INTRODUCTION
Innovation plays a prominent role in creating and sustaining firms’ competitive advantages, and thus executives are expected to engage in R&D activity. However, according to GE's Global Innovation Barometer 2014 report, only 39% of the 3,209 CEOs who said innovation was an important driver think that their company excels at delivering it. To reconcile this uncomformity, the role of executive managers in firm innovation has been broadly studied from the perspective of agency theory (e.g., Tong, He, He, & Lu, Reference Tong, He, He and Lu2014; Zhou, Gao, & Zhao, Reference Zhou, Gao and Zhao2017). Top managers such as CEOs exert significant influence over their firm's strategic choices in areas such as innovation (Hambrick & Mason, Reference Hambrick and Mason1984). Innovation is inherently risky and its returns are uncertain (March, Reference March1991). For example, out of every seven new product/service projects, only one succeeds (Cooper, Reference Cooper2011). According to agency theory, CEOs are reluctant to engage in innovation because as agents they are assumed to be self-interested and risk-averse (Jensen & Mecking, Reference Jensen and Meckling1976). While most researchers have taken the assumption of self-interest as a given and focused on designing governance mechanisms to mitigate untested self-interest problems, some other research findings have shown that managers are not all self-interested but are good stewards motivated to act in the best interests of their principals (Donaldson & Davis, Reference Donaldson and Davis1991). Thus, researchers have long called for the direct examination of the basic assumption of agency theory (Eisenhardt, Reference Eisenhardt1989).
In this study, we aim to directly test the assumption of self-interest by examining the motives of CEOs. Specifically, we argue that CEOs possess a mix of both self-preserving and other-regarding motives. This mix is rooted in their values in the self-other dimension. Self-regarding values involve the pursuit of one's own success, happiness, and wealth (Jensen, Reference Jensen, Coffee and Lowenstein1988), while other-regarding values emphasize the success, happiness, and well-being of others (Fu, Tsui, Liu, & Li, Reference Fu, Tsui, Liu and Li2010; Wood, Reference Wood1991). Self-interested CEOs, that is, CEOs with self-regarding values that outweigh other-regarding values, tend to be more risk-averse and self-preserving. Thus, they are less likely to put much effort into innovation.
Moreover, the literature has offered a limited understanding of the mechanism through which managerial self-interest affects firm choices and performance. Top managers such as CEOs incorporate their personal preferences into corporate policies, and this effect becomes stronger when the outcomes of efforts such as innovation are uncertain (Hambrick & Mason, Reference Hambrick and Mason1984; Wu, Levitas, & Priem, Reference Wu, Levitas and Priem2005). Thus, CEO values, as an attribute at the individual level, must affect the outcomes of firms via intermediate mechanisms at the firm level (Chin, Hambrick, & Treviño, Reference Chin, Hambrick and Treviño2013). We posit that CEOs with higher levels of self-regarding values are less likely to develop corporate strategies oriented toward the long term (Venkatraman, Reference Venkatraman1989; Wang & Bansal, Reference Wang and Bansal2012), and more specifically, that they are less likely to facilitate firm innovation by directing resource allocation toward future competitive advantages. In sum, we propose a mediation model to investigate how CEO values affect firm innovation through firms’ long-term orientation.
Furthermore, the effect of CEOs’ personal attributes on firm strategic decision-making is contingent on executive job demands (Hambrick, Finkelstein, & Mooney, Reference Hambrick, Finkelstein and Mooney2005). Specifically, CEOs tend to rely more on their psychological attributes when faced with high-level job demands. Thus, we propose that the mediation model varies with the individual (e.g., CEO tenure), organizational (e.g., CEO duality), and environmental (e.g., uncertainty) factors that alter the extent to which managers matter in firm innovation (Hambrick, Reference Hambrick2007).
We used a multi-wave large-scale survey research design in collaboration with China Entrepreneurs Survey System (CESS), and tested our hypotheses with a sample of 436 firms across different regions and industries between 2014 and 2016. We found that firms with CEOs who have high self-regarding values invest less in R&D and perform worse in new product sales, a relationship mediated by firms’ long-term orientation. CEO tenure, CEO duality, and environmental uncertainty further weaken the mediated main relationship.
Our study makes several theoretical contributions to the literature. First, it contributes to agency theory by directly examining the effects of self-preserving managers on firm innovation. Second, we add the new attribute of CEO values to the empirical testing of upper echelon theory, and echo the call of Finkelstein, Hambrick, and Cannella (Reference Finkelstein, Hambrick and Cannella2009) to investigate the relationship between executive values, firm outcomes, and boundary conditions. Third, the finding that long-term orientation mediates the relationship between CEO values and firm innovation demonstrates that managerial self-interest may hurt firm innovation by shortening the temporal orientation of firms, which provides a new explanation for how managers matter in firm innovation.
THEORETICAL BACKGROUND AND HYPOTHESES
Agency Theory and CEO Values
Agency theory addresses the principal-agent interest divergence where the principal (e.g., owners) delegates authority to the agent (e.g., top managers) who performs the managerial work (Jensen & Meckling, Reference Jensen and Meckling1976). Agency problems arise when ‘(a) the desires or goals of the principal and agent conflict and (b) it is difficult or expensive for the principal to verify what the agent is actually doing’ (Eisenhardt, Reference Eisenhardt1989: 58). The theory assumes that managers are self-interested, bounded rational, and risk averse, and substantial research has focused on determining the most efficient mechanisms governing the principal-agent relationship. In contrast, almost no studies have tried to examine this taken-for-granted assumption.
The untested assumption described above has incurred much criticism in the literature. For example, Ghoshal (Reference Ghoshal2005) argued that the assumption of agency theory is unrealistic and lacks both face validity and empirical support. Donaldson and Davis (Reference Donaldson and Davis1991) and Davis, Schoorman, and Donaldson (Reference Davis, Schoorman and Donaldson1997) contended that managers are not self-interested but rather good stewards of corporate assets because they are motivated by the need to achieve intrinsic satisfaction from successfully performing challenging work, and by the desire for authority and recognition. Bosse and Phillips (Reference Bosse and Phillips2016) challenged this assumption by arguing that ‘economic actors are not narrowly self-interested but boundedly self-interested’ (italics original: 276). Empirical evidence has also shown that managers’ self-interest is bounded by norms of reciprocity and fairness (Ariño & Ring, Reference Ariño and Ring2010; Fong, Misangyi, & Tosi, Reference Fong, Misangyi and Tosi2010).
Indeed, managers are not all self-interested. Lin, Lin, Song, and Li (Reference Lin, Lin, Song and Li2011) found that CEOs with different levels of education, professional experiences, and political connections vary in their levels of self-interest in terms of innovation. However, these superficial demographic attributes may reflect the extensive differences in the values of CEOs (Harrison, Price, Gavin, & Florey, Reference Harrison, Price, Gavin and Florey2002). Values, defined as a broad and relatively enduring preference for a specific state of affairs over others (Hofstede, Reference Hofstede1980), consist of ‘enduring beliefs that a specific mode of conduct is personally or socially preferable’ (Rokeach, Reference Rokeach1973), influencing the fundamental ways in which individuals perceive the environment (Meglino & Ravlin, Reference Meglino and Ravlin1998).
Many researchers have examined the effects of CEO values on firm behavior and performance. For example, O'Reilly and Pfeffer (Reference O'Reilly and Pfeffer2000) argued that CEO values affect how a firm deals with stakeholders such as employees, customers, and suppliers. Ling, Zhao, and Baron (Reference Ling, Zhao, Baron, Ling, Zhao and Baron2007) suggested that two values held by founders, collectivism and novelty, improve the quality of their decision making and enhance the implementation of selected strategies, thus improving entrepreneurial firm performance. Fu et al. (Reference Fu, Tsui, Liu and Li2010) demonstrated that CEO values moderate the effect of transformational leadership on middle managers, depending on the congruence between CEO values and transformational behavior. Chin et al. (Reference Chin, Hambrick and Treviño2013) found that unlike conservative CEOs, liberal CEOs exhibit greater corporate social responsibility (CSR) because they are more likely to perceive CSR as beneficial to the firm's owners and believe that CSR is intrinsically desirable.
As a broad construct, values include many aspects. We focus on the dimensions of ‘self-regarding’ and ‘other-regarding’ (Allison & Messick, Reference Allison and Messick1990; Griesinger & Livingston, Reference Griesinger and Livingston1973). Specifically, self-regarding values encompass the pursuit of success, happiness, and wealth for oneself and one's family, while other-regarding concerns enhancing the success and happiness of others, with particular importance placed on the common good. The topic of executive self- versus other-regarding values has attracted increased attention in the management literature. Some scholars have argued that firms perform better when CEOs are self-regarding and focus narrowly on profit maximization (Jensen, Reference Jensen, Coffee and Lowenstein1988). Such positive performance effects result from the efficiency gained from self-regarding. In contrast, other scholars have suggested that CEOs with other-regarding values are more likely to take a cooperative approach and seek a fit between the firm and its important external relationships, enhancing firm performance (Clarkson, Reference Clarkson and Preston1988; Miles, Reference Miles1987). In addition, other scholars have combined both perspectives and argued for a null or non-linear relationship between CEO values and firm performance. For instance, Agle, Mitchell, and Sonnenfeld (Reference Agle, Mitchell and Sonnenfeld1999) found no significant relationship between CEOs’ self-regarding values and firm performance because self-regarding values can improve efficiency in the firm-centered context but may affect the focal firm's relationship with broad stakeholders in the system-centered context. In contrast, other-regarding values enjoy the benefits of collaboration and cooperation in the firm-centered context while losing the efficiency gains in the firm-centered context.
Drawing on the literature, we argue that CEOs are neither narrowly self-interested, as agency theory assumes, nor purely altruistic (Ghoshal, Reference Ghoshal2005) and motivated to pursue pro-organizational behaviors (Davis et al., Reference Davis, Schoorman and Donaldson1997). Instead, CEOs possess a mix of self-regarding and other-regarding values. The strength of self-regarding values relative to other-regarding values captures the extent to which CEOs may be self-preserving, and thus affects whether they direct their firms toward innovation. Moreover, the literature has examined the effects of CEO attributes such as demographic characteristics (Barker & Mueller, Reference Barker and Mueller2002; Wu et al., Reference Wu, Levitas and Priem2005), hubris (Li & Tang, Reference Li and Tang2010), and overconfidence (Galasso & Simcoe, Reference Galasso and Simcoe2011) on firm innovation, but little has been done to underpin the effect of CEO values. We endeavor to fill this gap. In the next section, we develop our hypotheses to link CEOs’ self-regarding values and firm innovation via long-term orientation. Figure 1 depicts our theoretical framework.

