Introduction

Although corporate social responsibility (CSR) has emerged in recent decades as a central theme in the management literature (Caroll 1979; Campbell 2007), studies that explicitly integrate family business and CSR are more recent (Van Gils et al. 2014). While the behavioral agency model (Gomez-Mejia et al. 2007, 2019) has become one of the central theoretical foundation used to explain why family firms behave differently from nonfamily firms, empirical results on this issue are mixed, and no clear evidence has yet been found to indicate whether family firms are systematically more socially responsible (Bingham et al. 2011; Craig and Dibrell 2006; Dyer and Whetten 2006) or less socially responsible (Campopiano and De Massis 2015; Rees and Rodionova 2015) than nonfamily firms.

Moreover, most of these studies conceptualize family firms as a homogeneous group although some recent studies suggest that family firm heterogeneity in leadership accounts for the observed CSR performance (CSP) variance within family firms (Campopiano et al. 2014; Marques et al. 2014; Zattoni et al. 2015; Labelle et al. 2018). Moreover, few studies systematically investigate the full range of internal/external stakeholder-orientated CSR practices (Block and Wagner 2014a, b), while some authors (Cruz et al. 2014) suggest that family firms could be simultaneously more responsible toward external stakeholders and less responsible toward internal stakeholders than nonfamily firms. At last, studies following the behavioral agency model are based on a homogeneous and stable reference point, which is mainly used as a nonobservable variable to account for CSR differences between family and nonfamily firms (Nason et al. 2018). Recent studies challenge the conventional assumption of a static and stable reference point in family firms (Aranda et al. 2017), meaning that family firms attitudes toward CSR may change over time.

Following Nason et al. (2018), we expect that, as CEO succession concerns become more prominent, family firms shift their reference point slightly, thus moderating how they prioritize internal and external stakeholder demands and the associated CSR outcomes. In line with Strike et al. (2015), we argue that these changes due to transgenerational control and/or CEO succession priorities are associated with the incumbent CEO’s age. Our research question is as follows: To what extent does the type of CEOs in family firms influence external and internal stakeholder-related CSP as compared to nonfamily firms, and to what extent does the incumbent CEO’s age moderate these relationships?

We develop three categories of hypotheses related to the influence of family firm CEOs (postfounder family CEO and nonfamily CEO) on CSP and to the moderating role of the CEOs’ age. To test our hypotheses, we use CSR scores from Vigeo, Europe’s leading social rating agency. Vigeo provides us with three scores related to external stakeholder-related CSP (i.e., environment, business behavior, community involvement) and two scores related to internal stakeholder-related CSP (i.e., human resources, human rights).

Our results suggest that the type of family firm CEO is a significant factor in explaining variances in CSP as compared to nonfamily firms. First, firms managed by family CEOs appear to be positively associated with both external and internal stakeholder-related CSP. Second, postfounder firms with nonfamily CEOs seem less likely to invest in external and internal stakeholder-related CSR activities. Finally, we provide evidence that the CEO’s age moderates all the above relationships. This study contributes to the family business and CSR literature in three ways. First, we conclude that CSP depends more on the type of the leadership involved within family firms than on the duality family versus nonfamily firms or than on the dichotomy internal versus external stakeholder-related CSP. Second, we provide evidence that the CEO’s age moderates the relationships between the type of leadership and family firm CSR outcomes, pointing out the existence of shifting family priorities and a tendency toward CSR conformity as succession issues become more salient. Third, we contribute to a stream of research that highlights CEOs’ personal attributes as potential determinants of family CSP (Fabrizi et al. 2014; Oh et al. 2016).

The remainder of this paper proceeds as follows. The first section builds on prior research that links family firms and CSR outcomes to develop our hypotheses. The second section describes our data and statistical methods. The third section presents our results. Finally, the last section discusses the results, describes the study’s limitations, and outlines its implications for future research.

Literature and Hypothesis Development

Several recent studies link family businesses and CSR (Dyer and Whetten 2006; Gallo 2004). Most of these studies are based on specific definitions of “family firm” and “CSR.” Van Gils et al. (2014) claim that “of the varying conceptualizations of family and the family’s impact on the firm in relation to socially related inquiries, family involvement seems to be prevalent.” We follow this approach by defining a family firm as a postfounder firm with family ownership and family involvement in management and/or firm governance.

While many definitions and measurements of CSR have been proposed in the literature (Wood 2010), most management studies describe CSR as policies and/or activities developed by firms to improve their social, economic, and environmental impacts beyond legal requirements. We define CSR as a multidimensional construct that incorporates several internal and external dimensions of an enlarged CSR. This approach has been popularized through the creation of social agencies such as KLD in the US and Vigeo in Europe. Recent studies have used the social ratings of KLD (Barnett and Salomon 2012) and Vigeo (Girerd-Potin et al. 2014) to measure CSP. Following Cruz et al. (2014), we assess external stakeholder-related CSP by grouping the environment, business behavior, and community dimensions of CSR. Similarly, we assess internal stakeholder-related CSP by grouping the human resources management and respect for human rights dimensions of CSR.

Family Firms and CSR

Although researchers provide evidence that family firms behave differently from their nonfamily counterparts with respect to CSR, the literature on family firms and CSR offers contradictory arguments concerning whether family firms are more (Porter and Kramer 2002; Fombrun and Shanley 1990; Berrone et al. 2010; Sharma and Sharma 2011; Bingham et al. 2011; Gallo 2004; Dyer and Whetten 2006) or less responsible (Morck and Yeung 2003; Rees and Rodionova 2015; Cespa and Cestone 2007; Cruz et al. 2014). More generally, according to the proponents of the SEW (Socioemotional Wealth) approach (Gomez-Mejia et al. 2007), family firm CSR decisions (positive or negative) are driven by the nonfinancial benefits that family owners can derive from them (Berrone et al. 2010; Cennamo et al. 2012).

