Chair–CEO age dissimilarity and firms’ annual report tone – evidence from China

This study empirically investigates how the age dissimilarity between a Chair and a CEO impacts the tone of the firm's annual report. Prior research has established that top management team (TMT) diversity spans multiple dimensions, including age (Rabl & Triana, 2014), tenure (Keck, 1997), gender (Tonoyan & Olson-Buchanan, 2023) and educational background (Heenipellage, Fernando, & Gibbons, 2022). These dimensions affect various corporate outcomes, such as firm performance, information disclosure, financial reporting quality, corporate social responsibility and risk-taking tendencies (Komal et al., 2023). Empirical findings suggest that diversity yields advantages, such as enhanced innovation, improved decision-making and greater organisational adaptability (Wang, Duan, & Liu, 2021). Accordingly, Goergen, Limbach, and Scholz (2015) show that Chair–CEO age dissimilarity may increase firm value, while Trinh, Kara, and Elnahass (2022) report its potential to improve bank performance. Conversely, Zhou, Kara, and Molyneux (2019) document a decline in bank risk-taking behaviour, and Zhu, Gao, and Tan (2021) link greater age dissimilarity to heightened pay inequality. Hu, Bhuiyan, and Houqe (2025) examined the effect of Chair–CEO age dissimilarity on firm value. However, the effect of this dissimilarity on annual report tone (ART) remains largely unexplored, underscoring the significance of this study for corporate governance research.

The Chair–CEO relationship is crucial in organisational dynamics, as the Chair oversees long-term strategy and board governance (Parker, 1990), while the CEO implements this strategic vision through operational execution (Roberts & Stiles, 1999). The nature of these two roles is therefore intrinsically interdependent and complementary. The quality of their relationship can significantly influence organisational dynamics, generating positive or negative effects across the firm (Roberts & Stiles, 1999). Similarity-attraction theory suggests that individuals naturally gravitate toward those with similar characteristics, fostering greater interaction and communication (Byrne, 1997). Consequently, age dissimilarity between the Chair–CEO may produce both task and relationship conflicts within the TMT (Amason, 1996). Both types of conflict are destructive to firm performance (De Dreu & Weingart, 2003). When considered alongside Chair–CEO age dissimilarity, a strong working relationship may foster a positive internal business environment and a more favourable disclosure tone, whereas conflict or communication breakdown may harm firm performance and result in less positive reporting. Accordingly, the Chair–CEO dynamic is vital to enhancing the readability and tone of corporate disclosure.

Empirical research highlights that China's unique cultural context significantly shapes organisational behaviour, corporate policies, and firm outcomes, suggesting that the influence of Chair–CEO age dissimilarity may differ from that observed in other nations. Unlike Western cultures, which often adopt a direct approach to conflict resolution, Chinese cultural norms tend to favour conflict avoidance as a primary strategy (Tse, Francis, & Walls, 1994). Therefore, while Western research suggests that avoiding conflict may hinder innovation, many Chinese TMTs strongly favour this approach (Chen, Liu, & Tjosvold, 2005). This preference can be attributed to the collectivistic values that are deeply embedded in Chinese and broader Asian cultures, which emphasise interpersonal harmony and social cohesion (Morris et al., 1998). As a result, Chinese managers tend to avoid conflict more frequently than their Western counterparts (Fu, Yan, Li, Wang, & Peng, 2008). However, prolonged conflict avoidance could lead to the accumulation of latent tensions, which can eventually trigger the escalation of internal organisational conflicts.

The primary function of corporate annual reports is to convey operational information to external stakeholders, particularly those who do not have direct knowledge of the company. Therefore, the tone of annual reports is crucial for both firms and investors. Research suggests that the managerial tone used in annual reports assists investors in interpreting accounting figures and market returns (Loughran & McDonald, 2011). Studies have found that a positive tone provides valuable incremental information, serving as a signalling mechanism to communicate firms' future performance expectations (Xie & Lin, 2015) and influencing stock market returns (Jiang, Lee, Martin, & Zhou, 2019). A positive tone can effectively mitigate information asymmetry between firms and investors, enabling investors to assess the company's financial position more accurately, thereby increasing investment confidence (Feldman, Govindaraj, Livnat, & Segal, 2010). However, some scholars argue that a positive tone may not always be reliable as managers might deliberately adopt an optimistic tone to mislead investors and obscure unfavourable environmental and social performance. Ultimately, the tone of annual reports is shaped by corporate executives, whose individual characteristics influence the writing and editing process, thereby affecting the overall linguistic tone of the reports (Shiller, 2017).

