ENTERPRISE RISK MANAGEMENT AS A SOLUTION
TO THE FINANCIAL COMPLEXITY:
A STUDY ON EUROPEAN INSURERS
University of Amsterdam MSc Business Economics, Finance
Master Thesis
by Teun ten Broeke supervised by dr. R.I. Todorov
2
Acknowledgements
Foremost, I would like to take this time to express my sincere gratitude to the University of Amsterdam and in particular my supervisor, dr. Radomir Todorov, who supported me throughout my thesis with his support and enthusiasm. I attribute the level of my master thesis to his effort and excellent advice.
My sincere thanks also goes to Ron Martinek and Göran Vernooij for their counsel and feedback during my internship at PricewaterhouseCoopers.
I would also like to thank my girlfriend, Anne Sophie because she was always there for me in so many ways.
Last but not least, I would like to thank my parents, Herman and Ria, for supporting me spiritually throughout my life.
3
Abstract
The aim of this paper is to assess whether enterprise risk management (ERM) quality enhances firm performance. Several economic actors have stimulated the shifting from a traditional silo-based risk management approach to a more comprehensive one, the so-called enterprise risk management. Despite the increasing interest in risk management, academic research in this area is still limited and controversial. This paper studies the relationship between ERM quality and firm value and performance. Results show that ERM quality is significantly positive related to firm value, measured as Tobin’s Q. A one-level increase in ERM rating is associated with an increase in Tobin’s Q of approximately 1percent.
4
Table of Contents
Acknowledgements 2 Abstract 3 1. Introduction 5 2. Literature Review 72.1. Definition and History 7
2.2. Benefits of Enterprise Risk Management 9
2.3. Empirical Evidence on Enterprise Risk Management 10
3. Methodology and Hypothesis 11
4. Data and descriptive statistics 15
5. Results 19
6. Robustness checks 24
7. Conclusion and Discussion 26
8. References 29
5
1. Introduction
The term enterprise risk management (ERM) is relatively new in business, yet its attention in the media drastically increased. Especially in the aftermath of the recent financial crisis, which once again showed the drastic consequences of shortcomings in risk management, ERM is seen as a critical solution to the economic complexity. The crisis has reinforced the importance of risk management and substantial improvements in risk systems will follow (Jorion, 2009). Whereas risk formerly used to be managed in “silos” of risk categories, ERM provides a framework that integrates risk of all sources and consolidates risk management policies.
ERM is adapted by enterprises globally and various initiatives encouraged the adoption of it. In 2000, for example, the EU initiated the Solvency II Directive for the European insurance sector. The project is designed to ensure that capital requirements better reflect underlying insurer risks, to create a level playing field among insurers across different geographies and to protect policy holders’ interests. This directive has stimulated European insurers to improve risk management practices, however a delay in its initiation has prompted some insurers to reduce their efforts in developing ERM (S&P, 2013). In 2004, the Committee of Sponsoring Organization of the Treadway Commission (COSO) released the Enterprise Risk Management Integrated Framework, which established the first common framework for ERM and a guidance to assist companies in developing and improving their risk management activities. It defines ERM as a process, effected by the board, management and other employees, designed to identify potential events that may affect the organization. ERM assists in managing risk to be within its appetite and provides a sound assurance for attaining corporate goals (COSO, 2004). Therewithal, as of 2006 Standard and Poor’s expanded their credit rating with an ERM component. Companies spend heavily on implementing and optimizing ERM programs, universities give ERM related courses; its value for consolidating the financial system seems acknowledged.
Whereas the traditional risk management approach manages risk in separate silos, ERM is designed to integrate the management of risks in aggregate. Hereby it avoids duplication and coordinates risk management, which results in higher efficiency. Companies that succeed at ERM could have a long-term competitive advantage over those who manage risk individually. A firm may sharply increase her chances to achieve its strategic objectives by aligning the incentives of employees and by measuring and managing risk systematically (Nocco and Stulz, 2006). Shortly, awareness about aggregate risk facilitates better strategic and operational decision making (Meulbroek, 2002; Miccolis and Shah, 2000; Hoyt, Merkley, and Thiessen, 2001; Kleffner, Lee, and McGannon, 2003; Liebenberg and Hoyt, 2003; Beasley, Clune, and Hermanson, 2005). Other
6
benefits of this new framework are less volatile earnings, lower stock volatility, increasing capital efficiency, and reduced external capital costs (Miccolis and Shah, 2000; Cumming and Hirtle, 2001; Lam, 2001; Meulbroek, 2002; Beasley, Pagach, and Warr, 2008; Hoyt and Liebenberg, 2011).
Despite the increasing interest in risk management, academic evidence in this area is still limited. This is most probably due to the difficulty in developing a proper measure for risk management. Some researchers use the appointment of a chief risk officer (CRO) as a proxy for ERM implementation (Beasley, Pagach, and Warr, 2008; Hoyt and Liebenberg, 2010). Others develop their own index. Moreover, empirical evidence on the benefits of this new framework is limited to firms in the US financial sector. Hoyt and Liebenberg (2011) find a positive relation between firm value and the adoption of ERM, an premium of roughly 20%. Baxter et al. (2013) are the first to use Standard & Poor’s review of ERM ratings on insurance companies. They find that higher ERM quality is associated with improved accounting performance. This paper will also use the ratings of Standard and Poor’s as a proxy for risk management.
The empirical evidence on this relationship is limited, yet authorities increase the pressure
on firms to improve their risk management with ERM framework as global standard. Most literature in this field investigate whether the implementation of ERM benefits firm value in the U.S. financial industry. This research takes the quality of a firms’ risk management into account, besides that it targets the European insurance industry. Moreover, it tries to answer the crucial question whether these costly ERM systems indeed boost shareholders’ value. Furthermore, this research will contribute to the field of corporate governance; as ERM is an internal mechanism that firms can implement to mitigate the risk managers may take beyond the overall firm strategy (Bushman and Smith 2001).
This paper tries to assess how the quality of enterprise risk management is related to a firms’ value and performances. In order to investigate this relationship, a sample is created which encompasses all European insurers with coverage in the S&P Ratings Direct database, in a timeframe from 2009 to 2013. As there is no database collecting all ERM ratings collectively, they are manually extracted from company specific reports by S&P. All in all 103 insurers with ERM disclosers are identified, 58 observations are removed with missing Datastream and Orbis data. Fixed effects and random effects panel regressions will be performed to control for unobservable heterogeneity across insurers.
The remainder of this paper is organized as follows. The next section provides a literature review on the development of enterprise risk management, the history of risk management in general, and empirical evidence on the effectiveness of ERM. Thereafter, the methodology and
7
hypothesis of this research are clarified. Subsequently, the data collection procedure will be explained and the descriptive statistics are shown. After that the results of the research are presented, thereupon the robustness checks and additional results are clarified. Finally, the conclusion, implications for existing literature, and suggestions for future research are given.
2. Literature Review
2.1. Definition and HistoryEnterprise risk management is the name for a corporate integrated framework that manages business risk with an entire business risk approach. Enterprise risk management has a rich history of predecessors that assess more or less the same: internal risk management, integrated risk management, strategic risk management, corporate risk management, and business risk management. All these terms have a somewhat different focus, yet the general concept of dealing with corporate risk in aggregate is shared. The definition for ERM that is most commonly used in literature is given by the Casualty Actuarial Society (CAS): "The process by which organizations in all industries assess, control, exploit, finance and monitor risks from all sources for the purpose of increasing the organization's short and long term value to its stakeholders (Casualty Actuarial Society,2003, Overview of Enterprise Risk Management, page 8) ." This definition is extended by classifying risk into hazard, financial, operational, and strategic risk. Hazard risks are those risks that have conventionally been addressed by insurers, in example health and pensions, liability, business interruption or theft. Financial risks concerns potential losses due to changes in financial markets, including interest rates, foreign exchange rates, liquidity risks and credit risk. Operational risks includes a wide variety of situations, including trademark protection, customer satisfaction, product development and failure, management fraud and information risk. Strategic risks mainly concerns technological innovation and regulatory or political impairments. The spine of ERM is that firms should consider all these types of risk in aggregate, rather than assessing it individually (CAS, 2003).