Recent studies underline three major limitations of past studies linking family priorities and CSR: family firm heterogeneity may lead to differentiated priorities and differentiated CSR outcomes (Campopiano et al. 2014; Dou et al. 2014; Labelle et al. 2018; Le Breton-Miller and Miller 2013; Marques et al. 2014; Van Gils et al. 2014); multiple and conflicting external and internal stakeholder demands may lead to differentiated external or internal stakeholder-related CSP levels (Block and Wagner 2014b; Campbell 2007; Cruz et al. 2014; Mitchell et al. 2011); and shifting priorities may lead to differentiated CSP levels across time (Aranda et al. 2017; Miller and Breton-Miller 2014; Nason et al. 2018; Strike et al. 2015). However, no study explicitly integrates these three limitations into one single framework. This research intends to provide some theoretical arguments and empirical material to fill this gap. More precisely, this study explores a specific source of family firm heterogeneity depending on which type of CEO is leading the firm (i.e., family CEO, nonfamily CEO). Building on the work of Le Breton-Miller and Miller (2013), we expect that each type of CEO will engender both differentiated priorities and differentiated external and internal stakeholder-related CSP levels. Furthermore, following Nason et al. (2018) and Strike et al. (2015), we argue that associated with CEOs’ age, as transgenerational and/or CEO succession becomes more salient, family firms shift their reference point slightly, which moderates how they prioritize external and internal stakeholder-related CSP.

Family Firms Managed by Family CEOs

Firms that feature high family involvement are more inclined to pursue the family agenda before the business agenda, given that noneconomic profits balance economic costs and induce risks (Berrone et al. 2012). Family firms managed by family CEOs are likely to put the emphasis on specific family priorities associated primarily with family members’ identification with the firm, binding social ties, and intra-group norms (Nason et al. 2018).

External Stakeholder-Oriented CSP

The strong development of a sense of belonging centered around the family business leads to noneconomic family priorities based on the family’s core values, strong social ties, emotional attachments (Cennamo et al. 2012), family influence, close identification with the firm (Berrone et al. 2012), and a desire to protect family assets (Dyer and Whetten 2006). Family members’ strong identification with the family firm leads to stronger sensitivity regarding the firm’s external image (Micelotta and Raynard 2011; Gedajlovic et al. 2012). As binding social ties engender strong social ties with the community at large (Brickson 2007), these priorities may then lead to a stronger desire to appear legitimate to an enlarged set of stakeholders (Cennamo et al. 2012). Postfounder firms with high family involvement display a strong family desire to preserve their external reputation (Sharma and Manikuti 2005) and public image (Berrone et al. 2012; Craig and Dibrell 2006). The need for the family to improve its legitimacy (Cennamo et al. 2012) and gain community visibility and status (Miller et al. 2013) may also lead to higher CSR investments in activities related to external stakeholders such as environmental issues, community involvement, and responsible business behaviors (Berrone et al. 2010; Craig and Dibrell 2006). We therefore propose the following:

Hypothesis 1a

Compared to nonfamily firms, family firms managed by family CEOs are positively associated with external stakeholder-related CSP.

Internal Stakeholder-Oriented CSP

Family CEOs value social ties and are sensitive to family image, reputation, and social legitimacy (Choi and Wang 2009), which results in investments in an enlarged set of CSR activities, including human resources and human rights (Berrone et al. 2012). In this perspective, recent works developing justice-based arguments suggest that family influence in management limits the bifurcation bias (Verbeke and Kano 2012) and the perception of injustice between family and nonfamily internal stakeholders (Barnett and Kellermanns 2006). As family harmony matters (Miller et al. 2013), family CEOs seek to avoid conflicts within the family. Family CEOs may thus find themselves in a position of moral and/or emotional responsibility toward family members and specific employees with long-standing relationships with the firm. This could lead CEOs to develop a protectiveness concerning them in light of their past relationship with the former generation (Meier and Schier 2016). Finally, the development of long-term relationships with internal stakeholders should strengthen the family’s social capital and their firm’s resilience (Carney 2005). We therefore propose the following:

Hypothesis 1b

Compared to nonfamily firms, family firms managed by family CEOs are positively associated with internal stakeholder-related CSP.

Family Firms Managed by Nonfamily CEOs

We contend that nonfamily CEOs in family firms are less likely to invest in CSR than CEOs in nonfamily firms. Following Cui et al. (2018), our main argument, based on the behavioral agency model, relies on the idea that when families’ SEW preservation is not aligned with the CEOs’ interest, family firms are less likely to invest in CSR. The objective of maintaining the family SEW can be considered by nonfamily CEOs as a contingent loss because of its ambivalent impact on short-term financial performance. According to Cruz et al. (2010) and to Gomez-Mejia et al. (2001), nonfamily CEOs may indeed be more concerned with their “own market value for future employment” (Cruz et al. 2010) than with pursuing the family noneconomic agenda (Gomez-Mejia et al. 2001). To develop our argument further, we consider two stylized cases, the one of family firms with dispersed ownership (Schulze et al. 2003) and the one of family firms with a concentrated family ownership (Rees and Rodionova 2015). In the first case, nonfamily CEOs may be primarily inclined to reduce CSR investments associated with the family agenda in order to focus on short-term economic results and maximize family and nonfamily shareholders expectations (Miller et al. 2013). In the second case, the existence of strong family monitoring capacities may force nonfamily CEOs to limit CSR investments associated with a personal agenda (Surroca and Tribó 2008; Barnea and Rubin 2010) resulting in both cases in a limitation of CSR investments as compared to those in nonfamily businesses. The next two paragraphs detail these mechanisms for both cases of external and internal stakeholder-orientated CSP.