Using data from the China Stock Market & Accounting Research (CSMAR) database spanning the period between 2010 and 2023, this study analyses 26,876 firm-year observations to explore the relationship between Chair–CEO age dissimilarity and the tone of firms' annual reports. The results indicate a significant negative association between Chair–CEO age dissimilarity and the positive tone of firm's annual reports, highlighting its economic relevance. Specifically, a one-standard-deviation increase in Chair–CEO age dissimilarity (GAPS) is associated with a 0.48% decrease in the positive tone of the firms' annual reports (TONE) relative to the sample mean. In addition, a one-standard-deviation increase in unsigned Chair–CEO age dissimilarity (GAPU) is associated with a 0.42% decrease in the positive tone of the firms' annual reports (TONE) relative to the sample mean. Additionally, cross-sectional analyses were conducted to examine the contextual factors that could influence this relationship. The negative effect is particularly pronounced in firms characterised by a high presence of institutional investors, a low concentration of controlling rights, greater information asymmetry, heightened agency conflicts and family ownership structures. Furthermore, conflicts stemming from the Chair–CEO age dissimilarity are exacerbated under these conditions, becoming more pronounced and exerting a significant negative impact on the overall tone of the firms' annual report.

This study makes several contributions. This study focuses on highlighting the critical role of the Chair–CEO age dissimilarity in influencing the tone of annual reports. As the leader of the board, the Chair plays a central role in engaging shareholders and fostering positive relationships among board members (Amran, Yusof, Ishak, & Aripin, 2014). Our findings show that a larger Chair–CEO age dissimilarity reduces the positive tone of firms' annual reports. Additionally, the study offers valuable insights into how industry competitiveness, ownership structure and organisational ownership status moderate this relationship. Specifically, the link between the Chair–CEO age dissimilarity and ART is stronger in environments characterised by less concentrated controlling rights, stronger external monitoring pressures, higher information asymmetry, higher agency conflicts and family-owned and non-state-owned firms. These results highlight the complex nature of Chair–CEO age dynamics and their effects on the tone of firms’ annual reports across different organisational settings. Furthermore, the study found that Chair–CEO age dissimilarity influences the positive tone of annual reports through four different mediating channels: internal monitoring intensity, tenure overlap, free cash flow and earnings management. Our mechanism tests suggest that Chair–CEO age dissimilarity reduces the positive tone of annual reports, reflecting a disciplining effect on aggressive reporting behaviour.

This section has provided an introduction, and the remainder of the study is organised as follows. Section 2 reviews the existing literature, discusses the theoretical framework and presents the research hypotheses. Section 3 provides an overview of the data selection process, defines the variables and outlines the model construction. The results are then presented in Section 4, followed by robustness checks, cross-sectional tests, endogeneity tests, mediation tests and mechanism tests. Section 5 concludes the study with a summary of the findings.

Grounded in agency theory and Upper Echelon Theory (UET), this study examines how Chair–CEO age dissimilarity shapes disclosure outcomes. Agency theory highlights conflicts of interest and information asymmetry between shareholders and managers, suggesting that weak governance intensifies agency problems and leads to more cautious or conservative disclosure, reflected in fewer positive expressions and potentially negative signals in annual reports. Such disclosure may undermine investor confidence and firm performance, whereas strong governance mechanisms can mitigate agency conflicts and improve disclosure tone. UET posits that managerial characteristics influence cognition and strategic decisions. Age, in particular, shapes risk preferences and behavioural tendencies, affecting disclosure choices. Moreover, the quality of the Chair–CEO relationship influences governance effectiveness and, in turn, the tone of annual reports.

Based on UET, demographic characteristics are key factors influencing managers' behaviour, decision-making styles and cognitive structures (Hambrick & Mason, 1984). Bamber, Jiang, and Wang (2010) emphasise that the personal characteristics of managers play a profound role in shaping corporate accounting behaviour. Within this framework, the composition of the TMT (including age, professional experience, education and tenure) significantly affects financial reporting quality (Habib & Hossain, 2013), strategic decision-making (Iaquinto & Fredrickson, 1997), overall firm performance and the tone of annual report expressions (Wang, Holmes, Oh, & Zhu, 2016). Age, as a core demographic variable, reflects the accumulated experience of individuals throughout their careers (Herrmann & Datta, 2005). With increasing age, changes in personality traits lead to significant adjustments in behaviour patterns, values and strategic decision-making preferences (Tanikawa, Kim, & Jung, 2017). As TMT age increases, the readability of the Management Discussion and Analysis (MD&A) and annual reports may improve (Xu, Fernando, & Tam, 2018).

Internal conflicts within TMTs are inevitable during strategic decision-making, primarily due to the complexity and high uncertainty of such decisions (Simons & Peterson, 2000). Consequently, internal coordination and collaboration become critical factors influencing the corporate strategy for information disclosure in annual reports. As the proportion of younger managers in TMTs continues to rise, the issue of age dissimilarity within teams has garnered increasing attention. Age differences among TMT members may lead to disparities in work values, cognitive approaches and communication styles, often manifesting as generational effects (Hambrick & Mason, 1984), thereby impacting team collaboration and decision-making efficiency.

