CEO Emotional Intelligence and Firms’ Financial Policies. Bayesian Network Method

The aim of this paper is to explore the determinants of firms’ financial policies according to the manager’s psychological characteristics. More specifically, it examines the links between emotional intelligence, decision biases and the effectiveness of firms’ financial policies. The article finds that the main cause of an organization’s problems is the CEO’s emotional intelligence level. We introduce an approach based on Bayesian network techniques with a series of semi-directive interviews. The research paper represents an original approach because it characterizes behavioral corporate policy choices in emerging markets. To the best of our knowledge, this is the first study in the Tunisian context to explore this area of research. Our results show that Tunisian leaders adjust their decisions (on investments and distributions) to minimize the risk of loss of compensation or reputation. They opt for decisions that minimize agency costs, transaction costs, and cognitive costs.


Introduction
Recent research has focused on the importance of emotion as it relates to intellectual abilities, particularly in organizations that evaluate employees' abilities in terms of emotions rather than cognition (Brackett et al., 2006).
The importance of emotional intelligence is emphasized because human relations in organizations are affected by emotional factors more than rational factors. The emotional quotient is as important as the intelligence quotient; indeed, the emotional intelligence of individuals who carry out duties and play essential roles in ensuring organizational outcomes is quite significant. Therefore, successful organizations require employees who can communicate effectively, control their emotions, and demonstrate their technical abilities (Fiori, 2009 Hackbarth, 2009;Hawkins, Hoch, & Meyers-Levy, 2001;Ho, 2009;Malmendier, Tate, 2010).
The presence of these heuristics effects such as emotional bias pushes individuals to acquire emotional skills and regulate their emotions. Indeed, this research has encouraged researchers to use the concept of emotional intelligence to augment the unsatisfactory notion of IQ and to consider skills in emotional regulation as real capital. Mayer and Salovey (1997) showed that emotional intelligence plays an important role not only in regulating and controlling emotions but also in developing intellectual and cognitive processes (Lopes et al, 2005;Song et al, 2010). According to Anderson (1983), the absence of this skill implies an uncertainty that may lead an individual such as a CEO to react conservatively or refuse to make any decision (Trautmann, Veider, & Wakker, 2009) that is likely to alter his current status.
Our idea for this research was inspired by the behavioral approach, and the paper aims to highlight the role of emotional intelligence in minimizing the behavioral biases (optimism, loss aversion and overconfidence) and improving CEO financial policies (choice of capital structure, investment decisions and dividend policies).
The article is structured as follows: Section 2 presents the related literature and the theories that motivated the work, and Section 3 discusses the empirical strategies that were adopted. Section 4 discusses the main results, and Section 5concludes.

Emotional intelligence concept (IE)
The study of the positive role of emotions in the decision-making process leads us to the concept of emotional intelligence (IE). Indeed, emotional intelligence is at the heart of the skills portfolio of an effective leader. Some authors even consider it a key driver of organizational performance (Côté et al., 2010;Goleman, 2001;Kilduff, Chiaburu, & Menges, 2011;Song et al., 2010). In this section, we review the literature on emotional intelligence. Salovey and Mayer (1990), who originally used the term "emotional intelligence", initially defined it as a form of intelligence that involves the ability to monitor one's own and others' feelings and emotions, to discriminate among them and to use this information to guide one's thinking and actions (Salovey & Mayer, 1990).

Definition
Later, the authors revised their definition of emotional intelligence, and the current characterization is now the most widely accepted. Emotional intelligence is thus defined as the ability to perceive emotion, integrate emotion to facilitate thought, understand emotions, and regulate emotions to promote personal growth (Mayer & Salovey, 1997).
On the most general level, emotional intelligence is the ability to recognize and manage emotions in one and others (Goleman, 2001).