External Stakeholder-Oriented CSP

As nonfamily CEOs’ relationships with the firm are relatively distant (Chua et al. 2003), we can expect the level of SEW endowment to be lower than that in family firms managed by family CEOs. In the case of a fragmented ownership (Schulze et al. 2003), the identification with the firm may be abridged for family members with small stakes in the firm (Miller et al. 2013). With such lower identification comes a reduced need for external legitimacy, community visibility, or for the status associated with the family firm’s name and reputation. Consequently, nonfamily CEOs may have less incentive to invest in external CSR and may be inclined to focus on financial results (Minichilli et al. 2014). Furthermore, according to Dou et al. (2014), when families have fragmented ownership, they have less capability to control the firm’s strategic decision-making (Kabbach de Castro et al. 2017), and therefore would have more difficulty in having the firm to pursue the family noneconomic agenda through external stakeholder-orientated CSR investments. In the case of family firms with significant family ownership, family members with monitoring capacities (Rees and Rodionova 2015) are more likely to control nonfamily CEOs’ willingness to pursue personal branding by preventing/limiting coalition with external stakeholders (Cespa and Cestone 2007; Surroca and Tribó 2008; Barnea and Rubin 2010). This is especially crucial as otherwise, nonfamily CEOs may be inclined to overinvest in external CSR investments to serve their reputation and personal agenda (Bénabou and Tirole 2010). We then propose the following:

Hypothesis 2a

Compared to nonfamily firms, family firms managed by nonfamily CEOs are negatively associated with external stakeholder-related CSP.

Internal Stakeholder-Oriented CSP

Family firms with fragmented ownership (Schulze et al. 2003) are more inclined to ensure short-term performance, take less risk, and request high dividend payments. In this situation, family members are more inclined to see the firm as an economic resource designed to generate economic benefits for family members (Miller et al. 2013). Finally, based on the self-serving interest argument (Kellermanns et al. 2012), we expect that postfounder firms with nonfamily CEOs are less likely to invest in internal stakeholder-related CSR dimensions if they are not directly linked to gains in family members’ wealth. Nonfamily CEOs of family firms are more likely to feel beholden to shareholder demands (family and nonfamily) and to focus on maximizing shareholder value rather than employee interests (Mullins and Schoar 2016). Moreover, and using the same reasoning as previously, family members with monitoring capacities are more likely to control nonfamily CEOs (Burkart et al. 1997) to prevent them from developing entrenching strategies through internal coalition (Surroca and Tribó 2008) or from responding to internal stakeholder claims that could put the family control at risk (Cruz et al. 2014). In addition, several empirical studies found a negative relationship between the family firm and employee-related CSR activities (Chua et al. 2009; Gomez-Mejia et al. 2003). We therefore propose the following:

Hypothesis 2b

Compared to nonfamily firms, family firms managed by nonfamily CEOs are negatively associated with internal stakeholder-related CSP.

The Moderating Role of CEO’s Age

Building on the behavioral agency model, Gomez-Mejia et al. (2007) claim that family firms accumulate a socioemotional endowment that shapes firm preferences and strategic decision-making. According to Berrone et al. (2012), these preferences may be thought of as idiosyncratic noneconomic family priorities, such as family control and influence, identification with the firm, binding social ties, emotional attachment, and dynastic succession. These are integrated into consistent and auto-reinforcing reference points that are used when making strategic decisions such as those on CSR investments. Early studies usually refer to the existence of homogenous and stable reference points (Miller and Le Breton-Miller 2014).

Nason et al. (2018) challenge the conventional assumption of stable reference points by introducing a new classification incorporating a temporal dimension whereby reference points may be thought of as either backward-looking or forward-looking. Backward-looking reference points are characterized by a reinforcement mechanism based on an emphasis on past performance and/or the recall of past events, whereas forward-looking reference point are shaped by anticipating future events. According to Nason et al. (2018), while most family firm priorities may be classified as backward-looking, the prioritization of CEO succession and/or transgenerational control typically falls into the forward-looking category. We expect that family firms managed by family CEOs or nonfamily CEOs shift their reference points slightly as CEO succession and/or transgenerational control become key priorities, thus modifying how they prioritize internal and external stakeholder demands.

Strike et al. (2015) argue that the reinforcement of the transgenerational control and dynastic succession priority is associated with the family firm CEO’s age. As CEOs get older, the succession anticipation (forward-looking priority) becomes a predominant concern. Similarly, associated with the CEO’s age, we posit that family firms managed by nonfamily CEOs may alter their reference point with the increasing importance of the CEO personal agenda, modifying the way they address internal and external stakeholder demands.

Fabrizi et al. (2014) and Oh et al. (2016) have already incorporated CEO’s age as a potential determinant of CSR. However, contradictory results have been produced, such as a positive relationship in Fabrizi et al. (2014) and a negative relationship in Oh et al. (2016). These contradictory results may be explained not only by the direct effect of CEO’s age but also by the way CEO’s age interacts with the family firm CEO types involved.

Family Firms Managed by Family CEOs

When getting older, family CEOs may change their priorities from “control and ego affirmation” (Strike et al. 2015) to dynastic succession. While family CEOs focus on the firm’s image and reputation and the development of internal harmony and cohesion, leading to a positive relationship with both external and internal stakeholders around the succession time, family CEOs may anticipate a greater dependence on the appraisals of external analysts around the succession time (Miller et al. 2013). The focus shifts from preserving the firm’s image and reputation and developing a close identification with the firm to protecting family assets and wealth, strengthening family firm comparability, enhancing the firm’s external attractiveness, and reinforcing the business commitment of the new generation, even among members with smaller stakes.

Our argument is related to the career horizon problem (McClelland et al. 2012). A career horizon problem arises when the interests of shareholders and CEOs are not aligned due to a shorter CEO time horizon. Oh et al. (2016) show that CEO career horizon matters for CSR decisions, as older CEOs tend to invest less in CSR. Using Strike et al. (2015) argument, we posit that in family firms, family CEO’s age is specifically associated with a change in noneconomic priorities. Hence, the older family CEOs get, the more prospective their reference point gets (Nason et al. 2018), as they first and foremost think according to the family firm succession agenda (Strike et al. 2015). In order to ease their succession, that, in turn, becomes their noneconomic purpose priority, family CEOs aim at normalizing their business image by progressively reducing CSR investment, as the latter could be seen as an exclusive family agenda-oriented move (Miller et al. 2013). To put it simply, it is not the CEO’s age that is the reason for the CSR investment decrease, but rather the anticipation of the upcoming succession and the associated need for normalization.

