Discretionary Behavior with Respect to the Adoption of Sfas 142 and the Behavior of Security Prices

Transcription

1 Sngapore Management Unversty Insttutonal Knowledge at Sngapore Management Unversty Research Collecton School Of Accountancy School of Accountancy Dscretonary Behavor wth Respect to the Adopton of Sfas 142 and the Behavor of Securty Prces Yoonseok ZANG Sngapore Management Unversty, Follow ths and addtonal works at: Part of the Accountng Commons, Fnance and Fnancal Management Commons, and the Portfolo and Securty Analyss Commons Ctaton ZANG, Yoonseok. Dscretonary Behavor wth Respect to the Adopton of Sfas 142 and the Behavor of Securty Prces. (2008). Revew of Accountng and Fnance., 7(1), Research Collecton School Of Accountancy. Avalable at: Ths Journal Artcle s brought to you for free and open access by the School of Accountancy at Insttutonal Knowledge at Sngapore Management Unversty. It has been accepted for ncluson n Research Collecton School Of Accountancy by an authorzed admnstrator of Insttutonal Knowledge at Sngapore Management Unversty. For more nformaton, please emal lbir@smu.edu.sg.

2 Dscretonary Behavor wth Respect to the Adopton of SFAS No. 142 and the Behavor of Securty Prces Yoonseok Zang Assstant Professor School of Accountancy Sngapore Management Unversty 60 Stamford Road Sngapore Emal: Phone: Fax: November 2006 Ths study s adapted from my Ph.D. dssertaton at UC Berkeley. I would lke to thank my dssertaton char, Brett Trueman, for hs support and gudance durng the study. The study has also benefted from the comments of Sunl Dutta, Stephen Penman, Paul Ruud, Jacob Sag, Theodore Souganns, Xao-Jun Zhang, two anonymous revewers, and workshop partcpants at UC Berkeley, UC Rversde, the Unversty of Illnos at Urbana-Champagn, and Sngapore Management Unversty. The generous fnancal support of the Wharton-SMU Research Center and the Haas School Ph.D. Program at UC Berkeley s gratefully acknowledged.

3 Abstract Ths study examnes two sets of ssues n connecton wth SFAS No. 142, Goodwll and Other Intangble Assets. The frst s whether managers use dscreton n determnng the transtonal goodwll mparment loss (ntal mparment loss or IIL) n a strategc manner. The second s whether and how the market reacts to a dsclosed mparment loss and to the absence of goodwll amortzaton. I fnd that more hghly leveraged frms have a lower goodwll mparment charge, whch s consstent wth the strategc reducton of the goodwll mparment by management to avod the volaton of debt covenants. I also fnd that frms that have undergone recent management changes report greater mparment charges, whch s reflectve of the noton that new managers take a bg bath so that they can report hgher earnngs n the future. Regardng the market response, evdence shows that the stock return s negatvely assocated wth an unexpected IIL, wth the assocaton beng stronger for hghly leveraged frms, and that subsequent to the IIL announcement, analysts revse ther forecasts of the upcomng quarters earnngs downward n response to an unantcpated IIL. These fndngs are nterpreted as evdence that an unexpected IIL provdes value-relevant nformaton about a negatve vew of the future proft-makng potental of a frm or an adverse mpact on ts debt contracts. In contrast, the return s not assocated wth the absence of goodwll amortzaton, whch can be nterpreted as the market s antcpaton and adjustment or the lack of nformaton content n goodwll amortzaton. Ths research helps better understand the mportance of managers ncentves n determnng IIL as well as the stock market effect of the announcement of the IIL and the excluson of goodwll amortzaton. Keywords: SFAS No. 142; goodwll mparment; earnngs management; stock prce response. 1

4 I. Introducton After many years of controversy, the Fnancal Accountng Standard Board (FASB) fnally ssued Statements 141 and 142 on July 20, The two statements made sgnfcant changes n the accountng rules for busness combnatons and goodwll. The Statement of Fnancal Accountng Standards (SFAS) No. 141, Busness Combnatons elmnates the use of poolng-of-nterests method and allows only purchase method for busness combnatons. SFAS No. 142, Goodwll and Other Intangble Assets requres that goodwll and ntangble assets wth ndefnte useful lves cease to be amortzed, and that a transtonal goodwll mparment test n the adopton year and an annual mparment test n subsequent years be performed. 1 Ths paper examnes two research questons wth respect to the adopton of SFAS No The frst s whether managers manpulate the transtonal goodwll mparment loss (Intal Imparment Loss, or IIL hereafter), and f so, then what motvates the manpulaton. The second s how a frm s stock prce responds to the announcement of the IIL and a boost n earnngs caused by the elmnaton of goodwll amortzaton. The nvestgaton of earnngs management s motvated by the flexblty that management has n determnng the ntal wrtng-off of goodwll as a result of the adopton of SFAS No. 142 because the rule requres a sgnfcant amount of judgment by managers, especally n the adopton year. One example of such judgment s the allocaton of goodwll to reportng unts. Although SFAS No. 142 calls for a reasonable and supportable approach (SFAS No. 142 paragraph 32-35), managers can use subjectve judgment n assgnng goodwll to reportng unts because f a busness combnaton provdes synerges and benefts to other operatons, managers can assgn some of ts goodwll nto other reportng unts. Thus, f managers want to wrte off a large amount of goodwll n the adopton year (as mght be the case f they are lookng to take a bg bath), they may allocate goodwll nto the reportng unts for whch 1 The prevous rule, APB No. 17, presumed that goodwll and other ntangbles were wastng assets (that s, wth fnte lves), and thus the amounts that were assgned to them were amortzed over a maxmum of 40 years. Under the new rule, ntangble assets wth fnte useful lves wll contnue to be amortzed over ther useful lves, but wthout the constrant of an arbtrary celng. 1

