Economic Consequences of Financial Reporting ... - SSRN papers

0 downloads 0 Views 242KB Size Report
Christine I. Wiedman ... which suggests that managers trade off the costs of these actions against the .... These proposed disclosure requirements did not engender much .... that this company was willing to pay $1 million to restructure the entire ... and faced a possible decrease in its forecasted diluted EPS for 2004 of $1.00.
Economic Consequences of Financial Reporting Changes: Diluted EPS and Contingent Convertible Securities Carol A. Marquardt Baruch College – CUNY Christine I. Wiedman University of Waterloo

Abstract. This paper examines the economic consequences of changes in the financial reporting requirements for contingent convertible securities (COCOs). Using a sample of 199 COCO issuers from 2000-2004, we find that issuers are more likely to restructure or redeem existing COCOs to obtain more favorable accounting treatment when the financial reporting impact on diluted earnings per share (EPS) is greater and when EPS is used as a performance metric in CEO bonus contracts. These results provide new evidence that managers are willing to incur costs to retain perceived financial reporting and compensation benefits. We also present evidence of significantly negative stock returns around event dates associated with the financial reporting changes, consistent with investor anticipation of the agency costs associated with the rule change. Keywords: Economic consequences; agency costs; diluted EPS; executive compensation; convertible securities. JEL Classification: M41

We thank Ravi Arcot, CEO of Kynex, Inc., for providing convertible bond data. We also thank Eli Amir, Maya Atanasova, Donal Byard, Greg Clinch, Fabrizio Ferri, Richard Frankel, Cristi Gleason, Duane Kennedy, Ken Klassen, Henock Louis, Bin Ke, Zack Liu, Michel Magnan, Claudine Mangen, Jim McKeown, Mark Nelson, Mort Pincus, Joshua Ronen, Steve Ryan, William Scott, Charles Shi, Steve Umlauf, Joe Weber, Mike Willenborg, and participants at the MIT, Iowa, Penn State, London Business School, Baruch, Arizona, Connecticut, Purdue, UC-Irvine, Ohio State, York, Waterloo, Western Ontario, and Concordia accounting workshops, the 2005 CAAA Conference, and the 2005 AAA Annual Meeting for helpful comments and suggestions.

Economic Consequences of Financial Reporting Changes: Diluted EPS and Contingent Convertible Securities This paper examines the economic consequences of changes in the financial reporting for contingent convertible securities. Contingent convertibles (COCOs) are convertible securities that cannot be converted into shares of common stock until a prespecified stock price is reached; i.e., conversion is “contingent” upon reaching the price threshold. From the time of their first appearance in late 2000 through mid-2004, firms were able to postpone including the effects of the COCOs in diluted earnings per share (EPS) calculations until the necessary conditions for conversion were satisfied. Marquardt and Wiedman (2005) show that this perceived financial reporting benefit was the primary determinant of firms’ decisions to issue COCOs rather than traditional convertible bonds during this time period. However, in mid-2004, the Financial Accounting Standards Board (FASB) revised the financial reporting requirements for COCOs. Beginning with the first reporting period after December 15, 2004, COCOs would be included in diluted EPS calculations regardless of whether market-based contingences were satisfied.1 In this paper, we examine two dimensions of the economic consequences arising from this accounting change.2 First, we examine the responses of the COCO issuers to the new financial reporting requirements. In particular, we examine the determinants of managerial decisions to undertake costly restructurings or cash redemptions of outstanding COCOs that would mitigate the financial reporting impact of the new rule. Second, we examine the shareholder wealth effects associated with the rule change, including mean effects as well as cross-sectional variation in shareholder reactions. Our analysis is based on a sample of 199 COCOs issuers from 2000 to early 2004. Using multivariate probit analysis, we first examine managers’ decisions to either modify the terms of the COCOs to require cash rather than stock settlement of par values at conversion (which would effectively nullify the reporting effects of the new rule) or to 1

See EITF Issue No. 04-8, “The Effect of Contingently Convertible Debt on Diluted Earnings per Share.” Consistent with Holthausen and Leftwich (1983, p. 77), we view accounting choices as having economic consequences “if changes in the rules used to calculate accounting numbers alter the distribution of firms’ cash flows, or the wealth of parties who use those numbers for contracting or decision making.” 2

redeem the COCOs for cash prior to the adoption date. We find that the likelihood of these events is positively associated with the impact of the rule change on diluted EPS, which suggests that managers trade off the costs of these actions against the perceived benefits of reporting a higher diluted EPS figure; i.e., they “buy” a better diluted EPS number. We further find that the likelihood of a restructuring or redemption is positively associated with firms’ use of EPS as a performance metric in CEO bonus contracts, consistent with agency costs. In addition, firm size and lobbying efforts against the proposed ruling are positively associated with the likelihood of restructuring or redemption, while financial leverage is negatively associated, which suggests that highlylevered firms are less able to bear the costs associated with these decisions. We also examine stock returns around announcement dates of restructurings and redemptions and find that mean and median three-day abnormal returns are significantly negative, which suggests that shareholders view these managerial decisions as value-decreasing. We examine the shareholder wealth effects associated with rule change in our second set of tests. Using standard event study methodology (Schipper and Thompson 1983), we examine shareholder wealth effects around six separate dates related to the financial reporting changes for COCOs and document significantly negative effects around two of these dates. We also accumulate returns across all six event dates and report a significantly negative shareholder reaction of approximately three percent of firm value, which suggests that shareholders expect the potential responses of COCO issuers to the rule change to impose net costs on the firm. Using the estimation procedure presented in Sefcik and Thompson (1986), we also examine cross-sectional variation in shareholder reactions around the event dates. Consistent with our predictions, we find that the impact of the financial reporting change on diluted EPS is a significant determinant of shareholder reactions. We further find that shareholder reactions vary cross-sectionally with the disclosure quality of COCO-related information, whether EPS is used as a performance metric in CEO bonus contracts, the degree to which COCOs are out-of-the-money, and whether the COCOs are callprotected. Overall, the findings are consistent with investors viewing the rule change as resulting in increased agency and renegotiation costs for COCO issuers.

2

Our results contribute to the accounting literature in several ways. First, we extend the stream of literature that examines the effects of new accounting rules on managerial decision-making. For example, Imhoff and Thomas (1988) examine the effects of SFAS No. 13 on the capital structure decisions of lessees and document a systematic substitution from capital to operating leases and nonlease sources of financing following adoption of the standard. Mittelstaedt, Nichols, and Regier (1995) report that the financial reporting consequences of SFAS No. 106, which required recognition of post-retirement benefit (PRB) obligations, played a significant role in explaining reductions in these benefits, and Amir and Livnat (1996) find that late adopters of SFAS No. 106 were more likely to reduce PRB obligations by negotiating a plan amendment than early adopters. More recently, Venkatachalam, Rajgopal, and Choudhary (2005) show that firms accelerate vesting of employee stock options prior to the effective date of SFAS No. 123-R to avoid reporting an expense under the new standard, and Bens and Monahan (2005) provide evidence that U.S. banks restructure asset backed commercial paper conduits to avoid consolidation under FASB Interpretation No. 46. In empirically documenting the effect of the accounting change for COCOs on managerial decisions to restructure or redeem these securities, our study corroborates and extends previous findings in this stream of literature in a novel setting. However, a key distinction between our study and prior research is that the reporting changes we examine affect only diluted EPS; there are no income statement or balance sheet effects, as in the studies cited above. Our results thus provide compelling evidence that managers not only view diluted EPS as an important performance measure, consistent with Marquardt and Wiedman (2005), but will actually incur costs in order to report a higher figure. We are unaware of a similar finding in the accounting literature.3 We also contribute to the literature relating executive bonus compensation to earnings management. Previous research (Healy, 1985; Holthausen, Larcker, and Sloan, 1995; Guidry, Leone, and Rock, 1999) shows that managers will manipulate net income 3

Previous research documents that managers will incur costs to secure other financial reporting effects, such as reporting higher net income (e.g., Ayers, Lefanowicz, and Robinson, 2002; Matsunaga, Shevlin, and Shores, 1992; Erickson, Hanlon, and Maydew, 2004) or removing liabilities from the balance sheet (Engel, Erickson, and Maydew, 1999). Graham, Harvey, and Rajgopal (2005) comment on the difficulty of convincingly documenting this tradeoff using archival data.

3

in response to bonus contracts. We extend this literature by providing evidence that managers are more likely to incur debt restructuring costs to increase diluted EPS when their bonuses depend upon EPS, consistent with arguments forwarded by Watts and Zimmerman (1990). We also extend the literature on the shareholder wealth effects associated with new accounting standards or other regulatory changes. Numerous studies have examined these effects, including Collins, Rozeff, and Dhaliwal (1981); Dechow, Sloan, and Sweeney (1996); D’Souza (2000); and Weber (2004). However, net income is affected in each of the rule changes examined in these studies. Our findings are unique in this literature in that we document significant shareholder wealth effects related purely to the reporting of diluted EPS. Relatedly, because the changes we examine do not affect net income, the contracting cost arguments typically offered as explanations for shareholder reactions around reporting changes, such as debt or political cost hypotheses, are less likely to apply. Instead, we focus on the role of agency costs in explaining market reactions to the financial reporting changes for COCOs. Our results suggest that investors anticipate the agency costs associated with accounting rule changes and respond accordingly. The remainder of this paper is organized as follows. Section 1 provides background on the financial reporting issues surrounding contingent convertible securities. In section 2, we develop our hypotheses regarding the managerial responses to the new reporting requirements, describe our sample, and present our empirical results from our probit analysis. In section 3, we present our analysis of the shareholder wealth effects associated with the rule change, and in Section 4 we summarize our findings and conclusions.

1. Background COCOs are convertible securities that cannot be converted into shares of common stock until a pre-specified stock price, or conversion threshold, is reached. This is in contrast to traditional convertible securities, which may be converted into shares at any time. This convertible bond structure first appeared in late 2000 and quickly became a popular financing vehicle, with more than 300 firms issuing COCOs through the first half

4

of 2004. Reports in the financial press attributed this growing popularity of COCOs to their financial reporting advantages over traditional securities (see Henry, 2003). Marquardt and Wiedman (2005) provide a detailed discussion of the financial reporting effects associated with COCOs, which we briefly summarize here. Through mid-2004, SFAS 128 required that the effects of convertible securities be reflected in diluted EPS through application of the ‘if-converted’ method. Under this method, aftertax interest charges are added back to the numerator, and the debt is assumed to be converted as of the beginning of the accounting period, with the resulting common shares included in the denominator. If, however, the shares were “contingently issuable,” then the dilutive effects of the securities were only included if “the necessary conditions for issuance are satisfied” at the end of the accounting period (see paragraph 30). As a result of this loophole in SFAS 128, firms could simply add contingent conversion to a traditional convertible security and postpone including its effects in diluted EPS until the conversion threshold was reached. Exhibit 1 illustrates these effects using Omnicom’s 2003 financial statements. Omnicom Group Inc. issued three series of COCOs over 2001-2003 for aggregate proceeds of $2.3 billion. All three series of COCOs were zero-coupon, zero-yield bonds with thirty-year maturities. Because the contingencies of these COCOs were not satisfied by the end of 2003, the dilutive effects of the combined offerings were excluded from their reported diluted EPS of $3.59 for that year. Had these effects been included, Omnicom would have reported diluted EPS of only $3.26. Omnicom is thus able to increase its reported profitability by over 10% by simply adding contingent conversion features to traditional convertible bonds. Marquardt and Wiedman (2005) provide empirical evidence that these financial reporting effects are indeed the primary determinant in the decision to issue COCOs.4 They further show that firms bear minimal costs in choosing to issue COCOs rather than traditional convertible bonds, which is consistent with their growing popularity through the first half of 2004. In fact, by mid-2004, approximately 85% of all U.S. convertibles issuers included contingent conversion features. 4

In addition to the financial reporting benefits, Marquardt and Wiedman (2005) also show that the use of earnings-based CEO bonus plans, reputation costs, free cash flows, and tax factors play significant roles in the decision to issue COCOs rather than traditional convertible bonds.

