Summary of an Article 1 Each student is assigned a specified article and pages. Each student submits a typed written summary and analysis of the assigned article and pages by Nov. 5, 2013, at the beginning of the class. A group of students are assigned each article. Students assigned each article should coordinate the specified parts of the page where the assignment overlaps. This is an individual assignment and not a group assignment. 2 On the written assignment indicate your name, full citation of the article and pages assigned. 3 For the assigned pages indicate the main points, any statistical test(s) performed, method(s) used to support the position(s) taken, and conclusion. 4 Your comments on the article. this is the article i need you to summarize it which only pages from "PP. 396-399" Carcello, Joseph V., Terry L. Neal, Zoe Vonna Palmrose, and Susan Scholz (2011). CEO Involvement in Selecting Board Members, Audit Committee Effectiveness, and Restatements. Contemporay Accounting Research, v. 28, no. 2 (Summer), pp.396-430. Available at ABI/INFORM Complete Additional Requirements Min Pages: 2 Other Requirements: Summerize the article from pages 396-399,

CEO Involvement in Selecting Board Members, Audit Committee Effectiveness, and Restatements* JOSEPH V. CARCELLO, University of Tennessee TERRY L. NEAL, University of Tennessee ZOE-VONNA PALMROSE, University of Southern California SUSAN SCHOLZ, University of Kansas 1. Introduction We investigate whether involvement of the chief executive ofcer (CEO) in selecting board members reduces audit committee effectiveness. A decade ago, studies found CEO inuence reected in appointments to full boards and audit committees of inside and grey directors: directors who are employees or have business and family connections to the company and its ofcers (Klein 1998; Shivdasani and Yermack 1999). However, listing standards adopted subsequent to the 1999 Blue Ribbon Committee Report limited the ability of inside and grey directors to meet independence criteria, and the 2002 SarbanesOxley Act (SOX) proscribes inside and grey directors from audit committee membership.1 Nonetheless, these regulations do not, and arguably cannot, consider the CEO's myriad personal connections. So, the CEO's inuence can still be exerted through directors who appear independent, but are not independent in fact, if such directors can obtain board membership. One way for this to occur is through CEO involvement in the board selection process. Thus, when the CEO is involved in the board selection process, there is a greater risk that a director appears independent without being independent in fact. That is, although there are no statutory violations of director independence standards, there are other ties between the CEO and the director that may compromise the director's independence. In addition to potential social ties between the CEO and board members, CEOs may appoint outside directors who share similar demographic characteristics and who are therefore more likely to support the CEO than to monitor him or her (Westphal and Zajac 1995). Therefore, we expect that previously documented benets of maintaining an independent audit committee are decreased when the CEO is involved in selecting board members. In addition, we expect that audit committee nancial expertise will be less effective in these * Accepted by Michael Willenborg. We would like to thank Beth Bryant, Brian Carver, Stacy Mastrolia, and especially Carl Hollingsworth for their data gathering assistance. We also thank the associate editor, Michael Willenborg, two anonymous referees, Mark Beasley, Rebecca Hann, Dana Hermanson, Mingyi Hung, April Klein, and seminar participants at the University of Arizona, Boston Area Research Consortium, Colorado State University, University of Kansas, Miami University, Philadelphia Area Research Consortium, University of Toronto, and University of Virginia-Darden for helpful comments. 1. Section 301 of SOX requires all rms listed on a national securities exchange or association (e.g., NYSE, AMEX, NASDAQ) to maintain a 100 percent independent audit committee. Moreover, SOX Section 407 requires all public companies to disclose whether or not a nancial expert sits on the audit committee and, if not, why not. Contemporary Accounting Research Vol. 28 No. 2 (Summer 2011) pp. 396-430 CAAA doi:10.1111/j.1911-3846.2010.01052.x CEO Involvement in Audit Committee Effectiveness 397 situations, because the expert is less likely to be independent in fact.2 To investigate these expectations, we examine the relation between CEO involvement in board selection and nancial statement restatements.3 Past research has found that audit committee independence and nancial expertise are associated with a lower likelihood of restatements (Abbott, Parker, and Peters 2004; Agrawal and Chandha 2005). We nd some evidence that CEO involvement in the board selection process eliminates the benets of both an apparently independent audit committee and nancial expertise. In addition, we nd that more severe restatements appear to drive these results, suggesting that diminished effectiveness is associated with real economic costs to the rm. Finally, in considering these issues in the context of stock price reactions to restatement