Figure 1. Theoretical framework
CEOs’ Self-Regarding Values and Firm Innovation
We define innovation as a process that refers to the effort firms invest to produce, assimilate, renew, develop, and exploit value-added products, services, and markets (Crossan & Apaydin, Reference Crossan and Apaydin2010). Innovation is widely regarded as a critical source of competitive advantage and as contributing to a firm's long-term success (Ahuja, Lampert, & Tandon, Reference Ahuja, Lampert and Tandon2008; Tushman & O'Reilly, Reference Tushman and O'Reilly1996). Despite its high returns, the failure rate of innovation is also high. According to recent estimates, the failure rate of new product development ranges from 24% to 48% and varies across industries (Castellion & Markham, Reference Castellion and Markham2013). Indeed, R&D expenditures may not result in any payoffs or may only translate into profits after many years have passed.
In uncertain situations, CEO values play a critical role in determining resource allocation. Value fundamentally ‘influences the selection from available modes, means, and ends of action’ (Kluckhohn, Reference Kluckhohn, Parsons and Shils1951: 395). Following agency theory, we argue that CEOs’ self-regarding values are negatively related to firm innovation. When CEOs have high levels of self-regarding values, they care more about their employment security and income, which are tied closely to the firm (e.g., Beatty & Zajac, Reference Beatty and Zajac1994; Gray & Channella, Reference Gary and Cannella1997; Hirshleifer & Thakor, Reference Hirshleifer and Thakor1992). Allocating resources to innovation can be very risky. Self-regarded CEOs are less likely to put their job security, personal income, and reputation at risk.
Instead, managerial mischief is more likely to occur (Reilly, Souder, & Ranucci, Reference Reilly, Souder and Ranucci2016). Empirical evidence has suggested that self-regarded CEOs do not direct firms to engage in CSR even though they know that CSR improves performance in the long run (Agle et al., Reference Agle, Mitchell and Sonnenfeld1999). Similarly, there are often multi-year lags between R&D expenditures and their potential revenues. High self-regarded CEOs are tempted to attain short-term gains rather than expect uncertain long-term returns. In doing so, they can also rapidly enhance their reputations and preserve their personal well-being (Narayanan, Reference Narayanan1985). In sum, CEOs with a high level of self-regarding values lack the motivation to increase their innovation efforts. Thus, we posit the following:
Hypothesis 1:
CEOs with higher self-regarding values will be negatively related to firm innovation.
The Mediating Role of Long-Term Orientation
Firm temporal orientation is a subjective perspective of time reflected in the temporal depth of a firm's strategic decisions (Ancona, Okhuysen, & Perlow, Reference Ancona, Okhuysen and Perlow2001), and can range from short-term to long-term. Even though these are not mutually exclusive, the two types of temporal orientations often reflect different strategic priorities and require different organizational processes (Wang & Bansal, Reference Wang and Bansal2012). Strategic decisions with a short-term orientation emphasize efficiency, while decisions with a long-term orientation emphasize effectiveness (Venkatraman, Reference Venkatraman1989). Short-term orientation may also push managers to pursue a course of activities that is best for the short term but that is suboptimal over the long term (Flammer & Bansal, Reference Flammer and Bansal2017; Laverty, Reference Laverty1996; Martin, Wiseman, & Gomez-Mejia, Reference Martin, Wiseman and Gomez-Mejia2016). In contrast, long-term orientation prioritizes the long-range implications and effects of decisions that come to fruition after an extended period (Lumpkin, Brigham, & Moss, Reference Lumpkin, Brigham and Moss2010; Miller & Breton-Miller, Reference Miller and Breton-Miller2005).
CEOs’ self-regarding values are negatively related to firms’ long-term orientation. CEO values shape the cognitive processing in which CEOs selectively notice, encode, and retain information that is consistent with their values (De Dreu, Reference De Dreu2007; Kruglanski, Orehek, Dechesne, & Pierro, Reference Kruglanski, Orehek, Dechesne and Pierro2010). The time horizon is manifested through two prominent decision-making concerns. First, CEOs care about when exactly they can expect to see returns. Second, CEOs are uncomfortable about the variability of expected returns from long-term projects because uncertainty increases with the longer time horizon (Reilly et al., Reference Reilly, Souder and Ranucci2016). The ‘loss aversion’ of self-regarded CEOs reflects an inherent preference for avoiding risks by discounting losses in the future relative to losses in the present (Tversky & Kahneman, Reference Tversky and Kahneman1991), and thus they are myopic in terms of their preferences for an immediate payoff. Research has shown that self-regarded CEOs favor investments that pay off in the short term even at the expense of organizational long-term success (Stein, Reference Stein1988).
Thus, self-preserving motives compel the self-regarded CEOs to reduce their personal risks by shortening their firm's long-term orientation. For example, a survey conducted by Graham, Harvey, and Rajgopal (Reference Graham, Harvey and Rajgopal2005) provided evidence that 78% of the surveyed executives preferred to sacrifice projects with positive net present value (NPV) if adopting them resulted in the firm missing its quarterly earnings expectations. In contrast, CEOs with a high level of other-regarding values believe that personal needs can be met by working towards organizational and collective ends. Therefore, they make maximizing firm performance their first priority and accordingly pursue their firm's long-term development. For example, Ling et al. (Reference Ling, Zhao, Baron, Ling, Zhao and Baron2007) found that CEOs with other-regarding values care more about the opinions and gains of others and initiate more cooperation with other firms.
We further propose that strategic decisions with a long-term orientation are conducive to firm innovation. First, long-term orientation enables firms to aim for future returns and recognize the potential value of R&D investments. Flammer and Bansal (Reference Flammer and Bansal2017) found that an exogenous increase in long-term orientation leads to increased engagement in stakeholder relations, which is a critical resource for long-term success. Wang and Bansal (Reference Wang and Bansal2012) also argued that firms with a long-term orientation are more likely to draw value from stakeholder relationships that is implicit and difficult to identify. Second, firms with a long-term orientation can be seen as ‘probing into the future’ (Brown & Eisenhardt, Reference Brown and Eisenhardt1997: 16) and thus have more motivation to invest in innovation. Third, long-term orientation facilitates core strategic changes (Nadkarni & Chen, Reference Nadkarni and Chen2014; Ocasio, Reference Ocasio2011) and thus promotes the persistence and implementation of innovation. In contrast, short-term orientation often disrupts, delays, or even terminates innovation efforts. For example, Aghion, Van Reenen, and Zingales (Reference Aghion, Van Reenen and Zingales2013) demonstrated that shareholders with short-term orientation impede firm innovation.
Finally, we argue that strategic temporal orientation plays a mediating role in the relationship between CEO values and firm innovation. CEO values per se do not contribute to firm innovation, but the direction and activation of such values build the individualized constructs of CEOs regarding the environment (Hambrick & Mason, Reference Hambrick and Mason1984), infusing strategic decisions with a long-term orientation at the organizational level (Finkelstein et al., Reference Finkelstein, Hambrick and Cannella2009). These decisions then determine the direction of resource allocation and innovation efforts (Bower, Reference Bower1986; Flammer & Bansal, Reference Flammer and Bansal2017). This line of logic is consistent with the framework of the CEO ‘strategy formulator’ effect, which posits that CEOs produce effects through formulating firm strategies (Boone, Brabander, & Witteloostuijn, Reference Boone, Brabander and Witteloostuijn1996). Thus, long-term orientation primarily stems from strategies as conduits that channel the preferences and values of top managers (Finkelstein et al., Reference Finkelstein, Hambrick and Cannella2009) and that implement corporate policies in a top-down manner (Bower, Reference Bower1986; Floyd & Wooldridge, Reference Floyd and Wooldridge2000). This suggests that CEO values may not have a direct effect on firm innovation, but that they instead influence firm innovation via firms’ long-term orientation.
Some research, despite not focusing on the relationship between CEO values and firm innovation, has found that strategic decisions with different attributes mediate the relationship between CEO attributes and firm outcomes. For example, Nadkarni and Herrmann (Reference Nadkarni and Herrmann2010) found that strategic decisions with flexibility mediate the effects of CEO personality on firm performance. Boone et al. (Reference Boone, Brabander and Witteloostuijn1996) showed that strategic decisions between cost leadership and product differentiation mediate the relationship between the CEO locus of control and firm performance. Thus, we propose the following hypothesis:
Hypothesis 2:
Long-term orientation will mediate the relationship between CEOs’ self-regarding values and firm innovation.
The effects of managerial characteristics can only have an influence on strategic decisions if job demands are at a relatively high level (i.e., the degree to which CEOs experience their jobs as challenging; Hambrick et al., Reference Hambrick, Finkelstein and Mooney2005). CEOs are bounded rational. The higher the executive job demands are, the more remote strategic rationality becomes. In such circumstances, CEOs tend to economize their strategic decision-making process by relying on their psychological dispositions, which include their individual values (Hambrick & Mason, Reference Hambrick and Mason1984). Following this line of logic, we build a set of moderators at the individual (CEO tenure), organizational (CEO duality), and environmental (uncertainty) levels, which may either increase or decrease job demands and thus generate variations in the main relationship. In the following sections, we develop the boundary conditions that moderate the effect of CEO values on firm innovation via long-term orientation.
The Moderating Role of CEO Tenure
