Does Executive Compensation Affect Corporate Investments

  • Doc File 1,380.50KByte

STRONG BOARDS, management entreNchment, and Accounting ReSTATEMENT

William R. Baber

Sok-Hyon Kang

Lihong Liang

The George Washington University

School of Business

Government Hall 401

710 21st Street, NW

Washington, DC 20052

This version: July 2005

Baber: (202) 994-5089

Kang: (202) 994-6058

Liang: (202) 994-8346

We thank Chris Jones, Krishna Kumar, and seminar participants at the George Washington University, the 2005 European Accounting Association Meetings, the Investor Responsibility Research Center (IRRC), University of Minnesota, and Rutgers University for helpful comments. We are grateful to Carol Bowie, Robin Cowles, Annick Dunning, Sergio Shuchner, and Karen Taranto for their assistance with the data, and to Suhair Musa for her research assistance. The study is funded partially by John C. Richardson of JMR Financial, Inc. We also thank IRRC for granting access to portions of their databases.

STRONG BOARDS, management entreNchment, and Accounting RESTATEMENT


Using a large sample of widely-held U.S. firms, we investigate cross-sectional associations between the probability of financial restatements and contractual and legal provisions that restrict shareholder rights, while controlling for other key governance characteristics representing the strength of the Board of Directors, equity incentives of the CEO, and share ownership structure. Comparisons of 185 restatement firms with 1,487 non-restatement firms indicate that the vast majority of popular corporate governance indicators such as board and audit committee independence, CEO-chair separation, board size, whether or not the audit committee has an outside financial expert, director interlock, busy directors, equity incentives of the CEO, and share ownership of various claimants -- fail to explain the propensity of a financial statement restatement. We do find, however, that the probability of a restatement varies directly with restrictions against shareholder participation in the governance process.

STRONG BOARDS, management entreNchment, and Accounting RESTATEMENT


Despite serious agency problems that result from the power and information advantage enjoyed by corporate managers, shareholders of modern corporations typically lack both the ability and incentive to question management (Bainbridge 1995). Instead, shareholders delegate management oversight to the board of directors (the board), a group of agents empowered to monitor and discipline management on shareholders’ behalf. Corporate boards are often criticized as the source, rather than the solution, of corporate control failures, however. For example, Jensen (1993) notes that board members, often appointed by -- and therefore indebted to -- the CEO, are inclined to permit the CEO to advance management, rather than shareholder, interests. These circumstances promote a culture of acquiescence rather than effective oversight (Lorsch and MacIver 1989; Crystal 1991; Byrne 1996). Accordingly, effectiveness of corporate boards is a central issue in contemporary corporate governance literature.

When corporate boards fail to execute their fiduciary responsibilities, shareholders can intervene or even displace the management and the board. In practice, however, shareholders’ ability to do so differs considerably across firms (Gompers, Ishii, and Metrick 2003; hereafter GIM). In some cases, the legal and contractual environment makes it relatively easy for shareholders to replace the board and/or the management. In other cases, however, charter and bylaw provisions increase shareholder costs of challenging management, and generally restrict the ability of shareholder participation. Such provisions tilt the balance of power from shareholders to corporate insiders (GIM 2003).

The objective of this study is to compare and contrast associations between accounting restatements and shareholder rights, and between restatements and characteristics of the board of directors. A distinguishing feature of the study is the consideration of both mechanisms advocated by academics and policy-makers as remedies for control problems and statutory/chartered provisions that facilitate or restrict shareholder participation in the corporate governance process. This feature of the investigation permits us to consider the possibility that these mechanisms and provisions are compensatory as they relate to the incentives and consequences of accounting misstatements by corporate managers.

Similar to previous literature (Abbott, Parker, and Peters 2003; Argawal and Chadha 2005; Srinivasan 2005), the investigation is motivated by the supposition that weak corporate governance is at least partially responsible for recent high-profile financial reporting failures and accounting restatements (The U.S. General Accounting Office (GAO) 2002; Byrne 2002). Given this context, investigating the extent that corporate governance characteristics correlate with the incidence of accounting restatements potentially informs evaluations of the effectiveness of governance procedures often prescribed in response to accounting scandals and restatements (e.g., Sarbanes-Oxley Act (SOX) of 2002).

The analysis is interpreted in the context of two competing, but not necessarily mutually exclusive, characterizations of how corporate governance systems evolve and operate. First, the entrenchment hypothesis posits that managers design control systems that advance their private interests (Bebchuk, Cohen, and Ferrell 2004). More specifically, managers have incentives to construct governance systems that restrict shareholder participation – and thus, increase and preserve management’s influence -- in the policy-making process. Management’s ability to act on these incentives depends in part on legal and contractual provisions that are costly to modify and that define the role and influence of management relative to shareholders. These provisions, which may or may not be implemented by current management, distinguish firms cross-sectionally. If accounting misstatements indicate management negligence or malfeasance, then the entrenchment hypothesis predicts that accounting misstatements vary directly with contractual provisions that establish or preserve management’s dominant position.

Second, the substitution hypothesis presumes that firms, acting on behalf of all interested parties, choose among control alternatives to achieve a governance system – that is, a portfolio of control mechanisms and procedures – that minimizes combined control and agency costs (Agrawal and Knoeber 1996). Each entity faces unique control problems and circumstances, and therefore, the specific control procedures that shareholders use to monitor or discipline management vary cross-sectionally. This characterization implies that control systems need to be evaluated in the aggregate as a compensatory process, rather than as a collection of independent control procedures. Thus, to the extent that all firms choose an optimal mix of corporate governance mechanisms, systematic associations between a specific corporate governance mechanism and the incidence of accounting restatements are unlikely.[1]

To consider these two characterizations, we focus on potential tradeoffs between the strength of shareholder rights and the strength of the board, while controlling for other corporate governance mechanisms. We use the “Governance Index” (hereafter G-Index) -- developed by GIM (2003), along with derivative sub-indices advanced in other studies (e.g., Bebchuk, et al. 2004), as a composite measure of the strength in shareholder rights. For board strength, we consider a fairly comprehensive set of corporate board characteristics, including (but not limited to) independence of the board and its subcommittees, board size, whether or not the CEO sits on the nominating committee, whether an independent director serves as the board chair, the extent of interlocked directors, and the number of boards served on other companies (“busy” directors).

The extent that the members are independent monitors of management is of particular interest. This focus reflects the prominence of board independence in recently enacted or proposed rules governing public corporations.[2] We therefore extend existing research to consider independence of key oversight committees -- audit, nominating, and compensation committees -- in addition to considering independence of the full board. Recognizing that certain board strengths and other governance attributes are potential substitutes, we consider other governance characteristics, including the structure of CEO compensation, the age and tenure of the CEO, and ownership structure of the firm. In total, we consider in excess of twenty-five corporate governance attributes.

The empirical investigation focuses on accounting restatements announced by Standard & Poor’s (S&P) 1500 publicly-traded and most-widely held small, medium, and large U. S. firms from 1997 to 2004. These restatements relate to accounting errors that occurred during 1997-2002. The principal finding is that the vast majority of corporate governance indicators fail to play a significant role in the likelihood of an accounting restatement, either individually or in the aggregate. In particular, notwithstanding the popular enthusiasm for board independence as a mechanism for improving corporate governance, our analysis offers little support for the notion that board independence, or independence of subcommittees (audit, compensation, and nominating committees), is related to accounting restatements. Likewise, we find little support for associations between accounting restatements and governance mechanisms frequently advanced in prior studies and endorsed by governance experts and policymakers -- as examples, whether or not the audit committee has a financial expert, an audit committee comprised entirely of outside directors, board size, CEO-chair separation, the CEO’s equity incentives, ownership structure (equity ownership of the CEO, the board, blockholders, and institutions), whether an independent director serves as the board chair, the extent of the interlocked or “busy” directors.

The only governance characteristic that is reliably related to accounting restatements is the G-Index, an indicator of shareholder rights. In particular, the probability of an accounting restatement is greater for firms where legal and contractual provisions limit shareholder rights and thus encourage management entrenchment (Bebchuk et al. 2004).[3] This result is robust to considering the effects of industry, firm size, alternative governance procedures, and economy-wide changes in governance characteristics.

Finally, we find a positive association between the G-Index and a composite score constructed from nine board characteristics frequently considered as indicators of board effectiveness. Such evidence, which suggests a tradeoff between weaker shareholder protection and a stronger board, is consistent with the substitution hypothesis.

In sum, we find that legal and contractual provisions that influence shareholder rights are correlated with the incidence of accounting restatements. We can offer no evidence, however, that board independence or other recommended procedures alleged to improve corporate governance systems are related to restatements.

Failure to detect associations between corporate governance indicators and restatements can indicate inadequacies or measurement error in governance indicators or it can indicate that firms substitute one dimension of corporate governance with another. Regardless of which of these characterizations applies, the evidence documented in this study suggests the futility of incorporating a few dimensions of corporate governance into rules and regulations. If the measurement error explanation applies, then efforts to change some identifiable measure of corporate governance (such as board independence or requiring financial experts) lead to “apparent” but not substantive improvement in corporate governance. If the substitution explanation applies, then improvement on one dimension of corporate governance potentially leads to deterioration on another dimension, which implies a suboptimal and more costly mix of governance mechanisms. Thus, the study informs the debate about whether and how corporate governance can be legislated.

The remainder of the paper is organized as follows. Section II discusses the relevant literature and presents specific hypotheses. Section III explains the methodology, and Section IV contains data collection procedures. Results are reported and discussed in Sections V. Concluding remarks are in Section VI.


A. Financial Misstatements versus Restatements

The objective is to document associations between accounting misstatements and corporate governance characteristics. Since not all misstatements are detected and disclosed, previous studies adopt either of two sampling approaches. The first is to examine accounting restatements as in Abbott, et al. (2003), Kim (2004), Kinney, Palmrose and Scholz (2004), Srinivasan (2005), and Agrawal and Chadha (2005). The second approach is to explore cases of fraud alleged by the U. S. Securities and Exchange Commission (the SEC) identified during the Commission’s enforcement of Accounting and Auditing Enforcement Releases (AAER).[4] Studies in the second category are Beasley (1996), Dechow, Sweeney and Sloan (1996), Erickson, Hanlon, and Maydew (2004), and Farber (2005). In the case of AAER, willful misrepresentation is alleged by the SEC but not necessarily acknowledged by the company. In the restatement case, however, the firm acknowledges the prior financial reporting error, but not necessarily intent to mislead investors.

Both samples are noisy indicators of intentional misstatement of financial conditions, and likely partition the observations with error. First, some restatements can be “honest mistakes” rather than attempts to manage earnings. Second, since not all misstatements are discovered or investigated, undetected misstatements are classified incorrectly, leading to a sample which does not fairly represent the population. Although the AAER sample likely alleviates the first problem, it exacerbates the second. Furthermore, since the SEC can examine only a small subset of publicly traded firms based on less-than-transparent selection criteria, potential selection bias could be an issue.[5] Beasley, Carcello, and Hermanson (1997) report that firms investigated under AAER during 1987-1997 were typically small (total assets and sales well below $100 million), and 78% of them were not traded on large stock exchanges (that is, the NYSE and ASE). Thus, neither of these sampling approaches dominates the other.

Existing literature supports the notion that restatements are non-trivial events which usually trigger substantial declines in equity price (Palmrose, Richardson, and Scholz 2004). The GAO (2002) and Palmrose et al. (2004) report that firms restating the financial reports lose about 9 to 10% of shareholder value during the 2 or 3 trading days surrounding the announcement of the restatement. The magnitude is comparable to price reaction to AAER fraud announcements of -8.8% documented in Dechow, et al. (1996). Furthermore, similar to the case of AAER, restatements often trigger subsequent SEC investigation, lawsuits, removal of management, or bankruptcy (Kim 2004). Such evidence supports the use of published accounting restatements to study financial statement misrepresentations.

