Enterprise Risk Management Program Quality:

Determinants, Value Relevance, and the Financial Crisis

RYAN BAXTER, Bentley University
JEAN C. BEDARD, Bentley University and the University of New South Wales
RANI HOITASH, Bentley University
ARI YEZEGEL, Bentley University

Abstract
This paper investigates factors associated with high quality Enterprise Risk Management (ERM)
programs in financial services firms, and whether ERM quality enhances performance and
signals credibility to the financial markets. ERM, developed with the assistance of the
accounting profession, provides a framework and plan to integrate management of all sources of
risk. Challenged by measurement difficulties common to research on management control
systems, prior ERM studies present mixed findings. Using ERM quality (ERMQ) ratings of
financial companies by Standard & Poor’s, we find that higher ERMQ is associated with greater
complexity, less resource constraint, and better corporate governance. Controlling for such
characteristics, we find that higher ERMQ is associated with improved accounting performance.
Results show a market reaction to signals of enhanced management control from initial ERMQ
ratings and rating revisions, and a stronger response to earnings surprises for firms with higher
ERMQ. Focusing on the recent global financial crisis, our analysis suggests that there is no
relation between ERM quality and market performance prior to and during the market collapse.
However, returns of higher ERMQ companies are higher during the market rebound. Overall,
results reveal that firm performance and value are enhanced by high quality controls that
integrate risk management efforts across the firm, enabling better oversight of managers’ risk-
taking behavior, and aligning that behavior with the strategic direction of the company.
Keywords: Enterprise Risk Management; Firm value; Earnings Response Coefficient; Global
financial crisis.
JEL descriptors: G32; G21; G22; M41; M42

* We are grateful for comments and suggestions by Dorothy Feldmann, Steve Fortin, Dana Hermanson, Ronny
Hofmann, Udi Hoitash, Doug Prawitt, Zvi Singer, Monte Swain, Desmond Tsang, Ann Vanstraelen, Jeff Wilks,
David Wood, Bill Zhang, Mark Zimbelman, the anonymous reviewers of this Journal and seminar participants at
Bentley University, Brigham Young University, Maastricht University, McGill University, the University of New
South Wales, and Virginia Tech. We also thank the PricewaterhouseCoopers INQuires program and Bentley
University for research grant support, and Bentley graduate students Lindsay Bove, Dana Bogen, Tien-Shih Hsieh,
and Maximino Rivera for their excellent research assistance.

Electronic copy available at: http://ssrn.com/abstract=1684807

Enterprise Risk Management Program Quality: Determinants, Value Relevance, and the
Financial Crisis

1. Introduction

Enterprise Risk Management (ERM) programs are designed to integrate management of

risks from a wide variety of sources. ERM improves risk management by promoting awareness

of all sources of risk, and by aligning strategic and operational decision-making across the entity

with the company’s risk appetite (e.g., COSO 2004; Nocco and Stulz 2006). As such, ERM is a

corporate governance mechanism that constrains and coordinates managers’ behavior. While

potential benefits to firm performance and value (e.g., through improving efficiency and

reducing volatility) have been espoused (Beasley, Pagach, and Warr 2008; Hoyt and Liebenberg

2011) there is little available archival evidence on these benefits. Our investigation concentrates

on two principal questions. First, what factors are associated with high quality ERM programs as

measured by ratings produced by Standard & Poor’s (S&P)? Second, are the anticipated benefits

of high quality ERM for performance realized, and do the markets recognize those benefits by

rewarding program quality and valuing the reduction in uncertainty that better risk management

brings?

The criteria used by S&P to rate ERM quality (see Online Appendix 1) relate to effective

communication of strategy throughout the enterprise, appropriate project selection, improvement

of accounting-based returns and risk-adjusted market returns. ERM development involved the

accounting profession through the Committee of Sponsoring Organizations (COSO 2004). ERM

embeds the internal control framework and principles (COSO 1992) beyond the financial

reporting system, more broadly and systematically throughout organizations. 1 Prior research

1
ERM has been a primary focus of COSO since the early 2000s, largely attributable to its Enterprise Risk Management – Integrated Framework
(COSO 2004), which helped promote ERM particularly among U.S. companies.

Electronic copy available at: http://ssrn.com/abstract=1684807

addresses determinants and market implications of corporate governance choices such as internal

control effectiveness (e.g., Hammersley, Myers, and Shakespeare 2008; Ashbaugh-Skaife,

Collins, Kinney, and LaFond 2008) and expertise of personnel in governance (e.g., DeFond,

Hann, and Hu 2005). These issues have not been examined with respect to ERM quality.

This study contributes to the corporate governance literature by examining ERM, which

is among the ―internal‖ mechanisms that companies can adopt to reduce the risk that managers

undertake activities not well aligned with the overall firm strategy. Prior research investigates

how corporate control mechanisms affect allocation and utilization of economic resources

(Bushman and Smith 2001). For instance, high quality ERM may affect allocation of resources

through market participants’ perceptions of the reliability and persistence of accounting earnings.

However, Zimmerman (2001) notes that a key constraint on empirical research on management

control systems (MCS) is the lack of information on what corporations do internally. Concerns

over the quality of publicly available proxies for corporate governance quality are also expressed

by Larcker, Richardson, and Tuna (2007). Further, Davila and Foster (2005, 2007) and Ittner and

Larcker (2001) both note the difficulties of using manager perceptions as indicators of MCS

quality. S&P assessments of communication, action and evaluation processes within companies,

in forming their ERM ratings, address these measurement problems an independent, professional

view, with access to nonpublic information.

This study addresses two basic research questions using a sample of 165 firm-year

observations in the banking and insurance industries with S&P ERM ratings in 2006-08. We first

investigate company characteristics associated with variation in ERM quality, as measured by

S&P ratings. Our use of S&P ratings contributes to prior archival research that concentrates on

the existence of ERM programs (e.g., Liebenberg and Hoyt 2003; Hoyt and Liebenberg 2011),

2
uses survey data to investigate ERM program quality (e.g., Kleffner, Lee, and McGannon 2003;

Beasley, Clune, and Hermanson 2005; Beasley et al. 2008), or infers quality from publicly

available data (e.g., greater than expected sales; Gordon, Loeb, and Tseng 2009).2

We develop explanatory constructs using theories from the ERM and MCS literatures.

Consistent with ERM’s focus on integrating risk management across disparate parts of the

enterprise, we find that larger and more diversified entities have higher quality programs. We

also consider effects of volatility and risk of financial distress. If effective ERM helps protect

against lower tail outcomes, riskier companies may demand better programs, and yet those firms

may lack the personnel and systems resources to implement them. Our results support the latter

explanation: higher-risk companies have lower quality ERM, likely due to resource constraints

inhibiting the investment necessary for effective ERM.3 Lastly, we find higher quality ERM is

associated with better corporate governance (i.e., audit committees charged with direct oversight

of risk), less audit-related risk (i.e., stable auditor relationships and effective internal controls),

presence of risk officers/committees, and boards with longer tenure.

We address our second research question through tests investigating the association of

ERMQ with performance and market response. Limited prior research finds ERM quality

associated with firm value as measured by Tobin’s Q: Hoyt and Liebenberg (2011), who identify

ERM programs through Lexis-Nexis and SEC filings, and McShane, Nair, and Rustambekov

(2011) who use S&P ratings among insurance companies. But how does this greater valuation

occur? Are higher accounting returns achieved through better risk management? Does the market

anticipate superior future performance when ratings become available? Does the market have

2
Our sample is limited to the banking and insurance industries, which S&P has chosen to include in their ERM quality review. As noted later in
the paper, results from highly regulated industries might not apply to other sectors.
3
While we include a series of lagged control variables to rule out the alternative explanation that poor ERM results in higher risk, we cannot
entirely rule out this possibility.

3
greater confidence in the earnings reliability and persistence of high ERMQ companies, due to

reduction of uncertainty? We analyze each of these possible outcomes.

Results show that ERM quality is positively associated with operating performance. We

find significant market reaction to initial ratings disclosures preceding the announcement

(implying leakage of the ratings), and to ERM rating changes across various event windows. We

further find that ERM quality is positively associated with earnings response coefficients,

implying that the market places greater value on unexpected earnings of those firms. This builds

on prior research that investors’ perceptions of the credibility/persistence of earnings are affected

by governance factors such as auditor size (Teoh and Wong 1993), auditor industry

specialization (Balsam, Krishnan, and Yang 2003), and ineffective internal controls (Beneish,

Billings, and Hodder 2008).

Lastly, reflecting the purpose of ERM to add value by reducing costs from lower-tail

outcomes (Pagach and Warr 2008; Nocco and Stulz 2006), we examine the relation between

ERM quality and market returns around the Global Financial Crisis (GFC) of 2008-09. While we

find no relation prior to and during the market collapse, we find a positive association of ERM

quality with market performance during the market rebound. This suggests that following the

flight from financial services companies during the crisis period, information contained in ERM

quality ratings was used by investors to identify companies more likely to rebound.

The remainder of the paper is organized as follows. Section 2 reviews prior ERM

research and the empirical accounting literature on corporate governance and management

controls. Sections 3 and 4 discuss the empirical methods and results of our tests and Section 5

presents our conclusions and their implications for research.

2. Prior Research and Hypothesis Development

4
Factors Associated with Enterprise Risk Management Quality

The COSO (2004) ERM framework defines 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
the enterprise, designed to identify potential events that may affect the entity, and
manage risk to be within its risk appetite, to provide reasonable assurance
regarding the achievement of the entity’s objectives.‖

The literature on ERM comprises two basic streams, reflecting research on MCS in

general. One explores drivers and/or impediments to ERM program adoption, while the other

asks whether the presence of (or less frequently, the quality of) ERM provides information to

financial markets. Research on company characteristics associated with the presence or level of

implementation of ERM programs, considered in this sub-section, mirrors studies on

implementation of advanced MCS (Elbashir, Collier, and Sutton 2011), and on determinants of

effectiveness of internal auditing and control over financial reporting (e.g., Prawitt, Smith, and

Wood 2009; Ashbaugh-Skaife, Collins, and Kinney 2007; Doyle, Ge, and McVay 2007). Related

ERM studies measure adoption/implementation through surveys of company personnel, or infer

the status of ERM programs from publicly available information. Ittner and Larcker (2001) and

Larcker et al. (2007) discuss the difficulties associated with each source of information in the

MCS literature. Surveys of company personnel provide a valuable inside view of the firm, but

could be biased. Publicly available data are poor proxies for corporate activities, a key limitation

when program quality (not adoption) is the object of study. We address these limitations through

use of S&P’s evaluations of ERM quality.

In this discussion, we place studies examining company characteristics associated with

ERM adoption/implementation in the context of prior accounting research that investigates

characteristics associated with other internal corporate governance mechanisms involving

5
management control. First, since ERM is a risk management tool, prior research has investigated

the association of ERM adoption/implementation with two basic sources of risk. Several studies

find that more complex entities are more likely to adopt ERM (Kleffner et al. 2003; Beasley et

al. 2005; Hoyt and Liebenberg 2011). As Liebenberg and Hoyt (2003) note, such companies face

a broader scope of risks, and lack of coordination in response will lead to greater risk in the

aggregate. Thus, large and diversified companies have a greater incentive to move from siloed to

integrated risk management. Likewise, the literature on MCS (e.g., Elbashir et al. 2011; Davila

and Foster 2005) implies that complexity should influence development of advanced controls in

general, as larger and more diversified companies have greater information processing needs.