Existing research suggests that age diversity within TMTs typically leads to two main types of core conflicts: task conflict and relationship conflict (Jehn, 1997). Task conflict primarily arises from friction caused by differences in opinions during the decision-making process. It can allow for the integration of diverse perspectives and optimise decision quality and drive improvements in organisational performance (Van de Vliert & De Dreu, 1994). However, De Wit, Greer, and Jehn (2012) indicated that the positive effects of task conflict may be limited and may negatively affect team performance. In contrast, relationship conflict is more associated with distrust and antagonism among team members, damaging organisational coordination, hindering information exchange and weakening corporate performance (Parayitam & Dooley, 2009). Chair–CEO age dissimilarity may hinder the development of a close collaborative relationship, limiting information-sharing and thereby exacerbating relationship conflicts (Hu, Bhuiyan, & Houqe, 2026). Additionally, due to Chair–CEO age dissimilarity, the Chair and CEO’s cognitive approaches and values regarding corporate operations might diverge significantly, potentially intensifying internal conflicts, weakening strategic decision-making efficiency and ultimately resulting in a more negative tone in the annual reports.

Furthermore, research indicates that when both task and relationship conflicts are present within an organisation, the positive impact of task conflict on decision quality may be diminished or entirely neutralised (Korsgaard, Soyoung Jeong, Mahony, & Pitariu, 2008), while the negative effects of relationship conflict tend to dominate (De Dreu & Weingart, 2003). Therefore, implementing effective conflict management mechanisms can alleviate relationship conflicts and indirectly reduce the negative effects of task conflicts, thereby optimising the tone and readability of annual reports.

Ross (1977) introduced the theory of signalling and information asymmetry, suggesting that firms use signals to communicate positive or negative information about the company to stakeholders. Unlike financial data, the language of textual information, particularly the positive or negative tone in which it is conveyed, plays a significant role in conveying emotions and expressing attitudes (Loughran & McDonald, 2011). Tucker (2015) highlights that the influence managers have on the tone of disclosures surpasses their impact on numerical reporting. Managers utilise the tone of disclosures to convey information more qualitatively or subtly (Rogers, Van Buskirk, & Zechman, 2011). Stakeholders tend to interpret the tone as an indicator of managerial performance (Balsam, Boone, Liu, & Yin, 2016). Research has shown a significant association between the tone of annual reports and factors such as a company's future financial performance, volatility, fraud allegations and significant internal control deficiencies (Aly, El-Halaby, & Hussainey, 2018). This suggests that the information contained in annual reports is more valuable than that conveyed in earnings announcements (Loughran & McDonald, 2011).

Existing literature on the intent and function of report tone is divided into two opposing views: the impression management view and the information increment view. The impression management view argues that when a company's performance is poor and there is a need to boost its stock price, managers are motivated to manipulate information asymmetry by altering the tone (i.e. providing misleading information) (Osma & Guillamón-Saorín, 2011), thereby misleading investors. In contrast, the information increment view posits that the tone of an annual report plays a crucial incremental role, with managers using a positive tone to signal and reduce information asymmetry, while conveying information about the firm's future performance (Xie & Lin, 2015), profitability (Prencipe, 2004), stock market returns (Jiang et al., 2019), internal control quality (Loughran & McDonald, 2011), external financing needs (Lang & Lundholm, 1993) and the need to increase external investors' investment intensity (Feldman et al., 2010) and address risk features such as bankruptcy, operational and systemic risks (Dobler, Lajili, & Zéghal, 2011; Linsley & Shrives, 2006; Miihkinen, 2012). According to the information increment view, the positive tone of an annual report generates a favourable information increment effect, attracts the attention of external investors (Zhong, Dong, & Chen, 2020) and increases corporate value and alleviate financial constraints (Wang, Zhou, & Gan, 2024). Furthermore, when facing challenging environments, managers tend to reduce the positive reporting tone (Mayew, Sethuraman, & Venkatachalam, 2015).

In general, senior executives play a crucial role in the preparation of annual reports. Within the fields of accounting and finance, the concept of “tone at the top” is particularly significant, indicating that managers adopt a unique style of disclosure and that their disclosure choices are interconnected with investor sentiment (Bamber et al., 2010). Patelli and Pedrini (2015) assert that both the Chair and the CEO influence the tone of disclosures. Their empirical findings suggest that aggressive financial reporting is positively correlated with language that is assertive, syntactically complex and less engaging. Numerous studies have confirmed that board members' characteristics (such as gender, education, financial expertise and board turnover) are associated with more negative disclosure tones, while an older Chair is linked to fewer negative tones in disclosures (Martikainen, Miihkinen, & Watson, 2022).

Based on the previous discussion, team heterogeneity can generate both beneficial and detrimental effects (Milliken & Martins, 1996). Such heterogeneity may increase internal conflict during decision-making, lead to rejection among members, weaken board effectiveness and ultimately harm firm profitability (Abt & Knyphausen-Aufseß, 2017). Some studies report negative associations between age diversity and outcomes such as corporate social responsibility disclosure (Khan, Khan, & Senturk, 2019) and earnings management (Komal et al., 2023). From an agency theory perspective, the relationship between the Chair and the CEO is akin to a principal-agent dynamic, which in this case is influenced by differing risk preferences and time horizons due to their age gap (Hambrick & Mason, 1984). From a UET perspective, the Chair–CEO age dissimilarity can influence the decision-making process through cognitive and demographic factors (Zenger & Lawrence, 1989). A significant age difference may foster cognitive diversity and generational viewpoints (Wei & Wu, 2013). An older Chair nearing retirement may prioritise stability and short-term returns (Carmeli & Halevi, 2009), whereas a younger CEO may prefer higher-risk and growth-oriented strategies (McGuinness, 2021). Such differences in risk preferences and strategic horizons can increase disagreement regarding corporate communication and disclosure decisions. Consequently, greater Chair–CEO age dissimilarity may hinder consensus formation and reduce the positivity of ART.