According to Hambrick and Fukutomi (Reference Hambrick and Fukutomi1991), ‘there are discernible phases, or seasons, within an executive's tenure in a position, and [those] seasons give rise to distinct patterns of executive attention, behavior, and ultimately, organizational performance’ (719). In relation to our context, CEO tenure may affect the main relationship for two reasons.[Footnote 1] First, as tenure becomes longer, CEOs tend to increase their identification with their firms (Dukerich, Golden, & Shortell, Reference Dukerich, Golden and Shortell2002) and align their own interests with those of the firms (Simsek, Reference Simsek2007; Wu et al., Reference Wu, Levitas and Priem2005). CEOs with longer tenures encourage a merging of the individual ego and their firms’ goals (Donaldson & Davis, Reference Donaldson and Davis1991). Boivie, Lange, Mcdonald, and Westphal (Reference Boivie, Lange, Mcdonald and Westphal2011) empirically showed that CEO tenure enhances organizational identification, which further reduces the managerial mischief of CEOs that could harm the firm. Thus, a long tenure mitigates the tendency of CEOs to project their self-preserving motives onto their decision-making, resulting in an attenuated principal-agent contradiction.
Second, the level of job demands decreases as CEO tenure increases, due to learning effects (Nelson & Winter, Reference Nelson and Winter1982; Wu et al., Reference Wu, Levitas and Priem2005). CEOs with longer tenures have better relationships with employees and are more deeply embedded in the networks of other crucial stakeholders (Luo, Kanuri, & Andrews, Reference Luo, Kanuri and Andrews2014), increasing the availability of resources that can be used for solving problems. Learning from past success and failure also enriches the managerial experience of long-tenured CEOs (Wu et al., Reference Wu, Levitas and Priem2005). Thus, the increased knowledge and power render long-tenured CEOs more capable of dealing with their roles (Ahuja et al., Reference Ahuja, Lampert and Tandon2008), leading to reduced job demands. When job demands are lower, CEOs do not need to economize their strategic decision-making, and they rely on their personal values to a lesser extent. We thus propose the following hypothesis:
Hypothesis 3:
CEO tenure will moderate the indirect relationship between CEOs’ self-regarding values and firm innovation via long-term orientation, in such a way that the indirect negative effect will be weaker when CEO tenure is longer.
The Moderating Role of CEO Duality
CEO duality occurs when a firm's CEO also chairs its board (Hayward & Hambrick, Reference Hayward and Hambrick1997; Mizruchi, Reference Mizruchi1983), which is a significant structural feature of organizations in both the US and China (Boyd, Reference Boyd1995; Donaldson & Davis, Reference Donaldson and Davis1991; Li & Tang, Reference Li and Tang2010). There are conflicting views regarding the effects of CEO duality. According to agency theory, independent leadership structure (i.e., the absence of CEO duality) is important for board monitoring. When CEO duality is present, board vigilance is weak, and so is the monitoring of board members (Boivie, Bednar, Aguilera, & Andrus, Reference Boivie, Bednar, Aguilera and Andrus2016; Boyd, Reference Boyd1995). Agency problems then become more severe, and a chair-CEO is more likely to advance his or her personal preferences in a relatively unchecked manner (Li & Tang, Reference Li and Tang2010), such as through the decoupling of high CEO pay and generous perquisites from firm performance (Boivie et al., Reference Boivie, Lange, Mcdonald and Westphal2011). Following this line of logic, we observe that CEO duality strengthens the negative relationship between CEO values and firm innovation via long-term orientation.
However, CEO duality also suggests consolidated leadership, which is essential in achieving unity of action, coordination, and focus (Boyd, Reference Boyd1995). A strong and unified CEO/chairperson leadership can help firms adapt to environmental demands (Boyd, Reference Boyd1995), thus increasing their legitimacy and sending a signal to stakeholders that a firm has a clear sense of direction (Salancik & Meindl, Reference Salancik and Meindl1984). CEO duality also implies internal efficiencies through a unity of command and eliminates potential conflict between the CEO and the board chair, and between principals and agents on management teams (Daily & Dalton, Reference Daily and Dalton1997). CEO duality thus facilitates task completion and reduces the demands on the CEO. This reduced level of job demands attenuates the need to rely on personal values to make decisions. We should therefore find a weakened negative relationship between CEO values and firm innovation via long-term orientation.
Both claims relating to CEO duality and separation are reasonable, and Finkelstein and D'Aveni (Reference Finkelstein and D'Aveni1994) noted that both CEO duality and separation are favorable, depending on the circumstances (see also Boyd, Reference Boyd1995), so we develop competing hypotheses for the moderating effects of CEO duality:
Hypothesis 4a(4b):
CEO duality will moderate the indirect relationship between CEOs’ self-regarding values and firm innovation via long-term orientation, in such a way that the indirect negative effect will be stronger (weaker) when CEO also chairs its board.
The Moderating Role of Environmental Uncertainty
Environmental uncertainty captures the extent to which a technological or a market environment is changing rapidly and in unpredictable ways (Russell & Russell, Reference Russell and Russell1992). It describes the perceived inability to understand the direction of environmental changes and the potential impact of those changes on organizations (Milliken, Reference Milliken1987). It also reflects the extent to which CEOs need to make decisions in an unpredictable environment, which is significant to the survival of their firms (Finkelstein & Boyd, Reference Finkelstein and Boyd1998). The competition in the market becomes more unstable and unpredictable when environmental uncertainty is high (Ferrier, Reference Ferrier2001). The demand on CEOs appears to increase to a large extent. However, uncertainty also increases the level of means-ends ambiguity (Li & Tang, Reference Li and Tang2010), making it more difficult for owners to attribute the outcomes to their causes and to evaluate the link between firm performance and CEOs’ task completion.
An uncertain environment gives more freedom to CEOs to attribute their success and failure. Poor performance is more likely to be attributed to the external environment, while good performance is more likely to be attributed to their own managerial decisions (Clapham & Schwenk, Reference Clapham and Schwenk1991). When uncertainty is higher, the possibility of not being accountable for outcomes is also higher, with self-regarding CEOs tending to care less about the temporal orientation of strategic decisions. The stress and anxiety of failure may also constrain the range of a firm's attention, centralize control, and induce rigidity (D'Aunno & Sutton, Reference D'Aunno and Sutton1992). CEOs may then shift their attention to simplistic efficiency concerns in the internal environment (D'Aveni & MacMillan, Reference D'Aveni and MacMillan1990). Such centralized power makes it even easier for CEOs to complete their tasks, leading to reduced job demands.
We thus hypothesize the following:
Hypothesis 5:
Environmental uncertainty will moderate the indirect relationship between CEOs’ self-regarding values and firm innovation via long-term orientation in such a way that the indirect negative effect will be weaker when environmental uncertainty is higher.
METHODS
Data and Sample
This project was implemented through collaboration with the China Entrepreneurs Survey System (CESS), an organization under the State Council of China. Since 1992, endorsed by the State Council and with its help in improving response rate, CESS has carried out annual surveys of Chinese CEOs with the intention to understand the problems that firms encounter when facing market transitions, dynamic environments, and technological competition. The firms surveyed constitute a stratified random sample based on industry, location, ownership, and size. The data collected through the system has been previously used to investigate the influence of CEO hubris (Li & Tang, Reference Li and Tang2010), CEO self-enhancement versus self-transcendent values (Fu et al., Reference Fu, Tsui, Liu and Li2010), and CEO humility (Ou et al., Reference Ou, Tsui, Kinicki, Waldman, Xiao and Song2014) on firm behavior and performance.
We collaborated with CESS from 2014 to 2016 to design the questionnaires and collect data for the three phases. The sample frame for this study comprised more than 10,000 Chinese industrial and service firms stratified randomly in 2014. From August to October, CESS mailed questionnaires to these firms, and 2,446 usable responses were returned (out of 2,458).[Footnote 2] We used the Kolmogorov-Smirnov (K-S) test and Heckman selection models to assess sample representativeness. The K-S test showed that the respondents were not significantly different than the non-respondents in terms of any of the continuous variables measured with archival data such as firm age, firm size, return on assets, CEO age, and CEO tenure (see the description of control variables below). The t-test also revealed that the respondents were not significantly different than the non-respondents in terms of the dichotomous variables included in this study. The Heckman model, in which the dependent variable was the likelihood of responding to the survey, and the independent variables included archival data and those data that described survey characteristics (e.g., questionnaire return time), suggested that nonresponse bias was not a serious concern in our study.
We then applied a longitudinal design to collect information from the same 2,446 firms. From August to October 2015, CESS received 1,084 responses from the 2,446 firms, giving an effective response rate of 44.3%. The data used in this study were derived from the information collected in the survey. In the first phase in 2014, we collected individual-level data on CEOs’ demographic characteristics, cognitive incentives, and firm information such as firm background and performance. In the second phase in 2015, we designed questions focusing on firms’ innovation-related strategies. From August to October 2016, we followed these firms to collect more information for our robustness check. The time lags between the phases enabled us to mitigate the concern of reverse causality between CEO values, firm decision-making, and innovation strategies.
The focus of our study is firm innovation, so we only used respondents from manufacturing industries (accounting for more than 70% of the full sample each year). Manufacturing industries in China were further classified into 31 categories according to the new Classification of National Economic Industries (GB/T4754-2011, National Bureau of Statistics of China, 2011), and the number of firms in our sample ranged from 5 to 91, with an average of 25 firms from each industry. After screening and excluding missing values, 780 firms were retained. Given the assumption of the agency problem, we only focused on CEOs who were not the founders of focal firms (56%). Therefore, after excluding those respondents with significant missing values, the final effective sample for our study comprised 436 firms participating in both survey phases in 2014 and 2015.
Table 1 and Figure 2 show the geographic distribution of 436 firms, among which firms from Jiangsu, Zhejiang, and Shandong account for the largest part of our sample. Table 2 reports the industry distribution of sample firms.

