Financial and extra-financial determinants of CEOs’ sustainable discourse: A longitudinal analysis on the US market

Lachuer, Julien; Jost, Sébastien

doi:https://doi.org/10.59876/a-g895-raj9

Over the last decade, corporate social responsibility (CSR) has become a major issue in asset management. Asset managers are now meeting the new expectations of investors by adjusting their investment to the new standards of sustainable management (Amel-Zadeh & Serafeim, 2018). Between 2018 and 2020, the total US-domiciled sustainably invested assets under management, both institutional and retail, grew by 42% to $17.1 trillion, up from $12 trillion. This represents 33% of the $51.4 trillion in total US assets now under professional management.[1] Despite the growing popularity of sustainable investments, the positive impact of the level of sustainability on financial performance is still widely debated. The emergence of new sustainable financial products may be perceived as a commercial positioning rather than as a way to maximize shareholder value. From one side, through an instrumental view of CSR (Jones, 1995; McWilliams & Siegel, 2001; Surroca et al., 2010), environmental, social, and governance (ESG) criteria would increase shareholder value by using them as a defensive measure when allocating resources (Bass et al., 2017). In fact, CSR could reduce information asymmetry (Bouslah et al., 2013) as well as extra-financial risk (Archambeault et al., 2008). CSR can also increase a firm’s value (Fatemi et al., 2018) or enhance a firm’s image (La Leyva-de Hiz et al., 2019). On the other hand, CSR could also be seen as costly for shareholders, for example, when sustainability spending supports private objectives and may result in agency costs (Hasan et al., 2018). CSR may be considered as an agency conflict when firm’s involvement in such activities goes beyond a certain level (Benlemlih et al. 2021). More generally, the link between corporate social performance (CSP) and corporate financial performance (CFP) is driven by the type of CSR investment, sectorial effects, and/or indirect effects of moderating variables, such as innovation and reputation. (Grewatsch & Kleindienst, 2017). Therefore, diversified information, such as the level of CSR, is increasingly valued and requested by investors (Bender & Samanta, 2017; Ross, 2017).

In our study, we consider that chief executive officers (CEOs) may actually moderate the CSP—CFP relationship through their discourses. Prior research has shown that CEO values matter for a range of organizational outcomes (Chin & Semadeni, 2017; Gupta et al., 2018). The recent work of Hambrick and Wowak (2021) provides new insight into how CEOs’ words and actions are both a cause and a consequence of stakeholders’ sentiments toward the firm. When a CEO’s position is aligned with stakeholder interests, the relationship between the company and stakeholders is strengthened, which, according to stakeholder theory, is a precursor to value creation. Indeed, sustainable investments should be in line with the firm’s strategy and stakeholders to positively affect firm performance (Lachuer & Lilti, 2021). Therefore, the impact of CSR on financial performance is determined by managerial decisions through the ways in which they implement sustainability expenditures.

Nowadays, investors are increasingly seeking information on the level of sustainability of firms. Moreover, stakeholders increasingly expect CEOs to be more transparent on contested issues (Hambrick & Wowak, 2021). To satisfy these expectations, an increasing number of firms are disclosing their sustainability reports. CEOs usually introduce these reports by explaining the main guidelines of the firm’s sustainable strategy with the objective of attracting new investors. However, it is important to be cautious since firms’ objectives are not limited to disseminating information on financial markets but also consist of managing investors’ impressions. However, CSR disclosure is often seen as providing a biased picture of the firm’s activities, as companies do not report on all of their activities in a clear manner (Cho et al., 2010) . In their introductory discourses, CEOs have to decide how to treat CSR dimensions. According to Christensen et al. (2010) and Schoeneborn and Trittin (2013), there is an important link between communication and action. CSR approaches reflect firms’ values (Chin et al., 2013) and managers’ real motivations.

The aim of this study is to extend the current body of research on how CSR affects CFP, considering that sustainability discourse could be a moderating variable interrelated with financial and extra-financial variables. Therefore, we intend to study the content of CEOs’ discourse to determine whether there is a link between CEOs’ discourse and the level of financial and extra-financial performance. To do so, we first examine the relation between the issuing of sustainability reports and financial and extra-financial variables through the use of a logistic regression. In the second step, we perform a thematic lexical analysis of CEOs’ discourse. Thus, we conduct a longitudinal analysis over six years from 2014 to 2019 in three major sectors of the S&P500. In this way, we propose extending the preliminary results of Lachuer and Lilti (2021). The use of a US sample is motivated by the freedom US firms have in releasing their sustainability reports. Even though the Global Reporting Initiative (GRI) standards provide guidance on how reports should be framed, our findings could also be relevant to all nonregulated communication. The relatively long period analyzed refers to a period of stable economic growth in the United States that is suitable for controlling economic uncertainty.

