From dawn to dusk: The relationship between CEO career horizon and ESG engagement

International Review of Financial Analysis 93 (2024) 103200 Available online 15 March 2024 1057-5219/© 2024 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). From dawn to dusk: The relationship between CEO career horizon and ESG engagement David Cabreros * , Gabriel de la Fuente, Pilar Velasco Universidad de Valladolid, Spain ARTICLE INFO JEL codes: G30 G34 M12 M14 Keywords: ESG Prospect theory CEO career horizon Decision framing ABSTRACT This study investigates the influence of a CEO's career horizon on a firm's engagement in ESG. Given the private benefits which CEOs can extract from this strategy, we argue that a firm's willingness to pursue additional ESG efforts is likely to mirror CEOs' desire to strengthen their personal reputation and curb their perceived risk exposure at each stage of their career horizon. Drawing on the prospect theory, we theorize the decomposition of CEOs' career horizon into two counteracting forces (career concerns and legacy concerns) to explain how a CEO's effort in ESG could change depending on the length of the horizon and the influence of a reference point, such as their firm's performance relative to the industry benchmark. Using a sample of European listed companies from 2007 to 2018, our evidence supports an inverted U-shaped relationship between CEO career horizon and a firm's ESG engagement. This result shows the importance of a CEO's eagerness for personal reputation as a key antecedent of companies' ESG actions. However, a CEO's perceived risk exposure proves not to be a relevant driver of ESG, since we find that such a relationship is not sensitive to decision framing based on relative performance. 1. Introduction Prior literature extensively echoes the advantages of ESG (environmental, social and governance) strategy for firms and their stakeholders. Risk management is emphasized as a primary target (Koh, Qian, & Wang, 2014; Shiu & Yang, 2017), which emerges from the accrual of a firm's moral capital and legitimacy (Godfrey, 2005; Godfrey, Merrill, & Hansen, 2009) as well as stakeholder trust enhancement (Lins, Servaes, & Tamayo, 2017). Moreover, ESG engagement can reflect the private interests of certain stakeholders who have the power to shape corporate strategy. Among these, CEOs wield a major influence as top decisionmakers, such that their interests and personal traits are likely to play a central role in their corporate decisions. These interests and traits could offer alternative insights to enhance our understanding of the different ESG strategies observed across companies, since top managers bring their own personality, values and cognitive bases to deal with everyday problems and make strategic decisions accordingly (Hambrick, 2007; Hambrick & Mason, 1984). The so-called “CEO effect” is attributed a central role in firm strategy and its outcomes (Quigley & Hambrick, 2015). In this vein, Kang (2016) shows that CEOs who are nearing retirement reduce a firm's ESG efforts, although exactly how CEO willingness to engage in ESG might vary throughout their career remains an unexplored topic. This research seeks to fill this gap by investigating the relationship between CEO career horizon and a firm's level of ESG engagement. We posit that a CEO's career horizon sparks two opposite forces (career concerns and legacy concerns) which shape CEOs' willingness to pursue ESG actions. We focus on two private benefits that CEOs may extract from ESG: on the one hand, a stronger reputation as top leaders due to their alignment with ESG values and moral capital, and on the other, a reduction in firm risk, which decreases the exposure of their undiversified human capital investment. To elaborate our arguments, we examine how these private benefits for CEOs evolve throughout their career horizon. We argue that CEOs might prioritize gaining personal reputation to a different extent depending on which stage of their career they are at, as a result of the combination of career concerns and legacy concerns, which might have a different impact on their perceived need for personal reputation. In particular, we contend that a CEO's career and legacy concerns may involve two opposing forces that drive an inverted U-shaped relationship between CEO age and ESG engagement. Second, we delve into CEOs' private interest in alleviating their perceived exposure to a firm's risk through ESG engagement. Given such insurance benefits from ESG, this strategy's implementation is likely to depend on top managers' own risk perception. While previous research * Corresponding author at: Universidad de Valladolid, Faculty of Economics and Business, Avenida Valle del Esgueva 6, 47011 Valladolid, Spain. E-mail addresses: [email protected] (D. Cabreros), [email protected] (G. de la Fuente), [email protected] (P. Velasco). Contents lists available at ScienceDirect International Review of Financial Analysis journal homepage: www.elsevier.com/locate/irfa https://doi.org/10.1016/j.irfa.2024.103200 Received 24 July 2023; Received in revised form 2 December 2023; Accepted 7 March 2024 International Review of Financial Analysis 93 (2024) 103200 2 has applied prospect theory premises to other insurance strategies such as corporate diversification (Hutzschenreuter, Kleindienst, Groene, & Verbeke, 2014), whether it also applies to ESG still awaits further attention. From the prospect theory, individuals' interpretations are unavoidably affected by their framing, which leads them to assess potential outcomes of their decisions and perceived risk based on a neutral reference point. The prospect theory expects individuals to be more risk averse when outcomes represent gains (performance above the reference point) and to become more risk-seekers in contexts of losses (performance below the reference point) (Fiegenbaum & Thomas, 1988; Holmes, Bromiley, Devers, Holcomb, & McGuire, 2011). Consequently, such an inverted U-shape between CEO career horizon and ESG engagement is likely to be affected by the firm's performance relative to its industry benchmark (Hirshleifer, 1993). We argue that if a firm's relative performance shapes its CEO's risk aversion, the influence of career and legacy concerns over the propensity to implement additional ESG actions will differ between overperformers and underperformers. Should CEOs rely on ESG engagement to reduce their perceived risk exposure, they will be more likely to increase their ESG efforts when the firm outperforms its industry peers and to reduce them when a firm underperforms. Based on these premises, we hypothesize a shift in the inverted U-shaped curve. Our hypotheses are tested on a sample of listed European companies during the 2007–2021 period. Results support the curvilinear association between CEO age and ESG engagement: younger CEOs display ever increasing engagement to ESG up to a turning point in their age, after which CEO eagerness to pursue additional improvements in ESG is reversed. Maximum ESG engagement is reached when the CEO is around 53 years old. However, a CEO's perceived risk exposure motivation is not relevant and such risk behaviour in the ESG context shifts from conventional prospect theory contentions, since a firm's performance relative to its industry benchmark does not affect a CEO's ESG efforts. The remainder of the paper is structured as follows. Section 2 sets out the theoretical background from which our empirical propositions are derived. Section 3 describes the sample, variables and empirical models to test the hypotheses. Section 4 presents and discusses the main results and provides a number of sensitivity analyses. Finally, Section 5 concludes. 