5

However, during 2004, the FASB began to take notice of the COCO phenomenon. First, in early 2004, the FASB released Staff Position 129-a, “Disclosure Requirements under FASB Statement No. 129, Disclosure of Information about Capital Structure, Relating to Contingently Convertible Securities” in response to constituents’ concerns that COCO-related disclosures were inconsistent between companies or inadequate in communicating necessary information to users of financial statements.5,6 FAS 129-a proposed that COCO issuers disclose the significant terms of the conversion features to enable users to understand the circumstances of the contingency and the potential impact of conversion. It also required issuers to indicate whether issuable shares are included in diluted EPS and the reasons why or why not. These proposed disclosure requirements did not engender much controversy, if the comment letters are any indication. Only the Big 4 accounting firms offered comment letters, and all four voiced support for the proposed disclosures. At the end of 45-day comment period, the FASB posted the final draft of FAS 129-1, which became effective immediately, to its website. The mandated disclosure pronouncements were only the beginning of the FASB’s actions. In May 2004, the Emerging Issues Task Force (EITF) added Issue No. 04-8, “The Effect of Contingently Convertible Debt on Diluted Earnings per Share,” to its agenda for its next meeting, scheduled for June 30th and July 1st. Discussion materials distributed before the meeting expressed the view that the contingencies in COCOs do not have economic substance and are included solely to secure favorable accounting treatment under SFAS 128. On July 1, 2004, the EITF reached a tentative consensus that COCOs should be included in diluted earnings per share computations regardless of 5

Equity analyst reports during this time (see Gainey, 2004) similarly characterized overall COCO-related disclosure as “generally poor if the goal is to alert and inform analysts and investors of the existence of these instruments and their impact on the financials.” 6 In February 2003, the FASB introduced the FASB Staff Position (FSP) to issue application guidance like that previously found in Staff Implementation Guides and Staff Announcements. Before issuing an FSP, the FASB staff circulates a draft of a proposed FSP to Board members for review. If a majority of the Board does not object to the proposed FSP, it is announced at an open public meeting of the Board. Following the meeting, the proposed FSP is posted to the FASB website for a comment period and is announced in that day’s Action Alert. At the end of the comment period, the FASB staff drafts the final FSP. As with the proposed FSP, if a majority of the Board does not object to the final FSP, it is posted to the FASB website and announced in the Action Alert. The FSP is intended to ensure more timely and consistent communication about the application of FASB literature than the previous procedures that were in place (see www.fasb.org).

6

whether market-based contingencies have been met. In addition, the EITF proposed that prior period EPS amounts presented for comparison purposes be restated to conform to the new accounting treatment. The EITF’s decision received some coverage in the financial press, as well as in equity analyst reports. The Wall Street Journal reported on the proposed changes on July 6th and 8th, and a Bear Stearns equity research report dated July 2, 2004 (see McConnell et al., 2004), described the EITF’s tentative conclusion as “a surprise” and stated that confirmation of the tentative decision was expected at the next EITF meeting, scheduled for late September 2004. Furthermore, both sources predicted that the proposed accounting change would hurt the COCO market, at least temporarily while issuers waited for the FASB’s final decision. This prediction proved to be correct – Risk Magazine later reported on the downturn in the convertible bond market in a September 2004 article, noting in reference to investment bankers’ frequent exploitation of accounting rules that “loopholes can sometimes turn into nooses.” The FASB posted the draft abstract of Issue No. 04-8 to its website in mid-July 2004 and invited comments by constituents. Of the 32 comment letters posted to the FASB’s website, 31 expressed opposition to some aspect of the proposed accounting change, and only one, written by an equity analyst, praised the move. Despite the views expressed in the overwhelming majority of the comment letters, the EITF reaffirmed at their Sept. 30, 2004 meeting their earlier tentative decision that COCOs should be included in diluted EPS calculations regardless of whether market contingencies are satisfied. They also agreed that this treatment should be applied retroactively as of the adoption date of the consensus – i.e., the end of the first reporting period after December 15, 2004. The EITF also discussed at their September meeting what they termed the “unique” transition provisions for Issue 04-8, which would allow COCO issuers to avoid restatement or even the use of the if-converted method of calculating diluted EPS altogether: Diluted earnings per share of all prior periods presented for comparative purposes should be restated to conform to the consensus guidance. For instruments within the scope of this Issue whose terms have been modified prior to the date of adoption of this consensus, the consensus would apply

7

to terms of the instrument in place at the date of adoption, and diluted earnings per share of all prior periods would be restated based upon the modified terms. For instruments within the scope of this Issue that have been cash settled before the date of adoption, restatement of diluted earnings per share is not required; however, for instruments that have been converted or stock settled before the date of adoption, restatement of diluted earnings per share for all prior periods presented is required to conform to the consensus guidance. The above transition rules allow issuers to avoid restatement of diluted EPS if they redeem outstanding COCOs for cash prior to the adoption date. The transition provisions also allow issuers to modify the terms of the COCOs prior to the adoption date to reduce their dilutive effect. In particular, if the bond indentures are amended to require, upon conversion, cash settlement of par values and stock settlement of any excess over par (now referred to as “net shares” convertibles), then issuers would be able to apply the treasury stock method rather than the if-converted method of calculating diluted EPS to any outstanding COCOs, which would greatly reduce their effect on diluted EPS. The FASB ratified the EITF’s consensus decision in mid-October 2004. We use these events to separately examine 1) the debt restructuring decisions of managers and 2) the shareholder reactions to the change in financial reporting requirements. We present our hypotheses and empirical tests of the restructuring decision in Section 2 and follow with our predictions and analysis of shareholder wealth effects in Section 3.

2. Managerial responses to EITF 04-8 2.1 Hypothesis development – Restructurings and redemptions In this section we develop our hypotheses regarding managerial responses to the financial reporting changes for COCOs. Applying the if-converted method to COCOs regardless of whether market-based contingencies have been satisfied, as mandated by EITF 04-8, will decrease most COCO issuers’ diluted EPS figures for 2004 and thereafter, as well as for any prior periods in which COCOs were outstanding.7 However, as per the transition rules described above, issuers may avoid reporting a decrease in 7

In cases where the COCOs are anti-dilutive to EPS, or where the contingency has been met, the accounting change will have no effect on diluted EPS.

8

diluted EPS if they either redeem the COCOs for cash prior to the adoption date of the ruling, or if they amend the bond indentures to require cash settlement of par values and stock settlement of any excess over par at conversion. We consequently focus our analysis of managerial responses to the ruling on these two possible choices. Cash redemptions of COCOs are perhaps the most obvious solution to the financial reporting “problem” posed by the new accounting rule. Issuers could simply call the bonds and settle them for cash without issuing additional shares. Because the COCOs would no longer exist, they would not need to be included in diluted EPS for the current year; issuers would also avoid restatement of past periods. However, virtually all COCOs are call-protected for three to five years after issuance. Therefore, few COCO issuers would be able to take this course of action. In addition, in cases where call protection has expired (or will expire prior to the adoption date), this response is costly to implement, as it requires either the use of available cash or, if cash is not readily available, additional financing equal to the par value of the COCOs. Also, if it were optimal from a corporate financing perspective to redeem the COCOs for cash, the issuers should have already done so prior to EITF 04-8. These issues render it unlikely that many COCO issuers will opt for redemption in responding to the change in reporting requirements, though some may view the benefits of redemption as exceeding the costs. The transition rules of EITF 04-8 also allow issuers to avoid restatement and use of the if-converted method of calculating diluted EPS if they restructure their COCOs prior to the adoption date. Specifically, the bond indentures must be amended to require cash settlement of par values and stock settlement of any excess over par; issuers would then be able to apply the ‘treasury stock’ method of calculating diluted EPS to any outstanding COCOs.8 Because most COCOs are issued far out-of-the-money with high conversion premiums, treasury stock treatment will effectively exclude COCOs from diluted EPS until the conversion price is met, and even then will generally result in a much smaller impact on diluted EPS than under the if-converted method. 8

The impact of COCOs on diluted EPS under the treasury stock method depends on the company’s stock price. If the average stock price for the period is below the conversion price of the COCOs, then diluted EPS would be unaffected, as only COCOs that are “in-the-money” would have a dilutive effect on EPS. As the average stock price rises above the conversion price, the treasury stock method would require that shares be added to the denominator of diluted EPS, computed as the conversion value less the principal amount of the securities divided by the share price.

9

However, restructurings impose costs. In cases where the issuers may elect to settle in stock or cash, issuers must “irrevocably elect” to settle in cash, thereby relinquishing financial flexibility as they lose the option to settle the COCOs in stock.9 In cases where issuers had agreed to settle in stock, an exchange offer is necessary to amend the bonds to require cash settlement. In addition to any legal and underwriting fees associated with the transaction, the issuer also bears the costs of any “sweeteners” typically included in exchange offers to entice the security holders to complete the exchange. These sweeteners may include: cash payments (in the case of the COCO exchange offers, 25 basis points per $1000 face value bond is typical); additional put options; additional call protection; cash takeover protection; dividend protection; or a combination of the above. And, as in the above case of irrevocable elections, the issuer again loses financial flexibility in committing to cash settlement. The following is an example of a restructuring. On October 26, 2004, PPL Corp. announced that it was offering to pay $2.50 in cash for each $1,000 in principal amount of convertible notes to amend the provision of the original notes through an exchange offer, with the new securities requiring cash settlement of par values. Their news release stated: The accounting rule changes themselves do not change the underlying cash flows of PPL. However, given investors' valuation focus related to the diluted earnings per share calculation, PPL is seeking to modify the terms of the convertible notes to mitigate the dilutive earnings per share impact of these accounting rule changes. PPL believes that the proposed modifications are in the best interest of its investors, including the holders of the convertible notes. This suggests that this company was willing to pay $1 million to restructure the entire $400 million COCO issuance to avoid a decrease in reported diluted EPS (forecasted diluted EPS was predicted to decrease from $3.66 to $3.54). For firms with larger 9

Specifically, the FASB did not allow issuers to simply state an “intention” to settle conversions in cash rather than stock, as General Motors (GM) publicly announced it would on August 5, 2004. GM had $8 billion of COCOs outstanding and faced a possible decrease in its forecasted diluted EPS for 2004 of $1.00 from $7.00 if the proposed accounting change was ratified. The company announced that it would settle any conversions of its COCOs in cash rather than stock, thereby effectively nullifying the effects of the proposed change on its diluted EPS figure (Boudette, 2004). However, after the transition rules of EITF 048 were ratified, GM followed up its original announcement with an irrevocable election by the board directors on November 5, 2004 to settle par values in cash, thereby securing their right to apply the treasury stock method of calculating diluted EPS to their outstanding COCOs.

10

issuances, such as General Motors and Omnicom, the out-of-pocket costs would be significantly higher, and, of course, this cost estimate does not include any legal or underwriting costs, or the firm’s loss of financial flexibility. However, as PPL’s new release suggests, some firms clearly believe that the benefits of modifying the bond indentures is worth the associated cost.10 We examine managerial decisions to either redeem for cash or to restructure the COCOs as “net shares” convertibles using multivariate probit analysis. While both responses mitigate the financial reporting effects of EITF 04-8, they do not bear the same explicit costs, and redemptions could be undertaken for reasons other than the rule change.11 We therefore conduct our empirical analysis of managerial responses to the rule change by first examining only restructurings and then adding redemptions to the analysis.12 We predict that firms will be more likely to incur restructuring or redemption costs when the perceived financial reporting benefits provided by COCOs are greater, when the personal benefits to managers are higher, and when the relative costs are smaller.

We

consequently

expect

the

following

factors

to

relate

to

the

restructuring/redemption choice: 1. Financial Reporting Effects. We predict that managers are more likely to undertake costly restructurings or redemptions of COCOs when the financial reporting effects are relatively large, as perceived benefits are assumed to be increasing with the reporting effect. We measure financial reporting effects using two variables. The first is IMPACT, defined as the difference between firms’ First Call annual diluted EPS estimate for 2004 minus the comparable figure that would result if all outstanding COCOs were immediately included in diluted EPS, divided by share price.13 If the COCOs are anti-

10

Consistent with PPL’s reasoning, prior empirical research shows that stock prices reflect dilution of EPS (see Core, Guay, and Kothari, 2002; Huson, Scott, and Wier, 2001; or Jennings, LeClere, and Thompson, 1997). However, the negative restructuring announcement returns that we document later in the paper suggests that investors view these actions as value-decreasing, not increasing, as PPL’s press release suggests. 11 Brennan and Schwartz (1977) present conditions under which it is optimal to call convertible bonds. 12 Unfortunately, the small number of redemptions (six) prevents us from carrying out a separate analysis of this response. 13 We have forecasts of 2004 annual diluted EPS as of September 30, 2004. We divide the forecast by the share price as of February 18, 2004, that is, the share price one week before our first regulatory event occurs.