We thus propose the following:

Hypothesis 3a

The CEO’s age negatively moderates the relationship between family CEOs and both external and internal stakeholder-related CSP.

Family Firms Managed by Nonfamily CEOs

Previously, we maintained that the family’s self-serving interests (Kellermanns et al. 2012) and superior monitoring abilities (Rees and Rodionova 2015) explain why several recent studies using large samples find that public family firms are negatively associated with CSP (Cruz et al., 2014; Rees and Rodionova 2015). Following Surroca and Tribo (2008) and Barnea and Rubin (2010), we contend that a firm with an older nonfamily CEO will be more sensitive to the quest for internal and external legitimacy and thus more likely to invest in CSR as part of its personal agenda.

Hence, as nonfamily CEOs get older, their reference point increasingly shifts toward the end of their tenure (Nason et al. 2018). In this perspective, CEOs’ maximization of their market value and/or of their personal image as decision-makers gradually becomes a more and more prominent personal objective (Cruz et al. 2010). The CSR investments implemented by nonfamily CEOs thus become both a way to reinforce their entrenchment via coalition development with both internal and external stakeholders (Cespa and Cestone 2007), and personal agenda-related promotion objective via CSR investments supporting their personal image and reputation. The cumulative effect of these two patterns is a progressive reinforcement of CSR investments. Hence, rather than the sole CEO age factor, it is the degree of their entrenchment associated to their age and the looking-forward for the end of their tenure that lead nonfamily CEOs to be willing and able to reinforce their CSR investments. We thus propose the following hypothesis:

Hypothesis 3b

The CEO’s age positively moderates the relationship between nonfamily CEOs and both external and internal stakeholder-related CSPs.

Methodology

Sample and Data Collection

Our sample included all the European companies evaluated by Vigeo, the European rating agency, and that were publicly listed from 2008 to 2011.Footnote 1 As Vigeo does not always evaluate every firm every year, we only retained the firm-year observations containing at least two consecutive social ratings (i.e., CSR score t and CSR score t-1). Our final sample comprised 555 companies and 1093 firm-year observations. To supplement the CSR ratings supplied by Vigeo, we used the Bloomberg, Compustat, and Capital IQ databases to obtain accounting and financial data for the 555 companies covering 2008 to 2011. All the accounting data were taken from the end of the companies’ financial years. Following Cruz et al. (2014), we used the ORBIS (Bureau Van Dyck) database and the firms’ annual reports to identify family firms and focal families and Datastream to measure direct family ownership (Rees and Rodionova 2015). We manually examined the biographies of each CEO published by Capital IQ and Bloomberg, and we studied the annual reports of the firms in our panel for each year; these were considered in consecutive order to measure family board membership and to characterize each CEO with respect to our defined set of characteristics: nonfounder, belonging to the focal family, age, tenure, and duality.

Variables

Dependent Variables

Following Barnea and Rubin (2010), we measured CSP levels using general CSR scores provided by Vigeo. These scores vary from 0 to 100, and each score measures a specific CSR dimension.

External Stakeholder-Oriented CSP

To account for external stakeholder-oriented CSP, we averaged the following three CSR scores provided by Vigeo: a score linked to environmental performance (Environment), a score linked to professional ethical behavior toward customers and suppliers (Business Behavior), and a score relating to firms’ integration within their local communities (Community Involvement).

Internal Stakeholder-Oriented CSP

Similarly, we averaged the following two scores to evaluate internal stakeholder-related CSP: a score relating to human resources management (Human Resources); a score relating to human rights, such as the promotion of collective bargaining, nondiscrimination policies, and the prevention of degrading treatment (Human Rights). Vigeo has defined a list of criteria for each of these scores based on international norms and standards. For a given firm, each criterion is evaluated using three analysis themes: (1) the relevance of the policies/actions chosen, (2) the consistency of the actions/projects implemented, and (3) the observed/measured effectiveness. Within each theme, several items are defined and evaluated on a scale of 0 to 100. Vigeo independently assesses the items, and the scores are then aggregated using an in-house weighting system. The weightings, as defined by Vigeo, are based on a preliminary sectoral analysis of the critical issues related to CSR.

Independent Variables

We defined family firms using the ultimate owner criterion taken from the ORBIS database (Bureau Van Dyck). Consistent with Cruz et al. (2014), we imposed an ownership threshold of 20% to preselect companies with significant ultimate owners. We then used the annual reports of each firm for each year to identify companies with family ownership and identify the focal family, defined as the one with the highest percentage of shares (Cruz et al. 2014; Villalonga and Amit 2006). Following the method of Cruz et al. (2014), we classified a company as family-owned if an individual or family held more than 20% of the firm’s shares and if at least one family member sat on the board of directors. We also ascertained whether the CEO was a family member (in the case of family-owned companies). We identified all of the CEOs in our sample by their name; we manually examined the biographies published by Capital IQ and Bloomberg, and we studied the annual reports of the firms in our panel for each year, considered in chronological order, to characterize each CEO with respect to our defined set of characteristics.

Next, we defined two binary variables. The first variable, Family CEO, took a value of one if the firm was managed by a CEO who belonged to the focal family and if the CEO was not the founder, and zero otherwise. We defined the second binary variable, Nonfamily CEO, which took a value of one if the CEO of a family firm did not belong to the focal family and zero otherwise.

Moderator Variable

We identified CEO age in years. As each panel observation corresponded to a company and a year t, we determined CEO age in years for each observation as follows: age of the CEO in office at the end of year t. Thus, the variable CEO age is expressed as the logarithm of CEO age in years.