As mentioned earlier, ERM is a relative new approach toward risk management, therefore the academic literature on this subject of matter is rather limited. However, its fundamental can be found in the topic of risk management, which has been extensively discussed by academics. Modigliani and Miller (1958) were the first to consider risk management; they considered it as irrelevant, as it didn’t affect firm value under perfect market conditions. This “risk management irrelevance principle” provoked other academics which proved that markets are not perfect.
8
These imperfections1 bear costs to firms and could be minimized by risk management. This
traditional risk management (TRM) field focusses on the practices of hedging and corporate insurance, and its impact on firm value. Literature on hedging teaches us that it potentially reduces bankruptcy costs by limiting financial distress (Smith and Stulz, 1985). Furthermore, theories suggest that hedging moderates incentive conflicts and increases a corporation’s ability to exploit investment opportunities (MacMinn, 1987; Nance, Smith, and Smithson, 1993; Colquitt and Hoyt, 1997). Graham and Rogers (2002) test the relationship between corporate hedging and derivatives on firm value. They find that firm value increases with 1.1 percent as a result of increasing debt capacity due to hedging too. Other authors2 show a positive relation between firm
value and risk management. Guay and Kothari (2003) test whether hedging is beneficial for all companies, they find that only financial firms take positions in derivatives that are large enough to improve firm value. In line with this, Jin and Jorion (2006) discover that hedging does not affect a firm’s market value in the oil and gas industry. From a corporate insurance point of view theory suggests that corporate insurance reduces bankruptcy costs, agency costs, and taxes (Cole and McCullough, 2006). Empirically, evidence on corporate insurance is ambiguous; some researchers find evidence for the benefits of corporate insurance (Mayers and Smith, 1990; and Hoyt and Khang, 2000), while Regan and Hur (20007) do not.
Notwithstanding all the research in the field of risk management, the credit crisis of 2007 exposed its inadequacies anew. This the traditional risk management tools seem outdated and incapable in tackling the growing economic complexity. Companies, authorities, rating agencies, consulting firms, and universities seem to reach to a similar solution; enterprise risk management. Little by little, this relative new ERM approach is taking over the traditional way of risk management by taking operational and strategic risks into account. The goal is a corporate integrated framework that manages business risk with an entire business risk approach. Unlike the traditional silo-based approach, ERM aims to gain a systematic understanding of the interdependencies and correlations among risks. ERM takes the aggregate of all risk portfolios and then hedges the residual risk. This approach is more efficient and effective than dealing with risk independently (McShane et al., 2010).
A paper by D’Arcy and Brogan (2001) explains that the incentive for enterprise risk management arose when the risk manager and the financial risk manager began reporting to the same chief financial officer (CFO). Each risk department had its own peculiarities, obviously a
1 Imperfections as tax payments (Mayers and Smith, 1982; Smith and Stulz, 1985; Mian 1996; Ross, 1996; and
Graham and Rogers, 2002), financial distress (Mayers and Smith, 1982; Smith and Stulz, 1985; and Nance, Smith, and Smithson, 1993), asymmetric information (DeMarzo and Duffie, 1995), underinvestment (Myers, 1977; and Froot, Scharfstein, and Stein, 1993) and under-diversified stakeholders (Stulz, 1984; and Mayers and Smith, 1990).
9
common approach to risk management would be more effective. ERM manages such risks categories collectively and can be seen as a risk portfolio framework. Besides, other factors have contributed to the development and initiation of enterprise risk management. For example, innovations in computing powers allow us to perform comprehensive risk analysis for risks in interest rates movement or for hazard risks. The ability to store extensive amounts of data permit us to examine historical information to determine trends, correlations and other relationships among variables that is essential to enterprise risk management (D’Arcy and Brogan, 2001). Insurers are also more and more specializing on financial risk management. Some insurers are started to provide policies that coordinate financial and pure risk; one insurer has initiated a policy that provides protection against foreign currency losses (Banham, 1999).
2.2. Benefits of Enterprise Risk Management
In theory, the benefits of enterprise risk management range to both a macro and micro level. A major macro level benefit is the reduced probability of adverse cash flows shortfalls. Let’s say a firm expects a cash flow of €50 million but suffers a loss of €50 million instead, this adverse cash flow of €100 million has costs that go beyond this cash flow itself. This cash flow shortfall has implications for investors’ view on the future growth of the firm, and the reduced firm value most probably exceeds the adverse cash flow. As a result the company may need substantial funds. If the firm already has an optimal amount of leverage; both equity and debt need to be issued to maintain this. As issuing equity is a rather expensive way of raising new funds, it is often the case that firms cut on their investments. Enterprise risk management can reduce the probability of such an adverse cash flow as it assists firms in their risk-return decisions (Nocco and Stulz 2006). Firms may avoid losses and bankruptcy costs by mitigating the probability of detrimental financial events (Pagach and Warr 2008).
The traditional approach of managing risk individually and the lack of coordination between risk departments could lead to costly inefficiencies. ERM provides the possibility to deal with risk in aggregate by aligning these different departments and avoids expensive duplication of risk management expenditures. These benefits may result in better resource allocation, which increases capital efficiencies and return on equity. Moreover, ERM will help firms deciding in investment opportunities because of more accurate risk adjustments (Meulbroek, 2002). Another potential benefit of ERM is that it might improve a firm’s transparency and enables it to inform outside organizations of their risk profile. Consequently, it may reduce the expected costs of external capital (Meulbroek, 2002). If we look at the broader picture, risk management potentially increases the awareness about aggregate risk. This facilitates enhanced strategic and operational
10
decision making (Meulbroek, 2002). As a result, return increase and capital can be allocated more efficiently (Hoyt and Liebenberg 2011; McShane et al. 2011).
Briefly, the theoretical literature can be summed concisely as follows; enterprise risk management potentially benefits organizations by a decrease in earnings volatility and stock volatility, increasing capital efficiency, and reducing external capital costs (Miccolis and Shah, 2000; Cumming and Hirtle, 2001; Lam, 2001; Meulbroek, 2002; Beasley, Pagach, and Warr, 2008; Hoyt and Liebenberg, 2011).
2.3. Empirical Evidence on Enterprise Risk Management
One of the first papers that assesses the value of risk management is written by Smithson and Simkins (2005). The writers provide a comprehensive review of the relationship between risk management and firm value. They find that although prior literature investigates the effectiveness of various forms of hedging and insurance, no one investigated the relevance of enterprise risk management yet. Hereafter, the interest in the subject increased tremendously and two main streams in the enterprise risk management literature can be identified. One investigates the its benefits and company characteristics associated with the implementation of ERM, the other explores how risk management actually improves firm value and how the market reacts to ERM implementation. The first field of literature will be analysed in the next methodology section because it helps understanding the variable construction for the regression models, the latter will be discussed here.
