Master Thesis
Managerial overconfidence, risk-taking behavior, and internal control
Name: Yuhan Ji
Student number: 13218492
Thesis supervisor: Dr. Ana Mickovic Date: Jun 17, 2022
Word count: 10771
MSc Accountancy & Control, Specialization Accountancy Faculty of Economics and Business, University of Amsterdam
Statement of Originality
This document is written by student Yuhan Ji who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
Managerial overconfidence is a typical irrational behavior. Managerial overconfidence has this important effect on the risk-taking behavior of firms. Effective internal controls can have a positive impact on curbing managerial overconfidence and corporate risk-taking behavior. This paper examines the relationship between managerial overconfidence and corporate risk-taking behavior, and the moderating effect of internal control on their relationship, based on data from Chinese listed companies from 2012-2017. The finding shows managerial overconfidence promotes risk-taking behavior and that the positive relationship between overconfident managers and risk-taking behavior is weakened by high level of internal control. The results have important policy implications in the areas of selecting the right managers and improving the internal control mechanisms of the firm.
Keywords: managerial overconfidence, risk-taking behavior, internal control
Acknowledgements
I would like to genuinely express my gratitude to my supervisor Dr. Ana Mickovic, who gave me guidance and inspiration and helped me to improve my thesis. She guides me through my academic path with full patient and I am forever grateful for her kindness and mentorship. I would also like to thank my friends; we support each other to progress together. I would also like to thank my parents. In Chinese, we say 谁言寸草心, 报得三春晖, which translated to English is there is no way to repay the great kindness of parents who raised me. They were always unconditionally support me pursuing my studies, both financially and mentally. At last, I would like to give thanks to myself for all my effort spent during my masters. Your effort will finally payoff so let’s wait for that day to come.
Contents
1 Introduction ... 6
2 Literature review ... 9
2.1 Managerial Overconfidence ... 9
2.1.1 Definition ... 9
2.1.2 Consequences ... 10
2.2 Risk-Taking Behavior ... 11
2.2.1 Definition and Consequences ... 11
2.2.2 Influence Factors ... 13
2.3 Internal Control ... 13
3 Hypothesis Development ... 14
3.1 Theory ... 14
3.1.1 Control illusion theory ... 14
3.1.2 Cognitive theory ... 15
3.1.3 Behavioral finance theory ... 15
3.2 Hypothesis ... 15
3.2.1 Managerial Overconfidence and Risk-Taking Behavior ... 15
3.2.2 The moderating effect of Internal Control on the relationship of Managerial Overconfidence and Risk-Taking Behavior ... 17
4 Research Methodology ... 18
4.1 Research Design ... 18
4.1.1 Sample and Database ... 18
4.2 Variable Measurement ... 19
4.2.1 Managerial Overconfidence ... 19
4.2.2 Risk-Taking Behavior ... 21
4.2.3 Internal Control ... 22
4.2.4 Control Variable ... 22
4.2.5 Libby Box ... 24
4.2.6 Regression Model ... 24
5 Empirical Results ... 25
5.1 Main results ... 25
5.1.1 Descriptive statistics ... 25
5.1.2 Regression analysis ... 26
5.2 Additional tests ... 29
6 Conclusion and Future Research ... 33
6.1 Conclusion ... 33
6.2 Limitation and future research ... 34
6.2.1 Limitation ... 34
6.2.2 Future research ideas ... 34
7 Reference ... 36
1 Introduction
Traditional financial theory has assumed of "rational economic man" and has ignored the impact of the psychological characteristics of managers on the long-term health of the company.
Some economists have introduced the psychological characteristics of actors into the field of finance and found that there are a lot of irrational behaviors of managers in their daily economic decisions, among which, overconfidence of managers is one of the most typical irrational behaviors. From a psychological point of view, overconfident people are used to being overconfident in their own cognitive decisions. They tend to overestimate the likelihood of success and attribute success to their predictive ability. They rate their personal strengths and abilities more highly, leading to higher levels of risk-taking and improved future firm performance (Kim & Lu, 2011). However, it is argued that risk-taking has its limits (Wang, 2008). Risk-taking goes hand in hand with risky crisis. Overconfident managers are more prone to fall into the 'brave risk-taking trap' and take excessive risks, which is detrimental to firms developing and exploring new markets. Besides, some research shows risk-taking behavior does not pay off for stakeholders but with a greater problem of underfunding (Goldberg et al., 2020). So first I want to examine the relationship of managerial overconfidence and risk-taking behavior.
So far, I have seen a lot of research encouraging equity incentives for executives to promote greater risk-taking in firms (Certo et al., 2003), touting the positive effects of risk- taking in promoting investment in innovation and improving firm performance. This is used to address that under the 'rational man' assumption, managers tend to be risk-averse in order to maximize their self-interest. Malmendier and Tate (2005) argue that overconfident CEOs do not need incentives to maximize the market value of their companies because they believe they are already doing so. The financial crisis in 2008 led authorities around the world to recognize that many companies lost a lot of money because of excessive risk-taking. Seeking an external pressure to constrain managers at this point is a tricky issue. External governance mechanisms are macro in nature and difficult to change in the short term. Therefore, constructing internal governance mechanisms is particularly important. As an important part of the internal governance, internal controls play an integral role in regulating the behavior of employees at all levels of the business to avoid risk. Without effective internal controls to balance the risks, it is easy for a company to get into trouble.
Internal control can effectively reduce risks within a company. As a governance tool to achieve business objectives, internal control can strengthen the supervision of business activities at all levels of the enterprise. The Enron scandal in the US has increased awareness of the importance of internal controls in modern business. Governments in various countries have strengthened their attention to construct the internal control mechanisms for enterprises.
The US has enacted the SOX Act, which includes several provisions requiring companies to strengthen internal controls. The Corporate Internal Control Basic Standards in China issued by the Ministry of Finance of China clearly stipulates that the enterprises concerned must conduct self-evaluation of the effectiveness of their internal control and disclose it in a timely manner. In August 2011, the Research Group in China launched the Internal Control Index (ICI) of Chinese Listed Companies, which is an intuitive reflection of the effectiveness of internal control and risk control ability. Its launch fills the gap of lack of quantitative internal control evaluation in the Basic Code and the supporting guidelines. This also provides us with a great deal of convenience in studying internal control as a governance mechanism. Based on this, internal control is introduced to examine its moderating role on the relationship of managerial overconfidence and risk-taking behavior. Therefore, our research question is does internal control mechanism moderate between the effect of managerial overconfidence and corporate risk-taking behavior?
