Risk disclosure in family firms: Does ownership influence
risk disclosure?
by
Sybren van der Werff
Prof. dr. D. Swagerman, Supervisor
Thesis in Accountancy & Controlling, for the
requirements of the Degree of Master of Science
UNIVERSITY OF GRONINGEN
Faculty of Economics & Business
2016
2
Personal information
Name Sybren van der Werff
Date of birth 14/05/1992
Student number 2016257
E-Mail s.van.der.werff.2@student.rug.nl
E-mail (alternative) Sybren_werff@Hotmail.com
Phone +31 (0)6 13606190
Study program Accountancy & Controlling
Supervisor Prof. dr. D. Swagerman
Second assessor Prof. Dr. Van der Meer-Kooistra
Abstract
This research investigates the risk disclosure practices of family firms using a quantitative approach. When exploring the agency problems associated with family ownership,
questions arose regarding why a firm would disclose risk if shareholders have the incentives, opportunity, time and resources to monitor the firm without the publicly disclosed information. This research tests the differences in the risk disclosure of family firms and non-family firms in the Netherlands. The risk disclosure of these companies was tested based on quantitative and qualitative (readability) dimensions. The outcomes of this research are inconclusive and most hypotheses are not (fully) supported. In the end, the research question: ‘Do the different characteristics of family and non-family
firms have an effect on their risk disclosure practices?’ is answered negatively. Some
recommendations are given for further research and for businesses.
3 Acknowledgement:
I like to express my thanks to the people that helped me and support me in any way during the writing of my thesis. In this regard I’d like to thank Deloitte for offering me the resources to help me writing my thesis. Furthermore I’d like to thank my supervisor from the RUG Dirk Swagerman.
The help was appreciated and I hope this thesis will be interesting to read. Many thanks,
4 Contents 1. Introduction ... 5 Research question ... 7 Thesis outline ... 7 2. Literature review ... 8 2.1 Risk Management ... 8
2.2 Definition of the family business ... 10
2.3 Stakeholder theory ... 10
2.4 Agency theory and voluntary disclosure vs. mandatory disclosure ... 11
2. 5 Family businesses and the agency theory ... 12
2. 6 Risk disclosure ... 13
2.5.1 Effects of risk disclosure ... 14
3. Theoretical framework... 16
3.1 Research gap ... 16
4. Methodology ... 20
4.1 Sample and data collection ... 20
4.2 Dependent variables ... 21 4.2.1 Risk quantity ... 21 4.2.2 Risk quality ... 22 4.3 Independent variables ... 25 4.4 Control variables ... 25 4.5 Data analysis ... 26 5. Results ... 28 5.1 Descriptive statistics ... 28 5.2 Correlation ... 29 5. 3 Independent t-test: ... 31 5.4 Regressions ... 33
5.4.1 Dependent variable RiskRatio ... 33
5.4.2 Dependent variable RiskWords ... 36
5.4.3 Dependent variable for readability ... 40
6. Conclusion ... 43
7. Discussion ... 44
7.1 Implications for managers ... 44
7.2 Limitations and implications for future research... 46
7.3 Relevance in society and implications for regulators ... 47
5
1. Introduction
This research focuses on risk disclosure in family businesses. Family businesses have always been one of the drivers of the economy. For example, the Nyenrode Business University estimated that 69% of all businesses in the Dutch economy could be described as family businesses and that they account for more than 50% of the Netherlands’ GDP. They also account for almost 49% of the total jobs in the Netherlands. Research by PWC (2016) indicates that family businesses in the Netherlands created 8% more jobs while the total number of jobs in the Netherlands decreased by 0,4% over the period 2010-2014. In recent times, also investors have had an increased interest in family businesses because of their steady returns. (Lins, Volpin, Wagner, 2013).
Regarding family business and risk management, there are some interesting studies conducted by different public and private organizations. Erasmus University (2016) found that family businesses face above average risks, because they deal with problems of
succession. Additionally, a study by Deloitte (2016) found that family businesses view risk taking as one of the key subjects in their firms. Finally, PWC (2014) found that the
professionalization of family firms on risk management was one of the main priorities for the future of the firms.
Recent events like major corporate scandals have led to an increased interest in risk reporting (Oliveira, Rodrigues, and Craig 2013). By providing stakeholders with information, an organization can form a better connection with the key stakeholders and enhance
accountability. According to Beretta and Bozzolan (2004), “Disclosure is the communication
of information regarding a company’s strategies, characteristics, operations and other external factors that have the potential to affect the expected results”. In other words, risk
disclosure is the formal communication of information regarding the risk of a company. Researchers and regulators pointed to risk management as one of the most important tools for businesses to deal with uncertainty. Some scholars, such as Magnan and Markarian (2011), link risk disclosure with the aforementioned financial crisis. Risk disclosure is a way to restore stakeholders’ trust by disclosing (voluntary) information about risk and risk management (Deumes & Drechel, 2008). If for example shareholders and creditors do not observe companies’ risk management activities directly, they will tend to institute monitoring
6
systems to increase the flow of information about those activities, and to reduce uncertainty (Linsmeier et al., 2002).
Other researchers argue that the current form of risk disclosure is too general and of little use to investors (Abraham & Cox, 2007; ASB, 2009; Lajili & Zéghal, 2005; Linsley & Shrives, 2006; Solomon et al., 2000). Abraham & Shrives (2014) found that disclosures do not change over time and do not seem to have a relation to the actual risks that are faced by an
organization. Nonetheless, both legal and political entities have also stressed the importance of risk disclosure, resulting in the implementation of regulatory reforms by many government agencies and accounting standard makers in an effort to improve corporate governance and transparency. Many countries now have some type of formal requirement for increased disclosure of risk information. Despite the increase, it appears that there is still a demand for improved risk reporting (ICAEW 2011). Examples of these efforts in a Dutch context are the Dutch Corporate Governance Code (Commissie Corporate Governance, 2003) and the research by the AFM (2014).
The amount of risk disclosure in companies’ annual reports gives an indication of the amount of information that is communicated to the public and is open to all stakeholders. As it is reasonable to claim that companies have more internal risk information available in order to manage and control the business on a daily basis, it is interesting to ask why this information is not disclosed. Some argue that managers are inclined to disclose risk-related information to improve the corporation’s image and inform stakeholders about their managerial skills in managing risks (Iatridis, 2008). Others discuss how high levels of disclosure reduce agency costs, compliance costs and information asymmetry (Lundholm and Winkle, 2006; Skinner, 1994; Healy and Palepu, 2001). Abraham and Cox (2007), Deumes and Knechel (2008), and Lajili (2007) used agency theory to explain motivations for risk disclosure. For this research the agency theory and information asymmetry problems are highlighted in a family business setting. Compared to non-family firms, family firms in the US face less severe agency problems that arise from the separation of ownership and management (Type I agency problems). However, they are characterized by more severe agency problems that arise between controlling and non-controlling shareholders (Type II agency problems) (Gilson and Gordon, 2003; Ali, Chen, & Radhakrishnan, 2007). Overall Ali et al. (2007) found evidence to suggest that the severity in agency problems is a likely factor in the differences in
7
businesses, discussed in chapter 2 of this research, could cause the risk information disclosed in family businesses to be different from non-family businesses.
