read article and answer question

Contents lists available at ScienceDirect

Journal of Financial Economics

Journal of Financial Economics 98 (2010) 439–461

0304-40

doi:10.1

$ We

Johnson

The co

RiskMe

Univers

Pennsyl

America

fully ac

Rembe

Researc n Corr

E-m

journal homepage: www.elsevier.com/locate/jfec

Rating the ratings: How good are commercial governance ratings?$

Robert M. Daines a,n, Ian D. Gow b, David F. Larcker c

a Law School and Graduate School of Business, Rock Center for Corporate Governance, Stanford University, 558 Nathan Abbot Way, Stanford, CA 94305, USA b Kellogg School of Management, Northwestern University, USA c Graduate School of Business, Rock Center for Corporate Governance, Stanford University, USA

a r t i c l e i n f o

Article history:

Received 21 November 2008

Received in revised form

8 September 2009

Accepted 24 September 2009 Available online 23 June 2010

JEL classification:

G34

Keywords:

Corporate governance

Governance ratings

5X/$ - see front matter & 2010 Elsevier B.V.

016/j.jfineco.2010.06.005

would like to acknowledge the extensive hel

, Ravi Pillai, Michelle Gutman and Justin Han

mments of Abe Friedman, Kevin Murphy,

trics, and Audit Integrity, and workshop partic

ity, New York University, Stanford Univer

vania, University of Southern California, Yale

n Law and Economics Association annual co

knowledged. We would also like to thank th

Rock Center for Corporate Governance and Cor

h Program for financial support.

esponding author. Tel.: + 1 650 736 2684; fax

ail address: [email protected] (R.M. Daine

a b s t r a c t

Proxy advisory and corporate governance rating firms (such as RiskMetrics/Institutional

Shareholder Services, GovernanceMetrics International, and The Corporate Library) play

an increasingly important role in U.S. public markets. They rank the quality of firm

corporate governance, advise shareholders how to vote, and sometimes press for

governance changes. We examine whether commercially available corporate govern-

ance rankings provide useful information for shareholders. Our results suggest that they

do not. Commercial ratings do not predict governance-related outcomes with the

precision or strength necessary to support the bold claims made by most of these firms.

Moreover, we find little or no relation between the governance ratings provided by

RiskMetrics with either their voting recommendations or the actual votes by

shareholders on proxy proposals.

& 2010 Elsevier B.V. All rights reserved.

1. Introduction

Corporate governance advice is big business. RiskMetrics/Institutional Shareholder Services (ISS), the largest advisor, claims over 1,700 institutional clients managing $26 trillion in assets, including 24 of the top 25 mutual funds, 25 of the top 25 asset managers, and 17 of the top 25 public pension funds. ISS was sold in 2007 to RiskMetrics, a firm that has since gone public, for an estimated $550 million. GovernanceMetrics International (GMI) advises clients managing $15 trillion. These

All rights reserved.

p provided by John

sen on this project.

Roberta Romano,

ipants at Columbia

sity, University of

University and the

nference are grate-

e Arthur and Toni

porate Governance

: + 1 650 725 0253.

s).

governance ratings also serve as inputs into tradable indices created by ISS/FTSE Group and Standard & Poor’s (S&P)/Glass, Lewis & Co.

If these ratings identify corporate governance char- acteristics that lead to desirable or undesirable outcomes, users of these ratings may be able to earn superior risk- adjusted returns by either investing in firms with good governance or avoiding firms with poor governance. Governance advisory firms commonly make this claim explicit. ISS claims that its ratings ‘‘identify the worst corporate offenders’’1 and that ‘‘there is no doubt that [its] ratings could have helped some investment managers avoid the gigantic losses experienced during the corporate scandal era defined by meltdowns at Enron, Global Crossing and WorldCom.’’2 Similarly, The Corporate Library (TCL) says its approach ‘‘led to our successfully identifying the Enron, WorldCom, Global Crossing, HealthSouth, Kmart, Warnaco and DPL boards as likely

1 Institutional Shareholder Services, Solutions Overview: http://

www.issproxy.com/pdf/cgq.pdf. 2 ISS Web site: http://www.riskmetrics.com/issgovernance/esg/cgq.

html.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461440

to encounter problems well before those firms imploded, even while most other ratings systems awarded those boards generally high marks.’’3 GMI’s ‘‘premise is simple: companies that focus on corporate governance and transparency will, over time, generate superior returns and economic performance and lower their cost of capital. The opposite is also true: companies weak in corporate governance and transparency represent increased invest- ment risks and result in a higher cost of capital.’’4

These ratings also change firm practices when boards seek to increase their ratings. Aetna and GE reportedly hired ISS to recommend governance changes that would boost their ratings; the implemented changes lifted their ratings from 10% to more than 90%.5 Do such ratings- driven changes lead to better outcomes? The question is broader than the hundreds of similar firms that pay for advice on what they should change. In a recent survey, public firm directors listed corporate governance advisors as the third most influential institution on board, behind only institutional investors and analysts, and ahead of activist hedge funds or shareholder plaintiffs.6 Directors also said that a low governance rating is an important red flag that prompts them to increase their monitoring- falling just behind the firm’s missing analysts’ earnings estimates in importance.

Do these ratings identify important governance effects? Evidence about the value of these ratings and their ability to predict future events or performance is scarce and, when available, sponsored by the commercial companies themselves. If the (somewhat bold) claims of the rating firms are supported by rigorous empirical analysis, these commercial firms are providing valuable information to boards of directors and market partici- pants. However, if ratings have little predictive ability, it is not clear that boards of directors and shareholders should be concerned about their governance ratings or change their firms’ governance practices when pressed to do so by these governance monitors.

The purpose of this paper is to examine the association between the ratings produced by leading commercial corporate governance rating firms and subsequent undesir- able outcomes such as accounting restatements and share- holder litigation, as well as future operating performance, stock returns, and the cost of debt. In particular, we consider CGQ (the ‘‘Corporate Governance Quotient’’ calcu- lated by RiskMetrics/ISS), GMI (a measure of governance quality produced by GovernanceMetrics International), and

3 TCL Web site: http://www.thecorporatelibrary.com/Products-and-Ser

vices/board-effectiveness-ratings.html. 4 GMI Web site: http://www.gmiratings.com. 5 Wall Street Journal, ‘‘Making the grade: Want to lift your firm’s

rating on governance? Buy the test,’’ June 6, 2003. Reportedly hundreds

of other firms pay for similar advice on what they can change. This

potential conflict of interest (charging firms for advice on how to

increase the grades ISS assigns) was the subject of a recent Securities and

Exchange Commission (SEC) investigation. These ratings are also

included on the front page of RiskMetrics/ISS’s influential voting

recommendations to institutional shareholders, suggesting the ratings

may play a role in shareholder voting. 6 PricewaterhouseCoopers, 2008. What Directors Think. The Corpo-

rate Board Member/PricewaterhouseCoopers Survey.

TCL (a rating produced by The Corporate Library). We also examine the performance of AGR, a somewhat different measure of firm accounting practices and governance produced by Audit Integrity.

We find little evidence that the rankings are useful in predicting subsequent accounting restatements or share- holder litigation. While AGR and GMI have statistically significant relations with future restatements and AGR has with future class-action lawsuits, the improvement in prediction of such events from the use of these ratings is very modest. In terms of future performance, AGR predicts future improvements in operating performance, TCL has a positive relation with future Tobin’s Q, and AGR (and to a much lesser extent TCL and CGQ) has a positive relation with future alpha (excess stock price return). None of the ratings are able to predict the subsequent changes in a firm’s cost of debt, as measured by its credit rating. Thus, there seems to be something to AGR’s rating of accounting practices (which are probably best viewed as governance outputs rather than as measures of governance inputs, as explained in Section 3.4 below). However, the predictive ability of the leading commercial governance ratings (CGQ, TCL, and GMI) is well below the threshold necessary to support the bold claims made for them.

One especially interesting result is that CGQ (perhaps the most visible governance rating) exhibits virtually no predictive ability, and when CGQ is significant, more often than not it has an unexpected sign (e.g., higher CGQ seems to be associated with lower Tobin’s Q, and in some models more class-action lawsuits). In unreported regressions, we examine ISS’s ‘‘sub-scores’’ that rate the quality of firms’ audit reviews and boards of directors and find these are typically statistically significant, but again have an unexpected sign (better governance rating yields worse results).

The weak (and mixed) predictive results for CGQ, GMI, and TCL have several interpretations. It is possible that corporate governance is an endogenous choice by firms that optimally adjust the costs and benefits of these governance choices. If so, we should not observe an empirical association between firm performance and the governance ratings (assuming we have the correct func- tional form and the necessary control variables are measured without error). However, we should still observe a relation between governance ratings and out- comes such as restatements and shareholder litigation, as these variables are not net economic outcome variables such as profits or stock price. Moreover, since we find some relation between future operating performance and excess returns and the rating that focuses on accounting practices (AGR), it does not appear that the results for CGQ, GMI, and TCL are completely explained by concerns regarding the optimal endogenous selection of corporate governance.

Another interpretation of our weak and mixed results is that the commercial ratings contain a large amount of measurement error. It is well known that measurement error attenuates the estimated coefficients in simple regressions and produces mixed estimation results de- pending on the covariance structure of the variables included in multivariate regressions. It is instructive to

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 441

note that there is very little correlation among the ratings. This suggests that either the ratings are measuring very different corporate governance constructs or that there is substantial measurement error in at least some of the ratings. Since the firms use the same basic governance data, examine similar governance dimensions (e.g., anti- takeover provisions, board structure, and executive compensation), and claim to measure overall ‘‘corporate governance,’’ we believe that each firm is attempting to measure a similar corporate governance construct. Thus, we believe our results are produced by substantial measurement error in the commercial corporate govern- ance ratings.

The implications of this interpretation extend beyond the merits of these particular ratings. First, our results provide additional, if indirect, evidence regarding the merits of academic indices of corporate governance. If large commercial organizations with substantial expertise and extensive databases cannot devise reliable measures of corporate governance, it seems unlikely that the check- and-sum measures used by academic researchers have significantly better validity. Second, our results suggest a more fundamental issue with regard to the notion of ‘‘corporate governance’’ in general. The fact that experts cannot agree on the measurement of the quality of corporate governance structures highlights the need for future research on developing reliable and valid measures of the construct ‘‘corporate governance.’’ Finally, these results suggest that boards of directors should not implement governance changes solely for the purpose of increasing their ranking.

The remainder of the paper is as follows. Section 2 reviews prior research on corporate governance ratings and commercial corporate governance advisory firms. Section 3 reviews the commercial governance ratings examined in this paper. Section 4 provides descriptive statistics for our ratings data. Section 5 examines whether the ratings are useful in predicting future firm perfor- mance or outcomes of interest to shareholders. Section 6 examines the relation between the CGQ index and proxy recommendations by ISS and actual shareholder voting on proxy proposals. Section 7 concludes.

2. Prior research

There is a vast empirical literature examining the relations between selected corporate governance me- chanisms and firm performance. For example, Morck, Schleifer, and Vishny (1988) consider managerial owner- ship, Daines and Klausner (2001) examine takeover defenses, Fich and Shivdasani (2006) analyze the con- sequences of busy boards, Coles, Naveen, and Naveen (2008) consider board size, and Larcker, Richardson, and Tuna (2007) examine a variety of board and ownership variables and various firm outcomes. This literature yields mixed findings in terms of the relations between such corporate governance measures and firm performance.

Other academic researchers have attempted to com- bine these individual governance elements into a single metric or rating of the overall quality of a firm’s

governance. La Porta, Lopez-de-Silanes, Schleifer, and Vishny (1998) create an index of shareholder protections around the world and find that it correlates with economic growth and market capitalization. Gompers, Ishii, and Metrick (2003) create a governance score, the G-score, composed of mostly anti-takeover items and find that better-governed firms exhibit superior future share- holder returns. Although these academic indices have generated considerable research on the relationship between overall governance and firm performance, their validity is still an open question. For example, Core, Guay, and Rusticus (2006) report evidence that suggests that the G-score is not related to superior firm performance. Finally, Bebchuk, Cohen, and Ferrell (2009) find that an index based on six components of the G-score (which they call the E-index) produces large abnormal returns. How- ever, recent research by Johnson, Moorman, and Sorescu (2009) demonstrates that no abnormal returns are generated using the G-score or E-index when the bench- mark asset-pricing model is adjusted for industry cluster- ing. Thus, the research results linking individual governance indicators or indices to firm performance are quite mixed. Bhagat, Bolton, and Romano (2007) provide an excellent review of the theoretical and empirical issues associated with governance indices created by academic researchers.

Despite the extensive research on academic corporate governance indices, little scrutiny has been given to the governance ratings generated by commercial firms. There are several reasons to suspect that these commercial ratings might provide reliable and valid measures for the construct of corporate governance. First, firms selling ratings appear to be a commercial success, which suggests the possibility that the ratings are useful to their customers. Second, commercial ratings use proprietary, quantitative algorithms that presumably capture their extensive expertise regarding the relationship between governance choices and firm performance. In contrast, academic governance indices are generally calculated by simply counting the number of ‘‘good’’ or ‘‘bad’’ govern- ance mechanisms for each firm. This purposefully naive approach equally weights governance indicators that likely differ in importance and ignores the possibility that some provisions may be substitutes or complements (e.g., Larcker, Richardson, and Tuna, 2007). Third, commercial indices typically rate each firm relative to industry or size peers, whereas academic indices are usually absolute measures constructed without regard to variation in governance practices across industries. Fourth, commer- cial rating algorithms explicitly change each year to ‘‘take into account market trends,’’ whereas most academic ratings are calculated in the same way over time. Finally, commercial firms employ large, rich databases from multiple data sources, whereas typical academic govern- ance indices rely on relatively limited data sources such as the Investor Responsibility Research Center, Inc. (IRRC) data, which are heavily focused on takeover defenses.