announcements, we nd that the negative reaction is attenuated by audit committee independence, but only when the CEO was not involved in selecting board members. Our primary analyses focus on restatements announced from 1999 to 2001, the years immediately following the Blue Ribbon Committee Report, when there was both variation in practice and a fairly consistent regulatory regime. However, recognizing that our proxy for CEO involvement in the director selection process is likely subject to some measurement error, we extend our study to a separate sample of restatements of nancial statements for scal years from 2001 through 2003, which includes some post-SOX years. The Investor Responsibility Research Center (IRRC) database is available during this period, allowing analysis of additional observations, although variation in audit committee characteristics is more limited. We nd that our results apply in both time periods and analyses. In late 2003, U.S. listing requirements altered the formal involvement of CEOs in the board member selection process.4 Exchange regulations now require NYSE-listed companies to have a nominating committee comprising solely independent directors, and the NASDAQ's revised listing provisions require director nominees to be recommended or selected by either a nominating committee comprising solely independent directors or by the independent members of the full board of directors (Bebchuk and Fried 2004). Thus, by using samples prior to the enactment of these provisions, our study provides useful insights on the potential impact and efcacy of requirements imposed in the aftermath of SOX. Further, the question of whether CEO involvement in the director selection process reduces audit committee effectiveness remains important. For example, in Canada an independent nominating committee is not required for rms listed on the Toronto Stock Exchange (see Ontario Securities Commission 2004 National Policy 58-201, 1.1), allowing the CEO direct participation in the nominating process. In addition, U.S. public companies not traded on a national exchange (e.g., NYSE and AMEX) or quotation system (e.g., NASDAQ) and nonpublic U.S. companies likewise are not precluded from having CEOs involved in their director selection process. While recognizing that our sample does not include Canadian or U.S. private companies, our results may nonetheless inform consideration of CEO involvement in the nominating process in these settings. Finally, although U.S. companies traded via the NYSE, AMEX, or NASDAQ now have to select board members via an 2. 3. 4. Audit committee members are drawn from the full board, and prior research nds that board characteristics have a signicant inuence on audit committee characteristics (Klein 2002b). Further, independent audit committees are generally effective in monitoring nancial reporting and auditing (e.g., Klein 2002a; Carcello and Neal 2003; Abbott et al. 2004). Prior research also nds that audit committee nancial exper tise is associated with better nancial reporting quality (e.g., Anderson, Mansi, and Reeb 2004; Bedard, Chtourou, and Courteau 2004; Agrawal and Chadha 2005). Restatements for non-GAAP (generally accepted accounting principles) accounting are frequently used to assess nancial reporting and audit quality (e.g., see Public Oversight Board 2000; General Accounting Ofce 2002; Abbott et al. 2004; Kinney, Palmrose, and Scholz 2004; Agrawal and Chadha 2005). For example, see NYSE Listed Company Manual Rule 303A, AMEX Guide sec. 804, NASD Rule 4350, and Securities and Exchange Commission (SEC) Release No. 34-48745 (November 4, 2003). CAR Vol. 28 No. 2 (Summer 2011) 398 Contemporary Accounting Research independent nominating committee, or by charging the independent members of the board with this responsibility, the CEO still can potentially affect the board selection process to the extent that the nominating committee considers the preferences of the CEO before choosing board candidates. Our research suggests that CEO involvement in the director selection process reduces audit committee effectiveness, so regulators and auditors may want to monitor the effectiveness of, and the CEO's involvement in, the board selection process. In addition to the practical implications of our results, our paper contributes to the extant literatures on the board selection process, audit committee effectiveness, and restatement determinants. Prior research nds that: (1) CEO involvement in the director selection process leads to more insiders and grey directors being appointed to boards and audit committees (Klein 1998; Shivdasani and Yermack 1999), (2) CEOs who have greater power are more likely to succeed in appointing insiders and grey directors to the board (Baker and Gompers 2003; Boone, Field, Karpoff, and Raheja 2007), and (3) CEO involvement in the director selection process leads to higher CEO compensation and perquisites (Wade, O'Reilly, and Chandratat 1990; Westphal and Zajac 1995). Also, there is an extensive body of literature that generally nds that effective audit committees contribute to enhancing audit and nancial reporting quality (e.g., Carcello and Neal 2000, 2003; Klein 2002a; Abbott, Parker, Peters, and Raghunandan 2003). Our paper complements these