INTERPRETATION OF ABNORMAL AUDIT DELAYS: IMPLICATIONS FOR EARNINGS QUALITY AND FIRM VALUE 1. Inroducion. The purpose of his paper is o examine he associaion beween abnormal audi delays and earnings qualiy and is effec on he earnings valuaion coefficien. 1 Prior research (discussed in he nex secion) has examined he deerminans of audi delays. Audi delays can cause delay in annual accouning disclosures. Delayed earnings announcemens generally cause less marke reacion han early announcemens due o lack of imeliness or even negaive reacions as hey are likely o conain bad news. However, he research quesion ha abnormal audi delay, ha may be caused by maerial disagreemen beween he audior and clien regarding accouning pracices and/or calculaion of accouning numbers, may conain informaion abou qualiy of earnings beyond ha conveyed by earnings repor delay has no been invesigaed by exan research. Moreover, he marke may use his incremenal informaion abou earnings qualiy in he firm s valuaion process. The curren paper conribues o exising research in several ways. Firs, i esablishes a comprehensive model o explain audi delays and provides a ool o measure abnormal audi delays. Second, i provides evidence of inverse associaion beween abnormal audi delay and seven proxies of earnings qualiy. Third, he abnormal audi delay is shown o provide incremenal informaion abou earnings qualiy beyond ha conained in he earnings repor delay. Finally, he paper shows ha abnormal audi delay creaes skepicism among invesors abou earnings qualiy and hey value he disclosed earnings afer discouning for such delay. 1 Audi delay is defined as he lengh of ime from he firm s fiscal year-end o he dae of he audior s repor. Abnormal audi delay is he porion of he audi delay ha canno be explained by facors idenified in prior research ha deermine audi delay. Earnings valuaion coefficien is he pricing of $1 earnings per share in he deerminaion of he firm value.
This resul persiss even afer conrolling for he previously documened valuaion implicaions of delay in reporing earnings. Thus, he paper provides evidence o suppor he hypohesis ha abnormal audi delays signal poor earnings qualiy and ha invesors discoun he value relevance of such earnings when making resource allocaion decisions. These findings have implicaions for he role of independen audiors in he aesaion process and he informaional efficiency of he equiy markes. The res of he paper is organized as follows. The nex secion discusses he heory and hypoheses. Secion 3 presens he research design and secion 4 he sample selecion procedure. Finally, secion 5 discusses he empirical resuls and he las secion concludes. 2. Theory and Hypoheses Exising Research Several researchers have examined he deerminans of audi delays. Couris (1976), Gilling (1977), Davies and Whired (1980), and Garsombke (1981) find ha audi delays are inversely relaed o oal asses; Couris (1976) also repors ha financial firms have less delays han oher firms. Davies and Whired (1980) and Garsombke (1981) find longer delays for companies wih fiscal year-ends during he busy season. Givoly and Palmon (1982) look a relaionship beween audi delays and firm size, operaional complexiy, and qualiy of inernal conrols. Ashon e al. (1987) examine 14 deerminans of audi delays. In he mulivariae analyses, five of hese are significan. They find ha audi delay is posiively associaed wih naural logarihm of oal revenue and operaional complexiy; and negaively associaed wih publicly raded companies, qualiy of inernal conrols, and relaive mix of audi work performed a inerim and final daes. 2
Newon and Ashon (1989) examine audi delays among Canadian Big-Eigh firms. Conrary o heir expecaions, hey find ha srucured audi approaches lead o more audi delays han firms using unsrucured audi echnology. Ashon e al. (1989) find ha for a sample of Canadian firms, clien size, audior size, fiscal year ending in busy season, indusry classificaion, exisence of exraordinary iems, and sign of ne income have significan effec on audi delays. Carslaw and Kaplan (1991), in addiion o variables from prior research, add wo more variables for a sample of New Zealand firms- company conrol and deb proporion. Bamber e al. (1993) conclude ha audi delays are an increasing funcion of exen of audi work, decreasing funcion of incenives o provide a imely repor, and increasing funcion of he exen o which an audior employs a srucured audi approach. Kinney and McDaniel (1993) exend prior research by relaing audi delays o correcion of previous inerim earnings. They show ha audi delay is posiive for firms wih inerim oversaemens and declining earnings, and ha he audi delay increases wih he size of he oversaemen of inerim earnings. Ng and Tai (1994) exend prior findings on audi delays o Hong Kong firms. Lawrence and Glover (1998) repor ha, conrary o expecaions, mergers of audi firms did no lead o he expeced improvemens in operaional efficiency due o synergy. Knechel and Payne (2001) use proprieary daa o examine he effec of incremenal audi effor, resource allocaion of audi eam effor, and he provision of non-audi services on audi delays. Payne and Jensen (2002) and Johnson e al. (2002) examine audi delays in specific seings, such as municipal corporaions and local governmens. More recenly, Eredge e al. (2006) examine he impac of secion 404 of Sarbanes-Oxley Ac requiremens on audi delays and Tamara e al. (2007) examine he consequences of acceleraed filings required by SEC rule 33-8644. The laer find reducions (increases) in audi delay are associaed wih lower (higher) earnings qualiy. 3
Prior research cied above has primarily focused on he facors ha impac audi delays. Anoher sream of research has examined he effec of delay in disclosing accouning informaion on he informaion conen of he accouning disclosure. Chambers and Penman (1984), Givoly and Palmon (1982), Kross (1982), and Kross and Schroeder (1984) find ha lae earnings announcemens are associaed wih lower (even negaive) abnormal reurns han early announcemens. There is also evidence ha managemen may inenionally delay (speed up) he announcemen of bad (good) news (Givoly and Palmon 1982; Pasena and Ronen 1979; Paell and Wolfson 1982; Penman 1984; Ronen 1977; Verrechia 1983, ec.). Givoly and Palmon (1982) and Chambers and Penman (1984) argue ha he informaion conen of annual repors would deeriorae wih reporing delay as invesors gain informaion from alernaive sources of informaion, prevalence of leaks, exploiaion of inside informaion, volunary disclosures by firms, or hrough informaion ransfers from earnings repors released by oher firms (Foser 1981). This sream of lieraure deals wih inenional delay of bad news by managers or diminished informaiveness due o lack of imeliness of earnings disclosure. Limiaions of Exising Research and Incremenal Conribuion of Curren Paper A recen paper, Tamara e al. (2007) ha looks a a similar research quesion, uses he oal audi delay level and change in heir esimaion. They are, however, no conrolling for any facors ha deermine audi delay documened in prior research. Deviaion of acual audi delay from a normal audi delay (expecaion based on firm and audior characerisics), as defined in he curren paper, will yield more robus conclusions abou he relaionship beween audi delay and earnings qualiy. Tamara e al. (2007) also do no examine he incremenal informaion conained in audi delays over financial saemen reporing delays. The curren paper 4
