convincing the reasons for management practice, the more likely they will be adopted and institutionalized.
According to upper echelons theory, top managers play an important role in strategic decision-making and implementation [9,53]. CEOs, given their high level of discretion [54], represent the company in professional, legal, and social contexts more than any other employee [22]. CEOs have a critical role in addressing CSR issues and are one of the main factors influencing CSP [55][56][57]. To make the correct rhetorical decisions, CEOs have to determine the nature of stakeholder demands, as well as their intensity and urgency [58]. While several perspectives on the role of CEOs have been presented in the literature [47,59], little attention has been paid to how they persuade employees to engage in CSR activities effectively. We have focused on a new perspective, that is the rhetorical strategy of CEOs, through which they help employees better understand the CSR strategy, thus forming a sense of common understanding in order to improve CSP.
Rhetorical theory holds that rhetorical purposes can be achieved through pathos, ethos and logos. Pathos means drawing on abundant emotions to induce empathy from audience. Ethos refers to the personality, ability, and prestige of a speaker. Logos means using powerful logical argument. Managers may adopt different rhetorical strategies to improve the efficiency of communication. Hyland explored how CEOs use non-propositional materials to achieve rational, credible, and emotional appeals in company annual reports [60]. For Airbnb's intended users, the three types of rhetorical strategies have different effects on buying behavior [61]. New decisions may be effective only if rhetoric rationalizes their effectiveness [62]. From the perspective of rhetoric, this rationality is influenced by different rhetorical styles. So, what are the specific differences in the mechanisms and effects of these three rhetorical strategies on the audience? Will this be different when it comes to CSR activities?

Affective Appeals: Creating a Pathos
Pathos strategy, or values rhetoric, is "putting the audience in the appropriate mood, by playing on its feelings" [63], which aims at creating emotions and passions to induce the audience and helps to construct a set of shared values within and without firms [12]. Pathos indicates the elements that affect the recipients emotionally, and an emotional appeal is strong in persuasion [64]. Appeals to pathos are the "persuasive reasons in an argument that derive from a community's mostly deeply and fervently held values" [65].
Studies have shown that language can affect individual's emotions. In order to improve the persuasiveness of the information, speakers should use emotional expressions. The process of persuasion includes not only the cognitive response of the recipients but also the emotional response [66]. Burke emphasized the common feeling between the speaker and the audience [67]. Through the "field of pathetic expressions", a speaker can acquire the capacity to stir a whole crowd of people at the same time into a frenzy and then to enchant them with magic [13].
Emotions can influence an individual's behavior. Expressions with strong feelings can change people's attitudes and even influence people more than rationality [21]. Emotions expressed by role occupants influence the behavior of others [68]. Comments on views with emotions on the internet will form online public praise, which will have a significant impact on product sales [69,70]. Considering text emotion and other factors can also improve the accuracy of crowdsourcing project financing prediction partly due to its persuasive effect [71].
In this study, the pathos strategy used by the CEOs refers to the emotional expressions in the content of CSR, including quoting comments from stakeholders and using poems, metaphors, parallelism, exclamation, etc. That is to say, creating a strong ethical and moral emotional atmosphere that affects the emotions of stakeholders can improve cognitive legitimacy [19]. By using this strategy, CEOs attempt to create a powerful corporate culture around CSR values. CEO pathos strategy can lead to an irrational and unconscious emotional feedback among stakeholders, especially for the employees. The transfer of CSR strategy in this way will stimulate resonance, change conflicting feelings, and induce employees' organizational identity and enthusiasm for participation. Therefore, we predict the following: H1: CEO pathos strategy is positively related to CSP.