prior research streams by showing that CEO involvement in the director selection process eliminates the monitoring benets provided by audit committees that appear independent and nancially expert. Finally, our study also contributes to prior research investigating factors associated with the likelihood of restatement, including executive expertise and compensation, auditor specialization, nonaudit fees and other company characteristics (e.g., Kinney et al. 2004; Palmrose and Scholz 2004; Aier, Comprix, Gunlock, and Lee 2005; Burns and Kedia 2006; Chin and Chi 2009). The remainder of the paper is organized as follows. Section 2 provides further background and develops our hypotheses. The research design appears in section 3, and section 4 discusses sample selection and data. The basic results follow in section 5. Section 6 presents an analysis of a later sample of restatements, and section 7 contains a summary and discusses the implications and limitations of our ndings. 2. Background and hypotheses development The monitoring effectiveness of independent and expert audit committees is widely recognized by capital market participants as an important component of corporate governance that impacts nancial reporting quality. Our fundamental premise is that, as requirements in the United States have changed over the last decade, CEO involvement in the director selection process shifted from appointing nonindependent directors to appointing directors who appear independent, but are not in fact objective, thereby reducing the positive monitoring benets of an independent and expert audit committee.5 Thus, even though regulations such as SOX Section 301 may make it appear that all audit committees are equally independent, and presumably equally effective (ceteris paribus), 5. Prior research concludes that board independence is critical because it determines a director's willingness to monitor management (Hermalin and Weisbach 1998); and independence is viewed as the most important factor in determining board effectiveness (Hermalin and Weisbach 2003). Moreover, research nds that audit committee independence is associated with higher earnings quality (Klein 2002b; Bedard et al. 2004; Vafeas 2005), more appropriate reporting in cases of client nancial distress (Carcello and Neal 2000), fewer auditor dismissals after going-concern reports (Carcello and Neal 2003), and a lower incidence of fraudulent nancial reporting (Beasley, Carcello, Hermanson, and Lapides 2000). In addition, audit committee nancial expertise is associated with less earnings management (Bedard et al. 2004), a lower cost of debt (Anderson et al. 2004), and a positive abnormal stock price reaction upon appointment to the audit committee (DeFond, Hann, and Hu 2005). CAR Vol. 28 No. 2 (Summer 2011) CEO Involvement in Audit Committee Effectiveness 399 in substance important differences may remain. If de facto audit committee independence and benets of audit committee nancial expertise can be reduced by CEO involvement in the director selection process, key mechanisms for improving the quality of nancial reporting may be compromised. For example, Wade et al. (1990) nd evidence consistent with CEOs appointing outside directors who are sympathetic to the desires of the CEO. We examine the association between CEO involvement in board selection and nancial statement restatements. We expect audit committees that are independent, not just in appearance but also in fact, and that possess nancial expertise to be associated with a reduced incidence of restatements for a number of reasons. First, an independent and expert audit committee is likely to demand more extensive external audit procedures (Abbott et al. 2003), increasing the likelihood that material misstatements will be discovered prior to issuance of the nancial statements. Such a committee also strengthens the external auditor's hand in negotiations with management surrounding accounting estimates, judgments, and application of accounting principles (Deli and Gillan 2000; DeZoort and Salterio 2001; Ng and Tan 2003). Further, an audit committee that is independent and expert should be more effective in reducing the incidence of misstated nancial statements through its own efforts in overseeing management and in reviewing the nancial statements before issuance (SEC 1999; American Institute of Certied Public Accountants [AICPA] 2005; Beasley, Carcello, Hermanson, and Neal 2009). In addition, an independent and expert audit committee is likely to both demand improved internal controls and contribute to an improved control environment itself (Krishnan 2005), which should reduce the incidence of restatements (SEC 2003a). Finally, internal audit effectiveness is improved when the internal audit function reports to a completely independent and expert audit committee (Deli and Gillan 2000; Raghunandan, Read, and Rama 2001). Because we expect any overall benecial effect of an independent and expert audit committee to be weakened by CEO involvement in the board selection process, we rst demonstrate expected relations between audit committee independence and expertise and restatements, apart from any consideration of CEO involvement. Two prior studies examine the relations between audit committee independence