A small number of prior studies have examined ISS ratings or their inputs. Brown and Caylor (2004) report univariate results for one year suggesting that high CGQ scores are associated with higher current stock returns,

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461442

higher accounting returns, lower volatility, and higher dividends. However, this analysis is backward-looking and provides no evidence on the ability of CGQ to predict future firm outcomes. Brown and Caylor (2006) examine the relationship between Tobin’s Q and an index created from 51 governance variables collected by ISS (and identified as important elements of ISS’s rating). Their index is simply the sum of a variety of indicator variables that ISS considers consistent with good governance. They find that their own index is significantly related to contempora- neous Tobin’s Q for 2002, but do not report findings for the CGQ rating. Aggarwal and Williamson (2006) use ISS data to examine the relationship between firm value and 64 governance-related inputs to the ratings, but again do not examine the primary CGQ rating. Finally, Koehn and Ueng (2005) examine a sample of 106 large U.S. firms and find no statistically significant relationship between the CGQ scores and Audit Integrity’s measure of earnings quality.

Ashbaugh-Skaife and LaFond (2006) examine whether GMI’s governance ratings are related to cost of equity capital in research sponsored by GMI. In an executive summary of their findings, the authors report that higher GMI governance ratings were associated with lower cost of equity capital in 2004 and conclude that ‘‘GMI ratings are valid assessments of the strength (or weakness) of U.S. and non-U.S. firms’ governance.’’ However, they do not report whether current ratings predict future cost of capital. Similarly, Derwall and Verwijmeren (2007) find that GMI governance ratings for 2005 have a contemporaneous negative association with cost of equity capital and firm-specific and systematic risk.

Finally, Bhagat, Bolton, and Romano (2007) examine several ratings from TCL. Using multivariate analysis and simultaneous equations, they report mixed evidence about its ability to predict future operating performance and share price appreciation. To our knowledge, there is no third-party research examining the ability of Audit Integrity’s AGR to predict future performance outcomes.

In summary, there have been very few studies about whether commercial corporate governance ratings predict firm outcomes. Extant studies are generally backward- looking and thus, provide no evidence that the ratings predict future outcomes as they are frequently claimed to do. Moreover, a common comparative analysis for the three major commercial governance ratings has not been conducted. Given their practical importance, we conduct such an analysis in this paper.

7 After earlier drafts of this paper became public, RiskMetrics

announced that they would retire CGQ and instead provide 4 different

‘‘GRId’’ scores rating each firm’s board structure, shareholder rights,

compensation and audit. Though the name is different, the scores appear

to be based on similar factors and data as CGQ and its related sub-scores

which we find are often statistically significant but with an unexpected

sign (i.e. better scores are associated with worse outcomes). 8 Institutional Shareholder Services, 2003. ISS Corporate Govern-

ance: Best Practices User Guide & Glossary, Revision 2.4, October 8, 2003. 9 ISS Web site: http://www.isscgq.com/cgqratings.htm. 10 CGQ Corporate Governance Fact Sheet—November 3, 2006. 11 This sub-section was adapted from material found at http://www.

gmiratings.com. 12 GovernanceMetrics International, September 2006, Governance

and Performance: Recent Evidence.

3. Commercial corporate governance ratings

In this study, we evaluate governance ratings from three primary corporate governance rating firms: ISS Corporate Governance Quotient (CGQ), GovernanceMetrics International (GMI), and The Corporate Library’s TCL rating. As we describe below, these ratings differ in terms of focus, approach, and sample coverage, but each attempts to evaluate the corporate governance of a wide cross- section of public firms. We also examine the rankings produced by Audit Integrity (AGR), which differ in that they focus primarily on the risk of certain accounting and financial statement practices (governance outputs).

3.1. Institutional shareholder services’ CGQ ratings

The CGQ rating is produced by Institutional Shareholder Services (ISS), a division of RiskMetrics.7 The rating ‘‘evaluates the strengths, deficiencies and overall quality of a company’s corporate governance practices and board of directors’’ and ‘‘is designed on the premise that good corporate governance ultimately results in increased shareholder value.’’8 ISS reports two main ratings for each firm: CGQ_INDUSTRY, which gives a firm’s percentile standing within its Global Industry Classification Standard (GICS) industry group, and CGQ_INDEX, which gives a firm’s percentile within its index (e.g., S&P 500 for Microsoft). We focus on CGQ_INDUSTRY (hereafter simply CGQ), but virtually identical results are obtained when our analyses are conducted using CGQ_INDEX.

ISS ratings are based on data taken from public filings and company surveys in eight categories: board of directors (composition, independence), audit, charter, and bylaw provisions, anti-takeover provisions, executive and director compensation, progressive practices (such as performance reviews and succession plans), ownership, and director education. ISS conducts ‘‘more than 4,000’’ statistical tests using 16 measures of risk and performance to develop the optimal weighting of 64 governance variables in CGQ according to their correlation with firm risk and prior performance.9 The ratings were then back-tested and calculated for more than 9,000 companies. In addition, ISS states that it changes the ratings model and weights over time to ‘‘better reflect current market trends in corporate governance’’ and to align the rankings with ISS policies.10

ISS claims CGQ is a ‘‘reliable tool for identifying portfolio risk related to governance and leveraging governance to drive increased shareholder value’’ and emphasizes claims of a ‘‘very strong relationship between governance and firm value, using CGQ data.’’

3.2. GovernanceMetrics International’s GMI ratings11

GMI was ‘‘founded on the premise that the quality of corporate governance can add significantly to the risk- reward profile of credit and investment portfolios.’’12 GMI collects data on several hundred governance mechanisms (ranging from compensation to takeover defenses and board membership), as well as on firms’ compliance with securities regulations, stock exchange listing require- ments, and various corporate governance codes and

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 443

principles. In all, it collects ‘‘hundreds of metrics struc- tured in a manner that can only produce yes, no or not disclosed answers.’’ GMI develops a scoring model that examines each metric, weights it ‘‘according to investor interest,’’ and then calculates a rating on a scale of 1.0 (lowest) to 10.0 (highest). The GMI scoring algorithm rewards (or penalizes) ‘‘outliers’’ and ranks each firm relative to the other companies in the GMI sample. The GMI ratings are calculated for over 4,100 companies.

GMI says its ‘‘scoring algorithm has also been tested and validated by outside statistical experts and is patent pending.’’ Its materials tout the fact that ‘‘companies that emphasize corporate governance and transparency will, over time, generate superior returns and economic perfor- mance and lower their cost of capital,’’ suggesting that firms with high GMI scores will ‘‘generate superior returns.’’

3.3. The corporate library’s TCL ratings

Where the other ratings are the product of proprietary quantitative analysis, TCL’s ratings reflect subjective judg- ment and expertise. TCL analysts avoid data checklists and rely instead on their experience and private assessment of a firm’s governance quality. TCL analysts review four specific areas (the company’s board and succession planning, CEO compensation practices, takeover defenses, and board-level accounting concerns) and assign each firm a ‘‘grade’’ from A to F. Companies rated A or B do not exhibit significant risk in any of the four basic categories; C-rated companies exhibit risk in no more than one category; D-rated companies in two or more categories; and F-rated compa- nies were either bankrupt, delisted from an exchange, or described as companies ‘‘where management has achieved effective control over the company y and conducts its business with flagrant disregard for the interest of any minority public shareholders.’’ The analysts focus on ‘‘‘red flag’ indicators of board ineffectiveness and corporate mismanagement, supported by in-depth analysis and commentary by our senior research associates and analysts.’’

According to TCL, its ratings ‘‘have been proven to predict losses in shareholder value and the occurrence of securities class-action lawsuits’’13 and ‘‘have been tested against actual investment returns.’’14

3.4. Audit integrity’s AGR ratings15

In contrast to the three governance rankings described above, the Audit Integrity’s Accounting and Governance

13 http://www.thecorporatelibrary.com/UserFiles/Board_Analyst0907.

pdf. 14 http://www.thecorporatelibrary.com/info.php?id=53. 15 This section is a summary of the information provided by Audit

Integrity consisting of: (http://www.auditintegrity.com/documents/Audi

t_Integrity_Summary_Corp.pdf); Audit Integrity white paper, The Audit

Integrity AGR Model: Measuring Accounting and Governance Risks in

Public Companies (June 27, 2005), available at http://www.auditintegr

ity.com/documents/Audit_Integrity_AGR_White_Paper.pdf; The Audit

Integrity Multi-Factor Restatement Model: A Leading Indicator of

Financial Restatement (April 11, 2006), available at http://www.

auditintegrity.com/documents/Audit_Integrity_Restatement_White_Pa

per.pdf.

Risk (AGR) ranking is primarily focused on accounting practices. Audit Integrity examines 200 accounting and governance metrics and 3,500 variables designed to produce ‘‘an assessment of financial statement risk—the risk that financial statements do not accurately reflect a company’s true financial condition due to fraud or misrepresentation.’’ AGR seeks to identify ‘‘fraudulent patterns of behavior.’’ Although it is thus focused less on governance as an input, and more on trying to assess the quality of the firm’s financials as an ‘‘output’’ of govern- ance, we nevertheless examine AGR in part because it too includes some governance measures and also because, as will be seen, it provides a useful benchmark for the other three governance rankings.

AGR scores range from 0 to 100, corresponding to ‘‘Very Aggressive’’ (approximately 10% of all firms) to ‘‘Conservative’’ (approximately 15% of all firms). The ratings are objective and mechanical in that they are produced by statistical examination of financial data (such as changes and trends in revenue recognition variables) ‘‘without preconceived bias as to what defines fraud.’’ AGR scores are calculated for over 9,000 publicly traded companies.

Audit Integrity claims that its measure has been verified in ‘‘study after study’’ and that high-risk firms are more likely to be sued, to restate financials, to suffer large drops in share value, and earn lower returns.16 Its Web site claims that its ratings offer users the ability to ‘‘achieve excess returns,’’ ‘‘avoid companies at a high risk of litigation,’’ and ‘‘a great deal of predictive power concerning future corporate problems.’’

4. Governance ratings: data and descriptive statistics

Corporate governance ratings were compiled for U.S. firms from each of the four commercial rating services from a variety of public sources, research services, or from the advisory firms themselves. While we have data for each of the four ratings from late 2005 through to early 2007, most of our analysis focuses on the ratings available on December 31, 2005.17 Our primary sample consists of 2005 CGQ rankings for 5,059 firms, GMI rankings for 1,565 firms, TCL for 1,906 firms, and AGR rankings for 6,714 firms. There are 6,827 unique firms across the four commercial ratings. These sample sizes are consistent with the reported coverage universe for U.S. firms for these rating firms. Our sample also spans many economic sectors and closely mimics the industry distribution in Compustat (Table 1, Panel C).

As discussed in Section 3 above, the distribution of scores differs substantially for each governance rating (see Fig. 1). Since CGQ is expressed as a percentile, it is not surprising that it is approximately uniform between

16 Audit Integrity is careful to note that ‘‘behavior that matches past

patterns of fraud is not a guarantee of current fraudulent or misleading

behavior.’’ 17 We also have data for CGQ and GMI for earlier periods. The results

produced from this longer time series are discussed in later sections.

Analyses using 2006 ratings (and thus shorter periods for the outcome

variables) yield very similar results to those we provide here.

Table 1 Summary of governance ratings.

Panel A provides summary statistics for primary governance ratings of Audit Integrity (AGR), RiskMetrics/ISS (CGQ), GovernanceMetrics International

(GMI), and The Corporate Library (TCL). Panel B provides correlation statistics for the primary governance ratings. Pearson (Spearman) correlations

between governance ratings are presented above (below) the diagonal. Numbers on the diagonal represent correlation between 2005 and 2006 ratings for

firms in our sample. Panel C provides the percentage of each rating sample in each of 24 Global Industrial Classification System groups. AGR, CGQ, and GMI

are on a 0�100 scale. TCL is converted from an ‘‘A’’ to ‘‘F’’ grade to numerical values 1�5, where ‘‘A’’ equals 5 and ‘‘F’’ equals 1 (no ‘‘E’’). The governance

ratings are measured as of December 31, 2005.