Research question
Do the different characteristics of family and non-family firms have an effect on their risk disclosure practices?
Thesis outline
This thesis continues as follows. The literature review outlines the current views in the relevant research fields. Next, the theoretical framework outlines the research gap and the hypotheses that are used to answer the research question. The methodology chapter describes the variables used in the analysis and explains the reliability, validity and controllability of this research. Following this, the results of the research are presented. This thesis ends with a conclusion and discussion.
8 2. Literature review
This chapter provides an overview of the main theories related to risk disclosure and related variables. Furthermore, it elaborates on past research in the field related to risk disclosure and family ownership.
2.1 Risk Management
To disclose information regarding risks, an organization needs to have some kind of risk management in place. Every organization deals with risk in its everyday operations. Risk, as described by COSO (2004), is: “the possibility that an event will occur and adversely affect the achievement of objectives” (paginanummer toevoegen). Therefore, it is imperative to design, implement and maintain an effective risk management system, especially within the current challenging economic environment. This paper explores the various definitions of risk, but mainly focuses on the actual risk disclosed by organizations rather than looking for possible risks that are not disclosed. The concept of risk partially overlaps with the concept of uncertainty. It is impossible to view both terms as mutually exclusive, but risk has
measurable aspects, while uncertainty does not. By being able to manage risks, an
organization can effectively cope with the risks and strengthen its capacity to create value. This leads to the question of how to define risk management. According to Emanuels et al. (2010b, p. 150), it is “a system that enables management to identify, prioritize, analyze and control the relevant risks which may be threatening the organization’s ability to meet its objectives”. One of the most generally accepted definitions is provided by COSO (The Committee of Sponsoring Organizations of the Treadway Commission) (2004),1 which defines risk management as:
“A process, effected by an entity’s board of directors, management and other personnel, applied in strategy setting and across the enterprise, designed to identify potential events that may affect the entity, and manage risk to be within its risk appetite, to provide reasonable assurance regarding the achievement of entity objectives”.
COSO developed a wide spread model for risk management. The model offers guidelines for internal control and how to deal with risk. In addition to providing a definition, COSO (2004) formulated an eight step system for good risk management. For the purpose of this paper, step 7 ‘reporting’ is especially important, as the COSO model states that risk management
9
should be reported on. As may be deduced from COSO’s definition, the purpose of risk management is to enable an organization to manage all types of threats that could impede the attainment of its objectives.
Risk management can add value to a corporation’s key stakeholders, such as the shareholders and prospective investors, by improving the firm’s financial performance and enabling investors to make better valuation judgments. The potential benefit of risk management provides investors with an incentive to become informed about the internal risk management process. Their aim is to monitor management’s attempts to manage risks and to develop their own judgment about the effectiveness of the system (Deumes & Knechel, 2008; Michelon et al., 2009).
According to Deumes (2008, p. 122), being informed allows stakeholders “to make more
accurate corrections for risk when they value their investments, thereby preventing stock prices from becoming unrelated to the intrinsic value of the company”. However, it should be
noted that a stakeholder cannot directly observe an organization’s risk management, as it is an internal process accessible to and observable only to people within the organization (Deumes & Knechel, 2008; Michelon et al., 2009). Hence, it is vital for management to disclose any information related to the risk management system as well as the risk factors and their impact. Only through risk-related disclosure can external stakeholders be informed about the nature, extent and quality of the risk management system (Deumes & Knechel, 2008). The value of risk management for this paper is that companies often use a system for risk management that is based on or similar to COSO and companies should thus at least think about the reporting step in the COSO model. However, it should be noted that risk management could be a costly endeavor and therefore smaller companies often do not fully adopt the COSO or another risk management system.
10
2.2 Definition of the family business
There are many definitions of a family business. Chua, Chrisman and Sharma (1999) offer a comprehensive and thorough study on the definitions of family businesses. While some definitions might be stricter than others, the key distinctions are almost always based on ownership and control. While ownership is a fairly straightforward concept, control can be harder to define. Often this is based on the controlling ownership, which differs from pure ownership in that some shareholders might have more power than others. Other definitions often incorporate the family members being involved in the management of a business. This research takes to the definition of Flören (1998), because he highlights the importance of family businesses in the Netherlands. His definition uses three key variables. For a company to qualify as a family business, it must comply with at least two out of three variables: (1) More than 50% of the ownership is held by a single family; (2) A single family has a decisive influence on the decision-making process, such as strategy and succession, either managerial or executive; (3) A majority or at least two members of the daily management are members of one family.
2.3 Stakeholder theory
The focus of the modern stakeholder theory is that an organization is dependent on all its stakeholders and that it needs to disclose information to the stakeholders rather than only to the shareholders (Freeman, 1984). A narrow definition of stakeholder is “groups which the
organization is depending on for its continued survival” (Freeman, 1983). Later Freeman
(1984) provides a broader definition: “any group or individual that can affect or is affected
by the achievement of a corporation’s purpose”. The stakeholder theory states that groups or
individuals that could have an effect on an organization, or be affected by it, should be taken into consideration and the needs and wishes of these groups should be monitored by
management. This means an organization needs an explicit strategy for dealing with its stakeholders. The goal of stakeholder management in the stakeholder theory is that the interests of all stakeholders align and the shareholders are among the stakeholders (Freeman 2004). Finally, Freeman states that the concept of stakeholder should be used for the
formulation of a new non-shareholder theory of the organizations. Stakeholder theory is useful to consider in this research because of the concentration of ownership in family firms. This concentration could lead to a very shareholder-centric view of the firm. Stakeholder theory can help explain why these family firms should still consider risk disclosure. In other words, the stakeholder theory explains that there are parties other than the shareholders who
11
might have an interest in or interdependence with the business and would place a premium on risk disclosure, such as government bodies or environmental groups. Healy and Palepu (2001) further state that corporate disclosure is critical for the functioning of an efficient capital market; this can also be directed to stakeholders such as banks, rather than only the traditional shareholders (Deumes, 2008).
2.4 Agency theory and voluntary disclosure vs. mandatory disclosure
The agency theory explains the discrepancies in information between owner and principal. The agency theorists focus on identifying situations in which the principal and agent are likely to have conflicting goals and describe the governance mechanisms that limit the agent’s self-serving behavior (Eisenhardt, 1989).
Literature identifies two types of agency problems. The first type of agency problems arises from the separation of ownership and management (Type I agency problem). The separation of managers from shareholders may lead to managers not acting in the best interest of the shareholders. The second type of agency problems arises from conflicts between controlling and non-controlling shareholders (Type II agency problem). Controlling shareholders may seek private benefits at the expense of non-controlling shareholders.(Ali et al., 2007)
Executives (the agents) might act on their personal goals, which may not be perfectly aligned with the interests of the company and specifically of the investors (owners). To be able to reach these goals, it might be in the best interest of management to maintain the gap in information between the insiders and outsiders. The outcome of the agency problems is increased costs, called agency costs (Eisenhardt, 1989).