I investigated 679 Chinese listed companies from 2012 to 2017. Through empirical research, we find that (1) managerial overconfidence promotes risk-taking behavior and that (2) the positive relationship between overconfident managers and risk-taking behavior is weakened by high level of internal control. When a firm has a high internal control, overconfident managers are constrained to reduce the risk-taking behavior of the firm. When the level of internal control is low, overconfident managers will have a much stronger voice and will be more likely to influence the firm's decisions, thereby increasing risk-taking behavior. The results have important policy implications in the areas of (1) selecting the right managers and (2) improving the internal control mechanisms of the firm.
This paper adopted two approaches to conduct additional tests. The first is to change the variable settings. When calculating risk-taking behaviors, I changed the rolling window for adj.
ROA from 5 to 3 years for the regressions. The second is dividing the sample into SOE (state- owned enterprises) and non-SOE groups based on the nature of ownership to test two hypotheses. I find the conclusion still holds when changing to the alternative measure of risk- taking behaviors. However, I can only find empirical evidence that managerial overconfidence
increases risk-taking behavior in non-SOE group. The relationship between managerial overconfidence and corporate risk-taking in SOEs is not obvious. The moderating effect of internal control on the relationship of managerial overconfidence and corporate risk-taking can only be found in non-SOEs. Therefore, I conclude that overconfidence managers in non-SOE promotes risk-taking behaviors and non-SOEs have a sound internal control system to moderate the relationship between overconfidence manager and risk-taking behavior.
This paper contributes to the existing literature in two ways. Firstly it enriches the research on corporate risk-taking. While existing research on the influencing factors of corporate risk- taking has focused on corporate governance, managerial incentives, managerial power, managerial characteristics and cultural characteristics, this paper explores the influencing factors of corporate risk-taking from the perspective of managerial overconfidence, thus providing a new explanation for corporate risk-taking behavior. Secondly, it enriches the research on the economic consequences of internal control and provides empirical evidence on the effects of internal control in corporate financial decisions. The innovation of this paper is to introduce internal control as a governance mechanism to investigate the moderating role of internal control in managerial overconfidence and corporate risk-taking. Effective internal control as a governance mechanism can constrain high risk-taking behavior made by overconfident managers.
The structure of this paper unfolds as follows. Section 2 reviews previous relevant literature. Section 3 states the hypotheses based on relevant theory. Section 4 demonstrates the research design, the variable measurement, and the regression model. Section 5 reports the results of the empirical study and additional tests. Section 6 discuss the conclusions, study limitations and future research ideas.
2 Literature review
2.1 Managerial Overconfidence
2.1.1 Definition
The term "overconfidence" was first coined by psychologists. Overconfidence is a cognitive bias that arises mainly because people believe too much in their own abilities.
Overconfident people are overly optimistic about the probability of success of a project and underestimate the probability of their own failure, thus creating a bias against the probability of success and failure of the project itself (Wolosin et al., 1973; Langer & Ellen, 1975; Cooper et al., 1988). In 1980, Weinstein found in a large number of studies that people who were used in experiments were generally more likely than others to think they would live beyond 80 years of age. Very few thought they would get divorced, get cancer, and go through other bad things in the future.
As the economy develops and research progresses, traditional financial theory based on the assumption of 'rational economic man' is no longer able to explain the irrational behavior of managers. Overconfidence is one of the typical irrational behaviors of managers in making decisions. More confident people are more likely to sit in leadership positions because they appear more persuasive than others (Anderson et al., 2012). Weinstein's (1980) study found that managers in companies exhibited far more overconfidence than the general due to their higher positions and greater power (Cooper & Dunkelberg, 1988; Landier et al., 2004). Odean's (2001) study found that managers are generally more overconfident than the general. Moreover, overconfident managers typically have inflated estimates of expected project benefits and lower estimates of risks (Goel & Thakor, 2002).
Roll (1986) was the first to introduce the psychological term 'overconfidence' to the field of finance. He argued that managers are generally overconfident due to the control effect, which in turn influences investment decisions and leads to inefficient investments. He argued that the overconfidence is reflected in the overpriced acquisitions in M&A activities, where the acquirer is extremely optimistic about the economic benefits that the acquisition will bring to the company. Moore & Schatz (2017) proposed three approaches to the study of overconfidence, namely overestimation, overreplacement and overprecision. Here I apply three approaches to managers. Firstly, managers overestimate their own predictive abilities. Secondly, when compared to others, they perceive themselves as 'better than average'. Overconfident managers
perceive that their chosen projects are more likely to succeed than others. Thirdly, overconfident managers are too certain about the accuracy and outcomes of their decisions, so they tend to narrow their prediction range. As a results, they often overestimate the likelihood of project success and underestimate potential risks.
Currently, there is no single definition of overconfidence in behavioral corporate finance.
It is generally accepted that managerial overconfidence refers to the fact that managers always overestimate their own competence and knowledge levels and their ability to cope when making important decisions, overestimate the value of their firms, overestimate the firm's ability to cope with risk, and correspondingly underestimate the current risks faced by the firm (Malmendier & Tate, 2005a, 2005b).
2.1.2 Consequences
Some academics believe that managerial overconfidence can have adverse economic consequences. Overconfident people develop a cognitive bias towards risk, perceiving less risk than the actual project risk and pursuing riskier behavior (Simon & Houghton, 2003).
Kahneman and Lovallo (1993) argue that overconfident people perceive all the risks as only a fraction of the risks that actually exist, therefore they inadvertently take on more risk. They tend to make risky investments, and equity incentives encourage them to take risks (Seo &
Sharma, 2018). CEO overconfidence can distort a company's investment decisions for pension funds. Overconfident CEOs are more likely to invest their pension funds in more risky projects and thus take greater risks in managing their pension plans (Goldberg et al., 2020). Increased volatility in pensions raises the risk-taking of firms. CAPM financial model shows high risk is associated with high returns. However, Goldberg’s research demonstrates that high risk-taking does not generate the right returns for stakeholders. Firms led by overconfident CEOs experience greater underfunding due to aggressive investments than firms managed by more cautious CEOs. Yizhong Wang (2016) examine the moderating effect of managerial overconfidence on the relationship between inflation uncertainty and overinvestment using two measures of managerial overconfidence. The study demonstrates that managerial overconfidence is a stronger driver of overinvestment in Chinese SOEs (State-Owned Enterprise) than in non-SOEs. The result of overinvestment shows overconfident managers tend to make more adventurous decision based on self-perception bias and perceived project risk bias (March & Shapira, 1987; Cooper & Dunkelberg, 1988; Johnson & Fowler, 2011).