5 the mpled far value of the goodwll (that s, the far value of the reportng unt mnus the far value of net dentfable assets n the unt) s small. In contrast, f managers want to have lttle or no mparment n the adopton year, they may allocate as much goodwll as possble to reportng unts where an mpled far value of goodwll s suffcently hgh to pass the mparment test. Usng a sample of 870 frms that completed a transtonal goodwll mparment test n the adopton year, ths study nvestgates whether two manageral dscretonary varables frm leverage and change n management explan cross-sectonal varatons n the IIL after controllng for the actual economc mparment of goodwll, as proxed by varables such as ndustry, sze, the characterstcs of goodwll, and past asset or stock performance. Ths study frst predcts that as a frm nears ts debt covenant constrants, ts managers wll be more lkely to use ther dscreton to understate the IIL and thus create a debt covenant slack. Because a large IIL charge could tghten or cause the volaton of the constrants of debt covenants, frms wth a hgher leverage may be more reluctant to report a large IIL charge f leverage proxes for the exstence and tghtness of debt covenants. The second predcton s that a greater IIL charge wll be recorded by frms that have undergone recent changes n management than by frms n whch no such changes have occurred. New managers may tend to overstate the IIL n the transtonal perod of management change because by dong so they can reduce a possblty of future goodwll mparment that wll be charged aganst operatng earnngs. The emprcal results are consstent wth the two predctons, suggestng that the dscretonary behavor of managers plays a role n determnng the amount of the IIL. Ths study next examnes the behavor of the securty prces upon the dsclosure of the IIL charges and the elmnaton of goodwll amortzaton. The adopton of SFAS No. 142 has two sgnfcant effects on a frm s earnngs: t reduces earnngs f frms recognze the goodwll mparment loss, and ncreases earnngs as goodwll amortzaton s elmnated. 2 In event-study framework, I analyze the market effects of these two. Clearly, the IIL charge s a challenge for nvestors n assessng a frm s performance and 2

6 value because t s a bookkeepng adjustment that does not concde wth changes n the value of tangble assets or cash flows. The elmnaton of goodwll amortzaton may also confuse nvestors n ther assessment of frm value, especally n the adopton year, because the new accountng standard helps frms wth a substantal amount of goodwll to report a jump n earnngs, and because the market mght not be able to dstngush the cosmetc changes n earnngs caused by the new standard from other changes n earnngs. I predct that unantcpated IIL charges may have a sgnfcant and negatve mpact on stock prces, and provde two possble explanatons: sgnalng and contractng costs. Sgnalng suggests the conveyance of the prvate nformaton of managers to the market, n that an unantcpated IIL charge reveals a stuaton that s worse than expected by nvestors and thus changes the market s expectaton of a frm s future cash flows and rsk. Contractng costs ndcates that an unexpected IIL may have an adverse mpact on the debt contracts and other contracts of a frm (Watts and Zmmerman 1990). For example, f lowered earnngs caused by an unantcpated IIL result n more bndng debt covenants and further lead to sub-optmal frm decsons, then the unantcpated IIL wll cause a declne n market prce. However, I predct no market reacton to a boost n earnngs resulted from the elmnaton of goodwll amortzaton, because ths s a purely cosmetc and antcpated accountng number change that ratonal nvestors should be able to see through. Overall, the emprcal results reveal evdence consstent wth these predctons. Although stock returns have no sgnfcant assocaton wth an expected IIL or the elmnaton of amortzaton, they are negatvely assocated wth an IIL charge that s greater than expected, wth the negatve assocaton beng stronger for more hghly leveraged frms. Further analyss shows that subsequent to the IIL announcement, analysts revse ther earnngs forecasts for the upcomng quarter downward. Ths result s nterpreted as evdence that an unexpected IIL charge conveys value-relevant nformaton such as a negatve vew of future operatons or adverse effects on a frm s debt contracts. 2 One of the bggest IIL charges was AOL Tme Warner s 54 bllon-dollar charge (26% of ts total assets) n the fscal year of Had AOL Tme Warner adopted SFAS 142 n 2001, ts basc earnngs per common share would 3

7 Ths study contrbutes to the earnngs management lterature by examnng the dscretonary behavor of managers n determnng the amount of ther IIL upon the adopton of SFAS No The FASB contends that the beneft of wrtng down goodwll on an mparment bass, rather than on the bass of an arbtrary and systematc amortzaton, s that frms wll be more transparent to nvestors regardng the economc value of goodwll and the extent of ts mpact on earnngs. However, ths purported beneft wll be mtgated f managers use dscreton opportunstcally to mslead stakeholders about the underlyng economc value of goodwll. The results of ths study wll help accountng users to understand the opportunstc behavor by managers and ts effect on fnancal statements. In addton, ths study provdes new nsght nto research on the nformaton content of asset wrteoffs. Three prevous studes examne the nformaton content of goodwll wrte-offs n the pre-sfas No. 142 perod and reveal nconsstent results. Whereas Francs, Hanna and Vncent (1996) report no sgnfcant market reacton to goodwll wrte-offs, 3 Hrschey and Rchardson (2002, 2003) fnd a 2-3 percent negatve stock prce reacton to goodwll wrte-off announcements over a two-day wndow. Ths lack of consstency may be attrbutable to the dfferences n the samples of the goodwll wrte-offs examned or to the dffculty n dentfyng an approprate research desgn. By proposng a new research desgn that accounts for the possblty that nvestors wll use avalable frm performance nformaton to form expectatons of certan wrte-offs beforehand, ths study documents new and mportant evdence that only unantcpated portons of goodwll wrte-offs convey unfavorable news to the market, whereas the expected portons do not. Ths study also fnds evdence that analysts revse ther earnngs forecasts downward for upcomng quarters n response to an unantcpated IIL, whch s smlar to the market s response. Overall, gven that the test for the goodwll mparment became an annual routne event for frms adoptng SFAS No. 142, ths paper provdes tmely evdence that unantcpated goodwll have ncreased from $(1.11) to $0.31 wth the excluson of goodwll amortzaton. 3 Francs, Hanna and Vncent (1996) fnd that the market reacton to asset wrte-offs s negatve overall, but s condtonal on the type of asset wrte-off. They report a negatve reacton to nventory wrte-offs and a postve reacton to restructurng charges, but do not fnd a sgnfcant market reacton to goodwll wrte-offs, although goodwll wrte-offs are the most substantal n ther sample, both n terms of dollars per share and as a percentage of total assets. 4