Credibility Appeals: Creating an Ethos
Pathos is only one dimension of rhetorical strategies. A persuasive discourse also depends on the persuader's credibility. Ethos, or normative rhetoric, is "establishing the persuader's good character and hence credibility" [63]. Attention must be paid to establishing a competent, credible, authoritative, and honest personality by the persuader [72]. According to the power model framework proposed by Finkelstein, a CEO's personality can win social support and relieve the pressure of uncertainty caused by a complex external environment.
Ethos can form authority and build credibility. According to the persuasion model [21], the persuader is the first element of persuasion, so the objectivity and credibility of the persuader are the basic conditions of successful persuasion. The good character of the persuader builds credibility [64]. Persuaders need to demonstrate their reliability and authority to audiences, which can be shown through awards, education, professional level, working experience, and so on [72]. One must always behave ethically and responsibly towards participants. Keeping this in mind is crucial because of the strength of the rhetoric [73]. Research shows that ethos strategies have a greater impact on fundraising letters than any other rhetorical strategies [74].
Reliability can reduce information asymmetry and build trust relationships [75], thus affecting audiences' behavior. Highly trustworthy information often brings favorable associations [21], reduces refutation to positions [76], and influences attitudinal change [77] through related thinking [78]. If the messages are from trustworthy sources, they tend to be more influential [79]. Early studies on the effect of source credibility have found that speakers with high credibility are more persuasive than those with low credibility [80,81]. In the process of decision-making, people are accustomed to listening to the opinions of experts or authorities. In other words, individuals are more likely to be persuaded by the requester who is a domain expert [82]. If the project text highlights the expert status of the fundraisers, more support from investors will be obtained [83]. The higher the corporate credibility of sponsoring open source projects, the greater the internal incentives of source code authors and the contribution to open source projects [84]. Experiments show that using persuasion that emphasizes the credibility of the sender can significantly improve the questionnaire response rate [73].
In this study, the CEO ethos strategy refers to the obtaining of recognition and conviction by clarifying reputation, such as awards, professional certifications, personal qualities, good characters, evaluation ranking, etc. Responsibility not only forms a direct constraint on the employees but also brings about some restraints on managers. If CEOs only appoint employees without consideration of their own constraints, the role of responsibility will not be reflected, and the enthusiasm of workers cannot be mobilized. Research shows that when firms are under moral pressure from stakeholders, they will lobby stakeholders by resorting to ethos strategy to improve their moral legitimacy [19]. Through ethos strategy, CEOs can gain the trust of employees. The persuasion of the audience comes from the understanding of the persuader's personality, position, and professional ability. CEO ethos strategy will improve the moral legitimacy of firms and increase staff's sense of pride and organizational identity. Employees will have a deeper understanding of the CSR strategy and then participate in CSR activities more effectively. Therefore, we predict the following: H2: CEO ethos strategy is positively related to CSP.

Rational Appeals: Creating a Logos
Persuading the audience by stimulating the corresponding emotions or through personal charm are important means, but persuasion is essentially a process of reasonable argument. Besides appealing to emotion and personality, arguments should also be put forward to express the rational demands of the persuaders. Logos strategy, or instrumental rhetoric, is defined as logic, reasoning, or "rational argument" [63] and is further extended to the institutional level [85]. Institutional logic provides a "practical guide to action" [86]. Specifically, they encode standards of legitimacy through which the relationships among role identity, strategic behavior, and organizational form can be constructed and maintained [87,88].
Logos affects individual's cognition. In some formal cases, such as in philosophy and law, logos strategy is almost the only option [89]. Although scholars have been thinking for a long time about how managers establish a sense of mission, they neglect that this sense is achieved when multiple employees have the same understanding of the work objectives. Important responsibilities of managers include conveying vision and values to help employees understand the ultimate purpose of their work. Managers can use logos rhetorical strategies to achieve common perception among employees [90].
Cognition influences an individual's behavior. Speakers reach the conclusion by logical reasoning, which is an effective way to crush rumors [91]. In the analysis of the text content of charitable fundraising initiatives, it has been found that investors are more concerned about the text content than the emotion [92]. In a study on letter-raising, the frequency of emotional statement is higher than that of logical statement, whereas the impact is not [93]. Among the most successful crowdsourcing projects, the sponsors have adopted logical language in project descriptions to various extents [94], and logical appeals also have a significant impact on public service advertising when studying the effect of persuasion [95].
In this study, the CEO logos strategy refers to the use of a rational framework to convey CSR strategies. Through concrete ways of rational thinking, rigorous reasoning, and illustration, CEOs convey why firms implement CSR and what benefits CSR will bring to stakeholders. In particular, this logical approach can intuitively convey CSR strategies to employees and improve employees' pragmatic legitimacy [19], which contributes to employee participation and execution, making CSR implementation timelier and more effective. Therefore, we assume the following: H3: CEO logos strategy is positively related to CSP.