Control Variables

According to Fabrizi et al. (2014), the CEO’s individual power may have a positive influence on the level of CSP. The authors measure the degree of CEO power using the CEO’s length of tenure and duality (i.e., whether the CEO is also the chairman of the board). If CEO age and tenure are interdependent, the authors conclude that they influence firm outcomes differently and therefore should be decoupled (McClelland et al. 2012). We introduced CEO tenure as a control variable to distinguish between these effects (Weng and Zhiang 2014). Adopting the same approach as that described above, we determined the length of tenure of the CEO in office: year t less the appointment year for the CEO in office at the end of year t. We defined the CEO Tenure variable as the logarithm of the tenure length in years plus one. We then identified CEOs acting as both CEO and chairman of the board of directors. We defined CEO Duality as a binary variable, equal to one if the CEO held both functions and zero otherwise. According to Block and Wagner (2014a) and Labelle et al. (2018), family ownership and control influence family firms’ CSP. We used Bloomberg to collect information on board members for each founder and family firm and manually examined annual reports and the biographies for each board member published by Capital IQ and Bloomberg. We defined Family board membership as the percentage of family members on the board of directors. Following Rees and Rodionova (2015), we used the Datastream family ownership variable to measure direct family ownership and labeled it Family Ownership (Boyd and Solarino 2016).

We also included control variables used in previous empirical studies to influence our dependent variables, considering that a firm’s financial performance is presumed to affect its ability to invest in CSR (Campbell 2007). Consistent with Bingham et al. (2011), we used Tobin’s q ratio to capture the financial performance for each firm (Tobin’s q ratio). We determined Tobin’s q ratio for each firm and each year using the same measure as that used by Gompers et al. (2003). Firm risk levels are often controlled for in CSR studies, whereas the least risky companies are presumed to be both the best managed and the most inclined to invest in CSR activities. Investments in CSR can also be analyzed as a way of limiting a firm’s systematic risk (Roberts 1992). Therefore, we expected the risk level to be inversely correlated with the level of CSR investment. We controlled for the risk by using the beta coefficient from the market model as a measure of the company’s systematic risk (Beta). The beta coefficient is obtained using an estimated market model for each firm and for each year based on the regression over the last two years of the weekly returns of common stock against the weekly returns of the Euro Stoxx index (Source: Bloomberg). The company’s age may also affect its degree of CSR commitment, in so far as stakeholders’ CSR expectations may become entrenched over time and CSR reporting pressures may increase (Roberts 1992). We accounted for firm age (Age) using a variable that was the logarithm of one plus the number of years the company had been listed on the stock exchange (Oh et al. 2011). As size appears to be a significant determinant of CSP (Wu 2006), we expected it to be positively related to CSR, as larger companies are assumed to have more resources to make greater investments in CSR and comply with stronger external pressures in terms of CSR reporting (Wu 2006). Our variable, Size, was measured as the logarithm of each firm’s total assets (Barnea and Rubin 2010). A firm’s debt level is also considered to influence CSP. We thus expected debt-servicing obligations to limit CSR investments (Barnea and Rubin 2010); on the other hand, the CSR expectations and monitoring ability of creditors such as banks and other financial institutions may encourage firms to increase their CSR investments (Roberts 1992). Overall, however, previous studies point out a negative relationship between debt and CSP. We thus measured a company’s debt level as its ratio of total debt to total assets (Debt). Finally, earlier empirical studies exhibit strong industry, country, and year effects with respect to the CSR data (Waddock and Graves 1997). Thus, we controlled for industry, country, and year effects by using industry, country, and year dummies.

Statistical Methods

We used OLS regression models as presented in Table 2 after testing for the presence of random effects using a Breusch–Pagan Lagrangian multiplier (LM) test. The null hypothesis in the LM test was that there were no firm-specific intercepts (i.e., there were no panel effect). Tests results did not reject the null hypothesis for each of our specifications, indicating the absence of a panel effect. To account for the lack of independence of the error term for observations from individual firms, all calculated standard deviations were cluster robust. Following Barnett and Salomon (2012), we introduced a one-year lag of the dependent variable in each of our regressions. Incorporating linear autoregressive dynamics with a one-year lag in the dependent variable allowed us to account for within-firm persistence in CSR. Finally, following the recommendations of Barber and Lyon (1996), we used a winsorization procedure at the 1% threshold to reduce the effect of potential spurious outliers. The following variables were winsorized: Tobin’s q, Size, Debt, and Beta. Descriptive statistics are reported in Table 1. The moderate correlation among our variables (see Table 1) excluded the possibility of any multicollinearity problem in our data. Proceeding to alternative checks, we calculated variance inflation factors (VIFs) for all models. All the VIFs were below 3, confirming the absence of multicollinearity.

Table 1 Descriptive statistics and Pearson correlations
Full size table

Results and Discussion

Results

Table 2 analyzes the influence of our two types of leadership on both external and internal stakeholder-orientated CSP as well as the moderating role of the incumbent CEO’s age. As CEO age is supposed to have a direct influence on CSR (Fabrizi et al. 2014; Oh et al. 2016), we suspected that quasi-moderation was involved.Footnote 2 We therefore implemented a three steps procedure (Darrow and Kahl 1982). In step one, we implemented our OLS regressions with both independent and control variables only (Models 1 and 2). In step two, we added the moderator variable to our models (Models 3 and 4). In step three, we implemented our full interaction models (Models 5 and 6). Our results confirm that CEO age acts as a quasi-moderator. Following Carte and Russel (2003), we interpreted first-order effects within the interaction models only (models 5 and 6), as the interpretation of first-order effects outside interactions (models 1 to 4) was not relevant.

Table 2 OLS regressions
Full size table

All our interaction models (models 5 and 6) were statistically significant at the 1% level, with a F-test > 79 and a p value < 0.000 in all cases. In all models, our reference category referred to nonfamily firms (coded 0), meaning that the coefficients associated with our dummies (family CEO, and nonfamily CEO) can be interpreted as the direct effects of our corresponding categories when compared to the reference category (nonfamily firms). Interaction effects were measured by the interaction terms Family CEO * CEO Age and Nonfamily CEO * CEO Age. To enable a complete interpretation of the full effect (main effect + interaction effect), we plotted all the interactions in Figs. 1 and 2. Our two graphs aim at showing the full effect of CEO age on both external and internal stakeholder-oriented CSRs depending on whether the CEO is a family CEO or a nonfamily CEO as compared to nonfamily firms. The full effect (main effect + interactions) was plotted using the mean values of our control variables and taking a range of significant values for the age of the CEO (35 to 65). The line for nonfamily firms was plotted by setting the independent dummy variables to zero (family CEO, nonfamily CEO). The other two curves for family and nonfamily CEOs were plotted by setting these variables to one and/or zero, respectively.