The empirical evidence on the effect of enterprise risk management is rather limited; most probably due to the difficulty in developing a proper measure for risk management. Pagach and Warr (2010) try to measure the effect of ERM on firm performance by measuring returns on the announcement of hiring a CRO. Their results fail to find support for the proposition that ERM creates value. Gordon, Loeb, and Tseng (2009) create their own index and find that the effectiveness of ERM depends on the presence of certain firm specific factors. Beasley, Pagach, and Warr (2008) perform an event study on the market reaction of the appointment of senior executives to revise the firm’s ERM process. They find minor benefits for non-financial companies.
Three papers (McShane et al., 2010; Hoyt and Liebenberg 2011; and Baxter, et al., 2013) do find a significant positive relationship between enterprise risk management and firm value or firm performance. Hoyt and Liebenberg (2011) investigate the value associated with ERM, they use the treatment effects model to estimate the effect of ERM, a binary treatment, on Tobin’s Q. In this paper a detailed search of financial reports, newswires, and other media is analysed to
11
identify ERM activity. The authors also find that on average, insurers with ERM programs are valued approximately 4% higher than other insurers. Furthermore, ERM users are larger, less leveraged, less opaque, and have lower return volatility compared to companies without enterprise risk management. Two other two papers use the Standard and Poor’s ratings on risk management as a proxy for ERM and test its relation to firm value. McShane, Nais and Rustambekov (2010) find a positive relationship between traditional risk management capabilities and firm value, measured as Tobin’s Q. They conclude that market seems to value risk management quality to a certain level, but achieving “extreme” high levels ERM does not additionally increase firm value. Unfortunately, the data of this paper is limited to the year 2008. The research by Baxter et al. (2013) distinguishes between high quality ratings and low quality ratings. They find that ERM quality is associated with improved accounting performances. Their results show that both firm performance and value are improved by high-quality controls that integrate risk management efforts across the firm. Previous literature mainly focuses on pre-crisis years and is limited to the financial and insurance sector of the U.S. market. This paper investigates the relationship between ERM quality and the value and performance of European insurers. Moreover, it distinguishes between different levels of ERM quality and is the first to investigate how listed and non-listed firms are differently affected by this new framework.
3. Methodology and Hypothesis
In theory firms with higher ERM quality are more capable in identifying opportunities and risks which will result in lower stock volatility, decreased earnings volatility, and lower capital costs. If this theory holds in practice, one can infer that firms with higher enterprise risk management should have a higher market value than firms with relatively lower ERM quality. Accordingly the following hypothesis is set.
HYPOTHESIS 1: A firms’ ERM quality is positively related to its market value.
McShane et al. (2011) find that firm value increases with risk management quality to a certain degree. Hoyt and Liebenberg (2011) find a firms with an ERM system have higher Tobin’s Q ratio’s than ones without. Baxter et al. (2013) find significant evidence that risk management quality is associated with higher firm value for U.S. financial services firms. This paper will add to the current literature by researching an unexplored area, besides it seeks to answer whether ERM quality has similar effects on listed and non-listed firms.
12
performance by assisting in risk-return decision making which reduces the probability of adverse cash flow. Consequently, the benefits of ERM could enhance a firms’ accounting return. Baxter et al. (2013) is the only paper to find a significant positive association on this relationship. The following hypothesis is conceived.
HYPOTHESIS 2: A firms’ ERM quality is positively related to its accounting performance.
The first hypothesis tests the relationship between firm value and the quality of an insurers’ risk management. This can be examined by modelling the Tobin’s Q ratio as a function of ERM quality and other controlling variables. Tobin’s Q (TQ t) is the most commonly used proxy for
firm value in corporate finance, and for risk management studies in particular (Smithson and Simkins, 2005). Tobin’s Q does not require normalization or risk adjustments; therefore it dominates other performance measures on stock returns and accounting (Lang and Stulz, 1994). Therewithal, this ratio reflects the market expectations of a firm, which makes it insensitive for managerial manipulation of accounting information (Lindenberg and Ross, 1981). Tobin’s Q is calculated as the ratio between the sum of the market value of equity plus the book value of liabilities over the book value of assets. This version of Tobin’s Q is suitable for insurance companies in particular because the book value of an insurer’s assets is a good approximation of its replacement costs (Cummins, Lewis, and Wei, 2006; Hoyt and Liebenberg, 2010). Model 1 below shows the panel regression specifications.
The independent variable of interest is the Standard and Poor’s ERM t quality rating, which is
part over the overall credit rating. As of 2006 the rating agency evaluates enterprise risk management quality for firms in the U.S. financial industry. Shortly thereafter European financials and insurers were also provided with this rating. An insurers’ ERM is classified as (1) weak, (2) adequate, (3) adequate with strong controls, (4) strong, or very strong (5)3 based on 5
subfactors4. An insurer rated as weak has sporadic risk management with limited capabilities to
3This classification changed in 2010, formerly it was categorized as (1) weak – (2) adequate – (3) adequate strong
control – (4) adequate positive trend – (5) strong.
4 “The first and most important subfactor, risk management culture, focuses the importance accorded to risk and
ERM in all key aspects of the insurer's business operation and corporate decision-making. As risk management culture encompasses all aspects of the ERM framework and all the ERM subfactors are interconnected, it is difficult to evaluate this subfactor without reference to the others. For that reason, the analysis of the risk management culture subfactor focuses on the insurer's philosophy towards risk, especially its risk appetite framework, risk governance and organizational structure, risk communications and reporting, and the embedding of risk metrics in its
13
consistently identify, measure, and manage risk exposures. An adequate insurer can identify do this for key risks, but the process has not been extended to all significant risks facing the enterprise. Major risks may still be managed in silos instead of coordinated across the firm. The ERM program is rated adequate with a positive trend if it exhibits strong/excellent risk control systems, but still lacks a well-developed process for making coordinated risk/reward decisions that are necessary for effective strategic risk management. A strong ERM program has progressed beyond silo risk management to deal with risks in a coordinated approach, the capability to envision and handle emerging risks, and risk management is an important consideration in corporate decision-making. The very strong insurer’s risk control processes are leading edge, applied consistently, and executed effectively. The insurer continues to develop its risk control processes to integrate new technologies and adapt to the changing environment (S&P RatingsDirect: Enterprise Risk Management, 2010). The complete methodology for these definitions is provided in the appendix, in table 1.
In order to investigate the relationship between risk management quality and firm value, the model controls for variables found in literature which explain differences in firm performance and its value. Previous research suggests that larger firms (Size t) are more likely to employ ERM
programs (Yermack 1996; Colquitt, Hoyt, and Lee, 1999; Liebenberg and Hoyt, 2003; Kleffner et al. 2003; Beasley, Clune, and Hermanson, 2005). In general large and diversified firms have a greater incentive to integrate risk as they face a broader scope of risk. The potential benefits for these firms are considerably larger than for small companies.As larger insurers should be capable of utilizing economies of scale in underwriting insurance contracts, size is expected to be positively related to performance. Liebenberg and Sommer (2008) find that larger property-liability insurers generate higher returns on equity, and McShane and Cox (2009) find similar results for life-health insurers, which they attribute to the greater market power, economies of scale, and lower insolvency risk of larger insurers. However, Lang and Stulz (1994) and Allayannis and Weston (2001) find a significantly negative relation between size and firm value. As in
compensation structure (S&P, 2013). The second subfactor, risk controls, analyzes the processes and procedures insurers employ to manage their key risk exposures within the general areas of credit and counterparty risk, equity risk, interest risk, insurance risk (including reserving risk), and operational risk. The specific risks on which the analysis focuses are a function of the insurer's business and risk profiles. The emerging risk management subfactor analyzes how the insurer addresses risks that are not a current threat to creditworthiness, but could become a threat in the future. The analysis of risk models focuses on assessing the robustness, consistency, and completeness of the insurer's risk models, including, where relevant, its development and use of an economic capital model, and the processes for model governance and validation. Last, the strategic risk management subfactor assesses the insurer's program to optimize risk-adjusted returns and to evaluate and prioritize strategic options on a level playing field. Firms with a higher ERM rating should have an advantage in anticipating and dealing with the next big risk, lower volatility of earnings, and greater ability to allocate capital to attain higher risk-adjusted returns” (S&P RatingsDirect: Enterprise Risk Management, 2013).