A possible solution to close information asymmetry and counter agency problems is
voluntary disclosure (Healy & Palepu, 2001). Voluntary disclosure can be defined as all the disclosures that companies are not obliged to provide. For mandatory disclosure, the
minimum requirements are recorded in the laws and regulations. Corporations can also provide additional information through voluntary disclosure (Healy and Palepu 2001).
Voluntarily disclosing information reduces information asymmetries between the firm and its investors (Diamond & Verrecchia, 1991; Kim & Verrecchia, 1994). According to Chueng (2010), investors are more responsive to voluntary disclosure than mandatory disclosure. Ball et al. (2012) find that mandatory and voluntary disclosures are complementary. Strategic disclosure processes are complex, since voluntary and mandatory information have to be
12
combined as requested by the stakeholders. The amount of voluntary and mandatory
disclosure is therefore not static in time. Disclosing more information than strictly necessary regarding risk management can be classified as voluntary disclosure and the reasoning behind such disclosure should be researched.
2. 5 Family businesses and the agency theory
The agency theory is different if applied to the family firm. Several characteristics of family firms reduce the likelihood of managers acting in their own interest rather than the interest of the shareholders. Families tend to have large equity shares in the firms. This more or less eliminates the free rider problem that is often found in firms with small autonomous shareholders. Families, much like large block holders, have strong incentives to monitor managers and the firm as a whole (Demsetz & Lehn 1985). Therefore, managers are more inclined to act in the interest of the family to avoid the scrutiny of the largest shareholders. The statement made by Demsetz and Lehn (1985) implies that family owners are usually more actively involved in firm management, for example, by serving as executives or directors, which decreases type 1 agency problems. Thus, family owners have better access to information and can better monitor management, reducing the agency problem between management and shareholders. As the involvement and commitment of the family is known to other (non-related) shareholders, they could opt to free-ride on family owners’ monitoring of the firm and thus have lower demand for public information. Direct monitoring and corporate disclosure are substitutes in alleviating agency problems (Bushman et al., 2004). Furthermore, it is probable that the family has knowledge about the primary activities of the firm. This knowledge helps the family monitor the managers more effectively (Anderson & Reeb, 2003).
A characteristic of family firms is that they tend to have a longer investment horizon. The firm is meant to remain in the family rather than only provide high dividends to current shareholders. This characteristic reduces the riskiness of investments made by the managers (James 1999; Stein, 1989). Compared to non-family firms, family firms face less severe hidden-action and hidden-information agency problems due to the characteristics of the separation of ownership and management. However, concentrated ownership of founding families gives them power to seek benefits at the expense of other shareholders. Controlling shareholders can seek such private benefits by freezing out minority shareholders (Gilson &Gordon, 2003), by taking a larger share of the firm’s earnings in the form of special
13
dividends (DeAngelo & DeAngelo 2000), by engaging in related-party transactions that benefit the controlling party and not the firm as a whole (Anderson & Reeb, 2003a), or through managerial entrenchment (Shleifer & Vishny 1997). All of these factors could lead to family firms facing more agency problems from the conflict between controlling and non-controlling shareholders. This is why there could be a difference in the demand for
disclosure between the two parties. When families try to put their interests above other parties, at least economically, and the market finds out, they may incur substantial costs in the form of lower equity value following the distrust of non-family shareholders. This is especially interesting because families tend to have a large amount of equity and they tend to hold their firms’ equities for long periods (Shleifer & Vishney, 1997; La Porta et al., 1997, 1998, 2000).
There are arguments to be made for relatively more disclosure by family firms. Relative to other shareholders, family owners usually have large concentrated equity holdings and are less diversified, causing them to have more risk on their overall investments. Prior research demonstrates that voluntary disclosure leads to lower bid-ask spreads, lower information risk, and lower cost of raising capital. This suggests that family owners, who stand to gain greater benefits, would prefer more public disclosure to counter the risk of their less
diversified investment. Family owners are arguably more concerned with possible negative impacts on firm value. Therefore, they demand more disclosure to avoid costly litigation problems.
2. 6 Risk disclosure
Risk disclosure can be classified into three levels: internal, intermediate and external (Lajili & Zéghal, 2005). This research focuses on the external risk disclosure. Linsley and Shrives (2006) claim that risk disclosure constitutes informing the public of any opportunity or prospect, or of any hazard, danger, harm, threat or exposure that has impacted or may impact the company.
The three levels of risk disclosure are as follows. 1) Internal risk disclosure encompasses the sharing of information about risk identification, measurement, performance development and monitoring between management and employees. This type of disclosure generally occurs during informal internal meetings and aims to increase efficiency in the realization of an organization’s objectives. 2) Risk disclosure at an intermediate level occurs when
14
the board in their role of overseeing the risk management process. 3) External risk disclosure, also referred to as public disclosure, involves communication of risk information to the public, which generally occurs through periodic prospectuses or the organization’s annual report.
As mentioned above, disclosure can be categorized into mandatory and voluntary public disclosure. As reported by risk disclosure studies, such as Ashbaugh-Skaife et al., (2007) and Emanuels et al. (2010), the choice to voluntarily report risk information depends on the cost-benefit consideration.
2.5.1 Effects of risk disclosure
Prior studies on listed companies (e.g., Linsley & Shrives, 2000; Armitage & Marston, 2008; Farvaque et al., 2011) have presented several benefits and costs of risk disclosure. In the private sector, risk disclosure tends to provide investors with forward-looking information (Linsley & Shrives, 2000), which contains details about the organization’s future prospects. This is advantageous in that it reduces the information asymmetry between the corporation and the investors (Emanuels et al., 2010a). It may also enable investors to make a better stock valuation (Deumes, 2008). Linsley & Shrives (2000) argue that corporations that disclose risk-related information place themselves in a position of public scrutiny, attracting investors to carefully examine how effective the management is in dealing with risk. Accordingly, Hermanson (2000) proves that financial statement users perceive reporting on internal control as a stimulus for management to improve internal controls. Thus, one of the benefits that private corporations could attain through risk disclosure is the improvement, or at least a perceived improvement, of risk management. Which should be expected by every company that endulges in risk disclosure practices. (Linsley & Shrives, 2000)
An important aspect of the disclosure theory is that the decision of a manager to voluntarily disclose information is driven by the interpretation of investors. The choice to not disclose information by a manager can have the result that the investors will price protect themselves, because they have a lower amount of information available (Verrecchia, 2001; Hope, Kang, Thomas & Vasvari, 2009). Internal control has a positive effect on the quality of the financial reporting and adds to the reduction of the information risk. Investors are thereby reassured that the financial information is reliable as a result of the internal control system. This results in a lower incentive for the investors to protect themselves by asking for a higher reward (Ogneva, Subramanyam & Raghunandan, 2007).