Fig. 1
figure 1

Relationship between CEO age and external CSR

Full size image
Fig. 2
figure 2

Relationship between CEO age and internal CSP

Full size image

Model 5 is related to our hypotheses H1a, H2a and H3a,b. Family CEO is positively associated with external CSR (p < 0.01), which provides strong support for our hypothesis H1a, whereas nonfamily CEO is negatively associated with external CSR (p < 0.1), which corroborates hypothesis H2a. Moreover, both interaction effects are significant. The interaction term Family CEO * CEO Age is negatively associated with external CSR (p < 0.01) meaning that CEO Age negatively moderates the relationship between family CEO and external CSR, which provides supports for hypothesis H3a. This result means that the relationship between family CEO and external CSR strongly depends on CEO Age. We plotted the full interaction in Fig. 1. The line depicted for firms with family CEO is always above the line for nonfamily firms. The slope is negative, and we interpret it as a normalization move where, with CEO age, firms with family CEO tend to have the same level of external CSR commitment as nonfamily firms. In model 5, family firms managed by a nonfamily CEO are negatively associated with external CSR (p < 0.1), providing some support for hypothesis H2a. The interaction term nonfamily CEO * CEO Age is positively associated with external CSR meaning that CEO age positively moderates the relationship between nonfamily CEO and external CSR, providing strong support for our hypothesis H3b (p < 0.01). This interaction is also plotted in Fig. 1 where the line depicted for family firms with nonfamily CEOs is below the line for nonfamily firms when CEOs are their early age (before 50 years old). After that pivotal age, nonfamily CEOs in family firms tend to overinvest in external CSR as compared to nonfamily firms.

Model 6 is related to hypotheses H1b, H2b and H3a,b. Family CEO is positively associated with internal CSR (p < 0.01), whereas nonfamily CEO within family firms is negatively associated with internal CSR (p < 0.1). These results provide strong support for our hypotheses H1b and H2b. The interaction term Family CEO * CEO Age is negatively associated with internal CSR (p < 0.01) providing further support for our hypotheses H3a. Figure 2 plots the full effect (main effect & interaction effect) and shows that the line depicting family CEOs is also above the line depicting nonfamily firms meaning that family CEOs appear to be more committed to internal CSR than their nonfamily counterparts. Moreover, the slope of this line is also negative demonstrating a normalization move associated with CEO’s age. Similarly, the interaction term Nonfamily CEO * CEO Age is positively associated with internal CSR (p < 0.1) supporting our hypothesis H3b.

Robustness checks

We run alternative econometric specifications to verify the robustness of our results (Tables 3 and 4). Table 3 reports tests for alternative specifications. First, we used censored Tobit regressions (models 1 and 2) with cluster-robust standard deviations to account for the fact that our dependent variables had to lie between 0% (left-censoring limit) and 100% (right-censoring limit). Our findings are also similar and more significant than those of the OLS regressions, providing further support for hypotheses H1a, H1b, H2a, H2b and H3a,b.

Table 3 Robustness checks
Full size table
Table 4 Robustness checks
Full size table

Second, as we observed no difference in the relationship between the type of CEO in family firms and both external and internal CSR, we used a broad measure of CSR, averaging external and internal CSR ratings (model 3). These results were consistent with our previous results as family CEO was positively associated with our global CSR measure (p < 0.001) and nonfamily CEO (in family firms) was negatively associated with that measure (p < 0.05). Similarly, both interaction terms were significant (respectively with p < 0.001 and p < 0.05) showing a negative moderating effect of CEO age on the relationship between family CEO and CSR, and a positive moderating effect on the relationship between nonfamily CEO (in family firms) and CSR.

Third, we run OLS regressions on both external (model 4) and internal (model 5) CSR with a new independent binary variable labeled Family. This variable was equal to 1 if the firm was a family firm (with a family CEO or a nonfamily CEO) and zero otherwise. Our results show no significant relationship between this new Family variable and both external and internal CSR and no significant interaction effect (Family * CEO Age). We interpret these results as a strong robustness test as it allows us to demonstrate that the observed relationships in Table 2 are not associated with the family firms versus nonfamily firms duality but are rather strongly associated with both the type of CEOs within family firms and the interaction effect between the type of CEOs and CEOs’ age.

Table 4 reports tests on additional exogenous factors that could be associated with nonfamily CEOs’ behavior toward CSR.Footnote 3 First, we dealt with the origin of nonfamily CEOs, whether they were external or internal to the family firm. Indeed, according to Minicilli et al. (2014), the origin of CEOs may have an impact on the way CEOs embed the family agenda, external CEOs being less inclined to understand and preserve the family’s noneconomic priorities. This could in turn affect the way nonfamily CEOs prioritize CSR. We consequently introduced two new binary variables (models 1 and 2): Nonfamily external CEO and Nonfamily internal CEO. The first one took the value of one if the Nonfamily CEO had started his/her career before being appointed CEO outside the family firm, and zero otherwise. The second one took the value of one if the Nonfamily CEO was a member of the family firm before becoming the CEO. We used the definition of Borokhovich et al. (1996) stating that a CEO is categorized as external if he/she joined the company within the 12 calendar months prior to his/her appointment. Our results are qualitatively similar, though less statistically significant, which provides further support to our hypotheses.

Second, we controlled for the influence of newly appointed CEOs. As Gomez-Mejia et al. (2001) noted, family CEOs are typically replaced by new nonfamily CEOs only when the firm performance is very bad or the firm is in crisis, meaning that the newly CEO may have fewer resources to invest in CSR. We introduced two additional binary variables (models 3 and 4): Nonfamily CEO (Ten.\(\le\) 1) and Nonfamily CEO (Ten > 1). The first one took the value of one if the Nonfamily CEO had a tenure inferior or equal to one year, and zero otherwise. The second one took the value of one if the Nonfamily CEO had tenure superior to one year. Models 3 and 4 therefore controlled for a potential effect associated with newly appointed CEOs. Our results are qualitatively similar, though they are only statistically significant for Nonfamily CEO with tenure greater than one. These results provide some additional support to our hypotheses.