Tom (2020) argues that equity incentives for overconfident CEOs increase firm risk, indicating
that equity incentives for CEOs without overconfidence characteristics have less impact on the increase in firm risk. These CEOs are more disciplined in making decisions than overconfident CEOs. David Peo ́n (2015) designed a simulation game to study the impact of overconfidence on the quality of bank lending decisions. The simulation game proves that overconfidence leads to lower lending rates and more lending practices, indicating a significant decrease in the bank lending quality. Renneboog and Toblerc (2017) confirmed a positive relationship between overconfidence and investment choice under several direct measures. Stronger overconfidence leads to overinvestment.
There are some positive economic effects of managerial overconfidence. Russo&
Schoemaker (1992) stated, ‘Many of these people are distorting reality, yet their optimism has motivational value’. Some scholars have argued that managerial overconfidence helps to promote risk-taking in firms, which adds value to the firm (John, Litov and Yeung, 2008).
Galasso and Simcoe (2011) examine the correlation between CEO attitudes and firms' innovation performance. Overconfident CEOs are more biased towards innovation investment.
They are more bullish about the future benefits that innovation investment will bring to the firm and are more decisive in their innovation activities.
2.2 Risk-Taking Behavior
2.2.1 Definition and Consequences
Corporate risk-taking behavior is the risk decisions made by a company when faced with uncertainties such as market and R&D costs at the same time bear the cost of investment failures or the opportunity cost of resources invested (Lumpkin & Dess, 1996). Risk-taking behavior is the basis for corporate development (Boubakri et al., 2013). Today's rapidly changing markets places even tougher demands on the risk-taking capacity of firms. Managers, who are the decision makers of firms, determine the level of risk-taking of firms ultimately.
Their decisions reflect the firm’s risk appetite. As perfectly competitive markets do not exist, it more difficult for managers to make optimal decisions due to information asymmetry.
A higher level of risk-taking by a firm can help the firm achieve higher levels of financial performance and long-term growth (John, 2008). Traditional financial theory has followed the 'rational economic man' assumption, where rational managers prefer to allocate the firm's resources and make decisions to maximize their own interests, which may conflict with the interests of external shareholders (Jensen & Meckling, 1976). Agency problems are prevalent
in modern firms. Shareholders are seen as risk neutral because they can diversify their risk through a diversified portfolio. However, the investment decisions that managers can make are constrained by free cash flow, leverage, job careers, and many factors, resulting in managers being risk averse. Managers tending to forego risky but wealth-creating investment opportunities for shareholders (Smith & Stulz,1985; Guay, 1999) and select conservative investment strategies (John, 2008). Therefore, based on the 'rational economic man' hypothesis, scholars have investigated how to encourage risk-taking behaviors to enhance their positive effect on firms. Many studies have shown that giving CEOs equity incentives can promote risk- taking behaviors by CEOs who tend to be risk-averse, thereby boosting corporate performance and increasing shareholder wealth (Carpenter et al., 2003; Sanders & Hambrick, 2007; Devers et al., 2008). Kim and Lu (2011) point out that if a firm takes some risk proactively, it not only enhances the ability to invest in R&D, but also increases the firm's internal capital as the number of risky projects increases. These studies all support the assumption that under traditional financial theory, increasing the level of risk-taking can, to some extent, mitigate agency problems and improve corporate performance.
However, their own risk perception bias can lead them to inadvertently take too many risks. Overconfident managers perceive less risk than exists in the actual project and thus pursue more risk-taking behavior (Simon & Houghton, 2003). Managerial overconfidence is an irrational behavior, therefore the findings of traditional financial theory under which equity incentive promotes risk-taking by risk-averse managers are no longer applicable. Excessive risk-taking can have a number of negative effects, trapping firms in blind expansion and failed decisions that endanger their long-term survival and growth. The research on bank field have highlights a situation in which managers take too much risk, while no corresponding increase in performance (Asghar et al., 2016). Corporate risk needs to be balanced (Kemelgor, 2002).
Pieterse (2011) argues that while firms should not be deterred from taking risks in their growth, they should also not take excessive risks and fall into the 'brave risk-taking trap', i.e. both under- and over-risk-taking can lead to poor firm performance. In the aftermath of the global financial crisis, excessive risk-taking by overconfident CEOs did not generate the corresponding returns for shareholders. Those managed by prudent CEOs had the same fund investment returns as the firms led by overconfident CEO (Goldberg et al., 2020). As a result of higher risk-taking, companies led by these over-confident CEOs have in turn experienced more severe underfunding problems.
2.2.2 Influence Factors
The level of corporate risk-taking is influenced by many factors. Bargeron (2010) uses data from a sample of US, UK and Canadian companies from 1994-2006 to examine the impact of SOX on risk-taking in US listed companies. They found that risk-taking behavior in US listed companies decreased significantly after the implementation of SOX, as evidenced by an increase in corporate cash holdings and a decrease in investment spending. Faccio, Marchica and Mura (2011) find that firms controlled by more dispersed majority shareholders make more risky investment decisions than firms controlled by concentrated majority shareholders, suggesting that the degree of concentration of a firm's equity has an effect on the risk-taking.
John, Litov and Yeung (2008) argue that better investor protection allows firms to make risky but high firm value investments. In a study of the US restaurant industry, Seo and Sharma (2018) proves that overconfident CEOs are positively associated with risk-taking and are more inclined to make risky investments.
2.3 Internal Control
Managers are overconfident before making decisions that they believe are in the best interest of shareholders, such as lowering their assessment of project risk. However, the decisions made sometimes harm the stakeholders afterwards. Huang (2016) studied the lending behavior of overconfident CEOs before the 1998 Russian crisis and the 2008 financial crisis.
They proved that overconfident CEOs are more aggressive in relaxing lending standards, especially lowering lending rates and increasing lending volumes before a crisis, thus becoming more vulnerable to financial crisis when it occurs. Jensen and Meckling (1976) show that the governance structure of the firm directly influences the firm's investment decisions due to information asymmetry and flawed incentives. The financial crisis, which led to heavy losses, has led to the realization that many companies have taken excessive risks. Part of the reason why they took such high risks was due to the low quality of internal controls. It is important to note that internal controls play an integral role in regulating risk-taking behavior.