Data and Sample
Listed manufacturing firms in Shanghai and Shenzhen Stock Exchange from 2014 to 2016 were chosen as samples. There are several reasons for this. First, the samples selected from the manufacturing firms can guarantee the comparability and efficiency of the information and regulate the industry's impact. The same industry implies essentially the same external environment, and thus, the effect of CEO rhetorical strategies and other factors on CSP can be more appropriately measured. Second, manufacturing companies accounted for over 60% of the listed companies in China, according to industry classification indices, demonstrating its significant position. In addition, the reasons why we chose data for 2014-2016 are as follows: First, in 2006 and 2008, Shenzhen and Shanghai Stock Exchange respectively issued notices on listed companies' disclosure of social responsibility information. Since 2009, the disclosure of social responsibility information has been on the rise. Companies publishing reports on corporate social responsibility have gradually stabilized since 2014. Hence, we chose 2014 as the starting point for our studies. Second, taking into account the lagging effect of CEO rhetorical strategies on CSP, our CSP data lags behind CEO rhetorical strategy data by one year. CSP data have not yet been disclosed since 2017, so we chose 2014-2016 as the timeframe for our research. Finally, we measured the independent variables and control variables with t year, and measured the dependent variables with t+1 year. That is, independent variables and control variables were selected from 2014 to 2016, and dependent variables were selected from 2015 to 2017.
Three main data sources were used for this study: China Stock Market & Accounting Research Database (CSMAR), HeXun.com (http://stockdata.stock.hexun.com/zrbg/Plate.aspx), and Golden Bee (https://www.csr-china.net). CSR scores were obtained from Hexun, a leading Chinese professional financial company. CEO rhetorical strategies were coded from CSR reports listed on Golden Bee, and the missing reports were supplemented through firms' official websites. The main data involved in the control variables at the CEO-level and firm-level were obtained from CSMAR, and we supplemented the missing data by accessing different reliable database and websites, such as Wind and CNINFO indices. Following previous studies, we excluded ST and *ST companies and winsorized CSP scores at the upper and lower 1% tails of the distribution. After the above treatment, 738 observations were finally obtained.

Dependent Variables
We took CSP as our dependent variable. We adopted the professional CSP assessment system from the Hexun website. Hexun.com has a professional assessment system for CSP, and it is considered to be the latest and most authoritative database with appropriate rating information. Hexun scores CSP in five first-class indexes: (1) responsibility for shareholder; (2) responsibility for employees; (3) responsibility for suppliers, community, and consumers; (4) responsibility for environment; and (5) responsibility for society. Each first-class index also includes a corresponding second-class index and third-class index. This paper uses overall CSR scores as the measure of CSP.
When studying the relationship between CEO rhetorical strategy and CSP, CEO rhetorical strategy often fails to produce significant effects in a short time due to the lagging effect. Therefore, we selected the CSP of the following year as the research object to avoid the problem of reverse causality in empirical research [11,96]. The independent variables were consequently assessed by CSR reports in t year, whereas the dependent variables were assessed by CSR scores in t+1 year.