Emotional intelligence model
Each theoretical paradigm conceptualizes emotional intelligence from one of two perspectives: ability models or mixed models ( Bar-On, 2002;Goleman, 1997).
Ability models regard emotional intelligence as a pure form of mental ability and therefore as a pure intelligence. In contrast, mixed models of emotional intelligence combine mental ability with personality characteristics such as optimism and well-being (Mayer, 1999). Currently, the only ability model of emotional intelligence is that proposed by Mayer and Salovey. Two mixed models of emotional intelligence have been proposed, each with a somewhat different conception.
In this section, we present Mayer and Salovey model, upon which our empirical work is based.

Intelligence
Mayer and Salovey conception of emotional intelligence was included within a model of intelligence, that is, its goal was to define emotional intelligence emotion and cognition. The first branch, emotional perception, is the ability to be self-aware of emotions and to express emotions and emotional needs accurately to others. Emotional perception also includes the ability to distinguish between honest and dishonest expressions of emotion. The second branch, emotional assimilation, is the ability to distinguish among the different emotions one feels and to identify those emotions that influence one's thought processes.
The third branch, emotional understanding, is the ability to understand complex emotions (such as feeling two emotions at once) and the ability to recognize transitions from one emotion to another.
Finally, the fourth branch, emotion management, is the ability to connect or disconnect from an emotion depending on its usefulness in a given situation (Mayer & Salovey, 1997 Finally, some authors proposed that emotional intelligence is a better predictor of cognitive and professional success (Fiori & Antonakis, 2011;George, 2000;Goleman, 2001;Song et al, 2010).

Emotional intelligence, optimism and firm financial policies
Recently, more research on emotional intelligence has been implicated in developing international business capabilities and has been used to predict a person's performance in the workforce market or a non-work environment (Lin, Chen, & Song, 2012;Mount & Downton, 2006;Yoo, Matsumoto, & LeRoux, 2006).
Emotional intelligence has been found to be positively related to leader performance (Rosete & Ciarrochi, 2005) and commitment to the group or organization (Carmeli, 2003). Emotional intelligence allows managers to improve their skills in alternative assessments (strengths, weaknesses and characteristics of their companies). It reduces managers' over-or underestimates (overconfidence and optimism) about the value of their firms, which implies that emotional intelligence reduces the suggestibility of leaders with behavioral biases and improves firms' financial policies. Mavroveli et al. (2007) emphasized that a high level of emotional intelligence is positively associated with low CEO suggestibility to behavioral biases. The authors added that reducing the presence of emotion improves the effectiveness of decisions (financial policy). Mayer and Salovey (1997) (2008) demonstrated that emotional intelligence has a significant, direct influence on job satisfaction. Thus, well-managed emotions allow individuals to achieve the optimum use of their resources and capabilities. These skills are essential to adapt to specific situations. In other words, a high level of emotional intelligence improves cognitive flexibility of leaders, reducing their suggestibility to behavioral biases (loss aversion) and encouraging them to make less biased decisions. These factors imply that there is a positive correlation between emotional intelligence and the effectiveness of CEOs' financial policies.

H2:
The larger the decrease in CEOs' loss aversion (demonstrating a high level of emotional intelligence), the more effective a firm's financial policies.

Emotional intelligence, overconfidence and firm's financial policies
A growing body of empirical evidence suggests that emotional intelligence correlates robustly with a variety of outcomes that signal social emotional success, including more frequent positive effects, higher self esteem, greater life satisfaction, increased social engagement, and a greater sense of well-being (Goleman, 2001;Gond & Mignonac, 2002;Hess, 2003;Zeidner & Olnick-Shmesh, 2010). In other words, emotionally intelligent CEOs are less susceptible to the effects of emotional biases such as overconfidence.
Some studies have, suggested that individuals who are endowed with a high level of emotional intelligence are more aware of the factors that influence their positive and negative emotions (Karim, 2010;Rode et al., 2007), which reduces the presence of emotional bias and encourages effective strategies.