14
previous literature ( Liebenberg and Hoyt, 2003) the log of the total book value of assets is taken to control for size-related variation in firm value.
A leverage (Leverage t) variable is included in order to control for the relationship between
an insurers capital structure and its firm value. Up till now evidence this relation is mixed; on the one hand leverage implies greater default risk, then rational policyholders should pay lower prices for policies issued by more leveraged insurers (Sommer, 1996). On the other hand, financial leverage enhances firm value to the extent that it reduces free cash flow that might otherwise have been invested by self-interested managers in suboptimal projects (Jensen, 1986). Leverage is measured as ratio of total liabilities to total assets and the predicted sign is ambiguous.
A return on assets (ROA t) variable is included in order to control for profitability. Firms
with higher returns are more likely to be valued at a premium (Allayannis and Weston, 2001), therefore a positive relationship between profitability and firm value is expected. Furthermore, it is custom to control for the effect of future growth opportunities on Tobin’s Q. Previous articles do this by taking a sales growth variable (Allayannis and Weston, 2001; Bhagat and Black, 2002) and/or a capital expenditure over sales ratio (Yermack, 1996). Because of insufficient data on sales, Revenue(RevenueGrowth t-1) is used as an alternative. This variable takes the annual Revenue
growth in percentage over the previous year in order to control for future growth opportunities. Yermack (1996) finds evidence that smaller boards are more effective, resulting in higher Tobin’s Q ratios, therefore the variable LogBoardSize t is included. To control for potential differences in Q that are related to the industry sector in which an insurer operates, a dummy variable Life t is
included. This variable takes 1 for insurers that are primarily life insurers and 0 otherwise. The expected sign for this variable is ambiguous. An overview of all variable definitions, expected signs, and sources are available in the appendix, table 2.
The second hypothesis predicts a positive relationship between an insurers’ ERM quality and its accounting performance, measured as the return on assets (ROAt). In order to test this
relationship the same control variables are employed as in the first model. Clearly the ROA variable is removed as a control variable in this case. Model 2 below shows the regression specifications.
15
Before going into further detail about the regression estimations it is important to first address the fixed effect versus random effects debate. This discussion concerns the nature of unobserved effects and some features of the observed explanatory variable. In panel data the variable of interest in the regression model may contain observable variables that change across t but not i, variables that change across i but not t, and variables that change across both. Most research discusses whether these unobserved variables (also known as latent variable, unobserved component, and unobserved component) will be treated as random effects or fixed effects (Woolridge, 2010). In general fixed effects is considered a more convincing tool for estimating the ceteris paribus effect because it allows arbitrary correlation, whereas random effects does not. However when the variable of interest is rather constant over time, a fixed affect approach is not applicable. A solution to this problem is conceived by Hausman (1978), who was the first to come up with a test for significant differences in the coefficients on the time varying explanatory variables (Woolridge, 2010, page 493). Considering the low within variation of ERM ratings on insurers individually the Hausman test is applied on the first regression model. The Hausman (1978) test assumes to use random effects, yet a rejection of implies that fixed effects should be used. The Hausman test is performed for both regression models. For the first model no evidence is found to reject the Hausman (1978), therefore GLS random effects model will be used. Consequently the test is performed for the second model with ROA as key explanatory variable. In this case the random effects assumptions are rejected and the fixed effects estimates will be applied.
4. Data and descriptive statistics
This paper focuses on all European insurers with coverage in the S&P Rating Direct database, in a timeframe from 2009 to 2013. The ERM ratings are manually extracted from individual company reports by S&P, 104 insurers are identified. Datastream, Orbis, and annual financial reports are consulted for company specific information, 59 observations are removed with missing data. The final sample covers 45 firms. As the first research question concerns the effects of ERM quality on Tobin’s Q, solely the 23 publicly traded insurers are suitable. For testing the second hypothesis all 45 insurers will be used, additionally two subsamples are created. By distinguishing between listed and non-listed insurers, it can be investigate how the ERM quality of these subgroups affects ROA differently.
Table 3 summarizes the descriptive statistics of the variables used in the first regression model. Tobin’s Q represents the ratio of the market value of firms’ shares to its replacement cost of
16
physical assets. In an ideal state Tobin’s Q is in equilibrium and approximately equal to 1, consequently it can be concluded that this sample is in equilibrium with a Tobin’s Q mean of 0.9939. Insurers with a Tobin’s Q ratio higher than one should invest in the firm because the profits will exceed the cost of assets; vice versa firms with a ratio lower than one preferably sell their assets. The sample exists of 101 ERM ratings with a mean score is 3.1 on a scale from one to five. None of the insurers has a weak ERM, 30 are rated as adequate (ERM2), 32 as adequate with strong control (ERM3), 37 as strong (ERM4), and 2 as very strong (ERM5). Furthermore the LIFE variable of 0.722 indicates that circa 72 percent of the sample insurers are primary life insurers.
Table 3 - Descriptive statistics of model 1
Notes: This table summarized the descriptive statistics of the variables used in model 1. The table presents the amount of observations, mean, standard deviation, minimum and maximum of all variables.
Table 4 shows the Pearson correlations coefficients of the variables used in model 1. Remarkably size is negatively correlated to firm value, yet the relationship is rather insignificant. The remainder of the signs of the correlations between the independent variables and firm value are as expected. The correlations between two independent variables are slightly higher than 0.5; life and size (0.6424), and LogBoardSize and size (0.5237). Consequently a variance inflation factors (Belsley, Kuh, and Welsch, 1980) is computed and available in Table 5. With an average VIF of 2.18 multicollinearity seems inapplicable.
Variable Obs. Mean Std. Dev. Min. Max.
TQ 101 .9975073 .0321026 .9296797 1.123662 ERM 101 3.108911 .8590808 2 5 Size 101 18.12118 1.672195 14.53565 20.50624 Life 101 .7227723 .4498625 0 1 LogBoardSize 101 2.538823 .3684003 1.609438 3.295837 Leverage 101 .9621722 .1718944 .7654966 1.785385 ROA 101 1.411738 1.708182 -3.377425 9.792079 RevenueGrowth 101 .0525819 .1298577 -.21 .6 Beta 97 1.068357 .3977254 .329053 1.98 CR 83 5.204819 .6199732 2 6 ERM1 0 - - - - ERM2 30 - - - - ERM3 32 - - - - ERM4 37 - - - - ERM4 2 - - - -
17 Table 4 - Pearson Correlation Coefficient of model 1
Notes: This table provides the Pearson correlation coefficients for the variables used in model 1. This table includes the variables beta and credit rating, these variables will be used for robustness checks.