From the perspective of the similarity-attraction theory, individuals tend to associate with others who are similar to themselves as this fosters stronger relationships, greater cohesion and improved team performance. Greater internal consistency within TMTs helps avoid internal disruptions and enables faster and more effective strategic decisions (Hambrick & Mason, 1984). Accordingly, similarity between the Chair and CEO may facilitate more favourable feedback and reduce communication disputes (Lee, Lee, & Nagarajan, 2014). In contrast, age dissimilarity may intensify cognitive disagreement and trigger relational conflict, thereby hindering collaboration among senior executives (Schmid & Mitterreiter, 2020). Consistent commitment to strategic decisions within TMTs is crucial for firm performance (Pitcher & Smith, 2001). Similarly, Hu et al. (2026) found that CEO age dissimilarity increases conflict and reduces coordination efficiency, which negatively affects firm value and investment efficiency. Consequently, greater Chair–CEO age dissimilarity may reduce communication quality, limit information sharing, increase communication errors and internal conflict, lead to differing interpretations of corporate performance and ultimately impair the tone of firms’ annual reports.

Furthermore, in the Western cultural context, conflict arising from Chair–CEO age dissimilarity may stimulate the integration of strategic perspectives across generations, thereby positively promoting firm value (Goergen et al., 2015). In the context of corporate disclosure, such dynamics may influence how firm performance and prospects are framed in narrative reporting, thereby shaping the tone of annual reports. However, in the Chinese institutional context, cultural norms rooted in collectivism shape how leadership differences influence firm operations and information disclosure. Although Chinese corporate culture emphasises interpersonal harmony and conflict avoidance, these norms do not necessarily eliminate the influence of Chair–CEO age dissimilarity; rather, they shape how such differences are expressed within organisational communication. Prior research shows that collectivist cultures place a strong emphasis on maintaining harmonious interpersonal relationships and avoiding open confrontation (Morris et al., 1998). As a result, Chinese managers tend to adopt conflict-avoidance strategies more frequently than their Western counterparts (Fu et al., 2008). However, such approaches may allow latent tensions to accumulate, potentially leading to relational conflicts that affect narrative disclosure and reduce the positive tone of annual reports. Therefore, given that the ART is a collective communication outcome requiring coordination between the Chair and CEO, we expect the negative effects of age dissimilarity to dominate.

Although UET suggests that age diversity may generate broader perspectives and improve decision quality, the similarity-attraction paradigm and agency theory imply that substantial age differences are more likely to create relational conflict and coordination difficulties.

Therefore, we propose the following hypothesis.

This study begins with an initial sample comprising 5,649 firms (127,524 firm-year observations) listed on the Shanghai and Shenzhen Stock Exchanges between 2000 and 2023, obtained from the CSMAR database. The sample selection procedure is detailed in Table 1, Panel A, following the criteria used in prior literature (Gomariz & Ballesta, 2014).

Firstly, 66,246 firm-year observations are excluded, including observations from firms outside the A-share stock market, those before 2010, firms with CEO duality, financial institutions and firms in the conglomerate industry, which could have skewed the results. Subsequently, 34,402 observations with missing values for key variables are removed. Ultimately, the final sample consists of 3,026 A-share listed firms, encompassing 26,876 firm-year observations over the period 2010–2023. To mitigate the influence of outliers, all continuous variables are winsorised at the 1st and 99th percentiles.

The industry distribution of the final sample is presented in Panel B of Table 1. Firms are classified into 16 industries according to the Standard Industrial Classification (SIC) codes provided by CSMAR. The sample exhibits a well-balanced distribution, with the manufacturing sector representing a significant proportion of firms in the Chinese market.

1. Proxy for ART

Following Loughran and McDonald (2011), sentiment analysis is conducted using the Bag-of-Words (BoW) methodology. Initially, we rely on the “Chinese Sentiment Analysis Dictionary” from China National Knowledge Infrastructure and the financial sentiment lexicon provided by Loughran and McDonald (2011). Subsequently, we translate the English terms from the L&M lexicon into Chinese using the Youdao and Kingsoft dictionaries. The words contained in the annual reports are matched with those in this custom-built Chinese sentiment lexicon, yielding a list of positive and negative sentiment words.