The Ministry Finance of China issued the first internal control system, the Basic Standard for Enterprise Internal Control, in 2008 with the aim of giving enterprises standardized requirements for internal control operational processes. For the construction framework of internal control construction, both the COSO reports (COSO,1992,2004,2013,2017) and the Basic Standard for Enterprise Internal Control consider that a complete internal control system consists of five elements, namely control environment, risk assessment, control activities,
information and communication and monitoring activities. By building an internal control system, companies place management's business decisions in an open and transparent monitoring environment. Irrational behavior such as overconfidence of managers is restrained through internal controls to achieve as stable a level of risk-taking as possible. Bin Liu (2021) argues that there is a significant negative relationship between a firm's willingness to engage in internal control and managerial overconfidence, which means that increasing a firm's willingness to engage in internal control can help to restrain individual managers' overconfidence. Moreover, it has been shown that high internal control quality can have a dampening effect on managerial overconfidence (Xing & Song 2015, Zheng & Chen 2018).
High-quality internal controls can also effectively reduce the business risks of enterprises. As business risks increase, the inhibiting effect of internal control on business risks becomes more evident (Gao, 2010). SOX 404 requires the company's management and external auditors to evaluate and attest to the company's internal control over financial reporting. Bargeron (2010) concludes by examining the impact of SOX on risk-taking behavior in US listed companies that SOX significantly inhibited risk-taking, as evidenced by a significant reduction in investment spending and an increase in the firm's cash holdings. Therefore, strengthening internal controls helps to reduce corporate risk-taking behavior.
3 Hypothesis Development
3.1 Theory
3.1.1 Control illusion theory
Langer (1975) first introduced the control illusion theory. The illusion of control is an expectation that is biased compared to the reality. Individuals overestimate the probability of their own success compared to the probability of individual success under objective conditions.
In the daily operation of a company, managers generally have the final say. This leads managers to overestimate their control over decisions and thus to develop the illusion of absolute control.
Managers believe that they have a high degree of certainty about the company's ability and that they have a high degree of control over the company's decisions. This is the psychological basis for the study of managerial overconfidence.
3.1.2 Cognitive theory
Experimental studies by psychologists have shown that people are not completely rational throughout the decision-making process. Influenced by cognitive biases, emotions and feelings, people exhibit irrational behavior when making decisions (Tversky & Kahneman, 1974).
Cognitive biases also lay the foundation for behavioral finance. Cognitive biases consist of three main aspects, namely pre-, during and post-decision making. In the pre-decisional process, manager uses his or her own ability to sift through valid information. Habitual judgements are mixed in with the rational decision-making process. After the information has been processed managers are ready to start making decisions. An overconfident manager will assume that he or she can avoid most of the risks. Even decisions that have already been made can be remedied. Post-decisional cognitive biases manifest themselves as self-deception and self-attribution. Overconfident managers attribute good results to themselves and blame external factors and luck for failed decisions. Managerial overconfidence has been described as 'the most significant cognitive bias' (Kahneman, 2011), or even 'the mother of all decision- making biases' (Tenney et al., 2015).
3.1.3 Behavioral finance theory
Behavioral finance theory moves away from the 'rational man' assumption to re-examine the impact of the psychological characteristics of actors as market participants on financial markets. Overconfidence belongs to a branch of behavioral finance theory, based on the irrational perspective of managers. In contrast to irrational managers, rational managers may not act in maximizing the interests of the business for reasons of self-interest. Rational managers are more cautious in making investment decisions, as making riskier investments may affect the manager's personal reputation and job career. In the case of the overconfident manager in this paper, the manager makes decisions that invariably expose the firm to excessive risk-taking due to an underestimation of project risk and an overestimation of his or her own investment capacity.
3.2 Hypothesis
3.2.1 Managerial Overconfidence and Risk-Taking Behavior
On the one hand, overconfident managers usually have a high opinion of their personal strengths and abilities. Overconfident individuals generally tend to overestimate their own strengths (Larwood & Whittaker, 1997; Moore & Schatz, 2017). Managers with higher levels
of overconfidence usually believe that they have more knowledge, experience, and resources than their competitors in the same industry. As a result, they believe that the project they have chosen is the most promising. Overconfident managers believe they have absolute power to control the process of project. Sometimes overconfident managers tend to be very selective in their choice of projects. They believe that their excellent personal skills and superb problem- solving abilities need to be matched with complex and very challenging projects (Griffin &
Tversky, 1992). As a result, overconfident managers generally invest a lot in very risky projects (Hirshleifer, 2012). On the other hand, overconfident executives are overly optimistic about their projects. Moore and Kim (2003) found in an empirical study that overconfident managers tend to overestimate the returns and underestimate the risks of a project. That is, the greater the degree of overconfidence, the more optimistic the manager is about a project. They believe that the project will be profitable and are inclined to choose it even if it is a high-risk project. Overall, overconfident managers overestimate the ability of their business to withstand risk and therefore set a greater range of acceptable risks, promoting risk-taking behavior.
Wang et al. (2009) argue that the investment behavior of overconfident managers is more sensitive to the cash flows generated by financing activities. Overconfident managers will over- invest when their companies can generate sufficient cash flow from financing activities. With sufficient corporate cash flow, overconfident managers in their dominant position are more likely to overestimate the return on investment in projects, overinvest and increase corporate risk-taking behavior (Jensen & Meckling, 1986). Conversely, the positive relationship of managerial overconfidence and corporate risk-taking may not hold in the case of insufficient cash flow. Miguel and Pindado (2001) found that firms with inadequate internal free cash flow usually forego good investment opportunities and therefore generate underinvestment. Within the constraints of cash flow, the risk-taking of firms led by rational CEOs decreases accordingly.
The risk-taking of firms led by overconfident CEOs is likely to remain unchanged. From the above analysis we can see that the relationship of managerial overconfidence and risk-taking behavior is ambiguous but with a clear positive touch. This leads us to our first hypothesis.
H1: Managerial overconfidence promotes risk-taking behavior.
3.2.2 The moderating effect of Internal Control on the relationship of Managerial Overconfidence and Risk-Taking Behavior
Internal control is an important mean of disciplining the behavior of managers and help company achieve business objectives. Effective internal controls have a positive impact on financial performance (Eniola & Akinselure, 2016; Sahabi et al.,2017). Internal control is exercised by the board of directors, supervisory board, managers, and all employees.