15
Risk management disclosure can also increase accountability. Owusu-Ansah & Ganguli (2010) claim that increased transparency with regard to risk management inhibits
opportunism. By disclosing more risk, the managers open themselves up to more scrutiny. This leads to managers needing to justify and account for their actions more. The result would often be that management is less likely to pursue risky ventures that only relate to personal goals (Linsley & Shrives, 2000). Disclosure of risk-related information can also enhance organizational legitimacy. Organizational legitimacy is “the congruency
organizations seek to establish between the social values associated with or implied by their activities and the norms of acceptable behavior in the larger social system of which they are a part” (O’Sullivan & O’Dwyer, 2009). In other words, organizations “seek to ensure that they operate within the bounds and norms of their respective societies” (O’Sullivan & O’Dwyer, 2009). An organization’s legitimacy hinges on the ‘social contract’ between the organization and society. Social contract refers to the expectations society has about how an organization should operate (Deegan et al., 2002).
Failure to comply with these expectations may lead society to perceive the organization as illegitimate. As a result, public acceptance of the organization may decline. This may be evidenced by, for instance, the loss of customers, financiers reducing the supply of capital to the organization, or constituents lobbying the government for increased taxes, fines or laws to prohibit those activities that are unacceptable to society (Deegan, 2002). Legitimacy may be considered as an intangible asset important for the growth and survival of organizations (Oliveira et al., 2011b). This is especially true for organizations that are considerably more visible, such as large or listed companies. Hence, it is imperative to establish legitimacy, maintain it, and restore it when it is damaged through public disclosure (O’Sullivan & O’Dwyer, 2009; Oliveira et al., 2011b). Organizations that inform the public about their activities may change the perceptions around them, and as such improve their legitimacy (Dowling & Pfeffer, 1975; O’Sullivan & O’Dwyer, 2009). Being transparent helps organizations earn the trust of stakeholders. Thus, an organization that discloses risk
information builds a reputation for openness and stakeholder confidence, which in turn helps the organization gain and preserve its legitimacy (Deumes & Drechel, 2008).
While there are benefits to the disclosure of risk, there are also costs associated with it. According to Farvaque et al. (2011), corporate disclosure is generally associated with the costs of developing and presenting the corporate information. To communicate the information to the public, there must be a process involving the collection, verification,
16
analysis, quality control, and accurate presentation of information. Before an organization can disclose reliable information, all the costs associated with the implementation and functioning of the reporting process must be considered.
The implementation of a risk management system could also be related to the cost of the disclosure of risk and vice versa. Almost every company is required to issue an annual report containing detailed financial and non-financial information, however there are additional costs related to voluntary disclosure. Additionally, often risk disclosure provides forward-looking information to a corporation’s stakeholders. One concern is that it is harder to verify this information and thus such information could be unreliable, as opposed to more verifiable past information (Linsley & Shrives, 2000). This could lead to legal disputes between the company’s management and the investors (Emanuels et al., 2010a). This is true for both the organization and for the management of the company. Management puts its reputation on the line by disclosing certain information in addition to opening itself up to legal disputes if the information is later deemed wrong (Mcmullen et al., 1996).
Managers want to preserve or enhance their credibility and avoid potential damage to it (Graham et al., 2005). Damage to a manager’s reputation can occur when the manager only discloses favorable information; this results in a decrease in the effectiveness of the disclosed information and increases the information risk. Finally, an important consequence of the voluntary disclosure is that it also provides the organization’s competitors with proprietary information (Verrecchia, 2001). For example, the release of internal control information may include the main risks of a company and how they are managed (Deumes & Knechel, 2008). The capital needs theory goes deeper into the financing of an organization and the motivation for disclosure. The theory states that the need for capital is the primary reason for
organizations to voluntarily disclose information (Abd-Elsalam & Weetman 2003). More disclosure about the operations and organization reduces the uncertainty of external providers of capital and results in a lower cost of capital. As a result, an organization will have more possibilities to finance its operations (Abd-Elsalam & Weetman, 2003; Francis et al., 2005). In other words, disclosing risks could lead to a lower cost of capital, which will be desirable to every business, including family businesses.
3. Theoretical framework
17
Decisions about risk disclosures create a whole new set of challenges for firms. The amount of information disclosed and the specifics of the disclosure seem contingent on several factors. In fact, as risk disclosure involves substantial costs relating to litigation, copyrights, competition, regulation and taxation (Ntim, 2013; Lajili & Zeghal, 2005), it has been argued that in the absence of potential direct and indirect benefits, rational managers do not
voluntarily engage in risk disclosure (Lopes & Rodrigues, 2007). Thus, risk disclosure is dependent on the interests of owners and stakeholders.
The results of this research could also have implications for businesses and managers. Characteristics specific to the family firm have always had the interest of policy makers as family firms have always been a large part of the business world. Family firms often have a long term outlook (Miller & Le Breton-Miller, 2006), which would suggest a less risky approach and therefore less risk to disclose. Another interesting theoretical approach is that by disclosing information about risks and risk management, the company gives away information that could favour its competitors and be costly to the firm. Will the family firm, with its increased monitoring and lower risk profile, be inclined to disclose risk information? The implications of this research could provide interesting results regarding the usefulness of the risk disclosure of family firms for investors and capital providers. Do these parties get risk information from family firms in the same manner as from non-family firms or do they get risk information in another way? There are signs that point to extra investor interest in family businesses because of their disproportional returns and safety. This leads to the research question:
RQ: Do the different characteristics of family and non-family firms have an effect on their risk disclosure practices?
To answer this question, a correct methodology must be chosen. As seen in former research cited in chapter 2, risk disclosure is done to decrease agency problems. Miller & Le Breton-Miller (2006) concludes that when a family member is in charge of the
business there are significantly less agency costs involved. However, they also find that when control is too centralized there is a perceived opportunity to grant the family
excessive wealth over other stakeholders. Since agency problems based on the monitoring of companies are more limited for family firms, due to the family businesses greater incentive to monitor a company in the first place. Because the family has a greater understanding of the firm’s operation, it is feasible to say that the risks involved are also
18
better known. This smaller information gap raises the question of why a company would disclose more risk information than necessary. Research has shown that there are costs involved with disclosing risk, such as a potential for competition to take advantage of this information. Indeed, a vast majority of the studies empirically confirmed this predicted negative association between management ownership and the extent of disclosure (Eng & Mak 2003; Deumes 2004). However, Leng and Ding (2011) found no significant
association in this regard. This paper follows the suggested theoretical reasoning. Therefore, the first hypothesis is as follows:
H1: The quantity of risk disclosure is lower for companies that have a family member involved in daily management.