8 mparments are not trval economc events but rather have mportant mplcatons for ongong frm value. The remander of ths study s organzed as follows. Secton II dscusses the nsttutonal background to the accountng for goodwll ssue and relevant prevous studes, secton III outlnes the research hypotheses, and secton IV descrbes the sample selecton and data. Secton V detals the models and research methodology that are used to nvestgate the research questons. Secton VI reports the emprcal results, and secton VII concludes wth a summary. II. Background In formulatng SFAS No. 142, the FASB has attempted to brng a more realstc approach to goodwll accountng. Many had argued that amortzng and expensng a fxed amount of goodwll each year over ts useful lfe dd not accurately portray the way n whch the loss n ts economc value occurred, because when goodwll loses value t does so rregularly and to varyng degrees. The FASB asserts that the new approach of a goodwll mparment test combned wth the elmnaton of amortzaton more accurately assesses any actual declne n the value of goodwll and gves greater nsght nto ts economc value. Under SFAS No. 142, when assessng whether there has been an mparment of goodwll, frms must establsh one or more reportng unts and allocate all ther exstng goodwll as well as other assets and labltes nto these reportng unts. In the mparment test, the mpled far value of a reportng unt s goodwll s compared to the carryng amount of the goodwll. If the carryng amount exceeds the mpled far value, then the excess must be recognzed as an mparment loss. A recognzed mparment loss cannot be reversed f the far value of the goodwll subsequently ncreases. SFAS No. 142 came nto effect for frms wth fscal years that began after December 15, 2001, and early adopton was avalable for frms wth fscal years that began after March 15, Before the ssuance of SFAS No. 142, goodwll mparment was measured n the same way as other assets under SFAS No. 121, Accountng for the Imparment of Long-Lved Assets and for Long-Lved Assets To Be Dsposed of. Because goodwll s a unque asset that s not drectly assocated wth any specfc dentfable rghts and s not separable from the company as a whole, SFAS No. 142 was 5

9 formulated to provde specfc gudelnes for the goodwll mparment test. Under SFAS No. 121, an asset s consdered to be mpared f the undscounted future cash flows from the use of the asset are less than the carryng value. However, under SFAS No. 142, snce the present values of future cash flows s used n comparson wth the carryng value, goodwll s more lkely to be wrtten down than other long-lved assets (that s, t s a more conservatve measurement of goodwll). Accordngly, the effect of the IIL that results from the adopton of SFAS No. 142 on the fnancal statements of frms was substantal. The FASB states that an IIL charge at the date of the adopton (January 1, 2002 for calendar-year companes) s treated as the cumulatve effect of a change n accountng prncple. An advantage of ths treatment s that the IIL s not charged to operatng results, whle further future mparments n the subsequent perods are charged aganst operatng earnngs. A stream of lterature has documented evdence on the causes and effects of wrtng-off mpared assets. One vew of such wrte-offs s that management uses ther dscreton to manpulate earnngs ether by gnorng mparment when t occurs or by recognzng t only when t s advantageous to do so (Strong and Meyers 1987; Ellott and Shaw 1988). An alternatve pont of vew s that managers record wrte-offs not to manpulate earnngs, but prmarly to respond approprately to changes n the economc envronment of ther frms (Rees, Gll and Gore 1996). Francs et al. (1996) provde evdence that both manpulaton and economc mparment can drve wrte-off decsons. My study dffers from prevous studes of wrte-offs n that the focus s on the effects of adoptng the newly mandated accountng rule SFAS No. 142, whch establshes a dfferent set of standards for testng goodwll mparment. As an IIL n the adopton year can be reported as a cumulatve effect of accountng change, to whch analysts and nvestors often assgn a lower value weght than ncome tems from contnung operatons, the dscretonary behavor of management and the reacton of stock prces may substantally dffer from that reported n prevous studes. Several papers have studed the nformaton content of goodwll amortzaton. Vncent (1997) provdes evdence that nvestors adjust the accountng numbers of poolng and purchase frms to place them on an approxmately equal bass. By comparng the R 2 from annual regressons, Jennngs, LeClere, 6

10 and Thompson (2001) fnd that earnngs before goodwll amortzaton explan sgnfcantly more of the dstrbuton of share prces than earnngs after goodwll amortzaton. Lndenberg and Ross (1999) fnd that as goodwll amortzaton ncreases, the prce-to-earnngs multples ncrease to a suffcent extent to offset the mpact of the amortzaton. Unlke prevous studes, ths study examnes how the elmnaton of goodwll amortzaton and resultant earnngs ncrease convey nformaton to nvestors usng an eventstudy framework. The results of ths study wll help researchers better understand the nformatonal role of goodwll amortzaton. III. Hypothess Development Earnngs Management Hypotheses The adopton of SFAS No. 142 provdes an opportunty for earnngs management as the standard requres substantal judgment and estmaton on the part of management, especally n the adopton year. Ths ncludes, but s not lmted to, the dentfcaton of reportng unts; the allocaton of assets, labltes, and goodwll to these unts; the far value measurement of the reportng unts; and the measurement of the mpled far value of the goodwll. Ths flexblty, combned wth the potentally large sze of the ntal wrte-off, suggests that frms may use the adopton of SFAS No. 142 n an opportunstc way. Because managers have abltes to manpulate the IIL charge, gven the subjectvty of many estmates, ther preferences and ncentves wll nfluence the magntude of the IIL n a sgnfcant and drect way. In transtonal mparment tests, some frms may understate ther IIL and others may overstate t, dependng on the manageral ncentves. I hypothesze two ncentves assocated wth the IIL. The frst s that as a frm nears ts debt covenant constrants, management wll be more lkely to use accountng dscreton to understate the IIL, thereby creatng a debt covenant slack. Debt covenants, whch are wrtten manly n terms of accountng numbers, can be dvded nto affrmatve covenants and negatve covenants (Bartov 1993). Affrmatve covenants requre borrowng frms to mantan specfed levels of accountngbased ratos. Negatve covenants restrct the fnancng and nvestng actvtes of borrowng frms (for example, dvdend payments, merger actvty, and the ssuance of new debt) f any predetermned 7