Second, research on determinants of ERM adoption/implementation also recognizes

financial risk as a source of demand for ERM. Studies have considered both uncertainty (e.g.,

volatility in earnings or stock prices) and financial distress indicators as financial risk indicators,

with weak results. Liebenberg and Hoyt (2003) and Gordon et al. (2009) do not find significance

on measures of uncertainty in CRO appointment and ERM implementation. Pagach and Warr

(2007) find a significant, positive association between leverage and CRO appointment, but

Gordon et al. (2009) do not. This variability in findings may be due to the dual character of

financial risk measures. On one hand, companies with higher financial risk face higher likelihood

of lower-tail outcomes such as bankruptcy (Pagach and Warr 2007), and thus have more to gain

from programs aimed at reducing that likelihood. On the other hand, higher financial risk is

associated with fewer resources to devote to implementing advanced control systems (Davila and

Foster 2005; Elbashir et al. 2011). In the specific context of ERM, Liebenberg and Hoyt (2003)

note that ERM development is often stalled by the lack of technological tools necessary for

6
implementation. Thus, the direction of the effect of financial risk indicators on ERM quality is

unclear ex ante.

Third, the COSO (2004) framework suggests that risk management philosophy is

important in implementation of ERM programs. This implies that high quality ERM programs

should be associated with other indicators of effective corporate governance (e.g., Kleffner et al.

2003; Liebenberg and Hoyt 2003). Beasley, Branson, and Hancock (2009) report that

expectations of the board of directors are a key driver of ERM, due to increased demand for

greater risk transparency. Also, boards of companies with superior governance structures should

recognize the need for other mechanisms to control agency conflicts. Prior research shows that

ERM adoption is associated with greater board independence (Beasley et al. 2005), appointment

of Chief Risk Officers (CROs), and use of Big 4 auditors (Kleffner et al. 2003; Beasley et al.

2009). While we examine these governance mechanisms as determinants of ERM quality, we

also include corporate governance mechanisms specific to the auditing literature not addressed

by prior ERM research. These include internal control effectiveness, stability of audit

relationships, and presence of expert audit committee members.

In sum, prior research considers the roles of risk and corporate governance in

implementing ERM programs and MCS in general, but does not investigate these factors in

implementing high quality ERM programs. Our first research question is:

RESEARCH QUESTION 1. Is ERM program quality associated with company complexity,

financial risk/resources, and corporate governance?

The Association of ERM Quality with Performance and Market Response

Accounting Performance and Market Valuation

Our second set of tests examines outcomes associated with variation in ERM quality.

ERM programs are intended to add value by reducing costs from lower-tail outcomes (Beasley et

7
al. 2008; Nocco and Stulz 2006). By reducing the likelihood and impact of extreme, negative

financial events, firms avoid direct costs such as losses and bankruptcy, and indirect costs, such

as reputational effects with customers and suppliers (Pagach and Warr 2008). Yet, high quality

ERM programs should not only reduce the impact of negative events, but also help identify

untapped opportunities across the enterprise (COSO 2004). If so, capital should be allocated

more efficiently and returns increased (Hoyt and Liebenberg 2011; McShane et al. 2011). For

both reasons, ERM quality should improve performance and be valued in the financial markets.

Following implementation, ERM quality could enhance financial performance through

ability to both recognize opportunities on the upside and protect against lower-tail outcomes on

the downside, as suggested by the COSO definition cited above. If so, then companies with

better ERM quality should have higher accounting returns, but prior research has not examined

this association. Hypothesis 1(a) proposes a positive association of ERMQ and accounting

returns.

HYPOTHESIS 1(a). ERM program quality is positively associated with accounting

performance.
ERM quality could also enhance market value, but empirical evidence is very limited.

Hoyt and Liebenberg (2011) find that Tobin’s Q is positively associated with ERM program

existence, and McShane et al. (2011) find a positive association with ERMQ among insurance

firms. Hoyt and Liebenberg (2011) control for the possibility that this association might be

driven by endogeneity due to the non-random nature of ERM program quality, McShane et al.

(2011) do not control for endogeneity in their models. We extend prior research by first testing

Hypothesis 1(b) regarding the association of ERMQ with Tobin’s Q in our broader sample. We

then determine whether this result still holds after controlling for endogeneity.

HYPOTHESIS 1(b). ERM program quality is positively associated with market valuation.

8
Market Reactions to ERM Quality Ratings Disclosures and Revisions

If the market does value ERM quality, as predicted in Hypothesis 1(b), the question arises

as to whether market participants recognize the information content of the quality ratings at

announcement. Theory postulates that returns are a function of changes in market participants’

information sets. If the market perceives ex ante that higher ERM quality will increase firm

value, public announcements of ERM ratings should be associated with a market response.

Beasley et al. (2008) examine this issue indirectly, using CRO hiring announcements from 1992-

2003 as a proxy for ERM program adoption.4 They do not find an overall market reaction, but do

show a response contingent on industry and size. Beasley et al. (2008) suggest that lack of

overall response in the financial services industry may be due to regulatory pressure (i.e., those

firms ―may already have begun engaging in ERM before the CRO appointment‖). If so, the

market might react more strongly to information on ERM quality, which enables better

understanding of the level of commitment to ERM. Hypotheses 2(a) and 2(b) propose a market

response to variation in disclosed ERM quality ratings and ratings revisions, respectively:

HYPOTHESIS 2(a). Market reactions will be positively associated with the level of ERM
quality rating.
HYPOTHESIS 2(b). Market reactions will be positively associated with ERM quality rating
revisions.
ERM Quality and the Earnings Response Coefficient

Knowledge of relative ERM quality could assist investors by providing information that

helps assess the usefulness of financial reports in predicting future cash flows. Prior research

finds that investors’ responses to earnings surprises, measured using the earnings response

coefficient (ERC), varies by company characteristics providing information on financial

reporting quality. For instance, Teoh and Wong (1993) find stronger response to earnings
4
CRO announcements are an indirect proxy for ERM programs, because while CROs may lead ERM programs, other corporate officials may do
so. Further, ERM implementation and quality improvements may lag CRO appointment.

9
surprises among companies with large audit firms, and Balsam et al. (2003) find a similar result

among clients of industry specialist auditors. Anderson and Yohn (2002) find weaker response to

earnings surprises following restatements, and Francis, Schipper, and Vincent (2002) show that

the increase in earnings response over two decades is largely due to better disclosure of the

properties of earnings, implying that nonfinancial disclosures are used to assess earnings quality.

In the ERM context, firms with stronger programs have a more organized and structured

approach for capturing and managing risk across the entire organization (COSO 2004). Thus,

firms with superior ERM programs should have more persistent earnings, leading to greater

earnings credibility and thus a stronger market response to unexpected earnings. We test the

following hypothesis:

HYPOTHESIS 3. Earnings response coefficients will be positively associated with ERM

quality.

ERM and the Global Financial Crisis

One primary objective of ERM programs is to help companies avoid catastrophic loss in

profitability and market value. If so, the advantage of an effective risk management program may

be more pronounced at times of crises, when risk is of greater concern. The recent GFC of 2008-

09 provides a natural laboratory in which to examine market effects of ERM quality. While high

quality ERM programs may have protected companies during the crisis, given the depth of the

market crash, many investors may have divested their financial industry holdings regardless of

firm-specific factors. Further, as the market began to rebound, investors who desired to return to

financial firms might have considered evidence of effective risk management in deciding where

to initially invest. We propose the following research question:

RESEARCH QUESTION 2. Did firms with stronger ERM programs have better market
returns prior to, during, and/or after the global financial crisis?

10
3. Method

Sample Development

Our sample comprises financial services firms (banks and insurance companies) with

coverage in the S&P Ratings Direct database.5 From the 404 ERM disclosures during the period

2006-08 in the S&P database, we remove 157 observations without clear or matching company

identifiers, resulting in 247 firm-years. We eliminate 27 companies with missing Compustat

data, and an additional 28 companies with missing internal control disclosures in Audit Analytics.

We manually search Audit Analytics and proxy filings for data on company executives and

corporate governance, omitting 17 additional observations with missing data. Finally, we exclude

real estate investment trusts, shares of beneficial interest and depository units. Our final sample

contains 165 firm-year observations.6

ERM Quality

Our test variable is the S&P’s ERM quality rating, provided as part of credit rating

analysis for some companies since 2006. S&P analysts evaluate companies’ ERM programs and

classify program quality into four categories: Weak, Adequate, Strong, or Excellent. The ERM

rating ―includes the assessment of Risk Management Culture, Risk Controls, Emerging Risk

Management, Risk and Capital Models, and Strategic Risk Management‖ (Santori, Puccia,

Osborne, Jones, and Wright 2005, 4).7 Because the majority of ERM programs are categorized as

―Adequate‖, we developed a finer quality scale by dividing ―Adequate‖ ratings into three

5
At this point, coverage is only available in these industries. Using all available data, our final sample is composed of 28 observations in SIC 60
(Depository Institutions), 6 in 61 (Non-depository Credit Institutions), 11 in 62 (Security and Commodity Brokers, Dealers, Exchanges, and
Services), 119 in 63 (Insurance Carriers), and 1 in 67 (Holding and Other Investment Offices).
6
Depending on data availability, sample size varies slightly across analyses.
7
For insurance companies, S&P’s framework for assessing market, credit, operational, and insurance risks includes risk management culture,
emerging risk management, risk and economic capital models, strategic risk management, and general risk controls considerations. ERM
measurement criteria for financial institutions ―build on Standard & Poor’s Policies, Infrastructure, and Methodology (PIM) framework that was
constructed to assess the ERM practices of the trading operations‖ (Samanta, Barnes, and Puccia 2006, 2). Online Appendix 1 contains ERM
rating criteria for insurers. Criteria for financial institutions are similar, except ―emerging risk management‖ is replaced by ―extreme-event
management.‖

11
subcategories based on content of the analyst’s report. If that narrative includes more positive

than negative descriptions, we assigned a rating of ―Strong-adequate.‖ For example, one

description notes that a firm ―has an adequate enterprise risk management (ERM) process, with

many aspects viewed as strong‖ (Gaskel, Maher, and Wong 2007). Where descriptions note a

more negative tone, we assigned a rating of ―Weak-adequate.‖ For example, one report notes that

―the development of its ERM lags behind that of similarly rated peers.‖ Companies with

balanced narratives remain in the ―Adequate‖ category. 8 This process yields an ERM quality

scale ranging from one to six. 9 The numbers of observations in the resulting six levels are:

weak=1 (n = 6), weak-adequate=2 (n = 11), adequate=3 (n = 84), strong-adequate=4 (n = 22),

strong=5 (n = 35), excellent=6 (n = 7). Table 1 provides definitions for all variables used in the

models.

[Insert Table 1]

Factors Associated with ERM Quality

Our first research question asks whether company risk and corporate governance quality

are associated with ERM quality among firms rated by S&P. First, we consider company risk in

terms of complexity, predicted to be a driver of investment in ERM due to need to integrate risk

management across multiple and/or diverse operations. Because prior research finds that ERM

adoption is positively associated with complexity (Kleffner et al. 2003, Beasley et al. 2005; Hoyt

and Liebenberg 2011), we expect that more complex companies will build higher quality ERM

programs. To measure complexity, we include company size as MARKETCAP, SEGMENTS, an

indicator for GLOBAL companies, and an indicator for companies with FOREIGN operations.
8
Two of the authors read the section of each S&P credit rating report with an ERM program categorized as ―Adequate.‖ The coders agreed on
108 of the 117 reports (92.3 percent, Kappa = 0.834). For the remaining nine reports, the authors jointly reviewed the text, and agreed on its
classification.
9
We discuss the sensitivity of our results to other specifications in the following sections. S&P 2010 has also recently discussed the need for
more detailed description of firms initially rated as adequate. Accordingly, in future rankings they will consider two additional categories in the
―adequate‖ range: ―adequate with strong risk controls‖ and ―adequate with positive trend.‖

12
 Second, because ERM is intended to manage risk from all sources, companies with

greater financial risk might be more likely to have high quality programs. However, given threats

to solvency, such companies may lack the ability to invest in high quality ERM, making

direction difficult to predict ex ante. Following prior accounting research, we measure risk as

uncertainty and financial distress. We measure uncertainty using the standard deviation of stock

returns, STD-RET (e.g. Sengupta 1998). We also include the standard deviation of cash

generated from operating activities, STD-OP-CASH. Cash-flow volatility can increase the need

for external financing which necessitates better risk management (Froot, Scharfstein, and Stein

1993). Because companies with more frequent losses face financial distress, we control for

LOSS-PROPORTION with the proportion of loss years over the past five years. Further, because

ERM evaluations are done as part of the credit assessment, we control for credit risk with

CREDIT-RATING. Similar to Ashbaugh-Skaife, Collins, and LaFond (2006), we use seven

categories from the long-term issuer credit ratings, ranging from a high of AAA to a low of D.10

We also include cash flow from operations deflated by total assets (OP-CASH) and LEVERAGE.