and nancial expertise and restatements. Abbott et al. (2004) nd a higher likelihood of restatements when audit committees are less independent and lack nancial expertise, during a sample period from 1991 to 1999. Agrawal and Chadha (2005) also report that nancial expertise reduces the likelihood of restatement during 2000 and 2001, but they do not nd an association between independence and restatements. However, given the large body of literature that nds a relation between audit committee independence and improved monitoring of the nancial reporting and auditing process, we expect a negative relation between audit committee independence and restatements. This leads to our rst hypothesis (expressed in alternate form): HYPOTHESIS 1. Disregarding CEO involvement in board selection, there is a negative relation between audit committee independence and restatements. We also attempt to replicate the negative relation between audit committee nancial expertise and restatements found by both Abbott et al. 2004 and Agrawal and Chadha 2005. This leads to our second hypothesis (expressed in alternate form): HYPOTHESIS 2. Disregarding CEO involvement in board selection, there is a negative relation between audit committee nancial expertise and restatements. As previously noted, an audit committee member may meet the independence requirements established by the stock exchanges (independence in appearance) but he she may not be truly independent from management (independence in fact) and, therefore, may not be an CAR Vol. 28 No. 2 (Summer 2011) 400 Contemporary Accounting Research effective monitor of the nancial reporting process. For example, in many instances there is no prohibition against appointing personal friends and or business associates of the CEO or of other members of top management to the board (Glassman 2003; Cohen, Krishnamoorthy, and Wright 2004; Morck and Yeung 2006).6 Prior research nds that board members often have prior personal or business relationships with the CEO and senior management of the companies they serve, even though they are considered technically independent (e.g., see Beasley et al. 2009; Tanous 2003). Hermalin and Weisbach (1998: 111) observe that it is relatively easy for the CEO to choose directors who meet the statutory requirements of independence but who are not independent in fact. And Nelson (2006: 34) argues audit committee independence and hence effective oversight of the nancial reporting process can be compromised if management-dominated companies allow the CEO to select the directors who comprise the audit committee. Prior research indicates that under previous regulatory regimes board nominees were approved, if not chosen, by the CEO and other members of top management (e.g., see Lorsch and MacIver 1989). As stated by former SEC Commissioner Cynthia Glassman: ''At some companies, it appears that the CEO hires the Board instead of the Board hiring the CEO'' (Glassman 2003). Shivdasani and Yermack (1999) nd that fewer independent directors and more grey directors are appointed to the board when the CEO is involved in the board nomination process. They argue that directors who join the board under these conditions are less effective monitors (1831). Like Shivdasani and Yermack 1999, we consider grey directors to be nonindependent.7 We do not separately study grey directors, because studies nd that an audit committee must be completely independent to be effective (Deli and Gillan 2000; Bedard et al. 2004; Bronson, Carcello, Hollingsworth, and Neal 2009). In addition, Klein (1998) nds that, when the CEO is involved in the director selection process, rms have signicantly lower percentages of independent directors on their audit, compensation, and nominating committees. Thus, she argues that CEO involvement in the director selection process can lead to ''board captivity'' (285). Because audit committee members are selected from the full board (Klein 2002b), we empirically test these conjectures by examining whether CEO involvement in the director selection process affects the relation between audit committee independence and restatements.8 This leads to our third hypothesis (expressed in alternate form): HYPOTHESIS 3. The negative relation between audit committee independence and restatements is reduced when the CEO is involved in the director selection process. Similarly, the expertise of an audit committee nancial expert may not result in improved monitoring of the nancial reporting process if the CEO has participated in the director 6. 7. 