Data
Our empirical study is based on quantitative research, and we use a questionnaire as our method of data collection. Our questionnaire consists of four main parts based on the treated areas in the theories: • The first part aims to identify the company.
• The second part focuses on determining the CEO's emotional intelligence level.
• The third part determines the CEO' s loss aversion level. Our choice of listed companies is justified by the fact that these companies are assumed to be the most efficient, and they meet several conditions necessary for the reliability of our study, including diffuse shareholders and an important role for the board in the companies' ownership structure; these factors consequently increase the validity of our assumptions.
We decided to exclude financial firms: banks, insurance companies and investment companies, both for development and portfolio management. These companies have different characteristics from nonfinancial businesses and their exclusion avoids correlation effects specific to a specific sector.
To obtain a representative sample of the Tunisian market, we added other unlisted companies.

Variables' measurement
The objective of this section is to explain the measurement of the variables' .

Choice of capital structure
The Appropriate measures in the literature to evaluate three methods of financing are:

Internally generated resources (cash flow)
Studies within the framework of the financial theories of investments have resorted to many measures of internal resources. Cash flow (CF) represents the cash flow generated by a business' activities and, is one of the most appropriate measures (Lehen & Poulsen, 1989;Naoui, Elgaied, & Bayoudh, 2008).

Cash flow rate (RCF) = CF / Total Assets
To show whether leader chooses to use internally generated resources, we use the change in cash flow rate. A negative change indicates the use of internal resources.

Debt level
A variety of variables are used to measure the level of a company's debt. Measures such as the total debt service ratio have been used by several authors (A. Hovakimian, G. Hovakimian, & H. Tehranian, 2004). Others have used the debt ratio in the medium and long term (Myers, 2001). The debt ratio in the short term was also used by Titman (1984 As part of our analysis, we propose to use the debt ratio to measure the debt level. This ratio is calculated by:
To show whether a manager uses debt, we use the change in the debt ratio. A positive change indicates the use of debt.

Firms' investment decisions
The purpose of this article is to demonstrate the im-

Assets specificity
In this study, we use the degree of assets intangibility as a proxy of the specific investments. The degree of assets intangibility can be appreciated on many levels. In This measure was also used by Cazavan-Jeny (2004), Moussu and Thibierge (1997), and Thibierge (2001).

Investment level
In this study, we use the presence of free cash flow and growth opportunities as two indicators of overinvestment (low future investment opportunities and free cash flow) or underinvestment (low free cash flow and future investment opportunities). The literature differs on how to measure free cash flow as conceptualized by Jensen (1986). In general, however, free cash flow is defined as operating income before depreciation, interest expenses and taxes, as well as dividends paid (Gul & Tsui, 1998;Jaggi & Gul, 1999;Lehen & Poulsen, 1989) divided by book value of total assets to account for effects related to size (Lang, Schulz, & Walkling, 1991). Future investment opportunities are measured by Tobin's Q (Skinner, 1993). Tobin's Q is defined as the ratio of a firm's market value to the replacement value of its assets (Griliches, 1981;Lindenberg & Ross, 1981;Megna & Klock, 1993;Skinner, 1993). If the value of Tobin's Q is greater than one, the company has profitable investment opportunities and vice versa. In our study, we use an approximation of Tobin›s Q that is, calculated as follows (Chung & Pruitt, 1994): where MVS represents-the market value of common and preferred shares; D is the-book value of debt, defined as current liabilities plus long-term debt plus inventories minus current assets; and A is-total assets.

Investment horizon
Referring to the theory of agency, leaders have an obligation to obtain results in short horizons. Their wealth is tied to the firm's performance during the duration of their mission, which is the period when they run the firm. These leaders prefer short-term investment projects to quickly determine the performance of these investments and reduce uncertainty about their own value on the labor market.
In this study, we use the rate of investment operations (industrial and commercial assets) as an indicator of the investment horizon.

Choice of dividend policy
The variable used to measure the dividend level is the distribution rate (Agrawal & Jayaraman, 1994;Rozeff, 1982 information is provided in terms of retention of earnings and, therefore, whether the flow (the retention rate is equal to 100 in the payout ratio).

Emotional bias
The questionnaire focuses on evaluating and scoring the three emotional biases (risk aversion, optimism and overconfidence). The questions were inspired from the questionnaires formulated by the Fern Hill and Industrial Alliance companies.
Emotional bias has 2 possibilities: • 1 if the individual has a high level for each bias.
• 0 if the individual does not have has a high level for each bias.