Panel A: Descriptive statistics

Variable N Mean SD Min P25 Median P75 Max

Primary ratings

AGR 6,714 63.67 15.18 4.0 54.0 67.0 75.0 88.0

CGQ 5,059 51.61 28.50 0.4 27.1 52.0 76.2 100.0

GMI 1,565 7.08 1.22 2.5 6.5 7.0 8.0 10.0

TCL 1,906 3.22 0.90 1.0 3.0 3.0 4.0 5.0

Panel B: Correlation coefficients

AGR CGQ GMI TCL

AGR (0.558n) 0.005 0.031 0.063n

CGQ 0.029n (0.847)n 0.480n 0.005

GMI 0.048 0.484n (0.817)n �0.020

TCL 0.076n 0.016 �0.009 (0.613n)

Panel C: Industry composition (% of sample)

AGR CGQ GMI TCL Compustat

Energy 4.42 4.40 4.81 4.52 6.06

Materials 4.72 4.71 6.36 6.34 6.31

Capital goods 7.65 7.65 8.18 7.99 6.87

Commercial & professional services 3.33 3.25 3.25 3.25 3.37

Transportation 1.95 1.94 2.14 2.48 1.99

Automobiles & components 1.48 1.52 1.62 1.87 1.43

Consumer durables & apparel 4.29 4.40 4.61 4.80 3.78

Consumer services 3.56 3.63 4.03 4.08 3.40

Media 3.07 3.05 3.18 3.36 3.13

Retailing 5.21 4.98 6.04 6.39 3.77

Food & staples retailing 0.92 0.93 1.17 1.27 0.79

Food, beverage & tobacco 2.34 2.42 2.99 3.03 2.27

Household & personal products 0.76 0.80 0.84 0.99 0.91

Health care equipment & services 7.62 7.62 7.40 6.84 6.82

Pharma, biotech & life sciences 7.42 7.20 3.44 3.09 6.17

Banks 7.52 7.75 5.52 5.62 9.28

Diversified financials 2.57 2.53 3.18 3.03 2.84

Insurance 2.87 2.80 3.64 3.36 2.59

Real estate 4.16 4.29 3.64 3.53 3.21

Software & services 8.31 8.34 6.30 7.06 9.19

Technology hardware & equipment 6.99 6.99 7.34 7.00 7.24

Semiconductors (inc. Equipment) 3.99 3.95 4.16 4.08 2.67

Telecommunication services 1.72 1.63 1.17 1.32 2.66

Utilities 3.13 3.22 5.00 4.69 3.26

n Indicates statistically significant correlation at the 5% level (two-tail).

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461444

0 and 100. However, AGR and TCL have noticeable negative skewness, with many firms clustering at relatively high scores and a smaller number of firms forming a long tail to the left. The GMI scores are relatively symmetric. Clearly, AGR, CGQ, and TCL are not directly comparable even though each takes values between zero and 100.

Since the commercial firms use the same basic governance data, examine similar governance dimensions (e.g., anti-takeover provisions, board structure, and ex- ecutive compensation), and all claim to measure overall ‘‘corporate governance,’’ we would expect their ratings to be highly correlated.

However, one key finding is that, as illustrated in Panel B of Table 1, these four ratings are close to being

uncorrelated, with the exception of GMI and CGQ, which have a Pearson (Spearman) correlation of 0.484 (0.480). The Pearson (Spearman) correlations among the remain- ing five pairs range from –0.009 to 0.076 (–0.020 to 0.063). AGR in particular seems uncorrelated with most of the other ratings in our data set.

Consistent with this lack of correlation, many large firms with substantial investor followings and long track records receive wildly disparate grades from the various services: AT&T, General Electric, General Motors, and Safeway received nearly perfect scores from one rating firm (a 99 or 100 from ISS) and near-failing grades from another (a D from TCL). Notably, these firms are not obscure start-ups where appropriate governance

D en si ty

0 20 40 60 80 AGR

0

0.005

0.01

0.015

D en si ty

0 20 40 60 80 100 CGQ_INDUSTRY

0

0.01

0.02

0.03

0.04

0

0.1

0.2

0.3

0.4

D en si ty

2 4 6 8 10 GMI

0

0.1

0.2

0.3

0.4

D en si ty

0 1 2 3 4 5 TCL_RATING

Fig. 1. Distribution of governance ratings. Figure depicts histograms for primary governance ratings of Audit Integrity (AGR, 6,714 observations), RiskMetrics/ISS (CGQ, 5059 observations), GovernanceMetrics International (GMI, 1,565 observations), and The Corporate Library (TCL, 1,906

observations). AGR, CGQ, and GMI are on a 0�100 scale. TCL is converted from an ‘‘A’’ to ‘‘F’’ grade to numerical values 1�5, where ‘‘A’’ equals 5 and

‘‘F’’ equals 1 (no ‘‘E’’). The governance ratings are measured as of December 31, 2005.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 445

arrangements and practices are less standard. The fact that the governance ratings are so uncorrelated suggests substantial measurement error in one or more of the commercial ratings.

Finally, as might be expected, the ratings are positively correlated over time. Prior studies (e.g., Gompers, Ishii, and Metrick, 2003) have found that firm governance variables are quite stable over time. The correlations between 2005 and 2006 ratings range from 0.558 for AGR to 0.847 for CGQ.

5. Predictive ability of governance ratings

We evaluate the ratings by examining their ability to predict five important outcomes. These outcome variables are selected because either one or more of the rating firms claims that the ratings will predict the outcome or because prior literature has suggested a relation- ship between the outcome and quality of corporate governance. The first two outcomes, accounting restatements and class-action lawsuits, are relatively rare ‘‘bad’’ outcomes that one or more of the ratings should be expected to predict. We examine three outcomes that are traditional measures of corporate performance, namely accounting operating performance, Tobin’s Q, and excess stock returns (or alpha). Finally,

we examine the relationship between the governance ratings and cost of debt, which is affected by the quality of firm governance (Cremers, Nair, and Wei, 2007).

Our basic approach is to estimate a regression for each outcome variable on the ratings and perhaps a set of controls. Our analysis is conducted both with and without additional control variables (e.g., Larcker, Richardson, and Tuna, 2007). In assessing the impact of governance quality on economic outcomes, it may make sense to exclude the control variables to the extent that govern- ance quality affects the outcomes through its effect on the controls. For example, governance quality may affect the likelihood of restatements both directly and indirectly through its effect on a firm’s book-to-market ratio. Under this scenario, including the book-to-market ratio as a control will cause us to underestimate the total effect (direct and indirect) of governance quality on the like- lihood of accounting restatements. Similarly, to the extent that governance quality is persistent over time, it may affect operating performance in any given period both directly and indirectly through an impact on the prior period’s operating performance that persists over time. Thus, it is unclear whether the analysis should incorporate these control variables (a ‘‘conditional’’ analysis) or exclude them (an ‘‘unconditional’’ analysis). We therefore include both analyses.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461446

In addition, we aim not only to understand the relationships between governance quality (as measured by the ratings) and various economic outcomes, but also to assess the value of the ratings as predictive tools in their own right. In doing this, it is unclear whether the governance ratings are constructed to already capture the effect of the control variables (in which case, ‘‘uncondi- tional’’ analysis excluding these controls is appropriate) or the ratings are constructed from inputs distinct from the control variables (in which case, ‘‘conditional’’ analysis with the effects of the control variables included is appropriate). Thus, we perform each of our analyses of the outcome variable both with and without control variables to provide a more comprehensive analysis of the relationship between governance quality and economic outcomes.

In examining the predictive ability of the governance ratings, we focus on ratings available as of December 31, 2005, as this is the earliest point at which we have a sizable cross-section of ratings across the four rating firms. To facilitate the interpretation of the regression coefficients across ratings, we standardize each of the ratings to have zero mean and unit variance.

20 Consistent with most prior research, we winsorize our data to

eliminate unusual observations. Our results are substantially the same

without winsorization, but with somewhat less statistical significance. 21 ISS also provides CGQ sub-scores covering particular areas of focus,

for instance: ‘‘audit review’’ (CGQ_AUDIT), ‘‘board of directors’’

(CGQ_BOARD), ‘‘executive and director compensation and ownership’’

(CGQ_COMP), and ‘‘takeover defenses’’ (CGQ_TKOVER). Similarly, TCL

provides sub-scores, such as TCL_BP (percentage of certain ‘‘best practices’’

adopted by a company), TCL_ACCTG (‘‘accounting and auditing’’),

5.1. Accounting restatements

It is often claimed that accounting restatements are either evidence of, or caused by, weak governance. A number of papers predict that accounting restatements will be positively associated with poor governance and find support for this prediction (Farber, 2005; Beasley, 1996; Peng and Röell, 2008; Erickson, Hanlon, and Maydew, 2006). However, other papers find little evi- dence of a relationship between accounting restatements and governance (Larcker, Richardson, and Tuna, 2007). We expect that if the ratings predict restatements, higher ratings will be associated with fewer restate- ments.18

We obtain data on accounting restatements from Glass, Lewis & Co., which maintains a comprehensive database on restatement information obtained from SEC filings, press releases, and other public data. We focus on the indicator variable Earnings restatement, which takes the value of one for a firm Glass, Lewis & Co. identifies as making one or more accounting restatements in the period January 2006–December 2008 that relate to either revenue or expense recognition and affect fiscal years 2004–2008.19 Glass-Lewis identifies 595 such restate- ments, representing about 8.72% of our sample of 6,827 firms across the four ratings.

18 While the common assumption of a negative relationship

between governance quality and accounting restatements seems

intuitively plausible, it is important to note that accounting restate-

ments imply a minimal level of governance in that the mechanisms to

detect misstated financial statements must be in place for the

restatements to be observed by the researcher. 19 We exclude restatements affecting fiscal years 2003 and earlier

on the premise that these restatements are less likely to be of relevance

to current shareholders.

For each governance rating, we estimate logistic regressions with Earnings restatement as the dependent variable and either just the governance rating in question (unconditional analysis) or the rating and controls (con- ditional analysis) as independent variables. Based on the extensive research on restatements, we include the following controls: Leverage is calculated as the ratio of book value of debt (Compustat DLTT + DLC) to market value of common equity (PRCC_F�CSHO), and BM is the book value of common equity (CEQ) divided by the market value of common equity. Free cash flow is measured as the difference between operating cash flows (OANCF) and average capital expenditures over the prior three years (CAPX). External financing is total net external financing from debtholders and shareholders during the fiscal period. Acquisitions is cash spent on acquisitions (AQC) divided by market value of common equity. Log(Market value) is the log of market value of common equity. All control variables are measured in the latest fiscal year ending on or prior to September 30, 2005, allowing at least a three-month lag prior to the period over which we capture restatements so as to be confident that the controls are observable prior to this period. All controls are winsorized at the 2nd and 98th percentiles by fiscal year.20

Table 2 presents the results of this analysis. Two of the four primary ratings (CGQ and TCL) are not associated with restatements either conditionally or unconditionally. However, AGR and GMI exhibit a statistically significantly negative association with restatements, implying that higher ratings are associated with fewer future restatements.21 These results are robust to inclusion of controls (conditional analysis).

Assessing the economic (or substantive), rather than statistical, significance of these results in Table 2 is problematic in the absence of information about the loss function associated with Type I and Type II errors in predicting restatements. In order to provide some insight, we examine the ability of the ratings to improve the actual classification of outcomes. We focus on GMI, as this rating appears to have the greatest predictive (explana- tory) power for restatements in our sample.

TCL_BOARD (‘‘board composition’’), TCL_COMP (‘‘CEO compensation’’),

TCL_TKOVER (‘‘board effectiveness and shareholder friendliness in the area

of takeover defenses’’). TCL sub-scores take values of ‘‘very high concern,’’

‘‘high concern,’’ and ‘‘low concern,’’ which are re-coded as 1, 2, and 3,

respectively. In untabulated analysis, we examined the sub-scores from

ISS and TCL and find that most are not associated with restatements.

TCL_ACCTG is statistically associated with restatements, consistent

with the notion that this accounting-focused sub-score is somewhat

useful in predicting accounting problems. However, there are no

statistically significant results for the accounting-focused sub-score

CGQ_AUDIT.

Table 2 Governance ratings and future restatements.

This table reports the results of logit regressions where the dependent variable equals one if, in the three years after December 31, 2005, the firm

restates revenues or expenses for fiscal years 2004 or later, zero otherwise (data on restatements are obtained from Glass, Lewis & Co.). Numbers in

parentheses are standard errors clustered by two-digit SIC codes.

The independent variables included are a constant (unconditional analysis) or a constant plus controls (conditional analysis). Following Larcker,

Richardson, and Tuna (2007), the controls used are debt-to-market (Leverage), book-to-market (BM), External financing, log of market capitalization (Log

market value), cash spent on acquisitions (Acquisitions) and Free cash flow. All controls are measured for the latest fiscal year ending on or before

September 30, 2005. The governance rating variables are the AGR from Audit Integrity, CGQ from RiskMetrics/ISS, GMI from GovernanceMetrics

International, and TCL from The Corporate Library.

The governance ratings are measured as of December 31, 2005. Each governance rating is standardized to have a mean of zero and standard deviation of

one.