8. In fact, Glassman (2003) argues that ''personal relationships with the CEO living in the same community, kids at the same school, moving in the same social circle are just as likely to undercut independence''. And, as Morck and Yeung (2006) argue in a report commissioned by the Task Force to Modernize Securities Legislation in Canada, allowing public shareholders to nominate and elect a fraction of directors ''circumvents the problem of the CEO or controlling shareholder nominating 'independent' directors who meet the de jure denition, but who are not really de facto independent. Golng buddies are not independent'' (334). Shivdasani and Yermack (1999) dene grey directors as nonemployee directors who are retired employees, are relatives of the CEO, have business ties to the rm, or have interlocking directorships with the CEO. Although we use an absolute dichotomy of involved not involved to reect the CEO's involvement in the director selection process, we acknowledge that it is not possible to know with certainty that the CEO has no involvement in the director selection process (e.g., even if the rm has a completely independent nominating committee, the CEO may still exert some ''behind the scenes'' inuence on this process). Therefore, our indicator variable CEOINVOL may actually reect the extent of CEO involvement (i.e., more less involved) rather than an absolute measure of involvement. CAR Vol. 28 No. 2 (Summer 2011) CEO Involvement in Audit Committee Effectiveness 401 selection process. A nancial expert should be better able to assess the quality of a rm's nancial reports, presumably lessening the likelihood of a restatement, but we posit that an expert must be independent for the rm to realize the benets of the audit committee member's expertise. This leads to our fourth hypothesis (expressed in alternate form): Hypothesis 4. The negative relation between audit committee nancial expertise and restatements is reduced when the CEO is involved in the director selection process. 3. Research design We rst test the relation between audit committee independence and nancial expertise and restatements (Hypothesis 1 and Hypothesis 2). Because we match our restatement sample with nonrestatement rms, we use a matched-pairs (conditional) logistic regression model (e.g., see Hosmer and Lemeshow 2000). We estimate the following model: RESTATE b1 APPINDEP b2 ACEXPERT b3 ACMEET b4 ACSIZE b5 LNBDSIZE b6 BOSS b7 MVBV b8 NICHG b9 LNAGE b10 ACQUIS b11 LNTENURE b12 BIG5 b13 BODINDEP b14 STOCKOWN e: The variables are dened as follows: RESTATE: an indicator variable that captures whether the rm's annual or quarterly nancial statements led between 1999 and 2001 were restated during 2000 and 2001 due to non-GAAP reporting (1 = restatement due to non-GAAP reporting, 0 = no restatement). APPINDEP: an indicator variable that captures whether all of the rm's audit committee members appear to be independent during the rst misstated year (1 = 100 percent independent audit committee, 0 = otherwise).9 An audit committee member appears independent if the committee member's only tie to the rm is his her service as a board member.10 ACEXPERT: an indicator variable that captures whether at least one nancial expert sits on the audit committee (1 = at least one nancial expert sits on the audit committee, 0 = no nancial experts sit on the audit committee). We dene a nancial expert as an audit committee member who is now, or has been in the past, a certied public accountant, chief nancial ofcer, controller, treasurer, vice president-nance, investment banker, or venture capitalist. Our approach recognizes the conicting denitions of nancial expertise. All of the rms in our sample have an audit committee nancial expert as dened more broadly by the SEC's nal implementation guidance under section 407 of SOX (SEC 2003b), which essentially allows senior management (CEOs, in particular) who have supervised accounting functions (i.e., indirect accounting or nancial expertise) to count as nancial experts. In our model, the coefcient on the nancial expert variable indicates whether restatements are less likely when the audit committee has a member with direct accounting 9. 10. Klein (2002a, 377-78) discusses the new stock exchange listing standards requiring listed companies to maintain audit committees of at least three members, all of whom are independent. Although these rules were issued in 1999, they were not effective until June 2001. Therefore, a large number of our sample observations were not affected by this change in listing standards. We require 100 percent committee independence to be consistent with current exchange listing and SOX requirements. These requirements are supported by research indicating that even one nonindependent committee member can interfere with the free ow of information among committee members (Deli and Gillan 2000; Bedard et al. 2004; Bronson et al. 2009). CAR Vol. 28 No. 2 (Summer 2011) 402 Contemporary Accounting Research or nancial expertise (consistent with the SEC's original proposal for dening a nancial expert; SEC 2002). In addition to the test variables of interest, we control for the effects of other factors related to the incidence of restatements. These variables include audit committee characteristics, including size (ACSIZE) and number of meetings (ACMEET), and overall board characteristics, including board independence (BODINDEP), board size (LNBDSIZE), and whether the CEO is also the chairman of the board (BOSS). Company characteristics are the number of years the company has been public (LNAGE), growth (MVBV), merger or acquisition activity (ACQUIS), and the percentage of the company's stock owned by insiders (STOCKOWN). Auditor characteristics are audit rm quality (BIG5) and the length of the auditor-client relationship (LNTENURE). We also include a proxy that measures deterioration in the company's protability (NICHG). Unless noted, all variables are measured using values from the rst year misstated. Governance variables