The second research question addresses whether ERM quality influences firm performance. In this case, the complete sample group can be used. Moreover, by differentiating between listed an non-listed insurers, differences in the effect of ERM quality can be analysed. Table 5 shows the descriptive statisticsfor all variables used in the second model. Additionally, it presents the differences between ROA and ERM for the listed and non-listed firms. The average risk management rating for listed firms is slightly higher than for non-listed insurers, 3.08 and 2.76 respectively. Surprisingly, the mean ROA for non-listed insurers is approximately 2%,while the ROA for listed sample is just 1.3%. This is most probably due to the fact that larger firms are in general more stable than smaller ones. If the control variables are compared to those the descriptive statistics in the previous section, it can be concluded that listed the listed insurers are slightly smaller in terms of total assets and board size. Besides, they are lower leveraged and have lower revenue growth.
TQ ERM Size Life LogB. Lev ROA Turn. Beta CR
TQ 1.0000 ERM 0.1842 1.0000 Size -0.2427 0.3147 1.0000 Life -0.2255 0.0326 0.6424 1.0000 LogB. 0.0221 0.0949 0.5237 0.3040 1.0000 Leverage 0.1613 -0.0978 -0.1696 -0.3255 0.0257 1.0000 ROA 0.4504 0.0497 -0.4579 -0.3871 -0.1967 0.2645 1.0000 Revenuegr. 0.1587 0.0214 -0.2982 -0.2423 -0.1447 0.1788 -0.1788 1.0000 Beta -0.2449 -0.0499 0.4902 0.3182 -0.1707 0.1057 0.1057 -0.2249 1.0000 CR 0.0768 0.2440 0.0928 -0.1104 0.0561 0.2668 0.2668 -0.0260 0.0396 1.0000
18 Table 5 - Descriptive statistics of model 1
Notes: This table summarized the descriptive statistics of the variables used in model 2. The table presents the amount of observations, mean, standard deviation, minimum and maximum of all variables.
The Pearson correlation coefficients for model 2 are presented in table 6. Most surprisingly ERM quality is negatively correlated to ROA, which implies that risk management quality negatively influences a firms’ ROA. The correlation between these variables is quite little, therefore it is expected that ERM quality does not influence ROA significantly. Whereas the log of board size was positively correlated to Tobin’s Q, it is negatively correlated to the ROA. Again, size is negatively correlated to the dependent variable. The remaining variables show correlations as expected and the Pearson correlation coefficients do not give any reason to suspect multicollinearity.
Table 6 - Pearson Correlation Coefficient of model 1
Notes: This table provides the Pearson correlation coefficients for the variables used in model 2.
Variable Obs. Mean Std. Dev. Min. Max.
ROA 184 .0165244 .0216123 -.07426669 .109273 ERM 184 2.945652 .872959 2 5 Size 184 17.2779 2.142791 12.46579 20.50624 Life 184 .7119565 .4498625 0 1 LogBoardSize 184 2.359434 .4627093 1.098612 3.295837 Leverage 184 .9200832 1452811 .6637523 1.785385 RevenueGrowth 181 .0427667 .1350186 -.3 .68 Beta 127 1.098821 .4113264 .329053 1.98 CR 149 5.147651 .5498874 2 6 ERM-Listed 104 3.076923 .8667796 2 5 ERM-Nonlisted 80 2.775 .856472 2 5 ROA-Listed 100 .013092 .0175356 -.0337742 .0979208 ROA-Nonlisted 84 .0206106 .0251394 -.0742667 .109273
ROA ERM Size Life LogB. Lever. Revenu egr. ROA 1.0000 ERM -0.0007 1.0000 Size -0.2677 0.1831 1.0000 Life -0.3450 0.0878 0.4358 1.0000 LogB. -0.1882 0.0690 0.5289 0.2094 1.0000 Leverage 0.2118 -0.1131 -0.0999 -0.3528 0.0733 1.0000 Revenuegr. 0.2951 0.0112 -0.2201 -0.1806 -0.1567 -0.1634 1.0000
19 5.
Results
Hypotheses 1 predicts that ERM quality will be positively associated with firm value. Results of the GLS random effects model 1 are presented in Table 6; four different regressions are performed to explore this relationship. Column 1 shows a significant (p<0.05) positive ERM coefficient, in line with the hypothesis. This regression includes both time and country random effects. The coefficient of 0.0102 indicates that a one level increase in ERM rating is associated with a 1.02% increase in Tobin’s Q. Most results of the control variables are in line with the expectations. Leverage is significantly (p<0.01) positive related to TQ, most probably because leverage reduces a firms’ free cash flow that might mitigate manager to engage in self-interested investment. Insurers with higher profitability (ROA) are indeed valued at a premium. Moreover, the dummy variably life is positively associated with TQ suggesting that life insurers are slightly higher valued than non-life insurers.
Remarkably, size is negatively associated with TQ just as in the paper of McShane et al. (2010). A possible explanation might be that in general larger firms are more stable and less volatile, while investors prefer to devote their funds into riskier stock. Also against the expectations is the negative coefficient of the revenue growth. This might be explained by the weakness of the proxy, which was a results of limited available data. Ideally one controls for growth opportunities with capital expenditures, research and development, or sales ratios. However, due to limited data on these items the alternative revenue growth was created. A second explanation for this relation might be that insurers pursue to increase their market share at cost of firm value (McShane et al., 2010). Column 2, 3 and 4 use the same regression specifics, however for matters of validity the time random effects and country random effects are modified. All ERM coefficients remain positively significantly (p<0.05) related to firm value with 1.12%, 1.13%, and 0.95% respectively.
Model 1 investigated the average effect of the shift in ERM quality on firm value. A problem of the ERM quality ratings is the fact that this variable is an ordered response; a firm is rated as weak, adequate, adequate with strong controls, strong, or very strong. When these responses are given values (1,2,3,4 or 5) the outcomes are no longer arbitrary. Obviously, insurers rated as strong have better ERM systems than insurers rated as adequate. However, the rating itself only has ordinal meaning; it is unclear whether an increase in ERM rating from four to five is as important as an upgrade from one to two (Woolridge, 2010).
20 Table 6 - Results of estimating Model 1
(1) (2) (3) (4) (5) (6) VARIABLES TQ TQ TQ TQ TQ TQ ERM 0.0102** 0.0112** 0.0113** 0.00957** (0.00432) (0.00558) (0.00494) (0.00483) Size -0.00153 -0.00181 -0.00246 -0.00120 -0.00251 -0.00311 (0.00291) (0.00362) (0.00275) (0.00349) (0.00338) (0.00363) Life 0.00901 -0.0106 0.0112 -0.0115 0.0115 0.0118 (0.0109) (0.0134) (0.0113) (0.0131) (0.0111) (0.0112) LogBoardSize 0.00729 0.0165 0.0124 0.0152 0.00836 0.0121 (0.0144) (0.0132) (0.0139) (0.0130) (0.0150) (0.0156) Leverage 0.0532*** 0.0190 0.0542*** 0.0204 0.0573*** 0.0564*** (0.0148) (0.0146) (0.0141) (0.0150) (0.0178) (0.0173) ROA 0.585 0.544* 0.569* 0.545 0.539 0.539 (0.371) (0.293) (0.303) (0.364) (0.360) (0.353) RevenueGrowth -0.0223 -0.0239 -0.0244 -0.0223 -0.0200 -0.0203 (0.0215) (0.0195) (0.0198) (0.0207) (0.0231) (0.0228) ERM1 - - -0.0117 ERM2 (0.0102) ERM3 0.00620 (0.00838) ERM4 0.0227** 0.0148** (0.00919) (0.00670) ERM5 0.0199 0.0113 (0.0130) (0.0125) Constant 0.923*** 0.936*** 0.918*** 0.934*** 0.962*** 0.969*** (0.0633) (0.0634) (0.0561) (0.0693) (0.0616) (0.0621) Observations 101 101 101 101 101 101 R2-overall 0.6221 0.2760 0.5796 0.3155 0.5768 0.5886 Number of id 23 23 23 23 23 23
Time random effects YES NO YES NO YES YES
Country random effects YES NO NO YES YES YES
Notes: This table presents the results of estimating model 1. In regression five and six the ERM variable is decomposed into category dummies. ERM1 is set to 1 if the insurer’s S&P rating is adequate, and zero otherwise. ERM2 is set to 1 if the insurer’s S&P rating is adequate, and zero otherwise. And so on. The model is estimated with standard errors correcting for clustering at the year and firm level. Numbers in the cells are t statistics (coefficients) with significance denoted as *, **, ***, for ten percent, five percent, and one percent, respectively.