Independent Variables
CEO rhetorical strategies were measured by encoding CSR reports using text analysis. Text analysis, or content analysis, is defined as a systematic, replicable method to compress text phrases into classifications of content based on specific coding regulations [97][98][99]. All valuable communication content that can be recorded and preserved in text or non-text form can be the subject of content analysis [97]. Building a direct measure of CEO rhetorical strategy has been challenging for researchers, especially those who want to conduct statistical analysis. Variables must be built from information coherent across companies over time and suitable for quantitative measurement. There are many techniques for measuring the features of top managers, including standardized questionnaires. However, as CEOs are usually reluctant to engage in questionnaire surveys, the low questionnaire recovery level often leads to the deviation of the results [100]. For this reason, text analysis based on archival data has been regarded as an effective method to measure managerial characteristics. This method is based on the Sapir-Whorf hypothesis. Features can be represented by words used by individuals in social activities, and words can accurately reflect the individuals' inner thoughts and psychological cognition [101]. After years of exploration, it has been discovered that text analysis is highly reliable and valid and can be helpful in longitudinal research models [102]. Investigators compared human-coded content analysis to computerized coding of the same text communications. The results show that the two methods have the same effect [103].
Through a document or report written or approved by a CEO, his characteristics can be captured accurately. There is abundant evidence that the annual report of a listed company is a reasonable source of data for the measurement of executive cognition [104]. Similar to annual reports, CSR reports are existing text written or approved by CEOs, so they reflect the style of CEOs' speech to some extent. CSR reports are particularly designed to convey the managerial vision and approach toward CSR [99] and are one of the main means of CSR communication [105], and they can convince investors that the company is pursuing an effective strategy [106]. Compared with laws and regulations, the wording of CSR reports is more diverse and can largely reflect the rhetorical characteristics of the CEO. Measuring CEOs' rhetorical strategy through CSR reports contributes to the horizontal comparison among companies. Meanwhile, researchers can obtain CSR reports of the same company from different years, which is conducive to the construction of panel data. The taxonomy of CEO rhetorical strategy employed here is based on the classification defined by Aristotle. Text analyses were completed by two coders. At the training stage, the coders finished a subset of CSR reports and addressed the outcomes of their coding together. Then, both coders read each report and debated any disputes until they were settled. The following is the specific coding method for the three rhetorical strategies.
Pathos. The main means of pathos strategies are the use of poems, interjections, parallelism, metaphors, and other strong emotional statements. Through the use of these strong emotional words, employees can become more immersed, so as to achieve the effect of empathy and combine the CEO's goals with the employees' goals. Once one of the keywords or statements above was found in the CSR report, pathos was coded as "1".
Ethos. One of the approaches used in ethos strategy is to demonstrate the CEO's expertise, authority, personal qualities, and good characters to enhance their credibility. Another way is to show the success of the firm, such as honors, certifications, and awards that it has won. Showing corporate success is the easiest way to gain credibility, in which case CEOs may be able to use corporate resources to emphasize the authority of their claims. Once one of the keywords or statements above was found in the CSR report, ethos was coded as "1".
Logos. The main method of logos strategy is using logical statements. To a large extent, CEOs can state their views and construct arguments clearly through logical reasoning, such as adopting conjunctions (and, or, but, etc.), adverbs (however, thus, etc.), or prepositional phrases (due to, in spite of, etc.). Once one of the keywords or statements above was found in the CSR report, logos was coded as "1". Table 1 shows the specific explanations and examples of three rhetorical strategies. Table 1. Chief executive officer (CEO) rhetorical strategy classification.

Category Explanations Examples
Pathos Adopt emotionally appealing text to impress investors with strong emotions by using poems, exclamations, parallelism, metaphor, and so on.
"Haier should make due contributions to society and human beings. As long as our love for society and human beings is sincere forever, society will also recognize us forever, and Haier will be as eternal as the sea." (Haier, 2016) "Fulfilling social responsibility is the obligation and mission of an enterprise. If the development of an enterprise is separated from society, it will . . . Logos Convince the audiences with rigorous statements to gain support by using logical words, such as "in favor of", "therefore", and "only if".
"Society is the foundation of an enterprise. As a social citizen, an enterprise must integrate into social groups, take social responsibilities, and positively interact with various social organizations and individuals. Only by integrating social responsibility into corporate culture and transforming it into corporate values and business philosophy, can it create a better social atmosphere for the development of the enterprise and maintain its vitality for long-term sustainable development." (Shan Shan, 2014)

Control Variables
Taking into consideration the possible relationship between the CEO's background, the firm's characteristics, and CSP, we controlled for CEO-level and firm-level potential confounding factors to rule out alternative explanations. The description of each variable is presented in Table 2.