Finally, we run a sensibility analysis using a 50% ownership cutoff for the categorization of the family firm. Our results (available upon request) are largely unchanged and provide further support to our hypotheses.

Discussion and Conclusion

We examined family firm heterogeneity to develop our main argument that different types of CEO imply different noneconomic priorities, which leads to differentiated external and internal stakeholder-oriented CSP. This article provides indirect evidence that these priorities are evolving dynamically and that CEOs in family firms tend to gradually normalize their CSP (Miller et al. 2013). This move toward more conservative CSP (see Figs. 1 and 2) may be associated with different legitimization strategies, depending on who is leading the firm (Table 5).

Table 5 Family & Nonfamily CEO, Expected CSP, and Normalization moves
Full size table

First, we found evidence that family firms with family CEOs are strongly and positively associated with both external and internal stakeholder-related CSP. Our results are similar to the findings of Block and Wagner (2014b) regarding product-related CSP and diversity aspects, which is a factor similar to our human rights dimension. Our findings are also consistent with the view of Cennamo et al. (2012) that family firms are more inclined to develop strongly committed external and internal stakeholders. Furthermore, our findings are consistent with the view that family firms tend to invest in CSR to enhance their image and reputation within the broader community and develop trusting relationships with external stakeholders such as customers and suppliers. These findings are consistent with earlier research findings that family firms are more likely to invest in environmental activities as a response to institutional pressure and as a way to manage long-term risk and preserve the family name and image (Berrone et al. 2010). However, our results differ by suggesting that this finding may be true primarily for family firms with family CEOs. Our results appear to be strongly moderated by the CEO’s age, suggesting that family CEOs tend toward a gradual normalization of their CSP. The fact that CSR-overinvesting CEOs gradually reduce their CSR investments may indicate a need for family CEOs to normalize the external perception of the firm and thereby sends a signal that the firm’s initial orientation toward noneconomic family priorities is compatible with the objective of shareholder value maximization. When anticipating changes in family firm ownership and management, family CEOs may be inclined to objectify the family firm as a corporate enterprise, indicating that the family firm is an attractive prospect for nonfamily shareholders as well as for future next-generation family owners.

Second, our findings suggest that family firms with nonfamily CEOs may be negatively associated with both external and internal stakeholder-related CSP. As family firms with nonfamily CEOs are prominent within public family firms (accounting for 67% of our sample), this result is in line with recent studies using large samples that find that public family firms are negatively associated with CSP (Cruz et al. 2014; Rees and Rodionova 2015).

We also find a moderating effect of the nonfamily CEO’s age. This finding is consistent with previous studies’ finding that CSR may be associated with CEOs’ personal branding and reputation-building strategies (Barnea and Rubin 2010), entrenching strategy (Surroca and Tribó 2008), and power (Fabrizi et al. 2014). We interpret the moderating effect as a rebalancing of the forces between the family’s shareholding and its monitoring capacity (Cespa and Cestone 2007), combined with the gradual reinforcement of CEO power.

Implications for Research

Our findings contribute to the literature on the relationship between the family firm and CSR. This is the first study to systematically address how different types of family firm CEOs influence both external and internal stakeholder-related CSP (Bingham et al. 2011; Block and Wagner 2014a). By linking family CEOs and noneconomic priorities, we provide an integrated framework that contributes to the understanding of how heterogeneity in leadership influences the CSP of family firms.

By introducing the interaction between CEO’s age and CEO type, we introduce a temporal dynamic dimension in our analysis. Our results suggest that family firms priorities shift over time and that family CSR behavior is gradually adapted by CEOs to serve their objectives in terms of new legitimization needs (managing family social capital and/or personal agenda) and to prepare for succession (Chrisman et al. 2005; Steier 2001). Our research contributes to the growing literature that is challenging the conventional assumption of a stable reference point (Nason et al. 2018). Our evidence suggests that, if family firms priorities are a key driver of family CSP, the increasing importance of succession induces the consolidation of a new forward-looking reference point, which in turn modifies how family firms instrumentalize CSR investments. Our results call for more research on shifting noneconomic family priorities. If the incumbent CEO’s age (Strike et al. 2015) is an interesting variable that should be taken into account concerning the anticipation of future succession, other variables may be incorporated as well, such as the percentage of next-generation members on the board of directors for instance. Moreover, more research is needed to better understand how specific family firm priorities may be shifting during the leadership of a given type of CEO and, more specifically, how family firm priorities are shifting around specific events such as succession or the introduction of the new generation to firm governance or management.

This research also highlights the complexity of the notion of noneconomic objectives by distinguishing between those related to family interests (in line with the socioemotional wealth perspective) and those more related to business (in line with the stakeholder view) and by proposing a framework that explains to what extent the former influence the latter according to the type of CEO. Further research is then needed to understand whether and how other contextual factors (the number of generations, the participation of new generations, etc.) can influence the way noneconomic objectives are formulated in family businesses.

Managerial Implications

Our research makes several contributions to practice. First, we demonstrate that family CSR should not be analyzed as a homogeneous phenomenon. External analysts studying a specific family firm CSP dimension should be aware of how specific types of leadership shape the firm’s motivations and interests with respect to CSR. They should also be aware that CSR may be a lever by which family firm CEOs can pursue a specific agenda (family agenda or personal agenda). Moreover, a family CSR agenda may be better thought of as a legitimization lever than as a way to proactively respond to specific stakeholder needs. From this perspective, our work tends to support the instrumental view of CSR in family firms (Windsor 2006). This finding is of particular importance for any CSR executive in family firm seeking to implement specific CSR actions. We believe that these actions should be aligned with the family agenda, the firm’s specific type of leadership, and any anticipatable change in the patterns of future family involvement in firm management.