21
By using dummy variables for each ERM category it can be examined how a shift from one specific rating to another is valued by the market. The regression results of the decomposed ERM dummies5 are available in columns 5 and 6, the control variables are similar to those of
model 1. Column 5 takes the ERM2 ratings as a base, this implies that all ERM2 ratings are omitted in order to overcome the dummy variable trap. Results for the ERM3, ERM4, and ERM5 correlation coefficients are in relation to an ERM2 rating. A risk management rating of 2 signifies adequate ERM, with this category as a base it can be investigated how the market appreciates extraordinary risk management. Column 1 shows that the coefficients ERM3,ERM4 and ERM5 are positively associated with TQ, however ERM4 is the only coefficient with a statistically significant (p<0.05) relationship. In column 6 ERM3 is taken as a base, the results are rather similar. Again, ERM4 shows significant relevance (p<0.05) and ERM5 is positively associated compared to ERM3. Understandably, ERM2 is negatively associated with TQ with respect to the base of ERM3. The results of both columns imply that firm value will be increasing from ERM2 to ERM4, however achieving a very strong (ERM5) rating does not increase firm value significantly. The regression coefficients of all ERM categories show a pattern of market reaction to the quality of risk management; higher ratings suggest higher economic significance. Collectively, table 6 shows that ERM quality is positively related to firm value, measured as Tobin’s Q. On average a point increase in risk management will result in an increase in value of about 1%. The market seems to value the shift from a limited, but adequate, traditional risk management approach to the new framework of enterprise risk management. This market view contradicts the results of McShane et al. (2010), who find that increases in ERM rating beyond ERM3 do not add additional firm value. This difference could be explained by a switch in market perception. It is possible that investors started recognizing the importance of ERM quality. The second hypothesis predicts that the ERM quality of an insurer will be positively associated with firm performance, measured as the return on assets. Results of the fixed effects regression model are presented in table 7. Three different regressions are performed to investigate the effect of risk management quality on ROA, no significant relationship is found. As a matter of fact, the first column shows that ERM quality is negatively correlated to ROA. The board size of an insurer on the contrary seems to be a good predictor for ROA (p<0.05). The result of this regression coefficient is in line with the evidence of Yermack (1996) who finds that smaller boards are more effective, resulting in higher performance.
5The ERM variable is divided into dummy variables in the following way; ERM1 is set to 1 if the insurer’s S&P
rating is weak, and zero otherwise. ERM2 is set to 1 if the insurer’s S&P rating is adequate, and zero otherwise. ERM3 is set to 1 if the insurer’s S&P rating is adequate with a positive trend, and zero otherwise. ERM4 is set to 1 if the insurer’s S&P rating strong, and zero otherwise. ERM5 is set to 1 if the insurer’s S&P rating very strong, and zero otherwise.
22
Table 7 - Results of estimating Model 2
(1) (2) (3)
VARIABLES ROA ROA-Nonlisted ROA-Listed
ERM-Listed 0.00321 (0.00498) ERM-Nonlisted -0.00558 (0.00730) ERM -0.00297 (0.00383) Size 0.00569 0.0305 0.0164 (0.0158) (0.0180) (0.0120) Life - - - LogBoardSize -0.0305** -0.0547* -0.0162 (0.0136) (0.0294) (0.00978) Leverage -0.155 -0.568*** 0.0771 (0.255) (0.188) (0.200) RevenueGrowth 0.0168 -0.00558 0.0414 (0.0155) (0.0133) (0.0253) Listed - Constant 0.142 0.152 -0.329 (0.248) (0.201) (0.297) Observations 177 77 100 R-squared 0.138 0.376 0.219 Number of id 45 22 23
Time fixed effects YES YES YES
Country fixed effects YES YES YES
Notes: This table presents the results of estimating model 2. In columns 2 and 3 the sample is divided into non-listed and non-listed insurers. The model is estimated with standard errors correcting for clustering at the year and firm level. Numbers in the cells are t statistics (coefficients) with significance denoted as *, **, ***, for ten percent, five percent, and one percent, respectively.
In columns 2 and 3 the sample is divided into two subgroups, as explained in the methodology. The second column shows results for the relationship between the risk management ratings and firm performance for non-listed insurers. The coefficient of -0.00558 entails that ROA will drop roughly half a percent per level increase in ERM, however this relation does not show any statistical significance. Leverage seems to be an important determinant for the ROA for non-listed insurers, this variable shows a significant relation (p<0.01). In contrast to the first two regressions, column 3 predicts a positive, non-significant, relation between ERM quality and ROA. Apparently the negative association for non-listed firms (-0.00558) is stronger than the positive relationship for listed firms (0.00321), resulting in a collective negative association
23
(-0.00297). None of the control variables are significantly predictors of a firms’ ROA.
Next the ERM variable is decomposed similarly to the first model, consequently it can be investigated how the ERM categories differently affect ROA. Table 8 presents the regressions of model 2 with ERM category dummies, taking ERM2 as the base. The first column shows that ERM quality negatively influences a firms’ ROA to a certain degree. The coefficients of ERM3 and ERM4 show a negative relation compared to ERM2. However, having a superior risk management (ERM5) most probably leads to higher returns on assets in relation to adequate risk management (ERM2). Column 2 displays how risk management quality influences ROA for non-listed insurers. The coefficient of ERM4 shows that insurers with strong ERM have statistically significant (p<0.001) lower ROA. The last column on the other hand, describes that strong and very strong risk management is associated with higher levels of ROA for listed insurers.
Table 8 - Results of estimating Model 2 with decomposed ERM categories
Notes: This table presents the results of estimating model 2. The ERM variable is decomposed into category dummies. ERM1 is set to 1 if the insurer’s S&P rating is adequate, and zero otherwise. ERM2 is set to 1 if the insurer’s S&P rating is adequate, and zero otherwise. And so on. The model is estimated with standard errors correcting for clustering at the year and firm level. Numbers in the cells are t statistics (coefficients) with significance denoted as *, **, ***, for ten percent, five percent, and one percent, respectively.