CEO-Level Control Variables
Research has shown that views about the significance of CSR vary with age; compared with younger CEOs, older CEOs may place more emphasis on corporate social responsibility and will be more willing to assume more social responsibilities to meet the higher needs [107], so we controlled for CEO age. We also controlled for CEO education. With the improvement of a CEO's education level, his values are more easily affected by the concept of social responsibility [108,109]. Education was measured by a five-point scale, coding a PhD degree as "5", master's degree as "4", bachelor's degree as "3", junior college "2", and others or below as "1". Additionally, we controlled for CEO tenure and CEO duality (coded as a 1 if CEO is also a chairman of the board), which can reflect the actual power of the CEOs and may influence their ability to promote CSR projects [110]. CEO overseas background (coded as a 1 if CEO has an overseas background) was also included. The experience of studying or working abroad will affect the social cognitive orientation of executives [108]. People with overseas background will have a global awareness, pay more attention to the needs of others, and have a higher sense of responsibility [111]. Finally, we controlled for CEO government background (coded as a 1 if CEO has government background). Studies show that charitable donations in China are an important means for executives to establish and maintain relations with the government [112]. CEOs with government background tend to shoulder more social responsibilities. CEOs with government working experience are more likely to have an association with the government, and political connections are valuable intangible assets for firms [113,114], which significantly affect a firm's business decision-making and performance [115,116].

Firm-Level Control Variables
The following firm-level factors that might influence CSP were included as controls. The higher the current ratio, the smaller the leverage, and the lower the operating risk of the enterprise, the more funds are available for CSR activities. Therefore, we controlled for firm current rate and leverage. Compared with firms in poor condition, profitable firms are more capable of carrying out CSR activities [7], so return on assets (ROA) was controlled. In addition, large firms receive more social attention than small ones, and these managers tend to undertake more corporate responsibilities for the sake of corporate image [117], so we controlled for firm size, which is measured as the natural logarithm of assets. Firms with a short history may pay more attention to economic interests and ignore corporate social responsibility, while firms with a long history have formed their business philosophy, so we controlled for firm age, measured by years since firm's initial public offering. A dummy variable for state ownership was also included. In the first-stage equation of the Heckman model, another two factors were taken into consideration. The financial performance may influence the resources of a firm that can be used for CSR activities, so we controlled for prior year performance by including the firm's lagged ROA. To account for the possibility that CSP in previous years might have an impact on CEO rhetorical strategies, we included lagged CSP in the model as well.

Estimation Method
Keeping in mind the above analysis, the effect of CEO rhetorical strategies on CSP was further explored through the following equation, where Score is the independent variable and X is a set of control variables that are expected to have a bearing on CSP. Pathos, Ethos, and Logos are dummy variables that represent the different types of rhetorical strategies. β 1 -β 4 are the parameters to be estimated, and ε is an error term.
A fixed-effect model with panel data was used to solve the problem of model misspecification and examine whether CSR will affect CSP. One main question of interest is whether the coefficient of pathos, ethos, and logos measured a real level of the effect of rhetorical strategies on CSP. Because our sample is confined to CEOs that use rhetorical strategies, the test of the hypothesized relationship between rhetorical strategies and CSP was limited. Simply regressing with this sample of CEOs who use rhetorical strategies would not be suitable, because those CEOs may have distinct features from those who do not. Thus, it is possible that factors which affect whether CEOs use rhetorical strategies could also be related to the dependent variable, the CSP of the firm. This suggests that the coefficients of the terms associated with rhetorical strategies would be correlated with the error term, and thus, the ordinary least squares (OLS) estimates of those coefficients would be biased.
The effect of rhetorical strategies on CSP was estimated using the Heckman selection model to avoid such sample selection bias [118]. The Heckman selection model includes two equations: The first (selection) equation estimates the likelihood that CEOs use rhetorical strategies by applying a Probit model to the entire sample of firms. A term of adjustment called the "inverse Mills ratio" (λ) was calculated at this stage. In the second equation, the sample was restricted to the sample of CEOs using rhetorical strategies. The CSP model was re-estimated in this equation, with the "inverse Mills ratio" included as a control variable. The Heckman two-stage model thus corrected for sample selection bias based on information representing all companies in the population.