Because Liebenberg and Hoyt (2003) observe that financial distress is associated with ERM

implementation, we control for bankruptcy risk using a measure of Z-SCORE structured for

financial firms (Roy 1952).11

The remaining construct likely associated with ERMQ is the quality of company

governance. Companies can address risk by appointing a CRO and/or by assigning

responsibilities for risk management to board committees. Similar to Hoyt and Liebenberg

(2011), we collect data on the presence of CROs and other risk executives by searching Lexis-

10
Following are S&P ratings with their numerical equivalents: AAA (7), AA+, AA and AA- (6), A+, A, and A- (5), BBB+, BBB, and BBB- (4),
BB+, BB, BB- (3), B+, B, and B- (2), CCC+, CCC, CC, C, D or SD (1).
11
We also considered market beta, another measure of company risk, for use in the determinants model, but because it results in a small loss of
sample in this model, we do not include it here. When we include beta in the determinants model it is insignificant and other results are
unchanged.

13
Nexis and ABI-Inform for news reports and announcements. To identify risk committees, we

manually search Audit Analytics and company financial reports for the presence of a board

committee responsible for risk. 12 Combining this information, we define RISK-STRUCTURE,

which equals one in the presence of either a risk officer or a risk committee; zero otherwise.

The Conference Board (Brancato, Tonello, Hexter, and Newman 2006) suggests that

instead of creating a separate risk committee, companies might opt to assign this responsibility to

the audit committee (AC). We measure direct AC oversight of risk by manually collecting AC

charters from SEC filings, classifying AC-RISK-OVERSIGHT as one if the charter mentions that

the AC is responsible to oversee risk (i.e., not merely to discuss risk). Next, we also employ

common proxies for AC strength in the form of financial expertise and size. Past research finds

that expertise on the AC is associated with financial reporting quality (Abbott, Parker, and Peters

2004; Hoitash, Hoitash, and Bedard 2009). We measure accounting expertise with PAFE and

supervisory expertise with PSFE expecting that both will be associated with better ERM quality.

We also control for the size of the AC with ACSIZE, based on the argument that larger AC are

better equipped to handle diverse responsibilities and will have greater voting power within the

board.

We also study the association of board and management characteristics with ERM

program ratings. Prior research indicates a positive link between overall governance measures

and ERM adoption (Beasley et al. 2009; Beasley et al. 2005; COSO 2004; Kleffner et al. 2003),

as the board uses information produced by ERM programs to assess and address significant risks.

We control for board independence with BOARD-IND. Past research finds that more independent

boards are associated with better financial reporting quality (e.g., Klein 2002; Krishnan 2005),

12
As key words in this search, we use common risk committee titles, including: risk, risk and compliance, risk management, finance and risk
management, among others.

14
and with better ERM implementation (Beasley et al. 2005). We also include board tenure, as

longer tenure is associated with higher financial reporting quality (Beasley 1996; Bédard,

Chtourou, and Courteau 2004), and expect a positive association with ERMQ. The success of

ERM programs could also depend on top management commitment and willingness to be

monitored in that manner. Fama and Jensen (1983) suggest that separating the CEO and board

chair positions contributes to reducing the agency conflict. Accordingly, we include CEO-

DUALITY and expect an inverse association between CEO duality and ERMQ.

In addition, we include a measure of governance based on the external audit. Well-

governed companies should have both effective internal controls and a stable relationship with

their audit firm (Schwartz and Menon 1985). We define AUDIT-RELATED-RISK as an indicator

variable that considers both reported internal control weaknesses and auditor switches. We

predict a negative association between this variable and ERMQ.13 We further control for FIRM-

AGE, with no directional expectation. We also include NYSE, an indicator variable for firms

listed on the New York Stock Exchange, and expect a positive coefficient because NYSE

regulation could be associated with early adoption of ERM and greater time to facilitate an

increase in ERM program quality. All variables (with the exception of year and industry

dummies) are measured with one-year lags to ensure that the values correspond to information

available to S&P analysts at the time of their ratings.

Model 1, presented below, also includes year and industry fixed effects and is estimated

with standard errors corrected for clustering at the firm and year levels.

13
Because over 97 percent of our sample companies are audited by a Big 4 firm, we do not include auditor size as a control variable in our
models. Including a Big 4 indicator variable does not alter our results.

15
ERMQ= α +β1 MARKETCAPt-1+ β2SEGMENTS t-1 + β3 GLOBAL t-1+ β4FOREIGN t-1 (1)
+ β5 STD-RET t-1 +β6 STD-OP-CASH t-1 +β7 LOSS-PROPORTION t-1
+β8CREDIT-RATING t-1 + β9OP-CASH t-1+ β10LEVERAGE t-1+ β11 Z-SCORE
t-1 + β12RISK-STRUCTUREt-1+ β13AC-RISK-OVERSIGHTt-1+ β14PAFEt-1
+β15PSFEt-1+β16ACSIZEt-1+ β17 BOARDIND t-1+ β18BOARD-TENURE t-1
+ β19CEO-DUALITY t-1 +β20AUDIT-RELATED-RISK t-1 + β21FIRM-AGE t-1
+ β22NYSE t-1 + β23-24 YEAR_DUMMIES + β25-28 INDUSTRY_DUMMIES + e

ERM Quality, Performance and Market Response

Accounting Performance and Market Valuation

Hypothesis 1(a) predicts a positive association of ERM quality with accounting

performance. We measure accounting performance using ROA, the ratio of income before

extraordinary items divided by total assets. The ROA model controls for variables found to

explain differences in firm performance and value, including board size (Yermack 1996), board

independence (Rosenstein and Wyatt 1990), firm size and number of segments (Berger and Ofek

1995), operating profitability (Yermack 1996), institutional ownership (McConnell and Servaes

1990), leverage (Anderson and Reeb 2003), growth opportunities (SALES-GROWTH; Bhagat

and Black 2002) and CAPITAL-OVER-SALES (Yermack 1996), and the standard deviation of

ROA (Liebenberg and Sommer 2008). The regression specification is presented in Model 2

below.

ROA = α + β1ERMQ + β2BOARDSIZE + β3BOARDIND + β4 LOGASSETS + β5INST-OWN (2)

+ β6 SEGMENTS+ β7LEVERAGE + β8 SALES-GROWTH+ β9CAPITAL-OVER-
SALES + β10CREDIT-RATING + β11STDROA+ β12-13YEAR_DUMMIES
+ β14-17INDUSTRY_DUMMIES + e

Hypothesis 1(b) predicts a positive association of ERMQ with Tobin’s Q (the book value

of assets minus the book value of equity plus the market value of equity divided by the book

value of assets). We employ the same control variables as in the ROA analysis, adding ROA and

removing STDROA from the independent variables. Model 3 is as follows:

16
Q = α + β1ERMQ + β2BOARDSIZE + β3 BOARDIND + β4LOGASSETS + β5ROA (3)
+ β6INST-OWN β7SEGMENTS + β8LEVERAGE + β9SALES-GROWTH
+ β10CAPITAL-OVER-SALES + β11CREDIT-RATING + β12-13 YEAR_DUMMIES
+ β14-17INDUSTRY_DUMMIES + e

Market Reactions to ERM Quality Ratings Disclosures and Revisions

Hypothesis 2 proposes that equity market participants will react to information contained

in announcements of ERM quality ratings and revisions of those ratings. Using a standard event

study methodology, we examine the market reaction to the initial ERM S&P rating disclosures

and subsequent rating revisions.14 We define the event day (t) as the date that S&P publishes the

credit rating report containing the ERM quality rating. We compute abnormal return on day t as

ARjt=Rjt-Rrt, where Rjt is the return on security j on day t and Rrt is the return on the reference

portfolio on day t. Due to possible selection bias in the set of companies rated by S&P, we

primarily rely on a reference portfolio based on the value-weighted daily returns of all securities

in our sample to compute abnormal returns. However, we also test sensitivity to industry-, size-,

and market-based reference portfolios. Returns on the industry index are the value-weighted

returns of all securities within the relevant two-digit SIC codes. Size and market returns are

obtained from the CRSP index files. We expect a positive (negative) market reaction to ERMQ

disclosures in the strong and excellent (weak and adequate) ERM ratings. Similarly, an upgrade

(downgrade) in the ERM rating should be associated with a positive (negative) market reaction.15

To align the market reaction to positive and negative news, we multiply the returns in response

to low ratings and downgrades by negative one.

14
In order to ensure that the event study analysis is not influenced by contemporaneous credit rating changes, we remove ERM announcements
and revisions that take place within a ten day window surrounding credit rating changes. Our results are similar when we do not exclude these
observations.
15
Due to the limited number of ERM rating revisions, we collected additional S&P revision reports on sample companies that were published
after the end-date of our sample (until the end of the year 2009).

17
 Finally, following Beasley et al. (2008), we compute cumulative abnormal returns based

on the event windows (-2, +2), (-2, 0) and (0, +2) as follows: CAAR (T1 , T2 ) j  t 2 T AAR jt , where
T

AARjt is the abnormal return on security j on day t as defined above. T1 is the return accumulation

start day relative to the initial announcement or revision and T2 is the end day. We test for

significance based on parametric (Patell 1976) and nonparametric (Corrado 1989) tests.

ERM Quality and the Earnings Response Coefficient

Hypothesis 3 predicts that the security price reaction to earnings surprises will be greater

for firms with stronger ERM quality, as firms with superior ERM presumably have greater

credibility, less uncertainty and lower expected earnings volatility. We test this hypothesis by

estimating Model 4 using ordinary least squares (OLS) regression with firm and year clustered

standard errors:

CAR= α + β1UE + β2UE×STRONG_ERM + β3UE×CREDIT-RATING + β4NEG (4)

+ β5UE×NEG + β6UE×B/M + β7UE×BETA + β8UE×COV
+ β9-12INDUSTRY_DUMMIES + β13-29FISCAL_QUARTER_DUMMIES + e.

To measure the extent to which share prices respond to earnings surprises, we regress

cumulative abnormal returns for the three-day period centered on earning announcement dates on

unexpected earnings. We primarily rely on sample-adjusted (CAR_SAMP) abnormal returns, but

test sensitivity using industry- (CAR_IND), market- (CAR_MAR) and size-adjusted (CAR_SIZE)

abnormal returns. We measure unexpected earnings (UE) as the difference between reported

earnings and analysts’ earnings expectations scaled by price at the end of the previous fiscal

quarter. STRONG_ERM is an indicator variable that equals one for firms with ERMQ>3. Our

model follows prior ERC research in using the interaction of UE and STRONG_ERM as the test

variable for Hypothesis 3; we expect a positive sign. The model controls for variation in earnings

response due to other factors, including CREDIT-RATING. Following Hackenbrack and Hogan

18
(2002), we include an indicator for firms reporting negative income before extraordinary items

(NEG) and its interaction with UE. Consistent with the prior literature (Teoh and Wong 1993) we

control for growth (B/M), market risk (BETA) and analyst coverage (COV) by including the

interaction of these variables with UE. Finally, we include industry and fiscal quarter fixed

effects to control for variation in earnings announcement returns associated with industry and

period characteristics of the sample.

ERM and the Global Financial Crisis

Research Question 2 concerns the association of ERM quality ratings with market

performance before, during and after the recent financial crisis. We classify the period from

January 1st, 2008 to August 31st, 2008 as pre-crisis; the period from September 1st, 2008 to

February 28th, 2009 as the crisis; and the period from March 1st, 2009 to October 31st, 2009 as

post-crisis. 16 We obtain monthly return data from the CRSP monthly file (crsp.msf), and

compute buy-and-hold abnormal returns (BHAR) based on sample and industry reference

portfolios for the three crisis periods. We estimate the following regression model separately for

each period to test the predicted association of ERM quality with abnormal security performance.