are obtained from the relevant proxy statement led with the SEC, and nancial variables are obtained from COMPUSTAT's Research Insight. Further explanation of the measurement and rationale for these control variables is provided in the Appendix. The primary objective of our paper is to examine whether CEO involvement in the director selection process reduces the positive monitoring benets of audit committee independence (Hypothesis 3) and nancial expertise (Hypothesis 4). To test these hypotheses, we add an indicator variable that measures CEO involvement in the director selection process to our model. CEOINVOL is equal to 1 if the CEO is involved in the director selection process, and 0 if not. As stated previously, we measure whether all of the audit committee members are independent in the rst misstated year. To measure the effect of CEO involvement in the director selection process on audit committee effectiveness, we measure CEO involvement in the director selection process in the year before the rst misstated period. An audit committee member at the beginning of year t would have been appointed or reappointed to the audit committee at the previous year's annual meeting (year t ) 1). In addition, corporate board members typically are either appointed annually or are elected to three-year terms. For those board members who are joining the board for the rst time, or who are being reappointed to the board in year t ) 1, CEOINVOL also reects the role of the CEO in their appointment or reappointment to the board.11 We classify the CEO as being involved in the director selection process if there is an afrmative indication of the CEO's involvement in the nomination process. These afrmative indicators are: (1) the rm has a nominating committee and the CEO is on the committee12; (2) the rm does not have a nominating committee, but the proxy statement indicates that the entire board of directors is responsible for the director selection process; and (3) the rm does not maintain a nominating committee, but the proxy statement indicates that ''management is involved'' in the director selection process. 11. 12. We do not measure CEO involvement at the time each director rst joined the audit committee. Although it seems likely that rms with CEO involvement had similar involvement in the past, there is a risk that some companies coded as ''CEO involved'' were really ''CEO not involved'' at the time the director rst joined the audit committee. Any such misclassications would bias against us nding any results. Nonetheless, we perform a sensitivity test where we limit our sample to only those rms where each audit committee member was appointed after the current CEO assumed his or her position (see section 5). We focus on CEO rather than management involvement in the nominating process. However, there are four instances where the nominating committee has at least one insider but not the CEO. These four observations are classied as CEO involved. Two of the four have an inside chairman of the board on the nominating committee; the other two insiders are: (1) a company co-founder and senior executive vice president and (2) a company vice president and general counsel. Our results are not sensitive to excluding these four observations. CAR Vol. 28 No. 2 (Summer 2011) CEO Involvement in Audit Committee Effectiveness 403 We classify the CEO as not being involved in the director selection process if none of the above conditions exist. There are three possible states if we code the CEO as not involved in the director selection process. First, there may be a clear indication that the CEO is not involved, because the rm maintains a nominating committee composed entirely of independent directors. This state is observable, and it is likely that the CEO (or other management) is not directly involved in the director selection process. Unfortunately, as discussed in the next section, very few of our sample rms maintain a completely independent nominating committee, precluding direct tests of the effect of such a committee. Second, the rm may charge the independent members of the full board with the director selection process. If so, it is plausible to argue that the CEO is not involved in the director selection process. During our sample period this state is not necessarily publicly observable. None of the companies in our sample disclose that the independent members of the board are responsible for the director selection process. The third possibility is that the full board of directors, including the CEO, may be primarily responsible for the director selection process although this fact is not disclosed in the proxy statement. In our study, a rm in this third state would arguably be incorrectly classied as one where the CEO is not involved in the director selection process. Because this state is also unobservable during this period, we are unable to separate any misclassied companies from the third state from those in the second state that are properly classied. In sum, in the two last states, we assume the CEO is not involved absent information to the contrary. It is likely some rms are misclassied under this assumption. This is a potentially important limitation in the measurement of our CEOINVOL variable during the early sample period. However, during the later sample period more precise data are available, as described in section 6. Analyses using those measures are consistent with results reported here. We also note that erroneously including some rms with CEO involvement with the not-involved rms introduces noise to our analysis, biasing against nding signicant results for our hypotheses. 