disinguishes beween delay in reporing annual financial saemens versus audi delay; and beween diminished informaion conen due o lack of imeliness versus value relevance of earnings informaion. Even hough audi delay is he single mos imporan deerminan of he imeliness of he earnings announcemen (Givoly and Palmon 1982), audi delays explain less han 45% of he variabiliy in earnings announcemen delays; and on average, audi repors are signed more han seven days afer he earnings disclosure (see able 4). Thus, audi delays could provide incremenal informaion abou earnings qualiy beyond ha conveyed by earnings repor delays. Moreover, even-sudy mehodology ha looks a reurns/price volailiy around earning disclosure dae will no capure he influence of subsequen disclosures (such as 10-K repors) on earnings qualiy. Finally, prior sudies, such as, Chambers and Penman (1984) and Givoly and Palmon (1982) mainly focus on he imeliness aspec of he earnings disclosure. However, Saemen of Financial Accouning Conceps No. 2: Qualiaive Characerisics of Accouning Informaion, issued by he Financial Accouning Sandards Board in May 1980, liss several aribues of earnings qualiy, such as, accuracy, compleeness, verifiabiliy, neuraliy, comparabiliy, consisency, and predicive value, in addiion o imeliness. This paper uses a more comprehensive definiion of earnings qualiy han in prior research. By looking a he valuaion of earnings subsequen o he release of all relevan pieces of informaion (such as 10- K and proxy saemen), he findings shed ligh on a broader concep of earnings qualiy. The findings of his paper exend exan lieraure by examining he associaion beween audi delays and earnings qualiy and hen by looking a he effec of abnormal audi delays on he valuaion of earnings. 5
Hypoheses To he exen ha abnormal audi delays may be caused by disagreemens beween audior and clien on issues of accouning pracices, mehodology of compuaion, and accuracy of repored accouning numbers, such delays could reflec adversely on earnings qualiy. Thus, he hypohesis esed in he paper can be wrien in alernaive form as: Hypohesis 1: Ceeris paribus, abnormal audi delays are inversely associaed wih he qualiy of repored earnings. Chambers and Penman (1984), Givoly and Palmon (1982), Kross (1982), and Kross and Schroeder (1984) find ha delayed earnings announcemens are associaed wih lower (even negaive) abnormal reurns han early announcemens. Since audi delay is hypohesized o be associaed inversely wih earnings qualiy, any addiional delay before earnings repor is released is likely o be due o adminisraive reasons. If he earnings repor is released prior o he audi repor dae, he addiional delay afer he earnings repor release is expeced o provide incremenal informaion abou earnings qualiy. Thus, overall, audi delays are expeced o be a sronger predicor of earnings qualiy han earnings repor delay iself. This leads o he second hypohesis. Hypohesis 2: Ceeris paribus, abnormal audi delays provide incremenal informaion abou earnings qualiy over earnings repor delay. Since audi delays are easily observable in comparison o he qualiy of earnings, a semi-efficien form of equiy marke may use any delay ha is no explained by previously idenified facors as a proxy for poor earnings qualiy and price he repored earnings number accordingly. The las hypohesis can now be wrien in alernaive form as: 6
Hypohesis 3: Ceeris paribus, abnormal audi delays reflec adversely on he value- relevance of repored earnings. 3. Research Design The paper uses a wo-sage model o conduc he analyses. In he firs sage, a deailed model using deerminans from exan research (discussed in secion 2) ries o explain he audi delay. In he second sage, he unexplained delay from sage one is used in he associaion ess wih earnings qualiy. The following model is firs run: LADELAY = β 0 + Σ i=1-6 β i=1-6 DYR i=1-6 + β 7 DMANUF + β 8 DUTILITY + β 9 DFINANCIAL + β 10 ILADELAY + β 11 LSIZE + β 12 POWER + β 13 CURR2TA + β 14 LEVERAGE + β 15 LOSS + β 16 DISTRESS + β 17 CURRATIO + β 18 ROA + β 19 LOWNCONC + β 20 RETURN + β 21 EGROWTH + β 22 INVRATIO + β 23 SEGMENTS + β 24 SUBSIDIARIES + β 25 FOREIGNOPS + β 26 CONTINGENCY + β 27 BIG4/5 + β 28 EXPERT + β 29 TENURE + β 30 SWITCH + β 31 LNAFEE + β 32 OPINION + β 33 CITYDUM + β 34 BUSYSEASON + Є (1) The variable definiions and reasons for inclusion in he equaion are as follows. Consisen wih prior research (Johnson e al. 2002), LADELAY is defined as he naural logarihm of audi delay, measured as he number of calendar days from fiscal year-end o dae of audiors repor. 2 DYR i=1 6 are dummy variables for years 2001-06 o conrol for year specific commonaliies. DMANUF, DUTILITY, and DFINANCIAL are dichoomous variables ha represen cliens operaing in he manufacuring indusry (SIC codes 20-39), uiliy indusry (SIC codes 40-49), or financial indusry (SIC codes 60-69), and are included o represen he 2 This variable is obained from Audi Analyics daabase and abulaed as SIG_DATE_OF_OP_S minus FISCAL_YEAR_END_OP. 7
level of audi difficuly in hese indusries. There is empirical evidence ha some indusries are more difficul o audi han ohers (Simunic 1980; Turpen 1990; Pearson and Trompeer 1994). Financial and uiliy firms have relaively large asses bu have less invenory, receivables, or knowledge based asses and are easier o audi. On he oher hand, manufacuring firms do no have his advanage and are considered more difficul o audi. Couris (1976), Ashon e al. (1987) and Newon and Ashon (1989) find ha financial companies have shorer audi delays han oher companies. ILADELAY is he mean LADELAY for he wo-digi SIC indusry in ha year and is expeced o pickup up any common indusry-year effecs no capured by he previous variables. All he variables used in he paper are summarized in able 1. (inser able 1 abou here) LSIZE is he naural logarihm of firm s oal asses (Compusa daa #6). Clien size has been shown o be negaively associaed wih audi delay (Ashon e al. 1989). Ng and Tai (1994) argue ha larger companies have beer inernal conrols, allowing audiors o carry ou more inerim compliance and subsanive ess, hereby reducing year-end audi work. Addiionally, larger firms are under more pressure o release accouning informaion on a imely basis. POWER measures he clien s bargaining power and is calculaed as he naural logarihm of he audi fee paid by clien divided by he sum of he naural logarihm of audi fee paid by all cliens of he audior in ha indusry in ha year. Higher values of POWER imply greaer bargaining power for he clien (Casrella e al. 2004). Powerful cliens may be able o influence he audior o expedie heir audis. The nex four variables are included o conrol for he effecs of higher liigaion risk (Bamber e al. 1993; Francis and Wang 2005) and he audior may inves more ime audiing such firms o reduce he likelihood of liigaion. CURR2TA is he raio of curren asses 8