11 accountng-based condton n the covenant s not met. As t s costly to volate debt covenants, frms that are close to the constrants of ther covenants may take actons to manage earnngs to reduce the lkelhood of covenant volaton (DeFond and Jambalvo 1994; Sweeney 1994). A large IIL charge wll have a negatve mpact on the fnancal structure and debt covenant of a frm, because t reduces the amount of assets and at the same tme t flows through the ncome statement nto retaned earnngs, thus lowerng stockholder equty. Ths mpact s reflected n reports n the fnancal press. For example, n a Wall Street Journal artcle, Bll Cunnngham, the drector of credt strategy at J. P. Morgan Chase & Co., s quoted as sayng that he would absolutely expect some debt covenants to be volated as a functon of goodwll wrte-offs. 4 The artcle ndcates that the potental mpact of IIL charges on debt covenants s substantal, and mples that management may use ts dscreton to adjust the amount of the IIL to avod a negatve mpact on the debt covenant. To test ths predcton, prevous studes often used leverage ratos as a proxy for the exstence and tghtness of accountng-based covenants, largely because of the cost of accessng detaled debt covenant nformaton. Followng ths lterature, I adopt leverage (total debt deflated by total assets) as the proxy. 5 Accordngly, I predct that frms wth a hgher leverage wll be more reluctant to accept a large IIL charge, as they are more lkely to be constraned by debt covenants and to be closer to the volaton of the constrants of these covenants. Hypothess 1: Ceters parbus, the IIL charge wll be relatvely smaller for more hghly leveraged frms. 4 The artcle also clams that the loomng goodwll wrte-downs could be large enough to place companes n volaton of ther credt agreements, snce the wrte-downs of goodwll would slash a company s book value on whch the covenants are based, and that the wrte-offs could gve banks leverage to wn concessons from the borrowers, such as hgher nterest rates and fees, the pledgng of assets as securty or agreements by the companes to pay off ther loans sooner. See Rapoport and Wel, Goodwll as a Banker s Weapon, (Wall Street Journal, September 23, 2002) for detals. 5 The choce of leverage as a proxy has been justfed n prevous lterature. Duke and Hunt (1990) examne the relatonshp between leverage and the exstence and tghtness of several actual debt covenant restrctons for a random sample of frms, and fnd that leverage captures the exstence and tghtness of three common restrctons that are related to retaned earnngs, workng captal, and net assets. Press and Wentrop (1990) report that for frms wth accountng constrants n ther debt agreements, leverage s correlated wth proxmty to the actual constrants n the debt covenants. In a later secton, I perform senstvty checks on the effect of IIL on debt covenants and alternatve leverage proxes. 8

12 A recent change n management may be assocated wth the magntude of the IIL charge. Goodwll mparment charges may expose the management of the acqurer to allegatons of overpayment for acqustons. Even f such allegatons are proved untrue, the management s reputaton may reman tanted n the eyes of the captal markets communty. Such concerns could make exstng management reluctant to accept the IIL charge. In contrast, a new management may conscously overstate the IIL charge because by attrbutng the IIL charge to the poor decsons of ther predecessors they can reduce the possblty of future goodwll mparment, whch wll be charged aganst operatng earnngs after the transton perod. Moore (1973) and DeAngelo (1988) fnd that new managers have a tendency to take a bath by wrtng off assets and ncreasng the provson for future costs or losses. Strong and Meyer (1987) and Francs et al. (1996) also report that asset wrte-offs are more frequent and larger n magntude f there has been a recent change n management. A common vew n earler research s that a new management tends to attempt to blame the prevous management for current or past problems. Followng ths vew, I predct that a larger IIL charge wll be recorded among frms that have undergone a recent change n management, because the new managers may delberately overstate the IIL to reduce the possblty of future goodwll wrte-offs, whch wll be charged aganst operatng ncome. 6 Hypothess 2: Ceters parbus, the IIL charge wll be greater for frms n whch there has been a recent change n management than for frms n whch no such change has occurred. Stock Market Response Hypotheses Ths secton hypotheszes a short-term market reacton to the announcement of a decrease n earnngs nduced by the IIL charge on goodwll and a boost n earnngs caused by the elmnaton of goodwll amortzaton. It s expected that the market wll react more strongly to the IIL announcement the more unexpected the amount of the IIL. A frm that reports a larger IIL charge than expected, n essence, acknowledges that ts goodwll asset, and thus the frm tself, wll not generate as much cash from ts 9