11

dilutive, then IMPACT equals zero; IMPACT is thus always a non-negative amount. When the impact on diluted EPS is large, firms have greater incentive to maintain any financial reporting advantages gained from issuing the COCOs in the first place. We consequently expect a positive association between the likelihood of restructuring or redemption and IMPACT. The perceived financial reporting benefit provided by COCOs also depends on whether the securities’ conversion threshold has been reached. Once the issuing firm’s stock price exceeds the conversion threshold, the financial reporting advantages vanish because firms must now apply the if-converted method of calculating diluted EPS to the COCOs. We consequently predict that firms with stock prices in excess of the conversion threshold are less likely to restructure their COCOs. We create an indicator variable, TRIGMET, that equals one if the stock price threshold for conversion of the COCOs had been met as of July 1, 2004, and zero otherwise.14 We expect a negative association between TRIGMET and the likelihood of restructuring or redemption. 2. Bonus Contracts. We also expect compensation contracts to affect the restructuring or redemption decision. Given that Marquardt and Wiedman (2005) document that managers who are compensated on EPS are more likely to issue COCOs rather than traditional convertible bonds, we expect that these managers will be also eager to maintain at least some of the reporting advantages associated with COCOs by amending the conversion terms of the security.15 We create an indicator variable, BONUS, that equals one if EPS is explicitly mentioned in firms’ 2003 proxy statements as one of the determinants of annual cash bonuses for the CEO and zero otherwise.16 We predict a positive association between the likelihood of restructuring and the BONUS variable.

14

Reliable data on price thresholds is available only as of June 2004. In cases where sample firms have more than one COCO issue outstanding, we compare the average conversion trigger to the firm’s stock price in determining TRIGMET. 15 It is certainly possible that compensation committees could simply adjust for the rule change when setting bonus payments. To the extent that managers believe this is likely to occur, this effect will bias us against finding a positive relationship between BONUS and the restructuring decision. 16 We obtain this information from the “Report of the Compensation Committee” within the proxy statement. Ideally we would like to be able to determine whether the bonus plan relies on diluted EPS, as opposed to basic EPS, but the compensation reports do not provide this level of detail. Our assumption is that the term “EPS” refers to diluted EPS, as this is the figure that equity analysts forecast and investors use in valuation.

12

3. Lobbying Efforts. We also consider firms’ lobbying efforts against EITF 04-8 as a determinant of the restructuring or redemption decision. Prior research on corporate lobbying suggests that firms most affected by an accounting change are more likely to lobby against it. For example, Dechow, Sloan, and Sweeney (1996) find that the likelihood of submitting a comment letter opposing mandatory expensing of employee stock-based compensation is strongly related to the use of stock options in top-executive compensation, and Deakin (1989) finds that lobbying behavior around a series of changes to oil and gas accounting rule changes in the 1970’s is based on the expected economic effects of the changes on the firm or its management. Firms that expect to incur greater costs if EITF 04-8 is approved should be more likely to lobby against its ratification. We consequently predict a positive association between lobbying efforts against EITF 04-8 and the likelihood of COCO restructuring. We define an indicator variable, COMMENT, that equals one if the firm submitted a comment letter to the FASB opposing the ratification of EITF 04-8, and zero otherwise. We predict a positive association between COMMENT and the likelihood of restructuring. 4. Ability to bear costs of restructuring. We predict that the ability to bear the costs of COCO restructuring will also affect firms’ decisions to undertake these transactions. Because firms relinquish a degree of financial flexibility when they commit themselves to cash rather than stock settlement of par values, we predict that the likelihood of COCO restructuring or redemption is positively associated with free cash flows (FCF), and negatively associated with leverage (LEV), as firms with high free cash flows and low leverage are better able to give up this flexibility in financial planning. We define FCF as operating cash flows minus capital expenditures, divided by total assets, and LEV as total liabilities divided by total assets. Both are measured as of the end of fiscal 2003. In addition, exchange offers generally include a “sweetener” to entice the security holders to complete the exchange. In general, we expect more profitable firms to be better able to provide these inducements and therefore predict a positive association between the likelihood of COCO restructuring and firms’ return on assets (ROA), where ROA is defined as net income divided by beginning period total assets for fiscal 2003.

13

5. Expiration of Call Protection. In our analysis that includes the redemption decision, we control for the ability of the firm to redeem the COCOs for cash by including an indicator variable, CALLEXP, that equals one if the COCO security’s call protection period has expired prior to the adoption date of EITF 04-8, and zero otherwise. We expect CALLEXP to be positively associated with the likelihood of redemption. 6. Firm Size. Lastly, we include firm size as a control variable in our analysis of firms’ restructuring decisions. The costs of restructuring may be relatively smaller for large firms, as underwriting costs and other fees associated with the restructuring typically decrease as a percentage of the total dollar value of the offering. Therefore predict a positive association between firm size (SIZE) and the likelihood of COCO restructuring, where SIZE is defined as the log of total assets at the end of fiscal 2003. To summarize, we expect the likelihood of COCO restructurings and redemptions to be positively associated with IMPACT, BONUS, COMMENT, FCF, ROA, CALLEXP, and SIZE and negatively associated with TRIGMET and LEV. We present two models, as follows: Pr( RESTRUCTURE ) = β 0 + β1 IMPACT + β 2TRIGMET + β 3 BONUS + β 4 COMMENT + β5 ROA + β 6 FCF + β 7 LEV + β8 SIZE + ε

(1)

Pr( RESTRUCTURE / REDEEM ) = β 0 + β1 IMPACT + β 2TRIGMET + β 3 BONUS + β 4 COMMENT + β5 ROA + β 6 FCF + β 7 LEV + β8CALLEXP + β 9 SIZE + ε .

In equation 1, we define our dependent variable, RESTRUCTURE, as an indicator variable that equals one if the firm restructured at least one of its outstanding COCOs between July 1, 2004 and the adoption date of EITF 04-8, and zero otherwise. In equation 2, we define RESTRUCTURE/REDEEM as an indicator variable that equals one if the firm restructured or redeemed the COCOs for cash between July 1, 2004 and the adoption date of EITF 04-8, and zero otherwise. We obtained information regarding firms’ restructuring/redemption activities from press release data available from their corporate websites. If there was no press release announcing an exchange offer, an irrevocable election to settle COCO par values in cash, or a cash redemption of the COCOs, we then examined 8-K and registration statements for each firm through the SEC’s Edgar website 14

(2)

to ensure that the dependent variables are properly coded.17 All other variables are as defined above.

2.2 Sample Selection and Descriptive Statistics We obtained our convertible securities data from Kynex, Inc., a risk-management firm that maintains an extensive database of active convertible securities in the U.S. market. Our original sample included 234 firms that issued COCOs from the time of the initial Tyco offering in November 2000 through February 15, 2004, which is one week before the first event date included in our analysis of the shareholder wealth effects associated with the rule change. Of these firms, 207 were also covered in a Bear Stearns report that analyzed the impact of the COCOs on 2004 forecasted EPS and from which we obtain data to estimate our IMPACT and TRIGMET variables. Four firms were eliminated because Compustat company identifiers were not available, and 4 additional firms did not have price / returns data available on CRSP. Our final sample includes 199 firms. Table 1 presents descriptive statistics for the 199 COCO offerings in our sample. The mean (median) COCO offering is $372.4 ($201.0) million, representing 13.7% (9.7%) of total assets. COCOs have long maturities, with a mean (median) of 21.6 (20.0) years, and are typically call-protected for five years. The majority of offerings (91%) offer at least one put, on average 4.9 years from the issue date. Further, 85% offer a second put, on average 8.6 years from the issue date. Consistent with the relatively low interest rates over our sample period, mean (median) coupon rates are 2.4% (2.5%) and yields are 3.0% (2.9%). The mean (median) conversion premium of 39.5% (36.3%) is quite high relative to that of traditional convertibles, but is consistent with the addition of the COCO feature.18 Further, the contingency feature of the convertible security is 17

To the extent that firms respond to EITF 04-8 through means other than exchange offers or irrevocable elections to settle par values in cash rather than stock, there is potential for measurement error in our RESTRUCTURE variable. For example, firms may initiate or accelerate stock repurchase programs to offset the dilutive effects of EITF 04-8. A preliminary review revealed no evidence that such responses were taken by sample firms, perhaps because such action would impact basic as well as diluted EPS. In addition, any misclassification error will bias our tests against finding significant differences across the two groups. 18 Prior to EITF 04-8, a high conversion premium ensured that the dilutive effects of the COCOs would be excluded from diluted EPS over a longer time period, as the stock price would have to rise even higher before conversion conditions were satisfied.

15

expressed as a fixed percentage of the conversion price for 197 of the 199 sample firms, with the percentage ranging from ranging from 110 to 145% of the conversion price, but set to 120% for the majority of firms. In only two cases (Interpublic and Omnicom) did the trigger increase over time. Finally, 83% of the COCOs are 144a private placements. We present the industry membership of sample firms in Table 2. Our sample of 199 COCO issuers is well-distributed across 35 of 48 possible industry categories, as defined by Fama and French (1997). For comparison purposes, we also provide the distribution of all 2003 Compustat firms; in general, the distribution of COCO issuers appears representative of the greater Compustat population. This broad industry representation suggests that the potential problem of cross-sectionally correlated residuals, which is a particular concern in our analysis of the shareholder wealth effects, is less of a problem in our sample. In addition, the generalizability of our results is likely to be enhanced by wide industry membership.

2.3 Results – Restructurings and redemptions EITF 04-8 presented transition rules that allowed firms to redeem for cash or restructure their COCOs to qualify for preferred accounting treatment. Forty firms in our sample modified the terms of at least one outstanding COCO between July 1, 2004 and the adoption date of EITF 04-8. Further, twelve firms in the sample had COCOs outstanding where the call protection on the bonds had expired, and six of these twelve firms redeemed their debt for cash prior to the adoption date. While the restructurings are clearly linked to EITF 04-8, the rationale behind the cash redemptions is somewhat less clear. In light of this, we conduct two separate analyses. In the first, we exclude the redemptions and compare the 40 RESTRUCTURE firms with the 153 firms that did not restructure or redeem their bonds. In the second analysis, we include the redemptions with the restructurings and compare these 46 RESTRUCTURE/REDEEM firms to the 153 NON-RESTRUCTURE firms.19

19

To more rigorously examine alternative motivations for early redemption, it would be necessary to identify a control sample of non-contingent convertible securities, specify a prediction model of early redemption, and compare the probability of redemption across the two groups. As noted earlier, the very small number of redemptions prevents our undertaking a more thorough analysis.