Due to the different agency problems facing family businesses they are either less likely to disclose or more likely to disclose information. Ali et al. (2007) argue that the direction of the relationship is dependent on the severity of agency problems. Family businesses facing more severe type 1 agency problems are likely to disclose less. Whereas family firms with severe type 2 agency problems are expected to disclose more. It is feasible that the risk disclosure of family firms differs significantly from non-family owned companies. Stakeholder- and capital needs theory offer arguments as to why family firms have similar incentives to disclose information, as they have to manage stakeholders and creditors just as non-family counterparts. Finally, Zellweger and Kammerlander (2015) argue that united families are a myth and they tend to be much more heterogeneous. They present the argument that differing interest of individual family members create agency problems similar to non-family businesses. Based on the arguments presented above and in the literature review, the following hypotheses are formed:
H2: Family firms disclose a different amount of risk information than non-family firms
H2a: Family firms have less risk information disclosure than non-family firms (ratio and total)
H2b: Family firms have more risk disclosure than non-family firms (ratio and total)
Another issue is the difference in qualitative and quantitative disclosure. Skinner (1994) states that bad news disclosures tend to be qualitative statements, as opposed to quantitative good news disclosures. Additionally, Skinner (1994) mentions that voluntary disclosure is
19
often motivated by the fear of legal consequences. Voluntary disclosure reduces expected legal costs because it will be more difficult to argue that information is withheld, and because disclosing earlier rather than later reduces damages that a plaintiff could claim. Beretta and Bozzolan (2008) state that there is not a clear distinction between quantity and quality. It is generally assumed that the quality of disclosure can be determined by the quantity of
disclosure. However, Beretta and Bozzolan (2008) conclude that quantity is not a good proxy for the quality of the disclosure. Therefore, this study does not only test the quantity of the risk information, but also takes the readability as a measure of the quality of risk information. An increased readability adds to the understandability of a report. It is therefore feasible that the readability of a risk section is a good proxy for the quality of the risk section (Biddle, Hilary, & Verdi, 2009). To ensure that valid conclusions can be drawn regarding the main research question, this research also tests the quality of risk sections.
H3: The quality risk disclosure of family firms is different from non-family firms
H3a: The quality of risk paragraphs is higher for family firms
H3b: The quality of risk paragraphs is lower for non-family firms
Several prior studies provided evidence that high leverage is a determinant of the extent of risk information disclosure. Deumes and Knechel (2008) indicated that highly leveraged firms are more likely to report risk information. Additionally, Oliveira et al. (2011a) reported a positive association between high leverage and the amount of risk information disclosure. Several factors may lead an organization that relies heavily on external capital to disclose more risk information. The organization may be facing demands from investors for greater risk management transparency, or as Barako et al. (2006) argue, firms reliant on leverage disclose information to augment their chance to be eligible for loans and capital. Debt holders of highly leveraged firms may introduce more restrictive covenants into debt contracts, which will increase agency and monitoring costs. Risk information related to market, credit, and internal control risk may play a critical role in mitigating creditors’ concerns about the solvency of the firm and its capabilities to generate enough cash flows in the future (Rajab and Handley-Schachler, 2009).Furthermore, the capital needs theory also explains why companies in greater need of outside capital would disclose more information—either to lower the overall cost of capital or to attract external capital. From an agency theory perspective, highly leveraged companies induce higher agency costs and therefore need to
20
disclose more information to the creditors (Hunziker 2012). This leads to hypotheses 4 and 4a:
H4: The amount of leverage has a positive effect on the quantity of risk information disclosure
H4a: The amount of leverage has a positive effect on the relative amount of risk information disclosure
4. Methodology
This research uses a quantitative research methodology. Aangeven waarom voor een kwantitatieve onderzoeksmethodologie is gekozen. The methodology tries to enhance controllability and explain variables. Furthermore, it lists the statistical tests used and the operationalization of the variables.
Controllability is the prerequisite of the evaluation of the validity and the reliability of the study. The methodology chapter describes the tools used to make this study controllable. This is done by addressing the following questions: how were the data collected? How was the sample selected? Under what circumstances was the study executed? How were the data analyzed and conclusions drawn?
Reliability is created when the results of the study are independent of the particular
characteristics of the study and can therefore be replicated in other studies (Yin, 2003). The first potential bias in this study is that of the researcher. This bias was controlled for in this study by choosing set variables and using a sample that is not chosen by the researcher but is retrievable by any individual. While other forms of reliability, such as visual aids, are harder to control for in this research, they are addressed in the future research section of the
discussion chapter. The validity will be discussed by looking into the variables and
determining if the constructs are valid, furthermore the relationships between the variables are presented above.
4.1 Sample and data collection
To have a sample size that is robust and similar, the choice was made to limit the research to one country—the Netherlands. This also provides consistency in regulations. Furthermore, the economy of the Netherlands has a strong basis in family businesses. The Netherlands has both corporations listed on the Euronext in Amsterdam and a ready sample of large family
21
businesses, which insured that the research could be conducted in an efficient and timely manner.
To ensure the robustness of the sample size, the EURONEXT listings in Amsterdam were pulled from the official EURONEXT website. The family business sample consists of the 40 largest and 10 random, though sufficiently large, family businesses in the Netherlands. It should be noted that these two samples overlapped in a few instances and that the companies in both samples were only used once and not replaced by other entries. The sample of family businesses was retrieved from a (non-scientific) study conducted by Elsevier (2016) that investigated the largest family businesses in the Netherlands. The random businesses were added to ensure a similar range in the size of the family businesses as opposed to non-family businesses. Although it is not a scientific research, Elsevier used a clear definition of family businesses that is in line with this research.
Financial companies were eliminated based on industry codes found in the Orbis database. The financial industry is bound by different laws and regulations that would severely influence the outcome of the analysis. Furthermore, companies with revenues below 30 million Euros were also removed from the sample The final sample consists of 108 businesses. For further information on the sample, see the descriptive statistics section. For this sample the last published annual report was retrieved from either the website or the company info database. Eighty-six companies had already published an annual report on 2015 and for the remaining 22 companies the annual report of 2014 has been used. In order to analyze the text of the annual reports, Adobe DC was used for its copying and word count functions.
To enhance the controllability of this research, the entire sample is provided in an appendix A
4.2 Dependent variables
4.2.1 Risk quantity
The dependent variable in this study is the quantity of risk-related information disclosed in the annual reports of the sample organizations. To measure this variable, the content analysis approach is applied, which is one of the most widely used methods in risk disclosure studies (e.g., Linsley &Shrives, 2006; Abraham and Cox, 2007; Amran et al., 2009; Elzahar & Hussainey, 2012; Ntim et al., 2013). Content analysis involves classifying content units into various pre-defined categories (Beattie et al., 2004). Various content units (or units of
22
analysis) can be chosen, such as words, sentences, phrases, paragraphs, etc. (Unerman, 2000; Linsley & Shrives, 2006; Hooks & Van Staden, 2011; Ntim et al., 2013). In line with Ntim and Soobaroyen (2013) and Mahadeo et al. (2011), this study uses the number of words as the unit of analysis. Words are the smallest unit of analysis and therefore they offer the most robustness to this type of quantitative study (Wilmshurst & Frost, 2000). Using words can avoid inconsistency in measuring the extent of disclosure (Wilmshurst & Frost, 2000), which should enhance reliability (Ntim & Soobaroyen, 2013). Following the method of Beretta and Bozzolan (2004), the quantity of risk information disclosure is measured by the ratio of risk-related words to the total number of words in the annual report (RiskRatio).