Our research represents more than 291 CEO introductory discourses. Our qualitative research combines financial and extra-financial data retrieved from the Bloomberg professional database. Our results show that the release of a firm’s sustainability report is mainly positively and significantly determined by its governance quality rather than its CSR engagement. Moreover, we show that, depending on the sector, the content of CEOs’ discourses is interrelated with firms’ financial and/or extra-financial performance. Our results follow the alignment theory (Hambrick & Wowak, 2021), considering that CEOs’ sustainable discourses have to be aligned with sectorial strategic priorities to ensure CSR expenditures contribute to financial performance. These results suggest that discourses are not neutral, and integrating CEOs’ sensibility CSR into models, through their sustainable discourses, better explains the CSP—CFP link.

The rest of the paper is structured as follows. Section I provides a literature review and develops the hypotheses. Section II describes the sample, data, and methodology used. Section III presents our results and explores the mechanisms that drive them. Section IV concludes.

Literature Review and Hypothesis

Our research is based on two theoretical approaches that consider the relationship between CSP and CFP in the opposite way. First, the instrumental approach to CSR (McWilliams & Siegel, 2001; Surroca et al., 2010) suggests that the implementation of a CSR policy promotes firms’ financial performance. Indeed, incorporating the stakeholder theory of Freeman and Philips (2002), taking into account the expectations of primary and secondary stakeholders, will create moral capital (Ducassy, 2013), which will improve a firm’s long-term performance. Thus, stakeholders’ behaviors can help firms meet their objectives (Benabou & Tirole, 2010; Eccles et al., 2014; Hasan et al., 2018). Second, under the prism of agency theory (Jensen & Meckling, 1976), CSR can also be seen as costly for investors. According to Brown et al. (2006), agency costs play a prominent role in explaining corporate donations, which shows that firms with high debts demonstrate a lower level of philanthropic acts. However, managers can use CSR donations for their own benefits in order to improve their image at the cost of shareholders (Jiao, 2010). CSR corresponds, in that case, to the expression of managerial opportunism (Kim et al., 2012) and constitutes a significant cost for the firm (Cordeiro & Sarkis, 1997). This would result in managers overinvesting in CSR activities to improve their reputation as “good citizens” among stakeholders (Barnea & Rubin, 2010; Jo & Harjoto, 2011). Barnea and Rubin (2010) also consider that managers overinvest in CSR to enhance their personal reputation and increase self-satisfaction. These two contrasting theoretical approaches coexist in the current literature (Hambrick & Wowak, 2021) ever since the meta-analysis of Orlitzky et al. (2003) and Revelli and Viviani (2013). The results from meta-analyses indicate a significant but weakly positive effect of CSR on CFP, but research lacks the elements to characterize this relationship. The positive CSP—CFP link is the result of intermediary relations associated with intrinsic mechanisms from management science; due to this complexity, this link is indirect through intangible assets (Surroca et al., 2010).

Firms will decide to engage in CSR for various reasons, which can be instrumental, economic, efficiency-driven, reputational or normative (Aguinis and Glavas 2012). CSR activities are multidimensional and often represent a collection of uncoordinated initiatives, some of them are value-destroying for shareholders if they are relatively independent from the core business (Hasan et al. 2018). Indeed, the impact of CSR on financial performance is highly dependent on how financial resources are allocated to the strategic dimensions of CSR. Thus, the common factor, through the instrumental view of CSR and agency theory, is the way CSR investment are implemented. Shedding the light on the importance of managerial decisions. Depending on the CSR investment choice made by the CEO, the impact of CSR on financial performance may theoretically be in favor of financial performance or not. CEO state of mind on CSR is the key driver for a positive impact of CSR on financial performance. Which may lead to balance towards agency theory or stakeholder-instrument theory.

The involvement of the CEO

Executive and managerial involvement is critical for the conversion of CSR policies into future financial results. There exist several ways to invest in CSR activities but only proactive management and fully integrated CSR improve financial performance (Bocquet et al., 2017). If CSR investment choices are not aligned with the firm’s core business, the benefits of such investments will be overcompensated by the associated costs (Hasan et al., 2018; Kang et al., 2016). Thus, the positive impact of CSR on financial performance is more related to the ability of CEOs to understand how stakeholder satisfaction is an effective way to maximize firm value (Costa et al., 2015). According to Cardebat and Cassagnard (2011), managers use CSR activities and social dialogue to empower various risk takers in order to increase their legitimacy and build a positive reputation. Furthermore, specifically focusing on CEOs, research shows CEOs are a key variable in the CSP—CFP relationship due to the impact of their personal values on firms’ outcomes. Gupta et al. (2018) have shown that CEOs can act as political activists. Several recent studies have revealed that CEOs’ personal values affect organizational outcomes, such as CSR programs (Chin et al., 2013; Chin & Semadeni, 2017). Moreover, characteristics such as narcissism influence individuals’ behaviors, which, in the case of CEOs, may have strategic implications (Cragun et al., 2020). According to Petrenko et al. (2016), the impact of CSR expenditures diminishes with the greater narcissism of the CEO. Gupta et al. (2018) recognize that CEOs’ values impact the resource allocation of firms and also consider that CEOs’ values are amplified when aligned with the prevailing ideology among organizational members and, conversely, muted when misaligned. This is consistent with the alignment theory of Hambrick and Wowak (2021), considering that if a CEO’s own ideology is aligned with stakeholder concerns, it may have a positive impact on a firm’s financial outcomes. We believe that the CEOs’ discourse can be an interesting tool to understand CEOs’ real motivations in order to design the potential impacts of CSR on financial performance.