Drawing on existing literature, we define the tone index (TONE) as the density of positive tone words (Xu, Li, Hua, & Ren, 2024). TONE is calculated as the proportion of positive tone words per sentence in a firm's annual report. A higher TONE value indicates a greater proportion of positive tone words, reflecting a more optimistic emotional tone in the report, whereas lower values indicate the opposite. Additionally, industry heterogeneity may lead to differences in corporate ethical awareness (Brammer et al., 2006), thus significantly affecting the emotional tone of annual reports. Following Yuan, Shang, Ma, and Li (2022) and Qi, Chi, Yang, and Ouyang (2023), we introduce the tone peer index (TONE_P) as an alternative sentiment metric to evaluate the emotional tone of corporate annual reports. TONE_P is computed by subtracting the average TONE of firms within the same industry and year from the firm's own TONE. A positive TONE_P value indicates that the firm's proportion of positive tone words in its annual report exceeds the industry average, signalling a more positive tone. Conversely, a negative TONE_P value reflects a tone that is below the industry norm. Finally, inspired by prior research, we developed the tone standardised score index (TONE_S) (Mintzberg, 1978; Tang, Crossan, & Rowe, 2011; Ye, Shahab, Riaz, & Ntim, 2023), calculated by standardising each TONE index for industry and year (TONE_i,t; mean = 0, standard deviation = 1). A higher TONE_S value reflects a significant positive deviation from the industry norm, while a lower value indicates a negative deviation. In sum, this study utilises the TONE, TONE_P and TONE_S indices to assess the tone of annual reports.

1. Proxy for Chair–CEO Age Dissimilarity (Chair–CEO Age Gap)

The measured approach to age dissimilarity has yet to reach a consensus in the existing literature. Building on the conceptual framework developed by Goergen et al. (2015), this study uses two variables to quantify Chair–CEO age dissimilarity: GAPS (signed age dissimilarity) and GAPU (unsigned age dissimilarity). GAPS is measured as the arithmetic difference between the Chair's and the CEO's ages, generating either positive or negative values to indicate whether the Chair is older or younger than the CEO. This measure provides insights into potential cognitive conflicts or experiential divergences arising from age differences between the two executives. Meanwhile, GAPU is defined as the absolute value of GAPS, quantifying the extent of the age dissimilarity regardless of which individual is older.

1. Measurement of Control Variables

Consistent with prior studies, we control for various characteristics of the Chair, the CEO, the supervisory board and the firm that may influence the ART. Firstly, the Chair's and the CEO's characteristics are assessed by considering their respective power dynamics through variables such as replacement (CHANGE1/CHANGE2) and tenure (TENURE1/TENURE2) (Kaplan, Klebanov, & Sorensen, 2012). These variables are used to gauge the likelihood of either tacit understanding or potential conflict between the Chair and the CEO. Secondly, the supervisory board's characteristics are represented by board size (BODSIZE), which refers to the total number of directors on the board. Additionally, firm characteristics include financial leverage (LEVER), the firm's listing age (FIRMAGE), the firm size (FSIZE), the standard deviation of monthly stock returns during the year (STDRET), the sales growth rate (SALEGR), the percentage of independent directors (INDDIR%), the return-on-assets ratio (ROA), the occurrence of losses (LOSS), financial distress risk, as measured by the O-Score, and compensation of the top executives (SALARY) (Davis, Ge, Matsumoto, & Zhang, 2015; Qi et al., 2023; Xu et al., 2024). Appendix A provides further details regarding the definition of variables and the sources used in this study.

Based on prior literature, H1 posits an association between Chair–CEO age dissimilarity and ART. The following regression model is specified to test this hypothesis empirically. Equation (1) is as follows:

Equation (1) adopts the ordinary least squares (OLS) regression method to test H1; industry (INDUSTRY) and year (YEAR) fixed effects are incorporated. The main variable of interest is β₁GAP_i,t, where GAP_i,t serves as an index for the Chair–CEO age dissimilarity. As outlined in Section 2.3.2, this study uses two measures of age dissimilarity: GAPS (signed age dissimilarity) and GAPU (unsigned age dissimilarity). The dependent variable uses three primary metrics: TONE (tone index) (Xu et al., 2024), TONE_P (tone peer index) (Yuan et al., 2022) and TONE_S (tone standardised score index) (Qi et al., 2023). In equation (1), a positive (negative) coefficient of β₁GAP_i,t would suggest that age dissimilarity increases (decreases) the percentage of positive tone in the annual report.

Table 2 presents the descriptive statistics for the key variables utilised in this study. The average (median) of TONE is 3.41 (3.34), with a standard deviation of 0.74, indicating that each sentence in the annual reports of the sampled firms contains 3.41 positive-tone words on average, which is similar to that reported in existing research (Xu et al., 2024; Yuan et al., 2022). Within the sample, 49% of firm-year observations exhibit a more positive tone in their annual reports relative to their industry peers in the corresponding year. The average age difference between the Chair and the CEO (GAPS) is 3.45 years, while the absolute age difference (GAPU ^[1]) averages 5.68 years, which is close to the 5.93 years reported by Hu et al. (2025). The data further reveal that in 19% and 6% of observations, the Chair–CEO age dissimilarity exceeds 10 and 20 years, respectively, which is also similar to the 24% and 6% documented by Zhu et al. (2021). Observations involving a Chair (CEO) turnover during the year account for 18% (26%) of the sample. The average board size is 8.38 members. The average leverage ratio (LEVER) is 0.42, which is comparable to the 0.49 value reported by Zhu et al. (2021). The average firm listing age is 2.16 years, and the mean firm size is 22.08. The average sales growth rate is 17%, indicating that most firms are in an expansion phase with an overall upward trend in revenue generation. Overall, the descriptive statistics are consistent with findings from existing research, particularly studies conducted in the Chinese market (Goergen et al., 2015; Zhu et al., 2021).