Establishing an efficient internal control system can prevent fraud and business failure (Etengu
& Amony, 2016) and improve accountability mechanisms for management (Kaplan, 2008).
The five elements of COSO work together could reduce the impact of irrational decisions by managers on the company to an acceptable level. A firm's control environment helps managers develop the right mindset for decision making. The risk assessment is based on the company's risk appetite and risk tolerance. Establishing effective information communication channels within the enterprise restrains managers and hedges risks to a certain extent. The guidelines of the Corporate Internal Control Basic Standards in China No. 6 have specific provisions for the financial activities of an enterprise and the main types of risks and risk control measures.
This aims to prevent financial risks through regulations and to restrain the subjective decisions of managers from generating high risk-taking behavior for the company.
The moderating effect of internal control may be insignificant when the effectiveness of internal control regulations is less than the overconfident manager's own restraint power on risk-taking behavior. If constrained by internal cash flow, overconfident managers may choose to reduce the firm's investment activities and actively reduce risk-taking. Then the moderating effect of internal control, an internal corporate governance instrument, on the positive relationship between managerial overconfidence and corporate risk-taking may be insignificant.
Nevertheless, we expect internal controls to moderate the positive effect of managerial overconfidence on risk-taking behavior. I hope to find its inhibiting effect on excessive corporate risk-taking triggered by overconfidence of managers. The above analysis led us to our second hypothesis.
H2: The relationship between managerial overconfidence and risk-taking behavior is weakened by higher levels of internal control.
4 Research Methodology
4.1 Research Design
I use the archival research method to conduct this study and the data was analyzed using STATA17.0 software. This paper systematically presents a theoretical framework for the analysis of managerial overconfidence, risk-taking behavior and internal control. The relationship between managerial overconfidence and risk-taking is examined first, and then internal control is added to test its moderating effect on the relationship between managerial overconfidence and risk-taking behavior. Finally, the robustness of the empirical results is tested by changing variable setting of risk-taking behaviors and changing sample period, and splitting the whole sample into SOE (state-owned enterprises) group and non-SOE group.
4.1.1 Sample and Database
I download the public available data required for this paper from CSMAR database, DIB database. I choose a sample period of 2012 to 2017 for listed companies in China. This time period is the maximum time intervals can be observed because Internal Control Index which is a measure for internal control was released in 2011. Refers to John, Litov and Yeung (2008) and Faccio, Marchica and Mura (2011), I use five-year rolling window approach to calculate risk-taking behavior (𝑅𝑖𝑠𝑘!,#) based on the earnings volatility. The rolling window method uses observations from each window to construct an aggregated observation. I use this method when calculating risk-taking behavior of each firm on each year. The window here refers to the year (t). For example, if t=2012 and window length is five years, then we use the related firm data from 2012 to 2017 to roll forward the data by using STATA command to obtain an aggregated value which is the firm's risk-taking behavior in 2012. Rolling forward in order, risk-taking behavior in 2017 is calculated by using relevant data in 2017-2021. Finally, we end up with a risk-taking behavior for the period 2012-2017 because 2017 is the latest sample period I can observe.
Table 1 shows the sample selection process for regression analysis. Firstly, financial companies are excluded from the sample due to the specific nature of the financial sector and the different accounting standards adopted. Secondly, I exclude ST (Special treatment) companies, which are in an abnormal financial condition. Because I use the rolling-window approach to calculate risk-taking behavior and window length is 5 years, thus requiring five consecutive years of data for per company. Companies with any one or more missing values will be excluded entirely. Samples where any of the variables are missing are excluded as well.
This resulted in 4074 observations of 679 listed Chinese companies.
4.2 Variable Measurement 4.2.1 Managerial Overconfidence
One of the biggest difficulties in this study is how to measure managerial overconfidence.
There is not yet a universally accepted method of measuring managerial overconfidence. I analyzed measurements commonly used in the related papers and chose a relatively more suitable measurement for Chinese companies.
Malmendier and Tate (2005a) use whether there is a net increase in the number of shares or stock options held during the exercise period as a measure of managerial overconfidence.
The intrinsic logic is that CEOs will continue to hold options beyond the rational exercise threshold due to the psychological trait of overconfidence compared to rational CEOs. CEO delays in exercising options reflects over-optimism about future share price returns. Normally,
Table 1 Sample selection
Firm-year Observation
Listed Chinese companies dataset 19780
Less:
Financial companies (960)
ST companies (2700)
Missing data to calculate managerial overconfidence (2934) Missing data to calculate risk-taking behavior (4380)
Missing data to calculate internal control (1260)
Missing data to calculate control variables (3474)
Number of firm-years in the final sample 4074
Note: There are 4074 observations of 679 listed Chinese companies. After obtaining the raw data, financial companies, ST companies and companies for which data were not available were excluded from the paper.
options act as an incentive mechanism can motivate managers to work hard and produce better performance in order to increase the market share price. This is because inflating the share price allows them to maximize the cash proceeds when exercising their options. However, why some executives have not been exercising their options despite high share prices already exists?
Malmendier and Tate (2005a) interpret this as a psychological effect of over-optimism and use it as a proxy for managerial overconfidence.
The incomplete stock option system in China creates an obstacle to our use of the Malmendier & Tate (2005a) approach. The stock option system in western countries is supported by a multifaceted legal environment. Mature securities markets, market-based managerial remuneration and a liberal corporate capital regime also provide fertile ground for the implementation of stock option systems. However, we can find no data relating to the exercise of options by executives in the Share Incentive Exercise Schedule disclosed by listed companies in China. Chinese Company Law hinders the source of shares for stock options. The Company Law prohibits companies from reserving shares when issuing shares on incorporation or when issuing additional shares in the course of development and restricts companies from buying back shares in the secondary market. Besides, the shares held by an executive can only be sold six months after leaving or retiring from the company and cannot be realized for a long period of time. This significantly reduces the incentive effect of the option system on managers.
We have to say that there is a barrier to the transmission mechanism between option incentives and share price increases, which is in conflict with the intrinsic logic of Malmendier & Tate's (2005a) measure. Therefore it is not suitable as a measure of managerial overconfidence for this paper.