13 acqustons as was prevously projected. If an unexpected charge sgnals mportant changes n a frm s future proft-makng potental and the ntangble dmenson of a frm s value, then the frm s stock prce wll be affected because the forecasts that nvestors make of future cash flows, as foreshadowed by unexpected wrte-offs, wll have materally changed. The reacton hypothess s further supported by the fact that an unexpected IIL charge may cause some cash flow effects by changng the fnancal ratos and other measures that are used to montor complance wth varous contracts, such as regulatory captal requrements and debt covenants (Holthausen 1981; Watts and Zmmerman 1990). For example, f the IIL charge causes a frm to volate (or ncreases ts lkelhood of volatng) the constrants of ts debt covenants, then as long as the volaton s unexpected there wll be an adverse mpact on the frm s stock prce upon the announcement of the IIL. Takng these arguments together, I propose the followng hypothess: Hypothess 3: A greater (smaller) than expected IIL charge wll have a negatve (postve) mpact on a frm s stock prce. However, t s possble that there wll be no reacton to the announcement of the IIL charge. Ths may be the case f goodwll wrte-offs lack value-relevant nformaton (that s, are mere bookkeepng adjustments that have no mplcatons for the frm s value or cash flow), as some analysts and nvestment banks have argued n the fnancal press. 7 It may also be the case f the market perceves that nose n the measurement of an IIL swamps ts nformaton content (that s, that the potental manpulaton by an opportunstc management blurs ts value-relevant nformaton content). Lastly, I examne the market response to a boost n earnngs caused by the elmnaton of goodwll 6 Note that alternatve explanatons also exst for such a fndng. If t s the under performance of the frm that has prompted the change n management, then poor performance may be related to the IIL charge. Robustness tests are performed n a later secton to examne ths possble relaton. 7 Investment professonals sometmes argue that the accountng mpact of goodwll wrte-offs should be gnored, as s llustrated by the followng quotes. We beleve that there may be a large catch-up mparment wrte-off when SFAS No. 142 s frst adopted; however, t s mportant to understand that SFAS No. 142 adjustments are non-cash and do not reflect changes n busness fundamentals (Equty Research, Saloman Smth Barney, December 28, 2001). Wall Street analysts tend to encourage nvestors to overlook the one-tme charge as water under the brdge. And fundamentally, even huge goodwll mparment charges by themselves have no mpact on the sze of a company s cash flow streams or tangble assets (Wall Street Journal, January 3, 2001). 10

14 amortzaton on the dates of earnngs announcements. SFAS No. 142 stops goodwll amortzaton charges, and therefore frms that used to recognze a sgnfcant amount of goodwll amortzaton wll mprove ther earnngs substantally upon adoptng the standard. However, nvestors should not value such earnngs ncreases by dstngushng them from other real earnngs nnovatons because such earnngs ncreases are cosmetc changes n accountng numbers that do not concde wth changes n potental cash flows or rsk. Ths leads to the followng hypothess: Hypothess 4: A frm s stock prce s not sgnfcantly assocated wth a boost n earnngs caused by the elmnaton of goodwll amortzaton. Note that one could have dfferng vewponts. One artcle comments that the elmnaton of amortzaton may matter to nvestors because a rse n earnngs leads to a decrease n the prce-to-earnngs rato, whch may make stocks appear cheaper. 8 Nevertheless, the lowered prce-to-earnngs rato should not be valued f the market makes the necessary adjustments for dfferences n accountng methods based on the analyss of restated pro forma pror earnngs, rather than myopcally relyng on reported earnngs. The elmnaton of goodwll amortzaton may also have potental to reduce the debt-covenant volaton costs f the covenant s not revsed due to re-contractng costs. In such a case, one mght expect that the stock prce reactons wll be postve and ncreasng wth leverage. However, the elmnaton of goodwll amortzaton and the resultant earnngs boost are lkely to be fully or at least substantally antcpated through the prevous nformaton events, such as the release of drafts, publc hearngs, and the fnal announcement of SFAS No In turn, t s expected that ratonal nvestors consder the potental effect on the covenants n advance before the dates of earnngs announcements f such effect exsts IV. Data and Sample My ntal sample conssts of frms n the Compustat Industral database wth a goodwll balance of greater than 5% of total assets at the end of the 2000 fscal year (1,772 frms). Ths requrement s an attempt to guarantee that the adopton of SFAS No. 142 has a due nfluence on the sample frms. To 11

15 dentfy a sample of frms that have reported the adopton of SFAS No. 142, the result of a transtonal mparment test, and earnngs announcements followng the adopton, I searched the Dow Jones Interactve database for the perod March 2001 to July The search was conducted usng the stock tcker symbols of the sample frms and varous keywords that are typcally observed n such announcements, such as SFAS No. 142, goodwll mparment, and earnngs announcement. Ths search dentfed the announcements of 1,319 frms. Whle searchng, I found that among the frms wth a fscal year startng between July and December, 80 frms dd not adopt SFAS No. 142 untl June These frms are excluded from the sample, because the CRSP data that s necessary for stock market response tests n the adopton year were not avalable for these frms at the pont of data collecton. For the remanng frms, I retreved the IIL amount and the earnngs boost caused by the elmnaton of goodwll amortzaton from news releases. I obtaned more detaled note dsclosures about the mparment test, other fnancal accountng data, and stock prce data from EdgarScan TM, Compustat, and CRSP, respectvely. These procedures excluded 270 frms for whch relevant data for analyses were not avalable, whch left me wth a total of 969 sample frms. Table 1 reports the detals of the sample selecton. [INSERT TABLE 1 ABOUT HERE] Among the 969 fnal sample frms, 424 (44%) fnalzed ther ntal mparment test n the frst quarter of the adopton year, 446 (46%) fnalzed t n the second quarter, and 99 frms (10%) had not fnshed the procedure by the end of the second quarter. Among the 870 frms that fnalzed the procedure, 255 frms reported a non-zero IIL charge. [INSERT TABLE 2 ABOUT HERE] Panel A of Table 2 presents descrptve statstcs for the fnal sample of 969 frms. The mean (medan) amount of goodwll n the fnal sample s about $687.6 ($76.9) mllon, whch represents 20.8% 8 Elsten, Frms Fatten up Proft Outlooks on FASB Rule, (Wall Street Journal, October 21, 2001). 9 I use ths searchng perod because frms that adopted SFAS No. 142 between March 2001 and December 2002 (early adopters n 2001 and the others n 2002) were requred to complete the frst step of the transtonal goodwll mparment test before ther second quarter nterm reportng. Dow Jones Interactve ncludes news releases from Busness Wre, PR Newswre, Dow Jones Newswre, Southwest Newswre, Reuters, and Unted Press Internatonal. 12