Figure 2. Geographic distribution of the final sample across different provinces in China

Figure 3. Scatterplot for the direct effect of CEO self-regarding values on Firm long-term orientation and Firm innovation (N = 436)
Table 1. Geographic distribution of the final sample (N = 436)

Table 2. Industry distribution of the final sample (N = 436)

Measures
Dependent variable
Firm innovation
This was measured by the extent to which firms plan to increase or reduce their R&D input in the following year. In this survey, firm respondents were asked to rate the following items on a scale ranging from 1 (a large decrease) to 5 (a large increase): (1) general innovation input; (2) the ratio of R&D input to sales revenue; and (3) the ratio of employee training input to sales revenue. All three items were loaded on a single factor with a Cronbach's alpha of 0.87. Firm innovation was measured as the average of the three items that were collected in the second-phase survey. In the robustness tests, we also complemented the main analyses of firm innovation with an objective measure of new product sales. The details are given below.
Independent variables
CEOs’ self-regarding values
These involve the pursuit of success, happiness, and wealth for oneself and one's family, while other-regarding values emphasize enhancing the success and happiness of others. Following past research (Rokeach, Reference Rokeach1972), we used seven items related to self-regarding or other-regarding values and asked the CEOs to respond. The seven items are (1) a comfortable life (a prosperous life), (2) helpful (working for the welfare of others), (3) compassion (feeling empathy for others), (4) wealth (making money for oneself and one's family), (5) equality (brotherhood, equal opportunity for all), (6) loving (being affectionate, tender), and (7) pleasure (an enjoyable life). The CEOs were asked to rate all seven items on a scale ranging from 1 (strongly disagree) to 5 (strongly agree). Items 1, 4, and 7 represented self-regarding values, while items 2, 3, 5, and 6 reflected other-regarding values.
Table 3 gives the factor loadings of the seven items. We found that the items were separately loaded on two factors. Items 2, 3, 5, and 6 were loaded on factor 1, which measured other-regarding values and explained 34% of the total variance, while items 1, 4, and 7 were loaded on factor 2, which measured self-regarding values and explained 25% of the total variance. Following previous research (Agle et al., Reference Agle, Mitchell and Sonnenfeld1999), we first took the average of items representing self- and other-regarding values separately, with Cronbach's alphas of 0.63 and 0.75 respectively, and then used the difference between them as the indicator of CEOs’ self-regarding values. The data of CEO values were collected in the first-phase survey.
Table 3. Measurement and validity of CEO values (N = 436)