While estimating the model below we rely on the most recent ERM ratings available prior to the

beginning of each period of investigation.17

BHAR= α + β1ERMQ + β2CREDIT-RATING + β3TARP + β4BETA + (5)

β5MARKETCAP + β6B/M + β7PRET + β8E/P + β9ln(LEVERAGE)
+ β10-13 INDUSTRY_DUMMIES + ε

16
As a sensitivity analysis we replicated these tests using several alternative crisis begin dates including September 1 st, 2007, December 1st, 2007,
March 1st, 2008 and June 1st, 2008. The regression results based on these alternative crisis begin-dates were qualitatively similar. We find that the
ERMQ variable continues to be insignificant in the pre-crisis and crisis periods and significant during the post-crisis period.
17
For example, when investigating pre-crisis returns we use the most recent ERM quality rating before January 1st, 2008. Similarly we use the
most recent ratings before September 1st, 2008 and March 1st, 2009 when we study the crisis and post-crisis periods, respectively.

19
 The coefficient of ERMQ in this model measures the average change in market

performance associated with a one-level increase in the ERM rating.18 The model controls for

other factors that could affect abnormal returns, including the possible association with credit

ratings (discussed in the same report that analyzes a firms’ ERM quality) and returns by

including the S&P credit rating (CREDIT-RATING) in our regression model. The Capital Asset

Pricing Model posits expected returns to be a function of securities’ market risk. We control for

market risk by including the market model beta (BETA) estimated by regressing excess security

returns on excess market returns for the 60-month period ending one month before the period

beginning date. Prior literature (e.g., Fama and French 1993) shows that size and book-to-market

significantly explain variation in stock returns. Therefore, we include the natural logarithm of

size (MARKETCAP) and book-to-market ratio (B/M) in our regression model. We measure size

as share price times number of shares outstanding at fiscal-year-end and book-to-market as the

ratio of the book value of common equity and size using Compustat data. Jegadeesh and Titman

(1993) find that past winners outperform past losers. We control for the relation between future

returns and past returns by including the past six-month buy and hold return (PRET) as of the

period start date. Basu (1983) shows that the earnings-to-price ratio predicts future returns,

controlling for size and market risk. Therefore, we include earnings-to-price ratio (E/P)

computed as the ratio of earnings per share and price at fiscal-year-end. Because Bhandari

(1988) finds a positive relation between leverage and future returns, we control for leverage

using the natural logarithm of debt to equity ratio (ln(LEVERAGE)). We also include industry

fixed effects, based on the Fama and French (1997) 49 industry classification, to control for

possible variation in returns related to industry. Finally, Model 5 in the post-crisis period
18
SEC requires all firms to submit 10-K filings within 90 days. For large accelerated filers the deadline is 60 days. In order to ensure that
accounting data was available in event-time, we compute control variables using data from the most recent fiscal year that ended three months
before the period begin date.

20
includes a control variable for firms that received government assistance through the Troubled

Asset Relief Program (TARP) as such assistance is also likely to have affected market

performance in this period.

4. Results

Factors Associated with ERM Quality

Table 1 presents descriptive statistics for the dependent and independent variables. The

mean ERM rating is 3.54 on its scale from one to six, with 25.45 percent of companies having

―strong‖ or ―excellent‖ ratings. The mean credit rating is 4.39 on its scale from one to seven,

where a score below four represents the cutoff for investment grade rating.

Table 2 reports the estimation results of Model 1, which addresses Research Question 1

by investigating the determinants of ERMQ. Regarding complexity, results presented in Column

A show that larger companies as measured by MARKETCAP have higher quality ERM programs

(p<0.01). SEGMENTS, GLOBAL and FOREIGN are also positively associated with ERMQ

(p<0.05). These results are consistent with companies recognizing that operating complexity

creates greater risk inherent in a ―silo‖ management approach. Regarding uncertainty, Table 2

Column A shows no significance for the standard deviation of daily returns (STD-RET), standard

deviation of operating cash-flow (STD-OP-CASH) and proportion of loss years over the past five

years (LOSS-PROPORTION). These results suggest that firms with greater uncertainty do not

necessarily invest in higher quality ERM programs.

We also investigate the relation between ERM quality and financial distress. Results

show that credit rating (CREDIT-RATING, p<0.01) and Z-score (Z-SCORE, p<0.1) are

significant and positively associated with ERMQ, while LEVERAGE (p<0.01) is negatively

associated with ERMQ. Operating cash flow (OP-CASH) is not significant. Generally, these

21
results support the argument that less distressed firms have higher quality ERM programs, due to

availability of resources to invest in such programs, rather than the alternative proposition that

companies with greater risk of distress invest in high quality ERM to mitigate risk of lower-tail

outcomes.

[Insert Table 2]

We next investigate the relation between ERM program quality and corporate

governance. RISK-STRUCTURE, a variable combining the existence of a risk officer and/or a

risk committee is positively associated with ERMQ (p<0.01). The positive coefficient of AC-

RISK-OVERSIGHT indicates that firms with audit committees whose charters indicate direct

oversight over risk have higher ERMQ (p<0.10). Estimation results also show that audit

committees with a larger proportion of individuals having more upper management experience

(e.g. CEOs) have higher ERMQ (p<0.10), but audit committee size is not significant. We also

find a significant positive association of ERM quality with board tenure (BOARD-TENURE,

p<0.05), pointing to the importance of institutional knowledge for risk management, but not with

BOARDIND and CEO-DUALITY.19 Insignificant coefficients for other board and management

characteristics suggest that these very general governance attributes are poor proxies for board

oversight activities (e.g., Larcker et al. 2007). We find a negative coefficient on AUDIT-

RELATED-RISK (p<0.01), implying that companies with strong controls over financial reporting

and stable relationships with external auditors also tend to manage other risks well.20 Regarding

other control variables, we find that more mature companies in our sample are less apt to invest

in higher quality ERM (p<0.01), but the coefficient on NYSE listing is insignificant.

19
We also investigate board size, CEO tenure, and CEO age and found that these variables were not significant.
20
When we separate audit related risk into material weakness disclosure and auditor change the results remain unchanged and both variables are
negative and statistically significant (p<0.05).

22
 We perform several sensitivity analyses to test the robustness of our findings. First,

because ERMQ ranges from one to six, we use an ordered logistic regression. Results (presented

in Column B) are similar, as are those of a logistic regression with the dichotomous variable

STRONG_ERM (untabulated). In addition, we address possible nonlinearity in the association of

credit ratings categories with default probability by using rating factors from Moody’s and Fitch,

and default probabilities used by S&P. Analyses using these alternative measures (not tabled)

again yield results that are similar to those reported in Table 2. We also test for survivorship bias

by identifying and removing four companies that delisted after filing for bankruptcy during our

study period. Re-estimating Model 1, results are unchanged. Also, because S&P ERM ratings are

fairly new, we test for a learning curve effect by removing early ratings (those made in 2006) and

find consistent results.21

One empirical concern pertains to companies excluded from the sample. Not all

companies in the financial services industry are covered by S&P, and many that receive credit

ratings from S&P do not also receive ERM ratings. Selection of companies receiving S&P ERM

coverage is unlikely to be random, which could potentially impact the reliability and external

validity of our results. To test for this possibility, we extract data on 767 additional observations,

including companies within the same industries covered by our primary sample, which receive

S&P credit ratings but do not receive ERM ratings during the same time period. 22 We employ a

Heckman (1976) selection model to test for bias in OLS regression results. Based on prior

research investigating factors associated with coverage by rating agencies (e.g., Sufi 2009), we

include financial risk (MARKET-TO-BOOK), company size (MARKETCAP), maturity (FIRM-

21
Since our sample is not a balanced panel, as a sensitivity analysis we re-estimated Model 1 while including only the most recent observation per
company. Again, our results are qualitatively similar.
22
We perform a similar analysis that includes all companies in the same two-digit industry codes that are not necessarily covered by S&P (4,122
observations) and obtain similar results.

23
AGE) and leverage (LEVERAGE). As the instrument in the first-stage model of S&P credit rating

coverage, we use INST-OWN. As noted by the SEC (2003), debt issuers seek credit ratings to

improve marketability and pricing of their offerings. Greater penetration of a company into the

institutional market implies that the company will seek multiple ratings for its debt, thus

increasing the likelihood of rating by S&P. Estimating Model 1 with the Heckman procedure, we

observe (Panel B) that all of the selection variables affect the likelihood of being censored out of

the sample. However, the null hypothesis that rho=0 is not rejected (p=0.475), which indicates

that the OLS results in Panel A are not biased.23

In sum, results of Model 1 contribute to the literature by providing an initial view of

factors associated with ERM program quality. Prior research (Hoyt and Liebenberg 2011)

examines determinants of ERM program existence. Our analysis of ERMQ determinants

identifies some factors (e.g., size and leverage) that are also associated with program quality.

However, our results add new insight in showing that ERMQ is associated with publicly

available proxies for corporate governance, as well as audit risk and company complexity.

ERM Quality, Performance and Market Response

Accounting Performance and Market Valuation

Hypotheses 1(a) and 1(b) predict that ERM quality will be positively associated with firm

performance and value. To test whether ERM quality is associated with accounting performance,

we use Model 2, whose dependent variable is ROA. Results of the OLS regression are presented

in Table 3, Column A. The coefficient on ERMQ is positive and significant (p<0.01), supporting

Hypothesis 1(a). The coefficient of 0.0114 in the ROA regression implies that a one-level

23
The Dodd-Frank Act requires boards to consider risks that arise from incentive compensation. Incentive compensation could have also
influenced ERM ratings during our sample period. In additional analysis we control for risk taking incentives by manually collecting CEO
compensation to construct a variable that measures the ratio of incentive to total compensation. This variable is not significant while other results
remain qualitatively similar.

24
increase in ERMQ is associated with 1.14 percent increase in ROA from the sample mean of 1.9

percent (see Table 1).24 Hypothesis 1(b) predicts that ERM quality is positively related to market

valuation, measured as Tobin’s Q. Table 3 Column B presents results of Model 3, revealing a

positive and significant association between ERMQ and Tobin’s Q (p<0.05), which supports

Hypothesis 1(b). The coefficient on ERMQ implies that each increase in ERMQ level is

associated with an average increase of 0.0388 in Tobin’s Q. Compared to the sample mean of

1.14 (Table 1), this implies an increase of 3.40 percent in the typical firm’s Q value, which is

economically significant. Testing the sensitivity to using ERMQ_RAW (the original ratings by

S&P), we observe consistent results. 25 Our results also indicate no significant association

between credit ratings and Tobin’s Q.26

[Insert Table 3]

We perform supplemental analyses using 2SLS to determine whether results are sensitive

to controlling for endogeneity in the ERM rating. We choose instrumental variables from Model

1 that do not overlap with independent variables in the Tobin’s Q model and do not give rise to

overidentification: CREDIT-RATING, AUDIT-RELATED-RISK, and FIRM-AGE.27,28 Column C

shows that the coefficient of ERMQ remains significant in the ROA model.29 Similarly, Column

D reports the results of the second stage regression of the Tobin’s Q model, showing that as with

the OLS results, ERMQ is positive and significant (p<0.05). The Hansen J statistic for over-
24
Substituting ERMQ for ERMQ_RAW yields similar results. Also, our results remain unchanged when estimating this model without 2006
ratings, or using the most recent ERM ratings among unique firms.
25
Also, using the last available rating among unique firms as well as omitting ratings received in 2006 yields consistent results.
26
Variance inflation factors do not suggest multicollinearity as a concern in any of the regression results. MARKETCAP (2.28) and BETA (5.00)
in the Global Financial Crisis analysis and BOARD-SIZE (2.33) in the Tobin’s Q model are the only variables above two but still well below the
conventional cutoff of ten.
27
CREDIT-RATING is a valid instrument in this system, as results of Model 1 show that it is associated with ERMQ, but single-stage OLS
models of Tobin’s Q and ROA both show no significant effect of CREDIT-RATING. Because CREDIT-RATING and ERMQ are positively
correlated (0.49), it is possible that multicollinearity between the measures prevents significance on CREDIT-RATING. To investigate, we
regressed CREDIT-RATING on ERMQ, and substituted the residual from this model for CREDIT-RATING in Models 2 and 3. ERMQ remains
significant in the presence of the residual credit rating, and the residual credit rating is not significant in either model. Also, we exclude RISK-
STRUCTURE from the first stage as its inclusion causes overidentification.
28
In this model, CREDIT-RATING, is only included as instrument and is therefore removed from the Tobin’s Q regression model.
29
As a robustness test, we also estimate models wherein the dependent variables are as follows: (1) ROA based on operating income and (2) Net
cash flow from operating activities. In both regressions we observe that ERMQ is positive and significant (p<0.1, and p<0.05 respectively).