TEST
In this study, we generated a pool of 18 items (derived from Schutte et al, 1998, i.e.,the SSREI test) based on the theoretical model of emotional intelligence developed by Salovey and Mayer (1990). Each item selected for the initial pool of 18 items should reflect an adaptive tendency toward emotional intelligence within the model's framework. The respondents used a 5-point scale, whereby "1" represents "strongly disagree" and "5" represents "strongly agree, " to indicate the extent of the fit for each item described. The entire model is represented by the items. Each of the first four authors independently evaluated each item for fidelity to the relevant construct, clarity and readability. Some of the items were deleted, while others were added or revised before they were pilot tested by asking several individuals to complete the questionnaire and note any unclear elements. This process eventually resulted in a pilot-tested pool of 18 items.

Control variables
Static trade-off theory (STT) and pecking order theory (POT) are the theories that address the issue of a firm's financial decisions. The factors that explain a firm's financial decisions mainly concern the cost, size, level of risk, growth opportunities, and the structure of the assets and the business (Booth et al, 2001).
We include three control variables in our model that explain the effectiveness of choices regarding a compa-ny's financial structure. These variables are proxies for profitability, firm size and growth opportunities.
We include three control variables in our study that explain the effectiveness of the choice of the capital structure of a company. These variables are also proxies for profitability, firm size and growth opportunities.

Profitability
More profitable firms have, ceteris paribus, more internally generated resources to fund new investments.
If their managers follow a pecking order, they will be less likely to seek external financing (Fama & French, 2002). Thus, on average, these firms' leverage ratios will be lower. In trade-off models, on the other hand, this relationship is inverted. More profitable firms are less subject to bankruptcy risks, ceteris paribus.
Hence, their expected bankruptcy costs are reduced, and they can make more use of the tax shields provided by debt, thus choosing a position of greater leverage. We use the ratio of return on assets, ROA, to measure this variable: ROA=Earnings before interest, taxes, and depreciation divided by Total assets, lagged one year (9).

Firm size
Studies have suggested that the probability of bankruptcy is lower in larger firms; therefore, their debt capacity is higher than smaller firms, with all else being equal. However, fixed transaction costs can make new stock issues unattractive to small corporations, stimulating them to issue debt (Hovakimian et al, 2004;Rajan and Zingales, 1998).
Indeed, most studies have used total assets or turnover as a measure for firm size (Bujaki & Richardson, 1997). In this paper, we measure firm size through the log of the firm's total assets (LNSIZE).

Future investment opportunities
It has been argued that future profitable investment  face the need to issue undervalued securities to fund new projects. This process could, in turn, induce underinvestment. A more static version of the pecking order model, on the other hand, predicts that firms with more future opportunities will be more levered, ceteris paribus, because they need more external financing, and issuing debt is preferable to issuing new stock. (Booth et al, 2001;Naoui et al, 2008;Rajan & Zingales, 1998).
We use the Tobin's Q to measure this variable, estimated with the approximation formula proposed by Chung and Pruitt (1994): Where MVS represents the -market value of common and preferred shares; D is the -book value of debt, defined as current liabilities plus long-term debt plus inventories minus current assets; and A is total assets. For simplification purposes, the summary of each variable's range in the model, its name and its expected impact on the choice of capital structure are depicted in Table 2.

Bayesian Network Method
There are many versions of the definition of a Bayesian network, but the basic form (Pearl, 1986) can be stated as follows: a Bayesian network is a directed probability graph, connecting the relative variables with arcs, and this type of connection expresses the conditional dependence between the variables. The formal definition follows.
A Bayesian network is defined as the set of {D, S, P}, where.
(1) D is a set of variables (or nodes). In our case, D consists of the choice of capital structure, optimism, loss aversion, overconfidence, profitability, firm size and future investment opportunities.
(2) S is a set of conditional probability distributions In the Bayesian network, variables are used to express the events or objects. The problem can be modeled with the behavior of these variables. In general, we first calculate (or determine from expert experience) the probability distribution of each variable and the conditional probability distribution between them. Then, from these distributions, we can obtain the joint distributions of these variables. Finally, some deductions can be developed for some variables of interest using some other known variables.