CEOs’ discourse as a proxy for the alignment between CSR and financial objectives

Theoretically, CEOs’ involvement in firms’ CSR policies should be related to the discourse they use. CSR is a legitimized construction supported through communication (Chaudhri, 2016). Its purpose is to influence a firm’s image to obtain a “license to operate” (Henisz et al., 2014; Igalens, 2006). The discourse targets stakeholders to whom CEOs need to justify their actions (Preston & Post, 1981). According to Christensen et al. (2010) and Schoeneborn and Trittin (2013), there is an important link between communication and action. Regardless of the reliability of the discourse, the content and structure are likely to provide information. The way firms highlight their own CSR activities is therefore meaningful. In this regard, CEOs’ introductory discourses may be related to a firm’s level of global performance and reflect managers’ strategies and the values of managers (Chin et al., 2013; Gupta et al., 2018).

Firstly, the issuing of a sustainability report is the first step in identifying a CEO’s involvement; therefore, it can be interrelated with financial and extra-financial performance. following the information cost theory, firms that disclose information take into account the costs and benefits associate with disclosures (Verrecchia 2001). Therefore, we can expect that firms that disclose sustainability report are more likely to have a better responsible and financial performance. A firm’s disclosure is similar to a carefully scripted response to public pressure and, more specifically, to pressure from NGOs and institutions. Voluntary disclosure reduces information asymmetry. As a result, investors who focusing on firms which provide extensive disclosure can be relatively confident that any transaction is being conducted at a fair price, leading to increased liquidity in the shares of these firm (Husser and Evraert-Bardinet 2014).

In the context of no mandatory ESG disclosure, we support that firms that fail to provide a sustainability report would be firms with poor extra-financial performance. The release of sustainability reports reflects a willingness for transparency, which therefore minimizes agency costs by better engaging stakeholders and alleviating managerial opportunism (Benabou & Tirole, 2010; Eccles et al., 2014). In that sense, Dhaliwal et al. (2021) shows a positive and significant relationship between financial performance and CSR disclosure. Categorical variables, such as the presence or absence of ESG disclosures, may impact the CSP—CFP relationship by reducing information asymmetry (Zhang et al., 2021). Thus, before analyzing the content of a CEO’s speech, we hypothesize that a company that provides a sustainability report is more likely to be socially responsible and financially efficient.

H1a: The presence of a sustainability report is determined by the firm’s financial performance.

H1b: The presence of a sustainability report is determined by the firm’s corporate social performance.

Secondly, the sustainability report is a good way to identify the main motivations that drive CEOs’ CSR strategies because they address their discourses to all stakeholders in a very unrestricted way. Firms with good financial and extra-financial performance are likely to highlight their actions and link them to financial considerations. The choice of discourse is very subjective because firms strive to legitimize their actions by creating meaning. CEOs’ opening discourses are free and succinct, which forces them to make choices in the construction of their communication.

The content of the CEO’s discourse will be influenced by his desire to perform financially, which, according to stakeholder theory, requires taking into account the stakeholders. Such consideration will provide the opportunity to obtain social acceptability, involving legitimacy, credibility, and trust (Baba and Raufflet 2015).

Regarding the environmental dimension, Husser and Evraert-Bardinet (2014) found that improving environmental information enable investors to improve their understanding of the risks and reduce the costs related to information asymmetri. The quality of environmental information has a double positive effect: it increases the degree of certainty of investors as to the profitability of their financial investments as well as the degree of credibility of CEOs. Dardour and Husser (2016) found that environmental dimension seems to be directly related to executive incentive compensation costs. However, social and governance disclosure don’t seems to have an impact. Cowen et al. (1987) found that various organizations disclose distinct types of social performance indicators based on factors such as size, industry or the willingness to promote a certain image of themselves. In addition, Braune et al. (2021) put forward that CEOs will favor social aspects when they have access to free financial resources such as cash flow, and if they are left free from shareholder scrutiny. Within a broader context, while some firms choose not to report on some of their CSR commitments, others may overstate or even lie about their CSR activities in order to gain legitimacy (Berrone et al. 2017). In that sense, Conaway and Wardrope (2010) and Cho et al. (2010) show that firms with the lowest environmental performance highlight positive information while hiding information, which could be seen as negative.

Finally, we expect the content of a CEO’s opening speech and how certain dimensions of CSR are put forward in the discourse to be interrelated with their firm’s financial and extra-financial performance. This discourse should, through alignment theory, express the possible impact of CEO’s own ideology on CSR and financial performance. Considering that the results of Husser and Evraert-Bardinet (2014) have shown that multidimensional disclosure (both environmental and social) is appropriate to monitor the firm’s strategy and best explain its financial performance.

H2a: The content of CEO discourses is determined by a firm’s level of financial performance.