Malmendier and Tate (2005b) use media-based measure to measure managerial overconfidence. However, China does not have a well-developed media rating agency. The business evaluation of the manager was also not well announced. Doukas and Petmezas (2006) consider managers who make more than five acquisitions in a short period of time as overconfident. This approach has some limitations due to the restrictions on the number of M&A in China. Lin et al. (2005) use earning forecast deviation by managers to proxy for managerial overconfidence. Overconfidence is indicated when expected earnings exceed actual earnings. However, managers may falsify their early forecasts in order to raise capital. Data of earnings forecasts deviation for companies may not be reliable.
Therefore, we refer to Hayward and Hambrick’s (1997) study and choose to measure managers' overconfidence by using the relative proportion of executive compensation as a measure of managerial overconfidence. It has been shown that the higher the CEO's pay relative to other managers in the firm, the more important the CEO's position in the firm and the more likely he or she is to be overconfident (Hayward & Hambrick, 1997). There is a significant positive relationship between a manager's pay ratio and his control (Brown & Sarma, 2007).
However, listed companies in China only disclose the sum of the remuneration of the top three highest paid executives and the sum of the remuneration of all executives. Therefore, we draw on a modified Hayward and Hambrick (1997) measure for the Chinese context by Jiang et al.
(2009) and use the top three executives' compensation as a percentage of all executives' compensation as proxy for managerial overconfidence. We assume that the top three executives are the top managers of the company although the number of executives varies. It reflects to some extent the importance of the top managers in the overall management team, which is also in line with Hayward & Hambrick's (1997) thinking. If relative proportion of top three executives’ compensation is greater than the median, the variable equals 1.
We can clearly see that except media-based measures (Malmendier & Tate, 2005b), the other measures of managerial overconfidence are all binary measures of managerial overconfidence. Here we distinguish between managerial overconfidence and managerial confidence. Using binary measures indicates scholars’ intent to classify managers as either overconfident or non-overconfident, rather than measuring the level of managerial confidence.
4.2.2 Risk-Taking Behavior
Indicators that have been commonly used in the literature to measure risk-taking behavior include earnings volatility, stock return volatility, cash flow volatility and corporate R&D expenses. Due to the highly synchronized and volatile nature of the Chinese stock market, corporate cash flows are affected by seasonal factors. Corporate R&D expenses are more influenced by accounting standards and have more missing values in non-innovative firms.
Therefore, this paper draws on John, Litov and Yeung (2008) and Faccio, Marchica and Mura (2011) to calculate risk-taking behavior (𝑅𝑖𝑠𝑘!,#) based on the earnings volatility. Firm earnings volatility represents the active and voluntary risk-taking undertaken by firms through operational decisions, which subsequently leads to fluctuations in firm’s 𝑅𝑂𝐴. Here 𝑅𝑂𝐴 volatility is observed on a 5 rolling year basis. For example, the first period is 2012-2016; the second period is 2013-2017; the third period is 2014-2018, and so on. To eliminate the effect
of industry heterogeneity, we first adjust the 𝑅𝑂𝐴 of companies for each year using the industry average 𝑅𝑂𝐴. Then the standard deviation of 𝐴𝑑𝑗𝑅𝑂𝐴 is derived using every 5 years as the observation period for the study, resulting in a standard deviation that is a measure of risk- taking behavior. A larger standard deviation of 𝐴𝑑𝑗𝑅𝑂𝐴 indicates a more volatile level of profitability, i.e. a greater degree of risk-taking by the firm. To eliminate the effect of outliers, 𝐴𝑑𝑗𝑅𝑂𝐴!,$ is winsorized at 0.5% on both side of the sample distribution refers to John, Litov and Yeung (2008).
𝐴𝑑𝑗𝑅𝑂𝐴!,$ = %&'()*!,#
*++%(+!,# − ,
-#∑ %&'()*$,#
*++%(+$,#
-./, (1)
𝑅𝑖𝑠𝑘!,# = 901,, ∑0$/,(𝐴𝑑𝑗𝑅𝑂𝐴!,$−0,∑0$/,𝐴𝑑𝑗𝑅𝑂𝐴!,$)2|𝑁 (2) where i represents the sample firms, n represents the year of observation (n ∈ [1,5]), X represents the total number of all firms in the industry and k represents the 𝑘#3 firm in the same industry. In addition, N represents the length of the rolling time window, where N=5.
In the additional test section, I will adjust the rolling years to 3 years to further test whether the results of the empirical analysis remain appropriately robust as the variable setting change.
4.2.3 Internal Control
I use the Internal Control Index (ICI) to measure internal control. ICI is based on the five elements of internal control and the extent to which internal control objectives are achieved.
The company's own internal control evaluation report is further analyzed, after which a composite calculation is made based on the type of audit opinion and internal control deficiencies in the internal control report issued by the CPA in the company's financial report.
Therefore the ICI data we derived is reliable. I use the nature log of the ICI as the final measure of the internal control. After taking logarithm of the ICI, outliers can be removed, the absolute value of the data can be reduced and heteroskedasticity and covariance can be reduced accordingly. Taking logarithms also allows the internal control variables to conform to a normal distribution.
4.2.4 Control Variable
First, firm performance is included. Firms with more tangible assets are more able to invest for the venture (Malmendier et al., 2011), so tangibility is included as well. Firm
leverage can be seen as control variable because leverage has relation with corporate risk (Serfling, 2014; Hirshleifer et al., 2012). Another control variable is the growth (annual growth rate of operating revenue), controlling for its effect on corporate risk-taking decision (Jenson, 1993; Liu and Li, 2021). Considering corporate governance issues that may have an influence effect on our regression results, we also control the board size as refers to Liu and Li (2021).
CEO and chairman appointed by the same person can potentially increase manager's recognition of their own abilities and drive their tendency to show overconfidence in decision- making, so appointment is included as well. In order to reasonably control the impact of industry differences and year on the regression results, I add year and industry into control variables.
Note: Risk 2 is an alternative calculate measure of risk-taking behavior used in additional test.
Table 2 Variable Definition
VARIABLE CONCEPT DESCRIPTION
DEPENDENT
VARIABLE !"#$1!,# Risk-taking behavior Earnings volatility (5 rolling years)
!"#$2!,# Risk-taking behavior Earnings volatility (3 rolling years) INDEPENDENT
VARIABLE
()!,# Managerial
Overconfidence !"# %ℎ'(( ()(*+%,-(.′
*"0#(1.2%,"1
344 ()(*+%,-(.′ *"0#(1.2%,"1
MODERATOR *)!,# Internal Control ln(898)
CONTROL VARIABLE
+,-./-0123,!,# Firm Performance Firm’s ROA
4125"6"7"89!,# Firm Tangibility D(2' − (1F %21G,H4( 2..(%
(IIJ)
D(2' − (1F %"%24 2..(%
:,;,-15,!,# Firm leverage K"1G − %('0 F(H%
K"1G − %('0 2..(%
<-/=8ℎ!,# Firm growth Annual growth rate of operating revenue
?@!,# Board Size ln(%"%24 1+0H(' "L F,'(*%"'.) A+B!,# CEO and chairman
appointed by the same person
If CEO and Chairman are appointed by the same person，A+B!,#=1, otherwise 0.
