Master Thesis
Empirical Analysis of Enterprise Risk Management
and Firm Performance:
The moderating effect of firm variables
Author: ing. Jelmer H. Bulthuis Student number: S1808559 Email: j.bulthuis@student.rug.nl
Supervisor: dr. A.A.J. van Hoorn Co-assessor: mr. drs. H.A. Ritsema
Groningen, May, 2012
University of Groningen
Faculty of Economics and Business
Department of International Business and Management Nettelbosje 2
ACKNOWLEDGEMENTS
This Master Thesis is the final part of my Master in International Business and Management at the University of Groningen. It has been a long way to reach this point, but it was a great and interesting experience from a scientific perspective to go in-depth into the topic of Enterprise Risk Management. The writing of this Master Thesis would not have been possible without the help of a number of people. Therefore, I would like to take this opportunity to sincerely thank them.
First of all, I am very grateful to my supervisor, dr. A.A.J. van Hoorn, for all the valuable meetings, for his patience, his advice and suggestions, and for his knowledge sharing and support throughout the whole process of writing my thesis.
I also wish to thank the persons; F. Salverda, drs. J. Coes and ir. T. de Jong, for the time they spent providing helpful comments and given valuable suggestions on my thesis.
Finally, I wish to thank my family and friends for their support during this period. In particular, I am grateful to my parents Inge Coes and Nanning Bulthuis, not only for there support, faith, and encouragement during the process of writing this thesis, but also during my period as a student in general.
ABSTRACT
Enterprise Risk Management appeared in the mid-1990s as a new and holistic way of viewing and managing risks across firms. One of the main drivers of the increasing interest in Enterprise Risk Management lays in its ability to create value. The generalizability of previous empirical research, in terms of time span and differentiation of markets, is limited. Therefore, this thesis examines the moderating effect of four firm specific variables on the adoption of Enterprise Risk Management on the firm performance. The four firm specific variables are: (1) firm size, (2) firm complexity, (3) environmental uncertainty, and (4) legal origin.
This research is based on a sample of 103 publicly traded firms and 1,509 firm years, originated from the European Union or the United States of America with a time span from 1990 until 2010. The first appointment of a Chief Risk Officer is used as proxy for Enterprise Risk Management adoption. With a multicollinearity regression model the moderating effects of firm variables are tested.
In general, for all four firm variables significant evidence is found for their moderating effect on Enterprise Risk Management and firm performance. In terms of legal origin and firm complexity strong evidence is found for its moderating effect; the evidence for the moderating effect of firm size and environmental uncertainty however is limited.
TABLE OF CONTENT
LIST OF TABLES ... 5
LIST OF FIGURES ... 6
ABBREVIATIONS ... 7
1 INTRODUCTION ... 8
2 BACKGROUND & HYPOTHESES ... 12
2.1 TRADITIONAL RISK MANAGEMENT ... 12
2.2 THEORY OF ENTERPRISE RISK MANAGEMENT ... 13
2.3 EMPIRICS OF ENTERPRISE RISK MANAGEMENT & FIRM PERFORMANCE ... 16
2.4 THEORY & HYPOTHESES ... 17
2.4.1 FIRM SIZE ... 18
2.4.2 FIRM COMPLEXITY ... 19
2.4.3 ENVIRONMENTAL UNCERTAINTY ... 19
2.4.4 LEGAL ORIGIN ... 21
2.5 CONCEPTUAL MODEL ... 22
3 DATA & EMPIRICAL MODEL ... 23
3.1 PERIOD & SAMPLE SELECTION ... 23
3.2 DEPENDENT VARIABLES ... 24
3.2.1 TOBIN’S Q ... 24
3.2.2 RETURN ON EQUITY ... 25
3.2.3 MARKET TO BOOK RATIO... 26
3.3 INDEPENDENT VARIABLES ... 26
3.3.1 ERM ADOPTION ... 26
3.3.2 FULL ERM IMPLEMENTATION ... 28
3.5.2 ACCOUNTING YEAR ... 32 3.5.3 INDUSTRY ... 32 3.6 EMPIRICAL MODEL ... 32 3.7 DESCRIPTIVE STATISTICS ... 34 3.8 CORRELATION ... 35 4 RESULTS ... 37 4.1 BASIC RESULTS ... 37 4.1.1 MODERATING GRAPHS ... 40 4.2 ROBUSTNESS-CHECK ... 45
4.2.1 ALTERNATIVE PROXIES FIRM PERFORMANCE ... 45
4.2.2 ALTERNATIVE MEASURES FIRM SIZE & ENVIRONMENTAL UNCERTAINTY ... 46
4.2.3 ALTERNATIVE PROXY ERM IMPLEMENTATION ... 46
4.3 SUMMARY OF FINDINGS ... 47
5 DISCUSSION & CONCLUSION ... 48
5.1 DISCUSSION ... 48
5.2 IMPLICATIONS FOR PRACTICE ... 50
5.3 LIMITATIONS & FUTURE RESEARCH ... 50
LIST OF TABLES
Table 1: Sample statistics for industry and year of first CRO appointment ... 24
Table 2: Descriptive statistics ... 34
Table 3: Pearson correlation coefficient ... 36
Table 4: Regression analysis of Regression Model I ... 38
Table 5: Regression analysis of Regression Model II ... 39
Table 6: Firm list. ... 58
Table 7: Variable definitions. ... 62
Table 8: Regression analysis alternative performance measure of Regression Model I. ... 63
Table 9: Regression analysis alternative performance measure of Regression Model II. ... 65
Table 10: Regression analysis alternative moderating measures of Regression Model I ... 67
Table 11: Regression analysis alternative moderating measures of Regression Model II. ... 70
Table 12: Regression analysis alternative independent proxy of Regression Model I. ... 73
LIST OF FIGURES
ABBREVIATIONS
CEO Chief Executive Officer
COSO Committee of Sponsoring Organization of the Treadway Commission CRO Chief Risk Officer
ERM Enterprise Risk Management
EU European Union
RM Risk Management
TRM Traditional Risk Management
UK United Kingdom
1
INTRODUCTION
The global financial crisis of 2007-2009 has exposed the failure of Risk Management (RM) and is seen as one of the main causes of this event (The Financial Crisis Inquiry Commission, 2011; Miccolis and Goodman, 2012). In 2007, risk-models start to fail due to “known unknowns”, which have affected financial firms. The interconnections among these firms1 have triggered events that are described as “unknown unknowns” (Jorion, 2009).2 This caused a global financial meltdown, where financial firms, and also firms from other industries, have suffered significant losses. The magnitude of this event highlights the importance of proper risk management within firms for protecting and enhancing shareholders value. Therefore, it is crucial to study scientifically the evolution of RM, and the value creating ability that lies within proper RM.