(1) (2) (3)
VARIABLES ROA ROA-Nonlisted ROA-Listed
ERM1 - - - ERM3 -0.00366 0.000359 -0.000297 (0.00330) (0.00289) (0.00234) ERM4 -0.00308 -0.0284*** 0.00791 (0.00871) (0.00261) (0.00847) ERM5 0.00136 -0.00739 0.0126 (0.0110) (0.00956) (0.0131) Size 0.00926 0.0402* 0.0145 (0.0147) (0.0203) (0.0119) Life - - - LogBoardsize -0.0178 -0.0316 -0.0120 (0.0127) (0.0364) (0.00873) Leverage -0.164 -0.568*** 0.0613 (0.255) (0.162) (0.210) Revenuegrowth 0.0154 -0.0124 0.0421 (0.0166) (0.0152) (0.0269) Constant 0.0515 -0.0590 -0.282 (0.219) (0.212) (0.311) Observations 189 85 104 R-squared 0.121 0.357 0.221 Number of id 45 23 22
24
These results of tables 7 and 8 imply that better risk management is does not result in higher firm performance. A noteworthy result is the different effect ERM quality seems to have on listed and non-listed firms. Whereas the ROA of listed firms is positively correlated to ERM quality, it is negatively related for non-listed firms. The dissimilar relationships in this paper might be explained by the differences in ROA between the subsamples. The descriptive statistics showed that the ROA for non-listed insurers is much higher (35%) and shows more volatility than the listed samples’ ROA. One can question whether ROA is a solid proxy to measure firm performance in this specific case. One other paper by Baxter et al. (2013) tests the same relationship for U.S. listed financial firms. The authors find that one level increase in ERM quality significantly(p<0.01) ROA with 1.14 percent on average.
6. Robustness checks
Several robustness checks are considered in order to examine how the regression estimates behave when the regression model is modified. This section will amplify only the first model which showed a statistical significant (p<0.05) effect of risk management quality on firm value. The first four columns of table 9 correspond to those of the initial model in table 6, in the remaining columns the regression specifications are modified.
First of all, a crisis dummy is included in order to determine that the year 2009 in this panel dataset is not statistically different from the other years. It is highly possibly that this year negatively influences Tobin’s Q as the financial breakdown of 2007 had a massive impact on financial stability. In order to test for this relationship a dummy variable Crisis is added to the regression model. This variable takes the value of one for observations in the year 2009, and zero otherwise. As expected, the crisis dummy negatively influences firm value in comparison to the other years (column 5). Yet, the relation is rather small and insignificant, therefore it can be concluded that the year 2009 does not statistically differ from the other years. Secondly, some papers (McShane et al., 2010; and Baxter et al., 2013) in this field of literature proxy for a firms’ systematic risk with a Beta variable6. One could expect that an insurer
with a higher beta will discount expected cash flows at a higher rate, resulting in relatively lower firm value (Shin and Stulz, 2000). After including beta to the first regression model, the ERM coefficient remains significant (p<0.05). The magnitude of the risk management quality effects on firm value slightly drops from 1.02 percent to 0.99 percent per level increase in ERM.
6 The definition, expected sign, descriptive statistics, and Pearson correlation coefficients of this variable are available
25 Table 9 – Robustness checks for model 1
(1) (2) (3) (4) (5) (6) (7) (8) VARIABLES TQ TQ TQ TQ TQ TQ TQ TQ ERM 0.0112** 0.0113** 0.00957** 0.0102** 0.0102** 0.00987** 0.00884** 0.00908* (0.00558) (0.00494) (0.00483) (0.00432) (0.00432) (0.00449) (0.00409) (0.00460) Size -0.00181 -0.00246 -0.00120 -0.00153 -0.00153 0.00104 -0.00758 -0.00507 (0.00362) (0.00275) (0.00349) (0.00291) (0.00291) (0.00601) (0.00573) (0.00457) Life -0.0106 0.0112 -0.0115 0.00901 0.00901 0.0122 0.0209** 0.00171 (0.0134) (0.0113) (0.0131) (0.0109) (0.0109) (0.0149) (0.00891) (0.00987) LogBoardsize 0.0165 0.0124 0.0152 0.00729 0.00729 0.00752 0.0309** 0.0220 (0.0132) (0.0139) (0.0130) (0.0144) (0.0144) (0.0133) (0.0146) (0.0180) Leverage 0.0190 0.0542*** 0.0204 0.0532*** 0.0532*** 0.0691** 0.0462** 0.00819 (0.0146) (0.0141) (0.0150) (0.0148) (0.0148) (0.0324) (0.0190) (0.0222) ROA 0.544* 0.569* 0.545 0.585 0.585 0.611 0.539 0.663*** (0.293) (0.303) (0.364) (0.371) (0.371) (0.406) (0.411) (0.231) RevenueGrowth -0.0239 -0.0244 -0.0223 -0.0223 -0.0223 -0.0235 -0.00396 -0.0100 (0.0195) (0.0198) (0.0207) (0.0215) (0.0215) (0.0195) (0.0197) (0.0277) Crisis -0.00695 (0.00699) Beta -0.0182 0.0107 -0.000589 (0.0305) (0.0281) (0.0127) CR 0.0150** -0.00198 (0.00699) (0.00570) Constant 0.936*** 0.918*** 0.934*** 0.923*** 0.930*** 0.875*** 0.887*** 1.002*** (0.0634) (0.0561) (0.0693) (0.0633) (0.0609) (0.121) (0.0912) (0.0543) Observations 101 101 101 101 101 97 82 82 R2-adjusted 0.1862 R2-overall 0.2760 0.5796 0.3155 0.6221 0.6221 0.5980 0.6874 Number of id 23 23 23 23 23 22 20
Time fixed effects NO YES NO YES YES YES YES NO
Country fixed
effects NO NO YES YES YES YES YES NO
Notes: This table presents the results of estimating model 1 with additional robustness checks. The model is estimated with standard errors correcting for clustering at the year and firm level. Numbers in the cells are t statistics (coefficients) with significance denoted as *, **, ***, for ten percent, five percent, and one percent, respectively.
Furthermore, some researchers include the insurers’ credit rating7 (CR) in order to control for
factors that could affect returns. These numerical long-term credit ratings are divided into seven categories similar to previous research (Ashbaugh-Skaife, Collins, and LaFond, 2006). The ERM ratings of S&P are a component of their overall credit rating, therefore this variable is also taken from the S&P RatingsDirect database. It is expected that the credit rating will absorb some of the
7 The definition, expected sign, descriptive statistics, and Pearson correlation coefficients of this variable are available
26
magnitude ERM quality has on Tobin’s Q. The results in column 7 show that credit rating is significant and positively associated to Tobin’s Q. More importantly, the ERM coefficient remains significant (p<0.05). As projected the magnitude of this relationship abated slightly; a one-level increase in risk management quality is now related to an increase in firm value of 0.88 percent.
Lastly, model 1 variables is regressed using pooled OLS (POLS), including the beta and credit rating control variables. POLS does not differentiate between firms, time or countries; it treats the panel data observations as if it is one large cross section. The results are rather similar to the other outcomes (column 8). The coefficient of interest remains statistically significant (p<0.1). Collectively, the robustness checks showed that the effect of risk management quality on Tobin’s Q remains significant. Moreover, the magnitude of this relationship stays well-nigh the same.
7.
Conclusion and Discussion
The new framework of enterprise risk management has emerged as an important solution to the increasing economic and financial complexity. The recent financial crisis showed the major consequences of shortcomings in risk management agian. Whereas traditional risk management approaches deal with risk in “silos”, ERM provides a framework that integrates all risks of all sources and consolidates policies in risk management. It coordinates risk management and avoids duplication, which results in higher efficiency. ERM is adapted by enterprises globally and various initiatives encouraged the adoption of it. Despite the increasing interest in risk management, academic evidence in this area is still limited. This is most probably due to the difficulty in developing a proper measure for risk management.