16

Table 3, Panel A presents the results from univariate tests of differences in our variables of interest. We find that the RESTRUCTURE firms face a greater potential IMPACT on diluted EPS than non-RESTRUCTURE firms (p-values of 0.0815 and 0.0231 for the t-test and Wilcoxon tests, respectively) and that the stock prices of RESTRUCTURE firms are, on average, less likely to have met the COCOs’ conversion threshold (p-values of 0.0399 and 0.1100, for the t-test and Wilcoxon test on TRIGMET, respectively). Most notable, however, are the differences in BONUS, COMMENT, and SIZE. EPS is explicitly mentioned as a determinant of CEO annual cash bonuses for 57.5 percent of the RESTRUCTURE firms but only 34.0 percent on the non-RESTRUCTURE firms (p = 0.0064). Further, 27.5 percent of the RESTRUCTURE firms submitted comment letters to the FASB opposing EITF Issue No. 04-8 while only 3.3 percent of non-RESTRUCTURE firms did (p = 0.0019). SIZE is also significantly higher for RESTRUCTURE firms (p = 0.0016). Differences in ROA, FCF, and LEV do not attain statistical significance in the univariate tests. As shown in Table 3, Panel B, inferences from our univariate tests are unchanged when we include the firms that redeemed their COCOs for cash with the RESTRUCTURE firms. Multivariate tests of our empirical predictions are presented in Table 4. This table reports the estimated coefficients and significance levels from a multivariate probit analysis of the dependent variable RESTRUCTURE on the nine variables hypothesized to affect the restructuring decision. In Model 1, the six firms who redeemed their debt are excluded. In Models 2 and 3, these firms are included with the RESTRUCTURE firms. In general, our findings are consistent with our predictions. In Model 1, we find that the decision to restructure the COCO prior to the adoption date is positively related to IMPACT (p=0.0124), BONUS (p=0.0409), COMMENT (p=0.0040), and SIZE (p=0.0394) and negatively related to LEV (p=0.0409). As reflected in its very low pvalue, the variable COMMENT appears to have the greatest relative effect on the probability of restructuring – i.e., if the firm sent a comment letter on EITF 04-8 to the FASB, the predicted probability of restructuring increases by 1.1079. TRIGMET, ROA, and FCF are not significant at conventional levels, though we note that TRIGMET would

17

reach marginal significance in a one-tailed test. The pseudo-R2 for the model is 19.8 per cent.20 In Models 2 and 3, we include the six firms who redeemed their COCOs. Results are very similar to Model 1, except that the negative coefficient for TRIGMET becomes marginally significant at p=0.0866 and p=0.0775 for Models 2 and 3, respectively.21 Further, CALLEXP is significantly positive in Model 3 (p=0.0003) as predicted, since only those firms whose call protection had expired were able to redeem their bonds.22 CALLEXP also has the greatest relative effect on the predicted probability of restructuring, increasing it by 1.598 for those firms with expired call protection; COMMENT also retains its large effect on the dependent variable in Models 2 and 3, increasing the predicted probability of restructuring by 1.035 and 1.181, respectively. In summary, our results suggest that firms with greater incentives to maintain the perceived financial reporting benefits originally provided by COCOs, as reflected in IMPACT and BONUS, were also the ones more likely to bear the costs of restructuring these securities. Further, results for LEV provide evidence that firms with lower leverage were better able to bear the cost of reduced financial flexibility resulting from these restructurings. Our findings of a positive association between the decision to restructure and firm size are consistent with restructurings being less costly for large firms. These results also build on Marquardt and Wiedman (2005), who find that both IMPACT and BONUS were significant determinants of the original decision to issue 20

To assess potential multicollinearity for our multivariate analysis, we compute condition numbers of the matrices of explanatory variables (Greene 2003). Condition numbers between 30 and 100 indicate moderate to strong dependencies among the variables (Judge et al. 1985). The highest condition number for the matrix of explanatory variables in all three models is less than 18.0, indicating that multicollinearity is not a problem. 21 As previously indicated, for firms with multiple offerings, TRIGMET is calculated using the average conversion trigger for all of the COCOs outstanding (typically two) compared to firm’s stock price. For all but five firms in the sample with multiple offerings, either all of the COCOs had hit the trigger or none had. We assess the sensitivity of our results to two alternative definitions. First, we exclude five firms from the analysis where at least one COCO had hit the trigger and at least one had not. In the second set of tests we redefine TRIGMET to equal one if stock price threshold for conversion had been met for all of the outstanding COCOs. In both cases, results for all three models remain similar to those reported in Table 4, although the TRIGMET variable becomes somewhat less significant. 22 As an additional sensitivity test, we also exclude the firm 3M from the sample. This company did not restructure its COCO, but also did not comply with EITF 04-8. The company argued in its 2004 10-K that “due to the FASB’s delay in issuing SFAS No. 128R and the Company’s intent and ability to settle this debt security in cash versus the issuance of stock, the impact of additional diluted shares will not be included in the diluted earnings per share calculation until SFAS No. 128R is effective.” Results are very similar for all three models when 3M is excluded.

18

COCOs rather than traditional convertible securities. However, Marquardt and Wiedman (2005) also conclude that firms experience only minimal costs when issuing COCOs, while, in contrast, our analysis of COCO restructurings reveals that issuers incur significant costs in order to maintain the financial reporting benefits associated with COCOs. This finding complements recent survey evidence by Graham, Harvey, and Rajgopal (2005), who report that “managers are willing to sacrifice economic value to manage financial reporting perceptions” but also note that “it is difficult for empirical archival research to convincingly document such behavior” (p. 6). Further, to confirm that investors viewed COCO restructurings as costly, we examine shareholder returns around announcements of restructurings and report significantly negative market-adjusted returns in a three-day window around announcement dates. Mean (median) abnormal returns are –0.75% (-0.97%) with a pvalue of 0.0468 (0.0139). These findings of significantly negative shareholder reactions around restructuring announcement dates further suggest that investors viewed the costs of restructuring as exceeding any potential benefits, thereby representing an agency cost to shareholders. We examine the shareholder wealth effects associated with rule change more fully in the following section.

3. Shareholder wealth effects 3.1 Mean shareholder wealth effects – Hypothesis development Holthausen and Leftwich (1983) view accounting choices as having economic consequences if they affect “the wealth of parties who use those numbers for contracting or decision making.” Therefore, to complete our exploration of the economic consequences associated with the financial reporting changes for COCOs, we examine the investor reaction to events relating to the new accounting requirements.23 In particular, because our analysis of managerial responses in the previous section suggests that the rule change led to increased agency costs for COCO issuers, we are especially interested in determining whether investors were able to anticipate these costs when the rule changes were initially announced.

23

Ideally, we would also examine the bondholder wealth effects associated with EITF 04-8 but are unable because the overwhelming majority of COCOs are 144a private placements.

19

Table 5 provides the date for each regulatory pronouncement and summarizes the event. Events D1 and D2 relate to FASB Staff Position 129-a, which mandated footnote disclosure of COCO-related information, and events R1 through R4 relate to EITF 04-8 and the recognition of COCOs in diluted EPS.

3.1.1 Disclosure events We first explore the investor reaction to the release of FASB Staff Position 129-a and 129-1, which mandated disclosure of COCO-related information in firms’ financial statements. Because FSP 129 was initiated in response to FASB constituents’ concerns over poor disclosure, we expect that its release will increase the quality and amount of information available about firms’ outstanding COCOs. Specifically, the required disclosure should provide enough information to permit financial statement users to apply the if-converted method of calculating diluted EPS. If, as equity analysts argue, that investors were previously unable to perform this calculation, and, as shown empirically by Marquardt and Wiedman (2005), that managers’ primary motivation in issuing COCOs was to manage diluted EPS upwards, then the new disclosure requirements should render any past or future earnings management attempt through this means transparent to financial statement users. This increased transparency regarding earnings management could affect stock prices either positively or negatively. On the one hand, mandating full disclosure of COCO-related information could serve a corporate governance role, as managers may now be less likely to issue COCOs purely for earnings management motives. This suggests that improved disclosure of COCO-related information would be value increasing and that shareholders would react positively to news of the mandated disclosure requirements. Lo (2003) presents evidence consistent with this governance improvement hypothesis in his examination of the 1992 revision of executive compensation proxy disclosures. Alternatively, there may be costs associated with increased disclosure that are value decreasing. Grossman (1981) and Milgrom (1981) show that a full disclosure equilibrium is possible only in the absence of disclosure costs. The marked variability that Gainey (2004) documents in COCO-related disclosure practices prior to 2004

20

indicates that a full disclosure equilibrium was clearly not in place, consistent with the presence of such costs. For example, if investors were not aware that COCOs were excluded from diluted EPS calculations, or were uncertain about the extent of their impact on diluted EPS, then disclosures that reveal this information may be costly to COCO issuers. Also, prior to the mandated disclosure changes, COCO issuers had already determined their optimal disclosure level regarding COCO-related information. Requiring greater disclosure will result in suboptimal levels for at least some firms. This suggests that stock prices will react negatively to regulations requiring greater disclosure of COCO-related information. Because mandated disclosure has the potential to improve corporate governance through increased transparency or to impose additional disclosure costs on COCO issuers, we make no directional predictions regarding the stock price reaction around disclosure events D1 and D2.

3.1.2 Recognition events We next examine shareholder reactions to pronouncements related to the recognition of COCOs in diluted EPS calculations. We argue that shareholder reaction to these events will be related to the potential responses of COCO issuers. Firms with outstanding COCOs on July 1, 2004 will weigh the costs and benefits of each possible response and act accordingly. Possible responses include: (1) doing nothing; (2) redeeming the COCOs for cash; or (3) restructuring COCOs as “net shares” convertibles. Each response requires the firm to incur costs, which we describe below. It would seem that the response with the lowest costs is the first: do nothing, restate diluted EPS from prior periods, and apply the if-converted method of calculating diluted EPS going forward. While this response may require COCO issuers to bear contracting costs if debt or compensation agreements are predicated on diluted EPS levels, it appears unlikely that debt contracts rely on diluted EPS figures (see Begley and Freedman, 2004); Marquardt and Wiedman (2005), however, report that COCO issuers are more likely to compensate their CEOs based on EPS measures. In addition, while COCO issuers may suffer stock price declines associated with the rule change to the extent that investors functionally fixate on reported EPS measures, the increased

21

disclosure requirements that preceded events R1-R4 should moderate this effect. Issuers may also communicate the reasons for decreases in diluted EPS at relatively low cost (e.g., Omnicom addressed this issue in its second quarter 2004 investor presentation). It thus appears that response 1 is a reasonable choice for COCO issuers. However, managers may take a different view of EITF 04-8, particularly if their compensation depends on reported EPS figures. As we describe in the previous section, redeeming or restructuring the COCOs allows managers to mitigate the reporting effects of the new rule on diluted EPS, but at significant costs to shareholders. To the extent that investors can anticipate the agency costs associated with redeeming or restructuring COCOs in response to the rule changes, the reaction to events R1-R4 should be negative. Weber (2004) also makes agency cost arguments to explain shareholder reaction to SEC Staff Accounting Bulletin 96, which required managers to discontinue share repurchase programs in order to continue pooling-of-interest deals in progress. We expect shareholders to react negatively, on average, around recognition events R1 through R4. Their negative reaction around these events may be mitigated by the potential benefit of the rule change, namely, greater transparency in financial reporting. However, because recognition events R1-R4 are preceded by events D1 and D2, which mandated increased disclosure requirements for COCOs, we expect any benefit from greater transparency to be incorporated into stock prices around D1 and D2. We therefore predict negative mean shareholder reactions around recognition events R1-R4.

3.2 Cross-sectional predictions We expect shareholder reactions to events D1-D2 and R1-R4 to vary crosssectionally across COCO issuers. We examine seven variables, which are discussed and defined below. 1. Disclosure Quality. We consider the quality of COCO issuers’ pre-existing disclosure practices for COCOs in determining shareholder reactions to events D1 and D2. If these events increase disclosure costs for COCO issuers, we would expect shareholder reaction to be more negative for firms with poor COCO-related disclosure practices. In other words, firms that already provide a great deal of detailed information relating to their outstanding COCOs should not be penalized by investors around a

22

pronouncement mandating adequate disclosure, as these firms would not be affected by the new requirement. If, on the other hand, greater disclosure improves governance, then we would expect firms with poor COCO-related disclosure practices to benefit more than firms that already provide adequate disclosure, and the shareholder reactions should be more positive for this group of firms. In sum, we expect the shareholder reaction to be more pronounced for the group of firms with poor COCO-related disclosure, relative to other firms. We create an indicator variable, DISCLOSE, that equals one if a COCO issuer did not provide enough information in the footnotes of its 10-K filing from the year of issue to allow users to apply the if-converted method of calculating diluted EPS and zero otherwise. Application of the if-converted method requires information necessary to calculate the after-tax interest charges that need to be added back in the numerator, information necessary to calculate the number of shares issuable upon conversion that need to be added to the denominator, and information regarding the conversion conditions of the security. We examine firms’ EPS and long-term debt footnotes for these disclosures. If any required piece of information is not disclosed in either footnote, DISCLOSE equals one. A negative association between the shareholder response to events D1 and D2 and the variable DISCLOSE is consistent with FSP 129-1 increasing disclosure costs for COCO issuers, while a positive association is consistent with governance arguments. To the extent that shareholders can anticipate agency costs, variables found to be significant managerial incentives in the restructuring/redemption decision in the previous section, including IMPACT, TRIGMET, and BONUS may help explain cross-sectional variation in shareholder reaction to the recognition events R1 through R4. We also consider other factors relevant to the restructuring decision, including LEV, CALLEXP and SIZE. We do not include the variable COMMENT in this analysis, as the comment period falls in the middle of our event period and therefore information on comment letter submissions was not available to investors at the earlier event dates. 2. Financial Reporting Benefits. As outlined in the previous section, firms are more likely to undertake costly action to offset the financial reporting effects of EITF 048 when these effects are large.