Firms’ long-term orientation
This reflects a firm's tendency to prioritize the long-range implications and effects of decisions and actions that come to fruition after an extended period of time (Lumpkin et al., Reference Lumpkin, Brigham and Moss2010). Following Wang and Bansal (Reference Wang and Bansal2012), our respondents were asked about the approximate time frame in which they make future strategic plans using a 6-point scale (0 = no future plans, 1 = for temporal issues, 2 = for the next 3 years, 3 = for the next 5 years, 4 = for the next 10 years, and 5 = more than the next 10 years). The data were collected in the second-phase survey.
CEO tenure
To measure CEO tenure, we asked the CEOs how many years they had been working in their current position. We then calculated CEO tenure by taking the logarithm of the number of years since the CEO took the position.
CEO duality
This refers to the practice whereby a single person serves both as the CEO and the chairman of a firm. Accordingly, we coded CEO duality as ‘1’ if the CEO is also the chairman of the board, and ‘0’ otherwise.
Environmental uncertainty
This refers to the extent to which firms are surrounded by unpredictable and unstable environments (Finkelstein & Boyd, Reference Finkelstein and Boyd1998; Hambrick & Abrahamson, Reference Hambrick and Abrahamson1995). In this study, environmental uncertainty was measured based on firm respondents’ evaluation of the industrial environment in which they operated, in terms of the following aspects: (1) the intensity of competition, (2) bursts of new entrants, (3) technology dynamics, (4) technology importance, and (5) the speed of technology or product renewal (Desarbo, Benedetto, Song, & Sinha, Reference Desarbo, Benedetto, Song and Sinha2005; Voss & Voss, Reference Voss and Voss2000). Similarly, each item was rated on a scale ranging from 1 (strongly disagree) to 5 (strongly agree), and the five items were loaded on a single factor with a Cronbach's alpha of 0.68. Environmental uncertainty was measured as the average of the five items, the data of which were collected in the first-phase survey.
Control variables
To rule out alternative explanations, a set of variables on individual, firm, industry, and regional levels were included in our models. CEO age and CEO education were included because the literature has shown that CEOs’ demographic characteristics influence their decision-making (e.g. MacCrimmon & Wehrung, Reference MacCrimmon and Wehrung1990). We transformed CEO age by taking the logarithm to achieve normality. CEO education was measured by a categorical variable ranging from 1 (below high school) to 6 (doctorate), and each number indicated an ascending level of formal education.[Footnote 3] We also controlled for CEO ownership by the percentage of shares owned by the CEO because ownership is considered relevant to managers’ incentives to make long-term investments (Li et al., Reference Li and Tang2010). CEO gender was not included in our model because only 2.8% of the 436 CEO respondents were female.
We also controlled for variables that may affect firm innovation at the firm level. Firm age and Firm size, coded as the logarithm-transformed number of years from the founding year of a firm to 2014, were included, along with the number of employees. TMT size was also controlled in this study, as was a firm's previous R&D investment and innovation performance, to rule out alternative explanations. Specifically, we controlled R&D intensity, measured as the R&D investment percentage of total sales, and New product sales in the prior year to rule out the effect of a firm's historical innovation performance.
Firm innovation depends substantially on the region in which the firm is located. We included dummy variables representing different Provinces in China. All of the control variables at individual, firm, and regional levels were collected in the first-phase survey.
Construct Validity
We used confirmatory factor analyses (CFA) to assess the validity of the multi-item measures. We ran a confirmatory factor analysis for firm innovation, CEOs’ self-regarding values, other-regarding values, and environmental uncertainty. The 4-factor 15-item model reached a good model fit: χ2 = 234.546, d. f. = 84, RMSEA = 0.050, CFI = 0.945. Therefore, the overall measures met the recommended values, and the standardized item loadings on the hypothesized factors were significant. We examined internal consistency and convergent validity by calculating the construct reliability and average variance extracted (AVE). The results indicated that composite reliabilities and AVEs for all the multi-item constructs were greater than 0.70 and 0.50, respectively. Thus, we obtained adequate internal consistency and convergent validity.
We examined the discriminant validity of the measures in two ways (Anderson & Gerbing, Reference Anderson and Gerbing1988). First, we found that the AVE estimates of the two constructs were less than the square of the parameter estimate between the two latent variables, and the correlation among the indicators of each construct was greater than that between a construct and any other construct. Second, we further checked discriminant validity by using chi-square difference tests in which the correlation between pairs of constructs is freely estimated once and then constrained for unity. The hypothesized model performed better than all of the other competing models. These results show that our measurements have satisfactory reliability and validity.
Assessment of Common Method Biases
We took several steps to minimize the effect of common method variance (CMV) bias (Chang, van Witteloostuijn, & Eden, Reference Chang, Van Witteloostuijn and Eden2010; Podsakoff, MacKenzie, Lee, & Podsakoff, Reference Podsakoff, MacKenzie, Lee and Podsakoff2003). First, we collected data for independent and dependent variables with a one-year lag, which can reduce CMV bias to a large extent. Second, we improved the scale items by using multiple item constructs and different scale formats for predictors and criterion measures. Third, we counterbalanced the survey question order. Fourth, we used a linear regression model with interaction effects, which can reduce the likelihood of CMV because the respondents are unlikely to be guided by a cognitive map that includes difficult-to-visualize interactions. Finally, following Harman's single-factor test, we conducted a confirmatory factor analysis (CFA) with all of the variables used in this study, and the results showed that the single factor model did not fit well. We also performed an exploratory factor analysis (EFA) with all of the variables that yielded five factors with eigenvalues greater than one. The largest factor explained only 17.20% of the variance, and these test results confirmed that CMV bias did not pose a serious problem in this study.
To test the mediating effect of firms’ long-term orientation on the relationship between CEOs’ self-regarding values and firm innovation, we applied the bootstrapping method recommended by MacKinnon, Lockwood, and Williams (Reference MacKinnon, Lockwood and Williams2004) to generate asymmetric confidence intervals (CIs) for the indirect relationship. This approach generated a more accurate estimate of the indirect relationship because it produced asymmetric confidence intervals using the respective distributions of the two regression coefficients that comprised the term. Furthermore, to test the moderating effects of CEO tenure, CEO duality, and environmental uncertainty on this indirect relationship, we adopted the bootstrapping procedure recommended by Edwards and Lambert (Reference Edwards and Lambert2007).
RESULTS
Given that Firm innovation and Long-term orientation are both continuous variables with normal distribution, we used OLS regressions to examine our main hypothesis. All of the models were estimated using STATA Version 13.
Table 4 presents descriptive statistics and correlations for the variables. The key independent variable, CEOs’ self-regarding values, was distributed between −4.000 and 1.083, which was consistent with our assumption that the self-interest dimension of CEO values varied widely among individuals. We found from the table that CEOs’ self-regarding values were negatively correlated with firms’ long-term orientation and firm innovation, which also confirmed our theoretical hypotheses. A further inspection of the correlations did not reveal any serious multicollinearity, with a mean variance inflation factor (VIF) for each variable of less than 2. We mean-centered the variables involved in the interaction terms by subtracting the mean from each value (Aiken & West, Reference Aiken and West1994) to avoid possible collinearity among the interaction terms.
Table 4. Summary statistics and correlations (N = 436)