25
identification (Larcker and Rusticus 2010) is not significant (p=>0.52 for both), implying that the

instruments are valid.

Market Reactions to ERM Quality Ratings Disclosures and Revisions

We next present results relevant to testing Hypothesis 2, which predicts that market

reactions are associated with the initial ERM (Table 4 Panel A) and revised (Table 4 Panel B)

rating announcements. As previously noted, we primarily focus on results using sample-adjusted

abnormal returns. Panel A shows that the cumulative average abnormal return for the event

window (-2, +2) based on sample-adjusted returns is 1.62 percent, statistically significant using

parametric (p<0.01) but not using non-parametric statistics. This pattern suggests that the overall

market response could be driven by outliers. When we examine the market response preceding

the announcement date (-2, 0) we find that both parametric and nonparametric tests are

significant at least at p<0.05. However, in the post-announcement period (0, +2) we find only the

parametric test statistic to be significant. Examining individual observations, we find that the

post-announcement drift is driven by a single company, Colonial Bancgroup. When we exclude

this observation, the (0, +2) event window becomes insignificant using both parametric and non-

parametric tests. Taken together, these results imply that the market values ERM quality ratings,

but the information is already incorporated into share prices prior to the S&P announcement. The

remaining columns show that results of parametric testing of the pre-announcement response is

not sensitive to use of alternative reference portfolios, but the nonparametric result is only found

when the reference portfolio is formed within the financial services industry. Overall, these

results show that information contained within the revelation of ERM quality ratings is important

and is used by investors to revise their average expectations about firm value.

[Insert Table 4]

26
 Results of testing market reaction to changes in ERM quality ratings are reported in Table

4 Panel B. These statistics suggest a strong market reaction to changes in the ERM ratings. The

five-day market reaction (CAAR -2, +2) centered on the date of the ERM rating change, is 4.51

percent based on sample-adjusted abnormal returns. CAARs (-2, +2) computed using other

reference portfolios are also statistically significant using Patell and Rank test statistics across all

event windows. We conclude that Hypothesis 2(b) is supported, as changes in S&P’s ERM

ratings appear to convey significant information to markets and investors.

ERM Quality and the Earnings Response Coefficient

Hypothesis 3 predicts that the market reaction to earnings surprises is greater for firms

with higher ERM quality. Table 5 presents results of estimating Model 4, based on earnings

announcement cumulative abnormal returns relative to sample, industry, market and size

reference portfolios. The test variable for Hypothesis 3 is the interaction of unexpected earnings

(UE) and STRONG_ERM, which is positive and significant (p<0.01), supporting Hypothesis 3.

This indicates that markets react more strongly to earnings surprises of firms with higher quality

ERM programs, implying that investors perceive the earnings of those firms to be of higher

quality.30 Results are similar using industry-adjusted, market-adjusted, or size-adjusted returns to

measure market reaction to earnings announcements.31

Further, we examine whether our results are affected by nonlinearity, using the rating

factors of Fitch and Moody’s and default probabilities reported by S&P (not tabled), and find

similar results. Finally, we again test sensitivity to learning curve effects on the part of S&P

30
In additional analysis we control for corporate governance quality. To construct a measure of governance quality we sum of the quartile
rankings of the governance variables (INDEPENDENTBOARD, MEANTENURE, RISK, PAFE, PSFE, SIZEAUD, OVERSIGHT and
CEOCHAIR). Higher rankings on the governance variables, with the exception of CEOCHAIR signify higher governance quality. The ranking of
the CEOCHAIR variable was reversed and summed with the remaining variables to compute the final corporate governance score. We created an
indicator variable that accepts one for firms whose governance score was above the median, signifying strong governance, zero otherwise. We
find that the interaction between ERC and corporate governance is not significant, while other results remain qualitatively similar.
31
We repeat the earnings response coefficient analysis using an alternative STRONG_ERM variable that takes a value of one for firms with ERM
ratings greater than four and find that the primary interaction variable is statistically significant and that the remaining results are similar

27
analysts by including the interaction of year and UE X STRONG_ERM variables and find that the

coefficient of UE X STRONG_ERM is unaffected. In conclusion, results of earnings response

coefficient analysis are insensitive to the use of alternative ERMQ definitions, alternative credit

rating specifications and controlling for the timing of ERM ratings.

[Insert Table 5]

ERM and the Global Financial Crisis

Because the advantage of effective ERM is likely to be more pronounced at times of

greater risk, we investigate whether firms with higher quality ERM programs experienced

superior market returns in the periods surrounding the GFC (Research Question 2). Table 6

reports regression results based on pre-crisis, crisis and post-crisis return data, respectively.

Regression results based on pre-crisis period data (columns 1 and 2) indicate that the coefficient

on ERMQ is not significant. This implies no association of ERM quality with equity market

performance before the financial crisis, prior to the global financial crisis (GFC) when systemic

risk was relatively low. Results based on crisis period data, reported in the third and fourth

columns, show that ERMQ is also not statistically significant. Thus, as the crisis deepened,

financial services companies with stronger ERM programs fared no differently than those with

weaker programs. In contrast, the last two columns of Table 6 show a significant association

(p<0.01) of ERM quality and market returns in the post-crisis period. In the sample-adjusted

model, the coefficient on ERMQ is 0.2152 (i.e., a one-level variation in ERM quality is

associated with 21.52 percent change in post-crisis buy-and-hold abnormal return). 32 This

32
These results are obtained while controlling for credit ratings and several risk measures. The signs and significance of these variables differ by
phase of the crisis. For instance, BETA is negative and significant (p < 0.10) in the pre-crisis period and positive and significant (p < 0.05) in the
post-crisis period, consistent with higher variance in returns for higher-beta companies. Table 6 also shows that the receipt of TARP funds by
some companies in the sample is positively associated with returns in the post-crisis period (p < 0.05).

28
suggests that firms with superior ERM programs rebounded more swiftly from the crisis to

regain market value.33

[Insert Table 6]

To ascertain whether results reported in Table 6 are insensitive to methodological

choices, we conduct a number of robustness checks. First, replacing the ERMQ variable with

ERMQ_RAW, or measuring abnormal returns using size-decile and market reference portfolios,

leaves results unaffected. Replacing the credit rating variable CREDIT-RATING with Moody’s

and Fitch’s ratings factors and S&P’s default probability rates also yields similar results. Finally,

we estimate shifts in effects over time by including the interaction of ERMQ with year dummy

variables. The interaction variables are statistically insignificant and inferences are unchanged

from the original models. We test sensitivity to using the calendar-time portfolio approach by

constructing a portfolio that takes long positions in firms with ERM ratings of four or above and

short positions in firms with ERM ratings less than four. We regress the returns of this portfolio

on excess market returns and size, book-to-market, and momentum factor returns (four-factor

model). The intercept of the regression serves as an estimate of the average abnormal returns

associated with ERM quality. The estimation results (reported in Online Appendix 2) reveal

intercepts of 0.000, -0.001 and 0.002 for the pre-crisis, crisis and post-crisis periods,

respectively. Only the post-crisis intercept is statistically significant, suggesting that firms with

strong ERM programs outperformed other sample firms by an average daily return of 0.2 percent

during the post-crisis period. This corresponds to an annualized abnormal return of

33
We supplement our tests of the financial crisis period by testing whether firms with higher ERMQ are less likely to be affected by sharp market
fluctuations (i.e., have lower volatility of returns). We investigate this issue by regressing the standard deviation of daily returns within each crisis
sub-period on ERMQ and control variables. In contrast to our findings based on buy-and-hold abnormal returns, we do not find any significant
association between ERMQ and return volatility for any sub-period of the crisis.

29
approximately 40 percent. The results are largely similar when we use the three-factor model or

the Capital Asset Pricing Model to estimate abnormal returns.

5. Conclusions and Limitations

This paper investigates determinants of ERM program quality and the association of

ERM quality with firm performance and value, among financial services companies with ERM

quality ratings provided by Standard & Poor’s during 2006-08. As such, it follows a line of

research investigating why corporations adopt specific corporate governance mechanisms, and

whether those mechanisms achieve their goals. In addition to advancing knowledge relevant to

these questions in general, research in the context of ERM is also important because it is a

product of a joint effort by the professional and academic accounting communities, expanding

the COSO framework beyond the financial reporting system to management control more

broadly defined.

This research is also important due to increasing pressure on firms to invest resources in

improving risk management; for example, the SEC has considered requiring specific disclosures

on the risk qualifications of individual directors. While ERM presents the most widely accepted

framework that entities can use to manage risks, there is limited empirical evidence on the

relation between ERM and firm performance/value. Most prior research examines the drivers of

ERM adoption and its consequences, rather than ERM program quality. By using a direct

measure of ERM quality, we examine a set of companies known to have ERM programs, and

differentiate them by quality as determined by an independent rater. Given the challenges to

frequently used methods of measuring the presence or quality of corporate governance

mechanisms (e.g., surveys of company personnel or publicly available data), these quality ratings

provide a valuable view of the extent to which each ERM program has adopted and internalized

30
the integration of risk management across disparate parts of the entity, instead of a traditional

silo-based approach.

Prior to discussing conclusions, we note several limitations of our analysis. First, we

implicitly assume that S&P ERM ratings validly represent aspects of ERM quality, and that

program effectiveness increases in the ERM rating score. If this were not the case, it would

produce a bias against finding the associations we test, and our results are consistent with this

assumption. However, we recognize that there may be features of ERM quality that are not

measurable, which would reduce the power of our tests. Second, ratings agencies such as S&P

have been criticized for their ratings during the financial crisis, particularly their ratings of

financial products. If credit ratings were biased upward during our sample period, this bias may

also have applied to ERM quality ratings. However, the mean rating in our sample is at the

midpoint of the theoretical range, which does not suggest an upward bias. Third, we limit our

analysis to financial services companies, as S&P’s ERM ratings currently only cover those types

of firms. Future research could explore issues related to ERM quality in other industries as data

become available. Fourth, because S&P’s ERM ratings are part of the credit ratings process, their

analysts focus on the impact that ERM might have on the firm’s ability to repay debt. While

prior research investigates equity investors’ use of credit ratings in making market decisions, the

impact of ERM on the equity markets is less clear. Despite this limitation, we find strong

evidence that equity markets do find the ERM ratings useful.

Our first set of conclusions relates to determinants of ERM program quality. Davila and

Foster (2007), Elbashir et al. (2011; 157) note that studies of management control systems

typically examine adoption, but rarely explore ―the variation in quality or depth of the use of

MCS following that adoption.‖ Similarly, prior ERM research examines ERM adoption through

31
risk officer appointments (e.g., Liebenberg and Hoyt 2003) finding few significant determinants.