Panel A: Primary governance ratings, unconditional analysis

AGR CGQ GMI TCL

Governance rating �0.264** �0.0421 �0.383** �0.0164 (0.0469) (0.0777) (0.0994) (0.0828)

Constant �2.350** �2.319** �2.325** �2.205**

(0.130) (0.143) (0.202) (0.188)

Observations 6,554 5,003 1,565 1,903

Pseudo R2 0.0103 0.000238 0.0195 o0.0001

Panel B: Descriptive statistics, controls for conditional analysis

Variable N Mean Std. dev. P25 Median P75

Leverage 6,645 0.27 0.38 0.02 0.16 0.35

BM 6,551 0.47 0.36 0.21 0.41 0.65

External financing 6,551 0.02 0.25 �0.07 �0.01 0.07

Log market value 6,551 5.48 2.43 3.8 5.49 7.17

Acquisition 6,551 0.02 0.06 0 0 0

Free cash flow 6,518 �0.13 1.94 �0.06 0.03 0.07

Panel C: Primary governance ratings, conditional analysis

AGR CGQ GMI TCL

Governance rating �0.249** �0.0430 �0.352** �0.0709 (0.0451) (0.0829) (0.111) (0.0817)

Leverage �0.00821 �0.00900 �1.218 �0.402

(0.162) (0.271) (0.640) (0.471)

BM �0.420* �0.530** �0.304 �0.699*

(0.198) (0.168) (0.338) (0.334)

External financing �0.352 �0.348 �1.050 �0.298

(0.180) (0.207) (0.568) (0.455)

Log market value �0.0313 �0.0230 �0.0490 �0.190**

(0.0274) (0.0354) (0.0756) (0.0553)

Acquisition 1.557 2.416* 2.320 0.745

(0.924) (0.991) (2.048) (1.983)

Free cash flow 0.148 0.0906 �1.510* 0.551

(0.104) (0.0527) (0.703) (0.467)

Constant �2.005** �1.986** �1.605** �0.436

(0.152) (0.183) (0.596) (0.418)

Observations 6,416 4,979 1,562 1,897

Pseudo R2 0.0149 0.0071 0.0325 0.0128

*, ** Indicates significance at the 5% and 1% levels (two-tail), respectively.

22 It is important to note that this statistic implies that investors are

equally concerned about both kinds of classification errors, whereas they

may care more about reducing false negatives (i.e., owning stock in a firm

that later experienced a restatement) than the false positives (missing out

on firms predicted to restate that did not). Note that simply classifying all

firms as ‘‘not restate’’ would be correct for 90.59% ( = 1415–1562) of firms.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 447

For the 1,562 firms with GMI ratings and data to calculate our controls, 147 experienced a restatement in our test period. We estimate the predicted probability of restatement using the controls alone and setting a probability cutoff for predicting a lawsuit at 0.1; the estimated model classifies 574 firms into the ‘‘restate’’ category (but only 73 of these actually exhibit a restate- ment). Moreover, 74 firms that did restate are incorrectly classified as ‘‘not restate.’’ When we include GMI as an additional explanatory variable, 91 firms are correctly classified as restating firms, an improvement over the model with controls alone, and fewer firms that do not restate are misclassified—487 versus 501. Consistent with

this, the percent correctly classified increases from 63.19% to 65.24% with the inclusion of GMI.22 Thus, there is a modest within-sample improvement when GMI is included in the logistic regression. There is no evidence that other ratings could predict restatements in a meaningful way.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461448

We examine the sensitivity of these results to a number of variations. To allow for the possibility that governance only affects outcomes at the extremes (where firms have either very poor or very good governance), we run our analysis with the standardized governance ratings replaced by two indicator variables representing member- ship of the top or bottom deciles for each rating (if the rating does not allow for partitioning into deciles, we use the top and bottom category instead). We then examine the difference between the coefficients on these indica- tors. Statistically significant differences appear for pre- cisely the same ratings, with the same signs, as in our primary analysis. Given the differences in the sample size and composition across the ratings, we also perform analysis using a common sample across the ratings (there are 1,505 firms with ratings from each of the four rating firms and 1,504 meet the data requirements for our conditional analysis). Our inferences are identical in this case with one exception: CGQ becomes statistically significant at the 5% level in the predicted direction in both the unconditional and conditional analyses. We also allow for the possibility that many of the restatements in our sample are ‘‘innocuous’’ by including only those restatements associated with a negative return of 3% or more over either a three- or five-day window around the announcement of the restatement. Our results for this subset of observations are very similar to those reported in Table 2.

We also examine the relationship between changes in governance ratings and restatements. In this analysis, we find some evidence that improvements in CGQ and TCL (but not AGR and GMI) over the period from December 31, 2005 to June 30, 2006 are associated with lower probability of restatements occurring between July 2006 and December 2008. But neither relationship is robust to the inclusion of controls. Finally, the inclusion of industry fixed effects has virtually no impact on our inferences.

5.2. Class-action lawsuits

The second outcome we consider is whether the firm was the subject of a class-action lawsuit. Woodruff- Sawyer, an insurance brokerage and consulting firm, identifies 338 firms within our sample that are the subject of a class-action lawsuit from December 31, 2005 to June 30, 2009, representing 4.95% of our sample.23 We set the variable Lawsuit equal to one for these firms, and zero for the remaining firms. We again perform logistic regres- sions with Lawsuit as the dependent variable and either the governance rating in question (unconditional analysis) or the rating and controls (conditional analysis) as independent variables. We follow Rogers and Stocken (2005) in our list of controls: Size (log of market value of equity), Turnover (average daily turnover divided

23 We do not find any statistical difference in the overall frequency

of lawsuits between the overall sample and the AGR and CGQ samples,

but we do find a higher rate of lawsuits in the GMI (9.90%) and TCL

(9.22%) samples.

by average shares outstanding), Beta (the slope coeffi- cient from a regression of daily returns on the CRSP value-weighted index), Returns (buy-and-hold returns), Std dev(Returns) (standard deviation of daily returns), Skewness(Returns) (skewness of daily returns), and Min (Returns) (minimum value of daily returns). All controls are obtained from the Center for Research in Security Prices (CRSP) and measured over the year ending December 31, 2005.

Table 3 presents the results of our analysis. Higher AGR scores are statistically significantly associated with fewer future class-action lawsuits, both unconditionally and conditionally. Higher CGQ (TCL) scores are associated with more (fewer) lawsuits unconditionally, but this association disappears when the controls are included. GMI is not statistically associated with lawsuits unless the controls are included. Thus, other than AGR, there is no robust association between governance ratings and future shareholder suits.

To assess the economic significance of our findings, we use a similar approach to that used for restatements. In this case, we focus on AGR, as this seems to have the greatest power to predict lawsuits in our sample. Of the 5,304 firms with available data, 326 experience class-action lawsuits over the period we study. With a probability cutoff of 0.1, including AGR in the logistic model raises the number of firms with lawsuits that are correctly classified from 130 to 140, but at the expense of misclassifying 758 (an increase from 712) firms that do not have lawsuits. The reduction in ‘‘percent correctly classified’’ is from 82.88% to 82.20%. However, if the cost of misclassifying firms experiencing lawsuits is at least five times as great as that of misclassifying firms with no lawsuits, there is possibly an economic benefit from including AGR, as (140�130)�5�(758�712)40. But it should be recognized that this classification analysis is within the sample used for estimation, and thus likely represents an upper bound on the ability of AGR to predict ‘‘out-of-sample’’ observations.

As with the restatement analysis, we examine the effect of using indicator variables for the top and bottom category. In this analysis, only AGR has statistical significant differences between the coefficients on the two indicator variables in both the unconditional and conditional analyses. Looking at a common sample yields identical inferences (with lower significance levels in some cases). Focusing on changes in ratings from December 2005 to June 2006 yields consistently weaker results, with only TCL having a relationship and then only in the unconditional analysis.

5.3. Future operating performance

Following prior research (e.g., Gompers, Ishii, and Metrick, 2003), we assess whether governance ratings predict future operating performance by examining return on assets (ROA), measured as operating income (Compustat item OIADP) divided by average total assets (AT). Larcker, Richardson, and Tuna (2007) use only Log (Market value) and median industry ROA as controls

Table 3 Governance ratings and future class-action lawsuits.

Results are for logit regressions where the dependent variable (Lawsuit) equals one if a class-action lawsuit is filed against the firm after December 31,

2005 and before June 30, 2009 (the latest date on the Woodruff-Sawyer database of corporate litigation), zero otherwise. The independent variables are

the indicated governance rating as of December 31, 2005 and either a constant (unconditional analysis) or controls (conditional analysis). Numbers in

parentheses are standard errors.

Following Rogers and Stocken (2005), we use the following controls in Panel B: the natural log of the average market value of equity (Size), average

daily turnover divided by average shares outstanding (Turnover), the slope coefficient from a regression of daily returns on the CRSP value-weighted index

(Beta), buy-and-hold returns (Returns), the standard deviation, skewness, and minimum value of daily returns (Std dev(Returns), Skewness(Returns),

Min(Returns), respectively), and indicators for membership of the following industry groups: Biotechnology, Computer hardware, Electronic, Retailing,

and Computer software. All controls are measured over the year ending December 31, 2005.

The governance rating variables are the AGR from Audit Integrity, CGQ from RiskMetrics/ISS, GMI from GovernanceMetrics International, and TCL from

The Corporate Library. The governance ratings are measured as of as of December 31, 2005. Each governance rating is standardized to have a mean of zero

and standard deviation of one.

Panel A: Unconditional analysis

AGR CGQ GMI TCL

Governance rating �0.319** 0.551** 0.00689 �0.206** (0.0508) (0.0678) (0.0846) (0.0769)

Constant �2.989** �2.987** �2.208** �2.303**

(0.0583) (0.0699) (0.0846) (0.0804)

Observations 6,714 5,059 1,565 1,906

Pseudo R2 0.0140 0.0337 o0.0001 0.0061

Panel B: Descriptive statistics, controls for conditional analysis

Variable N Mean Std. dev. P25 Median P75

Turnover 5,329 0.0076 0.0156 0.0023 0.0048 0.0089

Size 5,329 19.77 1.91 18.43 19.66 20.99

Returns 5,329 0.11 0.54 �0.15 0.05 0.27

Std dev(Returns) 5,329 0.03 0.02 0.02 0.02 0.03

Min(Returns) 5,329 �0.10 0.08 �0.13 �0.08 �0.05

Beta 5,329 0.92 0.69 0.46 0.94 1.36

Skewness(Returns) 5,329 0.40 1.53 �0.13 0.30 0.84

Panel C: Conditional analysis

AGR CGQ GMI TCL

Governance rating �0.367** 0.0155 �0.211* �0.104 (0.0587) (0.0825) (0.0935) (0.0838)

Turnover 5.125 15.04** 37.71* 37.61**

(2.920) (4.496) (15.21) (11.80)

Size 0.435** 0.445** 0.625** 0.450**

(0.0381) (0.0456) (0.0758) (0.0658)

Returns 0.252** 0.244* �0.300 0.121

(0.0871) (0.0969) (0.280) (0.224)

Std dev(Returns) 6.987 �1.532 89.92** 41.24*

(8.730) (11.10) (32.33) (19.15)

Min(Returns) �1.635 �2.540 9.269* 4.861

(1.435) (1.660) (3.846) (3.004)

Beta 0.132 0.210* �0.112 �0.0280

(0.107) (0.125) (0.284) (0.218)

Skewness(Returns) �0.0439 �0.00765 �0.112 �0.109

(0.0612) (0.0682) (0.110) (0.100)

Constant �12.30** �12.60** �17.06** �12.86**

(0.860) (1.025) (1.865) (1.598)

Observations 5,304 4,326 1,563 1,871

Pseudo R2 0.110 0.108 0.100 0.0806

*, ** Indicates significance at the 5% and 1% levels (two-tail), respectively.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 449

reflecting a focus on measurement of corporate govern- ance quality. While current ROA seems to be a natural control, Larcker, Richardson, and Tuna (2007) argue that ‘‘to the extent that governance structures are stable over time y the inclusion of current operating performance is likely to remove the impact of governance that we are trying to estimate.’’ Given our interest in the governance ratings as both measures of governance quality and as

informative signals of future firm performance, we estimate regressions both with and without the prior period’s industry-adjusted ROA as an additional control.

Our outcome variable is industry-adjusted ROA, or the difference between ROA for a firm and the median ROA for its industry in that fiscal year (using two-digit Standard Industrial Classification (SIC) codes for industry

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461450

classification). We measure industry-adjusted ROA at the end of the fiscal year ending between June 2008 and March 2009, the latest data available on Compustat at the time of this study.

Table 4 presents the results from our analysis. We find that AGR is associated with future operating performance. This statistical result is robust to the inclusion of lagged ROA as a regressor. However, the strength of the relation is greater when lagged ROA is omitted, consistent with governance quality being relatively persistent and affecting ROA over multiple periods. For the remaining ratings, CGQ and TCL have significant coefficients with the predicted sign in both the contemporaneous and

Table 4 Governance ratings and future operating performance.

Results are for ordinary least-square (OLS) regressions where the dependent

year ending between June 2008 and May 2009 and the independent variables

median adjusted ROA for the latest fiscal year ending on or before December 3

(2007), the natural logarithm of market value in millions of dollars as of Decemb

is defined as using income from operations (Compustat OIADP) divided by avera

for the fiscal year ending between June 2008 and May 2009. ROA is winsorize

sample, a firm must have data for ROA on Compustat and be in an industry with

errors.

The governance rating variables are the AGR from Audit Integrity, CGQ from Ri

a The Corporate Library. The governance ratings are measured as of as of Decem

zero and standard deviation of one.

Panel A: Descriptive statistics

Variable N Mean Std.

Ind. adj. ROA2008 5,074 �0.0246 0.

Ind. adj. ROA2005 5,067 �0.0037 0.

ln(MV) 5,056 5.25

Panel B: Contemporaneous ROA

AGR

Governance rating 0.0372** (0.00329)

ln(MV) 0.0465**

(0.00144)

Constant �0.297**

(0.00908)

Observations 4,941

Adjusted R2 0.189

Panel C: Future ROA, without contemporaneous ROA

Governance rating 0.0371** (0.00335)

ln(MV) 0.0454**

(0.00147)

Constant �0.308**

(0.00924)

Observations 4,941

Adjusted R2 0.178

Panel D: Future ROA, with contemporaneous ROA

Governance rating 0.0133** (0.00263)

Ind. adj. ROA2005 0.641**

(0.0112)

ln(MV) 0.0156**

(0.00125)

Constant �0.117**

(0.00792)

Observations 4,941

Adjusted R2 0.504

*, ** Indicates significance at the 5% and 1% levels, respectively.

unconditional future ROA analysis, but once lagged ROA is included (Panel D), the coefficients are insignificant. In other words, neither CGQ, TCL, nor GMI predict future performance given current levels of performance.