2. Theoretical background and hypotheses Personal traits, values and background of top managers (e.g., CEOs) are reflected in their corporate strategies and outcomes (Hambrick, 2007; Hambrick & Mason, 1984; Wang, Holmes Jr., Oh, & Zhu, 2016). Since individuals have limited rationality, top managers draw on their cognitive base and values to perceive environmental and organizational stimuli (Hambrick & Mason, 1984). These are found to influence their strategic decisions, with ESG engagement being no exception (Arena, Michelon, & Trojanowski, 2018; Fu, Tang, & Chen, 2020; O'Sullivan, Zolotoy, & Fan, 2021; Sajko, Boone, & Buyl, 2021). A firm's engagement in ESG grants CEOs a number of benefits at the private level. On the one side, a stronger alignment of their firms with the values associated with ESG and moral capital help strengthen the professional and personal reputation of CEOs as top leaders of such firms (Borghesi, Houston, & Naranjo, 2014; Cai, Gao, Garrett, & Xu, 2020; Kang, 2016). On the other side, ESG sparks risk-management benefits (Godfrey, 2005; Godfrey et al., 2009; Jia, Gao, & Julian, 2020; Koh, Qian, & Wang, 2014; Shiu & Yang, 2017), which become particularly salient during crisis episodes (Lins et al., 2017). Given that CEOs have their undiversifiable human capital allocated in the firm (Wang, Barney, & Reuer, 2003), they are personally interested in mitigating a firm's risk in order to curb their perceived risk exposure. In fact, CEOs purposely draw on ESG in order to reap its insurance-like benefits (Jia et al., 2020). Consequently, ESG engagement might mirror CEO risk-taking (Borghesi et al., 2014; Chintrakarn, Jiraporn, & Treepongkaruna, 2021; He, Ding, Yue, & Liu, 2023). Stronger CEO risk appetite might lead them to underestimate risks and, therefore, weaken ESG efforts (McCarthy, Oliver, & Song, 2017; Sajko et al., 2021). In contrast, firms with less riskseeking CEOs are likely to exhibit stronger ESG engagement. 2.1. CEO career horizon and ESG engagement A CEO's career horizon is the “amount of time remaining until a CEO reaches retirement age” (Krause & Semadeni, 2014, p. 813). It is a key variable that determines their desire for a stronger personal and professional reputation and, consequently, their willingness to engage in additional ESG. Recent works such as Cai et al. (2020) go along with this idea by showing that CEOs are judged on the ESG-based reputation of their companies, which can enhance their professional prospects. We draw on Haans, Pieters, and He (2016) to delve into the rationale for the joint influence from two counteracting forces: CEO career concerns and CEO legacy concerns. 1 Fig. 1 illustrates the two forces and their additive effect. For the sake of easier interpretation, the X-axis is based on CEO age, which is an inverse proxy for career horizon. On the one hand, the force of career concerns reflects CEOs' worries about their labour market evaluation and career prospects (Antia, Pantzalis, & Park, 2010; Gibbons & Murphy, 1992; Kang, 2016; Nadeem, Zaman, Suleman, & Atawnah, 2021). Career concerns are greater the further the CEO is from retirement (Gibbons & Murphy, 1992), and peak at the beginning when they are younger. At this stage, CEOs are likely to display stronger ESG engagement as a strategic insurance tool to strengthen their professional and personal reputation as corporate leaders. As CEOs become older and reach later-career stages, lower career concerns might reduce their need to nurture their reputation and would therefore weaken additional ESG engagement. The career concerns curve will reach zero in the X-axis at retirement when career concerns no longer exist. Moreover, there might be increasing marginal reduction because a CEO's relief from career concerns is more prominent when near retirement (Kang, 2016; McClelland, Barker III, & Oh, 2012). Fig. 1. CEO career horizon (CEO age) and ESG engagement. 1 The approach by Haans et al. (2016) helps to theorize the counterbalancing forces whose addition produces a non-linear relationship. This approach aims to identify the two causal underlying mechanisms which drive the relationship between CEO career horizon and a firm's ESG engagement, and provide the theoretical rationale on which we build our arguments. D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 3 Fig. 1 depicts this negative association between CEO age and additional ESG engagement as engendered by career concerns. On the other hand, legacy concerns represent CEOs' worries about preserving a legacy of success as their lasting imprint within the firm (Bilgili, Campbell, O'Leary-Kelly, Ellstrand, & Johnson, 2020; Gao, Hua, Liu, & Yan, 2023; Kang, 2016; Matta & Beamish, 2008;). Stronger legacy concerns drive CEO willingness to avoid damaging their personal reputation and legacy associated with socially responsible actions (Kang, 2016; Matta & Beamish, 2008), which are likely to stimulate ESG engagement. Logically, these concerns increase with CEO age, although the increasing marginal effect might diminish when they are older due to the long-term nature of additional ESG engagement, as Fig. 1 shows. Considering these two counteracting underlying mechanisms jointly, Fig. 1 shows how the resulting added effect causes an inverted U-shaped relationship between CEO age and ESG engagement. Hence, we hypothesize: H1. CEO age displays an inverted U-shaped relationship with ESG engagement. 2.2. CEO career horizon and ESG engagement: The decision-framing effect Perceived outcomes from ESG (and in particular, CEOs' perceived need for risk management) may depend on corporate performance relative to a reference (or target) performance level (Holmes et al., 2011). Prospect theory states that decisions are reference-dependent and that decision-makers exhibit a mixture of risk-seeking and risk-averse behaviour (Fiegenbaum & Thomas, 1988; Kahneman & Tversky, 1979). They are expected to have risk-seeking choices when performing below the reference point, whereas they become risk averse when performing above the reference point (Fiegenbaum & Thomas, 1988; Holmes et al., 2011). In our research setting, we delve into how firm performance vis-` a-vis the industry performance benchmark shapes the career and legacy concern mechanisms discussed earlier, thereby moderating the CEO career horizon-ESG engagement relationship. Under an outperformance scenario, CEOs might become more risk averse according to the prospect theory (Fiegenbaum & Thomas, 1988), and might perceive the potential loss of current performance status as a more serious threat. Therefore, risk management will have greater importance for them, and would lead them to prioritize additional ESG efforts in order to safeguard both their career and legacy concerns. As Fig. 2 depicts, both forces shift upward, with the departing level of ESG for the career concern curve and the endlevel of the legacy curve increasing. In contrast, the two counterbalancing mechanisms might be influenced in the opposite way in an underperformance scenario as a result of greater CEO risk appetite and a lower incentive to implement additional ESG for risk management. Moreover, we contend that career concern and legacy concern mechanisms might not display the same degree of response to performance framing. We expect the legacy curve to be more sensitive to the decision-framing effect, with this curve displaying a more prominent movement upward (downward) in ESG engagement, under an outperformance (underperformance) scenario, as illustrated in Fig. 2. This is because, whereas young CEOs face long-term horizons to gradually adapt their strategic decisions to their risk appetite, older CEOs are subject to short-term horizons that demand a more substantial response. Overall, the greater sensitivity to the decision-framing effect for legacy concerns results in shifting the turning point of the added effect curve to the right (left) for outperformers (underperformers). Taking all these arguments together, we hypothesize: H2. Performance framing moderates the inverted U-shaped relationship between CEO age and ESG engagement such that the maximum of the curve shifts to the right (left) if the firm outperforms (underperforms) the reference point. 3. Empirical design 3.1. Data and sample Our sample consists of European publicly traded companies from 2007 to 2021. 