During the beginning of the 1990s, a variety of events has taken place in the global business environment, such as corporate scandals and financial collapses. These events have driven governments, mainly in western countries, to set up, or adjust more strictly, their corporate governance codes (e.g. Sarbanes-Oxley in the USA, Code-Tabaksblat in the Netherlands, Deutscher Corporate Governance Kodex in Germany, Turnbull Guidelines in the UK, and Codice di Autodisciplina in Italy). The purpose of corporate governance codes is that it should act as a protection mechanism for the firm’s shareholders and certain stakeholders.3 A combination of these events and the both newly or revised corporate governance codes have increased the expectations from the stakeholders in terms of effective risk management, and increasing volume and complexity of risks facing the firm (Beasley et al. 2006). This has led to an evolution of Traditional Risk Management (TRM) into a new risk management paradigm.
This new risk management paradigm is called Enterprise Risk Management (ERM)4 and emerged in the mid 1990’s. ERM is a more advanced and proactive way of managing risks, coordinated from an enterprise wide perspective. It identifies, quantifies, responds to and monitors the consequences of potential events and the interdependency between these
1 Also explained as Network Theory; which suggests that when an important part of the network fails, a chain reaction of failure can be triggered.
2 More familiar as “Black Swans” (Nassim Nicholas Taleb, 2007) Defined as events that are hard to predict, have a high consequence, and are retro perspective.
3
Depended on Anglo-American "model” or the coordinated or multi-stakeholder “model”.
events. The interest and adoption of this new paradigm continued growing after its emergence (Beasley, et al. 2008), and is currently still growing. For example, ERM has significantly increased, different consultancies have specialized themselves in the implementation of ERM and the amount of academic papers written about topics related to ERM also increased. Even some rating agencies have added ERM as part of their rating process (e.g. S&P Ratings Direct, 2008; Moody’s, 2004). Extended to this increasing interest of ERM adoption is the introduction of Basel II, Basel III and Solvency II.5 These new regulation standards give more restrictions to financial firms, which will increase the complexity of decision-making. Furthermore, the need of strategic and efficient allocating and managing of capital resources will increase by these stricter regulation standards. Modern management techniques, such as ERM, can be implemented to keep control on this increasing complexity. Also, ERM helps keeping the firm’s risk exposure to stay within the firm’s appetite for risk.
The ERM paradigm differs from the TRM perspective, because risks are managed individually in the TRM perspective. ERM is “the process of analysing the portfolio of risks facing the enterprise to ensure that the combined effects of such risks are within an acceptable tolerance” (Beasley, et al. 2008), and either protect or enhance the shareholders’ value or both. The argument that the implementation of ERM will increase the firm performance has become general accepted amongst executives (PWC, 2004; and Deloitte, 2009) and scholars (Stulz, 1996; Stulz, 2003; Lam, 2003; Barton, 2002; Nocco & Stulz, 2006; Hoyt & Liebenberg, 2011). This argument is mainly supported by the theory from Stulz (1996, p8), which proposes that the main focus of ERM is “the elimination of costly lower-tail outcomes”. Lower-tail outcomes are risks that threaten the firm earnings, and therefore can lead into negative consequences. The potential value creating ability of ERM lies in its ability to reduce or eliminate costly lower-tail outcomes (Stulz, 2003). Firms that are facing a higher degree of these lower-tail events, have potentially more benefit from the implementation of ERM. Yet, firms who are not facing such events will see no benefit (Stulz, 2003).
However, the amount of empirical evidence for this theory is limited, compared to empirical research of the relationship between TRM and firm performance. To my knowledge, Beasley et al. (2008), Gordon et al. (2009), Hoyt & Liebenberg (2011) and Pagach & Warr (2010), are the only ones who made an empirically contribution to this research area. Hoyt & Liebenberg (2011) found a positive relation of Tobin’s Q and the first
CRO appointment. Pagach & Warr (2010) found minimal evidence in the reduction of risk in the firm performance. Beasley et al. (2008) found that the stock market reaction is neither positive nor negative significant to the adoption of ERM. However, for non-financial firms certain firm variables are positively related to market reaction and ERM adoption, but this claim cannot be made for financial firms. Gordon et al. (2009) found that firm variables are contingent on the relationship between ERM and firm performance.
Even though empirically the support is growing for the argument that ERM protects and enhances shareholders value, different signs make it questionable if this really does. For example: Hampton (2009, p.66) argued that especially ERM could be blamed for the financial crisis. During this event a more advanced risk management program should have been able to better withstand such an event. However, this was not the case, because a gross of the financial firms that implemented ERM, were the ones that suffered most from this event. Furthermore, different regulatory regulations are drivers to implement ERM, such as: Sarbanes-Oxley, Basel III, Solvency II and Rating Agencies. If one of them is the main driver for implementation of ERM, than it is disputable that ERM adoption is in the best interest of shareholders. Based on the findings from Beasley et al. (2008) and Gordon et al. (2009), I assume that the value creating ability of ERM is more refined, and lies within the differences of firm variables.
Therefore, the primary goal of this thesis is to increase empirically the generalizability of the argument that ERM is related to firm’s value, but it is dependent on certain firm variables. For being able to find the dependency, the moderating effect is examined of firm related variables on ERM and firm’s value.6 The main research question is phrased as follow: “which firm variables have a moderating effect on ERM related to the firm’s value?” The firm variables that are included in this research are: (1) firm size, (2) firm complexity, (3) environmental uncertainty, and (4) legal origin.
This research is related to a combination of the following two papers: Beasley et al. (2008) and Gordon et al. (2009), in terms the contingency of firm variables on the relationship between ERM and firm performances. However, this research is still original, as none of the previous papers have examined the moderating effects of firm variables. Furthermore, firms from the European Union (EU) and the United States of America (USA) are included in the
sample. Extended, the time span is wider and includes an extreme uncertain event, namely the financial crisis in 2008.
The results in this thesis are based on 103 publicly traded firms from the EU and the USA. Publicly traded firms are more likely to announce ERM activities publicly, because corporate governance codes demand more transparency for public traded firms, compared to those who are not publicly traded. To limit the difference in industries, this research focuses on financial firms that are operating in the Major Banking-, Regional Banking-, Insurance-, and Investment-industries. To make the ERM adoption measurable, the firm’s Annual Reports and Lexis-Nexis are used as primary sources. For each firm, the earliest announcement or appointment of a “chief risk officer,” “director of risk management,” “CRO,” “head of global risk management,” and “head of Group Risk” is collected and numbered. The announcement or appointment of a Chief Risk Officer (CRO) can be seen as one of the first stages of ERM adoption. Beasley et al. (2005) found significant empirical evidence that the ERM implementation stage is positively related to the presence of a CRO. Additionally, for every single firm additional research is conducted to reduce the possible bias in the sample of the first CRO appointment.
As a preview of the results statistical significance is found for the moderating effect of all four firm variables on the relationship of ERM adoption on firm performance. Small banks that are highly complex and less shareholder-oriented do benefit most from adopting ERM. In case for financial firms in general, that are larger in size, active in a highly uncertain environment, and less shareholder-oriented will benefit the most from adopting ERM.
2
BACKGROUND & HYPOTHESES
This chapter discusses the theoretical background of this thesis, which is used for the hypotheses development. In the theoretical background, a historical overview is given of the evolution of risk management. Furthermore, an extended argument for the value creating capacity of ERM and an overview of related empirical research is summed-up. In the second part of this chapter the hypotheses are developed and the research model is presented.