Thus, only AGR appears to be robustly associated with future operating performance. To assess the economic significance of the coefficient on the AGR rating, we consider the shift in predicted lagged ROA associated with a one-standard-deviation shift in the AGR rating. Obvi- ously, these sensitivity analyses should be interpreted with caution because it is quite plausible that governance quality and operating performance are jointly determined in a manner that confounds any causal interpretation of

variable is industry-median adjusted ROA (Ind. adj. ROA2008) for the fiscal

are the indicated governance rating as of December 31, 2005, industry-

1, 2005 (Ind. adj. ROA2005), and, following Larcker, Richardson, and Tuna

er 31, 2005, ln(MV). Industries are defined using two-digit SIC codes. ROA

ge total assets (average of Compustat AT for current and prior fiscal year)

d to have an absolute value not greater than one. To be included in the

at least five observations on ROA. Numbers in parentheses are standard

skMetrics/ISS, GMI from a GovernanceMetrics International, and TCL from

ber 31, 2005. Each governance rating is standardized to have a mean of

dev. P25 Median P75

2709 �0.0321 0.0354 0.0986

2712 �0.0004 0.0488 0.1137

2.52 3.45 5.31 7.11

CGQ GMI TCL

0.00807** �0.00340 0.00884**

(0.00375) (0.00345) (0.00321)

0.0347** 0.0192** 0.0230**

(0.00170) (0.00263) (0.00220)

�0.216** �0.0876** �0.118**

(0.0108) (0.0211) (0.0172)

3,917 1,400 1,651

0.141 0.0378 0.0613

0.00816* �0.00287 0.00584

(0.00389) (0.00358) (0.00336)

0.0348** 0.0220** 0.0259**

(0.00176) (0.00273) (0.00230)

�0.234** �0.125** �0.156**

(0.0112) (0.0219) (0.0179)

3,917 1,400 1,651

0.134 0.0467 0.0703

0.00288 �0.000300 �0.000694

(0.00302) (0.00245) (0.00238)

0.654** 0.757** 0.739**

(0.0129) (0.0190) (0.0182)

0.0122** 0.00746** 0.00893**

(0.00144) (0.00191) (0.00168)

�0.0933** �0.0587** �0.0691**

(0.00917) (0.0151) (0.0129)

3,917 1,400 1,651

0.478 0.553 0.534

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 451

our regression coefficients. We first note that the 25th, 50th, and 75th percentiles for industry-adjusted ROA for the sample of firms with at least one rating are –6.3%, 0.2%, and 5.8%, respectively. Shifting up one standard deviation in terms of AGR is associated with a 3.7 percentage-point increase in ROA (using the coefficients on AGR in the unconditional regression). This shift seems economically significant, but is insufficient to move a firm across a full quartile. The coefficient on AGR when ROA is included in the regression is much smaller, but note that the inclusion of lagged ROA suggests that the appropriate benchmark is more appropriately measured by the one- year change in ROA. The 25th, 50th, and 75th percentiles for changes in ROA for the sample of firms with at least one rating are �5.9%, �0.9%, and 2.6%, respectively. This suggests that the shift of 1.33 percentage points asso- ciated with a one-standard-deviation shift in AGR is insufficient to move a firm across a full quartile in terms of change in ROA.

We also conduct the robustness checks described above. Focusing on a common sample of 1,348 firms with controls data and ratings from each of the four firms, there are few changes. Focusing on the top and bottom deciles, we find that AGR has a significant difference between the operating performance of firms in the top decile relative to those in the bottom decile in all three analyses. Looking at changes in ratings from December 2005 to June 2006, we find that improved governance ratings are not associated with higher levels of operating performance in fiscal 2008.

One possible explanation for the absence of robust relations between the three primary governance ratings and operating performance is that it may take more than three years for the effects of good or bad governance (i.e., 2005 ratings) to be observed in firm profits (i.e., 2008 ROA). In addition to the data in our primary sample, we have data on CGQ for 2004 and on GMI from 2003 to 2008. We conducted additional tests to evaluate the relation- ship between fiscal 2008 industry-adjusted ROA and the earliest rating available (2004 for CGQ and as early as 2003 for GMI). We find no statistically significant relation between either CGQ or GMI and operating performance in these tests. Given that 2005 CGQ predicts fiscal 2008 operating performance, but 2004 CGQ does not, we examine the relationship between 2006 ratings and fiscal 2008 performance and find that only the relationship between AGR and operating performance is robust. Thus, while 2005 CGQ predicts 2008 operating performance, neither 2004 nor 2006 CGQ does so.

24 At best, the market-to-book ratio captures average Tobin’s Q,

whereas the variable of interest is generally the marginal Tobin’s Q. The

Q results are reported in order to be consistent with prior literature, but

we believe that the more interpretable results are future operating

performance and excess stock price returns. 25 We examine the impact of winsorization of Tobin’s Q on our

results with the primary ratings by performing analyses without

winsorization. In the contemporaneous and unconditional analyses,

CGQ and TCL remain the only significant variables, and CGQ again has an

unexpected sign. None of the ratings has a statistically significant

coefficient in the conditional analysis.

5.4. Firm value

Tobin’s Q, typically measured using some variant of the market-to-book ratio, is commonly used as an indicator of firm value in accounting and finance research. However, since market-to-book ratios (or their inverse) are used as proxies for risk factors (Fama and French, 1993), account- ing conservatism (Roychowdhury and Watts, 2007), and investment opportunity set or future growth opportu- nities (Adam and Goyal, 2008), it is unclear whether the

market-to-book ratio captures the underlying theoretical construct of ‘‘firm value.’’24 However, in light of its continued popularity in academic corporate finance research, we also examine Tobin’s Q.

We measure Tobin’s Q as (TA + MVE–BVE)/TA, where TA is total assets (Compustat item AT), MVE is market value of equity (PRCC_F�CSHO), and BVE is the book value of equity (CEQ). To control for differences associated with industry rather than governance attributes of each firm, we include industry fixed effects in our regressions. Because Tobin’s Q, like measures of operating perfor- mance, is highly persistent, we follow the approach used with operating performance and estimate both regres- sions with industry fixed effects and the governance variables alone and regressions with these variables and lagged Tobin’s Q as an additional control. We measure Tobin’s Q at the end of the latest fiscal year ending on or before December 31, 2005 (contemporaneous Tobin’s Q) and at the end of the fiscal year ending between June 2008 and May 2009 (future Tobin’s Q), the latest data available on Compustat at the time of this study. We winsorize contemporaneous and future Tobin’s Q at the 2nd and 98th percentiles.

As shown in Table 5, when examining contempo- raneous Tobin’s Q (Tobin’s Q2005), we find that three of the four primary ratings have statistically significant coefficients, including two that are unexpectedly negative (AGR, CGQ) and one that is positive (TCL). Looking at future Tobin’s Q without Tobin’s Q in fiscal year 2005 (Tobin’s Q2005) as a control gives very similar results. But, when lagged Tobin’s Q is included, none of the ratings attain statistical significance.25

To assess the economic significance of the estimated coefficients, note that for the 5,053 firms with ratings from each of the four firms and data on Tobin’s Q, the mean value of Tobin’s Q is 1.82 and the 25th, 50th, and 75th percentiles are 0.95, 1.10, and 1.61, respectively. Thus, while the coefficient on TCL without lagged Tobin’s Q (0.119) is statistically significant, the effect is not sufficient for a one-standard-deviation shift in the ratings to be associated with a shift in Tobin’s Q across a full quartile.

Our somewhat inconclusive results with Tobin’s Q are quite sensitive to model specification. Placing firms in top and bottom deciles by rating, we find that TCL and CGQ are associated with Tobin’s Q at the 1% level in the con- temporaneous and unconditional analyses, but only the coefficients on TCL have the expected sign. Using a common sample of 1,349 firms with contemporaneous Tobin’s Q, none of the ratings is significant. For future

Table 5 Governance ratings and Tobin’s Q.

Results are for OLS regressions where the dependent variable is Tobin’s Q, defined as the ratio (TA + MVE�BVE)/TA, where TA is total assets (Compustat AT),

MVE is market capitalization (PRCC_F�CSHO), and BVE is the book value of equity (CEQ), each for the fiscal year ending between June 2008 and May 2009.

Tobin’s Q2005 is measured as of the end of the latest fiscal year ending on or before December 31, 2005. To be included in the sample, a firm must have data

for Tobin’s Q on Compustat and be in an industry with at least five observations. Industry fixed effects are not reported for reasons of space. Numbers in

parentheses are standard errors.

The governance rating variables are the AGR from Audit Integrity, CGQ from RiskMetrics/ISS, GMI from GovernanceMetrics International, and TCL from The

Corporate Library. The governance ratings are measured as of December 31, 2005. Each governance rating is standardized to have a mean of zero and

standard deviation of one.

Panel A: Descriptive statistics

Variable N Mean Std. dev. P25 Median P75

Tobin’s Q 5,048 1.82 2.49 0.95 1.10 1.61

Tobin’s Q2005 5,053 2.33 2.74 1.07 1.47 2.34

Panel B: Contemporaneous Tobin’s Q

AGR CGQ GMI TCL

Governance rating �0.136** �0.236** 0.0133 0.119** (0.0380) (0.0309) (0.0291) (0.0318)

Ind. fixed effects Yes Yes Yes Yes

Observations 5,000 3,922 1,402 1,653

Adjusted R2 0.0852 0.106 0.139 0.117

Panel C: Future Tobin’s Q, without contemporaneous Tobin’s Q

Governance rating �0.153** �0.228** 0.00943 0.0800** (0.0349) (0.0294) (0.0198) (0.0242)

Ind. fixed effects Yes Yes Yes Yes

Observations 4,995 3,920 1,399 1,649

Adjusted R2 0.0693 0.0776 0.162 0.162

Panel D: Future Tobin’s Q, with contemporaneous Tobin’s Q

Governance rating �0.0532* �0.0604** 0.00255 0.00951 (0.0211) (0.0201) (0.0125) (0.0154)

Tobin’s Q2005 0.731** 0.694** 0.529** 0.589**

(0.00795) (0.0104) (0.0118) (0.0121)

Ind. fixed effects Yes Yes Yes Yes

Observations 4,988 3,917 1,399 1,649

Adjusted R2 0.657 0.574 0.666 0.664

*, ** Indicates significance at the 5% and 1% levels, respectively.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461452

Tobin’s Q using the 1,346 firms with necessary data for all ratings, only CGQ is significant with the predicted sign in the analysis and only with contemporaneous Tobin’s Q.

Overall, we interpret our results as consistent with there being little systematic relationship between Tobin’s Q and the governance ratings. As before, we also examine the relationship between 2008 Tobin’s Q and the earliest available ratings for CGQ and GMI. In general, we find no statistically significant relations with the expected sign.

26 While Gompers, Ishii, and Metrick (2003) do find evidence

consistent with the market being surprised by the superior performance,

Core, Guay, and Rusticus (2006) demonstrate that this finding does not

extend beyond the sample period examined in Gompers, Ishii, and

Metrick (2003).

5.5. Stock returns

We conduct two different tests related to stock prices and governance ratings reflecting differing assumptions about how ratings might affect future firm performance and how stock markets form expectations about such effects. The first set of tests examines stock returns around rating changes. To the extent that favorable changes in governance ratings convey unexpected positive informa- tion to stock markets, we expect positive abnormal returns to firms that see their governance ratings improve around the time that these improvements occur. Additionally, even if the ratings per se do not convey information to

markets, to the extent that improvements in ratings reflect unexpected changes in underlying governance character- istics that are expected to result in increased firm value, we would expect changes in governance ratings to be associated with future abnormal returns to the extent that the governance ratings capture the improvements in governance on a timely basis.

The second set of tests examines the relation between corporate governance ratings and subsequent stock returns. Specifically, commercial governance rating firms frequently claim that their ratings are associated with subsequent stock performance. As discussed in Gompers, Ishii, and Metrick (2003), if markets correctly anticipate the effect of governance on firm performance, it is not clear that we should expect an association between stock returns and corporate governance.26 Furthermore, if better corporate governance implies a lower cost of capital, as argued by some corporate governance rating

Table 6 Stock returns around changes in governance ratings.

Results are the coefficients on the change in (unstandardized)

governance rating from OLS regressions of market- or size-adjusted

returns, expressed as percentages, over the indicated windows around

the date of the rating change (date 0). Sample includes rating changes

from public sources over the period from 2002 through September 2007.

Unchanged governance ratings are not included. Numbers in parenth-

eses are standard errors. These standard errors are clustered on event

dates for 3- and 5-day windows and on months for windows from prior

rating to the day after the rating change.

The governance rating variables are the AGR from Audit Integrity, CGQ

from RiskMetrics/ISS, GMI from GovernanceMetrics International, and

TCL from The Corporate Library. The governance ratings are measured as

of December 31, 2005.