C,1-!,# Dummy variable
*2DE#8-9!,# Dummy variable
4.2.5 Libby Box
4.2.6 Regression Model
To test hypothesis 1, I construct model 1, as shown in equation (1)，
𝑅𝑖𝑠𝑘!,# = 𝛼4+ 𝛼,𝑂𝐶!,#+ 𝛼2𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒!,#+ 𝛼5𝑇𝑎𝑛𝑔𝑖𝑏𝑖𝑙𝑖𝑡𝑦!,#+ 𝛼6𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒!,#+
𝛼7𝐺𝑟𝑜𝑤𝑡ℎ!,#+ 𝛼8𝐵𝑆!,#+ 𝛼9𝐴𝑃𝑀!,# + 𝑌𝑒𝑎𝑟 + 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 + 𝜀 (1) If the coefficients of 𝑂𝐶!,# are positive and significant (𝛼, > 0), then hypothesis 1 holds.
Managerial overconfidence is all positively related to risk-taking behavior, i.e. the more overconfident the manager, the higher the level of risk-taking of the firm.
To test hypothesis 2, I construct model 2, as shown in equation (2), 𝑅𝑖𝑠𝑘!,# = 𝛽4+ 𝛽,𝑂𝐶!,#+ 𝛽2𝛪𝐶!,#+ 𝛽5𝑂𝐶!,#× 𝛪𝐶!,#+ 𝛽6𝑃𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒!,#+
𝛽7𝑇𝑎𝑛𝑔𝑖𝑏𝑖𝑙𝑖𝑡𝑦!,#+ 𝛽8𝐿𝑒𝑣𝑒𝑟𝑎𝑔𝑒!,#+ 𝛽9𝐺𝑟𝑜𝑤𝑡ℎ!,# + 𝛽:𝐵𝑆!,# + 𝛽;𝐴𝑃𝑀!,# + 𝑌𝑒𝑎𝑟 +
𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦 + 𝜀 (2) If hypothesis 2 holds, the interaction coefficient of 𝑂𝐶!,#× 𝛪𝐶!,# should be significantly negative (𝛽5<0). That is, improving the level of internal controls can weaken the positive correlation between managerial overconfidence and risk-taking behavior.
I Managerial Overconfidence
I Risk-taking behavior
Earnings volatility Internal control
(ICI)
Relative compensation
Control Variables
5 Empirical Results
Firstly, I performed summary statistics on all the variables included in the regression model using STATA17.0. To circumvent potential multicollinearity between the variables in the regression model, I then performed a correlation analysis. Finally I conducted regression analysis and additional tests. The results are presented below.
5.1 Main results
5.1.1 Descriptive statistics
Panel A of Table 3 shows the statistical characteristics of the key variables. The mean of the risk-taking behavior (Risk1) is 0.0266, with a range of 0.0004 to 0.2229 and a standard deviation of 0.0261. This shows that there is little difference in risk-taking behaviors between companies. The mean value of managerial overconfidence (OC) is 0.5069 which indicating that overconfident managers account for nearly half of the sample. The internal control (IC) of companies as measured by the internal control index in the sample indicates that most listed companies in the sample have high quality and low variation in internal control.
Next, I analyze the control variables in this paper. Firm performance (ROA) has a mean value of 0.0877, ranging from -0.0521 to 0.2456, indicating a wide range of firm performance among the sample firms. Similarly, the mean value of Tangibility was 0.2466, ranging from - 0.0034 to 0.7097, indicating a wide variation in the proportion of tangible assets to total assets among the sample companies. Based on the range of Leverage (-0.2249 to 18.0179) we can see that the ratio of long-term liabilities to long-term assets of the sample companies fluctuates considerably, and that leverage varies greatly between companies. Also the standard deviation of the Leverage is 0.8105 indicates a high volatility between sample companies. The standard deviation of growth rate of operating revenue (Growth) was 111.8704. The min value of Growth is -837.7816 and the max value is 730.0777, which indicates that the percentage of annual growth rate of operating revenue varies greatly between the sample companies. The mean board size (BS) was 2.3437, with min and max values of 1.6094 and 3.2581 respectively, indicating a relatively small deviation of board size in the sample companies. The mean value of CEO and chairman appointed by the same person (APM) is 0.2101, with standard deviation of 0.4074, which indicates that in most of the sample companies CEO and chairman are not served by the same person and that the companies have a better governance structure.
Panel B in Table 3 shows the correlations between the main variables of the regression model. From the table we can see that the correlation coefficients between the variables are all less than 0.4, so there is no multicollinearity problem. There is a positive correlation between managerial overconfidence (OC) and risk-taking behavior (Risk1) and statistically significant, which initially verifies the positive relationship of our first hypothesis. Secondly, there is a negative relationship between internal control (IC) and managerial overconfidence (OC) and risk-taking behavior (Risk1), which tentatively confirms the inhibitory effect of internal control as a moderating variable on managerial overconfidence and risk-taking behavior.