2.1
Traditional risk management
The core elements of risk management find their origin back in the Renaissance.7 During this period the mathematics of probability were discovered. The book Against the Gods: The Remarkable Story of Risk, (Bernstein, 1996b) describes the evolution of risk management from the Renaissance until the end of the 20th century. It highlights how risk management has increased in dominancy within the business world by the changing business environment, and how it continues to change.
In the late 1940s and the early 1950s, the scope of risk management was limited to loss exposures. Insurances were used as the main financial instrument and later also derivatives8 to reduce and control uncertainties. This is also known as Traditional Risk Management (TRM). TRM is based on a “silo” approach to manage risks in firms. In this “silo” approach, risk classes are managed separately within different departments, where every department is specialized in a certain risk group. Within this approach, risks are controlled and diminished through insurances and derivatives. These two financial instruments are often used because of the need within firms to reduce uncertainties.
Insurance is the most frequently used tool in risk management. It allows transferring risks of unforeseen losses to a secondary party, who will bear a risk for a certain period, in return for money.9 In this way, firms are able to withstand large-scale economic losses due to natural disaster, have access to new financial liquidity after a loss and are less sensitive for withstanding financial shocks. There is a wide variety of insurances for reducing risk
7 In this period, from 1400-1700, scientist gained knowledge over the probability in games of change, which is the basic key principle for risk management.
exposure. For each firm the insurance profile is diverse, and should be adjusted to its appetite for risk.
In the 1970s, firms started to manage various financial risks with the use of financial derivative products. On their own, derivatives have no value, but the value changes through an increase or decrease of volatility in the price of the underlying asset. Volatility can affect the firm’s value, because it is facing interest rates-, commodity prices-, exchange rate-, and stock prices-exposure. There are two kinds of derivatives: first, there are futures, which are contracts to purchase a specific item at specified price at a specific time in the future, and second, there are options, which give the right to buy or sell to the other side at a specific time at prefixed price. These financial instruments can reduce the expected bankruptcy costs, tax burden, and costs of regulatory scrutiny. Furthermore, it will stimulate the attractive investment opportunities (Colquit & Hoyt, 1997).
The traditional risk management activities are well presented in the academic literature, in both theoretical and empirical sense (e.g. Smith & Stulz, 1985; MacMinn & Han, 1990; Allayannis & Weston, 2001; Nelson et al., 2005; Smithson & Simkins, 2005; Carter et al., 2006; and Nguyen & Faff, 2010). A main disadvantage of this “silo” approach of risk management when companies increase in size, is its ineffectiveness, due to the lack of coordination between the different departments of managing risk. During the mid-1990s, firms became aware that insurable risks and financial risks should be managed together (Dickson, 2001), because they are essentially used for the same goal. This goal is to reduce catastrophic losses, and so reducing the earnings volatility. The increase of this awareness stimulates that contingency planning should be managed in a more general way. Therefore, a more comprehensive way of internal system is needed to base a corporate strategy on. Finally, firms began to recognize the importance of intangible assets (Sanjay, 2010). TRM methodologies are not able to protect these intangible assets.
2.2
Theory of enterprise risk management
holistic point of view in the management of risks. This in contrary to TRM, where each risk concept is managed in an individual way, and so the interdependency of different risk concepts stayed unnoticeable. Furthermore, the ERM concept suggests an overall risk information system throughout the whole firm. This gives the ability for a more refined risk portfolio, on which management can base its risk management strategy and its corporate strategy. A frequently referred definition of ERM (e.g. Fraser, et al. 2008; Arena et al. 2010; Pagach and Warr, 2010, 2011), is from the COSO (2004), who formulated the definition of ERM as follows:
“Enterprise risk management is a process, effected by an entity’s board of directors, management and other personnel, applied in strategy setting and across manage risk to be within risk appetite, to provide reasonable assurance regarding the achievement of entity objectives (COSO, 2004 p.2).”
The committee suggests that the ERM paradigm is geared to achieve the following four firm objectives: (1) strategic: focusing high-level goals, aligned with and supporting its mission; (2) operations: effective and efficient use of its resources; (3) reporting: reliability of reporting, and (4) compliance: compliance with applicable laws and regulations (COSO, 2004). These objectives are used to construct a three-dimensional matrix, based on the narrower framework of COSO’s (1992), where the relationships between objectives, components of ERM, and an entity’s units are visualized into a ERM paradigm (see Figure 1).
One of the basic assumptions of ERM is the ability to create value for shareholders and stakeholders. Casualty Actuarial Society Committee defines clearly the same statement in their definition of Enterprise Risk Management:
“ERM is the discipline by which an organization in any industry assesses, controls, exploits, finances, and monitors risks from all sources for the purpose of increasing the organization’s short- and long-term value to its stakeholders (Casualty Actuarial Society Committee on Enterprise Risk Management, 2003, p.8).”
However, further development of risk management in firms can lead to conflict of interest between shareholders and company managers. Modern portfolio theory suggests that firm specific risks should not be a concern of the firm’s management. Shareholders are capable to adjust their portfolio and diversify all firm risks away (Markowitz, 1952), this in a relatively costless manner. For the implementation and the management of the ERM system, firm resources are used. Shareholders can see this as value destroying, because agency costs will increase by reducing individual risks.
Although, the above stated view relies on the idea that there are no frictions and imperfections within an efficient capital market, in the real world this is not applicable. Frictions and imperfections are present at the capital market, and within this lies ERM value creating capability. Based on this suggestion, Stulz (1996, 2003) and Nocco and Stulz (2006) assume that risk management can add firm value in combination with an increase of agency costs. The value creating ability of ERM lies in the reduction of costly lower-tail outcomes (Stulz, 1996, 2003). Lower-tail outcomes are risks that threaten the earnings, which can result in negative consequences for the firm. Therefore, the engagement to ERM could lead to value creation, if firms are enough exposed to lower-tail outcomes. This is supported by the argument from Beasley et al. (2008), who are stating that firms should benefit from adopting ERM when earnings are volatile, because expected result should smoother earnings and reduce the quantity of lower tail outcomes.
improve the transparency of the interconnection between risks, which in TRM would be undetected. This will also lead to less agency costs for reducing uncertainties, and can therefore be value creating. Finally, a new source of value is created of firms risk profile. This new firm risk profile can be seen as a source of risk information that continually receives new inputs. This information will help risk management to work out a more funded corporate strategy that fits into the firm’s appetite for risk. Furthermore, it gives the management the ability to respond quicker to changes in the business environment. What finally can transform into an advantage for the firm, if they are operating in a quickly changing environment, or if they base their corporate strategy on this kind of information.
According to COSO (2004), ERM systems are likely to vary among firms to increase its effectiveness. This argument is in line with the basic idea of contingency theory, which suggests that the effectiveness of certain management methods for organizations is contingent on internal and external environmental factors. This is also applicable in terms of risk management. Beasley et al. (2005) made the same consistent assumption that management and control is contingent. Therefore, it can be assumed that the effect of ERM on firm performance is dependent on certain variables.