Descriptive Statistics and Correlation Analysis
Tables 3 and 4 demonstrate descriptive statistics and correlation matrices for the major variables used in this study. Table 3 shows the variables used in the first-stage Probit model of the two-stage Heckman analysis. As is seen in Table 3, among the 738 observations of manufacturing firms, 89% were involved in rhetorical strategies. As would be expected, lagged ROA was positively related to the rhetoric choice dummy variable. The average age of the CEOs was 50.54 years old, the average educational background was bachelor's degree and master's degree, and the average tenure was 57.71 months. A total of 21.9% of CEOs concurrently served as the chairman, 6.2% had an overseas background, and 14.9% had a government background. Table 4 presents the descriptive statistics and matrix of correlation for the main variables used in the second stage of the Heckman analysis. As is seen in Table 4, among the 657 observations of manufacturing firms, the mean CSP was 35.90, and the standard deviation was 22.42, indicating that the overall CSP was not high and varied widely. On average, 36.4% of firms used pathos strategy, 98.3% used ethos strategy, and 29.4% used logos strategy. It can be seen that most companies used the ethos strategy, while relatively few of the other two rhetorical strategies were mentioned. The average age of CEOs was 50.55 years old, the average educational background was bachelor's degree and master's degree, and the average tenure was 59.19 months. A total of 22.1% of CEOs simultaneously served as the chairman, 6.4% had an overseas background, 14.5% had government work experience, and 59.4% of the sample firms were state-owned.    Table 5 shows the results of the first-stage Heckman selection model, which is a Probit regression of rhetoric choice against factors thought to influence whether a CEO decides to use rhetorical strategies. The dependent variable was a dummy variable indicating whether a CEO uses rhetorical strategies. A one-year lag was used between the dependent and the independent variables. The results presented are based on two first-stage rhetoric choice models. Model 1 was a baseline model that consisted of CEO-level variables and lagged ROA. Model 2 introduced lagged CSP as an additional independent variable, which was expected to have an impact on a CEO's rhetorical strategies. The first-stage analysis findings were compatible with expectations. In Model 1, firms with a higher level of ROA were found to be more likely to use rhetorical strategies. In Model 2, the coefficient on lagged CSP had a positive sign, as expected. The change in χ2, or the deviance, (6.63, p < 0.001) at the bottom of Table 2 indicates that Model 2 had a better fit. Thus, the results from Model 2 are used to formulate the inverse Mills ratio for the multiple regression estimates presented in the second-stage CSP model.   Table 6 provides the outcomes from the Heckman model's second-stage estimation using the inverse Mills ratio from the Probit model in Table 5 (Model 2) to account for selection bias in rhetoric choice dummy data. In order to test the hypothesized relationship between CEO rhetorical strategies and CSP, multiple regression analysis was introduced to the model. Because the scale and units vary across different variables, standardized coefficients were calculated to make them directly comparable. Model 1 reports the impacts of several fundamental control variables: CEO age, education, tenure, duality, overseas background, government background, firm current rate, leverage, return on assets, firm size, firm age, and ownership. Most of the variables have the expected signs. Tenure (b = −0.490, p < 0.001), leverage (b = −0.405, p < 0.05), ROA (b = −0.249, p < 0.05), firm size (b = −0.949, p < 0.05), and firm age (b = −0.725, p < 0.05) had negative and significant effects on CSP. This demonstrates that CSP was better explained by these five control variables. Adding these factors was useful for more correctly analyzing the relationship between rhetorical strategies and CSP. Age, overseas background, and government background were found to affect CSP positively but not significantly, while other control variables were found to affect CSP negatively but not significantly. The results above show that CEOs' overseas background, government background, and state-owned firms might be attached to higher CSP. In the subsequent three sets of models (Models 2-4), rhetorical strategies were added into the equation to test the main effect. Model 2 tested the main effect of the key independent variable of pathos strategy. The coefficient was positive and significant (b = 0.337, p < 0.001); this finding supports Hypothesis 1, which states that CEO pathos strategy contributes to CSP. This relationship was consistent in all of the models. Model 3 tested Hypothesis 2 by adding the independent variable of ethos strategy to Model 1. It turned out that the effect of ethos strategy was not significant, which means that the CEO's ethos strategy had no significant influence on CSP; thus, Hypothesis 2 was not supported. Model 4 tested the relationship between logos strategy and CSP. The coefficient of logos strategy was positive and significant (b = 0.206, p < 0.001), indicating that logos strategy contributed to CSP. The result supports Hypotheses 3, and in all models, this relationship was consistent. Model 5 was the full model including all the terms, showing that the effects of the control and independent variables in our research were generally consistent across the various models. Pathos and logos strategy showed significant and positive effect (b = 0.331, p < 0.001; b = 0.184, p < 0.001) on CSP, while ethos only showed marginal positive effect (b = 0.109, p < 0.05) on CSP, so our findings could be fairly credible.