H2b: The content of CEO discourses is determined by the firm’s level of corporate social performance.

These two hypotheses will allow us to determine whether CEOs’ discourses in sustainability reports are linked to tangible corporate situations.

Methodology and Sample Selection

Quantitative analysis

For our first hypotheses (H1a and H1b), we start use a univariate approach consisting in t-tests comparing the means between different sub-samples[2]. Then, we use a logit[3] model taking year (YearFE) and industry (IndustryFE) fixed effects into account. We also run the analysis with and without the ESG disclosure categorical variable (Disclosure), as its impact on the presence or absence of a report may be too strong.

Qualitative analysis

For our second set of hypotheses (H2a and H2b), we use a discourse analysis approach. This method has become a common approach in management science, but it is still emerging in accounting and finance. The article by Loughran and McDonald (2016) summarizes a broad range of qualitative methods that have been used in accounting and finance. Generally, we can find category analysis, targeted sentences, sentiment analysis, topic modeling, and measures of similarity. These approaches provide useful results for asset management, as demonstrated by Cohen et al. (2020). Here, we decide to use a more thematically structured approach. These “bag-of-words” techniques are used to categorize common themes in documents and reduce the dimensionality based on each word’s relation to latent variables (Loughran & McDonald, 2016). This method has multiple advantages and was first used in accounting and finance by Huang et al. (2018). This method seems to be the most relevant for the purpose of the present study because it focuses on the relations and place of words in relation to each other.

To analyze CEOs’ discourses, we employ the semantic analysis software “Alceste,” which uses the “bag-of-words” methodology. More specifically, Alceste is a word dictionary that also defines the root and frequency of the words used. The terms we will find in the classes correspond to the roots of several words. For example, product, production, productive, productivity, and products will give the root product and will be considered in the same “bag.” The software subdivides the corpus into homogeneous texts containing a sufficient number of words. This results in a classification of the segments according to the strongest opposition. In addition, Alceste performs a correspondence factor analysis (CFA) based on the results of the previous classification. This process makes it possible to take into account the relations of attraction or distance between classes, forms and classes and forms in the corpus. The software returns different lexical fields distributed in several classes, but no interpretation is made by Alceste beforehand. In order to make an interpretation, words are sorted by chi2 levels, and an examination of the terms with the highest level of chi2 allows us to interpret the classes defined by the analysis (Bart, 2011). Moreover, the names of the classes that will be presented are chosen by the researcher according to the distribution of the words present.

This software has the advantage of allowing us to give characteristics to the corpora. It is a “star line” in which we have defined the level of financial performance (return on asset [ROA], Tobin’s Q, Sharpe) and the level of extra-financial performance (CSR, disclosure, governance). These characteristics define the profiles relating to each class of speech. In fact, we combine lexical analysis with data relating to financial and extra-financial performance. The benefit of this method is to be able to analyze the lexical fields according to these present and absent characteristics and thus to identify markers not initially visible as objectively as possible (Pierrot & Spring, 2011). According to Seignour (2011), semantic software makes it possible to reveal variations that are not visible on reading; thus, it provides valuable help for analysis. Lastly, the characteristics in the star line cannot be raw data. We have therefore defined three levels of performance for each of the financial (ROA, Tobin’s Q, Sharpe) and extra-financial (CSR, disclosure, governance) variables as high, medium, or low. These levels are defined according to each sector: high for a performance higher than 30% of the firm in the sector, low below 30%, and medium for the others.

Sample

In order to conduct our qualitative analysis, we focus our analysis on the US market through the S&P500. In our opinion, the American market has two main advantages for our study. First, US firms have a great degree of freedom regarding their disclosures of nonfinancial information since CSR is still seen as a soft law in the US. Second, weak legislation on financial reporting does not exclude a certain level of maturity regarding CSR issues. For example, GRI standards are widely used by many US firms. Moreover, our data are constructed across three distinct sectors from the Global Industry Classification Standard (GICS): the energy sector (GICS 10), the sector related to industrial production (GICS 201), and the sector grouping industrial services (GICS 202) and transportation (GICS 203). These sectors are particularly interesting, as their associated activities are likely to generate significant negative externalities. The impact of sustainability measures, developed internally or imposed by external events, is likely to be important in these sectors. The choice of three sectors also allows us to derive sector-specific conclusions; it builds on a prior article published by Lachuer and Lilti (2021).

Thus, we collect and analyze 291 sustainability reports between 2014 and 2019 over the 477 observations. Table 1 provides information on the distribution of reports by year and sector. These reports have been retrieved directly from firms’ websites, the GRI website, or Bloomberg. We consider the availability of these reports essential because they provide indications of firms’ transparency. Our analysis period covers six years of relatively stable economic growth in the United States. Thus, the subprime crisis and the COVID-19 crisis have no impact on our variables. Moreover, it seems necessary to analyze the impact of CSR during periods of growth, as the literature has demonstrated the positive impact of CSR during periods of crisis, but this seems less significant during bullish periods.