2.3
Empirics of enterprise risk management & firm performance
In recent studies, a significant increase is found in the adoption of ERM (Beasley, et al. 2008), full implementations of ERM (Marsh & RIMS, 2009), and broadening of ERM’s scope (Deloitte, 2009). The Chief Executive Officer (CEO) is one of the main driving forces of ERM adoption, which is guided by two key factor regulations and the volatile economic situation (Acharyya & Jonson 2006). When the CEO is the main driver for the adoption of ERM, it will assume that the adoption will be beneficial for the firm’s value, based on the shareholders theory, which argue that CEO’s are acting in the best interests of their shareholders. Evidence for this assumption comes from PWC (2004), where 44 per cent of the CEO’s strongly agree that ERM provides the tools for managing risks and align them with the firm’s appetite for risk to create value, and Deloitte (2009), 72 per cent of the executives reported that the financial benefits exceed ERM implementation costs.
sample consisting of 117 USA insurers with data from 1998-2005 and they found a positive relation between first CRO appointment and Tonbin’s Q. Pagach & Warr (2010), however failed to find general support that the adoption of ERM increases firm performance. They only found minimal evidence in the reduction of risk in the firm performance with a sample of 106 firms over a period of five years, two years before and two years after the CRO announcement. Beasley et al. (2008), found that the stock market reaction is neither positive nor negative significant to the adoption of ERM. This research was conducted in a sample of 120 US firms in the financial, insurance, energy and miscellaneous industry, and with a time span from 1992 until 2003. However, for non-financial firms certain firm variables are positively related to market reaction and ERM adoption, but this claim cannot be made for financial firms. Gordon et al. (2009), found a positive relation between the adoption of ERM and firm performance, however, it is contingent on firm variables. These findings come from data from 112 firms, active in 22 different industries, in the year 2005.
This thesis is based on the findings of Beasley et al. (2008) and Gordon et al. (2009), who argue that firm variables can have influence on the relationship between the adoption of ERM and firm performance. However, they have not focussed on the single effect of these firm variables. Therefore, in this research the focus has been put on the specific effect of single firm variables, what is named in this research as the moderating effect, on ERM adoption in relation to firm performance. Furthermore, by using data, which not only include USA firms, like most previous studies, but also EU firms, and by broadening the time span of the data, it will help to increase the generalizability of the argument that ERM increases the shareholders’ value.
2.4
Theory & hypotheses
2.4.1 Firm size
In different business research topics (e.g. Lawrence & Lorsch, 1967), firm size is considered to be an important variable when performance is measured. Firms that are on average larger in size are frequently less flexible, due to the likeliness of more employees, greater amount of resources within the firm, and more financial reporting processes. This all makes it harder to quickly adapt to changing risk environments. Therefore, management systems, like ERM, play a key role in firms for being able to react in a proper way to these quickly changing risk environments. In accounting research the importance of firm size, like in measuring performance, is also an important characteristic to take in consideration when designing management systems (e.g. Shields 1995), and also for ERM (COSO, 2004). Larger firms are likely to have more available resources to support the implementation costs of managements systems, like ERM. This is a possible explanation for the findings of Beasley et al. (2005) and Hoyt & Liebenberg (2011), who found statistical evidence that the firm size is statistically significant and positively related to the implementation of ERM, for both financial and non-financial firms (Pagach & Warr 2011). Another possible explanation is that it is more likely that larger firms compared with smaller firms, are more inclined to implement ERM, based on the assumption that they benefit more from ERM adoption, in terms of firm’s value. This is in line with the shareholders theory, which assumes that the decisions from the board, like implementing ERM, are to increase the shareholders wealth. Therefore, larger firms have more potential to create value from implementing ERM, compared with smaller firms. The advantage for larger firms lies within the term economies of scale and the accessibility to the right resource. Furthermore, senior management, in larger firms, experience more noise in terms of information of risk, due to more hierarchical layers. ERM helps to reduce this problem by implementing a risk culture and IT risk management systems. Therefore, a more advanced way of detecting risks and more knowledge gathering about risks can lead to more reduction of lower tail outcomes, what can lead to more value creation. Gordon et al. (2009) found support for this assumption that the relation between ERM and firm performance is contingent on a proper match of firm size and other variables. Moreover, Beasley et al. (2008) found support that larger firms are more likely to benefit from risk management, compared with smaller firms.
therefore, expect that firm size has a moderating effect on the relation between ERM and firm performance.
Hypothesis 1. Firm size moderates the effect of ERM on firm performance.
2.4.2 Firm complexity
Firms, which have a higher degree of complexity, are more likely to benefit from the adoption of ERM. Complexity is mainly caused by a higher degree of diversification in the organization and it causes less integration. Less integration within a firm will assumable be harder to manage and will have more and broader arrays of risks. The adoption of ERM will help the internal risk management to synergize risk information throughout the firm. Furthermore, it can be used as a managerial tool and reduce costly lower tail outcomes. Hoyt & Liebenberg (2011) found a positive relation between firm complexity and adoption of ERM. According to Gordon et al. (2009) firm complexity is another variable that is contingent on a proper match between ERM and firm performance. Furthermore, Standard and Poor’s (2005) stayed that firms, which are relatively more complex, will probably benefit more from the adoption of ERM.
The above-mentioned literature and arguments suggest that it is assumable that there is a moderating effect between firm complexity and the adoption of ERM on firm performance. I, therefore, expect that firm complexity has a moderating effect on the relation between ERM and firm performance.
Hypothesis 2. Firm complexity moderates the effect of ERM on firm performance.
2.4.3 Environmental uncertainty
caused by organizational environment forces. Kreiser & Marino (2002) have conceptualized these environmental uncertainties into two main perspectives: information uncertainty theory, and resource dependence theory. The argument behind information uncertainty perspective is that environmental uncertainty is a shortage of information about the environment of the firm (López-Gamero et al. 2011). The resource dependence theory is that scarce resources come from its environments; firms are dependent on those to survive. Uncertainty arises for the firm when there is a lack of management control over the scarce resources in the specific environment (Kreiser & Marino, 2002).
In both cases, managing control systems can transform a certain amount of uncertainties into risks, which are quantifiable and measurable. ERM is managing a control system that will help mapping all risks that the firm is facing. Furthermore, it gives the ability to “avoid pitfalls and surprises along the way” (COSO 2004, p1), which can also be described as lower tail outcomes. These lower tail outcomes are more likely to occur when a firm is operating in an environment with greater uncertainties. Therefore, it can be assumed that in firms who are operating in greater environmental uncertainty, ERM can reduce lower tail outcomes, which can lead to an increase of firm performance. This is in line with the assumptions of Stulz (2003) that ERM value-creating ability lays within the reduction or elimination of costly lower-tail outcomes. Gordon et al. (2009) found that environmental uncertainty is contingent on a proper match between ERM and firm performance. On one side, I expect that firms with higher environmental uncertainty are in greater need for ERM adoption and will benefit more from its effects. This will suggest that firms that have implemented ERM probably have a higher environmental uncertainty. On the other hand, it is possible to experience more environmental uncertainty reduction, which can lead to an increase of performance. However, before ERM adoption, firms that adapt their risk profile according to their risk appetite, could, decide to operate in a more uncertain environment. This is because ERM already reduces a certain amount of risks, and the amount of risks facing the firm is not in line with the firm’s appetite for risks.