2 We gather financial data from Worldscope, and ESG data by Refinitiv, both of which were accessed through the Eikon platform. Refinitiv is considered to provide objective, auditable and systematic information on ESG initiatives. Data on corporate governance comes from NRG Metrics. Recent studies report that this latter database offers good coverage for large-cap indices worldwide, and indeed it has been drawing on a large team of expert analysts to review and hand-collect detailed information on corporate governance characteristics from firms' annual reports since 2007 (Attig, El Ghoul, Guedham, & Zheng, 2021). We exclude countries with excessively poor (or no) coverage of ESG data. The final sample comprises firms from 18 European countries: Austria, Belgium, the Czech Republic, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Poland, Portugal, Spain, Sweden, Switzerland, and the United Kingdom. We apply a number of selection criteria to build our final sample. First, we eliminate those firms operating in the finance and insurance sector (SICs from 6000 to 6499 and from 6700 to 6799), and those with negative common equity. Additionally, we convert ESG scores from a 100-point scale to a 10-point one so as to avoid heteroscedasticity problems (Cheng, Ioannou, & Serafeim, 2014; Fuente, Ortiz, & Velasco, 2022). Finally, we winsorize our variables (except for governance and ESG variables, which do not display extreme outlier observations) at the 1st and 99th percentiles. Observations with missing data on our key variables are also omitted. All of these sample selection criteria lead to a final sample of 7982 firm-year observations corresponding to 991 companies. Table 1 presents the sample distribution by year and country. The number of observations is seen to have grown in more recent years due to increasing ESG data coverage. The countries with the highest portion of observations in our sample are the UK (27.57%), France (14.73%), and Germany (12.63%). 3.2. Variables Table A.1. of the Appendix summarizes the definitions of our variables. 3.2.1. Dependent variable: ESG engagement Following previous literature (Cheng et al., 2014; Fuente et al., 2022; Gomes, 2019), we measure ESG engagement by the equally-weighted average of the scores of the three pillars of ESG (environmental, social, and governance) (ESGscore). Alternatively, we also use the average between the scores of the environmental pillar and the social pillar (ESscore) for additional robustness since some works do not consider the governance pillar to be a dimension of a firm's sustainability strategy (Lins et al., 2017). 3.2.2. Explanatory and moderating variables We proxy CEO career horizon by a firm's CEO age (Cho & Kim, 2017; McClelland et al., 2012). This is an inverse proxy since the older the CEO, the shorter their career horizon. As regards our moderating variable of a firm's relative performance, we base it on a firm's return on assets (ROA) relative to the median ROA of their industry peers each year. To identify the industry which each firm belongs to, we apply the 2 The time span of our sample seems of particular interest given the substantial changes in ESG regulation that have been implemented in Europe over the last few years, such as the Sustainable Finance Action Plan in 2018 and the European Green Deal in 2020. We thank an anonymous reviewer for raising this issue. D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 4 48-Fama and French (1997) industry classification scheme. We construct a dummy variable (dumROA_indFF) which equals one if a firm's ROA is above each year's industry median (outperformer), and zero otherwise (underperformer). 3 For robustness purposes, we alternatively use the 2-digit SIC classification, as well as the SIC division scheme 4 by the U.S. Department of Labor (Fuente et al., 2022) which classifies 2digit SIC codes into 10 divisions (dumROA_indDIV). These three alternative dummy definitions allow us to identify industry outperforming firms and industry underperforming ones to test Hypothesis 2. 3.2.3. Control variables Following earlier research (Fu et al., 2020; Jia et al., 2020; Oh, Cha, & Chang, 2017), we consider a number of controls about firm-level financial characteristics, which are expected to influence a firm's ESG engagement: financial performance (ROA), firm size (SIZE), firm leverage (LEV), cash reserves (CASH), asset tangibility (TANG), and investment opportunities (INVEST). Larger, better performing and richer cash companies are likely to allocate more resources to ESG actions due to the greater availability of financial resources and more visibility. In contrast, greater debt obligations reduce available resources and might curtail ESG. Asset structure is likely to affect firm investment such as ESG, and broader investment opportunities might add an opportunity cost to ESG actions. Moreover, we control for governance-related characteristics (Fu et al., 2020; Oh et al., 2017) since better governed companies usually display greater ESG engagement: board size (BOARDSIZE), board independence (BOARDINDEP), CEO duality (CEOduality), and institutional ownership (INSTOWN). Finally, we add year, industry (based on divisions by the U.S. Department of Labor), and country dummies. 3.3. Empirical models and estimation methodology Our equations are estimated by Tobit regression. This censored regression approach seems appropriate because our dependent variable (ESG engagement) is a continuous measure bounded at zero and 10 (Jia & Li, 2020). Such a range restriction makes ordinary least squares potentially unsuitable (Amore & Murtinu, 2021). Following Petersen (2009), we estimate a pooled Tobit and control for firm fixed effect (i.e. residual correlation) by clustering standard errors by firm. 5 Our baseline model to test the inverted U-shaped relationship between a firm's CEO age and ESG engagement (proposed in Hypothesis 1) is specified as follows: ESGi,t=β0+β1*CEOagei,t+β2*CEOage2 i,t+β3*CONTROLSi,t+YEARt +INDUSTRYj+COUNTRYk+ ε i,t (1) where i, t, j and k denote each firm, year, industry, and country, respectively. ESG i,t represents a firm's engagement in ESG, which is measured by ESGscore (and alternatively, ESscore for robustness). CONTROLS i,t is the vector of control variables. YEAR t , INDUSTRY j , and COUNTRY k are the sets of dummies to control for time-, industryand country-fixed effects, respectively. ε i,t is the error term. To test Hypothesis 2, we extend the baseline equation by adding the moderating role of a firm's relative performance to its industry peers on Fig. 2. CEO career horizon (CEO age) and ESG engagement: the decision framing effect. Note: dotted lines represent the baseline situation. Solid lines represent the situation of outperformance scenario (in the graph on the left) and the situation of underperformance scenario (in the graph on the right). 3 Robustness analyses were conducted by calculating this dummy with reference to each year's industry performance in each country. We decided to discard “country” when calculating industry performance in order to maximize sample size in each industry and reach more meaningful median performance values. Moreover, most listed European firms compete with industry peers worldwide. 4 Information about the correspondence of these industry divisions with SIC codes is provided in the official website: https://www.osha.gov/pls/imis/sic_ manual.html. 5 We opted not to apply a Tobit fixed-effect estimation because previous literature raises concerns that this might possibly drive an incidental parameter problem which produces biased and inconsistent estimates (Amore & Murtinu, 2021; Greene, 2004). D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 5 the CEO age-ESG engagement relationship (Eq. (2)): ESGi,t=γ0+γ1*CEOagei,t+γ2*CEOage2 i,t+γ3*CEOagei,t*dumROAindi,t +γ4*CEOage2 i,t*dumROAindi,t+γ5*CONTROLSi,t +YEARt+INDUSTRYj+COUNTRYk+ ε i,t (2) where dumROAind i,t denotes a firm's performance dummy relative to its industry benchmark (either dumROA_indFF or dumROA_ind3d). γ 1 and γ 2 capture the non-linear effect of CEO age on a firm's ESG engagement for below-median industry performance firms (i.e., underperformers), and by (γ 1 +γ 3 ) and (γ 2 +γ 4 ) for above-median industry performance firms (i.e., overperformers). Haans et al. (2016) distinguish two types of moderating effects in curvilinear relationships: curve flattening or steeping, and turning point shift. Hypothesis 2 proposes the latter type. Therefore, we first need to verify that γ 3 and γ 4 are not statistically significant, which implies that the shape (slopes) of the curvilinear relationship between CEO age and ESG engagement is not affected by the performance framing. This will serve to discard the presence of curve flattening or steeping. Second, we follow Haans et al.'s (2016) procedure to formally test whether the turning point shifts when the moderating variable (dumROAind) changes. 4. Results 4.1. Descriptive statistics Table 2 summarizes the descriptive statistics. On average, sample firms exhibit an ESG engagement score of about 5.35 out of 10, and average CEO age is about 54 years old. Therefore, if we assume a 70 year-old retirement age, our average sample CEO has a career horizon of 16 more years ahead. Table 3 presents the pairwise correlation matrix of our variables. CEOage is positively correlated with ESGscore (0.1188, p <0.01) and ESscore (0.1548, p <0.01), and negatively correlated with ROA (−0.0552, p <0.01). 