Note: |r|>0.08 is significant at p < 0.05.
Table 5 presents the OLS estimation results of CEOs’ self-regarding values and firms’ long-term orientation on firm innovation. Model 1 includes only control variables. The coefficients of control variables in Model 1 provide some implications. For example, the results showed that CEO ownership has a positively significant effect on firm innovation, confirming the argument of agency theory that the potential managerial mischief may be mitigated by the increasing stock shares hold by each CEO. Consistent with the findings in previous literature, the results showed that firm size has a positive and significant effect on firm innovation, while firm age is negatively related to firm innovation. CEO education has a positive and marginal significant effect on firm innovation. As to the relationship between new product sales (R&D intensity) and firm innovation, we proposed conflicting arguments along this effect. Because our dependent variable, firm innovation, was measured as the extent to which firms plan to increase or reduce their R&D input in the following year, previous remarkable performance (or R&D efforts) on one hand lays a basis encouraging future investment towards innovation, but on the other hand may constrain subsequent increase in investment.
Table 5. OLS regression of firm innovation on CEO self-regarding values (N = 436)

Notes: Province dummies were included in all models. Standard errors are in parentheses (se). P-values are in parentheses [p].
Consistent with our basic line prediction, the results of Model 2 in Table 5 indicated that CEOs’ self-regarding values had a significantly negative relationship with firm innovation (b = −0.138, p = 0.030), supporting Hypothesis 1. Holding other variables constant, a one-unit increase in CEOs’ self-regarding values resulted in a 0.138-unit decrease in firm innovation.
To test the mediating role of firms’ long-term orientation (Hypothesis 2), we first regressed long-term orientation on CEOs’ self-regarding values, and firm innovation on long-term orientation, respectively. Consistent with our predictions, Model 2 in Table 6 indicated that CEOs’ self-regarding values were negatively related to long-term orientation (b = −0.225, p = 0.005), and Model 3 in Table 5 showed a positive and significant effect of long-term orientation on firm innovation (b = 0.159, p = 0.001). After including both CEOs’ self-regarding values and firms’ long-term orientation in the regression (Model 4 in Table 6), the coefficient of the relationship between CEOs’ self-regarding values and firm innovation changed significantly from −0.138 in Model 2 to −0.106 in Model 4, which suggested that the negative relationship effect of CEOs’ self-regarding values on firm innovation was no longer significant after including long-term orientation in our model. Together, the results in Table 5 and Table 6 suggested that firms’ long-term orientation fully mediated the negative relationship between CEOs’ self-regarding values and firm innovation. A one-unit increase in CEOs’ self-regarding values led to a 0.225-unit decrease in long-term orientation, and a one-unit decrease in firms’ long-term orientation led to a 0.159-unit decrease in firm innovation. Hypothesis 2 was therefore supported.
Table 6. OLS regression of firm long-term orientation on CEO self-regarding values (N = 436)

Notes: Province dummies were included in all models. Standard errors are in parentheses (se). P-values are in parentheses [p].
Models 3–6 in Table 6 showed the moderating roles of CEO tenure, CEO duality, and environmental uncertainty on the indirect negative relationship between CEOs’ self-regarding values and firms’ long-term orientation. Specifically, in Model 3 we found that CEO tenure weakened the negative effect of CEOs’ self-regarding values on firms’ long-term orientation (b = 0.326, p = 0.005), which supported Hypothesis 3. A one-unit increase in CEO self-regarding values resulted in a 0.065-unit decrease in firm innovation via long-term orientation at low level of CEO tenure, and a 0.014-unit decrease at high level of CEO tenure. The difference was a 0.052-unit change in firm innovation.
As shown in Model 4, CEO duality also weakened the negative relationship between CEOs’ self-regarding values and long-term orientation (b = 0.338, p = 0.048), which was consistent with Hypothesis 4b rather than with Hypothesis 4a. A one-unit increase in CEO self-regarding values resulted in a 0.053-unit decrease in firm innovation via long-term orientation when CEO duality equals 0, and a 0.010-unit decrease when CEO duality equals 1. The difference was a 0.043-unit change in firm innovation.
The coefficient of the interaction term between environmental uncertainty and CEOs’ self-regarding values on long-term orientation showed that environmental uncertainty mitigated the negative effect of CEOs’ self-regarding values (b = 0.291, p = 0.023), supporting Hypothesis 5. A one-unit increase in CEO self-regarding values resulted in a 0.059-unit decrease in firm innovation via long-term orientation at low level of environmental uncertainty, and a 0.007-unit decrease at high level of environmental uncertainty. The difference was a 0.052-unit change in firm innovation.
We then applied Edwards and Lambert's (Reference Edwards and Lambert2007) procedure to examine the mediation effect and the moderated mediation relationship (i.e., the first-stage moderation model). The results based on the bootstrapping test in Table 7 supported Hypothesis 2 that firms’ long-term orientation mediated the relationship between CEOs’ self-regarding values and firm innovation (b = −0.034, p = 0.052) with a bias-corrected confidence interval excluding zero [−0.086, −0.010]. The difference between the indirect effect of CEOs’ self-regarding values at high and low levels (Mean+/− S.D.) of CEO tenure was positively significant (b = 0.052, p = 0.062), with a bias-corrected confidence interval of [0.009, 0.131], which provided additional support for Hypothesis 3. Similarly, the difference between the indirect effect of CEOs’ self-regarding values with and without (1/0) CEO duality was positive and significant (r = 0.043, p = 0.079), and the bias-corrected confidence interval was [0.011, 0.111], which supported Hypothesis 4b. The negative indirect effect of CEOs’ self-regarding values on firm innovation was also weakened at a high level of environmental uncertainty (b = 0.052, p = 0.058), with the bias-corrected confidence interval of [0.013, 0.127], which supported Hypothesis 5.
Table 7. The indirect effect of CEO self-regarding values on firm innovation via long-term orientation at different levels of moderators

Notes: Table 7 indicates that the negative indirect effect of CEO self-regarding values on firm innovation via long-term orientation was weaker at high (r = −0.014, p = 0.444) than low (r = −0.063, p = 0.022) level of CEO tenure, and the difference between the two effects was significant (r = 0.052, p = 0.062). In terms of CEO duality, the negative indirect effect was weaker when a CEO chairs the board (r = −0.010, p = 0.506) than not (r = −0.053, p = 0.047), and the difference between the two effects was significant (r = 0.043, p = 0.079). Similarly, the negative indirect effect was weaker at high (r = −0.007, p = 0.666) than low (r = −0.059, p = 0.047) level of environmental uncertainty, and the difference between the two effects was also significant (r = 0.052, p = 0.058).
Table 8 decomposes the effect of the moderators on the direct, indirect, and the total effects in the mediation models. The results show that the differences between the indirect effects of CEOs’ self-regarding values on firm innovation at high and low levels of CEO tenure, CEO duality, and environmental uncertainty were all significant. Specifically, regarding the moderating effect of CEO tenure, the results showed that the indirect effect of CEO self-regarding values on firm innovation is significant only at low level of CEO tenure, but is insignificant when CEO tenure is high. The insignificance could be explained by agency theory. As CEOs with longer tenures encourage a merging of individual ego and the organizational goals (Donaldson & Davis, Reference Donaldson and Davis1991), the aligned interest further mitigates the impact of CEO values. Moreover, the level of job demand decreases as CEO tenure increases (Nelson & Winter, Reference Nelson and Winter1982; Wu et al., Reference Wu, Levitas and Priem2005), which results in inhibited influence of CEOs on firm decision making.
Table 8. Effect decomposition results of the indirect effect of CEO self-regarding values on firm innovation