(Compusa daa #4) o oal asses (Compusa daa #6). LEVERAGE is he raio of oal deb (Compusa daa #9 + daa #34) o oal asses (Compusa daa #6). LOSS is a dichoomous variable wih value of one if clien has a negaive ne income before exraordinary iems, and zero oherwise. DISTRESS is he probabiliy of clien s bankrupcy based on Zmijewski s (1984) measure. The following five variables represen lower liigaion risk and herefore, reduced work for audiors (Bamber e al. 1993; Francis and Wang 2005). CURRATIO is he raio of curren asses (Compusa daa #4) o curren liabiliies (Compusa daa #5). ROA is he raio of ne income before exraordinary iems and cumulaive effec of accouning changes (Compusa daa #18) o oal asses (Compusa daa #6). LOWNCONC is he naural logarihm of owner concenraion. Owner concenraion is defined as he number of shares ousanding divided by he number of shareholders (obained from Compac-Disclosure daabase). The more concenraed he ownership of clien s shares (i.e., he less widely held he shares), he lesser he number of number of individual invesors relying on he clien s financial saemens. The lesser he reliance on clien s financial saemens, he lower he exposure o liigaion risk. RETURN is he annual reurn of he clien. EGROWTH is he percenage increase in he clien s earning from las year. Kinney and McDaniel (1993) argue ha declining earnings and declining sock prices may lead o he clien s failure, leading o quesions abou he adequacy of he audi and he reliabiliy of he financial saemens. This may increase he risk of he audior being sued. The audior would, herefore, spend exra ime on such firms in search of possible oversaemens. On he conrary, higher RETURN and EGROWTH would amoun o less ime being spen on he audi. 9
Cliens wih more complex audis would require exra effors on he par of he audiors (Ashon e al. 1987; Ashon e al. 1989; Cushing 1989; Ng and Tai 1994). The following variables proxy for audi complexiy. INVRATIO is he raio of invenory (Compusa daa #3) o oal asses (Compusa daa #6). SEGMENTS is he number of business segmens repored in COMPUSTAT segmen file. SUBSIDIARIES is he number of principal subsidiary companies held by he clien. FOREIGNOPS is a 0/1 dummy o capure he clien s operaions ouside he Unied Saes. CONTINGENCY is a dichoomous variable for he presence of coningen liabiliy (Compusa daa#327). The nex se of variables conrols for he effec of audior characerisics on audi delays. BIG4/5 is a dichoomous variable wih value of one if he audior (Compusa daa #149) is one of he Big 4 (or Big 5); and zero oherwise. Large audiors have been shown o complee heir audis faser due o more saff resources, beer experience, economies of scale (Ashon e al. 1989; Ng and Tai 1994). However, since large audi firms may have more srucured audis, hey may be slower (Newon and Ashon 1989). Also pos Enron and Andersen, such firms may be under more scruiny and may herefore be more horough, resuling in delays. Following Craswell and Taylor (1991), Craswell e al. (1995), and Casrella e al. (2004), EXPERT is coded as one if he audior has 20% or more marke share (raio of sum of naural logarihm of asses of audior s cliens in he indusry divided by he sum of naural logarihm of asses of all firms in he indusry) in he clien s indusry (wo-digi SIC classificaion). Indusry expers may audi faser due o greaer experience and economies of scale. Similarly, audiors wih greaer TENURE (coded one if audior has been wih clien for five years or more, zero oherwise) may complee heir audi sooner due o familiariy wih he clien s inernal conrols, accouning 10
sandards, ehical sandards, ec. 3 On he oher hand, as he enure increases, parner roaion required under he Sarbanes Oxley Ac may slow down he audi process. SWITCH is coded as one if he clien swiches o a new audior in he curren year and zero oherwise. A new audior would need exra ime o familiarize iself wih he clien s operaion, accouning procedures, and inernal conrol sysem (Ng and Tai 1994). Provision of non-audi services (measured as LNAFEE, he naural logarihm of non-audi fee) may speedup/delay he audi process (Knechel and Payne 2001). OPINION is a dichoomous variable wih value of one if he audi opinion (Compusa daa #149) is a qualified opinion; and zero oherwise. Cliens wih qualified opinions are likely o encouner audi delays (Ashon e al. 1989). Since a qualified audi repor conveys negaive informaion, cliens may ry o negoiae and/or delay is release by no cooperaing wih he audi process. Moreover, audiors may also spend exra ime on he audi procedures in order o reduce any uncerainies or disagreemens. CITYDUM is a dummy variable coded as one if he audi office is siuaed in one of he following coslies ciies (based on consumer price index numbers available from he Bureau of Labor Saisics): New York, Los Angeles, Chicago, San Diego, Boson, San Francisco, Philadelphia, Honolulu, and Tacoma. This variable conrols for any ciy-specific effecs on he audi process. December year-end is a busy season for audiors in USA. The audis of cliens wih fiscal year-ends during he busy season (proxied by variable BUSYSEASON ha is coded as a one/zero dummy) are likely o be delayed (Carslaw and Kaplan 1991). The error erm in equaion 1 is designaed as he abnormal audi delay, ABNDELAY. To es hypohesis 1 ha abnormal audi delay is associaed wih poor qualiy earnings, seven measures of earnings qualiy are esimaed. These are discussed below. 3 The resuls are no sensiive o he cuoff of 5 years. Using 3 or 10 years, yields similar resuls. 11