23

We employ two measures of financial reporting benefits. The first is an ex ante measure of the COCOs’ expected impact on diluted EPS (IMPACT), as previously defined. If the COCOs are anti-dilutive, then IMPACT equals zero; IMPACT is thus always a non-negative amount. We predict a negative association between IMPACT and the shareholder reactions to events R1-R4. Also, to the extent that disclosure was viewed by investors as a step towards recognition, IMPACT may also be negatively related to shareholder reaction to event D1. The financial reporting benefit provided by COCOs also depends on the degree to which firms’ COCOs are out-of-the-money (TRIGGER). Here we measure TRIGGER as the stock price conversion threshold for the COCO (i.e., the price at which holders may voluntarily convert the securities into stock) divided by the firm stock price. We use a continuous variable to measure the degree of moneyness here rather than simply an indicator variable as this better captures the differential impact of the rule changes for COCO firms, particularly when changes were first proposed. TRIGGER also serves as a proxy for investors’ expectations of TRIGMET. We expect TRIGGER will be negatively associated with shareholder responses to events R1-R4. 3. Bonus Contracts. Marquardt and Wiedman (2005) document that COCO issuers are more likely to base CEO annual bonuses on EPS performance metrics than are the issuers of traditional convertible securities. Further, our analysis of restructurings in the previous section indicates that firms are more likely to undertake costly restructurings of COCOs if their bonus contracts include EPS as a performance metric. We predict a negative association between BONUS and shareholder reactions around events R1-R4. 4. Leverage. Higher leverage firms are less likely to have the financial flexibility to pursue some of the alternatives available to COCO issuers, including restructuring and redeeming debt. Therefore, we predict a positive association between LEV and shareholder reactions around events R1-R4. 5. Expiration of Call Protection. As discussed in the previous section, the second possible response of COCO issuers to the change in diluted EPS calculations is to redeem the COCOs for cash. However, few COCO issuers are able to pursue this course of action because COCOs are typically call-protected for a five year period after issuance. We argue that shareholder reaction around events R1-R4 may vary cross-sectionally with the

24

availability of this response. Because cash redemption of the COCOs is costly, it is possible that shareholders might react negatively if call protection has expired. On the other hand, cash redemption might be viewed by investors as the most desirable of the possible issuer responses. We therefore make no prediction regarding the association between CALLEXP and shareholder reactions around R1-R4. 6. Firm Size. Shareholder reactions to events R1-R4 may vary cross-sectionally with firm size.24 Our results from Table 4 suggest that large firms are more likely to incur costs to restructure their COCOs. To the extent that investors can anticipate firm response based on size, we would expect a negative association between SIZE and returns. However, there is evidence that market anomalies such as post-earnings announcement drift are concentrated in small firms (see Bernard and Thomas, 1990), suggesting that any functional fixation by investors may be less likely for large firms. Also, it may be less costly for large firms to communicate the reasons for the change in diluted EPS calculations to analysts and investors than it is for small firms. Given these differing predictions, including the SIZE variable may provide us with some ability to discriminate between these competing hypotheses. Accordingly, we make no prediction on the sign of the association between this variable and stockholder reactions around events R1-R4. To summarize, we expect a more pronounced (either positively or negatively) relationship between DISCLOSE and shareholder reactions around events D1-D2; we expect a negative relationship between IMPACT, TRIGGER, and BONUS and shareholder reactions around events R1-R4; we expect a positive relationship between LEV and shareholder reactions around events R1-R4; and we make no prediction regarding the relationship between CALLEXP and SIZE and shareholder reactions around events R1-R4. Lastly, we recognize that because changes in accounting rules are often preceded by increases in disclosure requirements (notable examples include post-retirement benefits and, more recently, employee stock options), investors may reasonably view events D1 and D2 as a step toward eventual recognition of COCOs in diluted EPS 24

The positive accounting literature often argues that accounting changes that decrease reported earnings reduce political costs associated with regulatory pressures. This suggests that large firms, which are more sensitive to political costs, should experience muted shareholder reactions to such events. However, it is questionable whether regulators would focus on diluted EPS rather than, say, net income as a measure of firm profitability.

25

calculations. To determine whether this is the case, we include IMPACT and TRIGGER, which most directly capture the financial reporting benefits of COCOs, in our crosssectional tests of D1 and D2. We interpret a negative association between the shareholder reaction around events D1-D2 and IMPACT or TRIGGER as consistent with investors viewing these events as precursors to the eventual rule change.

3.3 Research design – Shareholder wealth effects 3.3.1 Mean effects Following standard event study methodology, we use a version of the Schipper and Thompson (1983) model to determine whether stock prices react significantly to events relating to the disclosure and recognition of COCOs in diluted EPS. We estimate the following regression for our sample of 199 COCO issuers: K

R% pt = α p + β p R% mt + ∑ γ pk Dkt + ε% pt

(3)

k =1

where R% pt is the return on portfolio p on day t; R% mt is the return on the CRSP valueweighted market return on day t; α p and β p are market model parameter estimates; Dkt is an indicator variable that equals one during the five-day period beginning the day before and ending three days after the announcement of event k and zero otherwise; γ pk is the estimated mean effect of event k for portfolio p; and ε% pt is an error term that is assumed to be independently and identically normally distributed. We use five-day windows in our tests because the events we examine were not widely covered in the financial press; it consequently may take longer for news of the regulatory changes to disseminate to investors than if the changes had been reported on immediately. For example, a search of ProQuest and Lexis-Nexis revealed no coverage of disclosure events D1 and D2, and the Wall Street Journal did not report on recognition event R1 until July 6th, the fourth day of our event window; a second story on COCOs appeared in the Wall Street Journal on July 8th, which falls outside the five-day window. We thus believe that a shorter event window has the potential to miss the investor

26

reaction we hope to capture.25 Collins, Rozeff, and Dhaliwal (1981) make a similar design choice, using one-week and two-week windows to examine shareholder wealth effects around event dates related to SFAS No. 19. We use two portfolio weighting schemes. In the first, we equally-weight the daily returns of each COCO issuer; in the second, we weight the returns using the diagonal elements of the covariance matrix of residuals from a first-stage firm-specific market model regression to mitigate potential cross-sectional heteroskedasticity. Day t runs from February 25, 2003, which is one year prior to our first event date, through December 31, 2004.26 We assess whether the disclosure or recognition of COCOs affects stock prices using t-tests of the estimated mean effects γ pk for events D1 and D2 and events R1 through R4, respectively. We note that t-statistics from the Schipper and Thompson (1983) procedure tend to be lower than from simple t-tests of abnormal returns in a particular event window because the latter method does not adequately control for the effects of multiple announcement events for a given policy change or for the high crosssectional correlation in stock return residuals due to each announcement event occurring on the same calendar date for all affected firms.

3.3.2 Cross-sectional effects To examine the cross-sectional determinants of event period reactions, we follow the three-step portfolio weighting procedure suggested by Sefcik and Thompson (1986) and summarized by Espahbodi et al. (2002). First, we form a matrix F that has a column of ones and P-1 columns of firm characteristics. Because P varies across the disclosure

25

In fact, using three-day windows beginning the day before each event yield results that are qualitatively similar but have reduced significance levels. These findings suggest that market participants needed a relatively longer period of time to learn of the rule changes and/or to interpret their effects on COCO issuers. However, to rule out the possibility that our results are driven by the longer return window, we perform a randomization test in which we randomly choose six dates during our event period, create threeday, five-day, or seven-day windows around each date, and run the time-series regression in equation 1. We repeat this procedure 100 times. Using an absolute t-statistic cutoff of 1.96, we expect the null hypothesis of no abnormal return to rejected approximately 5% of the time. Using the 600 estimated coefficients resulting from this analysis, we reject the null at rates of 6.2%, 5.2%, and 4.5% for the three-, five-, and seven-day windows, respectively. In addition, a test of the cumulative returns across all six dates is rejected at a 5% rate for all three window lengths. We thus conclude that our choice of a five-day event window will not bias our results. 26 As a sensitivity test, we also used a four-year and one-year event period; results are invariant to the length of the event period.

27

and recognition events, we estimate two separate models. For the disclosure event study, we examine the effect of three firm characteristics (DISCLOSE, IMPACT, and

TRIGGER); therefore, P = 4 in the disclosure model. In the recognition event study, we have predictions for six firm characteristics; therefore, P = 7 in the recognition model. We define the matrix F as follows:

F = 1 X 2 X 3 K X p 

(4)

where X p is an N × 1 vector of the pth firm characteristic and N equals the number of sample firms. Next, we create P sets of portfolio weights, as follows:

 W1′ W ′  2 W =   = ( F ′F ) −1 F ′  M   ′ W p 

(5)

and compute P separate time-series of portfolio returns for each set, as follows: R% pt = W p′Rit , p = 1,..., P, t = 1,..., T , i = 1,..., N

(6)

where W = P × N matrix of portfolio weights (P = 4 or 7 and N = 199 firms); Wp′ = pth row of portfolio weights;

F = N × P matrix as defined in equation 2; R% pt = return on portfolio p on day t; and Rit = N × 1 vector of individual firms’ stock returns on day t.

In the third and final step, we estimate P time-series OLS regressions of the form presented in equation 3, where t again runs from February 25, 2003 through December 31, 2004. For disclosure events D1 and D2, portfolios 2 through 4 correspond to the three firm characteristics for which we have made cross-sectional predictions; for recognition events R1 through R4, portfolios 2 through 7 correspond to the six firm characteristics we examine. The estimated mean effects γ pk reflect the effect of each firm characteristic on the stock market reaction to each event. As noted by Espahbodi et al. (2002), these estimates are the same as those in cross-section regressions of abnormal returns on firm characteristics.

However, the standard errors of the estimates in the Sefcik and

28

Thompson (1986) procedure account fully for the cross-correlation and cross-sectional heteroskedasticity in firm residuals. This is important, as even a fairly small amount of cross-correlation in data can lead to serious misspecification of test statistics in event studies. More particularly, the test will overreject the null hypothesis because the standard deviation of the cross-sectional distribution of returns will be downwardly biased. We test our cross-sectional predictions using t-tests on the estimated mean effects γ pk , as above.