Note: P-values are in parentheses [p].
In terms of the moderating effect of CEO duality, the results suggested that the indirect effect of CEO self-regarding values on firm innovation via long-term orientation is insignificant when the CEO also chairs the board. As we noted earlier, previous views on the effect of CEO duality are inconclusive. When CEO duality is present, agency problems become more severe, we therefore observed that CEO duality strengthens the negative relationship between CEO self-regarding values and firm innovation. However, CEO duality also suggests consolidated leadership, which facilitates task completion and reduces the demands on the CEO. The reduced job demand attenuates the need for CEOs to rely on personal values to make decisions. Following this argument, we should find a weakened negative relationship between CEO values and firm innovation via long-term orientation. The two competing logics may offset each other and result in a nonsignificant effect of CEO values on firm innovation for those firms under consolidated leadership.
Regarding the moderating role of environmental uncertainty, the results indicated that the indirect effect of CEO self-regarding values on firm innovation is insignificant when environmental uncertainty is high. To explain for this insignificance, we suggested that when uncertainty is high, CEOs with high self-regarding values can easily attribute potential poor performance to the external environment, making themselves less likely to be accountable. In this situation, CEOs will not rely on their personal values to make decisions. Therefore, the indirect effect of CEO self-regarding values on firm innovation via long-term orientation is insignificant.
In Figure 4–6, we used scatterplots to present the dispersion of the three moderators. Figure 4 plots the moderating effect of CEO tenure on the relationship between CEO self-regarding values and firm long-term orientation. The figure suggested that for cases with high CEO tenure, firm long-term orientation is less likely to be influenced by CEO self-regarding values. However, for cases low in CEO tenure, CEO self-regarding values have a more significant influence on firm long-term orientation. Similarly, Figure 5 shows that firm long-term orientation is more sensitive to the change of CEO self-regarding values for cases in which CEO duality equals 0 than 1; Figure 6 suggests that for cases facing high environmental uncertainty, CEO self-regarding values are less likely to influence firm long-term orientation.

Figure 4. Scatterplot for effect of CEO self-regarding values on Firm long-term orientation, moderated by CEO tenure (N = 436)

Figure 5. Scatterplot for effect of CEO self-regarding values on Firm long-term orientation, moderated by CEO duality (N = 436)

Figure 6. Scatterplot for effect of CEO self-regarding values on Firm long-term orientation, moderated by Environmental uncertainty (N = 436)
In Figures 7–9, we plotted the moderating effects of CEO tenure, CEO duality, and environmental uncertainty on the indirect relationship between CEOs’ self-regarding values and firm innovation, respectively. Figure 7 suggests that firm innovation decreased less with CEOs’ self-regarding values when the CEO has a long (Mean + S.D.) versus a short (Mean - S.D.) position tenure. Similarly, Figure 8 shows that firm innovation was less affected by CEOs’ self-regarding values when the CEO also served as the chairman (CEO duality = 1). Figure 9 shows that the negative effect of CEOs’ self-regarding values on firm innovation was weakened when environmental uncertainty was high (Mean + S.D.) versus low (Mean - S.D.). Thus, Hypothesis 3, 4b, and 5 were further supported by our results.

Figure 7. The indirect effect of CEO self-regarding values on firm innovation via firm long-term orientation at low and high levels of CEO tenure

Figure 8. The indirect effect of CEO self-regarding values on firm innovation via firm long-term orientation with and without CEO duality

Figure 9. The indirect effect of CEO self-regarding values on firm innovation via firm long-term orientation at low and high levels of environmental uncertainty
Robustness Tests
We conducted several robustness tests to explore the sensitivity of our results. First, we combined the data collected in 2016 to test whether the effect of CEOs’ self-regarding values on firm innovation still holds. In 2016, we surveyed all of the sample firms once again and asked the CEOs of these firms to respond to the questions on their firms’ R&D activities. We obtained a sample of 235 firms after combining three-wave data and excluding observations with missing values in the key variables.[Footnote 4]Table 9 presents the regression results of CEOs’ self-regarding values (collected in 2014) and firms’ long-term orientation (collected in 2015) on firm innovation (collected in 2016). The results again confirmed our hypotheses that CEOs’ self-regarding values were negatively related to firms’ long-term orientation (b = −0.316, p = 0.001), and that firms’ long-term orientation had a positive effect on firm innovation (b = 0.106, p = 0.053). The mediating role of long-term orientation still held, as the main effect of CEOs’ self-regarding values on firm innovation became insignificant after controlling for firms’ long-term orientation.
Table 9. Robustness test – Firm innovation in 2016 as the dependent variable (N = 235)

Notes: Region dummies were included in all models. We used region dummies rather than province dummies due to the restriction of the sample size after merging the data in 2016. Standard errors are in parentheses (se). P-values are in parentheses [p].
Second, we used an objective measure of firm innovation, New product sales, to examine the relationship between CEOs’ self-regarding values and new product sales. New product sales was measured as the ratio of new product sales to total sales. The results in Table 10 show that CEOs’ self-regarding values were negatively related to firms’ long-term orientation (b = −0.214, p = 0.007) and new product sales (b = −0.039, p = 0.003), and that firms’ long-term orientation was positively related to new product sales (b = 0.020, p = 0.058). In addition, after controlling for firms’ long-term orientation, the positive effect of CEOs’ self-regarding values on firm innovation became less significant, thus supporting our mediation hypothesis.
Table 10. Robustness test – New product sales as an objective measure of firm innovation (N = 436)