Hypothesis 3. Environmental uncertainty moderates the effect of ERM on firm performance.
2.4.4 Legal origin
In the period of colonization, roughly between 1500 and 1800, there were two dominant legal systems spread over the colonized countries, namely common-law countries and law countries. Nowadays, the differences between common-law countries and civil-law countries are still noticeable. La Porta et al. (1998, 2002) found evidence for this difference: civil-law countries have a weaker investor’s protection, in comparison to common-law countries. They proved that the relation between corporate valuation in a country and the origin of the legal system whereby the common law median of Tobin’s Q10
, is significantly higher.
The development of corporate governance codes is primarily based on the legal system of the country. Therefore, a clear distinction between corporate governance codes among countries is made in terms of investor’s protectionism. This argument is supported by the findings of Yoshimori (1995), who found empirical evidence based on a survey between CEO’s in the different orientations of countries. These different orientations are one of the main building blocks of the firms, and they are still noticeable when firms start to transform in global players. Therefore, I suggest that the different ways of corporate governance orientation are also a firm characteristic, but it is static.
The revised corporate governance codes from the last decade could be seen as one of the main drivers for ERM adoption (Beasely et al. 2005), together with leadership of the CEO (Acharry & Johnson, 2006) and encouragement from the board of directors (Kleffner et al. 2003). The CEO of a firm should justify the outcomes of the firm’s results to his investors. CEO’s of the firms, which are originating from countries with a higher investors protection, are assumable more focussed on creating value when ERM is adopted. Therefore, I assume that it will potentially create more value for investors, if it is compared with firms that origin from countries with lower investor protectionism. When the interest of shareholders plays a
less dominant role in the importance of decision making from the CEO, there will be less drive to create value for the shareholders.
The arguments, which are debated above, are showing that firms with a higher investor protection mechanism will have a moderating effect between legal origin and the adoption of ERM on firm performance. Therefore, I assume that the legal origin of the firm has a moderating effect on the relation between ERM and firm’s performance.
Hypothesis 4. Legal origin moderates the effect of ERM on firm performance.
2.5
Conceptual model
The main arguments of this research as described in paragraph 2.4, are presented in a conceptual model, which is illustrated in Figure 2.
Figure 2: Conceptual model
Enterprise Risk Management
3
DATA & EMPIRICAL MODEL
In this chapter the data and the empirical method of this research will be presented. First of all, a description of the sample period and the sample selection. Followed by the variables, which are dependent, independent, moderating and control variables. Subsequently the measurement of the moderating effects in the empirical model will be discussed, and finally the summary statistics will be presented.
3.1
Period & sample selection
In order to test the four hypotheses, a sample from the Forbes 200011 is selected. I have chosen for this sample because it contains not only USA, but also EU public traded firms, and they are grouped in different industries. As this research is limited in time, a selection of industries has been made. The sample selection will contain: (1) Major Banking-, (2) Regional Banking-, (3) Diversified Insurance- and (4) Investment Service-industries. Firms in these industries are facing a higher degree of risk, and thence are more likely to engage in ERM. The amount of capital that is circulating through financial firms is higher than the amount of capital for non-financial firms. Therefore, I assume that there is a higher need of a more advanced RM paradigm in these industries. This assumption has given a starting point of a sample of 170 firms. For all these firms, the earliest signal of ERM adoption is traced back, within a time span from 1990 to 2011. The starting point of the year 1990 is chosen to make sure that the earliest ERM adoption is included, and the final year of 2011 is chosen to capture the long-term effect. Firms where non-ERM adoption can be traced back, or where ERM adoption was disputable, are excluded from the sample.12 This has reduced the sample size to 103 firms. If there are missing variables in the data set, all the firm variables from that certain year are excluded. The overall sample size is 1,509 firm-years of observations. Table 1 contains an overview of the list of the earliest CRO appointments, sorted on years and industries. Appendix A, Table 6, contains a list of all firm names.
11
www.forbes.com. The world’s biggest public companies base on four metrics; sales, profits, assets and market value. The market value calculation is as of March 11 2011 closing prices, including all common shares outstanding.
Table 1: Sample statistics for industry and year of first CRO appointment
Major Banks Diversified
Insurance Investment Regional Banks Total
1990 - - - - - 1991 - - - - - 1992 - - - - - 1993 - - - - - 1994 1 - - 1 2 1995 1 - 2 - 3 1996 - - - - - 1997 1 1 - - 2 1998 2 1 1 1 5 1999 - - 2 - 2 2000 3 - 1 - 4 2001 1 1 - 4 6 2002 1 2 - 3 6 2003 3 2 2 4 11 2004 5 4 2 6 17 2005 3 2 3 4 12 2006 1 4 1 - 6 2007 2 - 2 1 5 2008 1 3 2 4 10 2009 - 1 1 - 2 2010 2 - - 2 4 2011 1 2 - 3 6 Total 28 23 19 33 103
3.2
Dependent variables
For this research three proxies of firm performance are used as dependent variables. Tobin’s Q is used as main dependent variable, return of equity and market to book ratio are taken as alternative dependent proxies to check the robustness of the main results.
3.2.1 Tobin’s Q
of measuring firm’s future investment, as the ratio between the market value of a firm to the replacement cost of a firm’s assets. Tobin’s Q is frequently used for measuring firm performance. Chen & Lee (1995) used Tobin’s Q as an alternate measure of firm performance, Lang & Stulz (1994) used Tobin’s Q as a measure of returns from diversification, Hoyt & Liebenberg (2011) used Tobin’s Q to determine the effect of ERM on firm value. Furthermore, Wernerfelt & Montgomery (1988) used Tobin’s Q to estimate the relative importance of industry, focus, and share effects in determining firm performance. During the last decades, complex proxies (e.g. Lindenberg & Ross, 1981) and more simple proxies (e.g. Chung & Pruitt 1994, Perfect & Wiles 1994) are presented to measure Tobin’s Q. Consistent with Bris & Cabolis (2008), Bortolotti et al. (2010), there is chosen for a simple proxy, where Tobin's Q ratio is the sum of market value of a firm and book value of debt divided by the book value of total assets, as shown below in equation (1). The data for the Tobin’s Q ratio are obtained from Datastream.
where, MVi,t = Market value of common equity (Datastream item WC08001) for firm i at year
t, DEBTi,t = the book value of debt for firm i at year t is calculated as book value of total
assets (Datastream item WC02999) minus the book value of equity (Data stream item WC03501), ASSETSi,t = firm i book value of assets (Datastream item WC02999) at year t.
When the ratio is low (<1.0), it implies that market value is undervalued. If the ratio is high (>1.0), it implies that the market value is overvalued.
3.2.2 Return on equity
where, ROE (Datastream item WC08301) is Net Profit for firm i at year t, divided by Common Shareholders Equity for firm i at year t.