3.4 Results – Shareholder wealth effects 3.4.1. Mean effects

We document mean shareholder wealth effects around our six regulatory event dates in Table 6. Using the Schipper and Thompson (1983) methodology, we report two portfolio weighting schemes: equally-weighted and variance-weighted. The latter method helps in mitigating potential cross-sectional heteroskedasticity. Our six events are represented as dummy variables, with the mean coefficients reported representing the mean excess return around each date. Events D1 and D2 relate to changes in disclosure requirements for COCOs (Panel A); events R1 through R4 relate to changes in the recognition of COCOs in the computation of EPS (Panel B). In Panel A, we present evidence of marginally negative abnormal portfolio returns around event D2, which corresponds to the final posting of FSP 129-1 to the FASB website on April 9, 2004, increasing disclosure requirements for COCO issuers. The mean daily return is -0.0026 for the equally-weighted portfolio and -0.0021 for the variance-weighted portfolio; both are significant at the 0.10 level. Because we use fiveday estimation windows, this indicates that the total abnormal portfolio returns around this date average -1.05 to -1.30 percent. These findings around event D2 are consistent with the argument that increased disclosure requirements are costly to firms, but also with shareholders’ interpreting event D2 as an intermediary step toward eventual recognition of COCOs in diluted EPS. In addition, we find no evidence of significant shareholder wealth effects around event D1, when the initial draft of FSP 129 was made public. In Panel B, we report significantly negative abnormal portfolio returns around event R1, which corresponds to the July 1, 2004 EITF tentative decision to require

29

inclusion of COCOs in diluted EPS. The mean daily effect is -0.0036 for the equallyweighted portfolio and -0.0023 for the variance-weighted portfolio (both significant at the 0.05 level), translating into total shareholder wealth effects of -1.15 to -1.80 percent in the five-day window around event R1.27 We also find significantly negative abnormal portfolio returns using the variance-weighted portfolio method around event R2, when the draft of EITF 04-8 was posted to the FASB website on July 19, 2004 for comment; portfolio returns for a combined window around all four recognition events are also significantly negative at the 0.10 and 0.05 levels for the equally-weighted portfolio and variance-weighted portfolio, respectively. The significance of events R1 and R2 is consistent with shareholder anticipation of costly responses to EITF 04-8 by COCO issuers. We also examine mean daily portfolio returns for a combined window including all six regulatory events (results not tabulated). The estimated coefficients of -0.0012 and -0.0011 are significantly negative at the 0.10 and 0.05 levels for the equally-weighted and variance-weighted portfolios, respectively. This translates into total shareholder wealth effects of -3.6% and -3.3% (the estimated coefficient multiplied by 30 days) across all six event dates. The strength of our results is comparable to that in other event studies employing the Schipper and Thompson (1983) methodology (see, e.g., Espahbodi et al., 2002 or D’Souza, 2000), though the regulatory events we examine affect only diluted EPS and not reported net income. Overall, we conclude that the regulatory changes relating to financial reporting for COCOs that occurred during 2004 have an overall negative effect on shareholder wealth.28

27

If we expand the window around R1 to include July 8th, when the Wall Street Journal published a second article on the EITF’s proposed change, total abnormal portfolio returns range from -1.47 to -2.16 percent and t-statistics are more highly significant for both the equally- and variance-weighted portfolios (-2.58 and -2.45, respectively). 28 To explore whether other events might affect our results, we examined quarterly earnings announcement dates during 2004 for our sample firms, as well as restructuring announcement dates. We find that 13, 3, 1, 64, 1, and 11 firms announced quarterly earnings within the five-day windows around D1, D2, R1, R2, R3, and R4, respectively. Of particular concern is event R2, where almost a third of the sample announced earnings and where we also find a significantly negative mean shareholder reaction. To determine whether these announcements are driving our results, we eliminate the 64 announcing firms from the portfolio and repeat our analysis. The t-statistic for the mean effect around R2 increases in its significance level when these firms are dropped (-2.49 versus the -2.21 reported in Table 3). For the restructuring announcements, one firm announced a restructuring within the R2 window, and two firms announced within the R4 window. Dropping these three firms from the analysis yields results that are virtually unchanged. We thus

30

3.4.2. Cross-sectional effects

We predict that the shareholder wealth effects around the six regulatory events will vary cross-sectionally with firm characteristics. Descriptive statistics for these characteristics are reported in Table 7, Panel A. All explanatory variables are winsorized at the 1 and 99 percent levels to mitigate the influence of outliers. The mean for the variable DISCLOSE is 0.2010, indicating that roughly 20 per cent of COCO issuers did not disclose enough information in their 10-K filing to allow users to apply the if-converted method of calculating diluted EPS to the outstanding COCOs. The mean (median) for the variable IMPACT, measured as the reduction in diluted EPS divided by price, is 0.0033 (0.0021). When expressed as a percentage of forecasted EPS, the mean (median) expected decrease to diluted EPS is 5.0 per cent (4.3 per cent) with a maximum (unwinsorized) decrease of a 25 percent. The mean (median) value for TRIGGER is 1.59 (1.43). Since these values are greater than 1.0, they indicate that the majority of firms had not hit their trigger as of the measurement date; only 12.1 percent of the sample had trigger values below 1.0. The reported mean for BONUS indicates that for 39.7 percent of sample firms, EPS was explicitly mentioned as a determinant of annual cash bonuses for the CEO. Mean (median) LEV is 0.6388 (0.6261). The mean CALLEXP is 0.06 indicating that the call protection on the COCOs had expired for only 12 of the 199 firms. Mean (median) SIZE, expressed as the log of total assets, is 8.02 (7.88). Mean (median) total assets (untabulated) is $18.1 billion ($2.6 billion). Pearson correlation coefficients for the explanatory variables are provided in Panel B of Table 7. SIZE is significantly correlated with both DISCLOSE, BONUS and LEV (ρ = 0.2101, 0.2302 and 0.4023 respectively), indicating that large firms are more

likely to engage in poor COCO-related disclosure practices, are more likely to use EPS as a performance metric in their CEO annual bonus contracts, and also have relatively more debt. SIZE is also significantly correlated with CALLEXP (ρ = 0.1580). BONUS is also significantly positively correlated with DISCLOSE (ρ = 0.1825), suggesting that firms that reward CEOs on EPS performance are less likely to provide adequate disclosure of

conclude that any confounding effects from earnings or restructuring announcements are unlikely to affect our inferences.

31

COCO-related information. In addition, IMPACT and TRIGGER are significantly negatively correlated (ρ = -0.2726), and IMPACT and LEV are significantly positively correlated (ρ = 0.1904). We report results from our cross-sectional tests of shareholder wealth effects in Table 8. We use the method suggested by Sefcik and Thompson (1986), as outlined previously, to control for cross-sectional correlation in errors, a problem that can create inference problems because of the alignment of events in calendar time (Bernard, 1987). In Panel A, we present results around disclosure events D1 and D2; in Panel B, we present results around recognition events R1 through R4. Each column reports the mean effect of each firm characteristic on abnormal portfolio returns.29 The first event for which there is evidence of significant cross-sectional variation in shareholder reaction is event D1, which corresponds to the date on which FSP 129-a was originally drafted and posted to the FASB website. While we did not find evidence of a significant shareholder reaction around event D1 for the entire portfolio of 199 COCO issuers (as shown in Table 6), we find in Panel A of Table 8 a significantly negative shareholder reaction around event D1 for the group of COCO issuers with poor COCO-related disclosure practices. This finding is consistent with shareholders anticipating increased disclosure costs for these firms. We do not, however, find evidence that the shareholder response around this event varies systematically with either IMPACT or TRIGGER, which suggests that shareholders did not at this point view the increased disclosure requirements of FSP 129 as a step toward recognition of COCOs in diluted EPS. This does not appear to be the case around the second disclosure event, D2, where IMPACT is significantly negatively related to abnormal portfolio returns. This finding is

consistent with shareholders viewing this event as a step toward eventual recognition and with their expectation that greater agency costs would be incurred by those firms where diluted EPS is most impacted by a change in the accounting rules for COCOs. The

29

We perform White’s (1980) test of heteroskedasticity for each column in Table 8. In every case, we cannot reject the null hypothesis of homoscedastic error terms at conventional levels of significance.

32

DISCLOSE variable is not significantly associated with shareholder effects around this

date.30 Results around the recognition events are presented in Panel B of Table 8. Consistent with our prediction, we find that abnormal portfolio returns are significantly more negative around event R1 and marginally more negative around event R2 for firms with COCOs that are far from reaching the conversion threshold (TRIGGER). However, we do not find evidence that our IMPACT variable is significantly associated with stock prices around this date, perhaps because shareholders had already responded to this information around the April 9th (D2) event. The expiration of call protection (CALLEXP) is significantly positively associated with shareholder returns, suggesting that investors view a firm’s ability to redeem the COCOs for cash as less costly than other possible responses to the rule change. We also find that BONUS is marginally negatively related to the returns around event R1. Lastly, we report a significantly negative shareholder effect around event R4, when the FASB ratified EITF 04-8 and provided guidance on transition rules, for firms with higher IMPACT values. The findings for IMPACT, TRIGGER and BONUS are consistent with agency cost arguments, namely that investors recognized that firms with the greatest financial reporting impact and personal incentives were most likely to incur costs to preserve their financial reporting benefits. These results also parallel the findings in Table 4 which show that these same variables significantly influence the decision to restructure COCOs. The SIZE variable, which offered some possibility of discriminating between functional fixation and agency cost arguments, is not significantly different from zero. In sum, our results collectively suggest that agency costs are behind the negative shareholder reaction to the financial reporting changes associated with COCOs.

4. Concluding Remarks

This paper examines two dimensions of the economic consequences of changes in regulation in financial reporting for contingent convertible securities. First, using a sample of 199 firms issuing COCOs from November 2000 through February 2004, we

30

As a sensitivity test, we also included the other cross-sectional variables from Panel B in our disclosure model. None were significantly associated with shareholder reactions around events D1 and D2.

33

examine issuer responses to the regulatory change. Out of our sample of 199 COCO issuers, 46 chose to restructure or redeem for cash their outstanding COCOs prior to the adoption date to qualify for preferred accounting treatment. Using multivariate probit analysis, we examine the determinants of firms’ decision to restructure their outstanding COCOs and find that the likelihood of costly restructuring is significantly positively associated with the financial reporting impact of EITF 04-8 on diluted EPS, the use of EPS as a performance metric in CEO bonus contracts, corporate lobbying efforts against the proposed change, and firm size. Restructurings are also negatively related to firms’ overall leverage. These findings suggest that managers trade off restructuring costs against the benefits of reporting a higher diluted EPS figure. Second, we examine shareholder wealth effects around six regulatory event dates during 2004 that relate to the financial reporting of these instruments. Overall, we document a negative shareholder reaction of approximately three percent of share value. Because the rule change we examine affects only diluted EPS, this finding suggests that investors recognize the importance of this financial measure to managers. In addition, cross-sectional tests are consistent with agency cost arguments. One limitation of the study is that we do not empirically examine all possible issuer responses to the new rule but instead confine ourselves to the two explicitly discussed in the transition rules for EITF 04-8 – cash redemptions and restructurings. However, as we previously note, misclassifying a firm that, say, decides to repurchase common stock as a response to EITF 04-9 as a “non-RESTRUCTURE” firm biases us against finding results in this analysis. Another caveat is that we classify six firms that redeemed their COCOs for cash as a “RESTRUCTURE” firm when it is possible that these firms decided to redeem their convertible bonds because it was optimal for them to do so from a financing perspective; i.e., EITF 04-8 may have had nothing to do with the redemption decision. To address this issue, we eliminate these firms from the restructuring analysis, and our results are robust to this adjustment. A second limitation is that we cannot conclusively rule out the effects of investor fixation on reported earnings in our shareholder reaction tests. For example, a few variables (namely, IMPACT and TRIGGER) in our cross-sectional tests could be as interpreted as consistent with both functional fixation and agency arguments. However,

34

the significance of the BONUS variable as a determinant of both the restructuring decision and shareholder reactions to financial reporting events provides some evidence in support of agency arguments. Thus, while the preponderance of our evidence suggests that anticipated agency costs explain the significantly negative reaction to announcements related to financial reporting changes for COCOs, we cannot completely rule out this alternative explanation. Despite these limitations, the paper contributes to the accounting literature in several ways. First, our results linking the likelihood of COCO restructurings to the impact of the new ruling on diluted EPS demonstrate that firms are willing to expend firm resources to procure financing reporting benefits, thereby adding to recent findings in the financial reporting cost literature. We also contribute to the literature on shareholder wealth effects associated with financial reporting changes. Unlike the vast majority of papers in this literature, the reporting change we examine does not affect net income but rather a single figure within the financial statements – diluted EPS. Our results demonstrate that investors recognize the relevance of diluted EPS to managers, a finding that reinforces the importance of the FASB’s current deliberations over appropriate measurement of EPS.