Table 1

**Composition of samples by sector**

Variables

To measure financial performance, we use the ROA, Tobin’s Q, and Sharpe ratio. We use both accounting-based measures and market-based measures since both approaches are considered to be often uncorrelated (Gentry & Shen, 2010; Venkatraman & Ramanujam, 1986). Indeed, accounting-based measures are generally seen as a reflection of past financial performance, while market-based measures are conceptualized as a reflection of future financial performance (Hoskisson et al., 1994). Market-based measures reflect investors’ expectations and are based on the market efficiency hypothesis, in which market prices fully reflect all available information in the market. According to Grewatsch and Kleindienst (2017), studies using accounting-based measures tend to demonstrate a stronger positive relationship between CSP and CFP than studies relying on market-based measures.

To take into account the different factors influencing financial performance and the level of CSR, we use several control variables (see Table 2 in the appendix section). The measurement of extra-financial performance will be approximated through three indicators: the level of CSR, the level of ESG disclosures, and the level of governance. The level of sustainability is approximated through the Sustainalytics CSR score obtained from the Bloomberg platform.[4]

Results and analysis

First, Table 4 reports the cross-correlations of the variables. We note that Tobin’s Q and ROA are significantly and negatively correlated with CSR. Moreover, Tobin’s Q and ROA are both negatively correlated with the presence of a sustainability report and the level of CSR performance. Nevertheless, the level of CSR is positively correlated with size, research and development, liquidity level, and dividends, and it is also very strongly correlated with disclosure level. Moreover, the presence of a sustainability report is positively correlated with several financial and extra-financial variables, including size, leverage, CSR, and ESG disclosure.

The determinant of the release

Our first approach is to examine the relationship between the presence of a sustainability report and the financial and extra-financial variables. We first propose a univariate analysis that divides our sample into two subsamples (i.e., firms providing a sustainability report and those not providing a sustainability report). Table 5 shows the average level of our variables according to the presence or absence of a sustainability report. Our t-tests compare the mean between the different subgroups. Our results suggest that the presence of sustainability reports is positively correlated with the level of CSR. This is also the case for governance in the industrial sector and in the transport and industrial services sector. Regarding the energy sector, the release of a sustainability report is positively correlated with ROA, while for the transport and industrial services sector, we find a negative correlation with the level of ROA and Tobin’s Q. These results suggest that the release of a sustainability report is a sign of a superior level of CSR performance and stronger governance but not necessarily a sign of higher financial performance. More surprisingly, the differences observed for the level of ESG disclosure are not significant for Energy and Industry sectors.

Table 5

**Average level of variables by sector and by presence or absence of a CSR report**

Further, we conduct a multivariate analysis using a logit model approach. Table 6 shows a logit model controlling for years and sector effects. The coefficients’ signs, as well as their relative magnitudes, provide a brief idea of the influence of our variables. These first results suggest that ROA and ESG disclosure positively influence the release of a sustainability report; this also holds for the CSR performance and governance variables when the ESG disclosure variable is removed from Model 5. However, in this last model, we observe that the impact of the level of governance appears to be stronger than the other variables. These results are confirmed by the marginal impact analysis conducted on each of these variables on the probability to release a report (see tables 7 and 8). In fact, the marginal effect analysis from Model 5 shows a positive and significant marginal impact of the governance and CSR performance levels on the probability to release a CSR report, while it shows a negative and significant marginal impact of cash flow levels. On the one hand, an increase in one unit’s level of governance increases the probability of a release by 2.6%, while one unit’s level of CSR performance increases the probability of a release by 0.4%. On the other hand, a one-unit increase in the cash flow level reduces the probability to release a CSR report by 0.4%. This implies that the release of a report seems to be mainly due to stronger governance through a willingness to be more transparent. The impact of the level of CSR remains very limited, and ROA is no longer a relevant variable. The positive impact of the level of governance on overall performance is not surprising. Indeed, a good level of governance will limit agency costs, and sustainable investments will be in line with maximizing shareholder value. Our impact study shows that the level of governance is a major determinant of the release of nonfinancial information.

Thus, our analysis seems to partially support hypothesis H1b, which states that the release of a sustainability report is driven by the level of extra-financial variables. However, our results do not provide a clear answer to hypothesis H1a; the results remain very weak and do not allow us to conclude a positive link. Thus, the release of a sustainability report is mostly driven by firms’ governance rather than their CSR levels.

The link between financial and extra-financial variables and the content of CEOs’ discourses

Finally, the purpose of the qualitative analysis is to determine whether there is any relationship between the content of CEOs’ discourses and financial and extra-financial variables. The analysis of the CEOs’ discourses on our three sectors gives us fairly high levels of initial context unit (ICU)[5] ranking, (i.e., a majority of the text could be related to a textual class). Therefore, here, we analyze the importance of given lexical fields in CEOs’ speeches. A more important presence of a lexical field is likely to reveal information about firms’ most important strategic objectives. We suggest making a first analysis by sector before giving a global synthesis. Indeed, the importance of specific fields of CSR in the discourse may depend on the area of operation. All three sectors studied remain inter-related.