Second, our research highlights that family firms have no general attitude to internal (nonfamily) or external stakeholders. This finding, consistent with the analysis of Morck and Yeung (2003), does not mean that family firms do not care about nonfamily stakeholders. Instead, our interpretation is more instrumental (Windsor 2006), in that our data suggest that the family stakeholder orientation is not homogenous but rather shifts over time and is shaped by family firm priorities and changing needs for legitimization. In this sense, the family firms stakeholders may benefit from internalizing these attitudes and developing relationships in a contractual mode regarding nonfamily CEO leadership, whereas they may be more inclined to develop relationships in a relational/emotional mode regarding the leadership of postfounder firms with family CEOs.

Third, our data suggest that, if nonfamily CEOs initially follow the family agenda and fulfill their priorities, gradual entrenching actions may, over time, balance the family’s monitoring activities, so that CSR investments become at once an entrenching strategy and a signal of CEO entrenchment. As CSR could be a source of conflict of interest between family owners and nonfamily CEOs, our results suggest that family firms should establish specific control mechanisms for monitoring and evaluating CSR activities.

Limitations and Future Research Directions

We acknowledge that the study has several limitations that point to future research avenues. The first limitation is related to the size and type of family firms examined in our sample. Because CSR rating data are only available to large public companies, our results should not be generalized to all types of family firms. Specifically, the dynamic pressure for normative CSP is more salient for public family firms that need to attract external financing and convince investors and other external regulatory bodies that the policies of the firm take into account the requirements and demands of the stakeholders; as such, the policies are not focused on the family agenda only. Moreover, as size is a key determinant of CSR investment, larger public family firms could exhibit behaviors with respect to CSR different from those shown by smaller private companies. Overall, if our theoretical arguments do not depend on the size or the public/private characteristics of family firms, our empirical results should not be generalized to all types of family firms and should be interpreted with caution. We encourage future research on private family firms as well as on small and medium-size family businesses (Herrero 2011). One of the main challenges for these studies would be to design and measure proxies of CSP for these companies, given that most social rating agencies tend to focus on large public firms.

Second, we develop arguments explaining how family priorities can shape family CSR strategies and how shifting priorities along with the emergence of transgenerational control and dynastic succession concerns modify family firms’ CSR outcomes. However, as in Cruz et al. (2014), we provide only indirect measures of these family priorities. These indirect measures do not allow us to study the direct relationships between specific family priorities and specific CSR activities. Moreover, in our sample, the leadership type remains stable at the firm level, which precludes us from exploring leadership variation over time. Future research should address this gap and develop methodological tools with which to identify and directly measure family priorities and to explore how changes in leadership types modify firms’ CSR outcomes.

A third limitation is related to our conceptualization of CSP. We adopt a broad view of CSP and use a general score instead of distinguishing between social initiatives and social concerns (Dyer and Whetten 2006). Using a general score has several advantages, as it allows us to proxy for CSR investment and commitment levels and to use a measure of CSP that is homogeneous across firms. Further research may consider divergent reference points of family CEOs and nonfamily CEOs when they face the decision of whether to invest in social initiatives or limit social concerns. Similarly, future research may investigate how different types of leadership may influence social initiatives and social concerns toward both external and internal stakeholders.

The fourth limitation relates to the importance of family governance and the personal characteristics of incumbent CEOs, as they may influence CSR decisions. Our research does not account for specific family governance bodies, such as family councils, family offices, or other control mechanisms, although family-related governance mechanisms can strongly influence the strategic decisions of family firms (Mustakallio et al. 2002). Further research may explore the ways in which incumbent-generation family members’ perceptions and monitoring abilities with respect to CSR influence CSR decisions in family firms. Moreover, we encourage researchers to consider more fine-grained characteristics of incumbent CEOs. Although our research offers controls for some personal characteristics (i.e., age, tenure, and CEO duality), additional factors relating to personal and professional experiences have been suggested in the literature as potential determinants of CSR investment decisions, such as CEO education, gender, compensation schemes, and the personal perceptions of CEOs with respect to ethics (Godos-Diez et al. 2011).

Lastly, following McClelland et al. (2012) and Strike et al. (2015), we used CEO Age as a proxy for succession concerns (Chua et al. 2003) in relation with CEO career horizon problem associated with CSR (Oh et al. 2016). We recognized that our proxy, CEO Age, might only be related with succession concerns but not with the effective succession process. As such, our results should be interpreted cautiously as it does provide direct evidences of shifting reference points around succession events. Future research should explicitly integrate this temporal dimension using for example a difference in difference approach in CSR performance around family CEO succession (Barber and Lyon 1996). Moreover, we call for longitudinal studies to the extent that the magnitude and speed of the shift in reference point may also depend on multiple contextual (number of generations, number of family branches present, etc.) and individual factors.

Beyond these limits, we propose three other avenues of research. First, we suggest developing the life cycle approach proposed by Le Breton-Miller et al. (2013) in order to further study how certain stages in the evolution of family businesses shape their noneconomic priorities and the different CSR strategies that can be associated with them. This would require tracking family firms at different stages (founding stage, postfounder GEN N/N + 1) and identifying different configurations (with/without family CEOs, number of family branches, number of generations on board, etc.) and measuring the differences in CSR strategy across these stages/configurations. Second, we propose to measure the potential change in CSR strategy around key issues other than CEO succession, such as a sharp decline in financial performance (Cruz et al. 2014), significant diversification or a major merger or acquisition, etc. Since key events are indeed likely to change the benchmark for family firms, we suggest that they could lead them to revise their CSR strategy (Berrone et al. 2012). As such, we suggest that these key events may also act as key drivers of family firms’ attitude toward CSR. Finally, we suggest investigating further potential differences between public and private family firms’ behavior toward CSR. As mentioned earlier, our sample only comprised large public firms with strong normative pressures (Miller et al. 2013). Following, Litz and Steward (2000), Ding and Wu (2014), and, Campopiano and De Massis (2015), we acknowledge that private family firms may behave differently from public family firms with respect to CSR. New researches should be done to sort out whether our results are specific to large public family firms or if they can be generalized to other types of family firms.