4.2. Baseline model: CEO career horizon and ESG engagement Table 4 reports the Tobit estimation results of Eq. (1) concerning the effect of CEO career horizon (CEO age) on a firm's ESG engagement. Panel A contains the estimations using ESGscore as the dependent variable proxy. Column (1) shows the estimates only with control variables included. As expected, superior financial performance, larger firm size, richer cash reserves, larger boards and boards with a stronger presence of independent directors have a positive impact on ESGscore. In contrast, higher leverage, stronger investment opportunities and CEO duality are negatively associated with ESGscore. Column (2) displays the results of the linear model. The effect of CEOage on ESGscore is not significantly different from zero (β 1 = −0.0030, p >0.10). Interestingly, when we also consider the quadratic term of CEOage in Column (3), results suggest an inverted U-shaped relationship between CEO age and ESGscore, supporting our Hypothesis 1. The estimated coefficient of the linear term CEOage is positive (β 1 = 0.1045, p <0.01) whereas that of the quadratic term CEOage_sq is Table 1 Sample distribution by year and country. Panel A: Distribution of firm-year observations by year Year No. of observations % of observations 2007 403 5.05 2008 418 5.23 2009 452 5.66 2010 485 6.08 2011 529 6.63 2012 532 6.66 2013 516 6.46 2014 513 6.43 2015 525 6.58 2016 534 6.69 2017 572 7.17 2018 758 9.50 2019 531 6.65 2020 608 7.62 2021 606 7.59 Total 7982 100.00 Panel B: Distribution of firm-year observations by country Country No. of observations % of observations Austria 162 2.03 Belgium 208 2.60 Czech Republic 33 0.41 Denmark 288 3.61 Finland 368 4.61 France 1176 14.73 Germany 1008 12.63 Greece 82 1.03 Ireland 129 1.62 Italy 510 6.39 The Netherlands 282 3.53 Norway 198 2.48 Poland 26 0.33 Portugal 90 1.13 Spain 187 2.34 Sweden 383 4.80 Switzerland 651 8.16 The UK 2201 27.57 Total 7982 100.00 This table shows the distribution of firm-year observations by year (Panel A) and by country (Panel B). The sample comprises an unbalanced panel of listed companies from 18 European countries during the 2007–2021 period. Table 2 Descriptive statistics. Variable N Mean SD p25 Median p75 ESG engagement ESGscore 7982 5.3553 1.9431 3.9556 5.4260 6.8816 ESscore 7982 5.3764 2.3097 3.5975 5.5275 7.2720 CEO career horizon CEOage 7982 54.2400 6.9712 50 54 59 Control variables ROA 7982 0.0556 0.0830 0.0280 0.0537 0.0872 SIZE 7982 15.139 1.590 14.059 15.095 16.152 LEV 7982 0.2548 0.1634 0.1330 0.2432 0.3604 CASH 7982 0.0918 0.0930 0.0316 0.0650 0.1203 TANG 7982 0.5025 0.3907 0.1915 0.3863 0.7644 INVEST 7982 0.0430 0.0374 0.0183 0.0333 0.0558 BOARDSIZE 7982 2.2783 0.3755 2.0794 2.3025 2.4849 BOARDINDEP 7982 0.6010 0.2225 0.4600 0.5700 0.7500 CEOduality 7982 0.2350 0.4240 0 0 0 INSTOWN 7982 0.1497 0.1677 0 0.0989 0.2429 This table summarizes the main descriptive statistics of the final sample. ESG engagement is captured by ESGscore (the average of the scores of the environmental, social, and governance pillars) and ESscore (the average of the scores of the environmental and social pillars), both of which are based on a 10-point scale. CEO career horizon is approximated by CEOage (a firm's CEO age in years). Control variables are: ROA (a firm's financial performance as return on assets), SIZE (a firm's size as the natural logarithm of the book value of total assets), LEV (leverage as the ratio of total debt to the book value of assets), CASH (cash reserves as the ratio of cash to the book value of assets), TANG (asset tangibility as the ratio of property, plant and equipment to the book value of assets), INVEST (investment opportunities as the ratio of CAPEX to the book value of assets), BOARDSIZE (the natural logarithm of the number of board directors), BOARDINDEP (the ratio of the number of independent directors to total board directors), CEOduality (a dummy variable equal to one if the CEO is also the board chair, zero otherwise), and INSTOWN (the percentage of shares held by institutional investors). All variables are winsorized at the 1st and 99th percentiles (except for ESG and governance variables, which do not display extreme outlier observations). D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 6 negative (β 2 = − 0.0010, p <0.01), with both being beyond the 1% level of statistical significance. At the bottom of this column, we see that the maximum value of the curve is reached at CEOage* =53.52 years old, which is within the range of our data and slightly below the sample average. This finding confirms that CEOs with a long career horizon (younger CEOs) are willing to pursue additional increases in their firms' ESG engagement as they get older. This result is consistent with younger CEOs –as they become older– increasing their legacy concerns at a faster rate than at which they reduce their career concerns. When some 16 years of career horizon remain until retirement age (broadly assumed to be 70 years old in previous works, e.g., Matta & Beamish, 2008), such behaviour reverses and CEOs exhibit a gradually diminishing propensity to implement additional ESG actions as they grow old. This evidence supports the hypothesis that –in the latter stages of a CEO's time in office– their career concerns decrease at a faster rate than at which their legacy concerns increase. Panel B of Table 4 summarizes additional analyses to verify the robustness of our findings. First, Column (4) repeats the estimations using ESscore as a dependent variable, with results remaining similar. Columns (5) to (7) are based on the ESG pillars, considering the environmental pillar (ENV), the social pillar (SOC), and the governance pillar (GOV) scores. Results confirm the inverted U-shaped relationship between CEO age and a firm's engagement in the environmental and social pillars, but not for the governance pillar. In regressions with GOV as the dependent variable, the linear term CEOage displays no statistical significance (β 1 =0.0170, p >0.10; β 2 = − 0.0003, p >0.10). Except for GOV, in the remaining cases the inverted-U association between CEOage and the alternative proxies for ESG is confirmed. The turning point of the curve is reached in the CEO age range of between 53 and 58 years old. This is consistent with prior evidence (Bannier, Bofinger, & Rock, 2022; Zhao, Fang, & Zhang, 2023) which attributes greater legitimacy and credibility to efforts made in the environmental and social pillars as a result of being associated with a broader range of stakeholders. 4.3. The influence of decision framing on the relationship between CEO career horizon and ESG engagement In this section, we perform a number of analyses to test Hypothesis 2 concerning the moderation of the inverse U-shaped relationship between CEO age and ESG engagement by a firm's relative performance (compared with its industry benchmark), which might cause a turning point shift. Table 5 reports the Tobit estimation results of Eq. (2) considering interaction effects between CEO age and the alternative dummies (based on different industry classification schemes) to identify industry overperformers and underperformers. The interaction effects display no statistical significance across all regressions. For example, when based on the 48-Fama-French industry classification in Column (1), the coefficient of CEOage×dumROA_indFF is negative (γ = − 0.0037, p >0.10) and the non-linear interaction effect is positive (γ =0.0001, p >0.10), although both lack statistical significance. Although the absolute value of the coefficient associated with the non-linear effect is lower for above-median industry performing firms (∑nonlinear =γ 2 +γ 4 = −0.0009) than for below-median industry performers (γ 2 = − 0.0010), we rule out the existence of curve flattening since the difference between the two coefficients cannot be considered statistically significantly different from zero (i.e., interaction effects are not significant). We follow Haans et al.'s (2016) procedure to formally test whether a moderation in the form of a turning point shift occurs in the inverted Urelationship. The turning point CEOage* of Eq. (2) is as follows: CEOage*=−γ1−γ3•dumROAind 2•γ2+2•γ4•dumROAind (3) which depends on the moderator (dumROAind). Taking the first derivative of the previous expression with respect to the moderating variable, we obtain: Table 3 Spearman's pairwise correlations. Variable 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 