Table 6

**Logit model of the determining variables in the realization of a sustainability report**

Table 7

**Average marginal effect of variables on the probability of release of sustainability report from Model 1**

Table 8

**Average marginal effect of variables on the probability of release of sustainability report from Model 5**

Energy sector

In the energy sector, the introductory discourses of directors are composed of three lexical fields. The first relates to pollution management, the second to the community, and the third to the promotion of sustainability management. These classes are fairly equally present in the speeches, representing 32%, 33%, and 35%, respectively, of the totality of the corpora.

The first lexical classes deal with polluting emissions through the presence of root words: “Emission, Reduce, Percent, Intensit, Carbon, Lower, etc.” In this sector, the strong presence of this lexical field is associated with a high level of financial accounting and market performance. Indeed, we notice the presence of characteristics related to a high level of Tobin’s Q and ROA. Moreover, the Sharpe ratio and ROA are never low. Concerning the extra-financial variables, this discourse reveals a moderately responsible firm whose levels of CSR disclosure and governance are low. This discourse is linked to firms that are financially performing but moderately sustainable. The second class of discourse deals with items related to the community with lexical roots: “Communit, Educat, Pandemic, Market, People, Famil, Price, Return, Business, Care, etc.” Although this discourse is generally present in all sectors, we note a particularity for the energy sector. Indeed, we find the presence of words such as “Market, Price, Return, Business,” which are associated with words related to the community. We believe that, for these firms, taking the community into account is intrinsically linked to financial performance. However, our analysis reveals that this discourse is associated with weak financial performance through the presence of low ROA and Sharpe ratio and the absence of high ROA and Tobin’s Q. These results follow the considerations of Odell and Ali (2016) and Hasan et al. (2018). At best, these firms manage to be in the average range for their Tobin’s Q level. The CEOs try to enhance their firms’ reputations to possibly deal with several stakeholder issues. Nevertheless, regarding extra-financial variables, these firms have a high level of governance but a sustainability score ranging around the sample mean. Finally, the last lexical field relates to the global approach to sustainability and governance. This reflects corporate culture with the presence of lexical roots: “Report, Sustain, Board, Perform, Transparency, Stakeholder, Share, Engage, Govern, Culture, ESG, etc.” This type of discourse is quite common, and the associated characteristics remain fairly average. These are firms in which the Sharpe ratio is median and whose market value is low due to the presence of low and never median Tobin’s Q. From an extra-financial point of view, these firms have a good level of governance, which may explain the place of this lexical field in the opening speech. This lexical field includes firms with high and low CSR levels. Thus, in the energy sector, firms with high overall performance are those for which the CEO emphasizes the limitation of polluting emissions and the implementation of sustainable practices that contribute to improved governance.

Industry sector

CEOs’ opening discourses in the industrial sector are the most extensive of the three sectors. Indeed, we find five classes of discourses related to emissions, community, governance—ethics, considerations regarding the future, and financial perspectives. These discourses are more or less present in our entire sample, with 26%, 12%, 30%, 23%, and 9% of the ICUs analyzed, respectively.

Table 9

**Synthesis of lexical fields present in the CEOs’ discourses in energy sector**

The first lexical class deals with emissions. This class is quite similar to the first class of the energy sector, although lexical roots, such as “Product, Lower, Plant, System, and Data,” are specific to the sector. The associated characteristics differ from those of the energy sector. We note that ROA and Tobin’s Q are rather low and that the Sharpe ratio is high and never low. Therefore, we have firms that are not very profitable, but compared with their risk levels, their financial performance can be considered good. Indeed, it would seem that, in this sector, taking emissions into consideration has a negative impact on a firm’s activity. Being responsible will generate additional costs; nonetheless, these costs are favorable since, at a constant level of financial performance, the risk is lower. This information is therefore essential for investors and follows the results of Bouslah et al. (2013). The second class deals with communities with lexical roots, such as “Educat, Veteran, Support, Student, Donat, Women, Science, Grant, and LGBTQ.” The presence of this type of discourse is highly correlated with high financial performance, as suggested by the high Sharpe ratio and Tobin’s Q, as well as the absence of a low Sharpe ratio, ROA, and Tobin’s Q. Nevertheless, the level of disclosure is low, and the level of sustainability is never high, only median. These results are, therefore, in contrast to those of the energy sector. The third class deals with governance and ethics through the presence of lexical roots, such as “Report, Corporate, Value, Ethical, Standard, and Safe.” These discourses are also contextualized by the importance of industry safety rules. The presence of this type of lexical field refers to firms with a high level of financial performance and high and never-low levels of Tobin’s Q and ROA. However, this improvement in performance suggests that it is linked to an increase in risk. Indeed, the level of CSR is never high, and the level of disclosure is never low. Moreover, even if the discourse is focused on governance, these firms have a low level of governance. The improvement in financial performance is therefore achieved at the expense of increased risk taking, which is confirmed by a Sharpe ratio that is never high; therefore, this is a more cosmetic approach. The fourth class includes textual elements related to future considerations through lexical roots: “World, Challenge, Future, Better, Safer, Changing, Planet, Nation, etc.” These discourses remain quite contextual through the presence of the YEAR_2019 feature. As sustainable reports are written after the end of the year, it is not uncommon for discourses to include some considerations inherent to the following year, notably here with the appearance of COVID-19. This lexical field concerns firms with a rather positive market valuation with a medium and never-low Tobin’s Q, as well as a never-low Sharpe ratio. However, the level of ROA is low. Finally, the last class of speeches deals with the financial perspective with the presence of lexical roots: “Market, Cash, Acquisition, Dividend, Margin, Debt, etc.” Even if CEOs talk about financial performance in their discourses, in practice, the characteristics associated with these firms indicate low and never-strong levels of ROA, Tobin’s Q, and Sharpe ratio. However, the levels of extra-financial variables are high. We suppose these discourses reveal that firms invest heavily in environmental considerations but are not related to financial performance, as was the case in class one on emissions. CEOs adopt discourses on financial considerations to justify the choices they make. Thus, the industrial sector proves to be quite particular because discourses turn out to be quite far from the realities (Cho et al., 2010). Therefore, it is necessary to be wary of this sector and select firms whose CEOs stress the consideration of emissions and communities.