Panel A: Market-adjusted returns

Event window AGR CGQ GMI TCL

(�1, + 1) 0.0036 0.0076 �0.0350* 0.1506

(0.0029) (0.0044) (0.0145) (0.1319)

(�2, + 2) 0.0048 0.0050 �0.0150 0.3328**

(0.0032) (0.0038) (0.0114) (0.1190)

(prior rating, + 1) 0.0287 �0.0182 �0.0910 �0.3846

(0.0154) (0.0108) (0.1118) (0.7423)

Num. of event dates 668 18 8 206

Number of months 56 18 8 32

Panel B: Size-adjusted returns

(�1, + 1) 0.0028 0.0069* �0.0270 0.1535

(0.0027) (0.0029) (0.0202) (0.1146)

(�2, + 2) 0.0044 0.0051 0.0003 0.3173**

(0.0032) (0.0042) (0.0126) (0.1096)

(prior rating, + 1) 0.0307* �0.0013 �0.0644 0.1303

(0.0144) (0.0110) (0.0983) (0.5173)

Num. of event dates 668 18 8 206

Number of months 56 18 8 32

Panel C: Descriptive statistics of governance rating changes

Mean �0.13 0.15 0.04 �0.19

Standard deviation 11.02 7.70 2.43 1.19

10th percentile �13.00 �4.29 �0.50 �1.00

25th percentile �6.00 �1.30 �0.50 �1.00

50th percentile 1.00 �0.40 0.50 �1.00

75th percentile 6.00 0.42 0.50 1.00

90th percentile 13.00 4.57 1.00 1.00

*, ** Indicates significance at the 5% and 1% levels, respectively.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 453

firms, we would expect subsequent stock returns to be lower for firms with better governance.

5.5.1. Stock returns around rating changes

In examining stock returns around rating changes, we examine two kinds of return windows, each consistent with one of the two reasons to expect an association between rating changes and stock returns. First, consis- tent with the idea that rating changes convey information to the market about changes in governance, we examine short-window (i.e., three- or five-day) returns around announcements of rating changes. Second, consistent with the possibility that rating changes do not convey information to markets but are driven by changes in governance also observed by the market, we also examine returns over the period from the prior rating to the issuance of a new, different rating. Note that the ratings differ markedly in terms of the frequency and clustering of updates. GMI updates its ratings quarterly and ISS updates CGQ monthly. In contrast, AGR and TCL update the ratings of individual companies on a less specific schedule. For these reasons, the rating changes for GMI (CGQ) are clustered into eight (18) dates, while we have 668 (206) dates with rating updates for AGR (TCL) spread over 56 (32) distinct months.

In conducting this test, we regress the market-adjusted or size-adjusted returns over a three-day or five-day window around the date of a rating change (day 0) on the change in the unstandardized ratings. In an attempt to capture returns over a longer time period, we also regress the adjusted returns over the time period from the prior rating to day + 1 on the change in rating. The results from these regressions are presented in Table 6. Overall, there is limited evidence of changes in corporate governance ratings being associated with contemporaneous stock returns. CGQ exhibits a very small positive coefficient using three-day size-adjusted returns. The primary exception is TCL, where changes appear to be associated with stock returns over a five-day window around the change with both market-adjusted and size-adjusted returns. The changes are arguably economically significant. For example, a change in rating from ‘C’ to ‘A’ would be associated with a market-adjusted return of 0.67% (0.33%�2) over a five-day period. However, only 12.9% of changes in TCL are of this magnitude, and 86.5% of changes in TCL are a change up or down one level.

5.5.2. Future stock returns

We also examine excess stock returns, Alpha, as an outcome variable. Specifically, for each firm in our sample, we obtain monthly stock returns (RET) from CRSP for the months January 2006 to December 2008. For each firm, we then estimate regressions of these returns on the standard Fama-French monthly factor returns (Mkt-RF, SMB, HML, and Mom) obtained from Ken French’s Web site for three overlapping periods: 12 months, 24 months, and 36 months (i.e., the full period). The estimated intercepts from these regressions form our estimates of Alpha. Since Alpha represents returns in excess of hypothesized risk factors, we do not include additional controls in our

subsequent regressions. As pointed out in prior literature (Gompers, Ishii, and Metrick, 2003; Larcker, Richardson, and Tuna, 2007), if stock prices incorporate rational beliefs about the costs and benefits of alternative govern- ance structures, we should expect no association between excess returns and the governance ratings. Any associa- tion is the result of either (i) inefficiency in the pricing of corporate governance, (ii) unexpected, systematic shocks in firm value either caused by or correlated with these measures of corporate governance, or (iii) an omitted risk factor that is correlated with the measures of corporate governance. Notwithstanding these arguments, it is frequently argued by the rating firms that the governance ratings will be positively associated with future returns. An alternative argument, though not one made by the rating firms, is that governance quality is associated with lower expected returns (i.e., lower cost of capital), in which case we might expect a negative relationship between governance quality and realized returns.

Table 7 Governance ratings and future stock performance.

Results are for OLS regressions where the dependent variable is Alpha, estimated as the residual from a four-factor Fama-French model with

a momentum factor estimated using returns over the 12, 24, or 36 months after December 31, 2005 (a minimum of 12 months of return data is required

for inclusion). Factor data are obtained from Ken French’s Web site.

The governance rating variables are the AGR from Audit Integrity, CGQ from RiskMetrics/ISS, GMI from a GovernanceMetrics International, and TCL from

The Corporate Library. AGR, CGQ and GMI ratings are on a 0–100 scale. TCL is converted from an ‘‘A’’ to ‘‘F’’ grade to numerical values 1–5, where ‘‘A’’

equals 5 and ‘‘F’’ equals 1 (no ‘‘E’’). The governance ratings are measured as of December 31, 2005. Each governance rating is standardized to have a mean

of zero and standard deviation of one.

Panel A: Alphas, 12 months after December 31, 2005

AGR CGQ GMI TCL

Rating 0.00449** �0.000795 �0.000563 �0.000162 (0.00101) (0.00107) (0.00118) (0.00124)

Constant �0.000135 0.00196 0.00260* 0.00256*

(0.000951) (0.00103) (0.00118) (0.00124)

Observations 4,970 4,063 1,525 1,808

Adjusted R2 0.00375 0.00011 0.00051 0.00054

Panel B: Alphas, 24 months after December 31, 2005

Rating 0.00318** 0.000369 0.000768 0.00108 (0.000458) (0.000496) (0.000527) (0.000559)

Constant �0.000669 �0.00118* �0.000211 0.000119

(0.000430) (0.000475) (0.000527) (0.000559)

Observations 4,975 4,066 1,526 1,809

Adjusted R2 0.00942 �0.00011 0.00074 0.00152

Panel C: Alphas, 36 months after December 31, 2005

Rating 0.00315** 0.00179** 0.000769 0.00150** (0.000446) (0.000486) (0.000527) (0.000562)

Constant �0.00244** �0.00280** 0.000964 0.000505

(0.000420) (0.000465) (0.000528) (0.000561)

Observations 4,977 4,068 1,527 1,810

Adjusted R2 0.00971 0.00307 0.000738 0.00336

*, ** Indicates significance at the 5% and 1% levels, respectively. Numbers in parentheses are standard errors.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461454

Table 7 presents the results of our excess returns analysis over, 12-, 24-, and 36-month periods. Over a 12-month period, only AGR has a statistically significant association with Alpha, with an additional excess return associated with a one-standard-deviation shift in AGR of 0.45% per month. Over a 24-month period, two of the four primary ratings, AGR and TCL, have a statistically significant positive association with Alpha. The additional excess return associated with a one-standard- deviation shift in AGR (TCL) is 0.32% (0.11%) per month. Over a 36-month period, the statistical significance of TCL increases and a statistically significant relation between CGQ and Alpha appears. These relations are arguably economically significant. For example, a shift in AGR of one standard deviation equates to Alpha over three years of 11.34% (36�0.315%). However, as these hypothetical excess returns do not account for transaction costs, it is not clear whether they could form the basis of a profitable trading strategy.

It is unclear why the statistical significance of the relationship increases over a longer period. One possibi- lity is that the market fails to appreciate the implications of good corporate governance, as measured by AGR, CGQ, and TCL, for future corporate performance and only corrects this failure over a period extending over three years. A second possibility is that the corporate govern- ance ratings are correlated with an omitted risk factor and increasing the sample period increases our ability to detect a relation. An implication of this alternative explanation is that governance ratings are negatively

associated with the portion of expected returns attribu- table to this hypothetical omitted factor. That is, better governance leads to a higher cost of capital, which contrasts with frequent claims that better governance lowers a firm’s cost of equity capital.

Another possibility is that there is something specific to the third year that affects the relations between the ratings and alphas. Given that the third year (January– December 2008) coincides with the recent financial crisis, we investigate this further. Examining governance ratings as of December 31, 2006 and Alpha over 24 months, we find statistically significant coefficients of similar magni- tudes on the same three variables (AGR, CGQ, and TCL). This is consistent with the ratings—and perhaps corporate governance—being more valuable in the recent crisis. However, given that the S&P 500 lost more than 38% of its value over this period, the 7.3% (untabulated coefficient on AGR of 0.306%�24) alpha from moving up a full standard deviation of AGR provides only very limited assurance even if these results are taken at face value.

We again examine the sensitivity of our results to the use of indicators for the top and bottom deciles in place of standardized ratings and the use of a common sample across the ratings. We focus on the full 36-month return period in these analyses. In the decile analysis, the difference between the coefficients on the indicators for AGR, CGQ, and TCL (i.e., the ratings with significant relations in Table 7) are both economically and statisti- cally significant. The mean monthly alpha for the top (bottom) decile based on AGR is 0.212% (�0.968%), with

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 455

the latter (and the difference) being statistically signifi- cant at the 1% level. The mean monthly alpha for the remaining 80% of observations is �0.156%, suggesting that the coefficient shown in Table 7 is primarily attributable to negative alpha in the lowest decile. The coefficients on the CGQ decile portfolios yield a statisti- cally significant difference (p-value of 0.02), but one that is less than half the size of the difference for the AGR portfolios. With a common sample of 1,500 firms across all ratings, only AGR remains statistically significant with a coefficient of 0.22%, slightly below that estimated above and consistent with the results above not being driven by differences in samples across the ratings.

We also examine (in untabulated analysis) the rela- tionship between changes in governance ratings and future outcomes. We measure the change in ratings over the period from December 31, 2005 to June 30, 2006 and examine each outcome (restatements, lawsuits, operating performance, Tobin’s Q, and credit ratings) over the next two or three years (depending on data availability). These results suggest even weaker ability for the ratings to predict future outcomes than that suggested by our primary analysis.

27 ISS Web site: http://www.issproxy.com/issgovernance/research/

recommendation.html. 28 ‘‘A proxy adviser’s two sides; some question work of ISS for

companies it scrutinizes,’’ The Washington Post, January 23, 2006. Also

see Alexander, et al. (2009). 29 See ‘‘ISS US Corporate Governance Policy, 2007 Updates,’’ available

at www.issproxy.com/pdf/2007%20US%20Policy%20Update.pdf. 30 ISS Web site: http://www.issproxy.com/issgovernance/research/

recommendation.html. 31 We exclude proposals by shareholders that do not receive

management support, as it is unclear how support for such proposals

will relate to the quality of a firm’s corporate governance.

5.6. Credit ratings

Our final outcome measure is credit rating, a measure of a firm’s cost of debt capital, which prior studies find is affected by the quality of firm governance (Cremers, Nair, and Wei, 2007). Table 8 presents results from three sets of analyses using ordered logistic regressions. Panel B of Table 8 reports the results from estimating the relationship between corporate governance ratings and contemporaneous credit ratings issued by Standard and Poor’s, after controlling for a number of variables shown to be related to credit ratings in prior research (e.g., Ashbaugh-Skaife, Collins, and LaFond, 2006). Only GMI exhibits a statistical association with contemporaneous credit ratings. Panel C of Table 8 provides results from regressions in which governance ratings are used as unconditional predictors of future credit ratings. Two ratings are associated with future credit ratings and with the expected sign. However, this unconditional analysis is presented mostly for completeness and comparability with our earlier analyses. Governance ratings are presumably only useful in predicting future credit ratings if they provide information of incremental usefulness given current credit ratings. Panel D of Table 8 presents the regression when the credit ratings at the end of 2005 are also included in the regression. In this analysis, none of the governance ratings has a statistically significant coefficient.

Using the top and bottom deciles of the ratings, we get similar results. In the unconditional analysis, CGQ and GMI are significantly (at the 1% level) negatively related, consistent with the idea that stronger governance may lead to weaker creditor protection, but TCL does not have a significant relation with future credit ratings. Using a common sample, inferences are almost identical to those from Table 8.

6. CGQ, ISS recommendations, and shareholder voting

ISS is unique among the firms we examine in that it also provides influential advice on shareholder voting. As the dominant player in this market, ISS provides clients with ‘‘comprehensive analyses of proxy issues and complete vote recommendations’’ on all shareholder votes.27 According to some reports, these recommenda- tions are followed by roughly 20% of its clients and are therefore influential in voting outcomes.28

In this section, we examine two questions. First, does ISS consider a firm’s CGQ when it evaluates whether to recommend shareholders vote for or against a particular proposal? Second, does CGQ appear to affect the share- holder support for a proposal? There are at least two reasons to expect a relationship between CGQ and ISS recommendations. First, and somewhat simplistically, ISS places CGQ scores on the cover of its voting recommenda- tions and reportedly reminds prospective clients of this when selling its services advising firms on how to increase CGQ scores. This suggests that ISS considers CGQ relevant to the voting decision. Second, there is substantial overlap between the inputs to CGQ and the inputs to voting recommendations. Indeed, ISS has ‘‘undertaken several steps to ensure that its voting policy and ratings criteria in CGQ are aligned.’’29 ISS also claims that its voting recommendations are ‘‘based on our benchmark policies, which leverage empirical research on the impact of proxy issues on shareholder value.’’30 With regard to voting outcomes, if CGQ provides useful information to share- holders, it seems plausible that CGQ would be associated with voting outcomes either directly, or indirectly through its effect on the voting recommendations of ISS.