4. Sample selection In our primary sample, we study restatements of previously issued nancial statements where the rst disclosure that the company may will restate its nancial statements occurred during 2000 or 2001 and involved correction of non-GAAP reporting on Form 10-Ks originally issued for 1999 or 2000 or Form 10-Qs originally issued for 2000 or 2001. We include restatements of 10-Qs because timely reviews of quarterly results were required during our sample period, so quarterly nancial statements clearly fall within the purview of the audit committee. We identied 157 rms announcing such restatements by searching the LEXIS-NEXIS News Library and Form 8-K le. Restatements were identied using keyword searches such as ''restat'', ''revis'', ''adjust'', and ''error''. Following similar restatement studies by Abbott et al. 2004 and Agrawal and Chadha 2005, we selected a matched sample of nonrestating rms. We matched on stock exchange, industry, and size. We measured rm size in the year preceding the restatement using the market value of equity, and matched within plus or minus 30 percent.13 We also veried that control rms were not involved in 13. We match on stock exchange because of differences among exchanges in requirements related to audit committee composition and audit committee nancial expertise during our sample period. Based on the conclusions of Bhojraj, Lee, and Oler 2003, we use the Global Industry Classications Standard (GICS) system to measure industry classication. Seventy-seven of the restatement rms were matched at the eight-digit GICS code level, 14 rms were matched at the six-digit GICS code level, eight rms were matched at the four-digit GICS code level, and ve rms were matched at the two-digit level. There was not a signicant difference between the mean market value of equity for the restatement and nonrestatement groups (p-value > 0.10), indicating that our size-matching procedure was effective. CAR Vol. 28 No. 2 (Summer 2011) 404 Contemporary Accounting Research accounting-related litigation and did not receive an Accounting and Auditing Enforcement Release (AAER) from the SEC between 1999 and 2004. Of the 157 restatement rms originally identied, 53 observations are lost because of missing data. Most of these (n = 32) are due to missing proxy statements and or Form 10-Ks necessary for hand-collecting governance data. Our nal sample consists of 104 restatement rms matched with an equal number of nonrestatement rms.14 5. Results Univariate analysis Table 1, panel A presents the distribution of restating and control rms across the CEO involved not involved classication categories described above. The number of restating and control companies is fairly even in each category. For restating (control) companies, the CEO is on the nominating committee in 11 (12) cases; for 20 (21) rms, the proxy discloses that the full board, which includes the CEO in every case, participates in nominations; and one (one) rm discloses that management is involved in the nominating process. For the latter, in both cases, the CEO is the only insider on the board. Thus, a total of 32 restating rms and 34 control rms are classied as CEO involved. Among the 72 restating (70 control) rms that are coded as CEO not involved, only eight (nine) report maintaining a 100 percent independent nominating committee. For the remaining 64 (61) rms, we assume the CEO is not involved, because the rm does not indicate that the CEO or management is involved in nominations. Table 1, panel B presents descriptive statistics for the independent variables by restatement and control rms. Based on prior research, we expect nonrestatement rms to have more independent audit committees and nancial expertise than restatement rms. Indeed, almost two-thirds of the nonrestatement rms have a nancial expert on the audit committee, compared to little more than half of the restatement rms (p-value 0.10). CAR Vol. 28 No. 2 (Summer 2011) CEO Involvement in Audit Committee Effectiveness 411 TABLE 3 Conditional logistic regression of restatements from 1999-2001 on CEO involvement in the nominating process, audit committee characteristics and control variablesa RESTATE b1 CEOINVOL b2 APPINDEP b3 CEOINVOLAPPINDEP b4 ACEXPERT b5 CEOINVOLACEXPERT b6 ACMEET b7 ACSIZE b8 LNBDSIZE b9 BOSS b10 MVBV b11 NICHG b12 LNAGE b13 ACQUIS b14 LNTENURE b15 BIG5 b16 BODINDEP b17 STOCKOWN e Predicted Relation Variableb CEOINVOL APPINDEP CEOINVOL*APPINDEP ACEXPERT CEOINVOL*ACEXPERT ACMEET ACSIZE LNBDSIZE BOSS MVBV NICHG LNAGE ACQUIS LNTENURE BIG5 BODINDEP STOCKOWN Number of Observations Pseudo R2 Estimated Coefcients Standard Errors z-statistic )1.054 )1.155 1.608 )1.047 0.478 )0.286 )1.211 )1.241 0.750 0.012 0.035 )0.200 0.783 )0.044 0.099 1.959 0.028 0.825 0.447 0.820 0.444 0.744 0.388 0.516 0.872 0.394 0.027 0.023 0.359 0.450 0.267 0.733 0.859 1.093 )1.28 )2.58*** 1.96** )2.36*** 0.64 )0.74 )2.34*** )1.42 1.91** 0.44 1.54 )0.56 1.74** )0.16 0.13 2.28 0.03 + ) + ) + ) ) + + + ? ) + ) ) ) ? 