Our results provide insight in showing that companies with superior ERM programs are more

complex, have greater financial resources, and better corporate governance as measured by

publicly available proxies. Similarly, prior research finds that effective internal controls (as

determined by auditors using the COSO framework) are associated with overall corporate

governance quality and audit committee financial expertise (Hoitash et al. 2009). Further, studies

find that resource constraints inhibit implementation of effective internal controls (e.g., Doyle et

al. 2007) and management control systems (Elbashir et al. 2011).34 In sum, our results provide

insights into the characteristics of financial services companies that allocate sufficient resources

to integrating risk management activities to achieve a high quality rating by S&P.

We also examine the proposition that high quality ERM programs enhance operating

performance and add value to companies, controlling for the characteristics identified in the

determinants analysis. As noted by Ittner and Larcker (2001), performance effects of

management control techniques are relatively unstudied. We find that firm performance as

measured by accounting returns, as well as market valuation using Tobin’s Q, are higher for

firms that invest in higher quality ERM, while controlling for possible endogeneity bias. These

results suggest that the improved risk management inherent in higher quality ERM programs

assists performance by helping to mitigate losses and/or to take advantage of opportunities.

Further, while results show higher market valuation for such companies, there are various ways

in which this could occur. We find that when ERM programs are initially rated by S&P, the

average market reaction is higher for strong/excellent ERM rated firms than for those with lower

34
S&P’s ERM rating criteria (e.g., Samanta et al. 2006; see Online Appendix 1) focus on company processes that are not externally visible, such
as independence between the risk management function and the business, the authority to advise the business if tolerances are exceeded and the
involvement of CROs in strategic planning. Our results imply that publicly available proxies, while recognized as limited in their portrayal of
underlying governance processes (e.g., Larcker et al. 2007), still explain some variance in ERM quality.

32
ratings. In addition, there is a positive and significant market reaction to revisions of ERM

ratings. These results imply market anticipation of better future performance by high-quality

ERM companies, while prior research does not find an overall market reaction to CRO

announcements as a proxy for ERM implementation (Beasley et al. 2008). Interestingly, the

market response to initial ratings appears prior to the announcement, implying that the ERM

information is incorporated into prices prior to the information becoming public. Further, we find

that the intensity of investors’ average reactions to earnings surprises increases for companies

with higher quality ERM. This is evidence of increased usefulness of accounting information,

consistent with market perceptions of greater reliability of earnings due to those companies’

better control over their future activities.

Finally, we examine ERM quality in the recent financial crisis, which our study period

encompasses. Because ERM programs are intended to protect against lower-tail outcomes, the

financial crisis provides a natural setting in which to examine this proposition. We consider how

relative ERM quality affected these financial industry firms during this very challenging time.

We find no association of ERM quality with abnormal returns in the sub-period preceding the

crisis (i.e., January through August 2008). While we cannot definitively explain this lack of

response, it may be that because market returns in that period were generally high, risk was not

as important an issue. We also find no association of ERM quality with returns during the crisis

(September 2008 through February 2009), suggesting that firms with higher-quality ERM were

not differentially protected from the sudden and catastrophic market declines experienced during

the crisis. In contrast, however, we find a strong association of ERM quality and returns in the

initial rebound period (March through October 2009). This result suggests that as the market

33
rebounded, investors looked to information such as ERM quality, which indicated that some

firms could address future risks in a more systematic and integrated manner.

34
 References

Abbott, L. J., S. Parker, and G. F. Peters. 2004. Audit committee characteristics and
restatements. Auditing: A Journal of Practice & Theory 23 (1): 69-87.
Anderson, K. L. and T. L. Yohn. 2002. The effect of 10K restatements on firm value,
information asymmetries, and investors' reliance on earnings. Working Paper, Georgetown
University. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=332380.
Anderson, R. C., and D. M. Reeb. 2003. Founding-family ownership and firm performance:
Evidence from the S&P 500. Journal of Finance 58 (3): 1301-1328.
Ashbaugh-Skaife, H., D. W. Collins, and R. LaFond. 2006. The effects of corporate governance
on firms’ credit ratings. Journal of Accounting & Economics 42 (1-2): 203-243.
Ashbaugh-Skaife, H., D. W. Collins, and W. R. Kinney. 2007. The discovery and reporting of
internal control deficiencies prior to SOX-mandated audits. Journal of Accounting &
Economics 44 (1-2): 166-192.
Ashbaugh-Skaife, H., D. W. Collins, J. W. R. Kinney, and R. LaFond. 2008. The effect of SOX
internal control deficiencies and their remediation on accrual quality. The Accounting Review
83 (1): 217-250.
Balsam, S., J. Krishnan, and J. S. Yang. 2003. Auditor industry specialization and earnings
quality. Auditing: A Journal of Practice & Theory 22 (2): 71-97.
Basu, S. 1983. The relationship between earnings' yield, market value and return for NYSE
common stocks : Further evidence. Journal of Financial Economics 12 (1): 129-156.
Beasley, M. S., B. Branson, and B. Hancock. 2009. Report on the Current State of Enterprise
Risk Oversight. Raleigh, NC: ERM Initiative at North Carolina State University (March).
Beasley, M., D. Pagach, and R. Warr. 2008. Information conveyed in hiring announcements of
senior executives overseeing enterprise-wide risk management processes. Journal of
Accounting, Auditing & Finance 23 (3): 311-332.
Beasley, M. S. 1996. An empirical analysis of the relation between the board of director
composition and financial statement fraud. The Accounting Review 71 (4): 443-465.
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 & Public Policy 24 (6): 521-531.
Bédard, J., S. M. Chtourou, and L. Courteau. 2004. The effect of audit committee expertise,
independence, and activity on aggressive earnings management. Auditing: A Journal of
Practice & Theory 23 (2): 15-37.
Beneish, M. D., M. B. Billings, and L. D. Hodder. 2008. Internal control weaknesses and
information uncertainty. The Accounting Review 83 (3): 665-703.
Berger, P. G., and E. Ofek. 1995. Diversification's effect on firm value. Journal of Financial
Economics 37 (1): 39-65.
Bhagat, S., and B. Black. 2002. The non-correlation between board independence and long-term
firm performance. Journal of Corporation Law 27 (2): 231.

35
Bhandari, L. C. 1988. Debt/Equity ratio and expected common stock returns: Empirical
evidence. Journal of Finance 43 (2): 507-528.
Brancato, C., M. Tonello, N. Hexter, and K. Newman. 2006. The Role of U.S. Corporate Boards
in Enterprise Risk Management. New York, NY: The Conference Board,
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=941179.
Bushman, R. M., and A. J. Smith. 2001. Financial accounting information and corporate
governance. Journal of Accounting & Economics 32 (1-3): 237-333.
Committee of Sponsoring Organizations of the Treadway Commission (COSO). 1992. COSO
Report: Internal Control – an Integrated Framework. New York: AICPA.
———. 2004. Executive Summary Enterprise Risk Management - Integrated Framework. New
York: AICPA.
Corrado, C. J. 1989. A nonparametric test for abnormal security-price performance in event
studies. Journal of Financial Economics 23 (2): 385-395.
Davila, A., and G. Foster. 2005. Management accounting systems adoption decisions: Evidence
and performance implications from early-stage/startup companies. The Accounting Review 80
(4): 1039-1068.
———. 2007. Management control systems in early-stage startup companies. The Accounting
Review 82 (4): 907-937.
DeFond, M. L., R. N. Hann, and X. Hu. 2005. Does the market value financial expertise on audit
committees of boards of directors? Journal of Accounting Research 43 (2): 153-193.
Doyle, J., W. Ge, and S. McVay. 2007. Determinants of weaknesses in internal control over
financial reporting. Journal of Accounting & Economics 44 (1-2): 193-223.
Elbashir, M. Z., P. A. Collier, and S. G. Sutton. 2011. The role of organizational absorptive
capacity in strategic use of business intelligence to support integrated management control
systems. The Accounting Review 86 (1): 155-184.
Fama, E. F., and M. C. Jensen. 1983. Separation of ownership and control. Journal of Law &
Economics 26 (2): 301-326.
Fama, E. F., and K. R. French. 1993. Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics 33 (1): 3-56.
———. 1997. Industry costs of equity. Journal of Financial Economics 43 (2): 153-193.
Francis, J., K. Schipper, and L. Vincent. 2002. Expanded disclosures and the increased
usefulness of earnings announcements. The Accounting Review 77 (3): 515-546.
Froot, K. A., D. S. Scharfstein, and J. C. Stein. 1993. Risk management: Coordinating corporate
investment and financing policies. Journal of Finance 48 (5): 1629-1658.
Gaskel, G., K. G. Maher, and P. Wong. 2007. Guardian Life Insurance Co. of America and
Guardian Insurance & Annuity Co. Standard and Poor’s Ratings Report.
Gordon, L. A., M. P. Loeb, and C. Tseng. 2009. Enterprise risk management and firm
performance: A contingency perspective. Journal of Accounting & Public Policy 28 (4): 301-
327.

36
Hackenbrack, K. E., and C. E. Hogan. 2002. Market response to earnings surprises conditional
on reasons for an auditor change. Contemporary Accounting Research 19 (2): 195-223.
Hammersley, J. S., L. A. Myers, and C. Shakespeare. 2008. Market reactions to the disclosure of
internal control weaknesses and to the characteristics of those weaknesses under Section 302
of the Sarbanes Oxley Act of 2002. Review of Accounting Studies 13 (1): 141-165.
Heckman, J. 1976. The common structure of statistical models of truncation, sample selection,
and limited dependent variables and a simple estimator for such models. Annals of Economic
and Social Measurement 5 (4): 475-492.
Hoitash, U., R. Hoitash, and J. C. Bedard. 2009. Corporate governance and internal control over
financial reporting: A comparison of regulatory regimes. The Accounting Review 84 (3): 839-
867.
Hoyt, R. E., and A. P. Liebenberg. 2011. The value of enterprise risk management. Journal of
Risk and Insurance 78 (4): 795-822.
Ittner, C. D., and D. F. Larcker. 2001. Assessing empirical research in managerial accounting: A
value-based management perspective. Journal of Accounting & Economics 32 (1-3): 349-
410.
Jegadeesh, N., and S. Titman. 1993. Returns to buying winners and selling losers: Implications
for stock market efficiency. Journal of Finance 48 (1): 65-91.
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 & Insurance
Review 6 (1): 53-73.
Klein, A. 2002. Audit committee, board of director characteristics, and earnings management.
Journal of Accounting & Economics 33 (3): 375-400.
Krishnan, J. 2005. Audit committee quality and internal control: An empirical analysis. The
Accounting Review 80 (2): 649-675.
Larcker, D. F., S. A. Richardson, and İ. Tuna. 2007. Corporate governance, accounting
outcomes, and organizational performance. The Accounting Review 82 (4): 963-1008.
Larcker, D. F., and T. O. Rusticus. 2010. On the use of instrumental variables in accounting
research. Journal of Accounting & Economics 49 (3): 186-205.
Liebenberg, A. P., and D. W. Sommer. 2008. Effects of corporate diversification: Evidence from
the Property–Liability insurance industry. Journal of Risk & Insurance 75 (4): 893-919.
Liebenberg, A. P., and R. E. Hoyt. 2003. The determinants of enterprise risk management:
Evidence from the appointment of chief risk officers. Risk Management & Insurance Review
6 (1): 37-52.
McConnell, J. J., and H. Servaes. 1990. Additional evidence on equity ownership and corporate
value. Journal of Financial Economics 27 (2): 595-612.
McShane, M. K., A. Nair, and E. Rustambekov. 2011. Does Enterprise Risk Management
increase firm value? Journal of Accounting, Auditing and Finance 26 (4): 641-658.
Nocco, B. W., and R. M. Stulz. 2006. Enterprise risk management: Theory and practice. Journal
of Applied Corporate Finance 18 (4): 8-20.