1.ESGscore 1 2.ESscore 0.9398*** 1 3.CEOage 0.1188*** 0.1548*** 1 4.ROA 0.0041 0.0164 −0.0552*** 1 5.SIZE 0.6060*** 0.6221*** 0.1543*** −0.0002 1 6.LEV 0.1004*** 0.1083*** 0.0421*** −0.1763*** 0.2289*** 1 7.CASH −0.1332*** −0.1518*** −0.0054 −0.1272*** −0.3401*** −0.2349*** 1 8.TANG 0.0953*** 0.1097*** −0.0331*** −0.0101 0.1779*** 0.1717*** −0.2530*** 1 9.INVEST −0.0372** −0.0393** −0.0936*** 0.0937*** 0.0274** 0.0661*** −0.1392*** 0.5661*** 1 10.BOARDSIZE 0.3684*** 0.4107*** 0.1301*** −0.0395*** 0.6011*** 0.1319*** −0.1960*** 0.1530*** 0.0518*** 1 11.BOARDINDEP 0.0801*** 0.0464*** −0.0758*** 0.0171 0.0112 −0.0547*** 0.0343*** 0.0055 −0.0322*** −0.2073*** 1 12.CEOduality 0.0175 0.0947*** 0.1267*** 0.0012 0.1194*** 0.0186* 0.0167 −0.0555*** −0.0309*** 0.1645*** −0.1223*** 1 13.INSTOWN −0.1436*** −0.1898*** −0.1327*** 0.0046 −0.2755*** −0.0595*** −0.0020 −0.0331*** −0.0242** −0.2459*** 0.0271** −0.1340*** 1 This table shows the pairwise Spearman correlation coefficients for our study's variables. All of them are winsorized at the 1st and 99th percentiles (except for ESG and governance variables, which do not display extreme outlier observations). ***, ** and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 7 ∂ CEOage* ∂ dumROAind =γ1•γ4−γ2•γ3 2• (γ2+γ4•dumROAind)2(4) Since the latter expression depends on the moderating variable dumROAind, we first need to assign specific values to this variable (namely zero and 1) and then conduct the test of whether Eq. (4) is significantly different from zero. 6 Should this be the case, it would imply that the turning point of the curve (CEOage*) depends on a firm's performance relative to its industry (dumROAind), which would support the idea that performance framing moderates the inverted U-shaped relationship between CEO age and ESG engagement by shifting its turning point. The results of this test are summarized at the bottom of Table 5 for the two possible values of the moderator (zero and 1). In all cases, we cannot reject the null hypothesis that Eq. (4) equals zero (p >0.10). Consequently, turning point is not affected by each firm's performance framing relative to the industry median, such that Hypothesis 2 is not supported. Additionally, we evaluate the moderating effect of performance framing by adopting a sample-split approach. Similar to prior works (e. g. Fuente et al., 2022), we estimate the baseline Eq. (1) for subsamples of industry outperformers and underperformers separately, and then assess whether the value of CEOage in the turning point of the curve for each subsample is statistically significantly different from that estimated in the baseline model for the full sample. Additionally, we perform analyses to examine whether there is a statistically significant difference between estimated extreme points between subsamples. 7 Table 6 reports these results. At the bottom of the table, p-values lead us not to reject the Table 4 CEO career horizon (CEO age) and ESG engagement. Panel A: Main analyses Panel B: Alternative measures of sustainability performance: Robustness analyses Dependent variable: ESGscore Dependent variable: ESscore Dependent variable: ENV Dependent variable: SOC Dependent variable: GOV (1) (2) (3) (4) (5) (6) (7) Constant −6.9610*** −6.8165*** −9.6485*** −13.1476*** −15.6946*** −10.8629*** −2.7741* (0.3829) (0.4457) (1.1337) (1.4437) (1.8006) (1.4650) (1.4235) CEOage −0.0030 0.1045*** 0.1501*** 0.1543** 0.1407*** 0.0170 (0.0046) (0.0395) (0.0510) (0.0646) (0.0510) (0.0500) CEOage_sq −0.0010*** −0.0013*** −0.0014** −0.0012*** −0.0003 (0.0003) (0.0004) (0.0006) (0.0004) (0.0005) Control variables ROA 0.7753** 0.7726** 0.7804** 1.3954*** 2.0570*** 1.2957*** −0.4417 (0.3238) (0.3235) (0.3236) (0.4106) (0.5321) (0.4552) (0.3853) SIZE 0.7929*** 0.7943*** 0.7927*** 0.8752*** 0.9831*** 0.7902*** 0.6284*** (0.0306) (0.0306) (0.0305) (0.0345) (0.0399) (0.0380) (0.0395) LEV −0.6877*** −0.6887*** −0.6987*** −0.6695** −0.8437** −0.5569* −0.7517*** (0.2394) (0.2394) (0.2397) (0.2859) (0.3471) (0.3015) (0.2872) CASH 1.2149*** 1.2113*** 1.2237*** 1.1020*** 0.5549 1.2654*** 1.4617*** (0.3303) (0.3306) (0.3296) (0.3977) (0.5093) (0.4397) (0.4133) TANG 0.4686*** 0.4704*** 0.4658*** 0.6472*** 0.9909*** 0.3252** 0.1044 (0.1289) (0.1292) (0.1287) (0.1514) (0.1845) (0.1599) (0.1658) INVEST −2.4309** −2.4855** −2.4174** −3.3655*** −4.1967*** −2.6422** −0.5045 (1.0470) (1.0512) (1.0447) (1.2830) (1.5545) (1.3105) (1.3472) BOARDSIZE 0.4489*** 0.4504*** 0.4364*** 0.6948*** 0.7702*** 0.6330*** −0.0807 (0.1331) (0.1332) (0.1325) (0.1491) (0.1769) (0.1598) (0.1692) BOARDINDEP 0.6683*** 0.6642*** 0.6419*** 0.3108 0.2422 0.4124* 1.3044*** (0.1712) (0.1714) (0.1698) (0.2011) (0.2358) (0.2180) (0.2188) CEOduality −0.2402*** −0.2376*** −0.2246*** 0.0543 0.0534 0.0630 −0.7812*** (0.0837) (0.0835) (0.0838) (0.0984) (0.1134) (0.1085) (0.1144) INSOWN 0.3240 0.3224 0.2928 0.0200 0.2207 −0.1352 0.8412*** (0.2231) (0.2231) (0.2226) (0.2580) (0.2947) (0.2912) (0.2926) Country fixed effects Yes Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes Yes No. of obs. 7982 7982 7982 7982 7982 7982 7982 Pseudo R2 0.1564 0.1564 0.1571 0.1538 0.1348 0.1178 0.0585 Estimated extreme point – – 53.5200 57.0646 56.4948 57.7688 – This table summarizes the Tobit estimation results of Eq. (1). The dependent variable is a firm's ESG engagement, which is measured by the ESGscore (the average of the scores of the environmental, social, and governance pillars) and ESscore (the average of the scores of the environmental and social pillars). Additional robustness analyses are shown for individual pillars by using ENV (environmental pillar score), SOC (social pillar score) and GOV (governance pillar score). CEO career horizon measure is CEOage (a firm's CEO age in years). Control variables are: ROA (a firm's financial performance as return on assets), SIZE (a firm's size as the natural logarithm of the book value of total assets), LEV (leverage as the ratio of total debt to the book value of assets), CASH (cash reserves as the ratio of cash to the book value of assets), TANG (asset tangibility as the ratio of property, plant and equipment to the book value of assets), INVEST (investment opportunities as the ratio of CAPEX to the book value of assets), BOARDSIZE (the natural logarithm of the number of board directors), BOARDINDEP (the ratio of the number of independent directors to total board directors), CEOduality (a dummy variable equal to one if the CEO is also the board chair, zero otherwise), and INSTOWN (the percentage of shares held by institutional investors). All regressions control for country, year and industry fixed effects. At the bottom of regressions considering the non-linear effect of CEOage, the estimated extreme point of the curvilinear relationship is included. Standard errors are shown in parentheses under coefficients. ***, ** and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. 6 As proposed by Haans et al. (2016), we apply the nlcom command in STATA to conduct this test. 7 We thank one of the anonymous reviewers for this suggestion. D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 8 null hypothesis that the turning point in each subsample is equal to that of the full sample of the baseline model. Furthermore, the turning points of the two subsamples exhibit no statistically significant difference between them. Again, this suggests there is no turning point shift in the two subsamples. Finally, Fig. 3 graphically plots the estimated results of our baseline model (Eq. (1)). We estimate this model for the subsamples of the top 25th percentile industry performers and bottom 25th percentile industry performers separately. Although the curve for outperforming firms is slightly above that of the underperformers, the turning point is not seen to change significantly in the X-axis. The lack of support for the moderating effect of performance framing could be explained by the intangible outcomes (e.g. legitimacy, trust) to emerge from ESG engagement. These outcomes do not occur in other risk-mitigating strategies such as corporate diversification. Such extrabenefits of ESG might not only lead to CEOs' risk appetite influencing a firm's ESG engagement but could also enhance their reputation, which in turn might translate into CEOs being held in greater esteem inside the company and thus being less likely to face dismissal (Shin, Lee, & Bansal, 2022), superior firm performance and access to wider future investment opportunities (Fuente et al., 2022). A firm's CEO reputation and legacy imprint are closely related to those outcomes (Kang, 2016), which could drive top managers' reluctance to allow their ESG engagement behaviour to be affected by performance targets. This explanation agrees with earlier literature suggesting the importance of ESG consistency being perceived as genuine by stakeholders (Wang & Choi, 2013). 