Table 3 presents the results of the mean difference test for key variables between two groups: those with a large age difference (GAPU > median) and those with a small age difference (GAPU ≤ median) between the Chair and the CEO. For firms with greater Chair–CEO age dissimilarity (GAPU > median), the three primary tone indices (TONE, TONE_P, TONE_S) are lower and the alternative measures of the dependent variable (TONEF, TONEF_P, TONEF_S) exhibit similar results. Consistently, the frequency of Chair and CEO position changes (CHANGE1, CHANGE2), board size (BODSIZE), financial leverage (LEVER) and the incidence of losses (LOSS) are higher. The reported variables are statistically significant at the 1% level, indicating thereliability of these tests. These findings suggest that the age difference between the Chair and the CEO significantly influences the tone of the company's annual report. This result provides preliminary support for our hypothesis.

Table 4 presents the pairwise correlations between the dependent and independent variables included in the baseline regression model. We observed a significant negative correlation between TONE/TONE_P/TONE_S and GAPU (correlation = −0.02, p < 0.01; correlation = −0.01, p < 0.05; correlation = −0.01, p < 0.05). Furthermore, TONEF/TONEF_P/TONEF_S also exhibited significant negative correlations with GAPS (correlation = −0.02, p < 0.01; correlation = −0.02, p < 0.01; correlation = −0.02, p < 0.01, respectively). Notably, both GAPS and GAPU show significant correlations with nearly all control variables, all statistically significant at the 1% level. Firms with larger board sizes (BODSIZE) or lower percentages of independent directors (INDDIR%) tend to exhibit a lower positive tone in their annual reports.

In Table 5, we investigate the relationship between Chair–CEO age dissimilarity and the ART. Columns 1 and 2 report the effects of GAPS and GAPU on TONE, respectively. Columns 4 and 5 analyse the impact of GAPS and GAPU on TONE_P. Lastly, the effects of GAPS and GAPU on TONE_S are examined in columns 7 and 8. There is a sensitivity test in columns 3, 6 and 9 including both GAPS and GAPSQU in the regression model (Haans, Pieters, & He, 2016). These models examine whether the association between Chair–CEO age dissimilarity and a firm's ART is nonlinear.

As anticipated, the results show negative coefficients for both GAPS and GAPU in relation to TONE (coefficients = −0.002, −0.002; t-statistics = −3.757, −2.906; p < 0.01), suggesting that the positive tone of the firm's annual report decreases with greater Chair–CEO age dissimilarity. Furthermore, we analyse the influence of Chair–CEO age dissimilarity on TONE_P and find that both GAPS and GAPU are negatively associated with it (coefficients = −0.002 and −0.002, t-statistics = −3.791 and −2.806), and these results are statistically significant at the 1% level. Moreover, GAPS and GAPU show a negative association with TONE_S, which is also statistically significant at the 1% level (coefficients = −0.003, t-statistics = −3.780; coefficients = −0.002, t-statistics = −2.829). These findings indicate that an increasing Chair–CEO age dissimilarity corresponds to a reduction in the positive tone of the firm's annual report, supporting H1. Moreover, there is no non-linear effect of Chair–CEO age dissimilarity on the positive tone of firms’ annual reports. Column 3 shows that the coefficient for GAPS is significantly negative, while the coefficient for GAPSQU is not statistically significant (−0.002, t-statistics = −2.239, p < 0.05; 0.000, t-statistics = −0.357, p > 0.10). Columns 6 and 9 show similar results. Regarding economic significance, a one-standard-deviation increase in GAPS is associated with a 0.48% decrease in TONE relative to the sample mean [(−0.002×8.23/3.41) = −0.0048]. This suggests that the effect of Chair–CEO age dissimilarity on the tone of the firm's annual report is economically meaningful. Similarly, a one-standard-deviation increase in GAPU is linked to a 0.42% decrease in TONE relative to the sample mean [(−0.002×7.09/3.41) = −0.0042], further demonstrating the economic significance of this effect.

Regarding the control variables, we observe that firms with longer Chair tenure (TENURE1), higher financial leverage (LEVER), stronger sales growth (SALEGR), higher return on assets (ROA) and higher executive compensation (SALARY) tend to have a more positive tone in their annual reports (Cheng, Dhaliwal, & Zhang, 2013; Khedmati, Sualihu, & Yawson, 2020). We conducted variance inflation factor (VIF) diagnostic tests to address concerns regarding multicollinearity in the regression model. The results indicate that all VIF values are below 10 (Marquaridt, 1970), suggesting that multicollinearity does not materially affect the findings. The adjusted R² values range from 2.28% to 9.66%.