Simultaneously, a robustness check is performed by using the absolute number of risk-related words (RiskWords). RiskRatio could be preferred over RiskWords, because the writing style of organizations may influence the results when determining the extent of the reported information (Abraham and Cox, 2007). A firm can disclose a substantial amount of risk information, but if the risk information is embedded in an even larger amount of non-risk information, the relevance of the risk information may be diminished. It could be argued that an overload of information diminishes the relevance of the risk information in an annual report. Therefore, the ratio of risk information could be deemed as the most relevant measure for quantity in this type of study. However, it could also be argued that the absolute amount of risk information (RiskWords) is important to take into consideration, because it is the best measure for absolute quantity of risk information. In this research, both RiskWords and RiskRatio are used.
4.2.2 Risk quality
The procedure for measuring RiskRatio comprises three steps. First, to perform a word count (WordCount), the PDF is converted to a MS Word document using the Adobe DC add-on. This ensures that the word count can be conducted without the annual report losing its layout. Second, the total number of words in the narrative part of the annual report is counted,
excluding the financial statements and the notes. Finally, the quantity of risk disclosure is calculated by dividing the number of words in the risk section by the total number of words in the narrative part, including the risk information, of the annual report. This yields the RiskRatio, which is employed in the main analysis of this research to determine the quantity of risk disclosure.
23
The second measurement used in this study is the readability of the risk section. There are two methods that are used to quantify the readability of textual content. Both methods use online tools that have algorithms that give an indication of readability. The first method is the Flesch Reading Ease Formula (FRE). This formula is designed for English texts and aims at quantifying the texts in difficulty levels for adults. The formula assigns a score to the readability of texts from ‘very easy’ to ‘very difficult’ (Li, 2008). A higher score indicates that the material is easier to read (see Figure 1).
The second method is the Gunning Fog Index (Gunning, 1952). This measuring method has often been used to measure the readability of financial information disclosure (Li 2008). This formula is used to determine the complexity of a text. When the results from this formula are higher, it means that it is more difficult (Figure 2). There has been some criticism of this method. Ploughman and McDonald (2014) argue that some of the words that are bound to be included in every annual report will be classified as difficult. They argue that the users of annual reports have these words in their vocabulary. General financial accounting theory also dictates that the usefulness of the annual report is not predicated on everyone understanding the report, but the informed reader should be able to understand it. This being said, it is argued by researchers that the Gunning Fog Index is not a good measure for the readability of annual reports on its own, but can be useful when combined with the Flesch reading index. Despite the criticism, the Gunning Fog Index is often used in recent research into the readability of annual reports (e.g., Li, 2008; Biddle, Hilary, & Verdi, 2009; Miller ,2010; Lehavy, Li, and Merkley,2011; Dougal, Engelberg, Garcia and Parsons 2012; Lawrence, 2013). Because the Gunning Fog Index is mostly used for the readability of annual reports, this method is used to measure the readability of the risk paragraphs. Because of the questions surrounding this method, both readability scores are used in order to establish commonalities in the effect. This would lead to hypotheses being rejected if the outcome of both the readability scores conflict with each other. The koRpus package, a program designed to analyse texts, is used to measure both scores.
24 Fog
Index
Examples
6 TV Guides, The Bible, Mark Twain 8 Readers Digest
8-10 Most popular novels 10 Time, Newsweek 11 Wall Street Journal 14 The Times, The Guardian 15-20 Academic papers
Over 20 Only government sites can get away with this, because you can’t ignore them.
Over 30 The government is covering something up
Figure 2 Gunning Fox Index
Periods that don’t function as the end of a sentence were replaced with a comma, because the program recognizes each period as an end of sentence, which influences the readability score. All texts are saved as ‘plain text’ (.txt) files before using the readability tool.
90-100 Very easy 80-89 Easy 70-79 Fairly easy 60-69 Standard 50-59 Fairly difficult 30-49 Difficult 0-29 Very confusing
25
4.3 Independent variables
The independent variables were chosen based on the research presented above. The independent variables in this study are as follows.
Family business or non-family business (FB):
Since this variable has a yes or a no outcome, the variable is recorded as 0 = non-FB, 1 = FB Management/board (DM)
The involvement in daily management by the family is also recorded as a binary variable, since there is no definitive evidence that there is a positive or negative difference between executive and non-executive involvement. A family-heavy board or a board in which the CEO is a family member is recorded as a 1. All other boards are recorded as a 0.
Leverage (Leverage)
This study controls for leverage, which is computed as the ratio of total debt to total assets (Barako et al., 2006; Bronson et al., 2006).
4.4 Control variables
Big 4 vs. non-Big 4 (AQ)
The choice of auditor can explain the variations in the extent to which organizations disclose information. Oliveira et al. (2011a) argued that companies with high agency costs employ large auditing firms, as they induce management to disclose more information so as to avoid reputation damage. Based on this notion, some authors found that non-finance companies that report more on risks are audited by large audit firms. Wang et al. (2008), consistent with Xiao et al. (2004), also detected an increase in disclosure when large and independent auditors are employed. They hereby hypothesize that in addition to the agency theory, the signaling theory can also help explain why companies employ a large audit firm. Based on the signaling theory, it can be argued that since large accounting firms demand transparency (Oliveira et al., 2011a), companies appoint large auditors to deliver a strong signal of their acceptance of high quality disclosure. The Big 4 audit firms (PWC, Deloitte, EY, and
KPMG) proxy for auditor type. Consistent with other studies (Oliveira et al., 2011a; Wang et al., 2008; Xiao et al., 2004), a value of 1 is assigned when the external auditor is one of the Big 4 audit companies and 0 otherwise.
26 Firm Size (Logsize)
Past research indicated that the extent to which risk information is reported is influenced by firm size. It is suggested that large organizations are more likely to disclose risk information (Bronson et al., 2006; Linsley and Shrives, 2006; Elzahar and Hussainey, 2012). There are various opinions on why this is the case. Some argue that larger organizations rely more on external financing and are more closely scrutinized by stakeholders than smaller
organizations. As such, larger organizations disclose more risk information in an attempt to enhance the confidence of lenders and stakeholder in their ability to manage any potential threats the organization may face (Bronson et al., 2006; Elzahar and Hussainey, 2012).
Others argue that the process of generating and communicating risk information is costly, and thus only large companies can implement and maintain such a process (Barako et al., 2006; Deumes & Knechel, 2008; Leng & Ding, 2011). This study controls for size, measured as the natural logarithm of total assets reported in the company’s balance sheet.