We focus on management proposals voted on at meetings in the years 2005–2007 for which we have prior CGQ ratings and use the most recently issued CGQ rating on the date of the meeting.31 We examine both a broad class of proposals (mostly director elections and auditor ratifications) and proposals related to employee compensation plans, as the latter are frequently closer votes. Using voting data provided by ISS, our sample includes 34,761 management-supported proposals be- tween 2005 and 2007 for which we have data on ISS recommendations. This sample includes 2,309 proposals on compensation plans, 27,243 director elections, 3,821 auditor ratifications, and 1,388 proposals on other matters, such as proposals to adopt majority voting or declassify the board.

Table 8 Governance ratings and credit ratings.

Results are for ordered logit regressions where the dependent variable is the Standard and Poor’s credit rating at either the end of the latest fiscal year

ending on or before December 31, 2005 (Panel B) or the end of fiscal 2008 (Panels C and D). In Panel B, the independent variables are the indicated

governance rating as of December 31, 2005, ln(MV), BM, ROA, and Leverage for the latest fiscal year ending on or before December 31, 2005, and Beta and

Volatility estimated using data for the 60 months prior to December 31, 2005. ln(MV) is the natural logarithm of market value expressed in millions of

dollars. BM is the book value of common equity (CEQ) divided by the market value of common equity. ROA is defined as using income from operations

(Compustat OIADP) divided by average total assets (average of Compustat AT for current and prior fiscal years). ROA is winsorized to have an absolute

value not greater than one. Beta is computed using firm-specific capital-asset-pricing model (CAPM) regressions and Volatility is the annualized standard

deviation of monthly returns in excess of the risk-free rate. In Panel D, we add the Standard and Poor’s credit rating at the end of the latest fiscal year

ending on or before December 31, 2005. Numbers in parentheses are standard errors.

The governance rating variables are the AGR from Audit Integrity, CGQ from RiskMetrics/ISS, GMI from GovernanceMetrics International, and TCL from

The Corporate Library. The governance ratings are measured as of December 31, 2005. Each governance rating is standardized to have a mean of zero and

standard deviation of one.

Panel A: Descriptive statistics, independent variables

Variable N Mean Std. dev. P25 Median P75

ln(MV) 1,310 8.19 1.49 7.18 8.16 9.39

BM 1,310 0.47 0.29 0.28 0.43 0.62

ROA 1,319 0.09 0.08 0.04 0.08 0.13

Leverage 1,319 0.30 0.21 0.16 0.27 0.41

Beta 1,201 0.99 0.80 0.43 0.83 1.29

Volatility 1,200 0.36 0.18 0.23 0.31 0.43

Panel B: Contemporaneous credit ratings

AGR CGQ GMI TCL

Governance rating 0.0599 �0.0578 0.159** 0.0953 (0.0601) (0.0676) (0.0668) (0.0631)

ln(MV) 1.213** 1.214** 1.089** 1.121**

(0.0555) (0.0585) (0.0740) (0.0699)

BM 0.149 0.0270 0.304 �0.0759

(0.219) (0.249) (0.324) (0.293)

ROA 1.028 1.910* 4.052** 3.587**

(0.773) (0.839) (1.027) (0.974)

Leverage �2.473** �2.507** �2.539** �2.699**

(0.303) (0.320) (0.406) (0.370)

Beta �0.0418 �0.0378 0.440** 0.297**

(0.0983) (0.112) (0.161) (0.133)

Volatility �7.866** �8.333** �12.48** �11.00**

(0.544) (0.602) (0.899) (0.759)

Observations 1,195 1,076 781 857

Pseudo R2 0.233 0.241 0.220 0.233

Panel C: Future credit ratings, unconditional prediction

Governance rating �0.0744 0.432** 0.521** �0.106 (0.0534) (0.0609) (0.0612) (0.0556)

Observations 1,342 1,185 832 918

Pseudo R2 o0.001 0.008 0.018 o0.001

Panel D: Future credit ratings, conditional prediction

Governance rating �0.0438 �0.0585 0.0256 �0.0381 (0.0555) (0.0625) (0.0657) (0.0581)

Credit rating2005 1.245** 1.270** 1.245** 1.202**

(0.0366) (0.0390) (0.0465) (0.0426)

Observations 1,217 1,098 785 863

Pseudo R2 0.331 0.335 0.306 0.307

*, ** Indicates significance at the 5% and 1% levels, respectively.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461456

As with the evaluation of the link between outcomes and ratings above, we consider both ‘‘conditional’’ and ‘‘unconditional’’ analyses. In so doing, we allow for the possibility that corporate governance factors captured by CGQ may have a direct impact on voting outcomes (independent of their impact on ISS recommendations) as well as an indirect impact through the ISS recommen- dation. Recognizing that shareholders may consider many factors other than CGQ or the ISS recommendation in their

assessment of how to vote on a proposal, we include a number of controls. For our analysis of all proposals taken together, we include excess returns over a 12-month period prior to the meeting date as a control to allow for the possibility that poor stock performance affects voting outcomes and is correlated with CGQ. For compensation proposals, we also include proposal dilution, burn rate, and overhang, as these variables are approximations of the factors that ISS explicitly considers in developing its

Table 9 CGQ, ISS recommendations, and shareholder voting.

Results in Panels A and B are for logit regressions where the dependent variable equals one if the ISS recommends a vote ‘‘for’’ a proposal. Results in

Panels C and D are for Tobit regressions (with bounds at zero and one) where the dependent variable percent_for is calculated as the numbers of votes for

a proposal divided by the sum of votes for, votes against, and abstentions. Shareholder voting data are for proposals receiving management support and

voted on at meetings in 2005, 2006, and 2007 for which we have prior CGQ ratings. Votes on compensation plans include votes on bonus, compensation,

stock and option plans, excluding non-employee plans. Recommendation and voting data were obtained from ISS. Director elections in Panels C and D are

restricted to those on the Equilar director file. Numbers in parentheses are standard errors clustered by firm.

Excess returns are the returns over the 12-month period ending two months prior to the meeting date, less the value-weighted CRSP returns over the

same period. Dilution measures are based on data supplied by Equilar. Proposal dilution equals shares requested under the proposal divided by shares

outstanding. Burn rate equals options granted in the prior fiscal year divided by shares outstanding. Overhang equals options outstanding divided by

shares outstanding. Recent restatement indicates that the firm had a restatement listed on the Glass-Lewis file over the 24-month period prior to the

meeting. Excess compensation is the residual, as a proportion of the fitted value, from a regression by two-digit SIC code of 2005 total direct compensation,

as defined by Equilar, on market value, sales, return on assets, and one-year total shareholder return. Panel B (and D) reports regressions estimating ISS

recommendations on (and results of) shareholder votes for particular directors. Chair, Vice chairman, Lead director, Outsider, Insider, and Female are

indicator variables for individual directors nominated for the board and are taken from the Equilar director file. Age (Tenure) is the director’s age (tenure

on the board) taken from the Equilar director file.

Panel A: CGQ and ISS recommendations

All proposals Compensation plans Compensation plans

CGQ 0.0282** 0.0241** 0.0243**

(0.0729) (0.0022) (0.0027)

Proposal dilution �9.5612**

(1.6661)

Burn rate �5.7680

(5.8100)

Overhang �2.8278

(1.7920)

Constant 0.4872** 0.1232 0.8802

(0.0013) (0.1397) (0.2218)

Marginal effect of change in CGQ on probability

that ISS recommends a vote ‘‘for’’ the proposal

(evaluated at the mean value for CGQ)

0.0022 0.0031 0.0031

Observations 34,761 2,309 1,527

Panel B: CGQ and ISS recommendations in director elections

Without controls With controls

CGQ 0.0278** 0.0248**

(0.00211) (0.00210)

Chair 0.233**

(0.0885)

Vice chairman �0.356

(0.227)

Lead director �0.210

(0.151)

Outsider 1.288**

(0.110)

Insider 0.0403

(0.110)

Female �0.0580

(0.112)

Age �0.00760

(0.00459)

Tenure �0.00248

(0.00433)

Constant 0.528** 0.413

(0.117) (0.307)

Marginal effect of change in CGQ on probability that ISS recommends a

vote ‘‘for’’ the proposal (evaluated at the mean value for CGQ)

0.00204 0.00165

Observations 13,011 12,698

Pseudo R2 0.0732 0.1180

Panel C: CGQ and shareholder voting outcomes

All votes Compensation plans

CGQ 0.0001** �0.0003** 0.00007** �0.0006**

(o0.0001) (o0.0001) (0.0001) (0.0001) ISS recommendation 0.1644** 0.1717**

(0.0054) (0.0107)

Excess return �0.0010 �0.0020

(0.0017) (0.0050)

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 457

Table 9. (continued )

Proposal dilution �0.4253**

(0.1066)

Burn rate �0.5960**

(0.2649)

Overhang �0.3919**

(0.0715)

Constant 0.9381** 0.8198** 0.8442** 0.7926**

(0.0026) (0.0062) (0.0087) (0.0151)

Observations 33,772 33,594 2,278 1,503

Panel D: CGQ and shareholder voting outcomes for director elections

All director elections Audit committee members

CGQ 0.0001 �0.000381** �0.000107 �0.000213**

(0.0001) (0.0001) (0.0001) (0.0000)

ISS recommendation 0.175** 0.227**

(0.00831) (0.0134)

Excess return 0.00515 0.00314

(0.00371) (0.00377)

Chair �0.00824** �0.0257**

(0.00217) (0.00649)

Vice chairman �0.00946 �0.00539

(0.00709) (0.00990)

Lead director 0.00175 0.00394

(0.00198) (0.00266)

Outsider 0.00206 0.00199

(0.00253) (0.00579)

Insider 0.0232** 0.0535**

(0.00326) (0.0166)

Female 0.00795** 0.00284

(0.00155) (0.00207)

Age 0.00004 �0.000106

(0.0001) (0.000109)

Tenure �0.000471** �0.000571**

(0.000108) (0.000154)

Committee chair �0.00262*

(0.00137)

Recent restate. �0.0127**

(0.00313)

Constant 0.943** 0.812** 0.959** 0.765**

(0.00385) (0.00978) (0.00437) (0.0145)

Observations 12,510 12,206 5,137 5,109

Governance and nominating committee members Compensation committee members

CGQ 0.000133 �0.000217** 0.0000 �0.000228**

(0.0001) (0.0001) (0.0001) (0.0001)

ISS recommendation 0.210** 0.231**

(0.0103) (0.0113)

Excess return 0.00216 0.00273

(0.00392) (0.00417)

Chair �0.0112* �0.00610

(0.00495) (0.00636)

Vice chairman �0.0237 �0.0315*

(0.0211) (0.0181)

Lead director 0.00386 0.00730*

(0.00247) (0.00287)

Outsider 0.00009 �0.00261

(0.00430) (0.00469)

Insider 0.0947** 0.139**

(0.0162) (0.0188)

Female 0.00824** 0.00537*

(0.00233) (0.00242)

Age 0.000143 0.00004

(0.000117) (0.000122)

Tenure �0.000787** �0.000624**

(0.000163) (0.000146)

Committee chair �0.00509** �0.000167

(0.00172) (0.00146)

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461458

Table 9. (continued )

Excess comp. 0.000006**

(0.0000)

Constant 0.935** 0.765** 0.944** 0.753**

(0.00517) (0.0121) (0.00529) (0.0131)

Observations 4,759 4,746 5,030 4,111

*, ** Indicates significance at the 5% and 1% levels, respectively.

Table 10 Summary of results.

This table summarizes the results reported in Tables 2–8, focusing only on the estimated coefficient for the governance ratings. An asterisk (*) signifies

a significant relationship with the expected sign (i.e., a high governance rating is related to fewer bad outcomes or superior performance). An ‘‘x’’ signifies

a statistically significant relationship with an unexpected sign. The governance rating variables are AGR from Audit Integrity, CGQ from RiskMetrics/ISS,

GMI from GovernanceMetrics International, and TCL from The Corporate Library.

Dependent variable AGR CGQ GMI TCL

Restatements No controls ** **

Controls ** **

Class-action lawsuits No controls ** xx **

Controls ** *

Operating

performance Contemporaneous ROA ** ** **

Future ROA, without contemporaneous ROA ** *

Future ROA, with contemporaneous ROA **

Tobin’s Q Contemporaneous Q xx xx **

Future Q, without contemporaneous Q xx xx **

Future Q, with contemporaneous Q x xx

Stock returns around Market-adjusted returns (�1, + 1) x

ratings changes (�2, + 2) **

(prior rating, + 1)

Size-adjusted returns (�1, + 1) *

(�2, + 2) **

(prior rating, + 1) *

Future stock performance Alphas—12 months **

Alphas—24 months **

Alphas—36 months ** ** **

Credit ratings Contemporaneous S&P rating **

Future S&P rating, controls

*(**) Indicates significance with the expected sign at the 5% (1%) level.

x (xx) Indicates significance with the unexpected sign at the 5% (1%) percent level.