Notes: Province dummies were included in all models. Standard errors are in parentheses (se). P-values are in parentheses [p].
DISCUSSION
Innovation contributes to a firm's long-term competitive advantages but also involves significant risk and uncertainty. Executive managers thus show a complex mix of attitudes towards R&D investments. Drawing on agency theory, researchers have contributed considerable insight to our understanding of CEOs’ role in firm innovation. However, this theory fails to directly test its basic assumption and uncover the mechanism by which managerial self-interest affects innovation. In this study, we build a mediation model in which CEOs’ self-regarding values affect firm innovation via firms’ long-term orientation. We find that CEOs with high levels of self-regarding values are negatively related to innovation efforts measured as R&D investment, which is a relationship mediated by firms’ long-term orientation. In addition, CEO tenure, CEO duality, and environmental uncertainty moderate the main relationship.
These findings in our study make several contributions to the literature. First, we contribute to agency theory by providing direct evidence that managers with self-regarding values hurt firms’ long-term performance in areas such as innovation. The sharpest criticism made against agency theory is from Ghoshal (Reference Ghoshal2005), who claimed that it held ‘both unrealistic and biased assumptions’ (77) regarding human self-interest. The core idea of the theory has even been labeled ‘the dumbest idea in the world’ (Denning, Reference Denning2011). Indeed, agency theory is deductive because it is based on untested assumptions and focuses on designing governance mechanisms to address agency problems. Our finding that CEOs’ self-regarding values negatively affect both innovation efforts and outcomes provides preliminary empirical evidence for agency theory, suggesting that managers do indeed have self-preserving motives that exert negative effects on firm performance.
We also demonstrate that not all CEOs are self-interested and that they have other-regarding motives. The combination of self-regarding and other-regarding values contributes to the recent extension of agency theory. For example, some researchers have argued that managers’ self-interest is bounded by norms of reciprocity, justice, and fairness (Ariño & Ring, Reference Ariño and Ring2010; Bosse & Phillips, Reference Bosse and Phillips2016; Fong et al., Reference Fong, Misangyi and Tosi2010). According to our results, managers’ self-interest may not necessarily be bounded by their beneficial behavior toward firms. Managers only engage in such beneficial behavior if their other-regarding values are stronger than their self-regarding values. Our results also inform stewardship theory, which highlights that managers are not self-interested but rather intrinsically motivated and able to act as good stewards (Davis et al., Reference Davis, Schoorman and Donaldson1997; Donaldson & Davis, Reference Donaldson and Davis1991). We show that managers simultaneously have self- and other-regarding values and that the relative strength of the two kinds of values matters.
Second, we add a new attribute of CEO values to the recent empirical test of upper echelon theory. It is surprising that researchers have investigated so many attributes of CEOs while neglecting the role of CEO values, which is a root construct central to understanding human behavior (Parsons & Shils, Reference Parsons, Shils, Parsons and Shils1951). Finkelstein et al. (Reference Finkelstein, Hambrick and Cannella2009) urged that exploring the effects of executive values in strategic decisions ‘should be a high priority’ (58). Our study fills this gap. We further demonstrate that CEO values, as individual attributes, affect firm outcomes through the mediation of long-term orientation at the organizational level. The effect of long-term orientation supports the recent critiques on firm short-termism (e.g., Flammer & Bansal, Reference Flammer and Bansal2017; Martin et al., Reference Martin, Wiseman and Gomez-Mejia2016) and provides more empirical evidence of the CEO ‘strategy formulator’ effect (Boone et al., Reference Boone, Brabander and Witteloostuijn1996; Nadkarni & Herrmann, Reference Nadkarni and Herrmann2010). The literature has identified direct (e.g., Lumpkin et al., Reference Lumpkin, Brigham and Moss2010) and moderating (Wang & Bansal, Reference Wang and Bansal2012) mechanisms through which long-term orientation affects firm performance. We contribute to this line of research by identifying a new mediating mechanism.
The findings related to moderators also generate some interesting insights. For example, we find that CEO tenure weakens the relationship between CEO values and firm innovation, which provides empirical evidence for Hambrick and Fukutomi's (Reference Hambrick and Fukutomi1991) seasons of a CEO's tenure. It shows that as the tenure becomes longer, CEOs increasingly identify with their firms and become more capable of performing their jobs, which is in line with the findings of Boivie et al. (Reference Boivie, Lange, Mcdonald and Westphal2011). CEO duality also weakens the main relationship, supporting the logic of job demands but rejecting the reasoning of agency theory, which provides more fuel for the debate on governance mechanisms such as CEO duality (Boyd, Reference Boyd1995; Finkelstein & D'Aveni, Reference Finkelstein and D'Aveni1994). Agency theory posits that CEO duality increases agency problems, whereas stewardship theory argues that it facilitates the completion of CEO tasks. Donaldson and Davis (Reference Donaldson and Davis1991) find that CEO duality provides additional benefits irrespective of the effects of the long-term compensation firms use to align the interests of the CEO and shareholders. Our study confirms that CEO duality may reduce job demands; and therefore, CEOs need not economize by relying on their values in strategic decision-making.
Third, this study contributes to our understanding of the innovation problem, especially in the context of China. China is surging forward on the most recent wave of technological innovation (The Economist, 2017). However, Chinese firms, especially state-owned firms, engage disproportionately in minor rather than major innovation (Tong et al., Reference Tong, He, He and Lu2014). The evidence is quite surprising because innovation benefits firms’ long-term performance (Porter, Reference Porter1992; Tushman & O'Reilly, Reference Tushman and O'Reilly1996), and because the secret of China's economic success lies in its culture, which has the highest level of long-term orientation (Hofstede & Minkov, Reference Hofstede and Minkov2010). We suggest that firms’ short-term tendencies regarding innovation are the product of CEOs’ personal values; CEOs’ self-regarding values might be harmful to firm innovation because they shorten the temporal orientation of firms. By addressing these issues, we provide some new insight for understanding how managers matter in firm innovation, thus enriching the innovation literature.
This study has important implications for ongoing managerial practice. CEOs may consider learning more about their influence on firm performance and understand that over-emphasizing their self-interests may impede firm innovation. The effect size that a one-unit increase in CEO self-regarding values resulted in a 13.5% decrease in firm innovation indicates a practically significant effect for firms. Thus, executive managers need to balance their personal interests and organizational objectives, rather than pursuing individual achievement at the expense of the firm's long-term success. Regarding board members, the contingent effects of CEO tenure and duality may inspire them to reconsider the appointment and replacement of members of the top management team. While CEOs who have worked in a certain firm for a long time are generally considered to be rigid and reluctant to pursue innovation, our findings suggest that CEO tenure significantly mitigates the negative effects of self-regarding values on firm innovation. In other words, newly appointed CEOs are more likely to be affected by their personal values when making strategic decisions, which could be harmful for firm innovation if those CEOs were self-regarded. We also provide evidence for the benefit of CEO duality in reducing the negative influence of CEOs’ self-interest on firm innovation. CEOs serving as chairs incorporate less personal values of self-regarding into their firm's strategic decisions, and as a result they are less likely to impede their firm's long-term development and innovation. Based on our results, a one-unit increase in CEO self-regarding values resulted in a difference of 5.2% change in firm innovation via long-term orientation between firms with CEO duality and those without CEO duality. Such significnant effects remind us that corporate governance structure can effectively monitor management team.
Limitations and Future Research Directions
Our current findings call for further exploration of the complex effects of CEO and environmental characteristics on firms’ strategic decisions. First, as we have extended the assumption of agency theory by demonstrating the combination of self-regarding and other-regarding values, the mixed motives of CEOs are of great interest and may affect a firm's strategic decisions in areas beyond innovation. Future research could further explore the effect of CEOs’ self-regarding and other-regarding values on other key decisions such as investments, merge and acquisitions, goal setting, and strategies for dealing with stakeholders. Second, we found that environmental uncertainty weakens the negative effect of CEOs’ self-regarding values on firm innovation, which supports our argument that uncertainty reduces rather than increases executive job demands. Prior work has argued that higher uncertainty exerts significant threats to firm performance and thus increases job demands. The focus is the objectivity of environment and its demands on executive jobs. However, it is also compelling that higher uncertainty gives managers more space to justify and frame their actions, decreasing job demands. This view emphasizes the subjective nature of job demands and the acting roles of managers to frame such demands. Future research can gain more insights by simultaneously examining the two competing mechanisms of uncertainty. Third, the moderating effect of CEO duality is in agreement with the logic of the job demands perspective rather than with agency theory, which suggests that duality is an effective way to mitigate the negative effect of CEOs’ self-regarding values on firm innovation. However, the comparable effects of these competing logics may change across different contexts. Future research could benefit from more exploration in this regard. Finnally, alternative or competing theories are in need to explain the unexplained variance of managerial features. For example, competing logics are supposed to exist concurrently to explain the nonsignificant relationship between CEO age and firm innovation in this study. On one hand, CEO age enhances organizational identification and further strengthens the alignment of their own interests and those of the firms, which facilitates innovation investment. On the other hand, as CEOs are aging, they are supposed to perform more conservatively towards strategic changes, such as innovation. These conflicting effects may offset each other and result in a nonsignificant relationship between CEO age and firm innovation. Moreover, TMT size shows a nonsignificant impact on firm innovation, consistent with the empirical findings of Ozer and Zhang's study (Reference Ozer and Zhang2015). Therefore, we speculate that when making innovation decisions, what really matters is the power and discretion of the top manager team rather than size. We call for more explorations and alternative theories about the unexplained variances.
Our study has several limitations and also suggests many opportunities for future research. First, most firms in our sample are large and established, which may prevent the generalizability of our findings to small and medium-sized enterprises (SMEs). However, we believe that in SMEs the relationship between CEO values and firm innovation may be more significant because CEOs have more influence over firm behavior and outcomes. Further investigation into the role of CEO values in SMEs may offer more insight.
Second, although we collected data with a one-year lag, the data were cross-sectional in nature, and thus it is still difficult for us to confirm the causality between CEO values and firm innovation. Future research could use experimental methods to help confirm the direction of causality proposed in our study. In addition, CEO values were measured based on subjective evaluation, which can be biased. Further research should attempt to use objective and other-rating measures to alleviate such potential biases.
The third concern comes from the responses of CEOs, which may have the potential for single-source bias. However, we may reasonably infer that such bias was not a serious concern because most of our measures consist of factual information about the firms. This information is generally concrete and specific. However, we admit that measures of CEO values, environmental uncertainty, and firm innovation may have single-source bias. The significance of these interactions is unlikely to be an artifact of the single-informant method, so the single-source bias may not be a serious concern for CEO tenure and CEO duality, for example. Future research may improve the accuracy of our measurement by asking multiple respondents to evaluate CEO values, environmental uncertainty, and other subjective variables.
Finally, although China is an appropriate setting for testing the generalizability of theoretical constructs such as CEO values, we should acknowledge that a single country study may limit the applicability of the results to other contexts. Future research should examine the roles of CEO values in other firm strategic choices. We also suggest more research to test our findings in other national contexts such as those of the US, Japan, and Europe.