3.2.3 Market to book ratio
Even as ROE, the Market to Book ratio is often used to measure the firm performance (e.g. Santo & Becerra, 2008). This performance variable is added as alternative performance measure in the robustness check. In this research Market to Book ratio, see equation (3), is the second alternative measure of firm performance for testing the robustness.
where, Market to Book ratio is Market Value For Company (Datastream item MVC) for firm i at year t, divided by Common Equity (Datastream item WC03501) for firm i at year t.
3.3
Independent variables
This research contains two independent variables. The first implies the start of ERM adoption that is used for the main analyses; the second is a proxy for the full implementation of ERM in a firm, which is used in the robustness check.
3.3.1 ERM adoption
al., 2009; Hoyt & Liebenberg 2011; Pagach & Warr, 2010; and Pagach & Warr, 2011), and surely the most workable method for this research.
Firms are not required to report the adoption of ERM. Therefore, the direct determination of ERM adoption within firms is often unreliable or inaccurate. Another proxy is therefore needed to determine if a firm has adopted ERM. During the emergence of ERM, a new role in the management is created. This role is given to the CRO, who is responsible for implementing and managing the ERM programs (Liebenberg & Hoyt, 2003). Beasley et al. (2005) found in their empirical research significant evidence that ERM implementation is positively related with the presence of a CRO. Therefore, to determine the adoption of ERM for a firm, the proxy of hiring and appointing announcements at enterprise-level of a CRO is used.
To gather the data of the earliest CRO announcement and appointments, the press releases from Lexis-Nexis Academic and annual reports are used. For each firm the sources are screened on following terms, to find the earliest evidence of a CRO appointment; “chief risk officer,” “director of risk management,” “CRO,” and “head of global risk management”, this in combination with the firm’s name. To make sure that the earliest CRO announcement is gathered, a period from 1990 until 2011 is used. The gathered data can be biased, because the first CRO appointment does not have to be the exact first appointment. For minimizing the bias, other sources like, the websites of Reuters, Bloomberg,13 and other media sources are used to verify the gathered announcements, or to look if there are any earlier announcements. Every earliest CRO appointment is, therefore, dated and coded.14
The presence of a CRO is used as a dummy variable, named as “ERM.” The ERM adoption in a particular year is expressed in equation (4), where the first ERM engagement is based on the first CRO appointment, during period 1990-2011.
where CRO is the year of the first CRO appointment, for firm i, and t is the accounting year.
For example, when a firm appoints its first CRO in 2004, the start from that year ERM equals 1 (ERM = 1). Prior to the first CRO appointment, ERM equals 0 (ERM = 0). For the years after the first CRO appointment ERM equals 1 (ERM = 1).
3.3.2 Full ERM implementation
With the implementation of ERM in a firm, it is possible that the firm cannot directly reap the full benefits from the implementation for this new paradigm. In the study of Pagach & Warr (2010), they argue that there is only minimal evidence found in the reduction of risk on the firm’s earnings, because it takes an extended period of time to fully operationalize ERM. Therefore, they cannot fully benefit from ERM adoption within the first years of adopting ERM. To control this argument, the year of ERM adoption is changed into the average time of implementing ERM, starting from the year of the first CRO appointment, see equation (5). This is measured in the robustness check of this research. In a survey from Fraser et al. (2008) is found that the average time to implement ERM is between three and four years. For this research, I selected an average for three years. This average is based on the assumption that the firm did already start its first engagement steps to ERM before the first appointment of CRO.
where CRO is the year of the first CRO appointment, for firm i, and t is the accounting year. For example, when a firm appoints its first CRO in 2004, then in 2007, three years later, Full ERM equals 1 (Full ERM = 1). Prior to the first CRO appointment ERM equals 0 (Full ERM = 0). For the years after Full ERM equals 1 (Full ERM = 1).
3.4
Moderator variables
3.4.1 Firm size
The measurement of the firm size for the main results is measured with the natural logarithm of the Total Assets. It is a commonly used method to measure the firm size (e.g. Gordon et al. 2009).
The other firm size measure, used in the robustness check, is based on the natural logarithm of the number of employees. This firm size measurement is also commonly used (e.g. Goetz Jr., et al. 1991). The variables are further defined in Table 7 of Appendix C.
3.4.2 Firm complexity
where firm complexity is one if firm i is a Major Bank, and zero if firm i is a Regional Bank. The data is extracted from the Forbes 2000.
3.4.3 Environmental uncertainty
For this study, environmental uncertainty is described in terms of environmental volatility as change variability. In the main analysis the volatility is measured by the historical beta, a commonly used method (e.g. Hoyt & Liebenberg, 2011). The variable is further defined in Appendix C, Table 7.
Organizational theorists as Tosi et al. (1973) argued that less fluctuation in volatility indicates a more stable environment. They have operationalized volatility with the use of three variables: (1) market volatility, the coefficient of variation of net sales; (2) technological volatility, the coefficient of variation of the sum of research and development and capital expenditures divided by total assets; and (3) income volatility, the coefficient of variation of profits before taxes. Due to a lack of available data in Datastream, only market volatility is used in the robustness check. Gordon et al. (2009) also used this measurement for environmental uncertainty. The coefficient variation is calculated over year period of t-5 until t, where t is the sample year, to provide a better measurement. Equation (7) shows how the firm’s market coefficients are computed.
where, , Xn,t = uncertainty i in year t, CV(Xi) = coefficient of variation of
uncertainty i, t = 1,2,...,21 to represent years 1990 – 2010, i = 1 to represent market (Datastream item WC01001), and mean of changes over five years of uncertainty i.
3.4.4 Legal origin
the stakeholder’s model. Hinterhuber et al. (2004) argued that USA and UK companies are both shareholders orientated and (mainland) European firms are more stakeholders orientated. This argumentation is used to make a distinction between firms with a higher degree of investor’s protection and a lower degree, see equation (8).
where, firms origin from the UK, Ireland and USA will set equal to one (Legal Origini,t = 1)
and firms origin from EU mainland will equal to zero (Legal Origini,t = 0). The data is
subtracted from the Forbes 2000.
3.5
Control variables
Control variables are used to minimize the bias in this research. Therefore, I included three control variables, (1) Leverage, (2) Accounting year and (3) Industries. They are discussed in the following three sub-paragraphs.
3.5.1 Leverage
where, Leverage (Datastream item WC08221) for firm i at year t.
3.5.2 Accounting year
To control the variability in variables between years, for every year a dummy variable is made, see equation (10).
where Accounting Yeari,t = is the fiscal year and n = 1990, 1991,..,2010, so for every year a
dummy variable is generated.
3.5.3 Industry
Previous research showed that industries could have a strong influence on performance (e.g. Wernerfelt and Montgomery 1988). Therefore, in this empirical research the industry is controlled by four dummy variables for each industry, see equation (11).
(11)
where n = 1,2,..,4 to represent Regional Banks, Major Banks, Diversified Insurance, and Investment Services industries (Industry classification from Forbes 2000).