In sum, we assume that the incidence of restatements is a noisy proxy of between-firm differences in the probability of an accounting failure. At the minimum, a restatement is a significant event which indicates low-quality financial reporting. In the extreme, it indicates an intent to mislead investors. In either case, it is evidence of mismanagement abetted potentially by weak corporate governance (GAO 2002).[6]

B. Existing Studies on Financial Misstatements and Restatements

Exhibit 1 summarizes prior studies that examine associations between corporate governance characteristics and the incidence of accounting failures, accounting restatements, or SEC AAER actions for alleged accounting violations. Evidence in Dechow, et al. (1996) suggests that 86 firms subject to SEC enforcement actions during 1982-1992 are more likely to have a board dominated by insiders, a CEO who also serves as the board chair or who is from the founding family; and are less likely to have an outside blockholder or an audit committee, than control firms that do not experience enforcement actions. Beasley (1996), who examines 75 SEC AAER firms during 1979-1990 in a multivariate (logit) context, reports that the probability of SEC action is unaffected by CEO-Chair separation or the existence of an outside blockholder, but increases with weaker board independence (measured as the percentage of independent and grey directors on the board) and absence of an audit committee.[7] Using a sample of AAER 87 firms during 1982-2000, Farber (2005) reports results consistent with those in Dechow, et al. (1996) with respect to board independence, CEO-Chair duality, and outside blockholder, but not audit committee independence. In contrast, Abbott, et al. (2003) find that, for 88 restatements during 1991-1999, the probability of accounting failure is unaffected by board independence, CEO-Chair separation, or existence of outside blockholder, but the probability varies positively with board size and inversely with audit committee independence and presence of an accounting expert. Agrawal and Chadha (2005), who compare 159 fiscal year 2000 restatement firms with industry- and size- matched firms, report that none of the key governance characteristics -- notably board and audit committee independence, blockholder, CEO-Chair separation, CEO ownership, and the provision of non-audit services by outside auditors -- is related to the probability of earnings restatements. The authors do find, however, that the probability of restatements is lower for companies where the board or audit committee has an independent director with financial expertise, and higher for companies with a CEO from the founding family. Finally, Larcker, Richardson, and Tuna (2004) use proprietary data to examine various aspects of managerial behavior and corporate performance (such as abnormal accruals, CEO pay, debt ratings, Tobin’s Q, investment, class action lawsuits, and accounting restatements) for fiscal year 2002.[8] The authors conclude that typical indicators of corporate governance used in academic research and institutional rating services have limited ability to explain managerial behavior and organizational performance. As a whole, results reported in these prior studies do not offer consistent evidence regarding associations between governance characteristics and the incidence of accounting irregularity.

This paper builds on previous studies in three dimensions. First, we consider a measure of corporate takeover defense, or the G-Index, which is often considered an indicator of management entrenchment (e.g., Bebchuk et al. 2004; Fahlenbrach 2004). Second, we consider a more comprehensive set of corporate governance mechanisms than those employed in most prior studies. Larcker, et al. (2004) suggest that previous studies do not permit a consistent and integrated set of inferences because they use different, relatively small sets of conveniently available indicators for corporate governance. This point is particularly important to the extent that governance mechanisms are substitutes. Finally, we address a larger sample of both restatement and non-restatement firms over a longer period (1997-2002) than most prior studies. The increased statistical power is useful because restatements occur infrequently.

C. Shareholder Rights and Governance Index

GIM (2003) construct a governance measure, designated the G-Index, from 22 firm-level charter and bylaw provisions and six state takeover laws. One point is added for each provision or state law that restricts shareholder rights; duplications among firm-level provisions and state laws yield 24 unique provisions. Thus, the index ranges from zero to 24, with zero being the most shareholder-friendly environment. GIM (2003) report that firms with low G-Index (designated “Democracy” firms) have higher long term shareholder returns, higher firm value (Tobin’s Q), greater accounting profits, and higher sales growth than firms with high G-Index (“Dictatorship” firms).[9] While the robustness of results reported in GIM (2003) is debatable (Bebchuk, et al. 2004; Core, Guay, and Rusticus 2005; Cremers and Nair 2005), a number of studies establish that the G-Index, or sub-indices derived from the G-Index, indicates the strength of corporate governance on at least some dimensions (Cremers, Nair, and Wei 2004; Cremers and Nair 2005; Klock, Mansi, and Maxwell 2004).

Much of the analysis is interpreted in the context of two, not necessarily mutually exclusive, perspectives about how the balance between the potentially conflicting interests of insiders and outside shareholders relates to accounting failures. First, the substitution hypothesis postulates that firms choose among control mechanisms to construct a portfolio of control procedures and mechanisms that minimizes combined control and agency costs (Agrawal and Knoeber 1996). As examples, Almazan and Suarez (2003) argue that weak boards and large severance pay are substitutes, and Cyert, Kang, and Kumar (2002) consider external takeover threats and internal governance mechanisms as substitutes. The logic behind the substitution hypothesis suggests that firms balance the costs and benefits of alternative control devices such that corporate governance mechanisms cannot be evaluated in isolation.

Second, the entrenchment hypothesis implies that limitations or restrictions of shareholder rights, once they are implemented for whatever reason, permit managers to behave opportunistically and to engage in wealth transfers from shareholders to managers - for example, by awarding themselves excessive compensation (Fahlenbrach 2004). Thus, the entrenchment hypothesis predicts an inverse association between shareholder rights and managerial malfeasance. We use these two perspectives to consider associations between governance characteristics and accounting restatements.

The first hypothesis considers cross-sectional differences in measures of shareholder rights during the period of financial statement misstatement.

Hypothesis 1: Restrictions on shareholder participation (as indicated by the G-index) and the incidence of financial reporting misstatements are unrelated.

Higher G-Index indicates more provisions that discourage shareholder participation, and therefore, rejecting the null in favor of a positive association between the G-Index and accounting restatement is consistent with management entrenchment.[10]

D. Board Independence and Board Strength

The G-Index indicates the relative power between shareholders and insiders (the management and the board of directors), but another potentially important issue is the extent that the board provides objective, disinterested monitoring and oversight of management (Fama 1980; Fama and Jensen 1983; Jensen 1993). Thus, the second hypothesis relates strong board (in the sense that the board of directors represents a strong corporate governance mechanism) with the likelihood of a firm misstating financial reports.

Hypothesis 2: Strong board characteristics and the incidence of financial reporting misstatements are unrelated.

Director independence is a popular gauge of board effectiveness. Despite its intuitive appeal, however, existing evidence is less than persuasive regarding the effectiveness of board independence as a control device (Bhagat and Black 2002). In particular, results are mixed regarding whether board independence is effective for disciplining management (Weisbach 1988), for increasing share values in the context of takeovers (Byrd and Hickman 1992; Brickley, Coles and Terry 1994; Cotter, Shivdasani, and Zenner 1997), for setting executive compensation (Core, Holthausen, and Larcker 1999; Cyert, et al. 2002), and for enhancing firm performance (Baysinger and Butler 1985; Hermalin and Weisbach 1991; Yermack 1996; Bhagat and Black 1999; 2002).

Recent public policy debate considers independence of not only the full board, but also the key oversight committees (audit, nominating, and compensation committees). As examples, the SOX Act specifically requires that listed firms have an audit committee composed solely of independent directors (one of whom is an expert in accounting or finance). Also, the NYSE, as well as various watchdog organizations and corporate governance experts, advocate that not only the audit committee, but also the compensation and nominating committees, be fully independent of management (e.g., California Public Employees Retirement System (CALPERS 1998)).

The literature also suggests that the following characteristics are relevant to board effectiveness beyond board independence: whether the CEO is also the board chair (Jensen 1993), whether the CEO is on the nominating committee (Jensen 1993), whether an independent director is the board chair (CALPERS 1998), the extent of interlocked directors and other company directorships (Core, Holthausen, and Larcker 1999), and the number of directors on the board (Jensen 1993; Yermack 1996). We consider the impact of these attributes on accounting restatements both individually and in the aggregate.

Since more than one attribute contribute to a strong board, we also consider a composite score – similar to the G-Index -- denoted B-Index, which is computed using nine board attributes. In particular, B-Index increases by one if: (1) more than 2/3 of the board is comprised of independent directors; (2) all audit committee members are independent directors; (3) all compensation committee members are independent directors; (4) all nominating committee members are independent directors; (5) the CEO is not the board Chair; (6) the Chair is an independent outside director.[11] B-Index also increases by one if (7) the board size is less than the median of the distribution for all firms (adjusted for firm size and time); (8) the board interlock is less than the median of all firms (adjusted for firm size and time); or (9) the mean number of other boards served by a director is less than the median of the distribution of means for all firms (adjusted for firm size and time).[12] We note that governance experts can disagree about whether some of these attributes used in the construction of the B-Index indicate effective governance. As examples, the consequences of changing board size, separating the duties of the CEO and the board chairman, or proscribing the number of directorship held by board members are not straightforward. Even so, we propose B-Index as a summary indicator of board strength that is comparable to the G-Index.

Strict application of the substitution hypothesis suggests no systematic association between specific corporate governance mechanisms and the incidence of accounting failures – to the extent that all firms choose an optimal, potentially unique, mix of governance mechanisms. We cannot be confident that all firms employ an optimal portfolio of corporate governance procedures, and therefore, the substitution hypothesis cannot feasibly be tested directly. Instead, we address a less ambitious proposition that tests whether strong boards and shareholder rights are complementary.

Hypothesis 3: Strong board characteristics and restrictions on shareholder participation are unrelated.

The G-Index serves as a proxy for shareholder rights, and the B-Index indicates the relative strength of the board. Thus, in the context of the substitution hypothesis, rejecting the null in favor of a positive association between the G-Index and B-Index supports the hypothesis that strong boards and shareholder rights are complementary.

III. methodology

A. Specification

We use the following logistic regression to address hypotheses 1 and 2 (time t subscripts omitted).

[pic], (1)

(+) (-)

where Restate = 1 if firm i restates its financial reports; 0, otherwise. Directions of associations between the probability of restatement and the independent variables under the alternative hypotheses are shown parenthetically. Variable definitions are in Exhibit 2. We emphasize that independent variables indicate circumstances at the time when the misstatement occurs, not the time when the misstatement is announced.

The G-Index (GIM 2003) indicates the extent that contractual and legal provisions restrict shareholder rights. Recently, Bebchuk et al. (2004) argue that only six of the 24 provisions included in the original G-Index are principally correlated with firm values.[13] Thus, we consider an alternative measure, designated the entrenchment index (E-Index), comprised of these six provisions.

The variables [pic] consider nine attributes of a strong board discussed before (N=9). The primary specification considers all nine measures individually and jointly. Recall that we also use a composite measure, designated B-INDEX, that is constructed from all nine board characteristics. Independence of the board (subcommittees) is measured as the fraction of independent directors on the board.[14]

We consider two modifications when estimating specification (1). First, because not all firms have separate nominating committees, we decompose NOM_INDEP (independence of the nominating committee) into two variables – an indicator variable (denoted DNOM), which equals one when the firm has a separate nominating committee (0, otherwise), and an interaction DNOM*NOM_INDEP. Accordingly, the estimate on DNOM indicates the incremental effect of the presence of a separate nominating committee, and DNOM*NOM_INDEP indicates the effect of nominating committee independence given that the firm has a nominating committee.