Table 10

**Synthesis of lexical fields present in the CEO discourses in industry sector**

Transportation and industrial services sector

The last sector has four lexical classes, which are partly present in the other two sectors. We note the presence of lexical fields relating to emissions, community, governance and ethics, as well as a fourth class specific to this sector, which covers employees. These classes are rather well distributed in the whole of the discourses, with respective compositions of 30%, 18%, 32%, 20%, and 20%.

The first class is related to emissions. The associated financial performance level is high, with a high and never-low ROA and Sharpe ratio. The level of Tobin’s Q is average. From an extra-financial point of view, these discourses refer to sustainable firms with high and never-low levels of CSR and disclosure. This lexical field related to emissions seems essential in this sector and underlines sustainable and financially efficient firms. The second class of discourses refers to the community. This lexical field is associated with an average level of financial performance. The level of Tobin’s Q is high and never low, but the level of ROA and the Sharpe ratio are never high. Thus, firms are overvalued by the market, and the considerations of the community can help. With regard to extra-financial variables, the levels of CSR and disclosure are high and never low. These discourses are related to sustainable firms but show weak governance. Therefore, the level of CSR would be to the detriment of shareholders and a high level of governance. The third class of discourse is related to the consideration of standards of disclosure and governance. We find lexical roots: “Sustain, Report, Respons, Stakeholder, ESG, Governance, Citizen, Practice, etc.” As in the industrial sector, this lexical field relates to financially successful firms with high ROA, Tobin’s Q, and Sharpe ratio. Unlike the previous sector, this type of discourse refers to firms with strong governance, although their levels of sustainability and ESG disclosure are very low. CEOs who mainly use this lexical field relate to firms with low sustainability. Finally, the last class relates to employees through the presence of the lexical roots “Safe, Employee, Workplace, Experience, Custom, Profit, Standard, Care, etc.” This lexical field is specific to this sector because of the importance of human resources. It is not surprising to see the lexical root “Profit” appear. The share of human resources in the creation of value is more important in this sector than in the other two sectors. Nevertheless, the emphasis on this lexical field reflects a rather weak financial performance, with a low Tobin’s Q and a Sharpe ratio that is never strong. The level of governance is high, but CSR performance remains average.

Synthesis

Thus, our qualitative analysis allows us to positively answer hypotheses 2a and 2b. Indeed, we validate the fact that the lexical fields present in CEOs’ discourses are strongly interrelated with financial and extra-financial characteristics, and the importance CEOs give to one or another CSR dimension is determined by financial and extra-financial performance. This is in line with the previous results of Lachuer and Lilti (2021). Moreover, the present study builds on prior findings and suggests that for a given discourse theme, the associated variables are not necessarily identical, depending on the sector. For example, we note that the considerations of emissions refer to firms that perform well financially in the energy sector, whereas in the industrial sector, the performance is put relative to the level of risk. As example the results of Soana (2011) show that community is positively correlated to firm financial performance in the industrial sector. Thus, in order to use the discourses in a sustainable and efficient asset selection strategy with positive screening, it is essential to make sector distinctions. These results follow the considerations of Price and Sun (2017), who consider that firms with different combinations of CSR activities yield different performance results. Relative to our results, we consider that this assertion is similar to CSR activities in sectors.

Conclusion

Our study provides a better understanding of CEOs’ critical impact on the relationship between CSP and CFP. Considering the interrelation between CEO sustainability discourses on the one hand and financial and extra-financial performance on the other hand, we aim to analyze which discourse(s) may shed light on the positive or negative impact of CSR on financial performance. Our contribution to the existing literature is manyfold.