16 (16.8%) of the book value of the total assets at the begnnng of the adopton year. The mean (medan) goodwll amortzaton expense n the frst quarter of the year before the adopton of SFAS No. 142 s $7.5 ($0.8) mllon dollars. The mean (medan) ncrease n earnngs caused by the elmnaton of goodwll amortzaton n the frst quarter of the adopton year s about 3% (0.7%) of sales revenue. Panel B of Table 2 provdes descrptve statstcs of the IIL of the 255 frms that completed the transtonal mparment test and reported a non-zero IIL durng the frst or second quarter followng the adopton. The mean before-tax (after-tax) IIL s (398.5) mllon dollars. The mean before-tax IIL represents 8.8% (or 46.3%) of the book value of total assets (or goodwll) at the begnnng of the adopton year, and the medan represents 5.6% (or 38.9%). The amount of the IIL was substantal for the majorty of IIL recognton frms. Whle 141 (55.3%) of the 255 IIL recognton frms recorded an IIL of more than 5% of the total assets, only 34 (13.3%) recorded an IIL of less than 1% of total assets. Panel C of Table 2 shows the ndustry dstrbuton of the 870 frms that completed the transtonal mparment test durng the frst quarter or the second quarter after adopton. The proporton of frms that recognzed a non-zero IIL s not evenly dstrbuted across ndustres. The proportons n the mnng/constructon, textles/publshng, chemcals, durables manufacturers, transportaton, and servces ndustres are hgher than the overall mean proporton (29%). In terms of the magntude of the IIL compared to total assets (goodwll), relatvely hgher numbers are found n the mnng/constructon, pharmaceutcals, and servces ndustres (mnng/constructon, chemcals, transportaton, retal, and computers ndustres). 10 On average, 2.6% (8.9%) of total assets or 13.6% (46.3%) of goodwll were wrtten off by all of the sample frms (non-zero IIL sample frms) as a result of IIL recognton n the frst and second quarter after adopton. V. Research Desgn Test for Earnngs Management 10 I repeat all of the tests n ths study after excludng sample frms n the utltes and bankng and fnancal servces ndustres, as these frms are subject to heavy government montorng and regulatons, but the nferences stay the same. 13

17 To test the earnngs management hypotheses, I estmate the followng tobt model. 11 IIL 0 + α RET1 + α RET 4 + α ROA1 + α ROA4 + α LEVERAGE + α ΔMGMT + ε 7 13 = α + α I = 1 1, I IND 8 I, + α SIZE + α GWA + α GWR + α IAPD + α IPR & D (1) Defnton of the varables Dependent varable: IIL : ntal goodwll mparment loss deflated by total assets at the begnnng of adopton fscal year for frm. IIL s measured pre-tax and s expressed as a postve number. Actual mparment proxy varables: Control varables for ndustry and sze IND I, (?): ndustry dummy varables. SIZE (?): log of total assets at the begnnng of adopton fscal year. Control varables for characterstcs of goodwll GWA (+): goodwll deflated by total assets at the begnnng of adopton fscal year. GWR (-): excess of the estmated mpled far value of goodwll over the book value of goodwll deflated by total assets. IAPD (+): ndustry-adjusted amortzaton perod of goodwll before the adopton of SFAS No. 142 (frm amortzaton perod mnus the ndustry average amortzaton perod). IPR&D (-): n-process R&D costs n M&A over the 5 years before adopton deflated by total assets. Control varables for past asset/stock performance RET1 (-): cumulatve market-adjusted return over the 1 year before adopton. RET4 (-): average annual cumulatve market-adjusted return over years -4 to -1 before adopton. ROA1 (-): ROA over the year before adopton. ROA4 (-): average annual ROA over years -4 to -1 before adopton. Dscretonary manpulaton proxy varables: LEVERAGE (-): total debt deflated by total assets at the begnnng of the fscal year. Δ MGMT (+): 1 f there was a change n key management durng the year before adopton, and 0 otherwse. The dependent varable IIL s the pre-tax value of the IIL charge for frm deflated by begnnng total assets. 12,13 The ndependent varables consst of two groups: actual mparment proxy varables and two 11 A tobt model s used for the multvarate analyss because the observatons on the dependent varables below zero are censored (unobservable), but the explanatory varables are avalable for all of the observatons (Maddala 1991). 12 Accordng to U.S. Tax Code Secton 197, only goodwll that was acqured after August 10, 1993 n connecton wth a trade or busness or n an nvestng actvty s allowed to be amortzed (over 15 years) n tax reportng. The IIL charge to deductble goodwll does not nduce early deductons for tax reportng purposes, but merely creates a temporary dfference between accountng and taxable ncome by brngng deferred tax assets and reduced ncome tax expenses nto the fnancal accountng n the adopton year. Measurng the IIL net of tax may ntroduce 14

18 dscretonary manpulaton proxy varables. The actual mparment proxy varables attempt to capture cross-sectonal varaton n the real mparment of goodwll before the attrbutes of manpulaton are explored. As Wlson (1996, 172) notes, the credblty of emprcal fndngs n asset wrte-off studes reles on the extent to whch the expermental desgn controls for actual mparment factors. Although each actual mparment varable s carefully defned, the hypothess test wll be based to the extent that the mssng actual mparment varables have an effect on the manpulaton proxes. The ndustry of frms s expected to create some common effects wth respect to goodwll mparment, as the IIL charge may be closely related to competton, deteroraton, or other economc factors n an ndustry. 14 SIZE s a comprehensve varable that proxes for varous aspects of a frm. For example, t can control for the frequency and amount of prevous acqustons. Large frms generally experence a greater number of mergers and acqustons, and have more complcated structures, and thus ther goodwll mparment pattern may be dfferent from that of small frms because ther operatons are affected by more frequent acqustons. It can also proxy for nformatonal effcency. If there are more analysts followng large frms and such frms receve greater publc attenton, then ths may lead to the more effcent processng of accountng nformaton for these frms and fewer ncentves for ther managers to manpulate the IIL. I nclude four varables to proxy for the characterstcs of goodwll. Frms wth a greater amount of goodwll n ther asset composton (frms wth a hgher GWA ) may ncur more goodwll mparment, because the relatve amount of goodwll that s exposed to the mparment test wll be greater. GWR measures the extent to whch the estmated mpled far value of frm-wde goodwll exceeds the book value of goodwll scaled by total assets, where the mpled far value of goodwll s estmated as follows. unnecessary nose n the assocaton of goodwll mparment wth the ndependent varables, because t may be sgnfcantly dfferent from the actual amount of goodwll wrte-off, dependng on the deductblty of goodwll. Hence, I use the before-tax IIL for all of the analyses n ths study. However, when both earnngs management and stock market response tests are performed wth the IIL net of tax, the man results are qualtatvely the same. 13 As the mpact of an IIL on earnngs s partly offset by the elmnaton of goodwll amortzaton, I also attempt an alternatve measure of IIL net of the reducton n goodwll amortzaton, and fnd that the results wth ths alternatve dependent varable are qualtatvely smlar. 14 Industry classfcaton s descrbed under Panel C of Table 2. 15