Second, recent recommendations are that boards have “supermajority” (e.g., more than 2/3) independent directors and that oversight subcommittees (such as the audit and the nominating committees) consist entirely of independent directors (e.g., Council of Institutional Investors 2003). Thus, we consider two alternative summary measures of independence. The first, denoted FULL_INDEP, is a dummy variable set equal to 1.0 (0, otherwise) when greater than 2/3 of the board members are independent and each member of the audit, the nominating, the compensation committees is independent. The second, denoted MAJ_INDEP, is a dummy variable set equal to 1.0 (0, otherwise) when the full board and each of the three oversight committees has majority independent directors (i.e., more than 50% of committee/board members are independent).

B. Other Governance Characteristics and Control Variables

Variables designated [pic] consider financial and corporate governance firm-specific characteristics that prior literature suggests can be related to the dependent and the primary explanatory variables. Some of these variables -- as examples, ownership structure and CEO compensation structure — can be construed as dimensions of the governance system.

Although the literature suggests that executive equity ownership aligns manager with shareholder preferences (Jensen and Meckling 1976), arguments and evidence advanced recently suggest incentive effects to the contrary. For example, Erickson et al. (2004) report that the probability of accounting fraud varies directly with the portion of executive compensation that is stock-based. Moreover, evidence in Peng and Roell (2004) suggests that compensation in the form of stock options increases the probability of securities class-action lawsuits and magnifies incentives to manipulate earnings. In general, such results imply that large management equity stakes, particularly in the form of stock options, can induce managers to maximize short-term earnings at the expense of long-run profitability.

Following Erickson et al. (2004), we consider STK_COMP, the ratio of equity-based compensation (stock option and restricted stock awards) to total current compensation.[15] Following Peng and Roell (2004), we include OPTVESTED, vested or exercisable stock options as proportion of total outstanding shares.[16] We do not include levels of compensation, however, since neither study finds levels to be a significant factor. Notwithstanding the findings of Peng and Roell (2004) and Erickson et al. (2004), the conventional view is that granting equity-based options aligns manager incentives to maximize shareholder wealth (Murphy 1995). Thus, we offer no predictions about the impact of CEO ownership (CEO_OWN), inside director ownership (EMPDIR_OWN), the outside director ownership (OUTDIR_OWN), and equity-based compensation (STK_COMP, OPTVESTED) on accounting misstatements. The literature is similarly ambiguous with respect to the effects of CEO’s age (CEO_AGE) (e.g., Gibbons and Murphy 1992) or director’s age (DIR_AGE), although some argue against boards comprised of elderly directors (e.g., Core, Holthausen, and Larcker 1999). Hence, we offer no predictions about directions of effects for these variables.

On the other hand, a widely-held view is that large, independent outside blockholders, and institutional investors, can effectively monitor and discipline management (Shleifer and Vishny 1986; 1997; Jensen 1993). Thus, as with Beasley (1996) and Dechow et al. (1996), we expect a negative association between the incidence of accounting misstatements and the number of outside blockholders (BLOCK) or institutional owners (INST_OWN).

Various studies report that the probability of accounting failure is less when the audit committee includes an accounting expert (Abbott, et al. 2003; Agrawal and Chadha 2005; Farber 2005). We therefore include a variable designated EXPERT to distinguish firms where an independent accounting expert serves on the audit committee. We anticipate that EXPERT varies negatively with the likelihood of a restatement.[17] Dechow et al. (1996) and Agrawal and Chadha (2005) report that the likelihood of an accounting failure increases when the CEO or the board chair is from the founding family; we use an indicator variable (FOUNDER) to distinguish such firms.

Other control variables follow previous studies (Abbott, et al. 2003; Richardson, Tuna, and Wu 2002; Larcker, et al 2004): BIG6 (large accounting firms), LOGSIZE (log of beginning total assets), GROWTH (growth rate of sales during the previous two years), DIV_YIELD (dividend yield: dividend per share/stock price per share), MARGIN (net income/sales), FREECF (free cash flow/total assets), and LEVERAGE (interest-bearing debt/total assets). All variables except for BIG6 are measured at the beginning of the misstatement year, and observations are removed when GROWTH, MARGIN, LEVERAGE, DIV_YIELD, or FREECF is in the top or the bottom one percent of the distribution for all sample firms.[18]

IV. Data and Sample Description

A. Data

The primary sample consists of 2,431 firms in the S&P 1500 stock indices (S&P 500, S&P MidCap 400 and SmallCap 600, hereafter the S&P 1500) during 1997-2002; 231 of these firms restate financial reports. Table I summarizes the sample selection procedure which is described in greater detail below.

[Insert Table I here]

The sample is the intersection of four data sets. The first, compiled by the Investor Responsibility Research Center (IRRC), contains Board of Director characteristics such as BOD_INDEP and BOD_SIZE. These data are taken from publicly-available 1997-2002 proxy statements issued in anticipation of the annual meetings for the S&P 1500 companies.

The IRRC is also the source of the G-Index measure, which is constructed using Corporate Takeover Defenses [Rosenbaum 1990, 1993, 1995, 1998, 2000, and 2002]. This publication covers S&P 500 firms as well as large U.S. corporations listed in Fortune, Forbes, and Business Week. These firms arguably represent more than 90 percent of the total capitalization of the combined New York Stock Exchange (NYSE), American Stock Exchange, and NASDAQ (GIM 2003). The G-index measure is relatively stable within each firm, although it varies considerably across firms (GIM 2003).[19]

The third dataset is from a study prepared and submitted to the U. S. Congress by the General Accounting Office (GAO 2002). The study lists 919 accounting restatements by 845 public companies during January 1, 1997 to June 30, 2002, although careful investigation reveals that some of the alleged restatements are inconsistent with the objectives of this study (see below).

These three datasets are combined and merged with the fourth, COMPUSTAT financial data, to obtain 2,431 firms; 231 of these firms disclose one or more restatements according to the GAO. We designate restatement firms as Test firms and the remainder as Control firms.

We find that some GAO restatements are early adoptions of accounting rule changes and opinions or revisions or those that yield no change in earnings. We exclude 43 such observations.[20] Finally, we remove 41 restatements where explanatory variables used in equation (1) are missing.[21]

Notice that restatement is an infrequent event, whereas the statistical power of the logit model increases in the frequency of occurrence. To enhance the statistical power, we hand-collect additional restatement announcements during the post-GAO period, that is, from July 2002 to December 2004. This effort yields additional 55 firms with complete data, and with 1997-2002 accounting errors announced during 2002-2004. Finally, we remove 17 firms with dual-class securities because G-Index for these firms may not be comparable to those with single-class security due to wide variety of voting and ownership differences (GIM 2003). In sum, we address a sample of 1,672 firms -- 185 firms that restate (Test sample) and 1,487 firms that do not restate (Control sample) financial reports.

An investigation of stock price response to the restatement announcement (not reported) indicates substantial reduction of shareholder value – sample firms lost, on average, 13.7% of equity value during the 30 days leading up to the announcement.[22] This evidence is consistent with a characterization that restatement disclosures, at a minimum, indicate serious mismanagement.

The potential consequences of focusing on the S&P 1500 firms are relevant for evaluating how the empirical results apply generally. More specifically, conclusions may not apply to all U.S. public firms to the extent that these firms differ from all firms. On the other hand, S&P 1500 firms are selected from large, medium, and small firms and represent approximately 85% of market capitalization of all publicly traded firms. Thus, they are economically significant. Furthermore, since S&P 1500 firms are more widely-held than other firms, they are of considerable interest to the investing public. Finally, the availability of computer-readable data for these firms permits investigation of a comprehensive set of governance characteristics.

Straightforward use of the pooled data is inappropriate because serial dependence in corporate governance data potentially violates i.i.d. assumptions.[23] Thus, we construct the Test sample using a single observation for each restatement firm. For Test firms where financial reports are restated in only one year, we use firm data for the restatement year. For firms with multiple years misstated, we consider the year prior to the most recent year that is misstated.[24] To identify the sample of Control firms, we select one fiscal year for each firm such that the fiscal years of the Control firms are closely matched to those of the Test firms.[25] Matching by time period considers significant time trends in corporate governance characteristics (to be described below), while maintaining the original sample of 185 Test firms and 1,487 Control firms.

As in Richardson et al. (2002), we prefer the full sample analysis to a matched-pair design, because the latter approach cannot dominate the former in terms of statistical power or representativeness of the sample.[26] Two variables, EXPERT and FOUNDER, are unavailable from the computer-readable database, however. Since these variables are hand-collected, we consider these variables using a matched-pair design following conventional procedures (e.g., Beasley 1996). More specifically, for each Test firm, we select a single companion firm that is similar in four dimensions: industry, fiscal year, stock exchange membership, and firm size (within 15% of the Test firm’s log total assets). This procedure yields 178 pairs, after losing seven test firms that cannot be matched using these criteria.[27]

B. Descriptive Statistics

Panel A in Table II shows descriptive statistics, and Panel B displays correlations between governance characteristics and financial variables.

[Insert Table II here]

Mean and median G-Index, 9.152 and 9.0 respectively, are in line with those reported in GIM (2003). Mean (median) board independence (row 2) is 63.3% (66.7%) -- the median coincides with the minimum independence level recommended by watchdog organizations (e.g., Council of Institutional Investors 2003). The degree of independence is greater for audit committees (AUD_INDEP) where the mean (median) is 86.2% (100%).

Information not tabulated indicates that all sample firms have an audit committee, and virtually all (98.9%) have a compensation committee, but only 62.9% have a separate nominating committee. Independent directors comprise the majority of the board and the three oversight committees (MAJ_INDEP=1) for 44.0% of sample observations. Instances where at least two-thirds of the Board is independent and all members of the three oversight committees are independent (FULL_INDEP=1) are 16.5% of the sample.

Finally, median board size (BOD_SIZE) is nine, and median audit committee size is three, the latter being consistent with the minimum audit committee size recommended by the Blue Ribbon Committee (1999). Duties of the CEO and the board chair are segregated for 27.3% of the sample (SEP_CHAIR=1). At least one financial expert (EXPERT=1) sits on the audit committee for 69.7% of sample observations.

[Insert Figure 1 here]

Figure 1 shows 1997-2002 cumulative changes in board independence (1A), board size (1B, 1C), frequency of board meetings (1D), and other board and director characteristics (1E). Data are aligned according to the year the annual meeting is held (which typically occurs two to three months after the end of the fiscal year).[28] Because the sample composition differs in each year, cross-sectional levels of governance variables are not comparable inter-temporally. Thus, we compute firm-specific annual percent change of each variable for each year, and then cumulate the mean changes through 2002. Except for the change in mean number of board directorships held (BUSYBOARD in Figure 1E), all cumulative increases or decreases during 1997-2002 are statistically significant ((( 0.01).

Some argue that board activity indicates board effectiveness – that is, effective boards (or committees) meet more frequently than ineffective boards (e.g., Lavelle 2002; Abbott, et al. 2003). Figure 1D indicates that board and committee activity increases from 1997 to 2002. Notice, in particular, that audit committee meetings more than double during the period. Also, audit committee meetings increase by 34% shortly after the issuance of the Blue Ribbon Committee report (1999) which recommends regular meetings between the audit committee and management. Related to this, and consistent with the supposition that restatements are significant events, we also find that boards and audit committees of restatement (Test) firms meet more frequently during and after restatement disclosures (result not tabulated). It is not surprising, however, that boards and audit committees meet more frequently when a restatement is detected or disclosed. Since the direction of the causality is unclear -- that is, whether board activity precipitates or responds to an accounting crisis -- we do not use the frequency of board meetings in the empirical analysis.

Overall, the evidence in Figure 1 indicates that boards of all firms are increasingly independent and active, and more concerned about financial reporting during the 1997 to 2002 period. This feature of the data, which is likely, in part, due to more stringent stock exchange listing requirements imposed during 2000-2002, underscores the importance of controlling for time trends in studies that use governance characteristics. The evidence also suggests that there has been a systematic and broad-based improvement in corporate governance characteristics well ahead of Sarbanes-Oxley Act of 2002.