DACC is he discreionary accruals o oal asses a he end of he fiscal year, where discreionary accruals are calculaed using he cross-secional version of he Jones (1991) model as in DeFond and Jiambalvo (1994) and he difference beween ne income and cash from operaions is he measure of oal accruals (Hribar and Collins 2002). 4 MISSED is a dichoomous variable equal o one if he repored EPS on I/B/E/S daabase falls shor of he mean analys forecas made during hiry calendar days preceding he fiscal year-end. TRANSITORY is a dichoomous variable equal o one if he sum of exraordinary iems (Compusa daa #66) plus disconinued operaions (daa #192) is differen from zero, and zero oherwise. VOLATILE is a proxy for he volailiy (lack of smoohness) of earnings and is he raio of sandard deviaion of earnings o sandard deviaion of cash flow from operaions for he period -6 o. TIMELY is a measure of he imeliness of he earnings announcemen. This is esimaed as he indusry mean (wo-digi SIC classificaion) of earnings repor delay minus he firm-specific earnings repor delay. Earnings repor delay is defined as he number of calendar days delay in disclosing he earnings (obained from he Quarerly Compusa) afer he end of he fiscal year-end. PERSIST is a measure of earnings persisence. This is he AR1 parameer from a firm-specific regression of curren earnings per share on lagged earnings per share for he period -6 o. PREDICT is he adjused r-square from he AR1 regression above. Then DACC, MISSED, TRANSITORY, and VOLATILE are expeced o be negaively associaed wih earnings qualiy and TIMELY, PERSIST, and PREDICT are expeced o be posiively associaed wih earnings qualiy. Since abnormal audi delay is hypohesized o be negaively associaed wih earnings qualiy (hypohesis 1), he following relaionships are expeced. ABNDELAY is posiively associaed wih DACC, MISSED, TRANSITORY and VOLATILE; and 4 Using Absolue DACC does no change he conclusions. 12
ABNDELAY is negaively associaed wih TIMELY, PERSIST, and PREDICT. In he univariae ess of hypohesis 1, Pearson correlaions coefficiens are repored beween ABNDELAY and he seven measures of earnings qualiy discussed above. Mulivariae analysis is also conduced wih he following regression. ABNDELAY = β 0 + β 1 DACC + β 2 MISSED + β 3 TRANSITORY + β 4 VOLATILE + β 5 TIMELY + β 6 PERSIST + β 7 PREDICT + Є (2) Hypohesis 1 predics ha coefficiens β 1, β 2, β 3 and β 4, > 0; and β 5, β 6, and β 7 < 0. To es hypohesis 2, LERDELAY (naural logarihm of number of calendar days beween earnings repor dae and fiscal year-end dae) is included as an explanaory variable in equaion 2. If hypohesis 2 is rue, hen β 1, β 2, β 3 and β 4, > 0; and β 5, β 6, and β 7 < 0 will coninue o hold even in he presence of LERDELAY. The paper also combines he seven measures of earnings qualiy, DACC, MISSED, TRANSITORY, VOLATILE, TIMELY, PERSIST, and PREDICT ino a single composie measure, E-QUALITY. All measures (excep MISSED, which is already a dichoomous variable) are convered ino dichoomous variables depending on wheher hey are above or below he median value in ha year. The 0/1 values are assigned such ha value of one signifies higher earnings qualiy. The seven variable are hen added o form E-QUALITY wih values beween 0 (low earnings qualiy) o 7 (high earnings qualiy) 5. Then, hypohesis 1 predics ha E-QUALITY is negaively associaed wih ABNDELAY. The paper uses a model based on Ohlson s valuaion model o es hypohesis 3. Assuming clean surplus accouning and auoregressive ime series srucure of abnormal 5 The resuls are no sensiive o he mehod of esimaion of E-QUALITY. Two oher mehods yield similar resuls. Mehod One: All he variables are arranged in increasing or decreasing order, and assigned ranks from 1 o N (sample size), such ha higher ranks imply beer earnings qualiy. These ranks are hen added (or log ransformed and added). Mehod Two: Each variable X is ransformed as follows-[(x-minimum value)/(maximum valueminimum value)]. The ransformaion is done so ha 0 (1) suggess low (high) earnings qualiy. These ransformed variable are hen added up. 13
earnings, Ohlson (1995) shows ha marke value is some linear combinaion of earnings and book value of equiy: MV = BV + θ (3) α 1 AE + α 2 where MV is marke value of equiy a he end of period, end of period, during period. BV is book value of equiy a he AE is abnormal earnings during period and θ is a vecor of oher informaion Furher, he abnormal earnings AE is defined as a funcion of observed earnings ( E ), he risk free rae (r f ), and book value, AE (4) = E r f BV 1 Subsiuing (4) ino (3), marke value of equiy is equal o MV BV E BV = + α 1 ( f 1 ) + α 2 r θ Rearranging erms, MV BV = E r BV θ (5) α 1 ( f 1 ) + α 2 To reduce cross-secional scale differences among he firms, all variables are scaled by BV 1, book value of equiy a he beginning of period (Chrisie 1987). Equaion (5) can be rewrien as: MV BV BV 1 = r E α 0 f + α1 + BV 1 α θ ' 2 (6) Define REV, a measure called relaive excess value, where, MV BV REV = (7) BV 1 and ROE, reurn on equiy, where E ROE = (8) BV 1 14
and θ is θ deflaed by prior period BV. Using annual dummies DYR o proxy for risk free rae (r f ), equaion 6 can be rewrien as REV = α DYR + α 1 ROE + α 2 θ (9) To es he hypoheses 3, equaion 9 can be rewrien as: 6 REV = α 0 + Σ i=1-6 α i=1-6 DYR i=1-6 + α 1 ROE + α 2 ROE * ABNDELAY (10) where REV = Relaive excess value, defined as he marke value of equiy a he end of quarer 1 afer he fiscal year-end minus he curren book value of equiy deflaed by he book value of equiy for he previous fiscal year. ROE = curren earnings (excluding exraordinary iems) deflaed by lagged book value of equiy If hypohesis 2 is rue and abnormal delay in he audi process is a signal of poor earnings qualiy, he invesors should value such earnings skepically. In oher words, α 1 > 0 and α 2 < 0. Prior research (Givoly and Palmon 1982; Chambers and Penman 1984) argues ha he informaion conen of earnings repors would deeriorae wih reporing delay. To es if he delay in he audiing process is merely mimicking he reporing delay in earnings, or conains incremenal informaion beyond delay in reporing earnings, equaion 10 is modified as follows. REV = α 0 + Σ i=1-6 α i=1-6 DYR i=1-6 + α 1 ROE + α 2 ROE * ABNDELAY + α 3 ROE * LERDELAY (11) Where LERDELAY is as defined above. Then if ABNDELAY conains incremenal informaion abou earnings qualiy beyond LERDELAY, hen α 2, α 3 < 0. Also, if ABNDELAY conains more useful informaion abou earnings qualiy han LERDELAY hen absolue (α 2 ) > absolue (α 3 ). 6 Subscrip is dropped for he sake of breviy. 15
4. Sample Sample selecion procedure is described in able 2. The daa used in he paper comes from four daabases: Compusa, Audi Analyics, Compac-Disclosure, and I/B/E/S. Complee daa is available for 5,298 firms for 22,492 firm-years. Table 3 presens he sample disribuion across indusries. 7 The larges number of firms belongs o he manufacuring indusry (39.05%) and he smalles number of firms o Oher indusries (0.19%). The disribuion of sample firms across he indusries is generally similar o ha of he Compusa populaion, wih he manufacuring firm having he bigges difference (over 6%). Table 4 shows he variable characerisics. The average sample size is $6.87 billion (median = $0.4568 billion). The mean audi delay (approximaely 54 days) is almos a week more he mean earnings repor delay (approximaely 47 days). Over 85% of he clien firms are audied by big4/5 audiors wih almos 73% cliens having December 31 fiscal year-end. Almos 27% of he cliens are locaed in large ciies and over 10% cliens swiched audiors during he sample period. (inser ables 2, 3, and 4 abou here) 5. Resuls Esimaes of equaion 1 are shown in able 5. The overall adjused r-square is almos 27%. DMANUF, ILADELAY, LSIZE, CURR2TA, LEVERAGE, LOSS, CURRATIO, INVRATIO, SEGMENTS, SUBSIDIARIES, BIG4/5, EXPERT, TENURE, SWITCH, OPINION, CITYDUM, and BUSYSEASON are significan in he expeced direcions. DUTILITY, DFINANCIAL, POWER, DISTRESS, LOWNCONC, RETURN, EGROWTH, FOREIGNOPS, and CONTINGENCY are insignifican. ROA is significan bu in a direcion opposie o expeced. One reason could be ha mean ROA during he sample period is negaive and herefore he prior findings may no apply. Moreover, since LOSS already capures he 7 Based on indusry classificaion of Dopuch e al. (1987). 16