4.4. Additional robustness analyses We perform an additional battery of robustness analyses. First, we use an alternative proxy for CEO career horizon: CEO life horizon, calculated as the yearly life expectancy 8 in a firm's country minus CEO age. This is a direct proxy for CEO time-decision horizon, since we are taking their vital horizon as a threshold of the final point. Table A.2. of the Appendix reports these robustness checks. Overall, results remain similar to those previously described. Second, we omit the governance pillar from ESG and repeat our analyses taking ESaggregated as the dependent variable. Alternatively, we also estimate our models by constructing an industry-adjusted ESG measure, following Hubbard, Christensen, and Graffin (2017). Third, we repeat the regressions to test Hypothesis 2 using one lag for the reference point measure of a firm's relative performance, since the latter's influence may take some time to influence a firm's decision-making. In addition, since CEO turnover might affect our findings, we conduct robustness analyses by dropping firm-year observations that have a CEO tenure below the fiveand ten-year thresholds. We repeat the analyses without excluding firms with negative common equity. All of these explained checks provide robust results. 9 Table 5 CEO career horizon (CEO age) and ESG engagement under different decisionframing scenarios: Interaction effects. Dependent variable: ESGscore (1) (2) (3) Constant −9.6467*** −9.6434*** −9.6465*** (1.1351) (1.1377) (1.1353) CEOage 0.1063*** 0.1034** 0.0994** (0.0402) (0.0401) (0.0401) CEOage_sq −0.0010*** −0.0010*** −0.0009** (0.0004) (0.0004) (0.0004) Interaction effects CEOage ×dumROA_indFF −0.0037 (0.0065) CEOage_sq ×dumROA_indFF 0.0001 (0.0001) CEOage ×dumROA_ind2d 0.0013 (0.0064) CEOage_sq ×dumROA_ind2d −0.0000 (0.0001) CEOage ×dumROA_indDIV 0.0064 (0.0066) CEOage_sq ×dumROA_indDIV −0.0001 (0.0001) Control variables ROA 0.6374* 0.7065** 0.2525 (0.3563) (0.3561) (0.3603) SIZE 0.7935*** 0.7930*** 0.7954*** (0.0305) (0.0305) (0.0305) LEV −0.6965*** −0.6981*** −0.6903*** (0.2399) (0.2397) (0.2392) CASH 1.2141*** 1.2202*** 1.2053*** (0.3289) (0.3296) (0.3274) TANG 0.4638*** 0.4649*** 0.4704*** (0.1287) (0.1289) (0.1289) INVEST −2.4367** −2.4195** −2.4661** (1.0449) (1.0449) (1.0472) BOARDSIZE 0.4380*** 0.4373*** 0.4419*** (0.1323) (0.1326) (0.1324) BOARDINDEP 0.6394*** 0.6416*** 0.6403*** (0.1697) (0.1698) (0.1697) CEOduality −0.2257*** −0.2253*** −0.2284*** (0.0837) (0.0837) (0.0838) INSOWN 0.2958 0.2950 0.3123 (0.2231) (0.2229) (0.2235) ∑linear effect 0.1026*** 0.1047*** 0.1058*** ∑nonlinear effect −0.0009*** −0.0009*** −0.0010*** Country fixed effects Yes Yes Yes Year fixed effects Yes Yes Yes Industry fixed effects Yes Yes Yes No. of obs. 7982 7982 7982 Pseudo R2 0.1528 0.1580 0.1577 ∂ CEOage / ∂ dumROA_ind if dumROA_ind =0 [p-value] 2.3815 [0.332] −0.2493 [0.934] −0.2800 [0.932] ∂ CEOage / ∂ dumROA_ind if dumROA_ind =1 [p-value] 2.8151 [0.427] −0.2407 [0.933] −0.2448 [0.930] This table shows the Tobit estimation results of Eq. (2). The dependent variable is a firm's ESG engagement, which is measured by the ESGscore (the average of the scores of the environmental, social, and governance pillars). CEO career horizon measure is CEOage (a firm's CEO age in years). A firm's performance relative to its industry peers is captured by three alternative proxies based on different industry classification schemes: dumROA_indFF (a dummy equal to one if a firm's ROA is above its 48-Fama-French industry median, and zero otherwise), dumROA_ind2d (a dummy equal to one if a firm's ROA is above its 2-digit SIC industry median, and zero otherwise), and dumROA_indFF (a dummy equal to one if a firm's ROA is above its U.S. division industry median, and zero otherwise). Control variables are: ROA (a firm's financial performance as return on assets), SIZE (a firm's size as the natural logarithm of the book value of total assets), LEV (leverage as the ratio of total debt to the book value of assets), CASH (cash reserves as the ratio of cash to the book value of assets), TANG (asset tangibility as the ratio of property, plant and equipment to the book value of assets), INVEST (investment opportunities as the ratio of CAPEX to the book value of assets), BOARDSIZE (the natural logarithm of the number of board directors), BOARDINDEP (the ratio of the number of independent directors to total board directors), CEOduality (a dummy variable equal to one if the CEO is also the board chair, zero otherwise), and INSTOWN (the percentage of shares held by institutional investors). All regressions control for country, year and industry fixed effects.∑linear effect tests the joint significance of the CEOage linear variable plus the interaction effect on the industry relative performance dummy. ∑non-linear effect tests the joint significance of the CEOage squared variable plus the interaction effect on the industry relative performance dummy. At the bottom of the table, we test the null hypothesis that the derivative of the turning point with respect to the moderator is equal to zero. Standard errors are shown in parentheses under coefficients. ***, ** and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. 8 This data was obtained from the World Bank database. 9 These results are available upon request. D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 9 Moreover, in order to alleviate potential omitted variable bias, we control for additional CEO characteristics which may also play a part in a firm's ESG engagement. Specifically, we add a binary variable which controls for whether a firm's CEO has financial expertise or not. We also account for CEOs' educational qualifications. We build a categorical variable which equals 1 to 3 based on whether the CEO holds a bachelor's degree, a master's degree, or a PhD, respectively. Table A.3. of the Appendix contains these robustness analyses. Column (1) corresponds with the regression to test Hypothesis 1, and columns (2) to (4) summarize the results of Hypothesis 2 based on different industry classification schemes to capture a firm's relative performance. Our evidence still holds after controlling for these additional CEO personal traits. 10 Following recent research (Bernile, Bhagwat, & Rau, 2017; Chen & Xie, 2022; Lee, Trzcinka, & Venkatesan, 2019; Shu, Tan, & Wei, 2023; Wang & Qiu, 2023), we conduct a placebo test to examine whether our empirical findings are likely to have been obtained by a random factor or Table 6 CEO age and ESG engagement under different decision-framing scenarios: robustness analyses by subsamples. Dependent variable: ESGscore Above-median industry performance (dumROA_indFF = 1) Below-median industry performance (dumROA_indFF = 0) Above-median industry performance (dumROA_ind2d = 1) Below-median industry performance (dumROA_ind2d = 0) Above-median industry performance (dumROA_indDIV = 1) Below-median industry performance (dumROA_indDIV = 0) (1) (2) (3) (4) (5) (6) Constant −10.2567*** −9.2660*** −10.1924*** −9.5550*** −10.4159*** −8.8376*** (1.3871) (1.4837) (1.3184) (1.5670) (1.3445) (1.4751) CEOage 0.1031** 0.1046** 0.0999** 0.1178** 0.1092** 0.0916* (0.0476) (0.0528) (0.0451) (0.0559) (0.0463) (0.0527) CEOage_sq −0.0010** −0.0010** −0.0009** −0.0011** −0.0010** −0.0008* (0.0004) (0.0005) (0.0004) (0.0005) (0.0004) (0.0005) Control variables ROA 1.9195*** −0.1634 2.0771*** −0.1221 1.5523** −0.5023 (0.6604) (0.4853) (0.6559) (0.4949) (0.6645) (0.5088) SIZE 0.7883*** 0.8163*** 0.7931*** 0.7999*** 0.7956*** 0.8040*** (0.0385) (0.0375) (0.0377) (0.0380) (0.0390) (0.0373) LEV −0.7505** −0.5985** −0.5718* −0.8218*** −0.6288** −0.7353** (0.3019) (0.3042) (0.2960) (0.3060) (0.3013) (0.3067) CASH 0.8220* 1.2230*** 0.9541** 1.1205*** 0.9865** 1.1026*** (0.4357) (0.4207) (0.4305) (0.4324) (0.4351) (0.4260) TANG −2.8195** −2.5149* −3.2074** −1.9482 −3.4748*** −1.5975 (1.2751) (1.3398) (1.2460) (1.3337) (1.2964) (1.3099) INVEST 0.5489*** 0.4089** 0.5505*** 0.4148** 0.6152*** 0.3645** (0.1559) (0.1589) (0.1531) (0.1615) (0.1589) (0.1538) BOARDSIZE 0.6757*** 0.1674 0.6454*** 0.2348 0.7507*** 0.0752 (0.1649) (0.1588) (0.1632) (0.1562) (0.1671) (0.1573) BOARDINDEP −0.1412 −0.3715*** −0.1624* −0.3543*** −0.1188 −0.4313*** (0.0988) (0.1065) (0.0970) (0.1096) (0.0988) (0.1098) CEOduality 0.7241*** 0.5378*** 0.7459*** 0.5172** 0.6108*** 0.7010*** (0.2145) (0.2077) (0.2082) (0.2177) (0.2084) (0.2148) INSOWN 0.4761* 0.1348 0.4758* 0.1309 0.5235* 0.0353 (0.2845) (0.2536) (0.2764) (0.2642) (0.2865) (0.2649) Country fixed effects Yes Yes Yes Yes Yes Yes Year fixed effects Yes Yes Yes Yes Yes Yes Industry fixed effects Yes Yes Yes Yes Yes Yes No. of obs. 4672 3310 4689 3293 4701 3281 Pseudo R2 0.1562 0.1677 0.1586 0.1645 0.1610 0.1648 Estimated extreme point 53.9908 52.4708 52.7556 54.3532 52.3318 55.2543 Equality of extreme value point between each subsample and full sample (p-value) 0.8804 0.7495 0.8149 0.7624 0.7023 0.6250 Equality of extreme value point between subsamples (p-value) 0.6270 0.6436 0.6247 0.5621 0.3471 0.4101 This table shows the Tobit estimation results of Eq. (1) by subsamples based on a firm's relative performance relative to its industry benchmark (proxied by dumROA_indFF, dumROA_ind2d, and dumROA_indFF depending on the industry classification scheme applied). The dependent variable is a firm's ESG engagement, which is measured by the ESGscore (the average of the scores of the environmental, social, and governance pillars). Control variables are: ROA (a firm's financial performance as return on assets), SIZE (a firm's size as the natural logarithm of the book value of total assets), LEV (leverage as the ratio of total debt to the book value of assets), CASH (cash reserves as the ratio of cash to the book value of assets), TANG (asset tangibility as the ratio of property, plant and equipment to the book value of assets), INVEST (investment opportunities as the ratio of CAPEX to the book value of assets), BOARDSIZE (the natural logarithm of the number of board directors), BOARDINDEP (the ratio of the number of independent directors to total board directors), CEOduality (a dummy variable equal to one if the CEO is also the board chair, zero otherwise), and INSTOWN (the percentage of shares held by institutional investors). All regressions control for country, year and industry fixed effects. At the bottom of the table, the estimated extreme point of the curvilinear relationship is reported, in addition to the p-values of the equality test of extreme points between each subsample and the full sample, and also the p-values of the equality test of extreme points between subsamples. Standard errors are shown in parentheses under coefficients. ***, ** and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. 10 We thank an anonymous reviewer for this suggestion. D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 16 (leverage as the ratio of total debt to the book value of assets), CASH (cash reserves as the ratio of cash to the book value of assets), TANG (asset tangibility as the ratio of property, plant and equipment to the book value of assets), INVEST (investment opportunities as the ratio of CAPEX to the book value of assets), BOARDSIZE (the natural logarithm of the number of board directors), BOARDINDEP (the ratio of the number of independent directors to total board directors), CEOduality (a dummy variable equal to one if the CEO is also the board chair, zero otherwise), INSTOWN (the percentage of shares held by institutional investors), FINEXPERTISE (a dummy variable equal to one if the CEO has financial expertise, and zero otherwise), and EDUCATION (a categorical variable which equals one if the CEO holds a bachelor's degree, two if they hold a master's degree, three if they hold a PhD, and zero if the CEO has no degree). All regressions control for country, year and industry fixed effects. ∑linear effect tests the joint significance of the CEOage linear variable plus the interaction effect on the industry relative performance dummy. ∑non-linear effect tests the joint significance of the CEOage squared variable plus the interaction effect on the industry relative performance dummy. At the bottom of the table, we test the null hypothesis that the derivative of the turning point with respect to the moderator is equal to zero. Standard errors are shown in parentheses under coefficients. ***, ** and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. References Amore, M. D., & Murtinu, S. (2021). Tobit models in strategy research: Critical issues and applications. Global Strategy Journal, 11(3), 331–355. https://doi.org/10.1002/ gsj.1363 Antia, M., Pantzalis, C., & Park, J. (2010). CEO decision horizon and firm performance: An empirical investigation. Journal of Corporate Finance, 16(3), 288–301. https:// doi.org/10.1016/j.jcorpfin.2010.01.005 Arena, C., Michelon, G., & Trojanowski, G. (2018). Big egos can be green: A study of CEO hubris and environmental innovation. British Journal of Management, 29(2), 316–336. https://doi.org/10.1111/1467-8551.12250 Attig, N., El Ghoul, S., Guedham, O., & Zheng, X. (2021). Dividends and economic policy uncertainty: International evidence. Journal of Corporate Finance, 66, Article 101785. https://doi.org/10.1016/j.jcorpfin.2020.101785 Bannier, C., Bofinger, Y., & Rock, B. (2022). Corporate social responsibility and credit risk. Finance Research Letters, 44, Article 102052. https://doi.org/10.1016/j. frl.2021.102052 Bernile, G., Bhagwat, V., & Rau, P. R. (2017). What doesn’t kill you will only make you more risk-loving: Early-life disasters and CEO behavior. The Journal of Finance, 72 (1), 167–206. https://doi.org/10.1111/jofi.12432 Bilgili, H., Campbell, J. T., O’Leary-Kelly, A., Ellstrand, A., & Johnson, J. (2020). The final countdown: Regulatory focus and the phases of CEO retirement. Academy of Management Review, 45(1), 58–84. https://doi.org/10.5465/amr.2016.0455 Borghesi, R., Houston, J. F., & Naranjo, A. (2014). Corporate socially responsible investments: CEO altruism, reputation, and shareholder interests. Journal of Corporate Finance, 26, 164–181. https://doi.org/10.1016/j.jcorpfin.2014.03.008 Cai, X., Gao, N., Garrett, I., & Xu, Y. (2020). Are CEOs judged on their companies’ social reputation? Journal of Corporate Finance, 64, Article 101621. https://doi.org/ 10.1016/j.jcorpfin.2020.101621 Chen, Z., & Xie, G. (2022). ESG disclosure and financial performance: Moderating role of ESG investors. International Review of Financial Analysis, 83, Article 102291. https:// doi.org/10.1016/j.irfa.2022.102291 Cheng, B., Ioannou, I., & Serafeim, G. (2014). Corporate social responsibility and access to finance. Strategic Management Journal, 35(1), 1–23. https://doi.org/10.1002/ smj.2131 Chintrakarn, P., Jiraporn, P., & Treepongkaruna, S. (2021). How do independent directors view corporate social responsibility (CSR) during a stressful time? Evidence from the financial crisis. International Review of Economics and Finance, 71, 143–160. https://doi.org/10.1016/j.iref.2020.08.007 Cho, S. Y., & Kim, S. K. (2017). Horizon problem and firm innovation: The influence of CEO career horizon, exploitation and exploration on breakthrough innovations. Research Policy, 46(10), 1801–1809. https://doi.org/10.1016/j.respol.2017.08.007 Cline, B. N., & Yore, A. S. (2016). Silverback CEOs: Age, experience, and firm value. Journal of Empirical Finance, 35, 169–188. https://doi.org/10.1016/j. jempfin.2015.11.002 Fama, E., & French, K. (1997). Industry costs of equity. Journal of Financial Economics, 43 (2), 153–193. https://doi.org/10.1016/S0304-405X(96)00896-3 Fiegenbaum, A., & Thomas, H. (1988). Attitudes toward risk and the Risk–Return paradox: Prospect theory explanations. Academy of Management Journal, 31(1), 85–106. https://doi.org/10.5465/256499 Fu, R., Tang, Y., & Chen, G. (2020). Chief sustainability officers and corporate social (Ir) responsibility. Strategic Management Journal, 41(4), 656–680. https://doi.org/ 10.1002/smj.3113 Fuente, G., Ortiz, M., & Velasco, P. (2022). The value of a firm’s engagement in ESG practices: Are we looking at the right side? Long Range Planning, 55(4), Article 102143. https://doi.org/10.1016/j.lrp.2021.102143 Gao, Z., Hua, M., Liu, H., & Yan, C. (2023). News sentiment and CEO retirement: The impact on firm performance and risk. Research in International Business and Finance, 66, Article 102031. https://doi.org/10.1016/j.ribaf.2023.102031 Gibbons, R., & Murphy, K. (1992). Optimal incentive contracts in the presence of career concerns: Theory and evidence. Journal of Political Economy, 100(3), 468–505. https://doi.org/10.1086/261826 