[Insert Table III here]

Table III Panel A indicates that the Test and the Control samples are comparably distributed in calendar year (the (2- = 2.39; p = 0.79). Of course, this similarity is by construction as control observations are selected to achieve a comparable distribution. Panel B indicates that the sample distributions are also comparable across stock exchanges ((2 = 4.70; p = 0.32), even though exchange membership is not used to match the samples. Panel C displays the GAO classification of the specific accounting error that precipitated the misstatement.[29]

V. results

A. Primary results

[Insert Table IV here]

Table IV reports estimates for four specifications of expression (1). The first two specifications differ in terms of how BOD/oversight committee independence is considered. In the first column, all four indicators of board/committee independence (BOD_INDEP, AUD_INDEP, COM_INDEP, NOM_INDEP) appear individually. In the second column, these four variables are combined into a single indicator variable, FULL_INDEP. Estimates not tabulated using MAJ_INDEP are comparable to those using FULL_INDEP. The specification in column 3 further combines the nine variables indicating board strength into a single index, B-Index. The specification in the last column is comparable to column 1 except that it includes three additional variables, INDEP_CHAIR, CEO_ON_NOM, and BUSYBOARD. Owing to lack of data for these three variables in 1997, specifications 3 and 4 can be computed only for the 1998-2002 observations. All specifications include dummy variables to extract mean effects of: (1) industry, where industries are delineated into 13 groups as in Barth, Beaver, Hand, and Landsman (1999);[30] (2) fiscal year;[31] and (3) stock exchange – NYSE, ASE, NASD, and OTC. For brevity, we do not report the estimates for these dummy variables. Statistical significance is evaluated using one-tailed levels when an alternative hypothesis is specified, and using two-tailed levels, otherwise.

Perhaps the most conspicuous aspect of the results in Table IV is the absence of statistically significant associations for governance variables that are frequently advanced in prior studies and commonly used in practice to evaluate board effectiveness. More specifically, measures of board and committee independence (BOD_INDEP, AUD_INDEP, COM_INDEP, NOM_INDEP, FULL_INDEP), stock ownership (CEO_OWN, OUTDIR_OWN, EMPDIR_OWN, BLOCK, INST_HOLD), CEO incentives (STK_COMP, OPTVESTED) and other board characteristics (SEP_CHAIR, INTERLOCK, BUSYBOARD, INDEP_CHAIR, DIR_AGE, CEO_ON_NOM) are not statistically significant.[32] Moreover, the positive association between B-INDEX and restatements contradicts expectations.

The large number of independent variables included in the specifications raises concerns about whether multicollinearity undermines the ability to achieve statistical significance. The evidence does not support this explanation, however. In particular, both variance inflation factors (VIF) and condition indices (Belsley, Kuh, and Welsch 1980) indicate that multicollinearity is not a concern.[33] Excluding specific variables from the specification, or combining variables with correlations greater than 0.40 (e.g., BOD_INDEP, AUD_INDEP, COM_INDEP, NOM_INDEP), yields comparable parameter estimates and significance levels. Stepwise regression indicates that only three governance variables and two control variables (G-INDEX, DIR_AGE, BUSYBOARD, GROWTH, and FREECF) are related to restatements at the five percent significance level. Finally, the low pseudo-R2 (less than 6%) for all specifications is comparable to evidence in Larcker, et al. (2004) who also report low explanatory power for governance indicators.[34]

In contrast with results for governance characteristics, all specifications reveal a positive association between the probability of accounting restatement and the G-Index, which indicates the extent that statutory or corporate provisions entrench management. The estimate on G-Index 0.108 (p = 0.001) in the first column suggests that, on average, the odds of accounting misstatement increase by about 11% (e0.108 = 1.114) for each additional provision restricting shareholder rights. Thus, accounting misstatements are more likely in firms where legal or contractual provisions restrict shareholder rights.

Note that estimates for financial control variables consistently indicate statistically significant associations between 1997-2002 accounting restatements and growth (GROWTH), free cash flow (FREECF), and size (LOGSIZE). That is, restatements are more likely to occur for large firms (which, to our knowledge, is not documented elsewhere) and for high growth and cash-strapped firms (as documented in Richardson, et al. 2002 and Larcker, et al. 2004). Finding statistical significance for financial control variables undermines arguments that failure to find statistically significant associations for corporate governance variables is attributable to lack of statistical power.

To summarize, results consistently indicate that accounting misstatements vary directly with the corporate and statutory restrictions that encourage management entrenchment, but misstatements are unrelated with corporate governance indicators commonly advanced both in academic research and practice. Moreover, associations are unlikely to be attributable to multicollinearity or lack of statistical power. Thus, based on the evidence in Table IV, we reject Hypothesis 1 in favor of the alternative hypothesis restatements are more likely in firms with provisions that restrict shareholder rights, but we cannot confidently reject Hypothesis 2 that restatements and board, director or firm characteristics commonly associated with strong governance are unrelated.

B. Matched-pair results

[Insert Table V here]

Estimates for the matched-pair sample, displayed as Table V, support these conclusions, although, perhaps owing to the smaller sample, significance levels are lower for G-Index and financial control variables.

Recall that we are particularly interested in associations for two variables – EXPERT and FOUNDER – which cannot be feasibly collected for the larger sample, but which are identified in previous studies. Estimates indicate that the probability of a restatement is related to neither the presence of a financial expert on the audit committee nor whether the CEO or the board chair is from the founding family.

Failure to find a statistically significant association for financial expert (EXPERT) is surprising, as recent evidence using the accounting expertise measure suggests otherwise (Abbott, et al. 2003; Agrawal and Chadha 2005; Farber 2005). We find that the proportion of restatement firms with an expert on the audit committee is nominally higher than the proportion of the Control firms (70.8% vs. 68.5%), but the difference is not statistically significant for univariate comparisons. To investigate further, we consider an alternative, more restrictive, definition of “accounting expert” proposed initially by the SEC during the SOX deliberations. Specifically, following DeFond, Hann, and Hu (2005), we define an accounting expert as a CPA, CFO, controller, or a chief accounting officer. Although the proportion of audit committees with this more narrowly-defined accounting expertise measure is lower for the restatement firms (24.7% vs. 29.8%), the difference is not statistically significant (t = 1.07 for a univariate comparison and p = 0.32 in the multivariate logit).

C. G-Index, E-Index, and sub-indices

Given consistent evidence that restatements are related to the G-Index, we investigate the extent that the source of the association can be attributed to specific provisions. GIM (2003) partition shareholder protection provisions into five categories: Delay (tactics designed to increase the cost of hostile takeover); Protection (protection of officers and directors from liability); Voting (limitations on shareholder voting rights); Other (direct takeover defense measures, e.g., poison pills); and State (state anti-takeover laws). Several provisions in the category designated State duplicate those in the other four categories. Thus, we do not examine the State category separately. We anticipate positive associations between accounting restatements and G-index components.

[Insert Table VI here]

Table VI displays a specification where G-Index components are considered. Control variables are included in these specifications but not reported in the table. Results for the E-Index, the composite governance measure advanced in Bebchuk, et al. (2004), are in the first row.

We find positive parameter estimates for all five indices, but only estimates for E-Index, Delay and Other are statistically significant (ά < 0.05). The index designated Delay (column 1, p-value=0.02) includes four provisions which delay takeovers or restrict the ability to hold shareholder meetings. The index designated Other (column 2, p-value=0.01) includes six provisions (anti-greenmail, poison pill, silver parachutes, pension parachutes, directors’ duties, fair price) designed to increase the cost of a takeover. Observe that the magnitudes of estimates for these two indices are greater than for the aggregate G-Index. For example, the estimate 0.283 (5th column) indicates that, on average, the odds of an accounting restatement increase by about 32.7% for each provision that is classified in the category designated Other. (Recall that the odds of accounting restatements increase by about 11% for each additional restriction on shareholder rights that is considered in the aggregate G-index.) Estimates for Protection and Voting indices are not statistically significant at acceptable confidence levels. Thus, provisions more directly related to delaying or increasing the costs of takeovers are more significantly associated with accounting failures than other limitations on shareholder rights or participation.

Finally, the magnitude of the estimate on the E-Index, although statistically significant (( ( 0.05), is neither as large as the estimate on Other nor as statistically significant as the G-Index (Table IV). The implication is that the E-Index does not dominate both the G-Index and other sub-scores for predicting accounting restatements.

D. Trade-offs between board characteristics and shareholder participation

Hypothesis 3 considers whether strong boards and the possibility of direct shareholder participation (takeover threats) are substitutes. If so, then we expect that G-Index and B-Index are positively correlated. Table VII shows estimates for a regression of B-Index on G-Index, after controlling for firm size, industry, year, and stock exchange. In column 1, B-Index is regressed on G-Index and firm size, defined as the log of beginning total assets. Industry, fiscal year, exchange membership, as well as size, are considered in column 2. Finally, column 3 considers a specification where the B-Index and G-Index are adjusted by subtracting the mean index for the firm’s corresponding size decile. In all three specifications, a positive correlation obtains between B-Index and G-Index, indicating that stronger board is correlated with stronger takeover defense (weaker shareholder rights).

We recommend a cautious interpretation of this result, as the evidence does not definitively support the substitution hypothesis. In particular, we lack the ability to ascertain whether the prevailing mix of governance mechanisms is optimal for a specific firm. Moreover, we cannot consider an exhaustive set of governance mechanisms, and the choice of measures used to construct the indices can be debated. Despite this caveat, the result suggests that the statutory/charter provisions and board strength function as compensatory mechanisms.

[Insert Table VII here]

E. Alternative Procedures

Additional procedures ensure that results are robust to alternative specifications and procedures.[35] First, we use security price reactions to restatement announcements to consider whether results are obscured by the inclusion of restatements that are either trivial or anticipated. In particular, restricting the analysis to 116 firms that experience negative three-day stock returns around the restatement disclosure (63.0% of the restatement sample) yields comparable results.

We also consider the following variables which arguably can be considered along with those addressed in Table IV -- the CEO’s tenure, length of time since the firm was on stock exchange, whether the CFO sits on the audit committee, and whether the CEO is on the compensation committee, and audit fees.[36] Incorporating these variables reduces sample sizes substantially, but does not materially influence interpretations of the results. Thus, we do not tabulate results for specifications that include these variables.


We investigate a comprehensive list of corporate governance characteristics and a large sample of accounting restatements. These restatements are significant events in that we find that sample firms experience 13.7% loss of shareholder value around the time of the restatement disclosure. Despite the significance of the event, our investigation offers relatively little evidence that accounting restatements are associated with remedies commonly endorsed by both academic and practicing corporate governance experts and, in some cases, imposed by policy-makers. Such results are consistent with at least three, not necessarily mutually exclusive, explanations.

The first explanation is that, in contrast with assertions that accounting restatements are attributable to weak corporate governance (GAO 2002, Byrne 2002), commonly recommended procedures (e.g., board/committee independence and separation of the CEO from the board chair) are ineffective remedies for curbing financial misstatement. A second explanation, which is suggested by the substitution hypothesis, is that the recommended practices and procedures are effective in some situations, but they are not appropriate for all firms. Given this context, mandating specific governance requirements potentially causes firms to reduce reliance on less costly governance mechanisms. A third explanation is that the effectiveness of corporate governance as a check on accounting misstatements is determined primarily by board or director attributes that are difficult to observe and measure – e. g., individual or group values and culture. That is, empirical measures of corporate governance fail to capture the richness and complexity of effective board oversight.

Regardless of which of these explanations applies, our evidence raises questions about whether effective corporate governance can be imposed through regulation (Bhagat and Black 2002; Anderson and Bizjak 2003; Homlstrom and Kaplan 2003; Romano 2004). The evidence also elevates concern about whether the public media create, and regulators respond to, political crises in ways that lack substance but advance their commercial or political objectives (Watts and Zimmerman 1979; Romano 2004).