impac of negaive earnings, a predominanly negaive ROA may no follow he prior expecaions. 8 (inser able 5 abou here) Panel A of able 6 presens correlaion ess. Pearson Correlaion coefficiens of ABNDELAY wih various proxies of earnings qualiy, along wih he composie measure E- QUALITY are repored. Thus, E-QUALITY, TIMELY, PERSIST, and PREDICT are decreasing in abnormal audi delay; DACC, DACC, MISSED, VOLATILE, and TRANSITORY are increasing in abnormal audi delay. All coefficiens are significan in he prediced direcions, supporing hypohesis 1. Panel B repors addiional porfolio ess. The sample is divided ino five porfolios formed on ABNDELAY. Porfolio 1 (5) comprises he smalles (larges) values of ABNDELAY. Mean E-QUALITY is decreasing wih porfolio number, implying ha earnings qualiy is a declining funcion of abnormal audi delay, furher supporing hypohesis 1. Tess of differences among means show ha average earnings qualiy for porfolios, 1, 2, and 3 are significanly higher han hose of porfolios 3,4, and 5, respecively. Figure 1 depics he decline in earnings qualiy wih abnormal audi delay. (inser able 6 and figure 1 abou here) Resuls of equaion 2 are repored in able 7. In panel A, ABNDELAY is regressed on various proxies of earnings qualiy. DACC, MISSED, TIMELY, PREDICT, and TRANSITORY are all significan in he prediced direcion. PERSIST and VOLATILE are insignifican. Panel B presens he same es bu wih he seven explanaory variables replaced wih he composie measure, E-QUALITY. Coefficien of E-QUALITY is significan and negaive as expeced. Overall, hese resuls suppor hypohesis 1. These resuls conradic he findings of Tamara e al. 8 For he purpose of esimaing ABNDELAY, regression 1 is esimaed separaely for each year (no repored) o allow he slopes o vary wih ime. 17
(2007). However, given he limiaions of Tamara e al. (2007) discussed earlier, he findings of he curren paper are more reliable. Panels C and D conduc he same es as in panels A and B, respecively, bu in he presence of earnings repor delay, LERDELAY. The overall conclusions are unchanged. Thus, abnormal audi delay conveys informaion abou earnings qualiy ha is incremenal o he informaion conveyed by delay in repored earnings. Resuls of panel C and D suppor hypohesis 2. (inser able 7 abou here) Esimaes of equaions 10 and 11 are presened in able 8. Version 1 conains only ROE as an explanaory variable. Version 2 adds ROE*ABNDELAY and version 3 adds ROE*LERDELAY o version 1. Finally, version 4 adds boh ROE*LERDELAY and ROE*ABNDELAY o version 1. ROE is always significan and posiive in all four versions, as expeced. ROE*LERDELAY is significan and negaive in versions 3 and 4; and ROE*ABNDELAY is significan and negaive in version 2 and 4. Adding ineracion of ROE wih ABNDELAY o version 1increases he adjused r-square from 0.0113 o 0.0123. The Vuong (1989) saisics (F=15.9341) is significan a less han 1% level suggesing a significan gain in explanaory power wih he addiion of ABNDELAY. Similarly, version 3 has significanly higher adjused r-square han version 1 (F=22.7699; p<0.0001). Finally, version 4 has significanly higher r-square han version 3 (F=32.2053; p<0.0001). This implies ha adding ineracion erm wih ABNDELAY in he presence of LERDELAY adds significan explanaory power o he model. These resuls sugges ha as he abnormal audi delay increases, he marke ges more conservaive / skepical in valuing a dollar of earnings. Moreover, ABNDELAY has a more criical role han LERDELAY in he valuaion process, since he coefficien of 18
ABNDELAY is more han 10 imes larger in absolue value han ha of LERDELAY (F=29.69; p<0.0001). Overall, his evidence suppors hypohesis 3. (inser able 8 abou here) Regression Diagnosics Various diagnosic ess are conduced on he regressions. Whie's (1980) es for heeroskedasiciy rejecs he null of homoskedasic errors for all he regressions. Heeroskedasiciy correced saisics are esimaed (no repored) bu none of he earlier conclusions are changed. Mulicollineariy checks are also conduced on all he regressions using Belsley, Kuh, and Welsch's (1980) procedure. Variance inflaion facors (no abulaed) are less han 5 and, hus, insignifican for all he regressions. Finally, ess for ouliers are conduced on all he regressions using Belsley, Kuh, and Welsch's (1980) procedure. Sudenized residuals are compued (wihou he curren observaion) and any observaion deviaing more han wo sandard deviaions from he mean sudenized residual is deleed. Resuls (no repored) do no change qualiaively when ouliers are removed. Thus, he resuls appear o be robus wih regards o heeroskedasiciy, mulicollineariy, and ouliers. 6. Conclusion This paper addresses he research quesions: 1) Does he abnormal delay in he audi process signal poor earnings qualiy? 2) Is his informaion abou earnings qualiy incremenal o ha conained in earnings repor delay? and 3) Does he marke use his informaion abou earnings qualiy in valuing he firm? To he exen ha unexpeced delays in he audi process may be caused by serious differences beween he independen audior and he clien over he 19
numbers o be repored, such delays may porend a lower earnings qualiy. Since audi delay is more readily observable han he elusive earnings qualiy, invesors may use his delay o form an opinion abou he qualiy of he repored earnings. In a perfec world, he independen audior would coninue o spend more ime on he audi ill he earnings qualiy was insured. However, audiors may have oher concerns driving heir decisions, such as lack of independence, fear of losing a major clien, limiaions of resources, such as, ime, manpower, ec. The earnings repored afer he exended audi process may sill suffer from deficiencies. Resuls repored in he paper are consisen wih abnormal audi delays being significanly associaed wih poor earnings qualiy. The informaion abou earnings qualiy conained in abnormal audi delays is incremenal o he informaion conveyed by delays in releasing earnings repors. When he marke values a dollar of repored earnings, i appears o discoun he valuaion by he exen of abnormal audi delay. This discouning is presen even afer conrolling for he negaive signaling abou earnings qualiy conveyed by earnings repor delays. These resuls are imporan for undersanding he mechanism of he audi process. Apparenly, he audiors are concerned abou earnings qualiy as hey appear o spend more ime on cliens ha have poor earnings qualiy. However, due o consrains in he conracing and audi process and due o limiaions of he accouning rules, a significan porion of he deficiencies in he earnings carries hrough o he repored earnings. The marke appears o behave efficienly o he exen ha i ries o adjus he valuaion of repored earnings based on he delays in he audi process. Wheher he adjusmen is complee or parial is beyond he scope of his paper, bu an ineresing issue for fuure research. 20