37
Pagach, D. and R. Warr. 2007. An empirical investigation of the characteristics of firms adopting
enterprise risk management. Working Paper, North Carolina State University.
———. 2008. The effects of enterprise risk management on firm performance. Working Paper,
North Carolina State University.
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1155218.
Patell, J. M. 1976. Corporate forecasts of earnings per share and stock price behavior: Empirical
tests. Journal of Accounting Research 14 (2): 246-276.
Prawitt, D. F., J. L. Smith, and D. A. Wood. 2009. Internal audit quality and earnings
management. The Accounting Review 84 (4): 1255-1280.
Rosenstein, S., and J. G. Wyatt. 1990. Outside directors, board independence, and shareholder
wealth. Journal of Financial Economics 26 (2): 175-191.
Roy, A. D. 1952. Safety first and the holding of assets. Econometrica 20 (3): 431-449.
Samanta, P., R. Barnes, and M. Puccia. 2006. Assessing Enterprise Risk Management Practices
of Financial Institutions. Standard & Poors RatingsDirect®.
Santori, L., M. Puccia, G. Osborne, R. Jones, and H. Wright. 2005. Evaluating the Enterprise
Risk Management Practices Of Insurance Companies. Standard & Poor’s RatingsDirect®
Schwartz, K. B., and K. Menon. 1985. Auditor switches by failing firms. The Accounting Review
60 (2): 248-261.
Securities and Exchange Commission (SEC). 2003. Report on the Role and Function of Credit
Rating Agencies in the Operation of the Securities Markets as Required by Section 702(b) of
the Sarbanes-Oxley Act of 2002. Washington, D.C.: SEC.
Sengupta, P. 1998. Corporate disclosure quality and the cost of debt. The Accounting Review 73
(4): 459-474.
Standard & Poor's (S&P). 2010. Expanded Definition of Adequate Classification in Enterprise
Risk Management Scores. Criteria Report.
Sufi, A. 2009. The real effects of debt certification: Evidence from the introduction of bank loan
ratings. Review of Financial Studies 22 (4): 1659-1691.
Teoh, S. H., and T. J. Wong. 1993. Perceived auditor quality and the earnings response
coefficient. The Accounting Review 68 (2): 346-366.
Yermack, D. 1996. Higher market valuation of companies with a small board of directors.
Journal of Financial Economics 40 (2): 185-211.
Zimmerman, J. L. 2001. Conjectures regarding empirical managerial accounting research.
Journal of Accounting and Economics 32 (1-3): 411-427.

38
TABLE 1
Variable Definitions and Descriptive Statistics
Mean (Median),
Variable Name Variable Definition [source]
S.D.
Panel A: ERM Quality Determinants (Model 1)
ERM quality measure ranging from 1 (low) to 6 (high) [S&P
ERMQ 3.545 (3), 1.134
Ratings Direct]
The raw ERM ratings published by S&P ranging from 1 (low) to 4
ERMQ_RAW 2.26 (2), 0.594
(high). [S&P Ratings Direct]
Natural log of the market value of equity [Compustat data PRCC_F 8.476 (8.362),
MARKETCAP
× data CSHO] 1.454
Sum of reported business and geographic segments [Compustat
SEGMENTS 5.388 (4), 4.957
Segment file]
=1 if the company operates in most countries [SEC filings]; 0
GLOBAL 0.103 (0), 0.305
otherwise
= 1 if the company has non-zero foreign currency translation; 0
FOREIGN 0.139 (0), 0.347
otherwise [Compustat data FCA]
Standard deviation of daily stock returns during the fiscal year. 0.033 (0.022),
STD-RET
[CRSP] 0.024
The standard deviation of operating cash-flows during the past five 0.026 (0.014),
STD-OP-CASH
years. [Compustat] 0.036
LOSS- 0.186 (0.071)
The proportion of loss years over the past five years.
PROPORTION 0.276
An ordinal variable based on credit ratings by S&P with values as
CREDIT- follows: AAA (7), AA+, AA and AA- (6), A+, A, and A- (5),
4.394 (4), 0.809
RATING BBB+, BBB, and BBB- (4), BB+, BB, BB- (3), B+, B, and B- (2),
CCC+, CCC, CC, C, D or SD (1) [S&P credit rating]
Cash flow from operations divided by total assets [Compustat data 0.063 (0.038),
OP-CASH
OANCF] 0.089
Ratio of total liabilities to total assets [ Compustat (data DLC+data
LEVERAGE 0.11 (0.059), 0.141
DLTT)/data AT)]
The sum of the mean rate of return on assets and the mean equity-
to-assets ratio divided by the standard deviation of the return on 22.191 (16.551),
Z-SCORE
assets. A minimum of four and maximum of 15 years of historical 22.787
data were required to compute this measure. [Compustat]
RISK- =1 if either RISK-OFFICER or RISK-COMMITTEE =1; 0
0.4 (0), 0.491
STRUCTURE otherwise.
AC-RISK- =1 if the audit committee has risk oversight responsibilities. [Proxy
0.273 (0), 0.447
OVERSIGHT statements]; 0 otherwise
Proportion of audit committee members that are accounting
financial experts (i.e. the biography indicates at least one of the
PAFE following: CPA, chief financial officer, auditor, chief accounting 0.273 (0.25), 0.242
officer, controller, treasurer or VP-Finance) [AuditAnalytics or
proxy statements]
Proportion of audit committee members that are supervisory
financial experts (i.e., the biography indicates at least one of the
0.383 (0.333),
PSFE following: CEO, chief operating officer, board chair or company
0.241
president, and individual is not an accounting financial expert)
[AuditAnalytics or proxy statements]
Number of members of the audit committee in 2004 [IRRC,
ACSIZE 4.2 (4), 1.122
AuditAnalytics or proxy statements]

39
TABLE 1 (continued)
Variable Definitions and Descriptive Statistics
Mean (Median),
Variable Name Variable Definition [source]
S.D.
Panel A: ERM Quality Determinants (Model 1) (Continued)
Percentage of outsiders as board members [IRRC, Audit 0.858 (0.889),
BOARDIND
Analytics or proxy statements] 0.083
BOARD- Average tenure of outsiders serving on the board [IRRC, Audit
7.842 (7), 3.83
TENURE Analytics or proxy statements]
= 1 if the CEO is also the board chair; 0 otherwise [IRRC,
CEO-DUALITY 0.582 (1), 0.495
Audit Analytics or proxy statements]
= 1 for companies that switched auditors in the current year, or
AUDIT-
reporting Section 302/404 material weakness; 0 otherwise 0.091 (0), 0.288
RELATED-RISK
[AuditAnalytics]
Natural logarithm of the number of years the firm has coverage 2.774 (2.773),
FIRM-AGE
by Compustat [Compustat] 0.646
=1 if the company’s shares are traded at the New York Stock
NYSE 0.442 (0), 0.498
Exchange; 0 otherwise. [CRSP]
Panel B: ERM Quality Determinants (Model 1)
MARKET-TO- Ratio of market value at fiscal year-end to book value of 1.455 (1.192)
BOOK common equity [Compustat data (PRCC_F × CSHO)/ CEQ] 1.323
(Book value of assets – (book value of equity plus the market
1.135 (1.026),
TOBIN’S Q value of equity)/ book value of assets) [Compustat data AT –
0.428
(data CEQ + data PRCC_F × CSHO) / data AT]
Logarithm of number of members serving on the board of 2.429 (2.485),
LOGBOARDSIZE
directors [IRRC, AuditAnalytics or proxy statements] 0.27
9.879 (9.586),
LOGASSETS Logarithm of total assets [Compustat data AT].
1.711
Income before extraordinary items divided total assets 0.019 (0.012),
ROA
[Compustat data IB / data AT] 0.053
Percent of shares owned by institutional investors [Compustat 0.687 (0.718),
INST-OWN
Institutional Ownership File] 0.242
Percentage growth in sales 0.006(0.015)
SALES-GROWTH
0.187
CAPITAL OVER Capital expenditure over sales [Compustat data CAPX/data 0.015(0.008)
SALES SALE] 0.025
The standard deviation of returns on assets calculated over a
0.018 (0.011),
STDROA period of no less than three years and no more than five years
0.021
[Compustat data IB / data AT]
Panel D: ERM and Earnings Response (Model 4)
Cumulative sample-adjusted returns for the three-day period 0.001 (0.005),
CAR_SAMP
centered on the earnings announcement date [CRSP] 0.072
Cumulative industry-adjusted returns for the three-day period 0.001 (0.002),
CAR_IND
centered on earnings announcement date [CRSP] 0.072
Cumulative market-adjusted returns for the three-day period 0.003 (0.005),
CAR_MAR
centered on the earnings announcement date [CRSP] 0.08
Cumulative size-adjusted returns for the three-day period 0.003 (0.004),
CAR_SIZE
centered on the earnings announcement date [CRSP] 0.079

40
TABLE 1 (continued)
Variable Definitions and Descriptive Statistics
Mean (Median),
Variable Name Variable Definition [source]
S.D.
(Reported earnings – analysts’ earnings expectation)/share price -0.009 (0.000),
UE
at the end of previous fiscal quarter [I/B/E/S and CRSP] 0.113
=1 if the ERM quality is 4 or above; 0 otherwise [S&P Ratings
STRONG_ERM 0.388 (0), 0.489
Direct]
Fiscal quarters for which income before extraordinary items are
NEG 0.07 (0), 0.255
negative [Compustat data IBQ]
Ratio of book value of common equity to market value at fiscal
B/M 0.73 (0.683), 0.341
year-end [Compustat data CEQ / (data PRCC_F × data CSHO)]
Market model beta estimated using 60 months of returns, ending 1.108 (0.992),
BETA
three months prior to the fiscal quarter end [CRSP] 0.551
Inverse of the number of analysts who made earnings forecasts 0.193 (0.143),
COV
for the fiscal quarter [I/B/E/S]. 0.193
BHAR_SAMP Sample-adjusted buy and hold return [CRSP] 0.07 (0.038), 0.734
0.141 (0.036),
BHAR_IND Industry-adjusted buy and hold return [CRSP]
0.759
An indicator variable that equals one for firms that received
0.055 (0.000),
TARP government assistance through the Troubled Asset Relief
0.229
Program.
Prior six-month buy and hold return as of the period begin date -0.229 (-0.152),
PRET
[CRSP] 0.296
Ratio of earnings per share to price [Compustat data EPFI / data 0.067 (0.085),
E/P
PRCC_F] 0.157
Notes: This table presents mean (median) and standard deviation of all variables. Each panel contains variables in the respective
models that are incremental to those previously described.

41
TABLE 2
Results of Estimating Model 1: ERM Determinants
(A) (B) (C)

Panel A: ERM Determinants Exp. Linear Ordered Heckman Two Stage

Sign Regression Logistic regression DV=ERMQ
DV=ERMQ Regression
DV=ERMQ
MARKETCAP + 0.2148*** 0.6894*** 0.1853
(3.43) (2.36) (1.14)
SEGMENTS + 0.0385** 0.1091* 0.0352*
(1.95) (1.78) (1.46)
GLOBAL + 0.6925** 2.2830** 0.6900***
(2.21) (1.65) (2.48)
FOREIGN + 0.3446** 0.7325* 0.3551*
(1.79) (1.29) (1.63)
STD-RET + 1.8087 3.0384 2.3232
(0.39) (0.23) (0.45)
STD-OP-CASH + 4.7052 13.9362* 4.9923*
(1.00) (1.45) (1.28)
LOSS PROPORTION + 0.4577 1.6692 0.3897
(1.00) (0.93) (0.56)
CREDIT-RATING + 0.2672*** 0.5936* 0.2599***
(3.25) (1.62) (2.39)
OP-CASH + 0.4302 1.4941 0.7316
(0.23) (0.51) (0.44)
LEVERAGE ? -1.2539*** -3.9950** -0.6363
(-3.40) (-2.20) (-0.20)
Z-SCORE + 0.0030* 0.0120* 0.0031*
(1.32) (1.38) (1.37)
RISK-STRUCTURE + 0.5470*** 1.7017*** 0.5320***
(4.51) (3.17) (2.88)
AC-RISK-OVERSIGHT + 0.2282* 0.6800* 0.2186*
(1.61) (1.32) (1.46)
PAFE + -0.0028 -0.1677 0.0617
(-0.01) (-0.16) (0.12)
PSFE + 0.6373* 1.5350 0.6528**
(1.59) (1.15) (1.88)
ACSIZE + -0.0468 -0.2012 -0.0413
(-0.74) (-0.82) (-0.59)
BOARDIND + 1.3264 4.4153* 1.1917
(0.98) (1.55) (1.08)
BOARD-TENURE + 0.0343** 0.1160** 0.0363**
(1.84) (1.81) (1.71)
CEO-DUALITY - 0.2229 0.5351 0.2210*
(1.44) (1.14) (1.52)