Godfrey, P. C. (2005). The relationship between corporate philanthropy and shareholder wealth: A risk management perspective. Academy of Management Review, 30(4), 777–798. https://doi.org/10.2307/20159168 Godfrey, P. C., Merrill, C. B., & Hansen, J. M. (2009). The relationship between corporate social responsibility and shareholder value: An empirical test of the risk management hypothesis. Strategic Management Journal, 30(4), 425–445. https://doi.org/10.1002/ smj.750 Gomes, M. (2019). Does CSR influence M&a target choices? Finance Research Letters, 30, 153–159. https://doi.org/10.1016/j.frl.2018.09.011 Greene, W. (2004). Fixed effects and bias due to the incidental parameters problem in the Tobit model. Econometric Reviews, 23(2), 125–147. https://doi.org/10.1081/ETC120039606 Haans, R., Pieters, C., & He, Z. (2016). Thinking about U: Theorizing and testing Uand inverted U-shaped relationships in strategy research. Strategic Management Journal, 37(7), 1177–1195. https://doi.org/10.1002/smj.2399 Hambrick, D. C. (2007). Upper echelons theory: An update. Academy of Management Review, 32(2), 334–343. https://doi.org/10.2307/20159303 Hambrick, D. C., & Mason, P. A. (1984). Upper echelons: The organization as a reflection of its top managers. Academy of Management Review, 9(2), 193–206. https://doi.org/ 10.2307/258434 He, F., Ding, C., Yue, W., & Liu, G. (2023). ESG performance and corporate risk-taking: Evidence from China. International Review of Financial Analysis, 87, Article 102550. https://doi.org/10.1016/j.irfa.2023.102550 Hirshleifer, D. (1993). Managerial reputation and corporate investment decisions. Financial Management, 22(2), 145–160. https://doi.org/10.2307/3665866 Holmes, R. M., Bromiley, P., Devers, C. E., Holcomb, T. R., & McGuire, J. B. (2011). Management theory applications of prospect theory: Accomplishments, challenges, and opportunities. Journal of Management, 37(4), 1069–1107. https://doi.org/ 10.1177/0149206310394863 Hubbard, T. D., Christensen, D. M., & Graffin, S. D. (2017). Higher highs and lower lows: The role of corporate social responsibility in CEO dismissal. Strategic Management Journal, 38(11), 2255–2265. https://doi.org/10.1002/smj.2646 Hutzschenreuter, T., Kleindienst, I., Groene, F., & Verbeke, A. (2014). Corporate strategic responses to foreign entry: Insights from prospect theory. The Multinational Business Review, 22(3), 294–323. https://doi.org/10.1108/MBR-06-2014-0026 Jia, J., & Li, Z. (2020). Does external uncertainty matter in corporate sustainability performance? Journal of Corporate Finance, 65, Article 101743. https://doi.org/ 10.1016/j.jcorpfin.2020.101743 Jia, Y., Gao, X., & Julian, S. (2020). Do firms use corporate social responsibility to insure against stock price risk? Evidence from a natural experiment. Strategic Management Journal, 41(2), 290–307. https://doi.org/10.1002/smj.3107 Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–292. https://doi.org/10.2307/1914185 Kang, J. (2016). Labor market evaluation versus legacy conservation: What factors determine retiring CEOs’ decisions about long-term investment? Strategic Management Journal, 37(2), 389–405. https://doi.org/10.1002/smj.2234 Koh, P. S., Qian, C., & Wang, H. (2014). Firm litigation risk and the insurance value of corporate social performance. Strategic Management Journal, 35(10), 1464–1482. https://doi.org/10.1002/smj.2171 Krause, R., & Semadeni, M. (2014). Last dance or second dance? Firm performance, CEO career horizon, and the separation of board leadership roles. Strategic Management Journal, 35(6), 808–825. https://doi.org/10.1002/smj.2139 Lee, J. H., Trzcinka, C., & Venkatesan, S. (2019). Do portfolio manager contracts contract portfolio management? The Journal of Finance, 74(5), 2543–2577. https://doi.org/ 10.1111/jofi.12823 Lins, K., Servaes, H., & Tamayo, A. (2017). Social capital, trust, and firm performance: The value of corporate social responsibility during the financial crisis. The Journal of Finance, 31(2), 1785–1824. https://doi.org/10.1111/jofi.12505 Matta, E., & Beamish, P. W. (2008). The accentuated CEO career horizon problem: Evidence from international acquisitions. Strategic Management Journal, 29(7), 683–700. https://doi.org/10.1002/smj.680 McCarthy, S., Oliver, B., & Song, S. (2017). Corporate social responsibility and CEO confidence. Journal of Banking & Finance, 75, 280–291. https://doi.org/10.1016/j. jbankfin.2016.11.024 McClelland, P., Barker, V., III, & Oh, W. (2012). CEO career horizon and tenure: Future performance implications under different contingencies. Journal of Business Research, 65(9), 1387–1393. https://doi.org/10.1016/j.jbusres.2011.09.003 Nadeem, M., Zaman, R., Suleman, T., & Atawnah, N. (2021). CEO ability, career concerns, firms’ lifecycle and investments in intellectual capital. International Review of Economics and Finance, 75, 237–251. https://doi.org/10.1016/j.iref.2021.04.023 Oh, W. Y., Cha, J., & Chang, Y. K. (2017). Does ownership structure matter? The effects of insider and institutional ownership on corporate social responsibility. Journal of Business Ethics, 146(1), 111–124. https://doi.org/10.1007/s10551-015-2914-8 O’Sullivan, D., Zolotoy, L., & Fan, O. (2021). CEO early-life disaster experience and corporate social performance. Strategic Management Journal, 42(11), 2137–2161. https://doi.org/10.1002/smj.3293 Peltom¨ aki, J., Sihvonen, J., Swidler, S., & V¨ ah¨ amaa, S. (2021). Age, gender, and risktaking: Evidence from the S&P 1500 executives and market-based measures of firm risk. Journal of Business Finance & Accounting, 48(9–10), 1988–2014. https://doi.org/ 10.1111/jbfa.12528 D. Cabreros et al. International Review of Financial Analysis 93 (2024) 103200 17 Petersen, M. (2009). Estimating standard errors in finance panel data sets: Comparing approaches. The Review of Financial Studies, 22(1), 435–480. https://doi.org/ 10.1093/rfs/hhn053 Quigley, T. J., & Hambrick, D. C. (2015). Has the ‘CEO effect’ increased in recent decades? A new explanation for the great rise in America’s attention to corporate leaders. Strategic Management Journal, 36(6), 821–830. https://doi.org/10.1002/ smj.2258 Sajko, M., Boone, C., & Buyl, T. (2021). CEO greed, corporate social responsibility, and organizational resilience to systemic shocks. Journal of Management, 47(4), 957–992. https://doi.org/10.1177/0149206320902528 Sargan, J. D. (1958). The estimation of economic relationships using instrumental variables. Econometrica, 26(3), 393–415. https://doi.org/10.2307/1907619 Serfling, M. A. (2014). CEO age and the riskiness of corporate policies. Journal of Corporate Finance, 25, 251–273. https://doi.org/10.1016/j.jcorpfin.2013.12.013 Shin, S., Lee, J., & Bansal, P. (2022). From a shareholder to stakeholder orientation: Evidence from the analyses of CEO dismissal in large U.S. firms. Strategic Management Journal, 43(7), 1233–1257. https://doi.org/10.1002/smj.3369 Shiu, Y., & Yang, S. (2017). Does engagement in corporate social responsibility provide strategic insurance-like effects? Strategic Management Journal, 38(2), 455–470. https://doi.org/10.1002/smj.2494 Shu, H., Tan, W., & Wei, P. (2023). Carbon policy risk and corporate capital structure decision. International Review of Financial Analysis, 86, Article 102523. https://doi. org/10.1016/j.irfa.2023.102523 Wang, G., Holmes, M., Jr., Oh, I., & Zhu, W. (2016). Do CEOs matter to firm strategic actions and firm performance? A meta-analytic investigation based on upper echelons theory. Personnel Psychology, 69(4), 775–862. https://doi.org/10.1111/ peps.12140 Wang, H., Barney, J., & Reuer, J. (2003). Stimulating firm-specific investment through risk management. Long Range Planning, 36(1), 49–59. https://doi.org/10.1016/ S0024-6301(02)00203-0 Wang, H., & Choi, J. (2013). A new look at the corporate social-financial performance relationship: The moderating roles of temporal and interdomain consistency in corporate social performance. Journal of Management, 39(2), 416–441. https://doi. org/10.1177/0149206310375850 Wang, Q., & Qiu, M. (2023). Minority shareholders’ activism and stock price crash risk: Evidence from China. International Review of Financial Analysis, 87, Article 102594. https://doi.org/10.1111/peps.12140 Wooldridge, J. M. (2019). Introductory econometrics: A modern approach (7th ed.). Cengage Learning. Zhao, X., Fang, L., & Zhang, K. (2023). ‘Online search attention, firms’ ESG and operating performance. International Review of Economics and Finance, 88, 223–236. D. Cabreros et al.