4.5 Data analysis
An MS Excel sheet containing all the variables was created. After looking up the company, a few steps were completed. The company was searched for on Google and official websites were quickly scanned. This was done to find the following information: family involvement statements, ownership statements, public disclosure, board of directors, and potential risk management statements. If the annual report was found on a public corporate website, this annual report was used. If the annual report was not found, the company.info database was used to retrieve the report. The company.info database was also used to find information on ownership, corporate structure, and direct and indirect controlling shareholders. After this information was retrieved, it was determined whether a company could be deemed a family business. Using both the dataset containing family businesses and publicly and semi-publicly available data helped triangulate the findings, thus controlling for the researcher’s bias. After completing the dataset, the Excel sheet was transferred to SPSS 23. SPSS is a widely used program for statistical research
In SPSS, the data was cleaned and missing values were added. After this, descriptive
statistics were extracted and presented in the result section. A test of correlations is presented and used to determine if there is multicollinearity between variables. After these steps, the hypotheses were tested. To address the hypotheses that deal with a question of difference (i.e., is the mean of a variable different for the control group and the reference group), an
27
independent t-test was used. This shows whether there is difference between the two groupings and whether this difference is deemed to be significant. The independent t-test is not a very sophisticated statistical test, but it is deemed best by the researcher for the purpose of this research in combination with the linear regressions performed. The linear regression is the last step in the statistical analysis. To use the linear regression to the fullest, variables were added to the regression independently and the outcomes were reported separately in tables. This was done to avoid the model being influenced by correlations due to the small sample size.
28 5. Results
5.1 Descriptive statistics
The final sample contains 109 companies. A total of 51 qualified as family businesses based on the criteria given in Section 2.2. A total of eighty-nine businesses provided some type of risk disclosure in their annual report. As shown in Table 1, the companies differed in total size based on assets and revenue. This difference was later reduced by taking the natural logarithm as described in the methodology. In Table 1, the descriptive statistics are given for the following variables: RiskWords, RiskRatio, Logsize, Leverage, Fog and FRE.
Furthermore, the table provides some statistics on the total assets, debt and the length of the annual report (WordCount), as these variables present some interesting information about the final sample.
Table 1 Descriptive statistics
Min Max Mean median
RiskWords 0,00 8839,00 2205,11 1793,00 WordCount 0,00 142800,00 47551,65 46108,00 RiskRatio 0,42% 26,59% 4,84% 4,46% Total debt (x€1.000.000) 3,80 62818,00 1879,94 384,60 Total assets (x€1.000.000) 9,70 64797,00 2902,85 747,65 Leverage 0,04 0,97 0,56 ,562 Logsize 16,09 24,89 20,41 20,432 Fog 11 25 18,47 18,3 FRE 1 37 21,00 20,3
The nature of this research dictates that, while the descriptive statistics provided above are interesting for further research and give an insight into the sample size, they lack true insight
29
into the final sample size without providing splits showing the differences in groups in the sample size . To show the splits of the data grouped by family/non-family businesses and the daily management, tables 3 and 4 provides independent paired t-tests.
Table 1 shows that RiskWords varies from 0 to 8839 with an average of 2205 risk words and a median of 1793. The average RiskRatio is 4,84% which means that businesses in the sample used on average 4,84% of their annual report for risk related disclosure. Other results show that the readability of the risk portions is limited. The score on the Fog index suggests that the average readability is on the level of an academic paper (Fog (18,4) = academic paper). The FRE shows a score that is listed as a very confusing text. To examine how different the variables are when divided into family and non-family groups, an independent t-test is used. This t-test compares the means of two different groups in the total sample size. The t-test provides insight into the difference in averages between two groups. The t-test results are presented in section 5.3.
5.2 Correlation
Table 2 displays the correlations between the main variables in this research. The first model used for determining the correlations is the Pearson model. The Pearson correlations are shown in the table above the diagonal. Because this research contains several binary (dummy) variables that are not on a normal scale, the Spearman model is also used for measuring correlations. In theory, the Spearman model is more reliable in measuring correlations of variables that are not divided on a normal scale (Field, 2009). The Spearman correlations are shown below the diagonal in Table 3.
The correlation table is used to find interesting results that would go unnoticed in a regression analysis. Furthermore, the correlation table is used to verify the Multicollinearity shown by the variables in this research. A strong relation between two variables could influence the outcomes of the linear regressions used to test the hypotheses. This could lead to false conclusions being drawn based upon the outcome of the variables. Multicollinearity is assumed when the correlation shown in the table surpasses 0,7. This is only true for
RiskWords and WordCount, which is to be expected since the amount of total disclosure is sure to effect the amount of disclosure on risk. RiskRatio is calculated based on: RiskRatio = RiskWords / WordCount. The correlation between WordCount and RiskWords implies that the risk ratio does not change much when either the RiskWords or WordCount changes.
30
Table 2 Correlation matrix
1 1 2 3 4 5 6 7 8 9 10 1. FB 1 ,515** ,462** -,254** -0,021 -,392** -,498** 0,091 -,245* -,292** 2. MT 0,650** 1 ,358** -,271** 0,018 -,420** -,516** 0,159 -0,128 -,361** 3. BH 0,385** ,237* 1 -,220* 0,019 -,434** -,601** ,202* -,427** -0,18 4. Risk -,318** -,271** 0,185801 1 0,025 0,041 0,028 0,023 0,004 ,228* 5. RiskRatio -0,084 -,067 -,090 ,029 1 ,392** -0,078 ,657** 0,057 -0,147 6. RiskWords -,532** -,468** -,463** ,046 ,600** 1 ,719** -0,061 ,461** ,287** 7. WordCount -,620** -,546** -,562** ,038 0,061 ,788** 1 -0,139 ,623** ,385** 8. Leverage 0,125 0,159 0,106 -0,054 0,026 0,003 0,0267 1 0,019 -,244* 9. Logsize -,223* -,148 -,483** ,016 0,092 0,003 ,611** 0,069 1 ,340** 10. AQ -,309** -,361** -,137 ,228* 0,009302 ,332** ,418** 0,075198 ,331** 1
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
31
5. 3 Independent t-test:
To show how different the variables are when divided into family and non-family groups, an independent t-test is used. This test compares the means of two different groups in the total sample size. The independent t-test shows whether the difference in the mean is statistically significant.
Table 3 highlights the difference in the mean of four of the dependant variables and WordCount. WordCount is shown because the correlation matrix suggests a strong
correlation between RiskWords and WordCount. The analysis of the independent t-test shows that RiskWords is significant for t (88) = -4,8979, p = ,000. This indicates that family firms (M = 1407,292) have fewer RiskWords in their disclosures than their non-family counterparts (M = 3075,455). This aligns with the finding of the independent t-test that WordCount is significant for t (89) = -6,8344, p = ,000. This statistically significant result indicates that family firms (M = 30934,787) disclose fewer words than their non-family counterparts (M = 65301,477). What could be stated from these outcomes is that the amount of disclosure and the risk disclosure do indeed seem to correlate with the family business variable.