42
TABLE 2 (continued)
Results of Estimating Model 1: ERM Determinants
(A) (B) (C)
Panel A: ERM Determinants
Linear Ordered Logistic Heckman Two
Exp.
Regression Regression Stage regression
Sign
DV=ERMQ DV=ERMQ DV=ERMQ
AUDIT-RELATED-RISK - -0.6698*** -2.1587*** -0.6608***
(-4.27) (-3.21) (-3.01)
FIRM AGE ? -0.2986*** -0.9606** -0.3681
(-2.80) (-2.11) (-0.88)
NYSE + -0.0101 -0.0681 0.0029
(-0.05) (-0.13) (0.01)
Constant -1.7214 -0.3278
(-1.23) (-0.07)
Industry and Year Dummies Included Included Included
Observations 165 165 932
(165 Selected)
Wald Chi-square/Adjusted R2, or 0.461 0.296 194.42***
Pseudo R2
Panel B: Results of First Stage Heckman Model
MARKET-TO-BOOK + -0.2214**
(-2.09)
MARKETCAP + 0.1556***
(4.69)
FIRM_AGE + 0.0187***
(2.86)
LEVERAGE ? -2.5181***
(-8.92)
Constant -1.4899***
(-5.58)
Rho Chi-square 0.51
(p=.475)
Notes: This table presents results of estimating Model 1, investigating company characteristics associated with greater
investment in ERM, as measured by S&P ERM program quality ratings. Model 1 includes year and industry fixed effects and is
estimated with standard errors corrected for clustering at the year and firm level. All variables are defined in Table 1. Numbers in
the cells are coefficients (t-statistics), with significance denoted as ***, **, * for one percent, five percent, and ten percent,
respectively. One-tailed tests are presented for directional expectations.

43
TABLE 3
Results of Estimating Models 2 and 3: ERM Quality, Firm Value and Financial Performance
OLS Regressions 2SLS Regressions
Exp. (A) (B) (C) (D)
Sign Model 2: Model 3: Model 2: Model 3:
ROA Tobin's Q ROA Tobin's Q
ERMQ + 0.0114*** 0.0388** 0.0102*** 0.1120*
(3.72) (2.25) (4.11) (1.56)
LOGBOARDSIZE - -0.0441*** 0.1074 -0.0466*** 0.0809
(-3.26) (1.27) (-4.06) (0.99)
BOARDIND ? -0.0127 0.1339 -0.0095 0.0570
(-0.53) (0.49) (-0.45) (0.44)
LOGASSETS ? -0.0064*** -0.0653* -0.0074*** -0.0722*
(-2.73) (-1.75) (-4.23) (-1.83)
ROA + 5.1311*** 4.8284***
(20.60) (84.66)
INST-OWN ? 0.0062 0.0467 0.0086 -0.0012
(0.50) (0.30) (0.87) (-0.01)
SEGMENTS ? 0.0020* 0.0015 0.0020** 0.0002
(1.90) (0.16) (2.29) (0.02)
LEVERAGE ? -0.0052 0.2331 -0.0067 0.2946
(-0.12) (0.36) (-0.19) (0.53)
SALES-GROWTH + 0.0774*** -0.1086 0.0774*** -0.0735
(2.78) (-0.33) (3.39) (-0.33)
CAPITAL OVER SALES + -0.5766*** 0.8074 -0.5774*** 0.5679
(-3.36) (0.31) (-4.06) (0.26)
CREDIT-RATING ? -0.0058* 0.0561
(-1.35) (1.07)
STDROA ? -0.5885 -0.6003
(-0.79) (-1.01)
Constant 0.2166*** 0.9156*** 0.2008*** 1.1629***
(6.96) (3.34) (5.22) (5.39)
Industry and Year Dummies Included Included Included Included
Observations 165 165 165 165
Adjusted R2 0.376 0.520 0.375 0.496
Notes: This table presents results of estimating Models 2 (ROA, column A) and 3 (Tobin’s Q, column B) as a function of ERMQ
and other company characteristics. Columns C and D, respectively, show results of testing sensitivity to use of two-stage least
squares; please see the text for details of model structure. Model 2 includes year and industry fixed effects and is estimated with
standard errors corrected for clustering at the year and firm level. All variables are defined in Table 1. Numbers in the cells are
coefficients (t-statistics), with significance denoted as ***, **, *, for one percent, five percent, and ten percent, respectively. One-
tailed tests are presented for directional expectations.

44
TABLE 4
Market Reaction to the Disclosure of ERM Ratings and Ratings Revisions
Event Exp. Sample-Adjusted Industry-Adjusted Market-Adjusted Size-Adjusted
Window Sign Returns Returns Returns Returns
Panel A: Initial Enterprise Risk Management Rating Announcement (N=107)
(-2,+2) + 1.62% 1.33% 0.74% 0.70%
3.440*** 2.939*** 0.455 0.166
0.151 -0.093 -0.980 -0.836
(-2,0) + 1.27% 1.07% 0.68% 0.63%
*** *** **
4.676 4.073 1.881 1.338*
1.896** 1.640* 0.527 0.544
(0,+2) + 0.68% 0.55% 0.22% 0.21%
1.702** 1.572* -0.316 -0.363
-0.814 -0.885 -1.069 -1.018
Panel B: Enterprise Risk Management Rating Revisions (N=21)
(-2,+2) + 4.51% 4.67% 3.52% 3.24%
*** ***
4.969*** 5.628 3.316 3.169***
2.105** 2.694*** 1.799** 1.690**
(-2,0) + 2.95% 3.03% 2.60% 2.37%
*** *** ***
4.433 4.727 2.457 2.241**
1.227 1.345* 0.894 0.701
(0,+2) + 3.60% 3.78% 2.48% 2.35%
*** *** ***
5.078 6.128 3.789 3.844***
2.273** 3.148*** 2.332** 2.375***

Notes: This table presents results of testing the significance of abnormal market returns around initial S&P ratings (Panel A) and
ratings revisions (Panel B). The initial S&P rating is the first ERM rating assigned by the S&P for a particular firm (weak,
adequate, strong, and excellent). Rating revision is the issuance of a new ERM rating by S&P that results in a change in the
previous ERM rating, for example, an upgrade (downgrade) accepts the value of one (minus one) when ratings are revised. To
align the market reaction to positive and negative news, we multiply the returns in response to low ratings and downgrades by
negative one. All variables are defined in Table 1. Patell and Rank test statistics are shown on the second and third rows, with
significance denoted as ***, **, * for one percent, five percent, and ten percent, respectively.

45
TABLE 5
Results of Estimating Model 4: ERM Quality and the Earnings Response Coefficient

(A) (B) (C) (D)

Sample- Industry- Market- Size-
Panel A: ERM and Exp. Adjusted Adjusted Adjusted Adjusted
Earnings Response Sign Returns Returns Returns Returns
(CAR_SAMP) (CAR_IND) (CAR_MAR) (CAR_SIZE)
UE + 2.6516** 3.2926*** 3.3433*** 2.9821***
(2.17) (3.23) (3.47) (2.70)
UE×STRONG_ERM + 3.6620*** 3.3059*** 3.6876*** 3.9342***
(3.58) (2.86) (2.81) (3.16)
UE×CREDIT-RATING ? -0.0010 -0.0007 -0.0013 -0.0012
(-0.69) (-0.51) (-0.88) (-0.87)
NEG - 0.0165*** 0.0123 0.0092 0.0104
(5.12) (0.94) (0.64) (0.73)
UE×NEG - 1.0970 0.8623 1.0250 0.8706
(0.99) (0.92) (0.91) (0.81)
UE×B/M - -0.4306 -0.5488 -0.4295 -0.3791
(-0.59) (-0.77) (-0.63) (-0.52)
UE×BETA - -0.7447 -0.7982* -0.8094* -0.6901
(-1.21) (-1.41) (-1.40) (-1.14)
UE×COV - -2.1822 -3.3850* -4.1488** -3.6990*
(-0.92) (-1.34) (-1.90) (-1.61)
Constant -0.0057 -0.0081 -0.0024 -0.0034
(-0.43) (-0.59) (-0.18) (-0.25)
Industry and fiscal
Included Included Included Included
quarter fixed-effects
N 544 544 544 544
R2 0.148 0.147 0.175 0.176
Notes: This table presents estimation results of Model 4, investigating whether the market response to company earnings is
greater for high ERMQ companies. The model is estimated with fiscal quarter and industry fixed effects. All variables are
defined in Table 1. Numbers in the cells are coefficients (t-statistics based on standard errors clustered by fiscal quarter and firm),
with significance denoted as ***, **, * for one percent, five percent, and ten percent, respectively. One-tailed tests are presented
for directional expectations.

46
TABLE 6
Results of Estimating Model 5: ERM Quality and Pre-crisis, Crisis and Post-crisis Returns
Pre-Crisis Period Crisis Period Post-Crisis Period
Sample- Industry- Sample- Industry- Sample- Industry-
Adjusted Adjusted Adjusted Adjusted Adjusted Adjusted
Exp. Returns Returns Returns Returns Returns Returns
Sign (BHAR_SAMP) (BHAR-IND) (BHAR_SAMP) (BHAR-IND) (BHAR_SAMP) (BHAR-IND)
ERMQ + -0.0075 -0.0106 -0.0169 -0.0196 0.2152*** 0.2149***
(-0.36) (-0.52) (-0.57) (-0.66) (2.51) (2.47)
CREDIT-RATING ? -0.0069 0.0007 0.0697** 0.0767** 0.0182 0.0083
(-0.21) (0.02) (2.17) (2.29) (0.17) (0.08)
TARP ? 0.7610** 0.7239**
(2.10) (1.93)
BETA + 0.0706* 0.0675* -0.0960** -0.1113** 0.8142** 0.8439**
(1.44) (1.38) (-1.74) (-1.95) (2.11) (2.13)
MARKETCAP - -0.0516*** -0.0519** -0.0664*** -0.0594** -0.0529 -0.0651
(-2.34) (-2.32) (-2.45) (-2.11) (-0.50) (-0.60)
B/M + 0.0840 0.0689 -0.2814*** -0.2475*** 0.5133 0.4756
(0.64) (0.52) (-3.27) (-2.80) (1.18) (1.08)
PRET + 0.4707*** 0.4567*** 0.1039 0.1210 -0.8640 -0.8291
(3.08) (3.00) (0.81) (0.96) (-1.20) (-1.12)
E/P + 0.5013 0.4093 0.0082 0.0034 -0.1812 -0.1796
(0.79) (0.63) (0.05) (0.02) (-0.19) (-0.18)
ln(Leverage) + 0.0127 0.0022 -0.1054*** -0.0930*** 0.2605* 0.2325*
(0.38) (0.07) (-2.93) (-2.57) (1.64) (1.44)
Constant 0.4901*** 0.4741*** 0.9669*** 0.7645*** -3.4225*** -2.9621***
(2.80) (2.65) (4.73) (3.58) (-3.57) (-2.99)
Industry Fixed-Effects Included Included Included Included Included Included
N 101 101 102 102 99 99
Adjusted R2 0.204 0.195 0.273 0.226 0.482 0.484
Notes: This table presents results of estimating Model 5, which investigates the association of buy-and-hold abnormal market returns with ERM program quality in three sub-
periods of the recent financial crisis (2008-09). All variables are defined in Table 1. Numbers in the cells are coefficients (t-statistics based on standard errors clustered by firm),
with significance denoted as ***, **, *, for one percent, five percent, and ten percent, respectively. One-tailed tests are presented for directional expectations.

47