References Ashon, R. H., P. R. Graul, and J. D. Newon. 1989. Audi delay and he imeliness of corporae reporing, Conemporary Accouning Research (Spring): 657-673. Ashon, R. H., J. J. Willingham, and R. K. Ellio. 1987. An empirical analysis of audi delay, Journal of Accouning Research (Auumn): 275-292. Bamber, E. M., L. S. Bamber, and M. P. Schoderbek. 1993. Audi srucure and oher deerminans of audi repor lag: An empirical analysis. Audiing: A Journal of Pracice and Theory (Spring): 1-23. Belsley, D. A., E. Kuh, and R. E. Welsch. 1980. Regression Diagnosics: Idenifying Influenial Daa and Sources of Collineariy. New York: John Wiley and Sons. Carslaw, C. A. P. N., S. E. Kaplan. 1991. An examinaion of audi delay: Furher evidence from New Zealand. Accouning and Business Research (Vol. 22, No. 85): 21-32. Caserella, J. R., J. R. Francis, B. L. Lewis, and P. L. Walker. 2004. Audior indusry specializaion, clien bargaining power, and audi pricing. Audiing: A Journal of Pracice and Theory (March): 123-140. Chambers, A. E., and S. H. Penman. 1984. Timeliness of reporing and he sock price reacion o earnings announcemens. Journal of Accouning Research (Vol. 22, No. 1): 21-47. Chrisie, A. 1987. On cross-secional analysis in accouning research. Journal of Accouning and Economics 9 (December):231-258 Craswell, A., and S. Taylor. 1991. The marke srucure of audiing in Ausralia: The role of indusry specicializaion. Research in Accouning Regulaions (Vol. 5): 55-77. Craswell, A., J. R. Francis., and S. L. Taylor. 1995. Audior brand name repuaions and indusry specializaions. Journal of Accouning and Economics (Vol. 20): 297-322. Couris, J. K. 1976. Relaionships beween imeliness in corporae reporing and corporae aribues, Accouning and Business Research (Winer): 45-76. Cushing, B. 1989. Discussion of he associaion beween audi echnology and audi delay. Audiing: A Journal of Pracice and Theory (Supplemen): 38-47. Davies, B., and G. P. Whired. 1980. The associaion beween seleced corporae aribues and imeliness in corporae reporing: Furher analysis, Abacus (June):48-60. DeFond, M.L., and J. Jiambalvo. 1994. Deb covenan violaion and manipulaion of accruals. Journal of Accouning and Economics 17: 145-176. 21
Dopuch, N., R.W. Holhausen, and R.W. Lefwich. 1987. Predicing audi qualificaions wih financial and marke variables. The Accouning Review 62 (July): 431-454. Eredge, M. L., L. Siu, and C. Li. 2006. The impac of SOX secion 404 inernal conrol qualiy assessmen on audi delay in he SOX era. Audiing: A Journal of Pracice and Theory (November): 1-23. Francis, J. R., and D. Wang. 2005. Impac of he SEC s public fee disclosure requiremen on subsequen period fees and implicaions for marke efficiency. Audiing: A Journal of Pracice and Theory (Supplemen): 145-160. Garsombke, H. P. 1981. The imeliness of corporae financial disclosure, in J. K. Couris, ed., Communicaion via Annual Repors, AFM Exploraory Series No. 11 (Universiy of New England, Armidale, N.S.W.): 204-218. Gilling, D. M. 1977. Timeliness of corporae reporing: Some furher commen, Accouning and Business Research (Winer): 34-36. Givoly, D., and D. Palmon. 1982. Timeliness of annual earnings announcemens: Some empirical evidence, The Accouning Review (July): 486-508. Hribar, P., and D.W. Collins. 2002. Errors in esimaing accruals: Implicaions for empirical research, Journal of Accouning Research 40 (March): 105-134. Johnson, L. E., and S. P. Davies, and R. J. Freeman. 2002. The effec of seasonal variaions in audior workload on local governmen audi fees and audi delay. Journal of Accouning and Public Policy (Vol. 21): 395-422. Jones, J. 1991. Earnings managemen during impor relief invesigaions. Journal of Accouning Research 29 (Fall): 193-228. Kinney, W. R., Jr. and L. S. McDaniel. 1993. Audi delay for firms correcing quarerly earnings. Audiing: A Journal of Pracice and Theory (Fall) 135-142. Knechel, W. R., and J. L. Payne. 2001. Addiional evidence on audi repor lag. Audiing: A Journal of Pracice and Theory (March): 137-146. Kross, W. 1982. Profiabiliy, earnings announcemen ime lags, and sock prices. Journal of Business Finance and Accouning (Spring): 313-328. Kross, W. and D. A. Schroeder. 1984. An empirical invesigaion of he effec of quarerly earnings announcemen iming on sock reurns. Journal of Accouning Research (Spring): 153-176. Lamber, T., J. Brazel, and K. Jones. 2007. Uninended consequences of acceleraed findings: Do changes in audi delay lead o changes in earnings qualiy? Working Paper Drexel 22
Universiy, PA. Lawrence, J. E., and H. D. Glover. 1998. The effec of audi firm mergers on audi delay. Journal of Managerial Issues (Summer): 151-164. Newon, J. D., and R. H. Ashon. 1989. The associaion beween audi repor echnology and audi delay. Audiing: A Journal of Pracice and Theory (Supplemen): 22-37. Ng., P. P. H., and B. Y. K. Tai. 1994. An empirical invesigaion of he deerminans of audi delay in Hong Kong. Briish Accouning Review (26): 43-59. Ohlson, J. A. 1995 Earnings, Book values, and dividends in equiy valuaion. Conemporary Accouning Research 11 (Spring): 661-687. Pasena, V., and J. Ronen. 1979. Some hypoheses on he paern of managemen s informal disclosures. Journal of Accouning Research (Auumn): 550-564. Paell, J. M.\, and M. A. Wolfson. 1982. Good news, bad news, and he inraday iming of corporae disclosures. The Accouning Review (July): 509-527. Payne, J. L., and K. L. Jensen. 2002. An examinaion of municipal audi delay. Journal of Accouning and Public Policy (Vol. 21): 1-29. Pearson, T. and G. Tromper. 1994. Compeiion in he marke for audi services The effec of supplier concenraion on audi fees. Conemporary Accouning Research (Vol. 11, No. 1):91-114. Penman, S. H. 1984. Abnormal reurns o invesmen sraegies based on he iming of earnings repors. Journal of Accouning and Economics (December): 165-183. Ronen, J. 1977. The effec of insider rading rules on informaion generaion and disclosure by corporaions. The Accouning Review (April): 438-449. Simunic, D. A. 1980. The pricing of audi services: Theory and evidence. Journal of Accouning Research Vol. 18, No. 1): 161-190. Turpen, R. A. 1990. Differenial pricing on audiors iniial engagemens: Furher evidence. Audiing: A Journal of Pracice and Theory (Vol. 9, No. 2): 60-76. Verrecchia, R. E. 1983. Discreionary disclosure. Journal of Accouning and Economics (December): 179-194. Vuong, Q. 1989. Likelihood raio ess for model selecion and non-nesed hypoheses. Economerica (Vol. 59, 2): 307-333 23