The independent t-test shows no significant results for both the readability scores and the risk ratio. This implies that while the mean can differ between both groups, the difference is not statistically significant. In other words, the only conclusion that could be drawn on the basis
FB N Mean T df Sig. (2-tailed)
RiskRatio 0 48 4,783 -,073 88 ,942 1 43 4,904 RiskWords 0 48 2937,021 -4,879 89 ,000 1 43 1406,455 WordCount 0 48 61836,68, -6,834 89 ,000 1 43 31605,558 Fog 0 48 18,57 ,975 89 ,332 1 43 18,27 FRE 0 48 21,64 -,683 89 497 1 43 20,39
32
of these outcomes is that there is no difference in the RiskRatio, Fog and FRE scores between family and non-family firms.
Table 4independent t-test for family business
DM N Mean T df Sig. (2-tailed)
RiskRatio 0 67 4,91 0,469 87 0,64 1 21 4,51 RiskWords 0 67 2458,691 2,999 87 0,004 1 21 1198,727 WordCount 0 67 53570,209 4,538 87 0 1 21 25393,273 Fog 0 67 18,88 3,105 87 0,002 1 21 17,73 FRE 0 67 20,7450 -,889 87 0,376 1 21 22,3552
The analysis of the independent t-test shows that the RiskWords variable is significant for t (87) = 2,99, p =,004. This indicates that firms with a family CEO (M= 1198,727) do disclose less RiskWords than their counterparts (M=25393,273). The independent t-test shows that the WordCount variable is significant for t (87) = 4,538, p =,000. Indicating that firms with a family CEO (M= 30934,787) disclose statistically significant less words than their non-family counterparts (M=53570,209). The major difference between table 3 and 4 is the Fog which shows a significant result with t (87) = 3,105, p= 0,002 which shows that the mean of the family CEO group (M=17,73) lower than the group without a family CEO (M= 18,88). The independent t-test shows no significant results for the FRE and the risk ratio. This implies that while the mean can differ between both groups that the difference in the mean is not statistically significant. In other words the only conclusion that could be drawn on the results presented in table 4 is: there is no difference in the RiskRatio and FRE scores between firms with family CEO’s and without.
33
5.4 Regressions
5.4.1 Dependent variable RiskRatio
RiskRatio = β0+ β1*Logsize + β2*AQ + β3*FB + β4*DM+ β5*Leverage + εi
Table 5 RiskRatio regression
0 1 2 3 4 5 6 constant 2,664 (0,515) 6,152 (0,00)** 0,804 (0,805) 1,217 (0,782) 1,0 (0,817) 0,902 (0,803) 0,014 (0,761) Logsize 0,106 (0,594) ,291 (0,198) 0,284 (0,69) 0,69 (0,198) 0,308 (0,186) 0,003 (0,768) AQ -1,498 (0,170) -2,919 (0,73) -2,293 (0,068) -2,306 (0,072) -2,254 (0,07) -0,024 (0,070) FB -,300 (,690) -0,002 (0,768) DM -,251 (0,741) 0,001 (0,906) Leverage -,007 (0,740) -0,006 (0,779) Adjusted R2 0,057 0,01 ,018 0,009 0,008 0,007 -0,003 R2 0,003 0,022 0,201 ,042 0,042 0,04 0,206 F-statistic 0,287 1,916 1,808 1,252 1,23 1,196 0,758
34
Table 5 shows the outcomes of the linear regression performed on the dependent variable RiskRatio. The constants vary between 0,9 – 6,152, which shows the average expected value for y = RiskRatio (%) with x = 0. Since the firm size variable is never equal to 0, the constant has no meaning in model 0. In the models containing dichotomous variables, the constant has more meaning. These dummy variables equal either 0 or 1. In these models, the constant represents the mean of the reference group.
Model 0, 1 and 2, control variables
In model 0, the control variable Logsize is tested. Model 0 shoes no significant result (b = 0,106, p >0,05), which means that there is no significant effect of the firm size on RiskRatio. The F-statistic of 0,287 confirms that the model is not significant. The adjusted R2 of 0,057 suggests that 5,2% of the variance in the dependent variable RiskRatio is explained by model 0. With the model not being significant on at least a 5% level, there is a reasonable chance that the difference in variance is explained by chance and not by the independent variable. Model 1 is the same, but now with only the AQ control variable in the model. This model is also not significant and the adjusted R2 shows that just 1% of the model is explained by this variable An interesting thing in this model is that the average of the reference group or the constant is 6,152, with the x (AQ) = 0. Model 2 has both control variables in the same model. The explanatory value of the model is higher, with an adjusted R2 of 0,018.
Model 3, H2, H2a, H2b
The model controls for the firm size and the AQ variable in this test, while the model shows no evidence for a significant effect of the control variables. The linear regression shows that the family business variable is not significant (p > 0,1) at the 10% level and that there is a negative effect (b = -,300) on RiskRatio. This shows that a significant conclusion cannot be drawn from this test but it does give an indication on the nature of the relationship. The entire model is not significant, with an F-statistic of 1,205, and shows a predictive value of 0,009. Model 4, H1
Model 4 has the same variables as model 3, however, this model uses the DM variable as the independent variable. The regression analysis shows that that the DM variable is not
significant (p = ,741) and that it has a negative effect (b = ,251) on RiskRatio. This further confirms the analysis shown with the independent t-test that the involvement of a family CEO
35
leads to less relative risk disclosure, while also confirming the negative effect of this variable. The F-statistic of 1,23 further confirms that the model is not statistically significant.
Model 5, H4
Model 5 includes the leverage in the linear regression to test the effect of leverage on the RiskRatio. The linear regression shows that the amount of leverage has a small negative effect on the amount of RiskRatio (b = -,007), however, this model is far from significant ( p = 0,704). Furthermore, the predictive value of this model is greatly diminished to an adjusted R2 of 0,007 with a not significant F-statistic of 1,196.
Model 6
Model 6 includes all the variables. The model is not significant and lacks a true predictive value.
36 5.4.2 Dependent variable RiskWords
RiskWords = β0+ β1*Logsize + β2*AQ + β3*FB + β4*DM+ β5*Leverage + εi
Table 6 RiskWords regression
0 1 2 3 4 55 constant -7476 (0,000) ** -6221,86 (0,003) ** -6461,3736 (0,007) ** -5746,326 (0,003) ** -7463,22 (0,000)** -4594,34 (0,021) ** Logsize 473,67 (0,000) ** 340,349 (0,001) 404,816 (0,001)** 419,748 (0,000)** 474,69 (0,000)** 343,966 (0,001)** AQ -676,36 (0,240) 80,034 (0,877) FB -896,557 (0,001)** -466,451 (0,389) DM -1313,65 (0,00) ** -1050,69 (0,007) ** Leverage -,505 (0,959) 8,092 (0,389) Adjusted R2 0,212 0,179 0,297 0,582 0,0008 0,298 R2 0,461 0,197 0,545 ,339 0,042 0,337 F 24,228*** 10,782** 18,834** 22,8** 1,23 8,629**
Table 6 shows the regression analysis on the dependent variable RiskWords. The constant in this model varies between -7476 and -5746,36. The value of this statistic is relatively low in