35

References

Amir, E., & Livnat, J. (1996). Multiperiod analysis of adoption motives: The case of SFAS No. 106. The Accounting Review, 71, 539-553. Ayers, B., Lefanowicz, C. & Robinson, J. (2002). Do firms purchase the pooling method? Review of Accounting Studies, 7, 5-32. Begley, J. & Freedman, R. (2004). The changing role of accounting numbers in public lending agreements. Accounting Horizons, 18, 81-96. Bens, D. & Monahan, S. (2005). Altering investment decisions to manage financial reporting outcomes: Asset backed commercial paper conduits and FIN 46. Working paper, University of Arizona. Bernard, V. (1987). Cross-sectional dependence and problems in inference in marketbased accounting research. Journal of Accounting Research, 25, 1-48. Bernard, V. & Thomas, J. (1990). Evidence that stock prices do not fully reflect the implications of current earnings for future earnings. Journal of Accounting and Economics, 13, 305-340. Boudette, N. E. (2004). Moving the market: GM to pay holders of ‘Co-Co’ bonds to buffer its EPS. Wall Street Journal, August 6, C3. Brennan, M. & Schwartz, E. (1977). Convertible bonds: Valuation and optimal strategies for call and conversion. Journal of Finance, 33, 1699-1715. Collins, D., Rozeff, M. & Dhaliwal, D. (1981). The economic determinants of the market reaction to proposed mandatory accounting changes in the oil and gas industry: A cross-sectional analysis. Journal of Accounting and Economics, 3, 37-81. Core, J., Guay, W. & Kothari, S.P. (2002). The economic dilution of employee stock options: Diluted EPS for valuation and financial reporting. The Accounting Review, 77, 627-652. Deakin, E.B. (1989). Rational economic behavior and lobbying on accounting issues: Evidence from the oil and gas industry. The Accounting Review, 64, 137-151. Dechow, P., Sloan, R. & Sweeney, A. (1996). Economic consequences of accounting for stock-based compensation. Journal of Accounting Research, 34 (Supplement), 120. D’Souza, J. (2000). The stock price impact of mandated accounting charges on rateregulated firms. Review of Accounting Studies, 5, 232-257.

36

Engel, E., Erickson, M. & Maydew, E. (1999). Debt-equity hybrid securities. Journal of Accounting Research, 37, 249-274. Erickson, M., Hanlon, M. & Maydew, E. (2004). How much will firms pay for earnings that do not exist? Evidence of taxes paid on allegedly fraudulent earnings. The Accounting Review, 79, 387-408. Espahbodi, H., Espahbodi, P., Rezaee, Z. & Tehranian, H. (2002). Stock price reaction and value relevance of recognition versus disclosure: The case of stock-based compensation. Journal of Accounting and Economics, 33, 343-373. Fama, E., & French, K. (1997). Industry costs of equity. Journal of Financial Economics, 43, 153-193. Gainey, P. (2004). They’re contingently dilutive: Risk of EPS dilution from CoCos. Valuation & Accounting, Citigroup Smith Barney Equity Research (January 23). Graham, J., Harvey, C. & Rajgopal, S. (2005). The economic implications of corporate financial reporting. Journal of Accounting and Economics, 40, 3-73. Greene, W., ed. (2003). Econometric Analysis. MacMillan: New York. Grossman, S. (1981). The informational role of warranties and private disclosure about product quality. Journal of Law and Economics, 24, 461-483. Guidry, F., Leone, A. & Rock, S. (1999). Earnings-based bonus plans and earnings management by business-unit managers. Journal of Accounting and Economics, 26, 113-142. Healy, P. (1985). The effect of bonus schemes on accounting decisions. Journal of Accounting and Economics, 7, 85-107. Henry, D. (2003). The latest magic in corporate finance: How contingent convertible bonds blindside shareholders. Business Week, September 8, p. 88. Holthausen, R., Larcker, D. & Sloan, R. (1995). Annual bonus schemes and the manipulation of earnings. Journal of Accounting and Economics, 19, 29-74. Holthausen, R., & Leftwich, R. (1983). The economic consequences of accounting choice: Implications of costly contracting and monitoring. Journal of Accounting and Economics, 5, 77-117. Huson, M., Scott, T. & Wier, H. (2001). Earnings dilution and the explanatory power of earnings for returns. The Accounting Review, 76, 589-612.

37

Imhoff, E., & Thomas, J. (1988). Economic consequences of accounting standards: The lease disclosure rule change. Journal of Accounting and Economics, 10, 277-310. Jennings, R., LeClere, M. & Thompson, R. (1997). Evidence on the usefulness of alternative earnings per share measures. Financial Analyst Journal, Nov/Dec, 2433. Judge, G., Griffiths, W., Carter Hill, R., Luetkepohl, H., Lee, T., ed. 1985. Regression Diagnostics Wiley: New York. Marquardt, C., & Wiedman, C. (2005). Earnings management through transaction structuring: Contingent convertible debt and diluted EPS. Journal of Accounting Research, 43, 205-243. Matsunaga, S., Shevlin, T. & Shores, D. (1992). Disqualifying dispositions of incentive stock options: Tax benefits versus financial reporting costs. Journal of Accounting Research, 30 (Supplement), 37-68. McConnell, P., Pegg, J., Senyet, C. & Mott, D. (2004). Convertible bonds II: FASB plays catch-up with market innovations. Accounting Issues, Bear Stearns Equity Research, April. Milgrom, P. (1981). Good news and bad news: Representation theorems and applications. Bell Journal of Economics, 17, 18-32. Mittelstaedt, H.F., Nichols, W. & Regier, P. (1995). SFAS No. 106 and benefit reductions in employer-sponsored retiree health care plans. The Accounting Review, 70, 535-556. Schipper, K., & Thompson, R. (1983). The impact of merger related regulations on the shareholders of acquiring firms. Journal of Accounting Research, 21, 184-221. Sefcik, S., & Thompson, R. (1986). An approach to statistical inference in cross-sectional models with security abnormal returns as dependent variable. Journal of Accounting Research, 24, 316-334. Venkatachalam, M., Rajgopal, S. & Choudhary, P. (2005). Accelerated vesting of employee stock options in anticipation of FAS 123-R. Working paper, Duke University. Weber, J. (2004). Shareholder wealth effects of pooling-of-interests accounting: Evidence from the SEC’s restriction on share repurchases following pooling transactions. Journal of Accounting and Economics, 37, 39-57. White, H. (1980). A heteroskedasticity consistent covariance matrix estimator and a direct test of heteroskedasticity. Econometrica, 48, 817-838.

38

Exhibit 1. The effect of contingent conversion on diluted EPS for Omnicom Group Inc. During 2001 – 2003, Omnicom Group Inc. issued three series of COCOs for aggregate proceeds of $2.3 billion. All three series of COCO’s were zero-coupon, zero-yield bonds with thirty-year maturities. Because the contingencies of these COCOs were not satisfied by the end of 2003, the dilutive effects of the combined offerings were excluded from reported diluted EPS for that year. Had they been included, diluted EPS would have decreased nine percent, from $3.59 to $3.26. Under the “if converted” method, after-tax interest of the convertible securities is added to the numerator (zero in this case as Omnicom’s securities were zero-yield), and the common shares issuable upon conversion are added to the denominator, as indicated in the calculation below.

Numerator Denominator (millions of $) (millions of shares) 2003 EPS

Reported Diluted EPS

Adjusted Diluted EPS (estimated under the ‘if-converted’ method) Feb-01 COCOs ($847.0 million principal) Mar-02 COCOs ($892.3 million principal) Jun-03 COCOs ($600.0 million principal) – 6 of 12 months

39

$677.0

188.7

$0.0 $0.0 $0.0

7.7 8.1 2.9

$677.0

207.4

$3.59

$3.26

Table 1. Descriptive statistics for COCOs

Variable

N

Mean

Std. Dev

1st Q

Median

Proceeds ($MM) Proceeds (as a percentage of total assets) Bond maturity (years) Call protection (years) First Put (years) Second Put (years) Bond coupon (%) Bond yield (%) Conversion premium (%) Trigger (% of conversion price) 144a private placements (1/0)

199

$372.4

$475.2

$135.0

$201.0

$450.0

199 199 199 181 170 199 199 199 197 199

13.7% 21.6 4.9 4.9 8.6 2.4% 3.0% 39.5% 119.2% 0.83

13.2% 6.6 1.3 1.9 2.9 1.8% 1.7% 15.6% 5.8% 0.37

3.5% 20.0 5.0 4.0 7.0 0.6% 1.8% 30.0% 120.0% 1

9.7% 20.0 5.0 5.0 10.0 2.5% 2.9% 36.3% 120.0% 1

18.8% 30.0 5.0 5.5 10.0 3.8% 4.0% 45.0% 120.0% 1

The sample includes 199 COCOs issued from November 2000 through mid-February 2004. Data were provided by Kynex Inc.

40

3rd Q

Table 2. The distribution of 199 sample COCO issuers by industry.

Industry Business Services Transportation Insurance Pharmaceuticals Banking Retail Trading Computers Electronic Equipment Healthcare Telecommunications Oil and Gas Utilities Medical Equipment Restaurants & Hotels Construction Machinery Electrical Equipment Autos and Trucks Steel Business Supplies Wholesale Recreational Products Consumer Goods Construction Materials Food Products Defense Coal Lab Equipment Chemicals Real Estate Entertainment Precious Metals Nonmetallic Mining Miscellaneous Other Total

Number of firms

% of Sample

% of Compustat Firms

23 16 16 15 10 10 10 10 8 7 7 6 6 5 5 4 4 4 4 3 3 3 2 2 2 2 2 2 2 1 1 1 1 1 1 0 199

11.56% 8.04 8.04 7.54 5.03 5.03 5.03 5.03 4.02 3.52 3.52 3.02 3.02 2.51 2.51 2.01 2.01 2.01 2.01 1.51 1.51 1.51 1.01 1.01 1.01 1.01 1.01 1.01 1.01 0.49 0.49 0.49 0.49 0.49 0.49 0.00 100.00%

13.01% 3.30 2.14 5.75 11.19 3.92 5.02 3.77 5.54 1.37 3.47 3.24 4.26 2.82 1.65 0.83 2.69 1.15 1.32 0.95 0.84 2.91 0.62 1.33 1.32 1.12 0.11 0.13 1.78 1.64 0.85 1.48 0.90 0.61 1.57 5.40 100.00%

The sample includes 199 COCOs issued from November 2000 through mid-February 2004. Data were provided by Kynex Inc. Individual industries are as defined by Fama and French (1997), based on Compustat data for fiscal 2003. “Other” industries include: Agriculture; Candy and Soda; Alcoholic Beverages; Tobacco Products; Printing and Publishing; Apparel; Rubber and Plastic Products; Textiles; Fabricated Products; Aircraft; Shipbuilding; Personal Services; and Shipping Containers.

41

Table 3. Comparison of firm characteristics of firms that restructured (and redeemed) outstanding COCOs between July 1, 2004 and the adoption date of EITF 04-8 with firms that did not. Panel A. Univariate differences between 40 RESTRUCTURE firms (redemptions excluded) and 153 non-RESTRUCTURE firms Variable

Mean

Median

Std. Dev.