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 459

recommendations (Morgan, Poulsen, and Wolf, 2006).32

These data were obtained from executive compensation research firm Equilar Inc.

Panel A of Table 9 shows an association between CGQ and ISS recommendations, but one that is surprisingly weak. For example, for an increase of one point in a firm’s CGQ rating, the change in probability that ISS recommends a vote in favor of a proposal is approximately 0.0022, which suggests that a one-standard-deviation increase in

32 We do not include governance variables considered in prior

research (e.g., staggered board or majority voting in Cai, Garner, and

Walkling, 2009 or Choi, Fisch, and Kahan, 2008), as these variables are

identified as inputs to the CGQ score. Including such variables would

understate the impact of CGQ if shareholders do not consider these

variables directly, but do rely on CGQ. However, the impact of CGQ may

be overstated in our analyses if shareholders rely on these variables,

either alone or in conjunction with CGQ, as by excluding them we will

attribute the explanatory power of all governance variables to CGQ.

CGQ (28.50 points) is associated with a 6.3 percentage- point increase in the probability of ISS favoring a proposal. This is rather odd, as it suggests that ISS does not place much weight on its own measure when developing voting recommendations. Panel B of Table 9 provides results from similar analysis of director elections. Again the relation between CGQ and ISS recommendations is statistically significant, but substantively small, with a one-point (one- standard-deviation) increase in CGQ translating into 0.17 (4.70) percentage-point increase in the probability that ISS recommends a vote for a director. Given that, in our sample, ISS recommends a vote in favor of more than 90% of management-supported director candidates, the effect of CGQ on ISS recommendations seems small in that the predicted probability of ISS support is high even if CGQ is very low. (Table 10)

Panel C of Table 9 shows the relationship between CGQ and shareholder voting outcomes, where the outcome is defined as the percentage of votes cast for a proposal or

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461460

candidate director. We first examine voting outcomes on all proposals, including director elections and proposed compensation plans, taken together. Excluding the ISS recommendation from the analysis, the estimated coeffi- cient on CGQ is very small (0.0001). When ISS recom- mendations are included in the analysis, the coefficient is actually negative, suggesting that the higher the CGQ rating, the lower the percentage of votes cast in favor of a proposal. Note that the coefficients on the ISS recommen- dation indicator variable are consistent with prior work, such as Bethel and Gillan (2002). Taken literally, these suggest that an ISS recommendation in favor of a proposal can sway more than 16% of the vote.

Focusing just on votes concerning compensation plans, including stock and option plans, CGQ again has an economically insignificant relation with voting outcomes when ISS recommendations and controls are excluded from the regression; but conditional on ISS recommenda- tions and proposal-specific factors likely to affect voting outcomes (namely, excess return, proposal dilution, burn rate, and overhang), our results suggest a negative (but substantively weak) relationship between CGQ ratings and shareholder voting outcomes.

Finally, Panel D of Table 9 examines the relationship between CGQ and shareholder voting outcomes for director elections. We conduct analysis both with and without additional controls and for various subsets of director elections, namely elections of members of the three major committees: audit, compensation, and nomi- nating and governance committees. Data on director characteristics come from the Equilar director file. In all cases, we find that either CGQ has no statistically significant relationship with voting outcomes or the relationship is of the ‘‘incorrect’’ sign, namely that higher CGQ is associated with lower shareholder support for the directors proposed by management.33

To evaluate the robustness of our findings, we examine the role of institutional shareholders, as these firms are the major clients of ISS, and CGQ rankings may matter when more shares are held by institutional investors. We estimate the regressions in Panel C of Table 9 with the inclusion of the variable Percent institutional holding, which represents the percentage of shares outstanding held by institutions making 13-F filings, interacting this variable with both the CGQ rating and the ISS recommen- dation. We did not find this variable or either of the interactions to be statistically significant or the inclusion of these variables to alter the basic finding that CGQ has a very small impact on voting outcomes.

We also examined the role of ‘‘extreme’’ ratings on both ISS recommendations and voting outcomes. In particular, we create indicator variables for a firm’s rating being in the top or bottom decile of CGQ ratings and used these two indicator variables in place of CGQ in the regressions tabulated in Table 9. The results from these regressions are consistent with tabulated results, except

33 Few director elections in our sample are contested elections and

our results are unaffected when we exclude elections that appear to be

contested.

that there appears to be a more significant relationship between CGQ and ISS recommendations for ‘‘extreme’’ ratings. For example, including controls, a firm with a CGQ in the top (bottom) decile has a 4 (12) percentage-point increase in the probability that ISS recommends for (against) a director. However, the impact of ‘‘extreme’’ CGQ scores on voting outcomes remains economically small, as the predicted probability of ISS supporting a management-supported director is still high, even when CGQ is very low.

7. Summary and concluding remarks

Shareholders, regulators, hedge fund managers, press commentators, board members, and policy makers in- creasingly stress the importance of good governance, arguing that it improves firm performance, shareholder welfare, and the health of the public markets. However, distinguishing good governance from bad has proved more difficult, especially given the great variety of corporate governance mechanisms (and combinations thereof) employed by firms. Several commercial firms now offer ratings of the quality of a company’s govern- ance. The providers of these ratings make strong claims regarding the ratings’ value in predicting future outcomes, such as accounting restatements, shareholder suits, operating performance, and stock returns. Directors also use these ratings as guides in organizing their firm’s governance arrangements and as a ‘‘red flag’’ that indicates how much they need to monitor.

We provide an independent assessment of prominent commercial corporate governance ratings. Prior evidence on individual ratings has generally been backward- looking, raising the distinct possibility that the ratings reflect past firm performance but are unable to predict future outcomes. We examine the ability of ratings produced by RiskMetrics/ISS, GovernanceMetrics Interna- tional, and The Corporate Library to predict future restatements, security litigation, and firm performance. We find that these governance ratings have either limited or no success in predicting firm performance or other outcomes of interest to shareholders. Moreover, even when there is a statistical association with future out- comes, the substantive economic effect is small. In contrast, we find somewhat stronger predictive evidence for the governance rating produced by Audit Integrity, AGR, which uses information in financial statements, rather than focusing on observable corporate governance mechanisms, such as board structure.

The fact that we find some relation between AGR and both future operating performance and excess returns, suggests that the results for CGQ, GMI, and TCL are not simply attributable to the confounding effects of the optimal selection of governance structures. Our view is that a more plausible interpretation of the weak and mixed results we find is that the commercial ratings contain a large amount of measurement error. Some support for this interpretation is found in the surprisingly small correlations among the ratings. This suggests that either the ratings are measuring very different corporate

R.M. Daines et al. / Journal of Financial Economics 98 (2010) 439–461 461

governance constructs or there is substantial measure- ment error in at least some of the ratings. Since the firms use the same basic governance data, examine similar governance dimensions (e.g., anti-takeover provisions, board structure, and executive compensation), and claim to measure overall ‘‘corporate governance,’’ we believe that each firm is attempting to measure a similar corporate governance construct. The absence of cross- sectional correlation is consistent with a high degree of measurement error in the rating processes across firms.34

These results suggest that boards of directors should not implement governance changes solely for the purpose of increasing their ranking.

An alternative explanation is that we do not have the right model for estimating the impact of firm governance or the right measures of firm performance. Ratings firms may object that, given the right model specification, their ratings are significant and informative. Again, the fact that we find support for AGR suggests that our analyses are not completely confounded by such concerns. But, to the degree that this objection is valid, it is incumbent on the ratings firms to explain how their ratings map into future performance, given the apparent difficulty of indepen- dently substantiating the claimed relations across a battery of standard tests using several important outcome variables. Such explanations would be consistent with the rating companies’ urging of transparency on the firms they rate. As stated on the RiskMetrics/ISS Web site,

As more and more investors, insurers and credit rating agencies recognize the link between corporate govern- ance performance and risk, the more important it is for companies to understand how their corporate govern- ance practices are measured. y We believe profoundly that transparency instills trust and, with trust comes confidence and more intelligent decisions.

References

Adam, T.R., Goyal, V.K., 2008. The investment opportunity set, its proxy variables. Journal of Financial Research 31, 41–63.

Aggarwal, R., Williamson, R., 2006. Did new regulations target the relevant corporate governance attributes? Unpublished working paper, Georgetown University.

Alexander, C.R., Chen, M.A., Seppi, D.J., Spatt, C.S., 2009. The role of advisory services in proxy voting. NBER Working Paper No. w15143.

Ashbaugh-Skaife, H., Collins, D.W., LaFond, R., 2006. The effects of corporate governance on firms’ credit ratings. Journal of Accounting and Economics 42, 203–243.

Ashbaugh-Skaife, H., LaFond, R., 2006. Corporate governance and the cost of equity capital: an analysis of U.S., non-U.S. firms’ GMI ratings. GovernanceMetrics International-sponsored research.

34 Given our results, an interesting question is why institutional

investors, shareholders, and other parties buy the ratings. It is difficult

for us to precisely answer this question. We interviewed executives at

several money management firms that purchase the commercial ratings.

One consistent explanation was that, while the ratings do not have

predictive value, purchasing the ratings is a cost-effective way to obtain

the underlying data. It is also not clear whether the market for

commercial ratings is highly profitable. For example, the majority of

profit earned by ISS/RiskMetrics is produced from their voting

recommendation and processing work.

Beasley, M.S., 1996. An empirical analysis of the relation between the board of directors composition, financial statement fraud. The Accounting Review 71, 443–465.

Bebchuk, L., Cohen, A., Ferrell, A., 2009. What matters in corporate governance? Review of Financial Studies 22, 783–827.

Bethel, J.E., Gillan, S.L., 2002. The impact of the institutional and regulatory environment on shareholder voting. Financial Manage- ment 31, 29–54.

Bhagat, S., Bolton, B.J., Romano, R., 2007. The promise and peril of corporate governance indices. ECGI-Law Working Paper No. 89/2007.

Brown, L.D., Caylor, M.L., 2004. The correlation between corporate governance, company performance. Institutional Shareholder Ser- vices White Paper.

Brown, L.D., Caylor, M.L., 2006. Corporate governance and firm valuation. Journal of Accounting and Public Policy 25, 409–434.

Cai, J., Garner, J.L., Walkling, R.A., 2009. Electing directors. Journal of Finance 64, 2389–2421.

Choi, S.J., Fisch, J.E., Kahan, M., 2008. Director elections and the role of proxy advisors. University of Pennsylvania, Institute for Law & Economics Research Paper No. 08-18.

Coles, J.L., Naveen, D.D., Naveen, L., 2008. Boards: does one size fit all? Journal of Financial Economics 87, 329–356.

Core, J.E., Guay, W.R., Rusticus, T.O., 2006. Does weak governance cause weak stock returns? An examination of firm operating performance, investors’ expectations. The Journal of Finance 61, 655–687.

Cremers, M., Nair, V.B., Wei, C., 2007. Governance mechanisms and bond prices. Review of Financial Studies 20, 1359–1388.

Daines, R., Klausner, M., 2001. Do IPO charters maximize firm value? Antitakeover protection in IPOs. Journal of Law, Economics, and Organization 17, 83–120.

Derwall, J., Verwijmeren, P., 2007. Corporate governance and the cost of equity capital: evidence from GMI’s governance rating. ECCE Research Note 06-01, European Centre for Corporate Engagement, version 2.0.

Erickson, M., Hanlon, M., Maydew, E., 2006. Is there a link between executive compensation and accounting fraud? Journal of Account- ing Research 44, 113–143.

Fama, E.F., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3–56.

Farber, D.B., 2005. Restoring trust after fraud: does corporate governance matter? The Accounting Review 80, 539–561.

Fich, E.M., Shivdasani, A., 2006. Are busy boards effective monitors? The Journal of Finance 61, 689–724.

Gompers, P.A., Ishii, J.L., Metrick, A., 2003. Corporate governance and equity prices. Quarterly Journal of Economics 118, 107–155.

Johnson, S.A., Moorman, T.C., Sorescu, S., 2009. A reexamination of corporate governance and equity prices. Review of Financial Studies 22, 4753–4786.

Koehn, D., Ueng, J., 2005. Evaluating the evaluators: should investors trust corporate governance metrics ratings? Journal of Management and Governance 9, 111–128.

La Porta, R., Lopez-de-Silanes, F., Schleifer, A., Vishny, R., 1998. Law and finance. Journal of Political Economy 106, 1113–1155.

Larcker, D., Richardson, S., Tuna, I., 2007. Corporate governance, accounting outcomes, and organizational performance. The Account- ing Review 82, 963–1008.

Morck, R., Shleifer, A., Vishny, R.W., 1988. Management ownership and market valuation: an empirical analysis. Journal of Financial Economics 20, 293–315.

Morgan, A., Poulsen, A., Wolf, J., 2006. The evolution of shareholder voting for executive compensation schemes. Journal of Corporate Finance 12, 715–737.

Peng, L., Röell, A., 2008. Executive pay and shareholder litigation. Review of Finance, 141–184.

Rogers, J., Stocken, P., 2005. Credibility of management forecasts. The Accounting Review 80, 1233–1260.

Roychowdhury, S., Watts, R.L., 2007. Asymmetric timeliness of earnings, market-to-book, conservatism in financial reporting. Journal of Accounting and Economics 44, 2–31.