Whie, H. 1980. A heeroskedasiciy-consisen covariance marix esimaor and a direc es for heeroskedasiciy. Economerica 48 (May): 817 38. Zmijewski, M. E. 1984. Mehodological issues relaed o he esimaion of financial disress predicion models. Journal of Accouning Research 22 (Supplemen): 59-82. 24
Variable ABNDELAY BIG4/5 BUSYSEASON CITYDUM TABLE 1 VARIABLE DEFINITIONS (in alphabeical order) Definiion LADELAY no explained by equaion 1 (ha is, he error erm) Dichoomous variable wih value of one if he audior (Compusa daa #149) is one of he Big 4 (or Big 5); and zero oherwise Dichoomous variable coded as a one for December 31 fiscal year-ends, zero oherwise Dichoomous variable coded as one if he audi office is siuaed in one of he following coslies ciies (based on consumer price index numbers available from he Bureau of Labor Saisics): New York, Los Angeles, Chicago, San Diego, Boson, San Francisco, Philadelphia, Honolulu, and Tacoma CONTINGENCY Dichoomous variable for he presence of coningen liabiliy (Compusa daa#327) CURR2TA Raio of curren asses (Compusa daa #4) o oal asses (Compusa daa #6) CURRATIO DACC DFINANCIAL DISTRESS DMANUF DUTILITY Raio of curren asses (Compusa daa #4) o curren liabiliies (Compusa daa #5) Raio of discreionary accruals o oal asses a he end of he fiscal year, where discreionary accruals are calculaed using he cross-secional version of he Jones (1991) model as in DeFond and Jiambalvo (1994) and he difference beween ne income and cash from operaions is he measure of oal accruals (Hribar and Collins 2002) Dichoomous variables ha represen cliens operaing in he financial indusry (SIC codes 60-69) Probabiliy of clien s bankrupcy based on Zmijewski s (1984) measure Dichoomous variables ha represen cliens operaing in he manufacuring indusry (SIC codes 20-39) Dichoomous variables ha represen cliens operaing in he uiliy indusry (SIC codes 40-49) DYR Dummy variables for years 2001-06 EGROWTH Percenage increase in he clien s earning from las year E-QUALITY Composie index of earnings qualiy based on DACC, MISSED, PERSIST, PREDICT, TIMELY, TRANSITORY, and VOLATILE EXPERT Dichoomous variable coded as one if he audior has 20% or more marke share (raio of sum of naural logarihm of asses of audior s cliens in he indusry divided by he sum of naural logarihm of asses of all firms in he indusry) in he clien s indusry (wo-digi SIC classificaion) FOREIGNOPS Proporion of a clien s operaions ouside he Unied Saes. ILADELAY Mean LADELAY for he wo-digi SIC indusry in ha year INVRATIO Raio of invenory (Compusa daa #3) o oal asses (Compusa daa #6) LADELAY Naural logarihm of audi delay, measured as he number of calendar days from fiscal year-end o dae of audiors repor (ADELAY) LERDELAY Naural logarihm of number of calendar days beween earnings repor dae and fiscal year-end dae (ERDELAY) 25
TABLE 1 (coninued) Variable Definiion LEVERAGE Raio of oal deb (Compusa daa #9 + daa #34) o oal asses (Compusa daa #6) LNAFEE Naural logarihm of non-audi fee LOSS Dichoomous variable wih value of one if clien has a negaive ne income before exraordinary iems, and zero oherwise LOWNCONC Naural logarihm of owner concenraion-defined as he number of shares ousanding divided by he number of shareholders (obained from Compac- Disclosure daabase) LSIZE Naural logarihm of firm s oal asses (Compusa daa #6) MISSED Dichoomous variable equal o one if he repored EPS on I/B/E/S daabase falls shor of he mean of analys forecass made during hiry calendar days preceding he fiscal year-end OPINION Dichoomous variable wih value of one if he audi opinion (Compusa daa #149) is a qualified opinion; and zero oherwise PERSIST AR1 parameer from a firm-specific regression of curren earnings per share on lagged earnings per share for he period -6 o POWER Naural logarihm of he audi fee paid by clien divided by he sum of he naural logarihm of audi fee paid by all cliens of he audior in ha indusry in ha year PREDICT Adjused r-square from he a firm-specific AR1 regression of curren earnings per share on lagged earnings per share for he period -6 o RETURN Annual reurn of he clien REV Relaive excess value, defined as he marke value of equiy a he end of quarer 1 afer he fiscal year-end minus he curren book value of equiy deflaed by he book value of equiy for he previous fiscal year ROA Raio of ne income before exraordinary iems and cumulaive effec of accouning changes (Compusa daa #18) o oal asses (Compusa daa #6) ROE Reurn on equiy, defined as curren earnings (excluding exraordinary iems) deflaed by lagged book value of equiy SEGMENTS Number of business segmens repored in COMPUSTAT segmen file SUBSIDIARIES Number of principal subsidiary companies held by he clien SWITCH Dichoomous variable coded as one if he clien swiches o a new audior in he curren year and zero oherwise TENURE Dichoomous variable coded as one if audior has been wih clien for five years or more, zero oherwise TIMELY Measure of he imeliness of he earnings announcemen; esimaed as he indusry mean (wo-digi SIC classificaion) of earnings repor delay minus he firmspecific earnings repor delay TRANSITORY Dichoomous variable equal o one if he sum of exraordinary iems (Compusa daa #66) plus disconinued operaions (daa #192) is differen from zero, and zero oherwise. VOLATILE Proxy for he volailiy (lack of smoohness) of earnings, measured as he raio of sandard deviaion of earnings o sandard deviaion of cash flow from operaions for he period -6 o 26
TABLE 2 SAMPLE SELECTION PROCEDURE Daa Sep Firm-Year Observaions Daa available on Compusa (2000-06) 203,697 Daa available on Audi Analyics 76,777 Daa available on Compac-Disclosure 45,108 Daa available on I/B/E/S 107,911 Complee daa available on all Daabases 22,492 Noe: The final sample perains o 5,298 unique firms. 27
TABLE 3 SAMPLE DISTRIBUTION ACROSS INDUSTRIES Indusry Sample Compusa Populaion 1. Agriculure, Foresry, and Fishing 0.28% 0.32% 2. Mining 3.99% 5.21% 3. Consrucion 1.02% 1.05% 4. Manufacuring 39.05% 32.84% 5. Transporaion and Uiliies 8.06% 9.04% 6. Wholesale 2.79% 3.45% 7. Reail 5.53% 5.39% 8. Financial Services 22.88% 23.66% 9. Services 16.20% 17.58% 10. Ohers 0.19% 1.45% Toal 100.00% 100.00% The indusry classificaion is based on Dopuch e al. (1987), and includes he following SIC codes: Agriculure, Foresry, and Fishing 100-999 Mining 1000-1499 Consrucion 1500-1999 Manufacuring 2000-3999 Transporaion and Uiliies 4000-4999 Wholesale 5000-5199 Reail 5200-5999 Financial Services 6000-6999 Services 7000-8999 Ohers < 100 and > 8999 28