RESTRUCTURE firms (N=40)

Mean

Median

Std. Dev.

t-test

non-RESTRUCTURE firms (N=153)

Wilcoxon p-value

IMPACT

0.0046

0.0030

0.0050

0.0031

0.0019

0.0039

0.0815

0.0231

TRIGMET

0.0500

0.0000

0.2207

0.1438

0.0000

0.3520

0.0399

0.1100

BONUS

0.5750

1.0000

0.5006

0.3399

0.0000

0.4752

0.0064

0.0068

COMMENT

0.2750

0.0000

0.4522

0.0327

0.0000

0.1784

0.0019

0.0001

ROA

0.0390

0.0346

0.0295

0.0234

0.0274

0.0858

0.0638

0.3236

FCF

0.0415

0.0415

0.0466

0.0453

0.0405

0.0648

0.6745

0.7638

LEV SIZE

0.6471 8.6900

0.6272 8.3638

0.1938 1.8343

0.6390 7.8063

0.6261 7.6746

0.2154 1.4680

0.8301 0.0016

0.7578 0.0065

Panel B. Univariate differences between 46 RESTRUCTURE firms (redemptions included) and 153 non-RESTRUCTURE firms RESTRUCTURE / REDEEM firms (N=46)

non-RESTRUCTURE / REDEEM firms (N=153)

p-value

IMPACT

0.0042

0.0028

0.0047

0.0031

0.0019

0.0039

0.0956

0.0403

TRIGMET

0.0435

0.0000

0.2062

0.1438

0.0000

0.3520

0.0174

0.0681

BONUS

0.5870

1.0000

0.4978

0.3399

0.0000

0.4752

0.0025

0.0028

COMMENT

0.2609

0.0000

0.4440

0.0327

0.0000

0.1784

0.0013

0.0001

ROA

0.0407

0.0351

0.0308

0.0234

0.0274

0.0858

0.0383

0.2234

FCF

0.0466

0.0429

0.0469

0.0453

0.0405

0.0648

0.8804

0.7815

LEV

0.6380

0.6178

0.1918

0.6390

0.6261

0.2154

0.9778

0.9302

CALLEXP

0.1957

0.0000

0.4011

0.0196

0.0000

0.1391

0.0052

0.0001

SIZE

8.7401

8.3673

1.7687

7.8063

7.6746

1.4680

0.0004

0.0017

42

RESTRUCTURE is an indicator variable that equals one if the company restructured at least one of its COCOs between July 1, 2004 and the adoption date of EITF Issue No. 04-8, and zero otherwise. Six firms redeemed their debt prior to the adoption date; we exclude them from the first analysis and include them with the restructure firms in the second. IMPACT is measured as the difference between the firms’ First Call annual diluted EPS estimate for 2004 minus the comparable figure that would result if all outstanding COCOs were immediately included in diluted EPS, divided by share price. TRIGMET is an indicator variable that equals one if the stock price threshold for the COCO (i.e., the price at which bondholders may voluntarily convert the bonds into stock) had been met as of July 1, 2004, and zero otherwise. BONUS is an indicator variable that equals one if EPS is explicitly mentioned as one of the determinants of annual cash bonuses for the CEO and zero otherwise. COMMENT is an indicator variable that equals one if the firm wrote a comment letter to FASB regarding EITF Issue No. 04-8, and zero otherwise. ROA is return on assets for the firm. FCF is free cash flows, defined as operating cash flows minus capital expenditures. LEV is total liabilities divided by total assets. CALLEXP is an indicator variable that equals one if the COCO security call protection period had expired prior to the adoption date of EITF Issue No. 04-8. SIZE is measured as the log of total assets. All variables are winsorized at 1% and 99%. Significance values are all based on two-tailed tests.

43

Table 4. Probit regression results for models of RESTRUCTURE choice for the sample of 199 COCO issuances Pr (RESTRUCTURE) = β0 + β1IMPACT + β2TRIGMET + β3BONUS + β4COMMENT + β5ROA+ β6FCF + β7LEV + β8SIZE Pr (RESTRUCTURE / REDEEM) = β0 + β1IMPACT + β2TRIGMET + β3BONUS + β4COMMENT + β5ROA+ β6FCF + β7LEV + β8CALLEXP + β9SIZE

Variable

Predicted sign

Constant term

Model 1

Model 2

Model 3

?

-1.9617 (0.0050)

-2.1730 (0.0014)

-1.8755 (0.0083)

IMPACT

+

71.8681 (0.0124)

69.7466 (0.0158)

75.9484 (0.0100)

TRIGMET

-

-0.6767 (0.1237)

-0.7473 (0.0866)

-0.8764 (0.0775)

BONUS

+

0.4824 (0.0409)

0.4863 (0.0308)

0.4888 (0.0384)

COMMENT

+

1.1079 (0.0040)

1.0353 (0.0060)

1.181 (0.0023)

ROA

+

2.5915 (0.2912)

2.4863 (0.2934)

3.2611 (0.2154)

FCF

+

-2.7536 (0.2902)

-1.7186 (0.4832)

-3.4167 (0.2087)

LEV

-

-1.4946 (0.0409)

-1.7277 (0.0167)

-1.5323 (0.0406)

CALLEXP

+

SIZE

+

N Log likelihood Pseudo R2

1.598 (0.0003) 0.1948 (0.0394)

0.2448 (0.0082)

0.1797 (0.0639)

193 -79.37 19.8%

199 -86.38 20.1%

199 -79.17 27.5%

RESTRUCTURE is an indicator variable that equals one if the company restructured at least one of its COCOs between July 1, 2004 and the adoption date of EITF Issue No. 04-8, and zero otherwise. In Model 1, the 6 firms that redeemed their COCOs for cash prior to the adoption date are excluded. In Models 2 and 3, these 6 firms are included with the restructure firms. IMPACT is measured as the difference between the

44

firms’ First Call annual diluted EPS estimate for 2004 minus the comparable figure that would result if all outstanding COCOs were immediately included in diluted EPS, divided by share price. TRIGMET is an indicator variable that equals one if the stock price threshold for the COCO (i.e., the price at which bondholders may voluntarily convert the bonds into stock) had been met as of July 1, 2004, and zero otherwise. BONUS is an indicator variable that equals one if EPS is explicitly mentioned as one of the determinants of annual cash bonuses for the CEO and zero otherwise. COMMENT is an indicator variable that equals one if the firm wrote a comment letter to FASB regarding EITF Issue No. 04-8, and zero otherwise. ROA is return on assets for the firm. FCF is free cash flows, defined as operating cash flows minus capital expenditures, divided by total assets. LEV is total liabilities divided by total assets. CALLEXP is an indicator variable that equals one if the COCO security call protection period had expired prior to the adoption date of EITF Issue No. 04-8. SIZE is measured as the log of total assets. All variables are winsorized at 1% and 99%. Significance values are all based on two-tailed tests.

45

Table 5. Regulatory pronouncements relating to the financial reporting of COCOs reported on www.fasb.org.

Event number Event date

Description

Expected shareholder reaction

Panel A: Disclosure events

D1

February 25, 2004

FASB releases Staff Position 129-a, “Disclosure Requirements under FASB Statement No. 129, Disclosure of Information about Capital Structure, Relating to Contingently Convertible Securities” for comment.

?

D2

April 9, 2004

FASB posts to its website the final draft of FAS 129-1, which is effective immediately.

?

Panel B: Recognition events

R1

July 1, 2004

EITF announces a tentative decision at its public meeting that COCOs should be included in diluted earnings per share computations, regardless of whether market-based contingencies have been met.

Negative

R2

July 19, 2004

FASB releases draft abstract for EITF Issue No. 04-8, “The Effect of Contingently Convertible Debt on Diluted Earnings per Share” for comment.

Negative

R3

September 30, 2004 EITF affirms its July 1 decision at its public meeting.

Negative

R4

October 13, 2004

Negative

FASB ratifies EITF 04-8, effective for fiscal years ending after December 15, 2004, and provides guidance on transition rules.

46

Table 6. Mean excess returns, calculated using a market model and five-day (-1, +3) event windows for the period 2/25/2003 through 12/31/2004, for event dates related to the financial reporting of COCOs for 199 COCO issuers.

Equally-weighted portfolio

Event number

Expected market reaction

Mean daily return

Variance-weighted portfolio

t-statistic

Mean daily return

t-statistic

Panel A: Disclosure Events

D1

?

0.0021

1.21

0.0014

1.28

D2

?

-0.0026

-1.71*

-0.0021

-1.88*

Panel B: Recognition Events

R1

Negative

-0.0036

-2.35**

-0.0023

-2.17**

R2

Negative

-0.0025

-1.61

-0.0024

-2.21**

R3

Negative

0.0005

0.31

0.0001

0.07

R4

Negative

-0.0010

-0.65

-0.0011

-0.99

Significant at the 10% (*), 5% (**), 1% (***) levels, based on two-tailed tests. Event numbers are described in Table 1.

47

Table 7. Descriptive statistics for selected variables of 199 COCO issuers.

Variable

DISCLOSE

IMPACT

TRIGGER

BONUS

LEV

CALLEXP

SIZE

0.3970 0.0000 0.4905

0.6388 0.6261 0.2097

0.0603 0.0000 0.2386

8.0222 7.8759 1.5878

Panel A: Descriptive statistics for firm characteristics Mean Median Std. Dev.

0.2010 0.0000 0.4018

0.0033 0.0021 0.0042

1.5860 1.4340 0.6885

Panel B: Pearson correlation coefficients between firm characteristics IMPACT

0.0049 (0.9456)

TRIGGER

-0.0337 (0.6364)

-0.2726 (0.0001)

BONUS

0.1825 (0.0099)

0.0280 (0.6943)

-0.0365 (0.6090)

LEV

0.0112 (0.8753)

0.1904 (0.0071)

-0.1217 (0.0868)

0.0983 (0.2020)

CALLEXP

0.0310 (0.6641)

-0.0705 (0.3224)

-0.0197 (0.7824)

0.0965 (0.1752)

-0.0461 (0.5177)

SIZE

0.2101 (0.0029)

-0.0998 (0.1607)

0.0079 (0.9118)

0.2302 (0.0011)

0.4023 (0.0001)

48

0.1580 (0.0258)

DISCLOSE is an indicator variable that equals one if the firm did not disclose sufficient information to apply the if-converted method of calculating diluted EPS, and zero otherwise IMPACT is measured as the difference between the firms’ First Call annual diluted EPS estimate for 2004 minus the comparable figure that would result if all outstanding COCOs were immediately included in diluted EPS, divided by share price. TRIGGER equals the stock price threshold for the COCO (i.e., the price at which bondholders may voluntarily convert the bonds into stock) divided by the firm share price. BONUS equals one if EPS is explicitly mentioned as one of the determinants of annual cash bonuses for the CEO and zero otherwise. LEV is total liabilities divided by total assets. CALLEXP is an indicator variable that equals one if the COCO security call protection period had expired prior to the adoption date of EITF Issue No. 04-8. SIZE is measured as the log of total assets. P-values are presented in parentheses.

49

Table 8. Estimated coefficients and t-statistics (based on Sefcik and Thompson 1986) from a cross-sectional analysis of event parameters, using a market model and five-day (-1,+3) event windows for the period 2/25/2003 through 12/31/2004, for event dates relating to the financial reporting of COCOs for 199 COCO issuers Event Constant number term DISCLOSE IMPACT TRIGGER BONUS LEV CALLEXP SIZE Panel A: Disclosure Events

D1

0.0038 (1.42)

-0.0037** (-2.03)

0.3058 (1.51)

-0.0014 (-0.95)

D2

-0.0006 (-0.23)

0.0003 (0.17)

-0.4587** (-2.26)

-0.0008 (-0.54)

Panel B: Recognition Events

R1

-0.0065 (-1.01)

0.0956 (0.48)

-0.0029** (-2.34)

-0.0027* (-1.86)

0.0027 (0.52)

0.0049** (2.43)

0.0008 (1.15)

R2

-0.0051 (-0.79)

0.1037 (0.52)

-0.0021* (-1.74)

0.0017 (1.23)

0.0013 (0.25)

0.0004 (0.21)

-0.0003 (-0.43)

R3

0.0046 (0.73)

0.0889 (0.44)

0.0013 (1.09)

-0.0008 (-0.59)

-0.0048 (-0.90)

-0.0020 (-1.02)

-0.0005 (-0.75)

R4

0.0066 (1.04)

-0.4390** (-2.19)

0.0001 (0.01)

0.0018 (1.28)

-0.0003 (-0.40)

0.0002 (0.11)

-0.0007 (-0.94)

Event numbers are described in Table 1. Significant at the 10% (*) or 5% (**) levels, based on two-tailed tests. DISCLOSE is an indicator variable that equals one if the firm did not disclose sufficient information to apply the if-converted method of calculating diluted EPS, and zero otherwise IMPACT is measured as the difference between the firms’ First Call annual diluted EPS estimate for 2004 minus the comparable figure that would result if all outstanding COCOs were immediately included in diluted EPS, divided by share price. TRIGGER equals the stock price threshold for the

50

COCO (i.e., the price at which bondholders may voluntarily convert the bonds into stock) divided by the firm share price. BONUS equals one if EPS is explicitly mentioned as one of the determinants of annual cash bonuses for the CEO and zero otherwise. LEV is total liabilities divided by total assets. CALLEXP is an indicator variable that equals one if the COCO security call protection period had expired prior to the adoption date of EITF Issue No. 04-8. SIZE is measured as the log of total assets. Significance levels are based on two-tailed significance levels.

51