First, our quantitative analysis reveals that the release of a sustainability report seems mostly to result from strong firm-level governance rather than CSR or financial performance. Even though our univariate analysis suggests a link between the sustainability report release and firms’ financial performance, our multivariate analysis does not confirm a statistically significant relationship between the release of a sustainability report and firms’ financial performance, contrary to the results of Zhang et al. (2021) and Dhaliwal et al. (2021). In that sense, our findings suggest that firms’ level of governance seems to be the main driver of their financial performance. Second, our findings show that the content of CEOs’ discourses is linked to levels of financial and extra-financial performance. The presence of a more or less important lexical field in CEOs’ introductory discourses is determined by financial and extra-financial variables. This is for instance illustrated within the energy sector, where the preponderance of the community dimension in discourses is related to firms with weak financial—and sustainable performance, while discourses focused on polluting emissions tend to be associated to firms with strong financial performance. Third, our research extends the existing literature by showing the specific links for each sector, shedding light on the fact that links between variables and discourse can be totally different across sectors. This is the result of either strong or weak alignment between the CSR strategy and firms’ key objectives (Hambrick & Wowak, 2021). For example, in the energy sector, a discourse on the community is negatively related to the level of financial and extra-financial performance, whereas in the industrial sector, this discourse is positively correlated with the level of financial performance. Similarly, a discourse on polluting emissions in the energy sector is positively related to the level of financial performance but not to the level of CSR, whereas this type of discourse reflects poorly performing firms in the industrial sector. This same type of discourse is positively related to financial and extra-financial performance in the transportation and industrial services sector. These results are in line with the findings from Hasan et al. (2018), who consider that CSR is value-destroying if it is not linked to firms’ core business. Finally, we find that depending on the sector, discourses may not depict reality, as illustrated within the industrial sector. CEOs accentuate their speeches to hide their weaknesses (Cho et al., 2010). Our results moderate prior findings from Christensen et al. (2010) and Schoeneborn and Trittin (2013), since they suggest that rhetoric is not always linked to actions. As observable within the industrial sector, CEOs do not always have a discourse in line with the level of CSR.

Our results show that CEOs’ discourses are important for a better understanding of the strategic resource allocation issues that firms face regarding their CSR activities in their own sector; thus, our results follow the considerations of Gupta et al. (2018). The lexical fields employed indicate the alignment of CSR practices and value enhancement, recalling that not all CSR strategies are equivalent and strongly dependent on sector in which they are implemented. This result follows the ones from Hudson and Descubes (2021), and suggests that integrating CSR into the DNA of a firm, enable it to move ahead from simple cost-cutting or basic pragmatic implementation, to more proactive and transformative actions, in line with the norms and expectations of the new generation. This is also in line with the notion of social acceptability developped by Baba and Raufflet (2015), considering that this acceptability is neither a result nor a product, but rather a dynamic process that takes place over time.

Our work also has multiple managerial implications. First, the issue of a sustainability report from firms cannot be entirely considered as a good sign of strong CSR performance. Even if a good level of CSR has a slight impact on the release, investors should mostly consider it a signal of good governance, which could limit agency costs and have an indirect impact on financial performance. Second, our results highlight that CSR’s impact is largely dependent on the CSR sector alignment strategy of CEOs’ with stakeholders and shareholders’ objectives (Bocquet et al., 2017). CSR must be a part of firms’ strategies and business models, just as CEOs emphasize the community dimension, which relates to companies with overall poor performance. These results are in line with the consideration of Porter and Kramer (2011) regarding the positive impact of CSR on financial performance, when CSR is implemented into the firms’ strategy. Third, the informational value of CEOs’ responsibility discourses could be used for sustainability portfolio management purposes. In fact, the “negative screening”[6] approach is still prevalent in socially responsible mutual funds (Kotsantonis et al., 2016), but such selections result in a limitation of the universe of asset investment that is contrary to the maximization of financial performance, according to traditional financial theory (Markowitz 1952). We consider that the analysis of CEOs’ discourses can lead to this objective by using a more “positive screening” approach. CSR dimensions promoted through CEOs’ discourses have a potential informational value on the potential financial performance gain, due to the alignment of CSR with the sector’s strategic objectives. This could result in the integration of dummy variables related to CEOs’ discourses in traditional models. Considering that positive screening is going to become increasingly important in the future (Amel-Zadeh & Serafeim, 2018), our results open opportunities for future research.

Nevertheless, our research is not exempt from several limitations. First, we can only partially characterize the impact of CSR themes on our financial and extra-financial variables. The qualitative approach provides only a partial and unquantifiable view. Furthermore, our extra-financial measures are subject to the bias of their specific constructs. Although they are widely used by managers, these measures relate only to the theoretical level of CSR and cannot reflect the tangible level of CSR. Finally, we have focused on the introduction discourses to the sustainability report made by the CEOs, but we are aware that this is not enough to understand the entirety of the CEOs’ thinking.

To overcome these limitations, further research could integrate CEOs’ discourse dimensions into quantitative models to measure more precisely the impact of a specific “bag of words” on financial and extra-financial variables. Furthermore, similar analyses could be performed with alternative extra-financial measures or with other CEO discourse types, such as specific press or video interviews.

Abstract

Résumé

Resumen