TABLE 4 VARIABLE DISTRIBUTION Variable Mean Sand-Devn Quarile 1 Median Quarile 3 To Asses ($bill) 6.8741 51.2570 0.1055 0.4568 1.8235 ADELAY 53.7750 20.1847 37 55 70 ERDELAY 46.6766 21.0151 30 43 60 BIG4/5 0.8546 0.3527 1 1 0 BUSYSEASON 0.7263 0.4459 1 1 0 CITYDUM 0.2694 0.4437 0 0 1 CONTINGENCY 0.0922 0.2894 0 0 0 CURR2TA 0.3973 0.3085 0.0835 0.4018 0.6538 CURRATIO 2.7388 17.5967 0.6351 1.6275 2.9986 DACC 0.2075 32.7295-0.0915 0.0310 0.2248 DFINANCIAL 0.2288 0.4201 0 0 0 DISTRESS 0.0240 0.1268 0 0.0001 0.0003 DMANUF 0.3905 0.4879 1 0 0 DUTILITY 0.0806 0.2723 0 0 0 EGROWTH 0.0036 2.0839-0.5257 0.0532 0.4370 EXPERT 0.5806 0.4935 1 1 0 FOREIGNOPS 0.2812 0.4496 0 0 1 INVRATIO 0.0896 0.1328 0.0001 0.0255 0.1356 LEVERAGE 0.2129 0.5634 0.0198 0.1536 0.3227 LNAFEE 12.1377 1.7962 10.9000 12.0901 13.3000 LOSS 0.2760 0.4471 0 0 1 LOWNCONC 7.9962 4.8530 7.7200 9.5862 11.3000 MISSED 0.2136 0.4099 0 0 0 OPINION 0.2938 0.4555 0 0 1 PERSIST 0.2334 0.5932-0.0869 0.2030 0.5350 POWER 1.2957 3.1553 0.8928 1.0910 1.3900 PREDICT 0.0255 0.3586-0.2290 0.1020 0.2020 RETURN 0.1093 0.4512-0.1733 0.0465 0.3538 REV 1.2368 3.3443 0.1169 0.8126 2.1060 ROA -0.0077 0.1568-0.0094 0.0203 0.0638 ROE 0.0274 0.3443-0.0023 0.0878 0.1623 SEGMENTS 1.8194 1.2313 1 1 2 SUBSIDIARIES 1.8805 3.4504 0 0 3 SWITCH 0.1083 0.3108 0 0 0 TENURE 0.4868 0.4998 0 0 1 TIMELY 8.3655 20.3539-3.4740 11.4687 22.3300 TRANSITORY 0.2132 0.4096 0 0 1 VOLATILE 1.2798 1.3655 0.5873 0.9239 1.4072 See able 1 for variable definiions. 29
TABLE 5 REGRESSION OF AUDIT DELAY ON ITS DETERMINANTS (Dependen Variable = LADELAY) Expeced Esimae Saisics p Value VIF Variables Sign Inercep ***1.9805 29.74 <0.0001 0.0000 DMANUF + ***0.0444 6.44 <0.0001 1.5976 DUTILITY -0.0067-0.58 0.5605 1.3859 DFINANCIAL -0.0127-1.23 0.2181 2.6571 ILADELAY + ***0.4580 26.43 <0.0001 3.3900 LSIZE ***-0.0490-20.81 <0.0001 3.5414 POWER -0.0012-1.45 0.1465 1.0275 CURR2TA + ***0.0991 6.66 <0.0001 2.9241 LEVERAGE + ***0.1462 11.35 <0.0001 4.7321 LOSS + ***0.0844 12.14 <0.0001 1.3441 DISTRESS + -0.0190-0.78 0.4349 1.2516 CURRATIO **-0.0014-2.16 0.0304 1.3292 ROA ***0.0134 11.29 <0.0001 4.2844 LOWNCONC -0.0009-1.55 0.1203 1.1221 RETURN -0.0009-0.14 0.8921 1.2363 EGROWTH 0.0003 0.22 0.8229 1.0247 INVRATIO + ***0.1691 7.24 <0.0001 1.3384 SEGMENTS + **0.0062 2.25 0.0242 1.6588 SUBSIDIARIES + ***0.0027 2.72 0.0065 1.6641 FOREIGNOPS + -0.0061-0.88 0.3770 1.3855 CONTINGENCY + 0.0116 1.24 0.2162 1.0658 BIG4/5 +/ ***0.0972 9.88 <0.0001 1.5169 EXPERT ***-0.0800-12.91 <0.0001 1.3060 TENURE +/ **0.0143 2.34 0.0191 1.3078 SWITCH + ***0.0364 3.62 0.0003 1.1718 LNAFEE? ***0.0196 8.30 <0.0001 2.5375 OPINION + ***0.0598 9.21 <0.0001 1.2449 CITYDUM + ***0.0533 8.76 <0.0001 1.0311 BUSYSEASON + ***0.0327 5.04 <0.0001 1.1840 Observaions 22,492 Adjused R-Sqr 0.2662 F Value 244.18 Probabiliy > F <0.0001 Whie s Chi-Sqr. 2884.77 Prob > Chi-Sqr. <0.0001 See able 1 for variable definiions. Annual dummies (DYR) are no repored for he sake of breviy. *** implies wo-sided significance a 1% and ** implies wo-sided significance a 5%. 30
TABLE 6 ASSOCIATION OF ABNORMAL AUDIT DELAY WITH EARNINGS QUALITY Panel A: Correlaion of Abnormal Audi Delay (ABNDELAY) wih Earnings Qualiy Variable Prediced Sign Pearson Correlaion Coefficien p-value E-QUALITY ***-0.1214 <0.0001 DACC + **0.0179 0.0145 DACC + **0.0168 0.0219 MISSED + ***0.0182 0.0061 TRANSITORY + ***0.0434 <0.0001 VOLATILE + ***0.0188 0.0083 TIMELY ***-0.2794 <0.0001 PERSIST ***-0.0342 <0.0001 PREDICT ***-0.0359 <0.0001 Panel B: Porfolio Tess on Abnormal Audi Delay (ABNDELAY) and Earnings Qualiy Porfolio of ABNDELAY Mean E-QUALITY Tes Beween Porfolios -Saisics I (Lowes) 4.4259 II 4.3593 I & III ***9.8992 III 4.1503 II & IV ***7.6918 IV 4.1225 III & V ***2.9815 V (Highes) 4.0633 See able 1 for variable definiions. *** implies wo-sided significance a 1% ** implies wo-sided significance a 5% 31
TABLE 7 REGRESSION OF ABNORMAL AUDIT DELAY ON COMPONENTS OF EARNINGS QUALITY Panel A: Regression on Componens of Earnings Qualiy Variables Expeced Sign Dependen Variable = ABNDELAY Esimae Saisics p Value VIF Inercep ***0.0527 11.25 <0.0001 0.0000 DACC + **0.0002 2.02 0.0433 1.0007 MISSED + ***0.0272 4.18 <0.0001 1.0076 TRANSITORY + ***0.0407 6.28 <0.0001 1.0043 VOLATILE + 0.0005 0.23 0.8144 1.0184 TIMELY ***-0.0063-45.72 <0.0001 1.0103 PERSIST -0.0009-0.18 0.8609 1.2772 PREDICT *-0.0170-1.92 0.0552 1.2809 Observaions 22,492 Adjused R-Sqr 0.1044 F Value 307.33 Probabiliy > F <0.0001 Whie s Chi-Sqr. 500 Prob > Chi-Sqr. <0.0001 Panel B: Regression on Earnings Qualiy Variables Expeced Sign Dependen Variable = ABNDELAY Esimae Saisics p Value VIF Inercep ***0.1339 15.48 <0.0001 0.0000 E-QUALITY ***-0.0321-15.81 <0.0001 1.0000 Observaions 22,492 Adjused R-Sqr 0.0133 F Value 249.83 Probabiliy > F <0.0001 Whie s Chi-Sqr. 12.10 Prob > Chi-Sqr. 0.0024 32
TABLE 7 (coninued) Panel C: Regression on Componens of Earnings Qualiy in he Presence of Earnings Repor Delay Variables Expeced Sign Dependen Variable = ABNDELAY Esimae Saisics p Value VIF Inercep ***-0.8649-19.83 <0.0001 0.0000 LERDELAY + ***0.2325 21.15 <0.0001 3.1306 DACC + **0.0002 2.00 0.0459 1.0007 MISSED + ***0.0254 3.95 <0.0001 1.0078 VOLATILE + -0.0002-0.08 0.9383 1.0184 TRANSITORY + ***0.0383 5.98 <0.0001 1.0046 TIMELY ***-0.0021-8.83 <0.0001 3.1391 PERSIST 0.0036 0.67 0.5012 1.2793 PREDICT ***-0.0248-2.83 0.0047 1.2832 Observaions 22,492 Adjused R-Sqr 0.1256 F Value 331.37 Probabiliy > F <0.0001 Whie s Chi-Sqr. 916 Prob > Chi-Sqr. <0.0001 Panel D: Regression on Earnings Qualiy in he Presence of Earnings Repor Delay Variables Expeced Sign Dependen Variable = ABNDELAY Esimae Saisics p Value VIF Inercep ***-1.1181-40.25 <0.0001 0.0000 LERDELAY + ***0.3053 47.16 <0.0001 1.0785 E-QUALITY ***-0.0068-3.39 0.0007 1.0785 Observaions 22,492 Adjused R-Sqr 0.1197 F Value 1252.04 Probabiliy > F <0.0001 Whie s Chi-Sqr. 185.43 Prob > Chi-Sqr. <0.0001 See able 1 for variable definiions. *** implies wo-sided significance a 1% ** implies wo-sided significance a 5% * implies wo-sided significance a 10% 33
TABLE 8 REGRESSION OF RELATIVE EXCESS VALUE ON ROE AND DELAY VARIABLES Variables Expeced Sign Dependen Variable = REV Version 1 Version 2 Version 3 Version 4 Inercep ***1.4878 ***1.4835 ***1.4857 ***1.4800 (24.93) (24.87) (24.91) (24.83) ROE + ***0.2132 ***0.2073 ***0.2639 ***0.2665 (3.35) (3.26) (4.11) (4.15) ROE*LERDELAY ***-0.0027 ***-0.0032 (-5.50) (-6.48) ROE*ABNDELAY ***-0.0308 ***-0.0385 (-4.79) (-5.88) Observaions 22,492 22,492 22,492 22,492 Adjused R-Sqr 0.0113 0.0123 0.0126 0.0141 F Value 37.85 36.02 36.95 36.74 Probabiliy > F <0.0001 <0.0001 <0.0001 <0.0001 Whie s Chi-Sqr. 1059.58 1059.98 1064.81 1066.46 Prob > Chi-Sqr. 0.0001 0.0001 0.0001 0.0001 Voung s F Tes 15.9341 22.7699 32.2053 Probabiliy > F <0.0001 <0.0001 <0.0001 See able 1 for variable definiions. Annual dummies (DYR) are no repored for he sake of breviy. *** implies wo-sided significance a 1% 34
FIGURE 1 PLOT OF EARNINGS QUALITY VERSUS ABNORMAL AUDIT DELAY See able 1 for variable definiaions. 35