The evidence does suggest that firms with significant restrictions against shareholder participation have greater propensity to commit accounting misstatements, however. This evidence – particularly when considered along with evidence reported elsewhere that firm performance varies inversely (GIM 2003), and CEO compensation varies directly (Fahlenbrach 2004), with the G-index – suggests that these restrictions reinforce management entrenchment.


Abbott, L., S. Parker, and G. Peters 2003. Audit committee characteristics and restatements. Auditing, A Journal of Practice and Theory 23 (1): 69-87.

Almazan, A. and J. Suarez 2003. Entrenchment and severance pay in optimal governance structures. The Journal of Finance 58: 519-548.

Altamuro, J., A. Beatty and J. Weber 2003. Motives for early revenue recognition: Evidence from SEC staff accounting bulletin (SAB) 101. The Accounting Review 80 (2): 373-401.

Anderson, R. and J. Bizjak 2003. An empirical examination of the role of the CEO and the compensation committee in structuring executive pay. Journal of Banking and Finance 27: 1323-1348.

Agrawal, A. and C. Knoeber 1996. Firm performance and mechanisms to control agency problems between managers and shareholders. Journal of Financial and Quantitative Analysis 31: 377-397.

Agrawal, A., and S. Chadha 2005. Corporate governance and accounting scandals. Journal of Law and Economics (Forthcoming).

Bainbridge, S. 1995. The politics of corporate governance. Harvard Journal of Law and Public Policy 671: 694-699.

Barth, M., W. Beaver, J. Hand, and W. Landsman 1999. Accruals, cash flows, and equity values. Review of Accounting Studies 4 (3): 205-229.

Baysinger, B. and H. Butler 1985. Corporate governance and board of directors: Performance effects of changes in board composition. Journal of Law, Economics, and Organization 1: 101-124.

Beasley, M. 1996. An empirical analysis of the relation between the board of director composition and financial statement fraud. The Accounting Review 71: 433-465.

Beasley, M.S., J.V. Carcello, and D.R. Hermanson 1997. Fraudulent Financial Reporting: 1987-1997. An analysis of U.S. public companies. Research commissioned by Committee of Sponsoring Organizations of the Treadway Commission.

Bebchuk, L., A. Cohen, and A. Ferrell 2004. What matters in corporate governance? (November). Harvard Law & Economics Discussion Paper No. 491.

Belsley, D. A., Kuh, E., & Welsch, R. E. 1980. Regression diagnostics: Identifying influential data and sources of collinearity. New York: John Wiley & Sons.

Bhagat, S. and B. Black 1999. The uncertain relationship between board composition and firm performance. The Business Lawyer 54: 921- 963.

Bhagat, S. and B. Black 2002. The non-correlation between board independence and long term firm performance. Journal of Corporation Law 27: 231-273.

Blue Ribbon Committee on Improving the Effectiveness of Corporate Audit Committees (Blue Ribbon Committee) 1999. Report and Recommendations. NYSE and NASDAQ.

Brickley, J., J. Coles and R. Terry 1994. The board of directors and the enactment of poison pills. Journal of Financial Economics 35: 371-390.

Bryan, S.; Hwang, L.; and, Lilien, S. 2000. CEO Stock-based compensation: An empirical analysis of incentive-intensity, relative mix, and economic determinants. Journal of Business (October): 661-93.

Byrd, J. and K. Hickman 1992. Do outsider directors monitor managers? Journal of Financial Economics 19: 589-606.

Byrne, J. 1996. The best and worst boards. Business Week. November 25.

Byrne, J. 2002. How to fix corporate governance. Business Week. May 6.

California Public Employees Retirement System Corporate Governance Core Principles & Guidelines (April 13) 1998.

Core, J., W. Guay, 2002. Estimating the value of employee stock option portfolios and their sensitivities to price and volatility. Journal of Accounting Research 40: 613-630.

Core, J., W. Guay, and T. Rusticus 2005. Does weak governance cause weak stock returns? An Examination of firm operating performance and investors’ expectations. The Journal of Finance (Forthcoming).

Core, J., R. Holthausen, and D. Larcker 1999. Corporate governance, CEO compensation, and firm performance. Journal of Financial Economics 51: 371-390.

Cotter, J., A. Shivdasani and M. Zenner 1997. Do independent directors enhance target shareholder wealth during tender offers? Journal of Financial Economics 43: 195-219.

Cremers, M., and V. B. Nair, 2005. Governance mechanisms and equity prices. The Journal of Finance (Forthcoming).

Cremers, M., V. B. Nair and C. Wei 2004. The congruence of shareholder and bondholder governance. NYU, Law and Economics Research Paper No. 04-007 (October).

Crystal, G. 1991. In search of excess: The overcompensation of American executives. (W.W. Norton Company. New York).

Cyert, R., S-H Kang, and P. Kumar 2002. Corporate governance, takeovers, and top management compensation: Theory and evidence. Management Science 48: 453-469.

DeFond, M., R. Hann, X. Hu 2005. Does the market value financial expertise on audit committees of board of directors? Journal of Accounting Research 43 (2): 153-193.

DeAngelo, H. and E. Rice 1983. Anti-takeover charter amendments and stockholder wealth. Journal of Financial Economics 11: 329-359.

Dechow, P., R. Sloan, and A. Sweeney 1996. Causes and consequences of earnings manipulation: An analysis of firms subject to enforcement actions by the SEC. Contemporary Accounting Research 13: 1-36.

Erickson, M., M. Hanlon and E. Maydew 2004. Is there a link between executive compensation and accounting fraud? Working paper, University of Chicago.

Fahlenbrach, R. 2004. Shareholder rights and CEO compensation. Working paper, University of Pennsylvania.

Fama, E. and K. French 1997. Industry costs of equity. Journal of Financial Economics 43 (2): 153- 194.

Fama, E. 1980. Agency problems and the theory of the firm. Journal of Political Economy 88: 715-732.

Fama, E. and M. Jensen 1983. Separation of ownership and control. Journal of Law and Economics 26: 301-305.

Farber, D. 2005. Restoring trust after fraud: Does corporate governance matter? The Accounting Review 80 (2): 539-561.

General Accounting Office (GAO) 2002. Financial Statement Restatements: Trends, Market Impacts, Regulatory Responses, and Remaining Challenges. Report GAO-03-138 (October).

General Accounting Office (GAO) 2003. Financial Statement Restatement Database. Report GAO-03-395R (January).

Gibbons, R. and K. Murphy 1992. Optimal incentive contracts in the presence of career concerns: Theory and evidence. Journal of Political Economy 100: 468-505.

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

Hermalin, B. and M. Weisbach 1991. The effects of board composition and direct incentives on firm performance. Financial Management 20: 101-112.

Holmstrom, B. and S. Kaplan 2003. The state of U.S. corporate governance: What’s right and what’s wrong? NBER Working paper.

Jensen, M. 1993. The modern industrial revolution, exit, and failure of internal control systems. The Journal of Finance 48: 35-44

Jensen M. and W. Meckling 1976. Theory of the firm: managerial behavior, agency costs and ownership structure. . Journal of Financial Economics 3: 305-360.

Kang, S-H, P. Kumar, and H. Lee 2005. Agency and corporate investment: The role of executive compensation and corporate governance. Journal of Business (Forthcoming).

Kim, I. 2004. The role of directors’ and officers’ insurance in financial reporting and class action securities litigation. Working paper, University of Michigan.

Kinney, W., Z. Palmrose, S. Scholtz. 2004. Auditor independence, non-audit services, and restatements: Was the U.S. government right? Journal of Accounting Research 42 (3): 561-588.

Klein, April. 2002. Audit committee, board of director characteristics, and earnings management. Journal of Accounting and Economics 33: 375-400.

Klock, M., A. Mansi, and F. Maxwell. 2004. Does corporate governance matter to bondholders? Journal of Financial and Quantitative Analysis (Forthcoming).

Larcker, D., S. Richardson, and I. Tuna. 2004. How important is corporate governance? Working paper, The Wharton School, University of Pennsylvania.

Lavelle, L. 2002. The Best & Worst Boards. Business Week. October 7.

Lorsch, J. and E. MacIver 1989. Pawns or Potentates: The Reality of America’s Corporate Boards (Revised Edition). Boston: Harvard Business School Press.

Mehran, H. 1995. Executive compensation structure, ownership, and firm performance. Journal of Financial Economics 38: 163-184.

Murphy, K. 1995. Politics, economics, and executive compensation. University of Cincinnati Law Review 63: 713-748.

New York Stock Exchange 2002. Report from the Corporate Accountability and Listing Standards Committee.

Palmrose, Z., V. Richardson, and N. Scholz 2004. Determinants of market reactions to restatement announcements. Journal of Accounting and Economics 37 (1): 59-90.

Peng, L. and A. Roell 2004. "Executive Pay, Earnings Manipulation and Shareholder Litigation" Working paper, CUNY-Baruch College.

Richardson, S., I. Tuna, and M. Wu. 2002. Predicting earnings management: The case of earnings restatements. Working paper, University of Pennsylvania (Wharton).

Raghunandan, K, W.J. Read, J.S. Whisenant. 2003. Initial evidence on the association between nonaudit fees and restated financial statements. Accounting Horizons 17 (3): 223-234.

Romano, R. 2004. The Sarbanes-Oxley Act and the making of quack corporate governance. Yale Law & Economics Research Paper 297.

Rosenbaum, V. 1990, 1993, 1995, 1998, 2000, 2002. Corporate Takeover Defenses. Investor Responsibility Research Center Inc., Washington, D.C.

Shleifer, A., and R. Vishny 1986. Large shareholders and corporate control. Journal of Political Economy 95: 461-488.

Shleifer, A. and R. Vishny 1997. A survey of corporate governance. The Journal of Finance 52: 737-783.

Srinivasan, S. 2005. Consequences of financial reporting failure for outside directors: Evidence from accounting restatements and audit committee members. Journal of Accounting Research 43 (2): 291-334.

Stein, J. 1988. Takeover threats and managerial myopia. Journal of Political Economy 96: 61-80.

Watts, R. and J. Zimmerman 1979. The demand and supply of accounting theories: The market for excuses. The Accounting Review 54 (2): 273-305.

Weisbach, M. 1988. Outside directors and CEO turnover. Journal of Financial Economics 20: 431-460.

Yermack, D. 1996. Higher market valuation of companies with a small board of directors. Journal of Financial Economics 40: 185- 211.