3.6
Empirical model
relation of ERM and firm performance. Testing the moderating effect of these variables, a multiple regression analysis is applied (Sharma et. al., 1981). Every independent continuous variable is mean centred, so that the effects are easier to interpret (Kreft et al., 1995). Furthermore, the main research is split-up into two regression models. This is necessary for the measurement of firm’s complexity; only Major Banking industry and Regional Banking industry are included. This is due to the lack of available data, in terms of firm complexity proxies. Regression model I, see equation (12), includes the firm complexity, but the Industry dummy is excluded, because the firm complexity is using the differences in industries, where n=975. For Regression Model II, see equation (13), the firm complexity is excluded in the model, but the industry dummy is included, where n=1,509. Therefore, this could result in a more detailed outcome, due to the increase of the sample size.
Regression Model I
Regression Model II
where Pi,t = firm performance (Tobin’s Q), ERMi,t = enterprise risk management adoption,
FSi,t = firm size (ln total assets), FCi,t = firm complexity (dummy international
diversification), EUi,t = environmental uncertainty (Beta). LOi,t = Legal origin (dummy
shareholders orientation) , Leveragei,t = leverage for firm i in year t, Accounting Yeari,t = for
every accounting year there is a year dummy for firm i in year t, Industryi,t = for every
3.7
Descriptive statistics
Table 2 contains the descriptive statistics of the total sample and this is broken down into ERM and non-ERM adoption for each variable, with extension of the sample comparison. Several differences are noteworthy; all the three dependent variables, which measure the firm performance, do not support directly the contention that ERM adoption enhances firm performance. This can be argued by the fact that there is a significantly lower value after adoption of ERM. These results are not contingent with the results of Hoyt & Liebenberg (2011), who found a significant increase of Tobin’s Q after the implementation of ERM. Furthermore, Pagach & Warr (2010) founded a positive change in ROE and market to book ratio, however these findings were not significant. A plausible explanation of this result is that the financial crisis of 2008 has affected the dependent variable negatively.
Table 2: Descriptive statistics
Total Sample ERM No ERM Differences
Mean SD Mean SD Mean SD Differences T-test of Means
Independent Variable
ERM Adoption 0.440 0.497 - - - -
Full ERM Implementation 0.270 0.444 - - - -
Dependent Variable
Tobin's Q 1.09 0.198 1.06 0.137 1.11 0.232 -0.0551 5.74 ***
Return on Equity 13.2 15.7 9.90 19.4 15.9 11.4 -5.98 7.05 ***
Market to Book Value 2.01 1.38 1.70 1.15 2.26 1.50 -0.557 8.17 ***
Moderating Variable Firm Size I 7.87 0.721 8.22 0.643 7.60 0.659 0.621 -18.4 *** Firm Size II 9.46 1.53 9.94 1.47 9.08 1.46 0.863 -11.3 *** Firm Complexity (n=975) 0.446 0.497 0.509 0.501 0.397 0.490 0.112 -3.50 *** Environmental Uncertainty I 1.04 0.590 1.14 0.634 0.966 0.541 0.170 -5.51 *** Environmental Uncertainty II 1.70 57.5 3.74 81.2 0.103 26.8 3.64 -1.11 Legal Origin 0.574 0.495 0.572 0.495 0.575 0.495 -0.0029 0.111 Control Variable Leverage 64.5 26.9 67.0 25.5 62.6 27.8 4.42 -3.21 ** Year 2,001.8 5.640 2006.1 3.28 1998.5 4.83 7.54 -36.0 ***
3.8
Correlation
Table 3 contains the inter-correlations for every variable. The results from Pearson Correlation analyses showed the alternative variables of firm performance measures are positive significant correlated with Tobin’s Q, where values for Return on Equity are (r = 0.213, p < 0.001), and for Market to Book are (r = 0.533, p < 0.001). It can be questioned if Return on Equity is a good replacement measure for Tobin’s Q. For the main firm size measurement and alternative firm size measurement a strongly positive correlation is found (r = 0.831, p < 0.001), so the alternative firm size measure can be assumed that it is a good replacement. The other alternative measure for environmental uncertainty is minimal negative and not significant (r = -0.029, p = ns). Also for this alternative environmental uncertainty measurement it can be questioned if it is a good replacement for the main environmental uncertainty measurement. Furthermore, there are some high correlations among the variables independent variables. To make sure that there is no multi-collinearity between the variables, the variables are tested for variance of inflation factors (VIF). The multi-collinearity for each independent variable the VIF is computed following equation (14).
The results show that there is no multi-collinearity between the variables, because VIF stayed under the four in every single regression model. This suggests that multi-collinearity is not a problem (Marquardt, 1970) in our regression analysis.15
4
RESULTS
This chapter contains the basic results of Regression Model I and II. These main results are checked for its robustness in three different ways. First, the proxies of dependent variables are changed; second, alternative proxies for the moderating variables of firm size and environmental uncertainty are used and finally the terms of the independent variable are changed.
4.1
Basic results
In these basic results the outcomes of both regression models are reviewed. Each regression model is split up in only single moderating effects, as well as in the overall moderating effect of the regression model. To get a better understanding of the real effect, all statistical significant moderating effect outcomes are visualized in different graphs.
Table 4 provides the basic results corresponding to Regression Model I, as discussed in chapter 3. Model 1 and 2 in Table 4 are showing the buildup of the baseline model, which includes independent-, moderating- and control-variables. The results of Model 1 show that ERM has a statistically significant negative β on firm performance. In Model 2 of Table 4 the dummy variables and the other independent variables are added to the model. Noteworthy is that the β of ERM on firm performance is almost reduced to zero and changed into statistically not significant. Furthermore, all the other independent- and control variables are statistically significant, except for firm complexity. Models 3 a,b,c,d of Table 4 provide the moderating terms separately, and Model 4 gives a total overview of the overall model. The overall model of Regression Model I increase in fit, compared with the baseline model. In terms of regression fit Adjusted R2 increased with more than 4%.