Define network variables and values
The first step in building a Bayesian network expert is to list the variables recursively, starting from the target variable to the causes. Thus, we present the variables in this order in table 3.

CEO Emotional Intelligence and Firms' Financial Policies. Bayesian Network Method
The relative weight scale is 0 to 1. Thus, Table 4 shows The relationship analysis shows a negative correlation between a leader's loss aversion level and his overconfidence (β = -0.1555).
Finally, the results also show a negative correlation between a leader's overconfidence and his optimism level (β = -0.0159).

Average target maximizing analysis
After presenting all of the explanatory variables for each category of the target variable, we now introduce the variables that maximize each modality of the target variable. Thus, we use the target dynamic profile capability software (Bayesialab) to determine an a posteriori maximization of the target average. This test shows the case required to maximize the value of the target variable. Table 6 presents the dynamic profile of the CEO's emotional intelligence level (EI).
The dynamic profile analysis of the CEO's emotional intelligence provides the following results: The increased preference for directing flow of 89.82% and decrease the level of overconfidence with probability 100% involves a decrease in the emotional intelligence level to 50.42%.
The lower level of a CEO's loss aversion of 100% is positively correlated with anincrease in the CEO's emotional intelligence score of 49.58%.

Conclusion
This research examines the determinants of firms' financial policiesfrom a behavioral perspective.
The theoretical analysis highlights the role of a CEO's emotional profile (emotional biases and emotional intelligence) in explaining the CEO's decisions ( Bar-On, 2002;Damasio, 1994;George, 2000;Goleman, 1997;Mayer & Salovey, 1997). Management theory has revealed the existence of strong emotional interactions between a companies' . Several studies have shown that emotions and moods play a vital role in organizations performance (Damasio, 1994). An individual's emotional state and his intuition, often addressed by one's past experiences, guide an individual in the decisionmaking process. Emotional intelligence is the ability to perceive, feel, understand and regulate one's emotions in the context of emotional and intellectual development (Mayer & Salovey, 1997). An emotionally intelligent leader has a wide perspective, an open, synthetic vision, and a global understanding of a situation. Such an individual is aware of the emotional states of his partners and uses this knowledge as a strategic hedge against the risk of loss of his position or reputation. Hence, the specific nature of the choice, the level of investment and the investment horizon(long term or short term) are guided by a CEO's level of emotional intelligence and his suggestibility to behavioral biases. Thus, an emotionally intelligent leader has low suggestibility and low levels of behavioral biases (optimism, loss aversion and overconfidence) and chooses specific assets, underinvestment and long-term investment projects.
Thee empirical analysis is based on a survey of CEOs of large private companies in Tunisia. The analysis reveals the importance of a CEO's emotional intelligence in explaining a firm's financial policies. Indeed, the empirical analysis of the relationship between a CEO's  emotional intelligence and the firm's choice of capital structure demonstrates the role of emotional intelligence in explaining managerial decisions. A leader's emotional intelligence is positively correlated with his overconfidence and optimism, but emotional intelligence is negatively correlated with a CEO's level of loss aversion. The increase optimism (and / or overconfidence) and decrease the level of aversion to loss of incentive to choose the combination to achieve flow more debt investment projects. This leader optimistic or overconfident opts for decisions specific investment to improve the competitiveness of its business and ensuring creation of long-term value. He prefers to underinvestment to limit the maximum recourse to external financing and how to hedge against the risk of loss of compensation or reputation.

EI = NO
Finally, financial policy analysis that integrates the behavioral dimension allows richer predictions from organizational theories: leaders adjust their decisions (on investments and distributions) to minimize their risk of loss of compensation or reputation. They opt for decisions that minimize agency costs, transaction costs, and cognitive costs.