A summary of published studies on accounting restatements and financial reports accused of fraud by the SEC (AAER: Accounting and Auditing Enforcement Releases)

| |Agrawal and |Abbott, | |Erickson, | |Dechow, |

| |Chadha (2005) |et al. |Farber |et al. |Beasley (1996) |et al. |

| | |(2003) |(2005) |(2004) | |(1996) |

|Nature of failures |Restatement |Restatement(a) |AAER |AAER |AAER |AAER |

|Sample Size |159 |88 |87 |50 |75 |86 matched pairs |

| |matched pairs |matched pairs |matched pairs |firms |matched pairs |(Univariate) |

| | | |(Univariate) | | | |

|Sample Year |2000-2001 |1991-1999 |1982-2000 |1996-2003 |1979-1990 |1982-1992 |

|Board independence |insignificant |insignificant |(-)** |---- |(-)** |(-)** |

|Audit committee independence |insignificant |(-)** |insignificant |---- |(-)**(b) |(-)**(b) |

|Board size |---- |(+)** |insignificant |---- |---- | |

|Equity compensation/total compensation | | | | | | |

| |---- |---- |---- |(+)** |---- |---- |

|Exercisable option holdings |---- |---- |---- |---- |---- |---- |

|CEO ownership |insignificant |---- |---- |---- |---- |---- |

|Employer director ownership |---- |insignificant |insignificant |---- |insignificant |---- |

|Outside director ownership |---- |insignificant |---- |---- |---- |---- |

|CEO=Chair |insignificant |insignificant |(+)** |insignificant |insignificant |(+)** |

|CFO serves on the audit committee |---- |---- |---- |---- |---- |---- |

|Audit committee has independent director|(-)** |(-)** |(-)** |---- |---- |---- |

|with accounting background | | | | | | |

|Outside blockholder |insignificant |insignificant |(-)** |---- |insignificant |(-)** |

|CEO tenure |---- |---- |---- |---- |insignificant |---- |

|Founder CEO or Chair |(+)** |insignificant |---- |---- |---- |(+)** |

|Non-audit fees/total fees |insignificant |---- |---- |---- |---- |---- |

|Big 8 auditor or Big 4 |insignificant |---- |(-)* |---- |---- |insignificant |

|Reference |Table 5 |Table 4 |Table 3 |Table 5 |Table 4 |Table 8 |

** significant at the five percent level, two-tailed. A positive sign indicates a positive relation with restatement.

a) Abbott, et al. (2003) also examine SEC’s AAER (Accounting and Auditing Enforcement Releases) firms which are accused of committing fraud.

b) Beasley (1996) and Dechow et al. (1996) consider whether the firm has an audit committee.


Variable Definitions

|Variable |Description |

|Entrenchment | |

|G-INDEX |governance index advanced by Gompers, Ishii, and Metrick (2003) |

|E-INDEX |entrenchment index advanced by Bebchuk, Cohen, and Ferrell (2004) (See note 1) |

|B-INDEX |composite score of nine board characteristics (See note 2) |

|Board Strength | |

|BOD_INDEP* |fraction of independent directors on the board |

|AUD_INDEP* |fraction of independent directors on the audit committee |

|COM_INDEP* |fraction of independent directors on the compensation committee |

|NOM_INDEP* |fraction of independent directors on the nominating committee |

|MAJ_INDEP |1 if the fraction of independent directors on the board, audit committee, compensation committee and |

| |nominating committee is greater than 0.5; zero, otherwise |

|FULL_INDEP |1 if the fraction of independent directors on the board is greater than 2/3 and all members on the |

| |audit committee, compensation committee and nominating committee are independent directors; zero, |

| |otherwise |

|BOD_SIZE* |number of board members |

|SEP_CHAIR* |1 if the CEO is not the board chair; zero, otherwise |

|INDEP _CHAIR* |1 if an independent outside director is the board chair; zero, otherwise |

|CEO_ON_NOM |1 if the CEO is on the nominating committee; zero, otherwise |

|INTERLOCK* |fraction of interlocked directors on the board |

|BUSYBOARD* |number of other company boards served by all directors, divided by number of board members |

|EXPERT |1 if an outside accounting and financial expert serves on the audit committee |

|Ownership | |

|OUTDIR_OWN |fraction of company stocks held by independent directors |

|EMPDIR_OWN |fraction of company stocks held by insider-directors, excluding the CEO |

|CEO_OWN |fraction of company stocks held by the CEO |

|BLOCK |number of outside blockholders. |

|INST_OWN |fraction of shares held by institutions |

|Compensation | |

|STK_COMP |the CEO’s total equity-based compensation (stock options, SARs, and restricted stocks: EXECUCOMP |

| |variables Blk_valu+Rstkgrnt) divided by the CEO total compensation (EXECUCOMP variable TDC1) |

|OPTVESTED |number of exercisable or vested options of the CEO divided by the number of shares outstanding |

| |(EXECUCOMP variable UEXNUMEX/Shrsout*1000) |

|Other | |

|CEO_AGE |age of the CEO |

|DIR_AGE |average age of the board members |

|FOUNDER |1 if CEO or the Board Chair founded the company; zero otherwise |

|Financial Control Variables | |

|BIG6 |1 if the firm is audited by a Big Six firm; zero, otherwise |

|LOGSIZE |log of beginning total assets |

|GROWTH |growth rate of sales during the two years preceding the event year |

|FREECF |beginning-of-the-year free cash flow /beginning total assets |

|DIV_YIELD |beginning-of-the-year dividend yield |

|MARGIN |beginning-of-the-year net income/sales |

|LEVERAGE |beginning-of-the-year interest-bearing debt/total assets |

1. E-Index includes six anti-takeover provisions distilled from the components of G-Index by Bebchuk, Cohen, and Ferrell (2004). (Classified Board, Bylaws, Charter, Supermajority, Golden Parachutes, Poison Pills).

2. B-Index is a composite score constructed from nine indicators of board characteristics denoted in the table by an asterisk (*). B-Index increases by one if: (1) more than 2/3 of the board is comprised of independent directors; (2) all of the audit committee members are independent directors; (3) all of the compensation committee members are independent directors; (4) all of the nominating committee members are independent directors; (5) the CEO is not the board Chair; (6) the Chair is an independent outside director; (7) if the board size is less than the median of all firms (adjusted for firm size and time); (8) the board interlock (INTERLOCK) is less than the median of all firms (adjusted for firm size and time); or (9) BUSYBOARD is less than the median of all firms (adjusted for firm size and time). To adjust for firm size and time, the composite measure is computed net of the median for each firm size decile identified for each year.


Sample selection procedure

|Description |Number |

| | |

|Firms included in both the IRRC (Investor Responsibility | |

|Research Center) corporate governance database and in the |2,431 |

|S&P COMPUSTAT database (1997-2002). | |

| | |

|Test Sample: firms that restated financial statements during 1997-2002 according to GAO | 231 |

|Restatements in anticipation of pending EITF or SFAS pronouncements, or involve no change in | (25) |

|earnings.(1) | |

|Restatements cannot be verified. | (14) |

|Restatements did not change or decreased earnings | (4) |

|Missing G-Index in the event year. | (20) |

|Missing governance data or financial data (in the event year for the test firm) | (21) |

| Remaining GAO sample (announced during 1/1/1997- 6/30/2002) | 147 |

| Hand-collected sample (announced during 7/1/2002- 12/31/2004) | 55 |

|Exclusion of firms with dual class securities | (17) |

|Total Test Sample | 185 |

|Total Control (Non-restatement) meeting the same data requirement as the Test sample. | 1,487 |

| Final sample | 1,672 |

1. Excluded restatements include those issued for Emerging Issues Task Force (EITF) Issue No. 00-10, on shipping and handling fees and costs, EITF 00-14 on certain sales incentives, EITF 00-22 relating to the classification of certain sales incentives, and EITF 00-25 "Vendor Income Statement Characterization of Consideration Paid to a Reseller of the Vendor's Products," and adoptions of SFAS 132, 133, and 142.


PANEL A: Descriptive Statistics for the Full Sample

| | Mean |Median | Min. | Max. |Std. dev |

|G-INDEX |9.152 |9.000 |2.000 |17.000 |2.646 |

|E-INDEX |1.603 |2.000 |0.000 |5.000 |1.101 |

|BOD_INDEP |0.633 |0.667 |0.100 |0.933 |0.179 |

|AUD_INDEP |0.862 |1.000 |0.000 |1.000 |0.207 |

|COM_INDEP |0.885 |1.000 |0.000 |1.000 |0.209 |

|NOM_INDEP |0.751 |0.750 |0.000 |1.000 |0.268 |

|MAJ_INDEP |0.440 |0.000 |0.000 |1.000 |0.496 |

|FULL_INDEP |0.165 |0.000 |0.000 |1.000 |0.371 |

|BOD_SIZE |9.436 |9.000 |3.000 |26.000 |3.156 |

|SEP_CHAIR |0.273 |0.000 |0.000 |1.000 |0.446 |

|INDEP_CHAIR |0.070 |0.000 |0.000 |1.000 |0.255 |

|CEO_ON_NOM |0.164 |0.000 |0.000 |1.000 |0.370 |

|INTERLOCK |0.013 |0.000 |0.000 |0.429 |0.041 |

|BUSYBOARD |0.868 |0.710 |0.000 |4.786 |0.717 |

|EXPERT (1) |0.697 |1.000 |0.000 |1.000 |0.460 |

|OUTDIR_OWN |0.012 |0.003 |0.000 |0.651 |0.035 |

|EMPDIR_OWN |0.020 |0.001 |-0.003 |0.668 |0.055 |

|CEO_OWN |0.035 |0.011 |0.000 |0.526 |0.065 |

|BLOCK |1.870 |2.000 |0.000 |9.000 |1.508 |

|INST_OWN |0.580 |0.607 |0.000 |0.994 |0.210 |

|STK_COMP |0.425 |0.438 |0.000 |1.000 |0.307 |

|OPTVESTED |0.007 |0.004 |0.000 |0.127 |0.011 |

|CEO_AGE |54.753 |55.000 |34.000 |91.000 |7.634 |

|DIR_AGE |58.355 |58.800 |40.800 |71.400 |4.227 |

|FOUNDER (1) |0.183 |0.000 |0.000 |1.000 |0.387 |

|B-INDEX |4.736 |5.000 |0.000 |9.000 |1.581 |

|BIG6 |0.920 |1.000 |0.000 |1.000 |0.272 |

|LOGSIZE |7.336 |7.101 |3.009 |13.592 |1.702 |

|GROWTH |0.212 |0.124 |-0.297 |2.930 |0.332 |

|FREECF |-0.009 |-0.005 |-0.450 |0.324 |0.110 |

|DIV_YIELD |0.012 |0.003 |0.000 |0.091 |0.017 |

|MARGIN |0.045 |0.058 |-1.710 |0.366 |0.176 |

|LEVERAGE |0.237 |0.226 |0.000 |1.616 |0.189 |

1) These variables address 178 matched-pairs only (356 observations).

Descriptive statistics are for the full sample of 1,672 firm-year observations (185 Test and 1,487 Control observations). G-INDEX = the governance index advanced by Gompers, Ishii, and Metrick (2003); BOD_INDEP = fraction of independent directors on the board; AUD_INDEP = fraction of independent directors on the audit committee; COM_INDEP = fraction of independent directors on the compensation committee; NOM_INDEP = fraction of independent directors on the nominating committee; MAJ_INDEP = 1 if the fraction of independent directors on the board, audit committee, compensation committee and nominating committee is greater than 0.5; zero, otherwise; FULL_INDEP = 1 if the fraction of independent directors on the board is greater than 0.667 and all members on the audit committee, compensation committee and nominating committee are independent directors; zero, otherwise; BOD_SIZE = number of board members; SEP_CHAIR = 1 if the CEO is not the board chair; zero, otherwise; INDEP_CHAIR = 1 if an independent outside director is the board chair; zero, otherwise; CEO_ON_NOM = 1 if the CEO is on the nominating committee; zero, otherwise; INTERLOCK = fraction of interlocked directors on the board; BUSYBOARD = total number of other companies’ boards served by all directors, divided by total number of board members; EXPERT = 1 if an outside accounting and financial expert serves on the audit committee; DIR_AGE = mean board member age; OUTDIR_OWN = fraction of company stocks held by independent directors; EMPDIR_OWN = fraction of company stocks held by insider-directors other than the CEO; CEO_OWN = percentage of company stocks held by the CEO; BLOCK = number of outside blockholders; INST_OWN = fraction of shares held by institutions; STK_COMP = the CEO equity-based compensation (stock options, SARs, and restricted stocks) divided by the CEO’s total compensation; OPTVESTED = number of exercisable or vested options of the CEO divided by the number of shares outstanding; CEO_AGE = age of the CEO; FOUNDER = 1 if CEO or the Board Chair founded the company; zero, otherwise; B-INDEX = composite score of 9 board characteristics; TENURE = CEO’s tenure; BIG6 = 1 if the firm is audited by a Big Six firm; zero, otherwise; LOGSIZE = log of beginning total assets; GROWTH = growth rate of sales during the previous 2 years; FREECF = beginning-of-the-year free cash flow /beginning total assets; DIV_YIELD = beginning-of-the-year dividend yield; MARGIN = net income/sales; LEVERAGE = interest-bearing debt/(Interest-bearing debt +market value of equity).