208 0.24 Notes: ** and *** indicate signicance levels of 0.05 and 0.01, respectively, based on one-tailed tests when a sign is predicted, two-tailed otherwise. Joint Tests of Coefcients: Predicted Relation Estimated Coefcients Standard Errors z-statistic (b2 + b3 = 0) ? 0.453 0.686 0.66 (b4 + b5 = 0) ? )0.570 0.665 )0.86 Coefcients a Observations are the 104 restatement rms in Table 1 plus the 104 matched, nonrestatement rms. b Variable denitions: RESTATE = 1 if the rm's annual or quarterly nancial statements led between 1999 and 2001 were restated during 2000 and 2001 due to non-GAAP reporting, else 0. (The table is continued on the next page.) CAR Vol. 28 No. 2 (Summer 2011) 412 Contemporary Accounting Research TABLE 3 (Continued) CEOINVOL = 1 if the CEO is involved in selecting board members, else 0. APPINDEP = 1 if the audit committee is composed of 100 percent independent directors, else 0. ACEXPERT = 1 if the audit committee has at least one nancial expert, else 0. ACMEET = 1 if the audit committee meets at least four times, else 0. ACSIZE = 1 if the audit committee has at least three members, else 0. LNBDSIZE = natural log of the number of directors on the full board. BOSS = 1 if the CEO is also the chairman of the board, else 0. MVBV = market value to book value at end of last clean year. NICHG = (net income at end of last clean year - net income at end of year prior to last clean year) (net income at end of year prior to last clean year). LNAGE = natural log of the age of company at end of rst misstated year. ACQUIS = 1 if there was an acquisition during the misstated period, else 0. LNTENURE = natural log of the tenure of auditor at end of rst misstated year. BIG5 = 1 if the auditor for the rst year misstated was a Big 5 auditor, else 0. BODINDEP = percentage of non-audit-committee board members who are independent. STOCKOWN = percentage of the company's stock that is owned by insiders. these results is consistent with our expectations. None of the other control variables is signicant at conventional levels. Restatement severity To provide some perspective on whether restatements associated with CEO involvement have meaningful economic effects on rms, we use two different measures to identify severe restatements. The rst denes a severe restatement based on both characteristics of the restatement and the market reaction to it. That is, we code restatements involving fraud, SEC AAERs, litigation, a change in net income revenue of more than )10 percent, or a market reaction to the announcement of more than )10 percent as severe (n = 69 restatements). The second measure denes severe restatements based only on reactions of market participants: those with litigation or market reactions of more than )10 percent (n = 56 restatements). Partitioning our sample into subsets (severe not severe) based on the two denitions of severe and applying the model in Table 3 reveals that our results are driven by the severe restatements (results not tabulated). That is, results for the two partitions of severe restatements are consistent with those reported in Table 3, while none of the variables are signicant in the two partitions of nonsevere restatements, nor is the overall model signicant. This suggests the diminishment in audit committee effectiveness with CEO involvement in the director selection process is associated with real economic costs to the rm. Stock market reaction to restatement announcement Prior research documents a signicant negative abnormal stock price reaction to the announcement of a restatement (Palmrose, Richardson, and Scholz 2004). We posit that an audit committee that is independent and expert might mitigate the negative stock price reaction to the restatement announcement, either because the market perceives that an audit committee with these characteristics will be more likely to thoroughly investigate and correct the non-GAAP accounting or because a restatement from a rm with CAR Vol. 28 No. 2 (Summer 2011) CEO Involvement in Audit Committee Effectiveness 413 otherwise good corporate governance is viewed as a less systemic failure. However, the market's perception of audit committee independence and nancial expertise might be colored by the involvement of the CEO in the director selection process. We investigate the effect of audit committee independence and nancial expertise on the stock market reaction and whether this effect differs depending on CEO involvement in the director selection process for our sample of restatements. The Center for Research in Security Prices (CRSP) market data are available for 89 of the 104 restatement announcement dates. The 89 restatement announcements in our market reaction analysis occur on 75 calendar days across two years (62 days have one announcement, 12 days have two announcements, and one day has three announcements). Abnormal returns are calculated using the market-adjusted method and an equally weighted index. The abnormal returns are cumulated over a two-day window beginning on the announcement date. On average, rms in our sample experience signicantly negative announcement reactions mean (median) of )8.3 percent ()4.3 percent) (pvalue