Discussion
Anchored in upper echelons theory [9] and stakeholder theory [24], our study sought to explore how CEOs persuade employees to take an active part in CSR activities by using relevant CSR rhetorical strategies. Generally consistent with our expectations, the results show that pathos and logos strategies contributed to CSP, whereas ethos strategy had no significant effect.
Pathos strategy can significantly improve CSP. Pathos strategy changes an individual's attitude through emotion, and attitude affects an individual's behavior. Compared with other rhetorical strategies, CEO pathos strategy leads to irrational and unconscious emotional feedback among stakeholders. CEOs use pathos strategy to create an intense emotional atmosphere of ethics and morality, which can lead to the resonance of employees. Emotional words can improve employees' acceptance of the CEO's views, change their past resistance, and induce their organizational identity and enthusiasm. Through pathos strategy, employees' attitudes towards CSR activities are changed.
Ethos strategy has no significant effect on improving CSP. Drawn from the previous discussion, ethos strategy is based on moral norms, which may alienate the audiences from the persuaders. If persuaders attempt to adhere to an ethical norm, it may be attributed to the fact that their organizations are still operating in a hierarchical, top-down stream of authority. The failure of ethos strategy is due to the lack of appropriate communication mechanisms within the company. If the CEO does not consider the position of employees, then the ethos strategy may not have a significant effect in disseminating CSR strategy.
Logos strategy contributes to CSP. Logic can affect people's cognition, and cognition will influence people's behavior. Persuaders should present sound arguments and express their rational appeal. Logical reasoning is used to analyze the cause and effect to make the listener think, understand, and agree with the speaker's point of view, so as to achieve persuasion effect. In wording on CSR, the CEO uses logical reasoning to convey CSR strategy, to tell the stakeholders why the firm implements CSR activities, and what specific benefits CSR activities will bring to stakeholders, so as to affect the cognition of employees with respect to CSR. After having a deep logical understanding of the importance and necessity of CSR activities, employees will be more committed to the practice of CSR activities.
On the issues of CSP and rhetorical strategies, previous studies have explored how CEOs apply different rhetorical strategies when confronted with diverse stakeholder pressures [19], whereas we focus on the impact of different rhetorical strategies on CSP, which makes a complement to this field. Consistent with the extant literature, our findings show that various rhetorical strategies influence employees via different mechanisms and affect CSP distinctly [61]. Our research provides a reference for CEOs in using rhetorical strategies to encourage employees to participate in CSR activities. In addition, our study performs a micro-analysis of the language in CSR reports, which responds to the call of Rajandran and Taib; they claim that the majority of existing CSR report analyses are macro-level, which usually focus on frequent stages or common topics and can improve our understanding of the structure of CSR reports but ignore their linguistic features, and micro-analysis of analyzing linguistic features is crucial to understanding CSR representation [119].