PANEL B: Pearson correlation



| | | | | | | | |

|Restatement sample |20 |36 |37 |40 |36 |16 |185 |

| |(10.8%) |(19.5%) |(20.0%) |(21.6%) |(19.5%) |(8.6%) |(100%) |

| | | | | | | | |

|Control sample |193 |284 |275 |293 |272 |170 |1,487 |

| |(13.0%) |(19.1%) |(18.5%) |(19.7%) |(18.3%) |(11.4%) |(100%) |

1. (2 statistic for the null hypothesis that event years are identically distributed has a p-value of 0.79 ((2 =2.39).

B. Distribution of the stock exchange membership2

| | | | | | | |

| |NYSE |NASDAQ |ASE |OTC |Other |Total |

| | | | | | | |

|Restatement sample |113 |66 |0 |4 |2 |185 |

| |(61.1%) |(35.6%) |(0%) |(2.2%) |(1.1%) |(100%) |

| | | | | | | |

|Control sample |980 |446 |17 |25 |19 |1,487 |

| |(65.9%) |(30.0%) |(1.1%) |(1.7%) |(1.3%) |(100%) |

2. (2 statistic for the null hypothesis that exchange membership is identically distributed has a p-value of 0.32 ((2 =4.70).

C. Restatements by the type of accounting irregularity

|Nature of Misstatements |Frequency |Percent |

|Revenue Recognition |83 |44.9 |

|Expense, or Revenue and Expense |92 |49.7 |

|Other |10 |5.4 |

|Total |185 |100.0 |


Cross-sectional logistic regressions of restatement probability

The dependent variable is set equal to one, if the firm restated financial statements during 1997-2002; zero, otherwise. Variable definitions are in Exhibit 2. Asterisks *, **, *** indicate significance at 10%, 5%, 1% confidence levels, two-tailed when there is no direction specified under the alternative hypothesis and one-tailed when there is a specified direction under the alternative hypothesis (as indicated by predicted signs). Specifications 1 and 2 address 185 Test firms and 1,487 Control firms. Specifications 3 and 4 address 165 Test firms and 1,278 Control firms during 1998-2002. Specifications also consider fiscal year, industry, and stock exchange listings (corresponding estimates not reported).

| | |1 |2 |3 |4 |

| |Pred. | |(2 | |(2- statistic | |(2- | |(2- statistic|

| |Sign |Estimate |statistic |Estimate | |Estimate |statistic |Estimate | |

|Intercept | |-4.304 |6.80*** |-3.533 |5.16** |-4.716 |7.85** |-5.480 |9.19*** |

|G-INDEX |(+) |0.108 |9.95*** |0.109 |10.55*** |0.093 |7.00*** |0.103 |8.09*** |

|B-INDEX |(-) | | | | |0.098 |3.13* | | |

|BOD_INDEP |(-) |0.138 |0.03 | | | | |0.010 |0.00 |

|AUD_INDEP |(-) |1.099 |4.10 | | | | |1.257 |4.47 |

|COM_INDEP |(-) |-0.444 |0.83 | | | | |-0.370 |0.48 |

|NOM_INDEP |(-) |0.414 |0.79 | | | | |1.194 |3.56 |

|FULL_INDEP |(-) | | |-0.223 |0.77 | | | | |

|DNOM | |-0.471 |1.53 |-0.117 |0.32 | | |-1.059 |3.51* |

|BOD_SIZE |(+) |-0.003 |0.01 |-0.017 |0.19 | | |0.013 |0.10 |

|SEP_CHAIR |(-) |0.338 |2.98 |0.328 |2.88 | | |0.310 |1.81 |

|INTERLOCK |(+) |2.132 |1.31 |1.627 |0.84 | | |2.101 |1.06 |

|DIR_AGE | |-0.042 |2.88* |-0.043 |3.14* |-0.027 |1.18 |-0.029 |1.27 |

|OUTDIR_OWN | |3.966 |4.55** |4.334 |5.65** |2.768 |1.93 |2.203 |1.07 |

|EMPDIR_OWN | |-2.064 |0.99 |-2.348 |1.38 |-1.895 |0.85 |-2.277 |1.01 |

|CEO_OWN | |0.506 |0.10 |0.567 |0.13 |-0.233 |0.02 |-0.222 |0.02 |

|BLOCK |(-) |-0.098 |1.94* |-0.086 |1.49 |-0.062 |0.74 |-0.065 |0.78 |

|INST_OWN |(-) |-0.019 |0.00 |0.073 |0.03 |-0.156 |0.11 |-0.181 |0.15 |

|STK_COMP | |0.132 |0.22 |0.102 |0.13 |-0.016 |0.01 |0.017 |0.00 |

|OPTVESTED | |7.874 |1.09 |7.342 |0.97 |8.383 |1.06 |8.115 |0.95 |

|CEO_AGE | |-0.011 |0.66 |-0.012 |0.89 |-0.006 |0.22 |-0.005 |0.14 |

|INDEP_CHAIR |(-) | | | | | | |0.179 |0.24 |

|CEO_ON_NOM |(+) | | | | | | |0.503 |2.43 |

|BUSYBOARD |(+) | | | | | | |-0.035 |0.05 |

|BIG6 | |0.920 |4.77** |0.843 |4.10** |0.846 |3.72* |1.012 |4.75** |

|LOGSIZE | |0.259 |11.14*** |0.278 |13.00*** |0.215 |8.93*** |0.220 |6.13** |

|GROWTH | |0.491 |4.59** |0.463 |4.10** |0.561 |5.75** |0.559 |5.44** |

|FREECF | |-2.538 |11.04*** |-2.537 |11.24*** |-2.760 |12.21*** |-2.783 |11.99*** |

|DIV_YIELD | |13.661 |3.40* |-11.016 |2.28 |-9.099 |1.59 |-11.154 |2.23 |

|MARGIN | |0.364 |0.48 |0.298 |0.32 |0.464 |0.65 |0.478 |0.71 |

|LEVERAGE | |0.422 |0.73 |0.381 |0.60 |0.411 |0.64 |0.516 |0.98 |

|Sample size Test/Control | |185/ 1487 |185/ 1487 |165/ 1278 |165/ 1278 |

|Pseudo R2 | |0.059 |0.055 |0.048 |0.057 |

|Likelihood Ratio (p - | |101.6 |95.4 |71.6 |84.0 |

|value)/ | |(0.001) |(0.001) |(0.001) |(0.001) |


178 Matched-pair cross-sectional logistic regressions of restatement probability

The dependent variable is one, if the firm restated financial statements during 1997-2002; zero, otherwise. Variable definitions are in Exhibit 2. Asterisks *, **, *** indicate significance at the 10%, 5%, 1% level, two-tailed when there is no alternative and one-tailed when there is an alternative hypothesis (as indicated by the predicted signs). The sample comprises 178 Test firms and 178 Control firms selected to match the Test firms with respect to the year, stock exchange, firm size, and industry. Specifications include controls for fiscal year, industry, and stock exchange listings (estimates not reported).

| | |1 |2 |3 |4 |

| |Pred. | |(2 | |(2- statistic | |(2- | |(2- statistic|

| |Sign |Estimate |statistic |Estimate | |Estimate |statistic |Estimate | |

|Intercept | |-0.201 |0.01 |0.191 |0.01 |0.362 |0.03 |0.253 |0.02 |

|G-INDEX |(+) |0.150 |8.46*** |0.147 |8.28*** |0.155 |9.38*** |0.144 |9.06*** |

|B-INDEX |(-) | | | | | | |0.123 |2.90 |

|BOD_INDEP |(-) |0.194 |0.03 | | | | | | |

|AUD_INDEP |(-) |0.908 |1.38 | | | | | | |

|COM_INDEP |(-) |-0.474 |0.44 | | | | | | |

|NOM_INDEP |(-) |1.430 |3.71 | | | | | | |

|MAJ_INDEP |(-) | | |0.841 |5.40 | | | | |

|FULL_INDEP |(-) | | | | |-0.465 |1.64 | | |

|DNOM | |-1.491 |4.91** |-0.955 |5.13** |-0.180 |0.28 | | |

|BOD_SIZE |(+) |0.022 |0.20 |0.032 |0.45 |0.029 |0.37 | | |

|SEP_CHAIR |(-) |0.566 |3.12 |0.599 |3.74 |0.527 |2.97 | | |

|INTERLOCK |(+) |1.471 |0.34 |0.656 |0.08 |-0.297 |0.02 | | |

|DIR_AGE | |-0.036 |0.93 |-0.036 |1.01 |-0.041 |1.30 |-0.036 |1.07 |

|OUTDIR_OWN | |1.878 |0.59 |2.022 |0.68 |2.390 |0.95 |2.256 |0.91 |

|EMPDIR_OWN | |1.770 |0.27 |1.648 |0.25 |1.343 |0.17 |1.695 |0.29 |

|CEO_OWN | |-1.707 |0.69 |-1.624 |0.65 |-1.279 |0.40 |-2.340 |1.46 |

|BLOCK |(-) |-0.075 |0.72 |-0.039 |0.20 |-0.041 |0.22 |-0.065 |0.63 |

|INST_OWN |(-) |-0.669 |0.86 |-0.743 |1.06 |-0.407 |0.32 |-0.481 |0.49 |

|STK_COMP | |0.193 |0.21 |0.236 |0.32 |0.184 |0.20 |0.105 |0.07 |

|OPTVESTED | |10.169 |0.61 |10.029 |0.59 |7.154 |0.31 |7.014 |0.33 |

|CEO_AGE | |-0.006 |0.11 |-0.006 |0.11 |-0.005 |0.10 |-0.006 |0.15 |

|INDEP_CHAIR |(-) |0.894 |2.18 |0.902 |2.30 |0.875 |2.16 | | |

|CEO_ON_NOM |(+) |0.689 |2.72* |0.482 |1.70 |0.079 |0.05 | | |

|BUSYBOARD |(+) |0.199 |1.39 |0.195 |1.40 |0.212 |1.64 | | |

|EXPERT | |0.053 |0.04 |0.109 |0.17 |0.064 |0.06 |-0.018 |0.01 |

|FOUNDER | |0.023 |0.00 |0.021 |0.00 |-0.077 |0.05 |0.059 |0.03 |

|BIG6 | |0.345 |0.53 |0.341 |0.53 |0.318 |0.47 |0.394 |0.81 |

|GROWTH | |0.591 |2.26 |0.607 |2.44 |0.565 |2.14 |0.660 |3.02* |

|FREECF | |-3.397 |7.20*** |-3.533 |7.89*** |-3.590 |8.35*** |-3.297 |7.31*** |

|DIV_YIELD | |-6.598 |0.44 |-5.612 |0.33 |-1.504 |0.02 |-5.379 |0.34 |

|MARGIN | |0.427 |0.21 |0.393 |0.18 |0.175 |0.04 |0.250 |0.07 |

|LEVERAGE | |0.281 |0.18 |0.322 |0.24 |0.175 |0.07 |0.249 |0.16 |

|Pseudo R2 | |0.140 |0.137 |0.151 |0.099 |

|Likelihood Ratio (p - | |53.7 |51.3 |47.4 |37.2 |

|value) | |(0.004) |(0.002) |(0.006) |(0.008) |


Cross-sectional logistic regressions using four governance sub-indices

E-Index includes six anti-takeover provisions distilled from the components of G-Index by Bebchuk, Cohen, and Ferrell (2004). (Classified Board, Bylaws, Charter, Supermajority, Golden Parachutes, Poison Pills); Delay includes four provisions designed to delay hostile takeovers (Blank Check, Classified Board, Special Meeting, Written Consent; titles from Gompers, Ishii, and Metrick (2003)); Protection includes six provisions designed to protect officers and directors from liability (Compensation Plans, Contracts, Golden Parachutes, Indemnification, Liability, Severance); Voting comprises six provisions limiting shareholder voting (Bylaws, Charter, Cumulative Voting, Secret Ballot, Supermajority, Unequal Voting); Other includes six provisions representing direct takeover defense measures (Anti-greenmail, Directors’ Duties, Fair Price, Pension Parachutes, Poison Pill, Silver Parachutes). Italics indicates χ2 statistic. Asterisks *, **, *** indicate significance at the 10%, 5%, 1% level, two-tailed when there is no alternative and one-tailed when there is an alternative hypothesis (as indicated by predicted signs). The sample comprises 185 Test firms and 1,487 Control firms selected to match the Test firms. Estimates not reported for control variables displayed in column 1 of Table IV.