This paper investigates how the quality of enterprise risk management affects firm value and performance. In order to investigate this relationship, a sample is created which encompasses all European insurers with coverage in the S&P Ratings Direct database, in a timeframe from 2009 to 2013. A panel dataset of 45 European insurers is created. Fixed effects and random effects panel regressions is performed to control for unobservable heterogeneity across insurers. Results show that ERM quality is positively related to firm value, measured as Tobin’s Q. A one-level increase in ERM rating is associated with an increase in Tobin’s Q of approximately 1 percent. A similar research by Baxter et al. (2010) find a similar and relation for U.S. financial firms; an increase in ERM quality is associated with an increase in firm value of 3.4 percent. The market seems to appreciate the shift from the traditional risk management approach to the new
27
framework of enterprise risk management. This market view contradicts the results of McShane et al. (2010), who find that increases in ERM rating beyond level 3 do not add additional firm value. This difference in these views could be explained by a switch in market perception; it is possible that investors started recognizing the importance of ERM quality after the financial crisis. The results of this paper do not support a positive relationship between risk management quality and firm performance. Remarkably, it is found that the ROA of listed firms is not positively correlated to ERM quality, while this association is negative for non-listed companies. One other paper by Baxter et al. (2013) tests the same relationship for U.S. listed financial firms and find a significant positive relationship. The authors conclude that a one-level increase in ERM quality is associated with higher a return on asset of 1.14 percent. By decomposing the ERM ratings into different quality categories it is examined how a shift from one specific rating to another is valued by the market. The results suggest that firm value will be increasing from ERM2 to ERM4. However achieving a very strong (ERM5) risk management rating does not increase firm value significantly. The regression coefficients of all ERM categories show a pattern of market reaction to the quality of risk management; higher ratings suggest higher economic significance.
Limitations of this paper include the relatively small sample size and the use of only one proxy for enterprise risk management quality. As a result the extent to which the results may be generalized is reduced. Hence, future research should focus on additional measures for ERM quality and samples should preferably be higher. Second, this paper is limited to the European insurance industry as Standard and Poor’s provide ERM ratings on financial and insurance companies only. In order to assess the effectiveness of enterprise risk management for non-financial firms, future research on other sectors is essential. It is important to examine the effectiveness of ERM in a broader sense as it is a generally accepted framework to fight the financial complexity. Third, the proxy for risk management quality in this research is the rating by S&P. The validity of rating products by S&P, and other rating agencies, have been highly criticized during the financial downturn. Such companies are suspected of implausible ratings, therefore the ratings used in this paper might be biased too. Again, the importance of developing additional measures of ERM quality should be considered in subsequent research.
The results of this paper imply that European insurers could enjoy significant increases in firm value by upgrading their enterprise risk management quality. The market positively values enhancements in ERM beyond the statutory. It seems that these costly enterprise risk management systems indeed boost shareholders’ value. For a society at large, results seem to confirm the shift from the traditional silo-based approach to a more comprehensive one. If this
28
shift is carried on, risk management practices for corporations worldwide could be drastically affected.
29 8.
References
Banham, R., 1999, Understanding the Skepticism about Enterprise Risk Management, CFO
Magazine. April 1.
Baxter, R., J.C. Pedach, R. Hoitash, and A. Yezegel, 2013, Enterprise Risk Management Program Quality: Determinants, Value, and the Financial Crisis. Contemporary Accounting
Research, 30: 1264-1295.
Beasley, M. S., D. Pagach, and R. Warr, 2008, The Information Conveyed in Hiring
Announcements of Senior Executives Overseeing Enterprise-Wide Risk Management Processes, Journal of Accounting, Auditing, and Finance, 23: 311-332.
Beasley, M. S., R. Clune, and D. R. Hermanson, 2005, Enterprise Risk Management: An Empirical Analysis of Factors Associated with the Extent of Implementation, Journal
of Accounting and Public Policy, 24: 521-531
Belsley, D.A., E. Kuh, and R.E. Welsch, 1980, Regression Diagnostics: Identifying Influential Data and
Sources of Collinearity, John Wiley: New York.
Bhagat, S., and B. Black. 2002. The non-correlation between board independence and long-term firm performance. Journal of Corporation Law 27 (2): 231.
Bushman, R. M., and A. J. Smith. 2001. Financial accounting information and corporate governance. Journal of Accounting & Economics 32 (1-3): 237-333.
Casualty Actuarial Society, 2003, Overview of Enterprise Risk Management. Casualty Actuarial Society Website
http://www.casact.org/area/erm/overview.pdf
Cole, C. R., and K. A. McCullough, 2006, A Reexamination of the Corporate Demand for Reinsurance, Journal of Risk and Insurance, 73: 169-192.
Colquitt, L. L., and R. E. Hoyt, 1997, Determinants of Corporate Hedging Behavior: Evidence Fromthe Life Insurance Industry, Journal of Risk and Insurance, 64: 649-671. Committee of Sponsoring Organizations of the Treadway Commission (COSO), 1992, COSO
Report: Internal Control – an Integrated Framework. New York: AICPA.
Committee of Sponsoring Organizations of the Treadway Commission (COSO). 2004. COSO
Report: Enterprise Risk management – integrated framework. New York: AICPA.
Committee of Sponsoring Organizations of the Treadway Commission website http://www.coso.org/documents/coso_erm_executivesummary.pdf
Cumming, C. M., and B. J. Hirtle, 2001, The Challenges of Risk Management in Diversified Financial Companies, FRBNY Economic Policy Review, 7: 1-17.
30
Korea 12.1: 207-228.
DeMarzo, P.M., and D. Duffie, 1995, Corporate incentives for hedging and hedge accounting,
The Review of Financial Studies 8, 743-771.
Froot, K. A., D. S. Scharfstein, and J. C. Stein, 1993, Risk management: Coordinating corporate investment and financing policies, The Journal of Finance 48 (5): 1629–58.
George A., and J.P. Weston, 2001, The use of foreign currency derivatives and firm value, The
Review of Financial Studies 14, 243-276.
Graham, J., and D. Rogers, 2002, Do firms hedge in response to tax incentives? Journal of Finance
57, 815–839.
Guay, W. and S. P. Kothari, 2003, How Much Do Firms Hedge With Derivatives?, Journal of
Financial Economics, 70: 423-461.
Hausman, J. A. (1978). Specification tests in econometrics. Econometrica: Journal of the Econometric
Society, 1251-1271.
Hoyt, R. E., and A. P. Liebenberg. 2011. The value of enterprise risk management. Journal of Risk and Insurance 78 (4): 795–822.
Hoyt, R. E., and H. Khang, 2000, On the Demand for Corporate Property Insurance, Journal of
Risk and Insurance, 67: 91-107.
Hoyt, R. E., B. M. Merkley, and K. Thiessen, 2001, A Composite Sketch of a Chief Risk
Officer, The Conference Board of Canada, Toronto, September.
Jensen, M. C., 1986, Agency Costs of Free Cash Flow, Corporate Finance and Takeover, American
Economic Review, 76: 323-329.
Jorion, P., 2009, Financial risk manager handbook (Vol. 116). John Wiley & Sons.
Kleffner, A. E., R. B. Lee, and B. McGannon, 2003, The Effect of Corporate Governance on the Use of Enterprise Risk Management: Evidence from Canada, Risk Management
and Insurance Review, 6: 53-73.
Lam, J., 2001, The CRO Is Here to Stay, Risk Management, 48: 16-20.
Lang, L. and R. Stulz, 1994, Tobin’s Q, Diversification, and Firm Performance, Journal of Political
Economy, 102: 1248-1280.
Liebenberg, A.P., and D.W. Sommer, 2008, Effects of corporate diversification: Evidence from the property-liability insurance industry, Journal of Risk and Insurance 75, 893-919.
Liebenberg, A.P., and R. E. Hoyt, 2003, Determinants of Enterprise Risk Management: Evidence From the Appointment of Chief Risk Officers, Risk Management and Insurance Review, 6: 37
52.