Contributions
Our study makes contributions to the literature in several ways. To begin with, our study investigates how CEOs affect CSP from a new perspective. Previous studies based on upper echelons theory have mainly focused on the demographic characteristics of CEOs [9,53], while besides demographic characteristics, CEOs' verbal characteristics are equally important. This study set out to examine the influence of a new independent variable, CEO rhetorical strategy. Based on the general idea that rhetorical strategies contribute to CSP, we identified three mechanisms and evaluated the various effects of particular rhetorical strategies on CSP. Moreover, we focus on the implementation stage of CSR strategy. Previous studies mainly concentrated on the formulation stage of CSR strategies, but the implementation after formulation is equally important in order to achieve a high CSP. CEOs impact not only strategy formulation but also strategy execution. We highlight the significant role of communication in the implementation stage, which fills the research gap. Last but not least, our study underlines the importance of employee engagement with respect to CSP. The corporate strategy needs to be implemented efficiently, so employees play an important role. Employee engagement directly impacts CSR implementation; therefore, studying how to boost employee enthusiasm is essential. We conclude that different rhetorical strategies affect employee participation through different mechanisms.

Limitations and Future Research Directions
There are some limitations in our research. First, the classification criterion of rhetorical strategies is set artificially; due to the lack of an objective, constant, and data-based standard, there may be certain subjectivity and randomness in the determination of rhetorical strategies. Meanwhile, the measurement of rhetorical strategies-using text analysis by manual word coding method-may be rough and ambiguous in the recognition of rhetorical strategies and may result in a certain degree of deviation in the statistical results of the data. Second, we coded rhetorical strategies as dummy variables and analyzed the impact of rhetorical strategies on CSP, without taking into account the frequency of use. The rhetoric strategy is coded as "1" once adopted, while in fact, the number of uses of rhetorical strategies may affect CSP to various extents. On the other hand, we analyzed the influence of different rhetorical strategies on CSP separately, without considering the interaction or the combination of different rhetorical strategies. Lastly, we only examined the main effect of rhetorical strategy on CSP but did not define the boundary conditions of such relationship. Factors that might strengthen or weaken the relationship have not been discussed sufficiently.
Further research could be applied to improve our study. To begin with, for qualitative analysis or quantitative measurement of rhetorical strategies, various research methods could be adopted according to the needs of the research issues. The latest text analysis methods and relevant tools such as machine learning can be followed up and applied to future research. Moreover, in addition to the Aristotle's classification of pathos, ethos, and logos, there are other complementary elements of rhetorical strategies (e.g., Kairos, exaggeration) [73,120], which have developed and enriched rhetoric theory, and future research can be conducted under different classifications. At the same time, the frequency of use of rhetorical strategies can be taken into account in further analysis, which may draw some interesting conclusions. On the other hand, the interaction between different rhetorical strategies can be considered to analyze the appropriate way to organize, apply, and arrange various rhetorical strategies, which can maximize the effect of persuasion. Moreover, the relationship between rhetorical strategies and CSP may be influenced by other factors. There may exist other factors that could influence the relationship between rhetorical strategies and CSP. Studies in different contexts will help better understand the conditions under which rhetorical strategies play an important role. In this sense, contingent factors (e.g., managerial discretion, governance structure) that are able to distribute rhetorical strategies and influence CSP can be explored in future studies.

Conclusions
CSP is crucial for firms to realize long-term values and sustainable development [25]; therefore, relevant research on the rhetorical strategy of CEOs toward CSP is of both theoretical and practical significance. Fixed-effect models were used to conduct regression analyses of the sampled panel data of 246 manufacturing firms listed on the Shanghai and Shenzhen Stock Exchange in China from 2014 to 2016, and the Heckman's two-stage method was adopted to solve sample selection bias. Several conclusions have been drawn. Pathos and logos strategies have positive effects on the CSP, whereas ethos strategy does not. The results provide practical implications for CEOs to improve the persuasion effect of internal CSR communication so as to achieve high CSP by using proper rhetorical strategy.