PANEL A: Logistic Regression Estimates

| | |1 |2 |3 |4 |5 |

| |Pred. | | | | | |

| |Sign |Estimate |(2 |Estimate |(2 |Estimate |

|Likelihood Ratio | |87.9 |87.0 |84.6 |85.2 |91.3 |

|(p - value) | |(0.001) |(0.001) |(0.001) |(0.001) |(0.001) |

PANEL B: Spearman Correlation

| | | | | | |


|E-INDEX |0.685*** | | | | |

|DELAY |0.608*** |0.579*** | | | |

|VOTING |0.476*** |0.675*** |0.350*** | | |

|PROTECT |0.509*** |0.080*** |0.012 |-0.016 | |

|OTHER |0.613*** |0.504*** |0.262*** |0.117*** |0.260*** |

|E-INDEX | | | | | |


Regression estimates for association between G-INDEX and B-INDEX controlling for firm size

B-Index is a composite score constructed from 9 indicators of board characteristics. B-Index increases by one if: (1) more than 2/3 of the board is comprised of independent directors; (2) all of the audit committee members are independent directors; (3) all of the compensation committee members are independent directors; (4) all nominating committee members are independent directors; (5) the CEO is not the board Chair; (6) the board chair is an independent outside director; (7) board size is less than the median of all firms (adjusted for firm size and time); (8) the board interlock (INTERLOCK) is less than the median of all firms (adjusted for firm size and time); or (9) BUSYBOARD is less than the median of all firms (adjusted for firm size and time). To adjust for firm size and time, each variable is net of the median for corresponding size decile and year. G-INDEXR denotes size-adjusted G-INDEX (deviation from the mean G-INDEX for corresponding size decile and year); similarly B-INDEXR deviation from the mean B-INDEX of each size decile.

Regression Estimates

B-INDEX (or B-INDEXR)i,t=β0+β1(G-INDEX or G-INDEXRi,t)+β2(LOGSIZE,t) + Controls +(i,t

| | |1 |2 |3 |


| |Pred. | | | | | | |

| |Sign |Estimate |t-stat |Estimate |t-stat |Estimate |t-stat |

| | |2.237 |25.24*** |0.685 |5.97*** |-1.770 |-16.31*** |

|Intercept | | | | | | | |

|G-INDEX | (+/-) |0.047 |7.35*** | | | | |

|G-INDEXR | (+/-) | | | 0.042 |7.77*** |0.041 |7.59*** |

|LOGSIZE | |0.140 |13.27*** |0.119 |11.14*** |0.002 |0.17 |

|CONTROL VARIABLES | |Not Included |Included |Included |

|(Industry, Year, Stock | | | | |

|Exchange) | | | | |

| | |0.021 | | |

|Adj-R2 | | |0.34 |0.33 |


[1] Recent papers that use the substitution-entrenchment characterization include Cyert, Kang, and Kumar (2002), Fahlenbrach (2004), and Klock, Mansi, and Maxwell (2004).

[2] See recommendations of the Corporate Accountability and Listing Standards Committee (NYSE 2002) and the Sarbanes-Oxley Act of 2002.

[3] As noted in Bebchuk, et al (2004), this view is not universally supported, and the effect can be context-specific, as in Cremers and Nair (2005).

[4] To ease the presentation, we do not address literature that considers associations between earnings management and corporate governance (e.g., Klein 2002).

[5] For example, controversial management practices or actions, which can be correlated with poor corporate governance, can elicit the attention or intervention by regulators.

[6] The GAO (2002) observes that “the recent increase in the number and size of financial statement restatements and accounting issues and irregularities underlying these restatements have raised significant questions about the adequacy of the current system of corporate governance and financial disclosure oversight (p. 7).”

[7] Both Beasley (1996) and Dechow et al. (1996) classify directors as insiders versus outsiders, where outsiders include “affiliated” directors (not generally considered independent in recent studies).

[8] Larcker, et al. (2004) is not summarized in Exhibit 1 because they use factor scores which are not comparable to individual governance characteristics. While Larcker, et al. (2004) seek to examine how much corporate governance indicators as a whole explain broader issues of management in terms of statistical explanatory power, our study examines directional impact of governance indicators.

[9] Some provisions restricting shareholders rights are initiated as anti-takeover devices which are characterized typically as inimical to shareholder interests, but which can also benefit shareholders. As examples, such provisions can encourage CEOs to pursue long-term investments (Stein 1988) or can be employed tactically in ways that extract higher takeover offers from bidders (DeAngelo and Rice 1983).

[10] Restatements initiated by the board itself can be construed as evidence of responsible board oversight. If this interpretation applies, then we expect positive empirical associations between restatements and indicators of strong corporate governance.

[11] Notice that if an outside director serves as the board chair, the B-Index increases by two.

[12] Size and time adjustments are appropriate because governance attributes are often related to firm size (e.g., larger companies tend to have more board members who more often serve on other boards) and because we find a time trend in board characteristics. To adjust for firm size and time, each variable is compared against the median for each firm size decile and time period.

[13] This measure is constructed from six provisions – four “constitutional” provisions that discourage shareholders intervention (staggered boards, limits to shareholder bylaw amendments, supermajority requirements for mergers, and supermajority requirements for charter amendments), and two “takeover readiness” provisions against a hostile takeover (poison pills and golden parachutes).

[14] The independent director designation excludes current and former employees, providers of professional services (legal consulting, etc.), customers and suppliers, family members of employees or other individuals not deemed independent, and employees of organizations that receive charitable gifts from the firm.

[15] Studies that use this measure include Mehran (1995), Bryan, Hwang, and Lilien (2000), and Kang, Kumar, and Lee (2005).

[16] OPTVESTED is included to approximate the sensitivity of option-related pay to firm value, which is the only variable consistently significant in Peng and Roell (2004). The authors report that results are comparable using the Core and Guay (2002)’s approximation of option delta, in lieu of this measure.

[17] Independent directors with financial expertise have a CPA, CFA, or expertise in corporate financial management (CFO, treasurer, controller, VP of Finance or Accounting, and CEO of financial institutions). Notice that the firm’s executives are excluded from this designation.

[18] Results are robust to alternative profitability measures (ROA, or EBITDA/Sales, rather than MARGIN), and to log transformation of BOD_SIZE.

[19] Mean change in the index between the publication dates (for which the average length is 2.38 years) is 0.34, and the index changes no more than +1 in 85% of the cases. To align the data, we use the G-Index that is published closest to the annual meeting of the fiscal year that is being restated.

[20] Seventeen restatements due to SAB (Staff Accounting Bulletin) 101 are included because there is evidence that SAB101 restatements reflect earnings management (Altamuro, Beatty, and Weber 2003). Inferences are comparable without the SAB101 restatements.

[21] To preserve sample size and statistical power, we hand-collect the test-sample data that are missing from the IRRC dataset. Even so, 41 observations lack the requisite data.

[22] The common stock price decline is 6.26% from day -1 to day +1, and 7.45% during -28 to day -2 (market-adjusted).

[23] Unlike stock returns, governance characteristics are correlated over time, and as such, considering observations for all years for a single firm potentially overstates statistical significance.

[24] Designating the event year as the first or the middle restated year, rather than as the year prior to the last restated year, does not alter the results materially.

[25] Specifically, for each of the 185 restatement firms, we select a firm-year observation for a non-restatement (control) firm that has required data for the same year. We iterate the process -- each iteration provides 185 firm-year control observations -- until we exhaust the set of candidate control observations for a test observation. In the final round, when we are unable to identify control observations that match the year for the test firm, we select a random fiscal year for the control firm. This process yields a control sample with a distribution across years comparable to the distribution for the test sample.

[26] If clustering occurs in the test sample (e.g., with respect to industry, size, or exchange membership), the matched-pair design can yield a sample that is not representative of the population.

[27] As seen in Table II panel B, firm size is correlated with most explanatory variables. Thus, our matching requirement (within 15% of the test firm log of assets) is more stringent than in previous papers (typically within 25%). Relaxing the size matching requirement to within 25% of Test firm log total assets increases the sample by only two, and thus, is not deemed worthwhile.

[28] Some information in proxy statements is forward-looking (such as the board and subcommittee composition, size of the board and the committee, whether the CEO is the board chair), whereas other information (such as the number of board and committee meetings) is backward-looking. We align the data accordingly.

[29] GAO (2002) identifies the party that instigated the restatement for 90 firms (SEC=33, auditor=6, the firm itself=51). We do not exploit this information because careful investigation of restatement announcements undermines our confidence that this feature of the data can be reliably verified using publicly available information. For example, a number of high-profile accounting irregularities such as Cendant, WorldCom, Waste Management, and Xerox are allegedly prompted by firm management. This observation leads us to suspect that credit for discovering and disclosing accounting errors is negotiated privately among the parties involved. Even so, analyses using the SEC-prompted sub-sample of 33 restatements indicate that, except for SIZE and CEO_AGE, no variables are statistically significant at conventional levels.

[30] Classification into 48 industries as in Fama and French (1997) does not alter conclusions. We do not use the Fama-French classification because it yields industries with as few as three firms.

[31] Although Control firms are matched to Test firms using fiscal year, we include fiscal year dummies as additional precaution.

[32] Statistical significance for stock ownership of outside directors (OUTDIR_OWN) for 1997-2002 observations appears to be attributable to 1997 observations. That is, OUTDIR_OWN is not significant for specifications that use 1998-2002 data.

[33] The recommended cut-off for variance inflation factors is 10. For condition indices, the recommendation is that at least two or more independent variables with variance proportions greater than 0.5, combined with a condition index greater than 30, potentially indicate multicollinearity.

[34] In the first specification, the pseudo-R2 excluding financial control variables is 3.2%.

[35] Robustness tests are executed using KLRqÎÐ 6 7 8 E F V \ ] € ? ? ‘ ¨ © Ì Í Î Ü Ý ì í ï ð H



õìæÛÔæÍæÃæ¶Ã°Ãæ¦Ã¦™Ã?æ‡t‡?‡¦g¦cæ]hj?‚CJhj?‚hj?‚5?Kspecification 1 in Table IV.

[36] Since audit fee data are available only for fiscal year 2000 to 2002, we estimate Specification 1 of Table IV including FEERATIO (defined as non-audit fee/total audit fee) for 2000-2002 observations. The coefficient for FEERATIO is negative, as opposed to positive, and insignificant (p=0.44). The lack of explanatory power for audit fees is consistent with existing studies (Raghunandan, Read, Whisenant 2003; Agrawal and Chadha 2005), although failure to find statistically significant associations can also indicate a small sample size.









Google Online Preview   Download