1. Alternative measure of ART

To test the robustness of the research findings (Shahzad, Rehman, Nawaz, & Nawab, 2018), we followed Qi et al. (2023) and used alternative measures of ART. We calculated the tone full index (TONEF) by summing the number of positive and negative words and dividing by the number of sentences in the annual report. In line with the primary measures of ART (TONE, TONE_P and TONE_S), we constructed three additional measures for the dependent variables (TONEF, TONEF_P and TONEF_S). Higher values of these alternative dependent variables indicate a more positive tone in ART.

The results are presented in Table 6. We replicate the analysis from equation (1). Columns 1 and 2, we examine the relationship between TONEF, GAPS and GAPU, respectively. In columns 3 and 4, the relationship between TONEF_P, GAPS and GAPU is explored. Columns 5 and 6 analyse the link between TONEF_S, GAPS and GAPU. All findings support H1. Furthermore, the results for the control variables align with the findings presented in Table 6.

Furthermore, to enhance the reliability of our findings, we conducted additional robustness tests by enriching the measurement of ART using an alternative textual sentiment indicator. Following Jian, Yang, Liu, and Liu (2024), we performed a word-frequency-based sentiment analysis of firms' annual reports using the National Taiwan University Sentiment Dictionary of Chinese Sentiment Polarity ^[2] (NTUSD). A higher value of National Taiwan University Positive Words (NTUPW) indicates a greater prevalence of positive tone words in annual reports. We replaced the baseline tone measure with NTUPW in Equation (1). As reported in columns 1 and 2 of Table 7, all results are consistent with baseline results.

Annual reports consist of multiple narrative sections, such as the MD&A and the Chief Executive Officer's Letter (CEO Letter). Among these sections, management enjoys greater discretion in both content organisation and language choice. To ensure comprehensive coverage of textual content, we separately extracted the MD&A section and the CEO Letter from firms' annual reports. We then conducted regression analyses using the positive tone measures constructed from each section individually. Specifically, we replaced the baseline tone variable with Management Discussion and Analysis Positive Words (MDAPW) and CEO Letter Positive Words (CEOLPW) to re-estimate the models. As shown in Columns 3 to 6 of Table 7, regardless of the specific section of the annual report, the results support H₁.

In addition to audited financial statements, earnings conference calls represent an important disclosure channel through which management communicates firm information to the public (Price, Doran, Peterson, & Bliss, 2012). During these calls, market participants can interact directly with the management, particularly during the question-and-answer (Q&A) session. The tone adopted by managers in earnings conference calls, especially during the Q&A segment, conveys valuable information to investors (Loughran & McDonald, 2011). Unlike annual reports, which are pre-written and highly standardised, conference call disclosures are more flexible and responsive to investor concerns (Larcker & Zakolyukina, 2012). Consequently, the Q&A session provides a more effective setting for validating inferences regarding managerial tone. Accordingly, following Qin, Xiao, He, and Meng (2025), we employ tone data for Chinese listed firms from the Earnings Conference Call Database (ECCD) compiled by the China National Research Data Service (CNRDS). We use the number of positive tone words used by management during the Q&A sessions of earnings calls (ECCCQASPW) as an alternative proxy for managerial tone. As reported in columns 7 and 8 of Table 7, the results are consistent with the prior results, further reinforcing our main results.

Prior dictionary-based approaches may place disproportionate emphasis on positive tone in annual reports while overlooking the overall sentiment composition of the text. To complement the dictionary method and better capture complex emotional expressions, we calculated the proportion of negative tone words and the proportion of other tone-related words and replaced the original dependent variables accordingly. Table 8 illustrates how Chair–CEO age dissimilarity affects the proportion of negative tone words and other tone-related words in annual reports. In contrast to the positive tone, Columns 1 and 2 show that Chair–CEO age dissimilarity is significantly positively associated with the use of negative tone words, while Columns 3 and 4 indicate a significantly negative association between Chair–CEO age dissimilarity and the use of other tone-related words. All results consistently support our hypothesis.

Although the Loughran and McDonald dictionary has been widely applied in financial text analysis, recent studies suggest that FinBERT, a finance-domain-specific language model, outperforms dictionary-based approaches in sentiment classification, particularly in capturing contextual nuances (Huang, Wang, & Yang, 2023). Built upon phrase-level sentiment analysis and advanced Chinese word-segmentation techniques, FinBERT accurately parses textual content before sentiment evaluation, thereby mitigating misclassification arising from character-level or partial-word interpretations (Huang et al., 2023). Moreover, FinBERT is particularly effective in identifying positive or negative sentiment in sentences that are often misclassified as neutral by other algorithms, likely due to its ability to leverage contextual information embedded in financial texts. This approach further integrates textual tone analysis with quantitative performance indicators such as accounting earnings and free cash flows to validate sentiment classifications and improve accuracy.