Table 4: Regression analysis of Regression Model I
Variables Model 1 Model 2 Model 3a Model 3b Model 3c Model 3d Model 4
Q Q Q Q Q Q Q Intercept 1.075 *** .982 *** .983 *** .987 *** .979 *** .9703 *** .986 *** (.00340) (.00951) (.00956) (.00956) (.00989) (.00942) (.0106) Independent Variable ERM -.0304 *** -0.00457 -.00414 -.0174 ** -.00427 .0223 *** -.00134 (.00516) (.00551) (.00556) (.00652) (.00552) (.00657) (.00854) Moderating Variable Firm Size I -0.0123 * -.0104 -.0138 * -.0123 * -.0139 * .00593 (.00555) (.00639) (.00554) (.00555) (.00542) (.00682) Firm Complexity -0.00564 -.0057 -.0163 * -.00592 -.0041 -.0283 *** (.00627) (.00628) (.0069) (.00628) (.00612) (.00759) Environmental Uncertainty I 0.0184 *** .0184 *** .0175 .0142 * .0164 ** .00998 (.00546) (.00546) (.00543) ** (.0067) (.00533) (.00678) Legal Origin 0.0333 *** .0331 *** .0344 .0336 *** .0584 *** .0573 *** (.00479) (.00481) (.00477) *** (.004801) (.00586) (.00582) Moderating effect
Firm Size I *ERM -.00369 -.0469 ***
(.0064) (.00947) Firm Complexity * ERM .0279 *** .0606 *** (.00772) (.0119) Environmental Uncertainty I * ERM 0.00955 .0105 (.00888) (.00995)
Legal Origin * ERM -0.0538 *** -.0506 ***
(.00758) (.00763)
Control Variable
Leverage -0.00143 *** -.00143 *** -.0014 *** -.00143 *** -.00139 *** -.0014 ***
(.000172) (.00017) (.000171) (.000172) (.000167) (.000165)
Indicator variable
Accounting Year No Yes Yes Yes Yes Yes Yes
Adjusted R2 0.0334 0.471 0.471 0.478 .471 .497 .513
Delta Adjusted R2 - .438 0.000 0.007 .0001 .0262 .0418
F-value 34.649 *** 34.365 *** 33.082 *** 33.999 *** 33.141 *** 36.690 *** 35.185 ***
Table 5: Regression analysis of Regression Model II
Variables Model 1 Model 2 Model 3a Model 3b Model 3c Model 4
Q Q Q Q Q Q Intercept 1.114 *** .961 *** .959 *** .952 *** .941 *** .928 *** (.0067) (.0238) (.0238) (.0246) (.02398) (.0249) Independent Variable ERM -.0551 *** -.021 -.0238 + -.0201 .0272 + .0296 + (.0102) (.0128) (.0129) (.0129) (.0162) (.0167) Moderating Variable Firm Size I -.0505 *** -.0618 *** -.0496 *** -.0506 *** -.0532 *** (.00993) (.0116) (.00995) (.00986) (.0118) Environmental Uncertainty I -.0138 -.0142 -.0251 *** -.0156 + -.0324 ** (.00888) (.00888) (.012) (.00882) (.0122) Legal Origin .0362 *** .0366 *** .036 *** .0745 *** .07602 *** (.0101) (.01005) (.01005) (.0127) (.0128) Moderating effect
Firm Size I * ERM 0.0259 + .00928
(.0139) (.0144)
Environmental Uncertainty I * ERM .0218 .03204 *
(.0158) (.0163)
Legal Origin * ERM -0.087 *** -.09103 ***
(.018) (.0185)
Control Variable
Leverage -.00214 *** -.00211 *** -.00217 *** -.00212 *** -.00215 ***
(.000263) (.000263) (.000264) (.000261) (.000262)
Indicator variable
Accounting Year No Yes Yes Yes Yes Yes
Industry No Yes Yes Yes Yes Yes
Adjusted R2 .018 .256 .257 .256 .267 .269
Delta Adjusted R2 - .238 .001 .0005 .011 .013
F-value 29.378 *** 19.534 *** 19.013 *** 18.938 *** 19.960 *** 18.880 ***
increases in fit, compared with the baseline model. In terms of regression fit Adjusted R2 increased with 1.3%.
Model 3a and 4, from Table 4, and model 3a and 4, from Table 5, show results to test Hypothesis 1, stating that firm size moderates the effect of ERM on firm performance. Model 3a from Table 4 indicates that the moderating effect of firm size on the relation between ERM and firm performance is negative and statistically not significant (β= -.00369, p = ns). Model 4 from Table 4, indicates a negative and statistical highly significant moderating effect (β= -.0469, p = < 0.001). Model 3a in Table 5, indicates a positive and statistical significant moderating effect (β= 0.0259, p < 0.1). Model 4 in Table 5, the moderating effect is positive but statistical not significant (β= 0.00928, p = ns).
Model 3b and 4, from Table 4, show results to test Hypothesis 2, stating that firm complexity moderates the effect of ERM on firm performance. Model 3b, from Table 4, shows a positive and statistical highly significant moderating effect (β= .0279, p < 0.001). Model 4, from Table 4, show the same statistical highly significant moderating effect (β = .0606, p < 0.001) of firm complexity on the relationship between ERM and firm performance.
Model 3c and 4, from Table 4, and Model 3b and 4, from Table 5, show results to test Hypothesis 3, stating that environmental uncertainty moderates the effect of ERM on firm performance. No moderating effect is found for Model 3c (β= 0.00955, p = ns), and 4 (β= .0105, p = ns), from Table 4, and Model 3b (β= .0218, p = ns), from Table 5. Only in Model 4, from Table 5, a positive statistical significant moderating effect is found (β= .03204, p < 0.05).
Model 3d and 4, from Table 4, and Model 3c and 4, from Table 5, show results to test Hypothesis 4, stating that legal origin moderates the effect of ERM on firm performance. In all the four models a negative and statistically highly significance is found for the moderating effect of legal origin. In Model 3b, from Table 4 (β = -0.0538, p < 0.001), Model 4, from Table 4, (β= -.0506, p < 0.001), Model 3c, from Table 5 (β= -0.087, p < 0.001), and Model 4, from Table 5, (β= -.09103, p < 0.001).
4.1.1 Moderating graphs
graphed. This is done for the single moderating effects as for the moderating effect in the overall regression model, which are statistical significant. The graphing method is based on Aiken & West (1991).
The moderating effect of the firm size on the relationship between ERM and firm performance, of the overall model of Regression Model I, is graphed in Figure 3. The graph displays a clear distinction between the adoption of ERM in a high firm size and in a low firm size. This distinction is that ERM is more beneficial to firms that are smaller in size.
Figure 3: The moderating effect of firm size on the relationship between ERM adoption and firm performance, of the overall Regression Model I.
The single moderating effect of the firm complexity of Regression Model I is graphed in Figure 4. The graph displays a clear distinction between the adoption of ERM in a high complexity and in a low complexity. This distinction is that ERM is more beneficial to firms with a high complexity.
Figure 4: The moderating effect of firm complexity on the relationship between ERM adoption and firm performance, of the single Regression Model I.
Figure 5: The moderating effect of firm complexity on the relationship between ERM adoption and firm performance, of the overall Regression Model I.
Figure 6: The moderating effect of legal origin on the relationship between ERM adoption and firm performance, of the single I.
In Figure 7 the moderating effect from the overall Regression Model I of legal origin is graphed. The slope of the high shareholders orientated firms is very similar with the single moderating effect in Figure 6. However, the slope of low shareholders orientated firms became slightly negative, compared to the slope of Figure 6.
The statistical significant moderating effect of the firm size, in Regression Model II, is graphed in Figure 8. The graph displays that the adoption of ERM in terms of firm performance is not beneficial for both smaller and bigger firms. For smaller firms it seems to be even more unbeneficial to implement ERM than for bigger firms.
Figure 8: The moderating effect of environmental uncertainty on the relationship between ERM adoption and firm performance, of the overall Regression Model II.
The moderating effect of legal origin in the overall Regression Model II, is graphed in Figure 9. This graph displays that the adoption of ERM is beneficial for firms in a low, as well in a high uncertain environment, in terms of firm performance.
Figure 9: The moderating effect of legal origin on the relationship between ERM adoption and firm performance, of the single Regression Model II.