19 Impled far value of goodwll of a frm = Far (market) value of the frm - Far value of the net dentfable assets of the frm Market Captalzaton (Net book value of the frm Goodwll amount). Ths measure of the mpled far value of goodwll assumes that the whole frm was purchased n a busness combnaton n whch the far value of dentfable assets and labltes was assumed to be equal to ther book values. 15 I expect a frm wth a hgher GWR to be lkely to ncur a smaller goodwll mparment. IAPD measures the ndustry-adjusted amortzaton perod of goodwll before the adopton of SFAS No. 142 (the frm amortzaton perod - the ndustry average amortzaton perod n the sample frms). 16 Before SFAS No. 142, the SEC ncreasngly recommended that the amortzaton perod of goodwll be related to the nature of the busness that was acqured. Accordngly, hgh-tech acqustons often used 5 to 7 year amortzaton perods and banks used 15 to 20 year perods. Occasonally, other frms set a longer amortzaton perod to reduce the effect of goodwll amortzaton on earnngs, whch caused a hgh varaton n the amortzaton perod wthn an ndustry. I predct that a frm wth a longer amortzaton perod than the ndustry average wll ncur a greater IIL, because ts amortzaton expense n the past s more lkely to have been understated compared to the reducton n the economc value of ts goodwll. IPR&D measures the n-process R&D costs of acqustons over the 5 years before the adopton of SFAS No Some frms wrte off a large porton of ther acquston costs as purchased R&D upon acquston, and there has been an ncrease n such wrte-offs durng the last decade, especally n the hgh-tech ndustres. Ths practce allows the acqurng frms to reduce the allocaton of goodwll and the subsequent amortzaton expense. The n-process R&D wrte-off s predcted to mprove the qualty of the goodwll of the acqurng frms (Deng and Lev 2001), and a frm that recognzed a greater amount of these wrte-offs n the past s expected to ncur a smaller IIL. 15 Although GWR s a drect measure of the IIL of frms, but t s nosy because of the assumptons that are nvolved. Because of ths, other actual mparment proxes are ncluded n the regresson. 16 The ndustry average amortzaton perod s calculated for each IND I category. The data of the amortzaton perods are obtaned from the note dsclosures on the goodwll accountng polcy of the sample frms. 16

20 The next set of varables controls for hstorcal stock or asset performance. Lke Francs et al. (1996), I control for long-term hstorcal performance (years -4 to 1 before adopton) and recent performance (year -1 to 0), because mparment may start to occur long before adopton. 17 RET1 and RET4 are ncluded because the IIL s lkely to be closely related to past stock prce performance. In determnng the far value of a reportng unt, frms frst look for the market prce of the unt n actve markets, and so the worse a frm s hstorcal stock performance, the more lkely ts goodwll wll be mpared. ROA1 and ROA4 capture the returns on assets (ROA). 18 Frms wth a hgher ROA are lkely to ncur a smaller IIL because the far value of ther assets wll be more apprecated. Fnally, two proxes of earnngs management ncentves are ncluded. LEVERAGE s the total debt deflated by total assets at the begnnng of the adopton year. 19 Ths s ncluded to test H1. ΔMGMT s a dummy varable to test H2, and has a value of 1 f any of the top three compensated postons changed over the year before adopton, and 0 otherwse. 20 Test of Stock Market Response Ths secton outlnes the examnaton of the market reacton to the announcement of the IIL charge and the earnngs boost caused by the elmnaton of goodwll amortzaton. In searchng Dow Jones Interactve, I found that the IIL charge was dsclosed n conjuncton wth the earnngs dsclosure for all but one of the sample frms. Thus, the announcement perod return s defned as the cumulatve return from one day before to one day after the earnngs announcement of the frst or second quarter n the adopton fscal year less the return on the CRSP value-weghted market portfolo. The ncluson of day-1 s meant to capture early news leakage, whereas the ncluson of day+1 ensures that the prce mpact of 17 Startng wth the 4 years before adopton s a judgment that s based on the study of Ayers, Lefanowcz and Robnson (2000). Usng the Securtes Data Corporaton s (SDC) M&A Database, they show that frm acqustons by purchase method became more frequent n number and more sgnfcant n acquston amounts from ROA s defned as operatng ncome before deprecaton (Compustat #13) dvded by average total assets (#6). 19 Total debt s defned as the sum of long-term debt (Compustat #9) and debt n current labltes (#34). In a later secton, a check for the senstvty of the results relatve to ths partcular defnton of leverage s performed. As the results are ndstngushable, I only report the results usng ths defnton. 20 Data on changes n management were obtaned from each frm s proxy statement. 17