To critically evaluate the effectiveness of tone measurement, we process firms' annual reports using the FinBERT model and classify textual content into multiple sentiment categories, including positive, negative, uncertain, neutral and irrelevant. Following the main empirical design, we separately construct dependent variables for positive, negative and neutral tone measures, yielding the FinBERT positive tone indices (FinBERT_TONE, FinBERT_TONE_P, FinBERT_TONE_S), FinBERT negative tone indices (FinBERT_NEGT, FinBERT_NEGT_P, FinBERT_NEGT_S), and FinBERT neutral tone indices (FinBERT_NEUTRAL, FinBERT_NEUTRAL_P, FinBERT_NEUTRAL_S).

Panel A of Table 9 presents the associations between Chair–CEO age dissimilarity and sentiment categories derived from the FinBERT framework. We re-estimate Equation (1), and Columns 1 to 3 show that Chair–CEO age dissimilarity significantly reduces the use of positive tone words in annual reports. Columns 4 to 6 indicate a significant positive association between Chair–CEO age dissimilarity and the use of negative tone words, while Columns 7 to 9 demonstrate a significant negative effect on the use of neutral tone words. In Panel B, we exclude 49 observations classified as uncertain and re-estimate the regressions reported in Panel A. The results remain qualitatively unchanged. Overall, all findings consistently support our hypothesis.

1. Alternative measure of Chair–CEO Age Dissimilarity

We develop four alternative measures of Chair–CEO age dissimilarity: GAP20, GAP15, GAP10 and GAP5. GAP20 is a dummy variable assigned a value of 1 if the age difference between the Chair and the CEO is 20 years and 0 otherwise. Similarly, GAP15, GAP10 and GAP5 represent age differences of 15, 10 and 5 years, respectively. Table 10 shows the relationships between Chair–CEO age dissimilarity (GAP20/15/10/5) and the primary annual reports’ tone indicators: TONE (columns 1 to 4), TONE_P (columns 5 to 8) and TONE_S (columns 9 to 12). These results further support H1.

1. Company fixed-effects model

The baseline models in this study incorporate industry and year fixed effects rather than firm fixed effects. Within a given firm, Chair–CEO age dissimilarity typically changes through Chair–CEO turnover and the joint ageing of the two executives. As a result, much of the identifying variation in Chair–CEO age dissimilarity is cross-sectional. Firms that systematically appoint Chairs and CEOs of similar ages may differ from other firms in a variety of unobservable characteristics. To alleviate this concern, we re-estimate our models using firm and year fixed effects as a robustness test.

The results are reported in Table 11. The coefficients on GAPS remain negative and statistically significant for TONE, TONE_P and TONE_S (coefficients = −0.002, −0.002 and −0.003; t-statistics = −3.005, −3.052 and −2.955, respectively; all p-values <0.01). Similar results were obtained when using GAPU as the measure of Chair–CEO age dissimilarity. Furthermore, the results reported in Columns (3), (6) and (9) provide additional evidence that no nonlinear relationship exists between Chair–CEO age dissimilarity and annual reports’ tone. Overall, these results provide further support for H1 and demonstrate that our findings are robust to the inclusion of firm and year fixed effects.

As outlined in the previous section, our research has revealed a distinct negative relationship between Chair–CEO age dissimilarity and ART. We conduct cross-sectional analyses in this section to better understand the underlying complexities of this association, particularly concerning the concentration of controlling rights, institutional investors, information asymmetry, agency conflicts, family-firm and state-owned enterprises’ (SOEs) dynamics.

1. Does controlling right concentration affect the association between the Chair–CEO age dissimilarity and ART?

Controlling rights are a key element in corporate governance (Becht, Bolton, & Röell, 2003). Cao, Di, and Sun (2025) suggest that controlling rights substantially impact the tone of a firm's annual reports. Firms with low controlling rights generally have a more dispersed ownership structure, increasing the potential for internal conflicts. In such firms, disagreements between the Chair and the CEO are more prominent, leading to greater managerial tensions. These conflicts can affect the tone of corporate disclosures, resulting in a more cautious or even negative sentiment in the annual reports, thereby reducing the proportion of positive tone words. In essence, in firms with high controlling rights, agency problems are less prevalent and relational conflicts arising from Chair–CEO age dissimilarity are less likely to influence the linguistic tone of corporate disclosures.

We use controlling rights to measure the extent of ownership concentration and of managerial decision-making authority. Higher (lower) controlling rights indicate stronger (weaker) decision-making power of the controlling shareholder, more (less) concentrated ownership and a lower (higher) likelihood of internal conflicts, leading to a more (less) positive tone in the annual reports. The sample is divided into low-control firms (CCS 50% = 0) and high-control firms (CCS 50% = 1) (Wu & Chang, 2024). Results are presented in Panel A of Table 12. Columns 1 to 6 represent high-control firms, while columns 7 to 12 correspond to low-control firms. Results suggest that the negative effect of Chair–CEO age dissimilarity on ART is mainly evident in firms with low controlling rights or dispersed ownership structures.