5.1.2 Regression analysis
Table 4 presents the OLS regression results for equation (1) and (2). Equation (1) was used to test hypothesis 1 about the relationship of managerial overconfidence and risk-taking behavior. We hope to find a positive relationship between two variables that managerial overconfidence promotes risk-taking behavior. In Equation (2), I was mainly examining whether higher level of internal control can weaken the relationship between managerial
Table 3 Descriptive Statistics Panel A: Summary statistics
Variables Obs Mean SD Median Min Max
Risk 4074 0.0266 0.0261 0.0186 0.0004 0.2229
OC 4074 0.5069 0.5000 1.0000 0.0000 1.0000
IC 4074 6.5175 0.1023 6.5261 6.1212 6.7575
Performance 4074 0.0877 0.0470 0.0821 -0.0521 0.2456
Tangibility 4074 0.2466 0.1777 0.2103 0.0034 0.7097
Leverage 4074 0.3075 0.8105 0.1450 -0.2249 18.0179
Growth 4074 -1.8457 111.8704 7.9902 -837.7816 730.8777
BS 4074 2.3437 0.2548 2.3026 1.6094 3.2581
APM 4074 0.2101 0.4074 0.0000 0.0000 1.0000
Panel B: Pairwise correlations
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9)
(1) OC 1.000
(2) Risk1 0.057* 1.000
(3) IC -0.032 -0.113* 1.000
(4) Perf. -0.034 0.083* 0.281* 1.000
(5) Tang. -0.033 0.087* -0.062* 0.234* 1.000
(6) Growth. -0.003 0.002 -0.009 0.005 -0.008 1.000
(7) Lever. 0.007 -0.110* 0.039 -0.191* -0.185* 0.001 1.000
(8) BS -0.077* -0.060* 0.027 -0.009 0.158* -0.009 -0.012 1.000
(9) APM 0.065* -0.007 -0.036 -0.009 -0.088* -0.004 -0.018 -0.138* 1.000 Note: *** p<0.01, ** p<0.05, * p<0.1
overconfidence and risk-taking behavior. In the regression analysis I control for the industry and year fixed effects on the regression results.
Based on the regression results of equation (1), the coefficient (𝛼,)of OC is equal to 0.0027 which is greater than zero, indicating a positive relationship between managerial overconfidence and risk-taking behavior. In addition, the p-value<0.001 means the positive relationship between managerial overconfidence and risk-taking behavior is significantly positive. This indicates that a 1% increase in managerial overconfidence is associated with a 0.27% increase in risk-taking behavior. Overconfident managers are more likely to increase the risk-taking behavior in their firms, supporting for Hypothesis 1.
Based on the regression results of equation (2), the coefficient (𝛽,) of OC is significantly positive (𝛽,=0.0026, p<0.001), shows that managerial overconfidence promotes risk-taking behavior. The interaction coefficient of 𝑂𝐶!,#× 𝛪𝐶!,# indicating a negative relationship and statistically significant (𝛽5=-0.0176<0, p<0.05). This regression result supports for Hypothesis
Table 4
Regression Results for Equation (1) and (2)
(1) (2)
H1 H2
OC 0.0027*** 0.0026***
(3.5684) (3.4683)
IC -0.0279***
(-6.6079)
OC*IC -0.0176**
(-2.3350)
Performance 0.0207** 0.0414***
(2.1221) (4.0334)
Tangibility 0.0005 -0.0026
(0.1661) (-0.8786)
Growth 0.0000 0.0000
(0.0294) (0.0294)
Leverage -0.0012*** -0.0010***
(-4.0898) (-3.4019)
BS -0.0045*** -0.0040**
(-2.7153) (-2.4057)
APM -0.0025*** -0.0027***
(-2.7216) (-2.9433)
_cons 0.0349*** 0.2144***
(8.3797) (7.7120)
Year effect Included Included
Industry effect Included Included
Observation 4074 4074
R-Squared 0.2168 0.2275
Note: Table 4 presents the OLS regression results for equation (1) and (2) to test hypothesis 1 and hypothesis 2.
Hypothesis 1 tests whether managerial overconfidence promote risk-taking behavior. Hypothesis 2 tests whether higher level of internal control weaken the relationship between managerial overconfidence and risk-taking behavior.
Industry effect and year effect are included in regression but not reported for brevity. Significance at the 10%, 5%, and 1% level are indicated by *, **, ***, respectively. t-value are in parenthesis.
2, that is improving the level of internal control can weaken the positive relationship between managerial overconfidence and risk-taking behavior. Internal control can be used as an inhibiting mechanism to curb the effect of managerial overconfidence on firm’s risk-taking behavior.
Figure1 visualizes the moderating effect of internal control on the relationship of managerial overconfidence and the corporate risk-taking behavior. We can see that the slope is steeper for low internal control conditions, indicating the positive effect of managerial overconfidence and the risk-taking behavior is more pronounced. Low internal controls will lack the discipline of overconfident managers and promote higher risk-taking. However the slope becomes flatter under high internal control conditions. This indicates that the positive effect of managerial overconfidence and risk-taking behavior is weakened under high internal control. Therefore, high internal control can act as a good corporate governance mechanism to dampen the high risk-taking behavior promoted by overconfident managers.
Figure1 Moderator Effect Graph
0.2 0.205 0.21 0.215 0.22 0.225 0.23
Low OC High OC
Risk-taking behavior
Low IC High IC
5.2 Additional tests
In order to ensure the reliability of the regression results, this paper adopted the following two approaches to conduct additional tests: 1) First is changing variable setting. Many Chinese scholars (Yu et al., 2013, Huang, 2019) use 3 years rolling window to calculate risk-taking behavior because the normal tenure for executives in China is generally three years. Therefore, I change the rolling window of adj. ROA from 5 to 3 years in calculating the risk-taking behavior for re-regression. However, this change would not conflict with the 5-year rolling window used for calculating risk-taking behavior in the main regression part, as many company managers are eligible for re-election. Due to the narrowing of the rolling window from 5 years to 3 years, I have accordingly extended the sample period by 2 years to ensure that this is the longest sample period can be observed. 2) Second is dividing the sample into SOE (state-owned enterprises) and non-SOE groups based on the nature of ownership to test two hypotheses. We mainly examine whether the results still hold under different ownership. The results of additional tests are shown in Table 5, Table 6.
Table 5 presents the first OLS additional regression test of the two hypotheses by setting the rolling-window to 5 years in calculating risk-taking behavior (Risk 2) and do additional test of two hypothesis. Column (1) shows that the regression result for hypothesis1. We hope to find a positive relationship between two variables that managerial overconfidence promotes risk-taking behavior. Column (2) shows the regression result for hypothesis 2 that mainly examines whether higher level of internal control can weaken the relationship between managerial overconfidence and risk-taking behavior. We can see from column (1) that the positive effect of managerial overconfidence on firm risk-taking level remains significant after extending the rolling window of adj. ROA (𝛼,=0.0014，p<0.05). Secondly, from column (2) we see that the interaction coefficient 𝑂𝐶!,#× 𝛪𝐶!,# remains negative and statistically significant (𝛽5=-0.0166<0, p<0.05). That is, managerial overconfidence promotes risk-taking behavior, while internal control can act as an internal constraint mechanism to moderate the positive relationship between managerial overconfidence and risk-taking behavior. This is consistent with the results we obtained in our regression analysis part and further supports H1 and H2.