21 any announcements after the market closed s captured. 21 The announcement perod returns (CAR) reflect the market s contemporaneous reacton to the news of earnngs n the quarter, and so I regress ths on the earnngs surprse (SURPRISE). For the frms (700 frms) for whch analyst forecasts are avalable n IBES, the earnngs surprse s defned as the actual pro forma IBES earnngs per share mnus the most recent mean IBES earnngs per share forecast before the earnngs announcement scaled by the share prce on day For the frms (269 frms) for whch IBES data are not avalable, t s defned as the change n quarterly earnngs per share excludng the IIL and goodwll amortzaton (seasonal random walk: current quarter mnus the same quarter of the precedng year) scaled by the share prce on day-2. In addton, four prevously documented determnants of stock returns are ncluded: the book-tomarket rato (BTM), whch s constructed as the book value of shareholders equty dvded by the market value of equty at the end of the fscal quarter; the sze (SIZE) of the frm as measured by the log of the market value of the equty at the end of the fscal quarter; beta (BETA), whch s estmated usng the daly returns over the year before the fscal quarter; and momentum (MOM), whch s calculated as the marketadjusted stock return for the sx months before the earnngs announcement. The frst three control varables are frequently used as rsk factors, whereas the last s often used to control for the so-called momentum anomaly (Jegadeesh and Ttman 1993). Fnally, I nclude n the regresson nformaton on the IIL charges and the boost n earnngs that s due to the elmnaton of goodwll amortzaton. Ths results n the followng two-quarter-pooled regresson. CAR = α + β SURPRISE + β BTM + β AMORT + DIIL ( β EIIL + β UIIL ) + ε β SIZE + β BETA + β MOM (2) After controllng for other contemporaneous nformaton, ths model tests the stock prce reacton to 21 As a senstvty test, the announcement perod returns are also measured usng a 2-day wndow (0 to1), and the overall result s consstent. 22 IBES EPS measures do not nclude the tems below ncome from contnung operatons, such as dscontnued operatons, extraordnary tems, and the cumulatve effect of accountng changes (whch ncorporates the IIL charge). IBES has mgrated ts estmates from usng a pre-sfas No. 142 bass that ncluded goodwll amortzaton 18

22 the nformaton of the IIL per share (parttoned nto expected (EIIL) and unexpected (UIIL) components) and the boost n earnngs (AMORT) per share that s due to the elmnaton of goodwll amortzaton, both scaled by the share prce on day-2. If the fnalzed IIL s announced n the frst (second) post-adopton earnngs announcement for a frm, then DIIL = 1 for the frst (second) earnngs announcement perod observaton and DIIL = 0 for the second (frst) earnngs announcement perod observaton for the frm. Although researchers often use the consensus analyst earnngs forecast as the proxy for expected earnngs, t s not straghtforward to fnd a reasonable benchmark for the market s expectaton of the IIL. Due to ths dffculty, most prevous studes of asset wrte-offs (for example, Francs et al. 1996; Ellott and Hanna 1996; Ellott and Shaw 1988) use the entre amount of the wrte-off as the ndependent varable, rather than estmatng the unexpected porton, and assume all wrte-offs to be purely unantcpated. However, as s dscussed n the next secton, frms that announced a non-zero IIL experenced deteroratng accountng or stock performance relatve to frms that announced a zero IIL before the adopton of SFAS No Therefore, t s unlkely that all of the charge was a complete surprse to the market. To the extent that the market partally antcpates the IIL before ts announcement, usng the entre IIL amount ntroduces a measurement error and nconsstent estmates of the market response. I therefore estmate two pars of the expected and unexpected IIL by usng Eq. (1). (1) EIIL1 = E[IIL ] usng only the actual mparment proxy varables n Eq. (1), UIIL1 = e. (2) EIIL2 = E[IIL ] usng both the actual mparment and the dscretonary manpulaton proxy varables n Eq. (1), UIIL2 = e. The frst par of estmates assumes that nvestors formulate ther expectatons of the IIL based on the frm s ndustry, sze, past stock or operatng performance, and frm-specfc ndcators of goodwll qualty as wth the actual mparment proxes n Eq. (1), and that nvestors are unable to see through the earnngs management ncentves. The second par assumes that nvestors take nto account the dscretonary to a post-sfas No. 142 bass that excludes goodwll amortzaton. IBES has announced that once a frm reports ts results as havng been affected by SFAS No. 142, the old bass estmates are automatcally moved to a backup table. 19

23 manpulaton ncentves n formulatng ther expectaton of the IIL. 23 The approprateness of these two pars of estmates depends on how closely the proxy varables n Eq. (1) approxmate the unobservable percepton of nvestors of mpared goodwll. However, the power of my tests wll be reduced to the extent that nvestors rely on non-accountng and non-stock-market nformaton or other management ncentves that are not reflected n Eq. (1). If the test results reveal a sgnfcant negatve coeffcent on UIIL and an nsgnfcant coeffcent on EIIL, then ths wll be consstent wth H3. An nsgnfcant coeffcent on AMORT would be consstent wth H4. VI. Emprcal Results Test of Earnngs Management Table 3 provdes descrptve statstcs for the varables that are used n the earnngs management test. The table presents the mean and medan values of the varables for the IIL frms (N = 255), zero-iil frms (N = 615), and all of the sample frms (N = 870) that announced the fnalzed IIL measurement n the frst or second quarter after adopton. The two rghtmost columns report the mean and medan dfferences between the IIL frms and the zero-iil frms usng a bnomal t-test and a nonparametrc Wlcoxon ranksum test, respectvely. [INSERT TABLE 3 ABOUT HERE] Overall, the IIL frms exhbt sgnfcantly lower RET1, RET4, and ROA1 than the zero-iil frms. Ths suggests that the hstorcal asset or stock performance of the IIL frms s nferor to that of the zero- IIL frms before adopton. In general, the IIL frms are sgnfcantly larger n sze and carry a hgher percentage of goodwll relatve to total assets than the zero-iil frms. A hgher mean GWR among the zero-iil frms ndcates that the mpled far value of goodwll exceeds the book value of goodwll to a greater extent among the zero-iil frms (and thus gves them a greater buffer aganst the IIL charge) than 23 EIIL and UIIL are deflated by total assets when estmated by Eq. (1). I convert them to per-share amounts by multplyng the total assets and dvdng them by the weghted average (basc) number of shares outstandng for the quarter, scaled by the share prce on day-2. For the lst of actual mparment proxy varables and dscretonary manpulaton proxy varables, please refer to the defnton of the varables n the secton V. 20